Reversible Protein Acetylation

Novartis Foundation Symposium 259

REVERSIBLE PROTEIN ACETYLATION

2004

REVERSIBLE PROTEIN ACETYLATION

The Novartis Foundation is an international scienti¢c and educational charity (UK Registered Charity No. 313574). Known until September 1997 as the Ciba Foundation, it was established in 1947 by the CIBA company of Basle, which merged with Sandoz in 1996, to form Novartis. The Foundation operates independently in London under English trust law. It was formally opened on 22 June 1949. The Foundation promotes the study and general knowledge of science and in particular encourages international co-operation in scienti¢c research. To this end, it organizes internationally acclaimed meetings (typically eight symposia and allied open meetings and 15^20 discussion meetings each year) and publishes eight books per year featuring the presented papers and discussions from the symposia. Although primarily an operational rather than a grant-making foundation, it awards bursaries to young scientists to attend the symposia and afterwards work with one of the other participants. The Foundation’s headquarters at 41 Portland Place, London W1B 1BN, provide library facilities, open to graduates in science and allied disciplines. Media relations are fostered by regular press conferences and by articles prepared by the Foundation’s Science Writer in Residence. The Foundation o¡ers accommodation and meeting facilities to visiting scientists and their societies.

Information on all Foundation activities can be found at http://www.novartisfound.org.uk

Novartis Foundation Symposium 259

REVERSIBLE PROTEIN ACETYLATION

2004

Copyright & Novartis Foundation 2004 Published in 2004 by John Wiley & Sons Ltd, The Atrium, Southern Gate, Chichester PO19 8SQ, UK 01243 779777 National International (+44) 1243 779777 e-mail (for orders and customer service enquiries): [email protected] Visit our Home Page on http://www.wileyeurope.com or http://www.wiley.com All Rights Reserved. No part of this book may be reproduced, stored in a retrieval system or transmitted in any form or by any means, electronic, mechanical, photocopying, recording, scanning or otherwise, except under the terms of the Copyright, Designs and Patents Act 1988 or under the terms of a licence issued by the Copyright Licensing Agency Ltd, 90 Tottenham Court Road, London W1T 4LP, UK, without the permission in writing of the Publisher. Requests to the Publisher should be addressed to the Permissions Department, John Wiley & Sons Ltd,The Atrium, Southern Gate, Chichester,West Sussex PO19 8SQ, England, or emailed to [email protected], or faxed to (+44) 1243 770620. This publication is designed to provide accurate and authoritative information in regard to the subject matter covered. It is sold on the understanding that the Publisher is not engaged in rendering professional services. If professional advice or other expert assistance is required, the services of a competent professional should be sought. Other Wiley Editorial O⁄ces John Wiley & Sons Inc., 111 River Street, Hoboken, NJ 07030, USA Jossey-Bass, 989 Market Street, San Francisco, CA 94103-1741, USA Wiley-VCH Verlag GmbH, Boschstr. 12, D-69469 Weinheim, Germany John Wiley & Sons Australia Ltd, 33 Park Road, Milton, Queensland 4064, Australia John Wiley & Sons (Asia) Pte Ltd, 2 Clementi Loop #02-01, Jin Xing Distripark, Singapore 129809 John Wiley & Sons Canada Ltd, 22 Worcester Road, Etobicoke, Ontario, Canada M9W 1L1 Wiley also publishes its books in a variety of electronic formats. Some content that appears in print may not be available in electronic books. Novartis Foundation Symposium 259 x+300 pages, 58 ¢gures, 7 tables

British Library Cataloguing in Publication Data A catalogue record for this book is available from the British Library ISBN 0 470 86261 0 Typeset in 101
2 on 121
2 pt Garamond by DobbieTypesetting Limited, Tavistock, Devon. Printed and bound in Great Britain by T. J. International Ltd, Padstow, Cornwall.

This book is printed on acid-free paper responsibly manufactured from sustainable forestry, in which at least two trees are planted for each one used for paper production.

Contents Symposium on Reversibleprotein acetylation, held atthe Novartis Foundation, London, 6^8 May 2003 Editors: Gregory Bock (Organizer) and Jamie Goode This symposium is based on a proposal made by Dalia Cohen EricVerdin

Chair’s introduction 1

Yanming Wang,Wolfgang Fischle,Wang Cheung, Steven Jacobs, Sepideh Khorasanizadeh and C. David Allis Beyond the double helix: writing and reading the histone code 3 Discussion 17 Gunnar Schotta, Monika Lachner, Antoine H. F. M. Peters and Thomas Jenuwein The indexing potential of histone lysine methylation Discussion 37

22

Danesh Moazed, Adam D. Rudner, Julie Huang, Georg J. Hoppe and Jason C.Tanny A model for step-wise assembly of heterochromatin in yeast 48 Discussion 56 Anastasia Wyce, Karl W. Henry and Shelley L. Berger de-ubiquitylation in gene activation 63 Discussion 73 Ronen Marmorstein Discussion 98

H2B ubiquitylation and

Structural and chemical basis of histone acetylation 78

Louis C. Mahadevan, Alison L. Clayton, Catherine A. Hazzalin and Stuart Thomson Phosphorylation and acetylation of histone H3 at inducible genes: two controversies revisited 102 Discussion 111 EricVerdin, Frank Dequiedt and Herb Kasler developing thymocytes 115 Discussion 129 v

HDAC7 regulates apoptosis in

vi

CONTENTS

TimothyA. McKinsey and Eric N. Olson control of cardiac growth 132 Discussion 141

Dual roles of histone deacetylases in the

David Ciccone and Marjorie Oettinger Chromatin modi¢cations as clues to the regulation of antigen receptor assembly 146 Discussion 158 General discussion I

Histone modi¢cation in X inactivation 163

Je¡ery J. Kovacs, Charlotte Hubbert and Tso-Pang Yao The HDAC complex and cytoskeleton 170 Discussion 178 Melanie Ott, Alexander Dorr, Claudia Hetzer-Egger, Katrin Kaehlcke, Martina Schnolzer, Peter Henklein, Phil Cole, Ming-Ming Zhou and EricVerdin Tat acetylation: a regulatory switch between early and late phases in HIV transcription elongation 182 Discussion 193 Wei Gu, Jianyuan Luo, Chris L. Brooks, Anatoly Y. Nikolaev and Muyang Li Dynamics of the p53 acetylation pathway 197 Discussion 205 Warner C. Greene and Lin-feng Chen acetylation 208 Discussion 218 General discussion II

Regulation of NF-kB action by reversible

p300 and DNA repair

223

Stephen B. Baylin Reversal of gene silencing as a therapeutic target for cancer  roles for DNA methylation and its interdigitation with chromatin 226 Discussion 234 Shaowen Wang,Yan-Yan Neale, Marija Zeremski and Dalia Cohen Transcription regulation by histone deacetylases 238 Discussion 245 Peter Atadja, Meier Hsu, Paul Kwon, NancyTrogani, Kapil Bhalla and Stacy Remiszewski Molecular and cellular basis for the anti-proliferative e¡ects of the HDAC inhibitor LAQ824 249 Discussion 266

CONTENTS

vii

Paul A. Marks,Victoria M. Richon,Wm Kevin Kelly, Judy H. Chiao and Thomas Miller Histone deacetylase inhibitors: development as cancer therapy 269 Discussion 281 General discussion III Index of contributors Subject index

291

PML-RARa hypermethylation in leukaemia 289

285

Participants David Allis Rockefeller University, PO Box 78, 1230 York Avenue, NewYork, NY 10021, USA Peter Atadja USA

Novartis Corporation, 556 Morris Avenue, Summit, NJ 07901,

Stephen Baylin Cancer Biology Division,The Sidney Kimmel Comprehensive Center atJohns Hopkins, Bunting Blaustein Cancer Research Building, 1650 Orleans Street, Suite 522, Baltimore, MD 21230, USA Shelley Berger TheWistar Institute, Gene Expression and Regulation Program, Rm 389, 3601 Spruce Street, Philadelphia, PA 19104, USA Vincent Castronovo General and Cellular Biology, Center for Experimental Cancer Research, Metastases Research Laboratory,Tour de Pathologie1, Bat B23, University of Lie' ge, 4000 Sart Tilman, Lie' ge, Belgium Lin-Feng Chen Gladstone Institute of Virology and Immunology, UCSF, PO Box 419100, San Francisco, CA 94141-9100, USA Dalia Cohen Novartis Institute for Biomedical Research, 100 Technology Square, Cambridge, MA 02139, USA Philip Cole Department of Pharmacology, School of Medicine, Johns Hopkins University, 725 N Wolfe Street, 316 Hunterian, Baltimore, MD 21205, USA John Denu Department of Biochemistry and Molecular Biology, Oregon Health and Science University, 3181 SW SamJackson Park Road, Portland, OR 97239-3098, USA Warner C. Greene Gladstone Institute of Virology and Immunology, 365 Vermont Street, University of California at San Francisco, San Francisco, CA 94103, USA viii

PARTICIPANTS

ix

Wei Gu Institute for Cancer Genetics, Columbia University Health Sciences Division, RBP 412C, 1150 St. Nicholas Avenue, NewYork, NY 10032, USA Michael Hottiger University of Zurich, Institute of Veterinary Biochemistry and Molecular Biology,Winterthurerstrasse 190, CH-8057 Zurich, Switzerland Thomas Jenuwein Research Institute of Molecular Pathology,TheVienna Biocenter, Dr Bohrgasse 7, A-1030 Vienna, Austria Saadi Khochbin Laboratoire de Biologie Mole¤ culaire du Cycle Cellulaire, INSERM U 309, Institut Albert Bonniot, Faculte¤ de Me¤ decine, Domaine de la Merci, F-38706 LaTronche Cedex, France En Li Harvard Business School, MGH GeneTargeting Core, Cardiovascular Research Center, 149 13th Street, Charlestown, MA 02129, USA Louis C. Mahadevan Nuclear Signalling Laboratory, Department of Biochemistry, University of Oxford, South Parks Road, Oxford OX1 3QU, UK Paul Marks Memorial Sloan-Kettering Cancer Centre, 1275 York Avenue, NewYork, NY 10021, USA Ronen Marmorstein The Wistar Institute, Structural Biology Programme, 3601 Spruce Street, Philadelphia, PA 19104, USA Danesh Moazed Department of Cell Biology, Harvard Medical School, 240 Longwood Avenue, Boston, MA 02115, USA Marjorie Oettinger Department of Molecular Biology,Wellman 10, Massachusetts General Hospital, Boston, MA 02114, USA Eric Olson Department of Molecular Biology, University of Texas, Southwestern Medical Center at Dallas, 6000 Harry Hines Boulevard, Dallas, TX 75390-9148, USA Melanie Ott Gladstone Institute of Virology and Immunology, 365 Vermont Street, University of California at San Francisco, San Francisco, CA 94103, USA Pier Guiseppe Pelicci European Institute of Oncology, Department of Experimental Oncology,Via G Ripamonti 435, I-20141 Milan, Italy

x

PARTICIPANTS

Edward Seto H. Lee Mo⁄tt Cancer Center & Research Institute, University of South Florida,Tampa, Florida 33612, USA BryanTurner Department of Anatomy,The Medical School, University of Birmingham, Birmingham B15 2TT, UK EricVerdin (Chair) Gladstone Institute of Virology and Immunology, 365 Vermont Street, University of California San Francisco, San Francisco, CA 94103, USA Tso-Pang Yao Duke University Medical Center, Department of Pharmacology and Cancer Biology, C325 LSRC/ Box 3813, Durham, NC 27710, USA Ming-Ming Zhou Department of Physiology and Biophysics, Mount Sinai School of Medicine of NY University, 1 Gustave Levy Place, Box 1677, NewYork, NY 10029, USA

Chair’s introduction Eric Verdin Gladstone Institute of Virology and Immunology and Department of Medicine, University of California San Francisco, PO Box 419100, San Francisco, CA 94103, USA

The human genome was published in its complete form a few weeks ago. There is currently a corresponding shift in biology, with the attention turning from genomics to proteomics. This is the ¢rst meeting devoted to protein acetylation, and as such it marks this transition in a timely fashion. We all owe a debt of gratitude to Dalia Cohen who proposed holding this meeting on protein acetylation and to the Novartis Foundation for organizing it. Forty years ago, Vince Allfrey, who had recently discovered the acetylation of histone, concluded a talk with the following statement: ‘The results are fully consistent with a view that acetylation of the histones diminishes their capacity to inhibit RNA synthesis in the nucleus. It remains to be seen whether this hypothesis will withstand further experimental tests, and whether other modi¢cations of histone structure, such as methylation, can also in£uence the speci¢city of the regulatory roles of these basic nuclear proteins’. This remarkable statement proposed a critical role for histone in transcriptional regulation. How was this model received? A recent search on Pubmed for publications containing the words ‘protein’ and ‘acetylation’ showed an average of 10 papers per year published between 1964 and 1996. The role of histone acetylation in transcriptional regulation remained controversial, despite a steady £ux of circumstantial supporting evidence. In 1996, two major papers reported the identi¢cation of the ¢rst acetyltransferase, GCN5, by David Allis and his group (Brownwell et al 1996) and the ¢rst histone deacetylase, HDAC1, by Stuart Schreiber and colleagues (Taunton et al 1996). The realization that these enzymes were homologous to previously identi¢ed yeast transcriptional regulators ¢nally placed this ¢eld on solid footing. The e¡ect was dramatic. In the past 5 years, more than 2000 papers have been published on protein acetylation. Many of these papers have focused on the identi¢cation and characterization of novel deacetylases and acetyltransferases, and their substrates. Acetylation is now recognized as a critical component of the transcriptional regulatory apparatus. Acetylation has also been detected in non-histone proteins suggesting a role outside of transcription as well. I suggest that we dedicate this meeting to the memory of Vince Allfrey in recognition of his profound impact on our ¢eld of research. 1

2

VERDIN

While this is an impressive achievement, more than 12 000 papers are published on protein phosphorylation each year. Is phosphorylation really a much more prevalent modi¢cation then acetylation? If yes, why? Nothing at this point allows us to predict whether acetylation will remain a relatively rare modi¢cation or whether it may one day rival phosphorylation in scope and importance. The purpose of this meeting will be to review what has been accomplished in recent years and to look into the future of protein acetylation. The meeting has been divided into a number of themes. We will be reviewing the role of histone acetylation and the histone code. We will discuss the enzymes involved in acetylation: the histone acetyltransferases and the deacetylases and their biology. We will hear recent advances on the acetylation of non-histone proteins, a subject that is likely to grow in coming years. Finally, despite the fact that this ¢eld is still in its infancy, e¡ective HDAC inhibitors have been developed and we will hear about their promise as anticancer agents. To ¢nish this introduction, I will outline brie£y some of the key questions that we should be considering in our discussions. The extent of reversible protein acetylation in cells will de¢ne the size of this ¢eld of research. Every month sees the identi¢cation of a new acetylated protein. So far most of them have been transcriptional regulators. How widespread is acetylation? Should the acetylation of the proteome be explored, and if yes, how should we go about this? Another question is how acetylation modi¢es protein function? Is it via changing protein^ protein interactions, inducing conformational changes or via coupling with other post-translational modi¢cations, such as methylation or ubiquitination? We will also be looking at histone acetylation and the histone code. Is there a histone code? If yes, what are its predictive rules? What is the role of global chromatin acetylation versus local acetylation? Are there di¡erences in stability or plasticity of the distinct histone modi¢cations? How is this code being read? The discovery of the bromodomain and chromodomain are two elegant examples, but there are likely to be additional proteins interpreting the code. Another major question relates to the relationship between histone modi¢cations and the molecular mark inherited during duplication of the genome, the epigenetic chromatin state? This is the transcriptional memory that forms the basis of organism cloning and imprinting. This major question remains unanswered. There will undoubtedly be many other questions. All invited participants to this meeting were selected for their broad experience in transcriptional regulation and protein modi¢cation. I look forward to an interactive and productive meeting. References Brownell JE, Zhou J, Ranalli T et al 1996 Tetrahymena histone acetyltransferase A: a homolog to yeast Gcn5p linking histone acetylation to gene activation. Cell 84:843^851 Taunton J, Hassig CA, Schreiber SL 1996 A mammalian histone deacetylase related to the yeast transcriptional regulator Rpd3p. Science 272:408^411

Beyond the double helix: writing and reading the histone code Yanming Wang, Wolfgang Fischle, Wang Cheung*, Steven Jacobs*, Sepideh Khorasanizadeh* and C. David Allis1 Laboratory of Chromatin Biology, The Rockefeller University, New York, New York 10021 and *Department of Biochemistry and Molecular Genetics, University of Virginia Health System, Charlottesville, Virginia 22908, USA

Abstract. Chromatin is the physiological carrier of not only genetic information, encoded in the DNA, but also of epigenetic information including DNA methylation and histone modi¢cations. As such histone modi¢cations are involved in many aspects of nuclear processes including gene regulation and chromosome segregation. Recently, a ‘histone code’ hypothesis was put forward to explain how patterns of histone modi¢cation may function in downstream processes. In support of the ‘histone code’ hypothesis, we found in vivo and in vitro evidence that e¡ector proteins, HP1 (heterochromatin protein 1) and Pc (Polycomb) can discriminate and ‘read’ histone methylation marks on K9 and K27, respectively. Moreover, we propose a ‘binary switch’ model and suggest that binding and release of e¡ector proteins to their cognate sites can be regulated by modi¢cations on adjacent or nearby residues. Thus, combinations of adjacent histone modi¢cations would function di¡erently from singular modi¢cation, and static modi¢cations (e.g. Lys methylation) may well be regulated by dynamic modi¢cations (e.g. phosphorylation). Finally, we describe a novel histone phosphorylation event linking the function of Mst1 kinase and H2B Ser14 phosphorylation with apoptotic chromatin condensation in vertebrates. As this modi¢cation is not found during mitotic chromosome condensation, these ¢ndings suggest the intriguing possibility that a unique ‘death’ mark exists for chromatin condensation during apoptosis. 2004 Reversible protein acetylation. Wiley, Chichester (Novartis Foundation Symposium 259) p 3^21

The human genome is estimated to contain 30 000^40 000 unique genes. The DNA sequence and the chromatic location of most of these genes has been determined and they are publicly available (Lander et al 2001, Venter et al 2001). The central challenge now facing the biomedical community is how to derive valuable medical knowledge about the function of these genes from DNA sequence data, and to 1This paper was presented at the symposium by David Allis, to whom correspondence should be

addressed. 3

4

WANG ET AL

THE HISTONE CODE

5

answer questions such as how the expression of these genes is orchestrated to carry out normal cellular functions and for responses to environmental and physiological changes. The genetic information encoded by the DNA sequence determines the sequence of RNAs and proteins. However, it is becoming clear that ‘epigenetic’ information plays a major role in determining when, where, and to what level the genetic information should be utilized. Epigenetic information refers to di¡erential and inheritable changes of gene expression potentials that are not caused by mutations in DNA itself (Jaenisch & Bird 2003). Recent studies have focused on several molecular mechanisms of epigenetic gene regulation that include DNA methylation, histone modi¢cations and small nuclear RNAs (SnRNAs) or RNA interference (RNAi). Whereas emerging evidence suggests that the three mechanisms are coordinated and a¡ect each other, the discussion below will mainly focus on histone modi¢cations as they relate to epigeneticbased forms of gene regulation. Histone proteins as ‘messengers’ of epigenetic information The human body contains multiple organs and diverse cell types, and every gene exists within every cell. However, only a small percentage of genes are activated in any given cell type, and each type of cell has its unique gene expression pro¢le. These di¡erent expression pro¢les are formulated during early development in a multicellular organism, when cell division, cell di¡erentiation, tissue and organ formation occur rapidly (Francis & Kingston 2001). Moreover, these gene expression potentials can be memorized and inherited after mitosis and even meiosis. To regulate this genetic information e⁄ciently and in an epigenetic manner, nature has evolved a sophisticated system that controls access to speci¢c genes. This system relies on packaging DNA into a DNA^histone complex called chromatin, which is the physiological substrate of all cellular processes involving the DNA (Felsenfeld & Groudine 2003). The dynamic change of the three dimensional architecture of chromatin makes certain genes more readily accessible to transcription factors and other machineries engaging the genetic template (Lomvardas & Thanos 2002). Because parental DNA and associated FIG. 1. (A) Genetic information on DNA is inherited by daughter nuclei through a semiconservative DNA replication process. Therefore, covalent modi¢cations carried by the DNA strand, such as the well-known DNA methylation, will be passed on to the daughter cells and copied to the newly synthesized DNA strands by the maintenance DNA methyltransferase. (B) Similarly, since nucleosomes are randomly separated to form replicated chromatin polymers packaging two duplicated DNA double helices, the location of histone modi¢cations (e.g. methylation) on chromatin can be inherited. The spreading of local histone modi¢cation by recruited histone modifying enzymes can allow a similar density of local modi¢cation after each cell cycle.

6

WANG ET AL

histones are divided and incorporated into the newly duplicated chromatin during S phase of the cell cycle, it is possible that the epigenetic modi¢cations carried by DNA and histones can be passed to the daughter cells after M phase and cell division, making DNA and histone proteins also the attractive messengers of epigenetic information (Fig. 1). Recently, the chromatin ¢eld has witnessed an explosion of literature documenting the involvement of various histone modi¢cations, such as methylation, phosphorylation, acetylation, ubiquitination and ADP ribosylation in essentially all DNA-templated processes. In the coming of a new epigenomic era, the regulation of the enzymes responsible for adding or subtracting these covalent marks are poised to take centre stage in the study of gene expression regulation, and understanding the molecular aetiology of human diseases such as cancer. Identi¢cation of altered DNA methylation and histone modi¢cation activities in a range of human cancers supports the involvement of epigenetic mechanisms during cancer development (Kondo et al 2003). Thus, it is important to investigate the role of epigenetic regulatory proteins and the way that epigenetic regulation works in order to get a more in-depth picture of pathways leading to oncogenesis and to assist the development of new therapeutic strategies. New insights into the ‘histone code’ hypothesis It is clear that the regulatory signals, either extracellular or intracellular, ultimately impinge on chromatin, which can be viewed as a gigantic signalling platform for integrating and recording these signalling events (Cheung et al 2000). The epigenetic information carried by the chromatin can in turn impact on most of the chromatin-templated processes with far-reaching consequences for cell fate decisions and for normal and pathological development (Jenuwein & Allis 2001, Fischle et al 2003). As mentioned above, epigenetic information is inheritable through the cell cycle and through meiosis from one sexual generation to the next. We and others have proposed that an epigenetic indexing system for our genome, a ‘histone’ or ‘epigenetic’ code, works as a fundamental regulatory mechanism in addition to the DNA and the genetic information itself (Strahl & Allis 2000, Turner 2000, Jenuwein & Allis 2001). The original histone code hypothesis proposed that ‘distinct covalent histone modi¢cations, acting alone, sequentially, or in combination, form a ‘‘histone code’’ that is then read by e¡ector proteins to bring about distinct downstream events’ (Strahl & Allis 2000). Although this hypothesis has received much attention and some strong experimental support (Agalioti et al 2002, Kanno et al 2004), it has been hard to derive de¢nitive rules from our current knowledge of the ‘code’. In this meeting, I will expand on this general concept by proposing the ‘methyl/ phos’ (methylation/phosphorylation) switch hypothesis with ‘predictive rules’

THE HISTONE CODE

7

FIG. 2. (A) High density of modi¢cations on the histone H3 tail. Shown here are the Nterminal 30 residues of the histone H3 tail, with at least four TK or KS sites (T3K4, K9S10, T22K23, and K27S28) that may be dually modi¢ed. (B) In this proposed ‘binary switch’ model, two adjacent residues can each exist as unmodi¢ed (0) state or modi¢ed (1) state. When the binary code is 00, there is no modi¢cation and no e¡ector protein binding. After one of these sites is modi¢ed, the binary code is now either 01 or 10. We propose that as these states are produced, either e¡ector protein 1, recognizing state 01, or e¡ector protein 2, recognizing state 10, will be recruited. However, if the two residues are simultaneously modi¢ed, the binding of either e¡ector protein is dramatically reduced leading to release of the e¡ector proteins (see text for details).

that may govern the binding and release of e¡ector proteins and complexes that engage the chromatin polymer. On the histone H3 tail, several clear examples of adjacent Lys residues and Ser/Thr residues exist, such as K9S10 and K27S28, that can be modi¢ed by methylation and phosphorylation, respectively (Fig. 2A). I will present recent work suggesting that methylation- and site-speci¢c e¡ector proteins exist, and their function is likely regulated by phosphorylation of adjacent residues (see below, Jacobs & Khorasanizadeh 2002, Fischle et al 2003a). Thus, these adjacent KT/S sites may form ‘binary switches’ to regulate the binding of e¡ector proteins (Fig. 2B and discussion below). Importantly, our ideas provide an explanation for several long-standing questions embedded in the existing literature, and are open to experimental tests. In addition, I will present a newly

8

WANG ET AL

discovered histone phosphorylation event and the responsible kinase, which link histone modi¢cation to apoptotic chromosome condensation.

Molecular basis for discrimination of repressive methyl-lysines in the histone H3 tail On the histone H3 tail, lysines 9 and 27 are well-known methylation sites, and are often associated with epigenetic repression (Lys 27) and heterochromatinmediated gene silencing (Lys 9) (Cao et al 2002, Jacobs et al 2001). Although these two sites are involved in di¡erent epigenetic events, it is remarkable that both ‘target’ lysines are embedded within a highly related sequence motif: TARK9S versus AARK27S (Fig. 2A). Moreover, as predicted by the histone code hypothesis, emerging evidence shows that Lys9 and Lys27 methylation sites are ‘read’ by distinct e¡ector binding proteins: heterochromatin protein 1 (HP1) and Polycomb (Pc), respectively. Both HP1 and Pc are the prototype proteins in which the chromodomain was identi¢ed (Singh et al 1991). Recent work suggests that chromodomains serve as methyl-lysine recognition and binding modules (Jacobs & Khorasanizadeh 2002, Nielsen et al 2002). Our knowledge of the organization and function of heterochromatin has been greatly advanced by the study of HP1 and the histone H3 K9 methyltransferase, Su(var)3-9. The fact that the chromodomain of HP1 can recognize and bind the K9 methyl site generated by the Su(Var)3-9 o¡ers a mechanistic insight on the epigenetic gene silencing phenomena associated with heterochromatin, such as position e¡ect variegation (PEV) (Lachner et al 2001, Bannister et al 2001, Jacobs et al 2001, Rea et al 2000). On the epigenetic gene expression side, recent studies found that the E(z) (Enhancer of Zeste) complex can methylate histone H3 K27, and H3 K9 to a lesser extent, in in vitro enzymatic assays (Cao et al 2002, Czermin et al 2002, Kuzmichev et al 2002, Muller et al 2002). Likewise, the chromodomain of Pc was proposed to be able to ‘read’ both of these methylation sites. However, these promiscuous activities are paradoxical to the in in vivo observation that Pc and HP1 are involved in di¡erent pathways, and that the chromodomain of HP1 and Pc is responsible to target these protein to di¡erent destination in the nucleus (Messmer et al 1992, Platero et al 1995). Here, we present new data to show that the chromodomain proteins Pc and HP1 are highly discriminatory for binding to these sites both in vivo and in vitro. Using newly developed methyl- and site-speci¢c antibodies, we showed that trimethylLys27 and Pc are colocalized and both excluded from heterochromatic areas that are enriched in di- and trimethyl-Lys9 and HP1 in Drosophila S2 cells and on polytene chromosomes. In addition, swapping of the chromodomain regions of Pc and HP1 is su⁄cient for switching the nuclear localization patterns of these

THE HISTONE CODE

9

repressors, indicating a role for their chromodomains in both target site binding and discrimination (Fischle et al 2003b). To better understand the molecular basis for the selection of methyl-Lys binding sites, we have recently solved the 1.8 — structure of the Pc chromodomain in complex with a trimethyl-Lys27 H3 tail and compared it with our previously determined structure of the HP1 chromodomain complexed with a trimethylLys9 H3 tail (Jacobs & Khorasanizadeh 2002). The structures show clear di¡erences in how two chromodomains that are highly related in sequence and structure e¡ectively distinguish methylation sites on the H3 tail (Fischle et al 2003). Whereas both the HP1 and the Pc chromodomains form aromatic ‘cages’ that bind the positively charged methylammonium ion, they distinguish the two binding sites by discriminating residues N-terminal to the common ARKS motif, which di¡er between the two target sites. The Pc chromodomain has evolved an extended ‘groove’ that provides more contact surfaces to engage ¢ve more residues to a modest degree in addition to the ARK27S. Together, these surfaces provide enough additional binding complimentarily to generate enhanced recognition and binding a⁄nity (Fischle et al 2003b). On the HP1 side, it seems the residue T6 in front of ARK9S is critical for the binding speci¢city (Jacobs & Khorasanizadeh 2002, Fischle et al 2003). HP1 and Pc proteins themselves have been implicated in fundamental nuclear processes including heterochromatin-mediated gene silencing, homeotic gene expression and chromosome dynamics (Simon & Tamkun 2002). The above studies of HP1 and Pc o¡er supportive evidence to a central tenet of the ‘histone code hypothesis’, that the covalent marks are docking sites for e¡ector proteins that in turn bring about distinct downstream events (Strahl & Allis 2000). It is quite intriguing from an evolutionary aspect that the two e¡ector proteins with the similar functional domain can recognize two binding sites embedded in the similar sequence context, and are evolved to participate in two di¡erent silencing pathways. Binary switches as part of the histone code? The density of modi¢able residues on the histone tail, for example H3, is very striking (Fig. 2A). Recently, mass spectrometry analyses suggest that modi¢cation of two adjacent sites does coexist (C. D. Allis, D. F. Hunt, unpublished data). Many site- and modi¢cation-speci¢c antibodies have been developed and have greatly bene¢ted the ¢eld to tackle the histone modi¢cation problem. However, as many immunological tools were developed against speci¢c histone modi¢cation sites, a recurring question to us and others is whether adjacent histone modi¢cations might a¡ect the epitope recognition by antibodies. Similarly, it is equally intriguing to know whether adjacent modi¢cations, if they

10

WANG ET AL

exist, may a¡ect the binding of e¡ector proteins that normally recognize single modi¢cations, such as HP1 and Pc. On the basis of the presence of close dual modi¢cation sites on the histone tails, we wish to extend the histone code hypothesis and to propose the concept of ‘binary switches’ (Fig. 2B). We hypothesize that binary switches in the histone tails regulate the ‘ON/OFF’ state for the binding of e¡ector proteins, such as HP1 and Pc. Speci¢cally, a ‘phos/methyl’ or ‘methyl/phos’ switch likely operates on the histone tails to regulate e¡ector binding at the correct time in the cell cycle or appropriate stage of development. The central tenet of this hypothesis is as follows: ‘on^o¡’ binding of e¡ectors is ¢rstly regulated by adding chemical moieties, such as the methyl groups, to their cognate site; secondly, the addition and subtraction of modi¢cation at the nearby or adjacent site can release and recruit binding e¡ectors, respectively, without changing the primary modi¢cation site. For example, it is conceivable that the K9S10 sites might form such a binary switch. Structural analyses of the chromodomain of HP1 bound to the H3 tail methylated on Lys9 (Jacobs & Khorasanizadeh 2002), argue whether mitosis-driven phosphorylation of Ser10 will signi¢cantly diminish the binding a⁄nity of this module. Mitotic phosphorylation of H3 at Ser10 and/or Ser28, catalysed by aurora Btype kinases, is well documented in organisms ranging from yeast to humans (Hsu et al 2000). If the general concept is correct, mitosis (or meiosis) may drive the phosphorylation side of the phos/methyl switch allowing for the release, and potential clearing, of chromatin e¡ectors (e.g. HP1) that dock on stable methylation marks during interphase. In support, a large portion of HP1 and Pc protein is released from the mitotic chromosome in early embryos and in the Drosophila S2 cells (Dietzel et al 1999, Kellum et al 1995 and Y. Wang, C. D. Allis, unpublished results). For the proposed binary phos/methyl switches to work, we predict that methyl-speci¢c chromodomains would change their binding a⁄nities to methyl target sites with adjacent phosphorylation marks. New binding data with the chromodomain of HP1 show that this is indeed correct; the binding a⁄nity is dramatically decreased when the chromodomain of HP1 was tested with a H3 peptide that is both tri-methylated at Lys9 and phosphorylated at Ser10 (S. A. Jacobs, W. Fischle, unpublished results). It is intriguing to consider to what range this concept might be applied. To our knowledge, a binding partner has yet to be identi¢ed that ‘reads’ the H3 (Lys4) methyl mark, a mark often associated with an ‘on’ or ‘competent’ transcriptional state (Santos-Rosa et al 2002). If such a protein exists, we would predict that its binding to the H3 Lys4 methylation site will be regulated by phosphorylation at Thr3, a mark which has recently been found to be a strong mitotic phosphorylation site in mammalian cells (C. Barber, F. Turner, and D. Allis, unpublished data). Similar arguments can be made for T22K23, although to our knowledge, it is not

THE HISTONE CODE

11

yet known whether Thr22 and Lys23 in the H3 tail are phosphorylation and methylation sites, respectively. On the other side, it is intriguing to postulate that the acetyl-Lys binding protein of the K23 site may be regulated by an ‘acetyl-phos’ switch. Nevertheless, we suggest that there may be at least four ‘binary switches’ operating on the H3 tail alone (Fig. 2A). In turn, we predict that the interaction of phos-binding e¡ectors (yet to be found) and their cognate sites may be regulated by nearby or adjacent ‘o¡’ methyl switches. Histone methylation on the Lys residues is relatively stable epigenetic mark, and no histone (Lys) demethylases have been identi¢ed so far. However, phosphorylation can be reversibly regulated by kinase/phosphatase. We note that the Ipl1/aurora kinase and type 1 protein phosphatase (PP1ase) have been identi¢ed as the mitotic kinase/phosphatase responsible for regulating H3 Ser10 phosphorylation as cells enter/exit mitosis (Hsu et al 2000). Interestingly, PP1ase was also identi¢ed in the same genetic screens in Drosophila as Su(Var)3-6 (Baksa et al 1993), suggesting that its activity facilitates silencing by unknown mechanisms. However, it is tempting to think about the suppressor function of PP1 in the context of the ‘methyl/phos’ binary switch. The switch model makes a clear and testable prediction regarding the role of Su(Var)3-6 in the above silencing pathway: one role, if not the major role, of PP1ase, Su(Var)3-6, is to remove phosphates at Ser10 on H3 at the end of mitosis. Thus, the released and dispersed HP1 (also identi¢ed in the same genetic screen as Su[Var]2-5) in the M phase cells can be recruited back to H3 Lys9 methylation sites, which are themselves added by a H3 (Lys9) methyltransferase, Su(Var)3-9. In summary, this model and its predictions provide new insights into a potential role for protein phosphatases, kinases, methyltransferases, and potentially histone demethylases in regulating the binding and release of critical e¡ector proteins. More importantly, since histone demethylation activity has not been discovered to date, to regulate histone methylation and the associated methyl binding protein by reversible phosphorylation is a means to dynamically regulate histone methylation.

A ‘life’ versus ‘death’ histone code? Histone phosphorylation is one of the best-characterized histone modi¢cations. The function of histone phosphorylation has been linked with many aspects of chromatin biology, including mitotic chromosome condensation, gene expression, dosage compensation in Drosophila, DNA double-strand breakage and repair (Fig. 3A). The linker histone H1 and histone H3 phosphorylation are well documented and are linked to both transcription regulation and mitotic chromosome condensation (Dou et al 2002, Hsu et al 2000). Much less well

12

WANG ET AL

FIG. 3. (A) The functions of histone phosphorylation (f P)in various cellular processes. (B) The N-terminal tail of H2B from several model organisms (hs, Homo sapiens; mm, Mus musculus; xl, Xenopus laevis; dm, Drosophila melanogaster). The S14 phosphorylation site is conserved in species from Xenopus to mammals, which is highlighted. Whether a functionally similar phosphorylation site(s) exists in invertebrates and lower organisms is unclear (see text for details).

characterized are molecular events that in£uence the remarkable changes in chromatin condensation that characterize dying cells. During apoptosis chromatin is digested into oligonucleosomal fragments and is condensed to form pycnotic chromatin bodies, two hallmark properties of this process in most cells (Wyllie 1980). Because of the intimate association between histones and DNA, histone phosphorylation was suggested to be involved in the change of chromatin integrity and compaction (Ajiro 2000). However, how histone phosphorylation is induced and involved in apoptosis remains poorly understood (Cheung et al 2000). Phosphorylation at the C-terminal tail of a relatively minor histone variant, H2A.X (at serine 139 in human), increases during early stages of DNA fragmentation in apoptosis (Redon et al 2002). However, H2A.X phosphorylation correlates with all known double-stranded DNA breaks suggesting that it acts more as a ‘DNA-damage sensor’ than a speci¢c chromatin mark linked to the apoptotic process. To further investigate the relationship between histone phosphorylation and mitotic and/or apoptotic chromatin condensation, we have generated a

THE HISTONE CODE

13

novel phos-speci¢c antibody against Ser14 residue on histone H2B (hereafter a-Phos[Ser14]H2B) (commercially available from Upstate Biotech. Inc.; Lake Placid, NY). Using this antibody in several vertebrate and mammalian models, we found that H2B Ser14 phosphorylation speci¢cally correlates with the onset of apoptotic chromatin condensation and DNA fragmentation in human cells (Cheung et al 2003). This correlation was also found in cells undergoing programmed cell death during Xenopus tail resorption. Using in-gel kinase assays, we detect and have identi¢ed an apoptotic-induced H2B (Ser14) kinase with a molecular weight of 34 kDa as the caspase cleaved form of Mst1 (Mammalian Sterile Twenty), which is a well studied kinase activated by multiple apoptotic stimuli (Feig & Buchsbaum 2002, Graves et al 2001). Interestingly, the Ser14 phosphorylation site is only conserved among vertebrates, ranging from frog to human (Fig. 3B), leaving open the intriguing possibility that additional apoptotic phosphorylation sites might exist on other sites of the histone tails of invertebrates. Nevertheless, these studies de¢ne what may be an apoptotic ‘histone code’ conserved among vertebrates, and cast new light on physiological substrates of Mst1 kinase. Mst1 is a member of sterile 20-like superfamily of which approximately 30 related kinases exist in humans (Graves et al 2001). Kinases contained in this superfamily are most often regarded as an upstream regulator of MAPK pathways with roles in cellular morphogenesis and cytoskeletal rearrangements, as well as apoptotic cell death (Feig & Buchsbaum 2002). Our ¢nding that the cleaved-form of Mst1 is likely a nuclear-bound kinase directly responsible for H2B (Ser14) phosphorylation, at least in higher eukaryotic cells under some inducing conditions, might shed new light on it as a potential drug target. Chromatin condensation and DNA fragmentation have been viewed as the last committed step of apoptosis. Considerable evidence exists suggesting that many cells die under stress by undergoing apoptosis (Wyllie 1980). However, using caspase inhibitors has not been very e¡ective to decrease cell death after the initial stress, such as ischaemia, has occurred (Natori et al 2003). Perhaps, after e¡ector caspases initiate the death pathway leading to de¢ned chromatin changes, caspases are no longer needed. It is possible that e¡ective prevention of cell death may be best brought about by combining caspase inhibitors with drugs that target downstream activities such as are caused by Mst1 to prevent chromatin changes during apoptosis. Unlike the H3 Ser10 and Ser28 phosphorylation, the Phos (Ser14) H2B mark is not detected in mitotic chromosomes, at least in the cell types that we have examined. Thus, we are intrigued with the possibility that there may be a nonoverlapping set of phosphorylation marks that discriminate ‘mitotic’ from ‘apoptotic’ chromatin. It is becoming well established that bromodomains ‘read’ acetyl-lysine mark on histones (and likely non-histone proteins), and

14

WANG ET AL

chromodomains ‘read’ methyl-lysine marks in a sequence context-dependent fashion (see above, Jenuwein & Allis 2001). However, it is not clear whether there are e¡ector proteins docking on phosphorylated histone tails peptides, which is currently pursued in my laboratory (C. Barber and C. D. Allis, work in progress). Epigenomics and human diseases It has been widely accepted that DNA methylation and histone modi¢cations serve as two major mechanisms for the function and inheritance of epigenetic information (Jaenisch & Bird 2003). Recent advance has suggested that DNA methylation and histone methylation are correlated in Neurospora crassa, plant and mammals (Tamaru et al 2003, Soppe et al 2002, Fuks et al 2003). Importantly, histone deacetylation, histone methylation, and DNA methylation are involved in the aberrant silencing of certain tumour suppressor genes in tumour cells (Bachman et al 2003, Kondo et al 2003). In the case of the p16INK4a tumour suppressor gene, Bachman et al (2003) have found that histone Lys9 methylation precedes DNA methylation. The promising result of applying HDAC inhibitors in the treatment of leukaemia is a harbinger for cancer treatment by interfering with epigenetic histone modi¢cations. Collectively, these exciting developments make a compelling argument for investments in developing new therapies centered in attacking epigenetic forms of gene regulation. We predict that chromatin modi¢cations will revolutionize our view of cancer as new mechanisms of ‘epigenetic’ carcinogenesis are discovered. Given the emerging link between histone modi¢cations and DNA methylation, it is conceivable that histone modi¢cations might be involved in the multiple diseases caused by epigenetic disorder, including deregulation of imprinted genes. In addition to the direct control of tumour suppressor gene expression, members of the Pc and HP1 family have been suggested to play a role in the cell proliferative capacity (Lessard & Sauvageau 2003, Varambally et al 2002, Kirschmann et al 2000). Furthermore, it remains to be discerned whether histone modi¢cations are involved in various genomic imprinting disorders, such as Beckwith-Wiedemann, Angelman, and Prader-Willi Syndromes (Wol¡e & Matzke 1999). Conclusions and perspective Chromatin is the physiological template of our genetic information. Well known is the understanding that this polymer is subject to a diverse array of posttranslational modi¢cations that largely impinge on histone N-termini, thereby

THE HISTONE CODE

15

regulating access to the underlying DNA. The combinatorial nature of histone Nterminal modi¢cations thus reveals a ‘histone code’ that signi¢cantly extends the information potential of our genetic code. As is well documented in the literature and in the meeting itself, this covalent modi¢cation-based histone code may well exist in non-histone proteins, suggesting a more universally applied protein code. Current evidence suggests that it is a fundamental regulatory mechanism that impacts on most, if not all, chromatin-templated processes with far-reaching consequences for cell fate decisions, and normal and pathological development.

References Agalioti T, Chen G, Thanos D 2002 Deciphering the transcriptional histone acetylation code for a human gene. Cell 111:381^192 Ajiro K 2000 Histone H2B phosphorylation in mammalian apoptotic cells. An association with DNA fragmentation. J Biol Chem 275:439^443 Bachman KE, Park BH, Rhee I et al 2003 Histone modi¢cations and silencing prior to DNA methylation of a tumor suppressor gene. Cancer Cell 3:89^95 Baksa K, Morawietz H, Dombradi V et al 1993 Mutations in the protein phosphatase 1 gene at 87B can di¡erentially a¡ect suppression of position-e¡ect variegation and mitosis in Drosophila melanogaster. Genetics 135:117^125 Bannister AJ, Zegerman P, Partridge JF et al 2001 Selective recognition of methylated lysine 9 on histone H3 by the HP1 chromo domain. Nature 410:120^124 Cao R, Wang L, Wang H et al 2002 Role of histone H3 lysine 27 methylation in Polycomb-group silencing. Science 298:1039^1043 Cheung P, Allis CD, Sassone-Corsi P 2000 Signaling to chromatin through histone modi¢cations. Cell 103:263^271 Cheung WL, Ajiro K, Samejima K et al 2003 Apoptotic phosphorylation of histone H2B is mediated by mammalian sterile twenty kinase. Cell 113:507^517 Czermin B, Mel¢ R, McCabe D, Seitz V, Imhof A, Pirrotta V 2002 Drosophila enhancer of Zeste/ ESC complexes have a histone H3 methyltransferase activity that marks chromosomal Polycomb sites. Cell 111:185^196 Dietzel S, Niemann H, Bruckner B, Maurange C, Paro R 1999 The nuclear distribution of Polycomb during Drosophila melanogaster development shown with a GFP fusion protein. Chromosoma 108:83^94 Dou Y, Bowen J, Liu Y, Gorovsky MA 2002 Phosphorylation and an ATP-dependent process increase the dynamic exchange of H1 in chromatin. J Cell Biol 158:1161^1170 Feig LA, Buchsbaum R J 2002 Cell signaling: life or death decisions of ras proteins. Curr Biol 12:R259^R261 Felsenfeld G, Groudine M 2003 Controlling the double helix. Nature 421:448^453 Fischle W, Wang Y, Allis CD 2003a Binary switches and modi¢cation cassettes in histone biology and beyond. Nature 425:475^479 Fischle W, Wang Y, Jacobs SA, Kim Y, Allis CD, Khorasanizadeh S 2003b Molecular basis for the discrimination of repressive methyl-lysine marks in histone H3 by Polycomb and HP1 chromodomains. Genes Dev 17:1870^1881 Francis NJ, Kingston RE 2001 Mechanisms of transcriptional memory. Nat Rev Mol Cell Biol 2:409^421 Fuks F, Hurd PJ, Deplus R, Kouzarides T 2003 The DNA methyltransferases associate with HP1 and the SUV39H1 histone methyltransferase. Nucleic Acids Res 31:2305^2312

16

WANG ET AL

Graves JD, Draves KE, Gotoh Y, Krebs EG, Clark EA 2001 Both phosphorylation and caspasemediated cleavage contribute to regulation of the Ste20-like protein kinase Mst1 during CD95/Fas- induced apoptosis. J Biol Chem 276:14909^14915 Hsu JY, Sun ZW, Li X et al 2000 Mitotic phosphorylation of histone H3 is governed by Ipl1/ aurora kinase and Glc7/PP1 phosphatase in budding yeast and nematodes. Cell 102:279^291 Jacobs SA, Taverna SD, Zhang Y et al 2001 Speci¢city of the HP1 chromo domain for the methylated N-terminus of histone H3. EMBO J 20:5232^5241 Jacobs SA, Khorasanizadeh S 2002 Structure of HP1 chromodomain bound to a lysine 9methylated histone H3 tail. Science 295:2080^2083 Jaenisch R, Bird A 2003 Epigenetic regulation of gene expression: how the genome integrates intrinsic and environmental signals. Nat Genet 33:245^254 Jenuwein T, Allis CD 2001 Translating the histone code. Science 293:1074^1080 Kanno T, Kanno Y, Siegel RM, Jang MK, Lenardo MJ, Ozato K 2004 Selective recognition of acetylated histones by bromodomain proteins visualized in living cells. Mol Cell 13:33^43 Kellum R, Ra¡ JW, Alberts BM 1995 Heterochromatin protein 1 distribution during development and during the cell cycle in Drosophila embryos. J Cell Sci 108:1407^1418 Kirschmann DA, Lininger RA, Gardner LM et al 2000 Down-regulation of HP1Hsalpha expression is associated with the metastatic phenotype in breast cancer. Cancer Res 60: 3359^3363 Kondo Y, Shen L, Issa JP 2003 Critical role of histone methylation in tumor suppressor gene silencing in colorectal cancer. Mol Cell Biol 23:206^215 Kuzmichev A, Nishioka K, Erdjument-Bromage H, Tempst P, Reinberg D 2002 Histone methyltransferase activity associated with a human multiprotein complex containing the Enhancer of Zeste protein. Genes Dev 16:2893^2905 Lachner M, O’Carroll D, Rea S, Mechtler K, Jenuwein T 2001 Methylation of histone H3 lysine 9 creates a binding site for HP1 proteins. Nature 410:116^120 Lander ES, Linton LM, Birren B et al 2001 Initial sequencing and analysis of the human genome. Nature 409:860^921 Lessard J, Sauvageau G 2003 Bmi-1 determines the proliferative capacity of normal and leukaemic stem cells. Nature 423:255^260 Lomvardas S, Thanos D 2002 Modifying gene expression programs by altering core promoter chromatin architecture. Cell 110:261^271 Messmer S, Franke A, Paro R 1992 Analysis of the functional role of the Polycomb chromo domain in Drosophila melanogaster. Genes Dev 6:1241^1254 Muller J, Hart CM, Francis NJ et al 2002 Histone methyltransferase activity of a Drosophila Polycomb group repressor complex. Cell 111:197^208 Natori S, Higuchi H, Contreras P, Gores GJ 2003 The caspase inhibitor IDN-6556 prevents caspase activation and apoptosis in sinusoidal endothelial cells during liver preservation injury. Liver Transpl 9:278^284 Nielsen PR, Nietlispach D, Mott HR et al 2002 Structure of the HP1 chromodomain bound to histone H3 methylated at lysine 9. Nature 416:103^107 Platero JS, Hartnett T, Eissenberg JC 1995 Functional analysis of the chromo domain of HP1. EMBO J 14:3977^3986 Rea S, Eisenhaber F, O’Carroll D et al 2000 Regulation of chromatin structure by site-speci¢c histone H3 methyltransferases. Nature 406:593^599 Redon C, Pilch D, Rogakou E, Sedelnikova O, Newrock K, Bonner W 2002 Histone H2A variants H2AX and H2AZ. Curr Opin Genet Dev 12:162^169 Santos-Rosa H, Schneider R, Bannister AJ et al 2002 Active genes are tri-methylated at K4 of histone H3. Nature 419:407^411 Simon JA, Tamkun JW 2002 Programming o¡ and on states in chromatin: mechanisms of Polycomb and trithorax group complexes. Curr Opin Genet Dev 12:210^218

THE HISTONE CODE

17

Singh PB, Miller JR, Pearce J et al 1991 A sequence motif found in a Drosophila heterochromatin protein is conserved in animals and plants. Nucleic Acids Res 19:789^794 Soppe WJ, Jasencakova Z, Houben A et al 2002 DNA methylation controls histone H3 lysine 9 methylation and heterochromatin assembly in Arabidopsis. EMBO J 21:6549^6559 Strahl BD, Allis CD 2000 The language of covalent histone modi¢cations. Nature 403:41^45 Tamaru H, Zhang X, McMillen D et al 2003 Trimethylated lysine 9 of histone H3 is a mark for DNA methylation in Neurospora crassa. Nat Genet 34:75^29 Turner BM 2000 Histone acetylation and an epigenetic code. Bioessays 22:836^845 Varambally S, Dhanasekaran SM, Zhou M et al 2002 The polycomb group protein EZH2 is involved in progression of prostate cancer. Nature 419:624^629 Venter JC, Adams MD, Myers EW et al 2001 The sequence of the human genome. Science 291:1304^1351 Wol¡e AP, Matzke MA 1999 Epigenetics: regulation through repression. Science 286:481^486 Wyllie AH 1980 Glucocorticoid-induced thymocyte apoptosis is associated with endogenous endonuclease activation. Nature 284:555^556

DISCUSSION Khochbin: Some of the acetyltransferases from the MYST family have a chromodomain, like Tip60 or Mof. Do you have any evidence for the targeting of methylated lysine by these chromodomains? Allis: That’s a great question. Just in case anyone is not following, the comment was that in the histone acetyltransferase (HAT) ¢eld, there is a group of HATs that are known collectively as the ‘MYST’ family. Some members of this family have chromodomains themselves, so it is tempting to think that there might be a methyl mark that might be read. One of the members of this family that you might be thinking about is Mof. This was originally discovered by John Lucchesi and his colleagues at Emory University in Atlanta (Hil¢ker et al 1997). That Mof is a famous HAT because it is well known to be the up-regulator for the £y male X chromosome. It puts on the Lys16 acetyl mark. I didn’t say anything about H4, but one of the conserved lysines in the H4 tail is Lys16. The H4 tail has lysines at K5, K8, K12 and K16. Most of these are separated by runs of glycine, until you get to this K16, where it then goes KRHRK. The K20 is a well known methyl site in histone H4. It might be attractive to have a chromodomain that could dock on K20 methyl near where a HAT would have to go to put on this acetyl site (Nishioka et al 2002). In fact, if you ask me why the H4 tail suddenly goes from lysines separated by Gly to KRHRK, I think this is another hot area for something to be binding. I suspect we will have bromodomains that will be docking on the acetylated form of this lysine. It would be very clever if nature is exploiting this tail a little more, by putting on HATs that are required through binding via chromodomains. If you search for the KRHRK in the available databases, it goes well past histones. These might be small protein modules that have been exploited in histones with their abundant posttranslational modi¢cations.

18

DISCUSSION

Verdin: How big is the chromodomain family? How many chromodomains are there in the whole genome? Allis: There are lots of chromodomains in the human genome. At least 40. I don’t think they will necessarily all be histone-docking motifs. The chromodomain of Mof that might be reading methyl K20, was originally described to be an RNA-binding motif. With the employment of small RNAs and guide RNAs in the nucleus, we should be open to other roles for chromodomains beside histone methylation. Moazed: I have a couple of questions concerning your methyl-phospho switch. First, is it known whether HP1 or homologues such as polycomb dissociate from chromatin during mitosis? Allis: This has been published by others in Drosophila and Schizosaccharomyces pombe. Moazed: How do you propose that the kinase can gain access to the serine, if HP1 is stably bound to the histone tail? Allis: It is hard for me to tell this from crystal structural information. The question would be can the kinase get at its target in the presence of what might be docking here? I think the cell can do it. These are all in giant complexes. New studies suggest that HP1 is ‘on and o¡’ chromatin in a very dynamic way (Festenstein et al 2003, Cheutin et al 2003). I am sure that it will prove to be more than just a recombinant kinase getting in here and doing the job. Moazed: Is it possible that there is a second step: that something else dissociates the complex? Allis: It is possible. This sort of two-step model is possible. All I can tell is that our mass spectrometry studies have no problem in picking up that di-modi¢ed state. How it is achieved, I am not sure. Marmorstein: If it binds to the HP1 chromodomain with some given dissociation constant, there are times when it is o¡, so the kinase could gain access then. Jenuwein: There is another explanation. If you look at the phosphorylation status of the enzyme, the Suv39h HMTase (histone methyltransferase) becomes phosphorylated during mitosis, and along with it it dissociates from mitotic chromatin (Aagaard et al 2000). One could therefore argue that there may be another model. The phosphorylation event of the Suv39h enzyme could attenuate the activity of the HMTase itself. Allis: That is another possibility: the methyltransferase is phosphorylated as part of a reaction that then may do many things we have yet to ¢gure out. Turner: The ideas you have presented are tremendously exciting, but they present us with enormous problems in terms of using the antibody. If we have an antibody to methyl K4, as many of us do, is that still going to bind when P3 is phosphorylated, and vice versa? This means we are in danger of getting a whole bunch of false negative results.

THE HISTONE CODE

19

Allis: We could spend some quality time discussing this. If you have a methyl site, get a great looking antibody, employ it and suggest that there is some very interesting cell cycle or developmental regulation, you could be totally fooled by what you don’t know is going on next door. This density of modi¢cations may almost make the antibodiesas beautiful as they have beenpotentially problematic as tools. Louis Mahadevan has all kinds of evidence that Ser10 phosphorylation in the immediate early mitogen response can work with Lys9 acetylation next door, so antibody recognition here with a Ser10-selective antibody is a concern. We are comforted a bit that we can do mass spec, which gets us away from the use of antibodies alone. Turner: It doesn’t have the £exibility of the antibodies. Louis particularly has attempted to raise antibodies that recognize two motifs. Is this what we are looking at: raising a batch of antibodies for every site? Allis: That is a reasonable way to go. Louis knows better than anyone that if we have a K9 acetyl in conjunction with a neighbouring Ser10 phospho, the dimodi¢ed antibody is a gorgeous reagent for the immediate early mitogen response. It doesn’t seem impossible to then reach for K9 methyl Ser10 phospho, and then play this out over and over again. Mahadevan: I think eventually we will have to raise highly speci¢c antibodies against all these combinations. These problems are accessible to much more speci¢c well characterized antibodies. We should also consider quantitative issues. When we talk about serine 10 phosphorylation, for example in mitotic cells, would you have any idea about the stoichiometry of phosphorylation? When we read about it we hear that it is very highly phosphorylated. Do you think it is completely phosphorylated? Allis: I wouldn’t want to say it is all phosphorylated. When we run acid-urea gels that permit us to separate unmodi¢ed from a mono-modi¢ed protein, the mitotic sample jumps up to a signi¢cant amount of the total. If you got a pretty respectable enrichment for mitotic samples, I’d say that over half is jumping up. Unless you want to say that is mono-acetyl (and I don’t think it is in mitotic samples), I would suggest over 50% of the molecules are phosphorylated. A healthy dose of the H3 picks up the phospho mark in a mitotic culture. Verdin: Given the fact that methylation apparently is not reversible, wouldn’t this be one of the best candidates in terms of maintaining a memory? Allis: I think that is right. Acetylation has beautiful reversibility. Could it be that the decision to methylate was very purposeful because of its permanent chemical nature, whether on DNA or histone. Berger: At one point you mentioned that you think you have e¡ectors that bind to phosphorylated Ser10 on histone H3. But you said that they wouldn’t bind to the histone when it is methylated. For a uni¢ed model, this presents a bit of a problem. If the methyl marks are permanent, then how does the binding come about?

20

DISCUSSION

Allis: If we have just a Ser10 phospho e¡ector (which I called ‘Y’), and we eject this by methylation on the adjacent Lys9or, to be fair, acetylationwhat happens? All I can say is that we tested our candidate Y with the methyl-phos, and it didn’t bind at all. But you are right, if you put on that methyl mark to be the ejector, how do you remove that? It is a problem. Some people will say that it is possible to dilute histones out of chromatin and replace them. Berger: Is there a di¡erence between mono-, di- and trimethylation in terms of that binding? Allis: It is modest. Whether the HP1 on K9 is mono-, di- or trimethylated only has a modest e¡ect. If you generate antibodies of Lys9 that are mono- versus diversus trimethylated, these are like di¡erent beasts. The regulation opportunities that are available to the cell from just two residues are awesome. If you look at the H3 tail I have been telling you about, in my early Tetrahymena postdoc days I found out that the Tetrahymena in one of its nuclei (the silent nucleus) quantitatively cleaves the H3 between residues 6 and 7. The only thing we knew at the time was that the H3 extreme N-terminus was AR. After you cleave o¡ this six residue piece, there is a new AR. For Tetrahymena this would remove the Lys4. This is the positive up-regulator that ties in to activation, especially if it is trimethylated. I think this may be something that Tetrahymena does purposefully in its silent nucleus. In yeast, their chromatin is clipped in the H3 region. We recently sequenced this when we knew that GCN5 liked to hit on H3, and found that there is a new AR utilization. We have now a little glimmer that mammalian cells are doing this too, in constitutive heterochromatin. The trimethylated K9 mark is very much the constitutive heterochromatin. If this is something that Tetrahymena and yeast do, and it might extend to mammalian cells, then all bets are o¡ on how unwanted methyl marks are removed. Do you demethylate them, clip them or replace them? Atadja: Have people really looked for demethylases? Allis: We have looked very hard and come up with nothing. But there is a breath of fresh air: we have made some progress on arginine demethylation. Arginine looks reversible. Li: Does the K9 methylation re-establish every cell cycle if you just focus on one particular gene locus? Allis: I don’t know whether I can answer that. Li: What about Ser10 phosphorylation? Is it directly involved in chromatin condensation? Allis: If we make that mutation in Tetrahymena it causes severe mitotic dysfunction. I’m not sure why this doesn’t happen in yeast, but redundancy could be the explanation. There is a published report that the Ser10Ala mutation in Tetrahymena had a severe mitotic failure of condensation with some chromosome segregation defects (Wei et al 1999).

THE HISTONE CODE

21

References Aagaard L, Schmid M, Warburton P, Jenuwein T 2000 Mitotic phosphorylation of SUV39H1, a novel component of active centromeres, coincides with transient accumulation at mammalian centromeres. J Cell Sci 113:817^829 Cheutin T, McNairn AJ, Jenuwein T, Gilbert DM, Singh PB, Misteli T 2003 Maintenance of stable heterochromatin domains by dynamic HP1 binding. Science 299:721^725 Festenstein R, Pagakis SN, Hiragami K 2003 Modulation of heterochromatin protein 1 dynamics in primary mammalian cells. Science 299:719^721 Hil¢ker A, Hil¢ker-Kleiner D, Pannuti A, Lucchesi JC 1997 mof, a putative acetyl transferase gene related to the Tip60 and MOZ human genes and to the SAS genes of yeast, is required for dosage compensation in Drosophila. EMBO J 16:2054^2060 Nishioka K, Rice JC, Sarma K et al 2002 PR-Set7 is a nucleosome-speci¢c methyltransferase that modi¢es lysine 20 of histone H4 and is associated with silent chromatin. Mol Cell 9:1201^1213 Wei Y, Yu L, Bowen J, Gorovsky MA, Allis CD 1999 Phosphorylation of histone H3 is required for proper chromosome condensation and segregation. Cell 97:99^109

The indexing potential of histone lysine methylation Gunnar Schotta1, Monika Lachner1, Antoine H. F. M. Peters and Thomas Jenuwein2 Research Institute of Molecular Pathology (IMP), The Vienna Biocenter, Dr. Bohrgasse7, A-1030 Vienna, Austria

Abstract. Diverse post-translational modi¢cations of histone N-termini represent an important epigenetic mechanism for the organization of chromatin structure and the regulation of gene activity. Within the last three years, great progress has been made in understanding the functional implications of histone methylation, in particular through the characterization of histone methyltransferases (HMTases) that direct the site-speci¢c methylation of lysine positions in the histone H3 N-terminus. Histone lysine methylation has been linked with pericentric heterochromatin formation, X-inactivation, Polycomb group (Pc-G)-dependent gene repression and epigenetic control of transcription units at euchromatic positions. Together, these regulatory roles have strongly established histone lysine methylation as a central epigenetic modi¢cation for the organization of eukaryotic chromatin. However, they also create a paradox: if histone lysine methylation is present at so many chromatin regions, how can it impart epigenetic information? We provide evidence that di¡erences in distinct methylation states (mono- vs. di- vs. trimethylation) and selective combinations of individually methylated lysine positions can indeed index chromatin regions, resulting in epigenetic landmarks for the partitioning of eukaryotic chromatin. 2004 Reversible protein acetylation. Wiley, Chichester (Novartis Foundation Symposium 259) p 22^47

In eukaryotic cells, DNA is packaged with histones and other chromosomal proteins to form a dynamic polymer called chromatin. The basic repeating unit of chromatin is the nucleosome, containing two copies each of the four core histones H2A, H2B, H3 and H4 wrapped by 147 bp of DNA. The £exible N-termini (‘tails’) of histones are subject to a variety of post-translational modi¢cations, including acetylation, phosphorylation, methylation, ubiquitination and ADP ribosylation 1

These authors made equal contributions to this manuscript. paper was presented at the symposium by Thomas Jenuwein to whom correspondence should be addressed.

2This

22

HISTONE LYSINE METHYLATION

23

(van Holde 1988). The discoveries of enzymes modifying histone N-termini and of chromatin-associated proteins that speci¢cally bind to selective histone modi¢cations gave rise to the ‘histone code’ hypothesis, proposing an epigenetic mechanism that may signi¢cantly extend the information potential of the genetic code (Strahl & Allis 2000, Turner 2000, Jenuwein & Allis 2001). Histone tail modi¢cations can be marks for both active or repressed chromatin and are also interdependent (Rice & Allis 2001, Zhang & Reinberg 2001, Kouzarides 2002). For histone lysine methylation, there are at least ¢ve methylatable positions in the N-termini of H3 (K4, K9, K27, K36) and H4 (K20). In addition, there is one position in the histone fold domain of H3 (K79). Methylation of H3-K4, H3-K36 and H3-K79 are mainly correlated with transcriptional stimulation, whereas methylation of H3-K9, H3-K27 and H4K20 comprise marks for transcriptionally silent chromatin (Vaquero et al 2003, Lachner et al 2003). Finally, histone lysine positions can be presented in mono-, di-, or trimethylated states, with signi¢cant di¡erences in biological output (Santos-Rosa et al 2002, Peters et al 2003). Here, we will focus on the repressive marks H3-K9 and H3-K27 methylation to illustrate the complexities of histone lysine methylation and to underline technical challenges (e.g. by developing highly selective methyl-lysine histone antibodies) in the analysis of position-speci¢c methylation states. For example, H3-K9 and H3K27 methylation have been associated with heterochromatin formation, X chromosome inactivation, Polycomb-mediated transcriptional memory and euchromatic gene repression (Lachner et al 2003) (see Fig. 1). How can these marks be present at so many di¡erent chromatin regions and still be able to discriminate functionally di¡erent silencing domains? We will provide evidence using highly speci¢c methyl-lysine antibodies, together with the de¢nition of enzymatic pro¢les for several histone lysine methyltransferases (HMTases) that the combination of distinct H3-K9 and H3-K27 methylation states can indeed index di¡erent chromatin regions, consistent with some of the predictions for the existence of an ‘epigenetic code’. Histone methylation can direct DNA methylation in mammals The mammalian Suv39h genes  homologues of the Drosophila suppressor of position-e¡ect variegation Su(var)3-9 (Tschiersch et al 1994)  were the ¢rst genes to be shown to encode HMTase activity (Rea et al 2000). This activity is speci¢c for the H3-K9 position and is mainly directed towards pericentric heterochromatin (Peters et al 2001). On the basis of ¢ndings in Neurospora crassa (Tamaru & Selker 2001) and Arabidopsis thaliana (Jackson et al 2002) that histone methylation can direct DNA methylation, we decided to investigate whether there is a similar link in mammals. We generated mouse embryonic stem (ES) cells that

FIG. 1. Multiple roles of histone lysine methylation. The ¢gure summarizes described roles of H3-K9 and H3-K27 methylation in major epigenetic paradigms. Formation of constitutive heterochromatin is a multistep process in which small heterochromatic RNAs (shRNAs) are proposed to trigger H3-K9 trimethylation by Suv39h/HP1 complexes, followed by DNA methylation. A major epigenetic mark in X-inactivation is Xist-dependent H3-K27 trimethylation by Ezh2/eed complexes. Euchromatic gene repression correlates with H3-K9 dimethylation (Ogawa et al 2002, Tachibana et al 2001), whereas in Pc-mediated gene silencing the combination of H3-K27 and H3-K9 methylation may be the discriminating mark. Additional functions for the proposed indexing potential of histone lysine methylation may be regulated by di¡erences in mono- di- and trimethylation states. H3-K9 methylation is depicted by vertical, H3-K27 methylation by diagonal, and DNA methylation by small hexagons (Me).

24 SCHOTTA ET AL

HISTONE LYSINE METHYLATION

25

are double-null (dn) for the Suv39h HMTases, and analysed them for defects in DNA methylation. Digestion of genomic DNA with the methylation-sensitive MaeII (5’-ACGT) restriction enzyme indicated reduced DNA methylation at major satellite repeats in Suv39h dn ES cells. The major satellite repeats (major sats) are localized at pericentric heterochromatin and neighboured by minor sats at the centromeric regions. The surprising discoveries of a functional link between components of the RNAi machinery and H3-K9 methylation (Volpe et al 2002, Hall et al 2002) prompted a subsequent analysis for possible transcription across the major sats. Indeed, major sat transcripts were detected by RT-PCR that appeared signi¢cantly elevated as compared to minor sat transcripts. Together, these ¢ndings indicate that H3-K9 methylation can direct DNA methylation also in mammals (Lehnertz et al 2003). However, the interrelations between histone and DNA methylation appear more complex, since there is unaltered DNA methylation at minor sats or at endogenous retroviruses in Suv39h dn ES cells. Despite these di¡erences, the data suggest a hierarchy of epigenetic mechanisms, in which bi-directional transcription across DNA repeats may trigger the generation of double-stranded RNAs that are then processed by the RNAi machinery. The focal presence of ‘small heterochromatic’ RNAs or protein components of the RNAi machinery could recruit the Suv39h HMTases to major sats, where H3-K9 methylation would provide chromatin a⁄nities for HP1 and associated DNA methyltransferases, such as Dnmt3a and Dnmt3b (Lehnertz et al 2003) (see Fig. 1). Distinct histone lysine methylation pro¢les of constitutive and facultative heterochromatin The complexities of three distinct methylation states (mono-di-tri) and the di⁄culty to discriminate H3-K9 and H3-K27 positions, which are embedded in the same amino acid sequence, ARKS, prompted us to re-investigate described histone lysine methylation patterns. Indeed, many, if not all, of currently available H3-K9 and H3-K27 methyl-lysine histone antibodies show various degrees of cross-reactivities (Perez-Burgos et al 2003). We therefore generated a new series of highly speci¢c antibodies that selectively recognize the trimethyl state of H3-K9 and H3-K27 methylation (Fig. 2A). These novel antibodies di¡er from previously generated ‘structural’ 4x-H3-K9 methyl antibodies, which had been characterized as being speci¢c for H3-K9 dimethylation, since they showed no cross-reactivity with unmodi¢ed or with H3-K27 dimethylated peptides (Peters et al 2001) (see Fig. 2A). Extended quality controls with a full peptide panel, however, indicate broad reactivity with many di- and trimethylated states, independent of the respective lysine position (Perez-Burgos et al 2003). We

FIG. 2. H3-K9 and H3-K27 trimethylation can distinguish constitutive and facultative heterochromatin. (A) Immuno-dotblots showing speci¢city of the 4x-di-meth H3-K9 (‘multi-methyl lysine’), 2x-tri-meth H3-K9 and 2x-tri-meth H3-K27 antibodies. The branched peptides used to raise these antibodies are depicted as inserts. Dot-blots were performed as described (Perez-Burgos et al 2003). (B) Indirect immuno£uorescence of mitotic chromosome spreads prepared from wild-type (wt) female mouse embryonic ¢broblasts (MEFs). Focal dots visualized by the 2x-tri-meth H3-K9 antibody represent pericentric heterochromatin, whereas the inactive X chromosome (Xi; indicated by an arrow) is labelled by the 2x-tri-meth H3-K27 antibody.

26 SCHOTTA ET AL

HISTONE LYSINE METHYLATION

27

therefore renamed the 4x-H3-K9 methyl antibodies as ‘multi-methyl lysine’ antibodies that have an apparent preference for recognizing a high density or clustering of methylated histone tails. Indirect immuno£uorescence (IF) of mitotic chromosomes from mouse embryonic ¢broblasts (MEFs) indicates a composite staining pattern for the ‘multi-methyl lysine’ antibodies, with focal signals at pericentric heterochromatin, decoration of the inactive X chromosome (Xi) and dispersed staining along the chromosomal arms (Fig. 2B). By contrast, the H3-K9 trimethyl speci¢c antiserum (#4861) selectively stains pericentric foci but fails to decorate the Xi. Surprisingly, the H3-K27 trimethyl speci¢c antiserum (#6523) detects methyl epitopes on the Xi, is negative for pericentric foci and displays broad staining along the chromosomal arms (see Fig. 2B). These results suggest that H3-K9 vs. H3-K27 trimethylation can discriminate constitutive and facultative heterochromatin. In this model, the Suv39h HMTases would induce H3-K9 trimethylation at pericentric heterochromatin, although there is some Suv39h-independent H3-K9 dimethylation at centromeric regions (Lehnertz et al 2003, Peters et al 2003). Thus, the combination of high-density trimethylation of the H3-K9 position, together with some H3-K9 dimethylation, would provide an indexing signal for constitutive heterochromatin (see model Fig. 5, below). H3-K27 trimethylation, on the other hand, would be a prominent mark for the Xi, and HMTases with an apparent preference for the H3-K27 position, such as the Ezh2/eed complex, are likely regulators to induce distinct histone lysine methylation patterns during Xinactivation (Plath et al 2003, Silva et al 2003). Since H3-K9 di-methylation is also present at the Xi (Boggs et al 2002, Peters et al 2002, Heard et al 2001), combined H3-K9 dimethylation/H3-K27 trimethylation appear as a discriminating signal for the Xi. The Suv39h enzymes are trimethylating H3-K9 HMTases We next wished to examine the reaction pro¢les and substrate speci¢cities of the Suv39h HMTases. In vitro analyses indicated that recombinant SUV39H1 can use unmodi¢ed H3 peptides (Rea et al 2000), although there is a much higher preference for H3-K9 monomethylated substrates (Peters et al 2003 and unpublished observations). To investigate in vivo methylation pro¢les, we generated a series of H3-K9 methyl-speci¢c antibodies that can discriminate mono- di- and trimethylated states (Fig. 3A). IF analyses of mitotic chromosomes from wild-type MEFs indicated a focal enrichment for H3-K9 trimethylation at pericentric heterochromatin, a two-dotted signal for H3-K9 dimethylation at the centromeres, but the absence of H3-K9 monomethylation at heterochromatin. Both H3-K9 mono- and dimethylation are broadly present along

FIG. 3. The SUV39H HMTases are trimethylating enzymes. (A) Immuno-dotblots showing speci¢city of the 2x-mono-meth H3-K9, 2x-dimeth H3-K9 and 2x-tri-meth H3-K9 antibodies. The branched peptides used to raise these antibodies are depicted as inserts. Dot-blots were performed as described (Perez-Burgos et al 2003). (B) HeLa cells were transfected with constructs expressing (myc)3-tagged wild-type SUV39H1 or a catalytically inactive SUV39H1H324L mutant (Rea et al 2000). Cells were co-labelled with mouse monoclonal a-myc antibody (9E10) and the respective rabbit polyclonal a-mono, a-di and a-tri-methyl H3-K9 antibodies. Cells with high levels of (myc)3-SUV39H1 show a reduction of H3-K9 mono- and H3-K9 dimethylation but reveal a signi¢cant increase in H3-K9 trimethylation. No di¡erences in H3-K9 methylation states are detected in HeLa cells overexpressing the inactive SUV39H1H324L mutant.

28 SCHOTTA ET AL

HISTONE LYSINE METHYLATION

29

the chromosomal arms. In Suv39h dn MEFs, pericentric H3-K9 trimethylation is selectively lost, concomitant with a signi¢cant increase of H3-K9 monomethylation in this region (Peters et al 2003). This surprising result suggests a mechanism in which a currently unde¢ned H3-K9 monomethylase would act synergistically with the Suv39h enzymes to induce and propagate H3K9 trimethylation at pericentric heterochromatin. The reaction pro¢les of the Suv39h HMTases can also be visualized upon overexpression of a (myc)3-tagged SUV39H1 enzyme in HeLa cells. As a control, we used a catalytically inactive point mutant (H324L) (Rea et al 2000). Overexpression results in a broad distribution of the ectopic enzyme in the interphase nucleus (Melcher et al 2000). In myc-positive cells, there is a signi¢cant reduction of H3-K9 mono- or dimethylation as compared to nontransfected cells (Fig. 3B). By contrast, cells with high ectopic (myc)3-SUV39H1 levels display a pronounced enrichment for H3-K9 trimethylation all across the interphase nucleus. Together, these data indicate that the mammalian Suv39h enzymes are H3-K9 trimethylating HMTases, and that their ectopic expression can signi¢cantly perturb the balance between mono-, di- and trimethylation patterns that are normally present in the interphase nucleus. A combinatorial indexing signal for Polycomb-mediated gene silencing? Establishment and maintenance of gene expression patterns during development is dependent upon the antagonistic functions of Polycomb (Pc-G) and trithorax (trxG) group proteins, in a process termed transcriptional memory (Lyko & Paro 1999). Intriguingly, the Pc-G proteins Enhancer of zeste [E(z)] and Polycomb (Pc) form a histone methylation system, in which a SET domain enzyme [E(z)] and a chromo-domain protein (Pc) have been shown to physically interact. The Drosophila E(z)^Esc complex, (Czermin et al 2002, Mˇller et al 2002) and its mammalian Ezh^eed counterpart (Cao et al 2002, Kuzmichev et al 2002) possess HMTase activity towards H3-K27 and, to a minor extent, also towards H3-K9. In E(z) mutants, both H3-K27 (Cao et al 2002) and H3-K9 (Czermin et al 2002) methylation at Polycomb response elements (PREs) appear impaired. Moreover, in vitro binding studies suggest that a H3-K27 di- or trimethylated position can function as a partial recruiting signal for Pc (Czermin et al 2002, Kuzmichev et al 2002). However, the true in vivo histone methylation marks for Pc-recruitment remained unresolved. Evidence for a potential involvement of H3-K9 methylation had previously been suggested by the interaction of the SUV39H1 HMTase and the Pc-G member HPC2 (Sewalt et al 2002). In these studies, overexpression of SUV39H1 in HeLa cells resulted in the redistribution of several Pc-G proteins to so-called

30

SCHOTTA ET AL

FIG. 4. Overexpressed SUV39H1 redistributes H3-K27 trimethylation to Polycomb bodies. HeLa cells as shown in Fig. 3 were co-labelled for the Pc-G protein BMI1 and for H3-K9 or H3-K27 trimethylation using mouse monoclonal a-BMI1 (Sewalt et al 2002) and rabbit polyclonal 2xtri-meth H3-K9 or 2x-tri-meth H3-K27 antibodies. In SUV39H1 overexpressing HeLa cells, BMI1 accumulates at three to four foci re£ecting pericentric heterochromatin at region 1q12 on human chromosome 1 (Sewalt et al 2002) and co-localizes with H3-K27 trimethylation (see arrows). No relocalization is detected in non-transfected HeLa cells or in HeLa cells carrying an inactive SUV39H1H324L mutant. The pericentric 1q12 region contains arrays of satellite 2 and 3 repeats (depicted below).

HISTONE LYSINE METHYLATION 31

32

SCHOTTA ET AL

FIG. 5. Combined histone lysine methylation marks can index distinct chromatin regions. For constitutive or pericentric heterochromatin, bidirectional transcription (dashed lines) across mouse major satellite repeats is proposed to give rise to small heterochromatic RNAs (shRNAs) that trigger H3-K9 trimethylation by Suv39h/HP1 complexes, followed by DNA methylation. A combinatorial mark for pericentric heterochromatin would be the clustering of H3-K9 trimethylation together with the presence of H3-K9 dimethylation. For X-inactivation, Xist-dependent recruitment of Ezh2-eed complexes induces H3-K27 trimethylation, which together with broad H3-K9 dimethylation can index the Xi. For establishment of Polycomb bodies at region 1q12 in human chromosome 1, an interplay between H3-K9 and H3-K27 trimethylation systems is proposed, resulting in a high-density accumulation of H3-K9 and H3-K27 trimethylation at 1q12 heterochromatin. Whether shRNAs will originate from the satellite 2 and 3 repeats at region 1q12 and may facilitate targeting of H3-K9 and/or H3-K27 trimethylation is currently not known.

HISTONE LYSINE METHYLATION 33

34

SCHOTTA ET AL

‘Polycomb bodies’ which coincide with the largest block of pericentric heterochromatin (1q12) present on human chromosome 1 (Sewalt et al 2002). Since all of the above analyses relied on methyl-lysine histone antibodies, most of which display cross-reactivity between H3-K9 and H3-K27 (Perez-Burgos et al 2003), we revisited histone methylation patterns and Polycomb recruitment in the SUV39H1-overexpressing HeLa cells with our new series of highly speci¢c H3-K9 and H3-K27 trimethyl antibodies. Similar to the IF staining described in Fig. 3B, we performed co-labelling experiments for H3-K27 trimethylation and BMI1, which is a prominent member of the human Pc complex. Whereas H3K27 trimethylation displays a dispersed and speckled distribution in HeLa nuclei, there is a dramatic relocalization upon (myc)3-SUV39H1 overexpression towards three to four focal dots representing 1q12 regions of chromosome 1 in these hypo-tetraploid HeLa cells. Signi¢cantly, these foci of H3-K27 trimethylation overlap with BMI1 signals (see arrows in Fig. 4). The induction of these ectopic H3-K27 trimethylation foci requires an active SUV39H1 enzyme. Since the Suv39h enzymes exclusively methylate the H3-K9 position (Rea et al 2000, Peters et al 2003), and H3-K9 trimethylation is prominently dispersed across the entire nucleus in SUV39H1-overexpressing HeLa cells, we propose that the combination of H3-K9 and H3-K27 trimethylation provides an in vivo indexing signal for recruitment of the mammalian Pc complex. Although this conclusion requires the analysis of histone lysine methylation at PREs under normal developmental decisions, there is genetic evidence in Drosophila that Su(var)3-9 overexpression can enhance Pc-dependent homeotic transformations (G. Reuter, personal communication). Conclusion The above examples highlight the exquisite complexity of histone lysine methylation. There are around 50 SET domain sequences (Kouzarides 2002) and 530-chromo domain genes in the mammalian genome. In addition, all of the currently known methylatable lysine positions in histone H3 and H4 aminotermini can be presented in mono-di-or trimethylated states. Despite these complexities, we are beginning to be able to make predictions about distinct methylation imprints and how combinatorial signals between selective lysine methylation marks may be used to index chromatin regions. Thus, pericentric heterochromatin, the inactive X chromosome and Pc-G mediated gene silencing may be distinguished by qualititative and quantitative changes between H3-K9 and H3-K27 methylation, followed by the recruitment of di¡erent chromatinassociated factors (e.g. HP1, an Xi-speci¢c silencing complex and Pc). Intriguingly, in all of these three paradigms, RNA molecules appear to be important primary signals for triggering the conversion of epigenetic

HISTONE LYSINE METHYLATION

35

modi¢cations to the underlying chromatin template. These insights allow for models that could discriminate various degrees of chromatin repression (see Fig. 5), although many more histone marks and epigenetic factors remain to be uncovered. In addition to the interplay between H3-K9 and H3-K27 methylation systems described here, there may well be other examples for the ‘plasticity’ of histone methylation pathways, particularly during developmental decisions or in comparing histone lysine methylation in normal vs. transformed cells. The Suv39h HMTases are expressed at only low to moderate abundance during mouse development (Peters et al 2001), and it is likely that changes in their expression levels will considerably perturb the balance of H3-K9 mono-di- and trimethylation and also of other histone lysine methylation marks. It is according to this view that aberrant histone lysine methylation patterns and altered nuclear architectures of ‘heterochromatic’ subdomains have been associated with tumour progression. For example, the PML-RAR fusion protein causes acute promyelocytic leukaemia (PML) that is connected with perturbed histone lysine methylation at RAR target promoters, presumably by dispersing the SUV39H1 HMTase to PML bodies (Di Croce et al 2002). Other examples include association of HMTases with the tumour suppressor Rb (Schneider et al 2002) or of histone lysine methylation with BRCA1 (D. Livingstone, personal communication), and EZH2 has been involved in the progression of prostate cancer (Varambally et al 2002). By contrast, transition of proliferating cells into the senescent state has been correlated with an increased appearance of ‘heterochromatin-like’ subnuclear structures (Narita et al 2003). Thus, the plasticity of histone lysine methylation systems and alterations in their associated pathways are predicted to have important implications for normal di¡erentiation and proliferation. A more detailed understanding of epigenetic networks therefore promises to yield new mechanistic insights into the biology of aging and cancer and may o¡er novel strategies for the reversion of aberrant development. Acknowledgements We would like to thank our colleagues Gunter Reuter, Frank Eisenhaber, David Allis, Dieter Schweizer, Genevieve Almouzni, Neil Brockdor¡, Arie Otte, En Li and Yoichi Shinkai for their expertise and help in the course of these projects. We would also like to acknowledge all the past and present members of the Jenuwein laboratory for their enthusiasm and contributions without which the de¢nition of histone methylation systems would not have been possible. We are particularly indebted to Judd Rice and David Allis for contributing to the characteriziation of the new series of a-2x-mono-di-tri H3-K9 methyl-speci¢c antibodies. Research in the laboratory of T.J. is supported by the I.M.P. through Boehringer Ingelheim and by grants from the Vienna Economy Promotion Fund, the European Union and the Austrian GEN-AU initiative, which is ¢nanced by the Bundesministerium fˇr Bildung, Wirtschaft und Kultur.

36

SCHOTTA ET AL

References Boggs BA, Cheung P, Heard E, Spector DL, Chinault AC, Allis CD 2002 Di¡erentially methylated forms of histone H3 show unique association patterns with inactive human X chromosomes. Nat Genet 30:73^76 Cao R, Wang L, Wang H et al 2002 Role of histone H3 lysine 27 methylation in Polycomb-group silencing. Science 298:1039^1043 Czermin B, Mel¢ R, McCabe D, Seitz V, Imhof A, Pirrotta V 2002 Drosophila enhancer of Zeste/ ESC complexes have a histone H3 methyltransferase activity that marks chromosomal polycomb sites. Cell 111:185^196 Di Croce L, Raker VA, Corsaro M et al 2002 Methyltransferase recruitment and DNA hypermethylation of target promoters by an oncogenic transcription factor. Science 295:1079^1082 Hall IM, Shankaranarayana GD, Noma K, Ayoub N, Cohen A, Grewal SI 2002 Establishment and maintenance of a heterochromatin domain. Science 297:2232^2237 Heard E, Rougeulle C, Arnaud D, Avner P, Allis CD, Spector DL 2001 methylation of histone H3 at Lys-9 is an early mark on the X chromosome during X inactivation. Cell 107: 727^738 Jackson JP, Lindroth AM, Cao X, Jacobsen SE 2002 Control of CpNpG DNA methylation by the KRYPTONITE histone H3 methyltransferase. Nature 416:556^560 Jenuwein T, Allis CD 2001 Translating the histone code. Science 293:1074^1080 Kouzarides T 2002 Histone methylation in transcriptional control. Curr Opin Genet Dev 12:198^209 [Erratum in Curr Opin Genet Dev 12:371] Kuzmichev A, Nishioka K, Erdjument-Bromage H, Tempst P, Reinberg D 2002 Histone methyltransferase activity associated with a human multiprotein complex containing the Enhancer of Zeste protein. Genes Dev 16:2893^2905 Lachner M, O’Sullivan RJ, Jenuwein T 2003 An epigenetic road map for histone lysine methylation. J Cell Sci 116:2117^2124 Lehnertz B, Ueda Y, Derijck A et al 2003 Suv39h-mediated histone H3 lysine 9 methylation directs DNA methylation to major satellite repeats at pericentric heterochromatin. Curr Biol 13:1192^1200 Lyko F, Paro R 1999 Chromosomal elements conferring epigenetic inheritance. Bioessays 21:824^832 Melcher M, Schmid M, Aagaard L, Selenko P, Laible G, Jenuwein T 2000 Structure-function analysis of SUV39H1 reveals a dominant role in heterochromatin organization, chromosome segregation, and mitotic progression. Mol Cell Biol 20:3728^3741 Mˇller J, Hart CM, Francis NJ et al 2002 Histone methyltransferase activity of a Drosophila Polycomb group repressor complex. Cell 111:197^208 Narita M, Nunez S, Heard E et al 2003 Rb-mediated heterochromatin formation and silencing of E2F target genes during cellular senescence. Cell 113:703^716 Ogawa H, Ishiguro K, Gaubatz S, Livingston DM, Nakatani Y 2002 A complex with chromatin modi¢ers that occupies E2F- and Myc-responsive genes in G0 cells. Science 296: 1132^1136 Perez-Burgos L, Peters AH, Opravil S, Kauer M, Mechtler K, Jenuwein T 2003 Generation and characterization of methyl-lysine histone antibodies. Methods Enzymol Vol 376, in press Peters AH, O’Carroll D, Scherthan H et al 2001 Loss of the Suv39h histone methyltransferases impairs mammalian heterochromatin and genome stability. Cell 107:323^ 337 Peters AH, Mermoud JE, O’Carroll D et al 2002 Histone H3 lysine 9 methylation is an epigenetic imprint of facultative heterochromatin. Nat Genet 30:77^80

HISTONE LYSINE METHYLATION

37

Peters AH, Kubicek S, Mechtler K et al 2003 Partitioning and plasticity of repressive histone methylation states in mammalian chromatin. Mol Cell 12:1577^1589 Plath K, Fang J, Mlynarczyk-Evans SK et al 2003 Role of histone H3 lysine 27 methylation in X inactivation. Science 300:131^135 Rea S, Eisenhaber F, O’Carroll D et al 2000 Regulation of chromatin structure by site-speci¢c histone H3 methyltransferases. Nature 406:593^599 Rice JC, Allis CD 2001 Histone methylation versus histone acetylation: new insights into epigenetic regulation. Curr Opin Cell Biol 13:263^273 Santos-Rosa H, Schneider R, Bannister AJ et al 2002 Active genes are tri-methylated at K4 of histone H3. Nature 419:407^411 Schneider R, Bannister AJ, Kouzarides T 2002 Unsafe SETs: histone lysine methyltransferases and cancer. Trends Biochem Sci 27:396^402 Silva J, Mak W, Zvetkova I et al 2003 Establishment of histone H3 methylation on the inactive X chromosome requires transient recruitment of Eed-Enx1 polycomb group complexes. Dev Cell 4:481^495 Sewalt RG, Lachner M, Vargas M et al 2002 Selective interactions between vertebrate polycomb homologs and the SUV39H1 histone lysine methyltransferase suggest that histone H3-K9 methylation contributes to chromosomal targeting of Polycomb group proteins. Mol Cell Biol 22:5539^5553 Strahl BD, Allis CD 2000 The language of covalent histone modi¢cations. Nature 403:41^45 Tachibana M, Sugimoto K, Nozaki M et al 2002 G9a histone methyltransferase plays a dominant role in euchromatic histone H3 lysine 9 methylation and is essential for early embryogenesis. Genes Dev 16:1779^1791 Tamaru H, Selker EU 2001 A histone H3 methyltransferase controls DNA methylation in Neurospora crassa. Nature 414:277^283 Tschiersch B, Hofmann A, Krauss V, Dorn R, Korge G, Reuter G 1994 The protein encoded by the Drosophila position-e¡ect variegation suppressor gene Su(var)3-9 combines domains of antagonistic regulators of homeotic gene complexes. EMBO J 13:3822^3831 Turner BM 2000 Histone acetylation and an epigenetic code. Bioessays 22:836^845 van Holde KE 1988 Chromatin. Springer Verlag, New York Vaquero A, Loyola A, Reinberg D 2003 The constantly changing face of chromatin. Sci Aging Knowl Environ 14:RE4 http://sageke.sciencemag.org/cgi/content/full/sageke;2003/14/re4 Varambally S, Dhanasekaran SM, Zhou M et al 2002 The polycomb group protein EZH2 is involved in progression of prostate cancer. Nature 419:624^629 Volpe TA, Kidner C, Hall IM, Teng G, Grewal SI, Martienssen RA 2002 Regulation of heterochromatic silencing and histone H3 lysine-9 methylation by RNAi. Science 297:1833^ 1837 Zhang Y, Reinberg D 2001 Transcription regulation by histone methylation: interplay between di¡erent covalent modi¢cations of the core histone tails. Genes Dev 15:2343^2360

DISCUSSION Verdin: I was intrigued by the experiment in which you overexpressed Suv39h and showed relocalization. Is this true relocalization? Do you mean to say that the Ezh complex together with Polycomb would be localized to the 1q12 region which would mean that overexpression of Suv39h would function at other loci? Is this something that you see?

38

SCHOTTA ET AL

Jenuwein: Perhaps I should have explained this di¡erently. It is clear that we get a relocalization, because the focal staining is so bright, and the normal staining of BMI or human Polycomb is rather dispersed. It is not that we see relocalization of the Ezh complex which remains more-or-less dispersed. It is really Polycombassociated proteins. One of the reasons I put our data up for discussion is that I think that what we are looking at is a cross-talk between histone methylation systems. This is important. We could easily have conditions when we compare normal cells under physiological conditions with transformed cells under cancer conditions, or stem cells versus di¡erentiated cells, where the levels of these enzymes di¡er. As a consequence you can get di¡erent marks being induced. We also know from chromatin immunoprecipitation (ChIP) data in the Suv39h double null cells that H3-K27 methylation is now present and can be detected in pericentric heterochromatin. This illustrates the tremendous plasticity and crosstalk that occurs between these histone methylation systems. Cohen: You referred to the di¡erent levels of enzymes between normal cells and transformed cells, stem cells and di¡erentiated cells. Have you got any data on these di¡erences? Jenuwein: For the Suv39h enzymes, we were surprised to see that they are moreor-less ubiquitously expressed at a moderate level. They may be modulated by phosphorylation, though. Other histone methyltransferases are much more dynamic in their expression. If we take a resting cell from the spleen and look at the level of Ezh2 it is very low, but it can be dramatically up-regulated when we activate the cells to proliferate. Initially we argued that one of the hallmarks of histone lysine methylation may be that it is much more stable than phosphorylation and acetylation. This lends it the more permanent mark. Work done in collaboration with Amanda Fisher (London) indicated that in these primary splenocytes there is almost no histone methylation at the resting state. As soon as the cells are activated there is a tremendous up-regulation for monodi- and trimethylation. This is surprising. Similar data have come from recent work done by Wolf Reik (Babraham) in collaboration with Eckhard Wolf (Munich) which shows that in nuclear transfer experiments looking at cloning e⁄ciency of bovine cells, the transfer of a somatic nucleus allows for a more permanent decoration of trimethyl marks, even upon nuclear transfer. As a consequence these cells do not clone as e⁄ciently as embryonic cells. The trimethyl state, be it on H3-K9, H3-K27, or other lysine residues could potentially give a more de¢ned mark, thereby attenuating nuclear reprogramming e⁄ciencies. Mahadevan: I’d like to ask you a technical question, but one that has a biological angle. This relates to Bryan Turner’s point about the phosphate on the adjacent residues to the ones you are looking at. Will your antibodies recognize that residue when the phosphate is on the adjacent serine? What is the e¡ect of the adjacent phosphorylation on the interactions with HP1?

HISTONE LYSINE METHYLATION

39

Jenuwein: We have not looked directly with peptides that would be methylated on H3-K9 and phosphorylated on the adjacent Ser10. But I know from our work with David Allis that he has been looking in extracts that were treated with phosphatase. This is not a direct experiment in answer to the question you are asking, but there was no signi¢cant di¡erence in detecting signal with our antibodies. To the best of our knowledge, these are the most speci¢c antibodies that have been raised to detect mono- di- and trimethylation states, and now we have to look into the biology and the physiological function, including the tests that you suggest. Mahadevan: That experiment can be interpreted in two ways. One is that the phosphate is invisible to your antibody. The other is that this combination does not exist. But the mass spectrometric analysis that you mentioned suggests that it does exist. Jenuwein: Not in every H3 molecule, but maybe on a fraction of the H3 tails coming out from the mitotic extract. Mahadevan: What about interaction with its biological partners  HP1 and Polycomb? Jenuwein: Again, we have not tested this directly. We have not o¡ered a peptide that will be methylated and phosphorylated on H3-K9/Ser10 for binding to Polycomb. Allis: I showed the data for HP1. Marks: Have you looked at whether there is variation in the pattern of methylation in cycling cells going through di¡erent stages, of the cell cycle, or cells going through apoptosis? Jenuwein: The best data on this are from a collaboration with Amanda Fisher (London), where we started from a resting spleen cell and then looked at the methylation pattern after cell activation. We start with very little methylation and see a dramatic up-regulation. H3-K9 trimethylation is prominent at pericentric heterochromatin. Marks: When you get the trimethylation, at what stage are the cells in terms of the division cycle? Jenuwein: That is a good question. It will take 24 h. I can’t say whether this is now triggered by the cell division or whether it is independent of it. Turner: I have a comment on X-inactivation. We have done a lot of staining of metaphase chromosomes with various antibodies, including those to methyl H3-K9. We have never seen any sign that the inactive X is particularly enriched in methylated H3-K9, and this ¢ts in with the current thinking. However, we have also been doing some immunoprecipitation, and we have done ChIP analysis of embryonic stem cells before and after X-inactivation. We ¢nd that di/ trimethylation of H3-K4 goes down as the cells di¡erentiate and X-inactivation takes place. This ¢ts in with what we would expect  demethylation of H3-K4.

40

DISCUSSION

However, surprisingly we ¢nd a big increase in H3-K9 methylation by this ChIP analysis. Although we don’t see anything by immuno£uorescence, when we start looking at particular X-linked genes by ChIP, we see this striking severalfold increase in H3-K9 di- and trimethylation. There may be a di¡erence between the global patterns of methylation that we look at by immuno£uorescence and what we see when we start looking at particular regions and particular genes. Jenuwein: For the antibodies, you are correct in raising these issues. Again, in order to get to some clearer answers one has to combine immuno£uorescence and ChIP analysis on di¡erent nuclear backgrounds, and then see what the consistent feature is. Just one or two assays will not be good enough. It is possible that with H3-K9 there are hotspots. I would not be surprised to ¢nd a mark there to discriminate how you can enter the pathway for facultative heterochromatin. Turner: We were surprised by this. We saw the immuno£uorescence and we weren’t expecting much to happen. But it really is very striking, and particularly so because it is going in exactly the opposite direction to di/trimethylation of K4, which is going down as you would expect. Li: Where is the H3-K9? Which region do you look at? Turner: We have looked at probably a dozen regions on the X chromosome. We have looked at promoter regions and coding regions. At the moment everything seems to be going in the same direction, but we are looking at genes and not intergenic regions or heterochromatin as yet. Li: Is it possible that H3-K27 methylation is more widely spread on the X chromosome, involving condensation of the chromosome, and H3-K9 is more speci¢c to gene expression? Turner: That would ¢t with the data so far. Li: We have looked at DNA methylation and X-inactivation. We used Dnmt3a/ Dnmt3b double mutant mice and ES cells. We found that DNA methylation is not essential for initiating the X-inactivation process. You can have normal Xinactivation without de novo DNA methylation after di¡erentiation of ES cells or in the embryo. Previously, we also showed that loss of maintenance of DNA methylation can a¡ect the stability of X-inactivation. This may be related to pericentric methylation. Castronovo: Is there any evidence that this di- and trimethylation can be reversible? Jenuwein: At the moment, no. But once it is there it can still be removed under physiological settings. There is one additional mechanism I would like to discuss. I can foresee that under certain physiological conditions, such as very early after fertilization or when the genome becomes reprogrammed and DNA methylation marks are being actively erased, that this may be a condition where there are intermediate enzymes that could destabilize the N^C bond by oxidation or

HISTONE LYSINE METHYLATION

41

radical attack. The enzymes are very tightly regulated and would only be around at that early stage during embryogenesis. Then, once you have an oxidation event at a di- or trimethyl-lysine mark, you could potentially actively demethylate it because it is destabilized. This is a model that one should also look at, particularly in these cells under these physiological conditions. Castronovo: Do you think such a process could be involved in cancer activation of oncogenes? Jenuwein: Yes. There is tremendous plasticity and cross-talk between histone methylation systems  this is something I would like to emphasize. This includes acetylation and phosphorylation. Any time you alter a physiological setting, be it during development or the transition from a normal to a cancer state, then marks are altered. In cancer they are set where they normally should not be and you completely perturb the nuclear architecture. Marks: Have you any sense of the extent to which di- and trimethylation is present in transformed cells versus their normal counterparts? Jenuwein: We have indirect evidence. We gave reagents to Scott Lowe (CSHL) and he has been looking at the senescent state  trying to interfere with it. He observed that the more senescent a cell gets, the more heterochromatin is being built up (Narita et al 2003). Then we also shared our antibodies with him, and surprisingly he sees an increase in H3-K9 trimethyl marks going along with the perturbed physiological state. Whenever the cell detours from the normal programme, there can be quite a dramatic redistribution of methyl marks and other epigenetic changes. Marks: Is this just in senescing cells? Jenuwein: Yes. In cancer cells we are just starting to collaborate with Steve Baylin, and we will look at how these marks are altered there. Verdin: The connection with the small interfering RNAs is a fascinating one. What are the hypotheses that you are working on linking the two mechanisms? Jenuwein: There are two. We were intrigued to see that the branched antibody is not speci¢c for position but for a high concentration of histone methylation marks. Could we use this antibody to detect any polymer that is methylated, be it peptide, DNA or RNA? One of our working hypotheses is that these short RNAs are being accumulated because the concentration of satellite repeats is so high at pericentric heterochromatin. There are thousands of copies. As a consequence, the small heterochromatic RNAs stay in the nucleus because they have an a⁄nity for the underlying abundant DNA repeats. But you would need an adaptor molecule that can see the RNA in conjunction with chromatin. In work done with Genevieve Almouzni (Paris) we have shown that the localization of the H3-K9 methyl mark at pericentric heterochromatin is sensitive to RNAse treatment. If you treat permeabilized cells with RNAse you lose the H3-K9 methyl mark

42

DISCUSSION

(Maison et al 2002). It is as if a RNA molecule is required to induce a distinct con¢guration of the nucleosomes at pericentric heterochromatin. Verdin: How can you exclude that your antibody is actually not detecting methylated RNA? Jenuwein: That is another possibility. We are not excluding this but it has to be tested. So far there is no evidence for this. The high concentration of the small heterochromatic RNAs could see an adaptor molecule possibly containing a chromodomain. Work has been published arguing that HP1 could also have an a⁄nity for RNA (Muchardt et al 2002). The same could hold true for at least one of the Polycomb members. It is intriguing to see that in constitutive heterochromatin and facultative heterochromatin; and for Polycomb recruitment, there is evidence that RNA molecules could be the primary trigger for epigenetic marking. Verdin: Have any of the known molecules involved with siRNA been obtained in screens for suppressors of variegation? Jenuwein: Robin Allshire (Edinburgh) has some evidence. He has identi¢ed one or two alleles in gene silencing screens in S. pombe. This is very nice because, unlike S. cerevisiae, S. pombe maintains gene products involved in RNAi interference and H3-K9 methylation. Robin has identi¢ed epi-alleles for argonaute, which is a component of the RNAi interference complex that can process the small heterochromatic RNAs. Allis: Thomas referred to Marty Gorovksy’s study where Tetrahymena beautifully displays the H3-K9 methyl mark right at a period where small RNAs dominate the life cycle. This process is dependent on the machinery that produces the small RNAs. One of the chromodomains in the protein that reads the H3-K9 is unusual: Tetrahymena has dropped the Drosophila Mof-like chromodomain that Peter Becker originally described as a RNA binder (Akthar et al 2000) into one of these DNA elimination proteins. It looks like it reads RNA and it doesn’t care about the lysine 9 mark. I wouldn’t rule out that chromodomains reading RNA will not be a repeating theme. Baylin: You mentioned the di¡erences between stem cells and di¡erentiated cells. Has this been carefully studied for steady-state levels of some of the HMTases, such as Suv39h, G9a and Ezh? Jenuwein: I can tell you this for Suv39h and Ezh, but only partially for G9a. The Suv39h enzymes are present ubiquitously. You can detect Suv39h all across the cell cycle: it does not £uctuate. But coming back to an earlier thought, what does appear to change are the post-translational modi¢cations of the enzyme itself, namely phosphorylation. This may dictate the enzymatic activity. Baylin: Is this the same going from a stem cell to a more di¡erentiated state as it is across the cell cycle?

HISTONE LYSINE METHYLATION

43

Jenuwein: For all the cells we have looked at  mouse embryonic ¢broblasts, embryonic stem cells and di¡erentiated ES cells  we see the Suv39h enzymes present all the time. The situation is completely di¡erent for Ezh. Ezh2 is very high in ES cells. As an aside, here, one should not forget that in mammals there are two Ezh loci, Ezh1 and Ezh2. So far, most studies have only looked at Ezh2 because it is much more highly expressed. Ezh2 is very dynamic. It can be downregulated when ES cells di¡erentiate, it can be dramatically up-regulated when resting cells go into an activated state and there are also £uctuating levels in disease states. Castronovo: Do you have any evidence that oxidative stress may alter the patterns of methylation? If so, could we revisit the concept that cancer may be due to just epigenetic modi¢cation? If this is correct, how would you explain the transmission of these epigenetic modi¢cations from one cancer cell generation to the next? Jenuwein: On the basis of what we have learned within the last three years, I would not be surprised to see methylation also playing a very important role in regulating epigenetic transitions. I would not rule this out. Allis: I tend not to retain all the metabolism I ever learned. But there is a review driven by a need to solve cancer, written by someone at Scripps called Shi Huang. He has a hypothesis stemming from work done in the 1940s, when rats were fed things that £uctuated SAM to S-adenosyl homocysteine. This produces tumours like crazy. His hypothesis is that this is all feeding back to these epigenetic histone and DNA marks (Huang 2002). Jenuwein: There are some other very important studies aimed at identifying small molecule inhibitors for HMTases, similar to what has been done for histone deacetylases (HDACs). Several high-throughput screens are currently underway in a variety of laboratories. I would predict that the de¢nition of speci¢c HMTase inhibitors would have a far-reaching potential for cancer therapy and altering stem cell plasticity. Zhou: I can make a connection with potential drug targets for SET domain proteins. This SET domain protein is called MLL (mixed lineage leukaemia), which has multiple modular domains. The MLL gene undergoes chromosomal translocation with other genes, and the resulting fusion proteins contain its SET domain. It has been shown that the SET domain activity is actually required for the transformation activity of the MLL protein. I wanted to raise the topic of this protein for another reason. People have tried hard to study the activity of the MLL SET domain. But it has very little activity in in vitro assays with histone peptides or proteins. It is only in vivo that the activity of this SET domain is detected. This study was a paper published in Molecular Cell a couple of years ago in which David Allis was a co-author (Milne et al 2002). The concern of the SET domain activity comes from your data showing that ‘histone coding’ complexity of H3-K9 methylation is possibly controlled by

44

DISCUSSION

more than one enzyme. I think this may also explain why we don’t see SET domain activity in vitro in this particular protein. There are two possibilities. The ¢rst one is that the SET domain may require a cofactor, or being complexed with other proteins. The second is simply that people haven’t used the right substrates. The right substrates could require speci¢c mono- or dimethylation. Moreover, it appears that there are a large number of SET domain histone lysine methyltransferases, which have been identi¢ed so far in the protein database, as compared to that of the histone lysine acetyltransferase family. This probably also indicates the complexity of methylation patterns in the ‘histone code hypothesis’. Have you identi¢ed any speci¢c SET domains that can only methylate lysines in histones that have already been mono- or dimethylated on the same lysine residues? Jenuwein: I totally agree with your argument. We were really surprised to see how we can alter the H3-K9 methylation states at pericentric heterochromatin in the absence of the Suv39h enzymes. We would like to know the identity of the putative H3-K9 monomethyl HMTase. Similarly, there is also evidence that there can be cross-talk between trimethylated H3-K9 with other methylation marks, and that a pre-modi¢ed histone N-terminus is required for this to happen. Again, this highlights the connection that appears to exist between distinct histone lysine methylation systems. However, what we see in mammals may not always be re£ected in S. pombe or Drosophila. We were arguing that a trimethyl state would be the prominent mark for constitutive heterochromatin, and the Suv39h enzymes would do this. Work done in Drosophila (in collaboration with Gunter Reuter, Halle/S.) and S. pombe (in collaboration with R. Allshire, Edinburgh) indicate that in these organisms, it is more the dimethyl and not the trimethyl state that seems enriched at heterochromatin. Allis: While I agree about the complexity and that we are limited by what substrates we throw at these enzymes, eventually Scott Briggs in my lab did get a terri¢c amount of MLL-SET domain activity out of the isolated recombinant domain with an unmodi¢ed H3 peptide. It took about a year to ¢gure out what was going on, though. This is published (Milne et al 2002). Turner: Am I right in thinking that MLL only goes up to the dimethyl level? Allis: We haven’t looked at that. It is a very good lysine 4-methyltransferase, but I have no idea where it stops or where it goes to. Verdin: Are there any known small molecule inhibitors for the methyltransferase activity of SET proteins? Jenuwein: This is clearly one of the most intriguing areas. I know that a variety of laboratories are working on inhibitors and have identi¢ed them for arginine methyltransferases. These inhibitors will hopefully be di¡erent for lysine methyltransferases, because we would like this speci¢city. We are interested in doing this. For the time being there exist only generic inhibitors, such as SAM

HISTONE LYSINE METHYLATION

45

analogues. Small molecule inhibitors can now be better predicted from the available 3D-structures of SET domains and one can limit input material for the high-throughput screens. It would be absolutely fantastic to have such speci¢c inhibitors and use them to investigate epigenetic transitions in stem cell plasticity, di¡erentiation, tumorigenesis and other paradigms. Berger: Can you comment on the order of histone and DNA methylation? It seems like in the model you showed that histone methylation was upstream of DNA methylation. Jenuwein: You are right. The picture I showed indicated that there is a directionality for histone methylation triggering DNA methylation. In reality it is much more complex, particularly in mammals. In our collaboration with En Li (Harvard) we showed that Suv39h enzymes selectively trimethylate H3-K9 at pericentric heterochromatin. The HP1 a and b adaptor molecules can then bind to this mark and also interact with DNA methyltransferase Dnmt3b. Independent of this mechanism there is H3-K9 dimethylation at the minor satellite repeats present at the centromeres. This is totally independent of the Suv39h enzymes. Then there are data from Steve Baylin’s laboratory indicating that you can alter DNA methylation pro¢les by interfering with 5’-azacytidine which, in turn, will change histone lysine methylation marks. We would argue that in mammals there is a feedback process: there is a Suv39h-dependent mechanism that recruits Dnmt3b for pericentric DNA methylation, and there is a Suv39h-independent mechanism that allows for DNA methylation at the centromeric regions. There, histone methylation can even be triggered by preceding DNA methylation. We are not excluding this. But we can clearly show, at least for the Suv39hdependent pathway, that histone methylation can trigger DNA methylation also in mammals. Gu: Related to this, I have a question. The DNA methylation pattern can be passed from one generation to the next, as occurs in imprinting. If your model is right, histone methylation can also be passed and contributes to histone memory. Have you any evidence to show that histone methylation can be passed from one generation to the next? Jenuwein: We are thinking about experiments to test this, but we have not done them yet. One possibility would be to take the Suv39h double null cells where we know there is no more decoration of pericentric heterochromatin, and come back with an experiment done earlier by Renato Paro (Heidelberg) looking for transcriptional memory conferred by a pulse of histone acetylation (Cavalli & Paro 1999). One could bring back into the Suv39h double null cells inducible HMTases that are speci¢c for trimethyl, such as Suv39h, or HMTases that are tailored to be only dimethyl speci¢c, such as taking the SET-domain of G9a and targeting it to pericentric heterochromatin with the chromodomain of

46

DISCUSSION

Suv39h. And one could even bring back an inducible enzyme that would be speci¢c for monomethyl. Then you set a pulse and induce the mark at pericentric heterochromatin, because all of these engineered enzymes would be targeted by the chromodomain of Suv39h, and you could follow the persistence of the mark with immuno£uorescence in the subsequent cell divisions. By doing this, one can ask whether trimethyl will be more stably propagated than mono- or dimethyl. I should also mention that Shiv Grewal and Amar Klar did a very nice experiment in S. pombe, where they raised the Swi6 level (the HP1 homologue of S. pombe) and looked for transcriptional memory. They did that with a pulse, inducing Swi6 for one or two cell divisions and then shutting it o¡. The pulsed cells had a much longer memory of the repressed state at the silent mating type loci that could be propagated for more than 30 generations and even during meiosis (Grewal & Klar 1996). We would now argue that this persistent memory would, at least in part, be the consequence of having an H3-K9 methyl group there with Swi6 protecting and perpetuating this epigenetic state across many cell divisions. Turner: To go back to the di/trimethylation story, your results are compelling and results from my lab also suggest that on X-linked genes in ES cells there are real di¡erences in the way tri- and dimethyl H3-K4 behave. But I think there is a caveat, and this is especially true in something like S. cerevisiae where you have only got one H3-K4 methyltransferase. It is di⁄cult to make a distinction there between tri and dimethylation. The fact is, if you bring in high concentrations of SET1 to a particular region of chromatin, it is going to drive the methylation up to trimethylation. This will re£ect nothing more than the fact that you have high concentrations of the relevant enzyme in that position. It need not be a signi¢cant change in itself. The problem with that correlation between trimethylation and transcription is that if you say that transcriptionally active regions have a higher concentration than normal of the enzyme SET1, then you will inevitably go to trimethylation because you have a higher concentration of enzyme. I think we always have issues like this bubbling away in the background that we need to take into account. We need to be sure that we have evidence that it really is a speci¢c e¡ect rather than just enzyme targeting, which would be a rather trivial biochemical e¡ect. Jenuwein: Again, we are just moving along at such a high pace that it is hard to understand what all of these marks and methylation states mean. I agree with you that the signi¢cance of having a focal detection of a trimethyl mark could be just that you have a high concentration of the enzyme being localized. But even if it only results that you can go on for one or two more cell divisions, it could make a di¡erence. There are so many exciting questions that need to be analysed  it is certainly a highly rewarding time to do epigenetic research right now.

HISTONE LYSINE METHYLATION

47

References Akhtar A, Zink D, Becker PB 2000 Chromodomains are protein^RNA interaction modules. Nature 407:405^409 Cavalli G, Paro R 1999 Epigenetic inheritance of active chromatin after removal of the main transactivator. Science 286:955^958 Grewal SI, Klar AJ 1996 Chromosomal inheritance of epigenetic states in ¢ssion yeast during mitosis and meiosis. Cell 86:95^101 Huang S 2002 Histone methyltransferases, diet nutrients and tumour suppressors. Nat Rev Cancer 2:469^476 Maison C, Bailly D, Peters AH et al 2002 Higher-order structure in pericentric heterochromatin involves a distinct pattern of histone modi¢cation and an RNA component. Nat Genet 30:329^334 Milne TA, Briggs SD, Brock HW et al 2002 MLL targets SET domain methyltransferase activity to Hox gene promoters. Mol Cell 10:1107^1117 Muchardt C, Guilleme M, Seeler JS, Trouche D, Dejean A, Yaniv M 2002 Coordinated methyl and RNA binding is required for heterochromatin localization of mammalian HP1alpha. EMBO Rep 3:975^981 Narita M, Nu•ez S, Heard E et al 2003 Rb-mediated heterochromatin formation and silencing of E2F target genes during cellular senscence. Cell 113:703^716

A model for step-wise assembly of heterochromatin in yeast Danesh Moazed, Adam D. Rudner, Julie Huang, Georg J. Hoppe and Jason C. Tanny Department of Cell Biology, Harvard Medical School, Boston, MA 02115, USA

Abstract. Gene silencing involves the assembly of DNA into specialized chromatin domains that are inaccessible to trans-acting factors and are epigenetically inherited. In the budding yeast Saccharomyces cerevisiae, silent heterochromatic DNA domains occur at telomeres, the silent mating type loci, and the rDNA repeats. At telomeres and the mating type loci, silencing requires the Sir2, Sir3 and Sir4 proteins, the conserved N-termini of histones H3 and H4, and a number of chromatin assembly factors. The Sir proteins form a multimeric complex that binds preferentially to deacetylated nucleosomes through the Sir3 and Sir4 subunits. The Sir2 subunit possesses an unusual NAD-dependent deacetylase activity that is required for silencing at each of the above loci. Recent studies have shown that silent chromatin domains are assembled in a step-wise manner involving sequential cycles of deacetylation and SIR complex binding. Sir2-dependent deacetylation is speci¢cally required for the spreading of the complex to regions beyond nucleation sites but not for its initial binding to DNA at the mating type loci and telomeres. A distinct Sir2 complex called RENT is required for silencing at rDNA. In contrast to telomeres and the mating type loci, Sir2 activity is not required for association of RENT with rDNA. 2004 Reversible protein acetylation. Wiley, Chichester (Novartis Foundation Symposium 259) p 48^62

Eukaryotic chromosomes are divided into two types of domains, heterochromatin and euchromatin. Heterochromatic regions comprise inaccessible portions of the genome that are transcriptionally silent. In contrast, euchromatic regions are more accessible and comprise the transcriptionally active portions of the genome. The distinction between these two types of domains was initially based on cytological evidence in plant and insect cells. In the 1920s, working on moss cells, Emile Heitz observed that some chromosome regions maintained a dark staining appearance, characteristic of condensed mitotic chromosomes, throughout the cell cycle. Heitz named these domains heterochromatin to distinguish them from the remaining chromosome regions that appeared to unravel and decondense during interphase. Studies in the ensuing decades have established that heterochromatic domains are transcriptionally inactive, recombinationally inert, and in general less accessible to molecular probes. These domains range in size from several kilobases 48

HETEROCHROMATIN ASSEMBLY

49

to several hundred kilobases and in the case of X-inactivation can encompass an entire chromosome. Furthermore, once heterochromatic domains are assembled, they are epigenetically inherited over many cell divisions (reviewed in Grewal & Elgin 2002, Moazed 2001). The assembly of DNA into heterochromatin regulates both gene expression and genome stability. Heterochromatin is often associated with repetitive DNA sequences at centromeres and telomeres and plays important roles in the functions of these chromosome structures. Moreover, the expression of developmental regulators such as the homeotic genes in Drosophila and the mating type genes in the budding and ¢ssion yeasts is controlled by the assembly of DNA into silenced heterochromatic structures. The heritable inactivation of developmental regulators outside of their proper domains of expression along the body axes is crucial for di¡erentiation and maintenance of cell identity. Assembly into epigenetically heritable heterochromatin provides a mechanism for the stable preservation of gene expression patterns during development. In the budding yeast Saccharomyces cerevisiae, silent chromatin domains share a number of properties with heterochromatin and provide one of the best-studied examples of regional control of gene expression. In this yeast, silent chromatin domains occur at the silent mating type loci, called HML and HMR or HM loci, telomeric DNA regions, and within the ribosomal DNA (rDNA) repeats (reviewed in Gartenberg 2000, Gasser & Cockell 2001, Rusche et al 2003). Silencing at the HM loci and telomeres requires the Sir2, Sir3 and Sir4 proteins (Aparicio et al 1991, Klar et al 1979, Rine & Herskowitz 1987). The Sir1 protein is required for silencing only at the HM loci (Aparicio et al 1991, Rine & Herskowitz 1987). In addition the conserved N-termini of histones H3 and H4 are required for silencing but not viability (Kayne et al 1988). In particular, deletions of the H4 N-terminus or mutation of H4 lysine 16 (Lys16) to residues that mimic the acetylated state abolish silencing (Johnson et al 1992). At the rDNA repeats, Sir2, but not Sir3 or Sir4, is required for silencing. Sir2 forms a complex with Net1 and Cdc14 called RENT, which is recruited to DNA by interactions with Fob1 and RNA polymerase I (Bryk et al 1997, Huang & Moazed 2003, Shou et al 1999, Smith & Boeke 1997, Straight et al 1999). A hallmark of silent chromatin domains is the hypoacetylation of histones (Braunstein et al 1993, Turner 2000, Jenuwein & Allis 2001). Studies in budding yeast provide a direct connection between the silencing machinery and hypoacetylation of histones in silent domains. The Sir2, Sir3 and Sir4 proteins associate to form the SIR complex (Moazed et al 1997, Moretti et al 1994, StrahlBolsinger et al 1997). Sir2 and Sir4 appear to form a stable subcomplex that when puri¢ed contains only trace amounts of Sir3 (Ghidelli et al 2001, Hoppe et al 2002). The Sir3 and Sir4 proteins bind to GST fusions containing the H3 and H4 histone N-terminal sequences (Hecht et al 1995) and the Sir2 protein is an NAD-dependent

50

MOAZED ET AL

deacetylase whose activity is required for silencing (Imai et al 2000, Landry et al 2000, Smith et al 2000). In vitro Sir2 preferentially deacetylates histone H4 peptides that contain acetylated Lys16 compared with identical peptides with acetylated lysine at positions 5, 8, or 12 (Imai et al 2000, J. C. Tanny, D. Moazed, unpublished observations). Lys16 is the histone H4 lysine that appears to be the most important for silencing in vivo (Johnson et al 1992). Together, these results suggest that the SIR complex directly deacetylates histones within silent chromatin domains. This review focuses on recent studies that address how the deacetylated state is generated and how silencing factors and deacetylation are propagated along the chromosome ¢bre.

The deacetylase activity of Sir2 Sir2, unlike conventional deacetylases, requires a co-factor, NAD, to function. The reaction mechanism involves the transfer of acetyl groups from the substrate to an NAD cleavage product. Therefore, concomitant with deacetylation, Sir2 cleaves NAD to release nicotinamide and produces a novel compound, 2’-3’-O-acetylADP-ribose (Sauve et al 2001, Tanner et al 2000, Tanny & Moazed 2001). One of the activities that accompanies deacetylation may also contribute to the silencing function of Sir2 (Gasser & Cockell 2001, Tanner et al 2000, Tanny & Moazed 2001). In support of this idea, a recent study found that chromatin assembled with chemically acetylated histones, that had been ¢rst deacetylated by Sir2, exists in a less accessible state than chromatin assembled using histones that were deacetylated with HDAC1, a NAD-independent deacetylase (Parsons et al 2003).

Role of the deacetylase activity of Sir2 in assembly of silent chromatin The substitution of histidine 364 with tyrosine (H364Y) in the catalytic core domain of Sir2 abolishes the enzymatic activity of Sir2 in vitro and all silencing in vivo (Tanny et al 1999, Imai et al 2000). This substitution, however, has no e¡ect on the stability of Sir2 or on its ability to associate with its silencing partners, Sir4 and Net1 (Tanny et al 1999). These results indicated that the enzymatic activity of Sir2 is required for silencing and allowed us to determine the role of this activity in the assembly of silent chromatin using in vivo assays such as chromatin immunoprecipitation (ChIP). Speci¢cally, we were interested in testing whether we could distinguish a di¡erence in the requirement for Sir2 activity in binding of silencing complexes to sites that nucleate silencing, such as silencers and chromosome ends, versus sites that are distal from nucleation elements, where binding would require the spreading of silencing complexes.

HETEROCHROMATIN ASSEMBLY

51

Although it had been previously proposed that nucleation and spreading are distinct steps in the assembly process (e.g. see Braunstein et al 1996, Moazed 2001), these steps had never been experimentally separated. In our experiments, we found that the Sir2-H364Y protein bound to silencers and sites close to telomeres e⁄ciently but failed to spread to regions distal from these nucleation sites (Hoppe et al 2002, Tanny et al 1999). Similarly, Sir2 activity was required for the spreading of Sir3 and Sir4 to regions beyond nucleation sites. In contrast, at the rDNA repeats, enzymatically inactive Sir2 bound to all DNA sites with a similar e⁄ciency as wild-type Sir2 (Hoppe et al 2002). These experiments showed that Sir2 activity was speci¢cally required for the spreading of silent chromatin and for the binding of Sir3 to chromatin at the HM loci and telomeres. They also indicated that distinct steps in the assembly process could be detected in vivo (see Fig. 1). We therefore used ChIP experiments to also explore the role of the Sir3 and Sir4 proteins in the assembly process. Role of individual Sir proteins Biochemical studies had suggested a Sir2/Sir4 subcomplex associates with Sir3 to form the SIR complex. In order to determine whether the Sir proteins could associate with chromatin independently of each other, we examined the association of each Sir protein with chromatin in cells that contained deletions of the SIR2, SIR3 or SIR4 genes. An unexpected outcome of these experiments was the ¢nding that Sir4 could bind to silencers and sites close to telomeres (0.3 kb from telomeric repeats) independently of Sir2 and Sir3, but that its spreading to regions distal from nucleation sites required the other two Sir proteins (Hoppe et al 2002). Furthermore, the binding of Sir2 and Sir3 to silencers and other chromatin regions absolutely required Sir4, and partial deacetylation of histone H4 near the silencer was observed in the absence of Sir3. These experiments again established the binding of silencing proteins to nucleation sites (silencers and adjacent to telomeres) and their spreading to nearby DNA regions as separable events. Based on these experiments, we proposed a step-wise model for the assembly of silent chromatin (Fig. 1). Similar results have been obtained by the Rine and Grunstein laboratories, and they have proposed similar models (Luo et al 2002, Rusche et al 2002). Role of Sir^Sir interactions The working model presented in Fig. 1 highlights the importance of three classes of protein^protein interactions in the assembly of silent chromatin domains. Class I interactions involve the stable association of the Sir2 enzyme with Sir4 (or Net1), which targets the deacetylase activity of Sir2 to di¡erent loci (either HM/telomeric regions or rDNA). Class II interactions involve the association of the Sir3 and Sir4

52

MOAZED ET AL

HETEROCHROMATIN ASSEMBLY

53

proteins with the deacetylated N-termini of histones H3 and H4. Sir2 is required for class II interactions to take place, presumably because they require deacetylated histone tails, but other aspects of Sir2 activity may also contribute to class II interactions. Class III interactions involve homo- and heterotypic interactions between the Sir3 and Sir4 proteins. Whereas class I interactions seem static, class II and III interactions appear to be highly regulated and dynamic, and their role is central to understanding how silent chromatin is assembled and epigenetically inherited. The extreme C-terminal domain of Sir4 contains the binding site for Sir3. The X-ray crystal structure of this domain was recently solved and showed that it forms an extensive parallel coiled-coil dimer. The surface of the coiled-coil dimer contains a number of hydrophobic residues and mutagenesis studies show that substitutions of at least two di¡erent positions on the surface of the coiled-coil, M1307N and I1311N, abolish the association of the GST^Sir4 fragments with Sir3 in in vitro pull-down assays (Chang et al 2003) and the association of full-length Sir3 and Sir4 in vivo (A. D. Rudner, D. Moazed, unpublished observations). Both substitutions also completely abolish silencing. ChIP experiments from yeast cells containing the above substitutions indicate that the ability of Sir4 to associate with Sir3 is crucial for the successful recruitment of Sir3 to chromatin and demonstrate that histone deacetylation is not su⁄cient for binding of Sir3 to chromatin (A. D. Rudner, D. Moazed, unpublished observations). These ¢ndings suggest that the epigenetic code for silent chromatin in yeast is de¢ned by speci¢c protein^protein interactions involving non-histone silencing proteins as well as speci¢c histone modi¢cations (the histone code). Chromatin templates and the deacetylase activity of native Sir2 complexes We are interested in understanding how Sir2-containing silencing complexes deacetylate histones in the context of chromatin. Sir2-like enzymes expressed and FIG. 1. Model for step-wise assembly of silent chromatin in budding yeast. (A) The Sir2, Sir3, and Sir4 proteins are recruited to sites that nucleate the formation of silent chromatin by the way of interactions with site-speci¢c DNA binding proteins (step 1). After deacetylation of histones, the SIR complex is stably localized to chromatin (step 2). Sequential cycles of deacetylation coupled to association of the Sir3 and Sir4 proteins with each other as well as with the deacetylated tails of histones results in the spreading of silencing to regions beyond nucleation sites. Only the initial two rounds of recruitment are illustrated. (B) The RENT complex containing Net1, Sir2, and Cdc14 is recruited to rDNA by association with the replication fork block protein (Fob1) or RNA polymerase I (Pol I)(step1), and deacetylated histones to promote silencing (step 2). Sir2-dependent deacetylation is not required for binding of the RENT complex to rDNA, although it is required for rDNA silencing. See text for discussion and references.

54

MOAZED ET AL

puri¢ed in E. coli or insect cells can e⁄ciently deacetylate histone peptides or free histones. However, the activity of native Sir2 complexes on either free histones or native chromatin templates has not yet been studied. We puri¢ed the Sir2/Sir4 and RENT complexes from yeast using the TAP a⁄nity tag and compared their NADdependent deacetylase activity with that of Sir2 (J. Tanny, D. Moazed, unpublished observations). For substrates, we used acetylated histone peptides and hyperacetylated histones puri¢ed from butyrate-treated HeLa cells, either before or after assembly into mononucleosomes. The results of the above studies indicated that free Sir2, Sir2/Sir4, and RENT have a similar preference for histone H4 peptides that are acetylated at Lys16 but that, compared to free Sir2, the complexes have signi¢cantly lower Km values for both peptide substrates and NAD (J. Tanny, D. Moazed, unpublished observations). Furthermore, we found that the assembly of histones into nucleosomes inhibited their deacetylation by Sir2 complexes. These results show that the a⁄nity of Sir2 for its substrates is modulated by association with its silencing partners and suggest that Sir2-containing silencing complexes may require additional accessory factors in order to deacetylate nucleosomal histones in vivo. Finally, in contrast to a recent study suggesting that RENT contained NAD-independent deacetylase activity (Ghidelli et al 2001), we found that the deacetylase activity of RENT was entirely NAD-dependent (J. Tanny, D. Moazed, unpublished observations).

Concluding remarks The step-wise assembly model presented in Fig. 1 provides a general mechanism for the spreading of silencing factors on chromatin. In this model, after recruitment of the Sir2/Sir4 complex to DNA (Fig. 1A, step 1), Sir2-dependent deacetylation of histones provides a binding site for the SIR complex on chromatin (Fig. 1B, step 2). Sequential rounds of deacetylation coupled to homo- and heterotypic interactions involving Sir3 and Sir4 then results in the recruitment of additional SIR complexes and spreading of these complexes along nucleosomal DNA (Fig. 1A, step 3). Although in budding yeast spreading is driven by deacetylation, the basic features of this model can be generalized to metazoans and ¢ssion yeast, where both deacetylation and histone methylation are required for heterochromatin formation (e.g. see Grewal & Elgin 2002). Spreading of silencing factors from nucleation sites to nearby DNA regions provides a mechanism for transition from sequence-speci¢c genetic control to sequence non-speci¢c epigenetic control. Thus, we believe that the step-wise spreading model presented here provides an explanation for one of the most fundamental properties of heterochromatin.

HETEROCHROMATIN ASSEMBLY

55

Acknowledgements This work was supported by grants from the NIH, the Ellison Medical Foundation, and the Stewart Family Trust (DM). ADR is a fellow of the Jane Co⁄n Childs Memorial Fund. DM is a Leukemia and Lymphoma Society Scholar.

References Aparicio OM, Billington BL, Gottschling DE 1991 Modi¢ers of position e¡ect are shared between telomeric and silent mating-type loci in S. cerevisiae. Cell 66:1279^1287 Braunstein M, Rose AB, Holmes SG, Allis CD, Broach JR 1993 Transcriptional silencing in yeast is associated with reduced nucleosome acetylation. Genes Dev 7:592^604 Braunstein M, Sobel RE, Allis CD, Turner BM, Broach JR 1996 E⁄cient transcriptional silencing in Saccharomyces cerevisiae requires a heterochromatin histone acetylation pattern. Mol Cell Biol 16:4349^4356 Bryk M, Banerjee M, Murphy MK, Nudsen KE, Gar¢nkel DJ, Curcio MJ 1997 Transcriptional silencing of Ty1 elements in the RDN1 locus of yeast. Genes Dev 11:255^269 Chang JF, Hall BE, Tanny JC, Moazed D, Filman D, Ellenberger T 2003 Structure of the coiledcoil dimerization motif of Sir4 and its interaction with Sir3. Structure 11:637^649 Gartenberg MR 2000 The Sir proteins of Saccharomyces cerevisiae: mediators of transcriptional silencing and much more. Curr Opin Microbiol 3:132^137 Gasser SM, Cockell MM 2001 The molecular biology of the SIR proteins. Gene 279:1^16 Ghidelli S, Donze D, Dhillon N, Kamakaka RT 2001 Sir2p exists in two nucleosome-binding complexes with distinct deacetylase activities. EMBO J 20:4522^4535 Grewal SI, Elgin SC 2002 Heterochromatin: new possibilities for the inheritance of structure. Curr Opin Genet Dev 12:178^187 Hecht A, Laroche T, Strahl-Bolsinger S, Gasser SM, Grunstein M 1995 Histone H3 and H4 N-termini interact with SIR3 and SIR4 proteins: a molecular model for the formation of heterochromatin in yeast. Cell 80:583^592 Hoppe GJ, Tanny JC, Rudner AD et al 2002 Steps in assembly of silent chromatin in yeast: sir3independent binding of a Sir2/Sir4 complex to silencers and role for sir2-dependent deacetylation. Mol Cell Biol 22:4167^4180 Huang J, Moazed D 2003 Association of the RENT complex with nontranscribed and coding regions of rDNA and a regional requirement for the replication fork block protein Fob1 in rDNA silencing. Genes Dev 17:2162^2176 Imai S, Armstrong CM, Kaeberlein M, Guarente L 2000 Transcriptional silencing and longevity protein Sir2 is an NAD-dependent histone deacetylase. Nature 403:795^800 Jenuwein T, Allis CD 2001 Translating the histone code. Science 293:1074^1080 Johnson LM, Fisher-Adams G, Grunstein M 1992 Identi¢cation of a non-basic domain in the histone H4 N-terminus required for repression of the yeast silent mating loci. EMBO J 11:2201^2209 Kayne PS, Kim UJ, Han M, Mullen JR, Yoshizaki F, Grunstein M 1988 Extremely conserved histone H4 N terminus is dispensable for growth but essential for repressing the silent mating loci in yeast. Cell 55:27^39 Klar AJS, Fogel S, MacLeod K 1979 MAR1  a regulator of HMa and HMa loci in Saccharomyces cerevisiae. Genetics 93:37^50 Landry J, Sutton A, Tafrov ST et al 2000 The silencing protein SIR2 and its homologs are NADdependent protein deacetylases. Proc Natl Acad Sci USA 97:5807^5811 Luo K, Vega-Palas MA, Grunstein M 2002 Rap1-Sir4 binding independent of other Sir yKu or histone interactions initiates the assembly of telomeric heterochromatin in yeast. Genes Dev 16:1528^1539

56

DISCUSSION

Moazed D 2001 Common themes in mechanisms of gene silencing. Mol Cell 8:489^498 Moazed D, Kistler A, Axelrod A, Rine J, Johnson AD 1997 Silent information regulator protein complexes in Saccharomyces cerevisiae: a SIR2/SIR4 complex and evidence for a regulatory domain in SIR4 that inhibits its interaction with SIR3. Proc Natl Acad Sci USA 94:2186^2191 Moretti P, Freeman K, Coodly L, Shore D 1994 Evidence that a complex of SIR proteins interacts with the silencer and telomere-binding protein RAP1. Genes Dev 8:2257^2269 Parsons XH, Garcia SN, Pillus L, Kadonaga JT 2003 Histone deacetylation by Sir2 generates a transcriptionally repressed nucleoprotein complex. Proc Natl Acad Sci USA 100:1609^1614 Rine J, Herskowitz I 1987 Four genes responsible for a position e¡ect on expression from HML and HMR in Saccharomyces cerevisiae. Genetics 116:9^22 Rusche LN, Kirchmaier AL, Rine J 2002 Ordered nucleation and spreading of silenced chromatin in Saccharomyces cerevisiae. Mol Biol Cell 13:2207^2222 Rusche LN, Kirchmaier AL, Rine J 2003 The establishment inheritance and function of silenced chromatin in Saccharomyces cerevisiae. Annu Rev Biochem 27:27 Sauve AA, Celic I, Avalos J, Deng H, Boeke JD, Schramm VL 2001 Chemistry of gene silencing: the mechanism of NAD+-dependent deacetylation reactions. Biochemistry 40:15456^15463 Shou W, Seol J, Shevchenko A et al 1999 The termination of telophase is triggered by Tem1dependent release of the protein phosphatase Cdc14 from the nucleolar RENT complex. Cell 97:233^244 Smith JS, Boeke JD 1997 An unusual form of transcriptional silencing in yeast ribosomal DNA. Genes Dev 11:241^254 Smith JS, Brachmann CB, Celic I et al 2000 A phylogenetically conserved NAD+-dependent protein deacetylase activity in the Sir2 protein family. Proc Natl Acad Sci USA 97:6658^6663 Strahl-Bolsinger S, Hecht A, Luo K, Grunstein M 1997 SIR2 and SIR4 interactions di¡er in core and extended telomeric heterochromatin in yeast. Genes Dev 11:83^93 Straight AF, Shou W, Dowd GJ et al 1999 Net1 a Sir2-associated nucleolar protein required for rDNA silencing and nucleolar integrity. Cell 97:245^256 Tanner KG, Landry J, Sternglanz R, Denu J 2000 Silent information regulator 2 family of NAD-dependent histone/protein deacetylases generates a unique product 1-O-acetyl-ADPribose. Proc Natl Acad Sci USA 97:14178^14182 Tanny JC, Moazed D 2001 Coupling of histone deacetylation to NAD breakdown by the yeast silencing protein Sir2: evidence for acetyl transfer from substrate to an NAD breakdown product. Proc Natl Acad Sci USA 98:415^420 Tanny JC, Dowd GJ, Huang J, Hilz H, Moazed D 1999 An enzymatic activity in the yeast Sir2 protein that is essential for gene silencing. Cell 99:735^745 Turner BM 2000 Histone acetylation and an epigenetic code. Bioessays 22:836^845

DISCUSSION Verdin: In this system is there any evidence for RNAs, similar to the ones that have been shown at the other regions, being involved in setting up the silencing mark? Moazed: Whether RNA is involved or not is not clear. The RNAi system is de¢nitely not because the RNAi pathway does not seem to be conserved in budding yeast. There is a suggestion based on data from Je¡ Smith’s lab at the University of Virginia that ribosomal RNAs themselves may be involved. That’s because RNA polymerase I mutant cells are defective for rDNA silencing, but it is

HETEROCHROMATIN ASSEMBLY

57

not possible right now to distinguish whether ribosomal RNAs or the polymerase itself is important. Allis: What is the business end of Sir3? Moazed: It is between the C-terminal and middle of the protein where the binding sites for Sir4 and also the histone tails are located. Sir3 also has a very conserved N-terminal domain called the BAH domain that is found in a lot of chromosome-associated proteins. There is some speculation that this domain might also be involved in binding histones. Allis: Is that BAH domain a bromo-associated homology domain? Moazed: Yes, it was initially found in bromodomain proteins. Allis: Is there any evidence that this domain itself binds the H4 tail? Moazed: No. Allis: It would be a dream come true if it bound the H4 tail but only when Lys16 was deacetylated. That would put a lot of pieces of the puzzle together. Atadja: Are complexes analogous to those you’ve described formed in mammalian cells? Moazed: In mammalian cells there are de¢nitely similar types of silencing complexes. The histone-binding protein HP1 co-precipitates with the methyltransferase enzyme, Suvar39h. It is not clear whether this is a direct interaction or whether there are other components of the complex. The function of Sir2 itself may be conserved. Even though S. pombe has an entirely di¡erent silencing system that requires methylation, S. pombe has a Sir2-like protein. In collaboration with Shiv Grewal (Cold Spring Harbor) we have shown that the S. pombe Sir2 protein is absolutely essential for silencing. It acts upstream of the Clr4 methyltransferase and is required for H3-K9 methylation. It is probably the H3-K9 deacetylase. The case for Drosophila and human cells is more ambiguous right now. Cole: I have a biochemical question. You said early on the Sir2/3/4 is not such a stable complex when you have tried to isolate it, but you see pretty stable interactions by Sir3 and Sir4 by these pull downs. Is it conceivable that your use of GST, which can dimerize, could be giving some misleading results? Moazed: No. We see a very stable Sir complex by a variety of di¡erent methods in yeast extracts. It accounts for about 50% of the total Sir2/3/4 proteins. This complex is stable to 2M salt. We don’t understand what regulates its formation. It may only get formed on chromatin but that is not clear right now. If we put pure full-length Sir2, 3 and 4 together, we don’t see any complex formation. The issue with the truncations is that Sir4 may have a domain that inhibits its binding to Sir3, which is removed in the truncations. Cole: Have you tried doing any binding without the GST present? Moazed: Yes. We get the same results when we co-IP untagged Sir proteins or when we use other tags.

58

DISCUSSION

Marmorstein: I may have missed this, but did you say that the Sir2/4/5 complex can’t bind chromatin, or is it only the 2/4 that can’t bind? Moazed: Our ChIP experiments show that the Sir2/4 complex can bind chromatin in the absence of Sir3, but it only binds to sites that nucleate silencing, that’s silencers and very close to telomeric repeats. Marmorstein: I thought you got no deacetylation of the 2/4 on chromatin, so it can bind but can’t deacetylate. Moazed: It binds to the silencer and it deacetylates nucleosomes very close to the silencer, but that deacetylation does not spread. You need Sir3 to recruit additional complexes and spread the modi¢cations. But in the absence of Sir3 just the 2,4 complex by itself binds the nucleation site and we see partial deacetylation of histones that correlates with this binding. Li: So in the absence of histone methylation in S. cerevisiae, do you feel the silencing is less stable compared with S. pombe? Moazed: Silencing in S. cerevisiae may be less stable, especially at telomeric regions, but there are di¡erences other than methylation between silencing in S. cerevisiae and S. pombe. The histone binding proteins are di¡erent. I’m not sure whether we know why the silencing is unstable and what the contribution of histone methylation might be. Atadja: When you say it is ‘unstable’, are you talking about loci such as the silent mating locus? Moazed: Silencing at the mating type loci is very stable. I was thinking about telomeric silencing, which switches back and forth between the on and o¡ states fairly frequently. Castronovo: I have a more general question. How do you integrate silencing chromatin to acetylation/deacetylation and methylation? Is it an integrated process, or is it one or the other? How do these two phenomena cooperate to silence chromatin? Moazed: With regards to methylation I have to point out that in S. cerevisiae all the histone methylation events that have been identi¢ed seem to be associated with gene activation. They mark euchromatin and are absent from silent chromatin domains. One possibility is that there are competing assembly processes. Obviously, the factors involved in silencing are recruited to DNA and can assemble a structure that is inaccessible to methyltransferases, like SET1, which are involved in transcriptional activation. Turner: Hasn’t H3-K4 methylation been associated with rDNA silencing? Moazed: There is some evidence for that, but I am not sure whether the role of H3-K4 methylation in rDNA silencing is direct. For example, ChIP experiments have failed to detect H3-K4 methylation in rDNA or other silent chromatin regions. In cerevisiae there is a competition between the assembly of euchromatin and heterochromatin. If the inhibitory marks, such as H3-K4 and H3-K79

HETEROCHROMATIN ASSEMBLY

59

methylation, are removed from euchromatic domains, this results in a defect in hetrochromatin assembly. That is because the Sir proteins bind non-speci¢cally throughout the genome and become limiting for assembly. The case for rDNA is more di⁄cult to address because H3-K4 methylation is associated with transcriptional activity and rDNA is very highly transcribed. H3-K4 may become methylated because the locus is highly transcribed. Atadja: Are there interactions between the Rpd3 system and the Sir2/3 systems? Moazed: That’s a good question. The Rpd3 system is required for assembly of heterochromatin in many organisms, but in S. cerevisiae it is not. If you delete Rpd3 in S. cerevisiae you get better silencing. The major deacetylase required for silencing in S. cerevisiae is Sir2. Verdin: There is reported evidence that other HDACs are recruited to similar chromosomal sites as Sir2. For example, Hst4 is recruited to the telomere and centromere, while HOS1 and HOS3 are recruited to the ribosomal genes. What do you think these add to Sir2? Moazed: There is no evidence that deletion of HOS1/2 interferes with rDNA silencing, although that is a good experiment. With Hst3 and 4 the situation is more complicated. The double mutant that has a silencing defect also has severe growth defects. It activates the DNA damage checkpoints. Every time this happens it disassembles the Sir complexes. So, some of the e¡ects of Hst3 and 4 mutants may be indirect. Li: Is Ser10 of H3 phosphorylated in S. cerevisiae? Moazed: Yes. Allis: It has a meiotic, mitotic and transcriptional phenotype, as well. Li: On the other hand you don’t have H3-K9 methylation. Allis: Absolutely correct, not in S. cerevisiae. Li: This would challenge your model about the binary code. Allis: Except there would be nothing to bind on the H3-K9 methyl mark, so in this case Serine10 has a di¡erent role. Turner: For your in vitro chromatin assembly experiments in which you showed there was an inhibitory e¡ect on Sir2 deacetylase, what was the source of the acetylated histones that you used? Moazed: The deacetylases were puri¢ed from butyrate-treated HeLa cells. Turner: Were they labelled in vivo? Moazed: The histones were not labelled, because we were looking at deacetylation either by Western blotting or nicotinamide release. Because the deacetylation is coupled to nicotinamide release we can use unlabelled histones. Turner: The reason I asked is because if they are in vivo HeLa histones they will primarily be labelled at K16, which is actually quite close to the DNA.

60

DISCUSSION

Moazed: This is the major substrate for Sir2, as well. It is the site that is most important for silencing. Turner: Is it? In that original paper (Imai et al 2000) they didn’t test a lot of peptides. It was active against a K16 peptide, but they didn’t do a comprehensive test. Moazed: The e¡ect is pretty dramatic. Turner: It is a very dramatic e¡ect. K16 is quite close to the DNA, so if there is going to be an e¡ect you would expect to see it with K16. You may have other evidence, but the original evidence in the Imai et al (2000) paper for K16 speci¢city was not strong. Moazed: We have evidence that there is K16 speci¢city on histone peptides. But if you look at deacetylation of a mixture of histones, almost every site is deacetylated with long incubation times. Verdin: When we isolate histones from HeLa cells in response to histone deacetylase inhibitors, we ¢nd that most histones become hyperacetylated on many residues, and not just K16. Turner: Were they butyrate treated histones? Verdin: Yes. Turner: K16 will be the most common. Denu: I’d like to probe this in more detail. It is possible that the substrates that you are o¡ering within the nucleosome represent a more arti¢cial system than yeast Sir2 would see at the silent loci. Is it possible all those modi¢cations that were due to butyrate-induced hyperacetylation are very poor substrates for Sir2 within the context of the nucleosome, since those would not be the likely physiological acetylation targets seen by Sir2 in vivo? It is a di⁄cult question. To address this question, one has to have a pure substrate that is speci¢cally modi¢ed in accordance with what is found at these silent loci. Moazed: There must be two types of substrates. One is the histones that Sir2 sees immediately after DNA replication during de novo assembly of silent chromatin. These histones will have some of the modi¢cations of an active chromatin domain. In S. cerevisiae you can also get chromatin assembly in cells that are arrested in the cell cycle. You can get assembly in the absence of DNA replication. I am not sure whether I am really getting to your question. Denu: I think it is an intriguing result, and I am ba¥ed that it doesn’t work at all. Turner: I was intrigued by the fact that it doesn’t work, too. We have found something similar with HDAC1/HDAC2 in mammals. If you test these against nucleosomal substrates you get very ine⁄cient deacetylation unless you add ATP, in which case it is much more e⁄cient. Moazed: Are you suggesting that it could be one of the heatshock proteins that co-puri¢es with HDACs in your experiments?

HETEROCHROMATIN ASSEMBLY

61

Turner: We have taken this further. We have been able to show that if the heatshock protein is not there, then you don’t get the ATP e¡ect. Moazed: We know that we have ATP in this system. Addition of ATP to this system doesn’t change anything. The only thing that has helped so far is DNase. Gu: You proposed that acetyl ADP ribose is a product of Sir2-dependent deacetylation. Has this been detected in vivo? Moazed: Not by us, but John Denu’s lab have reported this. Denu: We have preliminary evidence. Gu: Is it stable? Denu: There are enzymes in the cell that metabolize it. Gu: Is there an e¡ect if you treat the cell with acetylated ADP ribose? Denu: We didn’t douse cells with acetylated ADP ribose, but we have done a developmental assay using echinoderms. When we microinject acetylated ADP ribose it can have an e¡ect. It blocks cell division and oocyte maturation. It is di⁄cult to know whether this is physiologically relevant. This particular experiment was set up just to establish whether or not acetylated ADP ribose has bioactivity. We have unpublished preliminary results showing that it does exist in cells. Allis: Is there any HAT that might be able to use the acetyl group from that? This would be very interesting. Denu: In a number of mammalian cell lines we have detected an acetyltransferase that does use acetylated ADP ribose. The problem is that we don’t know what it is being transferred to. It doesn’t appear to be histones. Gu: Does yeast Sir2 form dimers? Mammalian Sir2 can form dimers. Moazed: The evidence for yeast Sir2 is ambiguous. Sir3 and Sir4 do form dimers. Khochbin: What do you think about the role of histone variants on preventing the spreading of this silencing? Verdin: Could you broaden this question to discuss all the mechanisms that are thought to limit the spreading of silencing? Moazed: There are at least two mechanisms that limit the spreading of heterochromatin in S. cerevisiae. One mechanism involves speci¢c boundary elements, which are usually strong promoters. Here competition between transcriptional activation, and perhaps histone modi¢cations that are incompatible with silencing, limits the spreading. The second type of mechanism is a more general competition between euchromatin and heterochromatin. It seems like every time there is a mutation that weakens euchromatin, there is an increase in the spreading of silent domains. It may be that when euchromatin is weakened, it presents less of a barrier to the spreading of silencing factors. With regards to histone variants, the H2A.Z variant, which is usually associated with active chromatin, appears to be part of an anti-silencing mechanism that stops the spreading of silent chromatin.

62

DISCUSSION

Khochbin: Can the Sir complex bind nucleosomes containing histone H2A.Z variant? Moazed: That is a very good question. We’d like to do this experiment. If I were to guess, I would say it will. Reference Imai S, Armstrong CM, Kaeberlein M, Guarente L 2000 Transcriptional silencing and longevity protein Sir2 is a NAD-dependent histone deacetylase. Nature 17:795^800

H2B ubiquitylation and de-ubiquitylation in gene activation Anastasia Wyce, Karl W. Henry and Shelley L. Berger1 The Wistar Institute, Gene Expression and Regulation Program, 3601 Spruce Street, Philadelphia, PA 19104, USA

Abstract. Previous models for the role of histone modi¢cations suggest that adding and removing modi¢cations, such as acetylation/deacetylation in gene regulation, are functionally antagonistic. We have investigated a transcriptional role of H2B Cterminal ubiquitylation and de-ubiquitylation in Saccharomyces cerevisiae. H2B ubiquitylation is required for optimal transcription of SUC2 and GAL1 genes. The ubiquitin hydrolase Ubp8 is a stable component of SAGA but not ADA complexes, and is not required for overall integrity of SAGA. Biochemical and genetic evidence indicates that Ubp8 targets H2B for deubiquitylation. The dynamic balance of H2B ubiquitylation/deubiquitylation is important for GAL1 transcription since either substitution of the ubiquitylation site in H2B (Lys123), or loss of Ubp8, lowers GAL1 expression. Further, this balance of ubiquitylation appears to set the balance of histone H3 methylation at Lys4 relative to Lys36. Thus, unlike acetylation/deacetylation whose functions are mutually opposing, both ubiquitylation and de-ubiquitylation are required for gene activation. These results suggest that ubiquitylation of histones has a unique role among histone modi¢cations, possibly to orchestrate an ordered pathway of chromatin alterations. 2004 Reversible protein acetylation. Wiley, Chichester (Novartis Foundation Symposium 259) p 63^77

Over the past decade, intensive study of covalent histone modi¢cations has yielded a wealth of information regarding the role of these modi¢cations in the regulation of chromatin structure and transcription. Speci¢c enzymes and target residues are now known for lysine acetylation, serine phosphorylation, arginine and lysine methylation, and lysine ubiquitylation (Berger 2002). The current model for the role of histone modi¢cations has been shaped by the role of histone acetylation, wherein acetylation activates transcription and deacetylation is generally

1This

paper was presented at the symposium by Shelley L. Berger to whom correspondence should be addressed. 63

64

WYCE ET AL

repressive (Kuo & Allis 1998). This transcriptional e¡ect occurs through alternative recruitment of co-activator or co-repressor complexes (containing the relevant enzymatic activities) by DNA-bound activator or repressor proteins, respectively (Berger 2002). In general, it is believed that the enzymatic addition or removal of a modi¢cation has antagonistic e¡ects. This model appears to hold true at least for acetylation (by acetyltransferases/deacetylases) and phosphorylation (by kinases/phosphatases), both of which are dynamic and have a role in transcription regulation (Kuo & Allis 1998). In contrast, lysine methylation on histones appears to be quite stable and thus may not be removed by an opposing enzyme, but rather lost due to proteolysis or protein turnover (Bannister et al 2002). As described in this report, we have uncovered another potential paradigm in our recent examination of histone ubiquitylation in gene activation. In the case of ubiquitylation, our data suggests that the sequence of adding and then removing ubiquitin (ub) from histone H2B is required for full transcriptional activation (Henry et al 2003). Ubiquitin is a highly conserved 76 amino acid protein initially associated with marking cellular components for proteasome-dependent degradation (Pickart 2001a). At least three enzymes are required for the protein ubiquitylation: The E1 ubiquitin-activating enzyme, the E2 ubiquitin-conjugating enzyme, and the E3 ubiquitin ligase enzyme. The E1 enzyme activates and covalently binds the terminal Gly76 residue of ubiquitin to its active site. Ubiquitin is then transferred to an E2 enzyme that, following association with an E3 enzyme (which determines substrate speci¢city), ubiquitylates a lysine residue on the target protein. In the case of polyubiquitylation, the speci¢c residue that is utilized for ubiquitin linkage determines the fate of the target protein. Addition of a new ubiquitin moiety to Lys48 of a previous ubiquitin is correlated with targeting for proteasomal degradation (Pickart 2001a,b). Use of Lys63 instead of Lys48 in polyubiquitin chain formation is associated with cellular signalling processes. Mono-ubiquitylation of proteins also exists, and like Lys63 linkage, has functions other than degradation such as protein sorting, endocytosis and gene regulation (Pickart 2001b). For over a decade, it has been known that ubiquitylation of histones (monoubiquitylation of H2B and poly-ubiquitylation of H2A) occurs and is associated, at least in higher eukaryotes, with transcriptionally active regions of DNA (Nickel et al 1989, Davie & Murphy 1990, Davie et al 1991, M. A. Osley, personal communication). Recently, it was found in Saccharomyces cerevisiae that ubiquitylation of H2B occurs on Lys123, mediated by the enzymes Rad6 (E2) and Bre1 (E3) (Robzyk et al 2000, Wood et al 2003, Hwang et al 2003). While histone ubiquitylation is associated with gene activation in higher eukaryotes, ub-H2B in S. cerevisiae was recently found to be involved in cell cycle progression and telomeric silencing (Robzyk et al 2000, Sun & Allis 2002). Additionally, the

H2B UBIQUITYLATION

65

ubiquitylation of histone H1 by TAF250 has also been observed and may function in the transcriptional activation of certain genes (Pham & Sauer 2000). Ubiquitin-dependent gene regulation is not limited to histone ubiquitylation. Other processes, both degradative and non-degradative, can be utilized (reviewed in Conaway et al 2002, Marx 2002, Pickart 2001b). The release of activators from non-nuclear sites, removal of inhibitory motifs, or degradation of cofactors are all known ubiquitin-dependent processes that a¡ect transcription (Ostendor¡ et al 2002, Pickart 2001b, Salghetti et al 2001). Additionally, there are other processes, such as recruitment of 19S proteasome subunits to activated promoters or the ubiquitylation of the Spt7 SAGA subunit, that require the ubiquitin pathway. However, their speci¢c role(s) in transcription remain unknown (Gonzalez et al 2002, Saleh et al 1998). Our current research focuses on the role of H2B ubiquitylation in the regulation of transcription in S. cerevisiae. Based on the correlation between histone ubiquitylation and gene activation in mammalian cells, we examined the role of H2B ubiquitylation in the activation of transcription in yeast. We found that a substitution mutation of the target residue (Lys123 to Arg; K123R) in H2B lowered transcription of several well-characterized, inducible genes. One example is the SUC2 gene, which encodes a protein involved in sucrose catabolism that is induced under low glucose conditions. Quantitative analysis of RNA levels indicated that the level of transcription dropped approximately twofold in the K123R mutant relative to wild-type (Fig. 1). These results suggested that ubiquitylation of H2B is involved in gene activation. Other genes that are a¡ected by the K123R mutation are GAL1, ADH2 and PHO5 (Henry et al 2003, C.-F. Kao, M. A. Osley, personal communication). We then developed a double chromatin immunoprecipitation (double ChIP) assay to directly examine ub-H2B at the level of the gene. We found, examining a time course during activating conditions, that the level of ub-H2B at the GAL1 promoter increased early during induction, but then decreased well before stable RNA had accumulated to a high level (Henry et al 2003). The reduction of ub-H2B at the GAL1 promoter prior to signi¢cant RNA accumulation suggested that there may be a requisite step of H2B deubiquitylation prior to full transcriptional activation. There are 17 putative ubiquitin proteases in S. cerevisiae. We focused on Ubp8 as a candidate ubiquitin protease for H2B because it was reported to associate with the SAGA (Spt-AdaGcn5-Acetyltransferase) histone acetylation complex (Gavin et al 2002, Ho et al 2002, Sanders et al 2002). SAGA is involved in transcriptional activation, in part through its histone acetyltransferase, Gcn5. Moreover, each of the Lys123dependent genes was previously shown to be SAGA-dependent for full activation (Pollard & Peterson 1997, Dudley et al 1999, Bhaumik & Green 2001, Chiang et al 1996, Barbaric et al 2003). We successfully con¢rmed the physical

66

WYCE ET AL

FIG. 1. SUC2 transcription is reduced by loss of the H2B ubiquitylation site (K123R). S1 nuclease protection assay of RNA isolated from wild type (WT), htb1-K123R, and gcn5D strains. RNA was collected from cells grown in non-inducing (
¼ 2% glucose) or inducing (+ ¼ 0.05% glucose) conditions for SUC2. RNA was visualized using a radiolabelled probe complementary to the 5’ end of SUC2 RNA, and quantitation was performed by phosphoimaging. tRNA levels were used as loading controls. The fold increase over noninducing conditions is shown for each strain. Mutation of the ubiquitylation site (K123R) reduces transcription approximately threefold, which is similar to the reduction observed upon deletion of GCN5.

association between Ubp8 and SAGA using protein a⁄nity puri¢cation and ion exchange chromatography to purify the SAGA complex (Henry et al 2003). We then found, using chromatin immunoprecipitation (ChIP), that Ubp8 is recruited to the GAL1 promoter with kinetics similar to Gcn5 (Henry et al 2003). We next tested the ability of Ubp8 to de-ubiquitylate H2B in vivo and in vitro. In a wild-type strain, we detected ub-H2B in bulk histones. As expected, this ubiquitylated species was absent in the K123R strain. Elevated levels of ub-H2B were found in the ubp8D strain, suggesting that ub-H2B is indeed an in vivo target of Ubp8 (Fig. 2). We also developed an in vitro H2B deubiquitylation assay. Using this assay, we

H2B UBIQUITYLATION

67

FIG. 2. H2B ubiquitylation increases in vivo in the absence of Ubp8. Strains containing FLAGtagged H2B in a wild-type or mutant background were subjected to anti-FLAG immunoprecipitation followed by FLAG Western blot to visualize levels of ub-H2B. Identi¢cation of ub-H2B was con¢rmed by ubiquitin Western (data not shown). Strains used were wild-type (WT), htb1-K123R, ubp8D and the htb1-K123R/ubp8D double mutant.

observed that SAGA puri¢ed from a wild-type strain was able to deubiquitylate ub-H2B in vitro, but SAGA isolated from a ubp8D strain lacked this activity (Fig. 3). Finally, using the double ChIP assay, we found that loss of Ubp8 led to increased levels of ub-H2B at the GAL1 promoter (Henry et al 2003). Taken together, these results show that H2B ubiquitylation is dynamic during gene activation and that Ubp8, functioning within the SAGA complex, de-ubiquitylates histone H2B. Since we had observed that H2B ubiquitylation has a role in SUC2 and GAL1 activation, we tested the transcriptional e¡ect of increased ub-H2B levels in the strain lacking Ubp8. Induction of the SUC2 gene was lower in the ubp8D strain compared to wild-type (Fig. 4). This was also the case for the GAL1 and ADH2 genes (Henry et al 2003). Taken together with the data from the K123R strain, our results suggest that both the addition and removal of ubiquitin are required for full gene activation, in contrast to the antagonistic e¡ects of acetylation and deacetylation. Since Ubp8 exists within the SAGA complex, we also assayed the role of the acetyltransferase Gcn5 in gene activation relative to Ubp8. Deletion of GCN5 resulted in a transcriptional defect more severe than that of ubp8D at the GAL1 and ADH2 genes. Transcription of these genes was further reduced in a gcn5D/ubp8D double mutant strain (Henry et al 2003). Similarly, mutation of the ubiquitylation site (K123R) combined with loss of Gcn5 also produced synthetic

68

WYCE ET AL

FIG. 3. H2B de-ubiquitylation in vitro by SAGA requires Ubp8. SAGA was puri¢ed from wild-type (SAGA) and ubp8D (SAGA ubp87) strains, and equivalent amounts of each complex were incubated with ub-H2B and H2B. The ub-H2B/H2B substrate was obtained by anti-FLAG immunoprecipitation from a ubp8D strain expressing HA-ub and FLAG-H2B. The mock-treated lane contained no enzyme. Western blotting was performed with FLAG (to detect FLAG-H2B and ub-H2B) and HA (to detect ub-H2B and cleaved ubiquitin) antibodies. The asterisk indicates a background band of unknown origin.

defects (C.-F. Kao, M. A. Osley, personal communication). These observations suggest that acetylation and ubiquitylation/de-ubiquitylation may work together to activate transcription at these genes. We then explored the molecular basis for the requirement for H2B ubiquitylation/deubiquitylation in gene activation. It was previously established that H2B ubiquitylation is required for histone H3 methylation on Lys4, but not on Lys36 (Sun & Allis 2002, Briggs et al 2002, Dover et al 2002). Both Lys4 and Lys36 methylation have been linked to RNA Polymerase II (Pol II) elongation, and both of the methyltransferases (Set1 for Lys4 and Set2 for Lys36) physically associate with RNA Pol II (Krogan et al 2003a,b, Ng et al 2003, Xiao et al 2003). It appears that Lys4 methylation may be involved in an early step in RNA Pol II elongation, whereas Lys36 methylation may be involved in a later step of RNA

H2B UBIQUITYLATION

69

FIG. 4. SUC2 transcription is reduced by loss of Ubp8. S1 nuclease protection assay was performed as in Fig. 1 using wild-type and ubp8D strains. Fold increase over non-inducing conditions is shown for each strain.

synthesis. Thus, we hypothesized that the sequence of ubiquitylation/deubiquitylation may be required for a transitional step from Lys4 methylation to Lys36 methylation. We tested this possibility by ChIP assay at the GAL1 promoter, using antibodies to me-Lys4 and me-Lys36. me-Lys4 was decreased when H2B could not be ubiquitylated, as is the case in the K123R strain. The opposite e¡ect was observed in the ubp8D strain, in which ub-H2B levels are higher. In contrast, meLys36 was increased by loss of the ubiquitylation site (K123R) and decreased by the absence of deubiquitylation (ubp8D). Thus, ubiquitylation/deubiquitylation may function to establish the correct balance between the levels of H3 Lys4 versus Lys36 methylation, perhaps by interfering with the recruitment or activity of the enzymes (Set1 or Set2) necessary to produce these modi¢cations. Our data suggest that ubiquitylation and de-ubiquitylation of H2B are both involved in transcriptional activation, functioning to orchestrate an ordered

70

WYCE ET AL

FIG. 5. Model for the interplay of H2B ubiquitylation/de-ubiquitylation with other histone modi¢cations. Solid circle represents repressive chromatin, dotted circle represents poised chromatin, and open circle represents active chromatin. TATA is the TBP binding sequence upstream of the ORF (gene). The enzymes and abbreviations for histone modi¢cations are the same as in the text.

H2B UBIQUITYLATION

71

sequence of chromatin modi¢cations at a regulated promoter. We propose a model, shown in Fig. 5, for this sequence of histone modi¢cations in the activation of SAGA-dependent genes. During the ‘off’ state the promoter and gene are in a repressed chromatin state (Panel A). Early in activation H2B is ubiquitylated by Rad6/Bre1, which triggers Set1-mediated H3 Lys4 methylation (Panel B). The next step is the recruitment of SAGA, which leads to Gcn5mediated H3 acetylation (Panel C). This is followed by Ubp8-mediated deubiquitylation of H2B, which then slows H3 Lys4 methylation and triggers the switch to Lys36 methylation (Panel D). These events ultimately lead to productive elongation by Pol II (Panel E). To date we have analysed only highly inducible, SAGA-dependent genes, but other genes may be regulated by H2B ubiquitylation and de-ubiquitylation. In addition, we have not yet determined whether there are additional ubiquitylated substrates acted upon by Ubp8. Ubp8 may also be a component of complexes other than SAGA, and in that context may target additional substrates. Future studies will determine the extent to which H2B ubiquitylation and the de-ubiquitylation activity of Ubp8 are utilized by the cell to regulate transcription. Acknowledgements Research is supported by research grants from the NIH (GM55360) and NSF (MCB-0078940) to S.L.B. NIH training grants supported A.W. (GM007229) and K.W.H. (CA09171, as well as an NRSA to K.W.H. (F32 GM069207).

References Bannister AJ, Schneider R, Kouzarides T 2002 Histone methylation: dynamic or static? Cell 109:801^806 Barbaric S, Reinke H, Horz W 2003 Multiple mechanistically distinct functions of SAGA at the PHO5 promoter. Mol Cell Biol 23:3468^3576 Bhaumik SR, Green MR 2001 SAGA is an essential in vivo target of the yeast acidic activator Gal4p. Genes Dev 15:1935^1945 Berger SL 2002 Histone modi¢cations in transcriptional regulation. Curr Opin Genet Dev 12:142^148 Briggs SD, Xiao T, Sun ZW et al 2002 Gene silencing: trans-histone regulatory pathway in chromatin. Nature 418:498 Chiang YC, Komarnitsky P, Chase D, Denis CL 1996 ADR1 activation domains contact the histone acetyltransferase GCN5 and the core transcriptional factor TFIIB. J Biol Chem 271:32359^32365 Conaway RC, Brower CS, Conaway JW 2002 Emerging roles of ubiquitin in transcription regulation. Science 296:1254^1258 Davie JR, Murphy LC 1990 Level of ubiquitinated histone H2B in chromatin is coupled to ongoing transcription. Biochemistry 29:4752^4757 Davie JR, Lin R, Allis CD 1991 Timing of the appearance of ubiquitinated histones in developing new macronuclei of Tetrahymena thermophila. Biochem Cell Biol 69:66^71

72

WYCE ET AL

Dover J, Schneider J, Tawiah-Boateng MA et al 2002 Methylation of histone H3 by COMPASS requires ubiquitination of histone H2B by Rad6. J Biol Chem 277:28368^28371 Dudley AM, Rougeulle C, Winston F 1999 The Spt components of SAGA facilitate TBP binding to a promoter at a post-activator-binding step in vivo. Genes Dev 13:2940^2945 Gavin AC, Bosche M, Krause R et al 2002 Functional organization of the yeast proteome by systematic analysis of protein complexes. Nature 415:141^147 Gonzalez F, Delahodde A, Kodadek T, Johnston SA 2002 Recruitment of a 19S proteasome subcomplex to an activated promoter. Science 296:548^550 Henry KW, Wyce A, Lo WS et al 2003 Transcriptional activation via sequential histone H2B ubiquitylation and deubiquitylation, mediated by SAGA-associated Ubp8. Genes Dev 17:2648^2663 Ho Y, Gruhler A, Heilbut A et al 2002 Systematic identi¢cation of protein complexes in Saccharomyces cerevisiae by mass spectrometry. Nature 415:180^183 Hwang WW, Venkatasubrahmanyam S, Ianculescu AG, Tong A, Boone C, Madhani HD 2003 A conserved RING ¢nger protein required for histone H2B monoubiquitination and cell size control. Mol Cell 11:261^266 Krogan NJ, Dover J, Wood A et al 2003a The Paf1 complex is required for histone H3 methylation by COMPASS and Dot1p: linking transcriptional elongation to histone methylation. Mol Cell 11:721^729 Krogan NJ, Kim M, Tong A et al 2003b Methylation of histone H3 by Set2 in Saccharomyces cerevisiae is linked to transcriptional elongation by RNA polymerase II. Mol Cell Biol 23:4207^4218 Kuo MH, Allis CD 1998 Roles of histone acetyltransferases and deacetylases in gene regulation. Bioessays 20:615^626 Marx J 2002 Cell biology. Ubiquitin lives up to its name. Science 297:1792^1794 Ng HH, Robert F, Young RA, Struhl K 2003 Targeted recruitment of Set1 histone methylase by elongating Pol II provides a localized mark and memory of recent transcriptional activity. Mol Cell 11:709^719 Nickel BE, Allis CD, Davie JR 1989 Ubiquitinated histone H2B is preferentially located in transcriptionally active chromatin. Biochemistry 28:958^963 Ostendor¡ HP, Peirano RI, Peters MA et al 2002 Ubiquitination-dependent cofactor exchange on LIM homeodomain transcription factors. Nature 416:99^103 Pham A-D, Sauer F 2000 Ubiquitin-activating/conjugating activity of TAFII250, a mediator of activation of gene expression in Drosophila. Science 289:2357^2360 Pickart CM 2001a Mechanisms underlying ubiquitination. Annu Rev Biochem 70:503^533 Pickart CM 2001b Ubiquitin enters the new millennium. Mol Cell 8:499^504 Pollard KJ, Peterson CL 1997 Role for ADA/GCN5 products in antagonizing chromatinmediated transcriptional repression. Mol Cell Biol 17:6212^6222 Robzyk K, Recht J, Osley MA 2000 Rad6-dependent ubiquitination of histone H2B in yeast. Science 287:501^504 Saleh A, Collart M, Martens JA et al 1998 TOM1p, a yeast hect-domain protein which medicates transcriptional regulation through the ADA/SAGA coactivator complexes. J Mol Biol 282:933^946 Salghetti SE, Caudy AA, Chenoweth JG, Tansey WP 2001 Regulation of transcriptional activation domain function by ubiquitin. Science 293:1651^1653 Sanders SL, Jennings J, Canutescu A, Link AJ, Weil PA 2002 Proteomics of the eukaryotic transcription machinery: identi¢cation of proteins associated with components of yeast TFIID by multidimensional mass spectrometry. Mol Cell Biol 22:4723^4738 Sun ZW, Allis CD 2002 Ubiquitination of histone H2B regulates H3 methylation and gene silencing in yeast. Nature 418:104^108

H2B UBIQUITYLATION

73

Wood A, Krogan NJ, Dover J et al 2003 Bre1, an E3 ubiquitin ligase required for recruitment and substrate selection of Rad6 at a promoter. Mol Cell 11:267^274 Xiao T, Hall H, Kizer KO et al 2003 Phosphorylation of RNA polymerase II CTD regulates H3 methylation in yeast. Genes Dev 17:654^663

DISCUSSION Verdin: Is there any evidence that Lys123 can also be acetylated? If so, its acetylation and deacetylation could represent a gate for ubiquitylation/ de-ubiquitylation. Berger: We don’t know. We have made an antibody and we are working on this now. The other way to address this would be to do this experiment in a HATdefective strain to see whether acetylation by SAGA is somehow important for the entry of ubiquitylation. Then you might imagine a series of steps where acetylation is required for ubiquitylation and then de-ubiquitylation occurs. We have asked whether ubiquitylation is required for acetylation, but we haven’t asked the reverse question yet. Gu: I have a question regarding double ChIP. Since you did £ag IP ¢rst and HA second, how do you exclude the possibility that this might immunoprecipitate di¡erent proteins? Berger: That is why I belaboured point of the K123R mutant control. This is essential as for any ChIP. In yeast, where the engineering is possible, it is incumbent on us to use the right controls, which is usually the mutational background. The double ChIP is no di¡erent. Thus, we have shown that when we substitute the K123 to alanine the ChIP signal is gone indicating that it is detecting a modi¢cation of K123 on H2B. Gu: According to your proposed a model, when the nucleosome without Ubp8 is ubiquitinated but may be methylated, it should be inactive. Can you prove that H2B is ubiquitinated but H3 is methylated from a single nucleosome, and ¢nd that the nucleosome is still inactive? Berger: We haven’t done that. I think David Allis’ lab did an experiment where they showed that the same nucleosomes are ubiquitylated and methylated showing that the modi¢cations are de¢nitely linked, but the Allis lab did not use a PCR at the end to determine where in the genome the linkage occurs. Yao: Do you see mono-ubiquitylated histone converted to poly in any of your experiments? Berger: In the mobility assay we do not observe any di- or tri-ubiquitylated H2B, although that does not rule out a very small amount of polyubiquitylation. Many labs have done similar mobility shift assays with consistent results showing only mono-ubiquitylated H2B. However, your point is well-taken, and it might be that ubiquitylation leads to histone degradation. The ChIP experiment for H2B

74

DISCUSSION

suggests this is not the case since we see that the level of H2B-FLAG is constant over the time that Ub-H2B-FLAG is decreasing. Furthermore, we observe an increase in H3 Lys-trimethylation over the same time course, suggesting that histone octamers are not being targeted for depletion. Marks: In the ChIP experiments in the promoter region, where you have increased H3-K4 trimethylation, you ¢nd that hypermethylation is associated with a block in transcription. What do you think is going on there? Berger: It is not just high trimethylation. There are also high levels of ubiquitylation that remain. One model that I have been thinking about was prompted by a recent publication showing that the Paf1 elongation complex may help to load SET1 and some of the other methylases that are involved in elongation (Xiao et al 2003, Ng et al 2003, Krogan et al 2003a,b). One possibility is that there is a form of ‘squelching’ or competition going on  that hyperloading of SET1 because of high level ubiquitylation in the absence of Ubp8 inhibits the loading of SET2 or Dot1. This is a model that we are testing. Turner: Can you make a double mutant of Ubp8 and SET1? Berger: This is in progress. It will be interesting to see whether this is suppressing, since excessive ubiquitylation in the Ubp8 deletion leads to high levels of SET1-mediated H3-K4 methylation. Allis: Related to Paul’s question, have you looked at the GAL1 promoter with acetyl H3-K9 or acetyl H3-K14? Berger: We haven’t done this yet. Perhaps there is hyperacetylation  a freezing in an activated state. The way we are thinking of this is kind of like a cell cycle. Ubiquitylation might be a checkpoint during transcription, so it comes on and then it is removed. Just as in the cell cycle, if you eliminate the checkpoint you can pass through, even if the correct events have not occurred. Jenuwein: Would you then argue that the ubiquitylation would be required more for the elongation process, rather than for silencing a promoter? Berger: It could be blocking elongation: that is a perfectly good model. There is now very good evidence that hypermethylation is associated with elongation, and in the 5’ end of the gene H3-K4 is high, and then drops through the gene. This is one of the reasons for speculating that it might be the balance between H3-K4 methylation and H3-K36 or H3-K79 methylation that is disturbed in the Ubp8 deletion. Jenuwein: This would o¡er a model where you could possibly discriminate the involvement of SET1 for transcriptional activation and repression at a ribosomal DNA, for example. There could be di¡erent elongation processes involved, and SET1 could have a di¡erent function dependent on which RNA polymerase is transcribing the locus. Berger: That’s an interesting idea.

H2B UBIQUITYLATION

75

Verdin: How do you envision the change of ubiquitylation at the promoter site a¡ecting elongation? Is it by recruiting a di¡erent polymerase complex? Berger: The idea would be that the PAF elongation complex is necessary to load SET1, SET2 and Dot1, and these are all involved in some aspect of elongation or capping, or polyadenylation. Then, if ubiquitylation is speci¢cally involved in SET1 loading, if too much of SET1 is loaded this then you could imagine an inhibition of the loading of the others. Thus, it appears that ubiquitylation is involved in more than just promoter-speci¢c activation. Marks: Are histone deacetylases (HDACs) recruited to the SAGA complex? Berger: There is no published evidence yet. We have found that there is a small amount of SAGA that has Sin3 present, which is a corepressor that interacts with HDACs. One model is that this is a transition complex between an active and inactive state. But it could also be that this complex is also is involved in the same cycle: acetylation/deacetylation; ubiquitylation/de-ubiquitylation during actual activation. It is possible that there is intimate association between these complexes, so that they have to coordinate their activities both for establishing and removing modi¢cations. Seto: Is Rpd3 part of the Sin3/SAGA transition complex? Berger: No, but we haven’t looked at whether any of the other HDACs are present. It is an interesting complex. It doesn’t have Tra1, which is on of the main activator interaction modules within SAGA. So, it has Sin3 but it doesn’t have the key protein (Tra1) that allows SAGA to be able to interact with activators. For this reason, we imagine that it is a transition complex between the active and inactive states. Seto: Is it possible that Sin3 functions as a structural, but not regulatory, protein in this complex? Berger: It might be, and it might be important in holding that complex together. Interestingly the complex is also lacking Spt20, which is a structural protein within SAGA, so one possibility is that Sin3 holds the complex together without Tra1 or Spt20. Gu: Can you detect the proteasome subunit within SAGA? Berger: We haven’t looked at that. It is good question. Experiments from Stephen Johnston and Tom Kodadek show that there is a portion of the proteasome that is present in the nucleus and is recruited to the GAL1 promoter. I believe that they think there is a di¡erent substrate than histone H2B. Turner: Going back to the role of trimethyl H3-K4 on elongation, what is your view from the results in the Ng et al (2003) paper? What they showed is that trimethyl H3-K4 is not particularly high at the promoter, and it was not particularly high in the 3’ downstream region of the gene. It doesn’t seem to be essential for initiation nor for elongation. Berger: Why isn’t it essential for elongation?

76

DISCUSSION

Turner: Because it is not there in the downstream region of the gene where elongation is progressing perfectly well. Berger: But it may be necessary for RNA polymerase II to clear the promoter for transition from initiation to elongation. Turner: Sure, but they only found it in that early part of the transcribed region. Almost by default, they suggested this may be some sort of short-term memory mechanism  a mark for genes that have recently been transcribed. What is your view? Berger: A number of groups have now shown that there is a high level of SET1 and H3-K4-methylation at the 5’ end. Kevin Struhl and Rick Young have showed this in whole genome analysis (Ng et al 2003). Stuart Schreiber showed last year that H3-K4 methylation is associated with a transcribed part of a gene (Bernstein et al 2002). There is quite a lot of evidence that this is true. We have seen the signal within the promoter, but we think it is bleed-over from the 5’ end of the gene. Turner: A paper that came out in the same issue as the Struhl lab paper (Ng et al 2003) showed the same thing at the 5’ end, but they also showed quite a lot of methylation at the promoter. Berger: We do too. I think all the evidence taken together suggests that polymerase is recruiting SET1 and Lys4 methylation through the PAF connection. Turner: But it doesn’t seem to be necessary for elongation per se. Berger: Perhaps some of these other methylation marks are involved in elongation, Lys36 or Lys79, or there is redundancy. Baylin: When you saw the ubiquitylation go up, did you map the dynamics across the promoter? Berger: We haven’t done that. Baylin: What would the pattern look like in the Ubp8 mutants? Berger: Those experiments are rather di⁄cult because the sensitivity is low. All of our experiments were done by real-time PCR. We would like to do a lot more ¢ne ChIP analysis in terms of location. Is it associated with the 5’ end of the gene or is it promoter associated? Greene: Is there any possibility that K123 is also the subject of sumoylation? Berger: We don’t think so.

References Bernstein BE, Humphrey EL, Erlich RL et al 2002 Methylation of histone H3 Lys 4 in coding regions of active genes. Proc Natl Acad Sci USA 99:8695^8700 Krogan NJ, Dover J, Wood A et al 2003a The Paf1 complex is required for histone H3 methylation by COMPASS and Dot1p: linking transcriptional elongation to histone methylation. Mol Cell 11:721^729

H2B UBIQUITYLATION

77

Krogan NJ, Kim M, Tong A et al 2003b Methylation of histone H3 by Set2 in Saccharomyces cerevisiae is linked to transcriptional elongation by RNA polymerase II. Mol Cell Biol 23:4207^4218 Ng HH, Robert F, Young RA, Struhl K 2003 Targeted recruitment of Set1 histone methylase by elongating Pol II provides a localized mark and memory of recent transcriptional activity. Mol Cell 11:709^719 Xiao T, Hall H, Kizer KO et al 2003 Phosphorylation of RNA polymerase II CTD regulates H3 methylation in yeast. Genes Dev 17:654^663

Structural and chemical basis of histone acetylation Ronen Marmorstein The Wistar Institute and the Department of Chemistry, University of Pennsylvania, Philadelphia, Pennsylvania 19104, USA

Abstract. Histones are the predominant protein components of chromatin and are subject to a variety of speci¢c post-translational modi¢cations that are correlated with transcriptional competence. Among these modi¢cations are reversible acetylation that is mediated by acetyltransferases that mediate transcriptional activation and deactylases that mediate transcriptional repression and gene silencing. Structural studies have provided important insights into the mechanism of substrate speci¢c binding and catalysis by the enzymes that mediate reversible acetylation. In this paper I will review structural work from my laboratory on histone acetyltransferases (HATs) and the Sir2 family of histone deacetylases (HDACs), with a speci¢c focus on catalysis and substrate-speci¢c binding by these enzymes. 2004 Reversible protein acetylation. Wiley, Chichester (Novartis Foundation Symposium 259) p 78^101

The enzymes that mediate post-translational modi¢cation of the N-terminal tails of the histone proteins that package DNA into chromatin play key roles in gene regulation (Grant 2001, Marmorstein 2001a, Zhang & Reinberg 2001). These enzymes carry out acetylation, phosphorylation, ubiquitylation (on the C-terminal tail), ribosylation and methylation. Although many of these modi¢cations have been correlated with both gene activation and repression, acetylation has been generally correlated with gene activation while deactylation has been correlated with transcriptional repression and gene silencing. The mechanism by which these modi¢cations regulate gene expression is not well understood, however it appears that these modi¢cations work in a combinatorial and coordinated fashion to elicit distinct biological responses (Schreiber & Bernstein 2002, Strahl & Allis 2000, Turner 2002). Of the enzymes that enzymatically modify chromatin, those that reversibly acetylate histones have been extensively studied at both the structural and biochemical level (Marmorstein 2001b,c). In this paper I will review structural work from my laboratory on the histone acetyltransferase (HAT) and Sir2 family 78

HISTONE ACETYLATION

79

of histone deacetylase (HDAC) enzymes, with a speci¢c focus on catalysis and substrate speci¢c binding by these enzymes. The histone acetyltransferase and deacetylase protein families There are now over 20 HAT proteins that have been identi¢ed from yeast to human. These proteins fall into distinct families with high sequence similarity within these families, but poor to no sequence similarity between them (Kuo & Allis 1998, Marmorstein 2001b) (Table 1). In addition, di¡erent HAT families have speci¢c and distinct histone substrate speci¢cities. Despite the di¡erences between the various HAT families, there are similarities. First, all of the HAT proteins that have been characterized in vivo are associated with large multiprotein complexes (Grant et al 1998). Second, although the recombinant proteins can acetylate free histones, nucleosomal acetylation occurs only in the context of in vivo HAT complexes (Sterner & Berger 2000), and substrate speci¢city is modulated in the context of these complexes (Grant et al 1999). To date, over 45 HDACs have been identi¢ed within yeast, Drosophila, maize, chicken, mouse and human and homologous proteins have also been identi¢ed in bacteria (Table 2) (de Ruijter et al 2003, Marmorstein 2001c). Together, the eukaryotic proteins fall into at least three distinct classes based on sequence homology. Members of the class I subgroup have a high degree of sequence homology to yeast Rpd3 (yRpd3) (Rundlett et al 1996) and are referred to as Rpd3-like, and members of the class II subgroup show homology to yeast Hda1 (yHda1) (Carmen et al 1996, Rundlett et al 1996). The Class I and II HDACs have some homology to each other and appear to di¡er in their cellular localization, where the class I proteins are largely nuclear and the class II proteins are either nuclear or cytoplasmic. Recent data also suggests that some class II HDACs do not have histones as their primary targets and may therefore not be bona ¢de HDACs (Hubbert et al 2002, Lemercier et al 2000). Yeast Sir2 (ySir2) is the founding member of the Class III HDACs and this family of proteins show signi¢cant sequence and functional divergence from Class I and II subgroups (Table 2) (Imai et al 2000). A functional distinction of the Sir2-like proteins is that their deacetylase activity is NAD+-dependent (Frye 1999, Imai et al 2000, Landry et al 2000). The Sir2-like proteins are extensive and broadly conserved from yeast to man (Frye 2000). Surprisingly, bacteria (that are free of histones) also contain proteins that have sequence homology to ySir2, suggesting that histones may not be the only targets of the eukaryotic Sir2 proteins. This is also correlated with recent ¢ndings showing that several eukaryotic Sir2 homologues contain nonhistone targets (North et al 2003, Vaziri et al 2001) and that several are cytoplasmically localized (Perrod et al 2001). Sequence homology among the Sir2 proteins is restricted to a roughly 270-residue domain that has been shown to

80

TABLE 1

MARMORSTEIN

HAT families and their transcription-related functions

HAT

Organism

Function

H3 (K14)

GCN5/PCAF family Gcn5 Yeast to human PCAF Human

Coactivator (adaptor) Coactivator

MYST family Sas2 Sas3 Esa1 MOF Tip60 MOZ HBO1 TAFII250 family CBP/p300 family

Silencing Silencing Cell cycle progression Dosage compensation HIV^Tat interaction Leukeomogenesis Origin recognition interaction TBP-associated factor Global coactivator

Yeast Yeast Yeast Fruit £y Human Human Human Yeast to human Worm to human

Histone*

SRC family Mice and human Steroid receptor SRC-1 coactivators ACTR/AIB1/pCIP/TRAM-1/RAC3 SRC-3 TIF-2 GRIP1 ATF-2 Yeast to human Sequence-speci¢c DNA-binding activator HAT1 family Yeast to human Replication-dependent chromatin assembly (cytoplasmic)

+non-histone proteins H4 (H3)

H3 All +nonhistone proteins H3/H4

H4, H2B H4

*Only preferred histone substrates are indicated.

be su⁄cient for catalytic activity (Landry et al 2000, Min et al 2001, Smith et al 2000, Tanner et al 2000, Tanny et al 1999).

Overall structure of HAT proteins My laboratory has determined the structure of HAT proteins from the Gcn5/ PCAF (Clements et al 1999, Lin et al 1999, Rojas et al 1999, Trievel et al 1999)

HISTONE ACETYLATION

TABLE 2

81

Classi¢cation of histone deacetylases

Class

Members

Properties

I. yRpd3-like

Rpd3, HDAC1, HDAC2 HDAC3, Hos3

II. yHda1-like

Hda1, HDAC4, HDAC5, HDAC6, HDAC7, HDAC8 Hos1, Hos2 Ia ^ SIR2, SIRT1, Hst1 Ib ^ Hst2, SIRT2, SIRT3 Ic ^ Hst3, Hst4 II ^ SIRT4 III ^ SIRT5 IV ^ SIRT6, SIRT7

^ TSA-sensitive ^ nuclear ^ histone deacetylases ^ TSA-sensitive ^ nuclear/cytoplasmic ^ protein deacetylases ^ NAD+-dependant ^ nuclear/cytoplasmic ^ protein deacetylase

III. ySir2-like

Only proteins from yeast (S. cerevisiae) and humans are shown here. A more complete list of homologues can be found in Frye (2000).

and MYST subfamily (Yan et al 2000, 2002) in various liganded forms, and Dutnall and co-workers have determined the structure of the cytoplasmic yeast HAT1 enzyme bound to coenzyme-A (Dutnall et al 1998). A comparison of these structures reveals structural homology within a central core domain and structural divergence in regions N- and C-terminal to this core domain (Fig. 1). The structure of ternary complexes between Tetrahymena Gcn5 (tGCN5) with coenzyme-A (CoA) and either an 11 or 19-residue histone H3 peptide centred around the preferred lysine-14 target reveals the mode of cosubstrate binding. The protein structure contains an L-shaped cleft, with CoA bound in the short segment and the histone peptide bond in the long segment of the L-shaped cleft. A correlation of the structure with related biochemical and mutagenesis studies reveals that the central domain plays a particularly important role in acetyl-CoA binding and catalysis, while the regions N-and C-terminal to the catalytic core domain play a particularly important role in histone substrate speci¢c binding within Gcn5 (Fig. 1A). Catalysis by HAT proteins Enzymatic studies have revealed that the Gcn5/PCAF family of HAT enzymes employ a ternary complex mechanism for catalysis (Rojas et al 1999, Tanner et al 1999, Trievel et al 1999). The proposed mechanism of catalysis is outlined in Fig. 2A. Brie£y, a conserved glutamate residue within the central core domain

82

MARMORSTEIN

HISTONE ACETYLATION

83

functions as a general base. The structure of the ternary complexes reveals that this glutamate is H-bonded to a very well-ordered water molecule, located between the carbonyl oxygen of the glutamate and the Nz nitrogen of the target lysine, implicating a role for this water mediating the deprotonation of the histone lysine by the enzyme glutamate. The conserved glutamate residue (Glu122 of tGcn5) is surrounded by several hydrophobic residues that likely shield it from solvent and help raise its pKa to facilitate proton extraction. Once the lysine proton is extracted, the acetyl group of the acetyl-CoA, which is hydrogenbonded to the backbone NH of Leu126 (in the tGCN5/acetyl-CoA structure), is transferred to the reactive Lys14 side chain of the histone. The backbone NH of Leu126 probably functions to polarize the carbonyl group of the thioester prior to nucleophilic attack of the amino group and stabilize the negative charge that develops on the oxygen in the tetrahedral transition state. The structure of the core domain of the Esa1 member of the MYST family of HAT proteins superimposes almost perfectly with the core domain of the Gcn5/ PCAF family and in addition a structural alignment of this domain reveals that glutamate 338 of Esa1 superimposes almost perfectly with the catalytic glutamate of the Gcn5/PCAF family. Despite this, structural and enzymatic studies reveal that Esa1 uses a ping-pong catalytic mechanism involving the formation of an acetyl-cysteine intermediate (Yan et al 2002). The catalytic mechanism that was derived (Fig. 2B) is based on the ability to structurally trap the acetyl-cysteine intermediate, the ability to form acetylated Esa1 enzyme in solution and to transfer the acetyl group from protein to histone peptide in solution, and to reveal a bisubstrate kinetic pro¢le that is consistent with a ping-pong catalytic mechanism. In this mechanism, Ac-CoA ¢rst binds to the protonated form of cysteine-304. Cysteine-304 is then deprotonated by glutamate-338, which promotes nucleophilic attack of the thioester of Ac-CoA by cysteine-304. Upon formation of the acetyl-cysteine-304 intermediate, glutamate-338 donates a proton to CoA and thereby regenerates its ability to function as a general base. The substrate lysine of the histone then binds Esa1 and the reactivated glutmate338 now deprotonates the lysine side-chain. Upon deprotonation, the lysine substrate carries out nucleophilic attack of the acetyl-cysteine-304 intermediate producing the acetyl-lysine reaction product. Since both cysteine-304 and glutamate-338 are strictly conserved in the MYST subfamily of HAT enzymes it

FIG. 1. Structure of histone acetyltransferases. (A) Schematic structure of the Gcn5/PCAF HAT domain. The core catalytic domain of tGcn5 and CoA are coloured in black with the tGcn5 domains N- and C-terminal to the core domain and the histone H3 peptide (thick loop) colored in grey. (B) Schematic structure of the yEsa1 member of the MYST HAT domain in complex with CoA. The shade coding is as indicated in 1a. (C) Schematic structure of the yHAT1 HAT domain in complex with Acetyl-CoA. The shade coding is as indicated in 1a.

84

MARMORSTEIN

FIG. 2. Enzymatic mechanism of histone acetyltransferases. (A) Proposed catalytic mechanism for the Gcn5/PCAF HATs. The histone substrate is colored in grey. The residue numbering is for tGcn5. (B) Proposed catalytic mechanism for the MYST HATs. The histone peptide is shaded in grey. The residue numbering is for yEsa1.

HISTONE ACETYLATION 85

86

MARMORSTEIN

is likely that that the catalytic mechanism that has been proposed for Esa1 also applies to other members of the MYST family. It is very surprising that the Gcn5/PCAF and MYST family of HAT enzymes, two functionally and structurally related enzyme families, use di¡erent catalytic mechanisms. Interestingly, a recent report from the Cole group shows that the p300 HAT, a member of yet another HAT family, appears to use yet a di¡erent catalytic mechanism (Thompson et al 2001). It is attractive to hypothesize that the di¡erence in catalytic mechanism between these di¡erent HAT families may be related to their di¡erent in vivo properties. Substrate binding speci¢city by HAT proteins The crystal structure of the ternary complex between tGCN5, CoA and an 11 residue histone H3 peptide (H3p11) reveals that the histone substrate sits in a cleft formed by the central protein core and the pantetheine arm of CoA, and £anked by N- and C-terminal HAT domain segments that mediate most of the tGCN5-histone H3 contacts (Rojas et al 1999). One surprising feature of the tGCN5/CoA/H3p11 structure is that most of the peptide^protein interactions are mediated through the backbone of the peptide, in its C-terminal region, whereas the N-terminal residues 9^13 are £exible with disordered side chains in the crystal structure (Rojas et al 1999). This result, coupled with the biochemical studies demonstrating that GCN5/PCAF family members exhibit a high degree of speci¢city for Lys14 on histone H3 (Kuo et al 1996, 1998, Schiltz et al 1999, Wang 1998) raises the possibility that there are other structural determinants for substrate binding speci¢city by the GCN5/PCAF HAT domain. In support of this, the PCAF HAT domain has a greater than 100-fold enhanced substrate speci¢city (kcat/Km) for a 19-amino acid H3-derived substrate peptide (residues 5^23; H3p19) when compared to H3p11 (Trievel et al 2000), with only a slightly elevated activity towards a 27-residue histone H3 peptide (1.9-fold relative to H3p19). Taken together, these studies suggest that the 19-residue histone H3 peptide is su⁄cient to mediate most, if not all, of the histone H3 interactions with the GCN5/PCAF HAT domain. A more recent structure of tGcn5 in ternary complex with CoA and a 19-residue histone H3 peptide shows that 15 of the 19 histone H3 residues are ordered with only two disordered residues at each end of the peptide (Clements et al 2003) (Fig. 3). Surprisingly, relative to the complex containing H3p11, the H3p19 complex reveals signi¢cantly more extensive protein^peptide interactions. Although the C-terminal half of H3p19 roughly follows the backbone of residues Gly13 through Gln19 in H3p11 and many of the van der Waals interactions are conserved in this region of the peptide, few of the tGCN5/H3 hydrogen bonds are identical between the two structures (Figs 3A,B). The greatest structural

HISTONE ACETYLATION

87

alterations between the two peptides occur in residues N-terminal to Lys14 as H3p19 makes more interactions in this region, where Lys9 and Arg8 appear to play particularly important roles. Most strikingly, a core of 12 residues of histone H3 cantered on Lys14, are well-ordered in the tGCN5/CoA/histone H3p19 complex and each of these residues make sequence-speci¢c contacts to protein regions N- and C-terminal to the catalytic core of tGCN5. Together, the more extensive interactions observed with the longer histone H3 peptide is consistent with the solution studies showing that the GCN5/PCAF HAT domain has *100-fold improved substrate speci¢city for H3p19 over H3p11 (Trievel et al 2000). Moreover, this comparison reveals that residues outside the core 12residue sequence of the histone H3 peptide anchor and reposition the core of histone H3 for more optimal enzyme^histone contacts. The protein regions N- and C-terminal to the catalytic domains of Esa1 and Hat1 show structural divergence to the corresponding regions of Gcn5/PCAF. Despite this, a supposition of the respective core domains of the two protein families reveals that the N- and C-terminal regions of Gcn5 that speci¢cally contact histone substrate have structurally overlapping counterparts in Esa1 and Hat1 (Fig. 3C). This suggests that the Gcn5/PCA and MYST HATs, and possibly other HAT families, may have a similar structural sca¡old for substrate speci¢c recognition and that the sequence divergence within this sca¡old may contribute to substrate-speci¢c binding. Overall structure of the Sir2 family of HDACs There are now several reported structures of the catalytic domain of homologues of the Sir2 family of HDACs (Avalos et al 2002, Chang et al 2002, Finnin et al 2001, Jacobs et al 2002, Min et al 2001), including the structure determined in our laboratory of the catalytic domain of the yeast Hst2 member in ternary complex with 20 -O-acetyl ADP ribose and a histone H4 peptide (Zhao et al 2003a) (Fig. 4A). The catalytic Sir2 core forms an elongated catalytic domain containing a large and structurally homologous Rossmann fold domain characteristic of NAD+ binding proteins, and a small, structurally more variable, domain containing a structural zinc ion. A series of loops traverse between the large and small domain forming a pronounced extended cleft between the two protein domains. The ternary yHst2/NAD+/histone H4 complex shows that the two substrates enter the protein through opposite sides of a cleft between the small and large domains of the catalytic core, and that the functional groups of both the protein and substrates are buried within a protein tunnel that harbours the region of highest conservation within the Sir2 proteins. The structure determined in our laboratory of the full-length yHst2 protein reveals that regions N- and C-terminal to the catalytic core domain play

88

MARMORSTEIN

HISTONE ACETYLATION

89

autoregulatory roles (Zhao et al 2003b) (Fig. 4B). Speci¢cally, a C-terminal helix overlaps with the NAD+ binding site and thus autoregulates NAD+ binding. In addition, an N-terminal strand sits in the acetyl-lysine binding site of a symmetry related subunit in the crystal lattice mediating formation of a homotrimer, and thus autoregulating acetyl-lysine binding. Solution studies correlate with these ¢ndings (Zhao et al 2003b). Whether or not this type of enzyme autoregulation occurs with other Sir2 proteins is unknown, however, the sequence divergence within regions N- and C-terminal to the catalytic core domain of the Sir2 proteins suggests that these interactions may di¡er in other Sir2 proteins, possibly correlating with the substrate speci¢c roles of di¡erent Sir2 proteins. Catalysis by the Sir2 HDACs The structure of the ternary yHst2/NAD+/histone H4 complex, in combination with the reaction products identi¢ed by Schramm and co-workers (Sauve et al 2001) and the mutational sensitivity of residues in the active site (Avalos et al 2002, Chang et al 2002, Finnin et al 2001, Min et al 2001) allow us to propose a detailed catalytic mechanism for the Sir2 proteins (Fig. 5A). In this mechanism, the carbonyl oxygen of acetyl-lysine carries out nucleophilic attack of the 1’carbon of the ribose ring nucleating the displacement of the nicotinamide ring and forming of an O-alkyl amidate intermediate. The imidizole nitrogen of His 135 then functions as a general base to directly deprotonate the 3’ hydroxyl of the ribose ring. We then propose that proton migration from the 2’ to 3’ oxygen of the ribose ring results in formation of a cyclic acyl-dioxolane involving the 1’ and 2’ oxygens of the ribose ring. We then propose that a well-ordered water molecule that is held in place by, and possibly also activated by Asn116, and within hydrogen bonding distance to the 3’ oxygen of the ribose ring, carries out nucleophilic attack of the cyclic dioxolane. This results in the collapse of the cyclic intermediate to the 2’-O-ADP ribose and lysine reaction products. A comparison of the ternary complex with the binary yHst2/NAD+ (Zhao et al 2003a) and nascent yHst2 structures reveals that the most structurally dynamic

FIG. 3. Substrate binding by HAT proteins. (A) tGcn5/CoA/histone H3 complex. The core catalytic domain of tGcn5 and CoA (stick ¢gure) are in black with the tGcn5 domains N- and Cterminal to the core domain and the 19-residue histone H3 peptide (stick ¢gure) in grey. (B) Schematic tGcn5-histone H3 interactions in the ternary tGcn5/CoA/histone H3 complex shown in 4a. Regions of the histone H3 peptide that are ordered in the ternary complex with the 19-residue peptide, but not ordered in the ternary complex with the 11-residue peptide are in grey. (C) Superposition of putative substrate binding sites of HAT proteins. The superposition is generated by superimposing the core domains from the MYST member, yEsa1 (black), and the Gcn5/PCAF member, hPCAF (light grey), with yHAT1 (dark grey). Only the core domain and CoA of yEsa1 (black) is shown for clarity.

90

MARMORSTEIN

HISTONE ACETYLATION

91

FIG. 4. Structure of a Sir2 histone deacetylase. (A) Schematic structure of the ternary complex between the catalytic core of the Sir2 protein, yHst2 in complex with NAD+ and an 11-residue histone H4 peptide. The protein is shown with the large domain shaded in grey, and the small domain and connecting loops between the small and large domains in black. The NAD+ and the histone H4 peptide cosubstrates are white stick ¢gures. (B) Schematic structure of the yHst2 homotrimer showing in grey the intramolecular contacts of the C-terminal a13 helix that autoregulates NAD+ binding and intermolecular contacts of the N-terminal extended loop that autoregulates acetyl-lysine binding.

region of yHst2 is the b1^a2 loop that is disordered in the nascent structure, but becomes ordered upon NAD+ binding and mediates nearly half of the NAD+ contacts. The rearrangement of the b1^a2 loop of yHst2 also appears to be correlated with a rigid-body rotation of the small domain relative to the large domain of the catalytic core for more optimal acetyl-lysine interactions, presumably also linking NAD+ binding to acetyl-lysine binding and catalysis.

92

MARMORSTEIN

B

HISTONE ACETYLATION

93

94

MARMORSTEIN

FIG. 5. Enzymatic mechanism and histone substrate binding of the Sir2 histone deactylases. (A) Proposed catalytic mechanism for the Sir2 HDACs. The acetyl-lysine substrate is in grey. The residue numbering is for yHst2. (B) Schematic of protein^histone interactions in the ternary yHst2/NAD+/histone H4 complex. Hydrogen bonds are indicated with a dashed line and van der Waals interactions are indicated with a half-moon symbol. The residues in rectangles and ovals highlight interactions with NAD+ that are conserved and non-conserved, respectively, with the protein^NAD+ interactions observed in the Af1-Sir2/NAD+ structure. (C) The p53 peptide (dark grey) from the Sir2-Af2/p53 peptide structure and the ‘pseudosubstrate’ from the nascent yHst2 structure (white) is overlaid with the histone H4 peptide (light grey) onto a surface representation of yHst2 from the ternary complex. Protein residues that make conserved interactions between the three substrates are in black and protein residues that mediate variable interactions are in grey.

Substrate binding speci¢city by the Sir2 HDACs There are over 60 identi¢ed Sir2 HDAC homologues from prokaryotes to multicellular eukaryotic organisms (Frye 2000). Interestingly, despite their implication in histone deacetylation, the only Sir2 homologue shown to have

HISTONE ACETYLATION

95

bona ¢de HDAC activity in vivo is the Sir2 protein from yeast. The true in vivo substrates for most of the Sir2 homologues have not yet been identi¢ed, and the handful of other Sir2 proteins that have been characterized have been shown to target non-histone proteins. To date, there is no structure available of a Sir2 protein bound to its cognate acetyl-lysine containing target, although the ternary yHst2/NAD+/histone H4 complex determined in our laboratory and the Af2-Sir2/ p53 peptide complex (Avalos et al 2002) does provide some insights into substrate recognition by Sir2 proteins. In the ternary yHst2 structure with acetylated histone H4, acetyl-Lys 16 makes the most extensive interactions with yHst2 (Fig. 5B). Speci¢cally, the aliphatic arm of this residue makes extensive van der Waals interactions with several hydrophobic residues that are highly conserved within the Sir2 proteins. The backbone amino group of acetyl-lysine 16 also makes b-sheet interactions with Sir2-conserved backbone residues of the protein. Taken together, these studies suggest that the observed interactions are a conserved feature of the Sir2 proteins. Consistent with this proposal, the structure of Sir2-Af2 bound to an acetyl-lysine p53 peptide shows nearly identical interactions in this region (Avalos et al 2002) (Fig. 4B). Although the yHst2 protein makes extensive interactions with the acetyl-lysine residue of the histone peptide, outside of acetyl-lysine 16, the yHst2-histone H4 peptide interactions are very limited and largely restricted to backbone interactions involving residues that are not conserved among the Sir2 proteins (Fig. 5B). These relatively sparse set of protein-peptide interactions probably re£ects the fact that histone H4 may not be the true in vivo substrate for yHst2 and/or that substrate speci¢city determinants may involve other regions of the Sir2 proteins and/or regions of the substrate that are distal to the acetyl-lysine site. Consistent with the latter argument, a superposition of the ternary complex reported here with the Sir2-Af2/p53 peptide complex and yHst2 bound to a ‘pseudo-substrate’ within the autoinhibited homotrimeric form, reveals that each of the acetyl-lysine substrates (or substrate mimics in the case of the autoinhibited yHst2) diverge greatly in path and region of contact on the protein directly outside of the acetyl-lysine and the two £anking residues of the substrate (Fig. 5C). Uncovering the mode of substrate speci¢city by the Sir2 proteins is one of the major hurdles before us in understanding the mechanism of action of these enzymes. Perspective Although considerable progress has been made towards understanding the mode of catalysis by HAT proteins and the Sir2 family of HDAC proteins, considerably more work has to be done. The ¢ndings that HAT proteins from two di¡erent HAT families use di¡erent catalytic mechanisms is particularly novel and intriguing and

96

MARMORSTEIN

a remaining question is whether other HAT families use di¡erent mechanisms, and how are these mechanisms related to the di¡erent biological roles of these enzymes. With regard to the Sir2 proteins, although the general scheme for the reaction mechanism has been established, many details of the unusual NAD+-dependent deactylation reaction still need to be established. Moreover, the biological role, if any, of the 2’-O-ADP-ribose reaction product needs to be established. The speci¢c roles of the regions N- and C-terminal to the catalytic core domain of other eukaryotic Sir2 proteins will also be of considerable interest. Substrate speci¢city by HAT and Sir2 HDAC proteins is even more poorly understood. Although it is clear that regions N- and C-terminal to the catalytic core domain of the HAT proteins play particularly important roles in substrate speci¢city, it is clear that there are other important contributions to substrate speci¢city that are not as well understood. For example, HATs are part of large multiprotein subunits in vivo and it is known that other subunits of these complexes in£uence substrate speci¢city. The way in which this occurs is of considerable interest and may be a critical factor in gene regulation. Substrate speci¢city by the Sir2 proteins is more poorly understood and it appears that regions outside of the catalytic core domain and/or regions from other associated Sir2 proteins may be involved. Additional structures of HATs and Sir2 proteins both alone and with relevant associated proteins in complex with cognate histone substrates should go a long way to resolving many of these issues. Acknowledgements I would like to acknowledge all past and present members of my laboratory who have contributed to the studies discussed in this chapter. In particular, I would like to thank (in alphabetical order) X. Chai, A. Clements, A. Poux, J. Rojas, R. Trievel, Y. Yan and K. Zhao.

References Avalos JL, Celic I, Muhammad S, Cosgrove MS, Boeke JD, Wolberger C 2002 Structure of a Sir2 enzyme bound to an acetylated p53 peptide. Mol Cell 10:523^535 Carmen AA, Rundlett SE, Grunstein M 1996 HDA1 and HDA3 are components of a yeast histone deacetylase HDA complex. J Biol Chem 271:15837^15844 Chang JH, Kim HC, Hwang KY et al 2002 Structural basis for the NAD-dependent deacetylase mechanism of Sir2. J Biol Chem 277:34489^34498 Clements A, Poux AN, Lo W-S, Pillus L, Berger SL, Marmorstein R 2003 Structural basis for histone and phosphohistone binding by the GCN5 histone acetyltransferase. Mol Cell 12:461^473 Clements A, Rojas JR, Trievel RC, Wang L, Berger SL, Marmorstein R 1999 Crystal structure of the histone acetyltransferase domain of the human P/CAF transcriptional regulator bound to coenzyme-A. EMBO J 183:521^3532 de Ruijter AJ, van Gennip AH, Caron HN, Kemp S, and van Kuilenburg AB 2003 Histone deacetylases HDACs: characterization of the classical HDAC family. Biochem J 370:737^749

HISTONE ACETYLATION

97

Dutnall RN, Tafrov ST, Sternglanz R, Ramakrishnan V 1998 Structure of the histone acetyltransferase Hat1: a paradigm for the GCN5-related N-acetyltransferase superfamily. Cell 94:427^438 Finnin MS, Donigian JR, Pavletich NP 2001 Structure of the histone deacetylase SIRT2. Nat Struct Biol 86:21^625 Frye RA 1999 Characterization of ¢ve human cDNAs with homology to the yeast SIR2 gene: Sir2-like proteins sirtuins metabolize NAD and may have protein ADP-ribosyltransferase activity. Biochem Biophys Res Commun 260:273^279 Frye RA 2000 Phylogenetic classi¢cation of prokaryotic and eukaryotic Sir2-like proteins. Biochem Biophys Res Commun 273:793^798 Grant PA 2001 A tale of histone modi¢cations. Genome Biol 2:REVIEWS0003 Grant PA, Sterner DE, Duggan LJ, Workman JL, Berger SL 1998 The SAGA unfolds: convergence of transcription regulators in chromatin-modifying complexes. Trends Cell Biol 8:193^197 Grant PA, Eberharter A, John S, Cook RG, Turner BM, Workman JL 1999 Expanded lysine acetylation speci¢city of Gcn5 in native complexes. J Biol Chem 274:5895^5900 Hubbert C, Guardiola A, Shao R et al 2002 HDAC6 is a microtubule-associated deacetylase. Nature 417:455^458 Imai S, Armstrong CM, Kaeberlein M, Guarente L 2000 Transcriptional silencing and longevity protein Sir2 is an NAD-dependent histone deacetylase. Nature 403:795-800 Jacobs SA, Harp JM, Devarakonda S, Kim Y, Rastinejad F, Khorasanizadeh S 2002 The active site of the SET domain is constructed on a knot. Nat Struct Biol 9:833^838 Kuo MH, Allis CD 1998 Roles of histone acetyltransferases and deacetylases in gene regulation. Bioessays 20:615^626 Landry J, Sutton A, Tafrov ST et al 2000 The silencing protein SIR2 and its homologs are NADdependent protein deacetylases. Proc Natl Acad Sci USA 97:5807^5811 Lemercier C, Verdel A, Galloo B, Curtet S, Brocard MP, Khochbin S 2000 mHDA1/HDAC5 histone deacetylase interacts with and represses MEF2A transcriptional activity. J Biol Chem 275:15594^15599 Lin Y, Fletcher CM, Zhou J, Allis CD, Wagner G 1999 Solution structure of the catalytic domain of Tetrahymena GCN5 histone acetyltransferase in complex with coenzyme A. Nature 400:86^89 Marmorstein R 2001a Protein modules that manipulate histone tails for chromatin regulation. Nat Rev Mol Cell Biol 2:422^432 Marmorstein R 2001b Structure and function of histone acetyltransferases. Cell Mol Life Sci 58:693^703 Marmorstein R 2001c Structure of histone deacetylases: insights into substrate recognition and catalysis. Structure 9:1127^1133 Min J, Landry J, Sternglanz R, Xu RM 2001 Crystal structure of a SIR2 homolog-NAD complex. Cell 105:269^279 North BJ, Marshall BL, Borra MT, Denu JM, Verdin E 2003 The human Sir2 orthologSIRT2is a NAD+-dependent tubulin deacetylase. Mol Cell 11:437^444 Perrod S, Cockell MM, Laroche T et al 2001 A cytosolic NAD-dependent deacetylaseHst2pcan modulate nucleolar and telomeric silencing in yeast. EMBO J 20:197^209 Rojas JR, Trievel RC, Zhou J et al 1999 Structure of the Tetrahymena GCN5 bound to coenzyme-A and a histone H3 peptide. Nature 401:93^98 Rundlett SE, Carmen AA, Kobayashi R, Bavykin S, Turner BM, Grunstein M 1996 HDA1 and RPD3 are members of distinct yeast histone deacetylase complexes that regulate silencing and transcription. Proc Natl Acad Sci USA 93:14503^14508 Sauve AA, Celic I, Avalos J, Deng H, Boeke JD, SchrammVL 2001 Chemistry of gene silencing: the mechanism of NAD+-dependent deacetylation reactions. Biochemistry 40:15456^15463

98

DISCUSSION

Schreiber SL, Bernstein BE 2002 Signaling network model of chromatin. Cell 111:771^778 Smith JS, Brachmann CB, Celic I et al 2000 A phylogenetically conserved NAD+-dependent protein deacetylase activity in the Sir2 protein family. Proc Natl Acad Sci USA 97:6658^6663 Sterner DE, Berger SL 2000 Acetylation of histones and transcription-related factors. Microbiol Mol Biol Rev 64:435^459 Strahl BD, Allis CD 2000 The language of covalent histone modi¢cations. Nature 403:41^45 Tanner KG, Trievel RC, Kuo M-H et al 1999 Catalytic mechanism and function of invariant glutamic acid-173 from the histone acetyltransferase GCN5 transcriptional coactivator. J Biol Chem 274:18157^18160 Tanner KG, Landry J, Sternglanz R, Denu JM 2000 Silent information regulator 2 family of NAD-dependent histone/protein deacetylases generates a unique product1-O-acetyl-ADPribose. Proc Natl Acad Sci USA 97:14178^14182 Tanny JC, Dowd GJ, Huang J, Hilz H, Moazed D 1999 An enzymatic activity in the yeast Sir2 protein that is essential for gene silencing. Cell 99:735^745 Thompson PR, Kurooka H, Nakatani Y, Cole PA 2001 Transcriptional coactivator protein p300Kinetic characterization of its histone acetyltransferase activity. J Biol Chem 276:33721^33729 Trievel RC, Rojas JR, Sterner DE et al 1999 Crystal structure and mechanism of histone acetylation of the yeast GCN5 transcriptional coactivator. Proc Natl Acad Sci USA 96:8931^8936 Turner BM 2002 Cellular memory and the histone code. Cell 111:285^291 Vaziri H, Dessain SK, Ng Eaton E et al 2001 hSIR2SIRT1 functions as an NAD-dependent p53 deacetylase. Cell 107:149^159 Yan Y, Barlev NA, Haley RH, Berger SL, Marmorstein R 2000 Crystal structure of yeast Esa1 suggests a uni¢ed mechanism of catalysis and substrate binding by histone acetyltransferases. Mol Cell 6:1195^1205 Yan Y, Harper S, Speicher DW, Marmorstein R 2002 The catalytic mechanism of the ESA1 histone acetyltransferase involves a self-acetylated intermediate. Nat Struct Biol 9:862^869 Zhang Y, Reinberg D 2001 Transcription regulation by histone methylation: interplay between di¡erent covalent modi¢cations of the core histone tails. Genes Dev 15:2343^2360 Zhao K, Chai X, Marmorstein R 2003a Structure of the yeast Hst2 histone deactylase in ternary complex with 2’-O-acetyl ADP ribose and histone peptide. Structure 11:1403^1411 Zhao K, Chai X, Clements A, Marmorstein R 2003b Structure and autoregulation of the yeast Hst2 homolog of Sir2. Nat Struct Biol 10:864^871

DISCUSSION Hottiger: If you could hypothetically add an acetyl group to the lysine that you modify in your structure, would there be enough room to keep a substrate bound to the acetyltransferase given that the speci¢city is determined by domains outside the catalytic domain? Marmorstein: We have solved the structure of an acetylated lysine bound to Gcn5. It is accommodated, but it has popped out enough that it disorders the structure on the outside. It is clearly less stably held than an unacetylated substrate. This explains why, once it is acetylated, it comes o¡. Castronovo: Can you compete with the binding of the enzyme to the natural substrate that is going to be acetylated using your peptide?

HISTONE ACETYLATION

99

Marmorstein: Yes, we can do that. This is essentially what we did with the competition assay using K4 as an inhibitor. We have this £uorogenic acetyllysine substrate, we throw in histone H4 and it competes. We can measure the apparent Km. Verdin: The mechanism here is reminiscent of what we have seen with SirT3, which is translated as a precursor containing a 100 amino acid N-terminal extension. This N-terminal extension mediates import of the protein into the mitochondria and is cleaved in the mitochondrial matrix by a speci¢c protease. We have observed that this proteolytic processing of SirT3 leads to its enzymatic activation. Has anyone looked to see whether Hst2 is di¡erentially active in the cytoplasm vs. the nucleus? Marmorstein: That is a good question. I don’t know whether anyone has looked at this. The problem with Hst2 is that the mutant doesn’t have a phenotype, so it’s di⁄cult to obtain a read-out from such experiments. Cole: In the Wolberger structure, she was concerned that the NAD and the peptide were somewhat overlapping. You obviously don’t see that. How do you rationalize those two results? Marmorstein: When Wolberger superimposed her enzyme with acetyl-lysine bound to another archaeal enzyme with NAD bound, they overlapped. But this is comparing two enzymes that have enough structural divergence that you can’t accurately superimpose the two enzymesthey can’t really be compared. The nice thing about Hst2 is that we are able to compare the same enzyme in di¡erent liganded states. I think they overlap in the Wolberger superposition because the alignment between Af1-Sir2 bound to NAD and Af2-Sir2 bound to acetyl-lysine was not accurate enough. Cole: Do you see any evidence for something that could stabilize an oxocarbenium ion in your structure? Presumably this would be important for catalysis. Marmorstein: No, I don’t. Moazed: In the earlier structure from the Wollberger paper (Avalos et al 2002), there was speculation that substrate binding induces major conformational changes, which was prompted by the comparison of di¡erent structures. Do you see any major structural changes? Marmorstein: This was from comparing di¡erent proteins. I don’t see major changes when we do the comparison. If you compare the Hst2 binary complex with NAD bound, and the Hst2 ternary complex with NAD and histone bound, the NAD superimposes perfectly. Gu: How many lysine residues are acetylated on your p53 peptide? Marmorstein: Are you talking about the Gcn5 studies? Gu: Yes, and also Sir2. Marmorstein: That is the other structure. Only Lys320 is acetylated.

100

DISCUSSION

Denu: Regarding this e¡ect of the C- and N-terminus of Hst2 on enzymatic activity, the e¡ects are not very dramaticit looks like only two or threefold. Do you think this is signi¢cant in the context of the cell? Marmorstein: Yes. Phosphorylation of Ser10 and acetylation of Lys14 is not a dramatic e¡ect either, yet it clearly has an in vivo e¡ect. These are isolated proteins, and the e¡ect could be enhanced if Hst2 is part of a complex in the cell. These are real di¡erences. It is important to realise that these are not on^o¡ switches but instead are regulatory switches that modulate activity, and so their e¡ects can be more subtle. Verdin: Are these regulatory regions potentially modi¢ed themselves? I am particularly interested in the region that sits in the catalytic site. Could an acetylated residue sit in there and function as a competitive inhibitor? Marmorstein: That is an excellent question. I don’t know the answer but it would be interesting to look at. Allis: Are you talking about the conserved loop? Verdin: Yes, the one that falls back at the N-terminus may actually be modi¢ed itself. Marmorstein: I don’t know that anyone has actually looked at modi¢cation of any Sir2 proteins. Verdin: We have preliminary evidence that SirT2 is phosphorylated. However, we do not know the role of this modi¢cation. Moazed: Yeast Sir2 is a phosphoprotein. Marmorstein: Where does it get phosphorylated? Moazed: Serine 27 on the N-terminus, I think. Turner: Can I worry about speci¢city again? You showed a dramatic e¡ect of the P10 S10 phosphorylation on the conformation of the tail. What e¡ect is acetylation of other sites going to have? Often the enzyme will be seeing a multiply acetylated tail. How is this going to a¡ect speci¢city? The other thing is the dramatic result that Danesh Moazed showed, which indicated that the Sir2 doesn’t work when you present it with its proper substrate, which is the histone tail on a nucleosome. Marmorstein: People have been focusing on K9. One observation we have made is that in unphosphorylated histone H3, K9 is pointing in, and when it is phosphorylated at Ser10 it is pointing out and accessible. Subtle e¡ects like this might have dramatic di¡erences. If phosphorylation has a small e¡ect this might be enhanced by modi¢cation of Lys9. Turner: What I am fretting about is the possibility that Sir2 may not be able to deacetylate a multiply acetylated tail, for example. It could be that this ends up being a very speci¢c enzyme in vivo, because it will only recognize certain combinations of modi¢cations, whereas it may not be able to touch others. This could be extremely signi¢cant.

HISTONE ACETYLATION

101

Marmorstein: It’s feasible. I don’t think we learn anything from our structural work that addresses this possibility. Cole: My understanding is that even acetyl-lysine alone is a substrate for Sir2, without any peptide. Marmorstein: I was unaware of that and we have never looked at that. Turner: Why do nucleosomes not work? Marmorstein: It could be that it just does not gain access. Denu: We have done some work with just the free amino acid, acetylated lysine: it is terrible. If you attach an aromatic group or additional amino acids you get extra binding a⁄nity. Cole: From your structure can you get a sense of how Danesh and Schreiber’s aldehyde might be binding to Sir2? They have a Sir2 inhibitor that they isolated from a screen. Marmorstein: We have not tried to use that inhibitor yet. Mahadevan: I have a question about the presentation of substrate to Gcn5. What do you make of the recent paper where Shogren-Knaak et al (2003) ligated a synthetic phosphorylated N-terminal peptide onto histone H3 and reconstituted it into nucleosomal arrays. These authors showed that there was no di¡erence between the phosphorylated and non-phosphorylated forms in their ability to be acetylated by Gcn5 in the SAGA complex. Marmorstein: These were Greg Peterson’s data where he made histone cores and linked on them N-terminal histone tails containing various modi¢cations. One modi¢cation he looked at was the e¡ect of Ser10 phosphorylation on acetylation. There are a couple of issues that need to be resolved here. First, he has phosphorylated every tail in the 10-mer nucleosome, which may not be physiological. Second, he saw pretty good acetylation of these nucleosomes by recombinant Gcn5. In general, recombinant Gcn5 doesn’t acetylate nucleosomes well in vivo. There is a disconnect that doesn’t make sense to me here. In this context he sees a di¡erence, but is this a physiologically relevant context? References Avalos JL, Celic I, Muhammad S, Cosgrove MS, Boeke JD, Wolberger C 2002 Structure of a Sir2 enzyme bound to an acetylated p53 peptide. Mol Cell 10:523^535 Shogren-Knaak MA, Fry CJ, Peterson CL 2003 A native peptide ligation strategy for deciphering nucleosomal histone modi¢cations. J Biol Chem 278:15744^15748

Phosphorylation and acetylation of histone H3 at inducible genes: two controversies revisited Louis C. Mahadevan, Alison L. Clayton, Catherine A. Hazzalin and Stuart Thomson Nuclear Signalling Laboratory, Department of Biochemistry, Oxford University, South Parks Road, Oxford OX1 3QU, UK

Abstract. The phosphorylation and acetylation (phosphoacetylation) of histone H3 tails concomitant with gene activation is now well established and has been observed at several inducible genes. However, two aspects of this response have been controversial. The ¢rst relates to the identity of the kinase that phosphorylates histone H3. Experiments with Co⁄n^Lowry cells purporting to show that Rsk2 was the histone H3 kinase have proven to be irreproducible. The second relates to the proposition that histone H3 phosphorylation and acetylation are ‘synergistic and coupled’ in mammalian cells. But here too, some of the experiments have not been reproducible and some of the key statements contaminated by issues of antibody speci¢city. More recent studies indicate that H3 phosphorylation and acetylation are independently targeted to the same histone H3 tail. 2004 Reversible protein acetylation. Wiley, Chichester (Novartis Foundation Symposium 259) p 102^114

I’d like to talk about Histone H3 phosphorylation and acetylation in the mouse system. This follows on nicely from what we have just heard about the structure of GCN5 and its interaction with the histone tail, and the suggestion that H3 phosphorylation and acetylation are linked. We are interested in the induction of immediate-early (IE) genes by diverse stimuli. We have studied the response to many physiological, pharmacological and stress stimuli, of ¢ve AP1 family genes, c-fos, fosB, c-jun, junB and junD that are di¡erentially induced (Fig. 1; Hazzalin & Mahadevan 2002). In control cells there is no signal, but in response to di¡erent stimuli these genes are quickly switched on. It is apt that we remember Vincent Allfrey at this meeting, because when he discovered the link between histone acetylation and transcription, c-fos was one of the genes he started to look at (Chen & Allfrey 1987). It’s a suitable choice, because it goes from the ‘o¡’ to the ‘on’ state very nicely. The IE gene response is involved in di¡erentiation, mitosis and diseases such as in£ammation and cancer (Herschman 1991, McMahon & 102

HISTONE H3 PHOSPHOACETYLATION

103

FIG. 1. Di¡erential expression of IE genes in response to various stimuli. Northern blot analysis of immediate-early (IE) gene RNA isolated from quiescent C3H 10T1
2 cells stimulated as indicated: TNFa (5 ng/ml,); bFGF (20 ng/ml); EGF (50 ng/ml); TPA (100 nM); or anisomycin (An; 10 mg/ml), for 15 to 60 minutes. C, control (unstimulated). The housekeeping gene glyceraldehyde-3-phosphate dehydrogenase (GAPDH) was used as a control for RNA loading. TNFa, tumour-necrosis factor a; bFGF, basic ¢broblast growth factor; EGF, epidermal growth factor; TPA, 12-O-tetradecanoylphorbol-13-acetate. Reproduced with permission from Hazzalin & Mahadevan (2002).

Monroe 1992). We will be hearing more about another one of these genes, nur77, in a later paper. I want to discuss two controversies (see Davie 2003) related to these issues, ¢rstly, the kinase that phosphorylates histone H3 at these genes, and secondly, the proposed link between phosphorylation and acetylation of the histone tail at all of these genes (discussed in Clayton & Mahadevan 2003). We now know that these genes are controlled by MAP kinase cascades. There are four well-characterized MAPK pathways (Fig. 2), the original ERK1/2 cascade, plus two more recently discovered, the JNK/SAPKs and p38, and the most recent addition, the ERK5 cascade (reviewed in Hazzalin & Mahadevan 2002). When cells are stimulated with many di¡erent agents, these cascades are di¡erentially switched on depending upon the stimulus, but in parallel, to

FIG. 2. Mammalian MAPK cascades and substrates. Schematic representation of mammalian MAPK cascades and target substrates, including transcription factors involved in the regulation of fos andjun immediate-early (IE) genes. The four characterized MAPKs comprise the extracellular signal regulated kinase (ERK/MAPK), c-jun amino-terminal kinase/stress-activated protein kinase (JNK/SAPK), p38 (reactivating kinase (RK)/ p40/cytokine-supressive anti-in£ammatory drug binding protein (CSBP)/Mxi2 kinase), big MAP kinase (BMK/ERK5) subtypes. MEK, MAPK/ ERK kinase; MKK/MAPKK, MAPK kinase; SKK, SAPK kinase. References can be found in Hazzalin & Mahadevan (2002). Reproduced with permission from Hazzalin & Mahadevan (2002).

104 MAHADEVAN ET AL

HISTONE H3 PHOSPHOACETYLATION

105

produce e¡ects in the nucleus. Upstream of these kinases there are several speci¢c activators linked to particular stimuli, which are outside the scope of this talk. We are interested in the downstream e¡ects. Downstream of the MAP kinases there are three clusters of substrates that are involved in gene induction. The ¢rst includes some more downstream kinases. The ERK1/2 and p38 cascades have several downstream kinases shown in Fig. 2, whereas so far in mouse cells, JNK/SAPK and ERK5 have no good downstream kinases. Note that there are some kinases that are downstream of both ERK1/2 and p38 cascades, and some that are only downstream of one of these cascades. There is therefore di¡erential activation of these kinases by these cascades. The second class of substrate consists of transcription factors. There are many of these shown in Fig. 2. The fos and jun family genes have multiple regulatory elements upon which these transcription factors sit. These transcription factors can be phosphorylated either by MAP kinases directly or in some cases by the downstream kinases (Fig. 2). All of this happens within minutes, which makes it a nice system to study. That brings me to the third class of substrates, the nucleosomal proteins, which are also targets of these cascades. If we do [32P]-phosphate labelling of cells stimulated with EGF or anisomycin, we see the H3 ladder on acid-urea gels with nice stimulation of H3 phosphorylation within minutes. The two residues phosphorylated are serine 10 and serine 28. In addition there is a nucleosomal binding HMG14 protein (current nomenclature HMGN1) phosphorylated on serine 6 (Barratt et al 1994a). These studies were done by metabolic labelling with [32P]-phosphate, but more recently we have used phospho-speci¢c antibodies. Phospho-serine 10 and 28 are detected in response to EGF, anisomycin and TPA. I should stress that unlike metabolic labelling, immunostaining with di¡erent antibodies is not a quantitative representation: there is actually more serine 10 phosphorylation than serine 28. The question, then, is what is the kinase that phosphorylates these sites? There were two contenders in the literature. One, from the work of David Allis and Paolo Sassone-Corsi was Rsk2, and this comes from genetic analysis of a human condition called Co⁄n^Lowry syndrome in which there is a defect in the RSK2 gene (Sassone-Corsi et al 1999). The second study was our own (Thomson et al 1999), and was more biochemical and cell biological, looking at the e⁄ciency with which the kinase phosphorylated H3 peptide. Comparison of the sensitivity of the nucleosomal response and the various potential kinases to inhibitors suggested that MSKs were the kinases for histone H3, work that has been published (Thomson et al 1999). Today I want to discuss some additional data on MSK1 and 2. Both of these kinases have been knocked out by homologous recombination in mouse cells. This was done by Simon Arthur’s group in Dundee where they knocked out both MSK1 and MSK2 at both alleles. When we analysed H3 phosphorylation in these cells, we found that the normal H3 phosphorylation response seen in the wild-type ¢broblasts was completely

106

MAHADEVAN ET AL

[Image not available in this electronic edition.]

FIG. 3. Histone H3 phosphorylation in Co⁄n^Lowry cells. Wild-type or Co⁄n^Lowry immortalized human ¢broblasts were serum starved overnight and then stimulated with anisomycin (10 mg/ml, 30 min) or TPA (400 ng/ml, 30 min). (A) Cells were lysed, and histone H3 was extracted from the pellet and analysed for the phosphorylation on Ser10 by immunoblotting. The soluble fraction of the cell lysate was analysed by immunoblotting for (B) phosphorylation of CREB and levels of (C) RSK2 and (D) total ERK1/2 (loading control). Reproduced with permission from Soloaga et al (2003).

missing in the knockout cells (Soloaga et al 2003). Signalling goes on normally in both cell types. In the double knockout cells we have used phospho-speci¢c antibodies to show that phosphorylation at both sites (Ser10 and Ser28) on histone H3 and at the single site (Ser6) on HMG-N1 are all missing. You can reintroduce MSK2 kinase into these cells and the response is recovered (Soloaga et al 2003). This ‘rescue’ experiment showed us that it is MSK kinases that are phosphorylating H3, and that the loss was not due to any other de¢ciency in these cells. This was so convincing that it now it became interesting to go back to the Co⁄n^Lowry cells and analyse why it was that the response was reportedly lost in those cells too. Unfortunately, we were unable to detect any loss of H3 phosphorylation in the Co⁄n^Lowry cells (Fig. 3). We have not been able to reproduce the defect of H3 phosphorylation in the Co⁄n^Lowry cells. We also looked at another reported defect in Co⁄n^Lowry cells which is the CREB phosphorylation defect. Again, this too was not reproducible (Fig. 3). This was done in Simon Arthur’s group at the MRC Protein Phosphorylation Unit in Dundee. We were so concerned about this that we were worried that we had perhaps been sent the wrong cells. So the blots were re-probed with anti-Rsk2antibodies. We saw Rsk2 in the wild-type cells and it is missing in the

HISTONE H3 PHOSPHOACETYLATION

107

Co⁄n^Lowry cells, con¢rming that these were indeed the right cells. This seriously compromises the evidence that Rsk2 is the histone H3 kinase. The second controversial issue I would like to address is the proposed link between histone H3 phosphorylation and its acetylation. I am now taking you through some general analysis of histone modi¢cations in the mouse nucleus. We have used pairs of plates of cells, control or stimulated, treated for increasing amounts of time with the histone deacetylase (HDAC) inhibitor trichostatin A (TSA). This work was originally done in the early 1990s by Michael Barratt using sodium butyrate and metabolic labelling with [32P]-phosphate (Barratt et al 1994b), and has since has been repeated with phospho-speci¢c antibodies. If you look at the coomassie-stained histone H3 ladder, there is substantial histone H3 acetylation even in the control cells. When we treat these cells with TSA over a period of 2 h it changes slightly but not markedly. Replicate blots were probed with phosphoacetyl histone H3 antibody, and over this period, in contrast to the stained bulk histone H3, this population of phosphorylated H3 becomes very highly acetylated. This indicates that a minute fraction of histone H3 that is phosphorylated must be hypersensitive to this acetylation, and that both of these modi¢cations are on the same histone tail. However, we found that although phosphorylation and acetylation are co-targeting the same H3 tail, the two modi¢cations were independent of each other (Barratt et al 1994b, Clayton et al 2000). More recently, structural work with GCN5 (Lo et al 2000) has led to the idea that phosphorylation and acetylation are ‘synergistically coupled’ and that this also happens at these IE genes in mouse cells, i.e. at the genes that we are looking at, phosphate is targeted to the H3 tail at these genes ¢rst and acetylation follows as a consequence (Cheung et al 2000). We categorically refute this idea. Histone H3 acetylation is targeted independently to these nucleosomes whether or not the phosphate is there. This can be shown by chromatin immunoprecipitation (ChIP) assays using anti-phosphoacetyl antibodies and the anti-acetyl histone H3 antibody (Clayton et al 2000, Thomson et al 2001). The important thing about these antibodies is that they are mutually exclusive. The batch of anti-acetyl-lysine-9/14 H3 antibody that we used, sold by Upstate, does not recognize the histone tail when phosphate is present at serine 10 (Thomson et al 2001). In fact, it is now clear that this antibody mainly recognizes acetyl-lysine 9, not acetyl-lysine 14. Conversely, the phosphoacetyl H3 antibody will recognize this tail only when it is both phosphorylated at serine 10 and acetylated at lysine 9 (Clayton et al 2000). This gives us the ability to isolate two mutually exclusive populations of nucleosomes, and we can now look at the distribution of these modi¢cations along the c-jun gene. In a ChIP assay we have looked at three regions on this gene, showing that at jun1 and jun3 there are substantial modi¢cations, whereas at jun2 there is very little modi¢cation (Thomson et al 2001). The negative controls we used were two regions of the b globin gene and

108

MAHADEVAN ET AL

a region of histone H4. The crucial thing is that the two groups of nucleosomes are mutually exclusive, with one group having exclusively acetylated and not phosphorylated H3. There is a lot of exclusively acetylated histone H3, but only a small amount of phosphoacetylated histone. The point about this is that it shows that we don’t obligately need phosphorylation to create acetylation at these residues. You might argue that this is because the phosphate was there before acetylation and now it has disappeared leaving only acetylated H3. This is not true: you can do an anti-phospho-H3 immunoprecipitation and this fails to pick up the appearance of phosphate before the appearance of phosphoacetyl H3 (Thomson et al 2001). In fact, we now know that we can abolish histone H3 phosphorylation using inhibitors of either ERK1/2 or p38 MAP kinases or MSK1/2 itself. This removes the phosphorylation signal at these genes but not the acetylation signal, which continues normally whether or not phosphorylation arrives at these genes (Thomson et al 2001). The acetylation is targeted by a di¡erent mechanism. One candidate at the moment is the JNK cascade which signals to the c-fos and c-jun genes (Alberts et al 1998). Finally, we can engineer a situation in these cells where there is no kinase activation and there is no transcription, and still get acetylation at these genes. If we treat these cells with TSA, an HDAC inhibitor, we see something interesting. Over a period of 2 h, at nucleosomes at all three positions of the c-jun gene we get very strong acetylation. This is a highly targeted event to these oncogenes. It is not found in many other genes, such as b globin and histone H4. It is also hypersensitive. That is, we can see this within 15 minutes, with highly targeted acetylation at these genes. We have done a wider analysis of this, looking at ¢ve regions along the c-fos gene and four regions along c-jun. The targeted e¡ect of HDAC inhibitors is not seen evenly across the whole gene, but only a¡ects parts of it. We have tested genes which are continuously transcribed in these cells and they are not subject to this HDAC sensitivity. HDAC sensitivity is a really interesting property that a¡ects these oncogenes. These genes are in the ‘o¡’ state in quiescent cells; they are not activated or undergoing transcription. But even under these conditions there is opposing histone acetyltransferase (HAT) and HDAC activity continuously turning over acetyl groups at these positions on these genes. This takes us away from the idea that the correlation between the gene that is ‘on’ and the gene that is ‘o¡’ is a change in the acetylation status. All our present evidence suggests that there is turnover of acetyl groups both when the gene is ‘on’ and when it is ‘o¡’. The ¢nal piece of evidence that acetylation is independent of phosphorylation comes from the MSK knockout cells. We have shown that H3 phosphorylation is defective in these cells (Soloaga et al 2003). However, analysis of H3 acetylation at c-fos and c-jun by ChIP assays in these cells shows that we still see acetylation at these genes (Fig. 4), con¢rming that prior phosphorylation is not required for

HISTONE H3 PHOSPHOACETYLATION

109

[Image not available in this electronic edition.]

FIG. 4. H3 acetylation at c-jun in Msk1/Msk2 double knockout cells lacking H3 phosphorylation. (A) Schematic representation of the c-jun gene indicating the regions ampli¢ed by the primer pairs used in the PCR step of the ChIP assay. (B) Wild-type and MSK1/MSK2 knockout ¢broblasts at passage 5 were serum starved overnight and then left unstimulated or stimulated with 25 ng/ml anisomycin for 30 min. Cross-linked chromatin was immunoprecipitated using antibodies that recognize histone H3 that is both phosphorylated on Ser10 and acetylated on Lys9 (pAcH3) or acetylated on Lys9 and Lys14 (AcH3). The cross-links on the precipitated DNA were then reversed and the DNA analysed for the presence of the c-jun gene sequences by radioactive PCR. Graphical representation of the c-jun gene fragments recovered with the (C) anti-phosphoacetyl histone H3 or (D) anti-acetyl histone H3 antibodies. The data are shown as an average of three (c-jun1) or two (c-jun3) independent experiments, with the PCR step being carried out at least twice for each experiment. (E) Wildtype and MSK1/2 double-knockout cells were serum starved overnight and then left untreated or stimulated with anisomycin (25 ng/ml) for 30 min. Immunoblotting with the antiphosphoacetyl histone H3 antibody was carried out. Reproduced with permission from Soloaga et al (2003).

110

MAHADEVAN ET AL

targeted acetylation. I will conclude by saying two things. I have shown you that MAP kinases can send a phosphorylation signal to chromatin at these genes, and that MSK1/2 is the major kinase that phosphorylates H3. The case for Rsk2 is compromised by lack of reproducibility of the Co⁄n^Lowry experiments. Second, I have shown you that far from being targeted by phosphorylation, there is continuous highly targeted turnover of acetylation by HATs and HDACs at these genes, and there is no necessity for additional signalling for that targeting to occur.

References Alberts AS, Geneste O, Treisman R 1998 Activation of SRF-regulated chromosomal templates by Rho-family GTPases requires a signal that also induces H4 hyperacetylation. Cell 92: 475^487 Barratt MJ, Hazzalin CA, Zhelev N, Mahadevan LC 1994a A mitogen- and anisomycinstimulated kinase phosphorylates HMG-14 in its basic amino-terminal domain in vivo and on isolated mononucleosomes. EMBO J 13:4524^4535 Barratt MJ, Hazzalin CA, Cano E, Mahadevan LC 1994b Mitogen-stimulated phosphorylation of histone H3 is targeted to a small hyperacetylation-sensitive fraction. Proc Natl Acad Sci USA 91:4781^4785 Chen TA, Allfrey VG 1987 Rapid and reversible changes in nucleosome structure accompany the activation, repression, and superinduction of murine ¢broblast protooncogenes c-fos and c-myc. Proc Natl Acad Sci USA 84:5252^5256 Cheung P, Tanner KG, Cheung WL, Sassone-Corsi P, Denu JM, Allis CD 2000 Synergistic coupling of histone H3 phosphorylation and acetylation in response to epidermal growth factor stimulation. Mol Cell 5:905^915 Clayton AL, Mahadevan LC 2003 MAP kinase-mediated phosphoacetylation of histone H3 and inducible gene regulation. FEBS Lett 546:51^58 Clayton AL, Rose S, Barratt MJ, Mahadevan LC 2000 Phosphoacetylation of histone H3 on c-fos- and c-jun-associated nucleosomes upon gene activation. EMBO J 19:3714^3726 Davie JR 2003 MSK1 and MSK2 mediate mitogen- and stress-induced phosphorylation of histone H3: a controversy resolved. Sci STKE 2003 PE33 Hazzalin CA, Mahadevan LC 2002 MAPK-regulated transcription: a continuously variable gene switch? Nat Rev Mol Cell Biol 3:30^40 Herschman HR 1991 Primary response genes induced by growth factors and tumor promoters. Ann Rev Biochem 60:281^319 Lo WS, Trievel RC, Rojas JR et al 2000 Phosphorylation of serine 10 in histone H3 is functionally linked in vitro and in vivo to Gcn5-mediated acetylation at lysine 14. Mol Cell 5:917^926 McMahon SB, Monroe JG 1992 Role of primary response genes in generating cellular responses to growth factors. FASEB J 6:2707^2715 Sassone-Corsi P, Mizzen CA, Cheung P et al 1999 Requirement of Rsk-2 for epidermal growth factor-activated phosphorylation of histone H3. Science 285:886^891 Soloaga A, Thomson S, Wiggin GR et al 2003 MSK2 and MSK1 mediate the mitogen- and stress-induced phosphorylation of histone H3 and HMG-14. EMBO J 22:2788^2797

HISTONE H3 PHOSPHOACETYLATION

111

Thomson S, Clayton AL, Hazzalin CA, Rose S, Barratt MJ, Mahadevan LC 1999 The nucleosomal response associated with immediate-early gene induction is mediated via alternative MAP kinase cascades: MSK1 as a potential histone H3/HMG-14 kinase. EMBO J 18:4779^4793 Thomson S, Clayton AL, Mahadevan LC 2001 Independent dynamic regulation of histone phosphorylation and acetylation during immediate-early gene induction. Mol Cell 8:1231^1241

DISCUSSION Allis: I thought your double null data were very impressive. Can you not rescue the MSK1 with a catalytically dead mutant? You showed us you can rescue with an active MSK2. Mahadevan: We have not tried to do this. Allis: What about RSK2? Have you tried to go back into those cells, and can you rescue with RSK2? Mahadevan: The endogenous RSK2 is normal in these cells, so that is unnecessary. Cohen: Is the MSK1/2 double knockout lethal? Mahadevan: No. We have also done single knockouts and these experiments suggest that MSK2 has a greater contribution than MSK1 (Soloaga et al 2003). The knockout mice appear normal at present. Olson: Does that TSA-sensitive region of the c-jun promoter map to the position of the MEK2 binding site? Mahadevan: Not strictly; it is not limited to one region, but goes on along the gene. We are not talking just about the promoter here, but both the coding region of the gene and the promoter. So far we cannot distinguish between the two. As we go along the gene this response drops o¡. This is all done using small fragments in the ChIP assay that are mostly about three nucleosomes long. Ott: Have you tried EGF in the Co⁄n^Lowry syndrome cells? Mahadevan: Yes, we have tried several stimuli in these cells. We cannot reproduce the loss of the H3 phosphorylation phenotype. Zhou: Do you have data on the methylation state of Lys9 or 27 in the Co⁄n^ Lowry cells? Mahadevan: No, this is the extent of what we have done in those cells. Atadja: Are you planning an A¡ymetrix-type experiment with your double knockouts, comparing stimulated and unstimulated cells? Mahadevan: I believe those experiments are being done in Dundee. Marks: Does the increase in transcription of these early genes correlate with the transient increase in accumulation of the acetylated histones in the promoter region? What is the temporal relationship there?

112

DISCUSSION

Mahadevan: It is more or less an overlap. This is because there are about 100 immediate early genes and they all come up over this 30 minute period and drop o¡ at di¡erent rates. If you take that class of 100 genes it does correlate reasonably well. Cohen: With regard to the question of the phenotype, what worries me is what is compensating for MSK1/2. Mahadevan: We don’t know that compensation is occurring. In the MSK knockout cells, when we went in with a ChIP assay there is a very small residue of H3 phosphorylation signal. We don’t trust this, because it is very low, but it is there. If we look at the phenotype in the Petri dish and look at immediate early gene induction, we see a marked di¡erence of e⁄ciency of induction, but there is not an ablation. All our data indicate that the transcription factor-type molecules and their phosphorylation may a¡ect on^o¡ states of these genes in a more profound way than histone modi¢cations. Ott: What are MSK1/2 levels like in Co⁄n^Lowry cells? And what are RSK levels like in your knockout cells? Mahadevan: In MSK1/2 knockout cells the RSK level looks normal for wild-type ¢broblasts. We wanted to look at MSK levels in Co⁄n^Lowry cells but we never went on to do it because we saw normal H3 phosphorylation. Ott: Have you done siRNA knockdown in human cells for MSK1/2? Mahadevan: No. Atadja: It will be important to do this because it will give a more acute e¡ect than creating knockout animals. Mahadevan: I agree. It would be nice to produce another angle to show that MSK knockdown is directly responsible for this. But there are other characteristics of that phosphorylation response. For example, the e⁄ciency of phosphorylation between MSK and histone H3 is extremely high when compared with other kinases and substrates. Cole: Is that phosphorylation enhanced when you acetylate Lys14? Mahadevan: We haven’t done that. Olson: Is there a speci¢c phosphatase for that site? Mahadevan: We don’t know of any speci¢c phosphatase for that site, but this response is transient. Furthermore, if you did your experiment without throwing in phosphatase inhibitors you would lose the response by the time you had lysed your cells and run your gels. Allis: At least for mitosis, it is probably PP1. In yeast it is Glc7. But this could be just for mitotic Ser10 phosphorylation. Verdin: I was intrigued by what you said in terms of acetylation not being correlated with transcription, but indicating transcription potential. What de¢nes many of these immediate early genes is that they are all poised for transcription such that transcription can be rapidly activated. Can you

HISTONE H3 PHOSPHOACETYLATION

113

speculate on the mechanism by which acetylation makes a gene poised for transcription? Mahadevan: We are studying a unique subset of gene regulation events: genes that are just sitting there and waiting to ¢re, and it is crucial that they ¢re immediately. The other point about this is that these genes don’t just switch on. There are stable quantitative di¡erences between these genes. People who do this regularly can look at the patterns of gene induction and tell what the stimulus is. It is not just that one of these genes switch on: they switch on to a speci¢c extent. We think these genes are characterized by having quantitative control which is linked to enzyme biochemistry so that you can produce a particular amount of mRNA. This is why these genes constitute a special subset. Our present model is that at these genes part of the mechanism that makes it respond in a £exible way is the pre-association of HATs or HDACs or both with these regions. This is the only way we can understand the TSA experiment  within minutes of treating with TSA, these genes go up. There is another point that I didn’t make. If you look over that period of 2 h at the gene, acetylation drops back down again. But if you look in the nucleus, overall acetylation continues to go up for 6 h. That indicates a transient local interaction. Afterwards TSA-insensitive HDACs are taking over or HATs are leaving. This has some relation to the histone code and the original ideas of Vincent Allfrey, who thought that there was a hypo-acetylated gene that is switched on and then enters an acetylation state, which was like a mark of the active gene. Our work suggests that the process is much more dynamic than that, and that there is turnover in both of these states. Then you can talk about the histone code, as well. How does having such diverse combinations of modi¢cations associated regionally with a gene play with the idea of something like the histone code? Li: What is the methylation status of responsive genes? Does this contribute to the £exibility you see? Mahadevan: We have started experiments looking at this, but we are not happy with the antibodies we are using at the moment. Turner: You treat with the factor and the activity goes up and then drops down. How long after that can the genes be restimulated? What is the lag period? Mahadevan: There is a well-studied phenomenon of desensitization in the signalling ¢eld. That involves regulation at many levels, but principally you can show it with receptors. For example, if you pretreat with EGF and then come back with EGF, there is reduced receptor at the cell surface. So it is hard to answer your question. There is a period in which the gene is not inducible, but we don’t know whether this is to do with the gene itself or desensitization. Denu: If you treat with TSA you get signi¢cant amounts of resetting and deacetylation. Can you e⁄ciently reacetylate those histones?

114

DISCUSSION

Mahadevan: If you switch the gene on with a stimulant such as anisomycin and then treat with TSA, you still get enhanced acetylation. We haven’t tried doing this 6 h later when it is properly silenced. But in the 2 h time course it doesn’t matter when you add TSA. Gu: Have you treated the cells with nicotinamide together with TSA? Mahadevan: No. Reference Soloaga A, Thomson S, Wiggin GR et al 2003 MSK2 and MSK1 mediate the mitogen- and stress-induced phosphorylation of histone H3 and HMG-14. EMBO J 22:2788^2797

HDAC7 regulates apoptosis in developing thymocytes Eric Verdin, Frank Dequiedt and Herb Kasler Gladstone Institute of Virology and Immunology and Department of Medicine, University of California San Francisco, PO Box 419100, San Francisco, CA 94141-9100, USA

Abstract. Central immune tolerance is established in the thymus for T cells via a complex selection process that involves interactions between CD4+CD8+ double-positive thymocytes and antigen-presenting cells. Cells that express antigen receptors interacting strongly with self peptide^MHC complexes are deleted from the repertoire via activationinduced apoptosis, a process termed negative selection. Cells that express an appropriate signal are positively selected and mature into single positive na|« ve T cells, either CD4 or CD8 positive. The balance between positive and negative selection is thought to play a critical role in the elimination of self-reactive clones and in the establishment of central immune tolerance. We have recently reported that HDAC7, a class II histone deacetylase, is highly expressed in CD4+CD8+ double positive thymocytes. HDAC7 inhibits Nur77 expression, an orphan receptor involved in antigen-induced cell death and in negative selection. The inhibitory e¡ect of HDAC7 on the Nur77 promoter is mediated via the transcription factor MEF2D. During T cell receptor activation, HDAC7 is exported from the nucleus leading to the derepression of Nur77 expression and the induction of apoptosis. These observations de¢ne HDAC7 as a regulator of Nur77 and apoptosis in developing thymocytes and indicate that HDAC7 is likely to play an important role in the control of central immune tolerance. 2004 Reversible protein acetylation. Wiley, Chichester (Novartis Foundation Symposium 259) p 115^131

All peripheral T cells, CD4+ T helper cells and CD8+ cytotoxic T cells, do not normally respond to self-antigens, a phenomenon referred to as self-tolerance. Self-tolerance may be induced either in the thymus (central tolerance) or in peripheral tissues (peripheral tolerance). Breakdown of either central or peripheral tolerance can result in the onset of autoimmune responses with potential pathological consequences (Anderson et al 2002, Kishimoto & Sprent 2001, Ohashi 2002, Ohashi & DeFranco 2002, Peterson et al 1998). Thymic developing T cells, also called thymocytes, are derived from bone marrow precursors, and seed the thymus. CD4
CD8
double-negative (DN) thymocytes represent one of the earliest steps after thymic emigration. These cells undergo a proliferative expansion and di¡erentiation that is driven by the pre-T 115

116

FIG. 1. details.

VERDIN ET AL

Di¡erentiation of T cells in the thymus: negative vs. positive selection. See text for

cell receptor (TCR), a surface receptor that consists of an invariant pre-Ta chain and clonotypic b-chain. This clonal expansion and di¡erentiation of DN thymocytes gives rise to the main population of CD4+CD8+ double-positive (DP) thymocytes. DP thymocytes undergo one of three fates: death by neglect (*95% of DN thymocytes), positive selection (*2.5%) and negative selection (*2.5%) (Fig. 1). Death by neglect is a passive form of cell death caused by the failure of a clonotypic abTCR expressed by the thymocyte to engage a peptide^MHC ligand. These cells do not receive a survival signal and undergo apoptosis. Positive selection occurs if the TCR of the thymocyte engages a peptide^MHC ligand with low a⁄nity, resulting in the transduction of a survival and di¡erentiation signal. Positively selected DP thymocytes further di¡erentiate into CD4+ or CD8+ single positive (SP) na|« ve T cells (Fig. 1). Engagement of a peptide-MHC class I ligand positively selects for MHC class I restricted, CD8+ cytotoxic T cells, whereas recognition of a peptide-MHC class II ligand positively selects for MHC class II restricted, T helper CD4+ T cells. Negative selection occurs when the TCR of a

HDACs AND THYMOCYTE APOPTOSIS

117

thymocyte engages a peptide-MHC ligand with high a⁄nity, leading to the triggering of apoptosis and death of the thymocyte. Negative selection therefore deletes potentially self-reacting thymocytes thereby generating a population of peripheral T cells that is largely self-tolerant. For cells that escape negative selection in the thymus, safeguard regulatory mechanisms function in the periphery to control autoreactive T cells, i.e. peripheral tolerance (for recent reviews on the subject, please see Healy & Goodnow 1998, Palmer 2003, Sebzda et al 1999, Starr et al 2003). Negative versus positive selection The observation that dramatically di¡erent outcomes, i.e. survival and di¡erentiation during positive selection vs. cell death during negative selection, can occur as a result of antigen receptor signalling has stimulated a considerable amount of interest and only partial answers (for recent reviews on the subject, see Healy & Goodnow 1998, Sprent & Kishimoto 2001). It is generally accepted that four types of extracellular input shape the response to antigen: . . . .

antigen concentration binding avidity of the antigen timing and duration of antigen encounter and the association of antigen with costimuli from pathogens.

It is generally accepted that signals accompanied by high antigen concentration, high avidity interactions and longer duration favour negative selection over positive selection. There is also growing evidence that intracellular signalling via the TCR in response to distinct ligands is not an all-or-none event and that extracellular signals modulate these signalling events both quantitatively (strength and duration of signal) and qualitatively (di¡erent intracellular mediators). For example, positive selection but not negative selection is inhibited by dominant negative mutations in the ras/MEK/ERK signalling pathway (Alberola-Ila & Hernandez-Hoyos 2003, Alberola-Ila et al 1996, Swan et al 1995). Conversely, several cellular factors have been uniquely implicated in negative selection. They include CD28 (Buhlmann et al 2003, Kishimoto & Sprent 1999, Punt et al 1994), CD5 (Page 1999), CD40 ligand (Foy et al 1995), the JNK and p38 MAP kinases (Rincon et al 1998, Sabapathy et al 2001, Sugawara et al 1998), PTEN (Suzuki et al 2001), and the orphan steroid receptors Nur77 and NOR1 (Calnan et al 1995, Cheng et al 1997, Woronicz et al 1994). Nur77 has received particular attention given its role in thymocyte apoptosis and in negative selection (Hazel et al 1988, Milbrandt 1988, Ryseck et al 1989). Constitutive expression of Nur77 in thymocytes results in a dramatic involution

118

VERDIN ET AL

of the thymus, while expression of a dominant-negative Nur77 interferes with negative selection (Calnan et al 1995, Woronicz et al 1994). The expression of Nur77 in response to antigen receptor signals is tightly controlled through two MEF2 binding sites in the Nur77 promoter (Woronicz et al 1995) (Fig. 2). Interestingly, MEF2 transcription factors appear bound constitutively to the Nur77 promoter, irrespective of its transcriptional activity. The regulation of Nur77 promoter activity operates at the level of the proteins bound by MEF2: co-repressors are bound to MEF2 when the promoter is silenced while coactivators, p300/CBP and ERK5, are bound to MEF2 during transcriptional activation (Fig. 2). MEF2-dependent transcription of Nur77 in T cells is activated by calcium signals and by ectopic expression of activated Cam kinase IV, calcineurin, or ERK5/BMK1 (Blaeser et al 2000, Kasler et al 2000, Youn et al 2000a) (Fig. 2). As discussed below, an isoform of MEF2, MEF2D, is expressed in thymocytes and associates with a histone deacetylase (HDAC), and with Cabin-1 under basal conditions in T cell lines (Dequiedt et al 2003, Youn et al 1999, 2000a,b, Youn & Liu 2000). These factors are thought to contribute to the inhibition of Nur77 transcription under basal conditions (Fig. 2).

Class II histone deacetylases: basic properties Eighteen distinct human HDACs are grouped into three classes based on their primary homology to three Saccharomyces cerevisiae HDACs. Class I HDACs (HDAC1, -2, -3, -8 and -11) are homologous to yRPD3, share a compact structure, and are predominantly nuclear proteins expressed in most tissues and cell lines (reviewed in Fischle et al 2001a). Class II HDACs are homologous to yHDA1 and are subdivided in two subclasses, IIa (HDAC4, -5, -7 and -9 and its splice variant MITR) and IIb (HDAC6 and HDAC10), based on sequence homology and domain organization (Fischer et al 2002, 1999, Gao et al 2002, Grozinger et al 1999, Guardiola & Yao 2002, Kao et al 2002, Miska et al 1999, Tong et al 2002, Wang et al 1999). Class III HDACs are homologous to ySIR2 and show no homology to class I and II proteins. The Class IIa HDACs, HDAC4, -5, -7 and -9 contain a highly conserved Cterminal catalytic domain and a regulatory N-terminal domain (see Fig. 5A for HDAC7 organization) (Fischle et al 1999, Grozinger et al 1999, Miska et al 1999, Verdel & Khochbin 1999, Wang et al 1999). The activity of the class IIa HDACs is regulated at several levels, including tissue-speci¢c gene expression, recruitment of distinct cofactors and nucleocytoplasmic shuttling (see Verdin et al 2003 for recent review on this family of proteins). HDAC7 is the only class II HDAC expressed at signi¢cant levels in developing thymocytes (Dequiedt et al 2003).

FIG. 2.

Model for the regulation of Nur77 gene expression by corepressors and coactivators. See text for details.

HDACs AND THYMOCYTE APOPTOSIS 119

120

VERDIN ET AL

Class II HDACs represent the catalytic components of large multiprotein complexes. They do not bind directly to DNA and are thought to be recruited to speci¢c promoters through their interaction with DNA sequence-speci¢c transcription factors. These include the MEF2 transcription factors (Dressel et al 2001, Kao et al 2001, Lemercier et al 2000, Sparrow et al 1999, Wang et al 1999), 143-3 proteins (Grozinger & Schreiber 2000, Kao et al 2001, McKinsey et al 2000a, Wang et al 2000, Zhang et al 2001a), calmodulin (CaM) (Youn et al 2000b), transcriptional corepressors SMRT and N-CoR (Downes et al 2000, Fischle et al 2001b, 2002, Grozinger et al 1999, Huang et al 2000, Kao et al 2000), and others (see Verdin et al 2003 for review) (see Fig. 2 for illustration). All class IIa HDACs shuttle between the nucleus and the cytoplasm (Dressel et al 2001, Grozinger & Schreiber 2000, Kao et al 2001, McKinsey et al 2000b, Miska et al 1999, 2001, Wang et al 2000, Zhao et al 2001). Class IIa HDACs bind to 14-3-3 proteins in a manner dependent on the phosphorylation of two or three conserved N-terminal serines in the N-terminus of class IIa HDACs. This binding mediates their cytoplasmic sequestration, probably by masking a nuclear import signal and activating a nuclear export signal (Grozinger & Schreiber 2000, Kao et al 2001, McKinsey et al 2001, Wang et al 2000, Wang & Yang 2001) (Fig. 2). The nucleocytoplasmic shuttling of class II HDACs regulates their activities as transcriptional repressor proteins. Overexpression of constitutively active CamKs or signal-dependent activation of kinase induces the relocalization of class IIa HDACs to the cytoplasm and suppresses their repressive activity (Lu et al 2000, McKinsey et al 2000a). In contrast, mutation of the phosphorylation sites of class IIa HDACs abolishes their cytoplasmic export and enhances their repressive e¡ects during muscle di¡erentiation and T cell apoptosis (McKinsey et al 2000c, Miska et al 2001). Cytoplasmic relocalization of class II HDACs removes these enzymes from their substrates, the histone proteins in chromatin, and dissociates them from the N-CoR/SMRT-HDAC3 complex, leading to their enzymatic inactivation (Fischle et al 2001b, 2002) (Fig. 2).

HDAC7 is speci¢cally expressed in the thymus and interacts with MEF2D to regulate the Nur77 promoter Analysis of human tissues with a human HDAC7 probe showed the gene to be mostly expressed in human thymus (Fig. 3A). Comparative analysis of relative expression levels of HDAC7, HDAC4 and HDAC5 levels by real-time PCR showed that HDAC7 was signi¢cantly more abundant than HDAC4 and HDAC5 in human thymus. In situ hybridization with HDAC7 antisense RNA probe showed that cells of lymphoid morphology containing high levels of HDAC7 transcripts were abundant in the cortical region of the thymus.

FIG. 3. Human HDAC7 is highly expressed in double positive thymocytes. (A) A multiple human tissue Northern blot was probed with a human HDAC7 32P-radiolabelled cDNA probe. The same membrane was hybridized with a probe corresponding to human GAPDH cDNA. Molecular size markers are indicated on the right. (B) Thymocytes were sorted based on surface expression of CD3, CD4 and CD8, into triple negative (TN), double positive (DP), and single positive (SP4 and SP8). Expression of class II HDACs was assessed in each subset by real-time RT-PCR and normalized to GAPDH expression. Class II HDAC mRNA abundance for HDAC4 (white bars), HDAC5 (grey bars) and HDAC7 (black bars) are plotted relative to mRNA levels in TN thymocytes. (Reproduced from Dequiedt et al 2003, with permission.)

HDACs AND THYMOCYTE APOPTOSIS 121

122

VERDIN ET AL

To determine the expression pattern of HDAC7 during thymocyte maturation, the relative abundance of HDAC4, HDAC5 and HDAC7 mRNAs was assessed by real time PCR in primary human thymocytes sorted by £ow cytometry based on the di¡erentiation markers CD3, CD4 and CD8. Low levels of HDAC7 mRNA were observed in immature triple-negative thymocytes (TN, CD3
CD4
CD8
) (Fig. 3B). HDAC7 expression was highly and transiently increased during the CD4+CD8+ DP stage and returned to lower levels in mature SP T cells (SP4 or SP8). These observations demonstrate the selective up-regulation of HDAC7 expression in DP thymocytes. To test the model that HDAC7 regulates Nur77 expression via MEF2D, we generated polyclonal cell lines stably expressing HDAC7 or HDAC7 mutants (H656A and H657A). Nur77 protein induction was strongly inhibited by wildtype HDAC7 and by the catalytically active H656A mutant, but not by the catalytically inactive H657A mutant (Fig. 4A). We also observed that endogenous MEF2D, a MEF2 protein highly expressed in the thymus, coimmunoprecipitated with endogenous HDAC7 in both T cell hybridomas and primary thymocytes (Fig. 4B). This interaction was mediated by a 17 amino acid stretch conserved within HDAC4 and HDAC5 (Fig. 4C, D) as shown by deletion and substitution analysis (Fig. 4D). Importantly, the HDAC7 mutants defective for MEF2D binding (HDAC7-DMEF and HDAC7-K86AK88A) had no e¡ect on Nur77 promoter activity in transfection assays (Fig. 4E). In contrast, both wild-type HDAC7 and the HDAC7-Q87AE91A mutant inhibited transcriptional activation of the Nur77 reporter (Fig. 4E). To further demonstrate the speci¢c recruitment of HDAC7 to the Nur77 promoter, we fused the N-terminus of HDAC7 to the herpes simplex virus VP16 transactivation domain. We observed that the endogenous Nur77 gene was activated in cells transfected with this construct (data not shown). The same construct lacking the MEF2D interacting domain had no e¡ect on Nur77 expression. Regulation of thymocyte apoptosis by HDAC7 Phosphorylation of serine residues in the amino terminus of class IIa HDACs is required for their nuclear export to the cytoplasm (Grozinger & Schreiber 2000, Kao et al 2001, McKinsey et al 2000a,b, Miska et al 2001, Wang et al 2000, Zhang et al 2001b). Each of these conserved serine residues were mutated to an alanine individually (S155A, S318A, and S448A) or as a combined triple mutant (DP) (Fig. 5A). We observed that the triple mutant (HDAC7-DP) almost totally abolished induction of Nur77 transcription by P/I treatment and behaved as a super-repressor, while single mutations inhibited Nur77 transcriptional activity slightly better than wild-type HDAC7 (not shown). In unstimulated cells,

FIG. 4. HDAC7 represses Nur77 promoter activity via MEF2D. (A) Total cellular lysates were prepared from polyclonal DO11.10 cells stably expressing HDAC7, HDAC7 H656A, HDAC7 H657A, or pCDNA3.1 (Mock); cells were treated with PMA/ionomycin (P/I) for 1.5 h. Cell lysates were analysed by Western blotting for Nur77, FLAG and actin. (B) Total cellular extracts from DO11.10 cells or primary thymocytes were immunoprecipitated with an antiserum speci¢c for HDAC7 or an isotype control. Immunoprecipitated material was analysed by Western blotting with an antiserum against MEF2D (a-MEF2D). (C) Sequence comparison of the MEF2D binding domain of HDAC4, HDAC5 and HDAC7 and HDAC7 mutants. (D) HDAC7 and mutants were immunoprecipitated (a-FLAG) and analysed by western blot for MEF2D (aMEF2D) and FLAG (a-FLAG). The lower band represents the light chain from mouse IgGs used in the immunoprecipitation. (E) D011.10 cells were co-transfected with pNur77-Luc and HDAC7 expression vectors and activated with P/I. Luciferase activity was measured in treated (+) and untreated (
) cells. The activity of the Nur77 promoter is expressed relative to its activity after P/I treatment in mock-transfected cells. (Reproduced from Dequiedt et al 2003, with permission.)

HDACs AND THYMOCYTE APOPTOSIS 123

FIG. 5. Super-repression of Nur77 expression by a nuclear-export defective HDAC7. (A) Schematic representation of HDAC7. The positions of phosphorylated serines are indicated. (B) Immuno£uorescence microscopy of DO11.10 cells transfected with HDAC7-GFP or HDAC7DP-GFP, untreated or treated for 2 h with P/I or aCD3. (C) Polyclonal DO11.10 cell lines, stably expressing HDAC7, HDAC7DP or the empty vector (Mock) were stimulated with aCD3 antibodies. Apoptosis was measured after 24 hours by Phycoerythrin-Annexin V staining and £ow cytometry analysis. The mean apoptotic rates from four independent experiments expressed relative to the apoptotic rate observed in the Mock-transduced cell line (100%) is shown. (D) Polyclonal cell lines transduced with shRNA for either HDAC7 or GL3 luciferase were incubated with aCD3 e antibodies (indicated concentrations) and assayed for apoptosis. The inset shows a Western blot for HDAC7 in cells treated with the shRNA for GL3 or for HDAC. (Reproduced from Dequiedt et al 2003, with permission.)

124 VERDIN ET AL

HDACs AND THYMOCYTE APOPTOSIS

125

HDAC7DP localized mainly in the nucleus, similarly to HDAC7. Wild-type HDAC7 was exported out of the nucleus following stimulation with P/I or aCD3 antibodies while HDAC7DP remained nuclear in response to the same treatments (Fig. 5B). Overexpression of HDAC7 weakly inhibited apoptosis induced by P/I or aCD3 antibodies, while the super-repressor HDAC7DP was associated with signi¢cantly stronger inhibition of apoptosis (Fig. 5C). Conversely, inhibition of HDAC7 expression using RNA interference resulted in increased apoptosis in response to a-CD3 antibody over a range of concentrations (Fig. 5D). These observations indicate that HDAC7 represses Nur77 expression via MEF2D and modulates the rate of apoptosis of T cells in response to TCR engagement. Previous work identi¢ed Cabin-1, a MEF2D binding protein, as an inhibitor of Nur77 expression and TCR-mediated apoptosis (Youn et al 2000b, 1999) (Fig. 2). However, deletion of the MEF2D interacting domain of Cabin-1 did not a¡ect T cell apoptosis in mice, an observation compatible with the existence of redundant regulatory processes, such as HDAC7 (Esau et al 2001).

Summary and conclusions According to current models of T cell maturation, thymic negative and positive selections are directly dependent on signals delivered by the activation of the TCR. Based on our observations, we speculate that HDAC7 plays a critical role in determining the threshold level at which a developing T cell undergoes positive vs. negative selection. By inhibiting Nur77-mediated apoptosis as shown above, HDAC7 could inhibit negative selection and indirectly promote positive selection (Fig. 1). It is likely that HDAC7 regulates the expression of other genes in developing thymocytes, either via MEF2D as shown for Nur77, or via other interacting proteins (see Verdin et al 2003 for review). According to this model, HDAC7 coordinately represses the transcriptional activity of a number of genes under basal conditions in developing thymocytes. During the maturation of DP T cells, signals emanating from the TCR lead to the displacement of HDAC7 from MEF2 and its shuttling to the cytoplasm (Fig. 2). The displacement of HDAC7 from the Nur77 promoter, and other promoters regulated in a similar manner, leads to the activation of Nur77 and apoptosis. Our observations, that overexpression of HDAC7 inhibits apoptosis and that inhibition of HDAC7 expression via RNA interference enhances apoptosis, support this model. Future experiments in animal models will directly test this hypothesis and further establish the role of HDAC7 in thymocyte development and immune tolerance.

126

VERDIN ET AL

Acknowledgements We thank John Carroll for graphics and Sarah Sande for administrative assistance. This work was supported in part by the NIH and by institutional funds from the Gladstone Institute of Virology and Immunology. F.D. is a research associate from the Belgian National Fund of Scienti¢c Research and was supported in part by a fellowship from the International Agency for Research on Cancer.

References Alberola-Ila J, Hernandez-Hoyos G 2003 The Ras/MAPK cascade and the control of positive selection. Immunol Rev 191:79^96 Alberola-Ila J, Hogquist KA, Swan KA, Bevan MJ, Perlmutter RM 1996 Positive and negative selection invoke distinct signaling pathways. J Exp Med 184:9^18 Anderson MS, Venanzi ES, Klein L et al 2002 Projection of an immunological self shadow within the thymus by the aire protein. Science 298:1395^1401 Blaeser F, Ho N, Prywes R, Chatila TA 2000 Ca(2+)-dependent gene expression mediated by MEF2 transcription factors. J Biol Chem 275:197^209 Buhlmann JE, Elkin SK, Sharpe AH 2003 A Role for the B7-1/B7-2:CD28/CTLA-4 pathway during negative selection. J Immunol 170:5421^5428 Calnan BJ, Szychowski S, Chan FK, Cado D, Winoto A 1995 A role for the orphan steroid receptor Nur77 in apoptosis accompanying antigen-induced negative selection. Immunity 3:273^282 Cheng LE, Chan FK, Cado D, Winoto A 1997 Functional redundancy of the Nur77 and Nor-1 orphan steroid receptors in T-cell apoptosis. EMBO J 16:1865^1875 Dequiedt F, Kasler H, Fischle W et al 2003 HDAC7, a thymus-speci¢c class II histone deacetylase, regulates Nur77 transcription and TCR-mediated apoptosis. Immunity 18: 687^698 Downes M, Ordentlich P, Kao HY, Alvarez JG, Evans RM 2000 Identi¢cation of a nuclear domain with deacetylase activity. Proc Natl Acad Sci USA 97:10330^10335 Dressel U, Bailey PJ, Wang SC, Downes M, Evans RM, Muscat GE 2001 A dynamic role for HDAC7 in MEF2-mediated muscle di¡erentiation. J Biol Chem 276:17007^17013 Esau C, Boes M, Youn HD, Tatterson L, Liu JO, Chen J 2001 Deletion of calcineurin and myocyte enhancer factor 2 (MEF2) binding domain of Cabin1 results in enhanced cytokine gene expression in T cells. J Exp Med 194:1449^1459 Fischer DD, Cai R, Bhatia U et al 2002 Isolation and characterization of a novel class II histone deacetylase, HDAC10. J Biol Chem 277:6656^6666 Fischle W, Emiliani S, Hendzel MJ et al 1999 A new family of human histone deacetylases related to Saccharomyces cerevisiae HDA1p. J Biol Chem 274:11713^11720 Fischle W, Kiermer V, Dequiedt F, Verdin E 2001a The emerging role of class II histone deacetylases. Biochem. Cell Biol 79:337^348 Fischle W, Dequiedt F, Fillion M, Hendzel MJ, Voelter W, Verdin E 2001b Human HDAC7 histone deacetylase activity is associated with HDAC3 in vivo. J Biol Chem 276:35826^35835 Fischle W, Dequiedt F, Hendzel MJ et al 2002 Enzymatic activity associated with class II HDACs is dependent on a multiprotein complex containing HDAC3 and SMRT/N-CoR. Mol Cell 9:45^57 Foy TM, Page DM, Waldschmidt TJ et al 1995 An essential role for gp39, the ligand for CD40, in thymic selection. J Exp Med 182:1377^1788 Gao L, Cueto MA, Asselbergs F, Atadja P 2002 Cloning and functional characterization of HDAC11, a novel member of the human histone deacetylase family. J Biol Chem 277:25748^25755

HDACs AND THYMOCYTE APOPTOSIS

127

Grozinger CM, Schreiber SL 2000 Regulation of histone deacetylase 4 and 5 and transcriptional activity by 14-3-3-dependent cellular localization. Proc Natl Acad Sci USA 97:7835^7840 Grozinger CM, Hassig CA, Schreiber SL 1999 Three proteins de¢ne a class of human histone deacetylases related to yeast Hda1p. Proc Natl Acad Sci USA 96:4868^4873 Guardiola AR, Yao TP 2002 Molecular cloning and characterization of a novel histone deacetylase HDAC10. J Biol Chem 277:3350^3356 Hazel TG, Nathans D, Lau LF 1988 A gene inducible by serum growth factors encodes a member of the steroid and thyroid hormone receptor superfamily. Proc Natl Acad Sci USA 85:8444^8448 Healy JI, Goodnow CC 1998 Positive versus negative signaling by lymphocyte antigen receptors. Annu Rev Immunol 16:645^670 Huang EY, Zhang J, Miska EA, Guenther MG, Kouzarides T, Lazar MA 2000 Nuclear receptor corepressors partner with class II histone deacetylases in a Sin3-independent repression pathway. Genes Dev 14:45^54 Kao HY, Downes M, Ordentlich P, Evans RM 2000 Isolation of a novel histone deacetylase reveals that class I and class II deacetylases promote SMRT-mediated repression. Genes Dev 14:55^66 Kao HY, Verdel A, Tsai CC, Simon C, Juguilon H, Khochbin S 2001 Mechanism for nucleocytoplasmic shuttling of histone deacetylase 7. J Biol Chem 276:47496^47507 Kao HY, Lee CH, Komarov A, Han CC, Evans RM 2002 Isolation and characterization of mammalian HDAC10, a novel histone deacetylase. J Biol Chem 277:187^193 Kasler HG, Victoria J, Duramad O, Winoto A 2000 ERK5 is a novel type of mitogen-activated protein kinase containing a transcriptional activation domain. Mol Cell Biol 20:8382^8389 Kishimoto H, Sprent J 1999 Several di¡erent cell surface molecules control negative selection of medullary thymocytes. J Exp Med 190:65^73 Kishimoto H, Sprent J 2001 A defect in central tolerance in NOD mice. Nat Immunol 2: 1025^1031 Lemercier C, Verdel A, Galloo B, Curtet S, Brocard MP, Khochbin S 2000 mHDA1/HDAC5 histone deacetylase interacts with and represses MEF2A transcriptional activity. J Biol Chem 275:15594^15599 Lu J, McKinsey TA, Zhang CL, Olson EN 2000 Regulation of skeletal myogenesis by association of the MEF2 transcription factor with class II histone deacetylases. Mol Cell 6:233^244 McKinsey TA, Zhang CL, Olson EN 2000a Activation of the myocyte enhancer factor-2 transcription factor by calcium/calmodulin-dependent protein kinase-stimulated binding of 14-3- 3 to histone deacetylase 5. Proc Natl Acad Sci USA 97:14400^14405 McKinsey TA, Zhang CL, Lu J, Olson EN 2000b Signal-dependent nuclear export of a histone deacetylase regulates muscle di¡erentiation. Nature 408:106^111 McKinsey TA, Zhang CL, Olson EN 2001 Identi¢cation of a signal-responsive nuclear export sequence in class II histone deacetylases. Mol Cell Biol 21:6312^6321 Milbrandt J 1988 Nerve growth factor induces a gene homologous to the glucocorticoid receptor gene. Neuron 1:183^188 Miska EA, Karlsson C, Langley E, Nielsen SJ, Pines J, Kouzarides T 1999 HDAC4 deacetylase associates with and represses the MEF2 transcription factor. EMBO J 18:5099^5107 Miska EA, Langley E, Wolf D, Karlsson C, Pines J, Kouzarides T 2001 Di¡erential localization of HDAC4 orchestrates muscle di¡erentiation. Nucleic Acids Res 29:3439^3447 Ohashi PS 2002 T-cell signalling and autoimmunity: molecular mechanisms of disease. Nat Rev Immunol 2:427^438 Ohashi PS, DeFranco AL 2002 Making and breaking tolerance. Curr Opin Immunol 14: 744^759

128

VERDIN ET AL

Page DM 1999 Cutting edge: thymic selection and autoreactivity are regulated by multiple coreceptors involved in T cell activation. J Immunol 163:3577^3581 Palmer E 2003 Negative selectionclearing out the bad apples from the T-cell repertoire. Nat Rev Immunol 3:383^391 Peterson P, Nagamine K, Scott H et al 1998 APECED: a monogenic autoimmune disease providing new clues to self-tolerance. Immunol Today 19:384^386 Punt JA, Osborne BA, Takahama Y, Sharrow SO, Singer A 1994 Negative selection of CD4+CD8+ thymocytes by T cell receptor-induced apoptosis requires a costimulatory signal that can be provided by CD28. J Exp Med 179:709^713 Rincon M, Whitmarsh A, Yang DD et al 1998 The JNK pathway regulates the in vivo deletion of immature CD4(+)CD8(+) thymocytes. J Exp Med 188:1817^1830 Ryseck RP, Macdonald-Bravo H, Mattei MG, Ruppert S, Bravo R 1989 Structure, mapping and expression of a growth factor inducible gene encoding a putative nuclear hormonal binding receptor. EMBO J 8:3327^3335 Sabapathy K, Kallunki T, David JP, Graef I, Karin M, Wagner EF 2001 c-Jun NH2-terminal kinase (JNK)1 and JNK2 have similar and stage-dependent roles in regulating T cell apoptosis and proliferation. J Exp Med 193:317^328 Sebzda E, Mariathasan S, Ohteki T, Jones R, Bachmann MF, Ohashi PS 1999 Selection of the T cell repertoire. Annu Rev Immunol 17:829^874 Sparrow DB, Miska EA, Langley E et al 1999 MEF-2 function is modi¢ed by a novel corepressor, MITR. EMBO J 18:5085^5098 Sprent J, Kishimoto H 2001 The thymus and central tolerance. Transplantation 72:S25^S28 Starr TK, Jameson SC, Hogquist KA 2003 Positive and negative selection of T cells. Annu Rev Immunol 21:139^176 Sugawara T, Moriguchi T, Nishida E, Takahama Y 1998 Di¡erential roles of ERK and p38 MAP kinase pathways in positive and negative selection of T lymphocytes. Immunity 9:565^574 Suzuki A, Yamaguchi MT, Ohteki T et al 2001 T cell-speci¢c loss of Pten leads to defects in central and peripheral tolerance. Immunity 14:523^534 Swan KA, Alberola-Ila J, Gross JA et al 1995 Involvement of p21ras distinguishes positive and negative selection in thymocytes. EMBO J 14:276^285 Tong JJ, Liu J, Bertos NR, Yang XJ 2002 Identi¢cation of HDAC10, a novel class II human histone deacetylase containing a leucine-rich domain. Nucleic Acids Res 30:1114^1123 Verdel A, Khochbin S 1999 Identi¢cation of a new family of higher eukaryotic histone deacetylases. Coordinate expression of di¡erentiation-dependent chromatin modi¢ers. J Biol Chem 274:2440^2445 Verdin E, Dequiedt F, Kasler HG 2003 Class II histone deacetylases: versatile regulators. Trends Genet 19:286^293 Wang AH, Yang XJ 2001 Histone deacetylase 4 possesses intrinsic nuclear import and export signals. Mol Cell Biol 21:5992^6005 Wang AH, Bertos NR, Vezmar M et al 1999 HDAC4, a human histone deacetylase related to yeast HDA1, is a transcriptional corepressor. Mol Cell Biol 19:7816^7827 Wang AH, Kruhlak MJ, Wu J et al 2000 Regulation of histone deacetylase 4 by binding of 14-3-3 proteins. Mol Cell Biol 20:6904^6912 Woronicz JD, Calnan B, Ngo V, Winoto A 1994 Requirement for the orphan steroid receptor Nur77 in apoptosis of T-cell hybridomas. Nature 367:277^281 Woronicz JD, Lina A, Calnan BJ, Szychowski S, Cheng L, Winoto A 1995 Regulation of the Nur77 orphan steroid receptor in activation-induced apoptosis. Mol Cell Biol 15:6364^6376 Youn HD, Liu JO 2000 Cabin1 represses MEF2-dependent Nur77 expression and T cell apoptosis by controlling association of histone deacetylases and acetylases with MEF2. Immunity 13:85^94

HDACs AND THYMOCYTE APOPTOSIS

129

Youn HD, Sun L, Prywes R, Liu JO 1999 Apoptosis of T cells mediated by Ca2+-induced release of the transcription factor MEF2. Science 286:790^793 Youn HD, Chatila TA, Liu JO 2000a Integration of calcineurin and MEF2 signals by the coactivator p300 during T-cell apoptosis. EMBO J 19:4323^4331 Youn H-D, Grozinger CM, Liu JO 2000b Calcium regulates transcriptional repression of myocyte enhancer factor 2 by histone deacetylase 4. J Biol Chem 275:22563^22567 Zhang CL, McKinsey TA, Olson EN 2001a The transcriptional corepressor MITR is a signalresponsive inhibitor of myogenesis. Proc Natl Acad Sci USA 98:7354^7359 Zhang CL, McKinsey TA, Lu J-r, Olson EN 2001b Association of COOH-terminal-binding protein (CtBP) and MEF2-interacting transcription repressor (MITR) contributes to transcriptional repression of the MEF2 transcription factor. J Biol Chem 276:35^39 Zhao X, Ito A, Kane CD et al 2001 The modular nature of histone deacetylase HDAC4 confers phosphorylation-dependent intracellular tra⁄cking. J Biol Chem 276:35042^35048

DISCUSSION Olson: Does calcineurin activate Nur77 in your system, and does cyclosporin block? Verdin: We have not tested those yet. Olson: Do you know the identity of the kinase? Verdin: No, but that is something we would love to know. We believe that this kinase is activated by the TCR and leads to the shuttling of HDAC7 out of the nucleus. The identi¢cation of this kinase should provide us with some important leads on the regulation of HDAC7 shuttling. Mahadevan: I noticed that you are producing your translocation by a combination of TPA and ionomycin. Can you produce that with either of those agents alone? Verdin: The TPA signal is the critical one, not the Ca2+ signal. Mahadevan: Do you think that the same kinase phosphorylates those three sites? Verdin: We don’t have any reason to doubt this. Marks: If you expose these cells to HDAC inhibitors, do you induce apoptosis? Verdin: Yes, but we have no evidence that apoptosis induction in response to TSA bears any mechanistic similarity to the apoptosis induction in response to CD3. We see apoptosis induction in most transformed cells. Most of the work I described was done in thymocyte cell lines. Berger: I am curious about the speci¢city of the HDACs. Have you used other HDACs in place of HDAC7 in some of those assays? Verdin: If you compare HDAC7 with other members of the class II family, we have no evidence that HDAC7 is able to do things that HDAC4, 5 and 9 cannot do since all class II HDACs share the same MEF2 binding domain. We believe that HDAC7 is the relevant HDAC in thymocyte development because it is the only class II HDAC expressed at signi¢cant levels in these cells.

130

DISCUSSION

Khochbin: You have previously published that HDAC4 and 5 don’t have catalytic activity. Is this also true for HDAC7? You also showed that the deacetylase activity comes from HDAC3: what is the role of HDAC3 in the regulation of Nur77? Verdin: We have shown that HDAC7 is exactly the same as HDAC4 and HDAC5 in this regard. We have shown that class II HDACs do not have intrinsic HDAC activity. They are able to recruit a corepressor complex SMRT/ N-CoR and this complex contains HDAC3. Our model is that this HDAC3 represents the bona ¢de HDAC activity in this complex. The same applies to HDAC7. One experiment we need to do is to see what would happen to Nur77 regulation after knockdown of HDAC3 via siRNA. Seto: Have you attempted ChIP assays with an anti-HDAC7 antibody on the Nur77 promoter? Verdin: We have been unable to conduct these experiments so far. Seto: Have you tried HDAC4 and HDAC5 in your system? Verdin: No, but others have. Seto: The assumption in the ¢eld that class II HDACs are tissue speci¢c is mainly based on Clontech Northern blots. Has anyone done a comprehensive study to examine class II HDAC protein expression levels in di¡erent tissues? Verdin: We have not done this at this point. Olson: The embryonic expression patterns of some of them are very speci¢c. Li: Is Nur77 the only target for HDAC7 and MEF2D? What happens if you overexpress Nur77? Verdin: There are a number of other genes regulated by HDAC7 in thymocytes. We have conducted microarray analysis and have identi¢ed a subset of genes that are regulated via the same mechanism. Li: Are those other genes also regulated by MEF2D? Verdin: Yes, most of them are. Gu: Does HDAC7 form a complex in vivo? The reason I ask is that many HDACs form complexes. The direct binding between HDAC7 and MEF2D doesn’t seem very strong. Is there any possibility that there is another protein that helps with the binding? Verdin: We have some evidence that HDAC7 interacts with HDAC3 in the context of the SMRT/N-CoR complex but we don’t have any evidence on the other proteins. Turner: Annick Harel-Bellan (CNRS, Villejuif, France) has done some nice immunoprecipitation with HDACs in 3T3 cells looking at changes on the stimulation. I had a student who spent some time in Annick’s lab and then came back and has done some ChIP on HDAC1, 2, 3 and 4 in embryonic stem (ES) cells. This worked. It may depend on the cell type. We didn’t pursue this because the results turned out to be quite boringthe HDACs were just everywhere at all

HDACs AND THYMOCYTE APOPTOSIS

131

stages of ES cell di¡erentiation. The actual ChIP procedure seemed to be working OK. Verdin: Did you have to do anything special to modify this ChIP assay? In general, it has not been easy to ChIP HDACs, especially the class II proteins. Turner: Everything we did was with class I. The only class II we tried was HDAC6 and this didn’t work. This might have been a problem with the antibody we were using. The antibodies we have to HDACs I, II and III are higher titre. Greene: Would you care to speculate about HDAC7 and autoimmunity? For example, have you looked at any of the mouse lines with autoimmune diseases to see whether there are abnormally high levels of expression of HDAC7? Verdin: That is an interesting idea, but we haven’t looked at this. Cohen: What do you think of the recent paper proposing the use of HDAC inhibitors for the inhibition of T cell proliferation and as an immunosuppressors? It claimed that they were better than cyclosporine A (Skov et al 2003). Verdin: If you use HDAC inhibitors in non-transformed cells you are inducing cell-cycle arrest. There is another element there. These inhibitors have a degree of non-speci¢city. For example, HDAC inhibitors also induce thrombocytopaenia. My opinion is that these are more likely to be non-speci¢c antiproliferative e¡ects rather than selective activities but only time will tell whether these agents can be successfully used as immunosuppressors. Cohen: You were asking about ChIP with HDACs. We have been able to do this with HDAC1, 2 and 6. In our hands, ChIP works with HDAC1 and HDAC2. Reference Skov S, Rieneck K, Bovin LF et al 2003 Histone deacetylase inhibitors: a new class of immunosuppressors targeting a novel signal pathway essential for CD154 expression. Blood 101:1430^1438

Dual roles of histone deacetylases in the control of cardiac growth Timothy A. McKinsey* and Eric N. Olson{1 *Myogen, Inc., 7575 West 103rd Avenue, Westminster, CO 80021, and {Department of Molecular Biology, The University of Texas Southwestern Medical Center at Dallas, 6000 Harry Hines Boulevard, Dallas, TX 75390-9148, USA

Abstract. Diverse aetiological factors, including myocardial infarction, hypertension and contractile abnormalities, trigger a cardiac remodelling process in which the heart becomes abnormally enlarged with a consequent decline in cardiac function and eventual heart failure. Pathological cardiac hypertrophy is accompanied by the activation of a fetal cardiac gene programme, which contributes to maladaptive changes in contractility and calcium handling. Traditional treatment for heart failure involves administration of drugs that antagonize early signalling events at or near the cell membrane (e.g. cell surface receptor or ion channels). Given the complexity and redundant nature of the signalling networks that drive cardiac pathogenesis, a potentially more e⁄cacious therapeutic strategy for disrupting the disease process would be to target common downstream elements in pathological signalling cascades. We have shown that class II histone deacetylases (HDACs) suppress cardiac hypertrophy, and mice lacking class II HDACs are sensitized to hypertrophic signals. Paradoxically, HDAC inhibitors also block cardiac hypertrophy and fetal gene activation. Based on these ¢ndings, we propose that distinct HDACs play positive or negative roles in the control of cardiac growth by regulating opposing sets of target genes via their interactions with di¡erent sets of transcription factors. 2004 Reversible protein acetylation. Wiley, Chichester (Novartis Foundation Symposium 259) p 132^145

Postnatal cardiac myocytes undergo hypertrophic growth in response to a variety of stress signals (reviewed in Frey & Olson 2003). The hypertrophic response is characterized by an increase in myocyte size and protein synthesis, assembly and organization of sarcomeres, and activation of a fetal gene programme. Although traditionally considered an adaptive response to pathological signalling, prolonged expression of fetal cardiac genes in the adult heart can result in maladaptive changes in cardiac contractility and calcium handling that 1This paper was presented at the symposium by Eric N. Olson to whom correspondence should

be addressed. 132

HDACs IN CARDIAC GROWTH

133

culminate in dilated cardiomyopathy, heart failure and sudden death from arrhythmias (Lowes et al 2002). Moreover, increasing evidence in rodent models indicates that cardiac function is preserved when hypertrophy is inhibited in the face of stress signalling, pointing to the potential therapeutic bene¢t of strategies for suppressing the hypertrophic process (Hill et al 2000, Harding et al 2001, Rothermel et al 2001, Antos et al 2002). Several calcium-dependent signalling molecules, including the calcium/ calmodulin-dependent protein phosphatase calcineurin and calcium/calmodulindependent protein kinases (CaMKs) have been implicated in the transduction of hypertrophic stimuli (Frey & Olson 2003). Recent studies suggest that these signalling pathways control the hypertrophic gene programme by modulating chromatin conformation via their e¡ects on the activities, subcellular localization, and protein^protein interactions of histone acetyltransferases (HATs) and histone deacetylases (HDACs). HATs activate transcription by acetylating nucleosomal histones, resulting in relaxation of nucleosomal structure. The actions of HATs are opposed by HDACs, which deacetylate histones, promoting chromatin condensation and transcriptional repression (Jenuwein & Allis 2001, Johnson & Turner 1999). The responsiveness of these chromatin-remodelling factors to hypertrophic signalling pathways provides a mechanism for linking stress signals to the fetal cardiac gene programme. Mammalian HDACs can be divided into three classes based on their similarity with the three yeast HDACs (reviewed in Gray & Ekstrom 2001). Class I HDACs (HDACs 1, 2, 3 and 8) are expressed ubiquitously and consist mainly of a deacetylase domain (Fig. 1). Members of class II (HDACs 4, 5, 7 and 9) are highly expressed in striated muscle and brain and have an extended N-terminus in addition to the catalytic domain. Class III HDACs resemble the yeast HDAC Sir2, which is activated by nicotinamide adenine dinucleotide. Class II HDACs interact with a variety of positive and negative cofactors as well as other HDACs through their N-terminal regions, which contain two conserved phosphorylation sites for CaM kinase and other kinases that confer responsiveness to calcium signalling (Fischle et al 2001) (Fig. 1). Phosphorylation of these sites creates docking sites for 14-3-3 chaperone proteins, which escort the HDACs out of the nucleus, resulting in activation of genes that would otherwise be repressed by these HDACs (Fig. 2) (Grozinger & Schreiber 2000, McKinsey et al 2001, Kao et al 2001). One of the important targets of HDACs in muscle cells is the MEF2 transcription factor (reviewed in McKinsey et al 2002). In the adult heart, MEF2 proteins exhibit only basal activity, which is dramatically enhanced by calciumdependent stress signals that induce hypertrophy (Passier et al 2000). The activation of MEF2 by hypertrophic signals can be explained, at least in part, by the phosphorylation-dependent dissociation of class II HDACs and subsequent recruitment of HATs by MEF2.

134

McKINSEY & OLSON

FIG. 1. Schematic diagrams of class I and class II HDACs. Class I HDACs 1, 2, 3 and 8 contain only a catalytic domain. Class II HDACs 4, 5, 7 and 9 contain a catalytic domain at their C-termini and an N-terminal extension that mediates interactions with the MEF2 transcription factor and other coactivators and corepressors. Binding of calcium/calmodulin (CaM) to the MEF2 binding domain of class II HDACs disrupts MEF2:HDAC interactions. The N-terminal extensions of class II HDACs also contain two signal-responsive serines that are bound by 143-3 upon phosphorylation. Binding of 14-3-3 to these sites triggers nuclear export by masking the nuclear localization sequence (NLS) and unmasking a cryptic nuclear export sequence (NES) near the C-terminus that associates with CRM1.

Several recent observations have implicated HDACs and HATs in the control of cardiac hypertrophy. First, expression of signal-resistant mutants of class II HDACs in primary cardiomyocytes silences the fetal gene programme and renders myocytes insensitive to hypertrophic agonists (Zhang et al 2002). Overexpression of class II HDACs also blocks MEF2 activation by hypertrophic stimuli (Lu et al 2000). Second, knockout mice lacking HDAC9 are super-sensitive to stress signals and develop massively hypertrophic hearts when stressed (Zhang et al 2002). Third, the HAT p300 associates with the MEF2 and GATA4 transcription factors, which regulate fetal cardiac genes, and enhances their transcriptional activity (Sartorelli et al 1997, Dai & Markham 2001). Finally, overexpression of p300 induces hypertrophy of primary cardiomyocytes (Gusterson et al 2003). Numerous pharmacological inhibitors of HDAC activity have been identi¢ed, including Trichostatin A (TSA), sodium butyrate (NaB), HC-toxin (HC), subercyl-anilide hydroxamic acid (SAHA) and Pyroxamide. Inhibiting HDAC activity with these agents alters gene expression; however, instead of simply

HDACs IN CARDIAC GROWTH

135

FIG. 2. Control of cardiac hypertrophy by signal-dependent translocation of class II HDACs from the nucleus to the cytoplasm. Class II HDACs interact with MEF2 and repress the fetal gene programme and remodelling of post-natal cardiomyocytes. Stress signals activate an HDAC kinase that promotes HDAC phosphorylation and nuclear export mediated by 14-3-3. Protein kinase C (PKC) is required to activate the HDAC kinase in response to some but not all stimuli. The calcineurin phosphatase stimulates HDAC kinase activity, likely through an indirect mechanism involving autocrine/paracrine pathways. MEF2:HDAC interactions are disrupted by binding of calcium/calmodulin to the MEF2 binding domain of class II HDACs and by association of 14-3-3 with the HDAC. HATs, such as p300, can dock on the region of MEF2 that also binds HDACs, resulting in activation of genes for cardiac remodelling. Factors that promote and repress cardiac remodelling are depicted in green and red, respectively.

activating gene transcription as would be expected to result from repressing the repressive activity of HDACs, these inhibitors result in increases as well as decreases in expression of speci¢c genes (Sambucetti et al 1999, Vogelauer et al 2000, Vigushin et al 2001). In several cell types, HDAC inhibitors have been

136

McKINSEY & OLSON

shown to block growth and promote quiescence and di¡erentiation. HDAC inhibitors have also shown promise as anticancer agents, which may re£ect their ability to selectively derepress expression of the cyclin-dependent protein kinase inhibitor p21 (reviewed in Hassig et al 1997). Results and discussion Exaggerated hypertrophy in HDAC9 mutant mice To determine the functions of class II HDACs in vivo, we have generated knockout mice lacking each of the four class II HDACs  HDAC4, 5, 7 and 9. Mice lacking HDACs 5 and 9 are viable and do not show obvious cardiac defects until late in life (Zhang et al 2002 and our unpublished results). However, if these mutant mice are subjected to cardiac stress, for example by expression of a cardiac-speci¢c calcineurin transgene, they develop profound cardiac hypertrophy that rapidly progresses to heart failure and sudden death (Fig. 3). The fetal cardiac gene programme is also super-activated in the hearts of HDAC9 knockout mice following exposure to cardiac stress. These ¢ndings have led us to propose a model in which class II HDACs function as stress-sensitive repressors of cardiac hypertrophy. According to this model, the four class II HDACs play redundant roles in the repression of hypertrophy, such that the remaining class II HDACs in HDAC5 or HDAC9 knockout mice would be less e¡ective in counteracting prohypertrophic signals.

FIG. 3. Exaggerated hypertrophy in HDAC9 mutant mice in response to calcineurin signalling. HDAC9 null mice have hearts of normal size, but in response to calcineurin signalling, evoked by cardiac-speci¢c expression of activated calcineurin, the mutant heart displays an exaggerated growth response (Adapted from Zhang et al 2002). Exaggerated hypertrophy is also observed in HDAC9-de¢cient mice subjected to thoracic aortic banding (not shown).

HDACs IN CARDIAC GROWTH

137

FIG. 4. Blockade to ANF expression by HDAC inhibitors. Primary cardiomyocytes were treated with PE (20 mM) in the absence or presence of HDAC inhibitors (TSA, 85 nM; NaB, 5 mM; HC-toxin, 5 ng/ml) and ANF mRNA was detected by dot blot analysis. The results are graphed as the means  S.D. from four independent experiments.

HDAC inhibitors block cardiac hypertrophy In light of the ability of class II HDACs to suppress cardiac hypertrophy, we sought to determine whether HDAC inhibitors would mimic the e¡ect of genetic deletion of HDAC5/9 and thereby promote cardiac hypertrophy. Paradoxically, we found that HDAC inhibitors prevent hypertrophy, sarcomere organization and activation of the fetal gene programme normally evoked by hypertrophic agonists (Antos et al 2003). The e¡ects of the HDAC inhibitors TSA, NaB, or HC-toxin on expression of the hypertrophic marker ANF in the presence of the hypertrophic agonist PE are shown in Fig. 4. Some reports have associated HDAC inhibitors with cell death (reviewed in Marks et al 2000). However, neither TSA nor NaB were cytotoxic to cardiomyocytes at concentrations in which they completely block hypertrophy, indicating that the suppressive e¡ects on hypertrophy of HDAC inhibition are not an indirect consequence of cytotoxicity. HDAC inhibitors not only prevented hypertrophy, but also stimulated expression of a-MHC, which is normally down-regulated during hypertrophy. These ¢ndings suggest that these inhibitors fully antagonize the hypertrophic program rather than selectively inhibiting only a subset of genes involved in this

138

McKINSEY & OLSON

FIG. 5. Opposing roles of class I and class II HDACs in control of the gene programme for pathological cardiac hypertrophy. (A) Hypertrophy of cardiomyocytes is blocked by class II HDACs or by HDAC inhibitors. The anti-hypertrophic action of HDAC inhibitors may be mediated by their e¡ects on class I HDACs. (B) The opposing activities of class I and class II HDACs are likely to re£ect their repression of anti-growth and pro-growth genes, respectively. A novel homeodomain protein, HOP, has been proposed to relay the e¡ects of Class I HDACs to serum response factor (SRF), which regulates anti-growth genes (Shin et al 2002, Chen et al 2002, Kook et al 2003).

process. A previous study reported that a MEF2 binding site upstream of the aMHC gene was required for the repression of the gene (Adolph et al 1993). Given that class II HDACs repress the activity of MEF2, up-regulation of a-MHC in the presence of HDAC inhibitors could re£ect derepression of MEF2 activity at this site or it could represent an indirect e¡ect of these inhibitors on other transcriptional regulators. How can the seemingly paradoxical e¡ects of HDAC inhibitors on cardiomyocyte hypertrophy be reconciled with our previous ¢ndings that class II HDACs block hypertrophy? One possibility is that di¡erent classes of HDACs suppress distinct sets of genes that in£uence the hypertrophic program (Fig. 5).

HDACs IN CARDIAC GROWTH

139

For example, class I HDACs might suppress expression of anti-hypertrophic genes, whereas class II HDACs suppress pro-hypertrophic genes. If the antihypertrophic gene products were dominant over the genes suppressed by class II HDACs, one would expect HDAC inhibitors to block hypertrophy. The ¢nding that HDAC inhibitors mimic the e¡ects of class II HDACs on hypertrophy also raises the possibility that the mechanism whereby class II HDACs suppress hypertrophy and the fetal gene programme is independent of deacetylase activity. In this regard, the deacetylase domains of class II HDACs are not required for transcriptional repression (Fischle et al 2002). Instead, these HDACs recruit other corepressors, as well as class I HDACs, to target genes. Consistent with this conclusion, MITR (MEF2-interacting transcriptional repressor), a splice variant of HDAC9 that lacks a deacetylase domain, is highly e¡ective in repressing MEF2 activity and cardiomyocyte hypertrophy. The HDAC inhibitors used in our studies inhibit both class I and II HDACs (Grozinger et al 1999). Little is known about possible selective e¡ects of HDAC inhibitors on individual HDAC family members. It is unlikely that class III HDACs are involved in the TSA e¡ect, because they are TSA-insensitive. We therefore hypothesize that one or more class I HDACs function upstream of class II HDACs in the hypertrophic pathway. What might be the gene targets for pro-hypertrophic HDACs? One possibility is that such HDACs are required for repression of one or more genes whose products repress hypertrophy. Accordingly, inhibition of these HDACs would result in derepression of such anti-hypertrophic genes and a consequent block to hypertrophy. Several genes have been shown to repress hypertrophy, including those encoding GSK-3b (Antos et al 2002), MCIP (Rothermel et al 2001) and class II HDACs (Zhang et al 2002). We have examined whether HDAC inhibitors might suppress expression of these genes, but have not observed signi¢cant inhibition. HDAC inhibitors have been shown to block tumour cell proliferation, in part, by causing cell death (reviewed in Butler et al 2000). While we cannot formally rule out the possibility that suppression of hypertrophy by HDAC inhibitors is a nonspeci¢c consequence of cytotoxicity, our results argue against this interpretation because multiple inhibitors suppressed hypertrophy at concentrations at least 10fold lower than those that caused cellular demise. Moreover, the concentrations in which these inhibitors suppressed hypertrophy are comparable to those reported in other studies to speci¢cally inhibit histone deacetylation and other cellular processes. HDAC inhibitors have shown e⁄cacy as anticancer agents in humans and animal models (Marks et al 2001, Johnstone 2002). It remains to be determined whether HDAC inhibitors will also prove to be e⁄cacious in suppressing hypertrophy in vivo and, if so, which types of pathological stimuli are coupled to

140

McKINSEY & OLSON

HDAC activation. Nevertheless, our results point to the potential usefulness of such inhibitors in suppressing pathological cardiac hypertrophy.

Acknowledgements E.N.O. was supported by grants from the N.I.H. and the D.W. Reynolds Clinical Cardiovascular Research Center.

References Adolph EA, Subramaniam A, Cserjesi P, Olson EN, Robbins J 1993 Role of myocyte-speci¢c enhancer-binding factor (MEF-2) in transcriptional regulation of the alpha-cardiac myosin heavy chain gene. J Biol Chem 268:5349^5352 Antos CL, McKinsey TA, Dreitz M et al 2003 Does-dependent blockade to cardiomyocyte hypertrophy by histone-deacetylase inhibitors. J Biol Chem 278:28930^28937 Antos CL, McKinsey TA, Frey N et al 2002 Activated glycogen synthase kinase-3b suppresses cardiac hypertrophy in vivo. Proc Natl Acad Sci USA 99:907^912 Butler L, Agus DB, Scher HI et al 2000 Suberoylanilide hydroxamic acid, an inhibitor of histone deacetylase, suppresses the growth of prostate cancer cells in vitro and in vivo. Cancer Res 60:5165^5170 Chen F, Kook H, Milewski R et al 2002 Hop is an unusual homeodomain gene that modulates cardiac development. Cell 110:713^723 Dai Y, Markham B 2001 p300 Functions as a coactivator of transcription factor GATA-4. J Biol Chem 276:37178^37185 Fischle W, Dequiedt F, Hendzel MJ et al 2002 Enzymatic activity associated with class II HDACs is dependent on a multiprotein complex containing HDAC3 and SMRT/N-CoR. Mol Cell 9:45^57 Fischle W, Kiermer V, Dequiedt F, Verdin E 2001 The emerging role of class II histone deacetylases. Biochem Cell Biol 79:337^348 Frey N, Olson EN 2003 Cardiac hypertrophy: the good, the bad and the ugly. Annu Rev Physiol 65:45^79 Gray S, Ekstrom T 2001 The human histone deacetylase family. Exp Cell Res 262:75^83 Grozinger CM, Schreiber SL 2000 Regulation of histone deacetylase 4 and 5 and transcriptional activity by 14-3-3 dependent cellular localization. Proc Natl Acad Sci USA 97:7835^7840 Grozinger CM, Hassig CA, Schreiber SL 1999 Enzymatic activity associated with class II HDACs is dependent on a multiprotein complex containing HDAC3 and SMRT/N-CoR. Proc Natl Acad Sci USA 96:4868^4873 Gusterson R, Jazrawi E, Adcock I, Latchman D 2002 The transcriptional co-activators CBP and p300 are activated via phenylephrine through the p42/p44 MAPK cascade. J Biol Chem 277:2517^2524 Harding VB, Jones LR, Lefkowitz RJ, Koch WJ, Rockman, HA 2001 Cardiac beta ARK1 inhibition prolongs survival and augments beta blocker therapy in a mouse model of severe heart failure. Proc Natl Acad Sci USA 98:5809^5814 Hassig CA, Tong JK, Schreiber SL 1997 Fiber-derived butyrate and the prevention of colon cancer. Chem Biol 4:783^789 Hill JA, Karimi M, Kutschke W et al 2000 Cardiac hypertrophy is not a required compensatory response to short-term pressure overload. Circulation 101:2863^2869 Jenuwein T, Allis CD 2001 Translating the histone code. Science 293:1074^1080

HDACs IN CARDIAC GROWTH

141

Johnson CA, Turner BM 1999 Histone deacetylases: complex transducers of nuclear signals. Semin Cell Dev Biol 10:179^188 Johnstone R 2002 Histone-deacetylase inhibitors: novel drugs for the treatment of cancer. Nat Rev Drug Discov 1:287^299 Kao HY, Verdel A, Tsai CC, Simon C, Juguilon H, Khochbin S 2001 Mechanism for nucleocytoplasmic shuttling of histone deacetylase 7. J Biol Chem 276:47496^47507 Kook H, Lepore JJ, Gitler AD et al 2003 cardiac hypertrophy and histone deacetylasedependent transcriptional repression mediated by the atypical homeodomain protein Hop. J Clin Invest 112:863^871 Lowes BD, Gilbert EM, Abraham WT et al 2002 Myocardial gene expression in dilated cardiomyopathy treated with beta-blocking agents. N Engl J Med 346:1357^1365 Lu J, McKinsey TA, Nicol RL, Olson EN 2000 Signal-dependent activation of the MEF2 transcription factor with class II histone deacetylases. Proc Natl Acad Sci USA 97:4070^4075 Marks P, Richon V, Rifkind R 2000 Histone deacetylase inhibitors: inducers of di¡erentiation or apoptosis of transformed cells. J Natl Cancer Inst 92:1210^1216 Marks P, Rifkind R, Richon V, Breslow R, Miller T, Kelly W 2001 Histone deacetylases and cancer: causes and therapies. Nat Rev Cancer 1:194^202 Marmorstein R 2001 Structure of histone deacetylases: insights into substrate recognition and catalysis. Structure 9:1127^1133 McKinsey T, Zhang C, Olson E 2001 Identi¢cation of a signal-responsive nuclear export sequence in class II histone deacetylases. Mol Cell Biol 21:6312^6321 McKinsey TA, Zhang CL, Olson EN 2002 MEF2: a calcium-dependent regulator of cell division, di¡erentiation and death. Trends Biochem Sci 27:40^47 Passier R, Zeng H, Frey N et al 2000 CaM kinase signaling induces cardiac hypertrophy and activates the MEF2 transcription factor in vivo. J Clin Invest 105:1395^1406 Rothermel BA, McKinsey TA, Vega RB et al 2001 Myocyte-enriched calcineurin-interacting protein, MCIP1, inhibits cardiac hypertrophy in vivo. Proc Natl Acad Sci USA 98:3328^3333 Sambucetti LC, Fischer DD, Zabludo¡ S et al 1999 Histone deacetylase inhibition selectively alters the activity and expression of cell cycle proteins leading to speci¢c chromatin acetylation and antiproliferative e¡ects. J Biol Chem 274:34940^34947 Sartorelli V, Huang J, Hamamori Y, Kedes L 1997 Molecular mechanisms of myogenic coactivation by p300: direct interaction with the activation domain of MyoD and with the MADS box of MEF2C. Mol Cell Biol 17:1010^1026 Shin, CH, Liu ZP, Passier R et al 2002 Modulation of cardiac growth and development by HOP, an unusual homeodomain protein. Cell 110:725^735 Vigushin D, Coombes R 2002 Histone deacetylase inhibitors in cancer treatment. Anticancer Drugs 13:1^13 Vogelauer M, Wu J, Suka N, Grunstein M 2000 Global histone acetylation and deacetylation in yeast. Nature 408:495^498 Zhang CL, McKinsey TA, Chang S, Antos CL, Hill JA, Olson EN 2002 Class II histone deacetylases act as signal-responsive repressors of cardiac hypertrophy. Cell 110:479^488

DISCUSSION Berger: Are you certain that histones are the target for the class II HDACs? Olson: That is not really our forte. We have shown that the fetal cardiac genes undergo acetylation and deacetylation as would be expected in response to hypertrophic signals. In this case acetylation is clearly involved. I don’t know whether there are other targets or mechanisms for the class II HDACs.

142

DISCUSSION

Berger: I am curious about the TSA result with regard to class II associated with class I HDACs. Olson: The TSA result is puzzling but highly reproducible. One possibility is that the target genes that are regulated by class I HDACs might be dominant over the ones that are regulated by class II HDACs. This is something we need to test. Atadja: Do you see any overlap between the genes that were modulated by TSA versus stress, when you do A¡ymetrix-type experiments? Olson: We haven’t done detailed A¡ymetrix experiments on TSA-treated cells. We have mainly focused on known genes in the hypertrophic pathway. They are all coordinately suppressed by TSA or by class II HDACs. Atadja: What are the expression levels of these class II HDACs in heart? Why do you think HDAC9 is the hypertrophy-sensitizing HDAC? Olson: There is a particular splice variant of HDAC9 called MITR which is very highly expressed in the heart. Interestingly, MITR doesn’t have a catalytic domain, but it is just as e¡ective as the other HDACs in blocking hypertrophy. This comes back to the possibility that class II HDACs might not be utilizing an HDAC function for this activity. The other HDACs are expressed in the heart but to varying degrees. Atadja: Going back to the TSA result, with some of our HDAC inhibitors we saw e¡ects on contractility of cardiac myocytes. When treated, there was a lag period and then we saw that the contractility was a¡ected. We could never explain this result. Do you see any relationship at all between the signal that probably mediates hypertrophy and these electrical events? Olson: We haven’t seen that, but I can’t say that we have looked very carefully. It is possible that some of these HDAC inhibitors have di¡erent targets in myocytes. One argument against that is that every HDAC inhibitor we have tried so far has had the same e¡ect. Marks: If I understand you correctly, you are postulating that HDACs function to repress the pro-growth stress genes that would be activated by MEF2. So, when you remove the HDACs functionally with the HDAC inhibitors, you are getting the same e¡ect. Olson: No, we see the opposite e¡ect. Marks: I thought you said that the HDAC inhibitors block cardiac hypertrophy. Olson: Yes, they block hypertrophy. Class II HDACs also block hypertrophy, so their genetic deletion causes it. Li: Have you looked at the target fetal growth promotion genes? Have you looked in the knockout and in the normal situation at how the class II HDACs regulate these genes? For example, is it histone modi¢cation? Olson: The fetal cardiac genes are super-activated by stress in these HDAC knockouts. Their activation is basically parallel to the extent of hypertrophy. We

HDACs IN CARDIAC GROWTH

143

have done the acetylation in cultured cells and this correlates. We haven’t been able to monitor the acetylation state of these genes in the intact myocardium. Verdin: You mentioned that calcineurin seems to be controlling the kinase. This would place the kinase downstream of a calcium signal. I was struck by this, because in T cells, HDAC7 nucleocytoplasmic shuttling is not modulated by calcium signal, but by PMA. Is there something about the kinase that could explain these discrepant results? Olson: The kinase is regulated by a number of upstream kinases. Calcineurin regulates the kinase but this experiment is a little complicated. It is done in a whole heart, chronically activated for calcineurin. So it doesn’t tell us that calcineurin is directly activating the kinase. More likely, calcineurin is transducing a signal and perhaps activating an autocrine loop of signalling that ultimately switches on the kinase. Clearly, the kinase is somewhere in that circuit. Yao: Does the same kinase that you identify in the heart hypertrophy also operate in the skeletal muscle? For example, during muscle di¡erentiation do you see activation of the kinase? Olson: We are looking at that currently. When skeletal muscle cells di¡erentiate these HDACs are exported through the same sort of process, and we think that kinase is probably involved. Baylin: Is your in vitro model using fetal cells? Olson: We use neonatal myocytes. They can go through one cell cycle after birth and then they stop dividing. Baylin: So when you challenge them in vitro to undergo hypertrophy, there is no cell cycling. Olson: That’s correct. Baylin: This has some connotations for what the class I block would do. If we think about some of the genes that are activated, such as p21, this might play a role in hypertrophy or cell cycle control issues. Olson: p21 probably is involved in hypertrophy as a suppressor. But these cells are clearly post-mitotic. Gu: Have you looked at p53? Olson: We haven’t looked at p53. Gu: I have a related question about Sir2. Have you considered the possibility that Sir2 is involved in this type of regulation? Olson: We haven’t looked at that yet. Greene: On a more clinical note, you showed that the HDAC inhibitor could block the development of cardiac hypertrophy in response to stress. What happens if you allow the hypertrophy to occur ¢rst? In other words, what if you allowed the heart to fail and then administered HDAC inhibitors? Is there any reversibility observed?

144

DISCUSSION

Olson: That is a good question. We have done that in two ways. One is to let the hearts hypertrophy and treat the animals with cyclosporin to shut o¡ the Ca2+ pathways. The other is to use a Tet-inducible cyclosporin line and switch it on and o¡. The hearts will hypertrophy, and when we terminate the signal the fetal gene programme will switch o¡ and the heart will shrink back to its normal size. But if the heart has undergone ¢brosis, which is sort of a terminal step in the pathway, this cannot be reversed. Cole: Have you looked at any of these HDACs in patients who have congestive heart failure, for polymorphisms or expression levels? Olson: We have looked in biopsy samples of failing human hearts at the intracellular distribution of HDAC5. In normal biopsies it is nuclear and in failing hearts it is totally cytoplasmic. Mahadevan: You mentioned earlier that you’d looked at the acetylation state in cultured cells. What are the stresses you used in the culture system? Have you looked at the program of transcription in the culture system in response to the kinase activation by stress, and also TSA? Olson: There are a number of stresses that can be used in the cultured cell system. These include agonists such as isoprotenerol and phenylephrine. Or, you can drive hypertrophy in vitro using viruses to deliver constitutively active kinases. The same fetal genes will be switched on both in vitro and in vivo. Mahadevan: What does TSA do to the genes in the in vitro system? Olson: In the culture system it shuts them all down. Khochbin: I believe that members of the 14-3-3 family are also important in the function of the class II HDACs. There are di¡erent isoforms of these proteins. Do you know if there are muscle or cardiac speci¢c members? Olson: I don’t think so. There are quite a few 14-3-3 proteins  at least eight. We have looked at six of them and they all interact equally. We haven’t seen any speci¢city at that level. Which ever 14-3-3s are around will recognize those phospho sites on HDACs, and they conform to the consensus site. Marks: Do you see any apoptosis? Olson: In the heart, apoptosis is probably the basis of cardiomyopathy. The heart dilates, and this is probably due to activation of apoptotic cell death. In the HDAC4 knockout mice which have massive brains that have grown out of control, these ultimately apoptose. They proliferate and when they can’t di¡erentiate they apoptose. Marks: When you treat the cells with HDAC inhibitors such as TSA, do you see apoptosis? Olson: No. But we haven’t looked at that carefully. Atadja: It should be concentration dependent. Marks: We ¢nd that certain cells are extremely sensitive to apoptosis. Olson: Have you looked in myocytes?

HDACs IN CARDIAC GROWTH

145

Marks: No, but we have looked at a fair range of other cells. Atadja: When we treat myocytes with our inhibitors, at higher concentrations we begin to see caspase cleavage and other indicators of apoptosis. Li: Do you know the upstream and downstream signalling components for HDAC7 in the endothelial cells? Olson: No, we are just looking at that. There are some obvious candidates, given all the information that is available on signals involving vascular endothelial development. Gu: Have you looked at the e¡ects of hypoxia? Olson: Not yet. But we are planning to use these mice for all sorts of experiments. Mahadevan: I have a question for both Eric Verdin and Eric Olson. This relates to the same gene, Nur77, in the two di¡erent systems that you described. In the one case, your programme of gene expression, including that gene, is suppressed by TSA. In the other case, that gene is actually activated by TSA. Has the gene been analysed in both studies under the same conditions? Is Nur77 poised in a di¡erent way with respect to it’s TSA-sensitivity in your distinct cell models? Olson: These are the sorts of things that we will have to look at now. Perhaps these HDACs are engaging di¡erent partners in a cell-speci¢c way.

Chromatin modi¢cations as clues to the regulation of antigen receptor assembly David Ciccone and Marjorie Oettinger1 Department of Molecular Biology, Massachusetts General Hospital, 50 Blossom Street, Boston, MA 02114, USA

Abstract. Changes in chromatin structure play a key role in the regulation of the mammalian genome, governing diverse processes including transcription, replication and recombination. In the earliest stages of antigen receptor assembly, D and J segments of the immunoglobulin heavy chain (IgH) and T cell receptor (TCR) b loci are recombined in B and T cells respectively, while the V segments are not. Distinct distribution patterns of various histone modi¢cations and the nucleosome-remodelling factor Brg1 are found at recombinationally ‘active’ (DJ) and ‘inactive’ (V) regions, o¡ering a means independent of transcription or DNAse I hypersensitivity to de¢ne chromatin domains at these loci. Within some inactive loci marked by H3-K9 dimethylation, two distinct levels of methylation are found in a non-random, genesegment speci¢c pattern. Brg1 is not localized to speci¢c sequences, as it is with transcriptional initiation, but rather associates with the entire active locus in a pattern that mirrors acetylation of histone H3. Distinct ‘hotspots’ of histone H3 dimethylated at lysine 4 are localized at the ends of the active DJ domains of both the IgH and TCRb loci, suggesting they may serve as important marks for locus accessibility. The speci¢c patterns of modi¢cation imply that the regulation of V(D)J recombination involves recruitment of speci¢c methyltransferases in a localized manner. 2004 Reversible protein acetylation. Wiley, Chichester (Novartis Foundation Symposium 259) p 146^162

The vertebrate immune system speci¢cally recognizes and responds to an enormous number of antigens, with a properly developed immune system critical for the survival and health of humans and other vertebrate species. The interaction with antigen is mediated by the immunoglobulin (Ig) and T cell receptor (TCR) molecules expressed by B and T lymphocytes respectively. The exons encoding the variable domains of Ig and TCRs are assembled during lymphocyte development 1This

paper was presented at the symposium by Marjorie Oettinger to whom correspondence should be addressed. 146

CHROMATIN MODIFICATIONS AND ANTIGEN RECEPTORS

147

from component gene segments by a series of DNA-breakage and rejoining events known as V(D)J recombination (reviewed in Bassing et al 2002, Gellert 2002). There are seven di¡erent loci (T cell receptor a, b, g and d and immunoglobulin H, k and l) that undergo V(D)J recombination. Three separate rearrangements are required to assemble the gene segments that encode the two protein chains that make up a mature antigen receptor. Tight regulation of recombination assures that these rearrangements occur in a lineage-speci¢c manner. Thus, Ig genes are only fully rearranged in B cells and TCR genes are only fully assembled in T cells. Rearrangement also usually occurs in a preferred temporal order with, for example, Ig heavy chain rearrangement generally preceding light chain assembly and IgH D^J joining occurring prior to V^DJ rearrangement. A fundamental issue in lymphoid development is how regulation of these DNA rearrangement events is achieved to prevent deleterious chromosomal instability while assuring proper ordered assembly of antigen receptor genes. As discussed below, because the same recombination machinery is used for all antigen receptor rearrangement, a crucial part of the regulation must be achieved by governing the accessibility of the locus to this machinery (Yancopoulos et al 1986), using alterations in chromatin structure to achieve this control. Regulation of antigen receptor assembly Recombination signal sequences (RSS) £ank all recombinationally competent Ig and TCR gene segments and they serve as binding sites for the lymphoid-speci¢c V(D)J recombinase, composed of the RAG1 and RAG2 proteins (for recent reviews of V(D)J recombination see Bassing et al 2002, Fugmann et al 2000, Gellert 2002). Expression of RAG1 and RAG2 is su⁄cient to induce V(D)J recombination of an arti¢cial substrate in non-lymphoid cells, although the regulated endogenous loci remain quiescent. Together the RAG proteins introduce a double strand break at the border between the signal element and the adjacent coding DNA giving rise to a blunt signal end and hairpinned coding end. The broken DNA ends are then rejoined signal end to signal end and coding end to coding end with the aid of double-strand break repair factors. In an alternative and potentially very deleterious reaction, the RAG proteins can mediate transposition, transposing signal ended DNA into unrelated target sequences (Agrawal et al 1998, Hiom et al 1998). V(D)J rearrangement is subject to many layers of regulation (reviewed in Bassing et al 2002, Hesslein & Schatz 2001, Krangel 2003), from the simplest level of protein^DNA recognition between the RAG proteins and the RSS to the highly complex coordination of lineage and stage-speci¢c accessibility of the endogenous loci. DNA methylation, asynchronous replication, RSS pairing restrictions, chromosome localization, transcription, chromatin structure and the

148

CICCONE & OETTINGER

RAG proteins themselves have all been implicated in at least some aspects of the regulation of V(D)J rearrangement. A broad range of genetic and biochemical evidence suggests that much of the developmental regulation of V(D)J joining is mediated by speci¢c alterations in chromatin structure that render regions within a locus accessible or inaccessible to the recombinase machinery. Information about the speci¢c alterations is limited (and is a focus of our current work) but these changes in chromatin structure can encompass large genomic regions, are stably maintained, and can occur even in the absence of the recombinase (Stanhope-Baker et al 1996). A role for cis-acting transcriptional regulatory sequences in establishing and/or maintaining these developmentally regulated changes in accessibility has been observed (reviewed in Hempel et al 1998, Hesslein & Schatz 2001, Sleckman et al 1996). However, it may be that localized changes in chromatin structure induced by cis-acting regulatory elements, rather than transcription itself, are what are crucial for establishing a recombination-accessible chromatin state (Sikes et al 2002 and references therein). In addition, the transcription factors E2A and EBF or HEB in conjunction with the recombinase are su⁄cient to permit particular endogenous k or TCRg/d gene segments to rearrange in a non-lymphoid cell line, an e¡ect that is likely due to changes in chromatin structure (Ghosh et al 2001, Romanow et al 2000). A strong correlation between histone acetylation (or factors that increase acetylation) and a recombinationally accessible chromatin state has been shown in vivo (reviewed in Bassing et al 2002, Gellert 2002, Krangel 2003, Muegge 2003). A role for acetylation is suggested by the increased levels of V(D)J recombination in cells treated with histone deacetylase inhibitors. Furthermore, acetylated histone H3 and H4 have been mapped to recombinationally accessible gene segments at a series of loci. In vitro, RAG proteins cannot cleave unmodi¢ed nucleosomal DNA (Golding et al 1999, Kwon et al 1998, McBlane & Boyes 2000), but acetylation and hSWI/SNF remodelling render nucleosomal DNA accessible for cleavage (Kwon et al 2000). This growing body of evidence clearly links chromatin modi¢cations to the regulation of V(D)J recombination. Chromatin modi¢cations at antigen receptor loci As a step toward understanding the molecular basis for the tissue-, lineage- and stage-speci¢c regulation of V(D)J recombination, we have used chromatin immunoprecipitation (ChIP) to map the distribution of histone modi¢cations and Brg1, the catalytic component of hSWI/SNF (Narlikar et al 2002), across Ig and TCR loci (see Fig. 1), focusing our initial e¡orts on the ¢rst locus pair that becomes accessible in each lymphoid cell type (the D and J loci of IgH in Pro B cells and TCRb in Pro T cells) (Morshead et al 2003). Studying D^J rearrangement

FIG. 1. Map of primer positions for the IgH and TCRb loci used for real-time PCR analysis. DNA recovered from each chromatin immunoprecipitation (ChIP) was analysed by real-time PCR using primer pairs speci¢c for a collection of V (black), D (green) and J (orange) gene segments as well as C gene segments and regulatory enhancers (E). These primer pairs were chosen with the Oligo4 program. All primer pairs used were shown to amplify a single band by radioactively labelled PCR reactions. A pair of primers speci¢c for a segment of the promoter of the ubiquitously expressed CAD gene was also included. Dots represent position of primer pairs. For details of primers and protocols see (Morshead et al 2003).

CHROMATIN MODIFICATIONS AND ANTIGEN RECEPTORS 149

150

CICCONE & OETTINGER

allows us to take advantage of cell lines derived from RAG-de¢cient mice. In the absence of recombinase activity, the Pro B and Pro T cells from RAG
/
mice are blocked in development just prior to this ¢rst stage of rearrangement. Three di¡erent types of murine cell lines, representative of di¡erent lineage and developmental contexts, were analysed: a RAG2
/
Abelson virus transformed Pro B cell line (Kirch et al 1998), a RAG1
/
p53
/
pro T cell line (Mombaerts et al 1995) and the NIH3T3 ¢broblast line. All Ig and TCR loci are refractory to rearrangement in NIH3T3 cells (Schatz et al 1992). As indicated above, in the absence of the recombinase, the D and J segments at the IgH locus in the Pro B cells and the TCRb locus in the pro T cells are poised to rearrange (‘active’) (Stanhope-Baker et al 1996), while VH, Vb and Ig light chain gene segments should be inaccessible in both pro B and pro T cell lines (‘inactive’). Indeed, upon reintroduction of the absent RAG gene, DJ rearrangement at the IgH and TCRb loci is observed in RAG-de¢cient pro B and pro T cell lines, respectively. The observations discussed below underscore the complexity of regulation at antigen receptor loci and suggest that the loci are not just uniformly ‘open’ or ‘closed’. Rather, the results indicate that both the maintenance of antigen receptor loci in an inaccessible state along with their conversion from silent to accessible conditions will involve di¡erent modi¢cations used in combination. Both the types of histone modi¢cations and the relative amounts of each modi¢cation are likely to be critical to the regulation of this process. Inverse correlation of H3-K9 acetylation and H3-K9 dimethylation As shown in Fig. 2A and B there is an inverse correlation between the localization of acetyl H3 lysine 9, an activating modi¢cation and dimethyl H3-K9, a repressive modi¢cation, at IgH and TCRb gene segments in the pro B and pro T cell lines (for a normalized graph comparing all the modi¢cations discussed here, see Fig. 5) (Morshead et al 2003). Such an inverse correlation between these two markers is in keeping with other studies at other loci and in other organisms (e.g. the chicken b-globin locus, Litt et al 2001a,b). In the Pro B cell lines, the accessible region of IgH, as de¢ned by hyperacetylation and low dimethyl H3-K9 extends across a large (*80 kb) region that contains the DH and JH segments as well as Em and Cm. This same region is found to be hyperacetylated in primary Pro B cells and other Abelson-transformed RAG de¢cient cell lines (Chowdhury & Sen 2001, Johnson et al 2003, Maes et al 2001). Hyperacetylation and low dimethyl H3-K9 is also seen at the two clusters of TCRb D and J elements in the pro T cell line. As expected, the IgH and TCRb loci in NIH3T3 cells are H3-K9 methylated but not acetylated. IgH and TCRb V segments are likely to be inaccessible in both the pro B and pro T cell lines and they are indeed H3-K9 methylated but generally not

FIG. 2. Chromatin modi¢cations and modifying activities at the IgH and TCRb loci. (A^D) Results from real-time PCR analysis are shown, with primer sites on the x axis and relative enrichment on the y axis. Cell types and modi¢cation are as indicated. Average results of three or more ChIPs are included for each data point.

CHROMATIN MODIFICATIONS AND ANTIGEN RECEPTORS 151

152

CICCONE & OETTINGER

hyperacetylated in both cell types providing a useful contrast to the active DJ loci. As expected, the active CAD gene is acetylated and undermethylated to a similar extent in pro B, pro T and NIH3T3 cells (Fig. 2; Morshead et al 2003). Dimethyl H3-K79 also marks the active domains of antigen receptor genes In addition to the observed high levels of H3 acetylation across the active domains, we ¢nd high levels of H3-K79 dimethylation at the active DJ regions (as de¢ned by histone acetylation), while inactive VDJ regions (de¢ned by H3-K9 dimethylation) have low dimethyl H3-K79, making H3-K79 another marker of euchromatin in mammalian cells (Ng et al 2003), data not shown. This is similar to what is found in yeast (S. cerevisiae), where the level of dimethyl H3-K79 is much reduced in heterochromatin and relatively constant over the rest of the genome, which, in yeast, is generally euchromatic (Ng et al 2003). Two distinct levels of dimethyl H3-K9 at inactive loci in a segment-speci¢c pattern By comparing the distribution of dimethyl H3-K9 across multiple loci, we are able to discern an unexpected pattern (Fig. 3 and Morshead et al 2003). Two distinct levels of dimethyl H3-K9 are found at certain inactive loci in a pattern that tightly correlates with segment type. In the pro B cell lines, V segments and the enhancer at TCRb are methylated at a ‘low inactive’ level, which is comparable to that of IgH V segments. In contrast, a ‘high inactive’ level of methylation is found at the J and C gene segments at TCRb. A similar pattern is seen at the inactive Igl locus in the pro B cell lines. Particularly striking is that this reproducible pattern is maintained despite the physical separation and intermingling of gene segments of di¡erent types across the TCRb and the Igl loci, suggesting that it may re£ect some underlying, but as yet unknown, layer of regulation. Brg1 is broadly distributed across accessible loci in both Pro B and Pro T cells In addition to examining the distribution of histone modi¢cations we considered the association of chromatin remodelling activities with the developing antigen receptor loci. Although nucleosome-remodelling complexes are often recruited to speci¢c target sites such as enhancers and promoters (Goldmark et al 2000, Hassan et al 2001a, Ng et al 2002, Peterson & Workman 2000), we ¢nd that that association of Brg1 at IgH and TCRb loci closely correlates with that of acetyl H3 (see Figs 2C and 5, and Morshead et al 2003) and hence spans essentially the entire recombinationally accessible domain. In both the pro B and pro T cell lines Brg1 is

CHROMATIN MODIFICATIONS AND ANTIGEN RECEPTORS

153

FIG. 3. Enrichment of dimethyl H3-K9 at repressed loci is segment type speci¢c. J and C segments at TCRb and Igl are shown with hatching, highlighting their ‘high inactive’ status.

enriched at rearranging D and J segments relative to non-rearranging V segments as much as at promoter elements in other studies. Such high levels of enrichment strongly suggest that Brg1 is present across the entire locus in most cells in the population, rather than at localized high levels in individual cells. The distribution of Brg1 is distinct from that previously seen in studies of transcriptional regulation (Goldmark et al 2000, Hassan et al 2001a, Ng et al 2002, Peterson & Workman 2000) and may re£ect the preference seen in vitro for Brg1-like complexes to bind to acetylated nucleosomes (Deckert & Struhl 2002, Hassan et al 2001b) that has been seen in vitro. Our results provide a physiological example of this in vitro phenomenon. The broad distribution of Brg1 suggests that this remodelling activity (and perhaps others) may play a role in achieving an open chromatin structure that permits RAG cleavage as suggested by in vitro analysis (Kwon et al 2000). The distinction from transcription is likely to re£ect the di¡ering requirements of the two processes. In the case of transcription, nucleosome remodelling locally around promoters may be su⁄cient to allow transcription complexes to initiate. In contrast, the recombinase machinery requires access to gene segments distributed over large regions at the antigen receptor loci, so nucleosome remodelling may have to be more widespread.

Peaks of dimethyl H3-K4 mark the edges of active regions Comparisons of the distribution of acetyl H3-K9, dimethyl H3-K79 and Brg1 show a striking correlation with each other across both TCRb and IgH loci and an inverse correlation with dimethyl H3-K9 (Figs 2, 5 and data not shown). In doing so they provide a clear picture of the extent of the active DJ domains of TCRb and IgH in pro T and pro B cell lines respectively and a rough location of where the transition from active to inactive chromatin occurs.

154

CICCONE & OETTINGER

FIG. 4. Fine-scale mapping of peaks of dimethyl H3-K4 at IgH. The position of primers relative to DFL16.1 (D) and Cm (C) shown.

FIG. 5. The normalized enrichments for the indicated modi¢cations are shown (normalized to high and low for each modi¢cation). Vertical grey bar represents the gap between V and D segments.

The distribution of dimethyl H3-K4, however, is quite di¡erent (Figs 2D, 5 and Morshead et al 2003). Pronounced hotspots are found at the ends of the active DJ loci, coincident with the locations where the other marks of an active locus decrease to their near inactive levels. As discussed below, the position of the dimethyl H3K4 peaks suggests that they may be marking the boundaries between active and inactive domains. At the IgH locus in Pro B cells, these high peaks of dimethyl H3K4 (30^40 fold enrichments) are located at the ¢rst (most 5’) DH segment, DFL16.1, and downstream just 3’ of the Sm switch recombination site (Fig. 2D, see asterisks). These hotspots are con¢ned to a comparatively small region; more detailed analysis for the IgH peaks with additional primers shows that the levels of methylation drop rapidly, within 2 kb of the central peak (Fig. 4 and Morshead et al 2003). A remarkably similar pattern of dimethyl H3-K4 hotspots is observed in the pro T cell line but now, in a lineage appropriate fashion, at the TCRb locus. At this

CHROMATIN MODIFICATIONS AND ANTIGEN RECEPTORS

155

locus, where there are two separate DJC clusters, three peaks are found that encompass these two regions (Fig. 2D, see asterisks); one at the 5’ end of the ¢rst cluster (at Db1, a 30-fold enrichment), the second, a 70-fold enrichment, starting at Cb1 and continuing into the 3 kb region that separates the two clusters, and the third (a 70-fold enrichment) downstream of the second cluster (starting at Cb2). One additional hotspot of dimethyl H3-K4 is found at the TCRb enhancer and this is seen in both pro B and pro T cells, suggesting it may play a di¡erent role. The localization pattern of dimethyl H3-K4 is largely distinct from any of the distribution patterns for this modi¢cation previously described in the literature (for example, see Litt et al 2001a, Noma et al 2001). At the globin locus, a correlation between dimethyl H3-K4 (using the same antibodies used here) and enrichment of acetylation has been observed (Litt et al 2001a). The peaks of dimethyl H3-K4 are not coincident with high levels of acetyl H3, but rather are found where the levels of acetyl H3 are at their minima in both loci. (However, the moderate elevation (*6^10 fold) of dimethyl H3-K4 seen here throughout active regions is more in keeping with this correlation of dimethyl H3-K4 with active loci.) Tracking chromatin modi¢cations as clues to the mechanisms of regulation of antigen receptor rearrangement Changes in chromatin structure play a key role in the regulation of the mammalian genome, with the largest body of work focused on transcriptional regulation. Recombination processes must also deal with unravelling compacted chromatin to gain access to their target sites. While there are likely to be common themes found between the regulation of transcription and recombination in mammalian cells, the manner by which chromatin structure impedes a recombination event and the requirements for relieving the inhibition are likely to display di¡erences as well. Particular aspects of the cleavage reaction may impact on the structural requirements for locus accessibility. For example, generation of the hairpin coding end by an inter-strand trans-esteri¢cation requires that the DNA undergo a signi¢cant amount of distortion. This poses a constraint for recombination signals within a nucleosome. Indeed, assembly of an RSS into a nucleosome inhibits V(D)J cleavage in vitro (Bassing et al 2002 and references therein), but this inhibition can be relieved by nucleosome remodelling with SWI/SNF (Kwon et al 2000). In another example, a full complement of RAG proteins initially assembles on a single RSS and then a second signal is bound (Gellert 2002). This pathway of assembling all the protein on one site and then searching for the partner DNA implies that the requirements for making the ¢rst RSS accessible for cleavage can be distinct from those for making the partner DNA available to interact with an already assembled protein DNA complex.

156

CICCONE & OETTINGER

Already, our analysis of the distribution of histone modi¢cations and modifying activities has provided a molecular explanation, consistent with a previous suggestion (Maes et al 2001), for why developing T cells permit a low level of rearrangement between IgH D and J segments. Marks of active chromatin, elevated acetyl H3 and dimethyl H3-K4, are present across the IgH DJC region in pro T cells, albeit at much lower levels than in pro B cells. (Elevated acetyl H4 has also been observed at JH segments in another T cell line (Maes et al 2001).) Unlike the Jb segments in Pro B cells which have a ‘high inactive’ level of methyl H3-K9, DH and JH gene segments in Pro T cells are associated with ‘low inactive’ methylation, analogous to the ‘low inactive’ level seen across the Igk locus in Pro B cells. Thus, the chromatin at the IgH D and J loci in Pro T cells appears to be in a partly recombinase-accessible state, with the di¡erences in modi¢cation levels compared to B cells probably translating into reduced accessibility and rearrangement. The observation of the novel discrete peaks of dimethyl H3-K4 raises new questions as to how antigen receptor assembly is regulated and suggests new avenues for exploration. The intriguing location of these peaks of dimethyl H3K4 suggests that these sites in the DNA may have particular importance in the regulation of rearrangement. Their striking correlation with the edges of active domains suggests that the observed chromatin modi¢cations are re£ecting a chromatin boundary or the edge of domain. The dimethyl H3-K4 modi¢cation or the site at which it is located, might serve as a boundary marker, preventing the spread of ‘repressive modi¢cations’ into the active region (or vice versa) or directing chromatin modifying activities to the locus in a developmentally appropriate manner. It remains to be seen how the discrete localization of dimethyl H3-K4 to these sites is achieved. One clear possibility is that speci¢c regulatory elements exist, perhaps analogous to the insulator elements of the b globin locus control regions (West et al 2002) that will be recognized by DNA binding proteins. These speci¢c DNA binding proteins would then recruit an H3-K4 methyltransferase to these positions. Alternative scenarios, where the complex arrangement of histone modi¢cations themselves serve to attract or exclude additional modifying activities are also possible. It remains for future work to determine how the developmentally regulated pattern of modi¢cations is achieved at the antigen receptor loci and the mechanism by which the loci are rendered functionally accessible to the RAG proteins. References Agrawal A, Eastman QM, Schatz DG 1998 Transposition mediated by RAG1 and RAG2 and its implications for the evolution of the immune system. Nature 394:744^751 Bassing CH, Swat W, Alt FW 2002 The mechanism and regulation of chromosomal VDJ recombination. Cell 109:S45^S55

CHROMATIN MODIFICATIONS AND ANTIGEN RECEPTORS

157

Chowdhury D, Sen R 2001 Stepwise activation of the immunoglobulin mu heavy chain gene locus. EMBO J 20:6394^6403 Deckert J, Struhl K 2002 Targeted recruitment of Rpd3 histone deacetylase represses transcription by inhibiting recruitment of Swi/Snf SAGA and TATA binding protein. Mol Cell Biol 22:6458^6470 Fugmann SD, Lee AI, Shockett PE, Villey IJ, Schatz DG 2000 The RAG proteins and VDJ recombination: complexes ends and transposition. Annu Rev Immunol 18:495^527 Gellert M 2002 VDJ recombination: RAG proteins repair factors and regulation. Annu Rev Biochem 71:101^132 Ghosh JK, Romanow WJ, Murre C 2001 Induction of a diverse T cell receptor gamma/delta repertoire by the helix^loop^helix proteins E2A and HEB in nonlymphoid cells. J Exp Med 193:769^776 Golding A, Chandler S, Ballestar E, Wol¡e AP, Schlissel MS 1999 Nucleosome structure completely inhibits in vitro cleavage by the VDJ recombinase. EMBO J 18:3712^3723 Goldmark JP, Fazzio TG, Estep PW, Church GM, Tsukiyama T 2000 The Isw2 chromatin remodeling complex represses early meiotic genes upon recruitment by Ume6p. Cell 103:423^433 Hassan AH, Neely KE, Vignali M, Reese JC, Workman JL 2001a Promoter targeting of chromatin-modifying complexes. Front Biosci 6:D1054^1064 Hassan AH, Neely KE, Workman JL 2001b Histone acetyltransferase complexes stabilize swi/ snf binding to promoter nucleosomes. Cell 104:817^827 Hempel WM, Leduc I, Mathieu N, Tripathi RK, Ferrier P 1998 Accessibility control of VDJ recombination: lessons from gene targeting. Adv Immunol 69:309^352 Hesslein DG, Schatz DG 2001 Factors and forces controlling VDJ recombination. Adv Immunol 78:169^232 Hiom K, Melek M, Gellert M 1998 DNA transposition by the RAG1 and RAG2 proteins: a possible source of oncogenic translocations. Cell 94:463^470 Johnson K, Angelin-Duclos C, Park S, Calame KL 2003 Changes in histone acetylation are associated with di¡erences in accessibility of VH gene segments to V-DJ recombination during B-cell ontogeny and development. Mol Cell Biol 23:2438^2450 Kirch SA, Rathbun GA, Oettinger MA 1998 Dual role of RAG2 in VDJ recombination: catalysis and regulation of ordered Ig gene assembly. EMBO J 17:4881^4886 Krangel MS 2003 Gene selection in VDJ recombination:accessibility and beyond. Nat Immunol 4:624^630 Kwon J, Imbalzano AN, Matthews A, Oettinger MA 1998 Accessibility of nucleosomal DNA to VDJ cleavage is modulated by RSS positioning and HMG1. Mol Cell 2:829^839 Kwon J Morshead KB Guyon JR Kingston RE and Oettinger MA 2000 Histone acetylation and hSWI/SNF remodeling act in concert to stimulate VDJ cleavage of nucleosomal DNA. Mol Cell 6:1037^1048 Litt MD, Simpson M, Gaszner M, Allis CD, Felsenfeld G 2001a Correlation between histone lysine methylation and developmental changes at the chicken beta-globin locus. Science 293:2453^2455 Litt MD, Simpson M, Recillas-Targa F, Prioleau MN, Felsenfeld G 2001b Transitions in histone acetylation reveal boundaries of three separately regulated neighboring loci. EMBO J 20:2224^2235 Maes J, O’Neill LP, Cavelier P, Turner BM, Rougeon F, Goodhardt M 2001 Chromatin remodeling at the Ig loci prior to VDJ recombination. J Immunol 167:866^874 McBlane F, Boyes J 2000 Stimulation of VDJ recombination by histone acetylation. Curr Biol 10:483^486

158

DISCUSSION

Mombaerts P, Terhorst C, Jacks T, Tonegawa S, Sancho J 1995 Characterization of immature thymocyte lines derived from T-cell receptor or recombination activating gene 1 and p53 double mutant mice. Proc Natl Acad Sci USA 92:7420^7424 Morshead KB, Ciccone DN, Taverna SD, Allis CD, Oettinger MA 2003 Antigen receptor loci poised for V(D)J rearrangement are broadly associated with BRG1 and £anked by peaks of histone H3 dimethylated at lysine 4. Proc Natl Acad Sci USA 100:11577^11582 Muegge K 2003 Modi¢cations of histone cores and tails in VDJ recombination. Genome Biol 4:213^218 Narlikar GJ, Fan HY, Kingston RE 2002 Cooperation between complexes that regulate chromatin structure and transcription. Cell 108:475^487 Ng HH, Robert F, Young RA, Struhl K 2002 Genome-wide location and regulated recruitment of the RSC nucleosome-remodeling complex. Genes Dev 16:806^819 Ng HH, Ciccone DN, Morshead KB, Oettinger MA, Struhl K 2003 Lysine-79 of histone H3 is hypomethylated at silenced loci in yeast and mammalian cells: a potential mechanism for position-e¡ect variegation. Proc Natl Acad Sci USA 100:1820^1825 Noma K, Allis CD, Grewal SI 2001 Transitions in distinct histone H3 methylation patterns at the heterochromatin domain boundaries. Science 293:1150^1155 Peterson CL, Workman JL 2000 Promoter targeting and chromatin remodeling by the SWI/ SNF complex. Curr Opin Genet Dev 10:187^192 Romanow WJ, Langerak AW, Goebel P et al 2000 E2A and EBF act in synergy with the VDJ recombinase to generate a diverse immunoglobulin repertoire in nonlymphoid cells. Mol Cell 5:343^353 Schatz DG, Oettinger MA, Schlissel MS 1992 VDJ recombination: molecular biology and regulation. Annu Rev Immunol 10:359^383 Sikes ML, Meade A, Tripathi R, Krangel MS, Oltz EM 2002 Regulation of VDJ recombination: a dominant role for promoter positioning in gene segment accessibility. Proc Natl Acad Sci USA 99:12309^12314 Sleckman BP, Gorman JR, Alt FW 1996 Accessibility control of antigen-receptor variableregion gene assembly: role of cis-acting elements. Annu Rev Immunol 14:459^481 Stanhope-Baker P, Hudson KM, Sha¡er AL Constantinescu A, Schlissel MS 1996 Cell typespeci¢c chromatin structure determines the targeting of VDJ recombinase activity in vitro. Cell 85:887^897 West AG Gaszner M and Felsenfeld G 2002 Insulators: many functions many mechanisms. Genes Dev 16:271^288 Yancopoulos GD Blackwell TK Suh H Hood L and Alt FW 1986 Introduced T cell receptor variable region gene segments recombine in pre-B cells: evidence that B and T cells use a common recombinase Cell 44:251^259

DISCUSSION Marmorstein: Do you know which HATs act on H3? Oettinger: I have no idea which HATs are doing any of this. Moazed: Along similar lines, do you have any idea what is recruiting the di¡erent modifying activities? Oettinger: Not yet. Historically, transcription has always been associated with V(D)J recombination. Thus there has always been an assumption that the enhancers and what is recruited for tissue-speci¢c expression should also somehow be involved. However, there are now multiple examples where the

CHROMATIN MODIFICATIONS AND ANTIGEN RECEPTORS

159

correlation between transcription and V(D)J recombination is not as tight as it was originally thought to be. You can have loci that are transcribed that cannot be recombined, and loci where there is no apparent transcription yet they are subject to recombination. I showed the K4 spike at the 5’ end of the DJ locus  there is no known promoter there and no transcription. Again, the sites that we have now identi¢ed are not overlapping with any sites that people have knocked out. I guess this is because there were no features there to make people interested. Jenuwein: Along these lines, for the peak that you referred to, can you do any ChIP experiments to see whether any components of the transcription apparatus would be associated with this? For example, ChIP for RNA polymerase or transcription-associated factors. You said there is no promoter-speci¢c sequence, but perhaps some factors are recruited. Oettinger: They could be. We have tried to ChIP with polymerase and analyse the C-terminal tail, but we haven’t been successful. It is not clear, though, that the experiments worked technically. Jenuwein: You also mentioned that the broad association with Brg1 or the Swi/ Snf-associated activity may be crucial in order to have the ‘super-open’ structure. Brg1 knockout mice have been generated. Have you done any mutant analysis? Oettinger: We are trying to get these mice so that we can do these experiments, both to see just what the locus looks like and whether or not any recombination is taking place. Verdin: One of the hallmarks of methylation is the idea that it is not removed. Have you looked in fully di¡erentiated T cells or B cells at whether the K4 methylation mark persists? Oettinger: We haven’t looked in fully di¡erentiated T cells. We have looked at more di¡erentiated B cells. They are transformed and look weird in a lot of di¡erent ways. This is something we are interested in. One complication is that the site of the upstream methyl mark will be lost as a result of rearrangement. The downstream situation becomes increasingly complicated because it is spliced out when switch recombination occurs. These sites are therefore absent on the rearranged alleles in many normal T cells and B cells. We may need to look at intermediate stages or with a more complicated set-up that allows us to distinguish alleles. Verdin: I have a broader question for the group. Are there many instances where people have seen such highly localized methyl marks? Baylin: We have found a very similar picture to this in the promoters of the genes that we have studied, with localized acteylation. The di¡erence is that they track a bit to K4. Oettinger: That’s what is seen in Gary Felsenfeld’s work, where there is an exact tracking of acetylation, as opposed to this very inverse tracking. There are a couple of primer sets that are in there. One of the downstream D segments is actually overlapping the DQ52promoter, and we don’t see an

160

DISCUSSION

accumulation of dimethyl K4 there. Also, the spikes that you are talking about are usually trimethylated. Baylin: Actually, we are using the di for the K4 changes. Berger: Have you looked at trimethylation? Does it correlate with the acetylation? Oettinger: Yes. Cole: Did you ¢nd it paradoxical that the in vitro studies show that p300 wasn’t important but you were getting this massive acetylation increase in chromatin in the cellular studies? Oettinger: Yes, we would have liked to see that the p300 did something. But the amount of acetylation that we get under those conditions is quite small. It may be that the role of the p300 is in recruiting Swi/Snf across the region. The acetylation could have two e¡ects. One is to help open up very compacted chromatin, and we don’t have that to any great extent in these assembled substrates. Berger: Is the system amenable to transfection? You might be able to alter the modi¢cation levels by putting in more of certain histone modi¢cation enzymes. Oettinger: Yes, we could either try to put in more or deplete what is there. We are going to try to do the depletion with RNAi. You’d probably need to use either stable cell lines or retroviral or lentiviral vectors to target enough cells. Zhou: Is the PHD domain related to the PHD ¢nger domain found in other proteins? Oettinger: Yes. It is a non-classical one, but structural analysis shows that it is a PHD domain. Zhou: Does it promote the organization of RET proteins? Oettinger: No one knows, because people haven’t been able to make the fulllength protein in vitro. Mahadevan: I have a general question about the antibodies for the acetylation work. Have you also looked with other antibodies to see whether enhanced acetylation occurs generally? Does it happen on H4 also, for example? Oettinger: Ranjan Sen’s group has looked across the Ig heavy chain locus and the TCRd locus. They have looked with both H3 and H4 antibodies and this is a general property. Mahadevan: If the antibody for acetlylysine 9/14 used is the one sold by Upstate, there are two issues here. One is that it doesn’t recognize H3 tails that are both phosphorylated and acetylated. The other is that there is a recently emerging story that this doesn’t recognize both acetyl-lysines 9 and 14, but just acetyllysine 9. Allis: I think it is true that it is mostly selective for 9. But the rabbits were actually immunized with 9 and 14. In yeast you can knock out these sites selectively.

CHROMATIN MODIFICATIONS AND ANTIGEN RECEPTORS

161

Mahadevan: This ¢ts with our experience with that antibody (Thomson et al 2001). Have you looked at phosphorylation across there? Oettinger: No. The original phosphorylation antibodies didn’t ChIP for us. We’d love to obtain anyone’s favourite antibodies that actually work in ChIP analysis to test across these. We have been limited by what ChIPs well. Zhou: Does the PHD domain from RAG2 contain E3 ligase activity? Oettinger: To my knowledge, it hasn’t been tested. Actually, the N-terminus of RAG1 has E3 ligase activity. Zhou: Does it ubiquitinate histone? Oettinger: I don’t know. We are beginning to have an idea of what the PHD domain is interacting with. I don’t think it is E3 ligase. Baylin: Are the RAG1 and RAG2 constitutively expressed at high levels in all cells, so that the level of control would all be at the post-transcriptional level? Oettinger: No, there is an added level of control, and it is quite tightly controlled. During development the RAG1 and RAG2 genes come on and o¡ during lymphocyte development. Afterwards there is the rearrangement and they turn o¡ while the cells undergo their burst of proliferation post V(D)J recombination and heavy chain rearrangement. Then they come back on again when it is time for light chain rearrangement. Once light chain rearrangement is completed and the antibody or TCR gets to the cell surface they come back down. This is one layer of regulation. RAG2 is cell cycle regulated: it is on in G1, then it is phosphorylated and moves into the cytoplasm. The other feature about having this ordered assembly, with ¢rst site being occupied and then the bound protein binding to a second site is that if a full complement of RAG proteins were to assemble on both 12 and 23 signals, this would be inhibitory: they can’t come together and can’t kick o¡ the protein. The cells need to be careful that they regulate the amount of the protein available and the number of accessible sites so that pairing between 12 and 23 signals is not blocked by the prior presence of RAG proteins on the companion sites. Berger: You mentioned that the ¢rst step is not as speci¢c as the second step of recombination. Then you showed that the acetylation is spread over the whole DJ region. When you get to this second step, is the V region acetylation much more restricted to these speci¢c V segments that will be used in the rearrangement. Oettinger: V segment rearrangement occurs in waves. There are groups further downstream that go ¢rst. In Pax5 mutants, this pattern is altered. Kathryn Calame has looked in more detail at acetylation patterns. We haven’t looked at this for a variety of complicated reasons. Marmorstein: Is the acetylation pattern susceptible to HDAC inhibitors? Oettinger: We haven’t done that. Anecdotally, yes, in the sense that if we treat cell lines with HDAC inhibitors there is an increase in V(D)J recombination. We haven’t looked at these cells.

162

DISCUSSION

Verdin: What is the pattern of HDAC and HAT expression in these cells? Oettinger: This hasn’t been studied properly. Jenuwein: Along these lines, we should mention the activity of the Enhancer of zeste complex and H3-K27 trimethylation, shown by Alexander Tarakovsky (Rockefeller) to trigger V(D)J rearrangement (Su et al 2003). In Enhancer of zeste mutant mice you severely impair V(D)J rearrangement and B-cell di¡erentiation. Oettinger: You get some speci¢c triggering but it is not global. References Su IH, Basavaraj A, Krutchinsky AN et al 2003 Ezh2 controls B cell development through histone H3 methylation and Igh rearrangement. Nat Immunol 4:124^131 Thomson S, Clayton AL, Mahadevan LC 2001 Independent dynamic regulation of histone phosphorylation and acetylation during immediate-early gene induction. Mol Cell 8: 1231^1241

General discussion I Histone modi¢cations in X-inactivation Turner: I will present some results on histone modi¢cations during Xinactivation in female mouse embryonic stem (ES) cells. They give some interesting hints about possible functional di¡erences between mono-, di- and trimethylated histone lysines and identify an epigenetic mark speci¢c for Xlinked genes in female cells. To do these experiments, we have raised antibodies to H3 mono-, di- and trimethylated at lysine 4 (H3me1K4, H3me2K4, H3me3K4). If we use these antibodies to stain metaphase chromosome spreads from human female lymphoblastoid cells, we see one pale-staining X chromosome with all three antibodies, and a distinctive banding pattern on the chromosome arms. Thus, there seems to be an overall absence of H3-K4 methylation (mono-, di- and tri-) on the female inactive X (Xi). This is consistent with previous studies on H3 dimethylated at K4 (Heard et al 2001, Peters et al 2002, Boggs et al 2002). In contrast, an antibody to H3 dimethylated at K9 (H3me2K9) labels both X chromosomes (i.e. Xi and Xa) equally well, with the same level of labelling as we see on the autosomes. We are still working to optimize immuno£uorescence labelling with antisera to H3 mono- and trimethylated at K9, but initial results suggest that H3me3K9 is depleted on Xi (in contrast to H3me2K9). We are very much aware that H3 serine 10 is phosphorylated at high frequency in mitotic cells and that this modi¢cation may interfere with binding of antibodies to the adjacent residue, H3-K9. For some years we have been working with female ES cells as a model system to study the progression of X-inactivation. Undi¡erentiated ES cells have two active X chromosomes (Xa). Di¡erentiation triggers the X-inactivation process and di¡erentiated ES cells have one Xa and one Xi (randomly chosen). It is a lovely model system for studying the progression of the events associated with Xinactivation. If we stain metaphase chromosome spreads from undi¡erentiated female ES cells with antibodies to acetylated histones, or H3 methylated at lysines 4 or 9, then both X chromosomes stain and are indistinguishable from the autosomes. To establish when, during di¡erentiation and X-inactivation, changes in histone modi¢cation are put in place, we have made chromosome spreads at days 0, 1, 2 and so on of di¡erentiation, immunostained and counted the number of spreads that show a pale-staining chromosome. The ¢rst result is that all the histones (i.e. H2A, H2B, H3 and H4) begin to deacetylate at exactly the same 163

164

GENERAL DISCUSSION I

FIG. 1. (Turner) All four core histones are deacetylated at the same stage of ES cell di¡erentiation. Metaphase chromosome spreads were prepared from female ES cells after di¡erentiation for up to 7 days. Chromosomes were immunostained with antibodies to acetylated H2A, H2B, H3 and H4 and spreads with and without a clear, pale-staining chromosome were counted. Panel A shows the results of such counts. Panels (B) (FITC, antibody stain) and (C) (DAPI counterstain) show an example of a spread with a pale-staining, hypo-acetylated X (arrows). See O’Neill et al (2003).

time, between days 3 and 4 of di¡erentiation (Fig 1 [Turner]). Deacetylation of the inactive X seems to have reached a plateau by day 7 of di¡erentiation (i.e. the frequency of spreads with a pale-staining chromosome stops increasing). How does this compare with methylation? The loss of dimethyl H3-K4 exactly tracks with the deacetylation, so this is happening at exactly the same time. However, if we look at trimethyl K4, we see that this is lost much earlier: there is a signi¢cant loss of trimethyl K4 by day 1^2 of di¡erentiation. It is happening 2 days before the loss of dimethyl K4, which is quite a striking result. With monomethyl K4 the results are more complex. Even at day 0, we see a proportion of cells with palestaining chromosomes. We are exploring the identity of these pale-staining chromosomes and how things change as di¡erentiation proceeds. To complement the immuno£uorescence studies, we have done chromatin immunoPrecipitation (ChIP) to quantitatively measure changes in histone modi¢cation associated with speci¢c X-linked genes during X-inactivation. With all the antibodies to acetylated histones, we get a drop in the acetylation of X-linked

GENERAL DISCUSSION I

165

genes in female cells of 40^50% by day 7, but no change in male cells. Thus, the ChIP assay gives us exactly the result we’d expect if one of the two X chromosomes in female cells were to be completely deacetylated by day 7 of di¡erentiation. What was less expected was the ¢nding that, consistently, X-linked genes in female ES cells, prior to di¡erentiation, showed 2^3 fold higher levels of acetylation of all four core histones when compared to autosomal genes or X-linked genes in male cells. Collaborative experiments on hybrid (Mus musculusM. castaneus) ES cells with Jeannie Lee (MGH, Boston, USA), have shown that both female Xs are equally hyperacetylated. The results with antibodies to dimethyl K4 and dimethyl K9 show that in females dimethyl K4 starts o¡ high and drops by at least 50% by day 7 of di¡erentiation. The dimethyl K9, on the other hand, goes exactly the opposite way, starting low and going much higher. H3 methylation of X-linked genes does not change with di¡erentiation in XY male cells. Recent ChIP experiments have shown that H3 trimethylated at K4 is also elevated on X-linked genes (compared to autosomal genes) in undi¡erentiated female ES cells, and falls by day 7 of di¡erentiation, whereas H3 monomethylated at K4 is not consistently elevated and shows no consistent drop on di¡erentiation. In summary, X-linked genes in female ES cells, prior to di¡erentiation, are marked by increased acetylation of all four core histones, increased di- and trimethylation of H3-K4 and decreased dimethylation of H3-K9. This mark is retained on genes on Xa for at last 7 days of ES cell di¡erentiation, but lost from genes on Xi, which show extreme underacetylation of core histones, undermethylation of H3-K4 and increased dimethylation of H3-K9. These ¢ndings di¡er, in some respects, to what we see by immuno£uorescence, where we ¢nd no evidence that dimethyl H3-K9 is higher on the inactive X chromosome than on the autosomes, and no consistent evidence that the female Xs label any more strongly than the autosomes with antisera to acetylated histones. This may re£ect a di¡erence between histone modi¢cations associated with bulk intergenic and non-coding chromatin (which will account for much of the immuno£uorescence signal) and those associated with coding and promoter regions. We think we have detected a signi¢cant epigenetic mark for X-linked genes in female cells. The mark may be something to do with priming female X-linked genes for the dosage compensation mechanism when that cuts in. This observation suggests that chromatin and histone modi¢cations not only play a crucial role in silencing of genes on Xi, but also serve to distinguish the X-linked genes of male and female cells prior to X inactivation. The marking of female Xlinked genes may be involved in the so-called counting mechanism, the means by which X-inactivation is restricted to cells with more than one X chromosome. (The results identifying this epigenetic mark have recently been accepted for publication, O’Neill et al 2003.)

166

GENERAL DISCUSSION I

Verdin: In this system is there cell division throughout this di¡erentiation process? Turner: Yes, they are growing very fast. You can see blood islands by day 5 of di¡erentiation. We are ignoring all these interesting things and are focusing on Xinactivation. Verdin: Do you have any hypothesis or model as to what is removing these methyl marks in the di¡erentiation process? Turner: I don’t know what is removing the methyl marks. I am impressed with the speed at which the trimethyl K4 is lost. I am also impressed that the trimethyl K4 is lost at a time when the dimethyl K4 isn’t. If there is a demethylase that goes tri to di then this is easily explained. If there isn’t, and none of us have been able to ¢nd one, then this becomes a little more di⁄cult to explain. The other odd thing is that if you add TSA at day 2, before the global deacetylation has started, the cells carry on dividing. This deacetylation is not blocked by TSA. Once again, this is reminiscent of some of the other results that we have heard. Jenuwein: I would like to talk about models. Do you think one could argue that the monomethylation is there to prepare or synergize with subsequent methylation states and are there any means to detect the co-existence of modi¢cations on similar nucleosomal regions? In other words, are these the same nucleosomes that are then processed into di- and trimethylation? Turner: I think that is possible. We are really looking at a later stage. We haven’t got the antibodies to monomethyl K9 to work terribly well. It is K9 where we are seeing the increase in methylation. There is a possibility that mono-K9 was priming for an increase in di-K9-methylation. All the other things that we are looking at, at this stage, are a loss of methylation of K4 and a loss of acetylation. Jenuwein: Let me rephrase my point. We can focus on H3-K4. When you see the decrease in the H3-K4 trimethylation, do you gain di- and monomethylation marks in these regions? Turner: Laura O’Neill has done a lot of timed immunoprecipitations. We can hopefully link these to the immuno£uorescence. As things stand, the ChIP is bearing out the immuno£uorescence: tri disappears earlier than the di. This says we have a tri to di switch. Is it on the same nucleosomes, or are some nucleosomes losing their tri and going right down to non- or monoacetylated? We can’t answer this important question at the moment. We are losing a lot of tri. From the immunoprecipitation, if you quantitate the ChIP results, the tri is completely gone from the inactive chromosome. Jenuwein: Can this be correlated with a decrease in transcriptional activity? Turner: Yes. The early stage is when the genes are being switched o¡, and when Xist expression is being up-regulated and when Xist RNA is coating the chromosome. The deacetylation and the loss of dimethyl K4 is a later event and

GENERAL DISCUSSION I

167

may be more involved in maintenance. This is the other thing that makes the loss of tri exciting because it is happening right at this crucial period. Li: With your ChIP assay, can you see a spreading from the proximal gene across the X-inactivation centre and then to the distal genes? Turner: I can’t answer that because we haven’t got all the ChIP results yet. We need the timing to do this. By the immuno£uorescence we have never seen any evidence of partial pale-staining chromosomes. It is possible we wouldn’t see them. Li: If one compares H3-K4 and H3-K9 methylation, one sees di¡erential accumulation in transcribed genes or across certain chromatin regions. But in in vitro assays do you see a mutually exclusive methylation between H3-K4 and H3K9? Perhaps H3-K4 mono-methylation is not blocking tri- or di-methylation of H3-K9. Has anyone done in vitro HMTase assays to see whether di- or trimethlyation of H3-K4 actually blocks H3-K9 methylation or vice versa? Jenuwein: I can comment in part. Danny Reinberg (Rutgers) and Craig Peterson (Worcester) are trying to reconstitute ‘designer nucleosomes’ in vitro by ligating di¡erentially methylated histone H3 N-terminal peptides to nucleosomal cores and then analysing transcriptional competence of these templates. They are currently testing cross-talks between distinct histone modi¢cations in vitro. Allis: We have begun to explore this, but I don’t know the answer. Castronovo: I have a more general question. What protects the other X chromosome during the inactivation process? Turner: The X-inactivation system is broken down into steps. The ¢rst step is counting: the cell has to know whether it has one or two X chromosomes. The hyperacetylation and hypermethylation that we see is present on both X chromosomes in female cells. It could be part of the counting mechanism: it is something that is only seen on cells with more than one X and says that these genes are susceptible to the dosage compensation process. After that you get a choice mechanism, so one of the two chromosomes is randomly chosen. This involves Xist and its antisense gene, Tsix. The factors involved in that choice mechanism are still not understood. All we can say is that the things we have looked at so far don’t seem to be directly involved in that choice system. They are downstream events. Castronovo: Have you looked in cells with three X chromosomes? Turner: No. Castronovo: When the choice is made, it seems the process is blocked in the second one which would be protected. What protects it? Turner: For 30 years people have been talking about a blocking factor that binds to one of the Xs and protects it from the inactivation system. If you have a blocking factor you don’t need to count. You have a factor that is only su⁄cient for one X.

168

GENERAL DISCUSSION I

Baylin: I am curious about the very nice demarcation you have. Where are you setting your primers for your ChIP? Turner: The ChIP is done using native chromatin. It is a bit of a laborious procedure, but it is extremely accurate. We are starting to get the RT-PCR system going, but for all these we have used slot blotting and Phosphor Imaging. It is a bit of a clunky old system but it is extremely accurate. We can detect twofold di¡erences with great reliability. The probes that we are using are mostly for coding regions, but also for promoter regions. For the PGK1 we had a promoter probe and a coding region probe. We also have coding and promoter probes for HPRT. We have not seen any signi¢cant di¡erences between coding and promoter regions. Baylin: For many genes that people are examining di¡erences are seen between the coding region and small regions upstream downstream. Turner: We are talking about small numbers: we have probably looked at 2 or 3 promoters. We have looked at probably 10 X-linked genes altogether, but among these we have probably only looked at 2 or 3 promoters. Atadja: I have a comment about the decrease in acetylation and supposed demethylation. Using your argument from earlier, could it be that something else is happening that is making the histones lose recognition for the antibody? It is not necessarily the loss of acetyl or methyl groups that could explain your TSA results. If the acetyl is not changing, TSA is not going to make any di¡erence. Turner: That is a crucial point. I don’t think the deacetylation is going to be a¡ected by that because we see deacetylation of everything at the same time. This can’t all be due to blocking. For K4 methylation it could disappear because something adjacent is being modi¢ed, such as serine 3, which is blocking the binding of the antibody. But it is di⁄cult then to explain the di¡erence between the tri-, di- and monomethylation, because it is the same residue. The tri is disappearing two days before the di. Khochbin: The fact that TSA treatment doesn’t induce hyperacetylation could be explained by the fact that histone acetyl transferases (HATs) are not present. They could be excluded from the inactive X chromosome territory. Have you checked this? Turner: We have thought about the HAT side but we haven’t done this yet. It is a striking result. If you add TSA from day 0 it blocks deacetylation and Xinactivation. If you give them two days to di¡erentiate and then add TSA, they just carry on. I am a little bothered by this. We are limited in the concentration of TSA we can add before they die. Although we can add enough to induce an increase in histone acetylation, it may be that we can’t add quite enough to inhibit the HDACs that are necessary for the deacetylation. I don’t think we can use our results to say that HDACs are not involved in that deacetylation, but it is a puzzling result.

GENERAL DISCUSSION I

169

Baylin: However, if you have DNA methylation in those regions you are probably not going to see anything just using the TSA alone. How quickly do the promoters get DNA methylated? Turner: It takes at least two weeks. These are almost mice by the time they get round to methylating their DNA. It is a very late event. This is true in embryos as well. Gu: What about TSA-insensitive deacetylation? Turner: Do you mean things like Sir2? This would explain it. A PhD student, Hugh Spotswood, has done some ChIP with antibodies to HDACs 1^4, looking for targeting of HDACs to X-linked genes as the cell di¡erentiates. He could ¢nd no evidence for any changes. Hugh left before we could get decent antibodies to Sir2. We should look at this. Marmorstein: Your data are consistent with there being a demethylase that takes tri to di. I know lots of people have looked for these unsuccessfully, but they might have used the wrong substrates. Maybe they need to start with a tri rather than a mono or di. Allis: We have looked only with di. That is a fair comment. Turner: We have looked with tri and haven’t found it. All we can say is that looking at those results, they are most easily explained by a tri to di demethylase. That is easy to say: the challenge now is to ¢nd it. We have looked, but it is di⁄cult to model these things in vitro. References Boggs BA, Cheung P, Heard E, Spector DL, Chinault C, Allis CD 2002 Di¡erentially methylated forms of histone H3 show unique association patterns with inactive human X chromosomes. Nat Genet 30:73^76 Heard E, Rougeulle C, Arnaud D, Avner P, Allis CD, Spector DL 2001 Methylation of histone H3 at lys-9 is an early mark on the X chromosome during X inactivation. Cell 107:727^738 O’Neill LP, Randall TE, Lavender J, Spotswood HT, Lee JT, Turner BM 2003 X-linked genes in female embryonic stem cells carry an epigenetic mark prior to the onset of X inactivation. Hum Mol Genet 12:1783^1790 Peters AHFM, Mermoud JE, O’Carroll D et al 2002 Histone H3 lysine 9 methylation is an epigenetic imprint of facultative heterochromatin. Nat Genet 30:77^80

The HDAC complex and cytoskeleton Je¡ery J. Kovacs, Charlotte Hubbert and Tso-Pang Yao1 Department of Pharmacology and Cancer Biology, Duke University, Durham, North Carolina 27710, USA

Abstract. HDAC6 is a cytoplasmic deacetylase that dynamically associates with the microtubule and actin cytoskeletons. HDAC6 regulates growth factor-induced chemotaxis by its unique deacetylase activity towards microtubules or other substrates. Here we describe a non-catalytic structural domain that is essential for HDAC6 function and places HDAC6 as a critical mediator linking the acetylation and ubiquitination network. This evolutionarily conserved motif, termed the BUZ domain, has features of a zinc ¢nger and binds both mono- and polyubiquitinated proteins. Furthermore, the BUZ domain promotes HDAC6 mono-ubiquitination. These results establish the BUZ domain, in addition to the UIM and CUE domains, as a novel motif that both binds ubiquitin and mediates mono-ubiquitination. Importantly, the BUZ domain is essential for HDAC6 to promote chemotaxis, indicating that communication with the ubiquitin network is critical for proper HDAC6 function. The unique presence of the UIM and CUE domains in proteins involved in endocytic tra⁄cking suggests that HDAC6 might also regulate vesicle transport and protein degradation. Indeed, we have found that HDAC6 is actively transported and concentrated in vesicular compartments. We propose that an integration of reversible acetylation and ubiquitination by HDAC6 may be a novel component in regulating the cytoskeleton, vesicle transport and protein degradation. 2004 Reversible protein acetylation. Wiley, Chichester (Novartis Foundation Symposium 259) p 170^181

Reversible histone acetylation has been studied for more than 30 years (Pogo et al 1966). A large body of research has established that histone acetylation plays an important role in the dynamic regulation of chromatin structure essential for biological processes such as gene transcription and the execution of DNA damage repair (Bird et al 2002, Wade et al 1997). Despite signi¢cant progress, our understanding of reversible acetylation, however, is almost exclusively in the context of chromatin function. Thus, a central question concerning acetylation biology is to decipher whether reversible acetylation serves as a general protein 1This

paper was presented at the symposium by Pang Yao to whom correspondence should be addressed. 170

THE HDAC COMPLEX AND CYTOSKELETON

171

modi¢cation that has the capacity to regulate diverse biological processes in addition to its well-characterized role in chromatin dynamics. Recent studies from several groups now provide initial insight into this critical issue. Histone acetylation is tightly regulated by the opposing activities of the histone acetyltransferases (HATs) and histone deacetylases (HDACs). Homology searches based on the founding members, mammalian HDAC1, and yeast RPD3 and HDA1, led to the identi¢cation of at least 11 HDAC family members (Gao et al 2002, Grozinger et al 1999, Guardiola & Yao 2001). Although it was initially assumed that these HDACs functioned to regulate histone acetylation, the growing complexity of subcellular localization and substrate speci¢city strongly suggests diverse functions for the various HDAC family members. The identi¢cation of HDAC6 as a cytoplasmic, microtubule-associated deacetylase provides the ¢rst experimental evidence indicating a much broader role for reversible acetylation outside the presumed functions in chromatin remodeling and histone metabolism (Hubbert et al 2002). Among all the HDACs identi¢ed so far, HDAC6 is unique as it is localized exclusively in the cytoplasm. Confocal images reveal that HDAC6 is concentrated in punctate structures and enriched in the perinuclear region and at the leading edge of the cell (Fig. 1). Intriguingly, this staining pattern is shared by

FIG. 1. Immunolocalization of HDAC6. A549 cells were immunostained for endogenous HDAC6. Note the punctate staining pattern concentrated in the perinuclear region and at the leading edge of the cell (arrows).

172

KOVACS ET AL

p150glued, a component of the microtubule-associated dynein motor complex. HDAC6 and p150glued co-localize in cells and their distinctive subcellular distribution patterns are dependent on an intact microtubule network. Although the functional signi¢cance of this co-localization remains to be established, we speculate that there is a potential link between HDAC6 and microtubuleassociated motor function. Interestingly, an almost perfect localization of HDAC6 to microtubules becomes apparent when cells enter quiescence (Hubbert et al 2002). These studies indicate that HDAC6 is a microtubule-associated deacetylase. The signi¢cance of this observation may lie in the ¢nding that microtubules are subject to di¡erential acetylation (L’Hernault & Rosenbaum 1985) and that HDAC6 can function as a microtubule deacetylase (Hubbert et al 2002). The exact function of tubulin acetylation remains to be ¢rmly established. However this evolutionarily conserved modi¢cation is tightly regulated and often observed in stable microtubules but absent in more dynamic microtubules, such as those found in neuronal growth cones or leading edges (Piperno et al 1987, Robson & Burgoyne 1989). Indeed, a role for HDAC6 in regulating the stability of a speci¢c pool of microtubules was recently reported (Matsuyama et al 2002). Consistent with the idea that HDAC6 regulates microtubule function, we have found that HDAC6 can modulate serum-induced cell motility, a process that requires dynamic reorganization of the cytoskeleton, including microtubules (Hubbert et al 2002). The ability of HDAC6 to enhance chemotactic movement requires its deacetylase activity. This result substantiates the idea that HDAC6 functions by deacetylating one or more substrates, such as microtubules, that are important in chemotaxis (Hubbert et al 2002). Beyond this observation, however, little of the mechanism has been elucidated to explain the biological functions that are unique to HDAC6. As the number of non-histone proteins subject to reversible acetylation is growing rapidly (reviewed in (Polevoda & Sherman 2002), another critical issue arises: how does acetylation regulate protein function? Reversible acetylation modi¢es the e-amino group of speci¢c lysine residues in protein substrates. One consequence of this modi¢cation is the loss of charge in the lysine acceptor. Indeed, Ren and colleagues have presented genetic evidence that acetylation regulates histone H2A.Z mainly by neutralizing the charge patch conferred by lysine residues. This neutralization likely results in the destabilization of the histone-DNA interaction (Ren & Gorovsky 2001). However, charge modi¢cation is not the only mechanism by which acetylation operates. While studying the acetylation of p53, we found that acetylation and ubiquitination occur on the same sets of lysine residues critical for p53 stability (Ito et al 2001). Since both acetylation and ubiquitination modify the e-amino group of lysines, we have proposed that genotoxic- or stress-induced acetylation competes with

THE HDAC COMPLEX AND CYTOSKELETON

173

ubiquitination and thereby controls p53 stability by preventing ubiquitinmediated degradation (Ito et al 2002, 2001). This observation suggests an intriguing possibility that protein acetylation may regulate protein ubiquitination. Here we will describe that HDAC6 has distinct structural and functional features that may serve to integrate the acetylation and ubiquitination networks. This link to the ubiquitination network could potentially provide an important framework for understanding the function of reversible acetylation. Results and discussion In addition to the tandem catalytic domains, homology searches have identi¢ed a conserved motif at the C-terminus of HDAC6 (Fig. 1). Interestingly, sequences that show signi¢cant homology to this motif are found exclusively in a number of deubiquitinating enzymes of both animal and plant origins. Sequence alignment of this motif from HDAC6 and multiple deubiquitinating enzymes demonstrates the presence of eight regularly spaced, invariant cysteine and histidine residues. In addition, many other conserved residues can be identi¢ed throughout this motif (Fig. 2). This signi¢cant sequence conservation and invariant cysteine and histidine residues within this motif indicate that it is likely a novel zinc ¢nger. The distinctive presence of this domain in de-ubiquitinating enzymes suggests that this novel zinc ¢nger is likely involved in the ubiquitination pathway. That HDAC6 also contains this motif suggests a potential functional interaction between HDAC6 and protein ubiquitination. To investigate the possibility that HDAC6 may have a functional link to protein ubiquitination, we examined whether the putative zinc ¢nger binds ubiquitin. To that end, a recombinant HDAC6 zinc ¢nger fused to glutathione transferase (GST) was tested for its ability to bind mono- and/or poly-ubiquitin by an in vitro pulldown assay. As shown in Fig. 3, the wild-type HDAC6 zinc ¢nger, but not GST alone, binds both mono- and poly-ubiquitin chains e⁄ciently. Interestingly, a related zinc ¢nger from the human de-ubiquitinase UbpM (Cai et al 1999) also binds mono- and poly-ubiquitin, indicating that the ubiquitin-binding capacity is a common property of this novel zinc ¢nger domain. We therefore name this motif binder of ubiquitin zinc ¢nger (BUZ ¢nger). Importantly, point mutation of the conserved cysteine 1145 completely inactivates the mono- or poly-ubiquitin binding capacity (Fig. 3). Furthermore, the ubiquitin binding activity of the BUZ ¢nger also requires zinc ions (our unpublished results). These results con¢rm the idea that the conserved BUZ domain is a novel zinc ¢nger that functions as an ubiquitin-binding domain. Importantly, full length HDAC6 has the capacity to bind cellular ubiquitinated proteins in a BUZ ¢nger-dependent manner (data not shown). These results indicate that HDAC6 is a deacetylase that can communicate with the ubiquitin network.

FIG. 2. HDAC6 and the BUZ ¢nger. (A) A schematic diagram of HDAC6. CAT, catalytic domain. (B) A compiled sequence of USP ¢ngers from HDAC6 and several other representative family members. Invariant cysteine and histidine residues that form the putative Zinc ion coordinating residues are marked in red. The other invariant residues are marked in blue. Highly conserved residues are listed in black and conserved hydrophobic residues are marked as F. The consensus sequence is listed at the bottom of the table. Residues underlined indicate a various length of insertion at these positions.

174 KOVACS ET AL

THE HDAC COMPLEX AND CYTOSKELETON

175

FIG. 3. The BUZ ¢nger is a ubiquitin binding motif and mediates HDAC6 monoubiquitination. (A) Sequence of the core BUZ Finger. The putative zinc coordinating cysteines (C) and histidines (H) are underlined. The cysteine to serine at residue 1145 and histidine to alanine at 1164 are marked. (B) Recombinant wild type GST-BUZ ¢nger from HDAC6, C!S point mutant or C!S/H!A double mutant were incubated with a mixture of mono and polyubiquitin chains branched at lysine 48. The bound fractions were visualized by immunoblotting with an antibody for ubiquitin. Note that wild-type BUZ ¢nger binds mono and polyubiquitin; however, C!S mutation completely inactivates this activity. (C) All these recombinant proteins are expressed at comparable level as judged by coomasie stain.

Structurally, HDAC6 contains two catalytic deacetylase domains. This con¢guration likely confers HDAC6’s distinct pharmacological property (Furumai et al 2001, Guardiola and Yao, 2001) and could be responsible for its unique tubulin deacetylase activity. The characterization of the BUZ ¢nger now identi¢es the second functional domain only found in HDAC6 but not any other members of the HDAC family. The signi¢cance of this domain is underscored by the observation that BUZ ¢nger is required for HDAC6 to enhance cell motility (our unpublished results). Thus, the ability to communicate with the ubiquitin network is as important as the catalytic activity for HDAC6 to regulate cell motility. Our conclusion that HDAC6 may regulate protein ubiquitination is

176

KOVACS ET AL

consistent with a previous report showing that HDAC6 associates with monoubiquitin and binds PLAP2 (Seigneurin-Berny et al 2001), a mammalian homologue of the yeast protein UFD3 involved in the ubiquitin pathway (Ghislain et al 1996). Together, these studies reveal a surprising role for HDAC6 in protein ubiquitination. Interestingly, we have found that HDAC6 is itself modi¢ed by monoubiquitination (our unpublished results). Although ubiquitination has been studied, by and large, in the context of poly-ubiquitination for protein degradation, recent studies have clearly demonstrated a critical function for mono-ubiquitination. Genetic and biochemical studies in yeast have established mono-ubiquitination as a critical modi¢cation in the regulation of endocytic tra⁄cking (Hicke 2001, Katzmann et al 2001). The fact that HDAC6 is itself modi¢ed by mono-ubiquitin suggests that HDAC6 might communicate with and regulate endocytic tra⁄cking machinery via the BUZ ¢nger. Indeed, the punctate appearance of HDAC6 in cells is reminiscent of vesicular compartments (Fig. 1). Supporting this idea, we have found that HDAC6 is present in certain endocytic vesicles (C. Hubbert, T.-P. Yao, unpublished results). A function for HDAC6 in the endocytic pathway is also consistent with its activity toward microtubules, which are the essential network for vesicle transport. If this hypothesis were true, it would indicate an interesting possibility that HDAC6regulated reversible acetylation may be a critical modi¢cation in vesicle sorting and transport. What is the signi¢cance of HDAC6 communication with the ubiquitination machinery? At least two possibilities exist. First HDAC6 may mediate cross talk between the acetylation and ubiquitination machinery by making acetylated lysine residues available for ubiquitination. This scenario is similar to that proposed for MDM2-HDAC1 complex in the regulation of p53, in which MDM2 and HDAC1 cooperatively deacetylate p53, promoting its ubiquitination and subsequent degradation (Ito et al 2002). HDAC6 may similarly deacetylate speci¢c substrates to allow for e⁄cient ubiquitination. Alternatively, HDAC6 might bind its ubiquitinated substrates through the BUZ ¢nger and regulate their acetylation status, which in turn may a¡ect protein function. In either case, the identi¢cation of speci¢c ubiquitinated proteins and/or substrates for HDAC6 will be essential to fully understand how acetylation and ubiquitination might be functionally coupled. Based on our current understanding of HDAC6, we propose that one main function of HDAC6 is to integrate the acetylation and ubiquitination machinery and thereby regulate vesicle tra⁄cking and protein degradation. If this hypothesis proves to be true, it could be the beginning of a journey that will lead us to discover many surprising roles for reversible acetylation in varying biological processes.

THE HDAC COMPLEX AND CYTOSKELETON

177

References Bird AW, Yu DY, Pray-Grant MG et al 2002 Acetylation of histone H4 by Esa1 is required for DNA double-strand break repair. Nature 419:411^415 Cai SY, Babbitt RW, Marchesi VT 1999 A mutant deubiquitinating enzyme (Ubp-M) associates with mitotic chromosomes and blocks cell division. Proc Natl Acad Sci USA 96:2828^2833 Furumai R, Komatsu Y, Nishino N, Khochbin S, Yoshida M, Horinouchi S 2001 Potent histone deacetylase inhibitors built from trichostatin A and cyclic tetrapeptide antibiotics including trapoxin. Proc Natl Acad Sci USA 98:87^92 Gao L, Cueto MA, Asselbergs F, Atadja P 2002 Cloning and functional characterization of HDAC11, a novel member of the human histone deacetylase family. J Biol Chem 277:25748^25755 Ghislain M, Dohmen RJ, Levy F, Varshavsky A 1996 Cdc48p interacts with Ufd3p, a WD repeat protein required for ubiquitin-mediated proteolysis in Saccharomyces cerevisiae. EMBO J 15:4884^4899 Grozinger CM, Hassig CA, Schreiber SL 1999 Three proteins de¢ne a class of human histone deacetylases related to yeast Hda1p. Proc Natl Acad Sci USA 96:4868^4873 Guardiola AR, Yao TP 2001 Molecular cloning and characterization of a novel histone deacetylase HDAC10. J Biol Chem 277:350^356 Hicke L 2001 A new ticket for entry into budding vesicles-ubiquitin. Cell 106:27^30 Hubbert C, Guardiola A, Shao R et al 2002 HDAC6 is a microtubule-associated deacetylase. Nature 417:455^458 Ito A, Lai CH, Zhao X et al 2001 p300/CBP-mediated p53 acetylation is commonly induced by p53-activating agents and inhibited by MDM2. EMBO J 20:1331^1340 Ito A, Kawaguchi Y, Lai C-H et al 2002 MDM2-HDAC1 mediated-deacetylation of p53 is required for its degradation. EMBO J 21:6246^6256 Katzmann DJ, Babst M, Emr SD 2001 Ubiquitin-dependent sorting into the multivesicular body pathway requires the function of a conserved endosomal protein sorting complex, ESCRT-I. Cell 106:145^155 L’Hernault SW, Rosenbaum JL 1985 Chlamydomonas alpha-tubulin is posttranslationally modi¢ed by acetylation on the epsilon-amino group of a lysine. Biochemistry 24:473^478 Matsuyama A, Shimazu T, Sumida Y et al 2002 In vivo destabilization of dynamic microtubules by HDAC6-mediated deacetylation. EMBO J 21:6820^6831 Piperno G, LeDizet M, Chang XJ 1987 Microtubules containing acetylated alpha-tubulin in mammalian cells in culture. J Cell Biol 104:289^302 Pogo BG, Allfrey VG, Mirsky AE 1966 RNA synthesis and histone acetylation during the course of gene activation in lymphocytes. Proc Natl Acad Sci USA 55:805^812 Polevoda B, Sherman F 2002 The diversity of acetylated proteins. Genome Biol 3:REVIEWS0006 Ren Q, Gorovsky MA 2001 Histone H2A.Z acetylation modulates an essential charge patch. Mol Cell 7:1329^1335 Robson SJ, Burgoyne RD 1989 Di¡erential localisation of tyrosinated, detyrosinated, and acetylated alpha-tubulins in neurites and growth cones of dorsal root ganglion neurons. Cell Motil Cytoskeleton 12:273^282 Seigneurin-Berny D, Verdel A, Curtet S et al 2001 Identi¢cation of components of the murine histone deacetylase 6 complex: link between acetylation and ubiquitination signaling pathways. Mol Cell Biol 21:8035^8044 Wade PA, Pruss D, Wol¡e AP 1997 Histone acetylation: chromatin in action. Trends Biochem Sci 22:128^132

178

DISCUSSION

DISCUSSION Allis: Is tubulin known to be ubiquitinated? Yao: That’s an interesting question. One of the ¢rst papers that characterized a ubiquitin antibody described ubiquitin staining lighting up the microtubule network. But that is the only paper that has shown that. I hadn’t, until recently, seen any follow-up on this. Work done by mass spectrometry on microtubules mentioned everything except ubiquitin, but in the last few months a paper has come out showing that tubulin can be ubiquitinated (Ren et al 2003). Allis: The acetyl site on a tubulin is around position 40. Is there any evidence that this is ubiquitinated? Yao: No. Olson: We heard earlier that other class II HDACs spend a lot of time in the cytoplasm but their function there is unknown. Is it possible they may share any functions with HDAC6? Yao: Yes. HDAC10, for example, is the closest in relation to HDAC6 and also shows cytoplasmic distribution. It is possible. There may be some functional redundancy. Olson: With respect to function, you overexpressed HDAC6 and it caused the neural ¢lipodium. If you eliminate HDAC6 by RNAi in that experiment, do you lose ¢lipodia? Yao: That is what we are now doing. The problem is that I don’t have the construct that will work in neurons. Ott: Is HDAC6 ubiquitously expressed? Yao: We did a protein tissue blot. Assuming that the extraction e⁄ciency is the same throughout the tissues, then it is most abundant in testis and brain. Ott: Do you see it everywhere? Yao: Yes. Atadja: There is vesicular sorting and tra⁄cking going on during protein maturation, from translation at the ribosomes through translocation to the membrane. Do you see HDAC6 involved in this? Yao: The short answer is yes. Castronovo: Have you looked at the colocalization of HDAC6 and the microtubules during mitosis? Yao: Yes. There is no phenotype in the cell cycle. We were surprised when we saw this because overexpression of HDAC6 causes a huge deacetylation of the microtubules. Castronovo: So do you think it might be restricted to the area where microtubules may be involved in vesicular tra⁄cking? Yao: I wouldn’t say that. In those particular areas, such as growth cones, it is probably required for modulating some kind of microtubular organization. As to

THE HDAC COMPLEX AND CYTOSKELETON

179

the function of HDAC6 in regulating microtubules, my bet is that it is regulating vesicular tra⁄cking Verdin: Do you actually see HDAC6 on the microtubules during cell division, on the spindle? We have looked for SirT2 and reported that it also binds microtubules. Yao: We have looked at this carefully, but we have not seen it. Khochbin: We got the same result. There is no association with the spindle. If you look at tubulin acetylation through the cell cycle, it is regulated. The spindles are heavily acetylated, as are the midbodies. This means that the activity of HDAC6 is regulated during the cell cycle. Somehow HDAC6 cannot deacetylate tubulin during spindle formation. It is also important to mention that ubiquitin binding modi¢es the properties of HDAC6. We showed that when HDAC6 is interacting with ubiquitin through its C-terminal zinc ¢nger (ZnF-UBP) the HDAC6containing complex is disrupted (Seigneurin-Berny et al 2001). So ubiquitin binding could modulate HDAC6 activity during cell cycle. Castronovo: Did you look at the phosphorylation of HDAC when you stimulated the cells? Yao: Yes, we have seen this before. It is quickly phosphorylated. In a couple of experiments it became phosphorylated in response to growth factor. Castronovo: In the resting cells, what is HDAC6 associated with? Yao: We can clearly see it redistribute to sit on the microtubule. I have no clue why this is, but I would love to know. Castronovo: Do you know why HDAC6 is inhibited by TSA but not by trapoxin? Yao: We have a hypothesis. We have shown that HDAC10 has one and a half catalytic domains. HDAC10 is sort of intermediate between HDAC6 and other HDAC family members in terms of inhibitor sensitivity. HDAC10 and is more resistant than other deacetylases to trapoxin B and sodium butyrate, but HDAC6 is almost completely resistant under normal situations. However, if you chop out the second half of the catalytic domain of HDAC10, this protein now becomes sensitive to sodium butyrate and trapoxin B as well. We think that this domain probably has some functional interaction and gives a particular type of conformation that changes the way the inhibitor can gain access to the active site. Gu: Do you have an in vitro puri¢ed system to prove that? Yao: No. We have a pretty crude system involving immunoprecipitation. Gu: What is the proportion of HDAC6 that was ubiquitinated? Yao: That is an important question, and I don’t know the answer. Gu: By comparing the Western analysis with anti-HDAC6 versus the analysis with anti-ubiquitin, if HDAC6 is mono-ubiquitinated you should see the size change. Yao: Yes. You can see a doublet if you run good gels. These are pretty big proteins.

180

DISCUSSION

Denu: Have you shown that the HDAC activity is required for the tra⁄ckingdependent e¡ects of HDAC6? Yao: It is not required. We know this from mutant experiments. We think there are three important properties associated with HDAC6. One is the receiving end that allows HDAC6 to recognize speci¢c signalling events and in response, it translocates to speci¢c subcellular domain/structure. The second property is that once it gets there it will deacetylate speci¢c substrates. The third property is that it will take cargo, which is the ubiquitinated protein. These three properties can be separated. Denu: What about the tubulin deacetylation? I don’t understand how tubulin deacetylation is linked to these events. Yao: We don’t yet know how to link these events. We do not know how to translate HDAC6 deacetylase activity into something that we can use to understand what HDAC6 is doing in various subcellular domains. Verdin: What do you think about the possibility that deacetylation might be the ¢rst step of ubiquitination for a number of proteins? It might be a way to regulate ubiquitination. Is there any kind of coupling between the two activities? Yao: We hypothesize this is the case for p53. There was a nice correlation between deacetylation and then ubiquitination. In the case of HDAC6, in some of the events that we are looking at here, there may be a coupling of deacetylation with ubiquitination. Until we know the substrate, however, there is no way that we can test this. Verdin: How are you proposing to look for the substrate? Yao: We have tried several methods already. We have tried treating cells with TSA versus trapoxin B, and then we have tried to partially purify it with an acetylated lysine antibody. We see bands but we don’t know what they are yet. Castronovo: Which antibody are you using? Yao: It’s a secret antibody from Dr Yoshida’s lab! Atadja: Which catalytic site is active? Yao: This is a controversial issue. The papers published so far all have di¡erent stories. In my hands the ¢rst catalytic domain, the N-terminal one, is not required. But if you chop it in half, there is no activity. I think Eric Verdin saw the same thing. Khochbin: We saw the same thing. If you mutate each domain individually, the whole activity has gone. Atadja: If I understand you correctly, you are saying that you get activity in HDAC10 by deleting the N-terminus? Yao: That is for HDAC10. In our hands HDAC6 won’t work because when we chop it in half we don’t see activity. Greene: I would like to suggest a somewhat wild idea. It turns out that cyclin B1 in interphase cells decorates the microtubule network. Then, as the cells enter G2,

THE HDAC COMPLEX AND CYTOSKELETON

181

the CDC2 kinase comes out of the nucleus leading to the assembly of cyclin B1 with CDC2 and the entry of this complex back into the nucleus just prior to nuclear membrane breakdown during mitosis. Is there any possible functional connection between microtubule-associated HDAC6 and the assembly of microtubule-associated cyclin B1 with CDC2 leading to nuclear import of this mitotic kinase complex? Yao: I would like that to be true, but the fact that we haven’t seen any obvious cell cycle e¡ect makes me doubt that this would be a major target. We have tried overexpressing it and knocking it down in synchronized cycling cells. Olson: Does it ever go in the nucleus? Yao: I haven’t seen it. References Ren Y, Zhao J, Feng J 2003 Parkin binds to alpha/beta tubulin and increases their ubiquitination and degradation. J Neurosci 23:3316^3324 Seigneurin-Berny D, Verdel A, Curtet S et al 2001 Identi¢cation of components of the murine histone deacetylase 6 complex: link between acetylation and ubiquitination signaling pathways. Mol Cell Biol 21:8035^8044

Tat acetylation: a regulatory switch between early and late phases in HIV transcription elongation Melanie Ott*{, Alexander Dorr{, Claudia Hetzer-Egger{, Katrin Kaehlcke*{, Martina Schnolzer{, Peter Henklein{, Phil Cole}, Ming-Ming Zhou} and Eric Verdin* *Gladstone Institute of Virology and Immunology, University of California, 365 Vermont Street, San Francisco, CA 94103, {Deutsches Krebsforschungszentrum (DKFZ), D-69120 Heidelberg, Germany, {Humboldt University, Institute of Biochemistry, D-10115 Berlin, Germany, } Johns Hopkins University School of Medicine, Baltimore, Maryland 21205 and }Mount Sinai School of Medicine, New York, NY 10029, USA

Abstract. The HIV transcriptional activator Tat enhances the processivity of RNA polymerase II by recruiting the CyclinT1/CDK9 complex to the TAR RNA element. In addition, Tat synergizes with the histone acetyltransferase p300 and is acetylated by p300 at a single lysine residue (K50) in the TAR RNA binding domain. We have recently reported that this post-translational modi¢cation is necessary for the interaction and transcriptional synergy of Tat with the transcriptional coactivator PCAF. We have further studied the relevance of Tat acetylation during HIV transcription and generated antibodies speci¢c for acetylated Tat (AcTat). Microinjection of anti-AcTat antibodies inhibited Tat-mediated transactivation in cells. Similarly, the speci¢c p300 inhibitor Lys-CoA and short inhibitory RNAs speci¢c for p300 suppressed Tat transcriptional activity. Full-length synthetic AcTat bound to TAR RNA and CyclinT1 with high a⁄nity, but formation of the Tat-TAR-CyclinT1 ternary complex was inhibited when K50 was acetylated. Our data collectively show that Tat acetylation by p300 de¢nes a critical step in Tat transactivation that serves to disrupt the Tat/TAR/CyclinT1 complex and helps in recruiting PCAF to the elongating RNA polymerase II. 2004 Reversible protein acetylation. Wiley, Chichester (Novartis Foundation Symposium 259) p 182^196

The rate of HIV transcription is a crucial determinant in the rate of progression towards immunode¢ciency in infected individuals. Upon entry into the cells, the HIV RNA genome is reverse transcribed to a double-stranded complementary DNA that, integrated into the host chromatin, forms the HIV provirus with one long terminal repeat (LTR) at each end. A concerted network of cellular transcription factors and the viral transactivator Tat regulates the activity of the viral promoter located in the 5’ LTR. Tat is a unique viral transactivator that 182

TAT ACETYLATION AND COFACTOR RECRUITMENT

183

binds to an RNA stem^loop structure called TAR, which forms spontaneously at the 5’ extremities of all viral transcripts (reviewed in Cullen 1998). In the absence of Tat, HIV transcription is highly ine⁄cient because the assembled RNA polymerase II complex (RNAPII) cannot elongate e⁄ciently on the viral DNA template. The binding of Tat to TAR stimulates the production of full-length HIV transcripts, and the integrity of the Tat/TAR axis critically determines the dynamics of viral replication in infected cells (Emiliani et al 1998, 1996). Chromosomal integration, an essential step in the HIV-1 life cycle, leads to the packaging of the proviral DNA into an array of precisely positioned nucleosomes (Verdin 1991). Tat activity induces the remodelling of a single nucleosome called nuc-1 located in the HIV promoter (Verdin et al 1993). The position of nuc-1 immediately after the transcription start site and the fact that it is remodelled during transcriptional activation suggest that it causes an elongation block during HIV transcription. Transcription elongation is further inhibited by the assembly of a poorly processive RNAPII complex, a process overcome by the Tat-mediated recruitment of the positive transcription elongation factor b (pTEFb). Binding of Tat to CyclinT1 in the pTEFb complex leads to the recruitment of CDK9 to the transcription initiation site. Binding of CDK9 close to this site leads to the hyperphosphorylation of the C-terminal domain of RNAPII and to an increase in polymerase processivity (Mancebo et al 1997, Zhu et al 1997). Tat interacts physically in cells with a histone acetyltransferase (HAT), the transcriptional coactivator p300 (Benkirane et al 1998, Hottiger & Nabel 1998, Marzio et al 1998), and the two proteins synergistically activate transcription of the HIV promoter (Ott et al 1999). We and others have shown that the p300 HAT activity acetylates not only histones, but also acetylates the Tat protein directly (Kiernan et al 1999, Ott et al 1999). The target of p300 acetylation is a highly conserved lysine (K50) in the arginine-rich motif (ARM) of Tat. Mutation of K50 to arginine, a conservative mutation that abolishes acetylation, inhibited the synergistic activation of the HIV promoter by Tat and p300 in transient transfection experiments, demonstrating that acetylation of K50 by p300 is critical for the transcriptional activation of the HIV promoter (Ott et al 1999). These experiments have de¢ned a novel post-translational modi¢cation in the Tat protein that a¡ects its transcriptional potential. Tat acetylation by p300 is important for Tat transactivation in vivo Three independent observations support the model that Tat acetylation by p300 de¢nes a critical step in the transcriptional activation of the HIV promoter (summarized in Fig. 1). First, we showed that nuclear microinjection of an antiserum speci¢c for AcTat suppressed Tat-mediated transactivation (Kaehlcke et al 2003). HeLa cells stably expressing a full-length Tat protein (HeLa-Tat) or

184

OTT ET AL

FIG. 1. Tat acetylation by p300 is critical for Tat transactivation. Nuclear microinjection of HeLa or HeLa-Tat cells with HIV LTR luciferase and CMV-GFP constructs (A) together with puri¢ed anti-AcARM antibodies (B) or the p300 HAT inhibitor LysCoA (C). Synthetic Tat protein was also coinjected directly into HeLa cells pretreated with siRNAs against p300 or GL3 luciferase (D). Details are described in Kaehlcke et al (2003). Reproduced from Kaehlcke et al 2003 with permission.

TAT ACETYLATION AND COFACTOR RECRUITMENT 185

186

OTT ET AL

control HeLa cells were microinjected with a HIV LTR-luciferase construct. A construct expressing enhanced GFP under the control of the CMV promoter (CMV-GFP) was coinjected to assess the number and viability of injected cells. Approximately 100^200 cells were microinjected per condition and luciferase values were on average 200-fold higher in HeLa-Tat cells than in control HeLa cells (Fig. 1A). Coinjection of puri¢ed anti-AcTat immunoglobulins into HeLa-Tat cells inhibited Tat transactivation, whereas preimmune immunoglobulins had no e¡ect (Fig. 1B). No inhibitory e¡ect was observed on Tat-independent promoters, such as the Rous sarcoma virus LTR or the 5UAS reporter activated by the Gal4-VP16 transactivator, demonstrating that the antiserum is speci¢c for Tat. These experiments directly establish that Tat acetylation at K50 is necessary for Tat transcriptional activity in vivo. Second, we observed that Lys-CoA, a speci¢c inhibitor of the HAT activity of p300, inhibited Tat transactivation (Fig. 1C). Lys-CoA is a peptide that when microinjected into nuclei of oocytes, blocks p300-dependent transcription of MyoD RNA (Lau et al 2000). Injected into HeLa-Tat cells, Lys-CoA e⁄ciently inhibited transactivation of the HIV LTR, but did not a¡ect the transcriptional activity of the CMV-GFP reporter. Microinjected into HeLa cells, it showed no e¡ect on the basal HIV LTR activity, indicating that the p300-HAT activity targets Tat directly. Third, we used RNA-mediated interference (RNAi) to suppress p300 expression. Suppression of p300 led to a reduction of Tat activity in microinjection and viral infection experiments (Fig. 1D). Interestingly, the short inhibitory RNA oligonucleotides used in these studies targeted only p300 and not the closely related CBP coactivator (Gronroos et al 2002). Whether p300 and CBP di¡er in their roles as Tat acetyltransferases in vivo remains to be determined. In addition, the in vivo contributions of other Tat K50 acetyltransferases di¡erent from p300 and CBP, i.e. human GCN5 (Col et al 2001) or PCAF (Dormeyer et al, unpublished observation) need to be established. Acetylated Tat interacts with the transcriptional coactivator PCAF via the PCAF bromodomain In the case of chromatin, the level of acetylation of distinct lysine residues in each histone protein is under the control of competing histone acetylases and histone deacetylases. While our data indicate that p300 acts as an important Tat acetyltransferase in infected cells, the nature of the Tat deacetylase remains unknown. Histone hypo-acetylation is generally associated with transcriptional repression while histone hyperacetylation has been correlated with transcriptional activation. Early models proposed that histone acetylation leads

TAT ACETYLATION AND COFACTOR RECRUITMENT

187

to a global neutralization of positive charges on histones and loosening of the histone/DNA interaction at transcriptionally active sites. However, recent data suggest that acetylated lysine residues on histone tails serve as a recognition code for the coordinated recruitment of speci¢c factors (the ‘histone code’ hypothesis) (Strahl & Allis 2000). According to this model, acetylated lysine residues in the histone tails interact with a specialized protein module, the bromodomain, present in many nuclear coactivators. To examine if acetylated Tat is required during HIV transcription because it recruits a novel cofactor via its acetyl-lysine residue, the interaction of acetylated Tat with recombinant bromodomains of several coactivators was examined. This analysis identi¢ed the bromodomain of the p300/CBP-associated factor (PCAF) as a protein module that speci¢cally recognized acetylated Tat (Mujtaba et al 2002). Tat and PCAF physically interacted and functionally synergized in cells to activate the HIV promoter (Dorr et al 2002). PCAF, like p300, exhibits HAT activity (Ogryzko et al 1996) and resides in cells in multiprotein complexes that include p300 (Ogryzko et al 1998). PCAF is also associated with the elongationcompetent hyperphosphorylated form of RNAPII while p300 is found bound at the transcriptional start site (Cho et al 1998). The PCAF bromodomain structure consists of a left-handed, four-helix bundle (helices aZ, aA, aB and aC) interconnected with loop sequences that compose the acetyl-lysine binding site (Dhalluin et al 1999). Structural analysis of the PCAF/Tat complex revealed that while the overall three-dimensional structure of the bromodomain was preserved, the ZA and BC loops underwent signi¢cant conformational changes when bound to Tat (Fig. 2A). The sidechain of acetyllysine 50 in Tat bound to a hydrophobic cavity formed by PCAF residues and made extensive contact with tyrosines 802 and 809 (Y802 and Y809) located in the acetylbinding site. In addition, peptide residues £anking acetyl-lysine 50 in Tat (Y47, R53 and Q54) interacted with PCAF residues V763 and E756 and de¢ned a highly selective association between both proteins (Mujtaba et al 2002). Mutation of each of these residues in either Tat or PCAF inhibited in a cumulative manner the Tat^PCAF interaction in vitro and in vivo and abrogated the synergistic activation of the HIV promoter by both proteins (Fig. 2B). Similarly, microinjections of a speci¢c antiserum recognizing only the bromodomain of PCAF, but not of other transciptional co-activators, inhibited Tat transactivation (Dorr et al 2002). These results demonstrate that the recruitment of PCAF by acetylated Tat plays an important role in the regulation of HIV transcription. Disruption of the Tat/TAR/CyclinT1 complex by Tat acetylation The concept of Tat as an adaptor molecule that recruits important cofactors to the HIV promoter is well established. Classically, Tat is thought to serve as a bridge to

FIG. 2. Tat acetylation generates a unique protein/protein interface with the PCAF bromodomain. Structural analysis of the Tat/PCAF interaction and the identi¢cation of residues tyrosine 47 (Y47) and arginine 53 (R53) in the Tat ARM as critical interacting residues (A). Mutation of these residues diminish Tat transcriptional activity and the synergy with PCAF in cotransfection experiments (B). Note the dominant-negative e¡ect of PCAFY809A on wild-type Tat activity. Experimental details are described in Dorr et al (2002), Mujtaba et al (2002).

188 OTT ET AL

TAT ACETYLATION AND COFACTOR RECRUITMENT

189

recruit CyclinT1/CDK9 as part of the pTEFb complex to the HIV promoter thereby facilitating transcriptional elongation (reviewed in Garber & Jones 1999). CyclinT1 directly binds to the transactivation domain of Tat (amino acids 1^48) while the neighbouring ARM domain in Tat (amino acids 49^58) serves as the RNA binding motif. Extensive biochemical and genetic evidence suggest that the recognition of TAR by the ARM domain of Tat is not su⁄cient for binding of Tat to TAR in vivo and that the transactivation domain of Tat is necessary both for binding to TAR and for transactivation. The cellular cofactor that binds the transactivation domain of Tat and promotes speci¢c Tat-binding to TAR is CyclinT1 (Wei et al 1998). The Tat/CyclinT1 complex binds to the TAR hairpin both via Tat and the TAR bulge and via CyclinT1 and the TAR loop. The assembly of the Tat/TAR/ CyclinT1 complex is transient since later during transcriptional elongation, Tat associates directly with the elongating RNAPII, a process that is independent of TAR RNA and CyclinT1 (Cujec et al 1997, Keen et al 1997, Mavankal et al 1996). To study whether acetylation of the ARM alters binding of Tat to TAR RNA, we performed RNA gel mobility assays using full-length synthetic Tat proteins (Fig. 3). In the absence of CyclinT1, AcTat bound to TAR with a⁄nities equal to that of unacetylated Tat. Addition of CyclinT1 to the reaction led to the formation of the ternary complex composed of Tat, TAR and CyclinT1, but only when K50 in Tat was not acetylated. AcTat was unable to form the ternary complex while individual TAR and CyclinT1-binding activities were unchanged (Kaehlcke et al 2003). These data suggest that Tat acetylation, by preventing CyclinT1 recruitment, leads to a dissociation of Tat from TAR in vivo and serves to load Tat onto the elongating RNAPII complex. Indeed, synthetic biotinylated AcTat bound with higher a⁄nity to the elongation-competent RNAPII than unacetylated Tat, supporting the model that Tat acetylation regulates the transition between early, TAR-dependent and late, TAR-independent steps of HIV transcription elongation (Kaehlcke et al 2003). A model of Tat acetylation as a molecular switch between early and late steps of HIV transcription elongation These observations lead us to propose the following two-step model (Fig. 4): the ¢rst stage is mediated by unacetylated Tat acting together with CyclinT1 through TAR RNA to recruit the pTEFb complex to the HIV promoter. When bound to TAR, Tat is subjected to acetylation by p300 bound at the level of the HIV promoter. Acetylation of Tat by p300 leads to the disruption of the Tat/ CyclinT1/TAR complex and to the high-a⁄nity recruitment of AcTat to the elongating RNAPII. Whether AcTat binds to RNAPII via PCAF present in the elongating complex or whether PCAF is recruited to the complex via AcTat,

FIG. 3. Tat acetylation prevents the formation of the Tat/TAR/CyclinT1 complex. RNA bandshift experiments with radiolabelled TAR RNA probes, synthetic full-length acetylated or nonacetylated Tat proteins, and recombinant CyclinT1 protein (Reproduced from Kaehlcke et al 2003, with permission).

190 OTT ET AL

FIG. 4.

A model of Tat acetylation controlling the transition from TAR-dependent to TAR-independent steps in HIV transcription elongation.

TAT ACETYLATION AND COFACTOR RECRUITMENT 191

192

OTT ET AL

remains unclear. However, the recruitment of the PCAF HAT activity close to nuc-1 might account for the speci¢c chromatin remodelling events that are associated with Tat activity in vivo. Future experiments will further examine this model and de¢ne the role of AcTat with respect to the elongation competence of the RNAPII complex and the chromatin organization of the integrated HIV promoter.

References Benkirane M, Chun RF, Xiao H et al 1998 Activation of integrated provirus requires histone acetyltransferase. p300 and P/CAF are coactivators for HIV-1 Tat. J Biol Chem 273:24898^ 24905 Cho H, Orphanides G, Sun X et al 1998 A human RNA polymerase II complex containing factors that modify chromatin structure. Mol Cell Biol 18:5355^5363 Col E, Caron C, Seigneurin-Berny D, Gracia J, Favier A, Khochbin S 2001 The histone acetyltransferase, hGCN5, interacts with and acetylates the HIV transactivator, Tat. J Biol Chem 276:28179^28184 Cujec TP, Cho H, Maldonado E, Meyer J, Reinberg D, Peterlin BM 1997 The human immunode¢ciency virus transactivator Tat interacts with the RNA polymerase II holoenzyme. Mol Cell Biol 17:1817^1823 Cullen BR 1998 HIV-1 auxiliary proteins: making connections in a dying cell. Cell 93:685^692 Dhalluin C, Carlson JE, Zeng L, He C, Aggarwal AK, Zhou MM 1999 Structure and ligand of a histone acetyltransferase bromodomain. Nature 399:491^496 Dorr A, Kiermer V, Pedal A et al 2002 Transcriptional synergy between Tat and PCAF is dependent on the binding of acetylated Tat to the PCAF bromodomain. EMBO J 21:2715^ 2723 Emiliani S, Van Lint C, Fischle W et al 1996 A point mutation in the HIV-1 Tat responsive element is associated with postintegration latency. Proc Natl Acad Sci USA 93:6377^6381 Emiliani S, Fischle W, Ott M, Van Lint C, Amella CA, Verdin E 1998 Mutations in the tat gene are responsible for human immunode¢ciency virus type 1 postintegration latency in the U1 cell line. J Virol 72:1666^1670 Garber ME, Jones KA 1999 HIV-1 Tat: coping with negative elongation factors. Curr Opin Immunol 11:460^465 Gronroos E, Hellman U, Heldin CH, Ericsson J 2002 Control of Smad7 stability by competition between acetylation and ubiquitination. Mol Cell 10:483^493 Hottiger MO, Nabel GJ 1998 Interaction of human immunode¢ciency virus type 1 Tat with the transcriptional coactivators p300 and CREB binding protein. J Virol 72:8252^8256 Kaehlcke K, Dorr A, Hetzer-Egger C et al 2003 Acetylation of Tat de¢nes a CyclinT1independent step in HIV transactivation. Mol Cell 12:167^176 Keen NJ, Churcher MJ, Karn J 1997 Transfer of Tat and release of TAR RNA during the activation of the human immunode¢ciency virus type-1 transcription elongation complex. EMBO J 16:5260^5272 Kiernan RE, Vanhulle C, Schiltz L et al 1999 HIV-1 tat transcriptional activity is regulated by acetylation. EMBO J 18:6106^6118 Lau OD, Kundu TK, Soccio RE et al 2000 HATs o¡: selective synthetic inhibitors of the histone acetyltransferases p300 and PCAF. Mol Cell 5:589^595 Mancebo HS, Lee G, Flygare J et al 1997 P-TEFb kinase is required for HIV Tat transcriptional activation in vivo and in vitro. Genes Dev 11:2633^2644

TAT ACETYLATION AND COFACTOR RECRUITMENT

193

Marzio G, Tyagi M, Gutierrez MI, Giacca M 1998 HIV-1 tat transactivator recruits p300 and CREB-binding protein histone acetyltransferases to the viral promoter. Proc Natl Acad Sci USA 95:13519^13524 Mavankal G, Ignatius Ou SH, Oliver H, Sigman D, Gaynor RB 1996 Human immunode¢ciency virus type 1 and 2 Tat proteins speci¢cally interact with RNA polymerase II. Proc Natl Acad Sci USA 93:2089^2094 Mujtaba S, He Y, Zeng L et al 2002 Structural basis of lysine-acetylated HIV-1 Tat recognition by PCAF bromodomain. Mol Cell 9:575^586 Ogryzko VV, Schiltz RL, Russanova V, Howard BH, Nakatani Y 1996 The transcriptional coactivators p300 and CBP are histone acetyltransferases. Cell 87:953^959 Ogryzko VV, Kotani T, Zhang X et al 1998 Histone-like TAFs within the PCAF histone acetylase complex. Cell 94:35^44 Ott M, Schn˛lzer M, Garnica J et al 1999 Acetylation of the HIV-1 Tat protein by p300 is important for its transcriptional activity. Curr Biol 9:1489^1492 Strahl BD, Allis CD 2000 The language of covalent histone modi¢cations. Nature 403:41^45 Verdin E 1991 DNase I-hypersensitive sites are associated with both long terminal repeats and with the intragenic enhancer of integrated human immunode¢ciency virus type 1. J Virol 65:6790^6799 Verdin E, Paras P Jr, Van Lint C 1993 Chromatin disruption in the promoter of human immunode¢ciency virus type 1 during transcriptional activation. EMBO J 12:3249^3259 Wei P, Garber ME, Fang SM, Fischer WH, Jones KA 1998 A novel CDK9-associated C-type cyclin interacts directly with HIV-1 Tat and mediates its high-a⁄nity, loop-speci¢c binding to TAR RNA. Cell 92:451^462 Zhu Y, Pe’ery T, Peng J et al 1997 Transcription elongation factor P-TEFb is required for HIV-1 tat transactivation in vitro. Genes Dev 11:2622^2632

DISCUSSION Allis: Is there any chance at all that the Lys50 gets methylated? This sequence looks suspiciously similar to Lys36 in the H3 tail. Ott: There have been some rumours in the ¢eld that Tat is methylated. It’s a possibility. Khochbin: Could you comment on the acetylation of Lys28? Several papers have described this. Ott: You are referring to the observation that there is another proposed acetylation site in Tat which actually maps to the cysteine-rich region. The cysteine-rich region also contains a double lysine motif  K28 and K29. Benkirane’s group has reported that PCAF acetylates K28. You have reported that human GCN5, which is very similar to PCAF, also acetylates K50/K51. We ¢nd the same thing  that PCAF acetylates K50 a little bit less e⁄ciently than p300, but there is no doubt that it does it. We have spent a lot of time looking at K28. We ¢nd that none of the lysines in that region are acetylated, but that the cysteines in this region are good acetyl acceptors in in vitro reactions. Thus we see a very strong non-enzymatic chemical acetylation of the peptides that contain this region. However, if we look by mass spectrometry we see at least four acetyl sites, and there are only two lysines in there. If these two lysines are mutated, Tat is still

194

DISCUSSION

fourfold acetylated. If the cysteines are protected acetylation is totally abrogated. We have obtained similar results with an antibody raised against acetyl-K28. Greene: Is there any possibility that the mechanism for recruitment of acetylated Tat and PCAF ends up protecting the C-terminal domain of RNA polymerase II from dephosphorylation, allowing the polymerase to elongate e¡ectively over the 9 kb viral genome? Ott: It is possible. We don’t know yet whether what we are seeing in our binding assays is direct binding to the CTD. We are pulling down the whole RNA PolII complex. Acetylated Tat might interact with another factor in the PolII complex which is present there when the CTD is hyperphosphorylated. We plan to study this. Greene: Is there any possibility that the CTD domain of RNA polymerase II might also serve as a substrate for acetylation by the recruited Tat/PCAF complex? Ott: We have started to look at this. Marmorstein: Does acetylated Tat interact with the bromodomain of p300? Ott: No. It is a very speci¢c interaction that is completely due to these additional interaction sites. It is not just a simple acetyl lysine-mediated interaction. As I understand it, every acetylated lysine can in some way interact with a bromodomain. The Tat protein binds to the PCAF bromodomain with a very high a⁄nity. This also makes this interaction unique and sets it apart from other acetyl lysine^bromodomain interactions. We are very excited about this structure as a potential drug target. Chen: Have you ever tried to microinject acetylated Tat together with this Lys CoA to see whether Lys CoA inhibits Tat activity? Ott: We are just doing this. Acetylated Tat is an interesting protein. It is very highly active if it is injected. According to our model, we think it is highly active because it becomes deacetylated in the cell. We know this because if we incubate acetylated Tat with nuclear extracts, we see rapid deacetylation of K50. If we inject K50 acetylated Tat protein, we think the activity  which is actually higher than the activity of the non-acetylated Tat  depends on the initial partial deacetylation of the protein. We are using this as a tool to identify the Tat deacetylase. Hottiger: If your hypothesis is correct you would expect acetylated Tat to be associated with downstream regions of the HIV genome. Have you done ChIPs using your antibody to test this hypothesis? Ott: We haven’t done this yet. Seto: Is it possible to mutate the K50 acetylated site on Tat and see whether it has an e¡ect on the viral life cycle? Ott: Yes, we have mutated it, and Bre' s and colleagues have mutated it in the context of an infectious virus (Bre' s et al 2002). They see a delay in viral replication with the K50-mutated Tat protein.

TAT ACETYLATION AND COFACTOR RECRUITMENT

195

Seto: How about in terms of looking at the e¡ect on RNA-binding: do you have more direct evidence? The experiment you did was putting in p300 and showing that acetylated Tat inhibits its ability to bind RNA. Is that correct? Ott: No, the point was that we are using synthetic acetylated proteins that carry one acetyl group at position K50, and that we are binding it to Tar RNA in a gel shift assay. We see that it binds perfectly. If we put a relevant cofactor for binding in vivo, such as CyclinT, we see that acetylated Tat still binds, but it does not form a complex with CyclinT. What I said about p300 is that if we use non-acetylated Tat, bind it to TAR and CyclinT and acetylate the complex with p300 HAT and acetylCoA, this disrupts the complex. This shows us that K50 is exposed or accessible for p300 when Tat is in a complex with TAR and CyclinT. Seto: Was this using puri¢ed Tat? Ott: This is synthetic Tat. Seto: Is it possible to locate in vivo some mutant of Tat that no longer gets acetylated, to determine whether it can still bind RNA? Ott: We can generate a mutant, but the TAR RNA mobility shift assays are not easy to do. You cannot use cell lysates, you have to use a puri¢ed Tat protein. Seto: Is this the ¢rst case where protein acetylation is demonstrated to be important for RNA (not DNA) binding? Ott: It is not abrogating RNA binding in vitro. This is discrepant from what other groups have reported. Kiernan and colleagues ¢nd that if they use an acetylated K50 protein they no longer see in vitro RNA binding (Kiernan et al 1999). Khochbin: Can acetylated Tat enter cells easily, like non-acetylated Tat? Ott: We are looking at this. So far, when we incubate Tat and acetylated Tat with cells and look for transactivation, we see similar results. But we don’t know what happens to the acetylated Tat in the cell. Khochbin: You could follow this with your antibody. Ott: The antibody isn’t good for immuno£uorescence. Li: Is the TAR sequence essential for HIV replication? Ott: Yes. Mahadevan: You mentioned in passing that when you treated the cells with TSA, they actually had higher levels of Tat. Could you comment on this? Ott: Many promoters are responsive to TSA, and are up-regulated. If you use a simple Tat expression vector with a CMV promoter you get higher Tat expression if you treat the cells with TSA. In the system I described, Tat drives its own production from the HIV-1 promoter. Also this process is sensitive to TSA, as Eric Verdin has shown, through hyperacetylation of nucleosomes but also probably through a Tat-associated mechanism that we haven’t identi¢ed. Berger: Have you looked at the interrelationship with histone acetylation?

196

DISCUSSION

Ott: Not yet. This will be our main focus, looking at this second step and how PCAF recruitment is relevant for binding of acetylated Tat to the elongating polymerase and how this a¡ects nuc-1 acetylation. References Bre' s V, Kiernan R, Emiliani S, Benkirane M 2002 Tat acetyl-acceptor lysines are important for human immunode¢ciency virus type-1 replication. J Biol Chem 277:22215^22221 Kiernan RE, Vanhulle C, Schiltz L et al 1999 HIV-1 tat transcriptional activity is regulated by acetylation. EMBO J 18:6106^6118

Dynamics of the p53 acetylation pathway Wei Gu, Jianyuan Luo, Chris L. Brooks, Anatoly Y. Nikolaev and Muyang Li Institute for Cancer Genetics, and Department of Pathology, College of Physicians & Surgeons, Columbia University, 1150 St. Nicholas Avenue, New York, NY 10032, USA

Abstract. The p53 tumour suppressor exerts anti-proliferative e¡ects, including growth arrest, apoptosis and cell senescence, in response to various types of stress. However, p53 is a short-lived protein and its activity is maintained at low levels in normal cells. Numerous studies indicate that CBP/p300-mediated acetyl-transferase activity is critical for its role in both catalysing p53 acetylation and activating p53-mediated function during stress response. Interestingly, two additional regulators have also been identi¢ed in the p53 acetylation pathway. PID/MTA2 is a p53-interacting protein that induces p53 deacetylation by recruiting the HDAC1 complex. Subsequent work has also identi¢ed Sir2a, a NAD-dependent histone deacetylase that can attenuate p53 transcriptional activity through deacetylation. The prominence of deacetylase activity on p53 certainly raises the de¢ning question of its physiological purpose. It is likely that deacetylation provides a quick acting mechanism to stop p53 function once transcriptional activation of target genes is no longer needed. We present data indicating that both HDAC1 and Sir2a are critical for p53-dependent stress response. Furthermore, we also try to de¢ne the functional consequence of p53 acetylation at the molecular level. Finally, we propose a model regarding the di¡erential roles of HDAC1 and Sir2a in the regulation of p53 function. 2004 Reversible protein acetylation. Wiley, Chichester (Novartis Foundation Symposium 259) p 197^207

Background and signi¢cance The p53 protein acts as a bona ¢de tumour suppressor that can induce cell growth arrest, apoptosis, and aging/cell senescence in response to various types of stress (Vogelstein et al 2000). Since p53 mutations are found in more than half of all human tumours, inactivation of p53 function has been regarded as one of the most critical steps in tumorigenesis. Accumulating evidence further indicates that, in the cells that retain wild-type p53, other defects in the p53 pathway also play an important role in tumorigenesis (Prives & Hall 1999). The molecular function of p53 that is required for tumour suppression involves its ability to act as a transcriptional factor in regulating endogenous gene expression. A number of genes which are critically involved in either cell growth arrest or apoptosis have 197

198

GU ET AL

been identi¢ed as direct targets of p53, including p21CIP1/WAF1, Mdm2, GADD45, Cyclin G, 14-3-3s, Noxa, p53AIP1, PUMA and p53RDL1 (reviewed in Tanikawa et al 2003). In unstressed cells, p53 exists as a latent form and is maintained at very low levels, mainly due to tight regulation by the Mdm2 oncoprotein through functional inhibition and protein degradation mechanisms (reviewed in Freedman et al 1999). Mdm2 functions as a p53-speci¢c E3 ligase and plays a major factor in the regulation of p53 stability (Haupt et al 1997). The physiological role of Mdm2 in regulating p53 is best illustrated by genetic studies in mice. Inactivation of p53 was shown to completely rescue the embryonic lethality caused by the loss of Mdm2 function in mice (Jones et al 1995, Montes de Oca Luna et al 1995). Tight regulation of p53 is essential for its e¡ect on tumorigenesis as well as maintaining normal cell growth. Although the precise mechanism by which p53 is activated by cellular stress is not completely understood, it is generally thought to involve post-translational modi¢cations including acetylation (Brooks & Gu 2003). Early studies demonstrated that CBP/p300, a histone acetyl-transferase (HAT), acts as a coactivator of p53 and potentiates its transcriptional activity as well as biological function in vivo (Gu et al 1997). Genetic studies have also revealed that p300 mutations are present in several types of tumours, and that mutations of CBP in human Rubeinstein-Taybi syndrome as well as CBP knockout mice lead to higher risk of tumorigenesis. All of these studies support an important role for this p300/CBP^p53 interaction in the tumour suppressor pathway (reviewed in Goodman & Smolik 2000). Signi¢cantly, the observation of functional synergism between p53 and CBP/p300 together with its intrinsic HAT activity led to the discovery of a novel FAT (transcriptional factor acetyl-transferase) activity of CBP/p300 on p53; this ¢nding also predicted that acetylation may represent a general functional modi¢cation for non-histone proteins in vivo (Gu & Roeder 1997). In fact, over the past several years, this notion has been veri¢ed for many other transcriptional factors (reviewed in Kouzarides 2000). The critical role of acetylation for p53 activation p53 is speci¢cally acetylated at multiple lysine residues (Lys 370, 372, 373, 381, 382) of the C-terminal regulatory domain by CBP/p300. Although the precise role of p53 acetylation in transcriptional activation is still not completely understood (Prives & Manley 2001), acetylation of p53 can dramatically stimulate its sequence-speci¢c DNA binding activity in vitro, possibly as a result of an acetylation-induced conformational change (Gu & Roeder 1997, Sakaguchi et al 1998, Liu et al 1999). By developing site-speci¢c acetylated p53 antibodies, CBP/ p300 mediated acetylation of p53 was further con¢rmed in vivo by a number of

REGULATION OF p53 FUNCTION BY ACETYLATION

199

studies (reviewed in Chao et al 2000, Ito et al 2001). In addition, p53 can also be acetylated at Lys320 by another HAT cofactor, PCAF, although the in vivo functional consequence of this event needs to be further elucidated (Sakaguchi et al 1998, Liu et al 1999). Signi¢cantly, the steady-state levels of acetylated p53 are stimulated in response to various types of stress, indicating the important role of p53 acetylation in stress response (reviewed in Ito et al 2001). By introducing a transcriptionally defective p53 mutant (p53Q25S26) into mice, it was found that the mutant mouse thymocytes and ES cells failed to undergo DNA damage-induced apoptosis (Chao et al 2000). Interestingly, this mutant protein was phosphorylated normally at the N-terminus in response to DNA damage but could not be acetylated at the C-terminus (Chao et al 2000), supporting a critical role of p53 acetylation in transactivation as well as the p53-dependent apoptotic response (Chao et al 2000, Luo et al 2000). Interestingly, acetylation of p53 has been found critical for e⁄cient recruitment in vivo of CBP and the PCAF complex to promoter regions, and a mutation at the p53 acetylation site attenuates its mediated transcriptional activation of in vivo target gene expression (Barlev et al 2001). Furthermore, it has been reported that oncogenic Ras as well as PML can upregulate the levels of acetylated p53 in normal primary ¢broblasts and also induce premature senescence in a p53-dependent manner (Pearson et al 2000, Ferbeyre et al 2000). p53 acetylation may also play a critical role in protein stabilization (Rodriguez et al 2000, Ito et al 2001, Li et al 2002). In addition, another independent study showed that acetylation, but not phosphorylation of the p53 C-terminus, may be required to induce metaphase chromosome fragility in the cell (Yu et al 2000). Thus, CBP/p300-dependent acetylation of p53 has already been implicated as a critical event in p53-mediated transcriptional activation, apoptosis, senescence, and chromosome fragility. Deacetylation of p53 inhibits its mediated biological function In contrast to acetylation, much less is known about the role of deacetylation in modulating p53 function. Under normal conditions, the proportion of acetylated p53 in cells remains low. This may re£ect the action of strong deacetylase activities in vivo. Indeed, the acetylation level of p53 is enhanced when cells are treated with histone deacetylase (HDAC) inhibitors such as trichostatin A (TSA). These observations led to identi¢cation of a HDAC1 complex which is directly involved in p53 deacetylation and functional regulation (Luo et al 2000); PID/ MTA2, a component of this HDAC1 complex, acts as an adaptor protein to enhance HDAC1-mediated deacetylation of p53, but this activity can be completely repressed by TSA (Luo et al 2000). In addition, Mdm2 also actively suppresses CBP/p300-mediated p53 acetylation, and this inhibitory e¡ect can be abrogated by the tumour suppressor p19ARF, suggesting that regulation of

200

GU ET AL

acetylation also plays a critical role in the p53-MDM2-p19ARF feed back loop (Ito et al 2001). The yeast silent information regulator 2 (Sir2) protein belongs to a novel family of histone deacetylases, which is involved in gene silencing, telomere position e¡ects, and cell ageing (reviewed in Guarente 2000). The NAD-dependent deacetylase activity of Sir2 is essential for its functions, and this activity also connects its biological role with cellular metabolism (Guarente 2000, Imai et al 2000). Recently, mammalian Sir2 homologues have been found to also possess the NAD-dependent HDAC activity (Imai et al 2000), further supporting the idea that the enzymatic activity is key to elucidate the molecular mechanisms of its functions. Thus far, most information about Sir2 mediated functions comes from studies in yeast (Moazed 2001). Among Sir2 and its homologue proteins (HSTs) in yeast, Sir2 is the only protein exclusively localized in nuclei, which is critical for both gene silencing and extension of yeast life-span (reviewed in Guarente 2000). Based on protein sequence homology analysis, mouse Sir2a and its human orthologue SIRT1 (or human Sir2a) are the closest homologues to yeast Sir2 (Imai et al 2000, Frye 1999), and both of them, like yeast Sir2, exhibit nuclear localization. Since homologues of Sir2 have been identi¢ed in almost all organisms examined including bacteria, which has no histone proteins (reviewed in Gray & Ekstrom 2001, Frye 1999), it is likely that Sir2 also targets non-histone proteins for functional regulation. This hypothesis led us to ¢nd: . that p53 strongly binds to mouse Sir2a as well as its human orthologue hSIRT1 both in vitro and in vivo . that p53 is a substrate for the NAD-dependent deacetylase of mammalian Sir2a . that the Sir2a-mediated deacetylation antagonizes p53-dependent transcriptional activation and apoptosis . that the Sir2a-mediated deacetylation of p53 is inhibited by nicotinamide both in vitro and in vivo . that Sir2a speci¢cally inhibits p53-dependent apoptosis in response to DNA damage as well as oxidative stress, but not the p53-independent, Fas-mediated cell death, and . that expression of a Sir2a dominant negative mutant (Sir2a-363Y) increases the sensitivity of cells in response to stress. These results are especially relevant to the multiple regulatory pathways of p53 in vivo and, since the acetylation levels of p53 are stimulated in response to various types of stress, to the role of mammalian Sir2a in stress response (Luo et al 2001). Interestingly, our results are also further supported by independent studies from other groups (Vaziri et al 2001, Langley et al 2002).

REGULATION OF p53 FUNCTION BY ACETYLATION

201

Coincidentally, in oncogene-induced premature senescence of cells, the p53 negative regulatory pathway controlled by Mdm2 appears to be also blocked (reviewed in Sherr & Weber 2000); however, in contrast to the DNA damage response, the Mdm2-mediated pathway is abrogated by induction of p14ARF (or mouse p19ARF) in these cells (Honda & Yasuda 1999). Furthermore, when primary ¢broblasts undergo senescence, a progressive increase of the p53 acetylation levels was observed in serially passaged cells (Pearson et al 2000). Oncogenic Ras and PML induce p53-dependent premature senescence and upregulate the p53 acetylation levels in both mouse and human normal ¢broblasts (Pearson et al 2000, Ferbeyre et al 2000). Thus, these studies together with our ¢ndings suggest that mammalian Sir2a-mediated regulation may also play an important role in oncogene-induced premature senescence as well as replicative senescence. Acetylation of p53 blocks its ubiquitination by Mdm2 As in the case of acetylation, the e-amino group of the substrate lysine residue is also the target for ubiquitination (Pickart 2001). Signi¢cantly, recent studies have indicated that the lysine residues at the C-terminal domain of p53, ¢ve of which are the acetylation sites, play a critical role in Mdm2-mediated ubiquitination and subsequent degradation (Rodriguez et al 2000). Furthermore, increasing the levels of p53 acetylation with deacetylase inhibitors in the cell also prevents p53 from degradation in vivo (Ito et al 2001). Therefore, it is reasonable to speculate that acetylation of p53 may directly regulate its ubiquitination levels in vivo. However, thus far, there is no direct evidence regarding how p53 acetylation a¡ects its ubiquitination, mainly because of technical di⁄culties. One of the major obstacles in elucidating the precise role of acetylation in the regulation of protein function is obtaining the pure acetylated form. Since the acetylated non-histone proteins are inseparable from the unacetylated forms on a regular SDS-PAGE gel, thus far, most of the functional studies on newly-identi¢ed acetylated substrates including p53 are derived from the ‘acetylated proteins’ without any quantitative analysis of their contents (Gu & Roeder 1997, Sakaguchi et al 1998, Liu et al 1999, Ito et al 2001). Therefore, it is very di⁄cult to provide consistent results about the functional di¡erences between the ‘unacetylated form’ and the ‘acetylated form’ of these proteins as the ‘acetylated form’ may be heavily contaminated with the ‘unacetylated form’. To provide direct evidence that acetylation of p53 can modulate its ubiquitination by Mdm2, we developed a method to obtain the pure form of acetylated p53 from cells. Furthermore, we have demonstrated for the ¢rst time that acetylation of p53 directly inhibits its ubiquitination-dependent proteolysis. More importantly, we have also provided evidence elucidating a novel mechanism for acetylation-induced e¡ects on p53

202

GU ET AL

ubiquitination: acetylation of p53 not only blocks the e-amino group of the acetylated lysine residues for ubiquitination, but also attenuates Mdm2-mediated ubiquitination of other unacetylated lysine residues, possibly through inducing a protein conformational change (Li et al 2002). Acetylation of p53 augments its site-speci¢c DNA binding in vivo Numerous studies indicate that the C-terminus of p53 acts as a critical regulator of p53 and negatively modulates its transcriptional activation. Deletion of the Cterminus, injection of antibodies speci¢c for the C-terminus (PAb421), singlestrand DNA, protein^protein interactions such as HMG-1, and posttranslational modi¢cations at this region all induce profound p53 transactivation abilities (Hupp et al 1995, Selivanova et al 1997). Consistent with this model, the acetylation of p53 can also dramatically stimulate its sequence-speci¢c DNA binding activity in vitro, possibly as a result of an acetylation-induced conformational change (Gu & Roeder 1997). These results were con¢rmed in several other similar studies (Sakaguchi et al 1998, Liu et al 1999). To further verify the notion that acetylation of p53 enhances its sequence-speci¢c DNA binding, we tested the DNA binding activity of highly puri¢ed acetylated p53 proteins on di¡erent binding templates. More importantly, we also performed the Chromatin Immunoprecipitation assay and demonstrated that acetylation of p53 augments the DNA binding activity to its endogenous target promoter (J. Luo & W. Gu, unpublished results). Conclusions Recent advances in understanding the transcriptional activation of p53 have added a layer of complexity to this pathway. The number of mechanisms used to enhance (e.g. CBP/p300) or down-regulate (HDAC1 and Sir2a) p53-mediated function implicates this process as a key target in the response to genotoxic stress. The functional consequences of p53 acetylation are diverse and include increased DNA binding, enhancement of stability, and changes in protein^protein interactions. The prominence of deacetylase activity on p53 certainly raises the de¢ning question of its physiological purpose. One possibility is that deacetylation provides a quick acting mechanism to stop p53 function once transcriptional activation of target genes is no longer needed. Targeted deacetylation has been shown to occur very quickly amidst a global equilibrium of genomic acetylation and deacetylation (Fig. 1). Restoration of this steady-state level at p53 target genes is crucial for cellular homeostasis once DNA repair is complete. Deacetylation could also serve as an important step in Mdm2-mediated p53 degradation (Fig. 1).

FIG. 1. The p53 acetylation pathway. Upon DNA damage, p53 is acetylated by CBP/p300 and PCAF. The acetylated p53 induces transcriptional activation of its downstream target genes, which lead to cell growth arrest, apoptosis, or cellular senescence. In contrast, Sir2a and PID/HDAC1 provide crucial p53 co-repressor function to shut o¡ p53-dependent transcription if p53 activation becomes unnecessary. Nicotinamide and TSA are inhibitors of Sir2a and PID/HDAC1, respectively. A, acetylation; TSA, trichostatin A.

REGULATION OF p53 FUNCTION BY ACETYLATION 203

204

GU ET AL

References Barlev NA, Liu L, Chehab NH et al 2001 Acetylation of p53 activates transcription through recruitment of coactivators/histone acetyltransferases. Mol Cell 8:1243^1254 Brooks CL, Gu W 2003 Ubiquitination, phosphorylation and acetylation: the molecular basis for p53 regulation. Curr Opin Cell Biol 15:164^171 Chao C, Saito S, Kang J, Anderson CW, Appella E, Xu Y 2000 p53 transcriptional activity is essential for p53-dependent apoptosis following DNA damage. EMBO J 19:4967^4975 Ferbeyre G, de Stanchina E, Querido E, Baptiste N, Prives C, Lowe SW 2000 PML is induced by oncogenic ras and promotes premature senescence. Genes Dev 14: 2015^2027 Freedman DA, Wu L, Levine AJ 1999 Functions of the MDM2 oncoprotein. Cell Mol Life Sci 55:96^107 Frye RA 1999 Characterization of ¢ve human cDNAs with homology to the yeast SIR2 gene: Sir2-like proteins (sirtuins) metabolize NAD and may have protein ADP-ribosyltransferase activity. Biochem Biophys Res Commun 260:273^279 Goodman RH, Smolik S 2000 CBP/p300 in cell growth, transformation, and development. Genes Dev 14:1553^1577 Gray SG, Ekstrom TJ 2001 The human histone deacetylase family. Exp Cell Res 262: 75^83 Gu W, Roeder RG 1997 Activation of p53 sequence-speci¢c DNA binding by acetylation of the p53 C-terminal domain. Cell 90:595^606 Gu W, Shi XL, Roeder RG 1997 Synergistic activation of transcription by CBP and p53. Nature 387:819^823 Guarente L 2000 Sir2 links chromatin silencing, metabolism, and aging. Genes Dev 14:1021^ 1026 Haupt Y, Maya R, Kazaz A, Oren M 1997 Mdm2 promotes the rapid degradation of p53. Nature 387:296^299 Honda R, Yasuda H 1999 Association of p19(ARF) with Mdm2 inhibits ubiquitin ligase activity of Mdm2 for tumor suppressor p53. EMBO J 18:22^27 Hupp TR, Sparks A, Lane DP 1995 Small peptides activate the latent sequence-speci¢c DNA binding function of p53. Cell 83:237^245 Imai S, Armstrong CM, Kaeberlein M, Guarente L 2000 Transcriptional silencing and longevity protein Sir2 is a NAD-dependent histone deacetylase. Nature 403:795^800 Ito A, Lai C, Zhao X et al 2001 p300/CBP-mediated p53 acetylation is commonly induced by p53-activating agents and inhibited by MDM2. EMBO J 20:1331^1340 Jones SN, Roe AE, Donehower LA, Bradley A 1995 Rescue of embryonic lethality in Mdm2de¢cient mice by absence of p53. Nature 378:206^208 Kouzarides T 2000 Acetylation: a regulatory modi¢cation to rival phosphorylation? EMBO J 19:1176^1179 Langley E, Pearson M, Faretta M et al 2002 Human SIR2 deacetylates p53 and antagonizes PML/ p53-induced cellular senescence. EMBO J 21:2383^2396 Li M, Luo J, Brooks CL, Gu W 2002 Acetylation of p53 inhibits its ubiquitination by Mdm2. J Biol Chem 277:50607^50611 Liu L, Scolnick DM, Trievel RC et al 1999 p53 sites acetylated in vitro by PCAF and p300 are acetylated in vivo in response to DNA damage. Mol Cell Biol 19:1202^1209 Luo J, Su F, Chen D, Shiloh A, Gu W 2000 Deacetylation of p53 modulates its e¡ect on cell growth and apoptosis. Nature 408:377^381 Luo J, Nikolaev AY, Imai S et al W 2001 Negative control of p53 by Sir2alpha promotes cell survival under stress. Cell 107:137^148 Moazed D 2001 Enzymatic activities of Sir2 and chromatin silencing. Curr Opin Cell Biol 13:232^238

REGULATION OF p53 FUNCTION BY ACETYLATION

205

Montes de Oca Luna R, Wagner DS, Lozano G 1995 Rescue of early embryonic lethality in mdm2-de¢cient mice by deletion of p53. Nature 378:203^206 Pearson M, Carbone R, Sebastiani C et al 2000 PML regulates p53 acetylation and premature senescence induced by oncogenic Ras. Nature 406:207^210 Pickart CM 2001 Mechanisms underlying ubiquitination. Annu Rev Biochem 70:503^533 Prives C, Hall PA 1999 The p53 pathway. J Pathol 187:112^126 Prives C, Manley JL 2001 Why is p53 acetylated? Cell 107:815^818 Rodriguez MS, Desterro JM, Lain S, Lane DP, Hay RT 2000 Multiple C-terminal lysine residues target p53 for ubiquitin-proteasome-mediated degradation. Mol Cell Biol 20:8458^8467 Sakaguchi K, Herrera JE, Saito S et al 1998 DNA damage activates p53 through a phosphorylation-acetylation cascade. Genes Dev 12:2831^2841 Selivanova G, Iotsova V, Okan I et al 1997 Restoration of the growth suppression function of mutant p53 by a synthetic peptide derived from the p53 C-terminal domain. Nat Med 6: 632^638 Sherr CJ, Weber JD 2000 The ARF/p53 pathway. Curr Opin Genet Dev 10:94^99 Tanikawa C, Matsuda K, Fukuda S, Nakamura Y, Arakawa H 2003 p53RDL1 regulates p53dependent apoptosis. Nat Cell Biol 5:216^223 Vaziri H, Dessain SK, Ng Eaton E et al 2001 hSIR2(SIRT1) functions as an NAD-dependent p53 deacetylase. Cell 107:149^159 Vogelstein B, Lane D, Levine AJ 2000 Sur¢ng the p53 network. Nature 408:307^310 Yu A, Fan H, Lao D, Bailey AD, Weiner AM 2000 Activation of p53 or loss of the Cockayne syndrome group B repair protein causes metaphase fragility of human U1, U2, and 5S genes. Mol Cell 5:801^810

DISCUSSION Baylin: What do you know about the speci¢city of the Sir2 family members? Gu: We have tried many others, but this Sir2 is the only that causes deacetylation of p53. We tried SirT1, T3, T4, T5 and T6. We haven’t checked T7 yet. Turner: I would like to ask about side e¡ects of nicotinamide. What e¡ects does this have on cell growth and other cell functions? Presumably you are using it for a fairly short treatment time. Gu: We normally treat the cells for 6 h. The e¡ect we have depends on what types of cell we used. The drug does have wide-ranging side e¡ects. The good thing about this drug is that the cell can tolerate it. It is not like TSA. Pelicci: Do you think that deacetylation might a¡ect DNA binding speci¢city? Gu: That is possible. This is something we are very interested in looking at. Only about 20^30% of p53 in the steady state is acetylated during the DNA damage response. This percentage pretty much correlates with the apoptotic cells. There is a possibility that acetylated p53 speci¢cally binds to a particular type of promoter. If you hyperacetylate p53 you can enhance it, but even if it is not acetylated you can still see it. It is not an absolute requirement. Mahadevan: Did you say that you have managed to detect the acetylated p53 by immunocytochemistry in apoptotic cells? Gu: No, that is just a number correlation.

206

DISCUSSION

Atadja: Is acetylated p53 di¡erentially oligomerized? Gu: No. Zhou: What is the molar ratio between acetylated and non-acetylated p53? Gu: It depends on how much p300 is expressed. If you have a lot of p300 in the cells the acetylated p53 can be enriched. Ott: In your puri¢ed protein, which lysines are predominantly acetylated? Gu: That’s a good question. Everyone talks about di¡erential acetylation activities on 373, 372 or 382 in the literature. We tested this and found that all four lysines are acetylated. Marks: Why do you need both Sir2 and HDAC? Gu: HDAC1 is overexpressed in tumour cells. The associated p53 adaptor protein MTA2, which is a metastasis-associated factor, is overexpressed in tumour cells. Thus, HDAC1 activity might be critical for tumorigenesis. Sir2 has a di¡erent kind of pathway. It is regulated by metabolism. If you look at Sir2a levels, they are almost even. In the cell line we have tested so far, Sir2 expression is quite even between the normal cells versus the tumour cells. Verdin: I can add to this. One other di¡erence is that HDAC1 seems to be pannuclear in terms of its distribution. Sir2a is enriched in the PML body, and so is p53. So is there a possibility that HDAC1 serves as a more global deacetylase, whereas Sir2a is more speci¢c? Gu: The HDAC pathway works mainly through a large protein complex, such as in the case of MTA2. Its biochemical activity is not super strong, but it is abundant and global. Sir2 activity is a¡ected by metabolic rate and Sir2a-knock down seems have no signi¢cant e¡ect on global histone acetylation levels. Thus it is possible that Sir2a may have more speci¢c roles on local promoters as well as non-histone proteins. Pelicci: I don’t think we know enough to think that the PML body can distinguish between the activity of HDAC and Sir2. P53 and HDAC are recruited within the nuclear body and most likely the active p53 then gets out together with HDAC and Sir. Khochbin: Is HAUSP the same protein as USP7? Gu: Yes, they are the same. The correct name is HAUSP. Khochbin: So this protein is interacting with ICP0. Gu: Yes, it is also in the PML body. Khochbin: So this means that during the reactivation of herpes virus, this could be a way to destabilize p53. Gu: Very possibly. Marmorstein: Does your set of four acetylation sites include 320? Gu: No. The antibody we are currently using only picks up the last four acetylation sites. We don’t know whether the lysine 320 is acetylated. Again, this is through p300 overexpression.

REGULATION OF p53 FUNCTION BY ACETYLATION

207

Castronovo: How is phosphorylation regulating this system of acetylation? Gu: We haven’t done too much on the phosphorylation issue. According to the literature, ATM and ATR induce N-terminal serine 15 phosphorylation, which can enhance the binding between p53 and p300. Thus, phosphorylation is playing a positive role in regulating p53 acetylation. Seto: One of the earliest studies shows that HDACs interact with p53 either directly or through Sir3. Have you seen this, or do you think the deacetylasesp53 connection is exclusively through Sir2? Gu: About three years ago we published a study in which we took all the HDAC complexes and put them on a p53 column, looking at which proteins could contact p53. The only protein we ¢shed out was MTA2/PID. Yao: I have a question about Parc. It has been shown in primary cells that in primary cells p53 shuttles actively and spends a lot of time in the cytoplasm. Does Parc play any role in this stage? Gu: It is possible. We are working on this. In heart tissue and muscles Parc is overexpressed. Yao: If you use RNAi for Parc, and p53 translocates to the nucleus, is that population of nuclear p53 activated? Gu: If you transfect p53 into cells without p300, you will still activate the downstream targets. If you accumulate enough of the p53 protein you will see activation. Mahadevan: Stress signalling was originally thought to be due to DNA damage. More recently, it has become clear that a lot of stress signalling actually originates in the cytoplasm. Do you have a feel for how much of what you are seeing is actually directly due to DNA damage, or due to stress acting on components in the cytoplasm? Gu: Every type of stress will induce p53 acetylation. In this sense, there is a possibility that this also happens in the cytoplasm. There may be some acetyltransferases sitting in the cytoplasm. Mahadevan: In signal transduction biochemistry, the key ¢nding is that some signalling e¡ects can be reproduced in enucleated cells. Gu: It’s going to be tough to do that assay.

Regulation of NF-jB action by reversible acetylation Warner C. Greene*{{ and Lin-feng Chen* *Gladstone Institute of Virology and Immunology and the Departments of {Medicine and {Microbiology and Immunology, University of California, San Francisco, San Francisco, CA 94141, USA

Abstract. While the proximal cytoplasmic signalling events controlling the activation of NF-kB are understood in considerable detail, the subsequent intranuclear events that regulate the strength and duration of NF-kB action remain poorly de¢ned. Recently, we have demonstrated that the RelA subunit of the NF-kB heterodimer is subject to reversible acetylation. The p300/CBP acetyltransferases play a major role in the in vivo acetylation of RelA principally targeting lysines 218, 221 and 310 for modi¢cation. Acetylation of these distinct lysine residues regulates di¡erent functions of NF-kB, including transcriptional activation, DNA binding a⁄nity, I-kBa assembly and subcellular localization. Speci¢cally, acetylation of lysine 221 enhances DNA binding and impairs assembly with I-kBa while acetylation of lysine 310 is required for full transcriptional activity of RelA independent of changes in DNA binding or I-kBa binding. In turn, acetylated RelA is deacetylated by histone deacetylase 3 (HDAC3). Deacetylation of lysine 221 promotes high-a⁄nity binding of RelA to newly synthesized I-kBa proteins whose expression is activated by NF-kB. I-kBa binding to deacetylated RelA promotes rapid nuclear export of the NF-kB complex. This export is dependent on CRM1 binding to a nuclear export signal present in I-kBa and promotes replenishment of the cytoplasmic pool of latent NF-kB/I-kBa complexes thus readying the cell for response to the next NF-kB inducing stimulus Together, these studies highlight how reversible acetylation of RelA serves as an intranuclear molecular switch promoting both positive and negative regulatory e¡ects on nuclear NF-kB action. 2004 Reversible protein acetylation. Wiley, Chichester (Novartis Foundation Symposium 259) p 208^222

The inducible NF-kB transcription factor complex plays a central role in regulating the in£ammatory, immune and anti-apoptotic responses in mammals (Baldwin 1996). The prototypical NF-kB complex is a heterodimer of p50 and RelA subunits that is chie£y sequestered in the cytoplasm through its assembly with a family of inhibitory proteins termed the I-kBs (Baldwin 1996). Stimulus-induced phosphorylation of two N-terminal serines in the I-kBs, mediated by the macromolecular I-kB kinase complex (IKK) (Karin 1999), triggers the rapid 208

NF-kB ACETYLATION

209

ubiquitination and subsequent degradation of this inhibitor by the 26S proteasome complex. The liberated NF-kB heterodimer rapidly translocates into the nucleus, where it engages cognate kB enhancer elements and activates gene expression (Baldwin 1996, Ghosh et al 1998). One of the cellular genes induced by NF-kB is I-kBa (Beg et al 1993, Brown et al 1993, Sun et al 1993) thus revealing a potentially novel mechanism for feedback regulation of the NF-kB transcriptional response. Newly synthesized I-kBa proteins shuttle between the cytoplasm and the nucleus and can remove NF-kB from DNA, promoting the return of the now inactive NF-kB^I-kBa complex to the cytoplasm. These events lead to the termination of the NF-kB transcriptional response (Arenzana-Seisdedos et al 1995, 1997). In addition to the recruitment and binding of NF-kB to its DNA enhancer, transcriptional regulation of NF-kB involves the actions of various co-activator and co-repressor proteins. For example, transactivation mediated by NF-kB is greatly increased by co-expression of p300/CBP co-activators (Gerritsen et al 1997, Sheppard et al 1999, Vanden Berghe et al 1999). Moreover, NF-kB physically interacts with both CBP and p300 (Gerritsen et al 1997, Na et al 1998, Sheppard et al 1999). Each of these enzymes contains an intrinsic histone acetyltransferase (HAT) activity. The interplay of RelA and p300 involves an interaction of both the N-terminal Rel homology domain and the C-terminal transcriptional activation domain of RelA with the amino-terminal CH1 domain of p300 (Gerritsen et al 1997, Sheppard et al 1999). SRC-1/N-CoA-1, a nuclear receptor co-activator within the p160 family, also participates as a co-activator with NF-kB (Na et al 1998, Sheppard et al 1999). A second p160 family member, SRC-3, may also exert coactivating e¡ects with NF-kB and interesting its nuclear localization is proposed to be regulated through IKK-mediated phosphorylation of unidenti¢ed phosphoacceptor sites (Werbajh et al 2000, Wu et al 2002). The mechanism by which p300/CBP enhances NF-kB transcriptional activity is likely multifactorial. Both p300 and CBP contain a HAT enzymatic activity that regulates gene expression in part through acetylation of the N-terminal tails of histones (Berger 1999). Acetylated histones are associated with transcriptionally active regions in the genome, while deacetylated histones accumulate in transcriptionally repressed regions (Cheung et al 2000, Imhof et al 1997, Kuo & Allis 1998). p300/CBP also acetylates a variety of transcription factors, including p53, GATA1, MyoD, TFIIEb and E2F (Boyes et al 1998, Gu & Roeder 1997, Imhof et al 1997, Mart|¤ nez-Balba¤ s et al 2000, Sartorelli et al 1999), leading to alterations in various functions, such as DNA binding, transcriptional activity, protein^protein interactions and protein stability (Chen et al 2001a, Mart|¤ nezBalba¤ s et al 2000, Sterner & Berger 2000). The RelA subunit of NF-kB is also subject to acetylation on at least three di¡erent sites. Acetylation of di¡erent lysine residues has now been implicated in the regulation of di¡erent biological

210

GREENE & CHEN

properties of NF-kB including DNA binding, transactivation, and assembly with I-kBa (Chen et al 2001b, 2002). In addition to binding to co-activators, NF-kB also interacts with a family of corepressors and speci¢c histone deacetylases (HDACs) (Ashburner et al 2001, Chen et al 2001a, Zhong et al 2002). These HDACs inhibit NF-kB action by directly modifying the Rel proteins themselves and by targeting the histone tails surrounding various NF-kB-regulated genes. The N-terminal region of RelA subunit of NF-kB selectively interacts with the N-terminal region of HDAC3, leading to the deacetylation of RelA (Chen et al 2001a). The association of NFkB with the HDAC1 and HDAC2 co-repressor proteins blocks expression of NF-kB-regulated genes, including the interleukin 8 (IL8) gene. In contrast to the mechanism underlying the action of HDAC3, HDAC1 and HDAC2 appear to operate at the level of chromatin structure (Ashburner et al 2001, Zhong et al 2002). RelA is acetylated by p300/CBP and deacetylated by HDAC3 Acetylation of RelA was ¢rst demonstrated in vivo with [3H]-acetate radiolabelling assay (Chen et al 2001b). Overexpressed RelA was acetylated in vivo by endogenous HATs. More importantly, endogenous RelA is similarly acetylated after stimulation with tumour necrosis factor (TNF)a or phorbol myristate acetate (PMA), indicating that acetylation occurs under physiological conditions and in a signal-coupled manner (Chen et al 2001b, Kiernan et al 2003). In vivo acetylation of RelA is also detectable with anti-acetylated lysine antibodies when cells are cotransfected with expression vectors encoding a relevant HAT. Both p300 and CBP, but not PCAF, acetylate RelA under these conditions. Not surprisingly, mutation of the acetyltransferase domain within p300 blocks p300 mediated acetylation of RelA (Chen et al 2002). Further, co-expression of either a HAT-de¢cient mutant of p300, or the E1A gene product of adenovirus (an inhibitor of endogenous p300/ CBP activity), blocks the acetylation of RelA induced by endogenous HAT(s). These ¢ndings suggest that p300/CBP plays a key role in the regulation of RelA acetylation in vivo. Although acetylated RelA is detectable in vivo, acetylated RelA is undetectable in vitro in assays utilizing recombinant p300 performed under the same condition in which p53 or histone are acetylated (L. F. Chen, W. C. Greene, unpublished data). Interestingly, RelA can be acetylated in vitro by p300 immunoprecipitated from 293T cells, raising the possibility that an unknown factor or factors co-immunoprecipitated with p300 are necessary for the e⁄cient in vitro RelA acetylation (L. F. Chen, W. C. Greene, unpublished data). Under these conditions, the level of RelA acetylation observed is comparable to that detected with p53. RelA is also subject to deacetylation through a selective action of HDAC3 in vivo and in vitro. Overexpression of HDAC3, but not HDAC1 or 2 deacetylates RelA

NF-kB ACETYLATION

211

and markedly impairs transactivation of kB enhancer-dependent gene expression (Chen et al 2001b, Kiernan et al 2003). Acetylated RelA is also deacetylated by HDAC3 in vitro (Kiernan et al 2003). These inhibitory e¡ects of HDAC3 are blocked by trichostatin A (TSA), a general HDAC inhibitor, indicating that the deacetylase function of HDAC3 is required (Chen et al 2001a, Kiernan et al 2003). Deletion and mutagenesis studies reveal that RelA acetylation occurs at three major sites: lysines 218, 221 and 310. Conservative substitution all three of these lysines with arginine residues (designated as RelA-KR), which preserve overall charge but block acetylation, markedly reduced but did not completely eliminate acetylation detected in [3H]-acetate radiolabelling assays. Thus, other lysine residues with RelA may also be subject to modi¢cation by acetylation. Sequence alignment of all of the mammalian Rel proteins reveals that lysines 218 and 221 are highly conserved, while lysine 310 is uniquely present in RelA. The two conserved lysines within p50 are also be acetylated by p300; however the biological consequences of this post-translational modi¢cation remain unknown (L. F. Chen, W. C. Greene, unpublished data). Acetylation of RelA and regulation of transactivation of NF-jB A potential role for acetylation in the regulation of NF-kB-mediated transactivation emerged with the ¢nding that TSA enhances kB-luciferase reporter gene expression in the presence of TNFa (Chen et al 2001a). TSA also potentiates TNFa-mediated activation of the IL6 promoter and HIV longterminal repeats (Vanden Berghe et al 1999, Quivy et al 2002). NF-kB activation is also enhanced by co-expression of wild-type p300/CBP. However, these e¡ects are not observed when a HAT-de¢cient mutant of p300 is tested (Chen et al 2002, Gerritsen et al 1997, Sheppard et al 1999). Instead, HAT-de¢cient mutant of p300 inhibits TNFa-induced NF-kB activation (Chen et al 2002). In addition, RelA-KR fails to display enhanced transcriptional activity when co-activator p300/CBP is coexpressed, suggesting that p300/CBP enhances RelA-mediated transcription, at least in part, by promoting direct acetylation of this subunit of NF-kB. Conversely, expression of HDAC3, but not of HDAC1, 2, 4, 5 and 6, inhibits TNFa-mediated activation of NF-kB. These inhibitory e¡ects of HDAC3 are not observed in the presence of TSA, indicating that the deacetylase activity of HDAC3 is required for the observed biological e¡ects (Chen et al 2001a). Mutagenesis studies suggest that acetylation of lysines 221 and 310 is important in regulating the overall transcriptional activity of NF-kB (Chen et al 2002). When the transcriptional activity of each single lysine-to-arginine mutant is analysed in a luciferase reporter assay, both the K310R and RelA-KR mutants display greatly impaired transactivation function, whereas the K221R mutant exhibits intermediate activity. These ¢ndings raise the possibility that acetylation at lysine

212

GREENE & CHEN

FIG. 1. Acetylation of di¡erent lysine residues regulates distinct functions of RelA. Acetylation of lysine 221 increases DNA binding a⁄nity for the kB enhancer and prevents the association of RelA with I-kBa; acetylation of lysine 310 likely controls the association of an as yet unknown factor that is required for full transactivation by RelA.

310, and to a lesser extent at lysine 221, is required for the full transactivation function of RelA (Fig. 1). As discussed below, acetylation of lysine 221 is also centrally involved in regulating the DNA binding and I-kBa assembly properties of RelA. Changes in these activities likely contribute to the observed decrease in RelA transcriptional activity. In contrast, mutation of lysine 310 to arginine does not alter DNA binding or I-kBa assembly but markedly inhibits transactivation. We speculate that lysine 310 may form a platform for recruitment of another yet unidenti¢ed factor that contributes to the RelA transcriptional response. Does this factor contain a bromodomain? This is an intriguing question, because bromodomain-containing proteins speci¢cally interact with acetylated lysine residues (Dhalluin et al 1999, Polesskaya et al 2001). Acetylation of RelA and regulation of DNA binding activity of NF-jB Blockade of HDAC activity by addition of TSA leads to the enhancement of TNFa-induced NF-kB DNA binding, suggesting that acetylation of RelA may regulate the DNA binding properties of NF-kB (Chen et al 2001a, Quivy et al 2002). Analysis of each lysine-to-arginine RelA mutant further supports the role of RelA acetylation in the regulation of DNA binding activity of NF-kB. With regard to RelA homodimers, mutation of the lysine 221 produces a sharp decline in overall DNA binding activity. Whereas, mutation of the two other major acetylation sites in RelA, lysines 218 and 310, does not (Chen et al 2002). When these RelA mutants are tested in the context of p50/RelA NF-kB complexes,

NF-kB ACETYLATION

213

steady-state DNA binding activity measured in electrophorectic mobility shift assays (EMSAs) is not markedly changed. However, when unlabelled kB enhancer DNA is added as a competitor, the o¡-rate of DNA binding to the kB enhancer is accelerated when acetylation is blocked either by introduction of the arginine-for-lysine mutation at residue 221 or by co-expression of the dominantnegative p300 HAT-de¢cient mutant. When biotinylated forms of the kB enhancer were used, acetylated forms of RelA were identi¢ed by their direct binding to kB enhancer DNA (Chen et al 2002). Together, these results suggest that the acetylation of RelA at lysine 221 enhances the binding a⁄nity of the NF-kB complex for the kB enhancer (Fig. 1). Inspection of the crystal structure of the RelA-p50-DNA complex revealed that lysine 221 participates in direct contact with the DNA backbone (Chen et al 1998). Thus, acetylation of lysine 221 may result in a conformational change in the protein that enhances binding to the kB enhancer (Chen et al 2002).

Acetylation of RelA and regulation of assembly with I-jBa Acetylation of RelA also plays a key role in regulating its assembly with the I-kBa inhibitor. In GST^I-kBa pull-down assays, the binding of RelA to GST^I-kBa is markedly diminished by p300-mediated acetylation of RelA, but is increased by coexpression of HDAC3, which promotes deacetylation of RelA. (Chen et al 2001b). Consistent with this ¢nding, wild-type p300, but not the p300 (HAT
) mutant, blocks the interaction of RelA with I-kBa in a mammalian one-hybrid system. Studies of the various lysine-to-arginine substitution mutants of RelA reveals that lysine 221 plays a key role in regulating I-kBa assembly (Chen et al 2002). In further support of this conclusion, the addition of recombinant GST^I-kBa signi¢cantly diminishes DNA binding with the mutant of lysine 221. Dual arginine substitution of lysine 218 and 221 or all three sites (RelA-KR) produces even greater levels of inhibition of DNA binding in the presence of GST^I-kBa (Chen et al 2002). These ¢ndings support the notion that acetylation of RelA at lysine 221, alone or in combination with lysine 218, impairs assembly with I-kBa (Fig. 1). In the crystal structure of RelA-p50-I-kBa (Huxford et al 1998, Jacobs & Harrison 1998), lysine 221 directly interacts with methionine 279 in the sixth ankyrin repeat of I-kBa. Acetylation of lysine 221 may, thus, result in a conformational change in RelA that both increases its a⁄nity for the kB enhancer and disables its interaction with I-kBa (Fig. 1). Alternatively, the more rapid o¡rate of kB enhancer DNA binding displayed by RelA K221R may derive in part from its greater a⁄nity for I-kBa.

214

GREENE & CHEN

Acetylation of RelA and regulation of nuclear export of NF-jB When overexpressed in HeLa cells, GFP^RelA fusion proteins principally localize in the nuclear compartment. However, when co-expressed with HDAC3, they are found exclusively in the cytoplasm. HDAC3 expression induces the translocation of RelA to the cytoplasm. This e¡ect is due to the deacetylation of RelA, which in turn promotes I-kBa assembly and I-kBa-dependent nuclear export of RelA. Consistent with this model, HDAC3 fails to promote RelA translocation into the cytoplasm in the I-kBa de¢cient MEFs (Chen et al 2001b). In addition, a GFP^ RelA fusion protein containing lysine-to-arginine substitutions at the three major acetylation sites (GFP-RelA-KR) principally localizes in the cytoplasm. However, when expressed in the I-kBa-de¢cient MEFs, the GFP-RelA-KR mutant protein appears predominantly within the nucleus. Further, the cytoplasmic expression pattern of RelA-KR mutant is dependent on the presence of the nuclear export signal within I-kBa. Expression of a mutant I-kBa, which lacks this nuclear export signal, produces a nuclear predominant pattern of expression for the RelA-KR in I-kBa de¢cient MEF cells. In contrast, wild-type I-kBa promotes e⁄cient translocation of GFP-RelA-KR to the cytoplasm in these cells (Chen et al 2002). These ¢ndings are consistent with the hypothesis that when lysine 221 in the RelA-KR mutant is deacetylated, the interaction with I-kBa is enhanced and the hypoacetylated RelA is exported from the nucleus in a manner that is principally dependent on the nuclear export signal IkBa.

Conclusions and future perspectives These studies revealed key regulatory roles played by acetylation in the biology of NF-kB. This post-translational modi¢cation a¡ects many di¡erent biological properties of the RelA subunit of NF-kB including DNA binding, transcriptional activity, assembly with the I-kB inhibitors, and subcellular localization of the RelA-containing complexes. Acetylation of lysine 310 of RelA is required for full transactivation by the NF-kB complex, most likely by recruiting an as-yet unidenti¢ed cofactor. Acetylation of lysine 221 enhances the binding of RelA to the kB enhancer, while acetylation of lysine 221, alone or in combination with lysine 218, impairs the assembly of RelA with I-kBa. Lysines 218 and 221 are highly conserved within all Rel family members, including Dorsal from Drosophila. The possibility that these evolutionarily conserved lysine residues participate as targets for reversible acetylation and contribute to the regulation of the biological functions of other Rel factors remains an intriguing but unanswered question. Both p300 and CBP, but not PCAF, can mediate the acetylation of RelA in vivo. These acetylated forms of RelA are deacetylated through speci¢c interaction with

NF-kB ACETYLATION

215

FIG. 2. Schematic model for the role of HDAC3-mediated deacetylation of RelA as an intranuclear molecular switch promoting I-kBa binding and I-kBa-dependent nuclear export of the NF-kB complex. This deacetylation-controlled response both leads to the termination of the NF-kB transcriptional response and aids in re-establishing latent cytoplasmic forms of NF-kB bound to I-kBa, thereby preparing the cell to respond to the next NF-kB-inducing stimulus.

HDAC3. Deacetylation of RelA promotes its e¡ective binding to I-kBa and leads to rapid I-kBa-dependent nuclear export of the NF-kB complex by a CRM-1dependent pathway. Deacetylation of RelA thus functions as an intranuclear molecular switch that both shortens the duration of the NF-kB transcriptional response and contributes to replenishment of the depleted cytoplasmic pool of latent NF-kB/I-kBa complexes, thereby readying the cell for the NF-kB-inducing signal (Fig. 2). How the interplay between RelA and HDAC3 is regulated however remains unknown. Future studies will help to further de¢ne the relationship between these intriguing reactions and how they contribute to the overall regulation of NF-kB action. In addition to acetylation, NF-kB can undergo other post-translational modi¢cations, such as phosphorylation. Phosphorylation at serine 276 by PKA enhances the binding of p300/CBP (Zhong et al 1998). RelA is also phosphorylated at multiple sites by di¡erent kinases within its C-terminal transactivation domain, a region that participates in the interaction with p300/ CBP. Thus, acetylation of RelA may be contingent on prior phosphorylation of RelA. Phosphorylation at one or more of these sites could thus trigger the recruitment of p300/CBP, leading to enhanced RelA acetylation. Precedence for

216

GREENE & CHEN

such a scenario is found with tumour suppressor p53 (Liu et al 1999, Sakaguchi et al 1998) and retinoblastoma tumour suppressor protein (Chan et al 2001).

Acknowledgements This work was supported by funds from the J. David Gladstone Institutes, P¢zer, Inc., and bene¢ted from core facilities provided through the UCSF-GIVI Center for AIDS Research (National Institutes of Health Grant P30 MH59037). The authors thank R. Givens for manuscript preparation and J. Carroll and C. Goodfellow for graphics assistance.

References Arenzana-Seisdedos F, Thompson J, Rodriguez MS, Bachelerie F, Thomas D, Hay RT 1995 Inducible nuclear expression of newly synthesized IkBa negatively regulates DNA-binding and transcriptional activities of NF-kB. Mol Cell Biol 15:2689^2696 Arenzana-Seisdedos F, Turpin P, Rodriguez M et al 1997 Nuclear localization of IkBa promotes active transport of NF-kB from the nucleus to the cytoplasm. J Cell Sci 110:369^378 Ashburner BP, Westerheide SD, Baldwin AS Jr 2001 The p65 (RelA) subunit of NF-kB interacts with the histone deacetylase (HDAC) corepressors HDAC1 and HDAC2 to negatively regulate gene expression. Mol Cell Biol 21:7065^7077 Baldwin AS Jr 1996 The NF-kB and IkB proteins: new discoveries and insights. Annu Rev Immunol 14:649^683 Beg AA, Finco TS, Nantermet PV, Baldwin AS Jr 1993 Tumor necrosis factor and interleukin-1 lead to phosphorylation and loss of IkBa: a mechanism for NF-kB activation. Mol Cell Biol 13:3301^3310 Berger SL 1999 Gene activation by histone and factor acetyltransferases. Curr Opin Cell Biol 11:336^341 Boyes J, By¢eld P, Nakatani Y, Ogryzko V 1998 Regulation of activity of the transcription factor GATA-1 by acetylation. Nature 396:594^598 Brown K, Park S, Kanno T, Franzoso G, Siebenlist U 1993 Mutual regulation of the transcriptional activator NF-kB and its inhibitor, IkBa. Proc Natl Acad Sci USA 90:2532^ 2536 Chan HM, Krstic-Demonacos M, Smith L, Demonacos C, La Thangue NB 2001 Acetylation control of the retinoblastoma tumour-suppressor protein. Nat Cell Biol 3:667^674 Chen FE, Huang DB, Chen YQ, Ghosh G 1998 Crystal structure of p50/p65 heterodimer of transcription factor NF-kB bound to DNA. Nature 391:410^413 Chen H, Tini M, Evans RM 2001a HATs on and beyond chromatin. Curr Opin Cell Biol 13:218^ 224 Chen L, Fischle W, Verdin E, Greene WC 2001b Duration of nuclear NF-kB action regulated by reversible acetylation. Science 293:1653^1657 Chen LF, Mu Y, Greene WC 2002 Acetylation of RelA at discrete sites regulates distinct nuclear functions of NF-kB. EMBO J 21:6539^6548 Cheung WL, Briggs SD, Allis CD 2000 Acetylation and chromosomal functions. Curr Opin Cell Biol 12:326^333 Dhalluin C, Carlson JE, Zeng L, He C, Aggarwal AK, Zhou MM 1999 Structure and ligand of a histone acetyltransferase bromodomain. Nature 399:491^496 Gerritsen ME, Williams AJ, Neish AS, Moore S, Shi Y, Collins T 1997 CREB-binding protein/ p300 are transcriptional coactivators of p65. Proc Natl Acad Sci USA 94:2927^2932

NF-kB ACETYLATION

217

Ghosh S, May MJ, Kopp EB 1998 NF-kB and Rel proteins: evolutionarily conserved mediators of immune responses. Annu Rev Immunol 16:225^260 Gu W, Roeder RG 1997 Activation of p53 sequence-speci¢c DNA binding by acetylation of the p53 C-terminal domain. Cell 90:595^606 Huxford T, Huang DB, Malek S, Ghosh G 1998 The crystal structure of the IkBa/NF-kB complex reveals mechanisms of NF-kB inactivation. Cell 95:759^770 Imhof A, Yang XJ, Ogryzko VV, Nakatani Y, Wol¡e AP, Ge H 1997 Acetylation of general transcription factors by histone acetyltransferases. Curr Biol 7:689^692 Jacobs MD, Harrison SC 1998 Structure of an IkBa/NF-kB complex. Cell 95:749^758 Karin M 1999 How NF-kB is activated: the role of the IkB kinase (IKK) complex. Oncogene 18: 6867^6874 Kiernan R, Bres V, Ng RW et al 2003 Post-activation turn-o¡ of NF-kB-dependent transcription is regulated by acetylation of p65. J Biol Chem 278:2758^2766 Kuo MH, Allis CD 1998 Roles of histone acetyltransferases and deacetylases in gene regulation. Bioessays 20:615^626 Liu L, Scolnick DM, Trievel RC et al 1999 p53 sites acetylated in vitro by PCAF and p300 are acetylated in vivo in response to DNA damage. Mol Cell Biol 19:1202^1209 Mart|¤ nez-Balba¤ s MA, Bauer UM, Nielsen SJ, Brehm A, Kouzarides T 2000 Regulation of E2F1 activity by acetylation. EMBO J 19:662^671 Na SY, Lee SK, Han SJ, Choi HS, Im SY, Lee JW 1998 Steroid receptor coactivator-1 interacts with the p50 subunit and coactivates nuclear factor kB-mediated transactivations. J Biol Chem 273:10831^10834 Polesskaya A, Naguibneva I, Duquet A, Bengal E, Robin P, Harel-Bellan A 2001 Interaction between acetylated MyoD and the bromodomain of CBP and/or p300. Mol Cell Biol 21:5312^ 5320 Quivy V, Adam E, Collette Y et al 2002 Synergistic activation of human immunode¢ciency virus type 1 promoter activity by NF-kB and inhibitors of deacetylases: potential perspectives for the development of therapeutic strategies. J Virol 76:11091^11103 Sakaguchi K, Herrera JE, Saito S et al 1998 DNA damage activates p53 through a phosphorylation-acetylation cascade. Genes Dev 12:2831^2841 Sartorelli V, Puri PL, Hamamori Y et al 1999 Acetylation of MyoD directed by PCAF is necessary for the execution of the muscle program. Mol Cell 4:725^734 Sheppard KA, Rose DW, Haque ZK et al 1999 Transcriptional activation by NF-kB requires multiple coactivators. Mol Cell Biol 19:6367^6378 Sterner DE, Berger SL 2000 Acetylation of histones and transcription-related factors. Microbiol Mol Biol Rev 64:435^459 Sun SC, Ganchi PA, Ballard DW, Greene WC 1993 NF-kB controls expression of inhibitor IkBa: evidence for an inducible autoregulatory pathway. Science 259:1912^1915 Vanden Berghe W, De Bosscher K, Boone E, Plaisance S, Haegeman G 1999 The nuclear factorkB engages CBP/p300 and histone acetyltransferase activity for transcriptional activation of the interleukin-6 gene promoter. J Biol Chem 274:32091^32098 Werbajh S, Nojek I, Lanz R, Costas MA 2000 RAC-3 is a NF-kB coactivator. FEBS Lett 485:195^199 Wu RC, Qin J, Hashimoto Y et al 2002 Regulation of SRC-3 (pCIP/ACTR/AIB-1/RAC-3/ TRAM-1) coactivator activity by IkB kinase. Mol Cell Biol 22:3549^3561 Zhong H, Voll RE, Ghosh S 1998 Phosphorylation of NF-kB p65 by PKA stimulates transcriptional activity by promoting a novel bivalent interaction with the coactivator CBP/ p300. Mol Cell 1:661^671 Zhong H, May MJ, Jimi E, Ghosh S 2002 The phosphorylation status of nuclear NF-kB determines its association with CBP/p300 or HDAC-1. Mol Cell 9:625^636

218

DISCUSSION

DISCUSSION Ott: Is HDAC3 itself up-regulated by NF-kB? Greene: No, we have no evidence that HDAC3 is induced by NF-kB. Seto: There are reports that NF-kB is deacetylated by HDAC1 and HDAC2, but not by HDAC3. How do you reconcile your results with these other ¢ndings? Greene: The study that you refer to was performed with a kB reporter gene stably integrated into a chromosome and thus was subject to regulation by the organization of the surrounding chromatin. We believe it is very likely that other HDACs besides HDAC3 exert regulatory functions in the context of such target genes. However, in terms of the deacetylation of acetylated forms of RelA, HDAC3 is the principal mediator of this reaction. Seto: Does acetylation of NF-kB always occur in the nucleus? Greene: The only way we have currently of looking at that is through the use of our antibody that speci¢cally reacts with acetylated lysine 310. Thus far, we have not detected acetylated RelA in the cytoplasm and thus feel it is very likely that RelA deacetylation is a nuclear event. This also ¢ts with the role of deacetylation of RelA at lysine 221 in controlling the assembly with I-kBa and the subsequent nuclear export of RelA in an I-kBa-dependent manner. Atadja: How relevant is your cell system to the function of HDAC? What cell types are you using? Greene: While many of our studies have involved transfection of RelA expression vectors with the attendant concerns regarding over-expression, we have also examined acetylation/deacetylation of endogenous RelA protein in HeLa cells and 293 cells. Thus far, we have not extended our studies of endogenous RelA to cells in the haemato-lymphoid system. One interesting new line of investigation involves the reconstitution of RelA
/
murine embryo ¢broblasts with the various RelA acetylation mutants to assess potential changes in RelA function. Marks: In your co-transfection experiments with HDAC1 and 3, what cells were these? Presumably they have endogenous HDACs 1^3. Greene: Yes. These studies were performed in HeLa cells which contain endogenous HDACs. Marks: You are essentially getting superexpression of HDAC3 or HDAC1 in the transfected cells. Greene: Yes. These are studies where we are directing overexpression of HDAC3. Marks: In our experience we ¢nd that the HDAC1 complexes with HDAC2 and sometimes HDAC3. Have you seen this? Greene: The key point here is what is in the complex. We don’t think that HDAC3 is acting entirely alone, but more likely in a larger complex perhaps

NF-kB ACETYLATION

219

containing such proteins as SMRT/N-CoR. In terms of other HDACs in the complex, we only know that overexpression of HDACs 1, 2, 4, 5 and 6 did not recapitulate the results obtained with HDAC3. Marks: What is the endogenous level of HDAC3 relative to, say, HDAC1 before you transfect these cells? Greene: They are fairly comparable, but low compared with what we achieve by transfection. Turner: Did you test the catalytically de¢cient HDAC3 in this assay? Is the catalytic activity required? Chen: The catalytic activity of HDAC3 is required. Olson: As you know, most of the components of the NF-kB signalling system are conserved in Drosophila. Is there any evidence that HDAC is genetically required for Dorsal activity in £ies? There has been a lot of experience recently of treating £ies with TSA. Does this evoke a Dorsal-like phenotype? Greene: This is an open question and certainly a potentially fertile area of future research. Zhou: You showed three lysines that are acetylated by p300. Do you have any data on the order of acetylation events? Greene: Initially we were intrigued by the notion that acetylation of Lys310 preceded acetylation of lysines 221 and 218. But later studies have not con¢rmed this notion. For example, we can mutate Lys310 and still observe acetylation of 218 and 221 in vivo using tritiated acetate labelling. Thus, based on the sensitivity of the current assays, we have no clear evidence to support sequential acetylation. Cole: Along those lines it looked like in each of the single mutants where you showed the tritiated acetylation that you got much more than a third of the acetylation diminished, even though there are three sites. I would have predicted that mutating one would only result in a 35% drop o¡ or so, and you were getting way over that for each of the mutants. Can you explain this? Greene: The data that you refer to were obtained using anti-acetyl-lysine antibodies for immunoblotting. As such, this antibody may not equivalently recognize each of the three acetylated lysines and thus the degree of drop in acetylation cannot be equated with interdependence. We have also repeated these types of studies employing [3H]-acetate labelling which should of course eliminate antibody bias in the recognition of the acetylated lysine sites. As you know, this is technically a more di⁄cult experiment. What I can say is that no single mutation removes all of the acetylation of RelA. However, single mutations appear to diminish acetylation by more than 33% which could be an argument for interdependence, possibly related to conformational changes in the protein. I would also note that the triple lysine mutant (KKK218/221/310RRR) eliminates 90% of the acetylation but some residual acetylation is observed. This result raises

220

DISCUSSION

the possibility that the three lysines we have identi¢ed as major sites of acetylation may not be exclusive sites. Moazed: I am beginning to wonder whether every transcription factor will turn out to be acetylated. Have you looked in cells labelled with tritiated acetate at how many proteins get labelled? Greene: We haven’t done this experiment. However, I suspect that acetylation/ deacetylation will likely play an important role in regulating a number of transcription factors. Allis: I can’t get anyone todo that any more inmylab. Ifyoudo that inan organism such as Tetrahymena, the histones are incredible sinks. But if you let the ¢lm expose it is amazing. No one will do the in vivo labelling: these are nasty experiments. If you just pop out nuclei from a mammalian cell and cook that as a nuclear prep in acetyl-CoA, this is a very robust ‘on’ system for whatever you have down at the bottom of the pellet. Most of those speci¢c activities are up at least 20-fold. If you then have an antibody, that is a great hook to come in. We call this ‘in nucleo’ labelling and people do this in my lab. This is a potentially powerful technology. Turner: I think this is a better way to do it. This started when we had a medical undergraduate in the lab. If you check your metabolic pathways, there isn’t any obvious route from tritiated acetate to acetyl-CoA to acetylated proteins. It is a complicated route  it is not a nice neat labelling system. What you are proposing is much more biochemically robust. Gu: You cannot label them for too long. Greene: We label for 30 min in the presence of cycloheximide to prevent migration of the 3H-label to proteins during their synthesis. Gu: One hour is OK. Allis: I am talking about a short 15 min exposure. Verdin: I have a question for the whole group about the analysis of all acetylated proteins in the proteome. How would one go about this? Could the approaches we have outlined be a way to get at this? Allis: It is a start. It is very complex. Verdin: We have thought about doing this with some of the antibodies. For example, some newer anti-acetyl-lysine antibodies recognize a large number of proteins. However, our preliminary experiments are discouraging. Most of the bands don’t respond either to TSA or nicotinamide. This would mean that they are recognizing some other epitopes which are not acetylated. Allis: We have done some experiments with methyl antibodies. There we don’t have the TSA double check. But in those cases you run a total nuclear extract, blot it with your favorite histones, and it comes up like a rocket. But there are always a couple of ‘mystery bands’. They are probably invisible high molecular weight proteins. We have now puri¢ed a couple, and they have perfect histone modules just sitting there in the middle of the protein. Some of these are

NF-kB ACETYLATION

221

very sexy cancer-related proteins with a little histone motif that I would bet is then modi¢ed with these covalent marks. Cole: Why don’t people use [14C]-acetate since it is a stronger emitter? Allis: For us it has just been money. Castronovo: Would it be possible to screen the protein which has the sequence motif, and pull out all the ones that have a consensus sequence? Allis: Youcanblastawayandyougetlotsofhits.Pluginany6-merandyouwillget lots of hits just by chance. However, the antibody adds another level of believability. Khochbin: What is limiting in cells is actually acetyl-CoA. We were thinking of expressing acetyl-CoA synthase in cells. Turner: Have you tried this? Khochbin: No. We were thinking of it. Turner: There has to be a better way than just piling in vast amounts of tritiated acetate. Marks: It is di⁄cult to analyse. You get a lot of bands. Allis: Plus you have to make sure it doesn’t get into the backbone of the proteins. Turner: You need to add cycloheximide. Atadja: Are people thinking about acetylation and interaction with bromodomains as another way of transducing signals, as happens with kinases and SH2 domains? What do people think of this? Stuart Schreiber alluded to this in his review when he talked about histone acetylation. Verdin: Warner Greene just showed us a nice example where the acetylation is stimulus-coupled. As far as I know this is one of the ¢rst examples of stimuluscoupled acetylation. This is one area where there is a huge gap in our knowledge. Cohen: It is also a ligase by itself. I don’t know how we put all these functions together. Ott: Warner, you showed everything with TNFa stimulation. Do you see any di¡erences if you activate the cells with anything else? Greene: We have used other stimuli including phorbol esters and the results are similar to that we obtain with TNFa. We are exploring stimulation with HTLV-I Tax but this inducer of sustained nuclear NF-kB expression gives a rather di¡erent pattern that could be interesting. Cohen: Have you used HDAC RNAi to block endogenous HDAC in your system? Greene: These studies are now underway. Cohen: This will give you the answer to the speci¢city question. Greene: I am very interested in the tritiated acetate issue. When we entered the ¢eld this was the gold standard. We only labelled for 30 min and we added cycloheximide to prevent migration of the label. But these are horrible experiments so we were delighted when we found an anti-acetyl-lysine antibody that allowed us to track acetylation by immunoblotting. We spent a lot of money

222

DISCUSSION

making our own site-speci¢c antibodies. I hope we never have to do another tritiated acetate experiment. Turner: As David Allis has said, if you do the radiolabelling you ¢nd new things. You are not going to identify new targets with an antibody. Castronovo: Is it not possible to generate more promiscuous antibodies? Allis: People are trying hard. Verdin: Has anyone done a systematic analysis of these antibodies? What is the number of amino acids they tend to recognize? Allis: It would be worth looking into. Berger: Thomas Jenuwein, would you comment on this because of the K9/K27 methylation issue and the similarity of these peptide sequences around the methylated residues? I was wondering about the number of amino acids that are recognized by the antibodies. Jenuwein: One can compare the length of peptides and also modulate the concentration of the methylated residues. We have done a semi-comparison of a range of antibodies raised against the H3-K9 methyl position. The branched antibody that we generated ¢rst and which is not so speci¢c for a given lysine position recognizes methyl when presented at high concentrations. In fact, we have done blots looking at nuclear extracts lacking the Suv39h enzymes and also overexpressing other HMTases. We can pick up some bands. It is going to be a limited approach and you have to try several antibodies, but it is a possible way of going to detect some methylated non-histone targets. Mahadevan: The other issue relevant here is the physical size of an IgG and whether the epitopes are brought together or not. Considering the structure of IgG, the size of what it can recognize is not more than 9 amino acid residues. Turner: I think it is less than that. There is a lot of work on using synthetic peptides and the ¢gure is usually just 4 or 5 residues. But the other thing is that if you have a peptide, it is possible that its conformation will be altered by distant residues causing it to fold in a certain way. Berger: If that were true, that it’s only 4 or 5 in the recognition, then you wouldn’t ever be able to generate an antibody that could detect methylation of Lys9 vs. Lys27. Turner: That is why it has been so problematic. Berger: So what do you need? Jenuwein: You have to use a longer peptide. Berger: I have talked to David extensively over the years about the optimal size of the epitope. Shorter peptides are sometimes better because then the background is reduced. Jenuwein: It is a trade-o¡. Allis: We haven’t gone shorter than 6 or 7.

General discussion II p300 and DNA repair Hottiger: I would like to try to address the question Eric Verdin raised as to whether there is a bias in ¢nding acetylated proteins of other cellular processes due to the fact that we are all working in the ¢eld of chromatin remodelling and/ or transcription. Some years ago we were surprised to read that p300 primary knockout MEFs taken into culture would stop to proliferate after three or four rounds of replication. Our initial hypothesis was to investigate whether p300 has a role in synthesis and/or genome maintenance. We pulse-labelled HeLa cells with thymidine. Simultaneously we UV-treated the cells, and subsequently performed a ChIP analysis by immunoprecipitating p300 and asking whether p300 was associated with labelled DNA while loading precipitated DNA on SDS-PAGE and exposing the gel to an X-ray ¢lm. We observed that there was a substantial amount of labelled DNA associated with p300. This was not the case when we either did not treat the cells with UV or when we treated the cells with FCS after having starved them for several hours to induce transcription of immediate early genes. Subsequently we started with an approach by immunoprecipitating p300 from non-synchronized cells and investigating complex formation of p300 with di¡erent proteins described earlier to be involved in DNA synthesis. We indeed found di¡erent proteins and have published a few of them so far. One is proliferating cell nuclear antigen (PCNA), which formed a complex with p300 throughout the cell cycle. PCNA was described as auxiliary factor of di¡erent DNA polymerases and is interacting with a large number of proteins (Hasan et al 2001a). PCNA/p300 complex formation was not S-phase speci¢c. In the same publication we identi¢ed XPA as binding to p300, which is involved in nucleotide excision repair pathway. Fen1 and polymerase b were both found to bind p300 and are involved the base excision repair pathway (Hasan et al 2001b, 2002). Some thoughts about the studies we have done. Fen1 was found to be acetylated in vivo using an anti-acetyl lysine antibody. We saw an increase in acetylation when we co-transfected p300 and Fen1. Using a deletion mutant of Fen1 which we know no longer interacts with p300 we saw that there is no Fen1 acetylation in vivo. When we UV treat these cells, we observed a tremendous increase in acetylation of Fen1, indicating to us that UV treatment a¡ects the equilibrium between histone acetyltransferases and deacetylases. We also performed HAT assays in vitro using HAT domains of p300, CBP and PCAF. They are all able to acetylate histones. 223

224

GENERAL DISCUSSION II

PCNA was not acetylated at all. However, we saw a striking di¡erence between Fen1 acetylation induced by p300, CBP or PCAF. Despite the fact that the expressed HAT domain from CBP was larger than the one expressed for p300, CBP was not able to acetylate Fen1. We didn’t publish these results because when we repeated the same experiment with full-length p300 and CBP proteins we didn’t see any di¡erence between p300 and CBP, but we lost Fen1 acetylation with PCAF. This strongly argues that many published experiments using only the HAT domains of these proteins have to be taken very cautiously. Furthermore, we investigated which lysine residues of Fen1 are acetylated. Mass spectrometry analysis revealed that the C-terminal stretch of Fen1 which also mediated the interaction with p300, harboured four lysine residues that were acetylated. Fen1 C-terminus contains a stretch full of lysine residues, but we don’t know so far what discriminates between the acetylation of one lysine over another. At this moment we don’t really understand why some lysine residues and not others are acetylated by p300. Acetylation of Fen1 inhibited its DNA binding activity, in comparison with transcription factors which are often enhanced. The analysis revealed that we can get up to 100% of a speci¢c lysine being acetylated in vitro. However, the percentage and distribution between these lysine residues di¡ers. We found di¡erent populations of acetylated Fen1, some of which had four lysines acetylated, one which had only two, and others with just one. To investigate whether DNA polymerase b is acetylated in vivo we couldn’t use the anti-lysine antibody. It wouldn’t work. In this case we had to perform an in vivo labelling experiment. DNA polymerase b could only be labelled when it was overexpressed because endogenous levels are so low. We quantitated the labelled amount to be approximately 16% of the overexpressed Pol b. Additionally, we saw that acetylation of Pol b in vitro inhibited one of its enzymatic activities (lyase activity). It was shown earlier by site-directed mutagenesis that lysine 72 is very important for this enzymatic activity. Using p300 and di¡erent mutants of Pol b we showed that indeed lysine residue 72 was acetylated. Taken together, these results indicate that p300 may not only be involved in transcription regulation by modifying transcription factors and chromatin but also DNA repair synthesis. We have showed that p300 a¡ects Pol b by acetylating it and inhibiting its enzymatic activity. Furthermore p300 interacted with PCNA and Fen1, and inhibited Fen1’s activity. Thus we think that p300 is also playing a major role by in£uencing these cellular pathways. Berger: Is there a di¡erence between p300 and CBP? Hottiger: Not when you use full-length protein. There is a di¡erence between acetylating proteins with just the HAT domain of p300 or full-length p300. The acetylation patterns of a given target protein induced by the HAT domain or the full-length p300 and subsequently analysed by mass spectrometry are quite di¡erent, not for all proteins we analysed but for some. Additionally, if you take

GENERAL DISCUSSION II

225

target proteins and express them in bacteria or with the baculovirus system, the pattern of acetylation is sometimes also di¡erent. It really depends on where you take your proteins from and what you acetylate them with. Moazed: Is there any evidence that HDAC inhibitors a¡ect DNA replication? Hottiger: We haven’t looked for this in close detail. The problem we have is that there aren’t any really good experiments looking at repair in the whole-cell context. We identi¢ed a HDAC that interacts with Fen1 biochemically and in the cell by colocalization. In addition, it was described earlier that treatment of one-cell mouse embryos with inhibitors accelerated the completion of replication indication that also in this case acetylation has a regulatory role (Aoki & Schultz 1999). Gu: Has anyone else shown that p300 is involved in DNA repair? Hottiger: There is another publication showing that TDG is also interacting with p300 (Tini et al 2002). The authors distinguished between the involvement of TDG in transcription and repair. There are other reports coming out showing that HATs regulate origin of replication (e.g. Datta et al 2001). One of the ideas we had originally was that the cell has to modify the chromatin for both transcription and repair/replication, so why not use the same factors? References Aoki E, Schultz RM 1999 DNA replication in the 1-cell mouse embryo: stimulatory e¡ect of histone acetylation. Zygote 7:165^172 Datta A, Bagchi S, Nag A et al 2001 The p48 subunit of the damaged-DNA binding protein DDB associates with the CBP/p300 family of histone acetyltransferase. Mutat Res 486:89^97 Hasan S, Hassa PO, Imhof R, Hottiger MO 2001 Transcription coactivator p300 binds PCNA and may have a role in DNA repair synthesis. Nature 410:387^391 Hasan S, Stucki M, Hassa PO et al 2001b Regulation of human £ap endonuclease-1 activity by acetylation through the transcriptional coactivator p300. Mol Cell 7:1221^1231 Hasan S, El Andaloussi N, Hardeland U et al 2002 Acetylation regulates the DNA end-trimming activity of DNA polymerase beta. Mol Cell 10:1213^1222 Tini M, Benecke A, Um SJ, Torchia J, Evans RM, Chambon P 2002 Association of CBP/p300 acetylase and thymine DNA glycosylase links DNA repair and transcription. Mol Cell 9:265^ 277

Reversal of gene silencing as a therapeutic target for cancer
roles for DNA methylation and its interdigitation with chromatin Stephen B. Baylin The Sidney Kimmel Comprehensive Cancer Center at Johns Hopkins, Room 544, The Cancer Research Center, 1650 E. Orleans, Baltimore, MD 21231, USA

Abstract. It has become apparent over the past several years that one of the attractive emerging prevention and therapy targets for cancer is the reversal of aberrant gene silencing mediated by epigenetic events associated with transcriptional repression. Integral to the possibilities for this targeting is the need to dissect the molecular mechanisms which underlie these transcriptional changes. At present, the best studied of the epigenetically silenced genes involved in cancer are those which harbour aberrant DNA promoter region methylation. This growing list includes almost half of all proven tumour suppressor genes and also a rapidly expanding list of genes with candidate roles for antitumour activities as well. Thus, one approach receiving much attention for restoring expression of abnormally silenced cancer genes for therapeutic purposes is utilization of agents such as 5-azacytidine (5-AzaC) and 5’-deoxy-azacytidine (DAC), which inhibit the DNA methyltransferases (DNMTs) that catalyse DNA methylation. Other approaches, and particularly the notion of utilizing inhibitors of histone de-acetylation, are being suggested by the exploding body of data concerning the role of histone modi¢cations in mediating gene expression status, and especially those parameters that participate in gene silencing. Importantly, these histone parameters, which are the focus of this Novartis Foundation Symposium, are not only tightly linked to mechanisms through which DNA methylation participates in gene silencing, but are also being integrally linked to how abnormal patterns of this DNA modi¢cation may actually arise in tumour cells. In this brief review, it is hypothesized that these interactions between DNA methylation and histone modi¢cations may have to be targeted to most e¡ectively exploit the potential for using reversal of gene silencing as an approach to cancer prevention and treatment. 2004 Reversible protein acetylation. Wiley, Chichester (Novartis Foundation Symposium 259) p 226^237

226

DNA METHYLATION IN CANCER

227

Relationships between abnormal promoter DNA methylation and histone deactylation in cancer

A complex alteration of DNA methylation takes place in cancer, beginning during early phases of tumour progression, in which this modi¢cation is lost from widespread areas of the genome where it should normally be present and increased in gene promoter regions where it should be absent (Jones & Baylin 2002). The promoter events involve CpG-rich regions, or CpG islands, which are present in almost half of genes and which are normally unmethylated regardless of gene expression status (Bird 2002). This DNA promoter methylation is but part of a complex series of local chromatin changes which initiate and maintain transcriptional silencing of the genes involved. A fundamental feature of this chromatin is the presence of multiple histone deactylases (HDACs) which are required to maintain key histone amino acid residues, such as lysines 9 and 14 of histone H3 (H3-K9^K14), in the deacetylated state prerequisite for transcriptional silencing. These HDACs are components of various protein complexes, including some harbouring DNA methyltranferases (DNMTs) (Rountree et al 2000, Bachman et al 2001, Fuks et al 2000, 2001, Robertson et al 2000, Burgers et al 2002) and others containing proteins which bind to methylated cytosines, or the methyl binding proteins (MBPs) (Bird 2002, Bird & Wol¡e 1999). These complexes also are involved with recruitment of transcriptional repressors to the regions and, in turn, can be recruited by such repressors (Bird 2002, Rountree et al 2000, Bachman et al 2001, Fuks et al 2000, 2001, Robertson et al 2000, Burgers et al 2002, Bird & Wol¡e 1999). With regards to the interaction of the above DNA methylation and HDACs, microarray-based screens for genes aberrantly silenced in cancer reveal that there are, at least, two types of promoter change which may involve 200 or so genes in a given cancer cell type (Suzuki et al 2002). The best studied change with respect to the consequences of losing function of the genes involved are those in which there is dense CpG island hypermethylation as previously noted. The second, for which the function of the involved genes merits much more intense investigation, involves those which lack such DNA methylation but for which silencing is reversed by cancer cell treatment using HDAC inhibitors alone (Suzuki et al 2002). The existence of these two types of aberrantly silenced genes in cancer (depicted in Fig. 1), and the role of histone deacetylation in each, must be considered for development of strategies to reactivate transcription for prevention or therapeutic purposes. This hypothesis is discussed in more detail below as other chromatin features of gene silencing are considered.

228

BAYLIN

FIG. 1. Targeting two categories of epigenetically silenced genes for cancer prevention and therapy. The promoter regions for the two classes of aberrantly silenced genes (X across arrow for transcription start sites) in cancer, which are discussed in the text, are depicted including those with densely DNA methylated CpG islands (gene A, black lollipops ¼ methylated CpG dinucleotides) and those without DNA methylation (gene B, white lollipops ¼ unmethylated CpG dinucleotides). Both promoters are encompassed in densely packed nucleosomes with deactylated histones (grey doublet ovals interspersed between CpG sites) and harbour protein complexes which contain histone deacetylases (HDACs). For gene A, the HDACs are recruited by methyl-cytosine binding proteins (MBPs) and by DNMTs while for gene B, other transcriptional repressor complexes recruit the HDACs. Gene A cannot be reactivated by HDAC inhibitors alone but can be reactivated by inhibitors of DNMTs and, synergistically, by a combination of the inhibitors. Gene B can be reactivated by HDAC inhibitors alone. This paradigm suggests why reversal of epigenetic gene silencing in cancer might best be targeted for prevention and/or therapy by a combination approach. The ? depicts the potential need to employ inhibitors of the histone methyltransferases (HMTs) which catalyse methylation of lysine 9 of histone H3 (asterisks within nucleosomes) to augment reversal of gene silencing or prevent it from recurring following employment of DNMT and HDAC inhibitors.

DNA METHYLATION IN CANCER 229

230

BAYLIN

Other chromatin changes involved with DNA methylation and gene silencing
the key role of histone methylation In addition to the involvement of histone deactylation in gene silencing associated with DNA methylation, one of the most exciting developments in the ¢eld has been the discovery that histone methylation, another feature of chromatin discussed in detail at this Novartis Foundation conference, is also integrally linked to the function and origins of this DNA modi¢cation. Thus, the promoters of genes which harbour DNA hypermethylation and su¡er inappropriate transcriptional silencing in cancer cells are associated with methylation of H3-K9 (see Fig. 1), a key gene-silencing mark, and depletion of methylated H3-K4 (Fahrner et al 2002, Kondo et al 2003, Nguyen et al 2002, Ghoshal et al 2002), a key histone modi¢cation marking active gene transcription (Fischel et al 2003, Lachner & Jenuwein 2002, Kouzarides 2002). These same genes, when not DNA methylated at the CpG island and when expressed in cancer cells, have acetylated H3-K9^14, methylated H3-K4, and depletion of methylated H3-K9 (Fahrner et al 2002, Kondo et al 2003, Nguyen et al 2002). These ¢ndings are particularly pertinent to recent ¢ndings for a key role of H3K9 methylation as a targeting mechanism for DNA methylation in Neurospora and Arabidopsis (Tamura & Selker 2001, Jackson et al 2002). Thus, in these nonmammalian organisms, mutations in histone methyltransferases (HMTs) which mediate methylation of H3-K9 abolish some or all DNA methylation. Recent data also link this histone modi¢cation to targeting of the promoter DNA hypermethylation in cancer cells. In a colon cancer cell-culture model, genetic disruption of two of three biologically active DNMTs results in transcriptional activation and loss of promoter DNA hypermethylation and H3-K9 methylation for the p16 gene (Rhee et al 2002). However, with continued passage of the genetically disrupted cells, there is slow re-silencing of the p16 gene and accompanying acceleration of cell growth (Bachman et al 2003). In this setting, re-methylation of H3-K9 is seen multiple passages before DNA hypermethylation returns to the gene promoter (Bachman et al 2003).

Dominance of DNA methylation for gene silencing As outlined above, multiple components of chromatin contribute to aberrant transcriptional silencing of genes in cancer, possibly including recruitment of DNA methylation to involved promoters. However, once present, this DNA modi¢cation appears to be a dominant force in maintaining the gene silencing. This is evident in several ways. Despite the fact that components of the DNA methylation machinery potentially recruit HDACs to the silent and hypermethylated promoters, silencing of the involved genes is generally not

DNA METHYLATION IN CANCER

231

relieved by treatment, alone, of cells with powerful HDAC inhibitors (Suzuki et al 2002, Cameron et al 1999). This is consistent with the latest ¢ndings that none of the histone parameters outlined above for such transcriptionally repressed genes, including the deacetylated status of H3-K9 and K14, are altered by the HDAC inhibitor, trichostatin A (TSA) (Fahrner et al 2002). In contrast, however, demethylation via 5-aza-2’-deoxycytidine (DAC) treatment, alone, leads to dramatic enrichment of not only acetylated H3-K9 and K14, but also depletion of methylated H3-K9 (Fahrner et al 2002, Kondo et al 2003, Nguyen et al 2002, Ghoshal et al 2002). The timing for this e¡ect suggests that demethylation precedes gene re-activation which may actually precede the histone post-translational modi¢cation changes (Ghoshal et al 2002). The mechanisms underlying these above e¡ects of the loss of DNA methylation are of great interest. Hypermethylated and silenced genes in cancer may constitute some of the best models for this purpose. One distinct possibility is that the changes in histone modi¢cation observed may re£ect an important mechanism recently associated with activation of gene expression in Drosophila cells (Ahmad & Heniko¡ 2002) that is also active in mammalian cells. In this scenario, the onset of transcription is accompanied by a DNA replication-independent replacement of histone H3 with the variant histone, H3.3. This appears to account for the loss of methyl H3-K9 and, by subsequent modi¢cation of the new histone, for gain of methyl H3-K4 and acetylation of H3-K9 (Ahmad et al 2002). Loss of methyl-H3K9 following induced loss of DNA methylation in Arabidopsis has been suggested to involve a similar process (Johnson et al 2002). If similar mechanisms are operative in the chromatin associated with silenced genes in cancer cells then the loss of the DNA methylation may trigger a series of changes in histone modi¢cations by ¢rstly directly allowing transcriptional activation. Such dynamics may help explain why a HDAC inhibitor, alone, cannot reactivate genes with dense promoter DNA methylation but will synergize with a low dose of a DNA demethylating agent to do so (Suzuki et al 2002, Cameron et al 1999). In other words, the onset of acetylation of newly positioned histone H3.3 following application of the demethylating agent would be aided by subsequent inhibition of HDACs. The chromatin of gene silencing in cancer
rami¢cations for clinical consideration The data discussed above emphasize that fully reversing epigenetically mediated gene silencing for cancer prevention and/or therapy may have to involve targeting of a complex process which initiates and maintains transcriptional repression. First, considering the dominant force of promoter DNA methylation for maintaining transcriptional repression, this DNA process may have to be inhibited. There appear to be many genes with this change for virtually each

232

BAYLIN

cancer type (Johnson et al 2002) and this type of silencing occurs early in tumour progression even at the premalignant stages (Nuovo et al 1999). Second, one must consider those genes which may be transcriptionally down-regulated in cancer in the absence of dense promoter DNA methylation. These may be good targets for the inhibitors of HDACs that were discussed in detail at this symposium. However, use of these HDAC inhibitors alone may not be optimally e⁄cacious for at least two reasons which are inherent to the discussions above in this chapter. These drugs will not, alone, re-activate the genes with dense promoter DNA hypermethylation even though the transcriptional silencing accompanying this change involves histone deacetylation as a key player. Furthermore, HDAC inhibitors can be synergistic with currently utilized inhibitors of DNA methylation, such as 5-azacytidine (5-AzaC) and DAC. For both reasons, these agents may be best utilized together. In this manner, two major categories of genes which are aberrantly silenced may be re-expressed. Also, this combination may also allow for lower doses of each agent to be employed. This may be especially important for use of drugs like 5-Aza-C and DAC since these produce unwanted toxicities, many of which may be secondary to the amount of drug used and to mechanisms other than block of DNA methylation (Jones & Baylin 2002, Herman & Baylin 2003). A ¢nal point concerns the potential need to develop drugs which inhibit the HMTs speci¢c for formation of methyl H3-K9. Since this amino acid change appears to play a key role in specifying where DNA methylation takes place, including where it will return during recovery of silencing following block of DNMTs (Bachman et al 2003), prevention of its formation could be critical for initiating and/or especially maintaining induction of gene re-expression in a clinical setting. There is most certainly an active ongoing enterprise to identify such HMT inhibitors. Thus, the next few years should see much activity in the exploration of not only the clinical e⁄cacy of drugs already available for reversing key chromatin components of gene silencing, but also the derivation of new drugs which can block vital steps of this dynamic process of heritable transcriptional repression. References Ahmad K, Heniko¡ S 2002 The histone variant H3.3 marks active chromatin by replicationindependent nucleosome assembly. Mol Cell 9:1191^1200 Bachman KE, Rountree MR, Baylin SB 2001 Dnmt3a and Dnmt3b are transcriptional repressors that exhibit unique localization properties to heterochromatin. J Biol Chem 276:32282^32287 Bachman KE, Park BH, Rhee I et al 2003 Histone modi¢cations and silencing prior to DNA methylation of a tumor suppressor gene. Cancer Cell 3:89^95 Bird A 2002 DNA methylation patterns and epigenetic memory. Genes Dev 16:6^21 Bird AP, Wol¡e AP 1999 Methylation-induced repression  belts, braces, and chromatin. Cell 99:451^454

DNA METHYLATION IN CANCER

233

Burgers WA, Fuks F, Kouzarides T 2002 DNA methyltransferases get connected to chromatin. Trends Genet 18:275^277 Cameron EE, Bachman KE, Myohanen S, Herman JG, Baylin SB 1999 Synergy of demethylation and histone deacetylase inhibition in the re-expression of genes silenced in cancer. Nat Genet 21:103^107 Esteller M, Corn PG, Baylin SB, Herman JG 2001 A gene hypermethylation pro¢le of human cancer. Cancer Res 61:3225^3229 Fahrner JA, Eguchi S, Herman JG, Baylin SB 2002 Dependence of histone modi¢cations and gene expression on DNA hypermethylation in cancer. Cancer Res 62:7213^7218 Fischle W, Wang Y, Allis CD 2003 Histone and chromatin cross-talk. Curr Opin Cell Biol 15:172^183 Fuks F, Burgers WA, Brehm A, Hughes-Davies L, Kouzarides T 2000 DNA methyltransferase Dnmt1 associates with histone deacetylase activity. Nat Genet 24:88^91 Fuks F, Burgers WA, Godin N, Kasai M, Kouzarides T 2001 Dnmt3a binds deacetylases and is recruited by a sequence-speci¢c repressor to silence transcription. EMBO J 20:2536^2544 Ghoshal K, Datta J, Majumder S et al 2002 Inhibitors of histone deacetylase and DNA methyltransferase synergistically activate the methylated metallothionein i promoter by activating the transcription factor mtf-1 and forming an open chromatin structure. Mol Cell Biol 22:8302^8319 Herman JG, Baylin SB 2003 Gene Silencing in Cancer in Association with Promoter Hypermethylation. N Engl J Med 349:2042^2054 Jackson JP, Lindroth AM, Cao X, Jacobsen SE 2002 Control of CpNpG DNA methylation by the KRYPTONITE histone H3 methyltransferase. Nature 416:556^560 Johnson L, Cao X, Jacobsen S 2002 Interplay between two epigenetic marks. DNA methylation and histone H3 lysine 9 methylation. Curr Biol 12:1360^1367 Jones PA, Baylin SB 2002 The fundamental role of epigenetic events in cancer. Nat Rev Genet 3:415^428 Kondo Y, Shen L, Issa JP 2003 Critical role of histone methylation in tumor suppressor gene silencing in colorectal cancer. Mol Cell Biol 23:206^215 Kouzarides T 2002 Histone methylation in transcriptional control. Curr Opin Genet Dev 12:198^209 Lachner M, Jenuwein T 2002 The many faces of histone lysine methylation. Curr Opin Cell Biol 14:286^298 Nguyen CT, Weisenberger DJ, Velicescu M et al 2002 Histone H3-lysine 9 methylation is associated with aberrant gene silencing in cancer cells and is rapidly reversed by 5-aza-2’deoxycytidine. Cancer Res 62:6456^6461 Nuovo GJ, Plaia TW, Belinsky SA, Baylin SB, Herman JG 1999 In situ detection of the hypermethylation-induced inactivation of the p16 gene as an early event in oncogenesis. Proc Natl Acad Sci USA 96:12754^12759 Rhee I, Bachman KE, Park BH et al 2002 DNMT1 and DNMT3b cooperate to silence genes in human cancer cells. Nature 416:552^556 Robertson KD, Ait-Si-Ali S, Yokochi T, Wade PA, Jones PL, Wol¡e AP 2000 DNMT1 forms a complex with rb, E2F1 and HDAC1 and represses transcription from E2F-responsive promoters. Nat Genet 25:338^342 Rountree MR, Bachman KE, Baylin SB 2000 DNMT1 binds HDAC2 and a new co-repressor, DMAP1, to form a complex at replication foci. Nat Genet 25:269^277 Suzuki H, Gabrielson E, Chen W et al 2002 A genomic screen for genes upregulated by demethylation and histone deacetylase inhibition in human colorectal cancer. Nat Genet 2002; 31:141^149 Tamaru H, Selker EU 2001 A histone H3 methyltransferase controls DNA methylation in Neurospora crassa. Nature 414:277^483

234

DISCUSSION

DISCUSSION Castronovo: In cancer there is an imbalance between the activation of oncogenes and inactivation of tumour suppressor genes. How do you see the demethylation that is induced in an oncogene that will be, as a consequence, activated? Is there a bias in that you are just looking at tumour suppressor genes? Are you expecting to have some bad e¡ect by reactivating some of the genes? Baylin: There are two papers that have recently been published in Science from Jaenisch’s group that are relevant here (Eden et al 2003, Gaudet et al 2003). They raise the possibility of deleterious e¡ects of reversing DNA methylation in patients including the potential for re-expressing an oncogene. But one must remember, genes which function as oncogenes tend to be constitutively expressed to start with in normal cells. Often, the oncogenicity of such genes results from changes such as point mutations as in Ras. Rarely is oncogenicity due to an overexpression. The promoters of these genes in normal cells, especially if they have CPG islands, are not going to be methylated to start with. Take the Myc gene as an example. When Adrian Bird ¢rst published that many CpG islands become hypermethylated in cultured cells, he theorized you would never see this change in a gene for which loss of function would cause a negative selection on the cell. He looked at a number of genes. You never see it. For these genes, demethylation is not a likely way that you are going to reactivate them if they have CPG islands. As you know, there are a number of genes that have more CPG-poor promoters, where it is becoming more of a paradigm that there are key methylation sites in those promoters that are methylated in normal tissues, and demethylation events seem to be important to the re-expression of that gene. These would be the genes that you would most likely a¡ect. But there have been very few that have been looked at so far where you actually see that kind of methylation corresponding to overexpression in a tumour and with a distinct methylation event that can be attributed to that. Rudi Jaenisch has postulated that a lot of the methylation in the genome is important for proper chromosome assembly and stability. He has wondered whether genetic instability can arise through the loss of methylation. In one paper he shows that when he crosses a hypomorphic animal (with 10^20% normal methylation) with the p53 knockout mouse, he sees increased genetic instability which he attributes to a combination of p53 depletion and hypomethylation. In another paper (Gaudet et al 2003) he shows that these animals develop thymic lymphomas at about 6 months of age. He saw Myc ampli¢cation (a region of the chromosome that was ampli¢ed contained Myc). This is an example of where hypomethylation might result in tumorigenesis. However, these are young animals and they have lost a lot of methylation in the embryo. The tumours were thymic lymphomas, and someone needs to study the thymic maturation in these mice. Perhaps an early change in thymic development allowed the tumour clones to develop.

DNA METHYLATION IN CANCER

235

Verdin: In the lab when one looks at TSA responsiveness in di¡erent cell lines, one sees a broad distribution of sensitivity. Some cell lines are incredibly sensitive to low concentrations of TSA alone. Are there tumour cell lines in which you see this type of suppression operating at low levels? Is there any correlation between sensitivity to TSA and what you have observed? Baylin: I am sure we are going to hear more about the phenotypic e¡ects of inhibiting HDACs. We haven’t looked closely at overall cellular phenotypic e¡ects with TSA, for example. For the gene events we give a very short pulse and then stop. We have not seen a big change for a given gene between cell lines. It looks pretty much the same to us. If a gene is densely hypermethylated in the promoter, this is the paradigm we get. I would also stress for the methylation ¢eld that we need to work out whether any phenomena are due to speci¢c reversal of these gene events. It is very hard to prove, and even to work out which genes have changed. Marks: In the patients that you are treating with 5-Aza-C, you are giving the leukaemia patients 50 mg/m2 and the patients with solid tumours 15 mg/m2. Why the di¡erence in the dose? Baylin: I don’t treat the patients. The studies are being done by my colleagues, Steven Gore and Michael Carducci. With the patients with solid tumours the clinicians wanted to go to a lower dose to see whether they could treat for longer to start with. Since 5-Aza-C has had some e⁄cacy in leukaemia, they started these patients at a standard dose. Marks: Have you looked at peripheral mononuclear cells to see whether you are altering the pattern of histone methylation or acetylation? Baylin: We have been looking at overall global acetylation. We can readily see changes in the blood cells of the patients with 5-Aza-C treatment. After we saw what we did with the individual genes, Steve Gore went back and looked during the time course before the HDAC inhibitor was administered. There is quite a change in global acetylation. Marks: How long does it last? Baylin: We don’t know that yet. One of the responders had a methylated Ecadherin gene that went away. This person has stayed in long-term remission. In others it has been harder to see speci¢c gene changes. We have a lot of kinetics to work out, for both global changes and speci¢c changes. Cole: Have you looked at any of the cells that are lacking the methyltransferases, treated them with 5-Aza-C and seen any e¡ects that might not be methyltransferasespeci¢c but which might be interesting in their own right? Baylin: We have done the latter. Rudi Jaenisch and others have said that the toxicity seen with 5-Aza-C parallels the amount of activity you have to start with in a cell. They think that this is due to 5-Aza-C incorporation into the DNA, resultant binding of the DNMT proteins to the DNA, and that this in itself

236

DISCUSSION

might be a mode of toxicity and a site for potential damage. We didn’t see this. The double knockouts had almost no DNMT activity. Yet if you treat them with 5Aza-C they look almost the same as the wild-type cells in terms of manifesting toxic e¡ects of the drug. Cole: In the study where you looked at many passages of cells and methylation of DNA came back, what do you thing is going on there? What is the methyltransferase? Baylin: This is where this will be an interesting model. The dogma is that the maintenance of DNA methylation (you put back on the daughter strand what was on the parent strand, during replication) is almost entirely done by DNMT1. This enzyme accounts for 90%+ of what is measured in the cell. The dogma is that putting back a de novo site is due to 3a and 3b. I don’t think it is this simple. We have good evidence that DNMT1 can de novo methylate also. We have been putting the DNMT genes back into the double knockout cells. Potentially, the DNMTs are multitasking proteins. Their N-termini can silence gene transcription in Gal4 tethering-type assays, they bind HDACs and they are a platform for certain corepressors. When we put them back we start to see silencing of certain genes. For example, 3b will silence MLH1, and when we put it back early on we don’t see methylation. It may be acting as a transcriptional repressor and building a complex. Very little is known about this. Li: In your NNK mouse model you use adult mice. Do you know any target genes that are potentially methylated in this case? Baylin: Yes, there are candidate genes. This raises an important point about epigenetics and cancer. These NNK-treated animals quickly methylate the p16 gene as shown by Belinsky and colleagues. The repair gene, O6-MGMT, which is another gene that gets silenced in human tumours, is also methylated. Many of these events are conserved. We didn’t monitor all along the way, so we don’t know whether those targets were quickly hypermethylated but they would be candidates. In many of the human cancers, some of these hypermethylation events occur very early onp16 silencing, for example, is seen in precancerous lesions. Li: In the control experiment, if you treat adult mice with 5-Aza-C and phenylbutyrate do you see tumours if you let the mice go for a longer time? Does hypomethylation lead to tumours in this kind of model? Baylin: We didn’t see lymphomas in these animals. Whether methylation was reduced in the whole animal is unclear, but we didn’t see other tumours in the autopsies. Turner: You showed a diagram suggesting that certain promoter regions were protected from DNA methylation. You get the methylation coming in from either side, and then the CpG island protects it. At the same time you have got increased K4 methylation and decreased K9 methylation. You have an epigenetic histone

DNA METHYLATION IN CANCER

237

mark there as well. Then we all throw in TSA and other deacetylase inhibitors: you have found, and we have found as well, that some genes simply don’t respond. Are we looking at a region where some promoters are protected generally against ongoing changes. They are protected against DNA methylation but they are also protected from the e¡ects of inhibitors of acetylation or methylation. If this is a general model, it impacts on the treatments we are trying to implement for cancer. Baylin: It is a chicken and egg type situation. Are some of the histone modi¢cations protective? Are they boundaries that help establish a region that stays protected from the DNA methylation? Adrian Bird has surmised that what happens to CpG islands is that early in development, once they are transcribed, the binding of the transcription factor complexes and the moulding of that region protects them from subsequently being methylated. It may be that at some point in the history of the gene the active transcription puts the promoter in some sort of a zone with those marks. It could be the DNA demethylase(s) that clears these regions, but no one has reliably found this, so one would predict that the chromatin is somehow holding this in a protective zone. But I don’t know which comes ¢rst. Turner: I think the idea of those protective zones is very important. Baylin: I should brie£y mention that Peter Jones has now done extensive analysis of the p16 gene and he saw the map we did. He only saw that zone of acetylationenrichment and K4 methylation-enrichment in the 5’ region. He has done screening of chromosome 21 by ChIP analysis with the antibodies. But then, instead of looking for speci¢c sequences he does a random prime. He looked for all sequences that were acetylated, and it was a tiny minority. He suggested that for that chromosome only very small regions are held in this kind of zone of acetylation. The mammalian genome might be built that way. Mahadevan: We have looked at acetylation of several oncogenes. There is a speci¢c acetylation of these genes that happens very quickly, but then disappears by 4^6 h. There are di¡erent patterns of sensitivity to acetylation, and these oncogenes respond in a hypersensitive fashion. We also looked at several other control genes, and none were a¡ected. Baylin: It may be di¡erent for di¡erent genes. It took us 48 h to see the distinct histone modi¢cation changes. For p16, Peter Jones looked at the loss of methyl with 5-Azo-C and observed loss within 12^24 h. References Eden A, Gaudet F, Waghmare A, Jaenisch R 2003 Chromosomal instability and tumors promoted by DNA hypomethylation. Science 300:455 Gaudet F, Hodgson JG, Eden A et al 2003 Induction of tumors in mice by genomic hypomethylation. Science 300:489^492

Transcription regulation by histone deacetylases Shaowen Wang, Yan Yan-Neale, Marija Zeremski and Dalia Cohen 1 Novartis Institutes for Biomedical Research, Inc, 100 Technology Square, Cambridge, MA 02139, USA

Abstract. Dynamic changes in the post-translational modi¢cation pattern of histories such as acetylation, deacetylation, phosphorylation, methylation and ubiquitination are thought to provide a code for correct regulation of gene expression by a¡ecting chromatin structure and interaction with regulatory factors. Our studies focus on the role of histone deacetylases (HDACs) in transcriptional regulation and addressing functional di¡erences of class I and class II HDACs. To identify genes that were transcriptionally regulated by speci¢c HDACs, genome scale expression pro¢les were performed in cancer cells following the inhibition of three HDAC family members by speci¢c oligonucleotides. The modulated genes identi¢ed in this study represented a wide range of modi¢cations in di¡erent cellular pathways. In addition, treatment of cancer cells with a HDAC inhibitor was found to induce the expression of the small GTPase RhoB through an inverted CCAAT box in the RhoB promoter. These studies identi¢ed a speci¢c transcription element involved in HDAC-mediated gene transcription and genes that are transcriptionally regulated by speci¢c HDACs, providing important insight into the potential therapeutic bene¢t of HDAC inhibition. 2004 Reversible protein acetylation. Wiley, Chichester (Novartis Foundation Symposium 259) p 238^248

Dynamic changes in post-translational modi¢cations of histones are thought to provide a code for correct regulation of gene expression by a¡ecting chromatin structure and interaction with regulatory elements. Our studies focus on the analysis of histone acetylation and the regulation of speci¢c gene expression. The acetylation status of histones is dynamically controlled by histone acetyltransferases (HATs) and histone deacetylases (HDACs). A large number of HATs have been identi¢ed as transcriptional coactivators, among them GCN5, CBP/p300, PCAF, SRC-1 and TAF250 (Chen et al 2001, Nakatani 2001). To date, 18 mammalian HDACs have been identi¢ed and found to be associated 1This paper was presented at the symposium by Dalia Cohen to whom correspondence should be

addressed. 238

HDAC REGULATION OF TRANSCRIPTION

239

with transcriptional repression. The HDACs are classi¢ed into three classes (Gray & Ekstrom 2001, Imai et al 2000, Khochbin et al 2001). Class I, encompassing HDAC1^3, 8 and 11 are related to the yeast Rpd3 histone deacetylase. Class II, which includes HDAC4^7, 9 and 10, have been found to be related to the yeast Hda1 histone deacetylase (Bertos et al 2001) and Class III, also known as the Sir2 family, consists of seven genes related to yeast Sir2, and possess nicotinamideadenine dinucleotide (NAD+)-dependent deacetylase activity (Vaziri et al 2001). HDAC inhibitors are potent inducers of growth arrest, di¡erentiation and apoptotic cell death in a variety of transformed cells in culture and in tumourbearing animals (Marks et al 2001). Hence, HDAC inhibitors are being evaluated as potent anticancer agents. Interestingly, the action of HDAC inhibitors on gene expression appears to be selective, altering the transcription of only a limited number of genes in cultured tumour cells (Van Lint et al 1996), but the basis for this selectivity is unknown. In addition, the biochemical events subsequent to HDAC inhibition that lead to cell cycle arrest, cellular di¡erentiation and apoptosis have not yet been fully identi¢ed. Trapoxin A (TPX) is a potent, non-competitive HDAC inhibitor that has been used to explore the mechanisms by which histone deacetylation alters transcription (Furumai et al 2001). Using TPX, we investigated the growth responses of cells to the HDAC inhibitor. In addition, we focused our studies on the elucidation of molecular mechanisms by which p21waf1 and RhoB genes were regulated by HDACs and the possible role of p21waf1 and RhoB in HDAC inhibitor-mediated apoptosis, cell proliferation and tumour suppression.

TPX selectively alters the activity and expression of cell cycle proteins and inhibits cell proliferation To reveal the molecular mechanisms by which HDAC mediates its antiproliferative e¡ects in transformed cells, three human tumour cell lines with di¡erent p53 status were investigated: a human non-small cell lung NCI-H1299 carcinoma (p537/7), a human ductal breast carcinoma MDA-MB-435 (Nakano et al 1997) expressing a mutant p53, and a human lung A549 (p53 wild-type) carcinoma. TPX treatment of H1299 and MDA-MB-435 cells led to the accumulation of cells in G1 and G2 phases and a decrease of cells in S phase, thus indicating that TPX induced cell cycle arrest in these two cell lines. However, a large fraction of the A549 cells was observed to be apoptotic. These results indicated that cell cycle arrest may be primarily responsible for the antiproliferative e¡ects of TPX in H1299 and MDA-MB-435, whereas apoptosis without cell cycle arrest is induced by TPX in the A549 cells (Sambucetti et al 1999).

240

WANG ET AL

TPX treatment also signi¢cantly induced the level of p21waf1 and, since activated complexes of cyclin-dependent kinases and cyclins are necessary for complex formation with p21waf1, the expression levels of cdk2, cdk4, cyclin A, and cyclin B were examined. No changes in the levels of cdk2, cdk4, and cyclin B were detected and a small decrease in cyclin A was noted. The elevated levels of p21waf1 also correlated with a decrease in Rb phosphorylation and a decrease in cdk2 activity in the cells. TPX activates p21waf1 transcription through a responsive region that is acetylated within the chromatin complex To determine whether the e¡ect of TPX on p21waf1 gene expression could be mapped to the p21waf1 promoter, the regulatory region of p21waf1 was cloned upstream of a luciferase reporter gene and the TPX-responsive region within the p21waf1 promoter was identi¢ed. The promoter region from 7168 to 793 containing Sp1 sites was found to be responsible for TPX mediated p21waf1 activation (Fig. 1A). Moreover, this activation correlated with acetylation of the chromatin complex in this promoter region and the level of acetylated H3 and H4 signi¢cantly increased following treatment with TPX (Sambucetti et al 1999). TPX induces RhoB expression through an inverted CCAAT box and its binding protein NF-Y To further identify genes and transcription factors that are mediated by HDAC inhibition, we determined the global gene expression pro¢les of TPX-treated H1299 cells by DNA microarray analysis. Among other genes, and similar to p21waf1, the small GTPase RhoB was signi¢cantly up-regulated in response to TPX (Wang et al 2003). The increased levels of RhoB mRNA in TPX-treated cells suggested that the RhoB gene was transcriptionally regulated by HDACs, thus linking RhoB activity and HDAC inhibition. In contrast, the transcription of two other closely related Rho GTPase subfamily members, RhoA and RhoC, were not altered in response to TPX-mediated HDAC inhibition. Chromatin immunoprecipitation experiments indicated that HDAC1 and HDAC2, but not HDAC6 were associated with the RhoB promoter. Consistent with this observation, overexpression of HDAC1 was found to speci¢cally repress the RhoB promoter, whereas HDAC6 had no e¡ect on RhoB expression. Furthermore, inhibition of HDAC1 with a speci¢c antisense oligonucleotide induced RhoB gene expression, providing independent evidence that HDAC1 mediated RhoB transcriptional repression. Similar to the approach taken with p21waf1, the TPX-responsive element within the RhoB promoter was mapped and found to be an inverted CCAAT box

HDAC REGULATION OF TRANSCRIPTION

241

FIG. 1. Schematic representation of human p21waf1 and RhoB promoters. Human p21waf1 and RhoB promoters are schematically depicted. The putative binding sites for various transcription factors are indicated. +1 ATG is the translation start site. (A) p21waf1 promoter. (B) RhoB promoter.

(Fig. 1B). Mutations in this CCAAT box were found to reduce the ability of HDAC1 to confer repression of RhoB promoter activity, and chromatin immunoprecipitation assays revealed that TPX increased the level of chromatin acetylation in the RhoB promoter containing the inverted CCAAT box. Furthermore, gel super-shift assays indicated that NF-Y complex binds to this element suggesting that NF-Y may be a component in the HDAC1/CCAAT complex (Fig. 2). RhoB suppression is associated with high proliferation rates and ectopic overexpression of RhoB promotes apoptotic responses in HCT116 and HeLa cells Recent ¢ndings indicated that altered expression or activity of Rho proteins might be crucial to cancer progression and therapeutic responses (Prendergast 2001, Sahai & Marshall 2002). As an important step toward understanding RhoB function, a survey of RhoB expression in transformed cells was performed by DNA microarrays. Signi¢cantly low levels of RhoB expression were observed in most

242

WANG ET AL

FIG. 2. Gel supershift analysis using the 0.1 kb RhoB promoter fragment as probe. An aliquot of 10 mg nuclear extracts from H1299 cells was analysed for DNA binding activity to the 32P-labelled 0.1 kb RhoB promoter fragment (RhoB5 probe) by gel retardation. For the supershift assay, RhoB5 probe was incubated with 10 mg nuclear extracts of H1299 cells alone (
); or pre-incubated nuclear extracts with 1 mg anti-rabbit IgG, anti-NF-YA, anti-NF-YB, antiCDP, anti-NF-1, before adding probe; or post-incubated with anti-NF-YA, anti-NF-YB, antiCDP, anti-NF-1, after adding probe, respectively. Arrows indicate the shifted bands.

of the transformed cells, suggesting that RhoB suppression is associated with high proliferation rates. In contrast, high expression of RhoA and RhoC, which are highly homologous to RhoB, were detected in most of the transformed cells. Furthermore, overexpression of RhoB, but not RhoA, led to an increase in the apoptotic index of HCT116 and HeLa cells (Fig. 3). These results suggested that RhoB may play an important role in HDAC inhibitor-mediated apoptosis and tumour suppression and may be a negative modi¢er in cancer (Chen et al 2000, Du & Prendergast 1999, Liu et al 2001). In summary, we have mapped regulatory DNA binding motifs in two genes that are important for transcription regulation mediated by HDAC. Interestingly, while Sp1 sites play a major role in the expression of p21waf1 by HDAC, this element has no apparent role in HDAC-mediated regulation of RhoB. This gene

FIG. 3. Overexpression of RhoB induces apoptosis in HeLa and HCT116 cells. HeLa and HCT116 cells transfected with pcDNA-RhoB, pcDNA-RhoA or pcDNA3.1 vectors were analysed for apoptosis. Cells exhibiting sub-G1-phase DNA were scored as apoptotic cells. The percentage of cells undergoing apoptosis in the total cell population was calculated, and the number is shown above the sub-G1 peak. (A) HeLa cells. (B) HCT116 cells. (C) Expression levels of RhoA and RhoB determined by immunoblotting in the cells used for apoptosis analysis.

HDAC REGULATION OF TRANSCRIPTION 243

244

WANG ET AL

was found to be transcriptionally regulated by HDAC1 and HDAC2 through an inverted CCAAT box. The analysis of global gene expression of cells treated with a HDAC inhibitor identi¢ed RhoB as a gene regulated by HDAC. This established for the ¢rst time, a link between RhoB expression and HDAC inhibition and o¡ers a novel mechanism to explain some of the tumour suppression e¡ects and induction of apoptosis in tumour cells by HDAC inhibitors. The approach of mapping HDAC function to distinct DNA elements in regulatory regions of speci¢c genes may have the potential for identifying speci¢c target proteins that mediate promoter repression by HDAC inhibition. Furthermore, the identi¢cation of p21waf1 and RhoB as well other genes (Park et al 2002, Jin & Scotto 1998, Butler et al 2002) that are preferentially regulated by Sp1 and CCAAT boxes might indicate that independent cellular signalling events are a¡ected by HDAC inhibition. These pathways can result in modulations of di¡erent transcription factors and chromatin accessibility to allow gene expression but further studies are needed to reveal the consequences of HDAC inhibition on chromatin structure, transcription factors and speci¢c gene regulation. References Bertos NR, Wang AH, Yang XJ 2001 Class II histone deacetylases: structure, function, and regulation. Biochem Cell Biol 79:243^252 Butler LM, Zhou X, Xu WS et al 2002 The histone deacetylase inhibitor SAHA arrests cancer cell growth, up-regulates thioredoxin-binding protein-2, and down-regulates thioredoxin. Proc Natl Acad Sci USA 99:11700^11705 Chen Z, Sun J, Pradines A, Favre G, Adnane J, Sebti SM 2000 Both farnesylated and geranylgeranylated RhoB inhibit malignant transformation and suppress human tumor growth in nude mice. J Biol Chem 275:17974^17978 Chen H, Tini M, Evans RM 2001 HATs on and beyond chromatin. Curr Opin Cell Biol 13:218^ 224 Du W, Prendergast GC 1999 Geranylgeranylated RhoB mediates suppression of human tumor cell growth by farnesyltransferase inhibitors. Cancer Res 59:5492^5496 Furumai R, Komatsu Y, Nishino N, Khochbin S, Yoshida M, Horinouchi S 2001 Potent histone deacetylase inhibitors built from trichostatin A and cyclic tetrapeptide antibiotics including trapoxin. Proc Natl Acad Sci USA 98:87^92 Gray SG, Ekstrom TJ 2001 The human histone deacetylase family. Exp Cell Res 262:75^83 Imai S, Armstrong CM, Kaeberlein M, Guarente L 2000 Transcriptional silencing and longevity protein Sir2 is a NAD- dependent histone deacetylase. Nature 403:795^800 Jin S, Scotto KW 1998 Transcriptional regulation of the MDR1 gene by histone acetyltransferase and deacetylase is mediated by NF-Y. Mol Cell Biol 18:4377^4384 Khochbin S, Verdel A, Lemercier C, Seigneurin-Berny D 2001 Functional signi¢cance of histone deacetylase diversity. Curr Opin Genet Dev 11:162^166 Liu AX, Cerniglia GJ, Bernhard EJ, Prendergast GC 2001 RhoB is required to mediate apoptosis in neoplastically transformed cells after DNA damage. Proc Natl Acad Sci USA 98:6192^6197

HDAC REGULATION OF TRANSCRIPTION

245

Marks PA, Rifkind RA, Richon VM, Breslow R, Miller T, Kelly WK 2001 Histone deacetylases and cancer: causes and therapies. Nat Rev Cancer 1:194^202 Nakano K, Mizuno T, Sowa Y et al 1997 Butyrate activates the WAF1/Cip1 gene promoter through Sp1 sites in a p53-negative human colon cancer cell line. J Biol Chem 272:22199^ 22206 Nakatani Y 2001 Histone acetylases  versatile players. Genes Cells 6:79^86 Park SH, Lee SR, Kim BC et al 2002 Transcriptional regulation of the transforming growth factor beta type II receptor gene by histone acetyltransferase and deacetylase is mediated by NF-Y in human breast cancer cells. J Biol Chem 277:5168^5174 Prendergast GC 2001 Actin’ up: RhoB in cancer and apoptosis. Nat Rev Cancer 1:162^168 Sahai E, Marshall CJ 2002 Rho-GTPases and cancer. Nat Rev Cancer 2:133^142 Sambucetti LC, Fischer DD, Zabludo¡ S et al 1999 Histone deacetylase inhibition selectively alters the activity and expression of cell cycle proteins leading to speci¢c chromatin acetylation and antiproliferative e¡ects. J Biol Chem 274:34940^34947 Van Lint C, Emiliani S, Verdin E 1996 The expression of a small fraction of cellular genes is changed in response to histone hyperacetylation. Gene Expr 5:245^253 Vaziri H, Dessain SK, Ng Eaton E et al 2001 hSIR2(SIRT1) functions as an NAD-dependent p53 deacetylase. Cell 107:149^159 Wang S, Yan-Neale Y, Fischer D et al 2003 Histone deacetylase 1 represses the small GTPase RhoB expression in human nonsmall lung carcinoma cell line. Oncogene 22:6204^6213

DISCUSSION Castronovo: In the experiment in which you study the e¡ect of TPX in reducing the length of the promoter for RhoB, I understand how you have your inverted box. In the plasmid you are transfecting, how do you visualize the inhibition of the acetylase? How is this linked with what is really going on in the cell? How in such a small sequence can TPX a¡ect the luciferase activity of your construct? Cohen: I agree with you. Whenever we have data like this we go to the endogenous gene, to see whether we can match induction of p21 or RhoB by TPX. The induction by TPX is completely obliterated when we mutate it. Castronovo: Can anyone explain the molecular mechanism? Berger: It could be another target. The transcription factor that is binding to the CCAAT box could be acetylated. Cohen: NFY has been shown to perhaps be acetylated. Castronovo: This could have nothing to do with the acetylation of histone, then. Berger: That is a possibility. Verdin: One thing we have seen is that almost every reporter plasmid that is transfected is responsive to TSA. This could indicate that these transiently transfected plasmids are packaged in some unique form of chromatin. Berger: That could still be through other factors. Castronovo: Then you are not looking at the same process than you are going to in vivo. Marks: We essentially have exactly the same results with SAHA. Doing promoter deletion we come out with the inverted CCAAT box as being the

246

DISCUSSION

critical site in the p21 promoter for the action of SAHA. I don’t think this is mutually exclusive from an e¡ect on histone acetylation. As Dalia points out, it is possible to demonstrate the accumulation of acetylated histone in the promoter region of this p21 gene. Berger: Those could be concerted events. The acetylation of the factor and the acetylation of the histone could be linked, and these might both be a¡ected. Mahadevan: Is this a general e¡ect to do with CCAAT boxes? Have you looked at other genes that are controlled in this way? Cohen: We identi¢ed several genes containing CCAAT inverted boxes that were regulated by HDAC inhibition. However, not all genes containing the CCAAT box are regulated by HDAC. Mahadevan: Did you also see those e¡ects in a TATA mutant when you mutated the TATA box? Cohen: The TATA box was mutated in the reporter gene, and this mutation had no e¡ect. Mahadevan: Can it work in TATA-less genes, considering some endogenous genes are TATA-less? Cohen: When we mutated the TATA we didn’t see any e¡ect. It is only when we have a reporter gene that contained the mutated TATA with the mutated CCAAT that we saw the e¡ect. Verdin: When you mutate each individual HDAC with your antisense RNA, did you see any of them inducing apoptosis? Cohen: Those experiments are extremely di⁄cult. We didn’t want to kill the cell. We wanted to make sure that we are not looking at cell death. The condition in which the experiments were done corresponded to the down-regulation of HDAC1 before cell death. Because this time course is long there is a question as to whether we are seeing secondary e¡ects. In my experience, using RNA oligo there is a big distance between down-regulation of the gene and down-regulation of the protein: HDAC1 has a long half-life and the experiments focused on identifying genes that are modulated when the protein is down-regulated. Verdin: How long were these genes suppressed? Cohen: Between 36 and 48 h. Verdin: In most cell lines apoptosis induction happens at 12^13 h, very precisely. Cohen: That’s a good point. We are using H199 cells that go into cell growth arrest for ever and don’t apoptose. They can stay in the dish for a long time. Atadja: Sometimes in our lab we see annexin-stained cells with the mismatch antisense controls, but this is very cell-line speci¢c. The relevant time point to gain insight into the e¡ect of these antisense oligos is between 13^16 h. Verdin: Isn’t one of the critical points to identify the inhibition of which particular HDAC that will lead to this apoptosis response?

HDAC REGULATION OF TRANSCRIPTION

247

Gu: I don’t think we should just consider one type of HDAC inhibitor alone, particularly in the case of p53. If we treat the cell with a HDAC inhibitor such as TPX you may only see a mild e¡ect on p53-dependent pathways. The reason is that in unstressed cells p53 is not highly acetylated at all. However, in response to DNA damage, p300 will be recruited. If you treated the cell now with HDAC inhibitors, you may see very strong e¡ects. The key question is: have you tried a combination of a DNA-damaging drug and a HDAC inhibitor? I would propose this cocktail treatment of the tumour cells with a DNA damage drug, Sir2 inhibitors and HDAC inhibitors. Cohen: This is probably the right way to go. Steve Baylin has already shown some promising combinations in his paper. Gu: Also, you can use minimal concentrations and still see robust activity. Li: I was surprised to see so many genes down-regulated. What happened there? Is it an indirect e¡ect? Cohen: I don’t know yet. We are now BLASTing the promoters of all those genes to look for any consensus sequences. It seems that there are some sequences that might be conserved in a subset of genes. Marks: When do you look for the gene expression with the microarrays? Cohen: 36^48 h. This is only to see whether the protein is really down-regulated. Li: So you haven’t done any ChIP analysis? Cohen: This is in process. We have 20 genes on which we are performing ChIP. Seto: Many genes have Sp1 and inverted CCAAT boxes. Why aren’t they activated by HDAC inhibitors? Cohen: This was a surprise for us. RhoB has a Sp1 site, but this Sp1 is not mediating RhoB induction. I don’t know the answer. It looks like there is a subset of genes which are mediated by Sp1, and some that are mediated by CCAAT. I believe that there will be other transcription factors which will be gene speci¢c. Ott: How well does your up-regulated and down-regulated gene pro¢le overlap with TSA? Is it all mediated by the HDAC catalytic activity? Cohen: The ¢rst analysis showed that there are many functionally related genes, but a more detailed analysis will be required. Mahadevan: I’m interested in what you consider to be primary and secondary e¡ects of treatment. If you are up-regulating a transcription factor gene, for example, these could be going on to modulate other events. Is there a way of discriminating this? Cohen: I don’t know if I can discriminate using these experiments. We have validated some of these A¡ymetrix data genes and are now doing a dose^ response study. If you compare the A¡y data from HDAC1, 3 and 7, it is very clear that there is a subset of up- and down-regulated genes that is not mediated by a mismatched or non-speci¢c oligo. The question of primary and secondary e¡ects remains to be answered.

248

DISCUSSION

Turner: You showed a nice Venn diagram indicating the genes regulated by the di¡erent HDACs. Is there any correlation between the genes you see in those di¡erent circles, and the genes that Mike Grunstein and others have identi¢ed in yeast, where they have done similar sorts of experiments? Have you been able to do that sort of comparison? Cohen: We have, and the answer is yes. There is an overlap, but not a large one. It is clear that although there is no overlap between genes that are regulated by HDAC1, 3 and 7, we were able to classify the genes by function and cellular pathways. Turner: Can we say that we have real evolutionary conservation here of the types of genes? Cohen: The genes found to be regulated by the three enzymes were assigned to major classi¢cation and cellular pathways and some are known to be conserved.

Molecular and cellular basis for the anti-proliferative e¡ects of the HDAC inhibitor LAQ824 Peter Atadja, Meier Hsu, Paul Kwon, Nancy Trogani, Kapil Bhalla* and Stacy Remiszewski Department of Oncology Molecular and Cellular Biology, Novartis Institutes for Biomedical Research, East Hanover, NJ 07936 and *Department of Interdisciplinary Oncology, Mo⁄tt Cancer Center and Research Institute University of South Florida, Tampa, Florida, USA

Abstract. We have developed a cinnamic hydroxamic class of histone deacetylase inhibitors of which a prototype was designated as NVP-LAQ824. NVP-LAQ824, inhibits histone deacetylase enzymatic activities in vitro and transcriptionally activated the p21 promoter in reporter gene assays. When tested on a variety of solid tumour cell lines, NVP-LAQ824 exhibited selective anti-proliferative e¡ects, inducing cell growth inhibition in some, while inducing cell death in others. To induce cell death, a minimum of 16 h exposure to NVP-LAQ824 is required. Flow cytometry studies revealed that both tumour cell lines and normal diploid ¢broblasts arrested in the G2/M phase of the cell cycle after compound treatment. However, an increased sub-G1 population at 48 h (reminiscent of apoptotic cells) was only observed in the cancer cell lines. Annexin V staining data con¢rmed that NVP-LAQ824 induced apoptosis in tumour cells, but not in normal cells. To relate HDAC inhibition to the antiproliferative e¡ects of NVP-LAQ824, expression of HDAC 1 was inhibited using antisense and this was su⁄cient to activate p21 expression, hypophosphorylate Rb and inhibit cell growth. Furthermore, tumour cells treated with NVP-LAQ824 caused acetylation of HSP90 and degradation of its cargo oncoproteins. Finally, NVPLAQ824 exhibited antitumour e¡ects in a xenograft animal model. To determine if NVP-LAQ824 inhibited histone deacetylases in vivo, tumours treated with the drug were immunoblotted with an antibody speci¢c for acetylated histones H3 and H4 and the results indicated increased histone H3 and H4 acetylation levels in NVP-LAQ824 treated cancer cells. Together, our data indicated that the activity of NVP-LAQ824 was consistent with its intended mechanism of action. This novel HDAC inhibitor is currently in clinical trials as an anticancer agent. 2004 Reversible protein acetylation. Wiley, Chichester (Novartis Foundation Symposium 259) p 249^268

Inhibition of histone deacetylation (HDAC) provides a novel approach for cancer treatment. Acetylation of histones is a major regulator of gene expression that acts 249

250

ATADJA ET AL

by altering the accessibility of transcription factors to DNA. In normal cells a balance exists between histone acetyltransferase and HDA activity that leads to cell-speci¢c patterns of gene expression. In cancer cells, perturbing this balance can result in silencing of tumour suppressor genes. Inhibiting HDA increases histone acetylation and activates gene expression from acetylation-sensitive promoters (Struhl 1998). Inappropriate expression of genes required for cell proliferation, di¡erentiation or tumour suppression has been linked to cancer. Several lines of evidence suggest that aberrant recruitment of histone deacetylases (HDACs) and the resulting chromatin modi¢cations may lead to changes of gene expressions seen in transformed cells: . silencing of tumour suppressor genes at the chromatin level is common in human tumours (Corn et al 1999, Domann et al 2000, Schagdarsurengin et al 2002, van Engeland et al 2002, Gasco et al 2002, Sharpless et al 2001, Yanagawa et al 2002) . HDAC-containing complexes have been shown to interact with proteins involved in tumorigenesis (Boivin et al 2002, Shinagawa et al 2001, Won et al 2002) . HDAC inhibitors (HDAIs) have been reported to produce signi¢cant antiproliferative e¡ects, such as promoting di¡erentiation, cell cycle arrest or apoptosis (Han et al 2000, Marks et al 2000, Greenberg et al 2001, Jaboin et al 2002, Furumai et al 2001, Fournel et al 2002) . induction of p21 gene expression, a key mediator of G1 cell cycle arrest and di¡erentiation, is observed when tumour cells are treated with HDAIs (Han et al 2000, Kim et al 2000, Richon et al 2001, Sambucetti et al 1999, Strait et al 2002). Pre-clinical experiments using small-molecule inhibitors of HDACs have been reported. Synthetic HDAC inhibitors, MS-275 and SAHA exhibited e⁄cacy against several human tumour xenografts in athymic mice (Saito et al 1999, Richon et al 2001). In addition, the natural product HDAC inhibitors of the trapoxin class and trichostatin were shown to activate the p21 promoter, increase p21 protein levels, inhibit cdk2 kinase activity, reduce Rb phosphorylation and cause cell cycle arrest or apoptosis in three human tumour cell lines (Sambucetti et al 1999, Strait et al 2002, Saito et al 1999, Lavelle et al 2001). Here we report the molecular and cellular e¡ects of a novel synthetic HDAC inhibitor NVPLAQ824 (Fig. 1). NVP-LAQ824 potently inhibited HDAC enzymatic activities in vitro, transcriptionally activated the p21 promoter, and selectively inhibited tumour cell growth at sub-micromolar concentrations. Our studies indicated that NVP-LAQ824 induced apoptosis in tumour cells whereas treatment of normal

HDAC INHIBITOR LAQ824

251

FIG. 1. Chemical structure of NVP-LAQ824.

¢broblasts caused cell cycle arrest. Furthermore, we con¢rmed the activity of NVP-LAQ824 against HDACs which resulted in increased histone acetylation in treated cancer cells. Speci¢c suppression of HDAC1 expression was su⁄cient to produce the e¡ects seen with NVP-LAQ824. We therefore conclude that the mechanism of drug action was, in fact, inhibition of HDAC. Materials and methods Materials NVP-LAQ824 was prepared in house as the lactate salt and dissolved in dimethyl sulfoxide (DMSO). MS-275 was prepared as described in Suzuki et al (1999) as a free base and was dissolved in DMSO. The ECF Western blotting reagent pack for mouse or rabbit was purchased from Amersham Pharmacia Biotech Inc. (Piscataway, NJ). Pre-cast NuPAGE gels were from Invitrogen Life Technologies (Carlsbad, CA). The CellTiter 96 AQueous Non-Radioactive Cell Proliferation Assay was from Promega (Madison, WI). The ApoAlert Annexin V Apoptosis kit was from Clontech Laboratories, Inc. (Palo Alto, CA). The HybondP PVDF membrane was purchased from Amersham Biosciences (Piscataway, NJ). Cell culture H1299, HCT116, A549, DU145, PC3, and MDA435 cells were obtained from American Type Culture Collection (Rockville, MD) and were maintained according to supplier’s instructions. Normal dermal human ¢broblast (NDHF) cells were obtained from Clonetics (San Diego, CA) and were maintained in Dulbecco’s modi¢ed Eagle’s medium supplemented with 15% fetal bovine serum, 100 units/ml penicillin and 100 g/ml streptomycin. Cell cycle analysis A549, HCT116, and NDHF cells (1106 each) were seeded on 10 cm2 tissue culture dishes and were allowed to grow overnight at 37 8C with 5% CO2. A549 and

252

ATADJA ET AL

NDHF cells were then exposed to NVP-LAQ824 at 0.5 mM or 0.1% DMSO and HCT116 cells were treated with 0.1 mM NVP-LAQ824 or 0.1% DMSO. After 24, 48 or 72 h, the cells were subjected to trypsin treatment, collected by centrifugation, and washed once with 10 ml of PBS. They were resuspended in ice-cold 70% ethanol, washed again with 5 ml of PBS, resuspended in PI solution (70 mM propidium iodide, 38 mM sodium citrate, 20 mg/ml RNase A), and incubated at 37 8C for 30 min. Flow cytometric analysis was performed on a FACSort instrument (Becton Dickinson Immunocytometry Systems, San Jose, CA), and the data were subsequently analysed using the ModFitLT software (Becton Dickinson). Apoptosis detection A549, HCT116 or NDHF cells (3105 each) were seeded in a 24-well plate (Costar Inc., Corning, NY). 24 h later, cells were incubated with NVP-LAQ824 at indicated concentrations or 0.1% DMSO as the control. 24 or 48 h post NVPLAQ824 treatment, the media was removed and the cells were incubated in the dark for 15 min with binding bu¡er containing 1mg/ml Annexin V-FITC and 2.5 mg/ml propidium iodide. The cells were observed with a £uorescent microscope using a dual ¢lter set for FITC and rhodamine. Monolayer growth inhibition assay Cell proliferation was measured using an adaptation of published procedures (MTS assay) as previously described (Suzuki et al 2000). The CellTiter 96 AQueous Non-Radioactive Cell Proliferation Assay (Promega) was performed according to manufacturer’s protocol. The average background value (treatment with media alone) was subtracted from each experimental well; triplicate values were averaged for each compound dilution. The following formulae were used to calculate percent growth. If X 4 T0 , % Growth ¼ ((X
T0 )=(GC
T0 ))  100 If X 5 T0 , % Growth ¼ (X
T0 )=T0  100 where: T0 ¼ average value of T0
background GC ¼ average value of untreated cells ðin triplicateÞ
background X¼average value of compound treated cells (in triplicate) ^ background The ‘% Growth’ was plotted against compound concentration and used to calculate the IC50 employing the linear regression techniques between data points to predict the concentration of compounds at 50% inhibition.

HDAC INHIBITOR LAQ824

253

In vitro histone deacetylase assay The assay was performed as previously described (Sambucetti et al 1999). p21 promoter activation assay The plasmid encoding the luciferase gene driven by the p21 promoter and its transfection into H1299 cells were described previously (Sambucetti et al 1999). Data were analysed by calculating mean fold activation (n ¼3) of compoundtreated cells compared to the average of DMSO-treated control (n ¼3). Percent activity for each concentration of NVP-LAQ824 was determined by comparison to the maximal fold activation obtained from a reference HDAC inhibitor, psammaplin A (L. Perez, unpublished data). The AC50 values were calculated using a linear regression calculation between data points to predict the concentration of compound needed for 50% activation relative to the reference. Western blot analysis HCT116 and A549 tumour cells (1106 each) and NDHF cells (2106) were treated with 200 mM NVP-LAQ824. Following 3, 6 or 24 h treatments, cells were washed once with 10 ml of ice-cold PBS and lysed in 500 ml of ice-cold triple detergent bu¡er (50 mM Tris-Cl, pH 8.0, 150 mM NaCl, 0.02% sodium azide, 0.1% SDS, 1% NP-40, 0.5% sodium deoxycholate, 10 ml/ml protease inhibitor cocktail) for 5 min. The lysates were cleared by centrifugation at 14 000 rpm for 10 min at 4 8C. Cell lysates (10 mg) were separated on a 4^12% NuPAGE gel by electrophoresis and transferred onto the Hybond-P PVDF membrane. The membranes were probed with the primary antibodies diluted in PBS containing 1% non-fat dry milk and 0.2% Tween-20 for 2 h at room temperature or overnight at 4 8C, followed by the appropriate secondary antibodies for detection using the ECF Western blotting reagent pack. Signals were detected by the Storm 860 PhosphorImager (Molecular Dynamics, Sunnyvale, CA). The anti-p21waf1 antibody (Oncogene Research Products, San Diego, CA) was diluted 1:100, and the anti-Rb antibody (Pharmingen, San Diego, CA) was used at 1:1000 dilution. The anti-acetylated H3 antibody (Upstate Technologies, Lake Placid, NY) was diluted 1:1000 and the anti-acetylated H4 antibody (Upstate Technologies, Lake Placid, NY) was diluted at 1:250. The anti-proliferating cell nuclear antigen antibody was used at 1:1000 dilution. Tumour xenograft implantation The studies were performed on-site, using outbred athymic (nu/nu) female mice (‘Hsd:Athymic Nude-nu’ from Harlan Sprague Dawley, Indianapolis, IN). Mice

254

ATADJA ET AL

were anaesthetized with Metofane (Mallinckrodt Veterinary, Inc., Mundelein, IL), and a cell suspension (100 ml) containing 1106 HCT116 cells was injected subcutaneously into the right axillary (lateral) region of each animal. Tumours were allowed to reach the volume of approximately 100^400 mm3. At this point, mice bearing tumours with acceptable morphology (non-necrotic) and of similar size range were selected and distributed into groups of six for the studies. NVPLAQ824 was dissolved in DMSO to create a stock solution, which was further diluted just before dosing with D5W to a ¢nal DMSO concentration of 10%. Tumour-bearing mice were treated with the compound by intravenous injection into the tail vein. NVP-LAQ824 was dosed once daily, 5 days/week for a total of 15 doses. 5-Fluorouracil was administered at 100 mg/kg in 0.9% saline 1 day/week for a total of three doses. The control groups were treated with the vehicle. Tumours were collected from the animals at the indicated time points. Results HDAC enzyme inhibition by NVP-LAQ824 Cellular HDAC enzymatic activity puri¢ed by ion exchange chromatography from H1299 human lung carcinoma cell lysates was incubated with [3H]-labelled acetylated histone-H4 peptide. Radiolabelled acetic acid released in the presence or absence of NVP-LAQ824 was extracted using ethyl acetate and counted. The IC50 value was determined for NVP-LAQ824 in this assay. As shown in Table 1, NVP-LAQ824 inhibited HDAC enzymatic activities in several independent experiments at low nanomolar concentrations. Thus, NVP-LAQ824 is a potent inhibitor of HDAC activity. p21 promoter activation and antiproliferative activity of NVP-LAQ824 One gene whose expression is activated by HDAC inhibitors is the one encoding the p21 protein (Sowa et al 1997). To determine the e¡ect of NVP-LAQ824 on p21 gene expression, H1299 cells were transfected with a plasmid construct of p21 promoter driving a luciferase gene and followed by compound treatment. A concentration-dependent increase in luciferase activity was observed. The concentration of NVP-LAQ824 producing 50% of the maximal promoter TABLE 1

IC50 of HDAC enzyme inhibition

Compound

IC50 (Puri¢ed HDAC)

NVP-LAQ824

0.04 +0.001 mM

Determined as described in materials and methods.

HDAC INHIBITOR LAQ824

255

TABLE 2 The concentration of NVP-LAQ824 producing 50% of the maximal promoter activation (AC50) relative to a reference natural product HDAC inhibitor Inhibition of monolayer cell proliferation IC50 mM

Compound

p21 promoter activation AC50 mM

NVP-LAQ824

0.300+0.020 0.150+0.015 0.010+0.002 0.018+0.011 0.023+0.009

H1299

HCT116

DU145

PC3

The p21 promoter assays were done in triplicate. Inhibition of cell growth in monolayer after 72 h compound treatment was measured in triplicate by MTS assays as described in materials and methods. All experiments were repeated at least three times.

activation (AC50) relative to a reference natural product HDAC inhibitor, psammaplin A, is shown in Table 2. These results indicated that low micromolar concentrations of NVP-LAQ824 were needed to induce the p21 promoter activity, implicating that the compound is likely to up-regulate endogenous p21 gene expression in cells. Previous studies have demonstrated that HDAC inhibitors induce tumour cells to undergo growth arrest, therefore, we examined the e¡ects of NVP-LAQ824 on the growth of tumour cell lines. Monolayer growth inhibition assays were performed. As shown in Table 2, results of these experiments revealed that continuous exposure to NVP-LAQ824 for 72 h inhibited the growth of a nonsmall-cell lung carcinoma (H1299) cell line, a colon carcinoma cell line (HCT116), two prostate tumour cell lines (PC-3 and DU145), as well as the breast cancer cell line MDA435. As shown in Table 2, low nanomolar concentrations of NVP-LAQ824 were su⁄cient to signi¢cantly inhibit the growth of these tumour cells.

NVP-LAQ824 selectively induces apoptosis in tumour cells but not in normal ¢broblasts For the potential use of NVP-LAQ824 as an anticancer agent, it must exhibit toxicity towards tumour cells but not normal cells at therapeutic doses (Richon et al 2001). To compare the e¡ect of NVP-LAQ824 on tumour and normal cells, normal human dermal ¢broblasts (NDHF), and HCT116 colon carcinoma cells were exposed to various concentrations of the compound, and cell viability was measured at di¡erent time points. LD50 and LD90 values were de¢ned as the concentrations of compound that caused 50% and 90% cell death respectively. The results of these experiments showed that continuous exposure for 72 h to NVP-LAQ824 produced LD90 values of 0.09 mM in HCT116 (Table 3). Interestingly, the LD90 value was not achieved in NDHFs when treated with up to 3.75 mM of NVP-LAQ824. Therefore, NVP-LAQ824 exhibited a selective

256

TABLE 3 LAQ824

ATADJA ET AL

E¡ects of continuous exposure of HCT11 and NDHF cells to NVP-

IC50 (mM) HCT11 24 h 48 h 72 h NDH 24 h 48 h 72 h

LD50 (mM)

LD90 (mM)

0.0 0.0 0.0

0.6 0.0 0.0

2.2 0.1 0.0

43.7 0.1 0.0

43.7 43.7 0.6

43.7 43.7 43.7

HCT116 and NDHF cells were treated with various concentrations of NVP-LAQ824 for the indicated amount of time before the removal of the compound and the addition of fresh media. Cell viability was measured at 72 h by MTS assays as described in Table 1. De¢nitions of IC50, LD50 and LD90 were described in material and methods.

toxicity towards the tumour cell lines while inducing only growth arrest without signi¢cant cell death in normal ¢broblasts. To con¢rm that NVP-LAQ824 induced cell death is by apoptosis, annexin V binding was used as a marker for the early stages of apoptosis. HCT116 and NDHF cells were treated with NVP-LAQ824 for 24 or 48 h, stained with annexin V and compared to cells treated similarly with vehicle (DMSO). Cells were examined by £uorescence microscopy. Those undergoing apoptosis exhibited green £uorescent membrane staining. Viability was assessed by the counterstain, propidium iodide. Cells detected by red £uorescence were not viable. As shown in Fig. 2, the majority of HCT116 cells exhibited cell surface staining with annexin V after 24 h exposure and after 48 h treatment, red propidium iodide staining was observed indicating that the compounds induced apoptotic cell death. In contrast, NDHF cells did not show noticeable annexin V staining after 24 h or 48 h exposure. These data con¢rmed that HCT116 cells but not normal diploid ¢broblasts underwent apoptotic cell death upon NVP-LAQ824 treatment.

Lack of G1/S arrest in tumour cells despite induction of p21 expression and Rb hypophosphorylation by NVP-LAQ824 Progression through the mammalian cell cycle is controlled by cyclin-dependent kinases (CDK). The product of the p21 gene which was found to be activated by LAQ824 is a CDK inhibitor which causes hypophosphorylation of cyclin/CDK substrates such as Rb (Fig. 3). Hypophosphorylation of Rb is associated with cell

HDAC INHIBITOR LAQ824

257

FIG. 2. Detection of apoptosis by Annexin V staining. HCT116 (0.1mM), and NDHF (0.5 mM) cells were treated with NVP-LAQ824 for 24 or 48 h. Cells were stained with Annexin V or propidium iodide as described in materials and methods. Experiments were repeated at least three times and representative results are shown.

FIG. 3. NVP-LAQ824 induces p21 protein expression and Rb hypophosphorylation. HCT116 cells were treated with NVP-LAQ824 at the indicated concentration for 48 h. Total cell lysates were analysed by Western blotting using antibodies for p21 (upper panel) and Rb (lower panel).

258

ATADJA ET AL

cycle arrest at the G1/S phase (Han et al 2000, Sambucetti et al 1999). Consistent with previous ¢ndings, we observed a dose-dependent increase of p21 protein in response to NVP-LAQ824 treatment. Concomitant to the increased p21 level, we detected an increase in the hypophosphorylated state of the Rb tumour suppressor (Fig. 4). Thus, like many other HDAC inhibitors, NVP-LAQ824 was able to alter the expression level and change post-translational modi¢cations of key cell cycle regulators. In light of the above ¢ndings as well as the observation that NVP-LAQ824 produced growth inhibition in cancer cell lines, we examined the cell cycle pro¢les of NDHF and HCT116 cells following compound treatment. DMSOtreated cells were used as the negative control. As expected, a G1/S and a G2/M arrest occurred in the normal diploid ¢broblasts, whether treated for 24 h or 48 h. However, No G1/S arrest was observed in the HCT116 colon carcinoma cell lines when treated with NVP-LAQ824 for 24 h and a G2/M arrest which was observed after 24 h treatment was lost after treatment for 48 h, with most cells accumulating in a sub-G1 phase, reminiscent of apoptosis (Fig. 5). Interestingly, no signi¢cant sub-G1 population was discernible in NDHF cells at 24 h, and its rise was much less dramatic at 48 h compared to the increases in HCT116 and A549 cells (Fig. 4). These results suggested that NVP-LAQ824 treatment breached/bypassed Rbassociated checkpoint mechanisms normally responsible for the G1/S arrest, as well as the G2/M checkpoints in the tumour cells examined. Nevertheless, these checkpoint controls appear likely to remain intact in treated normal ¢broblasts.

NVP-LAQ824 exhibited antitumour activities in HCT116 tumour xenografts in athymic nude mice To assess the ability of NVP-LAQ824 to inhibit tumour growth, we examined its e¡ects in subcutaneously-implanted tumours in athymic nude mice. HCT116 cells were implanted subcutaneously and when tumours reached an average size of 100 mm3, NVP-LAQ824 or vehicle was administered intravenously once daily (q.d.) for 5 days: a regime that lasted for a total of 3 weeks. Tumour sizes were measured every 7 days post-implantation. As shown in Fig. 5, NVP-LAQ824 treatment produced a dose-dependent inhibition of tumour growth; at 100 mg/ kg, its antitumour e¡ect was similar to that of 5-£uorouracil, a standard anticancer agent. Importantly, no signi¢cant changes in animal body weight were observed even at the highest compound concentration (data not shown), suggesting that the growth inhibitory e¡ect was tumour-speci¢c and that NVPLAQ824 did not produce a general cytotoxicity in mice.

FIG. 4. Cell cycle alterations in response to NVP-LAQ824 treatment. HCT116 (0.1mM), and NDHFs (0.5 mM) were treated with NVP-LAQ824. Cell cycle pro¢les were analysed by FACS at 24 or 48 h. The percentage of the cells in G0, G1, S and G2 phases were calculated by the ModFit program and are shown.

HDAC INHIBITOR LAQ824 259

260

ATADJA ET AL

FIG. 5. Antitumour activity of NVP-LAQ824 in tumour xenografts in nude mice. Bar chart of HCT116, human colon tumour, xenografts shown and expressed as mean tumour volume. The mice were treated with NVP-LAQ824 dissolved in the delivery vehicle 5% dextrose in water, 5-£uorouracil, and 10% DMSO/D5W at indicated concentrations. T/C represents the ratio of compound-treated tumour volume over vehicle-treated tumour volume, expressed as a percentage.

NPV-LAQ824 increased acetylation of histones H3 and H4 in tumours treated with NVP-LAQ824 While NVP-LAQ824 exhibited robust in vivo antitumour activity (Fig. 5) it was also essential to demonstrate that this was accompanied by inhibition of HDAC activity in the tumours. The levels of histone H3 and H4 acetylation were examined in tumours of athymic mice bearing HCT116 treated with the compound. Mice were dosed with 100 mg/kg of NVP-LAQ824 and the tumours were removed 3, 6, 16 or 24 h later. Histone acetylation levels detected with antibodies speci¢c for acetylated histone H3 and H4 were compared to PCNA, the levels of which did not change in response to the compounds in cell culture. As shown in Fig. 6, in vivo treatment NVP-LAQ824 resulted in consistent increases in histone-H3 and H4 acetylation after 3 h and this persisted up to 24 h. PCNA levels were essentially invariant. The increase in histone acetylation in treated animals coupled with their potent e¡ects on partially puri¢ed HDA, strongly suggested that these compounds were inhibiting the HDAC enzyme in HCT116 tumour xenografts.

HDAC INHIBITOR LAQ824

261

FIG. 6. NVP-LAQ824 treatment increased histone H3 and H4 acetylation in tumour cells. HCT116 tumour xenograft bearing mice were treated with vehicle or with 100 mg/kg of NVPLAQ824, the tumours harvested at the indicated time-points post-dose and acetylation levels of histones H3 (Ac-H3) and H4 (Ac-H4) were analysed by Western blotting. The amount of PCNA was examined to ensure equal loading of samples in each lane.

Suppression of HDAC 1 expression by antisense was su⁄cient to mimic antiproliferative e¡ects of LAQ824 To determine that the antiproliferative e¡ects seen with NVP-LAQ824 was due to inhibition of a HDAC, speci¢c antisense oligonucleotides were designed against HDAC1 (design and sequence of the HDAC1 antisense oligonucleotides will be described elsewhere). As shown in Fig. 7A, transfection of H1299 cells with the HDAC1 speci¢c antisense oligonucleotide but not a control oligonucleotide inhibited the expression of HDAC. Concomitantly, increased expression of p21 protein and hypophosphorylation of Rb was detected in the HDAC1 antisense transfected cells but not the control oligonucleotide transfected cells. To further determine whether suppression of HDAC1 resulted in inhibition of cell growth, H1299 cells were transfected with increasing concentrations of HDAC1 antisense oligonucleotide or control oligonucleotide and MTS assays performed 72 h later. As shown in Fig. 7B, transfection of the HDAC 1 antisense but not control oligonucleotide resulted in signi¢cant inhibition of cell growth. Thus inhibition of HDAC1 expression is su⁄cient to produce the antiproliferative e¡ects seen with the pan-HDAC inhibitor LAQ824 in vitro. Treatment of tumour cells with NVP-LAQ824 increases HSP90 acetylation and degradation of oncoproteins Acetylation of HSP90 in response to HDAC inhibitor treatment has been previously described (Yu et al 2002). To determine whether NVP-LAQ824 treatment results in HSP90 acetylation, K562 cells were treated with NVPLAQ824, lysed and HSP90 protein immunoprecipitated and subsequently

262

ATADJA ET AL

FIG. 7. Antisense against HDAC1 activates p21 expression, hypo-phosphorylates Rb and inhibits cell growth. (A) H1299 cells were transfected with HDAC1 antisense or a mismatch oligonucleotide (control). Cells were harvested 24 h later and immunobloted with p21 or Rb antibodies. (B) H1299 cells were transfected with increasing concentrations of the HDAC1 antisense or the control mismatch oligonucleotide and an MTS assay was performed 48 h later to determine e¡ects on cell proliferation.

immunoblotted with an antibody against acetyl lysine. As shown in Fig. 8A, increasing concentration of NVP-LAQ824 resulted in increased acetylation of HSP90. To assess the e¡ect on the chaperone e¡ect of HSP90, SKBR cells were treated with NVP-LAQ824 and immunoblotted with antibodies against oncoproteins known to be chaperoned and stabilized by HSP90. Figure 8B

HDAC INHIBITOR LAQ824

263

FIG. 8. Treatment of tumour cell lines with NVP-LAQ824 results in HSP90 acetylation and degradation of oncoproteins. (A) K562 cells were treated with increasing concentration of NVPLAQ824, cells harvested, lysed and immunoprecipitated with an antibody against HSP90. The immunoprecipitates were immunoblotted using a speci¢c antibody against acetyl lysine. (B) SKBR-3 cells were treated with NVP-LAQ824 and lysates were immunoblotted with antibodies speci¢c for the indicated HSP90 cargo oncoproteins.

shows that treatment with NVP-LAQ824 results in the degradation of the oncoproteins Her2/neu, Her3, and c-raf as well as the anti-apoptotic protein phosphoAKT. Therefore, the antitumour activity of NVP-LAQ824 may be due to decreased cellular levels of oncoproteins.

264

ATADJA ET AL

Discussion In this report, we describe the molecular and cellular e¡ects of the novel HDAC inhibitor NVP-LAQ824. The response of HCT116 and NDHF cells to the compound varied both in the concentration needed to inhibit growth as well as in their propensity to undergo apoptosis in response to the compounds. The sensitivity of HCT116 cells to the HDAC inhibitor appeared to correlate with apoptosis induced by the compound. Both cell types carry a wild-type p53 gene. Although this may play a role in the cellular response to these compounds, previous results did not ¢nd a clear correlation between sensitivity to hydroxamate HDAIs and p53 status (Saito et al 1999). Normal diploid ¢broblasts (NDHFs) were less sensitive to NVP-LAQ824 than either HCT116 or A549 cells. NDHF response to the compounds was characterized by complete loss of S phase and accumulation of cells at G1/S and G2/M phases, with minimal cell death, as compared to the tumour cell lines. Although, all normal and tumour cells demonstrated an initial increase in the percentage of cells in G2/M in response to the compound, both G1 and G2/M arrest appeared to be most complete in NDHF cells, suggesting that the inability of the tumour cells to stably arrest in G1/S or G2/M phases may trigger their apoptotic response. Preliminary studies from our laboratory have shown that di¡erential sensitivity towards NVP-LAQ824, as well as other HDAC inhibitors, was associated with di¡erential expression of mitotic checkpoint regulators (data not shown). Other reports have also suggested that malfunction of G2/M checkpoint mediator(s) in many transformed cell lines may result in the loss of growth arrest following HDAC inhibitor treatment (Yu et al 2002). It is conceivable that defects in the checkpoint mechanisms, a characteristic in malignancy, ultimately trigger the apoptotic response to NVP-LAQ824 exposure in these tumour cells. By the same token, checkpoint mechanisms are likely to remain intact in normal cells allowing them to undergo growth arrest without massive apoptotic cell death. If the response in NDHF is representative of other normal cell types, this may provide a basis for a favourable therapeutic index between normal and tumour cells in vivo. The increase in histone acetylation in treated animals coupled with their potent e¡ects on partially puri¢ed HDA, strongly suggests that these compounds inhibit the HDA enzyme in HCT116 tumour xenografts in mice. This is consistent with a mechanism of action where HDA inhibition may activate expression of tumour suppressors, such as p21, and gene products that produce apoptosis in tumours. To better de¢ne the mechanism of antiproliferative action, antisense techniques were employed to speci¢cally inhibit the expression of HDAC1. Surprisingly, suppression of this HDAC isoform alone was su⁄cient to induce antiproliferative e¡ects seen with the HDAC inhibitor. These results suggest that

HDAC INHIBITOR LAQ824

265

the antitumour e¡ects seen with NVP-LAQ824 might include inhibition of HDAC1 and that selective inhibition of tumour-relevant HDAC isoforms to achieve antitumour e¡ect might be possible. Furthermore, we have showed that the antitumour e¡ect of NVP-LAQ824 may occur through acetylation of the chaperone protein HSP90 leading to the degradation of its cargo oncoproteins. In conclusion, the novel HDAC inhibitor NVP-LAQ824 which produced antitumour e¡ects in preclinical animal models, o¡ers the potential to be a highly potent anticancer therapeutic.

References Boivin AJ, Momparler LF, Hurtubise A, Momparler RL 2002 Antineoplastic action of 5-aza-2’deoxycytidine and phenylbutyrate on human lung carcinoma cells. Anticancer Drugs 13: 869^874 Corn PG, Kuerbitz SJ, van Noesel MM et al 1999 Transcriptional silencing of the p73 gene in acute lymphoblastic leukemia and Burkitt’s lymphoma is associated with 5’ CpG island methylation. Cancer Res 59:3352^3356 Domann FE, Rice JC, Hendrix MJ, Futscher BW 2000 Epigenetic silencing of maspin gene expression in human breast cancers. Int J Cancer 85:805^810 Fournel M, Trachy-Bourget MC, Yan PT et al 2002 Sulfonamide anilides, a novel class of histone deacetylase inhibitors, are antiproliferative against human tumors. Cancer Res 62:4325^4330 Furumai R, Komatsu Y, Nishino N, Khochbin S, Yoshida M, Horinouchi S 2001 Potent histone deacetylase inhibitors built from trichostatin A and cyclic tetrapeptide antibiotics including trapoxin. Proc Natl Acad Sci USA 98:87^92 Gasco M, Sullivan A, Repellin C et al 2002 Coincident inactivation of 14-3-3sigma and p16INK4a is an early event in vulval squamous neoplasia. Oncogene 21:1876^1881 Greenberg VL, Williams JM, Cogswell JP, Mendenhall M, Zimmer SG 2001 Histone deacetylase inhibitors promote apoptosis and di¡erential cell cycle arrest in anaplastic thyroid cancer cells. Thyroid 11:315^325 Han JW, Ahn SH, Park SH et al 2000 Apicidin, a histone deacetylase inhibitor, inhibits proliferation of tumor cells via induction of p21WAF1/Cip1 and gelsolin. Cancer Res 60:6068^6074 Jaboin J, Wild J, Hamidi H et al 2002 MS-27-275, an inhibitor of histone deacetylase, has marked in vitro and in vivo antitumor activity against pediatric solid tumors. Cancer Res 62:6108^6115 Kim YB, Ki SW, Yoshida M, Horinouchi S 2000 Mechanism of cell cycle arrest caused by histone deacetylase inhibitors in human carcinoma cells. J Antibiot (Tokyo) 53:1191^1200 Lavelle D, Chen YH, Hankewych M, DeSimone J 2001 Histone deacetylase inhibitors increase p21(WAF1) and induce apoptosis of human myeloma cell lines independent of decreased IL-6 receptor expression. Am J Hematol 68:170^178 Marks PA, Richon VM, Rifkind RA 2000 Histone deacetylase inhibitors: inducers of di¡erentiation or apoptosis of transformed cells. J Natl Cancer Inst 92:1210^1216 Richon VM, Zhou X, Rifkind RA, Marks PA 2001 Histone deacetylase inhibitors: development of suberoylanilide hydroxamic acid (saha) for the treatment of cancers. Blood Cells Mol Dis 27:260^264 Saito A, Yamashita T, Mariko Y et al 1999 A synthetic inhibitor of histone deacetylase, MS-27275, with marked in vivo antitumor activity against human tumors. Proc Natl Acad Sci USA 96:4592^4597

266

DISCUSSION

Sambucetti LC, Fischer DD, Zabludo¡ S et al 1999 Histone deacetylase inhibition selectively alters the activity and expression of cell cycle proteins leading to speci¢c chromatin acetylation and antiproliferative e¡ects. J Biol Chem 274:34940^34947 Schagdarsurengin U, Gimm O, Hoang-Vu C, Dralle H, Pfeifer GP, Dammann R 2002 Frequent epigenetic silencing of the CpG island promoter of RASSF1A in thyroid carcinoma. Cancer Res 62:3698^3701 Sharpless NE, Bardeesy N, Lee KH et al 2001 Loss of p16Ink4a with retention of p19Arf predisposes mice to tumorigenesis. Nature 413:86^91 Shinagawa T, Nomura T, Colmenares C, Ohira M, Nakagawara A, Ishii S 2001 Increased susceptibility to tumorigenesis of ski-de¢cient heterozygous mice. Oncogene 20:8100^8108 Sowa Y, Orita T, Minamikawa S et al 1997 Histone deacetylase inhibitor activates the WAF1/ Cip1 gene promoter through the Sp1 sites. Biochem Biophys Res Commun 241:142^150 Strait KA, Dabbas B, Hammond EH, Warnick CT, Iistrup SJ, Ford CD 2002 Cell cycle blockade and di¡erentiation of ovarian cancer cells by the histone deacetylase inhibitor trichostatin A are associated with changes in p21, Rb, and Id proteins. Mol Cancer Ther 1:1181^1190 Struhl K 1998 Histone acetylation and transcriptional regulatory mechanisms. Genes Dev 12:599^606 Suzuki T, Ando T, Tsuchiya K 1999 Synthesis and histone deacetylase inhibitory activity of new benzamide derivatives. J Med Chem 42:3001^3003 Suzuki T, Yokozaki H, Kuniyasu H et al 2000 E¡ect of trichostatin A on cell growth and expression of cell cycle- and apoptosis-related molecules in human gastric and oral carcinoma cell lines. Int J Cancer 88:992^997 van Engeland M, Roemen GM, Brink M et al 2002 K-ras mutations and RASSF1A promoter methylation in colorectal cancer. Oncogene 21:3792^3795 Won J, Yim J, Kim TK 2002 Sp1 and Sp3 recruit histone deacetylase to repress transcription of human telomerase reverse transcriptase (hTERT) promoter in normal human somatic cells. J Biol Chem 277:38230^38238 Yanagawa N, Tamura G, Oizumi H, Takahashi N, Shimazaki Y, Motoyama T 2002 Frequent epigenetic silencing of the p16 gene in non-small cell lung cancers of tobacco smokers. Jpn J Cancer Res 93:1107^1113 Yu X, Guo ZS, Marcu MG et al 2002 Modulation of p53, ErbB1, ErbB2, and Raf-1 expression in lung cancer cells of depsipeptide FR901228. J Natl Cancer Inst 94:504^513

DISCUSSION Marmorstein: The inhibitor compound you described looks signi¢cantly di¡erent from SAHA and related compounds. What is the evidence that it directly binds the HDACs? Atadja: We have modelling evidence, and also we know that it maintains the salient features that allow SAHA to bind to HDACs. Marmorstein: Has anyone done a binding experiment to see if it actually binds? Atadja: Yes, our NMR people have done binding experiments to show that there is good interaction between the compound and the protein. Marmorstein: In the NMR experiments does it bind as well as SAHA and related compounds? Atadja: At least as well.

HDAC INHIBITOR LAQ824

267

Marks: In your experiments with the normal human diploid ¢broblasts, what happens if you expose the cells for longer than 48 h? And what happens if you wash out your HDAC inhibitor at 48 h? Are the e¡ects reversible? Atadja: Good questions. When we treat the normal cells for longer than 48 h, as long as the concentration is under 1mM, we don’t see signi¢cant cell death. When we wash the compound o¡ the normal cells, they grow ¢ne. With the tumour cells, when we treat them for less than 16 h we can reactivate growth. If we treat the for longer than 16 h this is not reversible. Castronovo: Is this at the same concentration? Atadja: Because of the obvious e¡ect, we push the normal cells harder than the tumour cells. At far lower concentrations the tumour cells will enter a point of no return at 16 h, whereas the normal cells are reversible. Ott: Do you know which HDAC deacetylates HSP90? Atadja: I can’t say. Turner: If I interpreted your diagram correctly, what you found was that at low dosage of the deacetylase inhibitor, you did see binding of acetylated HSP90 to the ATP sepharose, but at higher doses the binding disappeared. Atadja: That’s correct. Turner: This suggests to me that the acetylated HSP90 can bind to ATP sepharose, but as you increase the acetylation then the binding disappears. The fact that you did seem to see binding of the acetylated HSP90 to ATP rather puzzles me. Atadja: These observations are phenomenology at this point. Are we acetylating more molecules? We don’t know. Nor do we know how we are losing the binding to ATP. We know that ATP needs to bind to HSP90. The kinetics of this might be very di¡erent from the kinetics of acetylating the HSP90. Castronovo: I have a comment about your A¡ymetrix study that looked at genes overexpressed in cancer compared with normal tissue. You have to be cautious about this because there is a lot of cell heterogeneity between the tumour and the normal tissue. This can lead to misleading information. We have compared the expression of HDAC1 in normal tissue and prostate cancer by Northern blot and quantitative PCR. We found a very consistent overexpression of HDAC1 in prostate cancer. When we looked by in situ hybridization we saw the contrary. It is very misleading. Atadja: I agree. We have paired normal and tumour tissues from the same people and we assay by Western blot and immunohistochemistry, as well as doing the A¡ymetrix chips. In the majority of tissues we have looked at we see increased expression of HDAC1 in cancers. Cohen: The only way to do it is by microdissection. Now there are methods to perform expression pro¢ling starting with microdissection of a homogeneous cell population.

268

DISCUSSION

Verdin: Even if you were to isolate the correct area, could this result not just re£ect mitotic index? Do you know whether HDAC1 is more expressed in dividing cells versus resting cells, for example? If this were the case, if you were to look in any tumour you’d ¢nd more HDAC1 than in normal tissues just by this mechanism alone. Atadja: That’s a good question. When we look at HDAC expression through the cell cycle, it doesn’t vary that much. I ask myself this question as well. Mind you, the cell cycle time of a prostate tumour is not the same as that of a normal cell. Gu: In tissue culture, HDAC1 is induced by serum. Atadja: We haven’t seen that. Castronovo: To my knowledge, at this point no one knows at which level HDAC is regulated. Is it the transcription level of the protein? Or the stabilization of RNA? Or translation? Cole: I have a couple of pharmacological questions. I noticed that in the xenograft models that you were treating, you were using what I calculated to be about 10 mM and above in order to get these responses. That is a big di¡erence from the levels used in cell culture. Could you comment on this? From your days in Levitski’s lab you obviously recognize that these are Michael acceptors. Have you considered that you are either alkylating the protein and/or inducing phase 2 enzymes? Atadja: We know that we are not alkylating the proteins. I don’t think Michael reactions are happening. With regard to the levels we know that 100 mg per kg gives tumour inhibition, but we can titrate that down to 5 mg/kg and this gives a statistically signi¢cant inhibition. Cole: That is still 10 mM. Atadja: That aside, the compound is also protein bound (about 80%). We are looking for the e¡ect of the free drug. The free drug in the animal may not be the same as the free drug in tissue culture. This drug is sharply metabolized in plasma. It is the fact that it is accumulating in the tumour that gives us the e¡ect. Castronovo: I am always surprised at drugs which target a very general process and have few side e¡ects. In the preclinical development of your compound, did you look to see whether it was abortive and prevented trophoblastic invasion, which is a state quite close to cancer progression in terms of its biology? Have you tried to see whether your drug would prevent trophoblastic invasion? Atadja: We haven’t looked. The only thing we saw with one of the compounds that did not make it to development phase was an e¡ect on some reproductive processes.

Histone deacetylase inhibitors: development as cancer therapy Paul A. Marks, Victoria M. Richon*, Wm Kevin Kelly{, Judy H. Chiao* and Thomas Miller* Cell Biology Program, Sloan Kettering Institute, Memorial Sloan-Kettering Cancer Center, *Aton Pharma, Inc., Tarrytown, New York and {Department of Medicine, Memorial SloanKettering Cancer Center, New York, USA

Abstract. Histone deacetylase (HDAC) inhibitors represent a new class of targeted anticancer agents. A number of structural classes of HDAC inhibitors have been developed of which several are in clinical trials, including phenylbutyrate (PB) and related compounds; the hydroxamic acids, suberoylanilide hydroxamic acid (SAHA) and depsipeptide (FK-228); and the benzamides, MS-275 and C1-994. This review will focus on our studies with the hydroxamic acid HDAC inhibitors, of which SAHA is the lead agent. X-ray crystallographic studies with a HDAC homologue (HDLP) demonstrated that the hydroxamic acid group, most of the aliphatic chain and part of the phenyl amino group of SAHA inserts into the pocket-like catalytic site of the enzyme, at the base of which is a zinc molecule. SAHA inhibits the activity of class I and II HDACs and is selective in altering gene expression. SAHA is synergistic in its anticancer activity with radiation, kinase inhibitors, cytotoxic agents and di¡erentiating agents. In phase I clinical trial with orally administered SAHA the agent caused accumulation of acetylated histones in peripheral mononuclear cells and tumour cells, has excellent bioavailability and has shown antitumour activity in patients with haematologic and solid tumours. 2004 Reversible protein acetylation. Wiley, Chichester (Novartis Foundation Symposium 259) p 269^284

The base sequence of DNA provides the fundamental code for proteins and for the regulation of gene expression. However, epigenic factors play a major role in the control of gene transcription. Nucleosomal histones are major sites of epigenic gene regulation. DNA is packaged into nucleosomes, the repeating protein complexes in chromatin, composed of about 146 base pairs of two superhelical turns of DNA wrapped around an octamer core of pairs of histones H4, H3, H2A and H2B (Spotswood & Tuner 2002, Jenuwein & Allis 2001, Zhang & Reinberg 2001). The N-terminal tails of the histones are subject to posttranslational modi¢cation by acetylation of lysines, methylation of lysines and arginines, phosphorylation of serines and ubiquitination of lysines. HDACs and 269

270

MARKS ET AL

histone acetyl transferases (HATs) are the groups of enzymes that determine the pattern of histone acetylation. The pattern of acetylation, and that of methylation and phosphorylation of histones, is dynamic and appears to involve sequential, post-translational modi¢cations. These epigenetic events likely represent a ‘code’ that can be recognized by non-histone proteins resulting in the formation of complexes involved in regulation of gene expression. HDACs are also involved in the reversible acetylation of non-histone proteins such as p53 and tubulin. Thus, acetylation and deacetylation determine, at least in part, the activity of proteins regulating cell cycle progression, as well as, gene transcriptional activity (Polevoda & Shaerman 2002, Kouzarides 2000, Freiman & Tjian 2003). Alteration in HATs and HDACs occur in many cancers (Timmermann et al 2001, Wang et al 2001, Jones & Baylin 2002). Genes encoding HATs have been found to be translocated, ampli¢ed, overexpressed and/or mutated in di¡erent cancers. Speci¢c alterations in HDAC genes have not been reported in human cancers, but these enzymes have been found to be associated with oncogenes and tumour suppressor genes. Several small molecules have been discovered which inhibit class I and class II HDACs. HDAC inhibitors induce cancer cell growth arrest, di¡erentiation and/or apoptosis in vitro and in vivo (Marks et al 2001, Kelly et al 2002a). These agents act selectively in altering the transcription of relatively few of the expressed genes (Butler et al 2002, Suzuki et al 2002, Della Ragione et al 2001, Van Lint et al 1996). Several HDAC inhibitors are in clinical trials (Kelly et al 2002b). Histone deacetylases Mammalian HDACs have been ordered into three classes (Table 1). Class I deacetylases; HDACs 1, 2, 3 and 8 share homology in their catalytic sites and are related to yeast Rpd3 deacetylase with molecular weights of 42^55 kDa; class II deacetylases include HDACs 4, 5, 6, 7, 9 and 10 (Grozinger & Schreiber 2002, De Ruijter et al 2003). They have molecular weight between about 120 to 130 kDa and are related to the yeast Hda1 deacetylase. HDACs 4, 5, 7 and 9 share homology in two regions, the C-terminal catalytic domain and the N-terminal regulatory domain. HDAC11 contains conserved residues in the catalytic core regions shared by both class I and II mammalian enzymes. HDAC6 and 10 have two regions of homology with the class II catalytic site. The third class is the conserved NAD dependent Sir2 family of deacetylases (Grozinger & Schreiber 2002). Class I and II HDACs are inhibited by trichostatin A (TSA), suberoylanilide hydroxamic acid (SAHA) and related compounds. The deacetylase activity of the Sir2 family is not inhibited by these compounds. There is increasing evidence that several HDACs are not redundant in function. While the levels of HDACs 1, 2, 3, 5, 6, 7 and 10 are about the same in normal

HDAC INHIBITORS AND CANCER

TABLE 1

271

Histone deacetylases (HDACs)a

Class

Enzymeb

Size (amino acids)

Human chromosome locus

1 (Rpd3-like)

HDAC1 HDAC2 HDAC3 HDAC8 HDAC4 HDAC5 HDAC6 HDAC7 HDAC9 HDAC10

482 488 428 377 1084 1122 1215 855 1011 4700

1p34 6q21 5q31 xq13 2q37.2 17q21 Xp11.23 12q13.1 7p21-p15 22q13.31

II (Hda1-like)

a

Refer to text for references. HDAC11 has homologies to both HDAC class I and II HDACs.

b

tissues, HDAC4 is expressed in embryonic but not adult muscle tissue (Turner 2002, Khochbin et al 2001). Class I HDACs are found almost exclusively in the nucleus, while class II HDACs shuttle between the nucleus and cytoplasm in response to certain cellular signals. HDACs 1 and 3 can deacetylate all four core histones at all lysines tested, although with di¡erent degrees of e⁄ciency. HDAC6 has greater deacetylation activity of histone 4, lysine 5 and lysine 8. HDAC5 complexes with CREB1 and may be involved in long-term memory (Guan et al 2002). HDACs do not bind directly to DNA but are recruited to DNA by protein complexes which di¡er in their subunits (Khochbin et al 2001). For example, HDAC1 and 2 have been found in complexes with Sir3, NuRD and CoREST. The N-terminal tails of HDAC4, 5 and 7 interact with the MEF2 (myogenic transcription factor) which is involved in muscle di¡erentiation (McKinsey et al 2001). That the di¡erent HDACs have di¡erent functions, albeit not yet well deciphered, makes it desirable to develop selective inhibitors. The HDAC homologue (HDLP) structure (Finnin et al 1999) that interacts directly with TSA and SAHA is generally conserved in all HDACs. There is less conservation in surrounding residues, which may allow for the development of selective inhibitors. Notably, there is divergence in the region of tyrosine 91 of HDLP and this tyrosine is poorly conserved among the human HDACs. Interestingly, selective inhibitors of NAD-dependent Sir2 deacetylase activity have been discovered and found not to cause an accumulation of acetylated histone (Grozinger & Schreiber 2002). This ¢nding suggests that histones are not a primary substrate for this class of deacetylases.

272

MARKS ET AL

Histone deacetylase inhibitors Several classes of agents have been shown to inhibit the activity of puri¢ed HDAC enzymes. HDAC inhibitors (Table 2) reported to date can be divided into several structural classes including hydroxamates, cyclic peptides, aliphatic acids, benzamides and electrophilic ketones. TSA was the ¢rst natural product hydroxamate discovered to inhibit HDACs directly (Yoshida et al 1990). SAHA, which contains relatively less structural complexity, was found to be a nanomolar inhibitor of partially puri¢ed HDAC (Richon et al 1998). M-carboxycinnamic acid bishydroxamide (CBHA) has also been shown to be a potent HDAC inhibitor. Recently, several CBHA derivatives have been described. These include LAQ-824 and sulfonamide hydroxamic acids (Curtin et al 2002, Bouchain et al 2003). Other hydroxamates of interest include oxam£atin and scriptaid (Kim et al 1999, Su et al 2000). Cyclic tetrapeptides, which constitute the most structurally complex class of HDAC inhibitors, include depsipeptide, apicidin and the CHAPs, are active at nanomolar levels (Singh et al 2002, Furumai et al 2001, 2002). Their general utility in the treatment of disease remains unproven, with the possible exception of depsipeptide. The aliphatic acids are the least potent class of HDAC inhibitor, possessing millimolar levels of activity. These agents include valproic acid (VA) and phenylbutyrate (PB) (Boivin et al 2002, Phiel et al 2001). The benzamide class, which is in general less potent than the corresponding hydroxamates and cyclic tetrapeptides, include MS-275 and CI-994 (Saito et al 1999, Prakash et al 2001). These agents typically possess micromolar levels of HDAC inhibiting activity. CBHA-derived benzamides have recently been reported, which have about the same potency as other benzamides. The electrophilic ketone is a new and growing class of HDAC inhibitor. These agents include various tri£uoromethyl ketones and a-ketoamides. These agents, like the benzamides, possess micromolar level inhibitory activities of HDAC preparations (Frey et al 2002). The structural details of HDAC inhibitor/enzyme interactions have been elucidated by Finnin et al (1999). The crystal structure of HDLP, a homologue of mammalian HDAC, was solved with HDAC inhibitors TSA and SAHA. The structure^activity relationship (SAR) of the HDAC inhibitor classes reported to date further validate key features found in the X-ray crystal structure. Particularly, the direct interaction of the hydroxamic group, with the zinc at the base of the pocket-like catalytic site, appears to be a prerequisite to inhibitory activity. Given the X-ray crystallographic ¢ndings and the SAR of the various inhibitor classes, the structural characteristics of most active HDAC inhibitors reported to date can be summarized as follows: a hydroxamic acid group, a spacer

HDAC INHIBITORS AND CANCER

TABLE 2

Histone deacetylase inhibitorsa

Class

Compound

Hydroxamate

Trichostatin A (TSA) Suberoyl anilide hydroxamic acid (SAHA)

CBHA

LAQ-824

Sulfonamide hydroxamic acids

Oxam£atin

Scriptaid

Cyclic Tetrapeptide

273

Depsipeptide (FK-228)

Structure

274

TABLE 2

MARKS ET AL

(cont.)

Class

Compound

Cyclic Tetrapeptide

Apidicin

TPX-HA analogue (CHAP)

Aliphatic Acid

Valproic Acid

Phenyl Butyrate

Benzamide

MS-275

CI-994

Electrophilic Ketone

Tri£uoromethyl Ketones

Alpha-ketoamides

a

For references see text.

Structure

HDAC INHIBITORS AND CANCER

275

aliphatic chain of 5-6 carbons, a second carbonyl polar group and a protein surface recognition domain, such as a phenyl amino group. Activity of histone deacetylase inhibitors HDAC inhibitors cause the induction of di¡erentiation, growth arrest and/or apoptosis in a broad spectrum of transformed cells in culture and of tumours in animals, including both haematological cancers and solid tumours (Marks et al 2001, Kelly et al 2002a). These inhibitory e¡ects are believed to be due, in part, to accumulation of acetylated proteins, such as nucleosomal histones, which play a major role in regulation of gene transcription (Spotswood & Turner 2002, Jenuwein & Allis 2001, Zhang & Reinberg 2001). A model for the antitumour e¡ects of HDAC inhibitors is that the accumulation of acetylated histones leads to the activation (and repression) of transcription of genes, which causes inhibition of tumour cell growth. Expression pro¢ling of cells cultured with HDAC inhibitors support this model, in that they demonstrate that the expression of a small number of genes (2^5% of the expressed genes) is altered (activated or repressed) (Butler et al 2002, Suzuki et al 2002, Della Ragione et al 2001, Van Lint et al 1996). The mechanism of repression is not well understood. Gene repression may result from either direct or indirect e¡ects of histone acetylation or the increase in acetylation of proteins other than histones, such as, transcription factors. HDAC inhibitors do alter the acetylated state of proteins that are important regulators of gene transcription, such as the transcription factors NF-kB, GATA1, CREB and MyoD; regulators of cell cycle progression (e.g. pRB and p53); hormone receptors such as glucocorticoids and thyroid hormone receptors; and chaperone proteins such as HSP90 (Lagger et al 2002, Kouzarides 2000). Culture of normal and tumour cells with HDAC inhibitors causes the accumulation of acetylated histones, H4, H3, H2A and H2B (Richon et al 1998, Marks et al 2001). Tumour cells appear to be more sensitive to growth inhibition and apoptotic e¡ects of these agents than normal cells (Qiu et al 1999). Parenthetically, the accumulation of acetylated histone in peripheral mononuclear cells and in tumour tissue has been used as a marker of biological activity in clinical trials (Kelly et al 2002b). An important characteristic of HDAC inhibitors is their selectivity in altering gene expression in transformed cells. As indicated above, fewer than 2^5% of expressed genes are up- or down-regulated by more than twofold within 1^6 h of culture with agents such as TSA or SAHA. One of the most commonly induced genes by HDAC inhibitors is the cell cycle kinase inhibitor p21Waf1. Induced expression of p21Waf1 correlates with an increase in the acetylation of histones in the promoter region (Richon et al 2000), suggesting that p21Waf1 is a direct target

276

MARKS ET AL

gene for HDAC inhibitors. Further, HDAC1-de¢cient embryonic stem cells show reduced proliferation rates, which correlate, in part, with elevated levels of p21Waf1 (Lagger et al 2002). The induction of p21Waf1 is not required for HDAC inhibitor-induced apoptosis. Indeed, inhibition of SAHA induced p21Waf1 expression leads to increased apoptosis (Vrana et al 1999). In addition to p21Waf1, several genes that are commonly induced or repressed by HDAC inhibitors may play important roles in HDAC inhibitor induced antiproliferative response. De¢ning downstream e¡ects of HDAC inhibition could provide mechanistic rationale for combination therapies with HDAC inhibitors. Taken together, HDAC inhibition leads to changes in transcriptional activity and the activity of proteins through the increased acetylation of proteins. SAHA has been found to be synergistic or additive with a number of anticancer agents, including radiation, anthracyclines, £avopiridol, gleevec and all-trans retinoic acid (see reviews Marks et al 2001, Kelly et al 2002a) in tumour cells in culture. Animal studies show that HDAC inhibitors cause growth arrest of a wide variety of tumours in vivo without toxicity (Marks et al 2001). While the selective inhibitory e¡ects on transformed cells compared to normal cells are not completely understood, they do not appear to be due to a di¡erence in the ability to inhibit HDAC activity. Accumulation of acetylated histones occurs in both normal and transformed cells (Qiu et al 1999, Richon et al 2000, Scott et al 2002, Vrana et al 1999). HDAC inhibitors fall into a class of agents that target an activity (reversible protein acetylation) that occurs in all cells rather than targeting an abnormal process in the cancer cell. The therapeutic index may result from the di¡erential response of the cancer cells and normal cells to inhibition of HDAC activity.

HDAC inhibitors in clinical trials Several HDAC inhibitors have entered clinical trials. These include the short fatty acids phenylacetate (PA), PB, pivaloyloxymethylbutyrate and VA; the hydroxamic acid-based HDAC inhibitors SAHA, pyroxamide and LAQ824; the cyclic peptide, depsipeptide; the benzamide, MS-275; and N-acetyl amide (CI-994). PA and PB are two of the short fatty acids that have been extensively studied, however, only PA is approved for the use in the humans for certain non-cancer applications. In phase I studies in humans, high doses of PA produced reversible confusion and lethargy while lower doses were well tolerated and produced palliative e¡ects although no objective tumour regression were observed. Further, phase II studies in malignant glioblastoma have only shown modest activity in this disease (Chang et al 1999).

HDAC INHIBITORS AND CANCER

277

PB is a precursor of PA after oxidization in the liver and kidney. This formulation is better tolerated and has been shown to inhibit histone acetylation, modify lipid metabolism and activate peroxisome proliferation activator receptor (Kelly et al 2002a). Somnolence and confusion were also observed with prolonged intravenous infusion of PB (Carducci et al 1997). Clinical studies in patients with leukaemia and myelodysplastic disorder did have modest improvement in their disease and in patients with solid tumours, a delay in the progression of the cancer was observed as well as improvement in cancer-related symptoms (Carducci et al 1997). An oral formulation of PB showed excellent bioavailability (78%) and biologically active plasma concentrations (0.5 mM) could be easily achieved in patients (Gilbert et al 2001). Major adverse e¡ects of oral PB were nausea, vomiting, dyspepsia, confusion, peripheral oedema, fatigue and hypocalcaemia but were all manageable. Cytostatic e¡ects were observed that were dose dependent but no objective tumour regressions have been documented. Interestingly, PB has been used with all-trans-retinoic acid in patients with acute promyelocytic leukemia (APL) that are refractory to all-transretinoic acid to restore the sensitivity to retinoid resulting in clinical improvement in one of the ¢ve patients (Warrell et al 1998). PB is thought to inhibit the corepressor complex that contains HDAC for the oncoprotein that is encoded by one of the translocation-generated fusion genes in APL, PML-RAR. Other trials using retinoic acid and demethylation agents are ongoing to exploit the cell modulating e¡ects of PB. Pivaloyloxymethylbutyrate is a butyric acid derivative that has been evaluated in phase II studies. Forty-seven patients with refractory non-small cell lung cancer were treated with a 6 h infusion for three consecutive days every 3 weeks. Two patients had partial responses and 18 had stable disease (Keer et al 2001). Pivaloyloxymethylbutyrate was well tolerated without myelosuppression and combination trials with chemotherapy are now ongoing. Valproic acid is a well tolerated anti-epileptic agent and has recently been shown to be a potent inhibitor of HDACs (Gottlicher et al 2001, Phiel et al 2001). Little is known about the e¡ects of valproic acid in cancer and clinical trials are now ongoing. SAHA and Pyroxamide are two hydroxamic acid derivatives that are potent inhibitors of HDAC that have completed phase I trials and are entering phase II clinical testing. Results of the phase I trials with the intravenous formulation of SAHA established that SAHA could be administered safely, could inhibit HDAC in normal and malignant cells and had antitumour activity in solid and haematologic tumours (Fig. 1) (Kelly et al 2002b, 2003). An oral formulation of SAHA has good bioavailability and a dose proportional increase in AUC and Cmax was observed with increasing doses of SAHA. The presumed delayed absorption through the gastrointestinal tract has resulted in prolonged therapeutic plasma

278

MARKS ET AL

FIG. 1. Patient with metastatic larynx cancer to the lung showing tumour necrosis and regression after 4 months of oral SAHA therapy.

concentrations and accumulation of acetylated histones in peripheral mononuclear cells for up to 10 h has been observed (Kelly et al 2002a,b). Dose-limiting toxicities of anorexia, diarrhoea, dehydration and fatigue were encountered at higher doses when SAHA was administered continuously on a daily basis. However, lower oral daily doses of SAHA have been well tolerated in patients for over one year without loss of the biological activity of the inhibitor. Antitumour e¡ects with measurable disease regression and symptomatic improvement in cancer-related symptoms have been seen in patients with lymphoma, larynx cancer, renal cell carcinoma, bladder cancer and thyroid cancer including a complete response in a patient with B cell lymphoma. Phase II studies in head and neck cancers and mycosis fungoides are ongoing and preliminary results in patients with refractory T cell lymphomas have been encouraging. Depsipeptide given as a 4 h intravenous infusion on days 1 and 5 of a 21 day cycle has been well tolerated with fatigue, nausea, vomiting and thrombocytopaenia observed at higher dose levels (Sandor et al 2002). Minimal cardiac toxicity was observed even though pre-clinical models predicted this could be a dose-limiting toxicity (Sandor et al 2002). An accumulation of acetylated histones in mononuclear cells from patients was seen post-treatment at the maximally tolerated dose. One patient in this phase I trial with metastatic renal cell carcinoma had a partial response that lasted for 6 months. Additional studies in patients having T cell lymphomas have shown hyperacetylation of histones in primary lymphoma cells that were associated with a clinical improvement of the disease (Piekarz et al 2001). One patient with peripheral T cell lymphoma had a complete response while two patients with Se¤ zary syndrome and one with cutaneous T cell lymphoma had partial responses. These promising results have prompted phase II studies in patients with T cell lymphomas that are ongoing.

HDAC INHIBITORS AND CANCER

279

Acknowledgements Memorial Sloan-Kettering Cancer Center and Columbia University jointly hold patents on the hydroxamic acid-based hybrid polar compounds, including SAHA, pyroxamide and CBHA, which are exclusively licensed to Aton Pharma Inc. of which PAM is a founder and member of the Board of Directors. Both institutions and founder have an equity position in Aton Pharma. PAM and WKK’s research cited in this review was supported by grants from the National Cancer Institute (USA), The Japan Foundation for the Promotion of Cancer Research, The DeWitt Wallace Fund for the Memorial Sloan-Kettering Cancer Center, the Kleberg Foundation, CaP Cure, Susan and Jack Rudin Foundation and the David Koch Prostate Cancer Research Fund.

References Boivin AJ, Momparler LF, Hurtubise A, Momparler RL 2002 Antineoplastic action of 5-aza-2’deoxycytidine and phenylbutyrate on human lung carcinoma cells. Anticancer Drugs 13:869^ 874 Bouchain G, Leit S, Frechette S et al 2003 Development of potential antitumor agents. Synthesis and biological evaluation of a new set of sulfonamide derivatives as histone deacetylase inhibitors. J Med Chem 46:820^830 Butler LM, Zhou X, Xu W-S et al 2002 The histone deacetylase inhibitor SAHA arrests cancer cell growth, up-regulates thioredoxin-binding protein-2, and down-regulates thioredoxin. Proc Natl Acad Sci USA 99:11700^11705 Carducci M, Bowling MK, Eisenberger MA et al 1997 Phenylbutyrate (PB) for refractory solid tumors: phase I clinical and pharmacologic evaluation of intravenous and oral PB. Anticancer Res 17:3972^3973 Chang SM, Kuhn JG, Robins HI et al 1999 Phase II study of phenylacetate in patients with recurrent malignant glioma: a North American Brain Tumor Consortium report. J Clin Oncol 17:984^990 Curtin ML, Garland RB, Heyman HR et al 2002 Succinimide hydroxamic acids as potent inhibitors of histone deacetylase (HDAC). Bioorg Med Chem Lett 12:2919^2923 Della Ragione F, Criniti V, Della-Pietra V 2001 Genes modulated by histone acetylation as new e¡ectors of butyrate activity. FEBS Lett 499:199^204 De Ruijter AJ, Van Gennip AH, Caron HN, Kemp S, Van Kuilenburg AB 2003 Histone deacetylases: characterisation of the classical HDAC family. Biochem J 370:737^749 Finnin MS, Donigian JR, Cohen A et al 1999 Structures of a histone deacetylase homologue bound to the TSA and SAHA inhibitors. Nature 401:188^193 Freiman RN, Tjian R 2003 Regulating the regulators: lysine modi¢cations make their mark. Cell 112:11^17 Frey RR, Wada CK, Garland RB et al 2002 Tri£uoromethyl ketones as inhibitors of histone deacetylase. Bioorg Med Chem Lett 12:3443^3447 Furumai R, Komatsu Y, Nishino N, Khochbin S, Yoshida M, Horinouchi S 2001 Potent histone deacetylase inhibitors built from trichostatin A and cyclic tetrapeptide antibiotics including trapoxin. Proc Natl Acad Sci USA 98:87^92 Furumai R, Matsuyama A, Kobashi N et al 2002 FK228 (depsipeptide) as a natural prodrug that inhibits class I histone deacetylases. Cancer Res 62:4916^4921 Gilbert J, Baker SD, Bowling MK et al 2001 A phase I dose escalation and bioavailability study of oral sodium phenylbutyrate in patients with refractory solid tumor malignancies. Clin Cancer Res 7:2292^2300 Gottlicher M, Minucci S, Zhu P et al 2001 Valproic acid de¢nes a novel class of HDAC inhibitors inducing di¡erentiation of transformed cells. EMBO J 20:6969^6978

280

MARKS ET AL

Grozinger CM, Schreiber SL 2002 Deacetylase enzymes: biological functions and the use of small-molecule inhibitors. Chem Biol 9:3^16 Guan Z, Giustetto M, Lomvardas S et al 2002 Integration of long-term-memory-related synaptic plasticity involves bidirectional regulation of gene expression and chromatin structure. Cell 111:483^493 Jenuwein T, Allis CD 2001 Translating the histone code. Science 293:1074^1080 Jones PA, Baylin SB 2002 The fundamental role of epigenetic events in cancer. Nat Rev Genet 3:415^428 Keer H, Reid T, Sreedharan S 2001 Pivanex activity in refractory non-small cell lung cancer, a phase II study. Proc Am Soc Clin Oncol 314a Kelly WK, O’Connor OA, Marks PA 2002a Histone deacetylase inhibitors: from target to clinical trials. Expert Opin Investig Drugs 11:1695^1713 Kelly WK, O’Connor O, Richon VM et al 2002b A phase I clinical trial of an oral formulation of the histone deacetylase inhibitor of suberoylanilide hydroxamic acid (SAHA). 14th EORTCNCI-AACR, November 2002, Frankfurt (abstr 286, available in Eur J Cancer 38:88) Kelly W, Richon VM, O’Connor O et al 2003 Phase I clinical trial of histone deacetylase inhibitor: suberoylanilide hydroxamic acid (SAHA) administered intravenously. Clin Canc Res 9:3578^3588 Khochbin S, Verdel A, Lemercier C, Seigneurin-Berny D 2001 Functional signi¢cance of histone deacetylase diversity. Curr Opin Genet Dev 11:162^166 Kim YB, Lee KH, Sugita K, Yoshida M, Horinouchi S 1999 Oxam£atin is a novel antitumor compound that inhibits mammalian histone deacetylase. Oncogene 18:2461^2470 Kouzarides T 2000 Acetylation: a regulatory modi¢cation to rival phosphorylation? EMBO J 19:1176^1179 Lagger G, O’Carroll D, Rembold M et al 2002 Essential function of histone deacetylase 1 in proliferation control and CDK inhibitor repression. EMBO J 21:2672^2681 Marks P, Rifkind RA, Richon VM, Breslow R, Miller T, Kelly WK 2001 Histone deacetylases and cancer: causes and therapies. Nat Rev Cancer 1:194^202 McKinsey TA, Zhang CL, Olson EN 2001 Control of muscle development by dueling HATs and HDACs. Curr Opin Genet Dev 11:497^504 Phiel CJ, Zhang F, Huang EY, Guenther MG, Lazar MA, Klein PS 2001 Histone deacetylase is a direct target of valproic acid, a potent anticonvulsant, mood stabilizer, and teratogen. J Biol Chem 276:36734^36741 Piekarz RL, Robey R, Bakke S, Sandor V, Wilson W, Bates S 2001 Histone deacetylase inhibitor for the treatment of peripheral or cutaneous T-cell lymphoma. ASCO 232b Polevoda B, Sherman F 2002 The diversity of acetylated proteins. Genome Biol 3:REVIEWS0006 Prakash S, Foster BJ, Meyer M et al 2001 Chronic oral administration of CI-994: a phase 1 study. Invest New Drugs 19:1^11 Qiu L, Kelso MJ, Hansen C, West ML, Fairlie DP, Parsons PG 1999 Anti-tumour activity in vitro and in vivo of selective di¡erentiating agents containing hydroxamate. Br J Cancer 80:1252^1258 Richon VM, Emiliani S, Verdin E et al 1998 A class of hybrid polar inducers of transformed cell di¡erentiation inhibits histone deacetylases. Proc Natl Acad Sci USA 95:3003^3007 Richon VM, Sandho¡ TW, Rifkind RA, Marks PA 2000 Histone deacetylase inhibitor selectively induces p21WAF1 expression and gene-associated histone acetylation. Proc Natl Acad Sci USA 97:10014^10019 Saito A, Yamashita T, Mariko Y et al 1999 A synthetic inhibitor of histone deacetylase, MS27-275, with marked in vivo antitumor activity against human tumors. Proc Natl Acad Sci USA 96:4592^4597

HDAC INHIBITORS AND CANCER

281

Sandor V, Bakke S, Robey RW et al 2002 Phase I trial of the histone deacetylase inhibitor, depsipeptide (FR901228, NSC 630176), in patients with refractory neoplasms. Clin Cancer Res 8:718^728 Scott GK, Marden C, Xu F, Kirk L, Benz CC 2002 Transcriptional repression of ErbB2 by histone deacetylase inhibitors detected by a genomically integrated ErbB2 promoterreporting cell screen. Mol Cancer Ther 1:385^392 Singh SB, Zink DL, Liesch JM et al 2002 Structure and chemistry of apicidins, a class of novel cyclic tetrapeptides without a terminal alpha-keto epoxide as inhibitors of histone deacetylase with potent antiprotozoal activities. J Org Chem 67:815^825 Spotswood HT, Turner BM 2002 An increasingly complex code. J Clin Invest 110:577^582 Su GH, Sohn TA, Ryu B, Kern SE 2000 A novel histone deacetylase inhibitor identi¢ed by high-throughput transcriptional screening of a compound library. Cancer Res 60: 3137^3142 Suzuki H, Gabrielson E, Chen W et al 2002 A genomic screen for genes upregulated by demethylation and histone deacetylase inhibition in human colorectal cancer. Nat Genet 31:141^149 Timmermann S, Lehrmann H, Polesskaya A, Harel-Bellan A 2001 Histone acetylation and disease. Cell Mol Life Sci 58:728^736 Turner BM 2002 Cellular memory and the histone code. Cell 111:285^291 Van Lint C, Emiliani S, Verdin E 1996 The expression of a small fraction of cellular gene is changed in response to histone hyperacetylation. Gene Expr 5:245^253 Vrana JA, Decker RH, Johnson CR et al 1999 Induction of apoptosis in U937 human leukemia cells by suberoylanilide hydroxamic acid (SAHA) proceeds through pathways that are regulated by Bcl-2/Bcl-XL, c-Jun, and p21CIP1, but independent of p53. Oncogene 18:7016^7025 Wang C, Fu M, Mani S, Wadler S, Senderowicz AM, Pestell RG 2001 Histone acetylation and the cell-cycle in cancer. Front Biosci 6:D610^D629 Warrell RP Jr, He LZ, Richon V, Calleja E, Pandol¢ PP 1998 Therapeutic targeting of transcription in acute promyelocytic leukemia by use of an inhibitor of histone deacetylase J Natl Cancer Inst 90:1621^1615 Yoshida M, Kijima M, Akita M, Beppu T 1990 Potent and speci¢c inhibition of mammalian histone deacetylase both in vivo and in vitro by trichostatin A. J Biol Chem 265:17174^17179 Zhang Y, Reinberg D 2001 Transcription regulation by histone methylation:interplay between di¡erent covalent modi¢cations of the core histone tails. Genes Dev 15:2343^2360

DISCUSSION Pelicci: Did you see any correlation between the levels of histone acetylation in the peripheral blood and the responses? Marks: We have looked at that and we have not seen any striking correlation. Our database is still relatively small. We have a total of perhaps 100 patients. The other thing is that we don’t ¢nd this to be a very quantitative assay. Doing a repeat on exactly the same sample three or four times, there is su⁄cient variability that it is very di⁄cult. Pelicci: There are now antibodies that work nicely by FACS analysis. Marks: I know about them, and the company is using them. Pelicci: Do you see hyperacetylation in every treated patient?

282

DISCUSSION

Marks: Yes, at every dose. The clinicians started at 200 mg O.D. and at that dose there was detectable accumulation of acetylated histones in peripheral mononuclear cells at 2 h after ingestion of the pill. Castronovo: Is there a dose e¡ect in regression of the tumour in your in vivo animal models? Marks: Yes, there was a dose-related response to SAHA in tumour-bearing mice and rats. For example, in human prostate cancer xenografts in nude mice, we tried three doses. At 25 mg/kg there was a partial response. At 50 mg/kg, there was complete cessation of growth and regression of the tumour. Castronovo: In your animal models, if you stop the drug do the tumours start again? Marks: We started the clinical studies, and in these we have seen durable responses in the patients who do respond. The patient who has been on the drug for longest period has been taking it for 16 months (24 months in August 2003). It is well tolerated and the patient has stable disease. Castronovo: So do you see this as a chronic treatment, like anti-angiogenesis treatment might be? Marks: My personal opinion is that I do not see any reason to stop the drug as long as the patient is not experiencing toxicity and is responding. Mahadevan: We have heard at this conference about histone-mediated and nonhistone-mediated mechanisms by which acetylation can act. In the genes that you looked at, particularly TBP2, that comes up in 30 min, is it possible to correlate histone acetylation at the gene with that very quick activation? Marks: Preliminary data suggests there is no accumulation of acetylated histones in region of the promoter of the TBP2 gene. Mahadevan: What about at p21? Marks: For p21 we do see accumulation of acetylated histones in the promoter region within 2 hours. Dalia Cohen has also shown this. There are probably di¡erent mechanisms for altering gene expression in response to inhibition of HDACs. We have transformed cell lines in which we can induce apoptosis, in which there is no evidence of induction of p21. We have studied one cell line in which SAHA rapidly induces apoptosis in which we don’t see induction of p21 or TBP2. But we do see changes in expression of other genes that suggest a pathway for apoptosis. Mahadevan: This is in striking contrast to what we see with those oncogenes I referred to earlier. We see acetylation very quickly but we don’t see any transcriptional up-regulation. Marks: Are these oncognes that are silent to begin with? Mahadevan: Yes. We treat synchronized quiescent cells with TSA. Within 15 min the oncogenes become acetylated, but they don’t become activated. They need another signal to activate transcription, and that is the phosphorylation signal

HDAC INHIBITORS AND CANCER

283

delivered by MAP kinase cascades which we know produces phosphorylation of transcription factors, co-activators as well as histone H3. Atadja: Some of your clinical work, as well as the NCI work from Susan Bates’ group, has shown e¡ects on cutaneous T cell lymphomas. Is there any mechanistic reason why we are getting these results? Marks: We think there is. We have looked at cells cultured from some of these patients. In preliminary studies, these cells in culture rapidly undergo apoptosis. We are now looking to see what genes are altered in expression. In culture they are relatively sensitive to SAHA and undergo apoptosis. Ott: Can you translate directly your data from cell lines into predictions about which tumours are going to respond well to SAHA? For example, speci¢cally the neuroblastoma that you showed. Marks: We have some evidence to suggest that this may be the case. Unfortunately, we haven’t been able to do a clinical trial in children with neuroblastoma. We are hoping to start one. I can’t answer that question. For what it is worth we have seen anticancer activity in patients with bladder cancer which corresponds to the T24 cell lines. We have seen it in non-Hodgkins lymphoma where we have also looked at cell lines. Oettinger: With the mouse question, I can understand why you can’t continue to study with the wild-type animals, but why can’t you carry on with the ones that have been treated and gone into remission? You could then see what the e¡ects of treatment cessation would be. Marks: We plan to do such studies. Atadja: You need to have a control. Pelicci: In Phase I clinical trials the idea is to use these drugs in chronic treatment as opposed to giving maximum tolerated doses in a short period of time. Why do you think this chronic treatment is better? Marks: I can’t answer that with any precision. But if you want to translate the preclinical studies with these agents, there is an optimal concentration at which they will induce G1 arrest or apoptosis. If you give too little you don’t get the e¡ect; if you give too much it is toxic. Translating this, we were looking for a dose that had good antitumour e¡ects and was well tolerated. Having said this, the clinical trials are now going forward to try to determine the maximum tolerated dose. Baylin: Are there any toxicities? Marks: The toxicity seen includes dehydration, diarrhoea and a sense of fatigue. These are all reversible when the drug is stopped. The dehydration and the diarrhoea can be controlled. Verdin: Does the drug cross the blood^brain barrier? Marks: Yes. We have been doing some non-cancer applications. We have just published a study (Hockly et al 2003) with Gill Bates at Kings on her transgenic

284

DISCUSSION

mouse Huntington’s disease model, in which SAHA signi¢cantly slows the progression of the disease. Seto: You said that SAHA inhibits HDACs 1, 3, 4, 6 and 9. Does this mean SAHA does not inhibit the other HDACs? Marks: No, it means that we haven’t puri¢ed the others and tried them. We are working on the others. We have no evidence that SAHA is selective in inhibition of HDAC class I or class II. Seto: Your statement that only 2% of the genes are up-regulated by SAHA seems premature since you have not really tried di¡erent combinations of treatments with varying time point and dosage, etc. If you treat cells with TSA in combination with serum, for example, a totally di¡erent set of genes can be activated. As technology gets more sophisticated, I predict we will see a lot more genes that can be induced or regulated by SAHA. Marks: I have to qualify the statement that fewer than 2% of the genes are altered in their transcription, in terms of the time. We are looking at 30 min and 1 h. If you look at 6 h or 10 h you see many more genes altered in their expression. Whether this represents downstream e¡ects of the early genes, or delayed e¡ects of SAHA, is unclear. We are also using the arbitrary cut-o¡ of a twofold change in expression either increased or decreased. Verdin: This ¢gure of 2% re£ects the genes that are activated solely by HDAC inhibitors. From Warner Greene’s talk, for example, we saw that if we treat cells with TNF there is another layer of induction. This does not imply that HDACs or HATs are only involved in regulation of 2% of the genes. I think it is probably 99% of genes that are modulated one way or another by HDACs. Marmorstein: Do you have a feeling for what proportion of patients don’t respond to SAHA? Marks: In the phase I clinical trials we had 39 patients. This included patients in the trial who received what we presume are sub-therapeutic doses. If you take the patients who received 400 or 600 mg SAHA per day, it is probably 12 patients, and of these 4 or 5 showed signi¢cant anticancer e¡ects. They were all patients with very advanced tumours. These are drugs that seem to be well tolerated and have an activity against tumours that no longer respond to standard therapies. These drugs represent a potentially valuable addition to our armamentarium. Reference Hockly E, Richon VM, Woodman B et al 2003 Suberoylanilide hydroxamic acid, a histone deacetylase inhibitor, ameliorates motor de¢cits in a mouse model of Huntington’s disease. Proc Natl Acad Sci USA 100:2041^2046

General discussion III PML-RARa hypermethylation in leukaemia Pelicci: I would like to share our recent data on the mechanisms of leukaemogenecity of fusion proteins. In particular, I will focus on the PMLRARa fusion protein, which forms as a consequence of a translocation, and which is thought to be su⁄cient to initiate the leukaemogenetic transition in transgenic animal experiments. This protein is made of relevant domains of PML (a nuclear matrix protein) and RARa (retinoic acid receptor a, a ligand-dependent transcription factor). To understand the mechanisms by which this fusion protein induces leukaemia, we ¢rst characterized its biological activities by ectopic expression within normal haemopoietic precursors. First, PML-RARa prolongs survival, making cells resistant to stresses such as exposure to DNA damage agents. Second, PML-RARa-expressing haemopoeitic precursors have increased ability to self-renew. A third phenotype is that the fusion proteins are able to block haemopoietic di¡erentiation. Unfortunately, we don’t have individual mutants of PML-RARa which segregate these three functions, so we don’t know the relative contribution of these three phenotypes to the leukaemogenicity of the fusion protein. However, they are attractive phenotypes to justify the leukaemia phenotype, which consists of the accumulation of long-living haemopoeitic precursors blocked at certain developmental stages. Our work stems from an observation we made a couple of years ago that PMLRARa is complexed with HDAC. If we mutatgenize a couple of the amino acids in the RARa component we can generate a mutant that is unable to recruit the HDAC complex. The fusion protein is then unable to mediate any of the biological activities which I have just described. This suggests that the complex formation with HDAC is critical for the fusion protein’s properties. We went on to characterize this complex and we now know that other enzymes are recruited by PMA-RARa, including DNA methyl transferases (DNMTs) and histone methyl transferases (HMTs). The recruitment of these proteins is mediated by both the RARa and PML components of the fusion protein. Indeed, we have recently isolated the PML complex and found, in the same complex, HDAC, DNMTs and HMTs. Then we tried to determine the role of these enzymes with respect to the ability of the fusion protein to modify the biology of stem cells. We ¢rst looked at the e¡ect of PML-RARa on RA target genes, considering the fact that the fusion protein retains the DNA binding domain of RARa. 285

286

GENERAL DISCUSSION III

We studied RARba well known RA target geneand learned that PML-RARa binds to the RARb promoter, recruits DNMT and induces hypermethylation of the promoter, at speci¢c CpG islands. We have provided biological evidence that this hypermethylation is relevant to the ability of the fusion protein to produce leukaemia. Furthermore, we demonstrated recruitment of HMT in the PMLRARa target promoters and H3-K9 methylation at the same promoters. Recruitment of HMTs seems to be relevant to the ability of PML-RARa to repress transcription of this promoter, as shown by the fact that PML-RARa synergizes with HMT (but not with HMT mutated in the Set domain) in its ability to repress transcription. In summary, we think that the e¡ect of PMLRARa on the RA target genes is to induce heterochromatin formation, which is then responsible for transcriptional silencing and block of di¡erentiation. More recently, we have tried to identify genes which could be relevant targets for this e¡ect. We have identi¢ed a group of genes which are repressed by PML-RARa and are all involved in the control of di¡erentiation commitment in normal haematopoietic cells, thus suggesting that repression of these genes blocks the ability of the stem cell to commit to myelopoeitic di¡erentiation. However, PML-RARa alters expression of many other genes through indirect mechanisms, which are not mediated by the binding of PML-RARa to its direct target genes. We identi¢ed two mechanisms. One is the derepression of E2F/Rb target genes, by titration of HMTs, and the second is inhibition of p53. We were surprised to see that the stoichiometry of the interaction between PML-RARa and HMT is so high. In fact, if we ectopically express PML-RARa, HMT is localized within microscopic subdomains where typically PML-RARa accumulates. This suggests that PMLRARa could antagonize the function of HMTs. Since HMTs have been shown to participate in repression of E2F target promoters through Rb, we checked whether PML-RARa deregulates expression of Rb/E2F target genes. We have done an experiment on the cyclin E promoter which shows that PML-RARa indeed induces its activation. We then compared Rb/E2F and PML-RARa expression pro¢le screens and found about 40 genes that are down-regulated by Rb and are activated by E2F and PML-RARa. This suggests that PML-RARa activates the G1/S transition through the regulation of Rb/E2F target genes. This seems to be mediated by HMT sequestering. In fact, if we re-express HMT in leukaemic cells, we completely suppress the ability of these cells to self-renew and give inde¢nitive growth. This suggests that the functional de¢ciency of HMT is responsible for the extended self-renewal of these cells. In terms of p53, it is even easier. We demonstrated a couple of years ago that PML is a cofactor for p53 acetylation. Indeed, it forms a tri-complex with CBP and p300 in living cells. PML-RARa binds p53 through PML and this probably competes for p53 acetylation. It is indeed easy to co-precipitate PML-RARa and HDAC1, and expression of PML-RARa reduces acetylation and induces

GENERAL DISCUSSION III

287

destabilization of p53. This is biologically relevant. We do not see resistance to X-rays or other DNA-damaging agents in leukaemic cells which we generated in the PML7/7 context. If HDAC is so important for the mechanism of this fusion protein in generating leukaemias, then inhibitors of HDAC should revert to the leukaemic phenotype. Not surprisingly, we observed a striking e¡ect of HDAC inhibitors in vivo in the APL mice. As early as 48 h after injection of VPAa HDAC-inhibitorthere is massive apoptosis in the spleen. This drug prolongs survival of APL mice. The surprise, however, is that histone acetylase inhibitors do not a¡ect the biochemical and biological activities of PML-RARa. We reported in the last year that TSA does not a¡ect transcriptional repression induced by PMLRARa (Di Croce et al 2002). This is not a big surprise: PML-RARa induces hypermethylation, so why should transcriptional repression be sensitive to HDAC inhibition? The only e¡ect we see using the HDAC inhibitor is on p53. Indeed, in the presence of PML-RARa, treatment with HDAC inhibitor stabilized p53 and re-induced some acetylation of p53. In fact, there is a reactivation of the stress response. However, we do not think that this e¡ect of HDAC inhibitors on p53 is responsible for the apoptogenic activity of these drugs on APL cells. To show this we have generated APL mice in the p537/7 background. You can see that there is still apoptosis following VPA treatment. We searched for other drug targets, and identi¢ed a number of genes which encode for death receptors and their ligands. These genes therefore, could mediate apoptosis in these cells. Using dominant negative constructs against these molecules, we could prevent apoptosis induced by HDAC-inhibitors. I would like to emphasize that these genes are not a direct target of HDAC inhibitors. Indeed, we don’t see them being regulated in normal cells, and VPA does not induce any apoptosis in the normal haemopoietic precursors, which instead become highly sensitive in terms of apoptosis by VPA or TSA, following PML-RARa expression. There is no e¡ect on death receptors here. It is as if the transformation by PML-RARa induces sensitivity to VPA. We think this could be a more general mechanism that is also true for other tumour cells. Verdin: You are seeing the recruitment of a methylase by PML-RARa. Could this be one of the target genes that Steve Baylin was talking about? Is there any overlap between these target genes? Pelicci: I am sure it could be. Baylin: If you look at the promoters of those genes, do you know what you see in terms of speci¢c histone methyltransferases? Pelicci: Not yet. Baylin: So it could be a promoter e¡ect. Yao: Is the recruitment through the PML or RAR portion? Pelicci: It is through both. I think the high stability of this complex is due to the fact that there are multiple interactions. For example, if you try to precipitate

288

GENERAL DISCUSSION III

HDAC with the RARa, it is di⁄cult to see a complex, whereas it is easy to do this with PML-RARa. Verdin: Treatment with RA leads to the reformation of PML bodies. Have you looked at HDAC inhibition? If you use an HDAC inhibitor do you also see the same reformation of PML bodies? Pelicci: The story with RA is very important. It is telling us that the two mechanisms are completely di¡erent. RA destabilizes the PML-RARa^HDAC complex. Immediately the phenotype is reverted and the cells undergo di¡erentiation. The HDAC inhibitor has no e¡ect on this complex, which stays there: the PML nuclear bodies do not reform and the cells undergo apoptosis. I think RA targets the genetic lesion of APL mice, while HDAC inhibitors induce apoptosis through a totally di¡erent mechanism, which is downstream from the oncogene. Li: Does the fusion protein localize to the PML body? Pelicci: Yes, within so-called microspeckles, which are very small dots. Li: Does this disrupt the PML body? Pelicci: Yes. Li: Have you mapped in the fusion protein region(s) responsible for recruiting di¡erent components, such as the histone methylase? Pelicci: On the side of RARa, the AHT mutation abolishes its interaction with the complex. Within PML it is di⁄cult to identify the interacting region. As soon as you touch the coiled coil, for example, which is not directly involved in binding, you lose complex formation and don’t have any more recruitment. Reference Di Croce L, Raker VA, Corsaro M et al 2002 Methyltransferase recruitment and DNA hypermethylation of target promoters by an oncogenic transcription factor. Science 295:1079^1082

Index of contributors Non-participating co-authors are indicated by asterisks. Entries in bold indicate papers; other entries refer to discussion contributions.

F

A

*Fischle, W. 3

Allis, C. D. 3, 17, 18, 19, 20, 39, 42, 43, 44, 57, 59, 61, 74, 100, 111, 112, 160, 167, 169, 178, 193, 200, 221, 222 Atadja, P. 20, 57, 58, 59, 111, 112, 142, 144, 145, 168, 178, 180, 206, 218, 221, 246, 249, 266, 267, 268, 283

G Greene, W. C. 76, 131, 143, 180, 194, 208, 218, 219, 220, 221 Gu, W. 45, 61, 73, 75, 99, 114, 130, 143, 145, 169, 179, 197, 205, 206, 207, 220, 225, 247, 268

B Baylin, S. B. 42, 76, 143, 159, 160, 161, 168, 169, 205, 226, 234, 235, 236, 237, 283, 287 Berger, S. L. 19, 20, 45, 63, 73, 74, 75, 76, 129, 141, 142, 160, 161, 195, 222, 224, 245, 246 *Bhalla, K. 249 *Brooks, C. L. 197

H *Hazzalin, C. A. 102 *Henklein, P. 182 *Henry, K. W. 63 *Hetzer-Egger, C. 182 *Hoppe, G. J. 48 Hottiger, M. 98, 194, 223, 224, 225 *Hsu, M. 249 *Huang, J. 48 *Hubbert, C. 170

C Castronovo, V. 40, 41, 43, 58, 98, 167, 178, 179, 180, 207, 221, 222, 234, 245, 267, 268, 282 Chen, L.-f. 194, 208, 219 *Cheung, W. 3 *Chiao, J. H. 269 *Ciccone, D. 146 *Clayton, A. 102 Cohen, D. 38, 111, 112, 131, 221, 238, 245, 246, 247, 248, 267 Cole, P. 57, 99, 101, 112, 144, 160, 182, 219, 221, 235, 236, 268

J *Jacobs, S. 3 Jenuwein, T. 18, 22, 38, 39, 40, 41, 42, 43, 44, 45, 46, 74, 159, 162, 167, 222 K *Kaehlcke, K. 182 *Kasler, H. 115 *Kelly, W. K. 269 Khochbin, S. 17, 61, 62, 130, 144, 168, 179, 180, 193, 195, 206, 221 *Khorasanizadeh, S. 3 *Kovacs, J. J. 170 *Kwon, P. 249

D Denu, J. 60, 61, 100, 101, 113, 180 *Dequiedt, F. 115 *Dorr, A. 182 289

290

L *Lachner, M. 21 Li, E. 20, 40, 58, 59, 113, 130, 142, 145, 167, 195, 236, 247, 288 *Li, M. 197 *Lou, J. 197 M Mahadevan, L. C. 18, 38, 39, 101, 102, 111, 112, 113, 114, 129, 144, 145, 160, 195, 206, 207, 222, 237, 246, 247, 282, 283 Marks, P. A. 39, 41, 74, 75, 111, 129, 142, 144, 145, 206, 218, 219, 221, 235, 245, 247, 267, 269, 281, 282, 283, 284 Marmorstein, R. 18, 58, 78, 98, 99, 100, 101, 158, 161, 169, 194, 207, 266, 284 *McKinsey, T. A. 132 *Miller, T. 269 Moazed, D. 18, 48, 56, 57, 58, 59, 60, 61, 62, 99, 100, 158, 220, 225 N *Neal, Y. Y. 238 *Nikolaev, A. Y. 197 O Oettinger, M. 146, 158, 159, 160, 161, 162, 283 Olson, E. N. 111, 112, 129, 130, 132, 141, 142, 143, 144, 145, 178, 181, 219 Ott, M. 111, 112, 178, 182, 193, 194, 195, 196, 206, 218, 221, 247, 267, 283 P Pelicci, P. G. 205, 206, 281, 282, 283, 285, 287, 288 *Peters, A. H. F. M. 21

INDEX OF CONTRIBUTORS

*Richon, V. M. 269 *Rudner, A. D. 48 S *Schnolzer, M. 182 *Schotta, G. 22 Seto, E. 75, 130, 194, 195, 207, 218, 247, 284 T *Tanny, J. C. 48 *Thomson, S. 102 *Trogani, N. 249 Turner, B. 18, 19, 39, 40, 44, 46, 58, 59, 60, 61, 74, 75, 76, 100, 101, 113, 130, 131, 163, 166, 167, 168, 169, 205, 219, 220, 221, 222, 236, 237, 248, 267 V Verdin, E. 1, 18, 19, 37, 41, 42, 44, 56, 59, 60, 61, 73, 74, 99, 100, 112, 115, 129, 130, 131, 143, 159, 161, 166, 179, 180, 182, 206, 220, 221, 222, 234, 245, 246, 268, 283, 284, 287, 288 W *Wang, S. 238 *Wang, Y. 3 *Wyce, A. 63 Y Yao, T.-P. 73, 143, 170, 178, 179, 180, 181, 207, 287 Z

R *Remiszewski, S. 249

*Zeremski, M. 238 Zhou, M.-M. 43, 111, 160, 161, 182, 206, 219

Subject index Page numbers in italics indicate tables. A

B

acetyl-ADP-ribose 50, 61 acetyl^phos switch 11 ACTR 80 acute promyelocytic leukaemia histone lysine methylation 35 PB trials 277 ADH2 65, 67 AIB1 80 aliphatic acid 272, 274 Allfrey, Vince 1 alpha-ketoamides 272 structure 274 Angelman syndrome 14 antibodies acetylation work 160 problematic use 18^19, 38^39 antigen receptor assembly 146^158 chromatin modi¢cations 148^150, 155^156 regulation 147^148 antigen receptor genes, dimethyl H3-K9 152 antigen receptor signalling 117 apidicin 272 structure 274 apoptosis cardiomyopathy 144 chromatin 12^13 HDAC inhibitors 250, 270 histone phosphorylation 12 LAQ824 255^256 p53 acetylation 199 RhoB 241^242 stress 13 thymocytes 122, 125 ATF-2 80 autoimmunity, HDAC7 131 5-azacytidine 232, 235, 237

B cell lymphomas, SAHA 278 BAH domains 57 Beckwith-Wiedemann syndrome 14 benzamides 272, 274 binary switches, histone code 9^11 bladder cancer, SAHA 278 BRCA1, histone lysine methylation 35 Brg1 148^150, 152^153 BUZ ¢nger 173, 175 C Cabin-1 118, 125 calcineurin hypertrophy 133 kinase regulation 143 calcium/calmodulin dependent protein kinases, hypertrophy 133 calcium signalling, HDACs 133 calmodulin 120 cancer DNA methylation 227 drug combinations 247 epigenetic mechanisms 6, 236 gene silencing 14, 226^233, 250 HATs 270 HDAC inhibitors 14, 136, 139, 232, 235, 239, 250^251, 269^281 HDACs 227, 270 immediate-early genes 102 oncogenes 41, 234 pre-malignancy stages 232, 236 RhoB 241^242 cardiac hypertrophy, HDACs 132^141 caspase inhibitors, stress-related death 13 CBHA 272 structure 273 CBP 80 Fen1 acetylation 223^224 NF-kB 209 291

292

C BP (cont.) p53 198, 199, 200 RelA acetylation 210^211 tumorigenesis 198 CCAAT box 240^241, 246 CD5, negative selection 117 CD28, negative selection 117 CD40 ligand, negative selection 117 CDC2/cyclin B1 assembly, HDAC6 181 cell cycle H2B ubiquitylation 64 HDAC inhibitors 250 HDAC6 179 methylation 39 RAG2 161 trapoxin 239^240 cell death, HDAC inhibitors 137, 139 cell motility, HDAC6 172, 175 cell proliferation HP1 14 Pc 14 trapoxin 239^240 central tolerance 115 CHAP 272 structure 274 chromatin antigen receptors 148^150, 155^156 apoptosis 12^13 dissociation during mitosis, HP1 and Pc 17^18 epigenetic information 5^6 histone tail modi¢cations 22 mitotic 12^13 regulatory signals 6 silent domains 49^51 V(D)J recombination 148 chromodomains human genome 17 methyl lysine recognition and binding 8^9 MYST 16^17 chromosome fragility, p53 acetylation 199 CI-994 272, 276 structure 274 Co⁄n^Lowry cells 105, 106, 111 colon cancer, DNA methyltransferases (DNMTs) 230 congestive heart failure, HDAC5 144 CpG islands 227, 234, 237

SUBJECT INDEX

CREB phosphorylation 106 CREB1, HDAC5 271 cyclic tetrapeptides 272, 273^274 cyclin B1/CDC2 assembly, HDAC6 181 cyclin T1, Tat/TAR binding 187, 189 D DAC 231, 232 depsipeptide 272 clinical trials 276, 278^279 structure 273 desensitization 113 dimethyl H3-K9 150, 153^155 DNA-damage sensor 12 DNA methylation cancer 227 gene silencing 230^231 histone methylation 22, 24, 45, 230 promoter regions 236^237 X-inactivation 40 DNA methyltransferases (DNMTs) 227, 230, 235^236 DNMT1 236 Dot1 74, 75 double chromatin immunoprecipitation (double ChIP) 65, 73 E E1^3 enzymes 64 electrophilic ketone 272, 274 Enhancer of zeste 27, 162 epigenetic information 5 cancer 6, 236 heterochromatin 49 histone messengers 5^6 human disease 14 Lys27 8 silent chromatin 53 X-linked genes 165 ERK1/2 cascade 103 downstream kinases 105 ERK5 103, 105 Esa1 80 catalysis 83 euchromatin 48 heterochromatin competition 61 Ezh1 43 Ezh2 43 EZH2, prostate cancer 35

SUBJECT INDEX

F FAT, CBP/p300 on p53 198 Fen1 acetylation CBP 223^224 p300 223, 224 PCAF 223^224 ¢broblasts, p53 acetylation 199 G G2/M arrest 258, 264 GAL1 65, 67 Gcn5/GCN5 80 catalysis 81, 83 identi¢cation 1 phosphorylation/acetylation link 107 structure 80^81 substrate binding speci¢city 86^87 substrate presentation 101 gene expression HDAC inhibitors 135, 239 heterochromatin 49 pro¢le formulation 5 gene silencing cancer 14, 226^233, 250 DNA methylation 230^231 heterochromatin 8 histone methylation 230 Lys9 8 Polycomb 27, 34 Su(Var)3-9 8, 11 genetic instability, methylation 234 glioblastoma, PA trials 276 GRIP1 80 GSK-3b, hypertrophy repression 139 H haematologic tumours, SAHA 277 HAT1 80 structure 81 HBO1 80 HC-toxin (HC) 134, 137 HDAC inhibitors 134^136, 272^279 activity 275^276 apoptosis 250, 270 cancer therapy 14, 136, 139, 232, 235, 239, 250^251, 269^281 cardiac hypertrophy 137^140, 143^144 cardiac myocyte contractility 142 cell cycle 250

293

cell death 137, 139 clinical trials 270, 276^279 di¡erentiation 250, 270 DNA replication 225 gene expression 135, 239 gene repression 275 gene transcription 275 HSP90 261^262 immunosuppression 131 a-MHC expression 137^138 p21 136, 250 p21Waf1 275^276 phenotypic e¡ects 235 reversibility 143^144, 267 selectivity 275 sensitivity 275 structure^activity relationship 272, 275 T cell inhibition 131 tumour cell proliferation 139, 255, 270 HDAC1 271 complexes 271 deacetylation e⁄ciency 271 identi¢cation 1 NF-kB 210 p53 deacetylation 176, 199^200 prostate cancer 267 RhoB 240 tumorigenesis 206 HDAC2 271 complexes 271 NF-kB 210 RhoB 240 HDAC3 271 deacetylation e⁄ciency 271 HDAC7 interaction 130 NF-kB 210 RelA 210^211, 215 HDAC4 271 embryonic expression 271 MEF2 271 HDAC5 271 congestive heart failure 144 CREB1 271 MEF2 271 memory 271 HDAC6 271 acetylation/ubiquitination integration 176, 180 BUZ ¢nger 173, 175 catalytic deacetylase domains 175 cell cycle 179

294

cell motility 172, 175 conserved residues 173 HDAC6 (cont.) cyclin B1/CDC2 assembly 181 cytoplasm 171 deacetylation activity 271 endocytic tra⁄cking 176 expression 178 microtubules 172 mono-ubiquitination 176 p150glued co-localization 172 PLAP2 binding 176 trapoxin resistance 179 ubiquitination role 176 vesicular sorting and tra⁄cking 178^179 zinc ¢nger 173 HDAC7 271 autoimmunity 131 catalysis 130 HDAC3 interaction 130 in vivo complex formation 130 MEF2 120, 122, 271 thymocytes 118, 120, 122, 125 HDAC8 271 HDAC9 271 cardiac hypertrophy 134, 136 HDAC10 271 head and neck cancer, SAHA 278 Heitz, Emile 48 heterochromatin epigenetic inheritance 49 euchromatin competition 61 gene expression 49 gene silencing 8 histone lysine methylation 24^26 limiting spread 61 naming 48 senescence 35, 41 step-wise assembly 48^56 heterochromatin protein 1 (HP1) cell proliferation 14 chromatin dissociation during mitosis 17^18 chromodomains 8 methylation site reading 8^9 nuclear processes 9 RNA a⁄nity 42 histone acetylation 63, 78^98 Allfrey on 1 V(D)J recombination 148 histone acetyltransferases (HATs)

SUBJECT INDEX

cancer 270 cardiac hypertrophy 134 catalysis by 81, 83^86 families 79, 80 multiprotein complexes 79 nucleosomal acetylation 79 origin of replication 225 structure 80^81 substrate binding speci¢city 86^87 histone code 3^16, 22, 187 binary switches 9^11 histone deacetylases (HDACs) cancer 227, 270 cardiac hypertrophy 132^141 ChIP 130^131 class I (Rpd3-like) 79, 81, 118, 133, 239, 270, 271 class II (Hda1-like) 79, 81, 118, 120, 133, 139, 239, 270, 271 class IIa 118, 120 class IIb 118 class III 79, 81, 118, 133, 239, 270 see also Sir2-like HDACs complexes 271 conservation 271 human chromosome locus 271 inhibitors see HDAC inhibitors muscle cells 133 NF-kB 210 redundancy 270^271 SAGA 75 sizes 271 transcription regulation 238^245 tumorigenesis 250 histone H1 ubiquitylation 64^65 histone H2A.X phosphorylation 12 histone H2A.Z acetylation 172 histone H2B ubiquitylation 63^73 ADH2 65, 67 cell cycle 64 GAL1 65, 67 gene activation 65 Lys4 methylation 68^69 Lys36 methylation 68^69 PHO5 65 SUC2 65, 67 telomeric silencing 64 Ubp8 65^67 histone H3 acetylation and transcription potential 112^113

SUBJECT INDEX

MSK phosphorylation 105^106 phosphorylation/acetylation link 107^110 repressive methyl-lysines 8^9 histone hypoacetylation, silent chromatin 49^50 histone lysine methylation 21^37 BRCA1 35 constitutive and facultative heterochromatin 24^26 methylatable positions 22 mono-di-tri states 22, 24^26 stability 38, 64 tumour progression 35 histone methylation cross-talk 41 DNA methylation 22, 24, 45, 230 memory 19, 45^46 histone methyltransferases (HMTs) dynamic expression 38 PML-RARa 286 histone phosphorylation apoptosis 12 function 11 histone tail modi¢cation, marks for chromatin 22 histone ubiquitylation elongation 74^75 mono/poly ubiquitylation 64, 73^74 transcriptionally active DNA regions 64 see also histone H2B ubiquitylation histones epigenetic information messengers 5 modi¢cation 9 HIV transcription 182, 189, 192^193 HM 49 HML 49 HMR 49 HMT inhibitors 232 HMTases inhibitors 43 Rb 35 Suv39h 22, 26^27, 35 HSP90, HDAC inhibitors 261^262 Hst2 87, 91, 95, 99, 100 human genome chromodomains 17 sequencing 3 hydroxamates 272, 273 p53 status 264

295

I I-kB kinase complex (IKK) 208 I-kBa 209, 213 I-kBs 208 immediate-early genes 102 immunoglobulin heavy chain (IgH) 147, 148^150, 152, 153 immunosuppression, HDAC inhibitors 131 in nucleo labelling 220 in£ammation, immediate-early genes 102 J JNKs 103, 105, 108 negative selection 117 K a-ketoamides 272

structure 274 L LAQ824 272, 276 antiproliferative activity 256 apoptosis 255^256 G1/S arrest 256^258 G2/M arrest 258 HDAC inhibition 254, 261 histone H3 and H4 acetylation 260 HSP90 acetylation 261^262 oncoprotein degradation 263 p21 expression 256^258 p21 promoter activation 254^255 Rb 256^258 reversible e¡ects 267 structure 253, 273 tumour growth 258 larynx cancer, SAHA 278 leukaemia 5-azacytidine 235 epigenomics 14 histone lysine methylation 35 PB trials 277 PML-RAR 35 PML-RARa hypermethylation 285^288 lung cancer, pivaloyloxymethylbutyrate 277 lymphoma, SAHA 278 Lys-CoA 186 Lys4 methylation, H2B ubiquitylation 68^69 Lys9, gene silencing 8

296

Lys27, epigenetic repression 8 Lys28 acetylation 193^194 Lys36 methylation, H2B ubiquitylation 68^69 Lys123 acetylation 73 M MAP kinase pathways 13, 103^104 MCIP, hypertrophy repression 139 Mdm2 198, 200, 201^202 MDM2-HDAC1, p53 regulation 176 MEF2 class II HDACs 120 HDAC4 271 HDAC5 271 HDAC7 120, 122, 271 heart 133 hypertrophy 133 MITR repression 139 muscle cells 133 Nur77 binding sites 118 MEF2D 118, 120, 122 memory HDAC5 271 methylation 19, 45^46 methyl/phos switch 10, 17 a-MHC expression, HDAC inhibitors 137^138 microtubules, acetylation 172 mitotic chromatin dissociation of HP1 and Pc 17^18 histone phosphorylation 12^13 MITR, cardiac hypertrophy 139, 142 MLL SET domain 45^46 MOF 80 Mof 17 MOZ 80 MS-275 250, 272, 276 structure 274 MSKs, H3 phosphorylation 105^106 Mst1 kinase, drug target 13 MTA2 200 tumour cell overexpression 206 multi-methyl lysine antibodies 26 muscle cells, MEF2 133 mycosis fungoides, SAHA 278 myelodysplastic disorder, PB trials 277 MYST 80 catalysis 83, 86 chromodomains 16^17

SUBJECT INDEX

structure 81 substrate binding speci¢city 87 N N-CoA-1, NF-kB 209 N-CoR 120 NAD RENT 54 Sir2 50 Sir2-like HDACs 87, 89 Ner77, TSA sensitivity 145 NF-kB 208^217 co-activators 209^210 co-repressors 210 DNA binding activity 212^213 HDACs 210 nuclear export 214 p50 208 p300/CBP 209 phosphorylation 215 RelA subunit 208, 209, 210^214 SRC-1/N-CoA-1 209 SRC-3 209 transactivation 209, 211^212 NF-Y 241 nicotinamide 50 NOR1, negative selection 117 nuc-1, Tat activity 183 Nur77 Cabin-1 125 HDAC7 120, 122 negative selection 117^118 O O6-MGMT 236 on^o¡ binding 10 oncogenes acetylation 237 activation in cancer 41, 234 oxam£atin 272 structure 273 oxidation 40^41 oxidative stress, methylation 43 P p16 5-Azo-C 237 precancerous lesions 236 p16INK4a tumour suppressor gene 14

SUBJECT INDEX

p19ARF 200 p21 HDAC inhibitors 136, 250 LAQ824 254^258 p21waf1 HDAC inhibitors 275^276 trapoxin 240 p38 103 downstream kinases 105 negative selection 117 p50 208 p53 HDAC inhibitor sensitivity 264 MDM2-HDAC1 regulation 176 PML-RARa 286^287 p53 acetylation 197^205 apoptosis 199 CBP/p300 198, 199, 200 chromosome fragility 199 endogenous gene expression 198 ¢broblasts 199 Mdm2 regulation 198, 200 Mdm2 ubiquitination 201^202 PCAF 199 PML 199 protein stabilization 199 Ras 199 role 198^199 senescence 199 site-speci¢c DNA binding 202 stress response 199, 207 transactivation 199 tumour suppressor function 197^198 ubiquitination and 172^173, 180 p53 deacetylation 176, 199^201 p150glued, HDAC6 co-localization 172 p300 80 cardiac hypertrophy 134 catalysis 86 DNA repair 223^225 Fen1 223, 224 NF-kB 209 p53 198, 199, 200 PCNA complex 223, 224 polymerase b 223, 224 RelA 209, 210^211 Tat acetylation 183^186 TDG 225 tumours 198 XPA 223 p300/CBP-associated factor see PCAF

297

Parc 207 PB 276, 277 PCAF 80 catalysis 81, 83 Fen1 acetylation 223^224 p53 acetylation 199 structure 80^81 substrate binding speci¢city 86^87 Tat acetylation 186^187 pCIP 80 PCNA, p300 complex 223, 224 peripheral tolerance 115, 117 phenylacetate (PA), glioblastoma 276 phenylbutyrate 272 structure 274 PHO5 65 phos/methyl switch 10 PID 199 pivaloyloxymethylbutyrate 276, 277 PLAP2, HDAC6 176 PML, p53 acetylation 199 PML-RAR 35 PML-RARa hypermethylation, leukaemia 285^288 Polycomb (Pc) cell proliferation 14 chromatin dissociation during mitosis 17^18 chromodomains 8 gene silencing 27, 34 methylation site reading 8^9 nuclear processes 9 RNA a⁄nity 42 Polycomb bodies 34 polymerase b, p300 223, 224 position e¡ect variegation 8 PP1ase, suppressor function 11 Prader-Willi syndrome 14 prostate cancer EZH2 35 HDAC1 267 14-3-3 proteins 120, 144 proteome, acetylated proteins 220^221 PTEN, negative selection 117 pyroxamide 134, 276, 277 R RA target genes, PML-RARa 285^286, 288 RAC3 80 radical attack 41

298

RAG1 lymphocyte development 161 V(D)J recombination 147 RAG2 cell cycle 161 lymphocyte development 161 V(D)J recombination 147 Ras, p53 acetylation 199 ras/MEK/ERK pathway, positive selection 117 Rb HMTases 35 LAQ824 256^258 recombination signal sequences 147, 155 RelA acetylation 210^214 CBP 210^211 deacetylation 210^211, 218 HDAC3 210^211, 215 I-kBa 213 NF-kB 208, 209, 210^214 p300 209, 210^211 phosphorylation 215 renal cell carcinoma, SAHA 278 RENT 49, 54 RhoB apoptosis 241^242 cancer proliferation 241^242 Sp1 site 247 trapoxin 240^241, 245 ribosomal DNA 49, 58^59 ribosomal RNAs, silencing 56^57 RNA Polymerase II (Pol II) elongation 68 RNAs, silencing 56^57 Rpd3 Sin3/SAGA complex 75 Sir2/3 systems 59 Rsk2 105 RSK2 105 S Saccharomyces cerevisiae H2B ubiquitylation 64 silencing stability 58 silent chromatin domains 49 Saccharomyces pombe silencing stability 58 Sir2 57 SAGA HDACs 75

SUBJECT INDEX

proteasome subunit 75 Sin3/SAGA complex 75 Ubp8 65, 66, 67 SAHA 134, 250, 272 blood^brain barrier 283^284 clinical trials 276, 277^278 dose-related response 282 HDAC binding 266 structure 273 synergistic activity 276 tumour response prediction 283 SAPKs 103, 105 Sas2 80 Sas3 80 scriptaid 272 structure 273 self-tolerance 115 senescence heterochromatin 35, 41 p53 acetylation 199 Sir2a 201 SET domain proteins 43^45 Set1 68, 74, 75 set2 68, 74, 75 silent chromatin epigenetic code 53 histone hypoacetylation 49^50 Sir2 49, 50^51 Sin3/SAGA complex 75 SIR complex 49^50 Sir^Sir interactions 51, 53 Sir1, silent chromatin 49 Sir2 chromatin templates 53^54 deacetylase activity 50, 53^54 dimers 61 function conservation 57 NAD binding 50 Rpd3 59 silent chromatin 49, 50^51 Sir4 association 49, 51, 54, 58 speci¢city 205 Sir2-like HDACs 79, 81 autoregulation 89 catalysis 89, 91 conservation 79 NAD binding 87, 89 structure 87, 89 substrate binding speci¢city 94^95 Sir2a nuclear localization 200

SUBJECT INDEX

senescence 201 Sir3 BAH domain 57 dimers 61 Rpd3 59 silent chromatin 49 Sir4 association 49, 51, 53, 54 Sir4 dimers 61 silent chromatin 49 Sir2 association 49, 51, 54, 58 Sir3 association 49, 51, 53, 54 SIRT1, nuclear localization 200 SirT3 99 small heterochromatic RNAs 24, 41^42 small interfering RNAs 41^42 SMRT 120 sodium butyrate (NaB) 134, 137 solid tumours 5-azacytidine 235 PB trials 277 SAHA 277 Sp1 sites 240, 247 SRC family 80 SRC-1 80 NF-kB 209 SRC-3 80 NF-kB 209 stress apoptosis 13 p53 acetylation 199, 207 SUC2, H2B ubiquitylation 65, 67 sulfonamide hydroxamic acid 272 structure 273 Suv39h, HMTases 22, 26^27, 35 Su(Var)3-9, gene silencing 8, 11 T T cell inhibition, HDAC inhibitors 131 T cell lymphomas 278^279, 283 T cell receptor (TCR) b loci 148^150, 152, 153 TAFII250 family 80 TAR 183 Tat binding 187, 189 Tat acetylation HIV transcription 189, 192^193 Lys-CoA 186 nuc-1 183 p300 183^186 PCAF 186^187

299

Tat/TAR/cyclin T1 complex 187, 189 TSA 195 TDG, p300 225 telomeres H2B ubiquitylation 64 silent chromatin 49 Tetrahymena 19^20, 42 Thr3 10 thymocytes 115^116 apoptosis regulation 122, 125 death by neglect 116 HDAC7 118, 120, 122, 125 MEF2D 118, 120, 122 negative selection 116^118 positive selection 116, 117^118 thyroid cancer, SAHA 278 TIF-2 80 Tip60 80 TRAM-1 80 transcriptional memory 27 trapoxin A (TPX) 239 cell cycle 239^240 cell proliferation 239^240 p21 promoter 250 p21waf1 240 RhoB expression 240^241, 245 trichostatin A (TSA) 134, 272 cardiac hypertrophy 137 class I/class II HDAC association 142 DNA methylation 231 in vitro systems 144 Ner77 145 p21 promoter 250 sensitivity 234^235 structure 273 Tat expression 195 tri£uoromethyl ketones 272 structure 274 trithorax 27 tubulin, ubiquitination 178 tumorigenesis CBP 198 HDACs 250 tumour progression DNA methylation 227 EZH2 35 histone lysine methylation 35 LAQ824 258 tumour suppressor gene silencing 14, 250

300

SUBJECT INDEX

Enhancer of zeste 162 H3-K27 trimethylation 162 transcription 158^159

U ubiquitin 64^65 ubiquitination, acetylation link 172^173, 176, 180 Ubp8, ubiquitin protease for H2B 65^67 V valproic acid (VA) 272, 276, 277 structure 274 V(D)J recombination 147^148

X X-inactivation 39^40, 163^169 X-linked genes, epigenetic mark 165 Xist 166, 167 XPA, p300 binding 223