Recent Progress of Biochemical and Biomedical Engineering in Japan II

91 Advances in Biochemical Engineering / Biotechnology Series Editor: T. Scheper

Editorial Board: W. Babel · I. Endo · S.-O. Enfors · A. Fiechter · M. Hoare · W.-S. Hu B. Mattiasson · J. Nielsen · H. Sahm · K. Schügerl · G. Stephanopoulos U. von Stockar · G.T. Tsao · C. Wandrey · J.-J. Zhong

Advances in Biochemical Engineering/Biotechnology Series Editor: T. Scheper Recently Published and Forthcoming Volumes

Recent Progress of Biochemical and Biomedical Engineering in Japan II Volume Editor: Kobayashi, T. Vol. 91, 2004 Recent Progress of Biochemical and Biomedical Engineering in Japan I Volume Editor: Kobayashi, T. Vol. 90, 2004 Physiological Stress Responses in Bioprocesses Volume Editor: Enfors, S.-O. Vol. 89, 2004 Molecular Biotechnology of Fungal b -Lactam Antibiotics and Related Peptide Synthetases Volume Editor: Brakhage, A. Vol. 88, 2004 Biomanufacturing Volume Editor: Zhong, J.-J. Vol. 87, 2004 New Trends and Developments in Biochemical Engineering Vol. 86, 2004 Biotechnology in India II Volume Editors: Ghose, T.K., Ghosh, P. Vol. 85, 2003 Biotechnology in India I Volume Editors: Ghose, T.K., Ghosh, P. Vol. 84, 2003 Proteomics of Microorganisms Volume Editors: Hecker, M., Müllner, S. Vol. 83, 2003 Biomethanation II Volume Editor: Ahring, B.K. Vol. 82, 2003

Biomethanation I Volume Editor: Ahring, B.K. Vol. 81, 2003 Process Integration in Biochemical Engineering Volume Editors: von Stockar, U., van der Wielen, L.A.M. Vol. 80, 2003 Microbial Production of l-Amino Acids Volume Editors: Faurie, R., Thommel J. Vol. 79, 2003 Phytoremediation Volume Editor: Tsao, D.T. Vol. 78, 2003 Chip Technology Volume Editor: Hoheisel, J. Vol. 77, 2002 Modern Advances in Chromatography Volume Editor: Freitag, R. Vol. 76, 2002 History and Trends in Bioprocessing and Biotransformation Vol. 75, 2002 Tools and Applications of Biochemical Engineering Science Volume Editors: Schügerl, K., Zeng, A.-P. Vol. 74, 2002 Metabolic Engineering Volume Editor: Nielsen, J. Vol. 73, 2001

Recent Progress of Biochemical and Biomedical Engineering in Japan II Volume Editor : Takeshi Kobayashi

With contributions by T. Hanai · H. Honda · S. Iijima · A. Ito · M. Kamihira · M. Kino-oka · T. Kobayashi · K. Nishijima · K. Shimizu · M. Shinkai · M. Taya · N. Uozumi

2 3

Advances in Biochemical Engineering/Biotechnology reviews actual trends in modern biotechnology. Its aim is to cover all aspects of this interdisciplinary technology where knowledge, methods and expertise are required for chemistry, biochemistry, micro-biology, genetics, chemical engineering and computer science. Special volumes are dedicated to selected topics which focus on new biotechnological products and new processes for their synthesis and purification. They give the state-of-the-art of a topic in a comprehensive way thus being a valuable source for the next 3–5 years. It also discusses new discoveries and applications. In general, special volumes are edited by well known guest editors. The series editor and publisher will however always be pleased to receive suggestions and supplementary information. Manuscripts are accepted in English. In references Advances in Biochemical Engineering/Biotechnology is abbreviated as Adv Biochem Engin/Biotechnol as a journal. Visit the ABE home page at www.springeronline.com

ISSN 0724-6145 ISBN 3-540-20419-9 DOI 10.1007/b14093 Springer-Verlag Berlin Heidelberg New York Library of Congress Control Card Number 2004104392 This work is subject to copyright. All rights are reserved, whether the whole or part of the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation, broadcasting, reproduction on microfilm or in any other way, and storage in data banks. Duplication of this publication or parts thereof is permitted only under the provisions of the German Copyright Law of September 9, 1965, in its current version, and permission for use must always be obtained from Springer-Verlag. Violations are liable to prosecution under the German Copyright Law. Springer is a part of Springer Science+Business Media springeronline.com © Springer-Verlag Berlin Heidelberg 2004 Printed in Germany The use of general descriptive names, registered names, trademarks, etc. in this publication does not imply, even in the absence of a specific statement, that such names are exempt from the relevant protective laws and regulations and therefore free for general use. Typesetting: Fotosatz-Service Köhler GmbH, Würzburg Cover: KünkelLopka GmbH, Heidelberg; design & production GmbH, Heidelberg Printed on acid-free paper

02/3020mh – 5 4 3 2 1 0

Series Editor Professor Dr. T. Scheper Institute of Technical Chemistry University of Hannover Callinstraße 3 30167 Hannover, Germany [email protected]

Volume Editor Professor Dr. Takeshi Kobayashi Department of Biological Chemistry Chubu University 1200 Matsumoto-cho, Kasugai Aichi 487-8501, Japan [email protected]

Editorial Board Prof. Dr. W. Babel

Prof. Dr. I. Endo

Section of Environmental Microbiology Leipzig-Halle GmbH Permoserstraße 15 04318 Leipzig, Germany [email protected]

Faculty of Agriculture Dept. of Bioproductive Science Laboratory of Applied Microbiology Utsunomiya University Mine-cho 350, Utsunomiya-shi Tochigi 321-8505, Japan [email protected]

Prof. Dr. S.-O. Enfors

Prof. Dr. A. Fiechter

Department of Biochemistry and Biotechnology Royal Institute of Technology Teknikringen 34, 100 44 Stockholm, Sweden [email protected]

Institute of Biotechnology Eidgenössische Technische Hochschule ETH-Hönggerberg 8093 Zürich, Switzerland [email protected]

Prof. Dr. M. Hoare

Prof. W.-S. Hu

Department of Biochemical Engineering University College London Torrington Place London, WC1E 7JE, UK [email protected]

Chemical Engineering and Materials Science University of Minnesota 421 Washington Avenue SE Minneapolis, MN 55455-0132, USA [email protected]

VI

Editorial Board

Prof. Dr. B. Mattiasson

Prof. J. Nielsen

Department of Biotechnology Chemical Center, Lund University P.O. Box 124, 221 00 Lund, Sweden [email protected]

Center for Process Biotechnology Technical University of Denmark Building 223 2800 Lyngby, Denmark [email protected]

Prof. Dr. H. Sahm

Prof. Dr. K. Schügerl

Institute of Biotechnolgy Forschungszentrum Jülich GmbH 52425 Jülich, Germany [email protected]

Institute of Technical Chemistry University of Hannover, Callinstraße 3 30167 Hannover, Germany [email protected]

Prof. Dr. G. Stephanopoulos

Prof. Dr. U. von Stockar

Department of Chemical Engineering Massachusetts Institute of Technology Cambridge, MA 02139-4307, USA [email protected]

Laboratoire de Génie Chimique et Biologique (LGCB), Départment de Chimie Swiss Federal Institute of Technology Lausanne 1015 Lausanne, Switzerland [email protected]

Prof. Dr. G.T. Tsao

Prof. Dr. C. Wandrey

Director Lab. of Renewable Resources Eng. A.A. Potter Eng. Center Purdue University West Lafayette, IN 47907, USA [email protected]

Institute of Biotechnology Forschungszentrum Jülich GmbH 52425 Jülich, Germany [email protected]

Prof. Dr. J.-J. Zhong State Key Laboratory of Bioreactor Engineering East China University of Science and Technology 130 Meilong Road Shanghai 200237, China [email protected]

Advances in Biochemical Engineering/Biotechnology Also Available Electronically

For all customers who have a standing order to Advances in Biochemical Engineering/Biotechnology, we offer the electronic version via SpringerLink free of charge. Please contact your librarian who can receive a password for free access to the full articles by registering at: www.springerlink.com If you do not have a subscription, you can still view the tables of contents of the volumes and the abstract of each article by going to the SpringerLink Homepage, clicking on “Browse by Online Libraries”, then “Chemical Sciences”, and finally choose Advances in Biochemical Engineering/Biotechnology. You will find information about the – Editorial Board – Aims and Scope – Instructions for Authors – Sample Contribution at www.springeronline.com using the search function.

Attention all Users of the “Springer Handbook of Enzymes”

Information on this handbook can be found on the internet at www.springeronline.com A complete list of all enzyme entries either as an alphabetical Name Index or as the EC-Number Index is available at the above mentioned URL.You can download and print them free of charge. A complete list of all synonyms (more than 25,000 entries) used for the enyzmes is available in print form (ISBN 3-540-41830-X).

Save 15% We recommend a standing order for the series to ensure you automatically receive all volumes and all supplements and save 15% on the list price.

Preface

Preface

During World War II, industrial production of penicillin was started. This was the first example of aerobic microbial cultivation on an industrial scale, and many new techniques were developed to cultivate Penicillium chrysogenum in large fermenters. Since demands for penicillin production were urgent, microbiologists, biochemists and chemical engineers were in a great hurry to start production of penicillin and as a result, most early techniques were acquired through empirical procedures. However, this was the start of biochemical engineering, and the contents of almost all chapters of the book ”Biochemical Engineering” written by Shuichi Aiba,Arthur E. Humphrey and Nancy F. Millis (University of Tokyo Press, 1965) dealt with this subject and its later development under academic conditions. In 1957, glutamic acid production was started by Kyowa Hakko Co. Other amino acids were also produced industrially and industrial microbial cultivation was rapidly developed to an advanced state. The organisms were not limited to microorganisms, and mammalian cells and plant cells were then also applied to production of glycosylated proteins and complex secondary metabolites. In 1972, genetic engineering technology was developed and this technique had a drastic influence not only on basic biosciences but also on biochemical engineering. The wealth of information that has been accumulated on genetic engineering technology and as well as hybridoma technology has made it possible to produce various metabolites and proteins in microorganisms, mammalian cells and plant cells. Species barriers between microorganisms, animals and plants have been, in principle, eliminated. The areas dealt with in biochemical engineering have been expanded to many organisms. In 2003, the human genome project has been completed, and complete DNA sequences have been announced. The areas dealt with in biochemical engineering have been expanded to humans; bioinformatics and biomedical engineering are now parts of biochemical engineering. A major objective in editing this book has been to gather together the information of the expanding areas in biochemical engineering. Publication was motivated by the retirement of the editor after working at Nagoya University from April 1968 to March 2004. The editor asked his esteemed friends and colleagues doing active research in biochemical engineering in Japan to sum

X

Preface

up the information of these rapidly expanding areas in biochemical engineering. This book is not intended to be an introduction to biochemical engineering, but to serve as a reference that looks at the expanded field of biochemical engineering in Japan and also looks forward to future prospects. This book (two volumes) is composed of 15 chapters dealing with microbial cultivation techniques in biomedical engineering including tissue engineering and cancer therapy. Hopefully, this book will give readers a glimpse of the past and also a view of what may happen in biochemical engineering in Japan. Finally, I would like to thank all my friends and colleagues for their cooperation in publishing this book. I express my deepest appreciation to my wife, Noriko Kobayashi who endured the long evenings and weekends I devoted to working at Nagoya University. Nagoya, May 2004

Takeshi Kobayashi

Contents

Metabolic Flux Analysis Based on 13C-Labeling Experiments and Integration of the Information with Gene and Protein Expression Patterns K. Shimizu . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

1

Application of Knowledge Information Processing Methods to Biochemical Engineering, Biomedical and Bioinformatics Field T. Hanai · H. Honda . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51 Large-Scale Production of Hairy Root N. Uozumi . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75 Large-Scale Micropropagation System of Plant Cells H. Honda · T. Kobayashi . . . . . . . . . . . . . . . . . . . . . . . . . . . 105 Development of Culture Techniques of Keratinocytes for Skin Graft Production M. Kino-oka · M. Taya . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135 Transgenic Birds for the Production of Recombinant Proteins M. Kamihira · K. Nishijima · S. Iijima . . . . . . . . . . . . . . . . . . . . 171 Functional Magnetic Particles for Medical Application M. Shinkai · A.Ito . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 191 Author Index Volumes 51–91 . . . . . . . . . . . . . . . . . . . . . . . . 221 Subject Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 237

Contents of Volume 90 Recent Progress of Biochemical and Biomedical Engineering in Japan I Volume Editor: Takeshi Kobayashi

Recent Progress in Microbial Cultivation Techniques E.Y. Park Clarification of Interactions Among Microorganisms and Development of Co-culture System for Production of Useful Substances M. Taniguchi · T. Tanaka High Rate Production of Hydrogen/Methane from Various Substrates and Wastes N. Nishio · Y. Nakashimada Bacterial Capsular Polysaccharide and Sugar Transferases K. Miyake · S. Iijima Bacterial Sterilization and Intracellular Protein Release by Pulsed Electric Field T. Ohshima · M. Sato Cell-free Protein Synthesis Systems: Increasing their Performance and Applications H. Nakano · Y. Kawarasaki · T. Yamane Enzymatic Synthesis of Structured Lipids Y. Iwasaki · T. Yamane Bioprocess Monitoring Using Near-Infrared Spectroscopy K. Suehara · T. Yano

Adv Biochem Engin/Biotechnol (2004) 91: 1– 49 DOI 10.1007/b94204 © Springer-Verlag Berlin Heidelberg 2004

Metabolic Flux Analysis Based on 13C-Labeling Experiments and Integration of the Information with Gene and Protein Expression Patterns Kazuyuki Shimizu 1(✉), 2 1

2

Department of Biochemical Engineering & Science, Kyushu Institute of Technology, Iizuka, Fukuoka 820-8502, Japan Institute for Advanced Biosciences, Keio University, Tsuruoka, Yamagata 997-0017, Japan [email protected]

1

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

2 2.1 2.2 2.3 2.4 2.5 2.6

Metabolic Flux Analysis Using 13C-Labeling Experiment . . . . . . . Labeling Experiments . . . . . . . . . . . . . . . . . . . . . . . . . . . Sample Preparation and Analytical Procedure for NMR Analysis . . . Sample Preparation and Analytical Procedure for GC-MS Analysis . . Relations Between Mass Isotopomers and Metabolic Flux Analysis . . Mathematical Modeling and Computer Program for Flux Calculation Metabolic Flux Distribution of E. coli . . . . . . . . . . . . . . . . . .

. . . . . . .

. . . . . . .

8 8 8 9 9 10 11

3 3.1

. . .

15

3.2 3.3 3.4 3.5 3.6

Metabolic Regulation Analysis of E. coli in Protein Expression Level Protein Expressions and Enzyme Activities of E. coli under Different Culture Conditions . . . . . . . . . . . . . . . . . . . . . . . . . . . . Glycolysis and Anaplerotic Pathway . . . . . . . . . . . . . . . . . . Pentose Phosphate Pathway and Entner-Doudoroff Pathway . . . . Fermentative Pathway . . . . . . . . . . . . . . . . . . . . . . . . . . TCA Cycle and Glyoxylate Shunt . . . . . . . . . . . . . . . . . . . . Comparison of the Protein Expression Level and Enzyme Activity .

. . . . . .

. . . . . .

. . . . . .

15 20 23 24 26 27

4 4.1 4.2 4.3 4.4

Metabolic Regulation Analysis of pgi – Mutant E. coli by Gene Expression Glycolysis and Pentose Phosphate Pathway Genes . . . . . . . . . . . . . TCA Cycle, Glyoxylate Shunt, and Fermentation Genes . . . . . . . . . . . Heat Shock and Regulatory Genes . . . . . . . . . . . . . . . . . . . . . . Relationship Between Gene Expression and Enzyme Activities . . . . . .

. . . . .

29 29 35 35 36

. . . . . . . .

37 37 37

. . . . . .

40 42

Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

45

References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

46

. . . . . .

Integration of Information for Metabolic Regulation Analysis for Gene-Knockout E. coli . . . . . . . . . . . . . . . . . . . . . . . . . 5.1 Metabolic Regulation of pykF – Mutant of E. coli . . . . . . . . . . . . 5.1.1 Fermentation Characteristics and Enzyme Activities for Batch Culture 5.1.2 Metabolic Regulation Analysis by Metabolic Flux Distribution, Enzyme Activities, and Intracellular Metabolites Concentrations for Continuous Culture . . . . . . . . . . . . . . . . . . . . . . . . . . 5.2 Metabolic Regulation Analysis of gnd – Mutant of E. coli . . . . . . . .

. . . . . . .

4

5

6

2

K. Shimizu

Abstract The recent progress on metabolic systems engineering was reviewed, in particular focusing on the metabolic flux analysis (MFA) based on the isotopomer distribution obtained using NMR and/or GC-MS. After the brief explanation of how to estimate the metabolic flux distribution (MFD) based on 13C-labeling experiments, the metabolic regulation analysis was made based on the protein expression patterns obtained by two-dimensional electrophoresis (2DE) together with enzyme activity data, and the gene expression patterns obtained by RT-PCR analysis. The particular application was considered for Escherichia coli. The effect of culture conditions such as different carbon sources (glucose, gluconate, glycerol, acetate, etc.) and different dissolved oxygen (DO) concentration, etc. on the metabolism was investigated. The effect of some single-gene knockout such as pgi–, pyk–, and gnd– was also investigated. It was found to be quite useful to integrate the information obtained from metabolic flux analysis, gene and protein expressions as well as intracellular metabolite concentrations to understand the overall picture of metabolic regulation. Keywords Metabolic flux analysis · 13C experiment · Isotope distribution · Gene expression · Protein expression · Metabolic regulation · Escherichia coli List of Abbreviations and Symbols

Pathways EM pathway ED pathway PP pathway TCA cycle

Embden-Meyerhof pathway Entner Doudoroff pathway Pentose phosphate pathway Tricarboxylic acid cycle

Metabolites AcCoA AcP CIT DHAP E4P F6P F1,6BP FUM G6P GAP ICIT KDPG aKG MAL OAA 3PG 6PG PYR PEP R5P RU5P S7P SUC SucCoA X5P

Acetyl CoA Acetyl phosphate Citrate Dihydroxy acetone phosphate Erythrose-4-phosphate Fructose-6-phosphate Fructose1,6 bisphosphate Fumarate Glucose-6-phosphate Glyceraldehydes 3-phosphate Isocitrate 2-Keto-3-deoxy-6-phosphogluconate a-Ketoglutarate Malate Oxaloacetate 3-Phospho glycerate 6-Phospho glycerate Pyruvate Phosphoenolpyruvate Ribose-5-phosphate Ribulose-5-phosphate Sedoheptulose 7-phosphate Succinate Succinyl CoA Xylose 5-phosphate

Metabolic Flux Analysis Based on 13C-Labeling Experiments Genes AceA aceB aceEF ack acnA,B acs adh eno eda edd fba fbp fumA,B,C gapA glk gltA gnd icdA lpd mae mdh pckA pfk pfl pgi pgk ppc pps pta pts pyk sdh sucAB tal tkt tpi zwf

Gene coding for isocytrate lyase Gene coding for malate synthase Gene coding for pyruvate dehydrogenase Gene coding for acetate kinase Gene coding for aconitase Gene coding for acetyl-CoA synthetase Gene coding for alcohol dehydrogenase Gene coding for enolase Gene coding for 6PGDH Gene coding for KDPG aldrase Gene coding for fructose bisphosphate aldorase Gene coding for fructose bisphosphatase Gene coding for fumarate Gene coding for glyceraldehydes 3-phosphate Gene coding for glucokinase Gene coding for citrate synthase Gene coding for 6PGDH Gene coding for ICDH Gene coding for pyruvate dehydrogenase complex Gene coding for malic enzyme Gene coding for malate dehydrogenase Gene coding for phosphoenolpyruvate carboxykinase Gene coding for phosphofruct kinase Gene coding for pyruvate formate lyase Gene coding for phosphoglucose isomerase Gene coding for phospho gluctokinase Gene coding for phosphoenolpyruvate carboxylase Gene coding for phosphoenolpyruvate synthase Gene coding for phosphotrans acetylase Gene coding for phospho transferase system Gene coding for pyruvate kinase Gene coding for succinate dehydrogenase Gene coding for 2-Ketoglutarate dehydrogenase Gene coding for transaldolase Gene coding for transketolase Gene coding for triose phosphate isomerase Gene coding for G6PDH

Enzymes 6PGDH ADH FBPase G6PDH GAPDH Hxk ICDH Icl MDH Mez Pfk

6-Phospho gluconate dehydrogenase alcohol dehydrogenase Fructose bisphosphatase Glucose 6-phosphate dehydrogenase Glyceraldehyde 3-phosphate dehydrogenase Hexokinase Isocitrate dehydrogenase Isocitrate lyase Malate dehydrogenase Malic enzyme Phosphofructkinase

3

4

K. Shimizu

Pgi Pps Pyk Rpi Rpe Tal Tpi Tkt

Phosphoglucose isomerase Phosphoenolpyruvate synthase Pyruvate kinase Ribulose 5-phosphate isomerase Ribulose phosphate 3-epimerase Trans aldorase Triose phosphate isomerase Transketorase

Others FMDV IDV MDV MTBST-FA TBDMS

Fragment mass distribution vector Isotopomer distribution vector Mass distribution vector N-(tert-Butyldimethylsilyl)-N-methyl-trifluoroacetamide tert-Butyldimethylsilyl substituent

1 Introduction It is becoming more and more important to analyze the cell in vivo for the post genome research. For this, it is important to analyze the cell as a whole or as a system, and to investigate the global metabolic regulation based on gene and protein expressions as well as metabolic flux distribution. A wealth of information is available on the genetic regulation, biochemistry and physiology of cellular metabolism, but surprisingly little is known about the overall metabolic regulation. The central metabolic pathway has the anabolic and catabolic functions by providing cofactors and building blocks for macromolecular synthesis as well as energy production. The changes in cellular physiology such as redirection of intermediary metabolism in response to the specific gene knockout and/or the change in culture conditions will affect the metabolism. While some single-gene knockout mutations in central metabolism preclude growth on glucose, a majority of such variations can potentially be compensated for either using alternative enzymes or by the rerouting of carbon fluxes through alternative pathways. Many of these adjustments in metabolism are accompanied by the changes in gene and protein expressions (or enzyme activities), the intracellular metabolite concentrations, and metabolic flux distributions. One of the valuable methods to gain insight into the complex metabolic control mechanism of the whole cell is the metabolic flux analysis based on the metabolic flux distributions in the central metabolism. Metabolic flux analysis is in principle based on the mass conservation of key metabolites in the central metabolism, through which the intracellular fluxes are calculated from a few measured fluxes by applying mass balances to the intracellular metabolites [1, 2]. Since the limited number of measurable extracellular fluxes and stoichiometric constraints often lead to an underdetermined algebraic system, it is often necessary to include cofactor mass balances into the stoichiometric model or introduce the objective function for optimization. However, the correctness of

Metabolic Flux Analysis Based on 13C-Labeling Experiments

5

the calculated fluxes depends on the validity of these cofactor assumptions or depends on the appropriate choice of the objective function employed. The presence of unknown reactions consuming or producing the cofactor might invalidate the assumptions of cofactor mass balance, and the selected objective function is usually an ideal one in theory but hard to be attained in practice. An alternative approach is to use isotopic tracer data, which can overcome the inherent lack of measurable quantities related to the intracellular flux distribution. By introduction of isotopically labeled substrates to the biological system, the labeled carbon atoms are then distributed throughout the metabolic network. Then the final isotopic enrichment in the intracellular metabolite pools may be measured [3]. The resulting data may provide a large amount of information to quantitate the intracellular fluxes. Since the amino acids in biomass hydrolysates are much more abundant than the precursors in the central metabolism from which they are derived, it is easier to deduce the labeling pattern of the intracellular metabolites from the labeling patterns of the amino acids if the precursor-amino acid relationships are known [4]. Currently, this tracer technique in combination with direct extracellular flux measurements is considered to be the most powerful method for obtaining intracellular flux distribution with only a few modeling assumptions [5, 6]. The tracer experiments are made using either NMR spectroscopy or GC-MS. The 13C NMR technique, although powerful and attractive in metabolic flux analysis, has the disadvantage of relatively low sensitivity. The technique is not able to detect metabolites present at concentrations below 10–4 mol/l [7]. Therefore, large amount of the sample is a prerequisite to the NMR analysis. On the other hand, GC-MS method can easily analyze metabolites in 10–7 mol/l concentration. As a result, it may be considered as the suitable method in the flux analysis of the culture process with low biomass concentration [8]. Although highly sensitive and cheaper, GC-MS has attracted much less attention than NMR. At present, most work on metabolic flux analysis is carried out by NMR which is based on the measurement of positional 13C enrichment by onedimensional 1H NMR [5] or based on 13C quantification of isotopomers by onedimensional 13C NMR [9, 10] or two-dimensional [13C, 1H] COSY NMR [4, 11–13]. One reason may be that GC-MS-derived constraints (i.e., certain linear combinations of isotopomer fraction) on isotopomer pools are more difficult to interpret intuitively than those of NMR, and the original mass data require an additional algorithm for correction of naturally occurring elemental isotopes [14, 15]. It may be considered to combine NMR tracer data with MS data to improve the accuracy of the flux distribution [16, 17]. It should be noted that the metabolic flux distribution is computed based on the assumed metabolic pathways. Therefore, it should be made sure that the assumed pathways are active under the condition considered by checking the corresponding protein expressions or enzyme activities. As for the protein expression analysis, two-dimensional electrophoresis (2DE) was proposed as early as 1975 as the most efficient method of separating complex protein mixtures to analyze global patterns of gene expression at

6

K. Shimizu

the protein level [18]. Since then, this approach has been extensively adopted [19, 20]. One major outcome of the proteome studies is the establishment of 2DE databases for many organisms. These databases provide an easy access to the analysis of gene expression in response to genetic or environmental alterations at the protein level. The regulation of metabolic processes ultimately depends upon the control of enzyme activity. There are three general mechanisms by which the activity of enzymes can be regulated: control by reversible binding of effectors, by covalent modification, and by alteration of enzyme concentration. In the first case, the enzyme is activated or inhibited by binding of a signal molecule, which may or may not be the substrate or product of the enzyme reaction, to the specific regulatory site, producing a conformational change. Substrate effect and allosteric control may be the example. In the case of covalent modification, the structure of the enzyme can be altered by the action of other enzymes. There are a number of such examples. For instance, the regulation by phosphorylation, i.e., phosphate is incorporated into the enzyme by a protein kinase using ATP, and is removed by a protein phosphatase. The third mechanism which regulates the enzyme activity is the alteration of the concentration of enzyme protein in the cell. The concentration of a protein in the cell is governed by the rate of the synthesis and the rate of degradation. The rate of synthesis of a particular protein may be controlled at several different levels. The rate of transcription of the gene may be controlled. Other possible sites of control are the processing of the transcript to give mRNA, the transfer of mRNA out of the nucleus, the rate of degradation of mRNA in the cytoplasm, or the rate of translation of mRNA to make the protein on the ribosome. There is very strong evidence that the rate of transcription is under rigorous control, and this control is important in determining the enzyme profile of a particular cell type [21]. Therefore, to large extent, the enzyme activity reflects the protein expression level. Gene expression profiling is also useful for analyzing metabolic regulation at a genomic level. It can be used to compare global changes in gene expression that occur in response to the environmental stimulus or to compare the effects of genetic changes on gene expression. This analysis can provide important information about cell physiology and has the potential to identify connections between regulatory or metabolic pathways that were not previously known. The use of gene arrays to analyze gene expression has been used extensively for eukaryotic systems [22, 23]. Recently, its usefulness for analyzing gene expression has also been demonstrated for prokaryotes [24–27]. Since biotechnological research and its application require the knowledge on how genes work at the genomic level, DNA microarray has been extensively used during the past several years for this purpose to reveal global regulation. As illustrated in S. cerevisiae, cell metabolism adapts to the environment during growth in a chemostat culture [28]. The transcript levels of approximately 10% of the genes were changed more than twofold after 250 generations in the fermentor. This kind of investigation is useful in analyzing the physiological changes and detecting the effects of gene structure changes for strain improvement [29].

Metabolic Flux Analysis Based on 13C-Labeling Experiments

7

RT-PCR can be also used to investigate the gene expression. In RT-PCR, an RNA template is copied into a complementary DNA transcript (a cDNA) using a retroviral reverse transcriptase. The cDNA is then amplified exponentially using PCR. This product can be used to detect or quantify the expression [30, 31]. RT-PCR has been applied to detect changes in absolute and relative amounts of specific RNAs [32]. The latter is also called semi-quantitative RT-PCR. RT-PCR is more sensitive and easier to perform as compared with other RNA analysis techniques, including Northern blots, RNase protection assays, in situ hybridization, and S1 nuclease assays [33, 34]. Comparing with northern analysis, which has been used routinely to visualize specific mRNA levels, RT-PCR offers several advantages such as: (1) much less amount of total RNA is required; (2) the low abundance transcripts of interest, which cannot be detected by Northern blotting, can be quantified reproducibly by RT-PCR; (3) the variable expression of multiple mRNAs can in principle be analyzed simultaneously by RT-PCR, which is difficult to realize by the traditional methods; (4) the RT-PCR procedure requires only a few hours to perform. Generally, two types of RT-PCR approaches have been used until now. The most popular method involves the use of an internal standard to control variations in PCR amplification efficiency and to determine the amount of transcript in original samples [35]. Although this method can be used to detect small changes in mRNA levels, it requires not only the additional steps to prepare the internal standard, but also necessitates performing several PCR reactions to quantify each sample of mRNA. Therefore, this method is time consuming, complicated and not useful for rapid quantification of multiple samples. Another approach [36, 37] relies on adjustment of the amount of input RNA and the number of cycles of PCR to assure that measurement is done in the exponential phase of PCR when the signal is proportional to the amount of input template or the number of cycles. This method is reproducible for measurement of relative changes in mRNA levels if the following two conditions are met. First, tube to tube variation must be minimal so that a constant value can be assumed in all related PCR reactions. Second, all data must be obtained before the reactions begin to reach the plateau phase. Although this method requires a standard curve to determine the level of signal that corresponds to the specific amount of mRNA, it is rapid and simple to quantify the variable expression of mRNAs with sufficient resolution [38, 39]. In the following, the recent research progress on the metabolic regulation is explained from the viewpoints of gene and protein expressions as well as metabolic flux distribution in particular for the main metabolism of Escherichia coli.

8

K. Shimizu

2 Metabolic Flux Analysis Using 13C-Labeling Experiment 2.1 Labeling Experiments Isotope labeling experiments are usually applied for the continuous culture and are initiated after the culture reached a steady state, which may be inferred from the stable oxygen and carbon dioxide concentrations in the off-gas and stable optical density in the effluent medium for at least twice as long as the residence time. The feed medium with unlabeled glucose may be then replaced by an identical medium containing 10% of uniformly labeled glucose [U-13C], 10% of the first carbon labeled glucose [1-13C] and 80% of naturally labeled glucose per liter. Another combination of labeling substrate may be used, but its determination should be made based on the available measurement apparatuses and the cost of labeling substrate. Note that the accuracy of the metabolic flux distribution depends on the mixture of labeled substrate and the measurement apparatuses [40]. 2.2 Sample Preparation and Analytical Procedure for NMR Analysis At the end of labeling experiment, about 300 ml of the culture samples may be harvested followed by centrifugation at 10,000¥g for 10 min at 4 °C. The cell pellet is washed once with 20 mmol/l tris-HCl (pH 7.6), and hydrolyzed in 6 ml of 6 mol/l HCl for 12 h at 105 °C. In the resulting hydrolysate, 16 proteinogenic amino acids are present, since cysteine and trytophan are oxidized, while asparagine and glutamine are deaminated during the acid hydrolysis. The hydrolysate is filtered through a filter with 0.2 mm pore size, and evaporated to dryness. The dried material is dissolved in 700 ml of 20 mmol/l deuterium chloride (DCl) in D2O, filtered and used for the NMR measurement. The labeling patterns of amino acids in the hydrolysates can be determined by NMR spectroscopy. Two-dimensional proton-detected 1H-13C heteronuclear multiplequantum correlation (HMQC) spectra may be recorded. For each labeling experiment, two spectra are measured: one focused on aliphatic carbons with 13 C carrier set to 45 ppm, and the other for the aromatic rings with the 13C carrier set to 125 ppm. The measurement time of the aliphatic spectra may be 15.5 h. Before Fourier transformation, the time domain data are multiplied in t1 and t2 with the sine-bell windows shifted by p/2. The cross section along the 13 C axis through the most intense part of each cross peak in the 2D spectra can be used for spectra analysis. After baseline correction, the various peaks in the cross section are integrated to quantify the relative contributions of singlet, doublet, doublet of doublets, and triplet signals etc. to the overall multiplet pattern of the carbon signal.

Metabolic Flux Analysis Based on 13C-Labeling Experiments

9

2.3 Sample Preparation and Analytical Procedure for GC-MS Analysis For the isotopomer mass analysis of the cellular amino acids, about 30 ml of culture broth may be harvested and centrifuged at 12,000¥g for 10 min at 4 °C. The pellet is washed twice with distilled water and centrifuged again. Then about 20 mg of wet biomass is transferred to 1 ml of 6 mol/l HCl. The closed tube is heated for 24 h at 110 °C for complete hydrolysis, and after cooling to the room temperature, the solvent is evaporated by a vacuum dryer. After this, about 1 ml of distilled water is added to the dried hydrolysates, which is then filtered through a 0.2 mm pore size filter for separation of the cell debris. The filtrate is dried again and redissolved in 0.5 ml acetonitrile (chromatographic grade) for GC-MS analysis. For analytical procedure, 100 ml of acetonitrile containing biomass hydrolysate is added to 100 ml N-(tert-butyldimethylsilyl)-N-methyl-trifluroacetamide (MTBST-FA). Other derivatization may be considered. The mixture is incubated for 60 min at 110 °C for complete derivatization. After cooling to room temperature, aliquots of the solution containing the derivatives can be used directly for GC-MS. 2.4 Relations Between Mass Isotopomers and Metabolic Flux Analysis It has been shown that isotopomer distribution vectors (IDV), which contain mole fractions of individual isotopomers, significantly depend on intracellular flux distributions [41]. IDV in combination with isotopomer mapping matrices (IMM) allows the efficient use of matrices to handle the complex metabolic flux equation system [6]. The isotopomer distribution expressed as IDVs has no direct relation with measurement data, but the simulated 13C labeling patterns of biomass components can be calculated from IDVs because such labeling patterns are determined by the distribution of 13C isotopes in the molecules, thus facilitating the direct comparison between simulated quantities and experimental data so as to obtain the best fit flux distribution. While synthetic NMR signals are simulated in the form of relative contributions of the individual multiplet signals to the overall condition pattern in the 13C-13C COSY spectra of proteinogenic amino acids [42], the concept of mass distribution vectors (MDVs) and fragment mass distribution vectors (FMDVs) is used in simulation to deal with mass isotopomer distribution in metabolic flux analysis [43]. FMDVs are caused by the fragment mass distribution analysis of the compound of interest by EI (electron ionization) mass spectrometry. Such fragments can give valuable information for the estimation of metabolic fluxes in that they can increase the resolution of the labeling pattern. Incorporation of FMDVs into an appropriate model for metabolic flux analysis can provide a highly accurate methodology to determine the metabolic flux distribution provided that the labeled substrate inputs are restricted to the feasible mixtures by

10

K. Shimizu

way of specifying the corresponding mixture components [44]. The dynamic correction may also be made [45]. 2.5 Mathematical Modeling and Computer Program for Flux Calculation For every bidirectional reaction step, a concept of exchange flux viexch, according to Wiechert [46], may be used to represent a forward and backward reaction viÆ and viÆ by net flux vinet and viexch [6]. To circumvent the difficulties involved in the numerical treatment of high exchange fluxes, the non-linear mapping between viexch and exchange coefficient exch(i) may be employed [42]. As for the intracellular metabolite pools, the quasi-steady state is assumed, which leads to Avnet = 0

(1)

where vnet is the steady state vector for the net fluxes. A is the stoichiometric matrix. The rank of A is determined, and the number of degrees of freedom is determined, followed by the specification of the independent fluxes to determine the rest. The values of these free net fluxes are then varied within the simulation runs or obtained from extracellular measurements to obtain the overall flux distribution in the network. The algorithm for computing metabolic flux distribution using GC-MS data may be as follows: (1) skewing effects of natural isotopes (C-13, H-2, O-17, O-18, N-15, Si-29, Si-30) existed in MTBSTFA-derivatized amino acids can be corrected by the algorithm proposed by Paul Lee [14, 15]; (2) free net fluxes are randomly chosen, and these values as well as those for the exchange coefficients, are first given with arbitrary values by a specifically designed function. By rearranging and partitioning A and v, all the net fluxes are determined by vu = –A–1 u Amvm

(2)

where vm is the measured specific rate vector, and vu the rest of the flux value vector. Am and Auare the corresponding stoichiometric matrices. All the forward fluxes vÆ and backward fluxes v¨ are expressed by the net fluxes vnet and the exchange coefficients exch. (3) By using the concept of isotopomer mapping matrices, which was introduced by Schmidt [41], and the assumed values of vnet and exch, the corresponding steady-state isotopomer distributions in the intracellular metabolite pools are obtained by an iterative scheme [6]. (4) The obtained steady-state isotopomer distributions are transformed into simulated fragment mass isotopomer distribution based on precursor-amino acid relationships. (5) The simulated fragment mass isotopomer distributions are compared to the measured data. (6) The above steps may be incorporated into a hybrid algorithm by combining global search algorithms such as Monte Carlo method, evolutionary methods, and genetic algorithms (GA) [47] with local

Metabolic Flux Analysis Based on 13C-Labeling Experiments

11

search algorithms such as Levenberg-Marquardt algorithm (LMA) [48] to find a global minimum. The objective function to be minimized may be defined as:

$

% $

%

n Mi,meas – Mi,calc(n,I) 2 n Sj,meas – Sj,calc e (n,I) = Â 0002 + Â 002 di dj i=1 j=1

2

(3)

where n is the parameter vector containing free net fluxes and exchange coefficients to be optimized, and I is the isotopomer distribution vector of the input substrate. Mi,meas are the n individual labeling measurements obtained from mass isotopomer analysis and Mi,calc their corresponding simulated values computed in the above steps based on the assumed values in n. Sj,meas is a vector containing the measured values of m extracellular fluxes and Sj,calc a vector containing the simulated values of extracellular fluxes. di and dj are the corresponding measurement errors. To save the high computational effort caused by global search algorithms such as GA, a local search algorithm such as LMA may be used [13]. To evaluate the statistical quality of the flux estimates, a statistical analysis should be made to check the reliability of flux estimates and investigate the sensitivity of the estimated values to the measurement inaccuracies. Several hundreds of the simulated measurement data sets of mass distribution may be generated by addition of normally distributed measurement noise to the simulated measurement data set corresponding to the best fit flux distribution. The same optimization procedure as was used for the estimation of the best fit flux distribution can be applied to estimate the flux distribution from the simulated measurement data sets. Then, from the probability distribution of these flux distributions, confidence limits can be obtained for the estimated parameters [11–13]. Another approach of utilizing the analytical expression can be also useful [40, 49]. 2.6 Metabolic Flux Distribution of E. coli The best-fit flux distributions in acetate-limited and glucose-limited chemostat culture at dilution rates of 0.11 h–1 and 0.22 h–1 are given in Figs. 1 and 2 [13]. It can be seen that the flexible regulation mechanism existed in the several key junctions of the metabolic network of acetate metabolism. Although the experimental data of the specific acetate uptake rates, the oxygen consumption rates, and the carbon dioxide evolution rates increased in proportion to the cell growth rate, the flux distribution showed little change between the two different growth rates. For example, the flux through icdA was always regulated to be approximately 2.45 times the throughput of aceA, and the flux ratio of gltA to pckA was always maintained around 7.38. The pentose phosphate (PP) pathway, although an important route for glucose metabolism, appears to contribute little to the metabolism of acetate. A PP pathway flux was estimated to be below 2% of the acetate uptake rate and the cell growth rate had little influence on the

Fig. 1 Net flux distribution in acetate metabolism of Escherichia coli K12 in chemostat cultures at D of 0.11 and 0.22 h–; flux values at D of 0.22 h– are at upper values and those for 0.11 h– are at lower values

12 K. Shimizu

Metabolic Flux Analysis Based on 13C-Labeling Experiments

13

Fig. 2 Net flux distribution in glucose metabolism of Escherichia coli K12 in chemostat cultures at D of 0.11 and 0.22 h–; flux values at D of 0.22 h– are at upper values and those for 0.11 h– are at lower values

14

K. Shimizu

PP pathway relative flux. In the glucose metabolism, however, a significant change in flux distribution was found when the specific growth rate increased. It can be observed that the relative flux for PP pathway increased while that of TCA cycle decreased as the cell growth rate increased. A similar change in flux pattern of Bacillus subtilis has also been observed by Sauer [50]. Quite different from the glucose metabolism, the central metabolism of E. coli cells growing on acetate is characterized by a high activity of the TCA cycle, high activity of the glyoxylate cycle as the anaplerotic reaction, and PEP/pyruvate formation from oxaloacetate/malate. During the metabolism, acetate is fluxed via central metabolic pathways to the precursors required for the synthesis of biomass and also to generate the reducing power and ATP required to convert these precursors to biomass. Two important enzymes, isocitrate lyase (coded by gene aceA) and isocitrate dehydrogenase (coded by gene icdA) lead the carbon flux to two different cycles which are responsible for these two functions. From the mass isotopomer analysis of the entire cellular network, we could know that subtle regulation mechanism exists in certain key junctions of the network system. In the steady state, the flux through icdA is always regulated to be approximately 2.45 times that of the flux of aceA. In general, the flux through aceA generates the precursors used for biosynthesis while the flux through icdA is dedicated to the supply of reducing power and ATP for biosynthesis. The effect of the regulation system is to maintain the flux ratio of icdA to aceA so that the rate of supplying NAD(P)H and ATP is equal to the demands of biosynthesis. Thus, when growth rate increased from 0.11 h–1 to 0.22 h–1, with the increase in absolute flux through aceA to generate more precursors for biosynthesis, the flux through icdA also increased in proportion to produce more energy. From the flux distribution in the entire network system, we could also know that icdA is the major source of NADPH in acetate metabolism, causing the flux to oxidative PP pathway to be low. This is the main reason why slight change in flux distribution was observed for acetate metabolism at different growth rates. The fact that almost no change in the flux through PP pathway when the cell growth rate increased from 0.11 h–1 to 0.22 h–1 revealed that the main role of the PP pathway in acetate metabolism is to provide erythrose-4-P and pentose phosphates for biosynthesis of nucleotides and aromatic amino acids, and its function as the source of NADPH is negligible. The central glucose metabolism of E. coli cells is, however, characterized by a relatively lower activity of the citric acid cycle, the absence of glyoxylate cycle activity, and anaplerosis solely by carboxylation of PEP. From the mass isotopomer analysis, it can be seen that when the specific growth rate increased, a significantly increased PP pathway flux was observed for E. coli at the expense of the TCA cycle. It is obvious that when growing on preferred carbon sources such as glucose, icdA could not fulfill the NADPH demand for cell growth by itself, causing the oxidative PP pathway to be utilized to a larger extent to complement the NADPH demand in addition to its normal function for pentose formation.

Metabolic Flux Analysis Based on 13C-Labeling Experiments

15

3 Metabolic Regulation Analysis of E. coli in Protein Expression Level 3.1 Protein Expressions and Enzyme Activities of E. coli under Different Culture Conditions We cultivated E. coli K12 aerobically using several different substrates such as glucose, acetate, gluconate, and glycerol as the sole carbon source and microaerobically for the case of glucose used as a carbon source. The cells were harvested for enzyme activity assay and the proteome analysis by 2DE. We selected 52 detected proteins involved in glycolysis, PP pathway, Entner-Doudoroff (ED) pathway, TCA cycle, anaplerotic pathway and fermentative pathway to analyze their expression levels under different conditions. Figure 3 shows one 2DE data with the protein identification for the aerobic cultivation of E. coli using glucose as a carbon source. Simultaneously, 26 enzymes encoded by the corresponding genes were assayed for the activities [51]. Tables 1–4 show the enzyme activities involved in the central metabolic pathways of E. coli. The regulation ratios of the enzyme activities as the comparison with the control experiment for aerobic cultivation using glucose as a

Fig. 3 2-DE gel map of the total lysate of E. coli cells grown on glucose under aerobic condition

Aerobic

Activity

0.015±0.001 0.032±0.002 3.29±0.05 0.34±0.01 0.024±0.001 1.60±0.01 2.80±0.02 0.036±0.001 0.061±0.01 0.054±0.001 0.22±0.02 0.037±0.001 ND

DO level

Enzyme

PTS HEX PGI PFK FBPase FBA TPI GAPDH PGK PYK PPC PCK MEZ

1 1 1 1 1 1 1 1 1 1 1 1 ~

Control

0.017±0.001 0.039±0.001 4.66±0.04 0.54±0.02 0.033±0.001 1.92±0.01 3.50±0.02 0.051±0.002 0.080±0.01 0.084±0.001 0.13±0.01 0.040±0.003 ND

Activity

Microaerobic

1.13 1.22 1.41 1.64 1.37 1.20 1.25 1.40 1.32 1.56 0.59 1.08 ~

Ratio ND 0.022±0.002 2.88±0.02 0.19±0.02 0.052±0.001 0.93±0.02 2.27±0.02 0.018±0.001 0.28±0.01 0.016±0.001 0.050±0.001 0.11±0.03 0.008±0.001

Activity

Aerobic

Acetate

~ 0.71 0.62 0.56 2.16 0.58 0.81 0.51 0.46 0.29 0.23 2.97 ~

Ratio ND 0.027±0.002 2.86±0.05 0.17±0.02 0.12±0.02 1.35±0.04 3.57±0.02 0.039±0.002 0.068±0.01 0.058±0.001 0.26±0.02 0.066±0.003 ND

Activity

Aerobic

Gluconate

~ 0.85 0.87 0.70 5.00 0.84 1.02 1.05 1.11 1.07 1.17 1.78 ~

Ratio

ND 0.014±0.002 2.18±0.04 0.22±0.01 0.077±0.001 1.02±0.02 4.14±0.02 0.045±0.001 0.076±0.02 0.036±0.002 0.40±0.01 0.047±0.004 ND

Activity

Aerobic

Glycerol

~ 0.44 0.66 0.64 3.21 0.66 1.48 1.25 1.25 0.67 1.82 1.29 ~

Ratio

Abbreviations: PTS: Phosphotransferase system; HEX: Hexokinase; PGI: Glucose phosphateisomerase; PFK: Phosphofructosekinase; FBPase: Frucotose1, 6-biphosphatase; FBPase: Fructose biphosphate aldolase; TPI: Triose phosphate isomerase; GAPDH: Glycerodehyde-3-phosphate dehydrogenase; PGK: 3-Phosphoglycerate kinase; PYK: Pyruvate kinase; PPC: Phosphoenolpyruvate carboxylase; PCK: Phosphoenolpyruvate carboxykinase; MEZ: Malic enzyme. The unit of the enzyme activity is mmol min–1 mg (protein)–1. ND: not detected. All the measurements were performed in triplicate.

Glucose

C source

Table 1 Activities of glycolytic and anaplerotic enzymes in response to carbon sources and DO level. Ratio calculation was based on the control in the case of aerobic growth on glucose

16 K. Shimizu

Aerobic

Activity

0.35±0.02 0.28±0.02 0.45±0.02

DO level

Enzyme

G6PDH 6PGDH E-D Pathway

1 1 1

Control 0.22±0.01 0.21±0.01 0.15±0.02

Activity

Microaerobic

0.64 0.74 0.33

Ratio 0.14±0.01 0.19±0.02 0.26±0.01

Activity

Aerobic

Acetate

0.40 0.68 0.57

Ratio 0.50±0.01 0.45±0.02 3.01±0.02

Activity

Aerobic

Gluconate

1.42 1.61 6.69

Ratio

0.44±0.01 0.33±0.02 0.38±0.04

Activity

Aerobic

Glycerol

1.25 1.18 0.85

Ratio

Abbreviations: G6PDH: Glucose-6-phosphate dehydrogenate; 6PGDH: 6-Phosphogluconatedehydrogenate; E-D pathway: overall E-D pathway enzymes including 6-phosphogluconate dehydrate and 2-keto-3-deoxy-6-phosphogluconate (KDPG) aldolase. The unit of the enzyme activity is mmol min–1 mg (protein)–1. All the measurements were performed in triplicate.

Glucose

C source

Table 2 Activities of PP and E-D pathway enzymes in response to carbon sources and DO level. Ratio calculation was based on the control in the case of aerobic growth on glucose

Metabolic Flux Analysis Based on 13C-Labeling Experiments 17

Aerobic

Activity

0.48±0.02 0.66±0.05 0.005±0.002 ND

DO level

Enzyme

ACK LDH ADH PFL

1 1 1 ~

Control 0.75±0.01 1.22±0.04 0.062±0.04 0.021±0.02

Activity

Microaerobic

1.56 1.85 12.40 ~

Ratio 0.34±0.02 0.34±0.03 ND ND

Activity

Aerobic

Acetate

0.70 0.51 ~ ~

Ratio

0.55±0.04 1.01±0.05 0.009±0.002 ND

Activity

Aerobic

Gluconate

1.15 1.53 ~ ~

Ratio

0.13±0.02 0.61±0.04 0.005±0.002 ND

Activity

Aerobic

Glycerol

0.27 0.92 ~ ~

Ratio

Abbreviations: PTA: Phosphtransacetylase; ACK: Acetate kinase; LDH: Lactate dehydrogenase; ADH: Alcohol dehydrogenase; PFL: Pyruvate-formate lyase. The unit of the enzymeactivity is mmol min–1 mg (protein)–1. ND: not detected. All the measurements were performed in triplicate.

Glucose

C source

Table 3 Activities of fermentative enzymes in response to carbon sources and DO level. Ratio calculation was based on the control in the case of aerobic growth on glucose

18 K. Shimizu

Aerobic

Activity

0.051±0.000 1.15±0.02 0.013±0.002 0.022±0.002 0.061±0.004 0.056±0.00

DO level

Enzyme

GLT ICDH ICL a-KGDH FUM MDH

1 1 1 1 1 1

Control 0.0076±0.0001 0.14±0.02 0.006±0.002 ND 0.017±0.002 0.030±0.001

Activity

Microaerobic

0.15 0.12 0.46 ~ 0.28 0.54

Ratio 0.25±0.01 0.26±0.02 0.12±0.02 0.058±0.001 0.20±0.01 0.15±0.01

Activity

Aerobic

Acetate

4.90 0.23 9.23 2.91 3.27 2.68

Ratio

0.32±0.01 2.20±0.04 0.019±0.001 0.060±0.001 0.11±0.01 1.78±0.01

Activity

Aerobic

Gluconate

6.27 1.91 1.46 2.73 1.80 1.96

Ratio

0.23±0.02 1.88±0.04 0.017±0.003 0.065±0.001 0.10±0.01 0.10±0.01

Activity

Aerobic

Glycerol

4.51 1.63 1.31 2.95 1.64 1.78

Ratio

Abbreviations: GLT: Citrate synthase; ICDH: Isocitrate dehydrogenase; ICL: Isocitrate lyase; a-KGDH: a-Ketoglutarate dehydrogenase; FUM: Fumarase; MDH: Malate dehydrogenase. The unit of the enzyme activity is mmol min–1 mg (protein)–1. ND: not detected.All the measurements were performed in triplicate.

Glucose

C source

Table 4 Activities of TCA cycle and glyoxylate shunt enzymes in response to carbon sources and DO level. Ratio calculation was based on the control in the case of aerobic growth on glucose

Metabolic Flux Analysis Based on 13C-Labeling Experiments 19

20

K. Shimizu

carbon source were also calculated and listed in the tables. In order to see how the protein expression levels changed depending on the culture condition, the abundances of each protein under different culture conditions were divided by the corresponding one in the control experiment. Those ratios are shown in the metabolic pathway maps of Fig. 4 for the case of microaerobic condition as compared with the control experiment. The red lines imply up-regulation, while the blue lines imply down-regulation. The thickness of the arrow depends on the magnitude of the change in the protein expression level [51]. 3.2 Glycolysis and Anaplerotic Pathway An early investigation showed that the glucose transport genes exhibited high basal expression, and the ptsHI genes were positively stimulated by cAMP-CRP receptor protein and also by the growth on glucose, while crr promoters within ptsI may be negatively regulated by CRP-cAMP [52, 53]. More recently, Plumbridge et al. [54] reported that these genes are in the same operon, and the operon was known to be regulated by mlc. In the absence of glucose, mlc represses the operons. In the presence of extracellular glucose, the conformation of EIIBCglc protein is changed and bound strongly with mlc, which no longer represses the operon. On the other hand, an essential feature of the pts system is that the phosphoryl donor molecule is PEP, not ATP. The experimental results show that the glucose transport genes ptsHI-crr were slightly up-regulated under microaerobic condition in comparison to the aerobic growth on glucose, and significantly down-regulated by two- to fivefold when the carbon source was changed to acetate, gluconate, or glycerol. The enzyme activity of glucose: PEP phosphotransferase was only detected in glucose medium (see Table 1), confirming the efficient induction by glucose either aerobically or microaerobically. Glk encoding the first enzyme glucokinase which drives glucose enter the glycolytic pathway in the cell was kept relatively unregulated at more than about 1.2-fold in all cases. The enzyme activity of hexokinase was observed to increase under microaerobic condition in glucose medium, while decreased in acetate, gluconate or glycerol medium under aerobic condition compared with the case of glucose medium (see Table 1). This enzyme is thought to be inhibited allosterically by its product G6P. Most of the glycolytic genes pgi, pfkA, fba, gapA, pgk, eno, pykF were also observed to be up-regulated under microaerobic condition within twofold, indicating a “stepping-up” of anaerobic glucose utilization via glycolysis. The rapid fall of TCA cycle enzyme activities (see Table 4) and the significant increase in fermentative enzyme activities (see Table 3) in microaerobic condition indicate that glucose utilization was switched towards fermentation under microaerobic condition. In contrast, the common glycolytic genes pgi, pfkA, fba, gapA, eno, pykF, and anaplerotic ppc were highly repressed by about two- to fourfold in the acetate medium compared to those in the glucose medium. Simultaneously, the glu-

Metabolic Flux Analysis Based on 13C-Labeling Experiments

21

Fig. 4 The relative expression level of E. coli K12 proteins of central metabolic pathways under microaerobic growth on glucose as compared with aerobic growth on glucose was used as the control. The proteins were annotated by their coding gene names. The numbers below the gene names represent the regulation ratios of the protein expression levels as the comparison with the control, which were calculated by taking the ratios of the abundances of the proteins in the designated condition to the corresponding ones in the control experiment. The red arrows indicate the induced proteins, and the blue arrows indicate the repressed proteins. The thicker the arrows, the higher the proteins were regulated. The black arrows indicate the similar levels of the protein expression as the control experiment or undetected on the 2DE gels

22

K. Shimizu

coneogenic genes fbp, pckA, ppsA were observed to be highly induced about 2.5-fold, 3.7-fold, and 8.3-fold, respectively, which were thought to be subjected to positive regulation by the catabolite repressor/activator cra, a global catabolite repression regulator [55], and fructose repressor fruR, a global regulatory gene concerned with carbon utilization by transcriptional modulation of the target genes [56]. The enzyme activities given in Table 1 agree with the fact that carbon flow will be channeled through gluconeogenic pathway when acetate was metabolized, and the flux in gluconeogenic direction is much smaller than the glycolytic flux during growth on glucose. Interestingly, in the reaction between F6P and F1, 6BP, PEP and PYR, PEP and OAA, the involved enzymes were tightly regulated. The pfkA, coding for the main phosphofructokinase, Pfk-1 was repressed by about twofold, while pfkB coding for the minor phosphofructokinase Pfk-2 was kept nearly unchanged. Correspondingly, the activity of Pfk dropped 1.8-fold in acetate medium, while fbp gene was up-regulated 2.5-fold, and the enzyme activity of FBPase varied in a coordinate manner, and increased 2.2-fold (see Table 1). It has been proved that pfkA is subject to the regulation by the catabolite repressor/activator cra and fruR, while the enzyme Pfk-1 is allosterically activated by ADP and inhibited by PEP and ATP. The inhibitory effect of ATP is opposed by AMP, and intensified by citrate which acts as a signal of the availability of alternative sources of ATP [57]. This control might link to an increase in the rate of Pfk-1 synthesis under anaerobic condition. Pfk-2, insensitive to ATP inhibition, may serve as another role (for instance, maintain the futile cycle for the regulatory amplification) [58, 59]. The down-regulation of PykF, but not PykA, suggests that these two isoenzymes were differentially regulated. Indeed, pykF has been suggested to be negatively regulated by cra and fruR, the global regulatory genes. PykA may serve as a different role. The ppc gene was down-regulated threefold in the acetate medium.Although the regulation of this gene has not been reported, this result is consistent with the direction of the metabolic flux. The carbon flux from oxaloacetate (OAA) to phosphenolpyruvate (PEP) will be dominant in such a case, and it is catalyzed by pckA. In fact, the pckA gene was up-regulated by 3.7-fold, which is also modulated by fruR at the transcriptional level. The enzyme activities were coordinately regulated, and Pck increased nearly threefold, while Ppc decreased 4.3-fold (see Table 1). High level expression of both ppc and pckA led to a futile cycling. Moreover, the ppsA gene was also found to be highly induced by about 8.3-fold in acetate medium. In addition, malic enzyme, Mez, which converts malate to pyruvate, showed low activity only in acetate medium, and it was not detectable on the 2DE gel because of its low abundance. This result suggests that Pps and Mez may play an important role for the gluconeogenic flux during the metabolism of acetate. In fact, the induction of ppsA and mez genes of E. coli during growth in acetate medium has recently been confirmed by DNA microarray [60, 61]. In contrast, most of the glycolytic genes were not affected significantly during growth in gluconate medium. An exception is fbp gene, which was induced

Metabolic Flux Analysis Based on 13C-Labeling Experiments

23

by 4.5-fold. The enzyme activity of FBPase increased fivefold in gluconate medium (see Table 1). The high induction of fbp gene will drive part of GAP from ED pathway through gluconeogenesis to supplement G6P. Surprisingly, the pckA gene was found to be induced by 1.6-fold, and the enzyme activity of Pck varied in accordance (see Table 1). The same phenomenon with pckA appeared in cells grown on glycerol. However this induction is unexpected since the flux in Pck-mediated reaction is not needed in both cases. The pckA may be induced gratuitously by the increased level of cAMP during growth in nonglucose medium. It has been reported that the concentration of cAMP in glycerolgrown cells was much higher than that in glucose-grown cells of E. coli [62, 63]. During growth on glycerol, one significantly affected gene was pykA, which was induced by 2-fold. When E. coli was cultivated in glycerol, the flux from PEP to pyruvate is governed by pyruvate kinase rather than the phosphotransferase system. Therefore, it is reasonable to expect that one of the pyk genes is up-regulated to satisfy the significantly increased demand for the flux from glycolysis. Apparently the pykA gene rather than pykF serves this role. Since pykF is activated at the protein level by F1, 6BP, whose concentration is relatively low, is not a good indicator for the metabolic state during growth in glycerol [64]. Therefore, pykF may remain nearly inactive due to the low concentration of its allosteric activator. PykA, on the other hand, is activated by cAMP, which is higher in glycerol medium than that in glucose medium [62, 63], and its activity can still be mediated at the protein level. Therefore, pykA is a better choice during growth on glycerol. 3.3 Pentose Phosphate Pathway and Entner-Doudoroff Pathway The two enzymes involved in the oxidative PP pathway, glucose 6-phosphate dehydrogenase (G6PDH) encoded by zwf gene and 6-phosphogluconate dehydrogenase (6PGDH) encoded by gnd gene, were found to be down-regulated during microaerobic growth in glucose medium and aerobic growth in acetate medium. On the other hand, both enzymes were up-regulated when the cells grew in gluconate or glycerol medium. The enzyme activities varied in a coordinated manner (see Table 2). The ED pathway was shown to be present in all cases, and significantly induced to a higher level on gluconate based on the measurements of the overall activity of the ED enzymes and the 2DE result. Table 2 shows the activities of the two ED pathway enzymes in E. coli under the tested conditions. Both enzymes were present at high levels during growth on gluconate, where edd gene was highly up-regulated by 7.4-fold, and eda was induced by twofold as compared with the case of using glucose as the carbon source, and the overall activity of ED pathway enzymes coordinately increased by about sevenfold (see Table 2). The overall activity of the ED enzymes, including 6PG dehydratase (edd) and KDPG aldolase (eda), was assayed by measuring 6PG-dependent formation of pyruvate, which was determined colorimetrically as its dinitro-

24

K. Shimizu

phenylhydrazone [65, 66]. This assay procedure underestimates the level of KDPG aldolase, which is usually present in excess compared with 6PG dehydratase. It measures the rate-limiting component in the pathway, 6PG dehydratase. Both enzymes G6PDH and 6PGDH were known to be subject to the cellular growth rate regulation, which are proportional to the growth rate influenced by the medium [67, 68]. Indeed, the cell growth was quite slow in glucose under microaerobic condition and in acetate medium, while it was faster during growth on gluconate and glycerol as compared with the cell growth on glucose. In addition, an early study reported that the gnd enzyme was induced by gluconate [65]. The other genes involved in the non-oxidative metabolism such as rpe, rpi, tal, tkt did not differ significantly between growth conditions. It is known that the edd gene, containing a regulatory region, is induced by gluconate, and eda gene probably being induced by EDGP, the product of edd [69, 70], and both genes are subjected to negative control by fruR. This result indicates that ED pathway was predominant in dissimilation of gluconate in E. coli, while insignificant in the metabolism of acetate, glucose, or glycerol. 3.4 Fermentative Pathway Among the fermentative genes, pfl and adhE were dramatically induced by about 11.2-fold and 10.8-fold, respectively, in response to the oxygen level shift from aerobic to microaerobic condition in glucose medium (see Fig. 4). The enzyme activity of Pfl was only detected in the microaerobic condition and the activity of NADPH-dependent ADH increased by 12.4-fold under microaerobic condition as well (see Table 3). Regulation of pfl synthesis and activity was subjected to the control by the fnr protein and arcA/B two-component regulatory system. Fnr, an oxygen-sensing global regulator, which is an iron sulfur-dependent DNA-binding protein and recognizes a specific sequence motif found in the promoter regions of the genes it regulates, serves as an activator of the transcription of the anaerobically regulated genes [71, 72]. AcrA is a transcriptional repressor of genes encoding enzymes for aerobic metabolism, and arcA/B system functions in transcriptional regulation under both anaerobic and aerobic conditions. The significant anaerobic induction of the pfl operon was thought to be accounted for the fnr, and arcA and arcB which mediate the residual transcriptional activation of the operon [73, 74]. However, the expression of pfl was not changed much in acetate, gluconate or glycerol medium, although this fermentative pathway was not used when growing aerobically in acetate medium [74]. Expression of adhE was also strongly induced by microaerobiosis. However, this induction is independent of the fnr and arcA transcription factors. There appears to be a direct correlation between the NADH/NAD+ ratio and enzyme synthesis; the higher the ratio, the more ADH is synthesized. Moreover, the

Metabolic Flux Analysis Based on 13C-Labeling Experiments

25

ADH protein itself may exert a regulatory function, since the gene expression was enhanced dramatically in the adhE mutant [75]. Therefore, the induction of adhE is not surprising, since the previous study showed that the significant differences of the NADH/NAD+ ratios between aerobic and anaerobic cultures (0.02 and 0.75, respectively) have been observed as the DOT (dissolved oxygen tension) of the culture was decreased [73]. The ldhA gene, encoding NAD+linked enzyme LDH, was also highly enhanced by twofold during microaerobic fermentation of glucose. This result is consistent with the previous study by Mat-Jan et al. [76], who found that the fermentative LDH was cojointly induced by anaerobic condition and acidic pH [73]. Since E. coli satisfies energy requirements through glycolysis at the accelerated rate under anaerobic condition, the large amount of NADH produced must be re-oxidized to NAD+, which is required to maintain glycolysis since it is again the substrate for GAPDH in glycolysis. In the presence of oxygen, the oxidation of NADH occurs through molecular oxygen, while in the absence of oxygen it proceeds through reduction of an organic acid, and hereby the conversion of pyruvate to lactate by LDH is preferable. The LDH activity changed in response to carbon sources. Namely, it was repressed in acetate, slightly enhanced in gluconate, and almost unchanged in glycerol. The pta and ackA genes, involved in reversible acetyl CoA metabolism in the cell, were found to be up-regulated by 1.3-and 1.7-fold, respectively, in the case of the microaerobic condition (see Fig. 4). The enzyme activity of Ack in microaerobically grown cells was 1.56 times as high as that of the case grown aerobically in glucose medium (see Table 3), in accordance with the gene expression level. The pta and ackA genes are constitutively expressed and present in the same operon, but are regulated differentially through different promoters. It has been proposed that the intermediate of this pathway, acetyl phosphate, might be an important effector of gene regulation, while the levels of acetyl phosphate vary dramatically depending on the carbon source in the growth medium. For example, in the defined medium under limiting phosphate concentrations, very low levels of acetyl phosphate were observed when cells were grown on glycerol (<40 mmol/l), moderate levels on glucose (300 mmol/l), and high levels on pyruvate [77, 78]. During growth on glycerol, Ack was down-regulated by almost threefold (see Table 3). This result is consistent with the observation that acetate production in glycerol medium is much lower than in glucose medium. However, a relatively small change of pta was observed. It may perhaps ensure the production of acetyl phosphate for regulatory purposes. During growth on acetate, both pta and ackA were downregulated. This result suggests that other genes may be responsible for acetate uptake during growth on acetate. Indeed, it was proposed that the acs gene (coding for acetyl-CoA synthetase) is mainly responsible for acetate uptake, whereas the pta-ackA pathway is used for acetate excretion during growth on glucose [61]. In gluconate medium, both genes were observed to be up-regulated. The measured enzyme activities of Ack were roughly correlated with the protein expression levels.

26

K. Shimizu

3.5 TCA Cycle and Glyoxylate Shunt It was observed that there was a significant effect of the change in either oxygen level or carbon source on the TCA cycle genes. Under microaerobic condition, the TCA protein expressions fell down to around two- to tenfold as compared with those under aerobic condition. In contrast, if acetate, gluconate or glycerol was used as the carbon source instead of glucose, most of the TCA cycle genes were up-regulated by two- to sixfold. Citrate synthase, encoded by gltA gene, the first enzyme to enter TCA cycle, was found to be suppressed by oxygen limitation and excess glucose, but elevated by oxygen (fivefold) and other oxidized carbon sources (four- to sevenfold) such as acetate, gluconate, or glycerol. The a-ketoglutarate dehydrogenase complex was encoded by sucA and sucB, and shared the third gene lpdA with pdh complex. These genes were strongly suppressed by oxygen limitation and induced by acetate and other oxidized carbon sources. Indeed, the enzyme activity of a-ketoglutarate dehydrogenase was not detectable in glucose medium under microaerobic condition (see Table 4). These observations coincide with an early study of Amarasingham et al. [79]. At the important control point of the junction between TCA cycle and glyoxylate shunt, aceA and aceB were cojointly induced by more than tenfold in acetate medium. The enzyme activity of isocitrate lyase (Icl) increased more than ninefold in acetate medium (see Table 4). The highly regulated glyoxylate shunt was absolutely essential to direct the carbon flow through this bypass to generate four-carbon precursor for the biosynthesis in acetate metabolism (Chung et al. 1988). Measurements of TCA cycle enzymes (see Table 4) were shown to be dependent on the gene expression levels, to a large extent. These genes are known to be subjected to cAMP-CRP mediated catabolite repression and arcA, fnr, and soxRs oxidative stress regulatory system mediated aerobic-anaerobic repression [80].Although there was a general reduction in TCA cycle gene expression and enzyme activity related to microaerobic condition in glucose medium, the variations of individual genes were not coordinated, implying that these genes were subjected to different regulations and different regulatory mechanism. For instance, fumA, fumB, and fumC were regulated differently by the oxygen availability, while fumB was shown to be slightly abundant under microaerobic condition (1.4-fold) (see Fig. 4), since the expression of fumB gene is controlled by anaerobic transcriptional activator, fnr, whereas the expression of fumA and fumC decreased three- to fivefold when the oxygen level was reduced to microaerobic condition as compared with completely aerobic condition (see Fig. 4). It was proposed that anaerobic expression of fumA promoter was depressed by arcA and fnr, while expression of fumC was depressed only by arcA, but required the soxRs regulatory protein, which is a positive regulator for controlling synthesis of proteins in response to oxidative stress [81, 82]. In addition, the observation of fourfold higher expression of fumA than that of fumC

Metabolic Flux Analysis Based on 13C-Labeling Experiments

27

suggested that fumA was predominant under aerobic condition. However, fumC is relatively less affected by the type of carbon sources than fumA, indicating that fumA is shown to be under catabolite control. AcnA showed two- to threefold lower expression due to catabolite and anaerobic repression, but acnB expression was relatively constant, although threefold up-regulated in acetate medium since acnB is used preferably in glyoxylate shunt, indicating that acnB was regulated differently. AcnA appeared to be activated by the soxRs oxidative stress regulatory system and fnr [80]. It was demonstrated that arcA functions as a repressor of gltA expression during both aerobic and anaerobic growth through lacZ fusion study [83]. The relative pattern of carbon control of mdh, gltA, and sdhC gene expressions were similar. However, the magnitude of oxygen control of mdh and gltA gene expressions were larger than that of sdhC. The difference in magnitude may reflect the functioning of the mdh and gltA protein under both aerobic and anaerobic conditions, whereas the sdhCDAB proteins are primarily used for aerobic cell growth [71, 84, 85]. IcdA gene encoding isocitrate dehydrogenate, competing for the common substrate with isocitrate lyase at the node, was shown to be up-regulated by 1.5-, 2.0-, and 1.8-fold in acetate, gluconate and glycerol medium, respectively. Nevertheless, the measurement of the activity of isocitrate dehydrogenase did not agree with this result in the case of acetate, which reduced nearly 4.3-fold (see Table 4). This fact may be explained as follows: the icdA gene was induced by acetate, and in turn led to an increased amount of the enzyme protein. However, in order to prevent the excess flux of isocitrate through ICDH, the enzyme activity was regulated by reversible inactivation of ICDH, which is catalyzed by bifunctional kinase/phosphatase, encoded by ackK, since the phosphorylated form of ICDH has no activity [86].Although the regulations of the enzyme ICDH dehydrogenase and Icl activities were opposite in acetate medium, the absolute enzyme activity of ICDH which leads the carbon flow to TCA cycle was still dominant over the glyoxylate bypass directing enzyme, Icl (see Table 4) because of the balance of the rate of supply of NADPH and ATP with the demand for biosynthesis. 3.6 Comparison of the Protein Expression Level and Enzyme Activity One of the problems associated with the 2DE result is the accuracy of the spot identification. It may be considered to identify the protein using amino acid sequencer or TOF-MS etc. [87]. We also compared the 2DE result with the enzyme activity to show how those are correlated [51]. Figure 5 shows the correlation of the protein expression levels and the corresponding enzyme activities. The logarithmic ratio of the enzyme activity was plotted vs the logarithmic ratio of the protein expression level in the figure. As can be seen in Fig. 5, most of the spots are close to the 45° line except isocitrate dehydrogenase under aerobic acetate condition, indicating that the protein abundances revealed by 2DE are correlated, to a large extent, with the enzyme

28

K. Shimizu

Fig. 5 Comparison of the logarithmic protein expression ratios based on 2DE and the corresponding enzyme activity ratios. Each set of data was plotted on the figure by the logarithmic ratio valve

activity. However, it is obvious that there exist deviations from the linear correlation between the logarithmic protein expression ratios and the enzyme activity ratios. The deviations may be caused by the fact that protein products expressed from a single gene can migrate to multiple spots on 2DE gels, and also the proteins from multiple genes can run to the same coordinates on a gel because of differential protein processing and posttranslational or artifactual modifications [88, 89]. Both differential migration and comigration of proteins complicate comparative and quantitative pattern analysis of 2DE gels. On the other hand, although the enzyme activity is primarily determined by the protein expression level, subtle regulations were exerted by many effectors at the enzyme level to satisfy the need for the metabolism. Isocitrate dehydrogenase in acetate medium may be such a case, which was regulated by reversible inactivation catalyzed by bifunctional kinase/phosphatase, as discussed above. Moreover, the Ppc and Pck enzymes, involved in the anaplerotic and gluconeogenic pathways were also shown to be subject to fine control by some effectors.

Metabolic Flux Analysis Based on 13C-Labeling Experiments

29

4 Metabolic Regulation Analysis of pgi – Mutant E. coli by Gene Expression The pgi inactivation in E. coli forces glucose catabolism via the PP pathway. As a consequence, metabolic redirection in a pgi mutant overproduces NADPH, and thus may inhibit the cell growth due to NADPH imbalance. An interesting idea of incorporating NADPH consuming pathway such as poly(3-hydroxybutyrate) (PHB) synthetic pathway may then arise. For PHB synthesis, three enzymes are necessary from acetyl-CoA such as b-ketothiolase, acetoacetyl-CoA reductase, and PHB synthase. Among them, the reaction catalyzed by acetoacetyl-CoA reductase requires NADPH [90–92]. Thus, NADPH imbalance in pgi mutant may possibly be overcome by incorporating phb genes in E. coli.We analyzed 87 E. coli gene expressions in glycolysis, PP pathway, and the TCA cycle etc. by RT-PCR. Moreover, we measured some of the enzyme activities to see the metabolic regulation mechanism from the viewpoint of gene and protein expressions [93]. Some of the gene expressions for E. coli JM109 are shown in Fig. 6. A number of significant changes in differential gene expressions obtained by semiquantitative RT-PCR analysis were observed in pgi mutant E. coli DF11 and in recombinant pgi mutant E. coli DF11/pAeKG1 having phb genes as compared with those in E. coli JM109, and the results are shown in Figs. 7 and 8, respectively, where the values shown in the figures are the ratios of the gene expressions as compared with the corresponding gene expressions of E. coli JM109 [93]. 4.1 Glycolysis and Pentose Phosphate Pathway Genes Figures 7 and 8 show that the glycolysis and PP pathway genes were affected significantly by expression of phb genes in pgi mutant E. coli DF11/pAeKG1 as well as in pgi mutant E. coli DF11 without phb genes. The results show that the transcript abundance of several glycolytic genes such as pfkA, fbaA, tpiA, gapA, pgk, eno, pykA, ppsA, and pckA were increased 1.5- to 12.9-fold in recombinant pgi mutant E. coli DF11/pAeKG1 having phb genes compared to those in control E. coli JM109 (see Fig. 6). Since the first six of those genes catalyze reversible reactions, their upregulation enhances the flux in the glycolytic direction, which is much greater than the gluconeogenic flux during growth in glucose. The down regulation of pykF in both E. coli DF11 and E. coli DF11/pAeKG1, but not pykA (Figs. 7 and 8) suggests that these two isoenzymes were differentially regulated. Indeed, pykF, but not pykA, has been regarded to be negatively regulated by fruR [94], a global regulatory gene that was found to be upregulated in both E. coli DF11 and E. coli DF11/pAeKG1. Interestingly, however, ppsA (coding for phosphoenol pyruvate synthase) was upregulated 12.9-fold in E. coli DF11/pAeKG1 but was almost unchanged in E. coli DF11 compared to those in E. coli JM109.

30

K. Shimizu

Fig. 6 The transcript levels of central metabolic pathway genes for E. coli JM109. Symbols for genes inside the boxes follow the E. coli K-12 linkage map

Metabolic Flux Analysis Based on 13C-Labeling Experiments

31

Fig. 7 The transcript levels of central metabolic pathway genes for E. coli DF11 relative to those in E. coli JM11. Symbols for genes inside the boxes follow E. coli K-12 linkage map. Red represents upregulation, blue represents downregulation, and black represents no change

32

K. Shimizu

Fig. 8 The transcript levels of central metabolic pathway genes for E. coli DF11/pAeKG1 relative to those in E. coli JM11. Symbols for genes inside the boxes follow E. coli K-12 linkage map. Red represents upregulation, blue represents downregulation, and black represents no change

Metabolic Flux Analysis Based on 13C-Labeling Experiments

33

The gene glk was upregulated by 1.8 and 1.2-fold in E. coli DF11 and E. coli DF11/pAeKG1, respectively, while the glucose phosphotransferase system genes ptsG and ptsH were downregulated compared to those in E. coli JM109 (Figs. 7 and 8). The decreases in the ptsG and ptsH transcript levels were consistent with the reduced specific glucose consumption rates. The increase in glk transcript level corroborates a previous report which shows that glucokinase takes over the major role (from the glucose-PTS) to supply G6P under increased protein synthesis conditions [95], although the mechanism for this induction is still unknown. METAFoR analysis by NMR has shown that a minor fraction of glucose catabolism occurs via ED pathway in both wild type and pgi mutant strain. Examination of the operational catabolic pathways and their flux ratios using [U-13C6] glucose-labeling experiments and metabolic flux ratio analysis provide evidence for the PP pathway as the primary route of glucose catabolism in pgi mutant E. coli [96]. Among the PP pathway genes, most of the gene transcripts in both E. coli DF11 and E. coli DF11/pAeKG1 were upregulated except gnd gene compared to those in E. coli JM109 (Figs. 7 and 8). The expression of gnd, which is regulated by the growth rate [97] clearly decreased in pgi mutants as the growth rate reduced as compared with that of E. coli JM109. By contrast, talA was upregulated in pgi mutant E. coli, although talB, the isoenzyme of talA, showed no change. In addition, edd was upregulated in both pgi mutants, although eda was downregulated significantly as was shown in Figs. 7 and 8. Basically, reoxidation of NADPH can potentially be achieved by three reactions in E. coli: (i) the NADPH dependent malic enzyme; (ii) the membrane bound transhydrogenase pntAB [98]; and (iii) the soluble transhydrogenase udhA [99]. Significant involvement of the malic enzyme in NADPH reoxidation can be excluded, since virtually no pyruvate originates from malate [96], and this result was consistent with our measurement of no malic enzyme activity. However, expression of maeB was moderately downregulated in both pgi mutants E. coli compared to that in E. coli JM109 (Figs. 7 and 8) indicating that a different kind of regulation for maeB after transcription initiation. Similarly, pntAB is unlikely to catalyze sufficient reoxidation of NADPH since pgi pntAB double mutant grows as slowly as the pgi mutant [100], which means that additional knockout of pntAB does not further reduce the rate of glucose catabolism. The primary metabolic function of pntAB appears to be in the gene ration of NADPH, since zwf pnt double mutant grows slower than the zwf mutant [100], as is expected when pntAB produces NADPH in the absence of PP pathway flux (in the zwf mutant). Expression level of pntA and pntB were upregulated in both pgi mutants compared to those in control strain (see Figs. 7 and 8), indicating that inactivation of pgi leads to the overproduction of NADPH in PP pathway, which apparently disturbs the reducing power balance in pgi mutant and that ultimately reduces the biomass synthesis. Overexpression of soluble transhydrogenase udhA in pgi mutant E. coli increased specific growth rate by about 25% that evidenced the ability of the soluble transhydrogenase udhA in the reoxidation of NADPH [96]. The down regulation of

34

K. Shimizu

udhA gene expression in both pgi mutants indicates an insufficient reoxidation of NADPH. It was observed from our experimental results that the expression of phb genes in pgi knockout E. coli increased the biomass production as compared with the pgi mutant without phb genes. The concentration of NADPH was much higher in E. coli DF11 than those of E. coli JM109. However, the NADPH concentration became lower for E. coli DF11/pAeKG1 as compared with that of E. coli DF11, which clearly evidenced the vital involvement of phbB genes in NADPH reoxidation. Since glucose is metabolized exclusively via PP pathway in pgi mutant, the important glycolysis metabolite GAP is supplied from PP pathway. GAP is one of the important precursor metabolites for the synthesis of isopentenyl diphosphate (IPP) that acts as a carrier in the synthesis of a number of cell envelopes and extracellular polymers (e.g., peptidoglycan, lipopolysaccharide, techoic acids, etc.).As a result, large fraction of GAP is being consumed to maintain the glycolytic flux for biomass production and to provide the precursor metabolites for PHB synthesis. Therefore, the expression of the gene gapA (12.4-fold increase) in E. coli DF11/pAeKG1 might have been upregulated highly to enhance the rate of conversion from GAP to glycerate-1,3-diphosphate to provide increasing demand for acetyl-CoA, since the PHB synthesis pathway competes for acetyl-CoA with three other metabolic pathways [91, 101]. The possible reason of ppsA upregulation in E. coli DF11/pAeKG1 might be due to the shortage of PEP and GAP. Another gene pckA also upregulated in E. coli DF11/pAeKG1, which also implies the above idea. It may be noted that ppsA upregulation might also be due to the effect of global regulatory gene, fruR since ppsA was found to be positively regulated by fruR [94]. The ppc gene was downregulated in both E. coli DF11 and E. coli DF11/pAeKG1, whereas pckA gene was upregulated in both strains. The regulation of these genes has not been fully elucidated, although it has been shown that ppc overexpression affects acetate production [102]. Thus, the downregulation of ppc expression that occurs during PHB synthesis may result in an increase in the acetate production flux. Furthermore, during growth in glucose, ppc is the major route to the aspartate family of amino acids. Therefore, downregulation of ppc has to be compensated by another anaplerotic pathway. The pyruvate dehydrogenase complex is composed of multiple copies of three separate enzymes: pyruvate dehydrogenase (encoded by aceE), dihydrolipoate transacetylase (encoded by aceF) and lipoamide dehydrogenase (encoded by lpdA). The reactions of the PDH complex serve to interconnect the metabolic pathways of glycolysis and fatty acid synthesis to the TCA cycle as well as to the PHB synthetic route. Although the activity of the PDH complex is highly regulated by a variety of allosteric effectors and by covalent modification, it can be expected that a reduced level of expression of these genes, i.e., reduced level of synthesis of this complex would limit the accumulation of acetyl-CoA and thus PHB synthesis.

Metabolic Flux Analysis Based on 13C-Labeling Experiments

35

4.2 TCA Cycle, Glyoxylate Shunt, and Fermentation Genes Most of the TCA cycle genes were downregulated except icdA in both pgi mutants E. coli compared to those in E. coli JM109. This downregulation was expected since most of the TCA cycle genes such as gltA [103], sucA [104], sdhC [105], fumA [106], and mdh [107] are known to be regulated by several regulators such as fnr and arcA that were found to be upregulated in both pgi strains. The upregulation of icdA may be due to the positive regulation of fruR [94]. In addition, the genes involved in the glyoxylate shunt (aceA and aceB) were upregulated in both pgi mutants E. coli indicating the shortage of C4 metabolites such as OAA and thus need to activate the anaplerotic pathway as mentioned in relation to ppc gene. Among the fermentative genes, fdhF, coding for formyl dehydrogenase, increased in E. coli DF11, but not in E. coli DF11/pAeKG1, compared to that in E. coli JM109. The upregulation of ackA in both pgi mutants is of interest. Pta and ackA genes were thought to be involved in both acetate synthesis and consumption, and constitutively expressed in the cell [108]. The upregulation of ackA in pgi mutants indicates that although both genes are present in the same operon, they are regulated differentially through different promoters. The pflA, pflD, ldhA, and adhE transcripts were downregulated in both pgi mutants E. coli as compared with those in E. coli JM109 (Figs. 7 and 8). 4.3 Heat Shock and Regulatory Genes Metabolically engineered E. coli undergoes a variety of physiological changes upon accumulation of PHB, which is not a normal metabolite of E. coli and it is considered as a stress on the cells inducing heat shock response. Therefore, the transcript levels of many heat shock proteins such as dnaK, dnaJ, groS, grpE, htpG, ibpA, and ibpB were all upregulated significantly in recombinant pgi mutant E. coli DF11/pAeKG1 as compared with those in E. coli JM109 (Fig. 8). Moreover, although those heat-shock genes except ibpA and ibpB were also found to be upregulated in E. coli DF11 (see Fig. 7), the level of expression varied among the heat-shock genes. Furthermore, the transcript level of groL and lon were slightly upregulated in both pgi mutants E. coli. The expression level of the genes for the protein secretion and disulfide bond formation, secA, lepB, and dsbA, were increased significantly in E. coli DF11/pAeKG1 compared to those in E. coli JM109. This result implies that PHB accumulation in pgi mutant E. coli increases the activity for protein secretion and disulfide bond formation. The expression ratio of the translation elongation factor gene tufA was significantly higher in E. coli DF11/pAeKG1 as compared with those in E. coli JM109, which is consistent with their coordinated regulation with the ribosomal protein gene rpsA [109]. Expression of uspA (coding for universal stress protein) was upregulated in both pgi mutants

36

K. Shimizu

E. coli, whereas the expression of uspB was almost unchanged compared to those in E. coli JM109. The significant upregulation of the gene fkpA in E. coli DF11/pAeKG1 may be considered to be due to the fact that fkpA plays an active role either as folding catalysts or as chaperones in extracytoplasmic compartments [110]. Furthermore, the takeover of the cytosolic space by PHB granules may disturb the normal intracellular architecture such as chromosomal attachment and may consequently result in the heat shock response. Bacteria naturally accumulating PHB synthesize phasin protein that covers the surface of PHB granules. E. coli does not produce PHB naturally, and therefore does not have the phasin gene [111]. Therefore, the hydrophobic PHB granules are in direct contact with intracellular biomolecules including DNA, RNA, and proteins. Certainly, this will become a major stress on the cells for several possible reasons including protein denaturation on the surface of PHB granules. This unfavorable condition generated by PHB accumulation might have resulted in the highest expression of rpsA (coding 30S ribosomal protein S1) in E. coli DF11/pAeKG1 (see Fig. 8), and it may have caused further increase in the soluble protein concentration since S1 is the largest ribosomal protein [112], present in the small subunit of the E. coli 70S ribosome that has a pivotal role in stabilizing the mRNA on the ribosome for protein synthesis [113]. Moreover, protein S1 has been reported to be necessary in some cases for translation initiation [114] and for translation elongation [115]. However, expression of the gene rpsA could be the effect of acetate accumulation as was observed by Arnold et al. [116]. In our study, highest expression of the gene rpsA in E. coli DF11/pAeKG1 was not due to the effect of acetate accumulation, since if it were caused by acetate effect, its expression should be decreased. Therefore, the highest amount of soluble protein synthesis in pgi mutant having phb genes was due to the effect of PHB accumulation that ultimately resulted in the highest expression of the gene rpsA. This phenomenon implies that protein synthesis capacity in E. coli DF11/pAeKG1 was much increased in the presence of plasmid. 4.4 Relationship Between Gene Expression and Enzyme Activities Since the cell’s physiological state is dictated at the protein level, mRNA expression results should be complemented by the enzyme activities to provide a better understanding of the observed phenomenon considering that the enzymatic regulations and transcriptional regulations are different. The semiquantitative RT-PCR results reflect the relative abundance of the transcript, which is influenced by both transcriptional activity and the mRNA stability. In contrast, the activity of enzymes can be regulated by reversible binding of effectors, covalent modification and alteration of enzyme concentration. However, to a large extent, the mRNA expression level reflects the enzyme activity [117]. To check how gene expression and enzyme activities correlate each other, enzyme activities of glk, zwf, tpiA, fbaA, ldhA, gltA, aceA, mdh, and maeB were

Metabolic Flux Analysis Based on 13C-Labeling Experiments

37

Fig. 9 The correlation between enzyme activities and mRNA transcripts on a log-log scale

measured. We compared the enzyme activities with gene expressions for three strains. Figure 9 shows how those are correlated except for aceA. Those results show that gene expression data obtained by semi-quantitative RT-PCR were reasonably consistent with the data obtained from enzyme activities except for the gene aceA indicating a different level of regulation for that gene after transcription initiation. This phenomenon indicated that aceA was expressed in gene level but glyoxylate pathway enzyme was not active for E. coli JM109, while it became active for pgi mutants.

5 Integration of Information for Metabolic Regulation Analysis for Gene-Knockout E. coli 5.1 Metabolic Regulation of pykF– Mutant of E. coli 5.1.1 Fermentation Characteristics and Enzyme Activities for Batch Culture Aerobic cultivation was made to see the effect of the pykF– gene knock-out on the metabolism. The growth rate of mutant was a little lower as compared with that of the wild type E. coli K12. During exponential growth, acetic acid was

38

K. Shimizu

also produced but its formation rate was significantly lower as compared with that of the wild type E. coli K12. It was found that the CO2 evolution rate was lower for the pykF– mutant as compared with that of the wild type. These results are consistent with other researches made by Ponce et al. [118, 119] and Zhu et al. [120]. Several enzyme activities for glycolytic enzymes (Hxk, Pgi, Pfk, Tpi, and Pyk) were measured for both the wild type and pykF– mutant of E. coli. The enzyme activity of phosphoglucose transferase system was detected only in the presence of glucose, confirming its efficient induction by glucose. Hxk activity also increased when glucose was present. This enzyme is thought to be inhibited by its product G6P. High activities of other glycolytic enzymes (Pgi, Tpi, and Pyk) during cell growth phase indicate a “stepping up” of glucose utilization during fermentation. All these enzymes are under control of catabolic repression/activation mechanism.All these mentioned enzyme activities were repressed under acetate condition as compared with the case where glucose was present.As acetate was fully consumed, they recover from the repression again. It was found that the enzyme activities of glycolytic enzymes such as Hxk, Pgi, Pfk, Tpi, and Ppc were higher during early growth phase.When glucose was depleted, these enzyme activities were down regulated. As soon as the acetate was consumed, these enzymes were derepressed, and their activities were increased again. These results correspond to those observed by Oh et al. [61], where they showed by DNA microarray analysis that during acetate metabolism, the glycolytic genes such as pfk, fba, ppc, pyk, zwf and gnd were down regulated. These genes are thought to be regulated by catabolite repressor/activator, fruR or cra, a global catabolite repression regulator [55]. The anaplerotic pathway enzyme Ppc is usually active when glucose is present to supply OAA for the production of biosynthetic compounds. This enzyme activity was found to be higher during the initial growth phase in both wild type and pykF– mutant when glucose was present. Moreover, as compared with that of wild type, an increased activity was found in pykF– mutant during initial growth phase. As the acetate was formed, the enzyme activity of gluconeogenic enzyme, Pck became higher as compared with the case where only glucose was present. Moreover, it can be also seen that the higher activity of Pck was observed for pykF– mutant as compared with that of E. coli K12. The enzyme activities of G6PDH and 6PGDH in the oxidative PP pathway were higher in pykF– mutant as compared with that of the wild type, in particular, when glucose was present. The enzyme activity of Mez is usually very low in bacteria when glucose is present. But this enzyme activity increased threefold in pykF– mutant in the presence of glucose as compared with that of the wild strain. This high activity might be due to the high activity of pyruvate carboxylase. The activity of this enzyme was found to be higher when acetate was utilized as a carbon source in the wild type. The MDH activity was also increased in pykF– mutant, showing the similar trend as Mez. The citrate synthase, encoded by gltA, the entry enzyme to TCA cycle, was found to be suppressed by excess glucose, but elevated by oxygen and other

Metabolic Flux Analysis Based on 13C-Labeling Experiments

39

carbon sources such as acetate. During the initial growth phase, this enzyme was not so active, but this enzyme became more active when glucose-acetate cometabolism condition exits. As compared with pykF– mutant, this enzyme activity was higher in the wild type. The glyoxylate shunt enzyme, isocitrate lyase, encoded by aceA is usually activated by acetic acid to supply OAA needed for anaplerosis. Thus, this enzyme activity was found to be higher when acetate was utilized as a carbon source. The enzyme activities for other enzymes such as Ack and LDH were found to be higher in the presence of glucose. The Ack and LDH activities were found to be lower in pykF– mutant than that of the wild type. The enzyme activity of the gluconeogenic enzyme, Pck showed high activation when acetate was used a carbon source. The gene pckA may be induced by the increased level of cAMP. This type of trend was also observed for Mez. Oh et al. [61] obtained in their study that Mez and Pck correlated when acetate was utilized. They explained that the mutation of mez and pckA genes prevented growth on acetate.As the glucose is consumed and acetate accumulates, cells switch smoothly to cometabolism of utilizing both glucose and acetate. This switch involves the induction of the TCA cycle enzymes and glyoxylate bypass enzymes required to provide energy and to replenish intermediates used for amino acid biosynthesis [121]. We observed similar phenomena in case of citrate synthase which was more active when both glucose and acetate were utilized during cell growth phase. The isocitrate lyase also became more active during acetate condition. The TCA cycle genes are known to be subject to cAMP and CRP which mediate catabolic repression [122]. The activity of LDH increased during growth on glucose as a carbon source and repressed in acetate condition. The Ack was also found to be active under glucose condition and repressed under acetate condition. Intracellular metabolite concentrations are also quite important to investigate the metabolic regulation. We, therefore, measured the intracellular concentrations of glycolytic and oxidative PP pathway intermediates. The measured metabolites include G6P, F6P, 6PG, PEP, and PYR. It was found that the pyruvate concentration was lower and slightly higher in PEP concentration in pykF– mutant as compared with those of E. coli K12. It was also found that G6P and F6P concentrations were higher for pykF– mutant as compared with those of the wild type. We also studied how the enzymatic activity changes affect the intracellular metabolite concentrations. We found that when extracellular glucose was present, the intracellular concentrations for glycolytic intermediates and oxidative PP pathway intermediates such as G6P, F6P, and 6PG concentrations were higher. As the result of higher enzyme activity, these intracellular metabolites concentrations were reduced after several hours of growth. The PEP concentration was found to be not so high during initial cultivation, which indicates that most of the PEP may be utilized in the glucose transport system. In the case of pykF– mutant, it was found that the enzyme activities of glycolytic pathway enzymes decreased as compared with those of the wild type,

40

K. Shimizu

while the enzyme activities of oxidative PP pathway enzymes increased for the former. Ponce et al. [119] also found the increased activities of oxidative PP pathway enzymes in pykF– mutant. The most significant differences were observed between wild type and pyk F– mutant in Ppc and Mez activities. In the presence of glucose, these two enzyme activities were significantly increased in pykF– mutant. These can be explained that the blockage of PEP to PYR pathway activated both Ppc and Mez enzymes to supply the PYR by forming an alternative pathway. It was thought that blockage of PEP to PYR will increase the PEP pool in the mutant, which might inhibit the Pfk enzyme activity in the glycolytic pathway. A slight increase of PEP pool was observed for pykF– mutant during cell growth phase. The decreased activity of Pfk in turn increased G6P and F6P concentrations. The increased activity of oxidative PP pathway was observed in pykF– mutant. This might be due to the increased G6P concentration. Overproduced NADPH in PP pathway may be partly utilized through LDH reaction for pykF– mutant [17]. 5.1.2 Metabolic Regulation Analysis by Metabolic Flux Distribution, Enzyme Activities, and Intracellular Metabolites Concentrations for Continuous Culture The continuous cultivation was conducted at two different dilution rates such as 0.1 h–1 and 0.5 h–1. In order to obtain higher resolution on the intracellular isotope distribution, we used both GC-MS and NMR, where the mixture of uniformly labeled [U-13C] glucose, first carbon labeled [1-13C] glucose and naturally labeled glucose were used for the 13C labeling experiments. Figure 10 shows the determined intracellular flux distributions for the wild type (upper figures) and for pykF– mutant (lower figures). It can be seen from the flux distribution that in pykF– mutant, there is a high flux from PEP to OAA. The flux through anaplerotic reaction catalyzed by Ppc in pykF– mutant also became high at 44% in pykF– mutant as compared with 19% for the wild type strain. It should be noted that the flux value in < > (exchange flux) for Ppc is also high, which indicates that the flux through Pck is also high. The flux from PEP to PYR was 1.6% for pykF– mutant which is significantly lower as compared with 130% for the wild type, which is consistent with the enzyme activity measurement. It can be also seen that the flux through Mez is high about 21% for pykF– mutant, while it is very low about 3% for the wild type. Moreover, the glycolytic flux from G6P to F6P was 20% for the pykF– mutant, while it was 65% in the wild type strain. On the other hand, the flux through oxidative PP pathway became 79% for pykF– mutant, while it was 34% in the wild type. The flux from acetylCoA to acetate reduced to 0.82% in pykF– mutant as compared with 20% in the wild type. From the investigation on the continuous culture, the flux calculation result shows clearly that the disruption of pyk gene increases the flux through Ppc and Mez. The flux result also shows the reduced carbon flow through glycolysis and increased carbon flow through PP pathway, which is consistent with the mea-

Metabolic Flux Analysis Based on 13C-Labeling Experiments

41

Fig. 10 Metabolic flux distribution of the wild type (upper values) and pykF– mutant (lower values) at dilution rate of 0.1 h–. The red arrows imply up-regulation, while blue arrows imply down-regulation

42

K. Shimizu

surements of enzyme activities and intracellular metabolite concentrations. A higher flux was obtained for pykF– mutant through anaplerotic pathway. This may lead to ATP dissipation, since, Pck constitutes a futile cycle during growth on glucose in pykF– mutant.Although futile cycling could be induced by simultaneous overexpression of Ppc and Pck in E. coli [123], it is generally maintained at low level [123–125]. The enzyme activity result for Ppc and Pck supports this futile cycle result obtained by flux calculation, since in batch culture, these two enzyme activities were also found to be higher for pykF– mutant as compared with those of the wild type K-12. 5.2 Metabolic Regulation Analysis of gnd – Mutant of E. coli To investigate physiological effects of gnd knockout, the parent E. coli strain BW25113 and its gnd – mutant were grown first in batch culture. It was found that a single deficiency in gnd has little effect on the growth rate of E. coli on glucose, which is in accord with previous results. Contrary to the depression of glucose utilization caused by other gene deficiency, this mutant exhibited higher glucose uptake rate than that of the wild type. The specific oxygen consumption rate and CO2 evolution rate varied little between the two different cultures, indicating that there are only minor changes in respiratory metabolism.Although more glucose remained unutilized for the wild type, the final cell concentration was almost the same in these two cultures. That is, the cell yield on glucose was higher for the wild type in spite of very similar growth characteristics. This difference was attributed to the relatively higher acetate production rate for the mutant. Key enzymes participating in the metabolism of glycolytic pathway (Pgi), oxidative pentose PP pathway (G6PDH and 6PGDH), non-oxidative PP pathway (Tal), ED pathway (phosphogluconate dehydratase and 2-keto-3-deoxy-6-phosphogluconate aldolase) and TCA cycle (ICDH and MDH) were measured in the cell extracts from cells under the condition of the mid-exponential growth phase. The enzyme activities of Mez and Ppc as well as Pck were also measured. In spite of the similarity in the growth rates, the levels of some key enzymatic activities varied greatly in these two strains. In gnd – mutant, higher enzyme activity involved in the conversion of G6P to F6P (Pgi) was observed than in the wild type, indicating that more carbon substrates were directed to EmbdenMeyerhof pathway. The increase in the expression of Pgi would be expected to decrease carbon flow into the oxidizing arm of the PP pathway, and this is supported by the decrease in the level of G6PDH. ICDH and Mez activities were comparable for both strains, which is consistent with the finding in the phenotype analysis that there were only minor changes in respiratory metabolism when the gnd was knocked out. NAD+/NADP+ dependent malic enzymes were identified in the mutant which was accompanied by an up-regulation in Ppc activities and down-regulation in Pck. In contrast, Mez in the wild type was almost negligible. Significantly important, despite the inactivation of the

Metabolic Flux Analysis Based on 13C-Labeling Experiments

43

6PGDH, the non-oxidative pathway was still active, as shown in the Tal activity with 43.2% of the level detected in the wild type. Moreover, in gnd– mutant, high enzyme activity levels involved in the formation of pyruvate via ED pathway were present, whereas no activity was detected in the wild type. The existence of non-oxidative PP pathway in the gnd – mutant allows us to infer that backward reactions which converted F6P and GAP into R5P may occur so that the ribose was produced solely by the non-oxidative PP pathway in gnd– mutant. In normal cells, the ribose metabolism was mainly performed via oxidative PP pathway. This hypothesis will be tested in the metabolic flux analysis. For metabolic flux calculation, we combined GC-MS with NMR to get as much labeled information as possible from both the individual carbon and the carbon clusters, and the results are shown in Fig. 11 [16]. Due to the higher activity of Pgi, the gnd– mutant had a higher glycolytic flux input than the wild type strain. While the wild type strain directed 78.6% of total carbon flux through the first step of the EM pathway, the flux increased by 10.2% was found in the gnd– mutant.As for the PP pathway, the gnd– deficiency elicited three important differences between the two strains. First, the wild type directed 20% of total carbon flux through G6PDH, but the relative flux was decreased to the half in the mutant. This reduction corroborates the tendency observed in the analysis of Pgi and 6PGDH specific activities, although the ratio of the change could not be comparable between these two different analytical methods. Second, the block of the gnd pathway forces the activation of the non-oxidative PP pathway. The highly reversible reactions catalyzed by Tkt I, II, and Tal were found to be in an opposite direction to those of the wild type. Third, evidence for the presence of the enzymes of the ED pathway was confirmed by metabolic flux analysis. The existence of the ED pathway kept G6PDH active in the mutant so as to produce NADPH for biosynthetic reactions. The flux analysis showed only small differences in the carbon flux through TCA cycle between the wild type and the mutant strain. In contrast, the carbon fluxes producing OAA from PEP varied and the reaction oxidizing MAL to produce PYR, and NADPH/NADH via Mez was active in the mutant but almost negligible in the wild type strain. This induction of Mez by gnd knockout was accompanied by the down-regulation in Pck. The analysis of acetate production rate shows that, although the wild type and gnd– mutant had almost the same growth rate on glucose, the mutant directed a higher flux to the synthesis of the acetate. These results are in agreement with previous observations in the batch cultivation, demonstrating the low energetic efficiency caused by gnd knockout. It was found that gnd knockout did not affect the constitutivity of the other enzymes involved in both oxidative and non-oxidative PP pathway.As we know, catabolic enzymes are usually expressed by internally formed substrates and regulated by the reaction products. If the wild type metabolic network was applied to gnd– mutant, and assuming the inducer of the first two enzymes of the non-oxidative PP pathway (ribulose phosphate 3-epimerase and ribose-5phosphate isomerase) was RU5P, the levels of enzymes in the non-oxidative

Fig. 11 Metabolic flux distribution in chemostat culture of wild-type (upper values) and gnd mutants (lower values). Arrowheads indicate the primary direction of fluxes in wild-type cells. Negative values are fluxes whose directions are opposite to arrowheads. The feed glucose concentration was 4 g/l, and the dilution rate was 0.2 h–1

Metabolic Flux Analysis Based on 13C-Labeling Experiments

45

pathway should be low enough because no RU5P was produced by the reaction catalyzed by 6PGDH. Similarly, since no pathway existed for 6PG consumption, the level of G6PDH should also be extremely low due to product inhibition. Such an assumption was not, however, in accord with the enzyme activity analysis, in which the level of G6PDH and Tal only decreased to 70.1% and 43.2% of the level in the wild type, respectively. The fact that both oxidative and non-oxidative PP pathway were still active in the gnd– mutant demonstrated that the glucose metabolism must then proceed by way of flux rerouting/redirection. This hypothesis was confirmed by the metabolic flux analysis. The flux distribution clearly indicated that the ED pathway was activated in the mutant and the ribose-producing metabolism was different between the two strains. In the wild type, the ribose was mainly produced via oxidative PP pathway, while in the mutant, the non-oxidative PP pathway utilized intermediates of the glycolytic pathway to synthesize R5P and E4P. In cellular metabolism this mode of ribose formation only occurs when most of G6P is converted to F6P and GAP by the glycolytic pathway, which is quite consistent with our findings in enzyme and flux analysis. Besides the ribose formation by way of oxidative PP pathway, another important function for this oxidative branch is to generate NADPH serving as an electron donor in inductive biosynthesis. The inactivation of gnd and the decrease in flux through zwf reduced the NADPH production. As we know, these two reactions are the main sources for NADPH supply in the normal cell growth. Therefore, other NADPH-producing reactions were expected to be highly active to complement the reducing power for the cell growth. In the gnd– mutant, the NADP/NAD-dependent malic enzymes were activated to synthesizes NADPH using MAL, an intermediate of the TCA cycle. This effect could be regarded as a compensatory mechanism in gnd– mutant for the decreased synthesis of NADPH. This drain of carbon skeletons from the TCA cycle enabled the cells to respond to TCA carbon depletion by regulating the carbon flux through Ppc and Pck enzymes. This second compensatory mechanism was to respond to the depletion of OAA in the TCA cycle resulting from the increase of the carbon flux by way of Mez. To increase the net synthesis of OAA from PEP, Ppc was up-regulated and PckA was down-regulated in the mutant.

6 Conclusion We have done the similar analysis for the other gene-knockout E. coli such as pck– [126], zwf– [127], ppc–, pfl–, mutants etc. Moreover we also made the above analysis for the other microorganisms such as Cyanobacteria [128], yeast [129], etc. Holms has done the metabolic control analysis based on the conventional flux calculation. The present investigation indicates that it is quite important to combine information of gene and protein expressions as well as enzyme activities, intracellular metabolite concentrations, and metabolic flux distrib-

46

K. Shimizu

ution obtained based on 13C-labeled experiments to uncover the metabolic regulation caused by gene knock out or the change in culture condition. Even a single gene plays an important role, and it was found that the organism has the backup system to be robust for the genetic modification. Acknowledgements It is acknowledged that the research was supported in part by a grant by the Ministry of Education and Science under grant no. 14 350438. It is also acknowledged that the research was supported in part by a grant from New Energy and Industrial Technology Development Organization (NEDO) of the Ministry of Economy, Trade and Industry of Japan (Development of Technological Infrastructure for Industrial Bioprocess Project). It is further acknowledged that the research was conducted by the postdoctoral fellows and the PhD students in our laboratory.

References 1. Sauer U, Hatzimanikatis V, Hohmann HP, Mannebegr M, Bailey JE (1996) Appl Environ Microbiol 62:3687 2. Goel A, Ferrance J, Jeong J, Ataai MM (1993) Biotechnol Bioeng 42:686 3. Wiechert W, Graaf AA, Marx A (1995) Computer Applications in Biotechnology – CAB 6 4. Szyperski T (1995) Eur J Biochem 232:433 5. Marx A, Graaf AA, Wiechert W, Eggeling L, Sahm H (1996) Biotechnol Bioeng 49:111 6. Schmidt K, Nielsen J, Villadsen J (1999a) J Biotechnol 71:175 7. Lin YY, Cheng WB, Wright CE (1993) Anal Biochem 209:267 8. Christensen B, Nielsen J (2000) Biotechnol Bioeng 68:652 9. Jeffrey FMH, Rajagopal A, Malloy CR, Sherry AD (1991) TIBS 16:5 10. Phalakornkule C, Fry B, Zhu T, Kopesel R, Ataai MM, Domach MM (2000) Biotechnol Prog 16:169 11. Yang C, Hua Q, Shimizu K (2002) J Biosci Bioeng 93:78 12. Yang C, Hua Q, Shimizu K (2002) Metabolic Eng 4:202 13. Zhao J, Shimizu K (2003) J Biotechnol 101:101 14. Paul Lee W-N (1991) Biol Mass Spectrom 20:451 15. Paul Lee W-N, Bergner EA (1992) Biol Mass Spectrom 21:114 16. Zhao J, Baba T, Mori H, Shimizu K (2003) FEMS Microb Lett 220:295 17. Siddiquee K,Al Z,Arauzo-Bravo M, Shimizu K (2004) Appl Microbiol Biotechnol 63:407 18. O’Farrell PH (1975) J Biol Chem 250:4007 19. Anderson NL, Anderson NG (1998) Electrophoresis 19:1853 20. Gygi SP, Rist B, Aebersold R (2000) Curr Opin Biotechnol 11:396 21. Martin BR (1987) The regulation of enzyme activity. In: Martin BR (ed) Metabolic regulation. Oxford London Edinburgh, Boston Palo Alto Melbourne, pp 12–27 22. DeRisi JL, Iyer VR, Brown PO (1997) Science 278:680 23. Iyer VR, Eisen MB, Ross DT, Schuler G, Moore T, Lee JCF, Trent JM, Staudt LM, Hudson MS, Boguski D, Lashkari D, Shalon D, Botstein D, Brown PO (1999) Science 283:83 24. Arfin ST, Long A, Ito ET, Tolerri L, Riehle MM, Paegle ES, Hatfield GW (2000) J Biol Chem 275:29672 25. Richmond CS, Glasner JD, Mau R, Jin H, Blattner FR (1999) Nucleic Acids Res 27:3821 26. Tao H, Bausch C, Richmond C, Blattner FR, Conway T (1999) J Bacteriol 181:6425 27. Wilson M, DeRisi J, Kristensen HH, Imboden P, Rane S, Brown PO, Schoolnik GK (1999) Proc Natl Acad Sci USA 96:12833

Metabolic Flux Analysis Based on 13C-Labeling Experiments

47

28. Ferea TL, Botstein D, Brown PO, Rosenzweig RF (1999) Proc Natl Acad Sci USA 96:9721 29. Kao CM (1999) Biotechnol Prog 15:304 30. Dieffenbach CW, Dveksler GS (1995) PCR primer: a laboratory manual. Cold Spring Harbor Laboratory Press, Cold Spring Harbor, New York 31. Siebert PD, Larrick JW (1991) Nature 359:557 32. Sambrook J, Russell DW (2001) Molecular cloning: a laboratory manual, 3rd edn, Cold Spring Harbor Laboratory Press, Cold Spring Harbor, New York 33. Arcelana-Panlilio MY, Schultz GA (1993) Methods Enzymol 225:303 34. Foley KP, Leonard MW, Engel JD (1993) Trends Genet 9:380 35. Mocharla H, Mocharla R, Hodes ME (1990) Gene 93:271 36. Singer-Sam J, Riggs AD (1993) Methods Enzymol 225:344 37. Harrison DC, Medhurst AD, Bond BC, Campbell CA, Davis RP, Philpott KL (2000) Mol Brain Res 75:143 38. Kwon DH, Osato MS, Graham DY, El-Zaatari FAK (2000) Int J Antimicrob Agents 15:31 39. Akiyama H, Kanai S, Hirano M, Kumei Y, Shimokawa H, Katano H, Hara E, Song S, Mukai C, Nagaoka S (1996) J Biotechnol 47:325 40. Brucato S, Harduin-Lepers A, Godard F, Bocquet J, Villers C (2000) Biochem Biophys Acta 1474:31 41. Arauzo-Bravo MJ, Shimizu K (2003) J Biotechnol 105:117 42. Schmidt K, Nielsen J, Villadsen J (1997) Biotechnol Bioeng 55:831 43. Schmidt K, Norregaard LC, Pedersen B, Meissner A, Nielsen JQ (1999b) Metab Eng 1:166 44. Wittmann C, Heinzle E (1999) Biotechnol Bioeng 62:739 45. Möllney M, Wiechert W, Kowantzki D, Graaf AA (1999) Biotechnol Bioeng 66:86 46. Winden V, Schipper D, Verheijen P, Heijnen J (2001) Metab Eng 3:322 47. Wiechert W (2001) Metabol Eng 3:195 48. Davis L (1991) Van Nostrand Reingold, New York 49. Moré JJ (1977) The Levenberg-Marquardt algorithm: implementation and theory. In: Watson GA (eds) Numerical analysis. Springer, Berlin Heidelberg New York 50. Wiechert W, Siefke C, Graaf AA, Marx A (1997) Biotechnol Bioeng 55:118 51. Sauer U, Lasko DR, Fiaux J, Hochuli M, Glaser R, Szyperski KW, Bailey JE (1999) J Bacteriol 181:6679 52. Peng L, Shimizu K (2003) Appl Microbiol Biotech 61:163 53. Reuse HD, Danchin A (1988) J Bacteriol 170:3827 54. Fox DK, Presper KA, Adhya S, Roseman (1992) Proc Natl Acad Sci USA 89:7056 55. Plumbridge J (1999) Mol Microbiol 33:260 56. Saier MH, Ramseier TM (1996) J Bacteriol 178:3411 57. Ramseier TM, Bledig S, Michotey V, Feghali R, Saier MH Jr (1995) Mol Microbiol 16:1157 58. Martin BR (1987) The regulation of enzyme activity. In: Martin BR (ed) Metabolic regulation. Oxford, London, Edinburgh, Boston, Palo Alto, Melbourne, pp 12–27 59. Campos G, Guixe V, Babul J (1984) J Biol Chem 259:6147 60. Guixe V, Babul J (1985) J Biol Chem 260:1101 61. Oh MK, Liao JC (2000) Biotechnol Prog 16:278 62. Oh MK, Rohlin L, Kao K, Liao JC (2002) J Biol Chem 277:13175 63. Unden G, Duchene A (1987) Arch Microbiol 147:195–200 64. Epstein W, Rothman-Denes L, Hesse J (1975) Proc Natl Acad Sci USA 72:2300 65. Lowry OH, Carter J, Ward JB, Glaser L (1971) J Biol Chem 246:6511 66. Esienberg RC, Dobrogosz WJ (1967) J Bacteriol 93:941 67. Lessie TG, Whiteley HR (1969) J Bacteriol 100:878 68. Wolf RE Jr, Prather DM, Shea FM (1979) J Bacteriol 139:1093 69. Pease AJ, Wolf RE Jr (1994) J Bacteriol 176:115

48 70. 71. 72. 73. 74. 75. 76.

77. 78. 79. 80. 81.

82. 83. 84. 85. 86. 87. 88. 89. 90. 91. 92. 93. 94. 95. 96. 97. 98. 99. 100. 101. 102. 103. 104. 105. 106. 107. 108. 109. 110.

K. Shimizu Conway T (1992) FEMS Microbiol Rev 103:1 Sugimoto S, Shiio I (1987) Agric Biol Chem 51:1257 Gunsalus RP, Park SJ (1996) 12th Forum in Microbiology, pp 437–450 Kiley PJ, Beinert H (1999) FEMS Microbiol Rev 22:341 Alexeeva S, Kort B, Sawers G, Hellingwerf KJ (2000) J Bacteriol 182:4934 Sawers G, Suppmann B (1992) J Bacteriol 174:3474 Bock A, Sawers G (1999) Fermentation. In: Frederick C (ed) Escherichia coli and Salmonella: cellular and molecullar biology. AMS, Washington DC, Adobe and Mira Digital publishing, pp 1–2 Mat-Jan F, Alam KY, Clark DP (1989) J Bacteriol 171:342 Wanner BL, Wilmes-Riesenberg MR (1992) J Bacteriol 174:2124 McCleary WR, Stock JB (1994) J Biol Chem 269:31576 Amarasingham CR, Davis BD (1965) J Biol Chem 240:3664 Cronan JE Jr, Laporte D (1999) Tricarboxylic acid cycle and glyoxylate bypass. In: Neidhardt F (ed) Escherichia coli and Salmonella: cellular and molecular biology.AMS, Mira Digital Publishing, Washington DC, pp 1–16 Tseng CP, Yu CC, Lin HH, Chang CY, Kuo JT (2001) J Bacteriol 183:461 Park SJ, Gunsalus RP (1995) J Bacteriol 177:6255 Park SJ, McCabw J, Turna J, Gunsalus RP (1994) J Bacteriol 176:5086 Park SJ, Tseng CP, Gunsalus RP (1995) Mol Microbiol 15:473 Park SJ, Cotter PA, Gunsalus RP (1995) J Bacteriol 177:6652 Holms WH (1988) Biochem Soc Symp 54:17 Kabir M, Shimizu K (2003) Appl Microbiol Biotech 62:244 Gygi SP, Rist B, Aebersold R (2000) Curr Opin Biotechnol 11:396 Gygi SP, Corthals GL, Zhang Y, Rochon Y, Aebersold R (2000) PNAS 97:9390 Kidwell J, Valentin HE, Dennis D (1995) Appl Environ Microbiol 61:1391 Lee SY, Chang HN (1995) Can J Microbiol 41:207 Shi H, Nikawa J, Shimizu K (1999) J Biosci Bioeng 87:666 Kabir M, Shimizu K (2003) J Biotechnol 105:11 Saier MH Jr, Ramseier TM (1996) J Bacteriol 178:3411 Arora KK, Pedersen PL (1995) Arch Biochem Biophys 319:574 Canonaco F, Hess TA, Heri S, Wang T, Szyperski T, Sauer U (2001) FEMS Microbiol Lett 204:247 Pearse AJ, Wolf RE (1994) J Bacteriol 176:115 Clarke DM, Loo TW, Gillam S, Bragg PD (1986) Eur J Biochem 158:647 Boonstra B, French CE, Wainwright I, Bruce NC (1999) J Bacteriol 181:1030 Hanson RL, Rose C (1980) J Bacteriol 141:401 Lee SY, Lee KM, Chang HN, Steinbüchel A (1994) Biotechnol Bioeng 44:1337 Farmer WR, Liao JC (1997) Appl Environ Microbiol 63:3205 Park S-J, McCabe J, Turna J, Gunsalus RP (1994) J Bacteriol 176:5086 Park S-J, Chao G, Gunsalus RP (1997) J Bacteriol 179:4138 Park S-J, Tseng C-P, Gunsalus RP (1995) J Bacteriol 15:473 Park S-J, Gunsalus RP (1995) J Bacteriol 177:6255 Park S-J, Cotter PA, Gunsalus RP (1995) J Bacteriol 177:6652 Brown TD, Jones-Mortimer MC, Kornberg HL (1977) J Gen Microbiol 102:327 Grunberg-Manago M (1996) Regulation of the expression of aminoacyl-tRNA synthetases and translation factors. In: Neidhardt FC, Curtiss R III, IngrahamJL, Lin ECC, LowKB, Magasanik B, Reznikoff WS, Riley M, Schaechter M, Umbarger HE (eds) Escherichia coli and Salmonella: cellular and molecular biology, 2nd edn. ASM Press, Washington DC, pp 1432–1457

Metabolic Flux Analysis Based on 13C-Labeling Experiments

49

111. Missiakas D, Betton J-M, Raina S (1996) Mol Microbiol 21:871 112. Wieczorek RA, Pries A, Steinbüchel A, Mayer F (1995) J Bacteriol 177:2425 113. Wittmann HG (1974) In: Namura M, Tissieres A, Lengyel P (eds) Probosomes. Cold Spring Harbor Laboratory Press, Plainview, NY, pp 93–114 114. Sengupta J, Agrawal RK, Frank J (2001) Proc Natl Acad Sci USA 98:11991 115. Tzareva NV, Makhno VI, Boni IV (1994) FEBS Lett 337:189 116. Van der Werf MJ, Guettle MV, Jain MK, Zeikus JG (1997) Arch Microbiol 167:332 117. Potapov AP, Subramanian AR (1992) Biochem Int 27:745 118. Arnold CN, Mcelhanon J, Lee A, Leonhart R, Siegele DA (2001) J Bacteriol 183:2178 119. Martin BR (1987) The regulation of enzyme activity. In: Martin BR (ed) Metabolic regulation. Oxford, London, Edinburgh, Boston, Palo Alto, Melbourne, pp 12–27 120. Ponce E, Flores N, Martinez A, Bolivar F, Valle F (1995) J Bacteriol 177:5719 121. Ponce E, Martinez A, Bolivar F, Valle F (1998) Biotechnol Bioeng 58:292 122. Zhu T, Phalakornkule C, Koepsel RR, Domach MM, Ataai MM (2001) Biotechnol Prog 17:624 123. Clark DP, Cronan JE (1996) Two-carbon compounds and fatty acids as carbon sources. In: Neidhardt FC, Curtiss R III, Ingraham JL (eds) Escherichia coli and Salmonella. ECC, pp 343–357 124. Cronan JE Jr, LaPorte D (1999) Tricarboxylic acid cycle and glyoxalate by pass. In: Neidhardt FC et al. (eds), Escherichia coli and Salmonella molecular biology, 2nd edn. ASM Press, Washington DC, pp 343–357 125. Chao YP, Laio JC (1994) J Biol Chem 269:5122 126. Chambost JP, Frankel DG (1980) J Biol Chem 255:2867 127. Dalal F, Fraenkel DG (1983) J Bacteriol 153:390 128. Yang C, Hua Q, Shimizu K (2003) Biotechnol Bioeng 84:131 129. Zhao J, Shimizu K (2004) Metabolic Eng. (in press) 130. Yang C, Hua Q, Shimizu K (2002) Appl Microbiol Biotech 58:813 131. Zhang H, Shimizu K, Yao S (2003) Biochem Eng J 16:211 132. Holms H (1996) FEMS Microbiol Rev 19:85

Received: September 2003

Adv Biochem Engin/Biotechnol (2004) 91: 51– 73 DOI 10.1007/b94205 © Springer-Verlag Berlin Heidelberg 2004

Application of Knowledge Information Processing Methods to Biochemical Engineering, Biomedical and Bioinformatics Fields Taizo Hanai1 (✉) · Hiroyuki Honda2 1

2

Laboratory for Bioinformatics, Graduate School of Systems Life Sciences, Kyushu University, Hakozaki 6-10-1, Higashi-ku, Fukuoka 812-8581, Japan [email protected] Department of Biotechnology, School of Engineering, Nagoya University, Furo-cho, Chikusa-ku, Nagoya 464-8603, Japan [email protected]

1

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

52

2 2.1 2.2 2.3 2.4 2.5

Knowledge Information Processing Methods . Fuzzy Reasoning . . . . . . . . . . . . . . . . . Artificial Neural Network (ANN) . . . . . . . . Fuzzy Neural Network (FNN) . . . . . . . . . . Genetic Algorithm (GA) . . . . . . . . . . . . . Fuzzy Adaptive Resonance Theory (Fuzzy ART)

. . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

53 53 54 55 56 57

Application of Knowledge Information Processing Methods to Biochemical Engineering Field . . . . . . . . . . . . . . . 3.1 Modeling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.2 Physiological State Recognition . . . . . . . . . . . . . . . . 3.2.1 Example – VB2 Production . . . . . . . . . . . . . . . . . . . 3.3 Software Sensor . . . . . . . . . . . . . . . . . . . . . . . . . 3.3.1 Example – Activated Sludge Process . . . . . . . . . . . . . . 3.4 Process Control . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

58 58 58 58 60 60 64

Application to Other Fields . . . . . . . . . . . . . . . . . . . . . . . . . . . Biomedical Field – Prediction of Prognosis for Cancer Patients from DNA Microarray Data by FNN . . . . . . . . . . . . . . . . . . . . . . Bioinformatics Field – Gene Clustering for DNA Microarray Data by Fuzzy ART . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

65

Conclusion and Perspectives . . . . . . . . . . . . . . . . . . . . . . . . . .

71

References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

72

3

4 4.1 4.2

5

65 68

Abstract In biochemical and biomedical engineering fields there are a variety of phenomena with many complex chemical reactions, in which many genes and proteins affect transcription or enzyme activity of others. It is difficult to analyze and estimate many of these phenomena using conventional mathematical models. Recently some knowledge information processing methods, such as the artificial neural network (ANN), fuzzy reasoning, fuzzy

52

T. Hanai · H. Honda

neural network (FNN), fuzzy adaptive resonance theory (fuzzy ART) and the genetics algorithm (GA), were developed in the computer science field and have been applied to analysis in a variety of research fields. In this chapter, these methods will be briefly reviewed. Next, the application of these methods in the biochemical field will be introduced, instancing two examples in actual industrial processes. In addition, the application in the biomedical and bioinformatics field as another attractive field will be reviewed. Two examples are our research such as the prediction of prognosis for cancer patients from DNA microarray data using FNN and gene clustering for DNA microarray data using fuzzy ART. Keywords Knowledge information processing · Neural network · Genetic algorithm · Adaptive resonance theory · DNA microarray

1 Introduction Recently, many high throughput analysis methods, for example on-line sensors in the biochemical engineering field, DNA microarray and protein chips in the biomedical and bioscience field, have been developed. From these methods, many kinds and a large amount of data can be collected simultaneously, but it is difficult to extract important knowledge by only human sense from these many kinds and large amounts of data. On the other hand, some kinds of computational analysis methods called data mining and knowledge information processing have been proposed and have been studied in the computer science field [1–4]. These methods enable us to extract important causality or knowledge from many kinds and large amounts of data. Under such condition, applications of the analysis methods using data mining or knowledge information processing have mainly been studied in the biochemical engineering field over the last 20 years [5, 6]. There are many research papers on the information approach to knowledge and the method was applied to process operations in actual industrial processes in Japan [7]. In this chapter, we attempt to review the recent developments in knowledge information approaches for biochemical engineering problems to process design and operation. Some examples in actual industrial processes will be introduced. We have also studied in the field; process control in the Japanese sake mashing process [8–10], and estimation of the sensory evaluation in alcohol beverages [11, 12] using the artificial neural network (ANN) or fuzzy neural network (FNN). Recently, bioinformatics especially in the biomedical field have been expanded and developed almost daily. Analysis methods using knowledge information processing should be actively applied in such fields for identifying important biomedical information.We have also reported studies on the estimation of the severity of dementia of the Alzheimer type from electroencephalogram data by FNN [13] and an estimation model of interaction between major histocompatibility complex molecules and antigen peptides by FNN [14] and hidden Markov models [15].

Application of Knowledge Information Processing Methods

53

In this chapter, some methods in data processing and knowledge information processing, fuzzy reasoning, ANN, FNN and genetic algorithms (GA), are introduced and application in the biochemical, biochemical engineering and bioscience fields using our studies as examples is also introduced.

2 Knowledge Information Processing Methods 2.1 Fuzzy Reasoning Fuzzy reasoning was proposed by Zadeh in 1965 [1] and has been widely used by researchers both in biochemical engineering and in other fields. This method can transform human reasoning into rules suitable for process control and estimation of relationship between input and output. Various studies on the control of the biochemical engineering process and the based on fuzzy reasoning have been published [16, 17]. Production rules and membership functions are necessary for calculation of fuzzy reasoning. Figures 1 and 2 show examples of the production rules and the membership functions. A production rule represents expert knowledge in a form such as “IF A is B, THEN C should be D”, which are known as “IF-THEN” rules, and can readily be summarized in tabular form. A membership function is used to transform a numerical value of a variable (such as temperature or specific gravity) to its “grade”, which shows the degree to which the variable belongs to a certain class (such as “big”,“medium” or “small”).

Fig. 1 Example of membership function

54

T. Hanai · H. Honda

Fig. 2 Example of production rules for temperature control in fermenter. Notations: PB, positive big; PS, positive small; ZE, zero; NS, negative small; NB, negative big

2.2 Artificial Neural Network (ANN) ANNs, the multi-layered network models normally driven by a back propagation algorithm [2], have been developed from the perceptron, which was proposed as a model for information circulation in the brain. There are numerous applications of ANNs in engineering, since they are powerful in pattern recognition and learning. Figure 3 shows the structure of a three-layered ANN with input, hidden, and output layers. Outputs from the units in the output and hidden layer are calculated using a sigmoid function operating on the sum of the inputs to these

Fig. 3 Concept of artificial neural network (ANN)

Application of Knowledge Information Processing Methods

55

units. ANN can be applied to many types of data sets, in which a non-linear relationship between input and output exists. 2.3 Fuzzy Neural Network (FNN) Fuzzy reasoning has been used in many cases in which an expert’s knowledge can be extracted and reconstructed in a computer. However, it takes a relatively long time to tune membership functions by trial-and-error and the expert’s experience is still necessary to accompany any rules with fuzzy reasoning and assess the quality of any rules, which are produced during this refinement. An ANN can be constructed automatically if a large quantity of historical data is available. Nevertheless, it is difficult to interpret the link between input and output variables in the final model, because the structure of an ANN is very complex; in essence, it behaves like a black box. Recently, FNNs have been proposed as a tool for fuzzy modeling [3]. FNNs have neural network structures in which the connection weights have a direct interpretation in terms of fuzzy production rules and membership functions. FNNs have been applied to many processes in chemical and biochemical engineering. In this section,“Type I” FNNs, proposed by Horikawa et al. [3], which have been used in our study, are explained. The FNN realizes a simplified fuzzy inference of which the consequences are described with singletons. The inputs are non-fuzzy numbers. Figure 4 shows the structure of the FNN, in which the FNN has two inputs x1 and x2, one output y*, and three membership functions in each premise. The

Fig. 4 Concept of fuzzy neural network (FNN)

56

T. Hanai · H. Honda

circle and square in Fig. 4 indicate units of the neural network, while wc, wg, wf, 1, and –1 are the connection weights. The connection weights wc and wg determine the positions and gradients of the sigmoid functions “f ” in the units in (C)-layer, where sigmoid functions “f ” are defined as follows: f (x) = 1/[1 + exp{– wg(x + xc)}]

(1)

Each membership function consists of one or two sigmoid functions. In the FNN, the membership functions in the premises are tuned and the fuzzy rules are identified by adjusting the connection weights wc, wg, and wf through the back propagation learning algorithm [2]. 2.4 Genetic Algorithm (GA) A GA imitates some aspects of natural evolution [18], and is based upon the presumption that well-adapted life will survive. In the GA, the gene in chromosomes constitutes the genetic information and adaptation is brought about by crossover and mutation. This algorithm can search rapidly for the maximum or minimum value of a function and can be used for global searching and multivariable optimization in a variety of fields. The chromosomes are the line of figures derived from the sets of searching values. In the first generation, initial chromosomes are generated at random.

Fig. 5 Flowchart of genetic algorithm (GA)

Application of Knowledge Information Processing Methods

57

The usual procedure in a GA is that genes (figures) in a chromosome are changed by mutation and crossover to create a new chromosome. The objective function called “fitness function” is defined and the fitness between the target and the searching values are calculated by this function.After calculating the fitness of each chromosome, those with higher fitness are selected to survive in this generation. In the following step, genes are changed by mutation and crossover again as the gene manipulation in the next generation. These procedures are repeated until the maximum generation and they are shown as Fig. 5. 2.5 Fuzzy Adaptive Resonance Theory (Fuzzy ART) Adaptive resonance theory (ART) is a kind of neural network in knowledge information processing and is a model of cognitional information processing developed by Capenter and Grossberg [19]. Basically, ART is trained by unsupervised learning which does not use the teaching signal. ART is thought an effective method for the clustering and analysis of noised data. In ART families, there are ART 1[19] which handles only binary data,ART 2 [20] and fuzzy ART [4] which handle only real numbers, and ART 3 [21] which handles both binary data and real number. Figure 6 shows the concept of fuzzy ART.At first, the function is used for the calculation of the fitness of the input data to the cluster and the cluster with highest fitness for input data is selected. If the value calculated by the function is larger than the vigilance parameter, then the input data is used for obtaining the resonance in ART of the connection weight of the selected cluster. On the other hand, if the value calculated by the function is smaller than the vigilance parameter, then the new cluster is generated.

Fig. 6 Concept of fuzzy adaptive resonance theory (Fuzzy ART)

58

T. Hanai · H. Honda

3 Application of Knowledge Information Processing Methods to Biochemical Engineering Field 3.1 Modeling Bioprocess modeling is very useful for design, optimization, and control of biological process. Until now, many types of models have been reported, which can be categorized in various ways such as deterministic vs non-deterministic, linguistic vs mathematical equation base, data-driven vs knowledge-driven, and so on.As a recent topic, there is the metabolic modeling approach [22, 23]. Several attempts have been made in utilizing metabolic information for on-line identification and control. Takiguchi et al. reported the modeling of lysine fermentation based on a metabolic reaction model [24]. Shi et al. reported the metabolic pathway analysis of Escherichia coli using metabolic signal flow diagram [25]. Chauvatcharin et al. reported the physiological diagnosis of the acetone-butanol fermentation process [26].ANN is one of the other approaches, which is known as a black-box-type data-driven model, because the causal relationship between input variables and output variables cannot be explained by linguistic rules. ANNs have been the focus of much attention for the modeling of various bioprocesses [27]. The main advantage of ANN modeling is that one can obtain quite an accurate model without detailed knowledge of the system. 3.2 Physiological State Recognition If a model with high accuracy and high robustness is constructed, process optimization and control will be achieved. In several cases, however, state recognition during fermentation is needed because of difficulty in process modeling in one state. In each state, the strategy of optimal process operation is determined. Fuzzy inference is a powerful tool for monitoring physiological state [28–31]. Horiuchi et al. reported several laboratory-scale applications of fuzzy inference to a-amylase production by Bacillus species [30] and b-galactosidase production by recombinant E. coli [31]. In Japan, fuzzy inference for phase recognition has been applied to industrial scale production in some bioprocesses, such as vitamin B2 (VB2) [32] and pravastatin precursor [33]. VB2 will be introduced as one example of phase recognition and application to process control. 3.2.1 Example – VB2 Production Fuzzy inference system was applied to the on-line control of feed rate and pH for the fed-batch cultivation of Bacillus species to produce VB2 [32].

Application of Knowledge Information Processing Methods

59

Fig. 7a–c Results of fuzzy control of a fed-batch culture for recombinant VB2 production: a time courses of average adaptability of on-line data to the rule for each phase; b glucose feed rate and pH; c cell concentration and VB2 concentration

Microbial VB2 production has been developed by Nippon Roche, Japan and the product, single-step fermentative VB2 using a recombinant Bacillus strain, was effectively produced directly from glucose in a fed-batch operation. For the phase identification, four parameters (the culture time, CO2 evolution rate, total CO2 evolution, and DO) were selected as state variables on the basis of operating experiences and a simulation study. The fermentation period was successfully divided into four states, namely lag phase, growth phase, production phase 1, and production phase 2 (Fig. 7). The state at certain fermentation times was expressed as the grade of the state determined by fuzzy membership function since they have ambiguous and overlapping boundaries. The culture phase transitions from the lag phase to production phase 2 were properly recognized by the system. Identification of the culture phases by the inference system coincided almost exactly with identifications made by the experienced operators. From these results, the glucose feed rate and pH were controlled by the system. As a result of appropriate feeding and pH control, the final VB2 concentration reached the level of the maximum concentration achieved in the fed-batch culture manually controlled by experienced operators.

60

T. Hanai · H. Honda

3.3 Software Sensor In bioprocesses, all state and process variables are not always available. The concentration of glucose as a sole carbon source is sometimes unknown and the value is obtained as off-line data. The cell concentration is not known in the actual medium containing insoluble components. In such cases, the value of important variables should be estimated from the data measured by a limited number of reliable on-line sensors. For this purpose, knowledge information approaches, such as fuzzy inference, ANN, and so on, are frequently utilized and known as a software sensor [34–37]. Estimated value by a software sensor is utilized for process control of the bioprocess. Software sensing in activated sludge process and determination of operation condition will be introduced as an example [38, 39]. 3.3.1 Example – Activated Sludge Process The activated sludge process has been widely used for water treatment of both municipal and industrial wastewaters. Activated sludge can convert various organic compounds in wastewater to carbon dioxide by the oxidative activities of aerobic microorganisms. However, to date, many models have been proposed to describe the dynamic characteristics of the process, such as the activated sludge models (ASMs) no. 1 [40] and no. 2 [41]. These models are constructed based on Monod’s equation and there are 19 (ASM No. 1) and over 50 parameters (No. 2), respectively, and it is difficult to evaluate these parameters exactly. In order to avoid the problems of modeling and to propose a new modeling method, we applied a software sensor using knowledge information processing methods. In this study, effluent chemical oxygen demand (COD) value was estimated using a software sensor such as FNN model with a recursive learning data-renewing method, called as the RFNN model [38]. Moreover, in order to control effluent COD values to a desired level, we combined a search method with the RFNN model. This method combines with a search of dissolved oxygen concentration (DO) and mixed liquor suspended solid (MLSS) using GA with a reliability index (RIGA) proposed by us [39]. Figure 8 shows the flow chart of this study. In this study, we used the time series data of three months (April 1999 to June 1999) measured at the wastewater treatment plant “U”, of which a flow diagram is shown in Fig. 9. Thirty-five items, the candidates of input variables in the RFNN model for effluent COD value, are measured at the positions (a) to (g) in Fig. 9 every 5 min; for example, suspended sludge (SS), influent COD value, influent flow rate, DO, MLSS, pH, water temperature, etc. Based on the knowledge of the skilled operators, six measured items, COD at (a), influent flow rate at (a), DO at (b), MLSS at (b), water temperature at (d), and pH at (d), were selected as the input variables for FNN from 35 measured items.

Application of Knowledge Information Processing Methods

Fig. 8 Concept of application of FNN and GA to activated sludge process

Fig. 9 Flow diagram of activated sludge plant “U”

61

62

T. Hanai · H. Honda

Fig. 10 Estimation result for effluent COD by RFNN. Notations: open circle, actual value; closed circle, estimated value

Estimation of effluent COD value was carried out every hour using three months data (April 1999 to June 1999). Figure 10 shows a three-week section of results from the three months data and the average error was 0.46 mg/l, the maximum error was 3.59 mg/l, and the standard deviation of relative error was 4.30%. For all three months data, those values were 0.68 mg/l, 3.90 mg/l, and 5.04%, respectively. Around days 9 to 11 in Fig. 10 the measured values of effluent COD oscillated sharply and the estimated values were not close to the measured values. This is due to the heavy rain. If high accuracy is desired, it is necessary to collect enough data sets in heavy rain days and use these as the learning data or to construct another model such as an expert system only for heavy rain days. Reliability index (RI) [39] value is calculated by the equation below, which indicates the reliability of the estimated value for the point of interest: 1 3 log error 2k RI = – 3 Â 08 3 k=1 rk

(2)

where rk indicates the Euclid distance between the point of interest and k-th nearest data, and errork indicates the learning error of k-th nearest data. The higher the RI value of the point of interest is, the higher the accuracy of the simulation obtained. RI in RIGA enables GA to search DO and MLSS values from where the reliability is higher than elsewhere. Flow chart of RIGA is shown in Fig. 11. In RIGA, the following operations are different from GA; generation of initial population, calculation of fitness and RI, and reproduction. In generation of initial population, N data sets are generated randomly from where RI values are relatively high. The fitness and RI value of each individual is calculated. To implement the reproduction procedure, first the rank of each individual is determined for fitness and RI value, respectively, and then the total rank of each variable is determined from the sum of the fitness-rank and the

Application of Knowledge Information Processing Methods

63

Fig. 11 Flowchart of genetic algorithm with reliability index (RIGA)

RI-rank. Second, the survival probability of the total rank n-th individual is calculated using the linear ranking approach. The RFNN model combined with GA was used to search DO and MLSS values in order to control effluent COD value to the desired level. The measured value of effluent COD was set to the desired value of effluent COD.A search was carried out every hour and the results are shown in Table 1. As a result, the search for DO and MLSS values at various times using GA was carried out with Table 1 Searched result by GA and RIGA DO

MLSS

a

Search method

Average error [mg/l]

Maximum error [mg/l]

Standard deviation of relative error [%] a

GA RIGA

0.16 0.11

1.10 1.00

29.3 20.2

Search method

Average error [mg/l]

Maximum error [mg/l]

Standard deviation of relative error [%] a

GA RIGA

214 144

833 737

11.5 8.33

Relative error: | measured value – searched value |/measured value

64

T. Hanai · H. Honda

Fig. 12 Results of the search for DO and MLSS values by RIGA. Notations: open circle, actual value; closed circle, searched value

high accuracy. Next, the same search for DO and MLSS values using the RFNN model combined with RIGA was carried out under the same condition of GA and the results are shown in Table 1. Figure 12 shows a three week section of results from the three months data by RIGA. As shown in Table 1, accuracy of search using RIGA was much higher than that using GA. As a result, the search for DO and MLSS values using RIGA was carried out with higher accuracy. It was verified that RI had a great positive effect on search for DO and MLSS values using GA. 3.4 Process Control Process control is a final goal in the biochemical engineering field. As mentioned above, a knowledge information approach such as phase recognition by fuzzy inference and a software sensor by ANN is available for final process control. Since the bioprocess possesses various characteristics which confound the establishment of control systems such as nonlinear characteristics with time-dependent parameters, large time constants of the process dynamics, and so on, such knowledge information approaches have been popular for the

Application of Knowledge Information Processing Methods

65

establishment of sophisticated control systems. Utilizing phase recognition and a software sensor is indirect process control and there are some reports on direct control such as direct fuzzy reasoning and ANN. Some typical applications in Japan are fuzzy control of glutamic acid fermentation [42], sake mashing processes [16], and enzyme production [43]. In glutamic acid production, control of dissolved oxygen (DO) concentration was carried out by a fuzzy control schema with manipulation of the aeration and agitation rates taking into account the time increment of DO. In sake mashing process, temperature control rules were developed based on an interview with an expert. Many applications of ANN for the direct control of bioprocesses have been reported. Karim et al. [6] reported the application of ANN to the control of glutamine and glucose feeding for monoclonal antibody production in a hybridoma cell culture.As advanced knowledge information develops, the combination of fuzzy inference and ANN is proposed. Neuro-fuzzy or fuzzy neuro control schema have been applied to some bioprocesses. Ye et al. reported the application of fuzzy neural networks in recombinant cell culture [44]. Honda et al. reported the temperature control in Ginjo sake mashing process using fuzzy neural network [45]. FNNs have neural network structures in which the connection weights have a direct interpretation in terms of fuzzy production rules and membership functions and it is a white-box model, not a black-box model such as ANN model. One of the remarkable characteristics of FNN is an automatically modeling such as input variable selection and tuning of weight parameter. Direct process control based on knowledge information approaches will become powerful for bioprocess control in the future.

4 Application to Other Fields 4.1 Biomedical Field – Prediction of Prognosis for Cancer Patients from DNA Microarray Data by FNN Despite recent progress in clinical study and biological technology of cancer, the prognostic prediction of patients still remains difficult and inaccurate. Since DNA microarrays permit a simultaneous analysis of multiple genes, it has been used to profile gene expression which can categorize cancers into subgroups [46]. To analyze gene expression data, many statistical techniques have been used, but there would be relationships among genes that cannot be expressed statistically. Therefore, ANN including FNN is useful for finding relationships with high accuracy [47]. However, the immensity of data makes us spend much computational time to construct the ANN or FNN model. We applied fuzzy neural network (FNN) combined with SWEEP operator method [48] to accelerate the calculation speed about 30 times faster than the FNN modeling method using the back propagation leaning algorithm. The con-

66

T. Hanai · H. Honda

structed models achieve high accuracy for prognosis of patients. The results in modeling were evaluated to compare with those of Multiple Regression Analysis (MRA). In this study, we used gene expression data from a study of Alizadeh et al. [46]. The expression levels from 12,069 cDNA clones are measured for each of 40 patients who were also followed up for 4 years. Single spots or areas of the array with obvious blemishes were deleted. All other array elements for which the fluorescent intensity in each channel was 1.4 times greater than the local background were considered well measured. The ratio values were log-transformed (base 2). If genes for any patients are not well measured, they were omitted. With this preprocessing, 2862 genes were applied for the following analysis. In this study,“Type I” FNN proposed by Horikawa et al. [3] was also used. The grade of membership function for each rules are shown as m in Fig. 4. The membership function and the fuzzy rules are identified by adjusting the connection weight wc, wg, and wf through the back propagation leaning algorithm. If membership function parts are fixed, fuzzy rule parts of “Type I” FNN are regarded as MRA model of which input is m. Therefore, SWEEP operator method that can calculate the parameter of MRA with high speed was used for efficiently computation [48]. To identify input variables of the model from a candidate gene, the parameter increasing method (PIM) is also applied. SWEEP operator method combined with PIM selected four input variables from 2862 candidate genes quickly. We used a fourfold cross validation procedure to check the estimation ability. Results of the each model using four genes as input variable are shown in Table 2. The genes selected by the PIM are shown in the order selected. The P-value for each gene, calculated by the MannWhitney test, indicates the significance of expression differences between patients with four-year survival and those without. Using FNN models with only four genes, average correctness for predicting prognosis was 93%. One of the four selected genes included CD10 gene, the known maker of lymphoma. Another known gene, IRF-4, expression may confer a growth advantage on the lymphoma cells. Others are unknown genes. This may indicate that CD10, IRF-4, and another two genes are biologically important for prognosis. Furthermore, our method could show the possibility of discovering other novel makers in other diseases. MRA models selected four unknown genes, and their

Table 2 Four genes selected with the FNN model Order of selection

Selected genes

p-Value

Predictive accuracy

1 2 3 4

CD10 Unknown (AA807551) Unknown (AA805611) IRF-4

0.008 0.002 0.032 0.022

93%

Application of Knowledge Information Processing Methods

67

average correctness ratio was 78%. One of the MRA models was only 60% correctness ratio. These showed the advantage of FNN over MRA. From the constructed FNN model, the relationship between the input of four genes and the output of the survival score is described as a fuzzy rule, shown in Fig. 13. H and L refer to high and low expression level of each gene, respectively. Since the expression level of each gene can be divided into either high or low groups according to fuzzy reasoning, this model comprised 16 (=24) fuzzy rules. Hatched areas represent predicted poorer prognosis by the FNN. Filled areas represent the poorest prognosis. Numbers in each matrix cell are the patients’ numbers previously described by Alizadeh et al. Bold type numbers indicate the patients dead within four years, and italic type numbers alive. Patient numbers are placed in the matrix according to the expression levels of the four genes in that patient. Patient numbers in circles represent incorrect classification by the FNN. From this matrix, the following rules are obtained. Patients with low expression of CD10 were predicted to have a poor prognosis in the FNN model. A poor outcome was predicted particularly when CD10 expression was low and IRF-4 expression was high. Fourteen of the patients were identified as having poor prognosis on the basis of these two factors, which corresponds to 67% of all patients with poor prognosis. Furthermore, the FNN model also identified those cases with a poor prognosis despite a high expression ratio of CD10. The correct identification of these cases was obtained by adding the expression information on the other two genes. Our study attained a high prediction accuracy of 93% with the FNN model. This means that 3 out of 40 patients’ prognoses were incorrectly predicted. These three

Fig. 13 Relationship among four input genes and predicted outcome

68

T. Hanai · H. Honda

Fig. 14 Kaplan-Meier plot of the four-year overall survival for patients grouped by FNN score (0.5)

patients, numbers 5, 11, and 24, are indicated with circles in Fig. 13. Patients numbers 11 and 24 had intermediate levels of CD10 expression, indicating the patients were considered marginal for survival. Kaplan-Meier survival analyses indicated that the patients predicted alive by the FNN model showed longer survival than the patients predicted dead (Fig. 14). We constructed prediction models at high speed and with accuracy in spite of a large amount of gene expression profiling data. Our method can easily accommodate more immense and nonlinear feature of genetic data. 4.2 Bioinformatics Field – Gene Clustering for DNA Microarray Data by Fuzzy ART The recent advances of genome-scale sequencing and array technologies have made it possible to monitor simultaneously the expression pattern of 3000 to 20,000 independent genes. One of the goals is to discover or extract the information for the genetic networks by analyzing such massive data sets. It is, however, a hard problem to interpret the interaction between genes, because many different types of expression pattern may be present. Therefore, various clustering methods, i.e., hierarchical clustering [49] or self-organizing maps (SOMs) [50], have been examined and used to elucidate the fundamental or/and characteristic expression pattern. In this section, we applied a fuzzy adaptive resonance theory (Fuzzy ART) model, a type of unsupervised clustering method, to the experimental data. The results were compared with those of hierarchical clustering, k-mean clustering, and SOMs. We used expression data from a study of Chu et al. using DNA microarray [51]. Saccharomyces cerevisiae was synchronized by transferred them to sporulation medium (SPM) at t=0 to maximize the synchrony of sporulation. RNA was harvested at time t=0, 0.5, 2, 5, 7, 9, and 11.5 h after transfer to SPM. About 6100 genes of expression profiles are included in their data. Among them, we finally selected 45 genes (shown in Table 3), whose functions are biologically characterized by Kupiec et al. [52]. The learning procedure of Fuzzy ART has

Application of Knowledge Information Processing Methods

69

Table 3 Gene list using analysis and induction period and clustering result by Fuzzy ART Cat.

Name

Time

Cat.

Name

Time

1

DMC1 IME2 ME15 RED1 HOP1 MEK1 MSH4 MSH5 REC114 SPO11 SPO13 SPO16 ZIP1 CDC14 CDC23 IME4 ME14 MPS1 MSI1 POL30 RAD51 RAD54 RAP1

Early Early

3

REC102 REC104 RFA1 SAE3 SPO12 SPS19 YPT1 ZIP2 CDC16 DIT1 DIT2 CDC20 CDC5 ISC10 NDT80 SGA1 SPO20 SPR1 SPR3 SPR6 SPS1 SPS18

Early Early

2

3

Early Early Early

Early Early Early Early Early

Early Early

4

5

Middle

Mid-Late Mid-Late

Late Late Late Middle

been reported by Tomida et al. [53]. We decided the number of the clusters of genes is five, based on the agreement of the clustering result with the biological knowledge. In the other method such as hierarchical clustering, k-mean clustering, and SOMs, we selected the same cluster number as that of generated clusters using Fuzzy ART in order to compare the clustering results. Figure 15 and Table 3 show the clustering result of 45 genes by Fuzzy ART. Compared with the other clustering methods, only the clustering result of Fuzzy ART classified the “DIT1” and “DIT2” genes expressed in the Middle Late stage during sporulation into a cluster which is independent of the other genes expressed in the other stages. When we analyze a set of time series expression data, it seems important to consider the shapes of expression profiles not only as simply several dimensional inputs but also as the timely continuous data during a specific biological phase. From this point of view, we propose to analyze the similarity of profiles in terms of two-dimensional area, where axes of two-dimension are “time” and “expression level”. Then we define the “gap index” so as to evaluate the similarity of profiles as area between each profile and average profile during the temporal phase. An average profile for cluster n is defined as an average of all profiles of cluster n. The concept of gap is shown

70

T. Hanai · H. Honda

Fig. 15 Representative time course of five clusters

as a shaded area in Fig. 16. Figure 17 shows the gap index for each cluster in four clustering methods mentioned above. For each cluster, the gap index of Fuzzy ART was set to 100 and the other gap indexes were calculated as relative value against it. Since cluster 1 in hierarchical clustering contains only one gene, the gap index of cluster 1 in hierarchical clustering is void. It is clear that the average gap index of Fuzzy ART is the lowest. In order to compare the clustering repeatability, we generated five sets of randomly noised data. Generally, the fluctuation of microarray data is within about a twofold change, and we added a random value from –1.0 to 1.0 to the log2Rt value. Table 4 shows the results of clustering robustness using five sets of noised data in four clustering methods. In the case of Fuzzy ART, 225 (45 genes¥5 sets) genes were clustered into the same clusters as those using un-noised data. That is to say, 79% genes were preserved in terms of robustness after adding random noise.We defined the robustness ratio as the ratio of genes whose clustering result was coherent. In the cases of hierarchical clustering, k-means algorithm, and SOMs, robustness ratios were 73, 56, and 57, respectively. It is obvious that the clustering result by Fuzzy ART is the highest score. Table 4 Comparison of clustering results using noised data

Average of clustering robustness Average of correctness ratio

Fuzzy ART

Hierarchical clustering

K-means algorithm

SOMs

79.1

73.3

55.6

57.3

0.85

0.78

0.80

0.82

Application of Knowledge Information Processing Methods

71

Fig. 16 Concept of gap index

Fig. 17 Comparison of gap index

That means Fuzzy ART is also more useful than the other clustering methods in the case of noised data.

5 Conclusion and Perspectives In this chapter, the applications of the knowledge information processing methods to the biochemical engineering, biomedical, and bioscience fields using our studies as the example were introduced. These methods could analyze with satisfactory results many kinds of phenomena which the conventional statistical method or mathematical model could not analyze or simulate. In the near future, popularization of DNA microarrays, development of new through-

72

T. Hanai · H. Honda

put analysis methods for protein, and bio-combinatorial chemistry enable us to get more kinds and larger amounts of biological data. Application of information technologies, especially knowledge information processing methods, to the biochemical engineering, biomedical, and bioscience fields is becoming more important day by day.

References 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18. 19. 20. 21. 22. 23. 24. 25. 26. 27.

Zadeh LA (1965) Inf Control 8:338 Rumelhart DE, Hinton GE, Williams RJ (1986) Nature 323:533 Horikawa S, Furuhashi T, Uchikawa Y (1991) Proc IFES 91:562 Carpenter GA, Grossberg S, Rosen DB (1991) Neural Networks 4:759 Konstantinov K, Yoshida T (1992) Biotechnol Bioeng 39:479 Karim MN,Yoshida T, Rivera SL, Saucedo VM, Eikens B, Oh GS (1997) J Ferment Bioeng 83:1 Shioya S, Shimizu K, Yoshida T (1999) J Biosci Bioeng 87:261 Nishida Y, Fukaya I, Takahashi N, Hanai T, Honda H, Kobayashi T (1994) Seibutsu-kogaku 72:267 (in Japanese) Hanai T, Honda H, Takahashi N, Nishida Y, Fukaya I, Kobayashi T (1994) Seibutsu-kogaku 72:275 (in Japanese) Hanai T, Nishida Y, Ohkusu E, Honda N, Fukaya I, Kobayashi T (1995) Seibutsu-kogaku 73:283 (in Japanese) Kakamu A, Hanai T, Honda H, Kobayashi T (1995) Seibutu-kougaku 73:387 (in Japanese) Noguchi H, Hanai T, Takahashi W, Ichii T, Tanikawa M, Masuoka S, Honda H, Kobayashi T (1991) Kagaku Kougaku Ronbunshu 25:695 (in Japanese) Hibino S, Hanai T, Nagata E, Matubara M, Fukagawa K, Shirataki T, Honda H, Kobayashi T (2001) J Chem Eng Jpn 34:936 Noguchi H, Hanai T, Honda H, Harrison LC, Kobayashi T (2001) J Biosci Bioeng 92:227 Noguchi H, Kato R, Hanai T, Matubara Y, Honda H, Brusic V, Kobayashi T (2002) J Biosci Bioeng 94:264 Oishi K, Tominaga M, Kawato A, Abe Y, Imayasu S, Nanba A (1991) J Ferment Bioeng 72:115 Tsuchiya Y, Koizumi J, Suenari K, Teshima Y, Nagai S (1990) Hakkokogaku 68:123 (in Japanese) Holland JH (1975) Adaptation in natural and artificial systems. University of Michigan Press, Michigan Capenter GA, Grossberg S (eds) (1991) Pattern recognition by self-organizing neural networks. A Bradford Book. MIT Press, Boston Capenter GA, Grossberg S (1987) Appl Opt 26:4919 Capenter GA, Grossberg S (1990) Neural Networks 3:129 Bailey JE (1991) Science 252:1668 Stephanopoulos G, Vallino JJ (1991) Science 252:1675 Takiguchi N, Shimizu H, Shioya S (1997) Biotechnol Bioeng 55:170 Shi H, Shimizu K (1998) Biotechnol Bioeng 58:139 Chauvatcharin S, Siripatana C, Seki T, Takagi M, Yoshida T (1998) Biotechnol Bioeng 58:561 Karim MN, Revera SL (1992) In: Fiechter A (ed) Adv Biochem Eng Biotechnol, vol 46. Springer, Berlin Heiderberg New York, p 1

Application of Knowledge Information Processing Methods 28. 29. 30. 31. 32. 33. 34. 35. 36. 37. 38. 39. 40. 41. 42.

43. 44. 45. 46.

47. 48. 49. 50. 51. 52. 53.

73

Konstantinov K, Yoshida T (1989) Biotechnol Bioeng 33:1145 Konstantinov K, Yoshida T (1992) J Biotechnol 24:33 Horiuchi J, Kamasawa M, Miyakawa H, Kishimoto M (1993) J Ferment Bioeng 76:207 Horiuchi J, Kamasawa M, Miyakawa H, Kishimoto M, Momose H (1993) J Ferment Bioeng 76:382 Horiuchi J, Hiraga K (1999) J Biosci Bioeng 87:365 Hosobuchi M, Fukui F, Suzuki T, Yoshikawa H (1993) J Ferment Bioeng 72:115 Huang YW, Mithani R, Takahashi K, Fan LT (1993) Biotechnol Prog 9:401 Albiol J, Campmajo C, Casas C, Poch M (1995) Biotechnol Prog 11:88 Kobayashi T, Honda H, Hanai T, Ueda N (1996) J Environ Instrum Control Autom 1:106 Imanishi T, Hanai T, Aoyagi I, Uemura J, Araki K, Yoshimoto H, Harima T, Honda H, Kobayashi T (2002) Biotechnol Bioproc Eng 7:275 Tomida S, Hanai T, Honda H, Kobayashi T (2001) J Chem Eng Japan 34:369 Yoshikawa H, Hanai T, Tomida S, Honda H, Kobayashi T (2001) J Chem Eng Japan 34:1033 Hanze M, Grady CPL, Gujer W, Marais GR, Matuo T (1987) Activated Sludge Model No. 1. Scientific and Technical Report No. 1. IAWQ, London Hanze M, Gujer W, Mino T, Matsuo T, Wentzel MC, Marais GR (1995) Activated Sludge Model No. 2. Scientific and Technical Report No. 3. IAWQ, London Nakamura T, Kuratani T, Morita Y (1985) In: Johnson A (ed) Proceedings of IFAC Symposium on Modeling and Control of Biotechnology Processes. Pergamon Press, Oxford, p 211 Shiba S, Nishida Y, Park YS, Iijima S, Kobayashi T (1994) Biotechnol Bioeng 44:1055 Ye K, Jin S, Shimizu K (1994) J Ferment Bioeng 77:663 Honda H, Hanai T, Katayama A, Tohyama H, Kobayashi T (1998) J Ferment Bioeng 85:107 Alizadeh AA, Eisen MB, Davis RE, Ma C, Lossos IS, Rosenwald A, Boldrick JC, Sabet H, Tran T, Yu X, Powell JI, Yang L, Marti GE, Moore T, Hudson J Jr, Lu L, Lewis DB, Tibshirani R, Sherlock G, Chan WC, Greiner TC, Weisenburger DD, Armitage JO, Warnke R, Levy R, Wilson W, Greverss MR, Byrd JC, Bostein D, Brown OP, Staudt LM (2000) Nature 403:503 Khan J, Wei JS, Ringner M, Saal LH, Ladanyi M, Westermann F, Berthold F, Schwab M, Antonescu CR, Peterson C, Meltzer PS (2001) Nat Med 7:673 Ando T, Suguro M, Hanai T, Kobayashi T, Honda H, Seto M (2002) J Cancer Res 93:1207 Eisen MB, Spellman PT, Brown PO, Bostein D (1998) Proc Natl Acad Sci USA 95:14863 Tamayo P, Slonim D, Mesirov J, Zhu Q, Kitareewan S, Dmitrovsky E, Lander ES, Golub TR (1999) Proc Natl Acad Sci USA 96:2907 Chu S, Derisi J, Eisen M, Mulholland J, Botstein D, Brown PO, Herskowitz I (1998) Science 282:699 Kupiec M, Ayers B, Esposito RE, Mitchell AP (1997) The molecular and cellular biology of the yeast Saccaromyces. Cold Spring Harbor, New York Tomida S, Hanai T, Honda H, Kobayashi T (2002) Bioinformatics 18:1073

Received: September 2003

Adv Biochem Engin/Biotechnol (2004) 91: 75– 103 DOI 10.1007/b94206 © Springer-Verlag Berlin Heidelberg 2004

Large-Scale Production of Hairy Root Nobuyuki Uozumi (✉) Bioscience and Biotechnology Center, Nagoya University, Nagoya 464-8601, Japan [email protected]

1 1.1 1.2

Induction of Hairy Root from Plants . . . . . . . . . . . . . . . . . . . . . Transgenic Plant, Hairy Root . . . . . . . . . . . . . . . . . . . . . . . . . Strategy of Biochemical Mass Production in Hairy Root . . . . . . . . . .

2

2.4 2.5

Conductivity Measurement for Cell Growth and Bioreactor Development for Hairy Root . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Estimation of Hairy Root Cell Mass by Medium Conductivity . . . . . . Bioreactor Configuration . . . . . . . . . . . . . . . . . . . . . . . . . . Cell Yields Determination for Main Components and Fed-Batch Culture Using Sucrose as a Carbon Source . . . . . . . . . Fed-Batch Culture Using Monosaccharide . . . . . . . . . . . . . . . . . Yield Coefficient for Hairy Root Biomass and Maintenance Energy . . .

3

2.1 2.2 2.3

76 76 77

. . .

79 79 80

. . .

81 83 85

Kinetics of Hairy Root and Simulation . . . . . . . . . . . . . . . . . . . .

87

4

Release and Recovery of Enzymes and Secondary Metabolites from Hairy Root . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.1 Release of Biochemical Products from Hairy Root . . . . . . . . . 4.1.1 Release and Enhancement of Peroxidase Production and Excretion from Horseradish Hairy Root and Medium Control . . . . . . . . 4.1.2 Cultivation in Combination with Adsorption . . . . . . . . . . . 4.1.3 Repeated Batch Culture of Horseradish Hairy Root . . . . . . . . 4.2 Release of Pigments by Controlling the Airation and Development of Repeated Recovery System . . . . . . . . . . . . . . . . . . . . 5

Use of Photoautotrophic Hairy Roots for the Culture

. . . . . . . . . .

90 90

. . . . . . . . . . . . . . .

90 91 91

. . . . .

92

. . . . . . . . . . .

93

Production of Regenerated Plant from Hairy Root . . . . . . . Regeneration of Hairy Root . . . . . . . . . . . . . . . . . . . . The Use of Root Fragments for Encapsulation . . . . . . . . . . Encapsulation of Adventitious Shoot Primordia . . . . . . . . . Encapsulation of Plantlet Regenerated from Hairy Root . . . . Mass Production of Hairy Root Fragments Using the Blender . Auxin Supplementation Stimulates Hairy Root Growth Stage on Plantlet Formation . . . . . . . . . . . . . . . . . . . . . . 6.4.3 Cytokinin Supplementation Stimulates on Plantlet Formation . 6.4.4 Improvement of Frequency of Plantlet Development Through Plantlet Dehydration . . . . . . . . . . . . . . . . . . 6.5 Application of Transgenic Plant to the Artificial Seed Procedure 6 6.1 6.2 6.3 6.4 6.4.1 6.4.2

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

93 93 95 96 97 97

. . . . . . . . . . . .

98 98

. . . . . . 99 . . . . . . 100

References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102

76

N. Uozumi

Abstract Many products of interest are synthesized in organized tissues, but not formed in suspension or callus culture. Therefore, most attention has been focused on root cultures. The transgenic plant,“hairy root”, has brought us to dramatic improvements in growth rate and high content of desirable products. Since the roots are quite different from callus in morphology, the culture manner should be explored independently. By providing a growth environment, an elite hairy root can be a more attractive plant. Both of strain selection to generate more competent plants in breeding and engineering development are necessary to overcome various limitations. In this chapter the engineering issues involved in using hairy root culture are discussed, as follows. 1. Measurement of cell concentration on line, and a designing bioreactors for hairy root in liquid culture. 2. High cell density culture and its kinetic parameters. 3. Secretion of target products. 4. The micropropagation of the regenerated hairy root by means of artificial seed system. In some cases where callus and suspension culture show negligible productivity, organ culture will be necessary to achieve good formation. This study on hairy root culture indicates one of the best attempts to the recovery of products from the organ culture in plant biotechnology. Keywords Hairy root · Bioreactor · Culture · Regeneration · Artificial seed

1 Induction of Hairy Root from Plants 1.1 Transgenic Plant, Hairy Root Developments in plant cell and tissue culture technology have extended to the production of important phytochemicals. There are two primary requirements for the effective production of phytochemicals from a plant cell culture: selection of an elite plant which contains more of the target chemicals than other plants in combination with in vitro mutation, and the development of suitable bioreactor designs and culture techniques. In particular, the commercial use of plant cells requires the development of effective culture processes because of the difficulties associated with high plant cell density and the extraction of intracellular products, even if a superior plant is produced. The Agrobacterium plasmids as vector systems for plants is a natural progression of the organisms: first, on the plant disease associated with it, and second in respect of the similarity of the disease symptoms to certain cancers. Agrobacterium rhizogenes is responsible for hairy root induction in infected sensitive plants as shown in Fig. 1. The phenomenon is due to the transfer, integration, and expression in the plant cell genome of DNA (T-DNA) originating from large plasmids called Ri (root inducing) plasmids [1, 2]. The hairy root is similar to the original root in shape and properties. By means of phytophormone supplementation, root is induced from the originated plant tissue. Hairy root transgenic plant does not, however, require the phytohormone for its growth. The induced hairy root has some advantageous properties, such as

Large-Scale Production of Hairy Root

77

Fig. 1 Hairy root induced by the integration of T-DNA in Ri plasmid

the higher content of target secondary metabolites, higher genetic stabilities and growth rate increment. Transformed root “hairy root” provides a promising alternative to the biotechnological exploitation of plant cells. Genetic modification using A. rhizogenes plasmids as vectors seems to be feasible for the improvement of plant properties and for the production of transgenic plants. Hairy root is characterized by secondary metabolite production and inherent genetic stability reflected in stable productivity. The diversity of secondary metabolites in the root shows the immense biochemical potential contained in this organ. Since tissue specific behavior and production are organized, the specific tissue or organ (i.e., root) is suitable for production of useful chemicals. 1.2 Strategy of Biochemical Mass Production in Hairy Root There are two main strategies for obtaining biochemical products in hairy root – high density culture of hairy root by using bioreactor in combination with effective control technique, and the micropropagation of hairy root.

78

N. Uozumi

Fig. 2 Process of biochemical production from hairy root

The use of the cultured hairy roots has consistently been focused on the large scale production of useful products or secondary metabolites such as pigments and alkaloids. The optimization of fermentor-scale plant cell culture requires knowledge of substrate requirements and utilization rates. To obtain a high density culture of plant cells, the culture conditions should be maintained at the optimum level. To optimize the culture condition, the cell mass has to be monitored correctly. Conductivity in the medium has been found to be useful

Large-Scale Production of Hairy Root

79

for cell mass evaluation. From the viewpoint of process development, carbon utilization of hairy root in liquid culture to evaluate the metabolic energy costs should be characterized. The chemicals produced by plants are often localized at specific tissues, and some types of them are stored in roots. The excretion and recovery of the biochemicals have been required to attain the efficient culture (Fig. 2). The above studies including bioreactor development are used for recovery of interested products from hairy root. Artificial seeds are expected to be a reliable delivery system for clonal propagation of elite plants. The delivery system has the potential for genetic uniformity, high yield and low cost of production. The plant cells for artificial seeds require a good ability of regeneration and a high resistance against disease and mutation. From some species of hairy roots, it has also been observed that the regeneration occurs under light conditions. Successful regeneration of the whole plant from Ritransformed cells has been obtained with various species. In general, genetic improvement of plants through conventional breeding and selection methods takes a long time. Within an acceptable time period, new gene transfer technologies offer the opportunity to produce plants easily with desirable traits such as disease or herbicide resistance. Successful reports on elite transgenic plant cells and their advantageous properties stimulated interest in developing regeneration and delivery system of hairy roots. The proper system of plant regeneration is necessary to gain the transgenic plant from the hairy root efficiently. In particular, the production process should be constructed and improved to realize these plants to the artificial seed in industrial scale (Fig. 2).

2 Conductivity Measurement for Cell Growth and Bioreactor Development for Hairy Root 2.1 Estimation of Hairy Root Cell Mass by Medium Conductivity Biomass concentration is a crucial point for the state of any biological process. The most frequently used methods for determining biomass concentration are the microscopic measurement of the cell number, the number of cells in a defined volume, and the measurement of dry cell weight. Both methods require an aseptic sample from the bioreactor. A grave drawback of both methods is the time requirement: the values are available only a considerable time (up to several hours) after sampling. Thus, more sophisticated method should be developed. The methods generally adopted for cell mass determination in plant cell cultures are gravimetric and volumetric measurements on a wet or dry basis, or the microscopic counting of cells after protoplast-forming treatment. Hairy roots normally develop in branched, filamentous organs, which make it impossible to obtain a homogeneous sample of the roots, or to employ the routine methods mentioned above in reactor cultures and especially in immobilized cell systems.

80

N. Uozumi

Conductivity measurement of culture broth was developed as a convenient tool for cell mass determination in plant cell suspension (or callus) culture [3, 4]. The effect of environmental change on medium conductivity was almost negligible under plant cell culture conditions. Cell mass grown was proportional to the decrease in the medium conductivity during the culture of carrot, horseradish, and Cassia torosa hairy roots. For each root culture, straight lines were obtained as Dk = a(DX)

(1)

where X is dry cell mass concentration [g/l], a is empirical coefficient [S m2/g] and k is specific conductivity [S/m]. The values of a were 46, 47, and 37 S m2/g for carrot, horseradish and Cassia torosa, respectively. Thus, it is possible to estimate the cell mass concentration from conductivity measurement. The decrease in medium conductivity appeared to reflect the amount of electrolytic or inorganic nutrients (mainly NH4+ and NO3–) consumed by the cells. Therefore, the value has to be experimentally predetermined for individual plant cells. 2.2 Bioreactor Configuration The configuration of three types of bioreactors is depicted in Fig. 3. The turbine-blade reactor (Model TBR-2, Sakura Seiki Co., Tokyo) was developed for hairy root culture [5].A cultivation space was separated with an agitation space by a stainless steel mesh so that hairy roots were not in contact with the impeller [5]. An impeller with eight turbine blades was controlled at 200 rpm. Humidified air was introduced into the agitation space. The hairy roots showed the most efficient growth in the turbine-blade reactor. The maximum growth rate was 0.63 g/l per day and 10 g/l of dry cell mass was obtained after 30 days, whereas only 4 g/l was obtained in flask cultures. In a rotating-drum reactor shown in Fig. 3, a cylindrical glass reactor was used as a reactor. The drum reactor with 1 l medium was rotated at 5 rpm on a Cellrotator (Shibata Hario Glass Co., Tokyo). Humidified air was supplied to the medium through a submerged nozzle. The preliminary experiments showed that hairy root cells adhering to the wall of the reactor were lifted above the liquid medium and then dropped back into the medium as the drum rotated. Cell disruptions occurred due to these repeated drops and growth was slower. To improve the slow growth, a polyurethane foam sheet which served as a support for the cells was attached onto the wall of the reactor. The adhering hairy roots became entangled in the sheet and showed active proliferation from the beginning of cultivation without detachment from the support when lifted above the medium. With this modification, the growth rate was significantly improved and a maximum growth rate of 0.61 g/l per day as was comparable to that obtained in the turbine-blade reactor. A turbine-blade reactor and an immobilized rotating drum reactor were found to be advantageous for hairy root culture [6].

Large-Scale Production of Hairy Root

81

Fig. 3A–C Experimental apparatuses: A turbine-blade reactor: 1, conductivity cell; 2, conductivity meter; B rotating-drum reactor; C air-lift reactor

2.3 Cell Yields Determination for Main Components and Fed-Batch Culture Using Sucrose as a Carbon Source The optimization of the fermentor-scale plant cell culture requires knowledge of substrate requirements and utilization rates. In general, sucrose has been used as a carbon source for plant cell culture. Extracellular invertase hydrolyzes sucrose to monosaccharides, i.e., glucose and fructose. Invertase is excreted into the medium during plant cell growth and the monosaccharide concentrations change with cultivation time. Therefore, control of monosaccharide concentrations is very difficult. Taking this problem into account, as far as plant cells could consume the monosaccharides, these saccharides should be a more interesting carbon source and the control of carbon source would become easier. To obtain high density culture of plant cells, the culture condition should be maintained at the optimum level. From the view point of process development, characterizing carbon utilization of hairy root in liquid culture to evaluate metabolic energy costs is interested. Since the proliferation of hairy roots occurs only at the apical meristem and is expressed by a linear growth, the balance equation for hairy growth should be used to determine the values of maintenance coefficient and cell yield under carbon-limited culture. Carrot hairy root was cultivated in a batch system with a turbine-blade reactor (initial sucrose concentration: 20 g/l). The final cell mass reached 10 g-dry weight/l at 31 days. On the basis of the concentrations of the residual key components in the batch culture broth after 31 days, the cell yield for each component was determined

82

N. Uozumi

(Table 1). The maximum attainable cell mass (Xmax) could be obtained theoretically if the other nutrients were sufficiently supplied. As shown in Table 1, the values for the inorganic ion components, except for Ca2+ and Cu2+ ions, were from 10 to 14 g/l, while the values for Ca2+ and Cu2+ were much larger than 14 g/l. This shows that the proliferation of the roots ceased owing to the depletion of not only the carbon source but also of most of the inorganic ion components. For the fed-batch culture, the composition of a suitable medium for MS nutrients feeding was determined not to accumulate some components of the feeding medium in the medium and inhibit cell growth (Table 2). The conductivity of the culture medium corresponded to the concentration of the total components. The hairy root growth can be estimated from the de-

Table 1 Cell yield of carrot hairy root for main components in MS medium Components

Residual amounts of component (mg/l)

Content in MS medium (mg/l)

Cell yield (g-dry cell/g)

Xmaxa (g-dry cell/l)

K+ Mg2+ Ca2+ Mn2+ Cu2+ Zn2+ NO3– NH4+ PO43– Sucrose

134 7.0 81 2.2 3.6¥10–3 0.4 61 13 15 0

780 36 120 7.2 6.4¥10–3 2.0 2200 300 120 20,000

15.4 345 256 1961 3.57¥106 6250 45.5 41.7 100 0.5

12 12 31 14 23 13 10 12 12 10

a

Maximum cell density estimated from the content in MS medium and the cell yield.

Table 2 Composition of medium for fed-batch culture for carrot and Ajuga hairy root Components

Medium used in the fed-batch culture

MS medium

Solution A

(g/l)

(g/l)

Sucrose NH4NO3 KNO3 CaCl2 2H2O MgSO4 7H2O KH2PO4

20 1.8 (1.1) 2.0 0.15 0.27 0.2 (0.3)

20 1.65 1.9 0.44 0.37 0.17

Solution B

Other inorganic compositions in MS medium

The concentration of compositions for Ajuga hairy root is designated in parentheses.

Large-Scale Production of Hairy Root

83

crease of the conductivity, and it was also observed that the correlation between two values remained constant through culture.When the conductivity decreased to 15% from the initial level, solutions A and B were fed to keep the conductivity constant. Final cell mass, 17 g-dry weight/l of hairy root was obtained. The concentrations of the main components including sucrose were kept at almost a constant level. However, glucose and fructose accumulated up to 30 g/l, which may have adversely influenced cell growth [7, 8]. 2.4 Fed-Batch Culture Using Monosaccharide Sucrose has been used as a carbon source for plant cell culture. Extracellular invertase hydrolyzes sucrose to monosaccharides, i.e., glucose and fructose. Invertase is excreted into the medium during plant cell growth and the monosaccharide concentrations change with culture time and thus control of monosaccharide concentrations is very difficult. To date, there is limited information about sucrose uptake directly into plant cells. To obtain information on the monosaccharide utilization for hairy root, batch cultures supplemented with various single monosaccharides were carried out in shake flasks. At 33 days, 6.2 g-dry weight/l of the hairy root grown on sucrose was obtained. Preliminary analyses on utilization of various monosaccharides revealed that either glucose or fructose was a suitable carbon source. In particular, supplementation of medium with fructose instead of sucrose or glucose resulted in a higher final cell mass, 6.8 g-dry weight/l. The fed-batch culture of carrot hairy root was performed using fructose as a sole carbon source. The inorganic ion concentrations in solutions A and B were the same, except that sucrose was substituted by a concentrated fructose. When the medium conductivity decreased to approximately 15% from the initial level, solutions A without sucrose and B, together with a concentrated solution of fructose were added to the bioreactor. During this period, the concentrations of the major nutrients including fructose were kept constant. As a consequence, the final cell mass at day 38 reached 30.1 g-dry weight (approximately 436 g-fresh weight) per liter medium. The roots grown by fed-batch operation using fructose in the reactor at day 38 were condensed extensively (Fig. 4). They bound tightly to the stainless mesh cage which provided a support matrix for the roots and invaded into the central space encompassing the mesh and a bottom section of the impeller blades. Since the turbine-blade reactor involved a dead space where the blades rotated, the cell density was evaluated to be 35.4 g-dry weight/l based on the working volume. This is a marked improvement compared with the previous cultures. Air was supplied and the dissolved oxygen level in culture medium could be monitored accurately up to 20 days. After that period, a mixture of air and oxygen was supplied. When cell density increased above 20 g/l, oscillations of dissolved oxygen concentration were observed. In fact, we observed that the bubbles were extensively trapped at the interspace of the root materials in the bioreactor. These interspaces

84

N. Uozumi

Fig. 4 Hairy root and condensed carrot hairy root materials cultured in turbine-blade reactor by fed-batch culture using fructose as a carbon source

would become anaerobic and the cell growth decreased even if concentration of the main components were kept constant. The productivity of an insect-molting hormone, 20-hydroxyecdysone, synthesized from hairy root was about four times higher than one from the original plant Ajuga reptans [6]. The difference in cell growth and secondary metabolite, 20-hydroxyecdysone production between sucrose and glucose was not found in shake flasks. The growth rate and final cell mass supplemented with glucose and galactose were comparable with those of sucrose. Fructose was also utilized for growth, whereas neither lactose nor mannitol were consumed by Ajuga hairy root. There was no significant difference in 20-hydroxyecdysone

Large-Scale Production of Hairy Root

85

content in the hairy root among them. Hence, glucose was used in the subsequent experiments, which is less expensive than sucrose or galactose. In the fed-batch culture with Ajuga hairy root, concentrations of glucose, phosphate, ammonium and nitrate ions were almost kept at a constant levels. The final cell mass increased markedly, 27.2 kg-dry cell/lat day 39. Ajuga is well-known to produce the insect molting hormone, 20-hydroxyecdysone. This fed-batch culture exhibited approximately twofold increase in 20-hydroxyecdysone production, 44 mg/l of 20-hydroxyecdysone, compared with batch culture in a shake flask with sucrose. 2.5 Yield Coefficient for Hairy Root Biomass and Maintenance Energy To control the concentration of the components in the medium is vital for a fedbatch culture.As inorganic ions in the medium were taken up by the cells, there was a consequent gain in cell mass. The cell mass obtained was accurately proportional to the cell yield coefficient (YX/S) of each inorganic ion: YX/S = –dX/dS

(2)

where s and X are substrate carbon concentration and cell concentration, respectively. The YX/S values for PO43–, NH4+, and NO3– were 33, 32, and 4 g-dry cell/g, respectively. To maintain inorganic ions in the medium constant, fivefold concentrated modified MS medium based on the above values was fed as the hairy roots grew. On the other hand, the carbon source was not supplied in the same manner, since carbon is utilized to gain biomass and to maintain cell viability. Carbon dioxide released from the fermentor during the culture is produced through cellular respiration and the metabolic pathway to yield the maintenance energy. According to the following balance equation, metabolic energy costs can be assessed from carbon utilized for growth (cell yield, YX/S) and carbon used to maintain existing biomass (maintenance coefficient, m), which are deduced from the following relationship, where t=time: q = –(dS/dt)/X = (1/X)(dX/dt)/YX/S + m = m/YX/S + m

(3)

The specific growth rate (m) and specific net uptake rate of carbon substrate into the cells (q) can be estimated under condition where the growth rate is carbon limited. Continuous culture can offer the proper conditions for estimating the growth yield and maintenance coefficient of callus growth. However, hairy roots cannot be carried out in the continuous culture due to its morphology, and the linear growth phase is predominated as described above. For evaluation of kinetic parameters of hairy roots in linear growth phase, the relationship between hairy root growth and carbon source consumption should be rewritten in the following balance equation: –dS/dt = (dX/dt)YX/S + mX

(4)

86

N. Uozumi

The results were plotted from the experimental data using monosaccharides on Ajuga hairy root and on carrot hairy root. Since the growth pattern of hairy roots provided a linear growth phase (dX/dt=const.) after the exponential phase, Eq. (4) was applicable to the estimation of the cell yield and maintenance coefficient of hairy root. The cell yield and maintenance coefficient were calculated from the intercept and the slope, respectively. As shown in Table 3, the two values of Ajuga and carrot hairy roots were calculated from Eq. (4). The fed-batch culture could be employed to estimate the growth yield and maintenance coefficient of hairy root on the basis of the following reasons. (1) A linear growth phase is dominant over the culture. (2) For microorganism or callus, the growth yield and the maintenance coefficient can be evaluated from the data of continuous culture in which nutrients levels including carbon source are constant. Therefore, they should be also constant for hairy root culture, which can be achieved by adding the concentrated medium through the measurement of medium conductivity. (3) In the case of sucrose, glucose and fructose are accumulated in the medium, which makes it difficult to measure the concentration of carbon source precisely. Glucose or fructose was more adequate than sucrose for evaluating YX/S and m for the carbon utilization of the plant cell. (4) There is little evidence concerning metabolites inhibitive to cell growth in plant cell culture. The methodology using a fed-batch culture meets the requirements for estimating the kinetics parameters sufficiently. The estimated results are summarized and compared to values obtained from the literatures in Table 3. Ajuga hairy root and carrot hairy root had YX/S Table 3 Comparison of biomass yield from saccharide and maintenance coefficients for saccharide with published values Organisms

Substrate

Cell yield, YX/S [g/g]

Maintenance coefficient, m [g/g · d]

Glucose Fructose Sucrose Sucrose Sucrose Glucose Lactose

0.77 0.60 0.71 0.57 0.60 0.52 0.77

0.105 0.085 0.074 0.012 0.086 0.106 0.113

Glucose Glucose Glucose Methane Glucose

0.14 0.40 – 0.56 0.51

0.864 2.256 0.528 – –

Plant Ajuga hairy root Carrot hairy root Eschscholtzia californica callus Apple callus Nicotiana tabacum callus Nicotiana tabacum callus Medicago sativa callus Microorganism Saccharomyces cerevisiae Aerobacter cloacae Penicillium chrysogenum Pseudomonas methanica Candida utilis

Large-Scale Production of Hairy Root

87

values comparable with those of callus, whereas the cell yields of Ajuga and carrot hairy roots were slightly higher than those of microorganisms. These findings indicate that the energy produced from the carbon source in plant cells is converted to biomass more effectively than in microorganisms, despite the low plant proliferation.Although there is no significant difference in the maintenance energy requirement between the hairy roots and various calli, the requirements of microorganisms are some fivefold those of plant cells. This may be an indication of the difference in metabolism as a result of a different cellular structure and a different environment. In general, plant cells do not have mobility, and only the apical meristem of a hairy root retains high elongation activity; thus the loss of the energy obtained from the carbon source in a plant might be smaller than that of microorganisms. In plant culture, evaluation of YX/S and m is required to assess metabolic energy costs associated with cell age and environmental conditions. The information on these coefficients allows us to predict the optimum conditions for plant cell culture and to control automatically the environmental condition for the cell growth by feeding the inorganic components and monosaccharide [9].

3 Kinetics of Hairy Root and Simulation Two-step culture is one of the most appropriate strategies to pursue both high cell density and high secondary metabolite content in the optimized medium, since optimal growth and optimal secondary metabolite production cannot be achieved simultaneously during the conventional culture. The construct of the specialized medium and control strategies are required for two-step culture [10]. To separate the culture of hairy roots into growth phase and production phase, ammonium ion and nitrate ion was used as nitrogen source in the corresponding culture [11]. Ajuga hairy root that produces an insect-molting hormone, 20-hydroxyecdyson (20-HE) was used as a model plant cell. The auxin supplementation and intracellular phosphate control were tested for an increase of cell growth rate and 20-HE production. The fed-batch culture combined with auxin supplementation and the simulation data for intracellular phosphate content eventually gave the effective 20-HE production. When plant cell proliferates well, the secondary metabolites production and accumulation are normally low. Once the cell growth rate comes to be lower and cell expansion begins, production and accumulation of the secondary metabolite begins to occur at an appropriate medium. The PO43– depletion might imply the cease of cell growth since phosphorus is essential component in the synthesis of DNA, ATP, and other important biomolecules involving in cell growth. Other chemicals should be looked for to understand the mechanism of the 20-HE biosynthesis. For actual culture, control of the intracellular

88

N. Uozumi

Fig. 5 Fed-batch culture of Ajuga hairy root.All points shown represent experimental data. PO43– concentration in the medium (filled triangle); dry cell mass (filled circles); glucose concentration in the medium (open circles); intracellular PO43– content (open triangles); 20-HE content (filled squares).All solid lines depict the calculated results based on the kinetic model

Large-Scale Production of Hairy Root

89

phosphate ion is an inexpensive, convenient, and harmless strategy to shift from the cell growth phase to the secondary production phase. To date, control of PO43– in the culture has been reported. For plant culture, Taya et al. conducted survey of the components that render the pigment accumulation from the MS medium and found that phosphate depletion enhanced the pigment accumulation [12]. The pigment production was enhanced by using the medium without phosphate in batch culture. It is generally recognized that the cells store phosphorus in vacuoles or other locations in the form of polyphosphate, phytic acid and so on and that the availability of stored phosphorus compounds supports the growth of hairy roots even under phosphate-deficient medium. The secondary metabolite production is inversely related to the intracellular phosphate content. Lower concentration of PO 43– enhanced 20-HE content in the root. The auxin promotes horseradish hairy root growth due to the increase in the number of root apical meristems [13]. In Ajuga hairy root, NAA, IBA, and IAA provided higher growth rate. The growth rate of hairy roots depends not only on the number of root apical meristems, but also the root elongation rate. The number of root apical meristems increased with the increase of IAA taken by the root. An equation was applied for expressing the number of root apical meristems and for representing kinetic growth inhibition regarding IAA supplementation. In addition, auxin supplementation impaired the 20-HE production to some. To take advantage of auxin effect, auxin should be supplemented at the beginning of the culture, and the auxin should be depleted before the 20-HE production phase. In horseradish hairy root, auxin treatment induced aggregated form whose morphology was changed to adventitious roots by removal of auxin from the medium [14–17]. Fed-batch culture is one of effective culture to produce the secondary metabolites 20-HE in Ajuga hairy root. The intracellular PO43– content is the key point to increase the 20-HE production. However, the intracellular PO43– content cannot be monitored on line and it is time-consuming to measure the intracellular PO43– content in the root during the culture. Therefore, we developed the kinetic model on the basis of the previous reports and applied it to the fedbatch culture [18]. Since horseradish was used as hairy root in the previous reports, several parameters were added and modified for Ajuga hairy root. The auxin supplementation at the beginning of the culture can shorten the culture duration, phosphate is the important factor to determine the hairy root growth phase or 20-HE production phase, and the simulation is powerful to keep intracellular phosphate concentration below the level [19]. There are some differences between the experimental data and simulation line in this report. For example, the final 20-HE content was larger than the simulation in Fig. 5. It might be mainly due to the different culture condition (e.g., flask and fermentor) although it is very difficult for the simulation to match the experimental data perfectly. Nevertheless, the results shown in Fig. 4 strongly suggests the simulation on the basis of the kinetic model is useful for secondary metabolites production from hairy root culture in flask and in fermentor scale.

90

N. Uozumi

4 Release and Recovery of Enzymes and Secondary Metabolites from Hairy Root 4.1 Release of Biochemical Products from Hairy Root To develop an effective production system, alternative means of stimulating production and excretion of a useful enzyme and secondary compounds have been studied from hairy roots. In particular, a major limitation in continuous production of proteins and secondary metabolites from plant cells is that desirable products are stored intracellularly. Some excretion methods might have significant potential to improve the feasibility of processes for producing favorable products from plant cell tissue culture. These involve pH cycling, use of permeabilizing agents, ultrasound, and dissolved oxygen control. Such methods enhanced product release although the permeabilization mechanism was unknown, and they are not generally used for proteins and secondary products in diverse plant cells. The two successful studies on the release of plant-derived products from hairy roots are discussed. One is the enhancement of the production, excretion into the medium and absorption to resins of horseradish peroxidase (EC 1.11.1.7.), which is present in higher plants, and widely used in the colorimetric analysis of biological materials. The other is the release and recovery of pigments from red beets hairy roots by controlling oxygen concentration in the culture medium. 4.1.1 Release and Enhancement of Peroxidase Production and Excretion from Horseradish Hairy Root and Medium Control Polypeptone enhanced intracellular peroxidase activity was observed, but was not effective for excretion [20]. Taking the polypeptone content into consideration, the enhanced production was attributed to NaCl present in the polypeptone. NaCl was found to induce peroxidase excretion into the medium, rather than to activate peroxidase synthesis by expression of peroxidase gene. In contrast, light enhanced peroxidase content in the cell, but not peroxidase excretion [21]. The combined use of illumination with light, NaCl addition, and the adsorption of excreted peroxidase leads to peroxidase overproduction. At higher NaCl concentrations (50–500 mmol/l), the light effect on cell growth was not significant for the final cell mass. Light-grown culture increased approximately twofold peroxidase production compared with that from dark-grown culture. In the light condition, peroxidase activity of the culture containing NaCl increased drastically compared with culture without NaCl; approximately 2.4-fold for the culture medium containing 50 mmol/l NaCl and 3.4-fold for the culture medium containing 100 mmol/l NaCl at day 34.

Large-Scale Production of Hairy Root

91

To test whether the osmotic effect of the ion supplementation in the medium had any influence in peroxidase excretion, various concentrations of mannitol were supplemented to the medium. It is known that mannitol is not metabolized and does not support plant tissue growth. Osmotic pressures of mannitol at 99 mmol/l and 198 mmol/l corresponded to those of NaCl at 50 mmol/l and 100 mmol/l, respectively. Cell growth pattern of the culture supplemented with mannitol was similar to that with NaCl.At day 30, extracellular peroxidase activity of the root treated with mannitol was lower than that of the non-treated root. This indicated that the osmotic pressure has no effect on peroxidase excretion from horseradish hairy root. 4.1.2 Cultivation in Combination with Adsorption To prevent peroxidase degradation during the culture, excreted peroxidase should be recovered from the medium periodically. Recovering of peroxidase from the medium will also enhance the enzyme excretion by maintaining its concentration in the medium at a low level. The cultivation in combination with the resin adsorption was carried out at each several days. The culture was performed with MS medium supplemented with 2% sucrose and 50 mmol/l NaCl in light condition.When peroxidase activity in the medium was kept below the maximum level of the batch culture, the productivity after 20 days retained above 2.8 U/ml-day. Thus, the stepwise adsorption gave rise to a significant increase in the total amount of peroxidase produced. In addition, the adsorption caused slightly higher final cell mass. Final total peroxidase activity reached approximately 70 U/ml level. Enhanced enzyme production was observed during cultivation in combination with the resin adsorption of peroxidase. The adsorption seems to be useful to retain peroxidase production and to prevent peroxidase degradation. Metabolic by-products which inhibit growth of root cells may also be adsorbed, since the final cell mass with the adsorption operations was 1.3-fold higher than that in the batch culture. Controlling peroxidase concentration at a low level in the medium by the adsorption led to the stimulation of enzyme excretion, thereby providing a large concentration gradient of peroxidase across the cellular membrane. It has to be attempted to determine favorable conditions to elute the peroxidase from the resin. The best product recovery was 85% when 80% cold acetone as elution agent was used [21]. 4.1.3 Repeated Batch Culture of Horseradish Hairy Root It is observed that the root at the stationary phase retained the cell viability and had the potential for producing peroxidase. To facilitate the peroxidase production from the root, repeated-batch culture was carried out in a shake flask using the medium supplemented with 50 mmol/l CaCl2 [22]. CaCl2 was found to show the similar effect to NaCl for excretion of peroxidase. Fresh medium

92

N. Uozumi

with or without 50 mmol/l CaCl2 was substituted for the culture medium when the decrease in conductivity of the culture medium ceased. The final cell mass at 149 d reached 49.4 g-dry cell/l in medium supplemented with 50 mmol/l CaCl2and 66.6 g-dry cell/l in the medium without CaCl2 respectively. Green portions (shoot formation) on the root surface appeared after four cycles and these portions extended to the gas space. When the preiodical replacement of the medium (22–28 days/batch) was performed, peroxidase activity in medium reached a maximum (80 U/ml) at the third cycle and then was kept constant at the 60 U/ml level. The total extracellular peroxidase activity in the culture for 149 days reached 12,840 U, whose activity was more than 12-fold as high as that of the batch culture supplemented with CaCl2 for 40 days. Peroxidase productivity in repeated batch culture reached 0.36 U/ml/day, while the productivity of the non-treated root was 0.16 U/ml/day. This repeated batch culture allowed a continuous retention of cell viability and production of a large amount of the peroxidase in the medium [22]. 4.2 Release of Pigments by Controlling the Airation and Development of Repeated Recovery System Hairy root of red beet produces pigments, which are usually accumulated in the vacuole. Taya et al. found that the red beet hairy root released the pigments significantly into the medium during the cessation culture shaking. The growth of the hairy roots were lower, but the hairy roots can be proliferated by resumption of the shaking after the cells were subjected to shaking cessation. Figure 6 shows the release of pigments from hairy root in the medium. On the basis of their analysis of the release, they indicated that pigments release from the hairy

Fig. 6 Release of pigments from red beet hairy root. Before (left) and after (right) shaking cessation for 48 h

Large-Scale Production of Hairy Root

93

root was considered to occur when the cells were exposed to oxygen limitation or starvation under a very low dissolved oxygen level that may lead to the partial disruption or relaxation of cell membranes. Since the root tip can survive during the cessation of air supplement, the repeated recovery of pigments released from the hairy roots can be performed [23–30]. To absorb pigments, a column with a hydrophobic resin were incorporated into the culture system with a bioreactor. In the culture with an oxygen starvation time of 16 h, an extracellular production rate of pigments of 11.3 ¥ 10–6 g/l h was obtained on the average. The long-term culture of hairy root was shown with the repeated operations of cell growth, pigment release and recovery [31].

5 Use of Photoautotrophic Hairy Roots for the Culture Photoautotrophic cell lines of plants, that are able to grow with CO2 fixation through photosynthetic reactions in the absence of any organic carbon sources, have been established accompanying the reinforcement of their photosynthetic potentials of in vitro cultured plant cells are known to be affected by environment factors such as sugar concentration in a medium. CO2 concentration in gas phase and light quantity furnished to the cells. Light irradiation can induce the formation of chlorophyll, which shows green in color. The pak-bung and horseradish green hairy roots have been applied to the culture [32–35]. The green hairy root cultured under the light exhibited increase in branching by light irradiation, resulting in higher growth rate of the root materials, and enhanced activities of a series of detoxification enzyme to active oxygens, such as superoxide disumutase and peroxidase. The ability of photosynthesis supplies carbon source to the cells. The calculated maintenance energy for the cells was comparable or slightly lower than that for the roots derived from sugar consumption. The green hairy roots have metabolite productivities which were distinct from their respective original white roots since the exposure of hairy roots to light leads to alternations in the biosynthetic potentials of hairy roots. The green hairy root is thus one of promising materials for expanding the availability of compounds produced by hairy root culture [36–38].

6 Production of Regenerated Plant from Hairy Root 6.1 Regeneration of Hairy Root The delivery system using artificial seed is of great value in the production of useful chemicals and nutrients. Artificial seeds for plant propagation has been investigated in somatic embryogenesis from callus. The concept will extend

94

N. Uozumi

Fig. 7 Three routes for hairy root regeneration. (1) Hairy root fragment excised by blade. (2) Adventitious shoot primordia formed in the dark. (3) Plantlets produced from hairy root under the light conditions

to the hairy root because the successful result regeneration of the whole plant from Ri-transformed cells has been obtained with potato, apple tree, horseradish, and Allocasuarium verticillata, etc. The micro-propagation system using these hairy roots promises practical application in cellular biology, agriculture and bioengineering fields. From the viewpoint of bioengineering, the process for artificial seed system using elite transformed roots has been required to establish on an industrial scale [39–42]. Here we focus on the development of micropropagation procedure of the hairy root by using artificial seed system from the standpoint of bioengineering. The adventitious shoot primordium on the horseradish hairy root by microscopic morphological observation and alteration to plantlet at the exposure to light is identified. The candidate hairy root cells included in beads were classified into three categories (Fig. 7): – The root fragment cultured under dark condition – The adventitious shoot primordium formed under dark condition – Plantlet developed from the hairy root which was formed after transferred

into light. The adventitious shoot primordium offers an advantage for relatively higher synchrony in development at exposure to light, compared with two other cells. It is necessary to increase the number of the adventitious shoot primordium formation in an artificial seed system. In this study, the regeneration frequency from horseradish hairy roots is evaluated by using excision and encapsulation combined with supplementation of growth regulators to develop the manipulation of hairy roots for an artificial seed and to estimate the efficiency of hairy roots as a seed system.Artificial seeds are expected to be a reliable delivery system for clonal propagation of elite plants. The delivery system has the poten-

Large-Scale Production of Hairy Root

95

tial for genetic uniformity, high yield and low cost of production. The plant cells for artificial seeds require a good ability of regeneration and a high resistance against disease and mutation. From some species of hairy roots, it has also been observed that the regeneration occurs in the light. Successful regeneration of the whole plant from Ri-transformed cells has been obtained with various species. Genetic improvement of plants through conventional breeding and selection methods takes a long period. Within an acceptable time period, new gene transfer technologies offer the opportunity to produce plants easily with desirable traits such as disease or herbicide resistance. Successful reports on elite transgenic plant cells and their advantageous properties stimulated interest in developing regeneration and delivery system of hairy roots [43–47]. The proper system of plant regeneration is necessary to gain the transgenic plant from the hairy root efficiently. In particular, the production process should be constructed and improved to realize these plants to the artificial seed in industrial scale [48–50]. 6.2 The Use of Root Fragments for Encapsulation This method is the easiest procedure among the three routes for the procedure of encapsulated beads including the hairy root [51]. The portion of the root is strongly correlated with regeneration frequency.When the root was excised to various fragments in length from 1.0 to 10 mm and encapsulated with alginate, the shoot appeared somewhere after the root elongated at the apical meristem. In the case of the root fragment containing lateral root, shoot formation was observed at the center of the beads, and the shoot formation frequency was comparable with that of the apical meristem fragment. The shoot formation frequency of root fragments without root apical meristem and lateral root (intermediate portion) was significantly low. The minimum root length enabling shoot formation was evaluated, choosing the root fragment with apical meristem as the encapsulated root fragment. The shoot formation frequency increased with the increasing root length up to 5 mm. The fragment whose length was more than 5 mm was almost constant in the frequency, indicating that the fragment of 5 mm length was suitable for encapsulation. In encapsulation, carbohydrate should be contained only in the beads.Various concentrations of sucrose in the beads placed on the plastic sheet were tested without sucrose supplementation in agar. High shoot formation frequency of the root was observed when sucrose concentration in the beads was above 3%. Once leaf emerges from the root, the energy of differentiation and proliferation could be supplied by the photosynthesis. Thus supplementation with carbohydrate into the beads is necessary to grow plantlet from root fragment. Effect of segmentation of hairy roots on number of lateral root emergence and growth were evaluated [52, 53]. Auxin like 1 naphthaleneacetic acid (NAA) or indole-3-butyric acid (IBA) stimulated emergence of root apical meristem and lateral root. To harvest a large

96

N. Uozumi

amount of root fragments which have root apical meristem or lateral root portions from the whole root culture, the root can be treated with various concentrations of auxin and the randomly picked root fragments were encapsulated and transferred to MS medium in the light. The highest shoot formation frequency was obtained when the root was placed on the medium containing 0.1 mg/l of NAA. To explain the effect of auxin-treatment on shoot formation, the frequency of NAA-treated root with or without lateral root was also examined. There was apparent correlation between lateral root emergence and shoot formation frequency. The higher frequency of shoot formation on the 0.1–1 mg/l of NAA and 1 mg/l of IBA-treated horseradish hairy root was due to a large number of lateral root induced by NAA or IBA treatment.Healthy plantlets were also grown in the preculture with 0.1 mg/l and 1 mg/l of NAA, which appeared larger than non-treated plantlet. In contrast, abnormal morphologies of the plantlets were observed in the liquid preculture with 5 mg/l of NAA. From the plantlet development frequency and morphological observation, the optimum NAA concentration in the preculture was thereby determined to be at 0.1 mg/l. After encapsulation, NAA inhibited plantlet formation extensively. Hence, plantlet development required removal of NAA from the medium in the light after the preculture with NAA in the dark for the lateral root emergence. 6.3 Encapsulation of Adventitious Shoot Primordia Adventitious primordia are formed from horseradish hairy root and Ajuga hairy root on the root in dark culture [54, 55]. The adventitious primordia turn to green and to grow up to plantlet. After being transferred to light, the neoplasm from horseradish hairy roots or Ajuga hairy roots turned green, and shoots emerged a few days later. From this observation, the neoplasm was considered to be adventitious shoot primordia. At the beginning of culture, adventitious shoot primordia developed only in the center portions of whole roots in the flask, which, being apart from the root apical meristem, are composed of older cells. Over the course of the culture, primordia also emerged close to the root apical meristem. Although some of the primordia on horseradish hairy root developed into etiolated plantlets in the dark, most of them remained 0.5–3 mm in size during the subculture period (three weeks) in the dark. Excision of the root with adventitious shoot primordia facilitated the handling of the primordia for encapsulation.After adventitious shoot primordia formed in the dark were excised, encapsulated and put onto agar-medium in the light, plantlets arose out of the beads after a few weeks. Small primordia clusters on basal portions of lateral roots could be observed by scanning electron microscopical observations.An adventitious shoot primordia consisted of two leaves. The number of primordia increased and individual primordia grew larger as the root culture proceeded. The maximum plantlet development frequency was reached in primordia formed in the three-week culture. After six weeks,

Large-Scale Production of Hairy Root

97

approximately half of the adventitious shoot primordia of horseradish hairy root formed multiple shoots which are difficult to grow into healthy plants, while the number of adventitious shoot primordia and the plantlet development frequency in the 8-day and 18-day cultures were low. In the shorter cultures primordia were generally too premature to enable plantlets to develop, whereas the six-week culture was too long and the plantlet development frequency decreased. The primordia from the four-week culture were most suitable for use in artificial seeds, both in terms of number and plantlet development frequency, and hence these were used in the subsequent experiments. As root culture proceeded, the number of primordia increased and individual primordia grew larger. The number increased with the culture time up to approximately 20 days, and then reached a plateau. To test the effect of auxin and cytokinin on primordia formation from horseradish hairy root, roots were cultured in MS medium containing varying concentrations of NAA (0–5 mg/l) or benzyladenine (BA) (0–5 mg/l) in the dark. NAA supplementation stimulated the formation of callus which had low plantlet development capability. Neither auxin nor cytokinin in the dark improved the number of the adventitious shoot primordia and their regeneration frequency although cytokinin enlarged adventitious shoot primordia. In the case of embryogenesis formed through somatic adventitious shoots, root growth is sometimes carried out in vitro under a high auxin concentration. The hairy root adventitious shoot primordia overcome the difficulties of root emergence and elongation in a natural manner since the root elongation occurred after encapsulation.An important advantage of root regeneration from hairy roots is thus found when compared with the use of adventitious shoots derived from somatic embryos, in which the root elongation does not occur. 6.4 Encapsulation of Plantlet Regenerated from Hairy Root 6.4.1 Mass Production of Hairy Root Fragments Using the Blender Root fragmentation before plantlet formation is required to obtain plantlets suitable for use as artificial seeds. Mechanical fragmentation is preferable for large-scale production of artificial seeds [56, 57]. Fragmentation was carried out using a commercial blender with blades (Fig. 7). Horseradish hairy roots cultured for 35 days without addition of plant growth regulator were fragmented in the blender for 10, 20, 30, or 60 s. Approximately 54% of the root fragments on a fresh weight basis was recovered in the case of fragmentation for 30 s, which is the best frequency among different treatment. After fragmentation, the root fragments were transferred into plantlet formation medium without plant growth regulator. Plantlets were formed at day 4, after which secondary roots began to elongate from the plantlets. The number of plantlets formed from the root fragments cut manually with a razor was 1.5-fold as high as that formed from

98

N. Uozumi

the roots which were fragmented mechanically in the blender for 30 s. However, it took approximately 10 min to cut the roots manually to achieve a degree of fragmentation comparable to that achieved by fragmentation in the blender for 30 s. Although use of the blender for fragmentation resulted in a decrease of plantlet formation frequency as compared with that in the case of manual fragmentation, use of a blender is advantageous for rapid and large-scale root fragmentation. 6.4.2 Auxin Supplementation Stimulates Hairy Root Growth Stage on Plantlet Formation Auxin supplementation stimulates the emergence of root apical meristems in horseradish hairy roots, and led to an increase in growth rate as the above section [56, 57]. In addition, NAA supplementation led to an increase in plantlet formation frequency of horseradish hairy roots. After the hairy roots were cultured until the stationary phase in MS medium supplemented with various concentrations of NAA, the hairy roots fragmented by the blender for 30 s were cultured in the medium without plant growth regulator for 10 days. NAA supplementation (0.1 mg/l or 1.0 mg/l) led to an increase not only in the root growth rate but also in plantlet formation frequency compared with those of non-treated roots. The frequency of plantlet formation from NAA (1.0 mg/l)treated roots was increased about 1.2-fold as compared with that of non-treated roots. On the other hand, supplementation with 5 mg/l of NAA led to a decrease in plantlet formation frequency. The plantlet productivity of NAA (1.0 mg/l)treated roots was 79 l–1 day–1 which was 1.8-fold that of non-treated roots. Horseradish hairy roots should be cultured in the medium supplemented with 1.0 mg/l of NAA for 20 days at the hairy root growth stage. 6.4.3 Cytokinin Supplementation Stimulates on Plantlet Formation The effects of kinetin supplementation at the plantlet formation stage on plantlet formation were tested in order to stimulate plantlet formation from horseradish hairy roots. Kinetin supplementation (0.01 to 1.0 mg/l) to root fragment culture led to an increase in the number of plantlets as compared to that of root fragments cultured without kinetin supplementation. The highest plantlet formation frequency was achieved in the medium supplemented with 0.1 mg/l of kinetin. On the other hand, supplementation of more than 5 mg/l of kinetin resulted in inhibition of plantlet formation and supplementation of 10 mg/l of kinetin caused callus formation. Kinetin supplementation also resulted in an increase in plantlet size. The optimal plantlet size for encapsulation is considered to be 2–4 mm, and control of plantlet size by adjusting the kinetin concentration is necessary for efficient production of artificial seeds. Moreover, since plantlets with multiple shoots do not develop into healthy

Large-Scale Production of Hairy Root

99

Fig. 8 Production of plantlets derived from horseradish hairy roots fragmented in a blender

plants, plantlets with single shoots are suitable for production of artificial seeds. Supplementation of 0.1 mg/l of kinetin resulted in the largest number of plantlets 2–4 mm in size, and 91% of plantlets (2–4 mm) had single shoots. Morphological changes were observed during plantlet development stage from the root fragment to plantlet in the case of supplementation with 0.1 mg/l of kinetin. After four days of root fragmentation, development of two leaves from bud primordium was observed and chlorophyll was detected. At six days of culture, more extensive leaf development from bud primordium was observed [56, 57]. The above results summarized the procedure of the hairy roots should be cultured with 1.0 mg/l of NAA before root fragmentation, and the fragmented roots should be grown in the medium supplemented with 0.1 mg/l of kinetin to achieve optimal plantlet formation frequency. Fragmentation for 30 s resulted in formation of the highest number of plantlets 2–4 mm in size which had a single shoot, compared with that for 10, 20, or 60 s (Fig. 8). 6.4.4 Improvement of Frequency of Plantlet Development Through Plantlet Dehydration At the plantlet development stage, roots emerged, and two new leaves appeared. Frequency of plantlet development from encapsulated plantlets was only 10% at 15 days of culture and 58% at 45 days. In contrast, non-encapsulated plantlets developed into healthy plants at high frequency (92% at 30 days). These results indicated that encapsulation resulted in a decrease in plantlet development frequency.When the plants were observed under a microscope, leaves were found to be water-soaked, translucent and glassy. These abnormal morphologies are

100

N. Uozumi

characteristic of hyperhydroricity, and the inhibition of plantlet development from encapsulated plantlets was due to prevention of dehydration after encapsulation [56, 57]. Plantlets dehydrated by exposure to an air flow in a laminar flow cabinet were encapsulated and their plantlet development frequency was measured (rapid dehydration).After 3 h of dehydration in the air flow, the plantlet weight had decreased by 81% compared with the initial weight. Dehydration treatment improved plantlet development frequency in the case of a weight decrease to 35% or less of the initial weight. Dehydration treatment of longer than 2 h caused serious damage to the plantlets. The dehydration treatment was also performed for plantlets cultured in sealed Petri dishes under light conditions (slow dehydration). When plantlets were subjected to the dehydration treatment of seven days (23% decrease in weight), plantlet development frequency at 15 days culture reached the highest level (98%) and was higher than that of the plantlets subjected to the dehydration treatment in the laminar flow cabinet. Slow dehydration was suitable for achieving high plantlet development frequency. 6.5 Application of Transgenic Plant to the Artificial Seed Procedure This section shows that the procedure of artificial seed production as proposed above can be applied for the hairy root, Ajuga hairy root in which GUS gene is expressed [58, 59]. Ajuga hairy root produced an insect-molting hormone, 20-hydroxyecdysone at about four times higher concentration than the one from the original plant Ajuga reputans. Ajuga hairy root growth rate was evaluated by supplementation of IBA, NAA, and IAA, compared to that without auxin. Branching appeared vigorously in the culture with IBA, NAA, and IAA. With regard to regeneration frequency, the culture with NAA supplementation reached the highest number of shoot formation frequency, approximately 1.2-fold of that of non-treated roots. The 0.1 mg/l NAA resulted in the highest number of plantlets. Supplementation of auxin, NAA (0.1 mg/l) in the hairy root growth culture promoted the cell growth rate and the regeneration frequency at the next plantlet formation stage. The mechanical fragmentation for 10 s in the commercial blender enabled to produce a large number of root fragments efficiently among 5 s, 10 s, 15s, and 20 s. From the fragment of Ajuga hairy root, shoot seldom came out in the MS medium without any growth regulator. To enhance shoot emergence from the fragment, benzyladenine (BA) was applied to the medium as cytokinine. BA exhibited the dramatic increase in the number of plantlets from the root fragments produced by a blender. BA supplementation of 10 mg/l reached the highest number of plantlets. The plantlets were encapsulated with Ca-alginate containing the modified MS medium and glucose, and placed on the agar. The plantlets could not grow up and turned to a blown condition, whereas non-encapsulated plantlets could grow up to plant. The dehydration should be examined for improvement of regeneration frequency. The different size of plantlets was transferred onto the

Large-Scale Production of Hairy Root

101

solid medium to allow to grow to the plant stage. Not only the size of plantlet but also root emergence was one of the key factor to grow to plant properly. As the length of plantlet increased, the number of plantlets with root increased. The plantlets larger than 4 mm had roots and they grew to plants with relatively highly efficiency (86%). The transgenic Ajuga hairy root was constructed by means of A. rhizogenes-mediated co-transformation of b-glucuronidase (GUS) gene under the control of light inducible promoter, ribulose-1, 5-bisphosphate carboxylase/oxygenase small submit, rbcS and root-inducing gene on the Ri plasmid. The Ajuga hairy roots that can grow on the solid medium containing kanamycin were sampled and it was confirmed that the hairy roots had produced what was hoped for. One of hairy roots that shows GUS activity was cultured in the liquid medium. The hairy root was applied to produce the regeneration procedure developed in this study. The final cell mass cultured with 0.1 mg/l NAA increased by 3.1-fold. The hairy root was subjected to fragmentation in a blender and the root fragments were transferred to the medium with 10 mg/l BA. The number of the root fragments treated with NAA was approximately 1.6-fold as high as that without NAA. The NAA treatment enhanced an increase in cell mass and number of plantlets. GUS activity was detected in leaf by histochemical reaction. There was no significant difference in regenerated plant appearance between Ajuga hairy root and GUS-positive. The GUS-transformed Ajuga hairy root was regenerated well through the regeneration procedure determined as in the above. Since many hairy roots that are superior to the original plant have been induced so far, the artificial production system is strongly recommended for mass propagation of hairy roots. The systematic regeneration procedure of Ajuga hairy root described here is shown, through artificial seed production, to enable industrial plant production. The transformation to hairy root by A. rhizogenes can generate the plant cell with increasing secondary metabolites content and with extremely stable regeneration ability after long-term subculture. The above study showed that hairy root has a potential to be used as artificial seed, likewise somatic embryo. The regeneration routes from hairy roots proposed here are threefold, and is proposed as a basic procedure for regeneration of hairy root. The utilization of root fragments is the most convenient for encapsulated hairy roots. The adventitious shoot primordia offer an advantage for a high potential to regeneration and relatively higher uniformity to change green under the light conditions, compared with two other cells. It is necessary to increase the number of the shoot formations in an artificial seed system. Although the production of plantlets from hairy roots is required for long-term culture, the plantlets have the highest regeneration frequency. The third procedure shown in Fig. 7 was the most efficient for obtaining regenerated plants although the procedure depends on the properties of the hairy roots. The procedure can be shown to be more sophisticated by extensive studies. Process engineering – including bioreactor design, high cell density culture, image analysis with computer device and robotic system – can improve the

102

N. Uozumi

artificial seed production process. These techniques will be applicable to hairy root regeneration systems and will reduce the time taken by several time-consuming and labor-intensive processes involved in conventional micropropagation. Moreover, morphological monitoring of the embryo and hairy root has been studied thought image analysis [60–62]. The development of human-aid techniques contributes to mass production of regenerated plantlets.

References 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18. 19. 20. 21. 22. 23. 24. 25. 26. 27. 28. 29. 30. 31. 32. 33. 34. 35. 36. 37.

Tepfer D (1984) Cell 37:959 Tepfer D (1990) Physiol Plant 79:140 Taya M, Yoyama A, Kondo O, Kobayashi T, Matsui C (1989) J Chem Eng Jpn 22:84 Taya M, Hegglin M, Prenosil JE, Bourne JR (1989) Enzyme Microb Technol 11:170 Kondo O, Honda H, Taya M, Kobayashi T (1989) Appl Microbiol Biotechnol 32:291 Uozumi N, Kobayashi T (1994) In: Ryu DDY, Furusaki S (eds) Advances in plant biotechnology. Elsevier, New York, p 307 Uozumi N, Kohketsu K, Kondo O, Honda H, Kobayashi T (1991) J Ferment Bioeng 72:457 Kino-oka M, Taya M, Tone S (1991) J Chem Eng Jpn 24:381 Uozumi N, Kohketsu K, Kobayashi T (1993) J Chem Technol Biotechnol 57:155 Taya M,Yoyama A, Nomura R, Kondo O, Matsui C, Kobayashi T (1989) J Ferment Bioeng 67:31 Kino-oka M, Taya M, Tone S (1993) J Chem Eng Jpn 26:578 Taya M, Yakura K, Kino-oka M, Tone S (1994) J Ferment Bioeng 77:215 Nakashimada Y, Uozumi N, Kobayashi T (1994) J Ferment Bioeng 77:178 Repunte VP, Kino-oka M, Taya M, Tone S (1993) J Ferment Bioeng 75:271 Repunte VP, Shimamura S, Taya M, Tone S (1994) J Chem Eng Jpn 27:523 Repunte VP, Shimamura S, Taya M, Tone S (1995) J Chem Eng Jpn 28:847 Repunte VP, Taya M, Tone S (1996) J Chem Eng Jpn 29:874 Taya M, Kino-oka M, Tone S, Kobayashi T (1989) J Chem Eng Jpn 22:698 Uozumi N, Makino S, Kobayashi T (1995) J Ferment Bioeng 80:362 Taya M,Yoyama A, Kondo O, Honda H, Kobayashi T (1989) Plant Tissue Culture Lett 6:159 Kato Y, Uozumi N, Kimura T, Honda H, Kobayashi T (1991) Plant Tissue Culture Lett 8:158 Uozumi U, Kato Y, Nakashimada Y, Kobayashi T (1992) Appl Microbiol Biotechnol 37:560 Taya M, Mine K, Kino-oka M, Tone S, Ichi T (1992) J Ferment Bioeng 73:31–36 Kino-oka M, Hongo Y, Taya M, Tone S (1992) J Chem Eng Jpn 25:490 Kino-oka M, Mine K, Taya M, Tone S, Ichi T (1994) J Ferment Bioeng 77:103 Kino-oka M, Taya M, Tone S (1995) J Chem Eng Jpn 28:772 Hitaka Y, Kino-oka M, Taya M, Tone S (1997) J Chem Eng Jpn 30:1070 Kino-oka M, Hitaka Y, Taya M, Tone S (1999) Chem Eng Sci 54:3179 Hitaka Y, Kino-oka M, Taya M, Tone S (1999) J Chem Eng Jpn 32:370 Hitaka Y, Takahashi Y, Kino-oka M, Taya M, Tone S (2000) Biochem Eng J 6:1 Kino-oka M, Taya M, Tone S, (1995) Plant Tissue Culture Lett 12:201 Kino-oka M, Nagatome H, Taya M, Tone S (1996) J Chem Eng Jpn 29:1050 Nagatome H, Tsutsumi M, Kino-oka M, Taya M (2000) J Biosci Bioeng 89:151 Taya M, Sato H, Kino-oka M, Tone S (1994) J Ferment Bioeng 78:42–48 Ninomiya K, Nagatome H, Kino-oka M, Taya M (2001) J Chem Eng Jpn 34:1396 Ninomiya K, Oogami Y, Kino-oka M, Taya M (2002) J Biosci Bioeng 93:505 Ninomiya K, Oogami Y, Kino-oka M, Taya M (2003) J Biosci Bioeng 95:264

Large-Scale Production of Hairy Root

103

38. Ninomiya K, Tsushima Y, Kino-oka M, Taya M (2003) J Chem Eng Japan 36(6):725–729 39. Kitto SL, Janick J (1985) J Am Soc Hort Sci 110:277 40. Redenbaugh K, Paasch BD, Nichol JW, Kossler ME,Viss PR, Walker KA (1986) Bio/Technology 4:797 41. Redenbaugh K, Slade D, Viss P, Fujii J (1987) HortScience 22:803 42. Redenbaugh K, Walker K (1990) In: Bhojwani SS (ed) Role of artificial seeds in alfalfa breeding. Plant Tissue Culture, Amsterdam, p 102 43. Ooms G, Karp A, Burrell MM, Twell D, Roberts J (1985) Theor Appl Genet 70:440 44. Noda T, Tanaka N, Mano Y, Nabeshima S, Ohkawa H, Matsui C (1987) Plant Cell Rep 6:283 45. Brillanceau MH, Tempé J (1986) Plant Cell Rep 8:63 46. Lambert C, Tepfer D (1991) Bio/Technology 9:80 47. Phelep M, Petit A, Martin L, Duhoux E, Tempé J (1991) Bio/Technology 9:461 48. Kobayashi T, Uozumi N (1992) In: Furusaki S, Endo I, Matsuno R (eds) Biochemical engineering for 2001. Springer, Berlin Heidelberg New York, p 270 49. Uozumi N, Kobayashi T (1995) In: Bajaj YPS (ed) Biotechnology in agriculture and forestry, vol 30. Springer, Berlin Heidelberg New York, p 170 50. Uozumi N, Kobayashi T (1997) In: Doran PM (ed) Hairy roots culture and applications. Harwood Academic Publishers, Amsterdam, p 113 51. Uozumi N, Nakashimada Y, Kato Y, Kobayashi T (1992) J Ferment Bioeng 74:21 52. Kino-oka M, Hitaka Y, Ninomiya K, Taya M, Tone S (1999) J Biosci Bioeng 88:690 53. Ninomiya K, Kino-oka M, Taya M, Tone S (2002) Biochem Eng J 10:73 54. Uozumi N, Asano Y, Kobayashi T (1994) Plant Cell Tissue Organ Culture 36:183 55. Repunte VP, Taya M, Tone S (1995) J Ferment Bioeng 79:83 56. Nakashimada Y, Uozumi N, Kobayashi T (1995) J Ferment Bioeng 79:458 57. Nakashimada Y, Uozumi N, Kobayashi T (1996) J Ferment Bioeng 81:87 58. Uozumi N, Ohtake Y, Nakashimada Y, Morikawa Y, Tanaka N, Kobayashi T (1996) J Ferment Bioeng 81:374 59. Tanaka N, Uozumi N, Kobayashi T (1999) In: Bajaj YPS (ed) Biotechnology in agriculture and forestry, vol 45. Springer, Berlin Heidelberg New York, p 30 60. Uozumi N,Yoshino T, Shiotani S, Suehara K, Arai F, Fukuda T, Kobayashi T (1993) J Ferment Bioeng 76:505 61. Takahashi Y, Hitaka Y, Kino-oka M, Taya M, Tone S (2001) Biochem Eng J 8:121 62. Ninomiya K, Tsushima Y, Kino-oka M, Taya M (2003) J Biosci Bioeng 96(1):98–101

Received: October 2003

Adv Biochem Engin/Biotechnol (2004) 91: 105– 134 DOI 10.1007/b94207 © Springer-Verlag Berlin Heidelberg 2004

Large-Scale Micropropagation System of Plant Cells Hiroyuki Honda 1 (✉) · Takeshi Kobayashi 2 1

2

Department of Biotechnology, School of Engineering, Nagoya University, Furo-cho, Chikusa-ku, Nagoya 464-8603, Japan [email protected] Department of Biological Chemistry, Chubu University, 1200 Matsumoto-cho, Kasugai, Aichi 487-8501, Japan [email protected]

1

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106

2

Bioreactor Design for Large-scale Micropropagation . . . . . . . . . . . . . 108

3 Immobilization Technology in Large-scale Micro-propagation . . . . . . . . 113 3.1 Plantlet Regeneration from Immobilized Calli in Gel Beads . . . . . . . . . . 113 3.2 Plantlet Regeneration from Immobilized Calli in Polyurethane Foam . . . . 118 . . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

123 123 125 126

4 4.1 4.2 4.3

Image Analysis System for Larger-scale Micropropagation Selection of Embryogenic Callus . . . . . . . . . . . . . . Selection of Embryos . . . . . . . . . . . . . . . . . . . . Estimation of Shoot Length . . . . . . . . . . . . . . . . .

5

Advanced Cultivation Method for Callus Regeneration – Viscous Additive Supplemented Culture . . . . . . . . . . . . . . . . . . . . 129

6

Conclusion and Perspectives

. . . . . . . . . . . . . . . . . . . . . . . . . . 132

References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 132

Abstract Plant micropropagation is an efficient method of propagating disease-free, genetically uniform and massive amounts of plants in vitro. The scale-up of the whole process for plant micropropagation should be established by an economically feasible technology for large-scale production of them in appropriate bioreactors. It is necessary to design suitable bioreactor configuration which can provide adequate mixing and mass transfer while minimizing the intensity of shear stress and hydrodynamic pressure. Automatic selection of embryogenic calli and regenerated plantlets using image analysis system should be associated with the system. The aim of this chapter is to identify the problems related to large-scale plant micropropagation via somatic embryogenesis, and to summarize the micropropagation technology and computer-aided image analysis.Viscous additive supplemented culture, which is including the successful results obtained by us for callus regeneration, is also introduced. Keywords Plant cell culture · Callus · Micro-propagation · Bioreactor · Image analysis

106

H. Honda · T. Kobayashi

1 Introduction Plant micropropagation is an efficient method of propagating disease-free, uniform and massive amounts of plants in vitro [1]. The micropropagation from cells can be achieved by direct organogenesis via shoots and somatic embryogenesis. Organogenesis via shoots is considered as one of the most widely used commercial method of regeneration. However, the procedures are labor-intensive and more specialized due to many steps on manual manipulations involved and the low multiplication rates. Gradual acclimatization of plants to the greenhouse and then to the field is also needed. These numerous steps are accompanied by extensive costs and the commercialization has been limited to highly valuable crops [2]. The prospect of micropropagation through somatic embryogenesis can provide a valuable alternative to the above propagation system [3–6]. This is amenable to a higher degree of automation and holds much promise for the mass propagation of plants at low cost because very large numbers of somatic embryos can be produced in a short period of time in a limited volume of medium. This would help towards making such a system economically feasible for many plant species. The potential applications of somatic embryogenesis in plant breeding depend to a large extent on whether embryos develop through callus or directly from explant cells. Genetically modified (GM) plants, especially crops, which are used to make GM food, have been developed as a very important technology for overcoming the food crisis in the beginning stage. Recently, various GM plants have been bred and marketed because there is some perceived advantage either to the producer or consumer of these foods; a lower price needed during the period from germination to harvest, greater benefit in terms of durability or nutritional value or both, and so on. GM crops currently on the market are mainly aimed at an increased level of crop protection through the introduction of resistance against plant diseases. The development of GM plants is considered to be extended to various plants and by incorporating various new genes, and the GM plant production will increase year by year. In such a situation, GM plant developers have focused on the development of a sterile plant not to allow GM plant producers such as a farmer to propagate the GM plant by themselves. This is an inevitable technology to protect economically the GM plant developer and promote the development. GM plant, which cannot be propagated via seed formation, brings another merit on a public acceptance of GM plant. If the heredity such as male-sterile is steadily transduced in the GM plant, the GM plant can form no pollens or form sterile pollens. The GM plant can not only be propagated via mating but can also fertilize other conventional plant in the wild. As a result, outcrossing – that is the movement of genes from GM plants into conventional crops or related species in the wild – will not occur. If, in addition, the germination of the seed from GM plant is also completely suppressed, the outcrossing by fertile pollen from wild plant is also prevented. Prevention of outcrossing is a very important issue for public acceptance of GM plant. The

Large-Scale Micropropagation System of Plant Cells

107

genetic heredity will be delightedly acceptable by GM developer as a significant merit for the GM plant. For these purposes, large-scale micropropagation of the plant via somatic embryogenesis will be strongly required. GM plants will distribute to the farmers as plantlets germinated from somatic embryos, not seed. Therefore, the establishment of somatic embryogenesis in various species of plants is very important and attractive. Once the availability of embryogenic cell is demonstrated, the production of somatic embryos on a large scale will be carried out and the scale-up of the whole process will be established by means of an economically feasible technology (Fig. 1). Large-scale production of somatic embryos in appropriate bioreactors is essential if micropropagation and artificial seed systems are to compete with natural seeds. It is known that plant cells are relatively large in size and are relatively weak to hydrodynamic or shear stress. When cells in suspension are subjected to moderate levels of hydrodynamic or shear stress, they will tend to deform or rupture causing cell death [7]. Therefore it is necessary to design suitable bioreactor configuration, which can provide adequate mixing, and mass transfer while minimizing the intensity of shear stress and hydrodynamic pressure. Immobilization technology applied widely in the production of secondary metabolites by plant cells has been developed successfully [8–10]. Immobilization not only promotes cell-to-cell contact which can enhance the accumulation of secondary metabolites [11], but also protects plant cells from hydrodynamic and shear forces to improve production of embryogenic callus with high regeneration potency [12–14]. Combined with the rapid progression of bioreactor

Fig. 1 Schematic diagram of large scale micro-propagation system

108

H. Honda · T. Kobayashi

technology, some novel immobilized cell systems are being developed for the scale-up production process. For mass production of regenerated plantlets, calli with embryogenic ability should be selected. Manual screening by human experts is labor-intensive, time-consuming, and expensive. Therefore, some automatic selection systems using image analysis should be developed. Most embryo cultures give rise to a heterogeneous population including embryos at various stages of development, abnormal embryos, and callus. This population heterogeneity necessitates a sorting step to select for normal and mature embryos [15]. Further optimization of bioreactor operation is often produced by overcoming the difficulty in evaluating results. Typically an operator examines a population sampled from a culture microscopically and classifies the embryos into developmental classes and enumerates them. The process is subjective and tedious, and often results in a low degree of statistical confidence due to the fact that a small number of embryos are often classified and counted. An advanced image analysis and pattern recognition system is therefore needed to discern the development of somatic embryo automatically in the process control and optimization. In addition, plantlets via regeneration in a bioreactor need to be transferred onto a solid medium and acclimatized under light irradiation so as to obtain healthy plants. In this step, the transfer time is different for each plantlet; only plantlets with a long shoot, which confers high photosynthesis ability, should be transferred. An automatic selection system using image analysis is also desired. In this chapter, advanced methods for large-scale micro-propagation via somatic embryogenesis will be introduced in each step. Successful results on viscous additive supplemented culture for promoting callus regeneration are also discussed.

2 Bioreactor Design for Large-scale Micropropagation The design and operation of a bioreactor are mainly determined by biological needs and engineering requirements, which often include a number of factors: efficient oxygen transfer and mixing, low shear and hydrodynamic forces, effective control of physico-chemical environment, easy scale-up and so on. Because some of these factors can be mutually contradictory, it is difficult to directly employ a conventional microbial reactor to shear-sensitive plant tissue cultures. Different reactor configurations for plant cells, tissue and organ cultures can be found in previous publications by Prenosil and Pederson [16], Scragg and Flower [17], Panda et al. [18], Doran [19], and Payne [20]. Several kinds of bioreactors, such as (hollow paddle and flat blade impellers) stirred tank bioreactor, bubble column, (internal and external loops) airlift bioreactor, rotating drum bioreactor, stirred-tank with draft tube and mist bioreactor, have been attempted for plant cell, tissue and organ cultures (Fig. 2).

Fig. 2A–D Different types of bioreactors for plant cell, tissue and organs: A mechanically-agitated bioreactors, a; aeration-agitation, b; rotating drum, c; spin filter; B air-driven bioreactors, a; bubble column, b; draft tube, c; external loop; C non-agitated bioreactors, a; gaseous phase (mist), b; oxygen permeable membrane aerator, c; surface aeration; D light emitting draft tube

Large-Scale Micropropagation System of Plant Cells 109

110

H. Honda · T. Kobayashi

Fig. 3 Configuration of helical ribbon type bioreactor (HRI)

Among the bioreactors available for cultivation of shear-sensitive plant cells, some modified conventional stirring tanks, which are effective in the mixing of the contents, the suspension of cells, the break-up of air bubbles for improved aeration have been developed by employing a variety of impeller designs [21–24]. Consequently, these bioreactors have great potential when used at mild agitation intensity and they have received increasing attention. A bioreactor with a helical-ribbon impeller [25, 26] has been effectively used in high-density plant cell culture (Fig. 3). This bioreactor displayed good performance for growing high-density Catharanthus roseus suspension cultures with high cell viability and embryogenic cultures of a transformed Eschscholtzia californica cell line [25, 26]. In the bioreactor, larger double helical-ribbon impeller with the surface baffles exhibited uniform low-shear mixing of cultures with sufficient of oxygen supply without excessive foaming and biomass flotation.

Large-Scale Micropropagation System of Plant Cells

111

Recently, Wang and Zhong successfully designed a novel centrifugation impeller bioreactor (Fig. 4) for shear-sensitive cell systems [27, 28]. They proved that the new bioreactor possessed several advantages over a widely used cell-lift one. These include much higher liquid lift capacity, better mixing performance, lower shear stress and surface liquid turbulence, which can cause a serious loss of cell viability. In addition, a sintered stainless sparger with tiny pore size was installed about 4 cm off the center with respect to the draft tube of the impeller, a high value of oxygen transfer was obtained under low hydrodynamic forces. To achieve low shear stress, we developed a shaking flask vessel-type bioreactor (Fig. 5), which is not associated with impeller for mixing and itself is shaken in a rotational mode. This reactor has been successfully used for large-scale plantlet propagation from the fragmented chip of horseradish hairy root. Lower shear rate is generally exposed to the cultures compared with a conventional agitated vessel with impeller, and oxygen requirement by plant cells is supplied from the free surface of the medium so that the cultures are never exposed to a hydrodynamic stress by bubbling. Therefore, almost the same number of regenerated plants at a rotational speed of 120 rpm, was obtained

Fig. 4 Schematic diagram of centrifugal impeller bioreactor (5 l)

112

H. Honda · T. Kobayashi

Fig. 5 Schematic diagram of shaking-vessel type bioreactor

compared with that in the control culture using a flask. Furthermore, the possibility to scale up the bioreactor and operate it has been proved [29]. Considering bioreactor operation, such as medium exchange, a spin filter bioreactor has been recognized as one of the most suitable bioreactor for mass propagation via somatic embryogenesis [30, 31]. In the bioreactor, the spent medium was removed without cell washout through the spinning central filter, which causes agitation of the medium without generating the shear forces. In the meantime, fresh culture medium that stimulates embryo differentiation is supplied in the bioreactor. In order to realize continuous callus cell proliferation and embryo development in a large scale, two spin filter bioreactors were configured in series, as shown in Fig. 6. A more efficient strategy would be operated with the first stage bioreactor (cell proliferation) as a continuous culture. Continuous culture provides more homogeneous and constant culture conditions for the cells and maintains an actively-growing cell population available for embryo development at all times. The second stage bioreactor results in the periodic production of embryos. The embryo development could involve controlled nutrient requirements and air components.

Large-Scale Micropropagation System of Plant Cells

113

Fig. 6 Two spin-filter bioreactors configured in series for the growth of embryogenic cells and the promotion of somatic embryo maturation on a large scale. The stage 1 bioreactor is for cell proliferation and operates on a continuous culture mode; the stage 2 bioreactor is for somatic embryo maturation and operates in a perfusion mode

Carrot embryogenic suspensions and mature somatic embryos have been successfully grown using spin-filter bioreactors on a large scale. When moved to a growth medium, the embryos produced in the bioreactor converted into normal plants.

3 Immobilization Technology in Large-scale Micro-propagation 3.1 Plantlet Regeneration from Immobilized Calli in Gel Beads Immobilization is an important strategy for the removal of shear stress. Gel entrapment culture using calcium alginate gel has been reported by many researchers [32–34]. We also investigated immobilized gel beads culture of embryogenic celery callus. Our advanced idea was that celery embryos and plantlets were released in a culture of immobilized Ca-alginate gel beads in which celery callus was entrapped under regeneration conditions. Culture procedure and schematic diagram of embryo and plantlet regeneration from immobilized callus are shown in Fig. 7. The cells released from the gel bead were larger that those obtained in suspension culture. The optimal concentration of alginate gel for embryo and plantlet production was 2% for the immobilized

114

H. Honda · T. Kobayashi

Fig. 7 Culture procedure and schematic diagram of embryo and plantlet regeneration from the immobilized callus

cell culture. Considering the maintenance of the gel bead structure and detrimental effect of CaCl2 on plantlet development, 5 mmol/l CaCl2 supplementation gave the best result in term of the number of heart and torpedo embryos and plantlets. The ratio of the number of heart embryos, torpedo embryos and plantlets to total number of cells in the immobilized cell culture was higher than that in the suspension culture. Repeated batch culture with 5 mmol/l CaCl2 provided long-term (more than 154 days) embryo and plantlet production without gel beads disruption. Productivity of plantlets in the immobilized cell culture

Table 1 Total number of embryos and plantlets in the suspension culture and in the immobilized culture

Suspension (1/l) (42 days) b

Immobilization with CaCl2 (1/l) a 0 mmol/l (126 days) b

5 mmol/l (154 days) b

10 mmol/l (154 days) b

Heart and torpedo, embryos+plantlets c

6.1¥104

1.7¥105

2.1¥105

1.0¥105

Plantlets d

2.1¥104

1.4¥105

1.7¥105

7.2¥104

5.0¥102

1.1¥103

0.1¥103

4.7¥102

Plantlets per a

b c

d

e

day e

All of the repeated batch cultures were carried out in the medium of the first regeneration stage containing 0, 5, or 10 mmol/l CaCl2. The duration of the culture is shown in parentheses. Total number of heart and torpedo embryos and plantlets in the medium. The heart and torpedo embryos and plantlets were counted at the end of first regeneration stage. Total number of plantlets at the end of second regeneration stage. The cells obtained from the first regeneration culture were grown in the second regeneration culture. The number of plantlets was evaluated at the end of the second regeneration stage. The number of plantlets produced per day.

Large-Scale Micropropagation System of Plant Cells

115

with 5 mmol/l CaCl2 was 2.2-fold higher than that in the suspension (Table 1). The attractive immobilization release system was also used in regeneration of carrot cells, repeated batch culture for plantlet production continued for 245 days with no significant decrease in the productivity (1.6¥105 plantlets/ l-medium/day) (Table 2). In the above immobilized gel beads system, the cells were entrapped in alginate gel and protected from physical stress. The released cells were immediately recovered from the culture. Thus, the immobilized gel beads system was expected to be easily scaled up because the production of heart embryo, torpedo embryos, and plantlets could be improved by increasing the number of gel beads in a bioreactor. A rotating-mesh basket type bioreactor with a cyclone Table 2 Summary of long-term repeated batch culture using immobilized gel beads

Immobilized culture supplemented with CaCl2 0 mmol/l 5 mmol/l 10 mmol/l 15 mmol/l 20 mmol/l

Suspension culture

Average of value in one batch operation Relative value of released cell volume a (–)

0.8

0.8

0.7

0.5

0.3

1b

Relative value of regeneration ability a (–)

1.5

1.4

1.2

1.0

0.9

1c

Number of plantlets produced (plantlets/l)

2.1¥106

2.0¥106

1.4¥106

0.93¥106

0.58¥106

1.8¥106

Cultivation time (days)

161

245

259

161

161

Number of batch operations

13

19

20

13

13

Cumulative number of plantlets (plantlets/l-medium)

2.9¥107

4.0¥107

2.9¥107

1.3¥107

0.8¥107

Productivity (plantlets/l/day)

1.8¥105

1.6¥105

1.1¥105

0.81¥105

0.50¥105

a

b c

1.3¥105

Relative values were calculated with respect to the value in the suspension culture. Regeneration ability was determined with an inoculum size of 0.2 ml-pcv in the regeneration medium. The released cell volume was 153 ml-pcv/l-medium. The number of regenerated plantlets was 12,000 plantlets/ml-pcv.

Fig. 8 Schematic diagram of cyclone cell separator

116 H. Honda · T. Kobayashi

Large-Scale Micropropagation System of Plant Cells

117

type cell separator (Fig. 8) was designed and used in the immobilized gel beads system for the embryo production and continuous recovery of released cells [13]. Plantlet formation of the released cells from the rotating mesh-basket type reactor culture reached a level of 115% compared with that from the flask culture. When 110 ml/min of medium flow was passed through the separator, 90% of the suspension cells were collected after 40 min of operation time. At the rotation speed of 100 rpm, constant value of released cell volume of 3 ml-pcv/l-medium/day and plantlet number of about 7000 plantlets/greleased cells were obtained over 12 days (Fig. 9). The volumetric productivity was 21,000 plantlets/l-medium/day, which corresponds to the number of plantlets able to be sowed in a 0.1-ha field. This bioreactor system has great potential for the production for the mass-production of regenerated plantlets. Furthermore, large numbers of plantlets will be obtained, because the size of the mesh-basket and packed beads volume can be increased easily. When 0.4 l (3700 beads) beads per l-mesh-basket were packed in the bioreactor, the ability to form plantlets was similar to that in the experiment of 0.1 l-beads

16

Rotational speed 100 rpm

14 12 10 8

50 rpm

6 4 2 0

150 rpm 0

2.0 1.8 1.6 1.4 1.2 1.0 0.8 0.6 0.4 0.2 0 0

1

2

3

4

5

6

7

8

9 10 11 12 13 14 15

150 rpm 100 rpm

50 rpm 1

2

3

4

5

6

7

8

9 10 11 12 13 14 15

Cultivation time (d) Fig. 9 Time courses of a long term culture in immobilized gel beads system. Number of plantlet formed at 4 days of 150 rpm and that at 14 days of 50 rpm were not measured since volume of released cells was too small

118

H. Honda · T. Kobayashi

per l-mesh-basket and the total number of plantlet was about fourfold. From these results, the bioreactor system proposed by us was concluded to be application for plantlet production as a strong tool for micropropagation in plant factories. 3.2 Plantlet Regeneration from Immobilized Calli in Polyurethane Foam Embryogenic rice calli tend to form larger clumps during cultivation. Therefore, immobilization of the calli has hardly been carried out until now. Porous supports such as polyurethane foam have often been used for immobilization of mycelial cells [35, 36] and plant cells [37–39]. In almost all cases, effective production of biologicals by the immobilized cells has been reported. To avoid the damage based on the hydrodynamic stress, we proposed the immobilization culture of rice callus using macroporous urethane foam support. A turbine-blade reactor (TBR), which has been developed for hairy root culture, was also used in the culture. In the culture space, polyurethane foam was added as immobilization support. Embryogenic rice calli were induced from mature rice (Oryza sativa L., Sasanishiki) seeded in N6 medium according to the previous paper [40]. Callus induction was carried out at 30 °C in the dark for four weeks in an incubator. When 3-mm cubes of polyurethane foam with the pore size of 1.3 mm were used for immobilization, rice calli proliferated well and a total cell concentration of 6.8 g/l was obtained after a one-week culture. Rice calli were cultured in the TBR under various agitation speeds (200, 300, and 400 rpm) using the cube supports corresponding to 5 vol.% of 600 ml growth medium. The results are shown in Fig. 10 (triangles and broken lines). When the supports were used in the TBR, cell growth was not inhibited and there was no decrease in the regeneration frequency at high agitation speeds (300, 400 rpm), whereas cell growth inhibition was observed without supports. Support volume had a pronounced effect on immobilization frequency of rice callus. Maximum immobilization frequency was found at 60 ml. By three times repetition of the periodic operation (agitating at 300 rpm for 5 min and then 50 rpm for 2 min, and then 200 rpm of constant agitation speed during the remaining time), distribution of rice calli within the support became homogenous and immobilization frequency was improved as compared with that using the constant bioreactor operation at 200 rpm. However, the immobilization frequency decreased due to insufficient support for immobilization and the floating of the support by air flow.After the turbine blade bioreactor was modified by setting the air sparger at the bottom of cylindrical stainless mesh in order to reduce the floating of the support by air flow, the immobilization frequency increased further and reached 86.3% when we increased the support volume to 90 ml in the modified turbine blade bioreactor. The regeneration frequency of immobilized callus was not affected by the support volume increase, periodic operation and bioreactor modification when they were transferred on solid regeneration medium.

Large-Scale Micropropagation System of Plant Cells

119

Fig. 10A,B Effect of agitation speed on: A cell concentration; B regeneration frequency of rice calli in TBR without (empty circle, filled circle) and with (empty triangle, filled triangle) supports. Dotted lines show the results of the control culture

We developed an immobilized cell system for cultivation of embryogenic rice callus using polyurethane foam, and found that the immobilized callus maintained high regeneration ability because shear stress and hydrodynamic damage were avoided [41, 42]. In these studies, however, regeneration was carried out on the solid medium. The clumps of calli grew larger on the solid medium for regeneration and larger clumps were so fragile. Moreover, the ratio of the clumps with root and shoot were about 80%, not 100%. Therefore, it will be necessary to establish some sophisticated systems for handling of regenerated callus when those calli were finally transplanted to the paddy soil in practical use. From these considerations, we hit upon an idea of direct plantlet regeneration of rice calli immobilized in polyurethane foam in liquid medium. This way will be convenient because our subjects are the calli immobilized in

120

H. Honda · T. Kobayashi

foam, not fragile clumps, and the foams exit in the liquid medium and are easy to transport. On the basis of these results, the following novel regeneration system is proposed. Immobilization supports are transferred directly into the regeneration medium and the regenerated calli are obtained in situ in the supports. Regenerated calli in the supports with three to five germinated shoots are incubated in a preculture nursery and then transferred to a normal paddy. This system, that is in situ regeneration, is attractive since it is easily applicable to conventional agricultural procedures. An in situ regeneration system is composed of the callus growth stage and regeneration stage. In the callus growth stage, a 3-mm cube of with a pore size of 1.3 mm was employed. In the in situ regeneration stage, regeneration was carried out in a liquid regeneration medium and macroporous support suitable for immobilization in regeneration stage was investigated. Since it has been reported that the medium exchange regeneration culture was effective for in situ regeneration, the regeneration stage was divided into two periods. Schema of in situ regeneration is shown in Fig. 11. The following two mediums were used [43]: A1 medium –1/2 N6 medium (pH 5.8) supplemented with 30 g/l sorbitol, 10 g/l sucrose, 1.4 g/l proline, 2.0 g/l casamino acids, 0.4 mg/l NAA, 0.5 mg/l kinetin; A2 medium –1/2 N6 medium (pH 5.8) supplemented with 15 g/l sorbitol, 7.5 g/l sucrose, 1.0 g/l casamino acids, 1.0 mg/l NAA, 0.5 mg/l kinetin. The rice calli of seven-day culture using N6 medium, which are mainly 1–2 mm of clump size, were transferred to A1 liquid medium for the first stage of regeneration. After 20 days, the medium was exchanged to A2 medium and the liquid culture was continued as the second stage of regeneration in this experiment. In the first stage, the embryogenic calli began to differentiate, and the callus size ranged mainly from 1.0 mm to 2.8 mm. No shooting plantlets were observed in the first stage of regeneration although some green spots were observed. On the contrary, in the second stage, the callus size increased quickly. Rice calli between 2.8 to 4.0 mm and above 4.0 mm in diameter occupied 48.1% and 31.4% respectively after 40 days. Most of them were regenerated plantlets with shoots. From these findings, the callus immobilization was conducted in the second stage after the medium exchange. From size distribution, we prepared three kinds of polyurethane foam with larger pore size, 1.9, 3.6, 5.1 mm. A 10-mm support cube with pore size of 3.6 mm gave the highest efficient immobilization and in situ regeneration of rice callus. Rice calli after seven-day culture using N6 medium were transferred in A1 medium, and cultured for a given number of days. At that time, the medium was exchanged to A2 medium with the support cubes addition, and the calli were cultured for the rest of the 40-day period. The optimum duration of the first stage in A1 medium was 15 days and the percent of support cube with 3–5 plantlets reached 61.0% after 40 days. Compared with the result at the optimum time of medium exchange, the ratio of in situ regeneration decreased in other cases. This may be due to the fact that the amount of nutrients and osmotic pressure in the medium should be regulated on the different development stage of the regeneration

Fig. 11 Schematic diagram of in situ regeneration culture of rice callus

Large-Scale Micropropagation System of Plant Cells 121

122

H. Honda · T. Kobayashi

Fig. 12A,B Photograph of: A regenerated rice calli; B developed plantlets in 10-mm support cube with the pore size of 3.6 mm

phase [43], or some inhibitory factors that might be released from the embryogenic calli were removed [44]. When the rice callus was cultivated with support cubes into 60 ml of A2 medium in a 500-ml flask, 83% of immobilization ratio was accomplished and 82% of support cubes involved in 3–5 regenerated plantlets after 25 days. The percent of support cube with 3-5 plantlets was limited at a low exchanged medium volume of 30 ml. This may be due to the lack of nutrients. However, at a higher exchanged medium volume of 90 ml, the support cubes were submerged completely in liquid medium during the regeneration period, and the plantlet development was also inhibited. This might be related to the limit of oxygen transfer to the immobilized calli in the support cube. Shoot length of regenerated plantlets obtained from the in situ regeneration culture was longer than that from the suspension culture. In this experiment, the support cube with 3–5 regenerated plantlets from the 500-ml flask was transferred to 1/4 MS solid medium supplemented with 10 g/l sorbitol and 5 g/l sucrose (S medium). After ten days, three to five regenerated plantlets developed quickly into three to five plants with a length above 10 cm (Fig. 12). These healthy plants were easy to transfer to photosynthetic growth in low nutrient broth. These mean that the plants could be transferred to a normal paddy in the practical agricultural procedures after further acclimatization.

Large-Scale Micropropagation System of Plant Cells

123

4 Image Analysis System for Larger-scale Micropropagation The process engineering including bioreactor design, high cell density culture, image analysis with computer device and robotic system can improve the micropropagation process. To establish the process for plant micropropagation, some problems are still laid, except for bioreactor design. The following three subjects should be overcome by image analysis. One is the selection of embryogenic callus. Many researchers have reported the induction and propagation of somaclonal callus from various species of plants. Such callus clones, however, do not all exhibit embryogenicity. A certain percentage of embryogenic calli can not generate shoots and roots and never regenerate. The second is the selection of embryos. Embryogenic callus is never homogeneous, and contains various calli, which are in the different stages of regeneration or development. In general somaclonal embryos, the embryos can be classified three stages for development, globular, heart, and torpedo types embryos. The third is the decision of the timing of transferring to an acclimatization medium. Regenerated plantlets with enough length of shoot can be transferred to the low nutrient medium for photosynthetic growth. 4.1 Selection of Embryogenic Callus In the cultivation of plant calli on solid media, two kinds of calli such as compact and friable calli, which are a bright yellow and a whitish clump, respectively, are often obtained. Generally speaking, a bright yellow compact callus shows high embryogenic activity and a whitish callus low or none. Therefore, distinction of these calli is of much importance in the regeneration step. In the study, eight clones of sugarcane callus from three subtypes, four compact callus clones and four friable ones, were used. For automatic selection of calli, the image analysis system constructed by us was used, which consisted of a microscope, a Charge Coupled Device (CCD) camera, and a ring type light source. Calli in sterilized dishes were used in situ for image analysis. The original image of the callus was input to a computer via an image analysis board. Using Adobe Photoshop 3.0 J, the image of only the region of the callus was extracted. The brightness of yellow, Br(Y), or white, Br(W), was defined using Br(R), Br(G) and Br(B). Those were defined as follows: Br(Y)Minimum [Br(R), Br(G)] Br(W)Minimum [Br(R), Br(G), Br(B)] Here, Minimum [A, B, C] is a function to determine the smallest value among A, B, and C. The difference between Br(Y) and Br(W), Br(Y–W), which can be used to express the yellowish grade, was calculated.

124

H. Honda · T. Kobayashi

Fig. 13 Frequency distribution of Br(Y–W) for compact calli (filled circle) and friable calli (empty circle)

When Br(Y–W) was determined from all pixels of the original images of both calli, the compact calli were found to be clearly distinguished from the friable calli by the frequency distributions of Br(Y–W). The typical frequency distributions of Br(Y–W) for compact and friable calli are shown in Fig. 13. The average brightness center values of the 20 compact and friable callus images were 32.3 and 11.7, respectively. The value was found to be specific to the type of callus. The brightness center, CY–W, was calculated: CY–W = ∑ Br(Y–W) ¥ f(Y–W)/∑ f(Y–W) where f(Y–W) shows the frequency of y–w at Br(Y–W). In order to apply the average brightness center values, Av(CY–W), to callus selection, 50 to 60 images of each of 8 callus clones were collected in the same fashion. Among them, images of small or irregularly colored calli were removed and 20 images were selected, then the value of Av(CY–W) was calculated. It was found that the callus clone with less than 10 units of Av(CY–W) was never regenerated and a proportional relationship between Av(CY–W) and the regeneration frequency of the callus clone, Fr, was obtained in the form of the following equation: Fr(%) = 3 ¥ (Av(CY–W) – 10) These results suggest that embryogenic calli can be selected from a mixed population of calli since the yellowish compact callus generally shows high embryogenecity. The image analysis using Br(Y–W) proposed here is a useful tool for automatic selection of such calli in a plant factory [45].

Large-Scale Micropropagation System of Plant Cells

125

4.2 Selection of Embryos To decide the time to transfer the next culture stage in plant somatic embryo culture, the classification of embryos and non-embryos using image analysis with artificial neural network (ANN) was demonstrated [46]. In general, the healthy embryos can develop via three morphological types of cells; globular, heart, or torpedo. From microscopic observation, human experts can distinguish the embryos from the non-embryos. This is one of the most reliable decisions. Therefore, if the morphological characteristics of the healthy embryos are fairly extracted from image data and information analysis are also done, the computer vision system for screening will be established.ANN is an intelligent technology, which can be used for estimation and prediction instead of a human decision which has an analogy with the human brain. It has been said that ANN could provide solutions by learning complex relationships without requiring the knowledge of the model structure. Celery embryogenic callus was cultivated in SH base medium supplemented with 8% mannitol and 10 ng/l zeatin (pH 5.8). The suspension culture was diluted by water to avoid the overlapping of embryos and poured in a 96-well Petri dish for embryo classification. Digital images from the cell preparations were acquired via an optical microscope equipped with an RGB color CCD video camera. The three channels (red, green, and blue) from the video signal were fed to an image acquisition board installed in an image-processing computer. The acquired image was digitized to binary image using binary transformation on the basis of all pixels of a brightness level which was the average of three channel levels. From the vast image analysis database, four parameters (area, ratio of length to width, circularity, and distance dispersion) were selected as morphological characteristics on human decision. The structure of ANN was shown in Fig. 14. Among the four parameters, the use of the first three was satisfactory for the classification between embryos and non-embryos. After the neural network was trained with 67 plant cells (23 embryos and 44 non-embryos), 33 plant cells (19 embryos and 14 non-embryos) were tested to narrow down to parameters extracted from three kinds of embryo or non-embryo for sufficient classification performance of the neural network. The ability of the resultant network with four parameters, area, circularity, ratio of length to width, and distance dispersion was excellent: 89.5% of objects was correctly classified with a corresponding human recognition. Three parameters, area, circularity and ratio of length to width were necessary for high recognition efficiency, and the distance dispersion could be omitted from parameters without loss of recognition ability. For three kinds of embryo and non-embryo classification, area and circularity provided a relatively large contribution for correct classification. Using the ANN model acquired, the total of heart and torpedo embryos in practical suspension culture was estimated, that is 1140 at the end of the regeneration stage. After the culture was transferred into the maturation stage,

126

H. Honda · T. Kobayashi

Fig. 14 Structure of artificial neural network Table 3 Prediction of plantlet formation by neural network

Recognized embryos

Obtained plantlets

Globular

Intermediate

Torpedo

Total

660

840

300

1800

Abnormal plantlet

Normal plantlet

Total

575

1240

1816

the culture has generated 1240 plantlets (Table 3). The ANN results matched well the experimental results, thereby indicating a high potential of the image analysis for prediction of plantlets number. 4.3 Estimation of Shoot Length Regenerated callus obtained from bioreactor cultures need to be transferred onto a solid medium and acclimatized under light irradiation so as to obtain healthy plants. In this step, the timing of transfer is different for each plantlet: only plantlets with a long shoot, which confers high photosynthetic ability, should be transferred. Therefore, the development of an automatic selection system using computer vision system is desired. In the case of micropropagation of embryogenic rice callus, regenerated rice callus with a shoot more than 2 cm in length could be transferred from a medium supplemented with sucrose to a medium without sucrose for acclimatization (Fig. 15). Image analysis was applied for the automatic selection of regenerated rice callus for acclimatization.

Large-Scale Micropropagation System of Plant Cells

127

Fig. 15A,B Lengths of elongated shoots after transfer to acclimatization medium; A for a particular callus; B for callus in each length range

Embryogenic rice callus was induced from mature rice seeds (Oryza sativa L., Sasanishiki) harvested in 1992 on solid N6 medium (pH 5.8) supplemented with 1% sucrose, 3% sorbitol, 12 mmol/l proline, 0.1 g/l casein hydrolysate, 5 mmol/l MES, 4 mg/l 2,4-dichlorophenoxyacetic acid, and 2 g/l Gelrite. After two to three weeks culture, regenerated callus was transferred onto an acclimatization medium, consisting of MS medium supplemented with 5 mmol/l MES and 4 g/l Gelrite prepared in a plant pot [47]. Incubation was done under the same conditions as those used for regeneration. After three weeks, elongated shoots were straightened and the length of elongation was measured. To estimate a shoot length, extraction of the shoot region from the image data is necessary. Therefore, an identification model was constructed in which, for each pixel, one region – shoot, callus, or medium – could be identified. The original image was stored as data sets of R(red), G(green), and B(blue) brightnesses with 256 levels. For regeneration, one multi-regression analysis (MRA) model and two fuzzy neural network (FNN) models were compared with each other. FNN-B consists of three independent models, which are FNNs with three input units and one output unit. Different grades of fuzzy variables are obtained from the data sets of the i-th pixel in each of the three models. On the other hand, FNN-A is comprised of only one model with three input and three output units. The correctness of the predicted recognition against all data points in the two FNN models was determined. In both FNN models, the recognition correctness of all subjects was remarkably high. The correctness for medium was close to 100%, and 95% was obtained for shoot recognition (Table 4). In order to predict the shoot length, the shoot region was extracted from the trinary image after pretreatment, i.e., such as removal of small, isolated shoot regions and filling-up small, isolated non-shoot regions occurring in the shoot

128

H. Honda · T. Kobayashi

Table 4 Comparison of correctness in FNN-A and FNN-B modeling

Model

MRA FNN-A FNN-B

J value

0.0657 0.0369 0.0350

Correctness (%) Shoot

Callus

Medium

All

Average

– 95.0 95.0

– 83.9 86.7

– 98.9 98.9

– 92.6 93.5

– 92.6 93.6

Fig. 16 Relationships between actual lengths and lengths obtained by image analysis

region. After thinning and extraction of the longest path, the number of pixels constituting the remaining line was counted. The actual shoot lengths were measured when the original images were taken. The correlations between the actual lengths and the predicted ones are shown in Fig. 16. Data points plotted on the diagonal line mean that the shoot length was accurately predicted by the image analysis. Almost all data points are seen to be close to the line, demonstrating that the level of accuracy was significantly high. The average error was only 1.3 mm. In Fig. 15 it was shown that a difference of 5 mm should be able to be distinguished. Therefore, it was concluded that the computer-aided image analysis developed here can be useful for the prediction of shoot length and the automatic transfer of regenerated callus to an acclimatization medium [48].

Large-Scale Micropropagation System of Plant Cells

129

5 Advanced Cultivation Method for Callus Regeneration – Viscous Additive Supplemented Culture To overcome the negative effects of hydrodynamic stress, immobilization of plant cells has been studied [8–11]. We have reported the culture of celery [12] and carrot [13, 14] embryogenic calli immobilized in gel beads as mentioned above. Although immobilization can improve the physical environment, immobilized culture is not commonly used in industrial plant cell biotechnology due to the high cost of immobilization and the much greater risk of contamination during the cell immobilization steps compared with suspension culture. Therefore, a new method that does not require complicated procedures but which has the same merits as immobilization is desired. We have successfully developed an innovative and convenient culture method using a medium supplemented with a viscous additive, designated N medium (viscosity; approximately 3 mPa s) [49]. N medium, MS growth medium supplemented with various concentrations of CMC, ALG, or PEG, was prepared and carrot calli were cultivated. The calli obtained were transferred to the regeneration medium. After 12 days, the medium containing regenerated plantlets was examined. Many regenerated plantlets were present in the cultures supplemented with CMC or ALG, whereas the control culture cell suspensions contained a relatively large volume of small calli (Fig. 17A). With CMC and ALG, the fresh weight increased with increasing additive concentration and became maximal at 0.4% and 0.1%, respectively. In these cases, the results correlated fairly well with increasing viscosity of the medium (Fig. 17B). To determine a desirable time frame for the enhancement of plantlet production, calli obtained from the conventional medium without CMC were transferred to N medium and the regeneration abilities of calli obtained after being cultivated in repeated batch mode in N medium were compared. Plantlet regeneration was most fully enhanced after two cultivations in N medium, and plantlet regeneration significantly decreased after the first run (14 days) when cells obtained from N medium culture were inoculated into the medium without CMC. To clarify why plantlet regeneration is enhanced by a viscous additive, the size distributions of cell suspensions after 14-day growth culture – not 12-day regeneration culture – were investigated using stainless-steel mesh sieves. After the cultivation in N medium, calli smaller than 0.2 mm comprised 28%, compared with 49% in the case of the control culture. After the second run in N medium, only 17% of the calli were smaller than 0.2 mm. It has been reported that small-size carrot calli have a low regeneration frequency [14, 50]. It was thus confirmed that the upward shift in the callus size distribution was a major contributory factor to the high plantlet productivity in N medium. To investigate the possibility of long-term operation, repeated batch culture was carried out using the growth medium with or without CMC. In the first

H. Honda · T. Kobayashi Fig. 17A,B Regeneration enhancement of carrot callus using viscous additive supplemented medium; A effect of various additive concentration; B replot of graph A against medium viscosity

130

batch, plantlet production was approximately 2.5 times higher than that of the control culture (Fig. 18). In the control, plantlet production decreased after the third batch (42 days). Surprisingly, calli obtained from N medium culture continued to produce large numbers of plantlets though 18 batches with no significant decrease. This corresponds to 250 days, which is approximately six times longer than the control. In the control culture, the callus size distribution shifted downwards after the third batch, while the size distribution in N medium did not vary through 250 days. Since the embryogenecity of carrot calli eventually decreases after several subcultures in liquid culture, it is note-

Large-Scale Micropropagation System of Plant Cells

131

Fig. 18 Regeneration ability of Carrot callus in repeated batch culture Symbols: (filled circle), using N medium; (empty circle), using the control medium

worthy that the embryogenic ability was retained for such a long period in N medium. This enhancement of regeneration ability of callus obtained from viscous additive supplemented culture is considered as following. In the ordinary culture without viscous additive, turbulent flow will occur to mix the liquid medium and the callus will be exposed to the collision or turbulent eddy. However, turbulence is lowered in the N medium supplemented with viscous additive and the boundary layer around the callus become thick. Consequently, the callus will be relieved from the hydrodynamic stress (Fig. 19). Callus is a cell cluster and it can be formed by intercellular attachment. The formation of the

Fig. 19 Considerable diagram of lowering hydrodynamic stress

132

H. Honda · T. Kobayashi

cluster is generally believed to bring high regeneration ability or a high level of metabolite production from the plant cells. Since callus could enlarge in N medium with relatively little disruption, large cell clusters could be maintained even if the callus intercellular attachment was weakened by long-term culture. Thus, low hydrodynamic stress in N medium caused the production of callus with high regeneration ability during the long-term culture, and significantly extended the embryogenic lifetime of callus. It should be noted that the anthocyanin dye production from grape callus was significantly enhanced when the N medium was applied to grape callus culture. It was found that an approximately two times higher level of anthocyanin content was obtained and the productivity was found to continue at a high level during the 30th batch culture [51] and the enhancement was observed even in the air-lift bioreactor culture [52]. These results suggest that the viscous additive supplemented medium was applicable to various types of plant cell cultures and the possibility of the medium in industrial scale plant cell culture was strongly suggested.

6 Conclusion and Perspectives The proliferation of embryogenic calli offers a unique opportunity for the large-scale micropropagation of plants. Bioreactor for plant cell, tissue and organ culture, which maintains high levels of mixing and mass transfer but reduces the intensity of shear, is an effective tool in this endeavor. Immobilization and viscous additive supplemented medium become important strategies for the removal of shear stress to keep high regeneration ability of plant cells in the large-scale bioreactor cultivation. Image analysis with computer device and robotic system will become a powerful technology in the process optimization of the micropropagation process. Efficient combination of these developing technologies will be great to accelerate the process automation and commercially competitive. Larger-scale micropropagation will open up a bright future for the modernization of agriculture.

References 1. 2. 3. 4. 5. 6. 7. 8.

Murashige T (1974) Ann Rev Plant Physiol 25:135 Denchev PD, Kuklin AI, Scragg AH (1992) J Biotechnol 26:99 Ammirato PV, (1989) Newsl IAAPTC 57:2 Redenbaugh K (1990) Hortic Sci 25:251 William EG, Maheswaran G (1986) Ann Bot 57:443 Chen THH, Thompson BG, Gerson DF (1987) J Ferment Technol 65:353 Hooker BS, Lee JM, An G (1989) Enzyme Microb Technol 11:484 Lindsey K, Yeoman MM, Black GM, Mavituna F (1983) FEBS Lett 155:143

Large-Scale Micropropagation System of Plant Cells

133

9. Nakajima HK, Sonomoto H, Morikawa H, Sato F, Ichimura K,Yamada Y, Tanaka A (1986) Appl Microbiol Biotechnol 24:266 10. Facchini PJ, DiCosmo F (1990) Appl Microbiol Biotechnol 33:36 11. Hulst AC, Meyer MMT, Breteler H, Tramper J (1989) Appl Microbiol Biotechnol 30:18 12. Suehara K, Kohketsu K, Uozumi N, Kobayashi T (1995) J Ferment Bioeng 79:585 13. Suehara K, Nagamori E, Honda H, Uozumi N, Kobayashi T (1998) J Chem Eng Jpn 31:613 14. Nagamori E, Honda H, Kobayashi T (1999) J Biosic Bioeng 88:226 15. Chi CM, Vits H, Staba EJ, Cooke TJ, Hu WS (1994) Biotechnol Bioeng 44:368 16. Prenosil JE, Pesersen H (1983) Enzyme Microb Technol 5:323 17. Scagg AH, Fowler WM (1985) In: Vasil I (eds) Cell culture and somatic cell genesis of plants. Academic Press, London, p 103 18. Panda AK, Saroj M, Bisaria VS, Bhojwani SS (1989) Enzyme Microb Technol 11:386 19. Doran PM (1993) Adv Biochem Eng Biotechnol 48:117 20. Payne AK, Shuler ML, Brodelius P (1987) In: Lydersen BK (ed) Large scale cell culture technology. Hansen Publishers, New York, p 193 21. Hooker BS, Lee JM, An G (1990) Biotechnol Bioeng 35:296 22. Treat WJ, Engler CR, Soltes EJ (1989) Biotechnol Bioeng 34:1191 23. Zhong JJ, Yu JT, Yoshida T (1995) World J Microbiol Biotechnol 11:461 24. Chisti Y, Moo-Young M (1993) J Chem Technol Biotechnol 58:331 25. Jolicoeur M, Chavarie C, Carreau PJ, Archambault J (1992) Biotechnol Bioeng 39:511 26. Archambault J, Williams RD, Lavoie L, Pepin MF, Chavarie C (1994) Biotechnol Bioeng 44:930 27. Wang SJ, Zhong JJ (1996) Biotechnol Bioeng 51:511 28. Wang SJ, Zhong JJ (1996) Biotechnol Bioeng 51:520 29. Honda H, Hattori T, Uozumi N, Kobayashi T, Kato Y, Hiraoka S (1997) J Chem Eng Jpn 30:179 30. Styer DJ (1985) In: Henke RR, Hughes KW (eds) Tissue culture in forestry and agriculture. Plenum Publishing Corp, New York, p 117 31. Ammirato PV, Styer DJ (1985) In: Zaitlin M, Day P, Hollaender A (eds) Biotechnology in plant Science: relevance to agriculture in eighties. Academic Press, New York, p 117 32. Brodelius P, Deus B, Mosbach K, Zenk MH (1979) FEBS 103:93 33. Majerus F, Pareilleux A (1986) Plant Cell Rep 5:302 34. Rhodes MJC (1985) Top Enzyme Ferment Bioeng 10:51 35. Nakashima T, Kyotani S, Izumoto E, Fukuda, H (1990) J Ferment Bioeng 70:85 36. Wakisaka Y, Segawa T, Imamura K, Sakiyama T, Nakanishi K (1998) J Ferment Bioeng 85:488 37. Rhodes MJC, Smith JI, Robins RJ (1987) Appl Microbiol Biotechnol 26:28 38. Furuya T, Orihara Y, Koge K, Tsuda Y (1990) Plant Cell Rep 9:125 39. Corchete P, Yeoman MM (1989) Plant Cell Rep 8:128 40. Chu CC, Wang CC, Sun CS, Hsu C, Chu KC, Bi FY (1975) Sci Sin 18:659 41. Moon KH, Honda H, Kobayashi T (1999) J Biosci Bioeng 87:661 42. Liu CZ, Moon KH, Honda H, Kobayashi T (2000) Biochem Eng J 4:169 43. Kobayashi H, Okii M, Hirosawa T (1992) Jpn J Breed 42:583 44. Raghava-Ram NV, Nabors MW (1985) Plant Cell Tissue Organ Cult 4:241 45. Honda H, Ito T, Yamada J, Hanai T, Matsuoka M, Kobayashi T (1999) J Biosci Bioeng 87:700 46. Uozumi N, Yoshino T, Shiotani S, Seuhara K, Arai F, Fukuda T, Kobayashi T (1993) J Ferment Bioeng 76:505 47. Nursery Technology Co (1993) Kenkyusho houkoku, Nursery Technology Co, Tokyo

134

H. Honda · T. Kobayashi

48. 49. 50. 51. 52.

Honda H, Takikawa N, Noguchi H, Hanai T, Kobayashi T (1997) J Ferment Bioeng 84:342 Nagamori E, Omote M, Honda H, Kobayashi T (2001) J Biosci Bioeng 91:283 Iwai H, Kikuchi A, Kobayashi T, Kamada H, Satoh S (1999) Plant Cell Rep 18:561 Nagamori E, Hiraoka K, Honda H, Kobayashi T (2001) Biochem Eng J 9:59–65 Honda H, Hiraoka K, Nagamori E, Omote M, Kato Y, Hiroka S, Kobayashi T (2002) J Biosci Bioeng 94:135

Received: September 2003

Adv Biochem Engin/Biotechnol (2004) 91: 135– 169 DOI 10.1007/b94208 © Springer-Verlag Berlin Heidelberg 2004

Development of Culture Techniques of Keratinocytes for Skin Graft Production Masahiro Kino-oka · Masahito Taya (✉) Division of Chemical Engineering, School of Engineering Science, Osaka University, 1–3 Machikaneyama-cho, Toyonaka, Osaka 560-8531, Japan [email protected]

1

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 136

2 Development in Tissue-Engineered Skin Grafts . . . . . . . . . . . . . . . . 139 2.1 Architectures of Skin Grafts . . . . . . . . . . . . . . . . . . . . . . . . . . . 140 2.2 Commercialized Skin Grafts . . . . . . . . . . . . . . . . . . . . . . . . . . . 141 3 3.1 3.2 3.3

Kinetic Model for Culture of Anchorage-Dependent Cells . . . . . . . . Construction of Kinetic Model . . . . . . . . . . . . . . . . . . . . . . . Model Estimation in Cultures of Adult and Neonatal Keratinocytes . . . Effect of Culture Surface on Attachment and Growth in Cultures of Various Cell Lines . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . 143 . . 144 . . 146 . . 150

4 Cellular Senescence in Successive Culture of Keratinocytes . . . . . . . . . . 152 4.1 Image-Analyzing System . . . . . . . . . . . . . . . . . . . . . . . . . . . . 153 4.2 Evaluation of Cellular Senescence . . . . . . . . . . . . . . . . . . . . . . . . 155 5

Designs of Culture Operation and Bioreactor for Epithelial Sheet Production . . . . . . . . . . . . . . . . . . . . . . . . . 159 5.1 Bioreactor Accompanied with Image Analysis . . . . . . . . . . . . . . . . . 160 5.2 Automated Operation for Successive Cultures . . . . . . . . . . . . . . . . . 163 6

Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 167

References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 168

Abstract The in vitro cultures of human tissues have attracted a great deal of medical attention as a promising technique for repairing defective tissues in vivo. In the last decade many companies have been established for supplying the regenerated grafts by means of tissue cultures of skin, cartilage, bone and so on. From the viewpoint of biochemical engineering, however, the culture systems for these tissues are not so sophisticated nor so programmed as the submerged culture systems developed for microorganisms. In manufacturing skin grafts, for instance, the raw materials of cells harvested from patients are heterogeneous, and the products of cultured tissues vary in the required size for individual epithelial sheets. Therefore, a reliable and robust process is desired for the production of cultured tissues with high reproducibility and quality. This review focuses on the strategies for developing the culture processes of keratinocytes targeting the epithelial sheet production, including (i) the introduction of culture techniques for keratinocyte cells and survey of skin graft production

136

M. Kino-oka · M. Taya

as it is, (ii) construction of kinetic model of cell growth, (iii) evaluation of cell properties based on image-analyzing techniques, and (iv) design of bioreactor system. Keywords Tissue engineering · Manufacturing process · Skin graft production · Human keratinocytes · Bioreactor design Abbreviations and Symbols AC Adherent cell area Cd Confluence degree of cells on culture surface GAG Glycosaminoglycan HF Human fibroblast HK Human keratinocyte Na Number of adherent cells on grids Nd Cumulative number of population doublings NdF Final value of cumulative number of population doublings NP Number of passages NS Total number of grids in culture vessel Pra Placement probability of daughter cell PBT Poly-butylene terephthalate PEO Poly-ethylene oxide PGA Poly-glycolic acid PLA Poly-lactic acid t Culture time ta Adherent time td Doubling time tL Lag time tm Time of first medium exchange X0 Concentration of inoculated cells X0¢ Unit concentration of inoculated cells Xa Concentration of adherent cells Xaf Final concentration of adherent cells Y Growth index a Ratio of saturated value of Xa to X0 in Eq. (2) b Parameter in Eq. (4) g Parameter in Eq. (4) m Specific growth rate t Time constant for cell adhesion in Eq. (2) * Value at critical point

1 Introduction In the last decade, the advances in tissue engineering have offered the promising strategies for reconstructing and repairing defected tissues in vivo [1–4]. In particular, the clinical trials of wound healing have yielded successful outcomes, and several companies are now established as suppliers of tissue engineered products such as skin, cartilage, bone and so on [5–9]. From a viewpoint

Development of Culture Techniques of Keratinocytes for Skin Graft Production

137

of biochemical engineering, however, the culture systems for these tissues are not so sophisticated nor so programmed as compared with the submerged culture systems developed for microorganisms. In manufacturing the tissue engineered products, the raw materials and final outputs are cells obtained from patients (or donors) and cultures themselves, respectively. The raw materials have heterogeneity and individuality depending on the state of patients and location of excised tissues, and the products vary in size required for patient individuals. These features demand the development of a novel strategy in the manufacturing to obtain the maximum harvests with high quality from the limited supplies of raw materials. For the production of cultured tissues, moreover, the operations including isolation of cells, primary and secondary cultures, and the preparation of cultured tissues are manually performed with tedious and repetitious tasks. The diversities of both raw materials and products as well as the complexity in culture operations oblige us to have a tailor-made process that belongs to the batch operation with less reproducibility. Figure 1 conceptually compares the manufacturing processes for common and fine chemicals with that for cultured tissues. The manufacturing process for common chemicals is in principle based on large-scale production accompanied with chemical reaction (reaction process) and refining steps through unit operation (purification process). Furthermore, to ensure the purity of products, an additional step of stringent selection for raw materials (sorting

Fig. 1 Comparison of manufacturing processes between chemicals and tissue engineered products

138

M. Kino-oka · M. Taya

process) may be inserted into the manufacturing process especially for fine chemical production. On the other hand, the manufacturing process for autologous cultured tissues has to be constructed on the basis of a concept distinct from the conventional processes described above. The impurities in final products of cultured tissues are difficult to be stripped after reconstruction of tissue. Therefore, the production culture should be achieved with assurance of quality under aseptically controlled conditions. In the case of production for allogenic cultured tissues, to some extent, material selection for sorting might be introducible. For the manufacturing process for the cultured tissues, the engineering problems have not been tackled so far although the studies on kinetic analysis, growth-model development, culture operation, and bioreactor design

Fig. 2 Culture process for epithelial sheet production

Development of Culture Techniques of Keratinocytes for Skin Graft Production

139

should be required to perform the reliable and robust process for certification of high-product quality. Cultivation of human keratinocytes is one of the most important steps in the production of epithelial sheet. As shown in Fig. 2, the culture operations generally comprise of isolation of keratinocytes from skin biopsy, primary and maintenance cultures, and the sheet formation culture. These cultures accompany with the cellular events of adhesion on culture surface, acclimation after inoculation, growth by cell division, contact inhibition among cells and differentiation accompanied with stratification. The complicated nature owing to these operations and cellular phenomena led to the fluctuation in culture profiles every culture runs. This article focuses on the strategies for developing the culture process of keratinocytes aiming at the epithelial sheet production, including (i) the introduction of culture techniques for keratinocyte cells and survey of skin graft production as it is, (ii) modeling and kinetics of cell growth, (iii) evaluation of cell properties based on image-analyzing techniques, and (iv) construction of bioreactor system associated with automated operations.

2 Development in Tissue-Engineered Skin Grafts Skin transplantation is most intensively and extensively applied for clinical treatments. The preparations of skin grafts, using cultured keratinocytes with stratified organization (cultured epithelial autologous or allogenic skin grafts), emerged soon after a breakthrough in multilayer keratinocyte culture using serum-containing medium and murine fibroblast cells as a feeder layer [10]. This technique permitted clinical application of the skin grafts to burn wounds and chronic leg ulcers at a commercial level [11, 12]. However, this traditional culture method with serum and murine fibroblast cells has several problems to be solved in biological and clinical aspects. Production of multilayer skin grafts takes a long period of 3–4 weeks, being thereby cost-intensive [13]. In addition, wounds covered with the skin grafts often blister due to serum antigens incorporated into cells and immunogenic rejection after transplantation [14, 15]. Therefore, the development of serum-free medium has been challenged to the support culture of keratinocytes free from feeder layer of murine fibroblasts [16–18]. Wound healing is affected not only by state of the wounds but also by constitution of the skin grafts. Lots of trails of transplantation with cultured skin grafts were reported so far, but an unequivocal standard to evaluate wound healing after transplantation has not been established yet. The cellular activities of keratinocytes such as growth rate in culture are considered to be one of the important factors relating to wound healing.

140

M. Kino-oka · M. Taya

2.1 Architectures of Skin Grafts The development of tissue engineered skin has been intensively undertaken in many dedicated companies and institutions. Several skin substitute products are currently available in the clinical setting. Skin as a tissue is characterized as follows. In the adult body, the skin surface area is estimated to be approximately two square meters and it weighs over 10 kg. It accounts for about one-third of the blood circulation. The skin is composed of three primary layers: epidermis, dermis, and hypodermis or subcutaneous tissues. These layers each contain different cell types and are structured to express the specialized functions of the skin. In addition to protective functions as a barrier to foreign intruders, the skin absorbs radiation and achieves sensory, metabolic, and immunologic functions. Cell sources for tissue engineering fall into three categories: autologous cells (from each patient), allogenic cells (from human donor, but not immunologically identical), and xenogenic cells (from donor but not human). Each category may be further delineated in terms of stem cells (adult or embryonic) or “differentiated” cells obtained from tissue, where the cell population obtained from tissue dissociation comprises a mixture of cells at different maturation stages and includes rare stem and progenitor cells. Some approaches use the whole cell mix, whereas others rely on separation or enrichment of stem cells. Several skin products that have been approved by the FDA to date rely on a cell origin from neonatal foreskin. This starting material reduces variability in cell conditions depending on age, sex, and anatomical location, and provides a source with sufficient proliferation potential. There are two types of skin defects and the healing of each follows distinct processes. Skin defects may be in partial thickness including the epidermis, or full-thickness which requires reconstruction of the dermis as well. For partialthickness defects, various types of biological or occlusive dressings including cultured epithelial sheets are employed to support re-epithelialization. Some products are available as temporary skin substitutes for the treatment of fullthickness defects. This is particularly useful for minimizing the steps of surgical procedures required for ultimate treatment of the patient, as well as providing immediate access to materials that will protect the patient at the critical step prior to autologous grafting procedures. Numerous approaches have been developed to treat burns, skin ulcers, deep wounds, and similar injuries. Some of these approaches include the following. (i) Composite material – upper layer of silicone to prevent fluid loss, and lower layer of chondroitin-sulfate and collagen to induce blood vessel formation and encourage the ingrowth of connective tissue – essentially a new dermis. After three weeks the upper layer is replaced with an extremely thin epidermal graft. The procedure was further refined to eliminate the upper layer in composite materials. This was accomplished by seeding epidermal cells from a small graft onto the lower original graft layer prior to placement. (ii) In vitro culture of epidermal cells (keratinocytes). Biopsies of about 1 cm2 is harvested from burn pa-

Development of Culture Techniques of Keratinocytes for Skin Graft Production

141

tients and expanded by 10,000 fold using a feeder layer of irradiated NIH 3T3 fibroblasts and medium components to ensure rapid cell growth. This approach allows the coverage of very large wounds. However, three to four weeks are needed to expand the number of cells, for some patients, particularly those severely burned, which may be too long a time period. It is suggested that cryopreserved allografts may offer the solution to the problem found commonly in the treatment with fresh (non-cryopreserved) skin grafts mentioned above. (iii) Poly-glycolic acid (PGA) mesh seeded with human neonatal dermal fibroblasts.A uniform stock of fibroblasts can be developed because these cells are fairly and easily cryopreserved. The seeded PGA graft can be placed into deeper wounds, and the implant is a vascularized tissue resembling dermis. A hydrated collagen gel seeded with fibroblasts has also been used. The gel is digested with concurrent reorganization of collagen fibrils. Recent topics concern the combined usage of artificial materials like temperature-responding polymers [19] in cultures of keratinocytes for epithelial sheet production. In addition, Ito et al. reported that the cultured human amniotic epithelial cells are available to be feeder layer for epithelial sheet formation of keratinocytes [20]. 2.2 Commercialized Skin Grafts Commercial products and experimental models for dermal and/or epidermal repair have been configured to employ single and combined materials as shown in Table 1 [21, 22]. Advanced Tissue Sciences’ Dermagraft is a bioengineered human-derived dermis made by fibroblast cells, and is composed of collagen, extracellular matrix proteins, and growth factors [23]. Living dermal tissue capable of supporting the migration, proliferation, and stratification of an epidermis is formed in about two weeks. In addition, a product from Advanced Tissue Sciences, TransCyte is used to provide a protective cover and to maintain the integrity of the wound bed, which is a temporary covering for burns that was approved by the FDA in March 1997 for full-thickness or third degree burns and for partial-thickness or second degree burns in October 1997 [24]. In 1998, Genzyme’s Epicel was said to employ a proprietary process for culturing and expanding autologous skin grafts for the treatment of severe burns. The grafts require the culture period of about 16 days or longer depending on graft size, and they have a 48-h shelf life (www.genzyme.com) [25, 26]. HydroMed’s Hydron Wound Dressing is an FDA cleared 510(k) film-forming medical device which acts as a protective barrier for second and third degree burns, dermabrasion, and decubitus ulcers (www.hydromed.com) [27]. LifeCell’s Alloderm is designed for use as a skin replacement and is commercially available for use in reconstructive plastic, dental, and burn surgery applications (www.lifecell.com) [28].AlloDerm is created using a LifeCell’s proprietary process which excludes the immunogenic components of donor skin

142

M. Kino-oka · M. Taya

Table 1 Selected examples of engineered skin substitutes

Trademarked products

Company

Source Dermis

Epidermis

USA ApliGraf

Organogenesis, Inc.

Collagen gel_allo HF

Cultured allo HK

EpiCel

Genzyme

Allodermis

Cultured auto HK

Integra

Integra LifeSciences

Collagen-GAG & silicone

Thin autograft

AlloDerm

LifeCell Corporation

Acellular dermal matrix

Thin autograft

DermaGraft

Advanced Tissue Sciences

PGA/PLA_allo HF

Thin autograft

TransCyte

Advanced Tissue Sciences

Allo HF

BioBrane

ORCEL

Ortec International, Inc.

Collagen_allo HF

Cultured allo HK

n/a

Univ Cincinnati/ Shriners Hospital

CollagenGAG_auto HF

Cultured auto HK

LaserSkin

Fidia Biopolymers (Italy)

Hayluronic acid

Cultured auto HK

PolyActive

HC Implants (The Netherlands)

PEO/PBT_auto HF

Cultured auto HK

Bioseed-M

BioTissue Technologies (Germany)

–

Cultured auto oral mucosa

Other countries

HF=human fibroblast; HK=human keratinocyte; GAG=glycosaminoglycan; PGA=poly-glycolic acid; PLA=poly-lactic acid; PEO=polyethyleneoxide; PBT=polybutyleneterephthalate.

obtained from tissue banks. The patient’s own cells are used in the AlloDerm transplant as a template to remodel the missing skin. LifeCell began the sale of AlloDerm grafts as a dermal replacement in the grafting of third degree burns in December 1993. Business products for Integra LifeScience’s skin defects and burns include INTEGRA Artificial Skin approved by the FDA in March of 1996 (www.integrals.com) [29]. INTEGRA Artificial Skin is designed to enable the body to regenerate a permanent dermal layer of skin while minimizing scar formation and wound contracture in full thickness skin defects. Organogenesis’ Apligraf Human Skin Equivalent is a product of manufactured living skin, being composed of two layers with a structure similar to

Development of Culture Techniques of Keratinocytes for Skin Graft Production

143

human skin (www.organogenesis.com) [30]. The product is originated from cells of infant human foreskins. One specimen of foreskin can be utilized in the production of up to 200,000 units of Apligraf. The FDA gave an approval to Apligraf for marketing in the treatment of venous leg ulcers. Ortec’s Composite Cultured Skin, ORCEL, is a tissue-engineered dressing consisting of two layers of human derived skin cells (dermal and epidermal) retained within a porous collagen matrix, which is placed so as to cover the area needed for skin regeneration (www.ortecinternational.com) [31]. ORCEL produces a multitude of growth factors which appear to promote migration of the applied cells into the wound site, which results in stimulation of skin regeneration and wound healing. The company is targeting a cryopreserved ORCEL which is expected to have six months of shelf life and to improve significantly manufacturing and inventory management, thereby reducing overall production cost. Additionally, hospitals and clinics will expand stocking capability because access to ORCEL can be optional in both scheduled and urgent procedures. In 1996, the FDA approved the protocols for the use of ORCEL in treating non-healing skin ulcers for patients with Epidermolysis Bullosa. Nowadays, ORCEL certified with FDA approval is available to several types of skin diseases, leg ulcers, etc. To date, in Japan there are no companies which produce cultured skin commercially, although some of them have the facility (i.e. cell processing center) for manufacturing cultured skin. Japan Tissue Engineering (J-TEC) is the earliest-established venture company, founded in 1999, in corporation with Prof. Ueda in Nagoya University, aiming at the supply of epithelial sheets (www.jpte.co.jp/English). J-TEC proposed an automation system for culture operation including medium change to prevent from cross contamination and human errors. In addition, BCS supported by Prof. Inoguchi in Tokai University recently announced the manufacturing plan for Autograf toward commercialization in 2004. Emerging companies in Japan must encourage tissue-engineered products (www.bcsinc.co.jp).

3 Kinetic Model for Culture of Anchorage-Dependent Cells Model predictions were performed to analyze the complex dynamics of cell populations as a whole [32] and to determine how the observed proliferation rates are affected by culture conditions. The development of populations of anchorage-dependent cells on surfaces poses interesting problems in modeling the sequences of events including cell division and contact inhibition. The understanding of the dynamics of entire cell populations and the prediction of their behavior is very important. This would be especially advantageous for tissue constructs seeded with cells from patients or donors, because these constructs may exhibit substantial patient-to-patient individualities. In the process accompanied with inhomogeneity, instability, and fluctuation in conditions of

144

M. Kino-oka · M. Taya

the starting materials, a feasible methodology to evaluate and predict culture states is needed as a supporter for the operators who have to monitor and judge the courses of tissue cultures. 3.1 Construction of Kinetic Model A cell placement model for the growth in monolayer culture of keratinocyte cells was established, and growth parameters governing culture performance was evaluated based on the model [33]. For the model development, as shown in Fig. 3, the culture process of keratinocytes, a kind of anchorage-dependent cells, is considered to consist of cell adhesion phase, lag phase, exponential growth phase and stationary phase. The model was constructed based on the following assumptions: – The cell suspension used as inoculum is prepared by enzymatic stripping of

–

– – –

cells in the preculture, being a mixture of the viable and nonviable cells owing to damage during the preparation. After inoculation of cell suspension into a culture vessel, the nonviable cells are ruptured and the individual viable cells settle and anchor with a time lag of adhesion time (ta) on bottom surface (cell adhesion phase). After a given period of acclimation (lag time, tL), the adherent cell becomes a mother cell capable of dividing (lag phase). By means of cell division, the mother cell of interest generates a daughter cell onto the vicinal bottom surface. The repetitive cell division occurs every doubling time (td) that is a constant value during the culture under sufficient supply of nutrients without inhibitory effects of toxic metabolites (exponential growth phase).

Fig. 3 Schematic drawing of culture process of keratinocytes

Development of Culture Techniques of Keratinocytes for Skin Graft Production

145

– When no vacant bottom surface for placement of a daughter cell exists, the

mother cell ceases cell division due to cell contact inhibition under the condition of monolayer growth without cell differentiation (stationary phase). With respect to individual cells after inoculation, the calculation of overall growth in a culture vessel was conducted with cell placement on a two-dimensional grid. To perform monolayer growth simulation, a single cell is assumed to occupy one unit square equivalent to cell size (AC). As culture time proceeds, the individual cells inoculated are placed on the squares at various adhesion times (ta), and the adhesion probability (Pra(ta)) is employed to assess the individual cells adhering to the grid at culture time, t=ta. Then, whole squares on the grid are scanned to pick up the squares occupied by mother cells with ability of cell division at a given culture time (t=ta +tL +n · td, n: integer). The possibility of cell division is checked for in the neighboring eight unit squares around the mother cell. In the absence of a vacant unit square around the mother cell, the daughter cell cannot be born owing to cessation of cell division by contact inhibition. After scanning cells capable of dividing throughout the grid, the repetition of the scanning is conducted until the end of culture. The concentration of adherent cells (Xa) was obtained by counting adherent cells on the unit square grid on the bottom surface as follows: Xa =

Na 95 NSAC

(1)

where Na and NS are the number of adherent cells on the grid and the square grid on bottom surface in a culture vessel. The cell adhesion was a requisite phenomenon during the early times of anchorage-dependent culture. To estimate the cell adhesion phase, the keratinocytes derived from different donors were incubated at X0=1.0¥104 cells/cm2 for 36 h during which cell division could be substantially neglected. As shown in Fig. 4, the value of the adherent cell ratio (Xa/X0) increased rapidly with

Fig. 4 Time profile of concentration of adherent cells during adhesion phase. The cells were suspended at X0=1.0¥104 cells/cm2 in a flask. The line shows the fitted values using Eq. (2)

146

M. Kino-oka · M. Taya

elapsed time and approached a saturated value. This value could be regarded as a fraction of viable cells capable of propagating. According to the data shown in Fig. 4, the time profile of cell adhesion was expressed as the time response of first order lag as follows:




$ %

t Xa/X0 = a 1 – exp – 32 t

(2)

where a and t are the ratio of saturated value of Xa to inoculum size, and time constant, respectively. Using Eqs. (1) and (2), the adhesion probability (Pra(ta)) was given concerning respective mother cells as follows: Pra(ta) =

d(Na/NS) 9922 dta

(3)

To investigate the growth manner of adult keratinocyte cells, the monolayer cultures in flasks were conducted at various inoculum sizes (X0) of 5.0¥103, 9.8¥103, 1.6¥104, and 2.7¥104 cells/cm2. As shown in Fig. 5, in the culture at X0=9.8¥103 cells/cm2, the adherent cell concentration (Xa) increased sharply during culture time of t=0 to 15 h (cell adhesion phase), but the Xa value was kept at approximately 5.0¥103 cells/cm2 from t=15 to 45 h (lag phase). Over t=45 h, Xa value exponentially increased and reached Xa=ca. 4.5¥104 cells/cm2 (exponential growth phase), followed by the cessation of increase in Xa value (stationary phase). The similar growth manners were also observed in the cultures with the other values of X0, and the time profiles were considered to consist of the four phases although the periods of phases varied each other depending on the inoculum sizes. 3.2 Model Estimation in Cultures of Adult and Neonatal Keratinocytes To evaluate the relationship between lag time (tL) and inoculum size (X0), the data shown in Fig. 4 were applied to the growth model. Here, the doubling time (td) was given as 64.5 h from separate experiments. The fitted lines of the time profiles in the cultures at various X0 values are indicated in Fig. 4. The calculated results could express the profiles depicting cell adhesion phase, lag phase, exponential growth phase, and stationary phase. Moreover, the evaluated value of tL is plotted against the logarithmic value of X0 as shown in Fig. 6. The tL value decreased linearly with an increase in the logarithmic X0 value: tL = b ln(X0/X0¢) + g

(4)

By using the growth model, and the parameters listed in Table 2, the growth index was calculated to clarify the effect of inoculum size on growth of adult

Fig. 5 Time profiles of concentration of adherent cells in culture of adult keratinocytes at various inoculum sizes. The lines show fitted values based on the two-dimensional cell placement model

Development of Culture Techniques of Keratinocytes for Skin Graft Production 147

148

M. Kino-oka · M. Taya

Fig. 6 Relationship between inoculum size and lag time. The lines indicate the fitted values using Eq. (4)

Table 2 Parameter values for adult and neonatal keratinocytes

Ratio of saturated value, a [–] Adherent constant, t [h] Lag time, tL [h] Doubling time, td [h]

Adult cells

Neonatal cells

0.55 4.0 44 65

0.81 2.1 0 31

keratinocytes. The growth index (Y) was defined as the ratio of adherent cell concentration at final culture time to inoculum size (Xaf/X0). In the cultures with X0=5.2¥102 to 5.2¥104 cells/cm2 for 144 h, the calculated Y value is indicated as the solid line in Fig. 7. At X0=1.7¥104 cells/cm2 the maximum Y value was obtained as Y=1.8 and the value was 2.5 and 1.9 times larger than those at X0=5.2¥102 and 5.2¥104 cells/cm2, respectively. Therefore, the inoculum size was considered to be an important factor to realize the effective growth in the monolayer culture of keratinocytes. Moreover, the calculated value of Y in Fig. 7 agreed closely with the experimental values (correlation coefficient: 0.96). To evaluate the kinetic parameters, the model was applied to cultures of neonatal keratinocyte cells according to the procedure described above. As shown in Table 2, the td value was 38.3 h which was 0.59 times lower than that in the case of adult cells, which indicates that neonatal cells possessed the higher growth potential than adult cells. On the other hand, the values of t and a of neonatal cells were 8.8 h and 0.81, respectively, showing slightly longer period for cell attachment and higher yield of cells on inoculation as compared with those of adult cells. In addition, as shown in Fig. 5, the low tL values at respective inoculum sizes of neonatal cells were obtained, suggesting that the

Development of Culture Techniques of Keratinocytes for Skin Graft Production

149

Fig. 7 Relationship between growth index and inoculum size. The line shows the calculated value of growth index at t=144 h on the basis of the two-dimensional cell placement model

adaptability of neonatal cells for new culture condition after inoculation (acclimation) is high and immediate initiation of cell propagation occurs. From these results, it was revealed that this model could be a useful tool for the expression concerning cell adhesion phase after inoculation, lag phase, exponential growth phase, and stationary phase owing to contact inhibition at high cell density, and it was also available to monolayer keratinocyte cultures with various inoculum sizes and cell lines derived from different donors. As summarized in Fig. 8, the kinetic parameters of t, tL, and td in the model present criteria for the potentials concerning attachment on culture surface, accli-

Fig. 8 Conceptual illustration showing relationship between cellular potentials and growth parameters evaluated by the two-dimensional cell placement model in culture of keratinocyte cells

150

M. Kino-oka · M. Taya

mation for new culture condition, and cell proliferation, respectively. Moreover, the lag time until the onset of cell division is correlated with inoculum cell size, based on the parameters of b and g determined for different cell lines. 3.3 Effect of Culture Surface on Attachment and Growth in Cultures of Various Cell Lines A variety of cell lineages derived from in vivo tissues and organs of humans have been established by the successive cell proliferation on surfaces improved in culture vessels. In most cultured tissues, several times of successive monolayer cultures are required to obtain the cells enough to reconstruct the tissues as final products. To enhance the potential of cell growth, the coating of cell attachment factors including extracellular matrixes on the culture surface has been accepted as an ordinary method. The cell adhesion onto culture surfaces mainly occurs through the cell attachment caused by physicochemical interaction just after inoculation and the subsequent extension of cells for intracellular reconstruction of actin orientation. A quantitative parameter to evaluate the culture properties should be proposed to optimize the operational factors controlling the interactions between cell and culture surfaces for individual cells from different tissue origins. To investigate the cell adhesion on culture surface, keratinocyte cells were cultivated on the culture surface coated with collagen at its concentration of C=0 and 5.8¥10–3 mg/cm2, and the cell attachment test was conducted in terms of Xa/X0 value [34]. The saturated profiles of cell attachment during cell adherent phase in both cultures with and without coating were observed. The improved attachment was found in the culture with coating, and the Xa/X0 value at t=2 h was achieved to be 0.8 which was twice as high as that in the culture without coating.As shown in Table 3, the t values in the cultures with and without coating were determined as t=2.1 and 7.4 h, respectively. To verify the broad availability of the proposed parameter, the cell attachment tests were applied to the cultures of fibroblasts and chondrocytes at C=0 and 5.8¥10–3 mg/cm2, respectively. In each test, the saturated profiles of Xa/X0 Table 3 Parameter values estimated in cultures of various cell lines with and without collagen coating

Species

C [mg/cm2]

t [h]

tL [h]

td [h]

Human keratinocytes Human keratinocytes Human fibroblasts Human fibroblasts Rabbit chondrocytes Rabbit chondrocytes

0 5.8¥10–3 0 5.8¥10–3 0 5.8¥10–3

7.4 2.1 3.9 0.4 15 2.9

25 20 30 20 28 23

31 20 17 21 14 9.6

Development of Culture Techniques of Keratinocytes for Skin Graft Production

151

were obtained, similarly to the case of keratinocytes and the t value in the culture of human fibroblasts and rabbit chondrocytes with collagen coating were 0.2 and 1.4 times lower than that of keratinocytes. As the growth tests, keratinocytes were cultivated for 96 h on the culture surfaces with and without coating. The active growth was observed in the culture with coating and Xa/X0 was 2.4 at t=96 h, which was 1.2 times higher than that in the culture without coating. In addition, during the exponential growth phase, the td value in the culture with coating was estimated as td=20 h, the value of which was 65% reduced in comparison with the culture without coating (Table 3), suggesting that the growth enhancement occurred with encouraging cell attachment and cell division. The collagen effect mentioned above was applied to the cultures of fibroblasts for 96 h and chondrocytes for 48 h, respectively. The values of td in the cultures with coating were determined as 21 h (fibroblasts) and 9.6 h (chondrocytes), the values of which were 66 and 69% reduced, respectively. From the results, both values of t and td in the cultures of keratinocytes, fibroblasts, and chondrocytes with coating were lower than those without coating though the values were at inherent levels for the respective cell lines. To investigate the influence of culture surface on cell attachment and growth in detail, the cultures of keratinocytes, fibroblasts, and chondrocytes at various levels of collagen coating in the range of C=5.8¥10–16 to 5.8¥10–3 mg/cm2 were conducted. The relative kinetic parameters of Rt, RtL, and Rtd denoted the ratios of t, tL, and td values at a given C value to those at C=0, respectively. In the case of keratinocytes, as shown in Fig. 9, with increasing C value over C=5.8¥10–10 mg/cm2, the values of Rt and Rtd decreased, achieving to be 0.3 and 0.7 at C=5.8¥10–3 mg/cm2, respectively, while RtL value was kept constant. Moreover, in the cases of fibroblasts and chondrocytes, the respective manners of kinetic parameters (Rt, RtL, and Rtd) against C value had good agreement with those of keratinocytes. In the culture of anchorage-dependent cells like keratinocytes, the cells after inoculation settle on surface of culture vessel and undergo attachment, which means surface recognition at interface between cytoplasmic membrane and culture surface during the adherent phase, following the cellular extension with the orientation of actin as intracellular immobilization. After the orientation of actin during the lag phase, cells start to make cell division as programmed by cell cycle, which means the entrance into the growth phase. Here, the generated daughter cells are accompanied with the cell attachment and extension for further cell division. Then, the extent of cell adhesiveness is considered to affect cell adhesion and doubling times, and thereby the overall growth potential of anchorage-dependent cells.

152

M. Kino-oka · M. Taya

Fig. 9 Plots of Rt, RtL, and Rtd values against collagen concentration in culture of various cell lines. Rt; relative value of time constant, t, RtL; relative value of lag time, tL, Rtd; relative value of doubling time, td

4 Cellular Senescence in Successive Culture of Keratinocytes In recent years much attention has been paid to the relationship between the morphological changes and the intracellular phenomena of cells such as orientation of cytoskeleton and nucleus. The cellular morphology is considered to include fundamental information of cellular potential. Computers can be a tool for extracting images and viewing the motion of objects in a visual way, and accomplish the conversion of optical images to digitized ones, which can be converted to the numeral variables or parameters expressing the time-elapsed changes of objects. The computer-aided systems have been applied to the analysis of cellular motions in mammalian cell cultures, although they are rather complex and still less developed. Some researchers analyzed the phenotypic features of human

Development of Culture Techniques of Keratinocytes for Skin Graft Production

153

cells as a sequence of intracellular metabolisms, and proposed parameters for characterizing the morphological and physiological conditions of cells. For the production of epithelial sheets, a series of passages is usually conducted in the monolayer cultures of keratinocytes for the expansion of cell number [35]. To achieve the demanded area of epithelial sheets, however, the finite number of cell division, which depends on the states of origin such as donor age and location of harvesting, becomes an obstacle to large expansion of cell number [36]. Some challenges to the problem have been done with evolutional techniques such as the enhancement of telomerase activity with genetical and biological modification [37]. However, the inaccessible regulation of the enzyme activity in vitro and in vivo still remains to overcome some break-through points for the clinical application as the commercialized products. Thus, for practical manufacturing with successive cultures of normal keratinocytes, their characterization based on cellular life span is very important to realize the stable production of epithelial sheets. 4.1 Image-Analyzing System The system flow diagram for the image analysis developed by us in keratinocyte culture is shown in Fig. 10 [38]. In the hardware unit, a microscope with CCD camera is used and the original image is captured with a video digitizer card in a personal computer. The numbers of pixels are 512 in horizontal and 582 in vertical directions, and each pixel has 256 classified gray levels ranging from 0 (black) to 255 (white). The captured area of an original image is 0.83 mm2. In the software unit, the confluence degree is determined by using LabVIEW software (National Instruments Corp.) with add-on image processing software IMAQ Vision. The original image is converted into the binary image (gray level of projected area=1 and gray level of non-projected area=0) through detecting the cell edge detection by the non-linear high pass filter and setting a intensity threshold. To remove noise from this binary image, the primary morphological transformation is processed by combination of erosion and dilation, and particle is omitted by area threshold (object area<2.2 mm2). The confluence degree, Cd, in the processed image was obtained after filling the holes inside the cell region: Number of pixels with gray level = 1 Cd = 9999997 Total number of pixels in the image

(5)

Figure 10 shows examples of original culture image, edge-clarified image, and final binary image obtained thorough the processing. The image processing method was proved to be a tool available for estimation of the Cd value. The values of Cd at several positions and their average values were determined in flask cultures of keratinocytes. The positional variations in the confluence degrees were negligible throughout the culture. The standard deviations

154

M. Kino-oka · M. Taya

Fig. 10 Schematic drawing of procedure for image analysis and images obtained through processing

of the positional variations and of a set of four flasks were comparable. As shown in Fig. 11, a linear relationship between average Cd value and cell concentration was obtained. The values of the slope and correlation coefficient in flask were determined by fitting the experimental data (correlation factor: 0.90). Using this calibration, on-line estimation of cell concentration was made, and calculated cell concentration agreed well with the experimental cell concentration in flask culture. Of course, to evaluate cell concentration of keratinocytes in culture vessel, several methods are available including measurement of MTT (3,(4,5-dimethylthiazol-2-yl)-2,5-diphenyl tetrazolium bromide) activity and cell number counting by hemocytometer [39]; however, none of them is suitable for non-destructive and on-line measurements. Image analysis was applied to measure morphology and cell size of anchorage-dependent cells on flat surfaces [40–42], and on microcarriers [43]. In the case of traditional culture of keratinocytes using murine fibroblast feeder layer, keratinocytes grow in the colonies, yielding relatively large

Development of Culture Techniques of Keratinocytes for Skin Graft Production

155

Fig. 11 Relationship between cell concentration and confluence degree

error in evaluation of confluence degree in the image area. Moreover, the feeder layer hinders the detection of keratinocytes. Use of serum-free medium without a feeder layer results in a more uniform cell distribution, excluding the abovementioned problems. However, the positional deviation for cell concentration in culture vessel becomes relative large when the small area of image is captured. The relative errors in the cell concentration measured by image analysis were slightly higher than the errors by cell counting [40]. Therefore the increase in the number of captured images at various positions will be a choice to reduce the deviation. 4.2 Evaluation of Cellular Senescence The influence of passage number on the cell attachment and subsequent growth was investigated in a series of the successive cultures of keratinocyte cells isolated from neonatal foreskin [44]. Figure 12 shows the time constant for

Fig. 12 Time constant for cell attachment and adherent cell concentration at 96 h of culture time in successive cultures of neonatal foreskin cells. The error bars show the standard deviations (SDs). N.D.=not determined

156

M. Kino-oka · M. Taya

Fig. 13 Time profiles of cumulative number of population doublings, doubling time and cell area in successive culture of keratinocyte cells isolated from neofetal foreskin (Run 1)

cell attachment, t, and adherent cell concentration at 96 h of culture time, Xa (t=96), in the respective passages. The t value was determined in the cell attachment tests conducted at the beginnings of the second to fifth passages. The Xa value was recorded as an average of the estimated number of adherent cells on the surfaces. With increasing passage number, NP, the t value increased and on the contrary, the Xa (t=96) value decreased. The t and Xa (t=96) values at NP=5 were three times and one-twentieth those at NP=2, respectively. These results suggest that the attachment and growth potentials of the examined cells were getting low as the passage progressed approaching the end in cellular life span. Figure 13 shows the plots of cumulative number of population doublings, Nd, specific growth rate, m, and cell area, AC, against cumulative culture time, t, in a series of the successive cultures of the keratinocyte cells isolated from neonatal foreskin (Run 1). Here, the area of a single cell, AC, was calculated as follows: AC = Cd/Xa

(5)

Development of Culture Techniques of Keratinocytes for Skin Graft Production

157

Fig. 14 Plots of specific growth rate and cell area against remaining number of population doublings in successive cultures of Runs 1–3. Closed keys show the data at critical points

The final value of Nd at cessation of cell division was NdF=12.9 at the final culture time of tF=536 h, and the profiles of m value could be apparently divided into two phases, showing a steep drop in the m values after t=343 h onwards. The values of m and Nd at the critical point of t *=343 h in the serial culture of Run 1 were N d* =10.6 and m*=2.3¥10–2 h–1, respectively. As opposed to the m value, a remarkable increase in the AC value was observed after the critical culture time. The profiles of the m and AC values were also examined in the successive cultures of keratinocyte cells isolated from donors 22 and 70 years old (Runs 2 and 3, respectively). The NdF values were 12.2 at tF =557 h (Run 2) and 9.9 at tF =523 h (Run 3), and the profiles of the Nd, m and AC values in these cultures were depicted in similar manners to those in Run 1. Two culture phases as seen in Run 1 were also determined by setting t *=281 and 189 h for Runs 2 and 3, respectively. The values of m* and AC* in Runs 1–3 were found to be at approximately same levels, though the N d* values were different from each other. To evaluate the cellular life span of keratinocytes, the values of (NdF–Nd) were employed as a principal parameter in the serial cultures of Runs 1–3.As shown in Fig. 14, in the case of Run 1, the opposite profiles of the m and AC values were depicted in terms of the (NdF–Nd) value. The notable changes in the m and AC values were observed below (NdF–Nd) value of 2.3, which corresponded to the value of (NdF–N d* ) at the critical point. Similar results were also recognized in the cultures of Runs 2 and 3, and the (NdF–Nd) values in Runs 1–3 overlapped each other, suggesting that the value of AC can be a possible variable to correlate the m value as growth potential with the (NdF–Nd) value as a parameter predicting the remaining life span of the cells.

158

M. Kino-oka · M. Taya

In general, the human skin cells like keratinocytes and fibroblasts cease their growth in the cultures after the limited number of population doublings. To understand the cause of growth cessation in senescent cells, Mondello et al. [45] examined the variations in telomere length with progress of life span by using twelve clones of fibroblasts. The cytogenetic analyses revealed the appearance of chromosome anomalies, mainly telomeric association and unbalanced translocations, in relation to the clones’ life spans, although the number of population doublings and time required for entering senescence varied among the clones. Moreover, Sayama et al. [46] suggested that the genes encoding p16, p21, and p53 proteins were responsible for the cellular senescence via the regulation of cyclindependent kinase (CdK) activities which govern the progress of cell cycle. It was also reported that the induction of p16 protein contributed to the cellular differentiation, which causes the development of cornified envelope conformation via expression of marker proteins for differentiation such as involucrin and keratin [47]. It is known that the enlargement of keratinocyte cells during differentiation occurs arising from the conformation of cornified envelope [48, 49]. From these facts, it is most likely that the increase in the number of population doublings induces the differentiation with progress of successive cultures of the keratinocyte cells (Runs 1–3), and that after critical points of t=t *, at which the N d* values reach 60–85% of the NdF value, the senescent cells are triggered to form cornified envelope, thereby accompanying the notable increase of cell area. From a practical standpoint of epithelial sheet production, the finite cell division that depends on the states of origins such as donor age and location of harvesting biopsy becomes an obstacle to acquisition of sufficient cells in the serial cultures. The inhomogeneity found in the culture process forces us to employ batch operation with less reproducibility, the so-called “tailor-made process”, and it is desired to estimate the potentials of population doublings in the serial cultures. In particular, a feasible method to evaluate cell states is needed to support the operators who have to judge cellular life span from information obtained through daily observation. The principal parameter of the remaining number of population doublings was proposed, being correlated with cell area and specific growth rate. Both variables of specific growth rate and cell area could be determined by means of the non-destructive and non-invasive method capturing the image of culture bottom surface. Figure 15 presents the rearranged relations among these parameters in the cultures of Runs 1–3. In this figure, the additional data are included, which were obtained from the successive cultures of the keratinocyte cells conducted apart from the cultures of Runs 1–3. The data were classified into three regions denoted as proliferative and senescent phases as well as transitional phase. The data belonging to NdF–Nd ≥3.0 crowded in the proliferative phase, and the senescent phase included the data with 0≤NdF–Nd<1.5, showing that the data in the proliferative phase shifted dispersively to the senescent phase via the transitional phase with decreasing (NdF–Nd) value. These trends suggest that the culture containing senescent cells directs toward an unstable culture phase, while approaching to the end of cellular life span.

Development of Culture Techniques of Keratinocytes for Skin Graft Production

159

Fig. 15 Relationship between specific growth rate and cell area in successive cultures of Runs 1–3. The arrows show the data at critical points

For the characterization of in vitro cultured keratinocytes, Green et al. [6] proposed three types of clones denoted as the holoclone which includes undifferentiated cells with high proliferative ability, the paraclone which undergoes limited mitotic cycles with differentiation, and the meroclone which is in transition of them. The trends shown in Fig. 15 supported the classification of keratinocyte cells into the three types of clones. This data map between specific growth rate and cell area can be available for predicting cellular senescence in the serial cultures for the production of epithelial sheets.

5 Designs of Culture Operation and Bioreactor for Epithelial Sheet Production The purpose of automated skin graft production is to reduce cost and time in the manufacturing process. This will ultimately increase the number of burn patients that will benefit from the technology. Reducing production costs will facilitate the treatment of relatively small burns and re-constructive plastic surgery with tissue engineered skin products on a commercial basis. In developing a reactor system, the on-line observation of cultured cells is on imperative element. Image analysis has provided a useful technique to support the monitoring of growth and morphology [40, 50, 51]. In particular, in the case of anchorage-dependent animal cells including keratinocytes, this tech-

160

M. Kino-oka · M. Taya

nique was one of the most promising methods for on-line measurement of cell density in a culture vessel [41]. 5.1 Bioreactor Accompanied with Image Analysis A bioreactor system with a microscopic video for on-line cell growth monitoring was constructed (named as KERATOR) [52]. Figure 16 shows the schematic drawing of the system, the main constituents of which are culture module and operation unit. The culture module consists of special growth chambers stacked together. Each chamber contains a polymeric support film of 660 cm2 size on which the keratinocytes attach and proliferate. This film was made from the same materials as that of bottom in a Petriperm culture flask. The culture module can be mounted on the stage with varied numbers of the growth chambers. In this way, up to 5280 cm2 of surface area, keratinocytes can be cultured in one operational run. The CCD camera with a specially developed optical system was attached to the stage for monitoring the culture growth in the KERATOR.With the current optics, the total observation area is 2.1 mm2 of the culture surface. Using a remote control unit, the CCD camera can be moved horizontally to choose various observation positions and vertically to adjust the image focus. Two observation positions can be in advance selected using electric microswitches. The computer-captured images were used for the image analysis as described above. CO2 concentration in the gas phase of the culture module was measured by a gas analyzer and was maintained at 5% CO2 by a mass flow controller. For medium change, the stage was tilted and the medium from the modules was sucked out by a pump. To fill up fresh medium again, the direction of fluid flow was switched by pinch valves and changing the pump direction. Software was used for the programming and all culture operations, including inoculation and medium change, were performed automatically by the operation unit controlled by the computer installed a data acquisition board. In the culture of keratinocytes in the system, the cell concentration was calculated by using mean confluence degrees (see Eq. 4) from different positions in the growth chamber bottom of the culture module. The relation between average confluence degree and cell concentration in the system was obtained in advance with independent experiments.As shown in Fig. 17, from the on-line monitoring of growth, the specific growth rate (m) could be calculated throughout the culture without any sampling of the cells and three stages of growth kinetics were identified. During early culture time from 0 h to 22 h, the specific growth rate was very low or zero due to growth lag and cell adhesion on the film surface (i.e., lag phase). From t=22 to 94 h, the cells started to grow and the m value increased. After 94 h, the value of m maintained at high levels ranging from 7.6¥10–3 to 1.2¥10–2 h–1, which was almost equal to that in the control culture with the Petriperm (i.e., exponential growth phase), showing that the bioreactor culture operations have no adverse effects on cell vitality.

Development of Culture Techniques of Keratinocytes for Skin Graft Production

161

Fig. 16A–C Schematic drawings of: A bioreactor system; B culture module; C overview photograph. 1 Culture module, 2 Personal computer with data acquisition system, 3 Gas inlet, 4 Gas outlet, 5 Medium port, 6 Humidifier, 7 Mass flow controller, 8 CO2 cylinder, 9 Air cylinder, 10 Temperature equilibration of fresh medium, 11 Fresh medium reservoir, 12 Waste medium reservoir, 13 Seeding bottle, 14 Refrigerator, 15 Multichannel peristaltic pump, 16 Computer automated pinch valve, 17 Filter, 18 Medium and gas lines, 19 Control line, 20 Growth chamber, 21 Hydrophilic film, 22 Module stage

162

M. Kino-oka · M. Taya

Fig. 17 Cultures of keratinocytes in bioreactor and Petriperm. Close and open symbols show experimental and calculated values, respectively. The arrows show the time of medium change

Several designs of culture vessels for the culture of keratinocytes were presented in the past [53]. However, they are designed only in terms of configurations and offer few concepts for an entire culture system including control and operation strategies. The on-line monitoring technique proposed can cope flexibly with the quantitative evaluation of cell concentration and average confluence degree in each cell line (or each patient) even if the required cell size depends on the cell lines. In a clinical aspect, moreover, it may be possible to use the specific growth rate as an indicator of wound healing impact because keratinocytes from different patients have their own growth rates, namely cell vitality or proliferation ability. Furthermore, the on-line monitoring of the specific growth rate can be useful to predict the operations in bioreactor such as time points of medium change and termination of the culture.

Development of Culture Techniques of Keratinocytes for Skin Graft Production

163

5.2 Automated Operation for Successive Cultures The operations of cell detachment and attachment during successive cultures are the most troublesome for the construction of automated bioreactor system because most of employed procedures include the centrifugation step which is a big obstacle against the automation. The treatment of cells with trypsin and/or trypsin inhibitor was proposed to skip the centrifugation in the successive cultures of human keratinocytes with serum-free medium [54]. Figure 18 shows the comparison of cell attachment and growth of keratinocytes in various medium containing trypsin and/or trypsin inhibitor. Although the cells inoculated in the trypsin-containing medium suffered serious suppression in

Fig. 18 Influence of trypsin and/or trypsin inhibitor treatment on cell growth for t=24 and 48 h. N.D.: not detectable (Xa(t)<3¥102 cells/cm2). The error bars show the standard deviations (SDs)

164

M. Kino-oka · M. Taya

Fig. 19 Time profiles of concentration of adherent cells in first and second passages during successive cultures (X0,1=1.0¥104 cells/cm2 and tm=12, 24, 48 h)

the potentials of adhesion to culture surface and subsequent growth at early culture phase, the coexistence of trypsin inhibitor with trypsin in the medium resulted in the potential of cellular adhesion similar to that in the conventional medium without the treatment agents. Concerning the delay in cell growth, the cells treated with trypsin and its inhibitor showed doubling time of 23 h, the value of which was comparable to that of the untreated cells, in exponential growth phase after replacing the medium with conventional one. This indicated that the potential of cell growth was recovered by the medium

Development of Culture Techniques of Keratinocytes for Skin Graft Production

165

Fig. 20 Procedures of successive cultures accompanied by trypsin and trypsin inhibitor treatment with and without centrifugation

exchange to remove trypsin and its inhibitor. In the cultures with trypsin and its inhibitor, it was found that adherent cell concentration at 96 h was maximized when time of first medium exchange (tm) was 24 h, and that prolonged tm value caused an increase in lag time until cell division in a linear manner. The profile of cell growth was expressed employing a cell placement model in the successive cultures comprising of the first and second passages, and final cell concentration in the first passage and tm value in the second passage were determined so as to give an enhanced cell production rate under the examined conditions. Based on the calculated results, as shown in Fig. 19, the successive cultures were performed by means of subculture operation with the trypsin and trypsin inhibitor treatment, and the cell production rate could be obtained at the same level as the culture associated with centrifugation. From the results described above, we proposed the novel strategy for successive culture, as shown in Fig. 20, where the centrifugation step is skipped by regulation of cell detachment and attachment through separate additions of trypsin and trypsin inhibitor. Then, the bioreactor system, as shown in Fig. 21, was designed for the automated operations of successive cultures. In addition, on-line estimation of cell concentration was feasible during the multilayer culture of keratinocytes in the bioreactor by analyzing medium component. The successive cultures for the formation of epithelial sheets (including monolayer and multilayer cultures) in the bioreactor system offered the culture performance comparable to that in the conventional cultures accompanied with centrifugation.

166

M. Kino-oka · M. Taya

Fig. 21A–C Photographs of bioreactor system for successive cultures of keratinocytes: A top view of bioreactor; B vessel with two culture surfaces; C bottom view of bioreactor

Development of Culture Techniques of Keratinocytes for Skin Graft Production

167

6 Summary The advances in tissue engineering have encouraged the clinical application of skin grafts to wound healing, several companies now supplying the productions on a commercial level. From an engineering aspect, however, the culture processes for these products are rather domestic and associated with manual operations. In this context, a growth model of cell placement was proposed to express the profiles of keratinocyte cultures including cell adhesion phase after inoculation, lag phase, exponential growth phase, and stationary phase owing to contact inhibition at high cell density. Based on the model, the optimal inoculum size could be determined to realize the efficient growth of keratinocytes. In addition, model consideration revealed that the culture with collagen coating exhibited the enhancement of cell attachment and subsequent cell growth of keratinocytes as compared to the culture on the surface without collagen coating. A monitoring system was developed using a computer-aided image analysis, and this system realized the non-disruptive measurements of growth rate and confluence degree in real time in the cultures of keratinocytes. Based on the techniques, the evaluation of lifespan in the serial cultures of keratinocytes was constructed, and the data map was presented in terms of specific growth rate and cell area, which could be classified into three regions of proliferative, senescent and transitional phases. This map must support the operators who have to judge cellular life span from information obtained through daily observation. The monitoring system was applied to the culture in the bioreactor. An autonomous modular bioreactor, KERATOR, for large scale production of skin grafts up to 5280 cm2 size was constructed. It is computer controlled and the operations, such as cell seeding and medium change, are automated by incorporating a special microscopic video with a CCD camera into the reactor system. The culture conditions of keratinocytes with trypsin and/or trypsin inhibitor treatments were determined to achieve subculture operation without centrifugation in the successive passages. Although the cells inoculated in the trypsin-containing medium suffered suppression in the potentials of adhesion to culture surface and subsequent growth at early culture phase, the coexistence of trypsin inhibitor and trypsin in the medium led to the culture result similar to that in the medium without the treatment agents. Then, the automated bioreactor system was applied to the successive cultures of keratinocytes without centrifugation, and the culture performance was comparable to that in the conventional successive cultures with centrifugation. The bioreactor system will lead to the minimized risks of human errors and biological contamination as well as the reduced production costs, and ultimately the increased number of patients who enjoy a benefit from tissue engineered skin grafts. Biochemical engineering can contribute to the system development through the integration of the scientific and technological knowledge concerning ma-

168

M. Kino-oka · M. Taya

nipulation (culture operation including bioreactor), measurement (monitoring and estimation of culture process), and prediction (simulation based on growth kinetics). The integrated system enables us to establish the stable and reliable production process of tissue engineered skin grafts, and in a prospective aspect the culture system will be required to offer the standards for both quality of the products and curing outcomes after transplantation of the products.

References 1. Sipe JD, Kelley CA, McNichol LA (2002) (eds) Reparative medicine; growing tissue and organs. Ann NY Acad Sci, vol 961. New York Academy of Sciences, New York 2. McIntire LV (2003) Tissue Eng 9:3 3. Lysaght MJ, Reyes J (2001) Tissue Eng 7:485 4. Lalan S, Pomerantseva I, Vacanti JP (2001) World J Surg 25:1458 5. Bassett P (1999) Tissue engineering. Market & Drug Development Publication, Westborough 6. Green H, Kehinde O, Thomas J (1979) Proc Natl Acad Sci 76:5665 7. Barnes D (1987) Biotechniques 5:534 8. Brittberg M, Lindahl A, Nilsson A, Ohlsson C, Isaksson O, Peterson L (1994) N Engl J Med 331:889 9. Temenoff JS, Mikos AG (2000) Biomaterials 21:431 10. Rheinwalt JG, Green H (1975) Cell 6:331 11. O’Connor NE, Mulliken JB, Banks-Schlegel S, Kehinde O, Green H (1981) Lancet 1:75 12. Hefton JM, Caldwell D, Biozes DG (1986) Am Acad Dermatol 14:399 13. Munster AM (1996) Ann Surg 224:372 14. Woodley DT, Peterson HD, Herzog SR, Stricklin GP, Burgeson RE, Briggaman RA, Cronce DJ, O’Keefe EJ (1988) J Am Med Assoc 259:2566 15. Sheridan RL, Tompkins RG (1995) J Trauma 38:48 16. Maciag T, Nemore RE, Weinstein R, Gilchrest BA (1981) Science 211:1452 17. Lechner JF, LaVeck MA (1985) J Tissue Cult Methods 9:43 18. Boyce ST, Ham RG (1985) J Tissue Cult Methods 9:83 19. Yamato M, Utsumi M, Kushida A, Konno C, Kikuchi A, Okano T (2001) Tissue Eng 7:473 20. Ito A, Takizawwa Y, Shinkai M, Honda H, Hata K, Ueda M, Kuno N, Itakura A, Kobayshi T (2003) J Biosci Bioeng 95: 21. Boyce ST, Warden GD (2002) Am J Surg 183:445 22. Harding KG, Morris HL, Patel GK (2002) BMJ 324:160 23. Gentzkow GD, Iwasaki SD, Hershon KS, Mengel M, Prendergast JJ, Ricotta JJ, Steed DP, Lipkin S (1996) Diabetes Care 19:350 24. Noordenbos J, Dore C, Hansbrough JF (1999) J Burn Care Rehabit 20:275 25. Carsin H,Ainaud P, LeBever H, Rives J, Lakhel A, Stephanazzi J, Lambert F, Perrot J (2000) Burns 26:379 26. Wright KA, Nadire KB, Busto P, Tubo R, McPherson JM,Wentworth BM (1998) Burns 24:7 27. Fang CH, Alexander JW, MacMillan BG (1982) J Burn Care Rehabit 3:164 28. Wainwright DJ (1995) Burns 21:243 29. Kremer M, Lang E, Berger AC (2000) Br J Plast Surg 53:459 30. Guerret S, Govignon E, Hartmann DJ, Ronfarod V (2003) Wound Repair Regeneration 11:35 31. Bello YM, Falabella AF, Eaglstein WH (2001) Am J Clin Dermatol 2:305

Development of Culture Techniques of Keratinocytes for Skin Graft Production 32. 33. 34. 35. 36. 37. 38. 39. 40. 41. 42. 43. 44. 45. 46. 47. 48. 49. 50. 51. 52. 53. 54.

169

Miura T, Shiota K (2000) Anat Rec 258:100 Kino-oka M, Umegaki R, Taya M, Tone S, Prenosil JE (2000) J Biosci Bioeng 89:285 Yashiki S, Umegaki R, Kino-oka M, Taya M (2001) J Biosci Bioeng 92:385 Eric WJ, Susan FM, Christopher JR, Nancy LP (1992) In Vitro Cell Dev Biol 28A:429 Boisseau AM, Donatien P, Surléve-Bazeille JE, Amédée J, Harmand MF, Bézian JH, Maleville J, Taieb A (1992) J Dermatol Sci 3:111 Bodnar AG, Ouellette M, Frolkis M, Holt SE, Chiu CP, Morin GB, Harley CB, Shay JW, Lichtsteiner S, Wright WE (1998) Science 279:349 Kino-oka M, Prenosil JE (2000) Biotechnol Bioeng 67:234 Freshney RI (1994) Culture of animal cells, 3rd edn. Wiley-Liss, New York Tucker KG, Chalder S, Al-Rubeai M, Thomas CR (1994) Enzyme Microb Technol 16:29 Ruaan RC, Tsai GJ, Tsao GT (1993) Biotechnol Bioeng 41:380 Stachecki JJ, Yelian FD, Leach RE, Armant DR (1994) J Reprod Ferti 101:611 Pons MN, Wagner A, Vivier H, Marc A (1992) Biotechnol Bioeng 40:187 Umegaki R, Murai K, Kino-oka M, Taya M (2002) J Biosci Bioeng 94:231 Mondello C, Riboni R, Casati A, Nardo T, Nuzzo F (1997) Exp Cell Res 236:38 Sayama K, Shirakata Y, Midorikawa K, Hanakawa Y, Hashimoto K (1999) J Cell Physiol 179:40 Lee G, Park BS, Han SE, O JE, You YO, Baek JH, Kim GS, Min BM (2000) Arch Oral Biol 45:809 Sun TT, Green H (1976) Cell 9:511 Watt FM, Green H (1981) J Cell Biol 90:738 Treskatis SK, Orgeldinger V, Wolf H, Gilles ED (1997) Biotechnol Bioeng 53:191 Hirai H, Umegaki R, Kino-oka M, Taya M (2002) J Biosci Bioeng 94:351 Kino-oka M, Prenosil JE (1998) J Chem Eng Japan 31:856 Lyman G, Mathus G, Root D (1995) U.S. Patent No. 5466602 Kino-oka M, Umegaki R, Ogawa N, Taya M (2002) Tissue Eng 8:1156

Received: September 2003

Adv Biochem Engin/Biotechnol (2004) 91: 171– 189 DOI 10.1007/b94209 © Springer-Verlag Berlin Heidelberg 2004

Transgenic Birds for the Production of Recombinant Proteins Masamichi Kamihira (✉) · Ken-ichi Nishijima · Shinji Iijima Department of Biotechnology, School of Engineering, Nagoya University, Furo-cho, Chikusa-ku, Nagoya 464-8603, Japan [email protected]

1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 172 2 Bird Embryo Culture

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 174

3 Non-Viral Gene Transfer for Animal Cells and Bird Embryos Using DDAB Lipid Vesicles . . . . . . . . . . . . . . . . . . . . . . . . . . . . 178 4 Generation of Transgenic Birds Using Retroviral Vectors . . . . . . . . . . . . 182 5 Concluding Remarks

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 186

References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 187

Abstract Transgenic birds were expected to be an excellent transgenic bioreactor for the production of recombinant pharmaceutical proteins. However, the only successful transgenic bioreactors have been based on mammals. We have developed two key techniques for obtaining transgenic birds. For bird embryo culture, we identified that the low rate of hatchability of cultured embryos is caused by limited oxygen and calcium availability. In quail embryo culture using a chicken eggshell as a culture vessel, hatchability increased to 80% by the supplement of calcium lactate in addition to oxygen aeration. A fully artificial vessel for quail embryo culture using a gas-permeable Teflon membrane was also designed.Although the hatchability was lower than that of cultures using a surrogate eggshell, we succeeded in hatching of bird embryos using a fully artificial vessel. For transgene introduction, a replication-defective pantropic retroviral vector based on Moloney murine leukemia virus (MoMLV) pseudotyped with vesicular stomatitis virus G protein (VSV-G) was injected to laid embryos at the blastodermal stage, and the embryos were hatched in vitro to generate G0 birds. The viral vector sequence was detected in the tissues of all G0 birds. The germ-line transmission efficiency was more than 80%. Plural copies of the transgene were inserted into the genome of G1 transgenic progeny. Keywords Transgenic bioreactor · Chicken · Embryo culture · Retroviral vector · Recombinant protein

172

M. Kamihira et al.

1 Introduction With the advance in recombinant DNA technology, many therapeutic proteins were successfully produced by bacterial fermentation and animal cell culture using recombinant cells. Recently, transgenic animals such as goat and sheep were engineered to produce large quantities of recombinant proteins in their milk [1–3], and the idea is known as transgenic bioreactor. The main advantage of the system is high productivity. It was reported that 35 mg/ml of alpha1antitrypsin was produced in the milk of the transgenic sheep [1]. In addition, transgenic bioreactor offers a renewable source of highly modified proteins that cannot be produced by microorganisms. However, long breeding time-lines and vast area to maintain flocks are required in such mammals. To overcome the disadvantages, bird species such as chicken and quail have long held promise as alternative transgenic bioreactors [4, 5]. In the case of sheep, it takes 5 and 8 months for gestation and sexual maturation, respectively, while chickens hatch 20 days after oviposition and need 4–6 months for sexual maturation. Furthermore, a typical chicken egg white contains 3–5 g of protein, more than

Fig. 1 Egg formation in hens

Transgenic Birds for the Production of Recombinant Proteins

173

half of which is ovalbumin. Therefore, if ovalbumin in egg white can be partially replaced with a target protein by expressing the transgene in the oviduct of hens, high-yield and low-cost production can be expected. For the realization of this idea, there have been several obstacles including the inefficiency of gene transfer and germ-line transmission, the low hatchability of manipulated embryos, and the absence of an effective transgene expression system. The difficulty in the establishment of transgenic procedures for birds is also due to the complex process of egg formation in hens (Fig. 1). Gene transfer into the avian genome was first achieved in the freshly laid fertilized egg (stage X; blastoderm). Blastodermal cells are pluripotent and contains the precursor cells giving rise to the germline.Although blastodermal stage embryos are available in large quantity, the embryo already develops to ~60,000 cells. Thus, it is necessary to introduce a transgene into the large number of embryonic cells. For this purpose, bird-derived retroviral vectors have been frequently used for gene transfer. In this case, since the chickens manipulated are mosaic for the transgene, hemizygotes are obtained after mating (Fig. 2). However, the germ-line transmission efficiency of transgene was less than 10% possibly due to low viral titer. In birds, the new technologies for embryo manipulation including embryonic stem (ES) cell and cloning by nuclear transfer have not been established. Besides retroviral gene transfer to blastoderm, microinjection of transgene into singlecell stage embryos has been carried out to generate transgenic chickens. However, many problems to be solved exist. For instance, pronuclei cannot be readily visualized in the chicken single cell zygote because of underlying yolk and the opaqueness of the vitelline membrane. In addition, to obtain the freshly fertilized single cell stage embryo in magnum, the donor hen is usually sacrificed. The manipulated embryos must be cultured to hatch by ex vivo culture system using surrogate eggshells. Furthermore, the frequency of transgene insertion into the genome and germ-line transmission efficiency were also very low.

Fig. 2 Procedure for generation of transgenic birds

174

M. Kamihira et al.

Thus, it is necessary to develop robust and effective techniques to generate transgenic birds. Here we describe our approaches to establish a transgenic bird bioreactor for the production of recombinant pharmaceutical proteins.

2 Bird Embryo Culture For the generation of transgenic birds, an eggshell is bored and resealed for bird embryo manipulation, and manipulated embryos must be incubated or cultured to hatch. Hatchability decreases to less than 10% during incubation. Therefore, the development of an effective in vitro culturing method for bird embryos has been studied by several research groups [6–13]. Perry reported an in vitro culturing method for chicken embryos from the single-cell stage using a recipient eggshell as a culture vessel in her pioneering work [8]. According to this method, the culture was divided into three steps (Systems I–III; Fig. 3). First, a fertilized single-cell stage embryo containing yolk and dense albumen and lacking a shell was taken from the oviduct of a hen, and statically incubated in a glass vessel for 24 h to form a blastoderm (System I). The contents were transferred to a recipient eggshell with thin albumen and incubated with rocking until the 2–3 days stage (System II). Finally the contents were transferred to a large recipient eggshell leaving air space above the embryo and incubated until hatching (System III). The development of Perry’s in vitro culture method enabled a bird embryo to be manipulated from the singlecell stage. Using this method, an embryo can mature to hatching despite removal of its contents from the eggshell.When Perry developed the surrogated shell culture system, hatchability was 7%. Naito et al. [9] improved the method, and reported the hatchability up to 40–50%. They pointed out the importance

Fig. 3 Ex vivo culture method for chicken embryos

Transgenic Birds for the Production of Recombinant Proteins

175

of the position of embryo during culture in System II and III, and removed the thick albumen to keep the embryo at the top of yolk. However, viability still dropped in System III. Since more than 90% of fertilized eggs hatch under normal conditions with rocking, it was understood that some unknown but important conditions for hatching in culture still remained for adjustment. In fact, we could not obtain hatched quail embryos under the culture condition according to Perry’s method using a chicken eggshell, although Ono et al. [11, 12] reported 20–50% hatchability using the same method. In order to clarify this uncertainty, we measured the concentrations of cations such as Ca2+, Mg2+, K+, and Na+, and metabolites such as lactate, ammonia, urea, and uric acid in yolk, albumen, amniotic and allantoic fluids of developing embryos from normal incubation and Perry’s culture method. We observed substantial differences in lactate concentration in amniotic and allantoic fluids. Most of the dead embryos in Perry’s culture method accumulated large amounts of lactate in these fluids, whereas the lactate concentration was very low in normal controls. Another difference was observed in Ca2+ and Mg2+ concentrations in yolk. In normal incubation, the concentrations of these cations were approximately 800 ppm and 75 ppm, respectively. However, the concentrations decreased to 100–200 ppm and 30–35 ppm, respectively, at 5–7 days in cultured embryos. Although the concentrations of these cations in yolk recovered to the normal levels after the chorio-allantoic membrane extended inside of the eggshell, the deficiency of these cations in the early stage of embryonic development may have retarded growth in the late stage of development. The accumulation of lactate in cultured embryos strongly suggested the deficiency of oxygen in the embryo. Thus, we applied oxygen aeration to the quail embryo culture using a recipient chicken eggshell [13]. By an incremental increase in oxygen aeration, the cultured embryos were led to hatch with 35% hatchability (Fig. 4). As mentioned above, Ono et al. achieved relatively high hatchability in quail embryo culture using a recipient chicken eggshell without oxygen aeration. Therefore, we assume that some culture conditions such as thickness of the recipient chicken eggshell and the sealing procedure of the shell may have been different from ours. We have also experienced that the viability of embryo during the culture was dependent on the size of recipient eggshell. Differences among the recipient eggshells sometimes caused uncertainty in hatchability. In Perry’s embryo culture method for chicken, the use of a large recipient eggshell such as a shell from a double yolk egg was essential in System III. This again suggests that the improvement of oxygen supply through the eggshell is a very important factor for in vitro culture, and that the lot differences among recipient eggshells is a serious problem. Oxygen aeration can prevent the deficiency of oxygen caused by uncertainties in the recipient eggshell. Although we could consistently attain 35% hatchability by oxygen aeration, a decrease in viability was observed in the late stage of cultivation. We thought that a deficiency of minerals such as calcium and magnesium might be another

176

M. Kamihira et al.

Fig. 4 Quail embryonic development during embryo culture using surrogate eggshell (System III)

limiting factor. In addition to oxygen aeration, adding fine eggshell powder and calcium lactate as mineral sources to the albumen, dramatically improved hatchability (more than 80%) in the surrogated eggshell culture. These factors are closely related to the functions of the eggshell. A bird embryo is highly dependent on the eggshell during its development not only for protection from physical stress and microbial infection, but also for gas exchange, control of water content, and the supply of minerals (Fig. 5). The eggshell normally supplies 80% of the embryo’s calcium requirements [7]. In particular, the chorio-allantoic membrane, which lines the inside of the eggshell from days 7 to 12, is a very important membrane system directly related to gas exchange and calcium uptake from the eggshell. To achieve embryonic hatching, normal development of the membrane system is essential. We also succeeded in hatching quail embryos using an artificial vessel [13]. Many researchers have unsuccessfully attempted shell-less culturing. One reason we were able to achieve consistent hatchability was the use of a porous PTFE membrane. The porous PTFE membrane enabled gas exchange in every direction. Oxygen supply, a serious problem in surrogated eggshell culture, appeared to be sufficient without oxygen aeration in consideration of the low lactate concentration in the amniotic and allantoic fluids. Moreover, the hydrophobic surface of the Teflon membrane was suitable for normal development of a chorio-allantoic membrane to encapsulate the embryo during embryonic development. For this reason, the viability of embryo was 100% until 12 days.

Transgenic Birds for the Production of Recombinant Proteins

Fig. 5 Schematic drawing of developing embryo inside eggshell

Fig. 6 Chicken embryo culture using artificial vessel

177

178

M. Kamihira et al.

In terms of gas exchange and normal development of chorio-allantoic membrane, the PTFE membrane was a suitable substitute for the eggshell. However, viability sharply decreased after 14 days, and hatching could not be observed using the artificial vessel without mineral supplementation because the artificial vessel cannot supply minerals. By the addition of fine eggshell powder and calcium lactate, a certain percentage of the embryos hatched. However, in consideration of the average weight and size of the hatched embryos, the calcium supply did not appear to be sufficient. This may also cause the relatively low hatchability in the artificial vessel culture. Nevertheless, hatchability of 43% in shell-less culture is notable. The culture method can be applied for chicken embryo after improving the culture vessel. We also succeeded in embryonic chicken hatching using the artificial vessel (Fig. 6). The shell-less and surrogated culture methods will contribute to basic research in embryonic biology and to the study of transgenic birds.

3 Non-Viral Gene Transfer for Animal Cells and Bird Embryos Using DDAB Lipid Vesicles Since the first development in the late 1980s, cationic lipids have been widely used as a transfection reagent because of several advantages [14–18]. For example, it is safer and simpler than virus-mediated transfection [19–21], less toxic, and gave higher transfection efficiency than other non-viral techniques such as electroporation and calcium phosphate precipitation. However, serum-containing medium significantly impairs the transfection efficiency in cationic lipid-mediated transfection [22–25]. This makes it difficult to apply the method for in vivo gene transfer. To overcome this problem, some lipids or non-lipid chemical substances have been developed [26–32]. A cationic lipid reagent composed of 1,2-dioleoyl-sn-glycero-3-trimethylammonium-propane (DOTAP) mixed with cholesterol achieved a high level of in vitro transfection efficiency in the presence of serum [26]. Addition of polybrene to the medium also improved the decrease in transfection efficiency caused by serum using lipofection [27].A polyamine polymer with a glucose backbone, as an example of non-lipid reagent, gave a higher level of reporter gene expression in vitro and in vivo than 2,3-dioleyloxy-N-[2(sperminecarboxamido)ethyl]-N,N-dimethyl1-propanaminium trifluoroacetate (DOSPA) based cationic lipid [28]. There are many obstructive critical steps in exogenous gene expression (Fig. 7); entry into cytoplasm, degradation in endosome compartment, and transfer to nucleus. Most of the transfection reagents including those described above were only designed to work effectively at the entry into cytoplasm. However, it is necessary to treat the whole obstructive steps for the development of non-viral gene vehicle. Viruses can efficiently deliver their genes from the entry into a cell membrane to a nucleus and sometimes integrate them into a host genome. Capsid-

Transgenic Birds for the Production of Recombinant Proteins

179

Fig. 7 Critical steps in exogenous gene expression using DNA vectors

coated viral genome escapes from the endosome degradation. Viral matrix attachment protein, which includes a nuclear localization signal, enables a lentivirus to pass through the nuclear membrane. Retroviral genomes can be easily integrated into host chromosome by viral integrase.Viruses have evolved the specialized system for each obstructive step of gene transfer. Thus, it is interesting to incorporate a virus-mimic mechanism into non-viral gene transfer. Based on this concept, we developed lipid-based transfection methods [33–36]. In these methods, lipid vesicles composed of dimethyldioctadecyl ammonium bromide (DDAB) and Tween 80 could efficiently introduced foreign DNA into cultured cell lines [33]. To escape from the degradation in cytoplasm and to make it ease to pass through nuclear membrane, a sperm protein protamine was added to a lipid-DNA complex [34]. Protamine is highly basic and strongly binds to negatively charged DNA resulting in the protection of the delivered DNA from degradation by DNase. Furthermore, a nuclear localization signal of protamine make it possible for the foreign DNA to pass through a nuclear membrane. We also reported the incorporation of a retroviral integrase in DDAB lipid vesicles-mediated gene transfer to enhance the integration of transferred DNA into a host chromosome [35]. By this method, stable transfection efficiency was increased up to 15 times compared with DDAB lipid vesicles alone. Although these DDAB-based transfection methods mimic viruses in some points and are superior to the first generation of cationic lipid-mediated gene transfer, the transfection efficiency drastically decreased in the presence of

180

M. Kamihira et al.

serum in culture medium. Therefore, protamine was covalently bound to stearic acid, as the next approach, and the stearyl protamine was incorporated into the lipid vesicles to make stable complexes between DNA and DDAB lipid vesicles (Fig. 8) [36]. The protamine-modified DDAB lipid vesicles enabled efficient gene delivery in a serum-containing medium up to 10% serum. Comparative study using other cationic lipid reagents has revealed the effectiveness of protaminemodified lipid vesicles. In addition, in vivo gene transfer to hearts of quail embryos using protamine-modified lipid vesicles was successfully achieved. The reason why protamine-modified vesicles effectively worked in the presence of serum is not clear. A possible mechanism is that the stable association of the lipid vesicles with DNA was achieved by the protamine-modification of the lipid vesicles. It is well known that protamine is a highly basic protein and strongly binds DNA. Thus, the protamine-modification of lipid vesicles may cause increased +/– charge ratio and lead to stabilization of DNA/lipid vesicle complexes, which cannot be achieved by the simple addition of protamine in the presence of serum. Several studies using other cationic lipids support this hypothesis [27, 31]. In the case of DC-chol liposome, constituted with 3b-(N((N¢, N¢-dimethylamino) ethane) carbamoyl)-cholesterol and DOPE, the maximal transfection efficiency was achieved at a charge ratio of 2.0 in the absence of serum, whereas that of 4.0 gave the maximal efficiency in the presence of serum [31]. Addition of a detergent (1% octylglucoside) at the formation of (DOSPA/DOPE)/DNA complex to stabilize the complex improved the transfection in the presence of serum [32]. Moreover, several serum components that have inhibitory effects on transfection were identified [37]. For instance, oleic acid and heparin contained in serum displaced DNA from lipid/DNA complexes, and interfered with gene delivery. These studies suggested again that serumresistant ability of protamine-modified lipid vesicles is attributable to stabilization of lipid/DNA complexes caused by the high affinity of protamine with DNA. Protamine-modified lipid vesicles also enhanced the viral transduction efficiency. Addition of protamine-modified lipid vesicles to pantropic retroviral particles increased the viral titer 12-fold. One of important steps for viral

Fig. 8 Preparation of protamine-modified DDAB lipid vesicles

Transgenic Birds for the Production of Recombinant Proteins

181

infection is the attachment of viral particle to cellular membrane. It is known that polybrene mediates the association between viral particle and cellular membrane, which results in the enhancement of infection efficiency [38, 39]. Hodgson et al. showed that DOSPA/DOPE also increased viral transduction in an amphotropic retroviral system [40]. Reduction of negative charge repulsion between the cell and viral membrane phospholipids, which facilitates interaction of the envelope glycoprotein and cell surface, was a possible mechanism for the enhancement of the virus transduction by DOSPA/DOPE. In the case of protamine-modified lipid vesicle, similar mechanism can be assumed, although the enhancement of pantropic retroviral transduction was twofold higher than that achieved by DOSPA/DOPE. Retroviral infection for cultured cells is usually performed in a serum-containing medium. Since a protamine-modified lipid vesicle is highly positively-charged and stable in the presence of serum as indicated in the DNA transfection study, it could effectively mediate virus-cell interaction in the presence of serum and enhance the retroviral transduction. A lipid that mediates gene transfer in the presence of serum in vitro is not always effective in in vivo gene transfer. This makes the development of the gene transfer method more complicated. In the gene transfection study using a series of amphiphilic imidazolinium compounds which differ only in the structure of the hydrophobic acyl-chains, the myristoyl-form was the most effective in vitro, whereas the oleoyl-form gave the best results in vivo [30]. Protamine-modified lipid vesicles exhibited excellent transfection ability for in vivo gene delivery to quail embryos. In the study of transgenic birds, replication competent retroviruses have been often used, although it possesses some problems in terms of safety and reproducibility [41, 42]. To avoid these problems, several researchers attempted transfection of bird embryos using cationic lipid reagents [43–46]. However, faint or mosaic-like expression of reporter gene was observed in the previous methods. On the other hand, gross staining of the reporter gene expression was observed in hearts by the injection of protamine-modified lipid vesicle/DNA complexes into hearts of quail embryos (Fig. 9). Thus, this method may be used for site specific exogenous gene expression in vivo by direct injection. The gene transfer using protamine-modified lipid vesicles effectively utilized the function of protamine. As described in the previous studies, protamine condenses DNA, prevents DNA from DNase degradation, and enforces nuclear transport. By the addition of protamine, transfection efficiency was improved in cationic lipid vesicle-mediated gene transfer [34, 47]. Protamine might exhibit similar effects since protamine-modified lipid vesicles gave almost the same transfection efficiency as protamine-added lipid vesicles on serum-free condition. Furthermore, the gene vehicle acquired serum durability by the incorporation of protamine. In our studies, DDAB lipid vesicles were conjugated to cell specific ligands to improve the specificity [48], and also combined with viral integrase to enhance genome integration of transferred gene [35]. In the future, gene transfer using protamine-modified lipid vesicles might be established as a safe and efficient gene delivery by incorporating the other virus-mimic functions.

M. Kamihira et al.

Fig. 9 In vivo transfection into heart of quail embryos

182

4 Generation of Transgenic Birds Using Retroviral Vectors The generation of transgenic chickens has been reported by several groups [49–59]. In the 1980s, bird replication-competent retroviral (RCR) vectors were developed and used for the generation of transgenic chickens [49, 50].Although the viral vectors could be amplified in infected cells, the germ-line transmission efficiency was as low as 10%. Because of safety concerns over the replication competency, however, there has been a restraint on the use of viral vectors in the food and pharmaceutical industries. To overcome this, replication-defective

183

Fig. 10 Production of RDR vectors pseudotyped with VSV-G

Transgenic Birds for the Production of Recombinant Proteins

retroviral (RDR) vectors (Fig. 10) based on reticuloendotheliosis virus [51] and bird leukosis virus [52–55] were used to generate transgenic chickens. The viral vectors were injected into chicken embryos at the blastodermal stage in ovo or isolated primordial germ cells (PGCs) in vitro. In both cases, viral DNA sequence was transmitted to G1 progeny at a frequency of less than 10%. The viral titers used in these studies were at most 106 colony-forming units (cfu)/ml, since it had been difficult to obtain a high viral titer in the RDR vectors.

184

M. Kamihira et al.

Microinjection of a reporter plasmid into fertilized chicken embryos at the single cell stage was also tried to generate transgenic chickens [57]. However, since egg-donor hens were sacrificed to obtain a limited number of embryos from the oviduct, and the germ-line transmission efficiency was low (3.4%), this method is very laborious. Germ-line transmission efficiency was significantly improved by using the Drosophila transposable element mariner [58]. However, a foreign gene has not been inserted into the transposon as a vector. Thus, the low frequency of germ-line transmission of transgene has been a major obstacle to the generation of transgenic birds for practical use. In recent years, vesicular stomatitis virus G protein (VSV-G) has often been used for the pseudotyping of retroviral vectors [59, 60]. In pseudotyped viral vectors, the original envelope protein of the retroviral particle is replaced by VSV-G (Fig. 10). The vectors can infect many cell types because VSV-G interacts with phospholipid of the host cell membrane [59]. Furthermore, the viral titer can be increased up to 1000-fold by ultracentrifugation [60]. Because of these advantages, VSV-G pseudotyped vectors have been used for producing transgenic animals such as zebrafish, cow and monkey. We generated transgenic quails using a VSV-G pseudotyped retroviral vector based on MoMLV at a greater efficiency (more than 80%) compared with previous methods (Fig. 11) [56]. Although we chose quail as the host for gene manipulation due to its short sexual maturation and relatively high egg productivity, the technique can essentially be transferred to chicken with only minor modifications. We assume that the high frequency of germ-line transmission in our method is attributable to the high viral titer of 109 cfu/ml, which is about 1000 times that in previous studies using bird retroviral vectors. This titer corresponds to 40 infectious viral particles per cell under our conditions.

Fig. 11 Transgenic quails generated by retroviral vector

Transgenic Birds for the Production of Recombinant Proteins

185

Since VSV-G pseudotyped retroviral vectors can be concentrated by ultracentrifugation, they have excellent transduction efficiencies in the applications for transgenic animals and gene therapy. For the production of pharmaceutical proteins using transgenic animals, a high copy number of transgenes is preferable to attain a high level of expression of the target protein unless the position effect of an inserted gene is serious. A study of transgenic mice showed that the more copies of the alpha-lactoalbumin gene introduced, the higher the productivity [61]. In our method, multiple copies of the vector sequence were inserted into the genome; a G1 transgenic quail had three copies of the vector sequence in its genome, even when a G0 transgenic quail was mated with a non-transgenic one to produce the G1 transgenic offspring. Therefore, it is assumed that more than three copies of the transgene are inserted in PGCs of the G0 quail. The copy number of transgenes may be further amplified by mating among the transgenic quails. In our previous study [56], the retroviral vector encoded two reporter genes flanked by viral long terminal repeats (LTRs), and the gene expression is controlled by the LTR promoter and Rous sarcoma virus (RSV) promoter.While the expression of the reporter gene under the control of the internal RSV promoter was detected (Fig. 12), the expression of another reporter gene controlled by the MoMLV LTR promoter was not observed. MoMLV-based retroviral vectors have been often used for the generation of transgenic animals, but active gene

Fig. 12 Transgene expression in transgenic quails

186

M. Kamihira et al.

expression induced by the viral LTR promoter has not been reported. In recent years, there have been efforts to elucidate the mechanism of retroviral gene silencing, and several studies showed that the silencing correlated with de novo methylation of proviral sequences [62, 63]. These observations and the recent finding that methylation leads to chromatin condensation by the action of a histone deacetylase (HDAC) suggest that DNA methylation may have a primary role in the provirus silencing. In our study, however, distinct methylation was not detected in the sequences as far as we have studied in the G1 transgenic quails [56]. Other mechanism independent of both methylation and histone deacetylation may be responsible for the silencing observed in the transgenic quails. The inactivation of provirus by trans-acting repressors in ES cells has also been studied extensively and the repressors have been identified [64, 65]. Therefore, it is possible that a transcriptional repressor silences the LTR promoter in the transgenic quails. In general, retroviruses are highly susceptible to transcriptional silencing, which hampers the use of virus-based vectors for the establishment of transgenic livestock to produce protein products. Recently, several authors reported that the use of a chromatin insulator such as locus-control regions [66–68] and matrix-attachment regions [69], at least partly, prevented gene silencing and positional effects of the transgene. Furthermore, partial deletion or mutagenesis of viral sequences may reduce the extent of proviral silencing. A combination of the use of an insulator and the modification of viral sequences may yield high expression levels of transgene from an internal promoter. We achieved a high frequency of germ-line transmission of the vector DNA sequence using a VSV-G pseudotyped retroviral vector with high titer. The system will provide a safer and less laborious procedure for the production of transgenic birds.

5 Concluding Remarks Transgenic bird bioreactors possess great potential for the production of recombinant pharmaceutical proteins. In particular, chickens have been recognized as a promising transgenic bioreactor because of several advantages including high protein productivity in eggs, the ease and small space requirement for breeding, the similarity of the glycosylation pattern of proteins to that of human, and the absence of the prion problem. Since several venture companies entered the scene in the mid-1990s, a major effort has been done to develop robust methods for modification of the chicken genome. However, transgenic procedures for birds have lagged far behind those of other animals, and many hurdles still remain. Although the retroviral methods have exhibited a powerful and promising way, they might not deliver large vector constructs for high level, tissue-specific expression in oviduct. We recently found that embryonic retroviral silencing occurred in the early stage within a few days of being laid.

Transgenic Birds for the Production of Recombinant Proteins

187

The viral injection in the period free from silencing enabled G0 chickens and quails to express the target gene at a high level, more than 1 mg/ml in serum and egg white, although the G0 birds must be mosaic for the transgene. To date, no one has reported high-level expression in transgenic progeny. Thus, it is still necessary to develop innovative and efficient transgenic procedures for birds. Among those, the ES and PGC approaches may include after the major obstacle has been overcome. We have developed novel techniques necessary to generate transgenic birds. Transgenic chickens and quails could be steadily generated using retroviral vectors. Now an exogenous gene could be expressed in egg white in G0 transgenic birds at the commercially feasible level. Our efforts have exhibited the reality of the commercial production of pharmaceutical recombinant proteins using transgenic bird bioreactors. Acknowledgements This work was supported in part by a Grant-in-Aid for Scientific Research (no. 14350434) from the Japan Society for the Promotion of Science.

References 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17.

18. 19. 20. 21.

Rudolph NS (1999) Trends Biotechnol 17:367 Houdebine LM (2000) Transgenic Res 9:305 Dove A (2000) Nat Biotechnol 18:1045 Sang H (1994) Trends Biotechnol 12:415 Ivarie R (2003) Trends Biotechnol 21:14 Ono T, Wakasugi N (1983) Poult Sci 62:532 Rowlett K, Simkiss K (1987) Brit Poult Sci 28:91 Perry MM (1988) Nature 331:70 Naito M, Nirasawa K, Oishi T (1990) J Exp Zool 254:322 Nirasawa K, Naito M, Oishi T (1992) Jpn Poult Sci 29:139 Ono T, Muto S, Mizutani M, Agata K, Mochii M, Kino K, Otsuka K, Ohta M, Yoshida M, Eguchi G (1994) Jpn Poult Sci 31:119 Ono T, Murakami T, Mochii M, Agata K, Kino K, Otuka K, Ohta M, Mizutani M, Yoshida M, Eguchi G (1994) Dev Biol 161:126 Kamihira M, Oguchi S, Tachibana A, Kitagawa Y, Iijima S (1998) Dev Growth Differ 40:449 Felgner PL, Gadek TR, Holm M, Roman R, Chan HW,Wenz M, Northrop JP, Ringold GM, Danielsen M (1987) Proc Natl Acad Sci USA 84:7413 Malone RW, Felgner PL, Verma IM (1989) Proc Natl Acad Sci USA 86:6077 Gao X, Huang L (1995) Gene Ther 2:710 Caplen NJ, Alton EW, Middleton PG, Dorin JR, Stevenson BJ, Gao X, Durham SR, Jeffery PK, Hodson ME, Coutelle C, Huang L, Porteous DJ,Williamson R, Geddes DM (1995) Nat Med 1:39 Oudrhiri N,Vigneron JP, Peuchmaur M, Leclerc T, Lehn JM, Lehn P (1997) Proc Natl Acad Sci USA 94:1651 Kielian T, Hickey WF (1999) Nat Med 5:1237 Nabel EG, Gordon D, Yang ZY, Xu L, San H, Plautz GE, Wu BY, Gao X, Huang L, Nabel GJ (1992) Hum Gene Ther 3:649 Thierry AR, Lunardi-Iskandar Y, Bryant JL, Rabinovich P, Gallo RC, Mahan LC (1995) Proc Natl Acad Sci USA 92:9742

188

M. Kamihira et al.

22. Zabner J, Fasbender AJ, Moninger T, Poellinger KA, Welsh MJ (1995) J Biol Chem 270:18997 23. Senior JH, Trimble KR, Maskiewicz R (1991) Biochim Biophys Acta 1070:173 24. Zelphati O, Szoka FC (1996) Proc Natl Acad Sci USA 93:11493 25. Leventis R, Silvius JR (1990) Biochim Biophys Acta 1023:124 26. Crook K, Stevenson BJ, Dubouchet M, Porteous DJ (1998) Gene Ther 5:137 27. Abe A, Miyanohara A, Friedmann T (1998) Gene Ther 5:708 28. Goldman CK, Soroceanu L, Smith N, Gillespie GY, Shaw W, Burgess S, Bilbao G, Curiel DT (1997) Nat Biotechnol 15:462 29. Lewis JG, Lin KY, Kothavale A, Flanagan WM, Matteucci MD, DePrince RB, Mook RA, Hendren RW, Wagner RW (1996) Proc Natl Acad Sci USA 93:3176 30. Solodin I, Brown CS, Bruno MS, Chow C-Y, Jnag E-H, Debs RJ, Heath TD (1995) Biochemistry 34:13537 31. Yang J-P, Huang L (1997) Gene Ther 4:950 32. Hofland HEJ, Shephard L, Sullivan SM (1996) Proc Natl Acad Sci USA 93:7305 33. You J, Kamihira M, Iijima S (1997) Cytotechnology 25:45 34. You J, Kamihira M, Iijima S (1999) J Biochem 125:1160 35. Mizuarai S, Kamihira M, Nishijima K, Iijima S (1999) J Biosci Bioeng 88:461 36. Mizuarai S, Ono K, You J, Kamihira M, Iijima S (2001) J Biochem 129:125 37. Zelphati O, Uyechi LS, Barron LG, Szoka FC (1998) Biochim Biophys Acta 1390:119 38. Andreadis S, Palsson BO (1997) Hum Gene Ther 8:285 39. Sorokin LM, Morgan EH, Yeoh GC, Coelen R (1989) In Vitro Cell Dev Biol 25:63–68 40. Hodgson CP, Solaiman F (1996) Nat Biotechnol 14:339 41. Takeuchi JK, Koshiba-Takeuchi K, Matsumoto K, Vogel-Hopker A, Naitoh-Matsuo M, Ogura K, Takahashi N, Yasuda K, Ogura T (1999) Nature 398:810 42. Petropoulos CJ, Hughes SH (1991) J Virol 65:3728 43. Watanabe M, Naito M, Sasaki E, Sakurai M, Kuwana T, Oishi T (1994) Mol Reprod Dev 38:268 44. Rosenblum CI, Chen HY (1995) Transgen Res 4:192 45. Inada S, Hattori MA, Fujihara N, Morohashi K (1997) Reprod Nutr Dev 37:13 46. Muramatsu T, Mizutani Y, Ohmori Y, Okumura J (1997) Biochem Biophys Res Commun 230:376 47. Sorgi FL, Bhattacharya S, Huang L (1997) Gene Ther 4:961 48. You J, Kamihira M, Iijima S (1998) J Ferment Bioeng 85:525 49. Salter DW, Smith EJ, Hughes SH, Wright SE, Crittenden LB (1987) Virology 157:236 50. Salter DW, Smith EJ, Hughes SH, Wright SE, Fadly AM, Witter RL, Crittenden LB (1985) Poult Sci 65:1445 51. Bosselmen RA, Hsu RV, Boggs T, Hu S, Brusxewski J, Ou S, Kozar L, Martin F, Green C, Jacobson F, Nicolson M, Schultz JA, Seman KM, Rishell W, Stewart RG (1989) Science 243:533 52. Vick L, Li Y, Simkiss K (1993) Proc R Soc Lond B Biol Sci 251:179 53. Thoraval P,Afanassieff M, Cosset FL, Lasserre F,Verdier G, Coudert F, Dambrine G (1995) Transgenic Res 4:369 54. Harvey AJ, Speksnijder G, Baugh LR, Morris JA, Ivarie R (2002) Poult Sci 81:202 55. Harvey AJ, Speksnijder G, Baugh LR, Morris JA, Ivarie R (2002) Nat Biotechnol 19:396 56. Mizuarai S, Ono K, Yamaguchi K, Nishijima K, Kamihira M, Iijima S (2001) Biochem Biophys Res Commun 286:456 57. Love J, Gribbin C, Mather C, Sang H (1994) Biotechnology 12:60 58. Sherman A, Dawson A, Mather C, Gilhooley H, Li Y, Mitchell R, Finnegan D, Sang H (1998) Nat Biotechnol 16:1050

Transgenic Birds for the Production of Recombinant Proteins

189

59. Emi N, Friedmann T, Yee JK (1991) J Virol 65:1202 60. Burns JC, Friedmann T, Driever W, Burrascano M,Yee JK (1993) Proc Natl Acad Sci USA 90:8033 61. Stinnakre MG, Soulier S, Schibler L, Lepourry L, Mercier JC,Vilotte JL (1999) Biochem J 339:33 62. Challita P-M, Kohn DB (1994) Proc Natl Acad Sci USA 91:2567 63. Lorincz MC, Schubeler D, Goeke SC,Walters M, Groudine M, Martin DIK (2000) Mol Cell Biol 20:842 64. Yamauchi M, Freitag B, Khan C, Berwin B, Barklis E (1995) J Virol 69:1142 65. Cherry SR, Biniszkiewicz D, Parijs LV, Baltimore D, Jaenisch R (2000) Mol Cell Biol 20:7419 66. Rivella S, Callegari JA, May C, Tan CW, Sadelain M (2000) J Virol 74:4679 67. Ellis J, Tan-Un KC, Harper A, Michalovich D,Yannoutsos N, Philipsen S, Grosveld F (1996) EMBO J 15:562 68. Emery DW, Yannaki E, Tubb J, Stamatoyannopoulos G (2000) Proc Natl Acad Sci USA 97:9150 69. McKnight RA, Spencer M, Wall RJ, Hennighausen L (1996) Mol Reprod Dev 44:179

Received: September 2003

Adv Biochem Engin/Biotechnol (2004) 91: 191– 220 DOI 10.1007/b94212 © Springer-Verlag Berlin Heidelberg 2004

Functional Magnetic Particles for Medical Application Masashige Shinkai (✉)1 · Akira Ito2 1

2

Department of Chemistry and Biotechnology, School of Engineering, University of Tokyo, 7-3-1 Hongo, Bunkyo-ku, Tokyo 113-8656, Japan [email protected] Department of Biotechnology, School of Engineering, Nagoya University, Furo-cho, Chikusa-ku, Nagoya 464-8603, Japan [email protected]

1

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 192

2

Magnetic Separation for Purification and Immunoassay . . . . . . . . . . . 192

3

Drug Delivery and Targeting

4

Magnetic Resonance Imaging and Diagnosis

5 5.1 5.2 5.3 5.4 5.5 5.6

Hyperthermia . . . . . . . . . . . . . . . . . . . . . . . . . . . Hyperthermia Using Magnetite Cationic Liposomes . . . . . . . Hyperthermia Using Needle-Type Magnetite . . . . . . . . . . . Hyperthermia Using Antibody-Conjugated Magnetic Particles . Gene Therapy Using Magnetic Particles . . . . . . . . . . . . . Immunity Induction by Hyperthermia Using Magnetic Particles Heat Immunotherapy Using Magnetic Particles . . . . . . . . .

6

Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 217

. . . . . . . . . . . . . . . . . . . . . . . . . . 193 . . . . . . . . . . . . . . . . . 195 . . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

. . . . . . .

196 197 201 202 203 208 215

References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 217

Abstract Magnetic particles for medical applications have been developed by many researchers. Since magnetic particles have unique magnetic features not present in other materials, they can be applied to special medical techniques. Separation, immunoassay, magnetic resonance imaging (MRI), drug delivery, and hyperthermia are enhanced by the use of magnetic particles. Magnetite cationic liposomes (MCLs), one of the groups of cationic magnetic particles, can be used as carriers to introduce DNA into cells since their positively charged surface associates with the negatively charged DNA. MCLs can also be used as heat mediators for cancer therapy. Magnetic particles conjugated with tumor-specific antibodies have enabled tumor-specific contrast enhancement in MRI. In addition, antibody-conjugated magnetic particles were shown to target renal cell carcinoma cells, and are applicable to the hyperthermic treatment of carcinomas. It was also found that the hyperthermic treatment using magnetic particles induced an antitumor immunity. Thus, the use of magnetic particles with their unique features will further improve medical techniques. Keywords Magnetic particle · Magnetic resonance imaging · Hyperthermia · Drug delivery · Cancer immunotherapy

192

M. Shinkai · A. Ito

1 Introduction Magnetic particles from nanometer to micrometer size are involved in an increasing number of biological and medical applications. The unique feature of magnetic particles is their reaction to a magnetic force. Magnetic particles are attracted to high magnetic flux density and this feature is used for drug targeting and bioseparation including cell sorting. Hysteresis loss in the alternative magnetic field is a very important feature of magnetic particles, because it enables effective thermotherapy (hyperthermia). Of course, magnetic particles themselves generate a magnetic field and influence the local area around them. This feature is exploited in magnetic resonance imaging (MRI). In these applications, magnetic particles are given various names, e.g., magnetic microspheres, magnetic nanospheres, and ferrofluids among others. Here, we use the general term ‘magnetic particles,’ and summarize their medical applications.

2 Magnetic Separation for Purification and Immunoassay Separation is the most documented and currently the most useful application of magnetic particles. Many magnetic particles have been developed as magnetic carriers in separation processes including purification and immunoassay [1–4]. Specific ligands are bound to magnetic particles in the same way as to other carriers comprising a polymer matrix. Separation techniques are extremely important in process engineering including bioprocessing. For example, in the case of fermentation processes, it has been estimated that downstream processing represents 40% or more of the added value of a fermentation product [5]. In laboratory use, automatic DNA/RNA purification apparatuses are commercially available.With respect to medical applications, magnetic cell sorting was developed for cellular therapy. In the removal of a specific cell population from blood, other dominant blood cells interfere with conventional immunoaffinity column chromatography [6]. In blood purification, the elimination of specific cytokines such as interleukin-1 or tumor necrosis factor, the major pathogenic mediators in septic shock and multi-organ failure, by magnetic particles has been proposed [7]. Magnetic particles are also used in immunoassays [8, 9]. Immunoassays are widely used in the field of medical examination (e.g., detection of viruses or assay of various hormones) [10, 11]. Magnetic separation is a good way to shorten both adsorption and separation steps [12]. Recently, gene transfer using magnetic carriers has been proposed by some researchers [13]. In clinical use, somatic cells transformed to produce specific proteins require quick selection from the limited cell cycle. Therefore, quick selection of transformants among a transiently transformed population is important. Selection using antibiotic-resistance marker genes is, however, time

Functional Magnetic Particles for Medical Application

193

Table 1 Effect of magnetic separation on transformant selection

Cell line

Rat glioma cell T-9 Mouse fibroblast NIH/3T3 Monkey fibroblast CV-1 Monkey fibroblast COS-1 a

Specific luciferase activity (RLU/cell) Total cells before separation

Trapped cells

4.37 1.74 0.623 904

21.1 12.2 0.813 2830

Transformant selection index a

4.82 6.99 1.29 3.13

Transformant selection index = (activity of trapped cells)/(activity of total cells before separation). In the case of the mouse fibroblast NIH/3T3 and the monkey fibroblastlike cell line COS-1, high values of the index were obtained as was the case for T-9 cells. CV-1 cells did not exhibit a high specific luciferase activity. This may be due to the low transfection efficiency by MCLs, since the specific luciferase activity of total cells before separation was also low.

consuming and recovery of transformants cannot be guaranteed, because the transformation efficiency depends on the cell line used. Therefore, a more simple and quick selection method was desired. For such an application, our original magnetite cationic liposomes (MCLs) were used as carriers to introduce DNA into cells since their positively charged surface associates with the negatively charged DNA [14]. MCLs are cationic liposomes containing magnetite particles and have been developed to promote the uptake of magnetic particles by target cells [15]. We reported about the transfection of cells by MCLs associated with plasmid vector and selection of transformants by magnetic field [14]. When MCLs associated with plasmid vector were transfected into mouse NIH/3T3 cells, the specific activity of luciferase gene-transfectants was about seven times higher than that of the cells before separation by the magnetic field (Table 1). Some researchers have reported on the separation of transiently transfected cells using magnetic particles associated with antibodies [16, 17]. However, this method requires co-transfection of a gene for expression of a surface marker. In our method, gene expression of an additional marker is not required for separation and magnetic separation can be performed immediately after the transformation. This is a major advantage for separation of transformants.

3 Drug Delivery and Targeting The treatment of solid tumors using chemotherapy has been limited by systemic toxicity resulting in sub-optimal dosing and thus limited efficacy. Regional therapy achieved through targeted drug delivery could improve efficacy

194

M. Shinkai · A. Ito

by increasing the drug concentration at the tumor while limiting systemic drug concentrations. Magnetic drug targeting should be safe and effective, i.e., with the least amount of magnetic particles a maximum concentration of drug should be easily administered and transported to the site of choice. The use of magnetic particles for the delivery of chemotherapeutic agents has evolved since the 1970s. Widder et al. and Morimoto et al. developed albumin microspheres encasing anticancer drugs [18, 19]. The magnetic particles injected into an animal were retained at the magnet site depending on the magnetic field strength. It has also been shown that magnetic particles can be retained in other parts of the body depending on the placement of the external magnet [20]. Therefore, the magnetic force enhances the uptake of magnetic particles by tumor cells.We have reported the efficacy of magnetic force on the uptake of magnetoliposomes in tumor cells. When a magnet was placed under the cell culture dish, the magnetoliposome-uptake in the tumor cells was increased markedly. In a rat subcutaneous tumor, the uptake was also enhanced by using a magnet: the uptake ratio being 70.0% of the magnetoliposomes injected, compared to 27.4% when no magnet was used as shown in Fig. 1 [21]. We believe that the geometry of the magnetic field is very important. In the conventional magnetic field geometry, the magnetic force works as the magnetic particles are adsorbed on the surface of the magnet. Therefore, the administration method is limited to an artery close to the tumor or to some other target close to the tumor. A novel magnetic field geometry is needed to realize completely targetable magnetic drug delivery.

Fig. 1 Uptake of magnetoliposomes by subcutaneous tumors. Columns and bars represent the means ±SD of five rats

Functional Magnetic Particles for Medical Application

195

4 Magnetic Resonance Imaging and Diagnosis Magnetic resonance imaging (MRI) offers the advantage of high spatial resolution of contrast differences between tissues. Due to the unique function of this imaging modality, there is a need to develop effective contrast agents that will enhance and widen its diagnostic utility. Paramagnetic ion chelates and ferromagnetic or superparamagnetic particles have been investigated as contrast agents and used in clinical diagnosis.

Fig. 2 Schematic illustration of the magnetic field sensing of a subcutaneous tumor in a rat and the magnetic field distribution in the transplanted tumor. The shaded circle on the x-y plane indicates the projection of the tumor. The magnetic particles were directly injected into the tumor transplanted under the skin of the rat leg and magnetized by a static magnetic field generator at 1.8 T. By using the MI sensor, the magnetic field distribution of the MCLs was detected, and the distribution indicated that the MCLs were localized in the tumor

196

M. Shinkai · A. Ito

During recent years, there has been increasing interest in the use of paramagnetic contrast agents like dextran magnetite in magnetic resonance imaging (MRI) [22, 23]. Compared with paramagnetic ions, superparamagnetic iron oxide particles have higher molar relaxivities, and, when used as blood pool and tissue-specific agents, may offer advantages at low concentrations [24, 25]. Commercial contrast agents do not have targetability but are useful for enhancing contrast differences of the lymph nodes, liver, or spleen. We and other researchers have been studying the specific distribution of these agents by active targeting [26, 27]. We have also proposed a direct sensing method of magnetic particles for cancer diagnosis. Accurate sensing of brain tumors during brain surgery is important, and MRI, for example, has been used during surgical operations.We developed a small and highly sensitive magnetic sensor, the magneto-impedance (MI) sensor for the sensing of tumor tissue incorporating magnetic particles and tried to detect magnetite particles in a rat glioma, a malignant brain tumor which was transplanted under the skin of the rat leg (Fig. 2) [28, 29]. The magnetite cationic liposomes (MCLs) were directly injected into the tumor and magnetized by a static magnetic field generator at 1.8 T. By using the MI sensor, the magnetic field distribution of the MCLs was detected, and its distribution indicated that the MCLs were localized in the tumor. However, such apparatuses for the above methods are experimental and, unlike MRI, are not used in clinical practice.When these apparatuses are available for clinical use, the diagnosis of metastatic cancer with high sensitivity will be possible.

5 Hyperthermia Hyperthermia is one of the promising approaches in cancer therapy. Various methods are employed in hyperthermia such as the use of hot water, capacitive heating, and induction heating among others [30–33]. However, the inevitable technical problem with hyperthermia is the difficulty of the uniform heating of only the tumor region until the required temperature is reached without damaging normal tissue. Therefore, some researchers have proposed intracellular hyperthermia and developed submicron magnetic particles for hyperthermia [34–36]. This idea is based on the principle that a magnetic particle can generate heat by hysteresis loss under an alternating magnetic field (AMF) [33]. Even in the early days of hyperthermia therapy, the use of magnetic particles was proposed for localized hyperthermia treatment, and local temperature elevations were achieved [37]. Administered particles usually migrate to reticuloendothelial systems like the macrophages and Kupffer cells in the liver, spleen, and lymph nodes [38, 39]. Lymph nodes are sites in which metastatic tumors easily grow. However, the delivery of magnetic particles is passive and its control is very difficult.

Functional Magnetic Particles for Medical Application

197

In 1979, Gordon et al. first proposed the concept of intracellular hyperthermia using dextran magnetite nanoparticles [40]. They administered magnetic particles intravenously to Sprague-Dawley rats bearing mammary carcinomas and showed that AMF-induced heating occurred in their in vivo experiments. Jordan et al. have proposed magnetic fluid hyperthermia in several more comprehensive in vitro studies [41]. Regarding their application to hyperthermia, the important properties of magnetic particles are non-toxicity, biocompatibility, injectability, high-level accumulation in the target tumor, and effective absorption of the energy of AMF. Chan et al. reported on modified dextran magnetite and its hyperthermic effect using several human carcinoma cell lines in vitro [42]. The specific absorption rate (SAR), which indicates the heat evolution rate in hyperthermia, of the conventional dextran magnetite is low. Dextran magnetite behaves as a superparamagnetic particles rather than a ferromagnetic one due to its small size, so that its hysteresis loss is very low. Chan et al. controlled the oxygen concentration in the preparation of the dextran magnetite and selected particles of around 15 nm. We have reported that particle size is a critical factor in obtaining a high SAR value [43]. In the case of magnetite, the SAR of 35-nm particles was much higher than that of 10-nm particles. 5.1 Hyperthermia Using Magnetite Cationic Liposomes Accumulation of magnetic particles in tumor cells can be enhanced by their surface charge. We developed ‘magnetite cationic liposomes’ (MCLs) with improved adsorption and accumulation properties [15]. MCLs, which have a positive surface charge, show a tenfold higher affinity for glioma cells than neutrally charged magnetoliposomes as shown in Fig. 3. They have a sufficiently high SAR and a general biocompatibility that is comparable to that of dextran magnetite. We previously demonstrated the efficacy of hyperthermia using MCLs against animals having several types of tumor, such as B16 mouse melanoma [44], T-9 rat glioma[45], Os515 hamster osteosarcoma (unpublished results), and VX-7 squamous cell carcinoma in rabbit tongue [46]. In the experimental model using T-9 rat glioma [45], MCLs were injected into solid tumors formed subcutaneously in F344 rats and the rats were irradiated once (group II), twice (group III), and three times (group IV) for 30 min with an AMF. In control group (group I), rats were administrated with MCLs but not exposed to AMF. The temperature of the tumor was elevated rapidly by magnetic heating and reached over 43 °C after 15 min as shown in Fig. 4. In contrast, the rectal temperature or that in a tumor lacking the MCLs remained between 35 and 37 °C.When AMF irradiation was extended beyond 30 min, the tumor temperature at the outside skin continued to increase gradually. Table 2 shows the average tumor volumes and ratios of complete regression on the 30th day after the MCLs injection. In group II, the decrease in the average volume was not significant. However, in the case of group III and IV, in which the ani-

M. Shinkai · A. Ito Fig. 3 Illustration of magnetite cationic liposomes (MCLs) and comparison of magnetite uptake between MCLs and neutrally charged magnetoliposomes (MLs). Open circles: MLs, closed circles: MCLs. Each point represents the means ±SD of three independent experiments

198

mals received repeated hyperthermia, the average tumor volume decreased markedly. Complete regression was observed in one of the five rats in group II, three of the five in group III, and seven of the eight in group IV, confirming the effectiveness of multiple irradiations in attaining tumor regression. In group IV, only one tumor did not regress. Since the shape of this tumor was not ellipsoidal but irregular, it appears that the MCLs could not heat the whole of the tumor homogenously even when it was irradiated three times. To achieve complete tumor regression in such a case, multiple injections of MCLs may be necessary. Figure 5 shows the time courses of tumor growth of five rats in each group. In control animals (indicated by ‘no irradiation’ in Fig. 5), the tumor volume in

Functional Magnetic Particles for Medical Application

199

Fig. 4 Temperature increases at the outside skin of the tumor and in the rectum during magnetic field irradiation. Open circles: rectum, closed circles: outside skin of the tumor injected with the MCLs. Each point represents the means ±SD of five rats

Table 2 Hyperthermic effect on solid glioma tissue formed subcutaneously 30 days after the MCL injection Group

Treatment time

Average tumor volume (mm3)

Relative tumor volume a

Complete regression (%)

I II III IV

0 1 2 3

30400 28100 8710 2684

1 0.924 0.287 0.088

0 20 60 87.5

a

Relative tumor volume = (average tumor volume of each group)/(average tumor volume of group I).

each rat steadily increased with no evidence of regression. In contrast, complete tumor regression was observed in many of the rats subjected to three separate AMF irradiations. In most cases of complete regression, the tumor volume increased up to the 12th day, after which it began to decrease and the tumor finally disappeared. No regrowth of tumors was observed following complete regression over a period of three months. In the present study, 40–60% of the injected MCLs accumulated in the tumor tissue, whereas only 20–25% of the neutral magnetoliposomes were accumulated. This high accumulation of MCLs is considered to be due to the cationic charge on the surface of the liposome used, and a significant hyperthermic effect was achieved as shown in Fig. 5.

200

M. Shinkai · A. Ito

Fig. 5 The time courses of tumor growth of five rats in each group. In non-irradiated animals, the tumor volume in each rat steadily increased with no evidence of regression. In contrast, complete tumor regression was observed in 90% of the rats irradiated by the magnetic field. Regrowth of the tumor over a three-month period was not observed following complete regression

Figure 6a shows non-stained tumor sections, and MCLs are evident as black regions and Fig. 6b shows the same sections stained with hematoxylin-eosin. In the group I specimen, the MCLs are localized around the point where the needle was inserted. In contrast, the MCLs in the group II, III, and IV specimens are more widely distributed as a result of hyperthermia treatment. As shown in Fig. 6b, the necrotic tissue also became wide by the repeated irradiation. Therefore, the hyperthermic effect was greatly improved in the group III and IV rats, even though the initial distribution of the MCLs was localized. This feature is never found in hyperthermia treatment using implantable needles. These results indicate that MCLs are the superior tool for hyperthermia.

Functional Magnetic Particles for Medical Application

201

Fig. 6A,B Photographs of tumor specimens: A paraffinized; B histologically stained by hematoxylin-eosin of tissues removed from rats at 4th day after the MCLs injection. I, Group I (control); II (irradiated once for 30 min); III, group III (irradiated for 30 min twice, at 24 h intervals); IV, group IV (irradiated for 30 min three times, at 24 h intervals). N and V in photographs indicate necrotic tissue (pale region) and viable tissue (dark region), respectively

5.2 Hyperthermia Using Needle-Type Magnetite We have improved the method of administration of the magnetic particles. The needle-type magnetite, newly manufactured magnetite-product, consists of magnetite and carboxymethyl cellulose (CMC) (Fig. 7) and it is as easy to administer as an injection needle [47, 48].Various kinds of hyperthermic treatment for malignant glioma had been inhibited due to both their incomplete feverish action and strict cooling effect of the brain. We showed an effective result of hyperthermia for the treatment of malignant glioma in an in vivo study using needle type CMC-magnetite.A needle type CMC-magnetite was inserted into the T-9 glioma in the rat brain stereotactically, and the rats were exposed to AMF. The application time of AMF, which measured 88.9 kHz and 380 Oe, was 30 min a day. The rats were divided into three groups: three AMF applications (group I), one AMF application (group II), and no application but only

202

M. Shinkai · A. Ito

Fig. 7 Photograph of needle-type carboxymethylcellulose (CMC) magnetite. Average diameter is 0.6 mm

injection of CMC-magnetite (control). The time taken to administer two magnetite needles was only 3 min, although the injection time for the magnetite cationic liposomes (MCLs) was at least 120 min because the backflow of the MCLs had to be prevented. Temperature of subscalp, pericranial region over the CMC-magnetite in the brain, and that of the rectum, were measured by the optical fiber, and the temperature of the subscalp region increased up to 44.4 °C, despite the fact that that of the rectum did not change. The mean survival in days of these three groups measured 44.2±10.9 (group I), 17.0±1.5 (group II), and 14.4±1.5 (control). This investigation showed both significant effectiveness in attacking malignant glioma and significant prolonging of the survival time in rats. It is also a characteristic feature of the magnetite particles to spread through the tumor diffusely after three applications of AMF. This feature seemed to be one of the main factors that caused greater hyperthermic effect on glioma in this study. Moreover, the dispersed magnetic particles in tissue are removed eventually by the blood flow. Therefore, it is not necessary for the magnetite needles to be removed following therapeutic treatment. These results suggest that administration of the needle-type magnetite is a simple and effective method for mediating therapeutic particulate heating in hyperthermia treatment of malignant glioma. 5.3 Hyperthermia Using Antibody-Conjugated Magnetic Particles Administration of the MCLs and the needle-type magnetite is limited to direct injection into the tumor tissue. If magnetic particles adsorb to only target cells, the particles could be administered via blood vessels. This feature would be of great advantage in terms of the quality of life of patients and would make it possible to diagnose cancer by MRI and to conduct hyperthermia treatment (the proposed therapeutic strategy using magnetic particles is illustrated in Fig. 8). Conjugation of antibodies to magnetic particles is one approach to achieving such an aim.We have developed magnetic particles conjugated to the Fab¢ fragments of human MN antigen-specific antibody and their hyperthermia effects were demonstrated using a mouse renal cell carcinoma model [49, 50]. The Fab¢

Functional Magnetic Particles for Medical Application

203

Fig. 8 Schematic illustration of the therapeutic strategy using magnetic particles

fragment of the G250 antibody which binds to the MN antigen on many types of human renal cell carcinomas was cross-linked to N-(6-maleimidocaproyloxy)-dipalmitoyl phosphatidylethanolamine (EMC-DPPE) in the liposomal membrane. The targetability of the G250-Fab¢ fragment-conjugated MLs (G250-FMLs) was investigated using a mouse renal cell carcinoma (mRCC) and MN antigen-presenting cells, MN-mRCC. In an in vivo experiment using MNmRCC-harboring mice, the accumulation was 7.9 mg of the FMLs per gram of carcinoma tissue (the tumor weight was 0.19 g), which corresponded to approximately 50% of the total amount injected. This value was 27 times higher than that of the MLs. After injection of the FMLs, mice were exposed to intracellular hyperthermia using AMF irradiation. The temperature of the tumor tissue increased to 43 °C and the growth of the carcinoma was significantly arrested for at least two weeks. These results indicate that the G250-FMLs could target the renal cell carcinoma cells in vitro and in vivo, and are applicable to the efficient hyperthermia treatment of carcinomas. As shown in Fig. 9, the G250-FMLs could target the renal cell carcinoma cells, and are applicable to the hyperthermia of carcinomas (Fig. 10). 5.4 Gene Therapy Using Magnetic Particles Gene therapy offers a multitude of exciting opportunities of novel approaches for cancer treatment. One of the most powerful approaches to achieve this is to genetically target toxic therapeutic genes to cancer cells.An approach to target tumor is based on the use of chemically-induced promoters [51–54] and physically-inducible promoter [55]. The strategy using inducible promoter is critical for the use of potent, but cytotoxic cytokines that are normally of limited

Fig. 9 Illustration of G250-Fab¢ fragment-conjugated MLs (G250-FMLs) and their magnetite uptake for carcinomas and various organs. Animals in group I (open columns), the first control group, were transplanted with MN-mRCC and magnetoliposomes (MLs) were injected. Group II animals (shaded columns), the second control group, were transplanted with mRCC and the G250-FMLs were injected. Group III animals (closed columns) were transplanted with MN-mRCC and the G250-FMLs were injected. G250-FMLs or MLs (0.4 ml, net magnetite: 3 mg) containing 0.3 mol% Tween 20 were injected via the heart. The five mice in each group were sacrificed and the carcinoma and organ were removed to assess accumulation of magnetite at 48 h after G250-FML or ML injection. Each point represents the means ±SD of five mice

204 M. Shinkai · A. Ito

Functional Magnetic Particles for Medical Application

205

Fig. 10a,b Photographs of a kidney at 14 days after hyperthermia treatment (scale bar, 5 mm). At 48 h after the injection, the mice were anesthetized and subjected to magnetic field irradiation (118 kHz and 30.6 kA/m [384 Oe]). The mean initial carcinoma weight was 0.19 g. Treatment was repeated three times at 24-h intervals: a MN-mRCC/G250-FMLs; b mRCC/G250-FMLs. Carcinomas shown in a was only observed on the surface of the kidney and the mean weight gain of carcinomas was only 0.08 g. Since G250-FMLs could not be adsorbed on the carcinoma surface, the carcinomas shown in b were not heated, grew vigorously and their weight gain was 1.38 g over the course of 14 days

practical utility due to systemic toxicity concerns. An example of a cytokine with potential therapeutic application is tumor necrosis factor (TNF)-a, which exerts cytotoxic effects on a wide range of tumor cells by inducing apoptosis and which also activates the immune system. However, its clinical use in humans has been greatly limited by systemic toxicity. Recently, gene therapy approach based on hyperthermia has been studied for its efficacy in cancer treatment by a number of different groups. This approach is based on the heat shock response and stress response. In mammalian cells, this response is usually activated when the cells are exposed to 3–7 °C higher temperature than physiological norms [56]. During heat shock response, mammalian cells shut down the majority of protein synthesis and initiate the synthesis of a class of stress proteins such as heat shock proteins (HSPs) [56]. Much of the induction of the stress proteins is mediated at the transcriptional level by the promoters of these genes [57]. A typical promoter is the one derived from the Hsp70B. Huang et al. reported that the expression of cytokine gene placed under the control of Hsp70B promoter was induced over 33- to 835-fold in vivo by heating at 42 °C [58]. Most importantly, the background activity of this promoter was extremely low, but still resulted in such an extremely high level induction. We selected the gadd (the Growth Arrest and DNA Damage) 153 promoter as a physically-inducible promoter. The gadd153 gene was originally identified

206

M. Shinkai · A. Ito

by a DNA damaging regent [59–61]. Ropp et al. reported that heat and oxidative stress could be related to this effect [62]. Therefore, we investigated whether the gadd153 promoter could be used as a heat-inducible promoter [63]. The plasmid containing luciferase gene placed under gadd153 promoter was introduced into the human glioma cell line, U251SP. When the transfected cells were heated at 43 °C for 1 h or 45 °C for 1 h, inducible expression of the luciferase gene was observed. Interestingly, the pattern of enhancement of the gene expression varied with temperature. The luciferase activity at one day after heating to 43 °C was sevenfold that in the case without heat treatment (basal level), then the activity gradually decreased to near the basal level. On the other hand, the expression level gradually increased after heating to 45 °C, with a maximum value of 31-fold the basal level obtained at three days after the treatment. We assumed that these differences of gene expression pattern were due to the degree of cellular damage. Yonezawa et al. compared the manner of cell death induced by hyperthermia [64]. Apoptotic cell death was induced in malignant fibrous histiocytoma cells by hyperthermia treatment at 42 °C, whereas necrotic cell death was induced by hyperthermia at 44 °C. Because gadd153 gene expression is involved in a pathway of apoptotic cell death [65], the manner of cellular damage may relate to the gene induction. We chose TNF-a for therapeutic agent. One of the mechanisms of TNF-ainduced cytotoxicity is damage caused by reactive oxygen species [62]. The refore, we also investigated whether TNF-a itself induced the activity of the gadd153 promoter and enhanced TNF-a gene expression, in addition to the investigation of the heat-induced TNF-a expression [66]. The plasmid including the TNF-a gene under the control of the gadd153 promoter (pGadTNF) was introduced into the cells of a human glioma cell line, U251-SP cells. When the transfected cells were exposed to heat treatment, the TNF-a gene expression was induced. The heated pGadTNF-transfected cells were co-cultured with non-heated pGadTNF-transfected cells as bystander cells.At the 1:0 ratio (only heated cells) or the 0:1 ratio (only non-heated cells), low concentration was detected (46.2 and 13.7 pg/ml, respectively). In the case of the co-culture at the 1:1 ratio, total TNF-a gene expression was 172 pg/ml. This value was three times higher than the sum of the values of both the heated cells alone and non-heated cells alone. This result means that the TNF-a expressions of both cells were enhanced by each other expression. According to this cyclic induction by both heated and non-heated cells, a strong cytotoxic effect on the bystander cells was observed. As a result of this bystander-killing effect, it was concluded that the transfection of the TNF-a gene under the control of the gadd153 promoter into tumor cells would be a potent tool for hyperthermia. We combined TNF-a gene therapy driven by a heat-inducible promoter gadd153 with hyperthermia using MCLs as a novel tumor-target therapy [67]. The strategy was illustrated in Fig. 11. This system was applied to an in vivo experiment using tumor-bearing athymic mice. The heat stress caused by the magnetic particles resulted in a threefold increase in TNF-a gene expression driven by the gadd153 promoter as compared with that of non-heated tumor

Functional Magnetic Particles for Medical Application

207

Fig. 11 Strategy of heat-inducible gene therapy. Magnetite cationic liposomes (MCLs) and TNF-a gene under the control of the gadd153 promoter (pGadTNF) are introduced into tumor cells. The heat stress caused by the magnetic particles resulted in an increase of TNF-a gene expression driven by the gadd153 promoter

Fig. 12 U251-SP tumor volumes in athymic mice treated with the pGadTNF and hyperthermia using MCLs. Open circles; control group: tumors were injected with empty liposomes in PBS on day 0 and with MCLs on day 1, AMF was not applied. Closed triangles; gene therapy group: tumors were transfected with pGadTNF on day 0 and injected with MCLs on day 1, but without an AMF irradiation. Closed squares; hyperthermia group: tumors were injected with empty liposomes in PBS on day 0 and with MCLs on day 1 and an AMF (118 kHz and 30.6 kA/m [384 Oe]) was applied on day 2. Closed circles; combined therapy group: tumors were transfected with pGadTNF on day 0 and injected with MCLs on day 1 and an AMF was applied on day 2. Five mice in each group were used for monitoring tumor growth

208

M. Shinkai · A. Ito

cells. TNF-a gene expression was also observed in the peripheral area where the hyperthermic effect was not sufficient to cause cell death. The combined treatment of hyperthermia and gene therapy strongly inhibited tumor growth in athymic mice over a 30-day period (Fig. 12), suggesting great potential for cancer treatment. We also demonstrated a combined therapeutic effect with cobalt-60 gamma irradiation in an in vitro study [68]. The plasmid was introduced into the human glioma cell line, U251-SP. Without cobalt-60 gamma irradiation, no cytotoxicity against the transfected cells was observed. When the transfected cells were irradiated with 10 and 20 Gy, gadd153 promoter was highly induced and the expression level of TNF-a increased (30- and 100-fold increase, respetively). The cytotoxicities against the transfected cells at five days after irradiation with 10 and 20 Gy were 79% and 91%, respectively. These results suggest the effectiveness of combination of hyperthermia, radiotherapy, and gene therapy. 5.5 Immunity Induction by Hyperthermia Using Magnetic Particles Interestingly, hyperthermia using MCLs induced antitumor immunity [69].We demonstrated this using two tumors transplanted to both femurs of a rat. Although only one tumor was subjected to hyperthermia, the other tumor also disappeared completely (Fig. 13). Immunocytochemical assay revealed that both CD8+ and CD4+ T cells migrated in the tumors after the hyperthermia treatment (Fig. 14). These results suggest that our therapeutic magnetic particles are potentially effective tools for hyperthermic treatment of tumors, because in addition to the killing of tumor cells by heat, a host immune response is induced. To elucidate how tumor antigens are recognized by the host immune

Fig. 13A,B Rats photographed on the 28th day after the MCL injection. Rat glioma T-9 cells (1¥107 cells) were transplanted subcutaneously into the left femoral region of F344 rats. On the ninth day after transplantation into the left side, another aliquot of the T-9 cell suspension (1¥107 cells) was transplanted subcutaneously into the right femoral region. The MCLs were directly injected into the tumor on the 11th day after the first transplantation. The rats were anesthetized and subjected to magnetic field irradiation (118 kHz and 30.6 kA/m [384 Oe]). Treatment was repeated three times at 24-h intervals: A a non-irradiated rat; B an irradiated rat

Functional Magnetic Particles for Medical Application

209

Fig. 14 A,B Immunohistological staining for immunocytes. Anti-CD3+, CD4+, CD8+, and NK cells antibodies were used (scale bar, 50 mm): A tumor tissue in the left femoral region of the rats given the hyperthermia treatment; B tumor tissue in the right femoral region of the rats

system, we proposed heat shock proteins (HSPs)-mediated antitumor immunity as a mechanism. Hyperthermia is known to induce HSPs [70]. Because expression of HSPs protects cells from heat-induced apoptosis [71], HSPs expression has been considered to be a complicating factor in hyperthermia. On the other hand, recent reports have shown the importance of HSPs, such as HSP70, HSP90 and glucoseregulated protein 96 (gp96), in immune reactions, and investigators have suggested that HSPs chaperon tumor antigens [72, 73]. With regard to the mechanism of antitumor immunity induced by hyperthermia using MCLs, we have demonstrated two possible mechanisms of antigen presenting via HSP70 expression during hyperthermia [74, 75]. One possible mechanism is presentation of antigenic peptides via MHC class I by tumor cells themselves. Srivastava et al. proposed the following “relay line model” for tumor antigenic peptide transfer during antigen processing and presentation by HSPs [76–78]. 1. The peptides are first bound to HSP70 (or HSP90), which are responsible for carrying them to the endoplasmic reticulum (ER) through the transporter associated with antigen processing (TAP). 2. The peptides are transferred to gp96 in the lumen of the ER. 3. In the terminal step, gp96 transfers the peptides to the MHC class I b2

210

M. Shinkai · A. Ito

microglobulin complexes.Wells et al. demonstrated that stably transfected B16 melanoma cells that constitutively expressed the human HSP70 exhibited significantly increased levels of MHC class I antigens on their surface [79, 80]. We have revealed that the augmentation of MHC class I antigens on the tumor cell surface via HSP70 expression caused the immunity induction [74]. The HSP70 expression reached a maximum at 24 h after heating, and the augmentation of MHC class I surface expression started 24 h after heating and reached a maximum 48 h after heating as shown in Fig. 15. The expression of other immunologic mediators, such as intracellular adhesion molecule-1 (ICAM-1) and MHC class II antigens, did not increase. In an in vivo experiment using immunocompetent syngenic rats (F344), growth of the heated T-9 cells, with augmentation of MHC class I antigen surface expression, was significantly inhibited, while the cells grew progressively in nude rats (F344/N Jcl-rnu). Furthermore, compared with lymphocytes from non-immunized (only PBS injection) rats or rats injected with non-heated T-9 cells, the splenic lymphocytes of the rats in which the heated T-9 cells were injected displayed specific cytotoxicity against T-9 cells. These results suggest that HSP70 is an important modulator of tumor cell immunogenicity during hyperthermia. An alternative mechanism how antigens of tumor cells are recognized by the host immune system in hyperthermia is cross-presentation of antigenic peptides by professional antigen presenting cells (APCs). The HSPs-mediated antitumor immunity may be caused by a vaccine-like effect of HSP-peptide complexes released from dying tumor cells. The released HSP-peptide complexes encounter APCs that express specific receptors such as CD91 [81]. Interaction of HSP-peptide complexes with CD91 leads to receptor-mediated endocytosis, processing of the antigenic peptide by the endogenous MHC class I pathway and re-presentation on the cell surface to CD8 positive T cell receptor [81]. Additionally, HSPs functions as a direct activator of APCs, stimulating cytokine secretion from monocytes and inducing the maturation of dendritic cells (DCs) through CD14 and/or CD91 receptors [82]. This cytokine-like ability of HSP70 to stimulate the innate immune system is independent of the peptides they chaperone, suggesting that HSP70 is a natural adjuvant. We previously demonstrated that HSP70 expression following hyperthermia using MCLs induced antitumor immunity in T-9 rat glioma [75]. Our hyperthermia system using MCLs overcame thermotolerance and induced necrotic cell death that was correlated with HSP70 expression. We purified the HSP70-peptide complexes from the tumor after hyperthermia, and found that rats immunized with T-9-derived HSP70 strongly suppressed T-9 tumor growth, suggesting that HSP70 chaperons some antigenic peptides after hyperthermia (Fig. 16). When the necrotic cell death was induced by hyperthermia using MCLs, HSP70 was released from dying T-9 cells. In order to investigate the vaccine effect of the tumor cells killed via hyperthermiainduced necrosis in rats, tumor rejection assay after hyperthermia treatment of implanted T-9 cells was performed, and tumor growth was strongly suppressed and 50% of these rats were protected from challenge with T-9 cells

Functional Magnetic Particles for Medical Application

211

Fig. 15 Augmentation of MHC class I surface antigen on heated cells. Heated or non-heated cells were harvested at 6, 24, 48 and 72 h after heating, and flow cytometric analyses were carried out using isotype control antibody (Control), anti-MHC class I (MHC I), or antiICAM-1 antibody (ICAM-1). Dotted line: non-heated, solid line: 43 °C, 1 h

Fig. 16a,b Immunogenicity of tumor-derived HSP70-peptide complexes: a silver-stained SDS-polyacrylamide gels and immunoblots with the antiHSP70 antibody of purified HSP70-peptide complexes from T-9 tumor and liver; b antitumor effect of tumor-derived HSP70-peptide complexes. Rats were immunized with 20 mg of HSP70-peptide complexes derived from T-9 tumor and liver as indicated. The nonimmunized group was injected with PBS. Immunization was carried out in 200-ml volumes subcutaneously three times with weekly intervals. Rats were challenged with 1¥106 T-9 cells subcutaneously one week after the third immunization. Each line represents the kinetics of tumor growth in a single rat. P<0.01 for nonimmunized vs tumor-derived HSP70-treated groups. P<0.02 for liver derived HSP70-treated vs tumor-derived HSP70-treated groups. P=0.6 for nonimmunized vs liver derived HSP70-treated groups

212 M. Shinkai · A. Ito

Functional Magnetic Particles for Medical Application

213

as shown in Fig. 17. These results suggest that via HSP70-peptide complexes released during necrotic tumor cell death in vivo, our hyperthermia system confers antitumor immunity. This phenomenon, which may be called in situ vaccination, has important implications in the development of novel antitumor therapies. A proposed scenario in which HSPs function during successive stages of an antitumor response after hyperthermia was summarized and illustrated in Fig. 18. 1. A poorly immunogenic tumor cell with a low concentration of intracellular HSP-peptide complexes, decreased function of the endogenous antigen processing machinery and a very low level of MHC class I-peptide complexes at the cell surface. 2.A sublethal stress response induced by hyperthermia using MCLs results in raised levels of intracellular HSP-peptide complexes, enhanced processing of endogenous antigens and an increase in the density of MHC class I-peptide complexes at the cell surface. These tumor cells were then recognized directly by MHC class I-restricted, tumor-specific T cells. 3. Dying tumor cells, which are killed by the tumor-specific T-cells or by lethal hyperthermia treatment, release their intracellular contents, including HSP-peptide complexes. 4. The released HSPs and/or antigenic peptides activate neighboring monocytes to produce proinflammatory cytokines and recruit APCs. 5. The HSPpeptide complexes are taken up by DCs and in turn presented to tumor-specific T cells via MHC class I (cross-priming).

Fig. 17 Antitumor immunity induced via tumor cell death by hyperthermia. T-9 cells were treated with MCLs and implanted subcutaneously into rats, and irradiated with AMF. The temperature of the injected sites was maintained at 42 °C for 30 min. This hyperthermia treatment was carried out three times with weekly intervals. The nonimmunized group was injected with PBS. Rats were challenged with 1¥106 T-9 cells subcutaneously one week after the last immunization. Each line represents the kinetics of tumor growth in a single rat. *Animals cured of tumor, n

214

M. Shinkai · A. Ito

Functional Magnetic Particles for Medical Application

215

Recently, a vaccine consisting of autologous tumor-derived heat shock protein gp96-peptide complexes (HSPPC-96, Oncophage; Antigenics, Inc,Woburn, MA) has entered clinical trials, and the feasibility in metastatic melanoma patients was demonstrated [83]. Since the HSP-peptide complexes have to be extracted from tumors in the body, surgery is needed in this therapeutic protocol. If hyperthermia by means of MCLs is used, the surgery and extraction will be unnecessary. Udono et al. reported that vaccination effect of HSPpeptide complexes was directly dependent on the dose [84]. When the HSPs in the tumor are regarded as an antigen source, 1 g (approximately 108 cells) of tumor tissue may contain approximately 2 mg HSP70. This represents a much higher dose than that used for the clinical trials. In our hyperthermia system, the expression of HSP70 was enhanced, and the tissue lysis via necrosis was observed throughout the tumor. Our hyperthermia system, using MCLs, can produce HSPs-peptide complexes (possibly including HSPs such as HSP90 and gp 96), resulting in vaccination therapy. 5.6 Heat Immunotherapy Using Magnetic Particles As described in the previous section, hyperthermia (HT) and immunotherapy (IT) is congenial. Interleukin-2 (IL-2) augments the activity of cytotoxic T lymphocytes (CTLs) [85] and has a broad range of immunologic effects, such as inducing specific T helper cells, natural killer and lymphokine-activated killer cells [86]. Combination of IL-2 and hyperthermia is effective for our proposed scenario illustrated in Fig. 18. Due to these effects, IL-2 is widely used in cancer therapy to enhance cellular immunity and the cytotoxic activity of effector cells [87]. Granulocyte-macrophage colony-stimulating factor (GM-CSF) is a hematopoietic growth factor that stimulates proliferation and differentiation of granulocytic and monocytic progenitor cells [88–90]. Because of this effect on

Fig. 18 Hypothesis for the mechanism of antitumor immunity induced by hyperthermia. (1) A poorly immunogenic tumor cell with a low concentration of intracellular HSP-peptide complexes (HSP-PC) (Low [HSP]), decreased function of the endogenous antigen processing machinery and a very low level of MHC class I at the cell surface [74]. (2) A sublethal stress response induced by hyperthermia using MCLs results in raised levels of intracellular HSP-peptide complexes, enhanced processing of endogenous antigens and an increase in the density of MHC class I-peptide complexes at the cell surface (Augmentation of MHC class I) [74]. These tumor cells were then recognized directly by MHC class I-restricted, tumor-specific T cells. (3) Dying tumor cells, which are killed by the tumor-specific T-cells (2) or by lethal hyperthermia treatment using MCLs, release their intracellular contents, including HSP-PC (Tumor cell death) [75]. (4) The released HSPs and/or antigenic peptides activate neighboring monocytes to produce proinflammatory cytokines such as tumor necrosis factor-a (TNF-a) and interleukin-6 (IL-6) and recruit APCs such as macrophages and dendritic cells (Recruitment of APCs) [75, 95]. (5) The HSP-PC are taken up by APCs and in turn presented to tumor-specific T cells via MHC class I (Cross-priming)

216

M. Shinkai · A. Ito

hematopoietic cells, GM-CSF has been used to counter myelosuppression induced by high-dose chemotherapy, to bolster host defense in immunocompromised patients, and to accelerate recovery after bone marrow transplantation [91, 92]. It has also been reported that GM-CSF plays an important role in the activation of APCs [93]. Dranoff et al. [94] reported that vaccination with irradiated GM-CSF-transduced murine melanoma cells induced a potent systemic immune response against the melanoma cells. Therefore, combination of GM-CSF and hyperthermia is also effective for our proposed scenario illustrated in Fig. 18. From our and other many researchers’ results and some hypothesizes mentioned above, we propose an establishment of “Heat Immunotherapy”. When applying hyperthermia using MCLs as a tumor vaccine, administration of cytokines may supplement the vaccine-induced response against tumor cells. For this purpose, we used IL-2 and GM-CSF in order to establish a “Heat Immunotherapy” [95]. As shown in Fig. 19, complete regression of tumors was observed in mice treated with hyperthermia using MCLs plus IL-2 and GM-CSF (75% (6/8) and 40% (4/10) of the mice, respectively). This study supports the combined use of IT and IH using MCLs in patients with advanced malignancies.As mentioned above, professional APCs play an important role in eliciting a primary immune response to tumor antigens chaperoned by HSPs. Enhancement of APC function by cytokines can potentially occur at two levels; directly on APCs during generation and maturation, and indirectly by acting on the

Fig. 19 Time course of tumor growth in each group after hyperthermia. Group I (control, n=10), group II (hyperthermia, n=10), group III (IL-2, n=8), group IV (GM-CSF, n=10), group V (hyperthermia+IL-2), and group VI (hyperthermia+GM-CSF). Each line represents tumor growth kinetics of in a single mouse. Crosses (¥) indicate when each mouse died

Functional Magnetic Particles for Medical Application

217

elicited T cell response. The most important finding was that GM-CSF, a cytokine playing an important role in the maturation and function of professional APCs, proved to be a powerful immunostimulant after hyperthermia using MCLs. IL-2 is a cytokine known for its ability to elicit T cell response, and has been used in APC-based tumor vaccines [96]. Survival was prolonged by administration of IL-2, whereas no significant effect on survival was observed in mice by administration of GM-CSF. Because IL-2 has a broad range of immunologic effects, an effect on survival was observed even without vaccination by hyperthermia using MCLs. On the other hand, injection of GM-CSF alone had no effect on survival. When treatment with GM-CSF was combined with hyperthermia using MCLs, 40% (4/10) of mice survived for 120 days. These results suggest that the antitumor effects of GM-CSF may require (or be enhanced by) vaccination by hyperthermia using MCLs. The roles of IL-2 and GM-CSF in combination with IH, and the involvement of specialized APCs and elicited effector cells in this therapy, remain to be fully elucidated. In order to establish “Heat Immunotherapy”, further study should be needed to determine the optimal combination of hyperthermia using MCLs with cytokines or other immunostimulants.

6 Conclusion In this article we have only been able to highlight a limited number of medical applications using magnetic particles. These techniques are based on biochemistry, electronics and magnetics, and physiology. Currently, magnetic techniques are complementary to other methods presently used in medical applications and this combination therapy should result in more effective medical treatment. Greater understanding of the properties of magnetic particles will increase their potential for medical application.

References 1. 2. 3. 4. 5. 6. 7. 8. 9. 10.

Dunnill P, Lilly MD (1974) Biotechnol Bioeng 16:987 Mosbach K, Anderson L (1977) Nature 270:259 Kondo A, Kamura H, Higashitani K (1994) Appl Microbiol Biotech 41:99 Kondo A, Fukuda H (1997) J Ferment Bioeng 84:337 Stechell CH (1985) J Chem Technol Biotechnol 35B:175 Barclay AN, Brown ML, Beyers AD, Davis SJ, Somoza C, Williams AF (1993) The leukocytes antigen facts book. Academic Press, San Diego, CA Häfeli U, Shutt W, Teller J, Zborowski M (1997) (eds) Scientific and clinical application of magnetic carriers. Plenum Press, New York Guesdon JL, Thiery R, Avrameas S (1987) J Allergy Clin Immnol 6:23 Druet E, Mahieu P, Foidart JM, Druet P (1982) J Immnol Methods 48:149 Pazzagli M, Kohen F, Sufi S, Masironi B, Cekan SZ (1988) J Immunol Methods 114:62

218

M. Shinkai · A. Ito

11. Vonk GP, Schram JL (1991) J Immunol Methods 137:133 12. Shinkai M, Wang J, Kamihira M, Iwata M, Honda H, Kobayashi T (1992) J Ferment Bioeng 73:166 13. Bergemann C, Müller-Schulte D, Oster J, à Brassard L, Lübbe AS (1999) J Magn Magn Mater 194:45 14. Nagatani N, Shinkai M, Honda H, Kobayashi T (2000) Biotechnol Lett 22:999 15. Shinkai M,Yanase M, Honda H,Wakabayashi T,Yoshida J, Kobayashi T (1996) Jpn J Cancer Res 87:1179 16. Schneider S, Rusconi S (1996) Biotechniques 21:876 17. Padmanabhan R, Corsico CD, Howard TH, Holter W, Fordis CM,Willingham M, Howard BH (1988) Anal Biochem 170:341 18. Widder KJ, Senyei AE, Scarpelli DG (1978) Proc Soc Exp Biol Med 58:141 19. Morimoto Y, Okumura M, Sugibayashi K, Kato Y (1981) J Pharm Dyn 4:624 20. Widder KJ, Morris RM, Poore G, Howard DP Jr, Senyei AE (1981) Proc Natl Acad Sci 78:579 21. Suzuki M, Shinkai M,Yanase M, Ito A, Honda H, Kobayashi T (1999) Jpn J Hyperthermic Oncol 15:79 22. Magin RL,Wright SM, Niesman MR, Chan HC, Swartz HM (1986) Magn Reson Med 3:440 23. Gillis P, Koenig SH (1987) Magn Reson Med 5:323 24. Saini S, Stark DD, Hahn PF,Wittenberg J, Brady TJ, Ferrucci JT (1987) Radiology 162:211 25. Stark DD,Weissleder R, Elizondo G, Hahn PF, Saini S, Todd LE,Wittenberg J, Ferrucci JT (1988) Radiology 168:297 26. Josephson L, Groman EV, Menz E, Lewis JM, Bengele H (1990) Magn Reson Imaging 8:637 27. Suzuki M, Honda H, Kobayashi T, Wakabayashi T, Yoshida J, Takahashi M (1996) Brain Tumor Pathol 13:127 28. Bushida K, Mohri K, Katoh T, Kobayashi A (1996) IEEE Trans Mag 32:4944 29. Shinkai M, Ohshima A,Yanase M, Uchiyama T, Mohri K,Wakabayashi T,Yoshida J, Honda H, Kobayashi T (1998) Kagaku Kogaku Ronbunnshu 24:174 (in Japanese) 30. Cavaliere R, Ciocatto EC, Giovanella BC, Heidelburger C, Jonson RO, Margottini M, Mondovi B, Moricca BG, Rossi-Fanelli A (1967) Cancer 20:1351 31. Ikeda N, Hayashida O, Kameda H, Ito H, Matsuda T (1994) Int J Hyperthermia 10:553 32. Lin JC, Wang YJ (1987) Int J Hyperthermia 3:37 33. Stauffer PR, Cetas TC, Fletcher AM, DeYoung DW, Dewhirst MW, Oleson JR, Roemer RB (1984) IEEE Trans Biomed Eng 31:76 34. Takemori S, Tazawa K, Nagae H, Yamashita I, Kato H, Kasagi T, Maeda M, Honda T, Fujimaki M (1991) Drug Delivery System 6:465 35. Jordan A, Wust P, Fähling H, John W, Hinz A, Felix R (1993) Int J Hyperthermia 9:51 36. Mitsumori M, Hiraoka M, Shibata T, Okuno Y, Masunaga S, Konishi M, Okajima K, Nagata Y, Nishimura Y, Abe M et al. (1994) Int J Hyperthermia 10:785 37. Gilchrist RK, David VC (1957) Annal Surgery 146:596 38. Weinstock SB, Brain JD (1988) J Appl Phys 65:1811 39. Lonnemark M, Hemmingsson A, Bach-Gansmo T, Hagberg H, Magnusson A, Gundersen HG, Nyman R (1991) Acta Radiologica 32:232 40. Gordon RT, Hines JR, Gordon D (1979) Med Hypothesis 5:83 41. Jordan A, Wust P, Scholz R, Tesch B, Fähling H, Mitrovics T, Vogl T, Cervós-Navarro J, Felix R (1996) Int J Hyperthermia 12:705 42. Chan D, Kirpotin DB, Bunn PA (1993) J Magn Magn Mater 122:374 43. Shinkai M, Matsui M, Kobayashi T (1994) Jpn J Hyperthermic Oncol 10:168 44. Suzuki M, Shinkai M, Honda H, Kobayashi T (2003) Melanoma Res 13:129

Functional Magnetic Particles for Medical Application

219

45. Yanase M, Shinkai M, Honda H, Wakabayashi T, Yoshida J, Kobayashi T (1998) Jpn J Cancer Res 89:463 46. Matsuno H, Tohnai I, Mitsudo K, Hayashi Y, Ito M, Shinkai M, Kobayashi T, Yoshida J, Ueda M (2001) Jpn J Hyperthermic Oncol 17:141 47. Ohno T, Wakabayashi T, Takemura A,Yoshida J, Ito A, Shinkai M, Honda H, Kobayashi T (2002) J Neuro Oncol 56:233 48. Shinkai M, Ueno K, Honda H, Kobayashi T (2002) Jpn J Hyperthermic Oncol 18:191 49. Le B, Shinkai M, Kitade T, Honda H,Yoshida J,Wakabayashi T, Kobayashi T (2001) J Chem Eng Jpn 34:66 50. Shinkai M, Le B, Honda H, Yoshikawa K, Shimizu K, Saga S, Wakabayashi T, Yoshida J, Kobayashi T (2001) Jpn J Cancer Res 92:1138 51. Gossen M, Bujard H (1992) Proc Natl Acad Sci USA 89:5547 52. No D, Yao TP, Evans RM (1996) Proc Natl Acad Sci USA 93:3346 53. Tsai SY, O’Malley BW, DeMayo FJ, Wang Y, Chua SS (1998) Adv Drug Deliv Rev 30:23 54. Pollock R, Issner R, Zoller K, Natesan S, Rivera VM, Clackson T (2000) Proc Natl Acad Sci USA 97:13221 55. Joki T, Nakamura M, Ohno T (1995) Human Gene Ther 6:1507 56. Morimoto R, Tissieres A, Georgopoulos C (1990) (eds) Stress proteins in biology and medicine. Cold Spring Harbor Laboratory Press, Cold Spring Harbor 57. Dreano M, Brochot J, Myers A, Cheng-Meyer C, Rungger D,Voellmy R, Bromley P (1986) Gene 49:1 58. Huang Q, Hu JK, Lohr F, Zhang L, Braun R, Lauzen J, Little JB, Dewhirst MW, Li CY (2000) Cancer Res 60:3435 59. Park JS, Luethy JD,Wang MG, Fargnoli J, Fornace AJ Jr, McBride OW, Holbrook NJ (1992) Gene 116:259 60. Luethy JD, Fargnoli J, Park JS, Fornace AJ Jr, Holbrook NJ (1990) J Biol Chem 265:16521 61. Luethy JD, Holbrook NK (1992) Cancer Res 52:5 62. Ropp M, Courgeon AM, Calvayrac R, Best-Belpomme M (1983) Can J Biochem Cell Biol 61:456 63. Ito A, Shinkai M, Bouhon IA, Honda H, Kobayashi T (2000) Jpn J Hyperthermic Oncol 16:91 64. Yonezawa M, Otsuka T, Matsui N, Tsuji H, Kato KH, Moriyama A, Kato T (1996) Int J Cancer 66:347 65. Friedman AD (1996) Cancer Res 56:3250 66. Ito A, Shinkai M, Bouhon IA, Honda H, Kobayashi T (2000) J Biosci Bioeng 90:437 67. Ito A, Shinkai M, Honda H, Kobayashi T (2001) Cancer Gene Ther 8:649 68. Ito A, Shinkai M, Hakamada K, Honda H, Kobayashi T (2001) J Biosci Bioeng 92:598 69. Yanase M, Shinkai M, Honda H, Wakabayashi T, Yoshida J, Kobayashi T (1998) Jpn J Cancer Res 89:775 70. Schlesinger MJ (1990) J Biol Chem 265:12111 71. Mosser DD, Caron AW, Bourget L, Meriin AB, Sherman MY, Morimoto RI, Massie B (2000) Mol Cell Biol 20:7146 72. Ménoret A, Chandawarkar R (1998) Sem Immunol 25:654 73. Srivastava PK, Ménoret A, Basu S, Binder R, Quade K (1998) Immunity 8:657 74. Ito A, Shinkai M, Honda H, Wakabayashi T, Yoshida J, Kobayashi T (2001) Cancer Immunol Immunother 50:515 75. Ito A, Shinkai M, Honda H, Yoshikawa K, Saga S, Wakabayashi T, Yoshida J, Kobayashi T (2003) Cancer Immunol Immunother 52:80 76. Srivastava PK, Maki RG (1991) Curr Top Microbiol Immunol 167:109 77. Srivastava PK, Heike M (1993) Semin Immunol 3:57

220

M. Shinkai · A. Ito

78. 79. 80. 81. 82.

Srivastava PK, Udono H, Blachere NE, Li Z (1994) Immunogenetics 39:93 Wells AD, Malkovsky M (2000) Immunol Today 21:129 Wells AD, Rai SK, Salvato MS, Band H, Malkovsky M (1998) Int Immunol 10:609 Basu S, Binder RJ, Ramalingam T, Srivastava PK (2001) Immunity 14:303 Asea A, Kraeft SK, Kurt-Jones EA, Stevenson MA, Chen LB, Finberg RW, Koo GC, Calderwood SK (2000) Nat Med 6:435 Belli F, Testori A, Rivoltini L, Maio M, Andreola G, Sertoli MR, Gallino G, Piris A, Cattelan A, Lazzari I, Carrabba M, Scita G, Santantonio C, Pilla L, Tragni G, Lombardo C, Arienti F, Marchiano A, Queirolo P, Bertolini F, Cova A, Lamaj E, Ascani L, Camerini R, Corsi M, Cascinelli N, Lewis JJ, Srivastava P, Parmiani G (2002) J Clin Oncol 20:4169 Udono H, Srivastava PK (1993) J Exp Med 178:1391 Farrar WL, Johnson HM, Farrar JJ (1981) J Immunol 126:1120 Rosenberg SA, Grimm EA, McGrogan M, Doyle M, Kawasaki E, Koths K, Mark DF (1984) Science 223:1412 Hadden JW (1988) Cancer Detect Prev 12:537 Gasson JC,Weisbart RH, Kaufman SE, Clark SC, Hewick RM,Wong GG, Golde DW (1984) Science 226:1339 Horiguchi J, Warren MK, Kufe D (1987) Blood 69:1259 Burgess AW, Metcalf D (1980) Blood 56:947 Brandt SJ, Peters WP, Atwater SK, Kurtzberg J, Borowitz MJ, Jones RB, Shpall EJ, Bast RC Jr, Gilbert CJ, Oette DH (1988) N Engl J Med 318:869 Antman KS, Griffin JD, Elias A, Socinski MA, Ryan L, Cannistra SA, Oette D, Whitley M, Frei E III, Schnipper LE (1988) N Engl J Med 319:593 Steinman RM (1991) Annu Rev Immunol 9:271 Dranoff G, Jaffee E, Lazenby A, Golumbek P, Levitsky H, Brose K, Jackson V, Hamada H, Pardoll D, Mulligan RC (1993) Proc Natl Acad Sci USA 90:3539 Ito A, Tanaka K, Kondo K, Shinkai M, Honda H, Matsumoto K, Saida T, Kobayashi T (2003) Cancer Sci 94:308 Shimizu K, Fields RC, Giedlin M, Mulé JJ (1999) Proc Natl Acad Sci USA 96:2268

83.

84. 85. 86. 87. 88. 89. 90. 91. 92. 93. 94. 95. 96.

Received: September 2003

Author Index Volumes 51 – 91 Author Index Volumes 1–50 see Volume 50

Ackermann, J.-U. see Babel, W.: Vol. 71, p. 125 Adam, W., Lazarus, M., Saha-Möller, C. R., Weichhold, O., Hoch, U., Häring, D., Schreier, Ü.: Biotransformations with Peroxidases. Vol. 63, p. 73 Ahring, B. K.: Perspectives for Anaerobic Digestion. Vol. 81, p. 1 Ahring, B. K. see Angelidaki, I.: Vol. 82, p. 1 Ahring, B. K. see Gavala, H. N.: Vol. 81, p. 57 Ahring, B. K. see Hofman-Bang, J.: Vol. 81, p. 151 Ahring, B. K. see Mogensen, A. S.: Vol. 82, p. 69 Ahring, B. K. see Pind, P. F.: Vol. 82, p. 135 Ahring, B. K. see Skiadas, I. V.: Vol. 82, p. 35 Akhtar, M., Blanchette, R. A., Kirk, T. K.: Fungal Delignification and Biochemical Pulping of Wood. Vol. 57, p. 159 Allan, J. V., Roberts, S. M., Williamson, N. M.: Polyamino Acids as Man-Made Catalysts. Vol. 63, p. 125 Allington, R. W. see Xie, S.: Vol. 76, p. 87 Al-Rubeai, M.: Apoptosis and Cell Culture Technology. Vol. 59, p. 225 Al-Rubeai, M. see Singh, R. P.: Vol. 62, p. 167 Alsberg, B. K. see Shaw, A. D.: Vol. 66, p. 83 Angelidaki, I., Ellegaard, L., Ahring, B. K.: Applications of the Anaerobic Digestion Process. Vol. 82, p. 1 Angelidaki, I. see Gavala, H. N.: Vol. 81, p. 57 Angelidaki, I. see Pind, P. F.: Vol. 82, p. 135 Antranikian, G. see Ladenstein, R.: Vol. 61, p. 37 Antranikian, G. see Müller, R.: Vol. 61, p. 155 Archelas, A. see Orru, R. V. A.: Vol. 63, p. 145 Argyropoulos, D. S.: Lignin. Vol. 57, p. 127 Arnold, F. H., Moore, J. C.: Optimizing Industrial Enzymes by Directed Evolution. Vol. 58, p. 1 Autuori, F., Farrace, M. G., Oliverio, S., Piredda, L., Piacentini, G.: “Tissie” Transglutaminase and Apoptosis. Vol. 62, p. 129 Azerad, R.: Microbial Models for Drug Metabolism. Vol. 63, p. 169 Babel, W., Ackermann, J.-U., Breuer, U.: Physiology, Regulation and Limits of the Synthesis of Poly(3HB). Vol. 71, p. 125 Bajpai, P., Bajpai, P. K.: Realities and Trends in Emzymatic Prebleaching of Kraft Pulp. Vol. 56, p. 1 Bajpai, P., Bajpai, P. K.: Reduction of Organochlorine Compounds in Bleach Plant Effluents. Vol. 57, p. 213 Bajpai, P. K. see Bajpai, P.: Vol. 56, p. 1 Bajpai, P. K. see Bajpai, P.: Vol. 57, p. 213 Banks, M. K., Schwab, P., Liu, B., Kulakow, P.A., Smith, J. S., Kim, R.: The Effect of Plants on the Degradation and Toxicity of Petroleum Contaminants in Soil: A Field Assessment.Vol. 78, p. 75 Barut, M. see Strancar, A.: Vol. 76, p. 49

222

Author Index Volumes 51–91

Bárzana, E.: Gas Phase Biosensors. Vol. 53, p. 1 Basu, S. K. see Mukhopadhyay, A.: Vol. 84, p. 183 Bathe, B. see Pfefferle, W.: Vol. 79, p. 59 Bazin, M. J. see Markov, S. A.: Vol. 52, p. 59 Bellgardt, K.-H.: Process Models for Production of b-Lactam Antibiotics. Vol. 60, p. 153 Beppu, T.: Development of Applied Microbiology to Modern Biotechnology in Japan.Vol.69, p. 41 Berovic, M. see Mitchell, D.A.: Vol. 68, p. 61 Beyeler, W., DaPra, E., Schneider, K.: Automation of Industrial Bioprocesses. Vol. 70, p. 139 Beyer, M. see Seidel, G.: Vol. 66, p. 115 Beyer, M. see Tollnick, C.: Vol. 86, p. 1 Bhardwaj, D. see Chauhan, V.S.: Vol. 84, p. 143 Bhatia, P.K., Mukhopadhyay, A.: Protein Glycosylation: Implications for in vivo Functions and Thereapeutic Applications. Vol. 64, p. 155 Bisaria, V.S. see Ghose, T.K.: Vol. 69, p. 87 Blanchette R. A. see Akhtar, M.: Vol. 57, p. 159 Bocker, H., Knorre, W.A.: Antibiotica Research in Jena from Penicillin and Nourseothricin to Interferon. Vol. 70, p. 35 de Bont, J.A.M. see van der Werf, M. J.: Vol. 55, p. 147 van den Boom, D. see Jurinke, C.: Vol. 77, p. 57 Borah, M. M. see Dutta, M.: Vol. 86, p. 255 Brainard, A. P. see Ho, N. W. Y.: Vol. 65, p. 163 Brazma, A., Sarkans, U., Robinson, A., Vilo, J., Vingron, M., Hoheisel, J., Fellenberg, K.: Microarray Data Representation, Annotation and Storage. Vol. 77, p. 113 Breuer, U. see Babel, W.: Vol. 71, p. 125 Broadhurst, D. see Shaw, A. D.: Vol. 66, p. 83 Bruckheimer, E. M., Cho, S. H., Sarkiss, M., Herrmann, J., McDonell, T. J.: The Bcl-2 Gene Family and Apoptosis. Vol 62, p. 75 Brüggemann, O.: Molecularly Imprinted Materials – Receptors More Durable than Nature Can Provide. Vol. 76, p. 127 Bruggink, A., Straathof, A. J. J., van der Wielen, L. A. M.: A ‘Fine’ Chemical Industry for Life Science Products: Green Solutions to Chemical Challenges. Vol. 80, p. 69 Buchert, J. see Suurnäkki, A.: Vol. 57, p. 261 Bungay, H. R. see Mühlemann, H. M.: Vol. 65, p. 193 Bungay, H.R., Isermann, H.P.: Computer Applications in Bioprocessin. Vol. 70, p. 109 Büssow, K. see Eickhoff, H.: Vol. 77, p. 103 Byun, S.Y. see Choi, J.W.: Vol. 72, p. 63 Cabral, J. M. S. see Fernandes, P.: Vol. 80, p. 115 Cahill, D. J., Nordhoff, E.: Protein Arrays and Their Role in Proteomics. Vol. 83, p. 177 Cantor, C. R. see Jurinke, C.: Vol. 77, p. 57 Cao, N. J. see Gong, C. S.: Vol. 65, p. 207 Cao, N. J. see Tsao, G. T.: Vol. 65, p. 243 Carnell, A. J.: Stereoinversions Using Microbial Redox-Reactions. Vol. 63, p. 57 Cash, P.: Proteomics of Bacterial Pathogens. Vol. 83, p. 93 Cen, P., Xia, L.: Production of Cellulase by Solid-State Fermentation. Vol. 65, p. 69 Chand, S., Mishra, P.: Research and Application of Microbial Enzymes – India’s Contribution. Vol. 85, p. 95 Chang, H. N. see Lee, S. Y.: Vol. 52, p. 27 Chauhan, V. S., Bhardwaj, D.: Current Status of Malaria Vaccine Development. Vol. 84, p. 143 Cheetham, P. S. J.: Combining the Technical Push and the Business Pull for Natural Flavours.Vol. 55, p. 1 Cheetham, P. S. J.: Bioprocesses for the Manufacture of Ingredients for Foods and Cosmetics. Vol. 86, p. 83 Chen, C. see Yang, S.-T.: Vol. 87, p. 61 Chen, Z. see Ho, N. W. Y.: Vol. 65, p. 163

Author Index Volumes 51–91

223

Chenchik, A. see Zhumabayeva, B.: Vol. 86, p. 191 Cho, S. H. see Bruckheimer, E. M.: Vol. 62, p. 75 Cho, G.H. see Choi, J.W.: Vol 72, p. 63 Choi, J. see Lee, S.Y.: Vol. 71, p. 183 Choi, J.W., Cho, G.H., Byun, S.Y., Kim, D.-I.: Integrated Bioprocessing for Plant Cultures. Vol. 72, p. 63 Christensen, B., Nielsen, J.: Metabolic Network Analysis – A Powerful Tool in Metabolic Engineering. Vol. 66, p. 209 Christians, F. C. see McGall, G.H.: Vol. 77, p. 21 Chu, J. see Zhang, S.: Vol. 87, p. 97 Chui, G. see Drmanac, R.: Vol. 77, p. 75 Ciaramella, M. see van der Oost, J.: Vol. 61, p. 87 Contreras, B. see Sablon, E.: Vol. 68, p. 21 Conway de Macario, E., Macario, A. J. L.: Molecular Biology of Stress Genes in Methanogens: Potential for Bioreactor Technology. Vol. 81, p. 95 Cordero Otero, R.R. see Hahn-Hägerdal, B.: Vol. 73, p. 53 Cordwell S. J. see Nouwens, A.S.: Vol. 83, p. 117 Cornet, J.-F., Dussap, C. G., Gros, J.-B.: Kinetics and Energetics of Photosynthetic MicroOrganisms in Photobioreactors. Vol. 59, p. 153 da Costa, M. S., Santos, H., Galinski, E. A.: An Overview of the Role and Diversity of Compatible Solutes in Bacteria and Archaea. Vol. 61, p. 117 Cotter, T. G. see McKenna, S. L.: Vol. 62, p. 1 Croteau, R. see McCaskill, D.: Vol. 55, p. 107 Danielsson, B. see Xie, B.: Vol. 64, p. 1 DaPra, E. see Beyeler, W.: Vol. 70, p. 139 Darzynkiewicz, Z., Traganos, F.: Measurement of Apoptosis. Vol. 62, p. 33 Davey, H. M. see Shaw, A. D.: Vol. 66, p. 83 Dean, J. F. D., LaFayette, P. R., Eriksson, K.-E. L., Merkle, S. A.: Forest Tree Biotechnolgy. Vol. 57, p. 1 Debabov, V. G.: The Threonine Story. Vol. 79, p. 113 Demain, A.L., Fang, A.: The Natural Functions of Secondary Metabolites. Vol. 69, p. 1 Dhar, N. see Tyagi, A. K.: Vol. 84, p. 211 Diaz, R. see Drmanac, R.: Vol. 77, p. 75 Dochain, D., Perrier, M.: Dynamical Modelling, Analysis, Monitoring and Control Design for Nonlinear Bioprocesses. Vol. 56, p. 147 Dolfing, J. see Mogensen, A. S.: Vol. 82, p. 69 Drmanac, R., Drmanac, S., Chui, G., Diaz, R., Hou, A., Jin, H., Jin, P., Kwon, S., Lacy, S., Moeur, B., Shafto, J., Swanson, D., Ukrainczyk, T., Xu, C., Little, D.: Sequencing by Hybridization (SBH): Advantages, Achievements, and Opportunities. Vol. 77, p. 75 Drmanac, S. see Drmanac, R.: Vol. 77, p. 75 Du, J. see Gong, C. S: Vol. 65, p. 207 Du, J. see Tsao, G. T.: Vol. 65, p. 243 Dueser, M. see Raghavarao, K.S.M.S.: Vol. 68, p. 139 Dussap, C. G. see Cornet J.-F.: Vol. 59, p. 153 Dutta, M., Borah, M. M., Dutta, N. N.: Adsorptive Separation of b-Lactam Antibiotics: Technological Perspectives. Vol. 86, p. 255 Dutta, N. N. see Ghosh, A. C.: Vol. 56, p. 111 Dutta, N. N. see Sahoo, G. C.: Vol. 75, p. 209 Dutta, N. N. see Dutta, M.: Vol. 86, p. 255 Dynesen, J. see McIntyre, M.: Vol. 73, p. 103 Eggeling, L., Sahm, H., de Graaf, A. A.: Quantifying and Directing Metabolite Flux: Application to Amino Acid Overproduction. Vol. 54, p. 1 Eggeling, L. see de Graaf, A.A.: Vol. 73, p. 9

224

Author Index Volumes 51–91

Eggink, G., see Kessler, B.: Vol. 71, p. 159 Eggink, G., see van der Walle, G. J. M.: Vol. 71, p. 263 Egli, T. see Wick, L. M.: Vol. 89, p. 1 Ehrlich, H. L. see Rusin, P.: Vol. 52, p. 1 Eickhoff, H., Konthur, Z., Lueking, A., Lehrach, H., Walter, G., Nordhoff, E., Nyarsik, L., Büssow, K.: Protein Array Technology: The Tool to Bridge Genomics and Proteomics. Vol. 77, p. 103 Elias, C. B., Joshi, J. B.: Role of Hydrodynamic Shear on Activity and Structure of Proteins. Vol. 59, p. 47 Ellegaard, L. see Angelidaki, I.: Vol. 82, p. 1 Elling, L.: Glycobiotechnology: Enzymes for the Synthesis of Nucleotide Sugars. Vol. 58, p. 89 Enfors, S.-O. see Rozkov, A.: Vol. 89, p. 163 Eriksson, K.-E. L. see Kuhad, R. C.: Vol. 57, p. 45 Eriksson, K.-E. L. see Dean, J. F. D.: Vol. 57, p. 1 Faber, K. see Orru, R. V. A.: Vol. 63, p. 145 Fahnert, B., Lilie, H., Neubauer, P.: Inclusion Bodies: Formation and Utilisation. Vol. 89, p. 93 Fang, A. see Demain, A.L.: Vol. 69, p. 1 Farrace, M. G. see Autuori, F.: Vol. 62, p. 129 Farrell, R. L., Hata, K., Wall, M. B.: Solving Pitch Problems in Pulp and Paper Processes. Vol. 57, p. 197 Fellenberg, K. see Brazma, A.: Vol. 77, p. 113 Fernandes, P., Prazeres, D. M. F., Cabral, J. M. S.: Membrane-Assisted Extractive Bioconversions. Vol. 80, p. 115 Ferro, A., Gefell, M., Kjelgren, R., Lipson, D. S., Zollinger, N., Jackson, S.: Maintaining Hydraulic Control Using Deep Rooted Tree Systems. Vol. 78, p. 125 Fiechter, A.: Biotechnology in Switzerland and a Glance at Germany. Vol. 69, p. 175 Fiechter, A. see Ochsner, U. A.: Vol. 53, p. 89 Flechas, F. W., Latady, M.: Regulatory Evaluation and Acceptance Issues for Phytotechnology Projects. Vol. 78, p. 171 Foody, B. see Tolan, J. S.: Vol. 65, p. 41 Fréchet, J. M. J. see Xie, S.: Vol. 76, p. 87 Freitag, R., Hórvath, C.: Chromatography in the Downstream Processing of Biotechnological Products. Vol. 53, p. 17 Friehs, K.: Plasmid Copy Number and Plasmid Stability. Vol. 86, p. 47 Furstoss, R. see Orru, R. V. A.: Vol. 63, p. 145 Galinski, E.A. see da Costa, M.S.: Vol. 61, p. 117 Gàrdonyi, M. see Hahn-Hägerdal, B.: Vol. 73, p. 53 Gatfield, I.L.: Biotechnological Production of Flavour-Active Lactones. Vol. 55, p. 221 Gavala, H. N., Angelidaki, I., Ahring, B. K.: Kinetics and Modeling of Anaerobic Digestion Process. Vol. 81, p. 57 Gavala, H. N. see Skiadas, I. V.: Vol. 82, p. 35 Gefell, M. see Ferro, A.: Vol. 78, p. 125 Gemeiner, P. see Stefuca, V.: Vol. 64, p. 69 Gerlach, S. R. see Schügerl, K.: Vol. 60, p. 195 Ghose, T. K., Bisaria, V.S.: Development of Biotechnology in India. Vol. 69, p. 71 Ghose, T. K. see Ghosh, P.: Vol. 85, p. 1 Ghosh, A. C., Mathur, R. K., Dutta, N. N.: Extraction and Purification of Cephalosporin Antibiotics. Vol. 56, p. 111 Ghosh, P., Ghose, T. K.: Bioethanol in India: Recent Past and Emerging Future. Vol. 85, p. 1 Ghosh, P. see Singh, A.: Vol. 51, p. 47 Gilbert, R. J. see Shaw, A. D.: Vol. 66, p. 83 Gill, R.T. see Stephanopoulos, G.: Vol. 73, p. 1 Gomes, J., Menawat, A. S.: Fed-Batch Bioproduction of Spectinomycin. Vol. 59, p. 1

Author Index Volumes 51–91

225

Gong, C. S., Cao, N. J., Du, J., Tsao, G. T.: Ethanol Production from Renewable Resources. Vol. 65, p. 207 Gong, C. S. see Tsao, G. T.: Vol. 65, p. 243 Goodacre, R. see Shaw, A. D.: Vol. 66, p. 83 de Graaf, A. A., Eggeling, L., Sahm, H.: Metabolic Engineering for L-Lysine Production by Corynebacterium glutamicum. Vol. 73, p. 9 de Graaf, A. A. see Eggeling, L.: Vol. 54, p. 1 de Graaf, A. A. see Weuster-Botz, D.: Vol. 54, p. 75 de Graaf, A. A. see Wiechert, W.: Vol. 54, p. 109 Grabley, S., Thiericke, R.: Bioactive Agents from Natural Sources: Trends in Discovery and Application. Vol. 64, p. 101 Griengl, H. see Johnson, D. V.: Vol. 63, p. 31 Gros, J.-B. see Larroche, C.: Vol. 55, p. 179 Gros, J.-B. see Cornet, J. F.: Vol. 59, p. 153 Gu, M. B., Mitchell, R. J., Kim, B. C.: Whole-Cell-Based Biosensors for Environmental Biomonitoring and Application.Vol. 87, p. 269 Guenette M. see Tolan, J. S.: Vol. 57, p. 289 Gupta, M. N. see Roy, I.: Vol. 86, p. 159 Gupta, S. K.: Status of Immunodiagnosis and Immunocontraceptive Vaccines in India. Vol. 85, p. 181 Gutman, A. L., Shapira, M.: Synthetic Applications of Enzymatic Reactions in Organic Solvents. Vol. 52, p. 87 Haagensen, F. see Mogensen, A. S.: Vol. 82, p. 69 Hahn-Hägerdal, B., Wahlbom, C.F., Gárdonyi, M., van Zyl, W.H., Cordero Otero, R.R., Jönsson, L.J.: Metabolic Engineering of Saccharomyces cerevisiae for Xylose Utilization.Vol. 73, p.53 Haigh, J.R. see Linden, J.C.: Vol. 72, p. 27 Hall, D. O. see Markov, S. A.: Vol. 52, p. 59 Hall, P. see Mosier, N. S.: Vol. 65, p. 23 Hammar, F.: History of Modern Genetics in Germany. Vol. 75, p. 1 Hanai, T., Honda, H.: Application of Knowledge Information Processing Methods to Biochemical Engineering, Biomedical and Bioinformatics Field. Vol. 91, p. 51 Hannenhalli, S., Hubbell, E., Lipshutz, R., Pevzner, P. A.: Combinatorial Algorithms for Design of DNA Arrays. Vol. 77, p. 1 Haralampidis, D., Trojanowska, M., Osbourn, A. E.: Biosynthesis of Triterpenoid Saponins in Plants. Vol. 75, p. 31 Häring, D. see Adam, E.: Vol. 63, p. 73 Harvey, N. L., Kumar, S.: The Role of Caspases in Apoptosis. Vol. 62, p. 107 Hasegawa, S., Shimizu, K.: Noninferior Periodic Operation of Bioreactor Systems. Vol. 51, p. 91 Hata, K. see Farrell, R. L.: Vol. 57, p. 197 Hecker, M.: A Proteomic View of Cell Physiology of Bacillus subtilis – Bringing the Genome Sequence to Life. Vol. 83, p. 57 Hecker, M. see Schweder, T.: Vol. 89, p. 47 van der Heijden, R. see Memelink, J.: Vol. 72, p. 103 Hein, S. see Steinbüchel, A.: Vol. 71, p. 81 Hembach, T. see Ochsner, U. A.: Vol. 53, p. 89 Henzler, H.-J.: Particle Stress in Bioreactor. Vol. 67, p. 35 Herrler, M. see Zhumabayeva, B.: Vol. 86, p. 191 Herrmann, J. see Bruckheimer, E. M.: Vol. 62, p. 75 Hewitt, C. J., Nebe-Von-Caron, G.: The Application of Multi-Parameter Flow Cytometry to Monitor Individual Microbial Cell Physiological State. Vol. 89, p. 197 Hill, D. C., Wrigley, S. K., Nisbet, L. J.: Novel Screen Methodologies for Identification of New Microbial Metabolites with Pharmacological Activity. Vol. 59, p. 73 Hiroto, M. see Inada, Y.: Vol. 52, p. 129

226

Author Index Volumes 51–91

Ho, N. W. Y., Chen, Z., Brainard, A. P. Sedlak, M.: Successful Design and Development of Genetically Engineering Saccharomyces Yeasts for Effective Cofermentation of Glucose and Xylose from Cellulosic Biomass to Fuel Ethanol. Vol. 65, p. 163 Hoch, U. see Adam, W.: Vol. 63, p. 73 Hoffmann, F., Rinas, U.: Stress Induced by Recombinant Protein Production in Escherichia coli. Vol. 89, p. 73 Hoffmann, F., Rinas, U.: Roles of Heat-Shock Chaperones in the Production of Recombinant Proteins in Escherichia coli. Vol. 89, p. 143 Hofman-Bang, J., Zheng, D., Westermann, P., Ahring, B. K., Raskin, L.: Molecular Ecology of Anaerobic Reactor Systems. Vol. 81, p. 151 Hoheisel, J. see Brazma, A.: Vol. 77, p. 113 Holló, J., Kralovánsky, U.P.: Biotechnology in Hungary. Vol. 69, p. 151 Honda, H. see Hanai, T.: Vol. 91, p. 51 Honda, H., Kobayashi, T.: Industrial Application of Fuzzy Control in Bioprocesses. Vol. 87, p. 151 Honda, H., Kobayashi, T.: Large-Scale Micropropagation System of Plant Cells. Vol. 91, p. 105 Honda, H., Liu, C., Kobayashi, T.: Large-Scale Plant Micropropagation. Vol. 72, p. 157 Hórvath, C. see Freitag, R.: Vol. 53, p. 17 Hou, A. see Drmanac, R.: Vol. 77, p. 75 Hubbell, E. see Hannenhalli, S.: Vol. 77, p. 1 Huebner, S. see Mueller, U.: Vol. 79, p. 137 Hummel,W.: New Alcohol Dehydrogenases for the Synthesis of Chiral Compounds.Vol.58, p.145 Ikeda, M.: Amino Acid Production Processes. Vol. 79, p. 1 Imamoglu, S.: Simulated Moving Bed Chromatography (SMB) for Application in Bioseparation. Vol. 76, p. 211 Inada, Y., Matsushima, A., Hiroto, M., Nishimura, H., Kodera, Y.: Chemical Modifications of Proteins with Polyethylen Glycols. Vol. 52, p. 129 Iijima, S. see Kamihira, M.: Vol. 91, p. 171 Iijima, S. see Miyake, K.: Vol. 90, p. 89 Irwin, D. C. see Wilson, D. B.: Vol. 65, p. 1 Isermann, H. P. see Bungay, H. R.: Vol. 70, p. 109 Ito, A. see Shinkai, M.: Vol. 91, p. 191 Iwasaki, Y., Yamane, T.: Enzymatic Synthesis of Structured Lipids. Vol. 90, p. 151 Iyer, P. see Lee, Y. Y.: Vol. 65, p. 93 Jackson, S. see Ferro, A.: Vol. 78, p. 125 James, E., Lee, J. M.: The Production of Foreign Proteins from Genetically Modified Plant Cells. Vol. 72, p. 127 Jeffries, T. W., Shi, N.-Q.: Genetic Engineering for Improved Xylose Fementation by Yeasts. Vol. 65, p. 117 Jendrossek, D.: Microbial Degradation of Polyesters. Vol. 71, p. 293 Jenne, M. see Schmalzriedt, S.: Vol. 80, p. 19 Jin, H. see Drmanac, R.: Vol. 77, p. 75 Jin, P. see Drmanac, R.: Vol. 77, p. 75 Johnson, D.V., Griengl, H.: Biocatalytic Applications of Hydroxynitrile. Vol. 63, p. 31 Johnson, E. A., Schroeder, W. A.: Microbial Carotenoids. Vol. 53, p. 119 Johnsurd, S. C.: Biotechnolgy for Solving Slime Problems in the Pulp and Paper Industry. Vol. 57, p. 311 Johri, B. N., Sharma, A., Virdi, J. S.: Rhizobacterial Diversity in India and its Influence on Soil and Plant Health. Vol. 84, p. 49 Jönsson, L. J. see Hahn-Hägerdal, B.: Vol. 73, p. 53 Joshi, J. B. see Elias, C. B.: Vol. 59, p. 47 Jurinke, C., van den Boom, D., Cantor, C. R., Köster, H.: The Use of MassARRAY Technology for High Throughput Genotyping. Vol. 77, p. 57

Author Index Volumes 51–91

227

Kaderbhai, N. see Shaw, A. D.: Vol. 66, p. 83 Kamihira, M., Nishijima, K., Iijima, S.: Transgenic Birds for the Production of Recombinant Proteins. Vol. 91, p. 171 Karanth, N. G. see Krishna, S. H.: Vol. 75, p. 119 Karthikeyan, R., Kulakow, P. A.: Soil Plant Microbe Interactions in Phytoremediation. Vol. 78, p. 51 Kataoka, M. see Shimizu, S.: Vol. 58, p. 45 Kataoka, M. see Shimizu, S.: Vol. 63, p. 109 Katzen, R., Tsao, G.T.: A View of the History of Biochemical Engineering. Vol. 70, p. 77 Kawai, F.: Breakdown of Plastics and Polymers by Microorganisms. Vol. 52, p. 151 Kawarasaki, Y. see Nakano, H.: Vol. 90, p. 135 Kell, D. B. see Shaw, A. D.: Vol. 66, p. 83 Kessler, B., Weusthuis, R., Witholt, B., Eggink, G.: Production of Microbial Polyesters: Fermentation and Downstream Processes. Vol. 71, p. 159 Khosla, C. see McDaniel, R.: Vol. 73, p. 31 Khurana, J. P. see Tyagi, A. K.: Vol. 84, p. 91 Kieran, P. M., Malone, D. M., MacLoughlin, P. F.: Effects of Hydrodynamic and Interfacial Forces on Plant Cell Suspension Systems. Vol. 67, p. 139 Kijne, J.W. see Memelink, J.: Vol. 72, p. 103 Kim, B. C. see Gu, M. B.: Vol. 87, p. 269 Kim, D.-I. see Choi, J.W.: Vol. 72, p. 63 Kim, R. see Banks, M. K.: Vol. 78, p. 75 Kim, Y.B., Lenz, R.W.: Polyesters from Microorganisms. Vol. 71, p. 51 Kimura, E.: Metabolic Engineering of Glutamate Production. Vol. 79, p. 37 King, R.: Mathematical Modelling of the Morphology of Streptomyces Species. Vol. 60, p. 95 Kino-oka, M., Nagatome, H., Taya, M.: Characterization and Application of Plant Hairy Roots Endowed with Photosynthetic Functions. Vol. 72, p. 183 Kino-oka, M., Taya M.: Development of Culture Techniques of Keratinocytes for Skin Graft Production. Vol. 91, p. 135 Kirk, T. K. see Akhtar, M.: Vol. 57, p. 159 Kjelgren, R. see Ferro, A.: Vol. 78, p. 125 Knorre, W.A. see Bocker, H.: Vol. 70, p. 35 Kobayashi, M. see Shimizu, S.: Vol. 58, p. 45 Kobayashi, S., Uyama, H.: In vitro Biosynthesis of Polyesters. Vol. 71, p. 241 Kobayashi, T. see Honda, H.: Vol. 72, p. 157 Kobayashi, T. see Honda, H.: Vol. 87, p. 151 Kobayashi, T. see Honda, H.: Vol. 91, p. 105 Kodera, F. see Inada, Y.: Vol. 52, p. 129 Kolattukudy, P. E.: Polyesters in Higher Plants. Vol. 71, p. 1 König, A. see Riedel, K: Vol. 75, p. 81 de Koning, G. J. M. see van der Walle, G. A. M.: Vol. 71, p. 263 Konthur, Z. see Eickhoff, H.: Vol. 77, p. 103 Koo, Y.-M. see Lee, S.-M.: Vol. 87, p. 173 Kossen, N.W.F.: The Morphology of Filamentous Fungi. Vol. 70, p. 1 Köster, H. see Jurinke, C.: Vol. 77, p. 57 Koutinas, A. A. see Webb, C.: Vol. 87, p. 195 Krabben, P., Nielsen, J.: Modeling the Mycelium Morphology of Penicilium Species in Submerged Cultures. Vol. 60, p. 125 Kralovánszky, U.P. see Holló, J.: Vol. 69, p. 151 Krämer, R.: Analysis and Modeling of Substrate Uptake and Product Release by Procaryotic and Eucaryotik Cells. Vol. 54, p. 31 Kretzmer, G.: Influence of Stress on Adherent Cells. Vol. 67, p. 123 Krieger, N. see Mitchell, D.A.: Vol. 68, p. 61 Krishna, S. H., Srinivas, N. D., Raghavarao, K. S. M. S., Karanth, N. G.: Reverse Micellar Extraction for Downstream Processeing of Proteins/Enzymes. Vol. 75, p. 119

228

Author Index Volumes 51–91

Kuhad, R. C., Singh, A., Eriksson, K.-E. L.: Microorganisms and Enzymes Involved in the Degradation of Plant Cell Walls. Vol. 57, p. 45 Kuhad, R. Ch. see Singh, A.: Vol. 51, p. 47 Kulakow, P. A. see Karthikeyan, R.: Vol. 78, p. 51 Kulakow, P. A. see Banks, M. K.: Vol. 78, p. 75 Kumagai, H.: Microbial Production of Amino Acids in Japan. Vol. 69, p. 71 Kumar, R. see Mukhopadhyay, A.: Vol. 86, p. 215 Kumar, S. see Harvey, N. L.: Vol. 62, p. 107 Kunze, G. see Riedel, K.: Vol. 75, p. 81 Kwon, S. see Drmanac, R.: Vol. 77, p. 75 Lacy, S. see Drmanac, R.: Vol. 77, p. 75 Ladenstein, R., Antranikian, G.: Proteins from Hyperthermophiles: Stability and Enzamatic Catalysis Close to the Boiling Point of Water. Vol. 61, p. 37 Ladisch, C. M. see Mosier, N. S.: Vol. 65, p. 23 Ladisch, M. R. see Mosier, N. S.: Vol. 65, p. 23 LaFayette, P. R. see Dean, J. F. D.: Vol. 57, p. 1 Lammers, F., Scheper, T.: Thermal Biosensors in Biotechnology. Vol. 64, p. 35 Larroche, C., Gros, J.-B.: Special Transformation Processes Using Fungal Spares and Immobilized Cells. Vol. 55, p. 179 Latady, M. see Flechas, F. W.: Vol. 78, p. 171 Lazarus, M. see Adam, W.: Vol. 63, p. 73 Leak, D. J. see van der Werf, M. J.: Vol. 55, p. 147 Lee, J.M. see James, E.: Vol. 72, p. 127 Lee, S.-M., Lin, J., Koo, Y.-M.: Production of Lactic Acid from Paper Sludge by Simultaneous Saccharification and Fermentation. Vol. 87, p. 173 Lee, S. Y., Chang, H. N.: Production of Poly(hydroxyalkanoic Acid). Vol. 52, p. 27 Lee, S. Y., Choi, J.: Production of Microbial Polyester by Fermentation of Recombinant Microorganisms. Vol. 71, p. 183 Lee, Y.Y., Iyer, P., Torget, R.W.: Dilute-Acid Hydrolysis of Lignocellulosic Biomass.Vol. 65, p. 93 Lehrach, H. see Eickhoff, H.: Vol. 77, p. 103 Lenz, R. W. see Kim, Y. B.: Vol. 71, p. 51 Licari, P. see McDaniel, R.: Vol. 73, p. 31 Lievense, L. C., van’t Riet, K.: Convective Drying of Bacteria II. Factors Influencing Survival. Vol. 51, p. 71 Lilie, H. see Fahnert, B.: Vol. 89, p. 93 Lin, J. see Lee, S.-M.: Vol. 87, p. 173 Linden, J. C., Haigh, J. R., Mirjalili, N., Phisaphalong, M.: Gas Concentration Effects on Secondary Metabolite Production by Plant Cell Cultures. Vol. 72, p. 27 Lipshutz, R. see Hannenhalli, S.: Vol. 77, p. 1 Lipson, D. S. see Ferro, A.: Vol. 78, p. 125 Little, D. see Drmanac, R.: Vol. 77, p. 75 Liu, B. see Banks, M. K.: Vol. 78, p. 75 Liu, C. see Honda, H.: Vol. 72, p. 157 Lohray, B. B.: Medical Biotechnology in India. Vol. 85, p. 215 Lueking, A. see Eickhoff, H.: Vol. 77, p. 103 Luo, J. see Yang, S.-T.: Vol. 87, p. 61 Lyberatos, G. see Pind, P. F.: Vol. 82, p. 135 MacLoughlin, P.F. see Kieran, P. M.: Vol. 67, p. 139 Macario, A. J. L. see Conway de Macario, E.: Vol. 81, p. 95 Madhusudhan, T. see Mukhopadhyay, A.: Vol. 86, p. 215 Malone, D. M. see Kieran, P. M.: Vol. 67, p. 139 Maloney, S. see Müller, R.: Vol. 61, p. 155 Mandenius, C.-F.: Electronic Noses for Bioreactor Monitoring. Vol. 66, p. 65

Author Index Volumes 51–91

229

Markov, S. A., Bazin, M. J., Hall, D. O.: The Potential of Using Cyanobacteria in Photobioreactors for Hydrogen Production. Vol. 52, p. 59 Marteinsson, V.T. see Prieur, D.: Vol. 61, p. 23 Marx, A. see Pfefferle, W.: Vol. 79, p. 59 Mathur, R. K. see Ghosh, A. C.: Vol. 56, p. 111 Matsushima, A. see Inada, Y.: Vol. 52, p. 129 Mauch, K. see Schmalzriedt, S.: Vol. 80, p. 19 Mazumdar-Shaw, K., Suryanarayan, S.: Commercialization of a Novel Fermentation Concept. Vol. 85, p. 29 McCaskill, D., Croteau, R.: Prospects for the Bioengineering of Isoprenoid Biosynthesis. Vol. 55, p. 107 McDaniel, R., Licari, P., Khosla, C.: Process Development and Metabolic Engineering for the Overproduction of Natural and Unnatural Polyketides. Vol. 73, p. 31 McDonell, T. J. see Bruckheimer, E. M.: Vol. 62, p. 75 McGall, G.H., Christians, F.C.: High-Density GeneChip Oligonucleotide Probe Arrays. Vol. 77, p. 21 McGovern, A. see Shaw, A. D.: Vol. 66, p. 83 McGowan, A. J. see McKenna, S. L.: Vol. 62, p. 1 McIntyre, M., Müller, C., Dynesen, J., Nielsen, J.: Metabolic Engineering of the Aspergillus. Vol. 73, p. 103 McIntyre, T.: Phytoremediation of Heavy Metals from Soils. Vol. 78, p. 97 McKenna, S. L., McGowan, A. J., Cotter, T. G.: Molecular Mechanisms of Programmed Cell Death. Vol. 62, p. 1 McLoughlin, A. J.: Controlled Release of Immobilized Cells as a Strategy to Regulate Ecological Competence of Inocula. Vol. 51, p. 1 Memelink, J., Kijne, J.W., van der Heijden, R., Verpoorte, R.: Genetic Modification of Plant Secondary Metabolite Pathways Using Transcriptional Regulators. Vol. 72, p. 103 Menachem, S. B. see Argyropoulos, D. S. : Vol. 57, p. 127 Menawat, A. S. see Gomes J.: Vol. 59, p. 1 Menge, M. see Mukerjee, J.: Vol. 68, p. 1 Merkle, S. A. see Dean, J. F. D.: Vol. 57, p. 1 Meyer, H. E. see Sickmann, A.: Vol. 83, p. 141 Mirjalili, N. see Linden, J.C.: Vol. 72, p. 27 Mishra, P. see Chand, S.: Vol. 85, p. 95 Mitchell, D.A., Berovic, M., Krieger, N.: Biochemical Engineering Aspects of Solid State Bioprocessing. Vol. 68, p. 61 Mitchell, R. J. see Gu, M. B.: Vol. 87, p. 269 Miyake, K., Iijima, S.: Bacterial Capsular Polysaccharide and Sugar Transferases. Vol. 90, p. 89 Möckel, B. see Pfefferle, W.: Vol. 79, p. 59 Moeur, B. see Drmanac, R.: Vol. 77, p. 75 Mogensen, A. S., Dolfing, J., Haagensen, F., Ahring, B. K.: Potential for Anaerobic Conversion of Xenobiotics. Vol. 82, p. 69 Moore, J.C. see Arnold, F. H.: Vol. 58, p. 1 Moracci, M. see van der Oost, J.: Vol. 61, p. 87 Mosier, N.S., Hall, P., Ladisch, C.M., Ladisch, M.R.: Reaction Kinetics, Molecular Action, and Mechanisms of Cellulolytic Proteins. Vol. 65, p. 23 Mreyen, M. see Sickmann, A.: Vol. 83, p. 141 Mueller, U., Huebner, S.: Economic Aspects of Amino Acids Production. Vol. 79, p. 137 Mühlemann, H.M., Bungay, H.R.: Research Perspectives for Bioconversion of Scrap Paper. Vol. 65, p. 193 Mukherjee, J., Menge, M.: Progress and Prospects of Ergot Alkaloid Research. Vol. 68, p. 1 Mukhopadhyay, A.: Inclusion Bodies and Purification of Proteins in Biologically Active Forms. Vol. 56, p. 61 Mukhopadhyay, A. see Bhatia, P.K.: Vol. 64, p. 155 Mukhopadhyay, A., Basu, S. K.: Intracellular Delivery of Drugs to Macrophages. Vol. 84, p. 183

230

Author Index Volumes 51–91

Mukhopadhyay, A., Madhusudhan, T., Kumar, R.: Hematopoietic Stem Cells: Clinical Requirements and Developments in Ex-Vivo Culture. Vol. 86, p. 215 Müller, C. see McIntyre, M.: Vol. 73, p. 103 Müller, R., Antranikian, G., Maloney, S., Sharp, R.: Thermophilic Degradation of Environmental Pollutants. Vol. 61, p. 155 Müllner, S.: The Impact of Proteomics on Products and Processes. Vol. 83, p. 1 Nagatome, H. see Kino-oka, M.: Vol. 72, p. 183 Nagy, E.: Three-Phase Oxygen Absorption and its Effect on Fermentation. Vol. 75, p. 51 Nakano, H., Kawarasaki, Y., Yamane, T.: Cell-free Protein Synthesis Systems: Increasing their Performance and Applications. Vol. 90, p. 135 Nakashimada, Y. see Nishio, N.: Vol. 90, p. 63 Nath, S.: Molecular Mechanisms of Energy Transduction in Cells: Engineering Applications and Biological Implications. Vol. 85, p. 125 Nebe-Von-Caron, G. see Hewitt, C. J.: Vol. 89, p. 197 Necina, R. see Strancar, A.: Vol. 76, p. 49 Neubauer, P. see Fahnert, B.: Vol. 89, p. 93 Nielsen, J. see Christensen, B.: Vol. 66, p. 209 Nielsen, J. see Krabben, P.: Vol. 60, p. 125 Nielsen, J. see McIntyre, M.: Vol. 73, p. 103 Nisbet, L.J. see Hill, D.C.: Vol. 59, p. 73 Nishijima, K. see Kamihira, M.: Vol. 91, p. 171 Nishimura, H. see Inada, Y.: Vol. 52, p. 123 Nishio, N., Nakashimada, Y.: High Rate Production of Hydrogen/Methane from Various Substrates and Wastes. Vol. 90, p. 63 Nordhoff, E. see Cahill, D.J.: Vol. 83, p. 177 Nordhoff, E. see Eickhoff, H.: Vol. 77, p. 103 Nouwens, A. S., Walsh, B. J., Cordwell S. J.: Application of Proteomics to Pseudomonas aeruginosa. Vol. 83, p. 117 Nyarsik, L. see Eickhoff, H.: Vol. 77, p. 103 Ochsner, U. A., Hembach, T., Fiechter, A.: Produktion of Rhamnolipid Biosurfactants. Vol. 53, p. 89 O’Connor, R.: Survival Factors and Apoptosis: Vol. 62, p. 137 Ogawa, J. see Shimizu, S.: Vol. 58, p. 45 Ohshima, T., Sato, M.: Bacterial Sterilization and Intracellular Protein Release by Pulsed Electric Field. Vol. 90, p. 113 Ohta, H.: Biocatalytic Asymmetric Decarboxylation. Vol. 63, p. 1 Oliverio, S. see Autuori, F.: Vol. 62, p. 129 van der Oost, J., Ciaramella, M., Moracci, M., Pisani, F.M., Rossi, M., de Vos, W.M.: Molecular Biology of Hyperthermophilic Archaea. Vol. 61, p. 87 Orlich, B., Schomäcker, R.: Enzyme Catalysis in Reverse Micelles. Vol. 75, p. 185 Orru, R.V.A., Archelas, A., Furstoss, R., Faber, K.: Epoxide Hydrolases and Their Synthetic Applications. Vol. 63, p. 145 Osbourn, A. E. see Haralampidis, D.: Vol. 75, p. 31 Oude Elferink, S. J. W. H. see Stams, A. J. M.: Vol. 81, p. 31 Padmanaban, G.: Drug Targets in Malaria Parasites. Vol. 84, p. 123 Panda, A.K.: Bioprocessing of Therapeutic Proteins from the Inclusion Bodies of Escherichia coli. Vol. 85, p. 43 Park, E.Y.: Recent Progress in Microbial Cultivation Techniques. Vol. 90, p. 1 Paul, G.C., Thomas, C.R.: Characterisation of Mycelial Morphology Using Image Analysis. Vol. 60, p. 1 Perrier, M. see Dochain, D.: Vol. 56, p. 147 Pevzner, P. A. see Hannenhalli, S.: Vol. 77, p. 1

Author Index Volumes 51–91

231

Pfefferle, W., Möckel, B., Bathe, B., Marx, A.: Biotechnological Manufacture of Lysine. Vol. 79, p. 59 Phisaphalong, M. see Linden, J.C.: Vol. 72, p. 27 Piacentini, G. see Autuori, F.: Vol. 62, p. 129 Pind, P. F., Angelidaki, I., Ahring, B. K., Stamatelatou, K., Lyberatos, G.: Monitoring and Control of Anaerobic Reactors. Vol. 82, p. 135 Piredda, L. see Autuori, F.: Vol. 62, p. 129 Pisani, F.M. see van der Oost, J.: Vol. 61, p. 87 Podgornik, A. see Strancar, A.: Vol. 76, p. 49 Podgornik, A., Tennikova, T. B.: Chromatographic Reactors Based on Biological Activity. Vol. 76, p. 165 Pohl, M.: Protein Design on Pyruvate Decarboxylase (PDC) by Site-Directed Mutagenesis. Vol. 58, p. 15 Poirier, Y.: Production of Polyesters in Transgenic Plants. Vol. 71, p. 209 Pons, M.-N., Vivier, H.: Beyond Filamentous Species. Vol. 60, p. 61 Pons, M.-N., Vivier, H.: Biomass Quantification by Image Analysis. Vol. 66, p. 133 Prazeres, D. M. F. see Fernandes, P.: Vol. 80, p. 115 Prieur, D., Marteinsson, V.T.: Prokaryotes Living Under Elevated Hydrostatic Pressure. Vol. 61, p. 23 Prior, A. see Wolfgang, J.: Vol. 76, p. 233 Pulz, O., Scheibenbogen, K.: Photobioreactors: Design and Performance with Respect to Light Energy Input. Vol. 59, p. 123 Raghavarao, K. S. M. S., Dueser, M., Todd, P.: Multistage Magnetic and Electrophoretic Extraction of Cells, Particles and Macromolecules. Vol. 68, p. 139 Raghavarao, K. S. M. S. see Krishna, S. H.: Vol. 75, p. 119 Ramanathan, K. see Xie, B.: Vol. 64, p. 1 Raskin, L. see Hofman-Bang, J.: Vol. 81, p. 151 Reuss, M. see Schmalzriedt, S.: Vol. 80, p. 19 Riedel, K., Kunze, G., König, A.: Microbial Sensor on a Respiratory Basis for Wastewater Monitoring. Vol. 75, p. 81 van’t Riet, K. see Lievense, L. C.: Vol. 51, p. 71 Rinas, U. see Hoffmann, F.: Vol. 89, p. 73 Rinas, U. see Hoffmann, F.: Vol. 89, p. 143 Roberts, S. M. see Allan, J. V.: Vol. 63, p. 125 Robinson, A. see Brazma, A.: Vol. 77, p. 113 Rock, S. A.: Vegetative Covers for Waste Containment. Vol. 78, p. 157 Roehr, M.: History of Biotechnology in Austria. Vol. 69, p. 125 Rogers, P. L., Shin, H. S., Wang, B.: Biotransformation for L-Ephedrine Production. Vol. 56, p. 33 Rossi, M. see van der Oost, J.: Vol. 61, p. 87 Rowland, J. J. see Shaw, A. D.: Vol. 66, p. 83 Roy, I., Sharma, S., Gupta, M. N.: Smart Biocatalysts: Design and Applications. Vol. 86, p. 159 Roychoudhury, P. K., Srivastava, A., Sahai, V.: Extractive Bioconversion of Lactic Acid. Vol. 53, p. 61 Rozkov, A., Enfors, S.-O.: Analysis and Control of Proteolysis of Recombinant Proteins in Escherichia coli. Vol. 89, p. 163 Rusin, P., Ehrlich, H. L.: Developments in Microbial Leaching – Mechanisms of Manganese Solubilization. Vol. 52, p. 1 Russell, N.J.: Molecular Adaptations in Psychrophilic Bacteria: Potential for Biotechnological Applications. Vol. 61, p. 1 Sablon, E., Contreras, B., Vandamme, E.: Antimicrobial Peptides of Lactic Acid Bacteria: Mode of Action, Genetics and Biosynthesis. Vol. 68, p. 21 Sahai, V. see Singh, A.: Vol. 51, p. 47

232

Author Index Volumes 51–91

Sahai, V. see Roychoudhury, P. K.: Vol. 53, p. 61 Saha-Möller, C. R. see Adam, W.: Vol. 63, p. 73 Sahm, H. see Eggeling, L.: Vol. 54, p. 1 Sahm, H. see de Graaf, A.A.: Vol. 73, p. 9 Sahoo, G. C., Dutta, N. N.: Perspectives in Liquid Membrane Extraction of Cephalosporin Antibiotics: Vol. 75, p. 209 Saleemuddin, M.: Bioaffinity Based Immobilization of Enzymes. Vol. 64, p. 203 Santos, H. see da Costa, M.S.: Vol. 61, p. 117 Sarkans, U. see Brazma, A.: Vol. 77, p. 113 Sarkiss, M. see Bruckheimer, E. M.: Vol. 62, p. 75 Sato, M. see Ohshima, T.: Vol. 90, p. 113 Sauer, U.: Evolutionary Engineering of Industrially Important Microbial Phenotypes. Vol. 73, p. 129 Scheibenbogen, K. see Pulz, O.: Vol. 59, p. 123 Scheper, T. see Lammers, F.: Vol. 64, p. 35 Schmalzriedt, S., Jenne, M., Mauch, K., Reuss, M.: Integration of Physiology and Fluid Dynamics. Vol. 80, p. 19 Schmidt, J. E. see Skiadas, I. V.: Vol. 82, p. 35 Schneider, K. see Beyeler, W.: Vol. 70, p. 139 Schomäcker, R. see Orlich, B.: Vol. 75, p. 185 Schreier, P.: Enzymes and Flavour Biotechnology. Vol. 55, p. 51 Schreier, P. see Adam, W.: Vol. 63, p. 73 Schroeder, W. A. see Johnson, E. A.: Vol. 53, p. 119 Schubert, W.: Topological Proteomics, Toponomics, MELK-Technology. Vol. 83, p. 189 Schügerl, K., Gerlach, S. R., Siedenberg, D.: Influence of the Process Parameters on the Morphology and Enzyme Production of Aspergilli. Vol. 60, p. 195 Schügerl, K. see Seidel, G.: Vol. 66, p. 115 Schügerl, K.: Recovery of Proteins and Microorganisms from Cultivation Media by Foam Flotation. Vol. 68, p. 191 Schügerl, K.: Development of Bioreaction Engineering. Vol. 70, p. 41 Schügerl, K. see Tollnick, C.: Vol. 86, p. 1 Schumann, W.: Function and Regulation of Temperature-Inducible Bacterial Proteins on the Cellular Metabolism. Vol. 67, p. 1 Schuster, K. C.: Monitoring the Physiological Status in Bioprocesses on the Cellular Level. Vol. 66, p. 185 Schwab, P. see Banks, M. K.: Vol. 78, p. 75 Schweder, T., Hecker, M.: Monitoring of Stress Responses. Vol. 89, p. 47 Scouroumounis, G. K. see Winterhalter, P.: Vol. 55, p. 73 Scragg, A. H.: The Production of Aromas by Plant Cell Cultures. Vol. 55, p. 239 Sedlak, M. see Ho, N. W. Y.: Vol. 65, p. 163 Seidel, G., Tollnick, C., Beyer, M., Schügerl, K.: On-line and Off-line Monitoring of the Production of Cephalosporin C by Acremonium Chrysogenum. Vol. 66, p. 115 Seidel, G. see Tollnick, C.: Vol. 86, p. 1 Shafto, J. see Drmanac, R.: Vol. 77, p. 75 Sharma, A. see Johri, B. N: Vol. 84, p. 49 Sharma, M., Swarup, R.: The Way Ahead – The New Technology in an Old Society. Vol. 84, p. 1 Sharma, S. see Roy, I.: Vol. 86, p. 159 Shamlou, P. A. see Yim, S.S.: Vol. 67, p. 83 Shapira, M. see Gutman, A. L.: Vol. 52, p. 87 Sharp, R. see Müller, R.: Vol. 61, p. 155 Shaw, A. D., Winson, M. K., Woodward, A. M., McGovern, A., Davey, H. M., Kaderbhai, N., Broadhurst, D., Gilbert, R. J., Taylor, J., Timmins, E. M., Alsberg, B. K., Rowland, J. J., Goodacre, R., Kell, D. B.: Rapid Analysis of High-Dimensional Bioprocesses Using Multivariate Spectroscopies and Advanced Chemometrics. Vol. 66, p. 83 Shi, N.-Q. see Jeffries, T. W.: Vol. 65, p. 117

Author Index Volumes 51–91

233

Shimizu, K.: Metabolic Flux Analysis Based on 13C-Labeling Experiments and Integration of the Information with Gene and Protein Expression Patterns. Vol. 91, p. 1 Shimizu, K. see Hasegawa, S.: Vol. 51, p. 91 Shimizu, S., Ogawa, J., Kataoka, M., Kobayashi, M.: Screening of Novel Microbial for the Enzymes Production of Biologically and Chemically Useful Compounds. Vol. 58, p. 45 Shimizu, S., Kataoka, M.: Production of Chiral C3- and C4-Units by Microbial Enzymes. Vol. 63, p. 109 Shin, H. S. see Rogers, P. L.: Vol. 56, p. 33 Shinkai, M., Ito, A.: Functional Magnetic Particles for Medical Application. Vol. 91, p. 191 Sickmann, A., Mreyen, M., Meyer, H. E.: Mass Spectrometry – a Key Technology in Proteome Research. Vol. 83, p. 141 Siebert, P. D. see Zhumabayeva, B.: Vol. 86, p. 191 Siedenberg, D. see Schügerl, K.: Vol. 60, p. 195 Singh, A., Kuhad, R. Ch., Sahai, V., Ghosh, P.: Evaluation of Biomass. Vol. 51, p. 47 Singh, A. see Kuhad, R. C.: Vol. 57, p. 45 Singh, R. P., Al-Rubeai, M.: Apoptosis and Bioprocess Technology. Vol. 62, p. 167 Skiadas, I. V., Gavala, H. N., Schmidt, J. E., Ahring, B. K.: Anaerobic Granular Sludge and Biofilm Reactors. Vol. 82, p. 35 Smith, J. S. see Banks, M. K.: Vol. 78, p. 75 Sohail, M., Southern, E. M.: Oligonucleotide Scanning Arrays: Application to High-Throughput Screening for Effective Antisense Reagents and the Study of Nucleic Acid Interactions. Vol. 77, p. 43 Sonnleitner, B.: New Concepts for Quantitative Bioprocess Research and Development.Vol. 54, p. 155 Sonnleitner, B.: Instrumentation of Biotechnological Processes. Vol. 66, p. 1 Southern, E. M. see Sohail, M.: Vol. 77, p. 43 Srinivas, N. D. see Krishna, S. H.: Vol. 75, p. 119 Srivastava, A. see Roychoudhury, P. K.: Vol. 53, p. 61 Stafford, D.E., Yanagimachi, K.S., Stephanopoulos, G.: Metabolic Engineering of Indene Bioconversion in Rhodococcus sp. Vol. 73, p. 85 Stamatelatou, K. see Pind, P. F.: Vol. 82, p. 135 Stams, A. J. M., Oude Elferink, S. J. W. H., Westermann, P.: Metabolic Interactions Between Methanogenic Consortia and Anaerobic Respiring Bacteria. Vol. 81, p. 31 Stark, D., von Stockar, U.: In Situ Product Removal (ISPR) in Whole Cell Biotechnology During the Last Twenty Years. Vol. 80, p. 149 Stefuca, V., Gemeiner, P.: Investigation of Catalytic Properties of Immobilized Enzymes and Cells by Flow Microcalorimetry. Vol. 64, p. 69 Steinbüchel, A., Hein, S.: Biochemical and Molecular Basis of Microbial Synthesis of Polyhydroxyalkanoates in Microorganisms. Vol. 71, p. 81 Stephanopoulos, G., Gill, R.T.: After a Decade of Progress, an Expanded Role for Metabolic Engineering. Vol. 73, p. 1 Stephanopoulos, G. see Stafford, D. E.: Vol. 73, p. 85 von Stockar, U., van der Wielen, L. A. M.: Back to Basics: Thermodynamics in Biochemical Engineering. Vol. 80, p. 1 von Stockar, U. see Stark, D.: Vol. 80, p. 149 Straathof, A. J. J. see Bruggink, A.: Vol. 80, p. 69 Strancar, A., Podgornik, A., Barut, M., Necina, R.: Short Monolithic Columns as Stationary Phases for Biochromatography. Vol. 76, p. 49 Suehara, K., Yano, T.: Bioprocess Monitoring Using Near-Infrared Spectroscopy. Vol. 90, p. 173 Suryanarayan, S. see Mazumdar-Shaw, K.: Vol. 85, p. 29 Suurnäkki, A., Tenkanen, M., Buchert, J., Viikari, L.: Hemicellulases in the Bleaching of Chemical Pulp. Vol. 57, p. 261 Svec, F.: Capillary Electrochromatography: a Rapidly Emerging Separation Method. Vol. 76, p. 1

234

Author Index Volumes 51–91

Svec, F. see Xie, S.: Vol. 76, p. 87 Swanson, D. see Drmanac, R.: Vol. 77, p. 75 Swarup, R. see Sharma, M.: Vol. 84, p. 1 Tabata, H.: Paclitaxel Production by Plant-Cell-Culture Technology. Vol. 87, p. 1 Tanaka, T. see Taniguchi, M.: Vol. 90, p. 35 Tang, Y.-J. see Zhong, J.-J.: Vol. 87, p. 25 Taniguchi, M., Tanaka, T.: Clarification of Interactions Among Microorganisms and Development of Co-culture System for Production of Useful Substances. Vol. 90, p. 35 Taya, M. see Kino-oka, M.: Vol. 72, p. 183 Taya, M. see Kino-oka, M.: Vol. 91, p. 135 Taylor, J. see Shaw, A. D.: Vol. 66, p. 83 Tenkanen, M. see Suurnäkki, A.: Vol. 57, p. 261 Tennikova, T. B. see Podgornik, A.: Vol. 76, p. 165 Thiericke, R. see Grabely, S.: Vol. 64, p. 101 Thomas, C. R. see Paul, G. C.: Vol. 60, p. 1 Thömmes, J.: Fluidized Bed Adsorption as a Primary Recovery Step in Protein Purification. Vol. 58, p. 185 Timmens, E. M. see Shaw, A. D.: Vol. 66, p. 83 Todd, P. see Raghavarao, K. S. M. S.: Vol. 68, p. 139 Tolan, J. S., Guenette, M.: Using Enzymes in Pulp Bleaching: Mill Applications.Vol. 57, p. 289 Tolan, J. S., Foody, B.: Cellulase from Submerged Fermentation. Vol. 65, p. 41 Tollnick, C. see Seidel, G.: Vol. 66, p. 115 Tollnick, C., Seidel, G., Beyer, M., Schügerl, K.: Investigations of the Production of Cephalosporin C by Acremonium chrysogenum. Vol. 86, p. 1 Torget, R. W. see Lee, Y. Y.: Vol. 65, p. 93 Traganos, F. see Darzynkiewicz, Z.: Vol. 62, p. 33 Trojanowska, M. see Haralampidis, D.: Vol. 75, p. 31 Tsao, D. T.: Overview of Phytotechnologies. Vol. 78, p. 1 Tsao, G. T., Cao, N. J., Du, J., Gong, C. S.: Production of Multifunctional Organic Acids from Renewable Resources. Vol. 65, p. 243 Tsao, G. T. see Gong, C. S.: Vol. 65, p. 207 Tsao, G.T. see Katzen, R.: Vol. 70, p. 77 Tyagi, A. K., Dhar, N.: Recent Advances in Tuberculosis Research in India. Vol. 84, p. 211 Tyagi, A. K., Khurana, J. P.: Plant Molecular Biology and Biotechnology Research in the PostRecombinant DNA Era. Vol. 84, p. 91 Ukrainczyk, T. see Drmanac, R.: Vol. 77, p. 75 Uozumi, N.: Large-Scale Production of Hairy Root. Vol. 91, p. 75 Uyama, H. see Kobayashi, S.: Vol. 71, p. 241 VanBogelen, R. A.: Probing the Molecular Physiology of the Microbial Organism, Escherichia coli using Proteomics. Vol. 83, p. 27 Vandamme, E. see Sablon, E.: Vol. 68, p. 21 Verpoorte, R. see Memelink, J.: Vol. 72, p. 103 Viikari, L. see Suurnäkki, A.: Vol. 57, p. 261 Vilo, J. see Brazma, A.: Vol. 77, p. 113 Vingron, M. see Brazma, A.: Vol. 77, p. 113 Virdi, J. S. see Johri, B. N: Vol. 84, p. 49 Vivier, H. see Pons, M.-N.: Vol. 60, p. 61 Vivier, H. see Pons, M.-N.: Vol. 66, p. 133 de Vos, W.M. see van der Oost, J.: Vol. 61, p. 87 Wahlbom, C.F. see Hahn-Hägerdal, B.: Vol. 73, p. 53 Wall, M. B. see Farrell, R. L.: Vol. 57, p. 197

Author Index Volumes 51–91

235

van der Walle, G. A. M., de Koning, G. J. M., Weusthuis, R. A., Eggink, G.: Properties, Modifications and Applications of Biopolyester. Vol. 71, p. 263 Walsh, B. J. see Nouwens, A.S.: Vol. 83, p. 117 Walter, G. see Eickhoff, H.: Vol. 77, p. 103 Wang, B. see Rogers, P. L.: Vol. 56, p. 33 Wang, R. see Webb, C.: Vol. 87, p. 195 Webb, C., Koutinas, A. A., Wang, R.: Developing a Sustainable Bioprocessing Strategy Based on a Generic Feedstock. Vol. 87, p. 195 Weichold, O. see Adam, W.: Vol. 63, p. 73 van der Werf, M. J., de Bont, J. A. M. Leak, D. J.: Opportunities in Microbial Biotransformation of Monoterpenes. Vol. 55, p. 147 Westermann, P. see Hofman-Bang, J.: Vol. 81, p. 151 Westermann, P. see Stams, A. J. M.: Vol. 81, p. 31 Weuster-Botz, D., de Graaf, A. A.: Reaction Engineering Methods to Study Intracellular Metabolite Concentrations. Vol. 54, p. 75 Weusthuis, R. see Kessler, B.: Vol. 71, p. 159 Weusthuis, R. A. see van der Walle, G. J. M.: Vol. 71, p. 263 Wick, L. M., Egli, T.: Molecular Components of Physiological Stress Responses in Escherichia coli. Vol. 89, p. 1 Wiechert, W., de Graaf, A. A.: In Vivo Stationary Flux Analysis by 13C-Labeling Experiments. Vol. 54, p. 109 van der Wielen, L. A. M. see Bruggink, A.: Vol. 80, p. 69 van der Wielen, L. A. M. see von Stockar, U.: Vol. 80, p. 1 Wiesmann, U.: Biological Nitrogen Removal from Wastewater. Vol. 51, p. 113 Williamson, N. M. see Allan, J. V.: Vol. 63, p. 125 Wilson, D. B., Irwin, D. C.: Genetics and Properties of Cellulases. Vol. 65, p. 1 Winson, M. K. see Shaw, A. D.: Vol. 66, p. 83 Winterhalter, P., Skouroumounis, G. K.: Glycoconjugated Aroma Compounds: Occurence, Role and Biotechnological Transformation. Vol. 55, p. 73 Witholt, B. see Kessler, B.: Vol. 71, p. 159 Wolfgang, J., Prior, A.: Continuous Annular Chromatography. Vol. 76, p. 233 Woodley, J. M.: Advances in Enzyme Technology – UK Contributions. Vol. 70, p. 93 Woodward, A. M. see Shaw, A. D.: Vol. 66, p. 83 Wrigley, S. K. see Hill, D. C.: Vol. 59, p. 73 Xia, L. see Cen, P.: Vol. 65, p. 69 Xie, B., Ramanathan, K., Danielsson, B.: Principles of Enzyme Thermistor Systems: Applications to Biomedical and Other Measurements. Vol. 64, p. 1 Xie, S., Allington, R. W., Fréchet, J. M. J., Svec, F.: Porous Polymer Monoliths: An Alternative to Classical Beads. Vol. 76, p. 87 Xu, C. see Drmanac, R.: Vol. 77, p. 75 Yamane, T. see Nakano, H.: Vol. 90, p. 89 Yamane, T. see Iwasaki, Y.: Vol. 90, p. 135 Yanagimachi, K.S. see Stafford, D.E.: Vol. 73, p. 85 Yang, S.-T., Luo, J., Chen, C.: A Fibrous-Bed Bioreactor for Continuous Production of Monoclonal Antibody by Hybridoma. Vol. 87, p. 61 Yano, T. see Suehara, K.: Vol. 90, p. 173 Yim, S. S., Shamlou, P. A.: The Engineering Effects of Fluids Flow and Freely Suspended Biological Macro-Materials and Macromolecules. Vol. 67, p. 83 Zhang, S., Chu, J., Zhuang, Y.: A Multi-Scale Study on Industrial Fermentation Processes and Their Optimization. Vol. 87, p. 97 Zheng, D. see Hofman-Bang, J.: Vol. 81, p. 151

236

Author Index Volumes 51–91

Zhong, J.-J.: Biochemical Engineering of the Production of Plant-Specific Secondary Metabolites by Cell Suspension Cultures. Vol. 72, p. 1 Zhong, J.-J., Tang, Y.-J.: Submerged Cultivation of Medicinal Mushrooms for Production of Valuable Bioactive Metabolites. Vol. 87, p. 25 Zhuang, Y. see Zhang, S.: Vol. 87, p. 97 Zhumabayeva, B., Chenchik, A., Siebert, P. D., Herrler, M.: Disease Profiling Arrays: Reverse Format cDNA Arrays Complimentary to Microarrays. Vol. 86, p. 191 Zollinger, N. see Ferro, A.: Vol. 78, p. 125 van Zyl, W. H. see Hahn-Hägerdal, B.: Vol. 73, p. 53

Subject Index

Activated sludge process 60 adhE 24 Adventitious shoot primordium 94 Agrobacterium rhizogenes 76 Ajuga reptans 84, 87 Amino acids, MTBSTFA 10 Anaplerotic pathway 20 Anchorage-dependent cells 143 Antibody 202 Artificial neural network (ANN) 54, 125 Artificial seed 76 Artificial vessel culture 178 Automated bioreactor system 163 Auxin 89 Bacillus subtilis 14 Balance equation 81 Batch culture, repeated 91 Bioinformatics 52, 68 Bioreactors 80, 108, 159 –, transgenic 171 Bird embryo culture 174 Birds, transgenic 171 Blastoderm 173 Blender 97 13C

labeling, metabolic flux analysis 1, 8 Ca-alginate gel beads 113 Calcium lactate 176 Cancer 65 Cationic lipids 178 Cell placement model 144 Cell yield 82 Cellular morphology 152 Cellular senescence 152 Chicken 172 Chorio-allantoic membrane 175 CO2 93 Commercial products 141 Conductivity 80 Confluence degree 153 Culture, batch, repeated 91 –, fed-batch 81, 83

–, two-step 87 Culture surface 150 Cultured tissues 138 Dehydration 99 Dimethyl dioctadecyl ammonium bromide (DDAB) 179 Dissolved oxygen (DO) 2 DNA, T- 76 DNA microarray 68 Drug delivery 193 Drum reactor 80 Egg white 173 Eggshell 174 Embryo 125 Embryo culture 174 Embryogenic rice callus 118 Embryonic stem (ES) cell 173 Encapsulation 95 Entner-Doudoroff pathway 23 Enzyme activity 6 – –, comparison 27 Enzymes, activity 6 Epithelial sheet 159 Escherichia coli, metabolic flux distribution 2, 11 Escherichia coli K12 proteins, relative expression level 21 –, acetate metabolism 12 –, glucose metabolism 13 Fed-batch culture 81, 83 Fermentation 58, 65 Fermentation genes 35 Fermentative pathway 24 Fragment mass distribution vectors 9 Fragmentation 99 Functional magnetic particles 191 Fuzzy adaptive resonance theory 57 Fuzzy control 59, 65 Fuzzy neural network (FNN) 55 Fuzzy reasoning 53, 67

238 GC-MS analysis, sample preparation 9 Gene expression 5 – – profiling 6 Gene therapy 203 Gene transfer 178 Genetic algorism (GA) 56 Germ-line transmission 173 Glucokinase 20 Glucose, 1-13C 8 Glycolysis 20, 29 Glyoxylate shunt 26 Glyoxylate shunt genes 35 GM plant 106 gnd 42 Growth rate, specific 85 Hairy root 75, 77, 93 – –, clonal propagation 79 Heat shock 35 HMQC 8 Horseradish 89, 93 Horseradish green hairy roots 93 Hydrodynamic (shear) stress 111, 113, 118, 119, 129, 131 Hyperthermia 196 Image analysis 102, 123, 153 Immobilization 113 Immunity 208 Immunotherapy 215 In situ regeneration 120 Isocitrate lyase 14, 26 Isotope distribution 2 Isotopomer mapping matrices 9 Keratinocytes 146 Kinetic model 143 Knockout mutations 4 Knowledge information processing 52 Lipid vesicles 179 Lipofection 178 Liposomes 197 Magnetic resonance imaging 195 Magnetic separation 192 Maintenance coefficient 85 Manufacturing process 137 Mass isotopomers/metabolic flux analysis 9 Membership function 55 Metabolic flux analysis 2 Metabolic flux distribution (MFD) 2, 40 MFA (Metabolic flux analysis) 2 MFD 2, 40 Micropropagation 106

Subject Index Moloney murine leukemia virus (MoMLV) 171 Monitoring 160 Non-viral gene transfer 178 Oviduct 173 Oxygen 175 Pak-bung 93 Pantropic retroviral vector 171 Pentose phosphate pathway 11, 23 – – – genes 29 Peroxidase 90 pfl 24 pgi 29 Photoautotrophic 93 Polyurethane foam 118 Porous PTFE membrane 176 Primordial germ cells (PGCs) 183 Process control 52, 53, 58, 60, 64, 65 Production rule 55 Protamine 180 Protamine-modified lipid vesicles 180 Protein expression, E. coli, metabolic regulation analysis 15 Protein expression analysis 5 Protein expression level, comparison 27 Pseudotyped viral vectors 184 pyk 40 pykF– mutant, E. coli 37 Quail 172 Reactor, turbine-blade 80 Recombinant DNA technology 172 Recombinant proteins 172 Regulatory genes 35 Replication-competent retroviral (RCR) vectors 182 Replication-defective retroviral (RDR) vectors 182 Retroviral vector 171 Rice callus, embryogenic 118 Ribose-5-phosphate isomerase 43 Ribulose phosphate 3-epimerase 43 Root induction 76 RT-PCR 7 Seed, artificial 93 Shell-less culturing 176 Skin graft 139 Software sensor 60, 64, 65 Successive culture 152 Surrogated eggshell culture 176

Subject Index TCA cycle 26 – – genes 35 Therapeutic proteins 172 Tissue engineering 136 Tracer experiments, NMR/GC-MS 5 Transcript levels 30–32 Transfection reagent 178 Transgene 173 Transgenic animals 171, 172 Transgenic bioreactors 171

239 Vesicular stomatitis virus G protein (VSV-G) 171 Viscous additive 129 Vitamin B2 58 VSV-G pseudotyped vectors 184 Yield coefficient 85