CURRENT TOPICS IN DEVELOPMNTL BIOLOGY V6, Volume 6

CURRENT TOPICS IN

DEVELOPMENTAL BIOLOGY VOLUME 6

ADVISORY BOARD JEAN BRACHET

ERASMO M A R R ~

JAMES D. EBERT

JOHN PAUL

E. PETER GEIDUSCHEK

HOWARD A. SCHNEIDERMAN

EUGENE GOLDWASSER

RICHARD L. SIDMAN

PAUL R. GROSS

HERBERT STERN

CONTRIBUTORS JOHN TYLER BONNER

J. R. TATA

EDWARD C. CANTINO

LOUIS C. TRUESDELL

STUART KAUFFMAN

J. E. VARNER

A. A. NEYFAKH

LEWIS WOLPERT

H. YOMO

CURRENT TOPICS IN

DEVELOPMENTAL B I O L O G Y EDITED BY

A. A. MOSCONA DEPARTMENT OF BIOLOGY T H E UNIVERSITY OF CHICAGO CHICAGO, ILLINOIS

ALBERT0 MONROY C.N.R. LABORATORY OF MOLECULAR EMBRYOLOGY ARCO FELICE (NAPLES), ITALY

VOLUME 6

1971

@

ACADEMIC PRESS New York

London

COPYRIGHT 0 1971, BY ACADEMIC PRESS,INC, ALL RIGHTS RESERVED NO PART OF THIS BOOK MAY BE REPRODUCED IN ANY FORM, BY PHOTOSTAT, MICROFILM, RETRIEVAL SYSTEM, OR ANY OTHER MEANS, WITHOUT WRIlTEN PERMISSION FROM THE PUBLISHERS.

ACADEMIC PRESS, INC. 111 Fifth Avenue,

New York, New York 10003

United Kingdom Edition published by ACADEMIC PRESS, INC. (LONDON) LTD. 24/28 Oval Road, London NWl IDD

LIBRARY OF

CONGRESS CATALOG CARD

NUMBER:66-28604

PRINTED IN THE UNITED STATES OF AMERICA

List of Contributors

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

ix

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

xi

Contents of Previous Volumes

The Direction of Developmental Biology

JOHN TYLER BONNER.............................

XV

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

xxi

Errata Volume 5 CHAPTER

1 . The Induction and Early Events of Germination in the Zoospore of Blasfocladiella emersonii

LOUISC. TRUESDELL AND EDWARD C . CANTINO

.

I I1. I11. IV

.

V. VI . VII . VIII . I X. X

.

Introduction ........................................................ Structure of Nangerminating Zoospores ............................. Behavior of the Spore during Encystment .......................... Mechanics of Flagellar Retraction and Rotation of the Nuclear Apparatus .......................................................... Fine-Structural Changes in Encystment .............................. Punctuation on Three Important Structural Changes in Germination . Macromolecular Synthesis during Germination ...................... Environmental Influences on Zoospores ............................ Kinetics of Encystment ............................................. Concluding Remarks .............................. ............ ............ References ........................................

.

CHAPTER

1 3 6 8 10 17 21 23 35 39 43

2. Steps of Realization of Genetic Information in Early Development

A . A . NEYFAKH I. I1. I11. IV . V. VI .

............ Introduction ......................................... Transcription ........................................ Transport of RNA to the Cyboplasm ................................ Translation ......................................................... Regulation of Enzymatic Activity .................................... Conclusion .......................................................... References .......................................................... V

45 46 54 61

72 74 75

vi

CONTENTS

CHAPTER

3 . Protein Synthesis during Amphibian Metamorphosis

J . R . TATA I. I1. I11. IV . V. VI . VII .

Introducti.on ........................................................ The Role of Hormones in Amphibian Metamorphosis ................ Proteins Involved in Metammorphosis ................................ Regulation of Protein Synthesis during Metamorphosis ............... The Role of DNA Synthesis ........................................ Requirement of RNA and Protein Synthesis for Tissue Resorption ... Conclusions and Future Problems ................................... References ..........................................................

CHAPTER

79 80 81

83 99 100 105 107

4 . Hormonal Control of a Secretory Tissue

H . YOMOAND J . E . VARNER Text ....................................................................... References .................................................................

CHAPTER

5.

111 143

Gene Regulation Networks: A Theory for Their Global Structure and Behaviors

STUART KAUFFMAN I. I1. I11. IV . V. VI . VII . VIII . I X.

.

X XI . XI1. XI11 XIV . XV XVI .

. .

Introduction ........................................................ Global Behaviors of Gene Control Systems .......................... Homeostasis: Constrained Dynamic Behavior ....................... Model Systems ...................................................... One-Input Control Systems .......................................... Multiple-Input Control Systems ..................................... Forcing Structures in Switching Nets ................................ The Size of Forcing Structures as a Function of the Number of Inputs per Element in Model Genetic Control Nets ........................ Behavior as a Function of the Size of a Model Genetic Control System, and the Number of Control Inputs per Model Gene ................. Biological Implications .............................................. Expected Character of Forcing Struotures as a Function of the Number of Forcing Connections ............................................. Control Advantages of Forcing Structures ........................... Molecular Mechanisms .............................................. Additional Evidence for the Theory ................................. Alternative Theories ................................................ Conclusions and Summary .......................................... References ..........................................................

145 146 148 140 150 151 151 156 156 160 170 172 173 174 179 180 181

CONTENTS CHAPTER

vii

6 . Positional Information and Pattern Formation

LEWISWOLPERT I. I1. I11. IV .

v.

VI . VII . VIII . I X. X. X I. XI1. XI11. XIV .

xv .

XVI .

Introduction ........................................................ Pattern and Form .................................................. French Flag Problem ............................................... Pattern Regulation .................................................. Universality and Prepatterns ........................................ Model Systems and Mechanisms .... ............................ Polarity ............................................................ Intercellular Communication ......................................... Interpretation ....................................................... Precision ............................................................ Cell Movement ..................................................... Mosaic Development ............................................... Growth and Cell Divisiron .......................................... Spacing Patterns .................................................... Pattern Formation in Plants ........................................ Conclusion .......................................................... References ..........................................................

183 184 186 188 192 196 207 209 211 212 215 217 218 219 220 220 221

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

225

Author Index Subject Index

232

This Page Intentionally Left Blank

LIST OF CONTRIBUTORS Xumbers in parentheses indicate the pages on which the authors' contributions begin.

JOHNTYLERBONNER,Department of Biology, Princeton University, Princeton, New Jersey (xv) EDWARD C. CANTINO,Department of Botany and Plant Pathology, Michigan State University, East Lansing, Michigan (1) STUART KAUFFMAN, Department of Theoretical Biology, and Department of Medicine, University of Chicago, Chicago, Illinois (145) A. A. NEYFAKH, Institute of Developmental Biology, U S S R Academy of Sciences, Moscow, USSR (45) J. R. TATA,National Znstitute for Medical Research, Mill Hill, London, England (79) LOUISC. TRUESDELL, Department of Botany and Plant Pathology, Michigan State University, East Lansing, Michigan (1) J . E. VARNER, M S U / A E C Plant Research Laboratory and Department of Biochemistry, Michigan State University, East Lansing, Michigan (111)

LEWISWOLFERT,Department of Biology as Applied to Medicine, The Middlesex Hospital Medical School, London, England (183) H. YOMO,"M S U / A E C Plant Research Laboratory and Department of Biochemistry, Michigan State University, East Lansing, Michigan (111)

* Present address: Kitchawan Research Laboratory, The Brooklyn Botanic Garden, 712 Kitchawan Road, Ossining, New York. ix

This Page Intentionally Left Blank

CONTENTS OF PREVIOUS VOLUMES Volume 1

REMARKS Joshua Lederberg ON “MASKED” FORMS OF MESSENGER RNA IN EARLY EMBRYOGENESIS AND IN OTHERDIFFERENTIATING SYSTEMS A. S. Spirin THETRANSCRIPTION OF GENETIC INFORMATION IN THE SPIRALIAN EMBRYO J . R. Collier SOMEGENETICAND BIOCHEMICAL ASPECTSOF THE REGULATORY PROGRAM FOR SLIME MOLDDEVELOPMENT Maurice Sussman THHMOLECULAR BASISOF DIFFERENTIATION IN EARLY DEVELOPMENT OF AMPHIBIANEMBRYOS H . Tiedemann THECULTURE OF FREE PLANT CELLSAND ITSSIGNIFICANCE FOR EMBRYOLOGY AND MORPHOGENESIS F . C. Steward, Ann E . Kent, and Marion 0. Mapes GENETIC AND VARIEGATION MOSAICS I N T H E EYEOF Drosophila Hans Joachim Becker BIOCHEMICAL CONTROL OF ERYTHROID CELLDEVELOPMENT Eugene Goldwasser DEVELOPMENT OF MAMMALIAN ERYTHROID CELLS Paul A. Marks and John S. Kovach GENETIC ASPECTSOF SKINA N D LIMBDEVELOPMENT P. F . Goetinck AUTHORINDEX-~UB,JECT INDEX

Volume 2

THECONTROL OF PROTEIN SYNTHESIS I N EMBRYONIC DEVELOPMENT AND DIFFERENTIATION Paul R. Gross xi

xii

CONTENTS OF PREVIOUS VOLUMES

THE GENES FOR RIBOSOMAL RNA AND THEIRTRANSACTION DURING AMPHIBIANDEVELOPMENT Donald D. Brown RIBOSOME AND ENZYME CHANGES DURING MATURATION AND GERMINATION OF CASTOR BEANSEED Erasmo MarrB CONTACT AND SHORT-RANGE INTERACTION AFFECTINGGROWTH OF ANIMAL CELLSIN CULTURE Michael Stoker AN ANALYSISOF THE MECHANISM OF NEOPLASTIC CELLTRANSFORMATION BY POLYOMA VIRUS,HYDROCARBONS, AND X-IRRADIATION Leo Sachs DIFFERENTIATION OF CONNECTIVE TISSUES Frank K. Thorp and Albert Dorfman THEIGA ANTIBODYSYSTEM Mary A n n South, M a x D . Cooper, Richard Hong, and Robert A . Good TERATOCARCINOMA : MODEL FOR A DEVELOPMENTAL CONCEPT OF CANCER G. Barry Pierce CELLULAR AND SUBCELLULAR EVENTS IN WOLFFIAN LENSREGENERATION Tuneo Yamada AUTHORINDEX-SUBJECT INDEX

Volume 3 SYNTHESIS OF MACROMOLECULES A N D MORPHOGENESIS I N Acetabularza J . Brachet BIOCHEMICAL STUDIES OF MALEGAMETOGENESIS IN LILIACFDUS PLANTS Herbert Stern and Yasuo Hotta SPECIFIC INTERACTIONS BETWEEN TISSUES DURING ORGANOGENESIS Etienne Wolff LOW-RESISTANCE JUNCTIONS BETWEEN CELLSI N EMBRYOS AND TISSUE CULTURE Edwin J . Furshpan and David D . Potter COMPUTER ANALYSISOF CELLULAR INTERACTIONS F. Heinmets CELLAGGREGATION AND DIFFERENTIATION I N Dictyostelium Giinther Gerisch HORMONE-DEPENDENT DIFFERENTIATION OF MAMMARY GLAND in Vitro Roger W . Turlcington AUTHORINDEX-SUBJECT INDEX

C O N T E N T S O F PREVIOUS VOLUMES

...

Xlll

Volume 4

GENETICS AND GENESIS Clifford Grobstein THEOUTGROWING BACTERIAL ENDOSPORE Alex Keynan CELLULAR ASPECTSOF MUSCLE DIFFERENTIATION in Vitro David Yafle MACROMOLECULAR BIOSYNTHESIS I N ANIMAL C E L L S INFECTED W I T H CYTOLYTIC VIRUSES Bernard Roizman and Patricia G . Spear THEROLEOF THYROID AND GROWTH HORMONES IN NEUROGENESIS M a x H a m burgh INTERRELATIONSHIPS OF NUCLEAR AND CYTOPLASMIC ESTROGEN RECEPTORS Jack Gorski, G. Shyamala, and D . Toft TOWARD A MOLECULAR EXPLANATION FOR SPECIFIC CELLADHESION Jack E . Lilien THEBIOLOGICAL SIGNIFICANCE OF TURNOVER OF THE SURFACE MEMBRANE OF ANIMALCELLS Leonard Warren AUTHORINDEX-SUBJECT INDEX Volume 5 DEVELOPMENTAL BIOLOGY A N D GENETICS : A P L ~FOR A COOPERATION Albert0 Monroy REGULATORY PROCESSES IN THE MATURATION AND EARLY CLEAVAGE OF AMPHIBIANEGGS L. D.Smith and R. E . Ecker ON THE LONG-TERM CONTROLOF NUCLEAR ACTIVITY DURING CELL DIFFERENTIATION J. B. Gurdon and H . R. Woodland THEINTEGRITY OF THE REPRODUCTIVE CELLLINEIN THE AMPHIBIA Antonie W. Blackler REGULATION OF POLLEN TUBEGROWTH Hansferdinand Linskens and Marianne Kroh PROBLEMS OF DIFFERENTIATION IN THE VERTEBRATE LENS R u t h M . Clayton RECONSTRUCTION OF MUSCLEDEVELOPMENT AS A SEQUENCE OF MACROMOLECULAR SYNTHESES Heinz Herrmann, Stuart M . Heywood, and Ann C . Marchok

xiv

CONTENTS OF PREVIOUS VOLUMES

THESYNTHESIS AND ASSEMBLY OF MYOFIBRILS IN EMBRYONIC MUSCLE Donald A . Fischman THE T-Locus OF THE MOUSE: IMPLICATIONS FOR MECHANISMS OF DEVELOPMENT Salome Glueclcsohn- Waelsch and Robert P . Erickson DNA MASKING I N MAMMALIAN CHROMATIN : A MOLECULAR MECHANISM FOR DETERMINATION OF CELLTYPE 3. Paul AUTHORINDEX-SUBJECTINDEX

THE DIRECTION OF DEVELOPMENTAL BIOLOGY John Tyier Bonner DEPARTMENT OF BIOLOGY, PRINCETON UNIVERSITY PRINCETON, N E W JERSEY

It has been said many times (and that includes prefaces to previous volumes in this series) that the rise of molecular biology has opened up the exciting possibility of a deeper understanding of developmental biology. The central dogma has transformed genetics and now there is a great rush to have it transform development. Buk in the heat of the moment I sometimes wonder whether we might forget what are the main problems. Perhaps it would be fairer to say, that we tend to forget many of the main problems, for certainly one of the greatest among them is gene action, and every bit of effort that has been concentrated on its molecular mechanisms has been of immense importance. However, without detracting from the crucial significance of the problem of gene action it should be reiterated that it is not the whole problem of development, but only a part of it (see Lederberg, 1966; Grobstein 1969; Monroy, 1970). It must be understood that this is said with a basic assumption that there is not one, unique, key problem of development (as Mendel’s laws dominate the study of genetics) but many. I have made this point before and indicated that this may be one of the reasons why we feel that progress in our understanding of developmental biology has been so slow. I n our desire to simplify and solve problems, it is quite understandable that we should seek the controlling mechanism in development. There is undoubtedly an inner sense of satisfaction in the notion that if a process involves a whole series of complex steps, then there is a master step that somehow controls everything. The immediate temptation is to ascribe this role to the genome, and indeed the genome does come closest to such a control headquarters. But the danger of this total genetical approach to developmental biology is that it leaves us with an incomplete picture of how an organism grows and differentiates. By analogy, it is as though a military historian .described a great xv

xvi

THE DIRECTION OF DEVELOPMENTAL BIOLOGY

battle by giving us a detailed account of all the thoughts and orders given out by the opposing generals; to include transcription and translation, our historian also might have described how the generals wrote down their messages, how they were carried to the field commanders, and how the field commanders barked them out to the troops. But that alone would be a very peculiar histmy and would give us very little idea of the actual battle, even assuming there were no mistakes and every order was carried out perfectly. A good history, on the other hand, would include s complete description of all the troops; how many infantry, cavalry, artillery, and so forth; how these were deployed in space before and after the orders were received; how the terrain affected the movement of the troops; how the movement and actions of one group of soldiers affected the actions of neighboring groups. Armies are made up of men, horses, and machines, and each of these units is limited in what it can do, and in the specific ways its actions can be triggered or suppressed. To win a battle or produce an organism, generals and genomes need their armies ; alone they are but an unreal abstraction. Returning now to the problems of development, let us ask where we should look, in addition to the expected absorption ‘in the activities of DNA, RNA, and the synthesis of proteins. What do we need for a complete and whole history? First, we need to understand the role of all the other parts of the cell that, lie beyond the chromosomes. In most higher organisms this is the part that comes primarily from the mother, for the father’s contribution is relatively modest. One way this has been approached, with notable success, is to examine how the genome contributes to building up the egg cytoplasm. Another way has been to study the activity of the cytoplasmic DNA, which plays a sign’ificant role in cytoplasmic inheritance. But besides these established DNA-related processes, there are many organelles, including a variety of membranes, which are a “given” in the system. These organelles are capable of self-duplication and growth; to what extent they are gene-dependent is, for the most part, unknown. But the important point is that they are essential parts of the whole system-as essential as the chromosomes-and without describing their contribution, the analysis of development is incomplete. To put the matter another way, deveiopment involves cells, not just genomes. The cell is a unit that can be broken up by abstractions, but one cannot build an organism with cell parts. I n this respect the armies in a battle remain a useful analogy; isolated troops or solitary generals are by themselves quite useless. Having sa’id that the cell is a basic unit of development, let us now turn to the problems that are not directly included in the study of gene

THE DIRECTION OF DEVELOPMENTAL BIOLOGY

xvii

action. First, let me put them in the traditional words of the experimental embryologist, and then let me restate them in a more generalized way. As Ebert (1968) has stressed in his use of the word “interacting,” a developing organism is constantly producing substances in one part which affect the activities of another part; in some instances, the effect is directly on the genome, in other cases it is directly on the cytoplasm. This statement encompasses the whole range of interactions in development, starting with Spemann’s classical studies on embryonic induction at the beginning of this century; it includes the evidence of gradients in developing organisms, which were first revealed in remarkable detail by C. M. Child (1941) and others, such as Dalq and Pasteels; and it also includes modern work on enzyme induction by hormones in embryonic systems (Moscona, 1971). A related generalization is the notion of organization. This term has always disturbed me because it covers everything; what does not have organiza’tion? Yet there are spec’ific aspects of pattern in cells and in cell groups within developing systems which are of enormous importance, and in this case are often remarkably independent of the genome. One example is that the polarity of cortical structures in ciliate protozoa examined by Sonneborn (1970) ; the other is the recent work on polarity in Fucus eggs by Jaffe (1969), both elegant and important studies. Another ancient and consistently absorbing problem of development is the ability of developing organisms to restore lost parts, or rearrange existing parts into a new whole. We call this regeneration or regulation (respectively), and Driesch (1907) included it in his grandiose expression, “harmonious equipotential system.” The fact that Driesch was concerned with these matters already at the turn of the century serves to emphasize that the problems of development are not newly arisen. I could go on with examples, all of which would also be connected with the establishmen6 of pattern. But let me turn to more modern questions. How can we investigate these venerable problems in molecular terms so as to completely describe and interpret the history of egg to adult? I would like to answer this question on two levels. The first is that we must go beyond the study of the structure and reactivity of biologically significant compounds, to their physical chemistry as well. This means, for instance, that the rates of activity of enzyme reactions, and how they are affected by substrate and product levels must be analyzed in detail, for in this way all chemical reactions within a living system interact. A number of workers have considered this in the more generalized terms of orderly changes in a steady-state system, and B. E. Wright (1968, and in preparation) has applied this concept to specific substances in the development of cellular slime molds.

xviii

THE DIRECTION O F DEVELOPMENTAL BIOLOGY

Another physicochemical property of substances is their ability to form three-dimensional structures. The whole foundation of molecular genetics rests on the double helix of DNA; it is therefore hardly necessary to convince anybody of the need for this approach. But we want to know more; we want to understand the molecular structure of simple and complex organelles, from ribosomes to basal bodies; how do the chemical constituents come together in their specific form? Perhaps the greatest question of all is the formation of membranes which play so many vital roles in cells and organisms. There is no doubt in our minds that the form of these structures can be interpreted in terms of the properties of molecules, but our present answers to specific questions are either nonexistent or hopelessly rudimentary. A good example of the significance of the physical properties of molecules in development is seen in the very important work on adhesion between cells and its role in tissue formation, cell recognition, and the sorting out of cells (Moscona, 1965, 1968). Here, the problems go beyond identifying the cell surface molecules responsible for some cells adhering together and others not (the straight biochemical solution), and extend to the kinds of effects that different degrees of adhesion between various cell types have on the progress of development (Steinberg, 1970). To give a totally different kind of example, electrical curren8ts can be generated by gradients along the cell axes of developing Fucus eggs (Jaffe, 1969); furthermore, the current is sufficient to permit electrophoresis of, for instance, proteins to the two poles of the elongating zygote. It would be possible to give other examples, but these few cases should suffice to support the view that developmental biology urgently needs a large dose of physical biochemistry. There is a wholly different approach to developmental biology that could also play a significant role. This is theoretical biology. Mathematicians interested in biology have made, in some areas, very large contributions. One need only look a t population genetics, and the more modern population biology, to be convinced of this fact. The application of mathematical models to developmental problems is hardly new: Rashevsky and Turing were pioneers and, more recently, there has been great interest in this endeavor by a large number of individuals. I n a rough way one can classify this theoretical approach into three main categories: interactional, spatial, and temporal. Models involving complex systems which interact provide ways of conceiving of stability and orderly change in embryos. Stuart Kauffman, in this volume, gives a perfect example of the value of this approach. Spatial models go back to Rashevsky. He, and later Turing, showed how diffusion, in particular, could play a significant role 'in pattern for-

THE DIRECTION O F DEVELOPMENTAL BIOLOOY

xix

mation. Wolpert, also in this volume, discusses the kind of “positional” information necessary to account for regulation and proportional development. Finally, there has been much recent interest in temporal models. As Goodwin and Cohen (1969) have pointed out, time clocks in the form of internal oscillators not only could account for the timing of certain events during development, but such oscillators can be translated into pattern information as well. The reason that these mathematical models have not yet achieved in developmental biology the status that they have achieved in genetics and population biology is that there still seems to be a large gap between the theory and the facts. It is no fault of the theory; rather it is that we are short on the necessary details about developmental processes. The basic mechanisms involved can often be aptly described by numerous models which differ widely. For instance, models about pattern could (as we have already implied) involve diffusion mechanisms, postulated internal oscillators, or even electrical fields, or any combination of these. One immediately wants to know what method is, in fact, being used by the developing system (not forgetting that different systems might use different methods). This brings us back to the importance of understanding the physical chemistry of biological compounds. But, since there are so many things on that score of which we are ignorant, the immediate value of theoretical studies in developmental biology is t o provide new and imaginative ways of looking a t old problems. As a stimulus to experimentation it is an important adjunct, but it seems to me that a t this stage in the history of our science it can be no more than that; we are still too short on facts to produce anything more than the most tentative models. As we learn more molecular facts about embryos, faclts both physical and chemical, let us not overlook one overriding consideration which I have not yet mentioned. No matter how detailed our analysis of development becomes, it must be remembered that development is not just a series of quasi-stable states, one leading into another ‘in an orderly fashion, but that it has also arisen by natural selection. This means that the possibilities facing any developing organism are not limited solely by chemical and physical consiiderations, but by selective considerations as well. All the possible physicochcmical permutations are held tight within the confines of what is selectively advantageous. While this is an easy statement to make in a brief introductory essay, in fact, this matter has profound and yet unexplored implications for the understanding of the mechanisms by which an organism develops from a fertilized egg into a complex adult.

xx

T H E DIRECTION OF DEVELOPMENTAL BIOLOGY

REFERENCES Child, C. M. (1941). “Patterns and Problems of Development.” University of Chicago Press, Chicago, Illinois. Driesch, H. (1907). “The Science and Philosophy of the Organism.” Black, London. Ebert, J. T. (1968). Curr. T o p . Develop. Biol. 3, xv. Goodwin, B. C . and M. H. Cohen (1969). J . Theor. Biol. 25, 49. Grobstein, C. (1969). Curr. Top. Develop. Biol. 4, xv. Jaffe, L. (1969). In “Communicahion in Development” (A. Lang, ed.), p. 83. Academic Press, New York. Lederberg, J. (1966). Curr. T o p . Develop. Biol. 1, ix. Monroy, A. (1970). Curr. T o p . Develop. Biol. 5, xvii. Moecona, A. A. (1965). In “Cells and Tissues in Culture” (E. N. Willmer, ed.), p. 489. Aoademic Press, New York. Moscona, A. A. (1968). Develop. Biol. 18, 250. Mosoona, A. A. (1971). In “Hormones in Development” (M. Hamburgh and E. J. W. Barrington, eds.), p. 169. Appleton, New York. Sonneborn, T. M. (1970). Proc. Roy. SOC.Ser. B.176,347. Steinberg, M. S. (1970). J. Em. Zool. 173, 395. Wright, B. E. (1968). In “Systems Theory in Biology” (M. D. Mesarovic, ed.). Springer-Verlag, Berlin and New York.

Volume 5 p. 286, 6 lines from the bottom, the reference to the page number should read p. 304 instead of p. 16 p. 292, Fig. 3, Part (2), the labels

D and K on the right side of the figure should

be reversed p. 299, the first left-hand horizontal column should be labeled t', to match the label above the sixth vertical column on p. 298

This Page Intentionally Left Blank

CHAPTER 1

THE INDUCTION AND EARLY EVENTS OF GERMINATION IN THE ZOOSPORE OF Blastocladiellu emersonii Louis C. Truesdell and Edward C. Cantino DEPARTMENT O F BOTANY A N D PLANT PATHOLOGY, MICHIGAN STATE UNIVERSITY, EAST LANSING, MICHIGAN

I. Introduction.. . . . . . . . . . ............................... 1 11. Structure of Nongermin Zoospores.. .................... 3 111. Behavior of the Spore durin A. Flagellar Retraction and B. Vacuole Formation.. . . . C. Volume Changes during D. Reaumb of Major Structural Changes in Encystment.. ...... 8 IV. Mechanics of Flagellar Retraction and Rotation of the Nuclear 8 Apparatus ................................................ V. Fine-Structural Changes in Encystment. . . . . . . . . . . . . . . . . . 10 A. Formation of Myelinlike Figures.. ....................... 11 B. Structural Changes prior to Flagellar Retraction, . . . . . . . . . . . 11 C. Structural Changes during Flagellar Retraction. . . . . . . . . . . . . 14 D. Structural Changes after Flagellar Retraction. . . . . . . . . . . . . . 14 VI. Punctuation on Three Important Structural Changes in Germi17 nation .................................................... A. Changes in the Backing Membrane.. ..................... 18 B. Breakdown of Nuclear Cap Membrane.. . . . . . C. Role of the Gamma Particle in Cell Wall V I I . Macromolecular Synthesis during Germinati WIT. Environmental Influences on Zoospores. ...................... 23 A. Effects of Low Temperatures.. ........................... 23 B. Self-Inhibition in Spore Populations. ...................... 27 C. Alternative Means of Effecting Encystment.. . . . . . . . . . . . . . . 34 IX. Kinetics of Encystment. . . . . . . . . . . . . . 35 A. The Normal Distribu ...................... 35 B. Comparison of Induction Methods.. ...................... 38 X. Concluding Remarks. ...... . . . . . . . . . . . . . . . . . . . . . . . .39 References. ............................................... 43

1. Introduction

Interest in the fungus spore is legendary. In particular, considerable inquiry and debate has centered on diverse aspects of spore germinstion-how to assess it, the morphological and physiological changes asso1

2

LOUIS C. TRUESDELL AND EDWARD C. CANTINO

ciated with it, external factors affecting it, etc. For some species, much is known about the similarities and differences between the spore and its immediate “progeny,” the germling, and about many of the things connecting them in time. But, it is also common knowledge that the primary events responsible for initiation of germination in fungal spores remain elusive. An understanding of germination mechanisms among the zoospores of water fungi has been, in general, superficial, and it has lagged behind that for some of their nonmotile counterparts. Yet, i t seems to US, the very nature of germination in zoospores-accentuated, as it is, by the process of e n c y s t m e n t m a y provide a special sort of handle for studying certain aspects of the earliest events that trigger germination. I n this review, we shall try to illustrate this point, using the uniflagellate zoospores of the aquatic fungus Blastocladiella emersonii (Cantino and Hyatt, 1953). The zoospores of this fungus, being rather frail and ephemeral cells, do not possess any obvious resistive capacity against adverse environmental conditions. They have an endogenous Qo2which ranks well with that of many vigorously metabolizing organisms. There are probably two main reasons for this high respiratory rate: they spend much of their time swimming actively, and they are constantly battling with the aquatic environment to maintain osmotic balance because they are not provided with a cell wall. On these and other bases, i t could well be argued that they are not dormant cells. But, if we accept a definition of dormancy such as Sussman’s (1965, p. 934), namely, “any rest period or reversible interruption of the phenotypic development of the organism,” then the zoospores of B. emersonii have to be labeled dormant. They do not “grow,” they do not display pronounced morphological changes (other than amoeboid gesticulations), they do not synthesize detectable amounts of ribonucleic acid (RNA) , deoxyribonucleic acid (DNA), or protein (Lovett, 1968)-indeed, they actually consume protein, not to mention lipid and polysaccharide, as they navigate about (Suberkropp and Cantino, 1971). Thus, while keeping most if not all of their biosynthetic machinery shut down, the zoospores must simultaneously maintain energy-producing pathways in operative condition. The first gross morphological change in germination is the conversion of the zoospore into an essentially spherical, cystlike cell. This is accompanied by retraction of the single flagellum into the main body of the spore, loss of motility, extensive changes in fine structure (Lovett, 1968; Sol1 et al., 19691, increase in oxygen uptake (Cantino et al., 1969), and onset of macromolecular synthesis (Lovett, 1968; Schmoyer and Lovett, 1969). These transformations occur so rapidly that they can

1.

INDUCTION OF GERMINATION IN

B. emersonii

3

be measured conveniently in seconds or minutes. We refer to this process as “encystment.” A small germ tube subsequently emerges from the cell, and later gives rise to a branched rhizoidal system. The cyst itself enlarges and eventually becomes a large, multinucleate, coenocytic thallus. Numerous aspects of the developmental biology of the growing plants of B. emersonii have been reviewed (Cantino and Lovett, 1964; Cantino, 1966) ; in the present essay, we will concern ourselves only with encystment. First (Sections II-VII) , we examine structural changes and other coordinated events, which, togethcr, comprise encystment. Special attention is devoted to an interpretation of the interrelationships involved. Second (Section V I I I ) , we examine the influences of some exogenous factors on encystment, and discuss the insight they may provide about its mechanism. Attempts have been made to keep repetitious literature citations to a minimum. As a general rule, the reader should assume that undocumented statements about B. enaersonii are based either on our own published [reviewed in Cantino et al. (1968) or cited in Cantino and Truesdell (1970) ] or unpublishcd observations. We have tried, of course, to include pertinent references to the works of others in the appropriate places. II. Structure of Nongerminating Zoospores The first electron micrographs of thin sections through zoospores of B. emersonii were made jointly in 1963 by Lovett a t Purdue University, and by Lythgoe, Leak, and Cantino a t Michigan State University; this was followed quickly with additional descriptions of their fine structure by Reichle and Fuller (1967), Lcssie and Lovett (1968), Sol1 e t al. (1969), and us (references in Cantino and Truesdell, 1970). We will not undertake a detailed discussion of all these architectural details here. However, since structural changes play a very important role in spore germination, selected aspects of the internal make-up of zoospores must be summarized briefly a t the outset to provide essential background (see Fig. 1 ) . A zoospore is about 7 x 9 p in size, and propels itself with planar waves of lateral displacement (Miles and Holwill, 1969) along a single, posterior, whiplash flagellum. The cell does not possess a wall; rather, it is delimited by a single, continuous, unit membrane. Our measurements indicate that it is ca. 90-100 A thick, a value typical of many other plasma membranes (Fawcett, 1966). The lack of a cell wall and plasticity of its plasmalemma permit the zoospore to take on a continuum of quickly changing, indefinable shapes and amoeboid characteristics.

4

LOUIS C. TRUESDELL AND EDWARD C. CANTINO

FIQ.1. Diagrammatic composite view of a longitudinal section through a spore of Blastocladiella emersonii. For purposes of clarity, the relative proportions of some structures are exaggerated, while a few ingredients (e.g., the second, short centriole, satellite ribosome packages) are not included. Overall dimensions of spore are approximately 7 x 9 p. Abbreviations: V, vacuole; M, single mitochondrion; BM, backing membrane; SB, side body; L, lipid body; K, kinetosome; P, prop (following terminology of Olson and Fuller, 1968) ; F, flagellum; R, rootlet; G, gamma particle ; NC, nuclear cap ; N, nucleus.

1. T h e Nucleus and Nuclear Cap The nuclear cap (NC, Fig. 1 ) partially encircles the nucleus and is obviously the most massive single structure in the spore. It is an aggregate of basophilic particles (Turian, 1962) identified as ribosomes (Lovett, 1963), and is entirely delimited by a system of double membranes (i.e., two parallel unit membranes), parts of which are continuous with the nuclear membrane and other structures. 2. T h e Kinetosome, Flagellum, and Banded Rootlet Two centrioles of different lengths (Reichle and Fuller, 1967; Lessie and Lovett, 1968; Sol1 et al., 1969) are located a t the posterior end of the nucleus; the longer one, the kinetosome (K, Fig. l ) , is continuous

1.

INDUCTION OF GERMINATION IN

B. emersonii

5

with the flagellar axoneme. Perpendicular to the long axis of the kinetosome, and in intimate contact with it, is a banded rootlet (R, Fig. 1 ) . Although it had been thought (Reichle and Fuller, 1967) that three such rootlets were present, it now seems likely (Cantino and Truesdell, 1970) that there is only one, it being bent a t the point along its length where i t makes contact with the kinetosome, with its two “arms” extending into open-ended channels in the mitochondrion. 3. The Single Mitochondrion

The single mitochondrion (M, Fig. 1) is situated asymmetrically around part of the nuclear apparatus (Nucleus plus nuclear cap) where it also surrounds the kinetosome. Sometimes, especially in amoeboid spores, the portion of the mitochondrion nearest the spore’s anterior is flattened, almost devoid of cristae, and exceptionally rich in particles similar in size and staining properties to nuclear cap ribosomes; many of them are aligned along the inner mitochondrial membrane. Usually, such particles are also found in the mitochondrion where i t contacts lipid bodies (L, Fig. 1 ) .

4. Lipid Bodies, SB Matrix, and Backing Membrane The lipid bodies are dispersed along the outer surface of the long “arm” of the mitochondrion and are usually in intimate contact with it. Molded against them is an organclle bound by a unit membrane, the SB matrix (SB, Fig, 1 ) . Although originally viewed as a collection of individual SB bodies, serial sections suggest (Cantino and Truesdell, 1970) that the SB bodies are part of a continuous structure. The SB matrix is confined to a region around the lipid bodies, does not obstruct the openings to the two mitochondrial rootlet channels, and consists of a granular to amorphous substance of moderate electron density; its composition is unknown. A sheet of double membrane, the backing membrane (BM, Fig. l ) , covers the SB-lipid-mitochondria1 complex and is attached a t several places to the outer unit membrane of the nuclear apparatus [two such points are shown in Fig. 1; for a three-dimensional view, see Cantino and Truesdell (1970) 1. The inner portion of the backing membrane stains intensely with OsO,, UO,2+, or Pb2+ in those areas adjacent to the SB matrix; in other regions, it does not. Portions of the backing membrane also enter into and extend along the surfaces of the mitochondrial rootlet channels. 5. Gamma Particles

These cytoplasmic organelles (G, Fig. 1 ) undoubtedly correspond to the “gamma” particles first described (Cantino and Horenstein, 1956)

6

LOUIS C. TRUESDELL AND EDWARD C. CANTINO

and recently reinvestigated (Matsumae et al., 1970) by way of light microscopy. The gamma particle consists of two major components: a unit membrane (the gamma surrounding membrane, or GS-membrane) which encloses an ellipsoid, bowl-shaped matrix (gamma matrix) about 0.5 p in length. The matrix is tightly packed with amorphous and membranous osmiophilic material (Truesdell and Cantino, 1970). Cantino and Mack (1969) provide a detailed, three-dimensional description of the gamma particle. All available evidence (Myers and Cantino, 1971) suggests that this organelle contains DNA. 6. Other Cytoplasmic Inclusions

The cytoplasmic ground substance is homogeneous and contains numerous, rather evenly dispersed, polysaccharide particles of the sort pictured by Lessie and Lovett (1968), and a few vesicles. Aggregations of particles (“satellite ribosome packages”; Cantino, 1969 ; Cantino and Mack, 1969; Shaw and Cantino, 1969) identical in size and staining properties to nuclear cap ribosomes, and surrounded by a double membrane, appear with high frequency in amoeboid spores. 111. Behavior of the Spore during Encystment The end of a zoospore’s existence commences with the beginning of e n c y s t m e n t a rapid avalanche of a highly coordinated sequence of events, as follows.

A. FLAGELLAR RETRACTION AND ROTATION OF

THE

NUCLEARAPPARATUS

These two intimately linked phenomena, previously described in detail (Cantino et al., 1963, 1968), are summarized below and supplemented with observations not reported earlier. When the time for encystment draws near, the spore (Fig. 3A) gradually becomes spherical. During this time, the flagellum straightens out, vibrates rapidly, and stops, all usually within a few seconds. The nuclear apparatus shifts slightly toward the spore’s anterior, then begins to rotate in the direction of the short arm of the mitochondrion. As rotation proceeds, the extended flagellum sweeps into an arc. The nuclear apparatus continues turning until the entire axoneme has been retracted through a fixed locus, the original point of entry on the spore’s surface. Thus, the body of the spore does not rotate simultaneously; it simply becomes progressively more spherical until, eventually, i t resembles an encysted cell (Fig. 3 D ) . It may take only a few seconds or longer than a minute (depending on conditions) for the flagellum to disappear, and an additional 90 seconds or so for the cell to attain its final cystlike morphology. By the end of this short interval, we find, as do Lovett (1968) and

1.

INDUCTION OF GERMINATION IN €3.

emersonii

7

Soll et al. (1969), that an initial cyst wall is already detectable in electron micrographs (Fig. 4A, arrow). Some 10 minutes later, a small germ tube emerges from which the rhizoidal system eventually develops.

B . VACUOLE FORMATION I n actively amoeboid spores, vacuoles may be present along the nuclear cap or a t the posterior end of the cell. When the cell begins to lose its amoeboid features prior to flagellar retraction, vacuolelike structures become more numerous. As encystment progresses, the vacuoles move to the spore surface and appear to fuse with it. The fact that gamma particles are frequently observed in wiwo in close association with vacuoles led to early speculations that such vacuoles arose from gamma particles. This idea will be developed further in Sections V and VI,B. During encystment, cells take on adhesive properties (Cantino et al., 1968; Soll et al., 1969), i.e., spores begin to adhere to one another or to their containers. We find that this occurs a t about the time vacuoles migrate to the spore surface. If spore suspensions are agitated during this period, cumulative collisions among encysting spores give rise to increasingly large clumps that may contain up to 100 or more cells. After encystment, such clumps are not broken up by very strong agitation, even in the presence of high concentrations of urea, sodium chloride, or mercaptoethanol.

C. VOLUMECHANGES DURING GERMINATION When observing encystment through the light microscope, it was always our impression that spores decreased in volume (compare Fig. 3A ws. 3D). Yet, this might have been illusory because changes in spore shape were occurring simultaneously. Therefore, we present here some quantitative data which show that the presumed volume changes during germination are, in fact, real (Fig, 2 ) . With the population density and solutions used (legend, Fig. 2 ) , most spores encyst quickly. The figure displays the relative size distribution in a spore population after 5 and 15 minutes of incubation. At 5 minutes, no spores had encysted; a t 15 minutes, about 86% had encysted. The size difference is obvious. At 5 minutes, the mode for the population is 14.5 volume units; a t 15 minutes, it is 8.2 volume units. Thus, spore volume decreases ca. 43% during encystment. A small bimodal component in the 15-minute curve is positioned beneath the &minute peak. One of these minor modes, located a t 14.2 volume units, represents spores which did not encyst. The other minor mode is there because a few encysted spores adhered to one another;

8

LOUIS C. TRUESDELL AND EDWARD C. CANTINO

5.0 MINUTES

VOLUME (ARBITRARY UNITS)

Wo. 2. Change in cell volume during encystment, as measured with a Coulter Particle Size Distribution Plotter. Distribution of spore volumes at 5 and 15 minutes after inducing encystment by diluting a freshly harvested, spore population with a NaC1-KC1 solution (2 mM, final conc.) to 5.7 x lo4 spores/ml.

note that it is positioned at about twice the relative volume of encysted spores. The slight displacement of these two minor modes toward one another, and away from their theoretical values, results from the summing effect of the overlapping distributions.

D. RE SUM^

OF

MAJORSTRUCTURAL CHANGDS IN ENCYSTMENT

We regard the structural changes described in Section I I I A to I I I C as major transformations in zoospore encystment, i.e., retraction of the flagellum, formation of vacuoles, deposition of cyst wall material, change in cell shape, and decrease in cell volume. Certainly, these processes must be rooted in cellular chemistry and physics; but, to achieve some understanding a t these levels, the relevant structural interrelationships among them must first be comprehended. In the following section, we examine these events in more detail. IV. Mechanics of Flagellar Retraction and Rotation of the Nuclear Apparatus

After observing encystment in vivo in hundreds of spores of B. emersonii, and examining their fine structure in detail, we conclude that retraction of the flagellum and concurrent rotation of the nuclear ap-

1.

INDUCTION OF GERMINATION IN

B. emersonii

9

paratus can be rationalized as a purely mechanical process. Our explanation follows. The portion of the flagellum that extends outward from the rest of the spore body is composed of two major parts: the axoneme and the membranous axonemal sheath. During retraction and rotation, the axoneme coils up along the inside periphery of the encysting spore. This can be seen with phase optics under optimal conditions, but the most conclusive evidence has been seen in electron micrographs of newly encysted spores (Lovett, 1967b; Reichle and Fuller, 1967; Sol1 et al., 1969). After the axoneme has been withdrawn, the membranous sheath is no longer associated with it (Fig. 4A, arrow). Since the sheath is continuous-and probably identical-with the plasma membrane, it seems highly likely that the membranous sheath is retained as part of the plasma membrane after flagellar retraction. This conclusion is, we think, indirectly strengthened by the fact that the axonemal sheaths in some other water molds also seem to have a similar fate, judging from the observations of Meir and Webster (1954) who noted that hairs on the flagella of primary zoospores of some Saprolegniaceae seem to appear on the cysts derived from them, and the conclusion of Fuller and Reichle (1965) that laterally projecting “flimmer filaments” (mastigonemes) attached to the axonemal sheath of Rhizidiomyces apophysatus are subsequently observed on part of the cyst surface after flagellar retraction. It must also be noted that in B. emersonii the axoneme is in no way partitioned off from the inside of the main body of the zoospore, for the axoneme is a continuous cordage extending from one location inside the spore to another location inside the spore. I n fact, in the light of the argument we are trying to develop, it would be more logical to speak of axonemal translocation than flagellar retraction. I n B. emersonii, the axoneme is being translocated only when the nuclear apparatus is rotating-and vice versa. This indicates that the two structures probably remain connected by some means throughout the process (supporting evidence for this comes from electron microscopy: Section V ) . In such a linked device, the force responsible for the simultaneous translocation and rotation could theoretically be applied a t either “end” of the system, e.g., a force that causes the nuclear apparatus to rotate would, in effect, wind in the axoneme or, conversely, a force applied to the axoneme would, in turn, push the nuclear apparatus in its circular path. Our observations, partially detailed elsewhere (Cantino et al., 1968) and summarized below, support the latter interpretation. During encystment, the spore is assuming a spherical shape, i.e.,

10

LOUIS C. TRUESDELL AND EDWARD C. CANTINO

its surface area is being minimized with respect to its volume, and its volume is decreasing as well. Thus, forces are operating that oppose any extension of the cell surface, i.e., the membrane extending around the protruding axoneme. These forces, therefore, may cause the membrane around the flagellum to assume a new position confluent with the increasingly spherical contour of the cell. Axonemal translocation and rotation of the nuclear apparatus will follow. However, there is evidence for the presence of binding sites between the nuclear apparatus and the plasma membrane which must first be broken before rotation can occur. The plasma membrane is closely associated with the electron scattering portion of the backing membrane which, in turn, is continuous with the outer unit membrane of the nuclear cap; it is linked, as well, by a number of close associations to other parts of the nuclear apparatus (Cantino and Truesdell, 1970). Furthermore, an amorphous substance (P, Fig. 1) connecting the kinetosome with plasma membrane a t the base of the flagellum is visible in B. emersonii [as well as in B. britannica and other water fungi; Cantino and Truesdell (1971) and Olson and Fuller (1968) respectivelyl. If all these binding sites are broken, the nuclear apparatus-axoneme assemblage would presumably “float” without restraint and be free to rotate. If the causal force for rotation of the nuclear apparatus is transmitted via the axoneme, rotation should not be detected in spores which have had their flagella removed. Experiments (Cantino et al., 1968) with mechanically deflagellated spores have shown that this is indeed the case. Throughout the entire process of their encystment, there is little detectable movement of the nuclear apparatus except a slight rocking motion, which we interpret to be due to a breakdown of binding sites holding the nuclear apparatus to the plasma membrane. The changing shape of the spore during encystment might also cause slight movement. Deflagellated spores, incidentally, display the same viability (Cantino et al., 1968) and encystment kinetics (Sol1 and Sonneborn, 1969) as flagellated spores. Flagellar retraction is simply an integral part of the overall structural-mechanical process of encystment, and requires no special mechanisms, or forces, not already provided by the other associated events.

V. Fine-Structural Changes in Encystment The investigation of encysting spores by electron microscopy has done much to clarify and integrate the major structural events of encystment, especially formation of vacuoles and cyst walls. But before discussing these phenomena, i t may be useful to relate some findings that

1.

INDUCTION

OF GERMINATION IN

B. emersonii

11

haunted our early germination studies, namely, the genesis of fixationinduced myelin-like configurations.

A. FORMATION OF MYELINLIKE FIGURES Within the first few minutes after induction of germination, myelinlike figures were frequently found in the nuclear cap along the double membrane separating it from the nucleus. They were sometimes continuous with the inner membrane of the nuclear cap. As the time of flagellar retraction approached, however, myelinlike figures began to appear outside of the cap as well; most of these were now near the plasmalemma, frequently contained in vacuoles, and apparently about to break through the spore surface. This sequence of events led us to suppose that the myelinlike configurations represented the migrating, vacuolelike structures (Section II1,B) observed by phase microscopy in germinating spores. However, this notion was soon confused by the fact that no such configurations were ever observed leaving the nuclear cap for the cytoplasm. Finally, we learned that by extending the wash period normally used between glutaraldehyde and osmium tetroxide fixations to 24 hours, the incidence of myelin-like figures was greatly reduced. When new combinations of fixatives were used (Truesdell and Cantino, 1970), the figures were always absent. Consequently, we have now consigned these myelinlike bodies to the rank of artifact. Nevertheless, this dubious distinction should not be allowed to cloud their possible significance as indicators of real changes in spores prior to and during encystment; their formation could well result from phospholipid released during cytomembrane alterations. [For a discussion of the origin and significance of myelinlike bodies in other organisms, see Anderson and Roels (1967) and references therein.]

B. STRUCTURAL CHANGESPRIOR TO FLAGELLAR RETRACTION The first changes detected after induction of germination occur in the backing membrane (see Section 11,4, and Fig. 1 ) . It consists of two portions; one is undifferentiated (i.e., it looks like a typical double membrane) and the other is differentiated (the region between the two unit membranes is filled with an osmiophilic substance), The undifferentiated portion fragments rapidly until only vesicles remain (Fig. 3F). The differentiated portion does not break down as quickly, but it, too, is slowly consumed via vesicle formation. These vesicles are much too small, however, to be the ones observed (see Section II1,B) by phase microscopy.

12

LOUIS C. TRUESDELL AND EDWARD C. CANTINO

1.

INDUCTION O F GERMINATION IN

B. emerS0nii

13

The backing membrane is, as previously emphasized, one of the focal points envisioned as a possible "binding site" (Section IV) between the nuclear apparatus and plasma membrane. Thus, i t would have to be broken for rotation of the nuclcar apparatus to occur-and this is, in fact, observed. Neither the differentiated nor the undifferentiated portion of the backing membrane has ever been found after normal flagellar retraction. [This does not apply to flagellar absorption during cold-induced swelling (Fig. 3F) .] As the backing membrane is breaking down, pronounced changes occur also in the gamma particles (described in detail by Truesdell and Cantino, 1970). In brief, the GS-membrane (in contrast to its nearly spherical appearance in nongcrminating spores) becomes amorphic (Fig. 3E, F). At times, it extends so irregularly through the cytoplasm that it can be fully traced only via serial sections. Concurrently, numerous vesicles begin to appear in the cytoplasm around the gamma particles. The GS-membrane is continuous with some of these vesicles, as though they were being budded off from the GS-membrane. Others of these vesicles lie free in the cytoplasm, and frequently seem to be fusing with the plasma membrane. These events temporally parallel the in vivo observations on vacuole formation (Section II1,B) made by phase microscopy. ~

FIG.3. Selected views of the spore of Blastocladiella emersonii. (A) Phase contrast view of amoeboid spore. Solid arrow points to nuclear cap, which lies immediately above nucleus (light region containing approximately centered, darker nucleolus) ; broken arrow points to mitochondrion-SB-lipid complex (i.e., "side body"). (B) Phase contrast view of a spore swollen by incubation a t O-l"C, showing nuclear cap, nucleus with nucleolus, swollen mitochondrion (arrow) with refractile lipid bodies (white dots) on surface and numerous gamma particles (small dark dots i n peripheral cytoplasm). ( C ) Same spore seen in (B), but after flagellum was absorbed while in the cold; note shift in position of nuclear apparatus toward center of cell. (D) Phase contrast view of an encysting spore decreasing in volume just after flagellar retraction ; note typical vacuoles. Compare the appearance of this spore, in which the flagellum was retracted as an integral part of the encystment process, with the spore in ( C ) , in which flagellar retraction was not (see Section VII1,A) a part of encystment. (E) Gamma particles showing GS-membrane, gamma matrix (black areas), and GM-vesicles (arrow) ; fixed with osmic acid-uranyl acetate mixture (Fixation 11; Truesdell and Cantino, 1970). (F) Spore encysting after cold treatment; fixed 3.5 minutes after transfer from Oo-l"C to 22°C. Section shows typical features for this stage in development, except that flagellum was absorbed (axoneme; solid arrow) as a result of cold-induced swelling (see Section VII1,A). Undifferentiated portion of backing membrane has vesiculated, while the differentiated portion (broken arrow) remains. Note gamma particles with amorphic conformation of GS-membranes. Abbreviations : K, kinetosome with associated banded rootlet; L, lipid; SB, SB matrix. Fixed with glutaraldehyde and osmic acid (Fixation I; Truesdell and Cantino, 1970).

14

LOUIS C. TRUESDELL A N D EDWARD C. CANTINO

Along the gamma matrix, small vesicles (GM-vesicles; Fig. 3E, arrow) are released. Some of these fuse with the GS-membrane. Others become entrapped in the vesicles budded off by the GS-membrane; this results in formation of multivesicular bodies (MVB), i.e., single large vesicles containing smaller vesicles (Fig. 4A). Although the foregoing events are initiated prior to flagellar retraction, they occur more frequently during advanced stages of germination.

C. STRUCTURAL CHANGES DURING FLAGELLAR RETRACTION Only very infrequently will a micrograph be obtained that shows a zoospore actually in the process of retracting its flagellum; thus, most of our conclusions about the nature of this process have had to be deduced by extrapolation from studies of spores that had just recently retracted. The following three conclusions, however, are derived from sections through spores that were caught in the act. First, the nuclear cap, nucleus, and axoneme remain connected and are arranged along an arc that follows the contour of the spore surface. This is just the arrangement anticipated from the model of flagellar retraction we have presented (Section IV) . Second the plasma membrane of a retracting spore (as well as that of spores observed after retraction) is irregularly folded. This could reflect the actual in vivo state, or be an artifact resulting from hypertonic fixation. The latter possibility is less likely, however, because nongerminating spores fixed a t the same osmolarity possess relatively smooth surface membranes. Possibly, the folding is a manifestation of surface forces that come into play during retraction. The folding also suggests that the surface area may, a t least in part, be “minimized” (see Section IV) by a folding process (as contrasted with an actual reduction by uniform contraction). Third, during rotation of the nuclear apparatus, the banded rootlet remains in place in its mitochondria1 channel and connected to the kinetosome; therefore, to some degree a t least, the mitochondrion must also rotate. The extent to which the SB matrix and lipid bodies move with the mitochondrion during rotation is uncertain, but it is our impression that their movement is much more confined than that of the mitochondrion. The mitochondrion probably slides by them and, in the process, causes some mixing of the Gther two components as i t tends to carry them along as well.

D. STRUCTURAL CHANGES AFTER FLAGELLAR RETRACTION Most of the conspicuous structural changes associated with encystment occur after flagellar retraction, when various morphological units

1.

INDUCTION OF GERMINATION IN

B. emersonii

15

in the cell continue to break down and become more evenly distributed throughout it. By the time the germ tube is formed, the flagellar axoneme has usually disappeared, and profiles of lipid bodies, SB matrix, and the mitochondrion are found dispersed in the cytoplasm. The lipid bodies and profiles of SB matrix usually lie close to the sides of mitochondria1 profiles adjacent to the cell membrane. The SB matrix can remain intact a t least until midway between flagellar retraction and germ tube formation, but it probably breaks up into smaller units shortly thereafter. Various aspects of the foregoing structural changes occurring in the interim between flagellar retraction and germ tube formation have been examined by Lovett (1968), Soll e t al. (1969), and the authors. Immediately after retraction, traces of a densely staining amorphous substance appear a t the spore surface (Fig. 4A; see also, Soll e t al., 1969, Figs. 8 and 9) ; this is the first sign of the initial cyst wall, which continues to increase in thickness during germination. Many of the gamma particles are now oriented so that their major openings face the nuclear cap, and portions of their GS-membranes protrude into indentations in the nuclear cap double membrane (Fig. 4A) ; they commonly adhere in this manner for 1-2 minutes after retraction. Similarly, but less frequently (perhaps as a function of available surface area?), Gamma particles also tend to adhere to the nucleus and the retracted axoneme. Soll e t al. (1969) have commented (p. 201) that “. . . such “gamma particles’’ are impressive by their scarcity in sections of this and succeeding stages.” We call attention, therefore, to representative sections through encysted spores in which well defined gamma particles unquestionably exist (see Fig. 4A and B ; these are equivalent to the early Round Cell I and I1 stages, respectively, of Soll e t al., 1969). I n our experience, this situation is not a t all exceptional but, in fact, the rule. The MVB and numerous smaller vesicles are located in the vicinity of the gamma particles and near the cell wall. Presumably, both vesicle types have originated from the gamma particles; in any case, both are commonly seen (Soll et al., 1969; Truesdell and Cantino, 1970) apparently in the process of fusing with the spore surface. Shortly after flagellar retraction, the mitochondrion surrounds a greater area of the nuclear cap surface than it did prior to retraction. Mitochondria1 channels, like those housing the banded rootlet (see Fig. 3F), are no longer observed a t this stage. Conceivably, the mitochondrion partially subdivides itself along these channels to yield a many-armed yet single structure that spreads out over the nuclear cap. Such a process could be the source of the obviously bpanched (Lovett, 1968; Soll e t al., 1969) mitochondrion that appears as the time for germ tube formation is approaching; see belaw.

16

LOUIS C. TRUESDELL AND EDWARD C. CANTINO

1.

INDUCTION OF GERMINATION IN

B. emersonii

17

At this stage, too, the exposed double membrane around the nuclear cap begins to fragment. This continues until the membrane no longer confines the ribosomes of the cap, and they scatter throughout the cytoplasm (Lovett, 1968; Soll et al., 1969); Soll et al. (1969) have also noted the alignment of ribosomes along some of these membrane fragments. While the nuclear cap membrane is fragmenting, the gamma particles move away from their positions along the cap. Their GS-membranes expand extensively, and confine more and more of the GM-vesicles that have been released from the gamma matrix (Fig. 4B).As these vesicles are discharged, the gamma matrix becomes progressively thinner. After the ribosomes have dispersed, the mitochondrion begins to take on a more tubular aspect, and it becomes more extensively distributed throughout new regions in the cytoplasm. Finally, the GS-membranes of two or more gamma particles begin to coalesce with one another. The number of such multiple fusions is obviously limited by the number of gamma particles available, which averages ca. 12/spore but may in rare instances reach ca. 30/spore (Cantino and Horenstein, 1956; Matsumae et al., 1970). As a consequence, very large vesicles are produced that contain two or more of the decaying gamma matricies ; the latter eventually break up completely into smaller vesicular elements, some of which fuse with the membranes of the parent vesicle. This large expanded membrane, in turn, buds off vesicles that can migrate to the surface of the spore and fuse with it; they are detectable up to a t least 2 hours after encystment. VI. Punctuation on Three Important Structural Changes in Germination

The breakdown of certain, specific, cytoplasmic membranes is a crucial step in the germination of B. einersonii zoospores. Vesiculation FIG. 4. The spore of Blastocladiella emersonii after flagellar retraction. ( A ) . Section through spore fixed about 1 minute after retraction; note pair of axoneme cross sections (arrow), initial deposits of wall material along plasma membrane, one of the gamma particles seated at indentation along nuclear cap, and adjacent multivesicular bodies. Fixed with glutaraldehyde and osmic acid (Fixation I11 ; Truesdell and Cantino, 1970). (B). Section through spore fixed about midway between flagellar retraction and germ tube formation. Nuclear cap membrane has fragmented, and ribosomes have begun to scatter. Mitochondria1 profiles are still located along periphery of ribosomal mass. Note, particularly, the long, thin profile on left containing ribosome-like particles; such mitochondria1 sections are commonplace at this and previous stages. The gamma surrounding membrane (arrow) around decaying gamma matrix (black region) is greatly expanded and encloses numerous vesicles. The deposits of wall material are heavier than in (A) above. Fixed as in Fig. 3 (E).

18

LOUIS C. TRUESDELL AND EDWARD C. CANTINO

of the backing membrane seems to destroy a binding site (see Section IV) that holds the nuclear apparatus in place, thereby allowing it to rotate. Its disruption is also necessary if lipid bodies, SB-matrix, and the mitochondrion are to be released and distributed to the cytoplasm. Similarly, disruption of nuclear cap membranes is a prerequisite for the scattering of its ribosomes. Although there is little information available about the causes of these membrane alterations, their importance leads us to some speculations.

A. CHANGES IN THE BACKING MEMBRANE The undifferentiated portion (as contrasted with the densely staining section) of this membrane is one of the most labile structures in the spore. It is the first membrane to be disrupted any time fixation is below par. It is also the first to change in vivo during germination. There is even some indication that the undifferentiated portion of the backing membrane may become fractured in nongerminating spores if but a few hours have passed since sporulation. When this happens, the SB matrix, lipid bodies, and mitochondrion still maintain their relative positions to one another, indicating that there are attractive forces among them. Frequently, the lipid bodies and mitochondrion are so tightly molded to one another that the line of demarcation is not even visible. The densely staining portion of the backing membrane is much more stable than the undifferentiated section. I n this connection, we have made an interesting observation on the electron dense region between the unit membranes of this double membrane. If thin sections of zoospores are floated on 10 mM NaOH for 2 minutes and subsequently stained with P b citrate, the dense interior shows no staining whatsoever. Only one other structure in the spore of B. emersonii displays this characteristic behavior, namely, the gamma matrix. It could well be, therefore, that the densely staining amorphous material of the gamma matrix and the densely staining interior of the backing membrane are composed of the same substance. Indeed, this may help explain why decay of the gamma matrix and the electron dense portion of the backing membrane is initiated simultaneously. B. BREAKDOWN OF NUCLEAR CAP MEMBRANE

It can be logically argued that the disruption of the nuclear cap membranes and the scattering of its ribosomes may be necessary for efficient synthesis and distribution of proteins in the germinating spore. Gamma particles adhere to the nuclear cap membranes only minutes before these two events occur. The firmness with which they cling is dramatized by the distortions they seemingly produce in the usually

1.

INDUCTION OF GERMINATION I N

B. emersonii

19

smooth contour of the cap membranes. Thus, i t would be tempting to link such events to the breakdown of these membranes. However, two other facts complicate this interpretation. First, the gamma particles also seem to stick t o the nucleus, which does not fragment, and t o the retracted axoneme. Second, the backing membrane (an extension of the nuclear cap outer membrane) begins to rupture before this alignment stage occurs, and without any prolonged contact with the gamma particles. It may be more realistic, therefore, to view the adhesion of gamma particles to the membranes of the nuclear cap as a result, not a cause, of subcellular changes. Many modulations are taking place in the cell at this time: alterations in cell shape and volume, onset of cell wall synthesis, etc. These things may be associated with ionic shifts, which could cause new arrangements in attractive forces and give gamma particles their adhesive properties, albeit properties which may or may not serve specific functions. We suspect, in fact, that the mitochondrion is more involved in membrane breakdown than are gamma particles. It will be recalled (Section V,D) that it spreads over the nuclear cap immediately after flagellar retraction, and remains there until membranes have fragmented. Then, i t changes in form and extends to other places in the cell. Specific ionic uptake and release by mitochondria is so well known as to need no documentation. If specific ionic changes are required for membrane alterations (Kavanau, 1965), then the mitochondrion is a likely candidate to provide them. The flattening of the mitochondrion may generate increased surface area to maximize a membrane mediated flow of substances. When the double membrane around the nuclear cap fragments, its inner and outer unit membranes fuse to form vesicles derived from both membranes. Such a fusion of different unit membranes may be necessary for fragmentation. This could explain why there is no obvious morphological change in the outer membrane of the nucleus, even though this membrane is continuous (see Fig. 1 ) with the outer membrane of the nuclear cap. Specifically, the inner nuclear membrane is the only membrane that makes direct and extensive contact with the nucleoplasm, which may exert a controlling influence upon it. If, as a result, the inner nuclear membrane cannot fuse with the outer nuclear membrane, then vesiculation of this particular part of the double membrane around the nuclear apparatus cannot take place.

C. ROLEOF

THE

GAMMAPARTICLE IN CELLWALL FORMATION

Although the earliest, visible (via electron microscopy) evidence for the presence of a cyst wall is not found until the axoneme has been

20

LOUIS C. TRUESDELL AND EDWARD C. CANTINO

translocated, it is likely that wall formation is initiated prior to retraction. We have said that spore surfaces acquire adhesive properties leading to mutual attractiveness shortly before retraction; there is reason to believe this is a direct result of newly deposited wall material, a t least in those cells that still adhere to one another after flagella have disappeared. Electron micrographs show that the wall material around the cysts merges into an undifferentiated continuum in areas where cysts adhere to one another (this may explain why all attempts short of vigorous hydrolysis have failed to break up clumps of encysted cells). Also, zoospores acquire adhesive properties a t approximately the time that vacuoles are fusing with the plasma membrane. These vacuoles originate from gamma particles (Section V,B), as do, probably (Section V,D), the MVB observed (Sol1 et al., 1969) at the spore surface. These events establish a link between gamma particle decay and cell wall formation. Other correlations include the facts (Truesdell and Cantino, 1970) that (a) gamma particles always begin t o break down shortly before the cell wall is detected; (b) they always (and only) continue to break down as the cell wall is being deposited; and (c) electron dense material resembling the gamma matrix accumulates a t the cell surface while simultaneously the electron dense substance of the gamma matrix is disappearing. Finally, our argument finds support in the results of some enzymatic assays. Camargo et al. (1967) studied chitin synthetase in B. emersonii and established that, in spore homogenates, the activity of this enzyme was highest in fractions containing mainly smooth membranes. Their electron micrograph of this fraction also showed that it probably contained gamma particles. Work with cell-free preparations of gamma particles in our laboratory has demonstrated (Myers and Cantino, 1971) that smooth membranes can originate from decaying gamma particles during their isolation. Thus, the available evidence is a t least consistent with the idea that gamma particles may contain the “cell wall” enzyme chitin synthetase. The growing plants of B . emersonii produce chitin (Cantino et al., 1957), but they do not contain gamma particles (Lessie and Lovett, 1968). The swimming zoospores apparently do not produce chitin, but they do contain gamma particles. Perhaps the extraordinarily compact nature of the gamma matrix provides the mechanism whereby chitin synthetase activity is suppressed in the motile stage of this organism’s life cycle. Although it does not pertain directly either to encystment or uniflagellate fungi, a recent paper from Bracker’s laboratory (Grove et al., 1970) on the ultrastructural basis for hyphal tip growth in Pythium is especially relevant and must be cited here. I n discussing the literature

1.

INDUCTION OF GERMINATION IN

B. emersonii

21

and their own work on the deposition of hyphal wall material, the authors hypothesize a sequence of events that includes: ( a ) secretion of vesicles by dictyosomes, with eventual loss of the entire dictyosome cisterna; (b) migration of the vesicles to the hyphal tip, with increase in size and/or fusions among some of them to yield larger vesicles; and (c) fusion of the vesicles with the plasma membrane, and concomitant deposition of vesicle contents in the wall region. There are, of course, obvious differences in origin and structure between the dictyosonies of Pythium and the gamma particles of Blastocladiella; yet, their apparent functional activities have much in common. Indeed, this is an opportune time to reinforce our stand on behalf of gamma particles with an earlier commentary by Bracker (1967, p. 349) : “The apparent absence of dictysomes in so many fungi raises the question of what cell component, if any, carries out the expected functions of dictyosomes in cells where none are present . . . . It seems logical for this role to fall upon a membrane-bounded structure capable of packaging materials within a membrane for transport.” The gamma particle in the spore of B. emersonii, it seems to us, is just such a structure. The results of one other study can also be cited here t o advantage. Manton (1964) believes that vesicles beneath the plasmalemma in zoospores of the freshwater alga Stigeoclonium provide the first components of the cell wall. After flagellar retraction, the spore secretes a flocculent, adhesive material probably derived from the contents of superficial vesicles. Furthermore, as the cell becomes obviously walled, the numerous small vesicles previously around the Golgi bodies in the swimming spore are replaced by large swellings a t the edges of the Golgi cisternae. I n the uniflagellate spores of the fungus Olpidium, too, vesicles could conceivably be involved in cyst wall formation (see Temmink and Campbell, 1969a,b; Lesemann and Fuchs, 1970). We concur wholeheartedly with Bartnicki-Garcia’s observation (1968) : “Solutions to some of the most important problems of fungal morphogenesis probably depend on . . . answers to the following questions: where are cell wall structural polymers synthesized? Are they polymerized in some intracellular site . . . and somehow transported in an orderly way to . . . the cell wall?” VII. Macromolecular Synthesis during Germination

Only a few macromolecular components in the spores of B. emersonii have been studied. These include some detailed investigations (Camargo et al., 1969) of the regulation of glycogen synthetase. It was concluded that glycogen synthesis is regulated by intracellular concentrations of glucose 6-phosphate, which stimulates synthetase activity 90-fold in

22

LOUIS C. TRUESDELL AND EDWARD C. CANTINO

zoospores but only 4-fold in 3-hour plantlets. Unfortunately, these interesting comparative data cannot yet be integrated into our discussion of early germination stages. Extensive studies on regulation of nucleic acid and protein synthesis have come from Lovett’s laboratory a t Purdue University. Not until spores have germinated and produced germ tubes do measurable amounts of RNA begin to accumulate (Lovett, 1968; in the system used, the RNA synthesis begins ca. 40-45 minutes after encystment). This is followed 30-40 minutes later by synthesis of protein and, about 40 minutes thereafter, DNA. Thus, during the early stages of germination being emphasized in this review, there is no measurable net increase in either RNA or protein, but pulse-labeling experiments with uraci1-14C and leucine-14C showed that synthesis of RNA and protein does begin, a t very low rates, about the time of encystment. Actinomycin D, a t 25 pg/ml, inhibits detectable uracil incorporation in germinating spores (Lovett, 1968). Nonetheless, spores in contact with the antibiotic encyst and develop up to the time when the directly measurable increase in RNA should begin; then they stop growing. They are noticeably different from those formed in the absence of the drug in that ( a ) , the nucleolus fails to show the increase in size that normally accompanies germination, and (b) the shape of the primary rhizoid changes somewhat. Since early protein synthesis is not affected significantly by inhibition of RNA synthesis (assuming lack of uracil incorporation does in fact denote this), it would seem that the ribosomal-, transfer-, and messenger-RNA’s necessary for it are all present in ungerminated spores. Cycloheximide, at 20 pg/ml, inhibits protein synthesis in germinating zoospores. Spores encyst normally, nuclear cap membranes fragment, and ribosomes scatter throughout the cytoplasm as usual, but the germ tube does not form and the retracted flagellar axoneme does not disappear. Although germination does not proceed as far as it does in the presence of actinomycin D, the structural changes of encystment apparently require only the protein and RNA found in a nongerminating zoospore. One obtains the impression that the spore has all the necessary ingredients for encystment packaged and waiting, and that some signal is needed for it to start using them. I n summary, Lovebt’s experiments (and, more recently, those of Soll in Sonneborn’s laboratory a t Wisconsin; Soll, 1970) suggest: first, the spores of B. emersonii can consummate the structural changes associated with encystment (Section III,D) without apparent protein synthesis ; and second, neither the foregoing events nor subsequent ones essential for germ tube formation require concomitant synthesis of RNA.

1.

INDUCTION OF GERMINATION

IN

B. emersonii

23

Last, let us come to the interesting question: How is protein synthesis suppressed in the nongerminating spore? Schmoyer and Lovett (1969) investigated some factors responsible for regulation of protein synthesis in germinating zoospores. When ribosomes isolated from nuclear caps were combined with cell-free protein-synthesizing systems derived from young germlings, no synthetic activity was detected ; when such inactive ribosomes were mixed with active ones (obtained from growing cells), the latter were rendered inactive. However, inactive ribosomes were activated by washing with KCl. From the KCl extracts, a fraction with inhibitor properties was isolated. Its behavior on gel columns resembled that of an inhibitor fraction isolated from a nonnuclear cap portion of the spore. It was concluded that the spore possesses a protein-synthesis inhibitor located in its cytoplasm and bound to nuclear cap ribosomes, and that inhibition is probably not due to background nuclease activity. I n contrast to some of the relatively unheralded esoteric phenomena set forth in earlier sections of this review-and some yet to come-the triggering of protein synthesis in fungal spores is presently a burning issue in many laboratories. Numerous questions are being asked about i t ; some of them have been neatly packaged in a recent review (Van Etten, 1969). They are, of course, the same kinds of questions being asked about the zoospores of B. emersonii. Solely from the sheer power of this multifronted attack, it can be optimistically supposed that much more will soon be suspected, if not known, about this particular facet of the biology of a Blastocladiella spore. VIII. Environmental Influences on Zoospores

A. EFFECTS OF Low TEMPERATURES The behavior of the zoospores of B. emersonii at low temperatures (ca. O0-loC) differs markedly from that a t slightly higher temperatures. This is not only of theoretical interest but also of practical significance because some routine procedures (Lovett, 1967a; Cantino et al., 1968) include manipulation of zoospores in an ice-water bath. It appears that ca. 4OC may be the transition temperature for the behavioral change because oxygen consumption is not detectable a t 4OC (Cantino et al., 1969) ; above this point, it increases linearly with temperature. Thus, below 4OC, those energy-producing functions which consume oxygen must shut down. This contrasts sharply with the fact that this same cell may exhibit (at 3OOC) an endogenous Qo2 as high as ca. 50-100 (McCurdy and Cantino, 1960; Cantino and Lovett, 1960). Sol1 and Sonneborn (1969) report that spores maintained in an incubator a t 3O-4OC can eventually germinate, although it takes much

24

LOUIS C. TRUESDELL AND EDWARD C. CANTINO

longer than a t higher temperatures (e.g., 15OC). However, in an ice-water bath a t O0-loC, we do not observe encystment; in fact, spores that remain long enough under these conditions lyse. 1. Morphological Changes during Incubation at Oo-1 O

C

During incubations a t O0-loC, spores eventually swell 2- to 3-fold (compare Fig. 3A and B) and become spherical [this and some of the following observations were made in a 10°-120C cold stage immediately after spores were withdrawn from an ice-water bath (Cantino et al., 1968) 1. Small cytoplasmic particles exhibit a rapid, random (presumably Brownian) motion not observed in swimming and amoeboid zoospores. Evidently, water uptake during swelling appreciably lowers cytoplasmic viscosity. The single mitochondrion enlarges somewhat and may become globose. Both light micrographs (Fig. 3B and C ) and electron micrographs (Shaw and Cantino, 1969; Cantino et al., 1969) also indicate that the mitochondrion is swollen, the latter pictures also showing that the backing membrane can break and that some of the lipid bodies may become dispersed throughout the cytoplasm. At first, spores in this swollen state can swim and (if brought back to room temperatures) are initially 95-100%viable (Cantino et al., 1968; Deering, 1968). Eventually, however, flagellar activity becomes increasingly erratic a t O0-loC and finally stops as the flagellum straightens out and extends directly away from the spore. Then, a t the base of the flagellum, the membrane sheath pulls away from the axoneme, the flagellum wraps around the spore, and it is absorbed. The time required for complete absorption may vary from a few seconds to a few minutes for different spores in the same suspension; this seems to be directly related t o the rate at which a spore is swelling. The axoneme can usually be observed within the spore, pushing against the plasma membrane and distorting its spherical contour momentarily. As the spore continues to swell, its spherical shape is regained ; the swelling continues until the spore bursts. The cytoplasm is discharged quite violently into the suspending medium, and the flagellar axoneme uncoils from its position within the spore to become readily visible. It is our impression that the amount of swelling is limited by the amount of membrane contributed from the flagellar sheath ; if the flagellum were not absorbed, the spore would not enlarge as much and would burst sooner, The method of flagellar retraction described above for chilled swollen spores is similar to one described (and labeled “wrap around”) by Koch (1968) for nonchilled spores of B. emersonii. Koch also portrayed two other types of flagellar retraction displayed by nonchilled spores-“body twist” and “vesicular”--which we, too, have observed in chilled swollen

1.

INDUCTION OF GERMINATION IN

B . emersonii

25

spores. The pictures provided by Koch (1968, Figs. 20-28) to illustrate these three methods of absorption invariably show his non-chilled spores to be highly swollen. However, his pictorial evidence for the method of flagellar retraction described earlier (Section IV) , which involves rotation of the nuclear apparatus, shows spores that are not swollen. Zoospores of B. emersonii can swell for a variety of reasons besides cold shock, e.g., overheating, changes in osmolarity, even the kind of paper used to wipe a microscope slide. Flagellar absorption associated with such swollen spores differs importantly from flagellar retraction in nonswollen spores: in the former, the nuclear apparatus does not rotate, the spore increases in volume, and the processes of flagellar absorption and zoospore encystment are not intrinsically associated; in the latter, the opposites are true. 2. The Influence of Low Temperature Incubation on Encystment

Increasing the duration of cold incubation of a spore population increases the percentage of spores encysting after the population is removed from the cold (Cantino et al., 1969). This behavior is partially dependent on the nature of the suspending medium. In Fig. 5 , curve C represents a spore population suspended in a PIPES-buffered medium (see legend for details). Spores are very stable in this system and display no lysis during the entire 110 minutes in the cold. After an initial rise, there is no further increase in encystment capacity (as measured a t 22OC) until spores have been chilled for 70 minutes. When spores are cold incubated in medium GC (curve A, Fig. 5 ) , encystment percentage increases immediately and continues rising to a maximum a t ca. 90 minutes, after which it declines. Spores begin to lyse around the time of maximum encystment, and continue to do so thereafter. While they are in the cold, there is no way of distinguishing spores that have been triggered to encyst from those that have not; therefore, it is impossible to determine whether the former are the ones that are lysing. I n other words, we cannot determine whether cell lysis is the direct cause of decreased encystment. Yet, it is tempting to suppose that, for each spore, the triggering of encystment always precedes lysis. I n a typical population, this event would occur asynchronously t o the extent that, while one spore is being induced to encyst, another one may be lysing. I n this manner, maximum encystment would be limited by the synchrony of the particular population under consideration. I n populations displaying a highly synchronous response to a cold incubation, 100% encystment would be expected. When glutamate is omitted from medium GC (compare curve B with

LOUIS C. TRUESDELL AND EDWARD C. CANTINO

26

0

I /

I

50

I00

MINUTES

FIG.5. Influence of cold incubation period in four different media on encystment capacity of washed spores, i.e., encystment percentage after spores are transferred from O"-l"C to 22°C. Curve A : medium GC, containing 0.5 mM Na phosphate (pH 7.8), 0.2 mM CaClr, 5 mM KCl, and 1 mM Na glutamate. Curve B: medium GC minus glutamate. Curve C : medium GC with 2/6 of its KC1 replaced by 2 mM piperasine-N,NN'-bis(2-ethane sulfonate) at pH 7.0 (PIPES) (Good et al., 1966). Curve D : Na morpholinopropane sulfonate (MOPS) buffer (a GOOD buffer; CalBiochem, Los Angeles) at pH 7.8. Population density for curves A-D, 5 x 10" spores/ml.

curve A, Fig. 5 ) , spores become very sensitive to low temperatures. Lysis ensues within the first 40 minutes and is accompanied by a substantial decrease in the population's encystment capacity. Almost all cells lyse by 60 minutes. Although such lability evidently results from omission of glutamate in medium GC, work with other media demonstrates that glutamate per se is not essential for maintenance of zoospore integrity a t low temperature. Spores suspended in media composed entirely of MOPS buffer (Fig. 5 , curve D) are very stable; there is no lysis throughout the cold incubation. Spores swell more gradually and survive longer than in medium GC; suspensions have been kept in this medium for up to 3 hours with less than 1% lysis, and they can probably be kept up to 6 hours or more without much additional breakage. Spores suspended in Na or K phosphate buffers (1-4 mM, p H 6.0, 6.8, 7.8) display behavior patterns similar to that obtained in medium GC without glutamate; lysis usually begins after 30-40 minutes in the cold. It is evident that the behavior of cold-incubated spores is greatly

1.

INDUCTION

OF GERMINATION

IN

B. emersonii

27

dependent on the medium in which they are suspended. It is not yet possible to make generalizations about the influence of individual compounds or ions on either spore viability in the cold or cold induced encystment. However, the function of glutamate should be examined further, for it does appear to control lysis in medium GC. I n this connection, it is of interest that in another medium composed of 2 mM sodium MOPS (pH 7.8), 1 mM KC1, and 0.1 mM glutamate, spores are not very stable and lysis begins after only 70 minutes of cold incubation. This is unexpected, since spores are stable in medium GC, which contains both KCI and glutamate, and they are very stable in MOPS buffer alone. These results point to the interesting possibility, consistent with all available data thus far, that the presence of either glutamate or phosphate can lead to instability in the cold, while the presence of both together yields cold-stable spores. Such an interaction might be directly related to the metabolic mechanisms associated with spores a t low temperatures.

B. SELF-INHIBITION IN SPORE POPULATIONS When a population of spores, kept in medium GC a t O0-loC for 90 minutes, is brought to 22”C, its encystment percentage is inversely related to population density (Cantino et al., 1969). Such “self-inhibition” is readily apparent between loo and lo7 spores/ml, but it has not been precisely determined a t what density it is first detectable. However, Sol1 and Sonneborn (1969) have noted that, in their system, it begins a t concentrations as low as ca. 6 x lo5 cells/ml. This observation has obvious implications: ( a ) the density of a spore suspension is a variable that must be rigidly controlled to obtain reproducible results in germination experiments; (b) the population density can be regulated so as to yield populations of either almost wholly encysted or wholly nonencysted spores; and ( c ) of more theoretical interest, any insight into the mechanism of self inhibition might reveal important information about the mechanism underlying control of encystment. Consequently, some aspects of this phenomenon which have been investigated further are discussed below. 1. Inhibition of Encystment by Cell-Free Supernatants from Spore Suspensions

We have examined the supernatants of high density spore suspensions for inhibitory properties. A “cold supernatant” was obtained from suspensions maintained a t O0-loC for 90 minutes ; a “warm supernatant” was obtained from the same cold-incubated suspension, but after a secondary incubation a t 22OC for an additional 20 minutes. The results

28

LOUIS C. TRUESDELL AND EDWARD C. CANTINO

of a representative experiment are tabulated in Table I; other details are provided in the footnotes. The warm supernatant inhibits encystment a t both dilutions ( l a and l b, Table I ) . The cold supernatant, unlike the warm supernatant, does not suppress encystment. Therefore, the inhibitory properties must be bestowed upon the suspending medium after spores have been removed from the cold. Presumably, it could result from either release of inhibitory substances into the medium or utilization of substances present in the medium in limiting quantities and required for encystment. The latter alternative seems unlikely; there is no evidence for the presence of specifically required substances except oxygen (Cantino et al., 1968) , and the latter would not have been limiting in the warm supernatant because there was ample time and agitation for it to reach saturation levels before the assay. Changes in pH probably were not limiting because supernatants were within 0.2 pH units of one another, and encystment is unaffected by pH over the range 6.0-9.0 (Sol1 et al., 1969). We hypothesize, therefore, that an encystment inhibitor is released by spore suspensions of B. emersonii. These data, together with others not reported here, show that encystment values for warm supernatant fall within a few percent of one another for both dilutions of the supernatant, and regardless of the density of the suspension from which the supernatant, was obtained (range: 8.8-12.0 X lo6 spores/ml) . Why both dilutions of supernatant (Table 1) should possess comparable inhibitory properties is an interesting but moot point; perhaps encystment cannot be suppressed below a certain level (e.g., 30-3576; Table I), and sufficient inhibitor is present a t either dilution to reach this limit. Cold treatment may induce the rounding up of a spore irreversibly; if so, no level of inhibitor would stop an induced cell from encysting.

6. The Influence of Environmental Factors on Self-inhibition Self-inhibition of germination among other fungal propagules is well known (see review by Sussman, 1965; also, Blakeman, 1969; Garrett and Robinson, 1969; Fletcher and Morton, 1970; and references therein) ; its occurrence during encystment in B . emersonii seemed sufficiently important to warrant its further characterization. The influence of selected salts, temperatures, and p H levels on encystment a t various population densities was therefore examined, not with a view toward doing a comprehensive survey but, rather, to finding some appropriate set of conditions that would produce minimum or maximum inhibition and establish patterns for the effect of specific variables on self-inhibition. Nine different systems (combinations of variables) were investigated

1.

INDUCTION OF GERMINATION I N

B. emersonii

29

TABLE I

INHIBITION OF ENCYSTMENT BY CELL-FREESUPERNATANTS FROM

SPORESUSPENSIONS~~

No. la lb

Assay Medium sporesb GC (ml) (ml) 2.5 2.5 2.5 0.5 0.5 0.5

2.5 4.5 -

Cold SUP (ml)

Warm SUP (ml)

2.5

2.5 4.5

-

4.5 -

Final population density (spores/ml) 5 5 5

x x x

106 106 106 106 108 106

Number of spores Amount of encystedc (%) inhibition6 45.3 43.9 29.1 93.0 91.5 34.9

1.4 16.2

-

2.5 59.1

5 Suspensions of 8.8 X loe spores/ml were placed in medium GC (see legend, Fig. 5) a t 0’-1°C for 90 minutes. An aliquot was immediately centrifuged in the cold to remove spores, and the supernatant (“Cold sup”) was assayed. A second aliquot was kept for 20 minutes a t 22”C, chilled quickly, centrifuged, and the supernatant (“Warm sup”) was also assayed. Both supernatants were stored ca. 2 hours at 0’-1°C before assay. The supernatants were tested for inhibitory properties against a suspension of lo7 “assay spores”/ml (also previously incubated in medium GC at O0-1”C for 90 minutes); spores were mixed with either Cold sup, Warm sup, or medium GC (control) in the proportions shown, and immediately brought to 22°C for encystment. Average of three assays. Percent encystment in sample less percent encystment i n control.

(Table 11) ; spores were subjected to each one for 20 minutes at O0-loC before encystment was initiated by bringing the spores to either 15O or 22OC (hereafter referred to as the “secondary incubation”). Within minutes after the temperature is raised, spores begin encysting; after 30 minutes of secondary incubation, almost all spores which can encyst have done so (30 minutes is not necessarily adequate for the maximum encystment percentage in other systems; see Soll et al., 1969; Soll and Sonneborn, 1969). The data displayed appreciable scatter; therefore, straight lines corresponding to Eq. (1) were fitted to the data by regression techniques; R=a+bp

(1)

their corresponding correlation coefficients and confidence intervals were determined by conventional techniques (e.g., Goldstein, 1964). I n Eq. ( l ) , R = decimal fraction of encysted spores in the suspension; p = population density in spores/ml ( x 10-O) ; a and b are constants determined

w

0

TABLE I1 REGRESSION EQUATIONS FOR SELFINHIBITION IN DIFFERENT ENVIRONMENTAL SYSTEMS System No.

Suspending medium

1

GCQ

Regression line R=a+bp R = a' b' In p

Temperature Sample ("C) points 22

21

la

GC

15

26

2

GMb

22

18

2a

GM

15

29

3

Na phosphate, 1 mM, pH 7.8

22

18

3a

Na phosphate, 1 m M , pH 7.8

15

15

4

Na phosphate, 2 mM, pH 7.8

22

18

5

Na phosphate, 2 mM, pH 6.1

22

17

6

K phosphate, 2 mM, pH 7.8

22

26

R R R R R R R R R R R R R R R R R R

+

Correlation coef. ( r )

tZ

0.898 - 0 . 0 8 1 ~ 1.055 - 0.378 In p 0.900 - 0 . 0 7 9 ~ 1.012 - 0.349 In p 0.991 - 0 . 0 6 1 ~ 1.007 - 0.203 In p = 0.985 - 0 . 0 5 1 ~ = 1.060 - 0.230 In p = 0.989 - 0 . 0 9 8 ~ = 1.157 - 0.447 In p = 0.885 - 0 . 0 7 9 ~ = 0.929 - 0.295 In p = 0.702 - 0 . 0 5 7 ~ = 0.807 - 0.267 In p = 0.769 - 0.048~ = 0.835 - 0.216 In p = 0.757 - 0 . 0 4 1 ~ = 0.794 - 0.167 In p

-0.884 -0.898 -0,770 -0.774 -0.863 -0. 800 -0.774 -0.761 -0.905 -0.912 -0.752 -0,738 -0.946 -0.927 -0.762 -0.752 -0.782 -0.759

0.781 0.805 0.593 0.599 0.745 0.640 0.599 0.579 0.812 0.832 0.566 0.545 0.895 0.859 0.581 0.566 0.612 0.576

=

= = = = =

Medium GC is defined in legend to Fig. 5. Medium GM differsfrom medium GC in that Ca is replaced by an equivalent amount of Mg.

1.

INDUCTION OF GERMINATION IN

B. emersonii

31

by the regression techniques. The absolute value of b is referred to as rate of inhibition because it is the rate a t which encystment decreases as population density increases; specifically, it is a measure of the decrease in R for every increase of lo6 spores/ml. Curves corresponding to Eq. (2) were also fitted to the data by regression techniques.

R

=

a'

+ b' In p

(2) The correlation coefficients for the linear and logarithmic equations were similar. This may mean that the true relationship between percent encystment and population density is not given by either relationship but, rather, that both relationships are almost equally good approximations. Overall, the linear equation yielded a slightly better fit and, therefore, will be used for the following comparisons among the test systems. The equations and their correlation coefficients ( r ) are tabulated in Table 11. I n every system examined, inhibition of encystment increases as the population density increases ; this relationship is reflected in the fact that r is not zero. However, there is always the possibility that this might have occurred by chance, i.e., that a population of unrelated sample points might have fallen by chance into a linear pattern. The probability of this event occurring was tested (Goldstein, 1964) ; it was much less than 1% for every system. Thus, self-inhibition apparently occurs under all conditions (Table 11) investigated. The regression lines for the nine systems are plotted in Fig. 6. In each graph (A, C, D, El F ) , two systems are compared which differ with respect to only one variable (except B, which makes two comparisons). a. The Influence of Secondary Incubation Temperatures. Preliminary work had already suggested (Cantino et al., 1969) that the relationship between population density and encystment was not sensitive to the secondary incubation temperature-a phenomenon which, if put upon an unassailable foundation, would be intrinsically very interesting. The influence of this secondary incubation temperature was further tested for medium GC, GM, and sodium phosphate by comparing systems 1 and l a , 2 and 2a, and 3 and 3a, respectively. (identified in Table 11). Two temperatures, 22O and 15OC, were compared in each case. It is obvious (Fig. 6A, 6B) that, in all three media, the 7OC difference in secondary incubation temperature had little effect on the regression lines. However, it might also be argued that significant differences associated with changes in secondary incubation temperatures are reflected in the squared correlation coefficients ( r 2 ) since they are significantly less for the lower temperatures (0.59 to 0.57 us. 0.81 to 0.75). To a small degree,

32

LOUIS C. TRUESDELL AND EDWARD C. CANTINO

lOOr

C

100

r

50.

. F

FIG.6. Effect of environmental variables on self-inhibition in spore populations. Curve numbers correspond to system numbers in Table 11. Vertical axes: percent encystment; horizontal axes: spores/ml ( X Straight lines that best fit the data were established by standard regression technique ; see text (Section VIII,B, 2) for details.

1.

INDUCTION O F GERMINATION IN

B . emersonii

33

this difference may reflect increased scatter a t 15OC, but most likely it results primarily from an alteration in the functional relationship between R and p ; i.e., a linear relation between R and p a t 15OC is not as good an approximation of the actual relationship between these variables as it is at 22OC. b. The Relative Effects of Ca and Mg. Since medium GC (systems 1 and l a , Table 11) differs from medium GM (systems 2 and 2a) only in the substitution of Ca for an equivalent amount of Mg, the behavior of spores in these two media can be compared to determine the relative effects of Ca and Mg on self-inhibition. Regression lines from systems 1 and 2 (Fig. 6C) are well displaced from one another with respect to percent encystment. A very small overlap of confidence intervals (not plotted) between these two systems supports the reality of this difference. It can be concluded with a fair degree of certainty, therefore, that less self-inhibition will occur with Mg than Ca for any specific population density between the limits of lo6 and 10' spores/ml. A similar conclusion can be drawn from the comparison of systems l a and 2a (not plotted; same media, but with a 1 5 O secondary incubation temperature). c. The E f f e c t of Different Concentrations of N u Phosphate. A comparison of systems 3 and 4 (Table 11, Fig. 6D) reveals the effect of doubling the concentration of sodium phosphate a t pH 7.8. Although the rates of inhibition (slopes of regression lines) differ greatly, this difference tests with a confidence of only ca. 70% (test for nonparallelism in Goldstein, 1964). However, the probability that the difference is real is strengthened by the fact that all three systems in 2 mM phosphate buffers (Table 11, systems 4, 5, and 6) have considerably lower rates of inhibition than the systems in 1 mM concentrations (Table 11, systems 3,3a ) . d . Comparison of Phosphate Buffers Containing N a and K . In system 6, K is substituted for the Na in system 4. A comparison of regression lines for these two systems (Fig. 6F) shows significant displacement along the vertical axis, thus signifying consistently less inhibition with K (within the range lo6 to lo7 spores/ml). I n addition, the value of r2 changes from ca. 0.9 to 0.6 (Table 11); such a lowered value of r2 indicates that the linear function, Eq. ( l ) ,is not a very good approximation of the real relationship of R and p in the system containing K. e. Comparison of N a Phosphate at p H 7.8 and 6.1. In systems 4 and 5, sodium phosphate buffers at pH 7.8 and 6.1 are compared. Again, a shift along the vertical axis is notcd (Fig. 6 E ) . Although not as great as that obtained in a comparison of systems 4 and 6, it is nonetheless significant, displaying only a slight overlap of 90% confidence intervals.

34

LOUIS C. TRUESDELL AND EDWARD C. CANTINO

Thus, in Na phosphate there is less self inhibition a t pH 6.1 and 7.8. At the lower pH, the value of r2 is also smaller (ca. 0.6 vs. 0.9; Table 11).Thus, a linear relationship between R and p constitutes a relatively poor approximation for describing self inhibition in the p H 6.1 system. Interestingly, the change from pH 7.8 to 6.1 yields a behavior pattern similar to that induced by substituting K for Na. f. Peroration. The results of the foregoing comparisons among nine systems examined for self inhibition in the range lo6 to lo7 cells/ml may be summarized as follows: 1. I n every system, inhibition of encystment increased as the population density was raised. 2. I n keeping with preliminary observations, the assay temperature had surprisingly little effect on inhibition, although a small change was reflected in the comparative values of r2. 3. Substitution of Mg for Ca lowered self inhibition. 4. Doubling the concentration of Na phosphate from 1 m M to 2 mM lowered the rate of inhibition. 5 . The substitution of K for Na in 2 m M Na phosphate a t p H 7.8 lowered self inhibition and the value of r2. 6. A change of pH from 7.8 to 6.1 in a Na phosphate system lowered self inhibition and the value of r2. C. ALTERNATIVE MEANSOF EFFECTING ENCYSTMENT Thus far, two methods for regulating zoospore germination in B. emersonii ha.ve been discussed in some detail: cold treatment and control of population density. Below, we outline briefly additional ways of doing it. 1. E f f e c t s of Some Inorganic Salts

The chlorides of K, Na, Rb, and Cs elicit encystment of zoospores under certain conditions (Soll and Sonneborn, 1969); Br- and I- are equally suitable as counterions for K+ and Na+, but their R b and Cs salts have not been tested. The K and Na salts of more complex anions such as so,'- or Mood2-,which induce encystment to varying degrees, are not as effective as the halides (Soll, 1970). Neither LiCl nor NH,Cl cause encystment; on the contrary, in every instance tested, LiCl strongly (but reversibly) inhibits encystment previously (or simultaneously) induced by other salts. CaC1, and MgCl, initiate encystment, but with inhibition of germ tube formation.

1.

INDUCTION

OF GERMINATION

IN

B. ernersonii

35

2. Eflects of Sulfonic Acid Axo Dyes

Sulfonic acid azo dyes such as Biebrich Scarlet and Methyl Orange trigger encystment quite effectively. For example, concentrations of !m i induce 25% encystment in otherwise Biebrich Scarlet as low as 0.05% nonencysting spore suspensions. Also, spore populations chilled to O0-loC, mixed with the dye, and then brought back to some higher temperature (e.g., 22OC), display much greater encystment percentages than unchilled spores that are simply mixed with the dye a t the higher temperature. Neither of the above dyes seems to stain markedly any particular spore structure or organelle. But, when added to Difco PYG agar media a t concentrations sufficient to induce encystment, they inhibit growth after the early germling stage. Although a wide range of sulfonic acid azo dye structures apparently induce encystment, no attempt has yet been made to correlate structure with effectiveness. We can state, however, that the relatively simple structured Na benzene sulfonate is without effect, and that most cationic dyes cause swelling and lysis within a few minutes. IX. Kinetics of Encystment Soll and Sonneborn (1969) have made a detailed study of the influence of cellular and environmental variables on the germination kinetics of B . emersonii zoospores, and many of their results have already been cited in previous sections of this review. In their first paper (Soll et al., 1969) methods were presented for interpreting germination curves. Before considering them, we will discuss an alternative approach. When induced by any of the methods heretofore described, encystment of spore populations follows a characteristic pattern of kinetics. If plots of encystment percentage us. time are constructed, the curves are sigmoid, always display a slight lag before encystment is detected, and reach final eneystment-percentage plateaus which are frequently below 100%. All evidence to date indicates that once a spore population reaches a plateau, there will be no additional encystment until the population is again induced to encyst. To illustrate: a typical spore population was induced with Biebrich Scarlet; encystment leveled off within 30 minutes, after which the plateau held steady for a t least 8 hours.

A. THENORMALDISTRIBUTION AS A MODEL The sigmoid shapes of the curves described above indicate that the frequency of encystment may be normally distributed about a mean time. This hypothesis can be tested by plotting the data with modified

36

LOUIS C. TRUESDELL A N D EDWARD C. CANTINO

coordinates such that encystment percentage is transformed (see, for example, Goldstein, 1964) into normal equivalent deviates (N.E.D.) or probits (N.E.D. 5 ) . Alternatively, the data can be plotted on normal probability paper. In any event, sigmoid curves will now map linearly if encystment is normally distributed with respect to time. This is, in fact, what happens when encystment curves for spore populations which have attained 100% encystment are so plotted. However, for a population which exhibits a plateau under loo%, the mapping is not linear unless all encystment-percentage points are calculated relative t o the final encystment percentage eventually attained by the population. I n effect, this treats the encysting spores as though their behavior were independent of nonencysting spores; the implications of this will be discussed shortly. It should be noted that the treatment of encysting spores as a separate population independent of nonencysting spores is applicable to the limiting case where 100% of the spores encyst. Thus, this is the most general and comprehensive way by which to view the kinetics of encystment. Since the normal distribution is an appropriate model for describing the time course of encystment of the zoospores of B. emersonii, the parameters which characterize a normal distribution are sufficient to characterize an encystment curve. There are only two such parameters, the mean and the standard deviation. The mean, in this instance, is the arithmetic average of the time it takes each encysting spore to encyst after induction; it will be referred to as the “mean time of encystment,” T. It also equals, as a result of the normal distribution symmetry, the time a t which 50% of the encysting spores have encysted, and it may be read directly from encystment curves. The other parameter, the standard deviation (s), can be used to completely and quantitatively represent the synchrony of an encysting population. It is a measure of the dispersion of individual spore-encystment times about the mean time of encystment. For a population of n spores with individual encystment times T i , it may be specifically defined as:

+

I

n

However, it is not essential to determine s with this equation because the slope of a transformed (i.e., to linear form) encystment curve is and s can easily be determined from transequal to l/s. Thus, both formed encystment curves (see curve 6, Fig. 7 ) . With the aid of these two parameters, any information about the time course of encystment can be calculated. Several illustrations follow.

1.

INDUCTION

OF GERMINATION

IN

B. emersonii

37

MINUTES

FIG.7. Time course of encystment for spore suspensions induced with cold (curve l), Biebrich Scarlet combined with cold (curve 3), and KC1 (curve 6). The curve numbers correspond to experiments 1, 3, and 6 in Table 111. Induction is defined in the footnote to Table 111. A method for determining methodology T and s is illustrated with curve 6. Data have been transformed to linear mappings by using a normal probability scale. Suppose it is desired to determine how long it takes 97.576 of a spore population to encyst after induction. It is only necessary to look up 0.975 in a table of areas under the standard normal density function and read off the corresponding distance from the mean in terms of standard deviations ( = 1.96). Therefore, if, for example, s = 6.9 minutes (as in Expt. 6 , Table 111), it would take 6.9 X 1.96 = 13.52 minutes to go from T to 97.5% encystment. Since T = 28.9 (Table III), the total time required for 97.5% encystment = 28.9 13.5 = 42.4 minutes. Alternatively, if it is desired to determine how long it takes a spore population to go from X % to Y % encystment, i t is only necessary to establish the distance of each from F in terms of standard deviations as determined above, and then to calculate the interval in minutes. The foregoing methodology for analyzing and characterizing germination curves offers some distinct advantages over that employed by 8011 et al. (1969) in their recent interesting work with B . emersonii. While they recognized that encystment was normally distributed, they did not choose to exploit this fact; rather, they used the sigmoid curve per se and zoospore “major to determine two basic parameters, zoospore “T5011 slope.” The T,, was defined as the time necessary for 50% of the ZOO-

+

38

LOUIS C. TRUESDELL AND EDWARD C. CANTINO

spores to become round cells (i.e., encyst). The major slope was the slope of the straight line used to approximate the "major rise (or fall) portions" of encystment curves, and was used as a measure of synchrony. However, in comparison with s, the major slope is an inferior measure of synchrony for three reasons: first, i t rests on the assumption that encystment is uniformly distributed over a selected interval, which it is not; second, it can be greatly influenced by the magnitude of the specific interval about T,, through which the approximating line is drawn; and third, it is affected by the degree of uniformity with which data points are distributed within this specific interval. Such difficulties are further complicated by the fact that the combination of T,, and the major slope docs not fully characterize the germination curves; to extract additional information from them, new parameters, such as the initial lag period, T,,, and T,,,must be utilized.

B. COMPARISON OF INDUCTION METHODS I n our studies of the kinetics of encystment, three inducers were used: cold incubation, sulfonic acid azo dyes, and KCl. The data were plotted and characterized by the standard methodology described in the preceding section. Representative linear mappings, derived by transformation of the kinetics data, are delineated in Fig. 7; additional s, and maximum percent encystexamples, summarized in terms of ment, are listed in Table 1 1 1 .

r,

TABLE I11 COMPARISON OF DIFFERENT INDUCTION METHODS Exp . No. 1 2 3

Method of inductiona 2.5 hr a t O0-1"C 2.0 hr a t O"-l°C Biebrich Scarlet (1 mM) with 15 min a t O"-l"C Biebrich Scarlet (1 mM) Methyl Orange (0.5 mM) KCl (50 mM) KC1 (50 mM) KCl (25 mM)

Population density Maximum % (spores/ml; x 10-6) encystment

F !

s

(min) (min)

4.04 2.18 1.9

31.4 22.3 100.0

13.1 14.2 13.7

5.4 3.8 4.5

4.3 4.7 3.6 2.7 4.1

43.7 14.3 100.0 55.0 16.0

12.7 14.2 28.9 29.6 30.7

4.2 4.8 6.9 8.9 9.1

aFor Exp. 1 and 2, the spore suspension was incubated a t O0-1"C and then brought to 22°C (zero time); for Expt. 3, it was pre-chilled to O"-l"C, mixed with prechilled dye, and then brought to 22'C after 15 minutes (zero time) ; for Exp. 4-8, the cold incubation was omitted (zero time measured from time of addition of dye or salt to spore suspension). I n all cases spore suspensions were harvested from Difco PYG agar cultures in Na MOPS buffer (2 mM, p H 7.8).

1.

INDUCTION OF GERMINATION I N

B. emersonii

39

The tabular data reveal that neither mean time of encystment nor synchrony are dependent on the maximum percent encystment. This observation holds true regardless of the method by which maximum percent encystment is regulated-for example, by altering the time spores are incubated at O0-loC for the cold-induced encystment; by including (or not including, as the case may be) a short cold-incubation (see Section VIII,C, 2) with the azo dye induction, or by substituting one dye for another (i.e., Methyl Orange for Biebrich Scarlet) ; or by varying KC1 concentrations. It is obvious from these observations that the interplay between endogenous and exogenous factors which regulate the fraction of the spores in a population that encysts does not affect the time it takes a spore to encyst after induction. If the process of induction is conceived of as a trigger for the succeeding processes of encystment, then for each specific method of induction the trigger is (for each spore) an all-or-none event that does not affect the rate of encystment. Within this conceptual framework, an additional observation must be rationalized. Spores induced to encyst by cold and sulfonic acid azo dyes have a mean time of encystment about half of that for spores induced by KC1. Apparently, the rate of the triggering process will not account for the differences in F , for i t has been determined (Soll, 1970) that as little as 30 seconds of contact with 50 m M KC1 is sufficient to trigger complete encystment in a population of zoospores. Thus, the ! for KC1-induced cells must result from differences higher value of ? in subsequent (i.e., secondary) events leading to encystment. One might speculate that the increase in results from an inhibitory effect of KCl on these subsequent processes, but this does not seem likely because variations in KC1 concentrations do not affect F. The triggering of encystment with cold or azo dyes, as compared with the triggering by KCl, must increase the average rate and/or decrease the number of secondary processes which lead to cyst formation. X. Concluding Remarks The spore of Blastocladiella emersonii contains an exceptionally tight and highly ordered arrangement of membrane-bound organelles. During encystment, this subcellular assemblage is swiftly disorganized by a cascade of changes: the flagellum is retracted; the nuclear apparatus rotates; the cell becomes spherical, loses volume, and forms a cyst “wall.” Substantial vesiculation accompanies the process. This succession of events is temporally coordinated and spatially integrated. Axonemal translocation and rotation of the nuclear apparatus results from structural-mechanical transformations requiring no special energy sources or mechanisms other than those needed for the other

40

LOUIS C. TRUESDELL AND EDWARD C. CANTINO

processes associated with encystment. I n the more camouflaged operation of cell wall synthesis, decay of gamma particles generates vesicles that fuse with the plasma membrane and thereby alter it; presumably they bring enzymes and/or structural components to the cell surface to lay down the foundation for synthesis of the cyst wall. Moreover, vesicles fusing with the plasmalemma contribute, on the one hand, to the volume decrease associated with encystment; on the other hand, by depositing new cyst wall material, they may be generating the surface forces needed for the change in cell shape and translocation of the axoneme. But, from all this there also stems a cardinal question: By what means are such events prevented in a nonencysting spore? There must be interlocking ways by which a zoospore keeps some things shut down. A simple on-and-off inhibitor-mediated process could underlie one of them, for Schmoyer and Lovett (1969) provide direct evidence for an internal inhibitor that suppresses protein synthesis, and we have indirect evidence that a substance released by spores is capable of regulating encystment. Yet, satisfying though it is to achieve simple answers to complex problems, the existence of such chemical agents provokes new questions. One of them is especially meaningful : Should the unidentified material which decreases percent encystment in a zoospore population be viewed as an encystment inhibitor or as a zoospore stabilizer? The significance of this question goes well beyond the problem in semantics that it poses. Suppose it could be shown unequivocally that the encystment “inhibitor” actually stabilizes zoospores in the sense of buffering them against adverse environments. For example, if the “inhibitor” prolonged the time that swimming spores could withstand some new external stress, or the proportion of them that lysed after the environmental change took place, would not “stabilizer” be the more appropriate term to use? Already there is some evidence to suggest that the stabilizer concept is preferable to the inhibitor concept, a t least insofar as it applies to those properties of spore supernatants that prevent encystment. We find that cells most recalcitrant to encyst upon induction are also the ones that most frequently exhibit the greatest resistance against cold-induced lysis. Along another tack, Sol1 (1970) reported that zoospores derived from plants grown under relatively crowded conditions, as compared to cultures a t lower population density, are more resistant to KC1-induced encystment and survive longer in balanced salt solutions. Hopefully, these suggestive observations will stimulate more investigations in this direction. I n any case, returning to the question of causality, the intricate character of a zoospore’s internal architecture may also be operating to prevent encystment; unfortunately, it is exceedingly hard to demon-

1.

INDUCTION OF GERMINATION IN

B . emersonii

41

strate. The notably compact nature of the gamma matrix could conceivably render impotent any enzyme complement contained in its interstices. The GS-membrane around it might also serve a regulatory function, i.e., via selective permeability. However, it is also clear that subcellular compartmentalization need not always provide a limitative function; the membrane surrounding the ribosomes in the nuclear cap is obviously not needed to inhibit protein synthesis, for this can be accomplished by way of the inhibitor. It may be, therefore, as Schmoyer and Lovett (1969) suggested, that the function of the nuclear cap membrane is to protect inactive ribosomes from degradation in a nongerminating zoospore of B. emersonii. But even this seemingly logical answer may be a trou-de-loup, for there are other uniflagellate fungi [e.g., Rhizidiomyces and Monoblepharella (Fuller and Reichle 1965, 1968) ] in which nonencysting zoospores carry around a caplike ribosomal aggregate not “protected” by a surrounding membrane. I n fact, in some zoospores [e.g., Olpidium (Temmink and Campbell, 1969a) 1, ribosomes are evenly scattered throughout the cytoplasm. The answer may simply be that the primary-if not the only-role of the nuclear cap membrane in the zoospore of B. emersonii is t o provide an immediate source of endoplasmic reticulum for early protein synthesis. This inquiry into the enigmatic nature of encystment also calls for brief consideration of some aspects of the kinetics of encystment. Spores induced with low temperatures and sulfonic acid azo dyes display similar kinetics, while spores induced with KC1 show a much greater mean time of encystment. The evidence suggests that the difference in behavior must result from an increase in the number and/or the average rates of encystment “processes” which ensue after triggering. When the reason for this difference is uncovered, an important aspect of encystment will have been resolved. But in the meantime, a comparison of similarities among induction methods may also be fruitful. First, as far as we can tell, they all yield essentially the same sequence of structural changes associated with encystment, from which it can be argued that all induction methods must affect pathways that converge a t some common step or process. Second, they all involve the placement of a spore under great stress. During cold incubation, the cell eventually becomes precariously balanced on the verge of lysis as it takes up water and approaches its “elastic limit.” In sulfonic acid azo dyes, the effects are not so dramatic, but experience shows that spores in contact with these dyes are more labile to the effects of other environmental factors, such as cold incubation and fixation. I n KC1, spores are also being pushed toward their limit; in 50 mM KCl, there is a little lysis, and a t higher concentrations lysis is substantial.

42

LOUIS C. TRUESDELL AND EDWARD C. CANTINO

One way of harmonizing these observations into a unified concept of the trigger mechanism is to conceptualize the induction step as a perturbation of the delicate balance of cellular controls in a spore. A momentary imbalance could evoke emergence of the new set of interrelated processes which comprise encystment. If this accurately represents what takes place, then it would be most enlightening to find answers to the question: what specific cellular controls will, when disturbed, lead to breakdown of other control mechanisms? Diverse induction methods may disrupt different control mechanisms. Disengagement of some of them will be sufficient to induce encystment; disengagement of others will not, and may, instead, cause autolysis or cell death. For example, cold incubation may be the type of trigger that interferes in a nondiscriminating fashion with control processes in the zoospore, If it modulates those critical things that suppress encystment without tampering with those that cause cell death, the spore will be induced to encyst. We can also suppose that when the different induction methods act on dissimilar controls, the results may pivot the breakdown of others. Depending on what is first attacked, the sequence of subsequent disjunctions will probably vary, as will their rates. Thus, the identity of the first control mechanism to be attacked will establish the rate of subsequent events; this will be reflected in the value of for the particular induction method used. Finally, we may well ask, what does all this have to do with the real world of aquatic fungi, and the more “natural” agents that induce their motile propagules to encyst? Mycologists and phytopathologists have recognized for many years that the element of change plays an importantalbeit a poorly understood-motivating role in germination. This knowledge is reflected in generalizations of the following sort: “Undey circumstances which provide for prolonged motility . . . encystment can be readily induced, often quite quickly, by changing some existing environmental factor, or introducing a new one, be it physical or chemical” (Hickman and Ho, 1966). The fact that we have disturbed spores in the laboratory with treatments harsher than some of those generally encountered in nature should not mask the utility of these laboratory techniques for uncovering fundamental aspects of the encystment process. It was demonstrated for the three induction methods examined that the severity of the perturbation only affects zoospore stability (whether this be measured as the number of spores that encyst or the number that lyse) ; it has no demonstrable effect on the processes associated with encystment. Although, ideally, less drastic methods of induction might seem to be preferable, there are practical reasons for continuing to use these experimental methods. Primarily, less drastic

1.

INDUCTION

OF GERMINATION IN

B. emersonii

43

methods do not yield the high levels of encystment or synchrony needed for certain kinds of work. Secondarily, the high population densities required for many biochemical and other kinds of experiments markedly inhibit encystment; more stringent methods of induction are therefore required. I n our experience with B. emersonii, what may be the most “natural” method of inducing encystment is simply to dilute a zoospore suspension to a lower population density. Unfortunately, this procedure is much less effective than the other methods we have used; furthermore, it results in spore suspensions of lower density. Our preliminary investigations indicate that its kinetics may be different than that for the other method of induction. Further studies with this form of induction are certainly needed. ACKNOWLEDGMENTS The results of unpublished work by the authors reported herein was sustained in part by general research support grants AI-01568-12, 13, 14 from the National Institutes of Health and GB-4967 from the National Science Foundation. REFERENCES Anderson, 0. R., and Roels, 0. A. (1967). J. Ultrastruct. Res. 20, 127. Bartnicki-Garcia, S. (1968). Annu. Rev.Microbiol. 22, 87. Blakeman, J. P. (1969). J . Gen. Microbiol. 57, 159. Bracker, C. E. (1967). Annu. Rev. Phytopnthol. 5, 343. Camargo, E . P., Dietrich, C. P., Sonneborn, D., and Strominger, J. L. (1967). J . Biol. Chem. 242, 3121. Camargo, E . P., Meuser, R., and Sonnehorn, D. (1969). J. BioE. Chem. 244, 5910. Cantino, E . C. (1966). In “The Fungi” ( G . C. Ainsworth and A. S. Sussman, eds.), Vol. 2, p. 283: Academic Press, Xcm York. Cantino, E . C. (1969). Phytopathology 59, 1071. Cantino, E . C., and Horenstein, E. A . (1956). Mycologia 48, 443. Cantino, E. C., and Hyatt, M. T. (1953). Anonie van Leeuwenhoek, J. Microbiol. Serol. 19, 25. Cantino, E . C., and Lovett, J. S. (1960). Ph~/siol.P l n ~ t 13, . 450. Cantino, E . C., and Lovett, J. S. (1964). Atlvrrn. i1Io1,phog.3, 33. Cantino, E. C., and Mack, J. P. (1969). Nova Hedicigiri 18, 115. Cantino, E. C., and Truesdell, L. C. (1970). Mycologia 62, 548. Cantino, E. C., and Truesdell, L. C. (1971). Trans. Brit. Mycol. Sac. 57, 169. Cantino, E. C., Lovett, J., and Horenstein, E . A. (1957). Amer. J. Bat. 44, 498. Cantino, E. C., Lovett, J. S., Leak, L. V., and Lythgoe, J. (1963). J. Gen. Microbial. 31, 393. Cantino, E. C., Truesdell, L. C., and Shaw, D. S. (1968). J. Elisha Mitchell Sci. Soc. 84, 125. Cantino, E. C., Suberkropp, K. F., and Truesdell, L. C. (1969). Nova Hedwigia 18, 149. Cotter, D. A., and Raper, K. B. (1968). J. Bacterial. 96, 1680.

44

LOUIS C. T R U E S D E L L A N D EDWARD C. CANTINO

Deering, R. A. (1968). Radiat. Res. 32, 87. Fawcett, D. W. (1966). “An Atlas of Fine Structure.” Saunders, Philadelphia, Pennsylvania. Fletcher, J., and Morton, A. G. (1970). Trans. Brit. Mycol. Sac. 54, 65. Fuller, M. S., and Reichle, R. (1965). Mycologia 57, 946. Fuller, M. S., and Reichle, R. E. (1968). Can. 1.Bat. 46, 279. Garrett, M. K., and Robinson, P. M. (1969). Arch. Mikrobiol. 67,370. Goldstein, A. (1964). “Biostatistics.” Crowell-Collier, New York. Good, N. E., Winget, G. D., Winter, W., Connolly, T. N., Izawa, S., and Singh, M. M. (1966). Biochemistry 5, 467. Grove, S. N., Bracker, C. E., and MorrB, D. J. (1970). Amer. J . Bat. 57, 245. Hickman, C. J., and Ho, H. H. (1966). Annu. R e v . Phytopathol. 4, 195. Horgen, P.A. (1971). J . Bacterial. 106,281. Kavanau, J. L. (1965). “Structure and Function of Biological Membranes,” Vol. 2. Holden-Day, San Francisco, California. Koch, W. J. (1968). Amer. J. Bat. 55, 841. Lesemann, D.E., and Fuchs, W. H. (1970). Arch. Mikrobiol. 71,9. Lessie, P. E., and Lovett, J. S. (1968). Amer. J . Bat. 55, 220. Lovett, J. S. (1963). J . Bacterial. 85, 1235. Lovett, J. S. (1967a). I n “Methods in Developmental Biology” (F. H. Wilt and N. K. Wessells, eds.), p. 341. Crowell-Collier, New York. Lovett, J. S. (196713). N A S A Contract. Rep. NASA CR-673,165. Lovett, J. S. (1968). J . Bacterial. 96,962. McCurdy, H. D., Jr., and Cantino, E. C. (1960). Plant Physiol. 35, 463. Manton, I. (1964). J . Ezp. Bat. 15, 399. Matsumae, A., Myers, R. B., and Cantino, E. C. (1970). J. Gen. Appl. Microbial. 16, 443. Meir, H., and Webster, J. (1954). J . Exp. Bat. 5, 401. Miles, C. A., and Holwill, M. E. J. (1969). J. Exp. Biol. 50,683. Myers, R. B., and Cantino, E. C. (1971). Arch. Mikrobiol. 78, 252. Olson, L. W., and Fuller, M. S. (1968). Arch. Mikrobiol. 62,237. Reichle, R. E., and Fuller, M. S. (1967). Amer. J . Bat. 54, 81. Reinert, J. and Ursprung, H. (eds.) (1971). “Origin and Continuity of Cell Organelles.” Springer-Verlag, Berlin and New York. Schmoyer, I. R., and Lovett, J. S. (1969). J. Bacterial. 100,854. Shaw, D. S., and Cantino, E. C. (1969). J . Gen. Microbial. 59, 369. Soll, D. R. (1970). Ph.D. Thesis, University of Wisconsin, Madison. Soll, D. R., and Sonneborn, D. R. (1969). Develop. Biol. 20, 218. Soll, D. R., and Sonneborn, D. R. (1971). Proc. Nat. Acad. Sci. U S . 68, 459. Soll, D. R., Bromberg, R., and Sonneborn, D. R. (1969). Develop. BioZ.20, 183. Suberkropp, K.F., and Cantino, E. C. (1971). Unpublished data. Sussman, A. S. (1965). I n “Handbuch der Pflanzenphysiologie” (W. Ruhland, ed.), Vol. 15, p. 933. Springer-Verlag, Berlin and New York. Temmink, J. H. M., and Campbell, R. N. (1969%). Can. J. Bat. 47, 227. Temmink, J. H. M., and Campbell, R. N. (196913). Can. J . Bat. 47, 421. Truesdell, L. C., and Cantino, E. C. (1970). Arch. Mikrobiol. 70,378. Turian, G. (1962). Protoplasma 54, 323. Van Etten, J. L. (1969). Phytopathology 59, 1060.

CHAPTER 2

STEPS OF REALIZATION OF GENETIC INFORMATION IN EARLY DEVELOPMENT A. A. Neyfalch INSTITUTE OF DEVELOPMENTAL BIOLOGY USSR ACADEMY OF SCIENCES, MOSCOW, USSR

I. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11. Transcription . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A. Changes in the Quantity of Templates. .................... ........... B. Onset of RNA Synthesis in Nucl ’ C. Rate of RNA Synthesis in Early

45 46 46

................... IV. Translation. .

asm. .........................

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

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

54 61 61 62

thesis in Early Development. .... C. Dependence of Protein D. Nonnuclear Control at V. Regulation of Enzymatic Activity VI. Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74 References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75

1. Introduction

One of the most important problems pertaining to the mechanisms of development is to find out which events take place between the time a gene comes into operation and its phenotypic expression. I n bacteria, the entire sequence of genetically regulated events is chiefly controlled a t the transcriptional level. I n eukaryotes, other points of control are the transfer of mRNA from the nucleus to the cytoplasm and the events preceding translation. These problems acquire special importance in embryonic development. Indeed, i t is evident that the regulation of the rate of each one of these processes may play an important role in differentiation. Although the mechanisms of regulation are genetically controlled, the regulation of many developmental processes is controlled by the genotype via some cytoplasmic organization. The true picture seems to be still more complicated. The mRNA molecules which have been transcribed in the oocyte at the lampbrush stage are partially translated during oogenesis, partially during cleavage, 45

46

A. A. NEYFAKH

and partially during later stages of development. This may be true of similar mRNA molecules, i.e., synthesized on one and the same gene, as well as of different mRNA molecules that are translated at different stages of development. And vice versa, a t any one moment of development, mRNA molecules transcribed during oogenesis, as well as those synthesized in the nuclei of the embryo a t a recent previous stage of the embryogenesis, are translated a t the same time. An attempt is made herein to analyze these problems during early stages of development. But similar situations may be observed at later differentiation stages, when the function of the nuclei has terminated earlier than morphogenesis. Erythropoiesis may be a good example of such differentiation. One may suggest that a cytoplasmic program of realization of gene expression is a general property of all the processes of differentiation. II. Transcription A. CHANGES IN

THE

QUANTITY OF TEMPLATES

In the course of oogenesis and early development, the quantity of DNA which may serve as templates for transcription increases. I n the early oogenesis of amphibians and many other animals, amplification of ribosomal genes occurs, which results in a some 1000-fold increase of the templates for rRNA synthesis. After this increase, the quantity of ribosomal nonchromosoinal DNA in the oocyte nuclei is twice as great as of chromosomal DNA (Brown and Dawid, 1968; Evans and Bernstiel, 1968). I n the course of development of the oocyte, the quantity of mitochondria containing DNA also increases. If in the small eggs of sea urchin the amount of this DNA is as low as 6 pg (Pik6 et al., 1967), in amphibian eggs it is 1000-fold greater (Dawid, 1965, 19661, and in birds it is a million times greater, than the amount of DNA in the haploid set of chromosomes (Shmcrling, 1965). When development starts, chromosomal DNA doubles at each division and increases in proportion with the increase in the number of cells. At the same time, the number of mitochondria (Abramova et al., 1966) and, respectively, the mitochondrial DNA, does not practically change. Therefore, the proportion of cytoplasmic DNA diminishes, and when, for example, an amphibian embryo approaches the gastrula stage (40,000 cells), the mitochondrial DNA in every cell and, correspondingly, in the whole embryo docs not exceed 1-2%, i.e., the value for the adult tissues.' As the mass of the cmbryo practically does not change in early devel-

2.

STEPS OF REALIZATION O F GENETIC INFORMATION

47

opment, the volume of cytoplasm per nucleus rapidly decreases as cleavage proceeds. This means that, assuming a constant rate of transcription, the capacity to provide the protein-synthesizing machinery with new mRNA molecules increases. Hence, the choice of units for evaluating the intensity of the RNA synthesis depends on whether one is interested in the absolute rate, in which case DNA should be taken as a reference; on the other hand, if the nuclear-cytoplasmic interrelations are the issue in question (i.e., to what degree synthesis of RNA provides for synthesis of new proteins), then the mass of cytoplasm, or the whole embryo, is a more appropriate reference.

B. ONSETOF RNA SYNTHESIS IN NUCLEI As early as 1959 a method was worked out whereby the onset of nuclear activity controlling morphogenesis (i.e., the so-called morphogenetic function of the nuclei) could be demonstrated (Neyfakh, 1959). At the time the experiments were started, available data on the arrest of development in lethal hybrids, in mutants, and in enucleated embryos suggested that in the amphibian and fish embryo nuclear function begins to manifest itself at the time of gastrulation, and in the sea urchin a t the mid-blastula (before hatching) stage. The method employed in our work was to inactivate the nuclei, either chemically or by means of heavy X-irradiation a t successive stages of development, and t o see to what stage such functionally enucleated embryos were able to develop. If the embryos irradiated a t two successive steps of development ceased to develop a t the same stage, one could conclude that between the two steps nuclei were exerting no morphogenetic function. On the other hand, if the development of the embryo irradiated later was arrested a t a later stage, then this was taken as evidence of nuclear morphogenetic function. The function is the stronger, the higher the ratio of time-gap between arrested stages to that between irradiated stages (Neyfakh and Rott, 1968). As a result of the application of this method, it was shown that the morphogenetic function of the nucleus begins in the fish embryo at the mid-blastula stage (6 hours a t 2 l o C ) , in amphibians a t the late mid-blastula (stage 8-9), and in the sea urchin a t early blastula (approximately 128 cells) (Neyfakh, 1964). An obvious suggestion is that the onset of the nuclear morphogenetic function might coincide with the initiation of mRNA synthesis. This assumption was admittedly not cntirely justified, as we know very little about the functional role of different mRNA’s or, specifically, whether RNA’s synthesized at any one stage serve a morphogenetic, or any other, role in the embryonic cells. Yet, when methods of determining the mRNA

48

A. A. NEYFAKH

synthesis became available, it was shown that in the loach (Misgurnus fossilis) (Kafiani and Timofeeva, 1964; Kafiani e t al., 1969) and also in the frog (Bachvarova and Davidson, 1966a,b) there is good correlation between initiation of RNA synthesis and the stages at which morphogenetic function of the nuclei had been predicted to begin. Later, a similar coincidence was revealed in the sea urchin embryo (Timofeeva et al., 1968). In recent years, however, evidence has been obtained in the trout (Donzova e t al., 1970a) and in the axolotl that synthesis of mRNA begins some time before the morphogenetic function of the nuclei. One possibility is that these mRNA’s are meant not for a morphogenetic function, but for the synthesis of, for example, nuclear proteins. I n fact, at all stages of development, including the ones before cleavage, a low level of RNA synthesis can be detected. It might be argued that a t the early stages of development the very low level of synthesis is due to the small number of nuclei, and that the increasing synthesis a t the mid-blastula stage (in loach) is related to the progressively increasing number of nuclei. However, there are facts that contradict this suggestion: 1. The kinetics of the increase of incorporation of radioactive precursors does not correlate with the increase in the number of nuclei. Between hour 2 and hour 5 of development of the loach, when the number of cells increases about 200-fold, the rate of incorporation increases not more than 2-3 times. On the contrary, between hour 6.5 and hour 8.5, the number of cells in the embryo increases less than 4 times, but the rate of synthesis increases more than 10-fold. Thus, if in the early stages of development of the loach the rate of incorporation into RNA is calculated per nucleus, the erroneous conclusion may be drawn that within the first 6 hours the rate of synthesis drops drastically (Kafiani e t al., 1969). 2. Autoradiographic studies of amphibians, fish, and echinoderms do not reveal any significant ~ r i d i n e - ~ H incorporation during the early stages of cleavage. I n the sea urchin, RNA synthesis reveals itself only at the 16-cell stage (Czihak, 1965); in the loach, a t the mid-blastula stage (Kostomarova and Rott, 1969) ; in amphibians, at the late blastula stage (Bachvarova and Davidson, 1966a,b). At these stages incorporation increases rapidly from very low to rather high values. 3. Incorporation of the precursors into RNA of the embryos functionally enucleated by irradiation of the gametes prior to fertilization or right afterward, or incorporation in the cytoplasm of the egg, makes up a considerable proportion of the total embryonic incorporation (Kafiani e t al., 1969).

2.

STEPS OF REALIZATION OF GENETIC INFORMATION

49

The above observations suggest that, at least i n fish and amphibians at the cleavage stages, nuclear RNA synthesis, if i t exists, is not very pronounced. This agrees well with the fact that a t these stages the GI and G, phases are practically absent and the S phase is extremely short (Rott, 1970), i.e., DNA synthesis proceeds at a high rate. But the most recent data by Timofeeva et al. (1972) show that in loach a t the early blastula stage (4.5-5 hours of development) there is a relatively weak synthesis of tRNA precursors. The question of the beginning of synthesis of RNA in the sea urchin embryo is more controversial. For example, it was shown by us that in Strongylocentrotus nudus intensification of RNA synthesis occurs after the onset of the fifth cleavage, i.e., approximately at the onset of morphogenetic function of the nucleus. Our autoradiographic investigations did not reveal considerable RNA synthesis in the nucleus of the early embryo, although one would have expected very intensive incorporation in the small number of nuclei that the embryos possess a t these early stages. GliHin and GliHin (1964) showed that during the early cleavage stages only terminal tRNA exchange occurs whereas a t the 32-cell stage true RNA synthesis may be observed. Czihak (1965) by an autoradiographic method was able to reveal RNA synthesis in the micromeres of the 16-cell embryo. Later there were reports of RNA synthesis a t still earlier stages : a t the four-blastomere stage (Slater and Spiegelman, 1970; Nemer, 1967); a t the two-blastomere stage (Kedes and Gross, 1969); and even prior to the first cleavage (Wilt, 1964; Rinaldi and Monroy, 1969) . Recently, synthesis of heterogeneous, nonribosomal RNA was detected in anucleated fragments of activated eggs (Chamberlain, 1970). A study of the hybridization properties of this RNA revealed it to be complementary to the mitochondrial, but not to the nuclear, DNA (Craig, 1970). The rate of synthesis in these fragments proved to be similar to that in the intact early embryos, suggesting that the early RNA synthesis may be cytoplasmic. Hybridization experiments also indicate that a t later stages, when the nuclei are known definitely to synthesize RNA, mitochondrial synthesis makes up a major part of the total RNA synthesis (Hartman and Comb, 1969). According t o Wilt (19701, in the sea urchin synthesis of nuclear RNA becomes detectable a t the 16-blastomere stage. I n conclusion, much has still to be learned about the timing of the early RNA synthesis in sea urchin nuclei, which, in fact, may differ by several cleavages in various species. The possibility that there may be differences among the blastomeres in the different territories of the cleaving embryo has been suggested by Czihak (1965) ; this, however, has not been confirmed (Hynes and Gross, 1970).

50

A. A. N E Y F A K H

C. RATEOF RNA SYNTHESISIN EARLY DEVELOPMENT The increase in the rate of RNA synthesis in the early development of the embryo is a sum of three components: change in the rate of synthesis in the nuclei (this may result from an increase in the rate of transcription and/or of the number of the genes transcribed) ; increase in the total number of nuclei in the embryo; and, finally, an increase of the proportion of RNA-synthesizing nuclei. It was shown by autoradiography that, if the isolated loach blastoderm at the mid-blastula stage is incubated with ~ r i d i n e - ~ H incorpora, tion can be detected only in the cells of the basal layer, which is normally adjacent to the yolk. At the later stages, more distant layers become involved in the synthesis; and a t the ninth hour (late blastula), many TABLE I

PERCENT OF SYNTHESIZINQ CELLSIN EARLY DEVELOPMENT IN INTACT BLASTODERM A N D I N DISSOCIATED CELLS ~

~~

Stages of development (hours)

Intact blastoderm

After dissociation of blastoderm in separate cells

Mid-blastula 7 9 Late blastula Early gastrula 10

14 21 29

39 48 57

more blastoderm cells are actively incorporating (Kostomarova and Rott, 1969). The increase in the percentage of the synthesizing nuclei is shown in Table I. That these regional differences do not depend on the rate of penetration of the precursor is demonstrated by experiments in which the embryos were cut tangentially and the animal and the vegetal zone were incubated separately. Under these conditions, the earlier beginning of the RNA synthesis in the vegetal fragment was still retained. It was shown by biochemical methods that a t the seventh hour the rate of RNA synthesis in the vegetal fragments was severalfold higher than in the animal fragments (per milligram of protein). Finally, if the cells were dissociated and incubated with ~ r i d i n e - ~ H the , percentage of incorporating cells increased with the age of the blastoderm a t the time of dissociation (Table I ) . However, in this case the percentage of synthesizing cells is larger than when the whole blastoderm is incubat,ed; hence it is likely that dissociation removes some of the regional differences. Besides the differences in RNA synthesis, i t has been shown that the basal cells contain more ribosomes (Kostomarova and Nechaeva, 1970)

2.

STEPS O F REALIZATION O F GENETIC INFORMATION

51

and display a higher rate of protein synthesis (Kostomarova and Burakova, 1971). Regional differences in thc RNA synthesis were also revealed in amphibians (Bachvarova and Davidson, 1966b) and in sea urchin embryos (Markman, 1961). They are likely to reflect the oooplasmic segregation which occurs in oogenesis prior to the first cleavage. I n teloblastic eggs, such as thosc of loach, the differences may depend also on the relationships with the yolk which lies underneath. Knowing the number of cells a t different stages, the change in the number of the synthesizing cells, and the rate of synthesis in the whole

Total RNA synthesis in embryo

a"

30 20 50 00

50

50 -x

I

2

4

8

6

40

42

Hours

FIG.1. Constituents of total RNA synthesis. 0-0, Total RNA synthesis per - -, Cell number; x - x, percentage of RNA-synthesizing cells; embryo ; 0 . . .,. RNA synthesis per cell.

-- -

-

embryo, one can evaluate the change of genome activity in development. The relevant data are shown in Fig. 1. For example, within the short period of time between the seventh and ninth hour of development, a 6- to 9-fold increase in RNA synthesis was observed in the loach. (The accuracy of this value depends upon thc correct estimation of the cytoplasmic synthesis of RNA.) Within the same time, the number of cells in the embryo increases from 1700 to 5900 (Rott and Sheveleva, 1968), i.e., 3.5-foldJ and the number of synthesizing cells increases from 14% to 21%, i.e., 1.5-fold. This means that the synthesizing activity per nucleus increases only 1.4- to 1.7-fold. Similar values are obtained by the autoradiographic method, which is more direct but less accurate.

52

A. A. NEYFAKH

After the tenth hour almost all the blastoderm cells are involved in RNA synthesis and the increase in total synthesis becomes proportional to the increase in the number of cells in the embryo. This means that the synthesizing activity per genome does not increase. I n amphibians RNA synthesis sharply increases between stages 8 and 9; in this time interval, the rate of increase in the number of cells is lower than that of the RNA synthetic activity. On t.he other hand, between stages 9 and 10 the increase in the rate of synthesis becomes equal to the rate of the increase in the number of cells; i.e., the rate of transcription per nucleus remains constant (Davidson, 1969). We have shown that in Strongylocentrotus nudus between the early blast,ula (128 cells) and the mid-blastula (about 500 cells) stages the total synthesis of high-molecular RNA increases by 5- to 6-fold; hence the rate of synthesis per nucleus increases by 1.5-fold (Timofeeva et al., 1968). The data of Wilt (1970) are consistent also with an increase of the synthetic activity per nucleus. I n the following hours, the increase of RNA synthesis in the embryo can be accounted for by the increase in the number of nuclei; in fact, a t this time the synthesis of RNA per nucleus actually decreases (Kijima and Wilt, 1969). The experiments carried out in our laboratory showed that in fish a t the gastrula stage, and in the sea urchin a t the stages following hatching, the overall RNA synthesis temporarily decreases (Krigsgaber et al., 1968). Since the number of cells is known to increase a t this time, the observation implies quite a sharp decrease in transcriptional activity in nuclei. No such decrease for the same species was reported by other authors; this may be due to the fact that the stages compared were quite far apart. However, no final conclusion as to whether or not the observed decrease is a real one can be drawn, short of information about changes of the pool of uridine triphosphate andlor permeability changes. Thus, in fish and amphibians, and probably in sea urchin, the period of very low RNA synthesis during cleavage is followed by a relatively short period when the intensity of synthesis per nucleus somewhat increases (not more than 2-fold) ; then the rate of synthesis per nucleus either remains constant or decreases. However, since the cell size decreases, if reference is made to unit of cytoplasm mass, then the RNA is found actually to increase; i.e., there is an increase of the transcriptional efficiency.

D. REGULATION We shall consider only one example of regulation of gene activity in development, i.e., the onset of the RNA synthesis a t the blastula

2.

STEPS OF REALIZATION OF GENETIC INFORMATION

53

stage in loach embryos. The specific questi’on concerns the factor(s) responsible for the simultaneous switching-on of RNA synthesis on many genes within the very short time of 0.5 hour. At the onset of synthesis is strictly related to a certain stage of development, the search should be restricted t o the change in the course of development and may serve as a “clock” for the activation of the genes. First of all, it is interesting to find out whether the clock operates in the embryo as a whole or whether it functions independently in every cell. To answer this question RNA synthesis in the whole embryo and in the dissociated cells was compared. In sea urchin embryos it was shown that dissociation a t the late blastula stage does not prevent the synthesis of rRNA from being switched on (Giudice and Mutolo, 1967). We have shown that, in early embryos of sea urchin and loach, intensification of mRNA synthesis may occur in the dissociated cells as well (Donzova and Neyfakh, 1969; Donzova et al., 1970b). Kafiani e t al. (1971) compared the rate of RNA synthesis in isolated loach nuclei a t the early blastula stage (prior to the intensification of the synthesis in the embryo) and a t the late blastula stage (8 hours) when the synthesis is intensive enough. The intensity of RNA synthesis in the isolated nuclei was found to display the same differences as the RNA synthesis by the whole cells. These differences could be due either to changes of the chromatin itself or of the enzyme. I n the presence of an excess of exogenous RNA-polymerase from Escherichia coli, RNA synthesis sharply increases and reaches the same level in the nuclei obtained from the early and late blastula stage. From this result the authors concluded that the intensification of the RNA synthesis in the late blastula depends on the greater activity of the RNA-polymerase in the nuclei. The degree of specificity of E . coli RNA-polymerase for the loach chromatin should, however, be verified. It should be mentioned in this context that the injection of the u-subunit of the E. coli polymerase into oocytes or early embryos of Xenopus drastically changes the pattern of RNA synthesis (Crippa and Tocchini-Valentini, 1970). Kafiani has also attempted to show that the major gene-activating factor is the change in the ionic composition of the loach embryos and, in particular, the change in the K : N a ratio (Beritashvili e t al., 1969). It is known that, by changing this ratio, it is possible to cause the appearance of new puffs in the polythene chromosomes (Kroeger and Lezzi, 1966). Indeed, as the development of the loach embryos proceeds, the Na+ concentration somewhat decreases while that of K+ considerably increases. There is no evidence so far that alterations in the experimental conditions, resulting in anomalous values of ion ratio, can change the time of the onset of RNA synthesis.

54

A. A. NEYFAKH

Shiokawa and Yamana (1967) reported having found a low molecular factor in amphibian eggs, which suppresses the rRNA synthesis from the early stages. These data were not confirmed in later investigations, although it is quite evident (and is also clear from the experiments on transplantation of nuclei) that the mechanism of the switch-on of RNA synthesis should involve some cytoplasmic factor. For example, Crippa (1970) has succeeded in isolating from Xenopus oocytes a protein factor which specifically suppresses rRNA synthesis by interacting with the rRNA cistrons. Furthermore, from the data cited above (Crippa and Tocchini-Valentini, 1970) i t follows that the regulation of the RNA synthesis depends on the RNA-polymerase. However, it is unlikely that this is the clock mechanism on which the onset of RNA synthesis in development depends. I n the course of clcavage, the size of cells progressively decreases while the amount of DNA in every nucleus remains constant. This means that the ratio between the quantity of DNA and the mass of cytoplasm in the cells changes in the course of the early development: could this be a signal for switching-on the genes at a certain stage of development? This question has received a positive answer in experiments carried out in our laboratory in which the onset of the morphogenetic function of the nucleus was compared with DNA synthesis in haploid and diploid loach embryos. The experiments showed that in the haploid embryos these processes begin one division later than in diploids (Rott and Kostomarova, 1970). An additional division of haploids having only half of the DNA causes a 2-fold decrease in cell sizcs; as a result, the nucleocytoplasmic ratio becomes equal to that of diploids a t the time of the onset of RNA synthesis. This correlation may be considered to be indirect experimental evidence in favor of the above hypothesis. 111. Transport of RNA to the Cytoplasm

The transfer of RNA to the cytoplasm has two most important peculiarities: it is preceded by a partial degradation of large RNA molecules in the nuclei and it proceeds in time. The known example of the changes RNA undergoes in the nuclei is the processing of the high-molecular precursors of ribosomal RNA and precursors of mRNA carrying information about hemoglobin in erythroblasts. That only the RNA molecules which have undergone processing pass to the cytoplasm and that the time of departure is comparable to, or longer than, that of the processing, indicates that these two phenomena are related; i.e., the rate of the RNA transfer from the nucleus seems to be determined by the rate of its maturation in the nucleus.

2.

STEPS OF REALIZATION OF GENETIC INFORMATION

55

It is still a difficult task to give an accurate kinetic description of the process of RNA transfer from the nucleus to the cytoplasm. Four processes which are very difficult to differentiate occur simultaneously in the cell: synthesis of RNA, degradation of its major part in the nucleus, transfer of the remaining RNA to the cytoplasm, and subsequent degradation of this portion in the cytoplasm. The rates of all four processes are different for different kinds of RNA. For example, in the case of rRNA the rate of 18 S rRNA transfer to the cytoplmm is higher than that of 28 S rRNA. It is highly probable that for different kinds of mRNA these rates are also different. The process of RNA transfer from the nucleus still cannot be adequately described by a system of differential equations, as we do not know what model to choose to describe this mechanism. A concentration dependence, when the rate of RNA transport to the cytoplasm is proportional to its concentration in the nucleus would be the simplest decision. At the other extreme would be the assumption that the time of transfer of every kind of molecule from the nucleus is predestined. Differences observed in the processing of RNA molecules of different kinds (rRNA and mRNA of hemoglobin) and in the fraction of RNA released to the cytoplasm in different tissues may be interpreted to mean that, in the course of embryonic development, the character of RNA transfer should be changing. Experimental evidence supporting this suggestion is not plentiful. On the one hand, some authors showed that in sea urchin embryos some portion of the label is found on the polysomes shortly after the beginning of synthesis (Kedes and Gross, 1969). At the same time a substantial portion of the newly synthesized RNA remains in the nucleus for quite a long time (at least 2 hours). After gastrulation, the labeled RNA leaves the nucleus much faster (Kijima and Wilt, 1969). Direct measurements of the RNA in the cytoplasm show that in the early stages of development of sea urchin this process is rather slow (Aronson and Wilt, 1969), but as development proceeds the rate of release increases 10- to 15-fold (Singh, 1968). Thus, in the course of development, the fraction of RNA undergoing degradation, the time the departing molecules are held in the nuclei, and their rate of transfer change. We studied RNA transfer from the nucleus to the cytoplasm in the loach embryo by autoradiographic and biochemical methods. However, an estimate of the rate of RNA transfer to the cytoplasm proved to be difficult, as chase experiments could not be done owing to the large pool of radioactive precursors that accumulated in the embryonic cells within a short time. Incorporation of the label in RNA gradually de-

56

A. A. NEYFAKH

creases, the deceleration being difficult to estimate accurately because the rate of RNA synthesis rapidly changes in ithe course of development. RNA synthesis in the nuclei of the loach embryo begins a t the midblastula stage (6 hours at 21OC). Autoradiographic data show that no transfer of this RNA to the cytoplasm occurs during the first hours (Table 11).It is likely that part of this RNA undergoes degradation within the nuclei while part accumulates, to be transferred to the cytoplasm in the course of the following hours of development. At the later stages (early gastrula, 10 hours), when transfer has started, only light RNA (2-10 S) leaves the nuclei while the heavier fractions (more than 30 S) remain in the nuclei (Rachkus et al., 1971). TABLE I1 TRANSITION OF RNA

FROM

OF

NUCLEITO CYTOPLASM AT DIFFERENT STAQES EARLYDEVELOPMENT Number of grains

Stages (hours) Mid-blastula 7 8 Late blastula Late blastula 9 Early gastrula 10

Nucleus 71 86 108 115

f3 f2 f4 f4

Cytoplasm 1 2 19 f 1 20 f 3

Whole cell 72 88 127 135

f3 f2 ?c 4

f5

Percent of transition 1 6 15 15

Another difficulty encountered in the quantitative autoradiographio estimation of the rate of RNA transfer from the nuclei to the cytoplasm in early development is that in the course of the experiment the cells divide one or several times. The result,ing distribution of the label between the daughter cells, and the decrease in the size of the cells and their nuclei, require the introduction of some corrections that diminish the accuracy of measurement. Nevertheless, some quantitative conclusions can be drawn. Figure 2 shows the results of an experiment in which, after a 2-hour incubation with uridineJH, the embryos were transferred to the “cold” medium; the synthesis of the labeled RNA continued although its rate decreased 4- to 5-fold. Two hours after the beginning of the “chase,” the rate of RNA synthesis and the rate of RNA transport from the nuclei become similar, and this is reflected in the fact that the quantity of labeled RNA in the nucleus is almost equal throughout the several hours that follow. At the same time, the cytoplasmic label rapidly grows.

2.

STEPS OF REALIZATION OF GENETIC INFORMATION

57

Assuming that RNA in the cytoplasm does not undergo degradation during the experiment, one can calculate that within the first 2 hours of synthesis about 30% of the total labeled RNA goes to the cytoplasm, although some part of it (about 5-10%) does so as late as 6 hours after being synthesized. Without going into the details of this calculation, it is clear that the labeled RNA synthesized from the remnants of the Chemical ossay

26C

200 Lo

._

30 Cytoplasm

L

/PI,

1oc

2-n

c .> ._ c

z //

Cytoplasm

0

.0

L

10

B

v)

7

9

14

13

Hours of development

15

120 180 240 Incubtion (minutes) 13 14 15 46 47 Development (hours)

( 5 60

FIG. 2. Transition of RNA from nuclei to cytoplasm. Autoradiographic data (left) and biochemical assay (right). Blastoderms of loach embryos were incubated with uridine-'H: for autoradiography, 2 hours a t mid-late blastula; for biochemistry, 15 minutes at early gastrula. Then blastoderms were washed with an excess of cold uridine and chased with it. Silver grains were counted .in autoradiographic preparations over nuclei and over cytoplasm ; the grain count was corrected for volumes of nuclei and cells, taking into account cell divisions during incubation. For biochemical assay, nuclei were sedimented from homogenates by routine procedures, and the specific activity of RNA was determined in nuclear Nuclei; O---O, cytoplasm; X =X , cell and cytoplasmic fractions. u, (nucleus plus cytoplasm).

radioactive pool of the precursors during the later hours is insuEficient to account for the increase in the quantity of radioactive RNA in the cytoplasm if the major portion of labeled RNA synthesized earlier has not contributed to it. If part of RNA in the cytoplasm undergoes degradation, the part of RNA which leaves the nuclei late will be still greater. The late transport of some part of RNA may be explained by two models. The first, a stochastic model, suggests that the transport of

58

A. A. NEYFAKH

RNA depends only on its concentration within the nucleus. I n this case, the rate of transport of labeled RNA after pulse labeling should decrease asymptotically. If some portion of RNA does not go out of the nucleus or does so very slowly, one has a situation resembling the experimental evidence: some decrease in the rate of transfer and some decrease in the content of the labeled RNA in the nucleus. An alternative model, a regulated one, suggests that the RNA synthesized is gradually released to the cytoplasm. If many kinds of RNA have been synthesized in the nucleus, and they have different rates of transfer, the release of the label will be asymptotic, or close to it. A more elaborate control may be that different kinds of RNA go to the cytoplasm one after another, thereby determining the sequence of translation of various proteins. Thus, according to the data cited above, in loach the RNA synthesized in the mid-late blastula goes to the cytoplasm a t various stages of gastrulation and the initial stages of formation of the axial organs. There are a few facts supporting the second model: for example, absence of the RNA transfer a t the early stages of synthesis, transfer only of the RNA of the 2 S to 10 S class a t the early stages of development, the slow dynamics of transition, and other indirect data. Evidence that different RNA’s synthesized simultaneously, but transferred early and late, carry different information would be unequivocal proof of this model. Passive transfer of RNA to the cytoplasm requires only restricted permeability of the nuclear membrane. For active, controlled transfer, a mechanism of higher complexity is needed, which should particularly include a binding, chemical affinity of nuclear RNA to the nuclear structures-chromatin or the membrane. The existence of the binding is confirmed by the possibility of thermal fractionation of RNA in the course of phenol extraction (Georgiev and Mantieva, 1962). The RNA fractions obtained in accordance with Georgiev’s method a t low and high temperatures agree with the concept of the nuclear and cytoplasmic RNA. In the nuclei there is a small amount of readily extracted RNA that might correspond to the molecules going out of the nucleus. Our investigations on RNA migration during mitosis (Neyfakh and Kostomarova, 1971; Neyfakh e t al., 1971) also testify to the existence of binding between RNA and the nuclear structures. I n these experiments loach blastoderms were pulsed with ~ r i d i n e - ~ H and then chased with “cold” uridine in the presence of high concentrations of actinomycin D. Within an hour almost all the cells underwent mitotic division. I n the dividing cells the label was dispersed over the whole cell whereas in the interphase cells it was concentrated in the nuclei. Hence, the conclusion is that during mitosis the newly synthesized RNA goes out

2.

STEPS OF REALIZATION OF GENETIC INFORMATION

59

of the nucleus and afterward it returns to the nuclei of the daughter cells. Some experiments were made with the culture of fibroblasts from Chinese hamster, in which RNA transfer from the nuclei occurs relatively quickly. The monolayers of fibroblasts growing on the glass were incubated with ~ r i d i n e - ~for H 40 minutes, the last 30 minutes in the presence of colcemid. Then the round metaphase cells were treated with trypsin for a short time and then shaken off the glass (Stubblefield et al., 1967). I n the cells synchronized in this way, the metaphase cells amounted to 90%. A portion of the metaphase cells was fixed immediately, and another portion was placed in the medium without colcemid but with actinomycin. Under such conditions cell division could continue, but de n o w RNA synthesis was inhibited. For control, the interphase cells were used which were taken from the glass after the metaphase cells had been removed and transferred to the medium with actinomycin. During mitosis, almost all the label is evenly distributed in the cell; its concentration on the chromosomes is somewhat greater than in the cytoplasm (Fig. 3 ) . On the other hand, in the interphase cells only 18% of the labeled RNA is found in the cytoplasm, although those cells that have not undergone mitosis have more time for synthesis (40 minutes). I n 1 hour, when all the cells isolated in the metaphase are about to end mitosis, most of the labeled RNA is again found in the nuclei of the daughter cells, and only 13% of the grains are to be found in the cytoplasm. Most of this is most likely RNA which had gone out of the interphase nuclei either before mitosis or immediately after it. By this time, in the interphase cells which have not undergone division about 28% of the labeled RNA has moved to the cytoplasm (Fig. 3 ) . Thus, in this case too, most of the newly synthesized RNA, if not all of it, goes out of the nuclei for the duration of mitosis and then returns, almost completely, to the nuclei of the daughter cells. Hence, cellular division is rather a retardation than ail enhancement of the normal transfer of RNA to the cytoplasm. The migration of RNA from the cytoplasm to the nucleus a t the end of mitosis requires some explanation. In the work of Goldstein et al. (1969), it was shown that the nucleus of an amoeba labeled in its RNA, when transplanted to a nonlabeled amoeba, loses part of its RNA, which moves to the nucleus of the host cell. Thus, RNA can migrate through two nuclear membranes; since about half of all RNA is found in the nuclei and half in the cytoplasm, the nuclear RNA must reversibly bind to the nuclear structures. If these data are valid also for higher animals, the behavior of RNA

60

A. A. NEYFAKH

in mitosis and interphase may be explained in terms of affinity to the nuclear structures (chromatin seems to be the most likely candidate, although the inner surface of the nuclear membrane cannot be ruled out). During cell division the structure of chromatin changes sharply; Dividing cells

8 Nondividing cells

!oo%

too %

.#&, .,..... .*.

42..

82%

.. ....::. *?:.,;*!

18%

:a:.

.;%;y. J.:,

*.

'

.*.

.

'4;.,

72010@*,*:.;;*I:.

...' . ..*.*.,:

,:; 28%

.

0.

*:

FIa. 3. Distribution of labeled RNA between nuclei and cytoplasm in dividing and nondividing cells (scheme). Fibroblasts of Chinese hamster in monolayer culture were incubated with uridine-'H (40 minutes) and Colcemid (the last 30 minutes). Then the metaphase and interphase cells were shaken from the glass in succession and incubated separately without Colcemid but with actinomycin (2 pg/ ml). Silver grains over nuclei and cytoplasm (in metaphase, over chromosomes and out of them) were counted. Correction was made for self-adsorption and cell area. For details, see Neyfakh et al. (1971).

it becomes spiralized into chromosomes. Its affinity to the nuclear RNA decreases so much that a t last RNA disperses in the cell. When mitosis is over and the chromosomes become despiralized, their ability to bind RNA is restored and the nuclear RNA concentrates in the nucleus. This takes place either before the nuclear membrane is formed after it, but then it is no obstacle to the nuclear RNA, as follows from the experiments with amoeba.

2.

STEPS OF REALIZATION OF GENETIC INFORMATION

61

I n the interphase nucleus the adsorption properties of chromatin do not change, but the properties of the RNA molecules change as a result of processing and the monocistronic fragments of mRNA go to the cytoplasm. Thus, both in mitosis and in interphase, the release of RNA from the nuclei is due to the decreased affinity to chromatin; in mitosis these changes occur in chromatin, and in the interphase in RNA. Making the above suggestion still more speculative, one may assume that the sites of RNA responsible for binding with chromatin are localized in the fragments of this molecule which undergo degradation; and it is the separation of mRNA from these fragments that is the mechanism of RNA departure from the nuclei. We do not consider here the question of the role of proteins that are bound to RNA in the nuclei (Georgiev and Samarina, 1969). These proteins can be responsible not only for RNA cleavage [they have been proved to possess endonuclease activity (Niessing and Sekeris, 1970) 1, but possibly for binding with chromatin. Thus, the study of the behavior of RNA in mitosis may be a handy model for the elucidation of the mechanisms controlling the fate of RNA after it has been synthesized. There are still many things to be clarified about the step (realization of hereditary information) which follows transcription. We are quite sure now only about the existence of processing and also that RNA release from the nucleus could be relatively slow, especially in the embryonic cells. How the process is controlled is still obscure. However, regardless of whether there is an active control over the pattern of RNA release from the nuclei or whether this is merely the outcome of the structure of the nuclei and the properties of RNA, the time of release is extremely important in embryonic development. The fact that mRNA synthesized at the blastula stage and responsible for gastrulation is released from the nuclei throughout this period of morphogenesis is a good illustration. IV. Translation

A. INFORMOSOMES mRNA transferred from the nuclei to the cytoplasm of the embryonic cells becomes incorporated in the informosomes and polysomes ; these differ chemically in the amount of protein they bind, and hence the difference in their buoyant densities. I n CsCl density gradients, polyribosomes band as one peak with a density of 1.51 maximum; informosomes give a peak a t 1.40 (Spirin, 1966). While ribosomes and polyribosomes make up a part of the cell mass and may be detected by optical density, no optically measurable amount of informosomes has

62

A. A. NEYFAKH

been obtained so far, and informosomes are usually detected by the radioactivity of this RNA. Informosomes were found and described in detail first in loach embryos (Spirin et nl., 1964) and then in sea urchin embryos. There have been a few reports about informosomes in tissues of adult organisms. Some properties make informosomes similar to the ribonucleoprotein complexes found in the nuclei, which givcs grounds for believing that the two groups of particles are identical (Georgiev and Samarina, 1969). Even in the early reports it was suggested that informosomes are an intermediate stage between RNA synthcsis in the nucleus and its translation, i.e., that mRNA, either before or upon being released from the nucleus, associates with proteins giving rise to the informosome particles, which then go to polyribosomes where the RNA is translated (Spirin, 1966). Thus, these particles can be instrumental in regulation a t the translational level. It is not surprising, therefore, that descriptions of informosomes arouse great interest. The existence of informosomes has been questioned by several authors, the main objection being the possibility of formation of RNA-protein artificial aggregates during homogenization. However, recent work from Spirin’s laboratory (Ovchinnikov et al., 1969; Spirin, 1969) gives new data supporting the concept of informosomes as real subcellular particles. The question of the role of informosomes in the transfer of mRNA to polyribosomes is still an open one. About 80% of the labeled RNA transferred from the nuclei to the cytoplasm of loach is associated with informosomes, i.e., it is in the area of the 1.4 g/cm3 peak, with only 20% in the polysome peak (1.51 g/cm3). This ratio does not change after prolonged incubation (Spirin, 1969). One may assume that at each moment only 20% of mRNA is translated, and 80% remains in a “masked” form, and informosomes continuously exchange mRNA ; other explanations are also possible. Therefore, until this question is resolved, the role of informosomes remains obscure, although i t is evident that the fact that a considerable part of the mRNA is associated with informosomes should be taken to indicate their great importance in the regulation of translation in the cell.

B. THE INTENSITY OF PROTEIN SYNTHESISIN EARLY DEVELOPMENT In the course of early development, the rate of protein synthesis increases in all animals to reach a maximum a t a stage which is different at the blastula stage in the sea urchin in different organisms-e.g., (Neyfakh and Krigsgaber, 1968), in the loach at the end of gastrulation (Krigsgaber and Neyfakh, 1968). The rate of synthesis decreases in the sea urchin during hatching and in the loach a t the beginning of

2.

STEPS OF REALIZATION OF GENETIC INFORMATION

63

organogenesis. Later, another increase in the rate of protein synthesis takes place. Incorporation of labeled amino acids in protein depends not only on the rate of synthesis, but also on the rate of penetration of amino acids into the cell and on the size of the amino acid pool. Therefore, for an accurate determination of the rate of synthesis, the specific activity of the amino acid pool should be estimated a t every stage investigated, and this makes the study much more complicated. I n the preliminary experiments with loach it was shown that, a t the stages in which an increased incorporation of arginine into protein was observed (at the end of gastrula), both permeability to arginine and the size of the arginine pool increase. No appreciable change in the specific activity of arginine was observed, hence the increased incorporation of labeled arginine in the protein may be interpreted as being due to an increase in the rate of protein synthesis (Kukhanova and Neyfakh, 1968). It is likely that the greater permeability to amino acids of the cells of the loach blastoderm ensures an intensified transport of amino acids from the yolk to the blastoderm, which may result in a larger amino acid pool, which in turn provides the conditions for a higher rate of protein synthesis. A comparison of the curves of protein synthesis a t various stages of development prompts some speculations-e.g., that in the sea urchin an increased protein synthesis a t the blastula stage is necessary for hatching and gastrulation; or that a slowing down of the rate of synthesis prior to gastrulation (in sea urchin) or organogenesis (in loach) is associated with the switching of synthesis from “old” to “new” proteins. I n fact, there are very few data on such “switchover” and, more generally, on the role of the proteins synthesized in the early development. Fractionation of proteins or immunological methods have shown rather minor changes in the patterns of the proteins synthesized. This, however, is not surprising if one considers the very small amounts of proteins that are likely to be synthesized that are beyond the sensitivity of the methods a t present available. These minor proteins may be responsible for morphogenesis. Besides, in the majority of experiments the analyses deal with soluble proteins, whereas differentiation is most likely to be associated with proteins of the cell surface and other structural, rather insoluble, proteins. I n early development, nuclear proteins make up a considerable portion of the newly synthesized proteins. The mass of the nuclei increases many times more rapidly than that of the embryo, hence it is not surprising that in sea urchin and loach nuclear proteins amount to one-fourth to one-half of all the proteins synthesized. This is why increased protein synthesis a t the early stages is to some extent indicative of an increase

64

A. A. NEYFAKH

in the mass of the nuclei, However, this phenomenon cannot account for all the fluctuations in the rate of synthesis. It may seem surprising that the templates for histones, which are not highly specific nuclear proteins, are not entirely provided for during oogenesis, but, rather, are partially transcribed during early development (Kedes et al., 1969). The data of these authors suggest that RNA synthesis and nuclear control of morphogenesis are not necessarily synonymous.

C. DEPENDENCE OF PROTEIN SYNTHESIS ON RNA I n the first experiments of Gross on protein synthesis in actinomycintreated sea urchin eggs, it was shown that early total protein synthesis does not depend on that of RNA (Gross and Cousineau, 1963). However, the recent papers of that laboratory report that a t the early stages actinomycin changes the share of synthesis on small polyribosomes, where histones are synthesized on newly formed templates (Kedes e t al., 1969). Evidently, a decreased level of synthesis on the new templates, entailed by the suppressed RNA synthesis, is compensated for by the higher level of synthesis on mRNA produced during oogenesis. We studied protein synthesis in sea urchin (Neyfakh and Krigsgaber, 1968) ,and loach (Neyfakh e t al., 1968b) embryos either treated with actinomycin or submitted to heavy X-irradiation a t various stages of early development. As shown in Fig. 4, 00th treatments produce a similar effect on protein synthesis, although the mechanisms of action are very different. Actinomycin interferes with RNA synthesis, whereas heavy doses of X-rays partially decrease the level of RNA synthesis but inhibit an increase in the synthesis during development. Heavy doses of radiation seem to impair transcription in such a way that the RNA synthesized after irradiation cannot serve as a template for protein synthesis. It is essential that in both cases the doses used ensured complete inactivation of the genetic function of nuclei (Neyfakh and Krigsgaber, 1968). Actinomycin- (or X-ray-) induced inactivation of sea urchin nuclei during the earliest postfertilization stages of development does not change the pattern of protein synthesis. Thus, the eggs were treated with actinomycin prior t o fertilization (Gross and Cousineau, 1963), and in the first hours after fertilization protein synthesis went on increasing, achieving the early blastula level. This can also be seen in Fig. 4 (curve 2 ) , when actinomycin or X-rays were used 2.5 hours after fertilization (for details, see Neyfakh and Krigsgaber, 1968; Krigsgaber and Neyfakh, 1972). This increase means that protein synthesis a t these stages occurs chiefly on the preexisting templates stored during oogenesis. Thereafter, however, the rate of protein synthesis markedly decreases ;

2.

STEPS OF REALIZATION OF GENETIC INFORMATION

65

this means that a t these stages in normal development the molecules of mRNA being translated are those synthesized in the nuclei of the embryo. Inactivation of nuclei a t all following stages also causes a rapid decrease in protein synthesis which is obviously due to the decay of the short-lived mRNA. Nevertheless, the level of protein synthesis, always lower than that of the control, is maintained for a long time, until the death of the embryos. Protein synthesis seems to continue on the long-lived templates synthesized during oogenesis or in the nuclei 200

Moment of irrad iat ion

Control

/'

Moment of

1O(

400

c

6 eavage Blastula hatching

12 48 Mesenchyme blastula Gastrula

6 Cleawge Blastula hatching

12

18 Mesenchyme blastula Gastrula

FIG.4. Protein synthesis in sea urchin eggs after X-irradiation (left) or actinomycin treatment (right). The embryos were X-irradiated (10 krad) or actinomycin was added (25 pglml) a t moments of development indicated by arrows. Aliquots of embryos at different stages of subsequent development were taken and incubated with "C-labeled amino acids for 1 hour; the specific activity of protein was determined.

of the embryo. Thus, protein synthesis in early development may occur on three kinds of templates: the ones stored during oogenesis, and the short- and long-lived ones synthesized in the embryo. The data on the dynamics of protein synthesis after nuclear inactivation a t different stages has enabled us to estimate the quantity of mRNA and the time of its appearance and disappearance-i.e., the character of genetic control over protein synthesis. Of course, one should not expect complete correlation between the rate of synthesis and the quantity of templates. It is possible, for example, that the translation on the old templates (when the synthesis of the new ones is blocked) will be unnaturally long, and their role in normal development will be erroneously overestimated.

66

A. A. NEYFAKH

Figure 5 shows similar data obtained with early (left) and later (right) loach embryos inactivated with heavy doses of X-rays. One can see that the inactivation of nuclei at any moment during the first 6 hours of development (2lOC) causes the same effect; i.e., protein synthesis slowly increases, achieving the level of the ninth hour of normal development. Morphologically, the irradiated or actinomycin-treated em3000

2500

t

2500

I 0

\

t I ,

3 Cleavage

6

9 1 2 1 5 Blastula Gastrulation

5 9 Blostulo

12

15

18

20

Gastrulation

Hours of development

FIG.5 . Protein synthesis in loach embryos irradiated at different developmental stages. The embryos were X-irradiated (20 krad) at moments indicated by arrows. The blastoderms were isolated from the yolk a t different stages of subsequent development and incubated for 1 hour with "C-labeled amino acids; the specific activity of protein was determined. Rates of protein synthesis after irradiation a t different moments during early development (left) fit one curve.

bryos also stop developing at the ninth-hour stage and then die (Neyfakh, 1964); apparently in the loach the templates providing protein synthesis and development until late blastula are stored during oogenesis, as in the sea urchin. The effects of nuclear inactivation a t subsequent times during development (Fig. 5, right) strongly depend on time of irradiation-the later the embryos were irradiated, the more protein did they have time to synthesize on the templates produced before the moment of inactivation.

2.

STEPS O F REALIZATION O F GENETIC INFORMATION

67

For instance, the difference in protein synthesis between curves 2 and 3 may be due to the appearance of the templates synthesized between the stages indicated by arrows 3 and 2, i.e., within 1 hour between hours 7.5 and 8.5 of development. A comparison of the right and left diagrams (Fig. 5) makes it clear that genetic control over protein synthesis in the loach is not realized until the sixth-hour stage, begins at this stage, and coincides with the onset of RNA synthesis. It is clear from Figs. 4 and 5 that mRNA synthesized within a very short period of time is being translated for a long time during subsequent stages of development. And, vice versa, a t any moment during early development, translation occurs both on newly synthesized templates and on templates formed during oogenesis. Unfortunately, not only is our knowledge about the changing types of proteins synthesized in development confined to a small number of soluble proteins present in high concentrations, but it now appears that the data on mRNA mostly refer to the RNA’s transcribed from repeated sequences. Therefore, we cannot say to what extent the pattern of mRNA synthesis and the pattern of translation predetermine the precise timetable or program of appearance of proteins responsible for differentiation. R N A synthesis and its release into the cytoplasm is the necessary prerequisite for the synthesis of the proteins; but this does not mean that the program for their synthesis depends entirely on the synthesis of RNA. For example, wc do not know to what degree the lifetime of mRNA of different types is predetermined by their structure and to what degree by the ability of cytoplasmic structures to distinguish between different types of mRNA in such a way as to accomplish their selective degradation in the cytoplasm. T o evaluate the role of both mRNA and the cytoplasmic components in the control of translation, one should consider the facts demonstrating independence of some parameters of translation from nuclear control.

D. NONNUCLEAR CONTROL AT

THE

TRANSLATIONAL LEVEL

A classical example of the independent character of protein synthesis and the changes it undergoes in the course of development is the activation of synthesis after fertilization of the sea urchin egg. Without going into the details of the rich literature dealing with this problem, the mechanism of this process can be described as the activation of ribosomes and “unmasking” of mRNA stored during oogenesis (see review by Felicetti et al., 1971). This is, however, not a unique example, as a t the subsequent stages the rate of synthesis has been also proved to be controlled by cytoplasmic

68

A. A. NEYFAKH

mechanisms. This can be seen in the data illustrated. After late nuclear inactivation, the level of protein synthesis is lower as compared to controls, but the patterns of the curves tend to be similar to those of the control. For example, in sea urchin embryos irradiated or treated with actinomycin before hatching, protein synthesis continues to decrease, as in the control, but the rate of decrease is more pronounced (Neyfakh and Krigsgaber, 1968). However, a t the stages corresponding to the mesenchyme blastula and the onset of gastrulation, the rate of protein synthesis increases just as in the control, although the level of the control is not reached, This is due entirely to the cytoplasmic mechanism, as the nuclei in these embryos are completely inactivated (Fig. 4 ) . I n the case of loach (Fig. 5) a similar situation is observed (Neyfakh et al., 1968b). Protein synthesis increases in the first hours of development up to mid-blastula, although the nuclei have not yet begun to synthesize new templates. If the nuclei have been inactivated a t the later stage (Fig. 5, right), protein synthesis still increases although no new templates capable of being translated are formed. This phenomenon is still more significant if irradiation has been carried out a t the end of gastrulation (16 hours). Protein synthesis in the control decreases temporarily and then goes up again. In irradiated embryos with inactivated nuclei, the same effect is observed; i.e., protein synthesis goes down (the process is more rapid than in the control a t the expense of degrading short-lived mRNA) and rises, as in the control. Being a t all times lower than in the control, protein synthesis achieves the level of the moment of irradiation. In other words, in loach and sea urchin, intensity of protein synthesis depends not only on the presence of the RNA templates but also on the cytoplasmic regulatory mechanisms which tend to maintain synthesis a t the level characteristic for the given stage of development. Some information about this process may be obtained from the evidence about the maintenance of the normal level of total protein synthesis of sea urchin embryos treated with actinomycin prior to fertilization. At the stages when part of the synthesized protein is being normally translated on the new templates, the synthesis of these proteins, which occurs mainly in the light polyribosomes, is inhibited in the achinomycintreated embryos; in a compensatory way there is an increase in protein synthesis in large polyribosomes, on the templates stored during oogenesis (Kedes et al., 1969). One may believe that embryonic cells possess a specific regulatory feedback mechanism which maintains in a nonspecific manner the level of protein regardless of what proteins are being produced. It would be simpler to assume that protein is limited by some substance;

2.

69

STEPS O F REALIZATION OF GENETIC INFORMATION

then the absence of some templates would stimulate translation on others. However, if the concentration of this limiting substance changes regularly in the course of development, it means that we are dealing with a cytoplasmic regulatory mechanism at the translational level. Two more examples will demonstrate that in embryonic development the quantitative relationship between RNA and protein synthesis is not very strict. When loach eggs or spermatozoids are irradiated with heavy doses of X-rays before fertilization, one obtains haploid embryos which seem

A 16 (4 -

2000':

RNA

1500 -

12 10 10oo -

86 -

500

4-

-

2 -

-

I

5

6

7

0

I

,

9 6 Hours of development

L

9

12

45

FIG.6. RNA and protein synthesis in diploid and haploid loach embryos. Haploid embryos werc obtained by fertilizing the eggs with sperm irradiated with heavy doses of X-rays (50 krad). RNA and protein synthesis was determined after 1 hour of incubation with uridinc-'H or I4C-labeled amino acids and expressed as specific activity (RNA, cpm/mg RNA x loz; protein, cpm/mg protein).

to develop normally until late organogenesis and often form abnormal but moving larvae. At the mid-late blastula stage, the haploid embryos synthesize half as much RNA (per nucleus) as the diploid embryos (Fig. 6 ) . At later stages in haploid embryos, a compensatory increase of the number of cells occurs and RNA synthesis in the embryos approaches that of the diploid controls (Kafiani et al., 1968). Yet in these embryos protein synthesis is a t all times maintained a t a normal level; i.e., it does not differ from the diploid controls (Neyfakh et al., 1968b). Apparently, the 2-fold decrease in the quantity of the templates formed at the blastula stage is compensated for by their more efficient translation.

70

A . A. NEYFAKH

A different situation is observed in nucleocytoplasmic haploid loach x goldfish (Carassius auratus) hybrids. If irradiated loach eggs are fertilized with goldfish sperm, they develop only to the late blastula stage, i.e., in the same way as the anucleated embryos obtained by irradiation of both gametes. On the other hand, if nonirradiated loach eggs are fertilized with goldfish sperm, the result will be well-developed

t

2400 -

Protein

2100 1800

haploids

-

1500 E 1200 .(u c

a e 900

Haploid hybrids

-

600 300

7

9

11

13

15

*

Hours of development

FIG.7. RNA and protein synthesis in haploid loach embryos, haploid androgenetic (nucleocytoplasmic) hybrids of loach x goldfish, and “anucleated” loach embryos. Irradiated loach eggs (20 krad) were fertilized with normal sperm of loach, with sperm of goldfish, or with irradiated (50 krad) loach sperm. Assav of RNA and protein synthesis as in Fig. 6.

hybrids possessing the traits of both parents. Hence, the goldfish nucleus in the irradiated loach cytoplasm cannot support development (Neyfakh and Radeievskaya, 1967). And yet these embryos have a normal level of RNA synthesis which is just as intensive as in loach haploids (Neyfakh et al., 1968a). Unlike the case of loach haploids, protein synthesis in nucleocytoplasmic hybrids proceeds in a way similar to that in anucleated embryos, i.e., slowly reaches the 9-hour stage (late blastula) and then ceases to grow (Fig. 7 ) . Thus, it seems that, in the absence of new loach RNA, goldfish templates cannot be translated in the loach cytoplasm. This experiment demonstrates that the presence of mRNA

2. STEPS

OF REALIZATION OF GENETIC INFORMATION

71

is not a sufficient condition to allow normal translation to take place. If this is so, translation control is, in turn, under some kind of nuclear control. Thus, in embryogenesis translation is also a point of regulation in the pathway of realization of genetic information. Nuclear control might be exerted via the composition and quantity of mRNA on both rate and types of proteins synthesized. However, the fact that half the amount of mRNA of a haploid embryo may be translated a t the same rate as in a diploid one, should be taken to mean that synthesis normally occurs not a t its highest possible rate. Only degradation of a major fraction of short-lived mRNA, as is the case after actinomycin treatment, may cause a decrease in the rate of protein synthesis. The remaining long-lived mRNA continues to be translated at a submaximum rate, and a t a certain stage of development synthesis on the same templates may increase considerably. One of the mechanisms involved in the regulation of the rate of protcin synthesis is the change in the amount of ribosomes taking part in the translation, i.e., the quantity of polyribosomes. This is the case in the early development of sea urchin, but i t is quite possible that there are other mechanisms and that the rate of translation itself may be regulated. The lifetime of mRNA is of importance for the character of protein synthesis. All we know is that some mRNA’s function in the cell for a short time, and others much longer. The differences between long-lived and short-lived mRNA have not been elucidated, but they seem to be inherent in the structure of RNA, i.e., determined by nucleus. Finally, direct control over the composition of the proteins synthesized may consist in the selection of the types of mRNA being translated. There is a t least one example of such regulation. In the course of spermatogenesis, RNA synthesis is arrested until meiotic divisions (in Drosophila) (Henning, 1968) or right after them (mouse) (Monesi, 1965). However, it is only after this that differentiation of the spermatozoid begins. Evidently it occurs without direct nuclear control, a t the expense of mRNA synthesized before meiosis (Hess, 1967). One possibility is that during the entire period of spermiogenesis one and the same set of proteins is being synthesized, although i t may seem unlikely that the entire sequence of all these complex events is determined by the same proteins. However, it was shown by cytochemical and autoradiographic investigations that in the maturing sperrnatozoids of Drosophila the lysine-rich histones are replaced by the arginine-rich histones (Das et al., 1964). This process is sensitive to puromycin; i.e., i t is a true de novo protein synthesis. No such synthesis was observed throughout

72

A. A. NEYFAKH

spermiogenesis ; con8equently1 the translation of the mRNA coding for the specific proteins of the spermatozoid head begins as late as several days after it has been synthesized.

V. Regulation of Enzymatic Activity I n this section we consider two examples demonstrating the stage of realization of genetic information following translation; i.e., the enzymatic function of the protein synthesized may also involve regulatory mechanisms. It should be noted that by regulation we mean, not the maintenance of a constant level of enzymatic activity (of the type of allosteric inhibition), but rather the control over the changes in the enzymatic activity in the course of the embryonic development which is realized independently of genomic control. I n the early development of loach embryos, the rate of glycolysis increases to a maximum a t the end of gastrulation. An analysis of the activity of all glycolytic enzymes led Milman and Yurowitzki (1966, 1967) to the conclusion that the increase in rate of glycolysis cannot depend on an increased enzyme activity. It turned out that, in fact, in the course of development the activity of the enzymes of gluconeogenesis, i.e., glucose-6-phosphate dehydrogenase, fructose diphosphatase, and phosphoenolpyruvate carboxylase, gradually decreases. The decrease in the activity of these enzymes is not associated with the formation of some inhibitor, but rather it depends on the degradation of the enzymes-alternatively, on the rate of degradation prevailing over that of synthesis. Such a simple mechanism makes it possible to intensify the respiration of the embryo as developed proceeds. Whether or not this mechanism is genetic is just a question of terminology. On the one hand, the intensification of glycolysis occurs within the functioning enzymatic system, and the changes in its functions do not require additional nuclear control or macromolecular synthesis. On the other hand, the short lifetime of the enzymes of gluconeogenesis as compared to the enzymes of glycolysis is likely t o be genetically programmed. However, we suggest that the processes that can take place autonomously in the cytoplasm should be considered as %ongenetic,” or “cytoplasmic.” We have also revealed some changes in the activity of aspartate aminotransferase in the sea urchin (Botvinnik and Neyfakh, 1969; Abramova and Neyfakh, 1971) ; this is also an example of nongenetically regulated alteration of enzymatic activity in early development. This enzyme changes its activity in a rather characteristic way which is difficult to associate with the morphological or biochemical events of early

2.

73

STEPS OF REALIZATION OF GENETIC INFORMATION

development (Fig. 8 ) . It is interesting that its activity is not affected by either actinomycin or puromycin. A sufficiently high concentration of puromycin is known to inhibit completely even cleavage; nevertheless, the behavior of aspartate amhotransferase in such an uncleaving, but

22 o c

'1

2

4

2

8

-

BlQtaneres

i6

4

0

6

Hours

t 0

0

Strmgylmntrntus droebOch/nsis. 8 OC

2

2

4

8

Blastomeres

4

6

8

Hours

c

Time of development

FIG.8. Changes in aspartate amhotransferase activity in early development in the sea urchins Strongylocentrotus nudus (top) and 8. droebachiensis (bottom). Control; 0- 0, in the Enzyme activity is expressed in arbitrary units. 0-0, presence (top) of actinomycin (25 p g l r n l ) or (bottom) of puromycin (100 pg/ml).

-

fertilized, egg is absolutely similar to that in the normally developing embryos of the control. We had only limited success in elucidating the mechanism of the above activity changes. They may be partially accounted for by the concentration of the cofactor, pyridoxal fi-phosphate, in the egg, although some of the changes seem to be connected with the enzyme protein itself. If this is the case, the significance of the changes in enzymatic activity is difficult to evaluate. But i t is quite apparent that the changes do

74

A. A. NEYFAKH

not depend on direct genetic control, synthesis of macromolecules in general, or even morphological manifestations of development. VI. Conclusion

A good analogy between ‘(regulation” and ‘(program” may be found on the highway: regulation implies driving along the road; reaching the place of destination by sticking to one’s route is a program. I n the cells of the adult organism, regulation is predominant ; in cell differentiation, a program prevails. Maintenance of the constant rate of translation or enzymatic activity is realized by the mechanisms of regulation; but the regular changes in the transcription pattern or those in the rate of protein synthesis in the course of development are carried out in accordance with a program. The mechanisms ensuring regulation and programming are, evidently, basically different, as a regulatory apparatus must respond to a system’s deviation from a given parameter, whereas the program implies alteration of the parameters in time. A programming unit should, therefore, include a time-meter-for example, an indicator of distance on the road or of developmental stage in the embryo. The mechanisms of realization of regulation and of the program, however, may be similar: a turn of the wheel in the car or a change in the number of active ribosomes in the course of protein synthesis. Usually only one aspect of the application of the program in development is analyzed, i.e., differential gene activity. The present paper attempts to show that the later steps of the realization of genetic information are, or may be, the points of application of the development program. For the step of mRNA transfer from the nuclei, only indirect evidence is available; but in the case of translation, the data described above are unequivocal. The term “program” does not always imply the existence of complex systems. When we deal with selective transport or translation of mRNA molecules, this mechanism should involve recognition, i.e., be rather complicated. In other cases, simple degradation of some enzymes may be responsible for the changing intensity of glycolysis with time. Why is such an intricate hierarchical system of control and realization of the program, involving four (or more) steps, necessary? This is the consequence of the duration and complexity of the process of realization of genetic information during development. At every moment of development, the nucleus yields the information which should be realized a t different moments of the following process of development. And, vice versa, a t every moment information is being realized that was retrieved minutes, hours, and even months, ago. It is the coordination

2.

STEPS OF REALIZATION OF GENETIC INFORMATION

75

of these complex relationships in time that requires the corresponding control systems. The concept of development as a hierarchy of controllcd steps is more complex than t h a t of the on/off switching genes, but we hope that i t is truer to reality. REFERENCES Abramova, N. B., Likhtman, T. V., and Neyfakh, A . A. (1966). Fed. Proc., Fed. Amer. SOC.Exp. Biol., Traizsl. Szcppl. 25, 489. Abramova, N. B., and Neyfakh, A. -4.(1971). Onlogenesis 2, 71. Aronson, A. I., and Wilt, F. H. (1969). Proc. N a t . Acnd. Sci. U S . 62, 186. Bachvarova, R., and Davidson, E. H. (1966a). Proc. N o t . Acad. Sci. U.S. 55, 538. Bachvarova, R., and Davidson, E. H. (196613). J . Exp. Zool. 163, 285. Beritashvili, D. R., Kvavilashvili, I. S., and Kafiani, C. A. (1969). Exp. Cell Res. 56, 113. Botvinnik, N. M., and Neyfakh, A. A. (1969). Exp. Cell Res. 54, 287. Brown, D. D., and Dawid, J. B. (1968). Science 160, 272. Chamberlain, J. P. (1970). Biochim. Biophys. Acla 213, 183. Craig, S. P. (1970). J. Mol. Biol. 47, 115. Crippa, M. (1970). Nature (London) 227, 1138. Crippa, M., and Tocchini-Valentini, G. P. (1970). Nntzoe (London) 226, 1243. Czihak, G. (1965). Natzir2uissenschafteII 52, 141. Das, C. C., Kaufmann, B. P., and Gay, H. (1964). Ezp. Cell Res. 35, 507. Davidson, E. H. (1969). “Gene Activity in Early Development.” Academic Press, New York. Dawid, I. B. (1965). J . Mol. B i d . 12, 581. Dawid, I. B. (1966). Proc. Nut. Acad. Sci. U.S. 56, 269. Donzova, G. V., and Neyfakh, A . A. (1969). Dokl. Akad. Nazik SSSR 184, 1253. Doneova, G. V., Ignatieva, G. M., Rott., N. N., and Tolstorukov, I. I. (1970a). Ontogenesis 1, 474. Donzova, G. V., Tolstorukov, I. I., and Neyfakh, A. A. (1970b). Ontogenesis, 1, 602. Evans, D., and Bernstiel, M. (1968). Biochim. Biophys. Acta 166, 274. Felicetti, L., Gambino, R., Metafora, S., and Monroy, A. (1971)’. Symp. Sac. Exp. B i d . 25, 183. Georgiev, G. P., and Mantieva, V. L. (1962). Biokhiniiya 27, 949. Georgiev, G. P., and Samarina, 0. P. (1971). Advnn. Cell Biol. 2 (in press). Giudice, G., and Mutolo, V. (1967). Biochim. Biophys. Acta 138, 607. GliBin, V. R., and G l i b , M. V. (1964). Proc. Nat. Acad. Sci. U.S. 52, 1548. Goldstein, L., Rao, M. V. N., and Prescott, D. M. ( 1969). Ann. Embryal. Morphogen S z ~ p p l 1, . 189.

Gross, P. R., and Cousineau, G. H. (1963). Exp. Cell Res. 33, 368. Hartman, I. F., and Comb, D. A. (1969). J . Mol. B i d . 41, 155. Henning, W. (1968). Proc. Nat. Acnd. Sci. U S . 38, 227. Hess, 0. (1967). Exp. Biol. Med. 1, 90. Hynes, R. O., and Gross, P. R. (1970). Develop. B i d . 21, 383. Infante, A. A,, and Nemer, M. (1967). Proc. Not. Acad. Sci. U S . 58, 681. Kafiani, C. A., and Timofeeva, M. J. (1964). Dokl. Akad. Nauk SSSR 154, 721. Kafiani, C. A,, Timofeeva, M. J., Melnikova, N. L., and Neyfakh, A. A. (1968). Biochim. Biophys. Acta 169, 274.

76

A. A. NEYFAKH

Kafiani, C. A,, Timofeeva, M. J., Neyfakh, A. A., Melnikova, N. L., and Rachkus, J. A. (1969). J . Embryol. Exp. Morphol. 21, 295. Kafiani, C. A,, Gasarian, K. G., and Akhalkazi, R. (1971) Ontogenesis (in press). Kedes, L. H., and Gross, P. R. (1969). J. Mol. Biol. 42, 559. Kedes, L. H., Gross, P. R., Cognett, G., and Hunter, A. L. (1969). J . Mol. Biol. 45, 337. Kijima, S., and Wilt, F. H. (1969). J . Mol. Biol. 40,235. Kostomarova, A. A., and Burakova, T. A. (1972). Ontogenesis (in press). Kostomarova, A. A,, and Nechaeva, N. V. (1970). Ontogenesis 1, 391. Kostomarova, A. A., and Rott, N. N. (1969). Demonstrations presented in Znt. Embryol. Conf., Bth, 1961, p. 43. Krigsgaber, M. R., and Neyfakh, A. A. (1968). Dokl. Akad. Nauk SSSR 180, 1259. Krigsgaber, M. R., and Neyfakh, A. A. (1972). J. Embryol. Exp. Morphol. (in press). Krigsgaber, M. R., Ivanchik, T. A,, and Neyfakh, A. A. (1968). Biokhimiya 33, 1214. Krigsgaber, M. R., Kostomarova, A. A,, Terehova, T. A., and Burakova, T. A. ('1971). J. Embryol. Exp. Morphol. 26 (in press). Kroeger, H., and Lezzi, M. (1966). Annu. Rev. Entomol. 11, 1. Kukhanova, M. K., and Neyfakh, A. A . (1968). Unpublished information. Markman, B. (1961). Exp. Cell Res. 23, 118. Milman, L. S., and Yurowitzki, Yu. G . (1966). Dokl. Akad. Nauk SSSR 170, 721. Milman, L. S., and Yurowitzki, Yu. G. (1967). Biochim. Biophys. Acta 148, 362. Monesi, V. (1965). Exp. Cell Res. 39, 197. Nemer, M. (1967). Progr. Nucl. Acid Res. Mol. Biol. 7, 243. Neyfakh, A. A. (1959). J . Embryol. Exp. Morph. 7, 173. Neyfakh, A. A. (1964). Nature (London) 201, 880. Neyfakh, A. A., and Kostomarova, A. A. (1971). Exp. Cell Res. 65, 340. Neyfakh, A. A,, and Krigsgaber, M. R. (1968). Dokl. Akad. Nauk SSSR 183, 493. Neyfakh, A. A., and Radzievskaya, V. V. (1967). Genetika 3, 88. Neyfakh, A . A., and Rott, N. N. (1968). J . Embryol. Exp. Morphol. 20, 129. Neyfakh, A. A., Timofeeva, M. J., Krigsgaber, M. R., and Svetajlo, N. A. (1968a). Genetika 4, 90. Neyfakh, A. A., and Krigsgaber, M. R., and Il'in, M. J. (1968b). Dokl. Akad. Nauk SSSR 181, 253. Neyfakh, A. A., Abramova, N. B., and Bagrova, A. M. (1971). Exp. Cell Res. 65, 345. Niessing, J., and Sekeris, C. E. (1970). Biochim. Biophys. Acta 209, 484. Ovchinnikov, L. P., Avanesov, A. C., and Spirin, A. S. (1969). Mol. Biol (USSR) 3, 465. Perry, R. P., Cheng, T.-Y., Freed, J. J., Greenberg, J. R., Kelley, Dl. E., and Tartof, K . D. (1970). Proc. Nat. Acad. Sci. U.S. 85, 609. Pik6, L., Tyler, A., and Vinograd, J. (1967). Biol. Bull. 132, 68. Rachkus, J. A., Kafiani, K. A., and Timofeeva, M. J. (1971). Ontogenesis 3, 263. Rinddi, A. M., and Monroy, A. (1969). Develop. Biol. 19, 73. Rott, N. N. (1970). Znt. Congr. Anat., Oth, 1970, Thesis, p. 151.

2.

STEPS OF REALIZATION O F GENETIC INFORMATION

77

Rott, N. N., and Kostomarova, A. A. (1970). In “Intercellular Interactions in differentiation and Growth,” p. 24. “Nauka,” Moscow. Rott, N. N., and Sheveleva, G. A. (1968). J . Embryol. Exp. Morphol. 20, 141. Shiokawa, K., and Yamana, K. (1967). Develop. Biol. 16, 389. Shmerling, Zh. G. (1965). Usp. Sovrem. Biol. 59, 33. Singh, U. N. (1968). Exp. Cell. Res. 53,537. Slater, D. W., and Spiegelman, S. (1970). Biochim. Biophys. Actu 213, 194. Spirin, A. S. (1966). Curr. T o p . Develop. Biol. 1, 1. Spirin, A. S. (1969). Europ. J . Biochem. 10, 20. Spirin, A. S., Belitsina, N. V., and Ajtkhozin, M. A. (1964). Zh. Obshch. Biol. 25, 321; see Fed. Proc., Fed. Amer. SOC.Exp. Biol. 24, Transl., T907-T915 (1965). Stuffllefield, E., Klevecz, R., and Deaven, L. (1967). J. Cell. Physiol. 69, 345. Timofeeva, M. J. (1972). Mol. Biol. ( U S S R ) (in press). Timofeeva, M. J., Ivanchik, T. A., and Neyfakh, A. A. (1968). Dokl. Akud. Nauk SSSR 184, 1014. Timofeeva, M. J., Solovjeva, J. A,, and Sosinskay, J. J. (1972). Ontogenesis (in press). Weinberg, R. A., and Penman, S. (1970). J . Mol. Biol. 47, 169. Wilt, F. H. (1964). Develop. Biol. 9, 299. Wilt, F. H. (1970). Develop. Biol. 23, 444.

This Page Intentionally Left Blank

CHAPTER 3

PROTEIN SYNTHESIS DURING AMPHIBIAN METAMORPHOSIS J. R. Tata NATIONAL I N S T I T U T E FOR M E D I C A L R E S E A R C H , M I L L HILL, LONDON, E N G L A N D

I. Introduction ............................................... 11. The Role of Hormones in Amphibian Metamorphosis.. ......... 111. Proteins Involved in Metamorphosis. ......................... IV. Regulation of Protein Synthesis during Metamorphosis., ........ A. Current Concepts of Regulation of Protein Synthesis in Animal Cells .................................... B. Regulation of Protein Synthesis in the Developing Tadpole Hepatocyte during Metamorphosis. ....................... V. The Role of DNA Sy VI. Requirement of RNA VII. Conclusions and Fu References .................................................

79 80 81 83

83 85

107

1. Introduction Metamorphosis is one of the most dramatic developmental changes in late embryonic life. Some of the classical notions of functional maturation, as in the case of acquisition of urea formation in Amphibia, have emerged from a biochemical study of amphibian metamorphosis (see Weber, 1967; Cohen, 1970). The initiation and completion of metamorphosis in both amphibians and insects is under obligatory hormonal control which distinguishes it from other hormone-dependent late developmental changes. It is well known that an anuran tadpole or an insect larva will never turn into its adult form if the respective endogenous metamorphic hormones, thyroxine and ecdysone, were withdrawn or prevented from reaching the target tissues. Another feature that distinguishes both amphibian and insect metamorphosis from ordinary adaptational responses is that the process is begun and completed in anticipation of a change in environment. The hormone serves t o trigger a predetermined program of developmental changes in order to prepare the embryo for an environment suitable for adult life. Since this chapter will deal with induction of new functions and proteins, it is important to note that the hormone in no way directs cells to acquire differentiative characters but allows already differentiated but immature cells to acquire adult functions and structures. 79

80

J. R . TATA

Ever since the discovery by Gudernatsch (1912) that exogenous thyroid hormone will cause the precocious onset of metamorphosis in tadpoles, hormonal induction of the process is a well established procedure of studying sequential biochemical changes in a variety of amphibians and insects. In most studies little has been learned about the mechanism of action of the hormone, the latter usually serving as convient tool to induce or delay developmental processes in free-living embryos. I t is in this context of a dcvclopmental tool that thyroid hormones will be treated in this article which is restricted to amphibian metamorphosis. There arc great similarities in the types of phenomena involved in insect metamorphosis which has been reviewed by many authors (Wyatt, 1962; Karlson, 1963; Schneidcrman and Gilbert, 1964). The following two main facets of the regulation of protein synthesis leading to a well-ordered and sequential acquisition of new functions or structures during metamorphosis will be considered separately : (1) the regulation of RNA synthesis, formation of enzymes, and alteration in cellular structures in tissues, such as the hepatocyte, that undergo further developmcnt or functional maturation ; (2) the importance of new or additional protein Synthesis in tissues, such as the tail, that are programmed for death or regression. I n accordance with the aims of this publication the account given below is based on personal interests and ideas and is not meant to be an exhaustive review of the subject. For the latter, the reader is referred to several reviews that have appeared in the last few years (see Bennett and Frieden, 1962; Wcber, 1967; Frieden, 1967; Cohen, 1970; Tata, 1970a, 1971a; Frieden and Just, 1970). II. The Role of Hormones in Amphibian Metamorphosis

Not long after the discovery by Gudernatsch (1912) that feeding tadpoles on mammalian thyroid tissue induced precocious metamorphosis, it was found that thyroidectomy of larvae (or feeding larvae on antitliyroid drugs) prevented their development into adult frogs or toads (Allen, 1916; see Etkin, 1964). The dormant thyroid tissue of the developing larva is activated into producing and secreting the two thyroid hormones, L-thyroxine and 3,3',5-triiodo-~-thyronine(Ts), by thyrotropic hormone (TSH) produced by the anterior pituitary (see Sax6n et al., 1957; Etkin, 1964, 1968). The larval pituitary itself is under neural control. Thus eventually it is some environmental factor, such as illumination, temperature, salinity, that acts as the initial trigger for metamorphosis. Although a hypothalamic tripeptide, thyrotropinreleasing factor ( T R F ) , of the kind recently described in mammals has not yet been discovered in amphibian larvae, there is much indirect

3.

METAMORPHOSIS

81

evidence to suggest that a similar neurochemical link may exist between external signals and the production of endogenous thyroid hormones (Etkin, 1964, 1968). Of great interest in considering hormonal control of metamorphosis is the relatively recent discovery in amphibians of “juvenilizing” factors of the type known for insects (Berman e t aZ., 1964; Bern et aZ., 1967; Etkin and Gona, 1967; Bern and Nicoll, 1969). The level of insect juvenile hormone is well known to determine whether or not or how the larval target tissues will respond to the metamorphic stimulus of ecdysone (see Williams, 1961 ; Schneiderman and Gilbert, 1964). It now seems that prolactin obtained from mammalian pituitaries exerts a similar juvenilizing effect on tadpoles in that its administration a t the same time of spontaneous or thyroid hormone-induced metamorphosis arrests, delays, or modifies the process. In salamanders and newts (urodeles), which do not undergo the same type of metamorphosis as in frogs and toads (anurans) , prolactin causes the terrestrial forms to return to water, a phenomenon known as “water drive” or “second metamorphosis” (see Bern and Nicoll, 1969). 111. Proteins Involved in Metamorphosis

Accompanying the dramatic visible morphological changes during metamorphosis (tail regression, limb emergence, positioning of eyes, etc.) are profound functional and biochemical changes in most tissues. Many of the changes or acquisition of new functions necessary for the organism for a terrestrial life are a consequence of the induction or preferrential synthesis of proteins. The induction of the enzymes of urea cycle leading to the switch from ammonotelism to ureotelism during amphibian metamorphosis is now indeed a classic of the biochemical basis of development (Cohen, 1966, 1970). Table I lists some of the functions and the main proteins underlying these that are involved. Only a few salient features emerging from Table I need be mentioned a t this point. First, virtually every type of cell in the embryonic tissue is subjected to the action of thyroid hormones, and none fails to respond in some important way. (This is not too surprising if one is dealing with a rapid transition from an aquatic to a terrestrial life.) Second, the hormone acts locally and directly on different cells, as for example Kollros (1942) has shown that only the part of that tail to which thyroxine was applied underwent lysis, leaving the rest of the tail intact. I n a series of elegant experiments, Wilt (1959; Ohtsu et al., 1964) had shown that only that eye to which thyroxine was applied developed rhodopsin, leaving the other with the larval visual pigment, porphyropsin. There is thus no indirect systemic action of the hormone although an alteration in the metabolic pattern of one tissue is eventually bound

TABLE I SOMEFUNCTIONS A N D PROTEINS UNDERLYING TEEMTEATARE CHARACTERISTIC OF NATURAL AND PRECOCIOUS THYROID IN ANURAN TADPOLES HORMONE-INDUCED METAMORPHOSIS Tissue

Change in function

Induction or preferential synthesis of proteins

Liver

Structural maturation, urea formation, composition of blood proteins, energy metabolism

Urea cycle enzymes (carbamyl phosphate synthetase, ornithine carbamyl transferase, arginasc, etc.) ; serum albumin, mitochondria1 proliferation

Tail, gut, gills

Resorption

Hydrolases (nucleases, cathepsin, 8-glucuronidase, collagenase, etc.)

Limb buds, lung, bone

Growth, organogenesis, adaptation to atmospheric oxygen

All components of muscle, bone, nerve

Skin (epithelial cells)

Hardening, adaptation against dehydration, salt movement

Collagen; Na+ I(+ATPase

Erythroid cells

Adaptation to higher oxygen tension

Replacement of fetal hemoglobin with adult hemoglobin

Visual cells

Photo pigment conversion (change in Replacement of porphyropsin by rhodpsin (enzymes for utilization of visual function unknown) vitamin A)

I-'

p

3.

METAMORPHOSIS

83

to affect the activity of another via secondary adaptational routes. Third, the types of proteins whose synthesis is induced is determined by the nature of the cell, that is to say that there is qualitatively no hormonedetermined protein synthesis. In some cases, the positioning of some types of cells determines the response. For example, epithelial cells of the tail synthesize collagenase while those on the skin accumulate collagen under the influence of thyroid hormones (Gross, 1964) or that thyroxine may provoke regression as well as growth in adjacent cells of Mauthner’s neurons (Weiss and Rossetti, 1951; Kollros, 1968). The last point is of considerable importance in the definition of the role of developmental hormones in the genetic control of protein synthesis. The multiplicity of responses to thyroxine in the same organism makes it highly unlikely that the hormone has any inherent informational content to determine, say, via an interaction with genes or repressors, to induce the synthesis of a fixed number of proteins. When one extends the above list to the very different actions of thyroid hormones in fish, bird, and mammals, it is only reasonable to conclude that once a cell has the receptor to recognize the hormone then it makes use of the hormone as a trigger mechanism to initiate processes determined, but not expressed, during early differentiation.

IV. Regulation of Protein Synthesis during Metamorphosis

A. CURRENT CONCEPTS OF REGULATION OF PROTEIN SYNTHESIS IN ANIMAL CELLS

It is now widely accepted that, although some fundamental concepts of DNA transcription and messenger RNA translation are universal, regulation of protein synthesis in nucleated cells of higher organisms involves additional control steps not described in microorganisms. Several reviews have been devoted to this topic, and the reader’s attention is drawn to two publications edited by San Pietro et al. (1968) and Wolstenholme and Knight (1970). The features mentioned briefly below are important for the interpretation of the work to be described later. Translational Control The contrast between long-lived proteins synthesized on unstable bacterial messengers and some animal proteins of short half-life synthesized on relatively stable messengers has prompted many investigators to propose an extranuclear regulatory mechanism in higher organisms (see Tomkins et al., 1969). Much of the earlier evidence for a translational control of protein synthesis was based on indirect manifestations of inhibitors of RNA and protein synthesis and did not establish the 1.

84

J. R. TATA

exact level at which such a control could be effected. The explanation of Wool et al. (1968) for the anabolic actions of insulin and of Korner (1970) for that of growth hormone are based on an almost direct modification of the functioning of ribosomal subunits. But the currently most favorable hypothesis of translational control is that put forward by Tomkins (Tomkins et al., 1969) to account for the induction of tyrosine aminotransferase in rat hepatoma cells. Tomkins has proposed the existence of a cytoplasmic repressor whose function is to control stability or availability of messenger RNA, for translation. It is the synthesis of such a repressor that is thought to be under hormonal control. No such evidence is yet available for metamorphosis or similar late embryonic developmental systems, and direct translational control, important as it may eventually turn out to be, will not be discussed in this article. 2. Nuclear Restriction and Breakdown of R N A

A substantial amount of RNA synthesized in the nucleus is rapidly turning over, heterodisperse and of high molecular weight (see Harris, 1964; Georgiev, 1966; Schcrrer and Marcaud, 1969). This RNA is not a precursor of rRNA, mRNA, or tRNA, and DNA-RNA hybridization studies have shown that many of the species of RNA within the nucleus do not appear in the cytoplasm (Georgiev, 1966; Shearer and McCarthy, 1967). Britten and Davidson (1969) have suggested that the rapidly turning over intranuclear RNA hybridizes very rapidly with DNA and may be a product of repeating DNA sequences or “redundant” DNA. Although a precise role for this RNA is far from established, they and others (Schcrrer and Marcaud, 1969; Kijima and Wilt, 1969) have thought it may be somehow important in differentiation and growth. 3. Selective Transfer of RNA f r o m A’ucleus t o Cytoplasm

An intranuclear restriction of one whole class of RNA poses the question of what mechanisms control a selective transfer of RNA from the nucleus to the cytoplasm. The transfer of both ribosomal and messenger RNA seems to require the formation within the nucleus of ribonucleoprotein particles which are then incorporated into cytoplasmic polysomes. Several workers have now clctected informosome-like particles both in the nucleus and the cytoplasm (Spirin, 1969; Samarina et al., 1968; Henshaw, 1968; Olsnes, 1970; Spohr et al. 1970). The role of the protein associated with such informational particles in the polysomes may be an important factor in the availability or rate of mRNA translation. Very little is yet known about the nature of these proteins, their

3.

85

1IETAMORPHOSIS

site of synthesis, how they react selectively with mRNA, etc., but any developmental process must depend on their ready availability.

4.Structural Requirement for Protein Synthesis This article deals in some detail with an important question which has not received much attention concerning regulation of protein synthesis in animal cells. It relates to the association between cellular structure and protein synthesis, particularly the attachment of ribosomes to membranes of the eridoplnsmic reticulum. Such an attachment is known to affect protein synthesis, but so far the main role for the structural association is thought to be that of secretion of proteins (see Palade, 1966; Hendler, 1968; P. N. Campbell, 1970). Howcver, the special role of membrane-bound ribosomes in bacteria (Schlessinger, 1963; Chefurka e t al., 1970) and the presence of highly active membrane-bound ribosomes in predominantly nonprotein secreting tissues (Andrews and Tata, 1968, 1971; Priestley et nl., 1968) suggests some other function for the membrane-ribosome attachment. It is of particular interest to note that the ribosomes and membranes to which they are attached are turning over continuously (Siekevitz et al., 1967) and that there exists a tight coordination between the proliferation of membranes and ribosomes when additional demands are made for protein synthesis during growth and development (Tata, 1970b,c; 197111). OF PROTEIN SYNTHESISIN B. REGULATION HEPATOCYTE DURING METAMORPHOSIS

THE

DEVELOPING TADPOLE

The liver of the tadpole has been intensively studied in the laboratories of Frieden (1967; Frieden and Just, 1970) and Cohen (1966, 1970) with respect t o the synthesis of proteins that characterize amphibian metamorphosis, such as urea cycle enzymes, serum albumin, and adult hemoglobin. These and other workers had firmly established, especially in the bullfrog, Rana catesbeinnn, that administration of thyroid hormones to premetamorphic tadpoles causes a de nowo synthesis of these proteins. Because of this firm biochemical background, the author’s laboratory has over the last sewn years investigated the formation and turnover of nuclear and cytoplasmic RNA in the premetamorphic bullfrog hepatocyte a t different stages after hormonal induction of metamorphosis in order to understand the nature of the process of induction (Tata, 1965, 1967a, 1970a,b, 1971). Cohen’s group have also been extensively investigating the metabolism of RNA i n viwo and in isolated preparations of tadpole liver (Nakagawa and Cohen, 1967 ; Nakagawa et al., 1967; Blatt et al., 1969; Cohen, 1970).

86

J. R. TATA

1. Thyroid Hormone-Induced Formation of N e w or Additional Proteins

I n bullfrog tadpoles (Rana catesbeiana) a rather long lag period of about 6 days elapses after exogenous thyroid hormone administration before new or additional proteins, characteristic of the metamorphic change in the function of the liver, could be detected (Paik and Cohen, 1960; Metzenberg et al., 1961; Nakagawa et al., 1967; Tata, 1967a). The increase of urea cycle enzymes or the appearance of serum albumin in the blood is preceded by a day or two by a rather abrupt increase

z 0

I00

200

300

Time after tri- iodothyronine (hours)

FIG.1. Schematic representation of the lag period for the induction by triiodothyronine of de novo protein synthesis and the stimulation of amino acid incorpo- 0,Specific ration into hepatic protein of Rana catesbeiana tadpoles. 0 radioactivity of protein recovered in the liver microsomal fraction 40 minutes after the administration of a mixture of “C-labeled amino acids; 0-0, carbamyl specific activity of cytochrome oxidase in mitophosphate synthetase ; A-A, , serum albumin accumulation in blood. Data compiled chondrial fraction ; from Tata (1965, 1967a).

.---.

---

in the rate of amino acid incorporat,ion into protein in vivo per unit of ribosomal RNA (see Fig. 1, Tata, 1967a). The decline in the rate of incorporation as seen in Fig. 1 is only an apparent one caused by the progressive changes in levels of free amino acid as regression of organs like tail, intestine, and gills gets under way. To some extent the lag period preceding the increase in protein synthetic rate represents the time for additional RNA to be synthesized and processed in the nucleus (see Section IV,B, 2 below). Some of the RNA synthesized during the lag period is certainly important for the de novo synthesis of hepatic

3.

METAMORPHOSIS

87

metamorphic proteins since actinomycin D, administered with or soon after thyroxine will prevent the rise in carbamyl phosphate synthetase (Nakagawq et al., 1967; Kim and Cohen, 1968). However, hormonal control of transcription is not the exclusive mechanism for the induction of these proteins, and it is thought that thyroid hormones are also required for a sustained translation of messenger RNA or for some process of maturation of the protein molecules. Cohen and his colleagues have concluded that thyroxine exerts a dual action in regulating the rate of formation of the key urea cycle enzyme, carbamyl phosphate synthetase (Shambaugh et al., 1969) as well as for glutamate dehydrogenase (Balinsky et al., 1970). It seems that part of the induction is due to control of transcription which involves all species of RNA and that almost simultaneously the hormone controls the activation of an inactive form of the enzyme. These results were based on the discrepancy between the detection of the enzyme by immunochemical methods and measurement of enzyme activity. Part of the inactive enzyme was preformed and part synthesized following hormone administration. Shambaugh et al. (1969) have further found that thyroxine also stimulates the synthesis of the inactive form of the messenger for carbamyl phosphate synthetase, but it is not certain whether or not it involves mechanisms based on the inactivation by the hormone of a cytoplasmic translational repressor of the type described by Tomkins for explaining the induction of tyrosine aminotransferase in cultured rat hepatoma cells by cortisol (Tomkins et al., 1969). How or a t what rate-limiting step of translation the hormone exerts an effect on protein synthesis is not clear although a few observations may be relevant. Unsworth and Cohen (1968) reported a n enhanced activity of hepatic aminoacyl tRNA transferase during induced metamorphosis of bullfrog tadpoles, a phenomenon that is frequently observed in many rapidly developing systems. Tonoue et aE. (1969) found an altered pattern of leucyl tRNA charged in vivo in a number of tadpole tissues during metamorphosis. It could also be argued that the enhanced uptake of amino acid under the influence of thyroid hormones is another way in which translational process is facilitated in a nonspecific way (Eaton and Frieden, 1969). Perhaps it is not necessary to separate translational and transcriptional phenomena from one another, but to consider that the two processes are coupled and integrated into a well coordinated regulatory complex. 2. Nuclear R N A Synthesis

It can be seen in Fig. 2 that there occurred, well within the latent period of 5-6 days for new proteins to be detected (see Fig. l ) , an

88

J. R. TATA

2

200 Time after tri-iodothyronine (hours)

.-.,

FIG.2. Schcmatic representation of the stimulation of nuclear RNA synthesis and its turnover into the cytoplasm of liver of Rntin cnlesbeinrin tadpoles after induction of metamorphosis with triiodothyronine. 0- - - - 0, Specific activity of nuclear RNA labeled with uridine-3H not corrected for changes in distribution of radiospecific activity of nuclear RNA activity in the acid-soluble fraction; after correction for changes in acid-soluble radiosctivity ; A-A, specific activity of RNA recovered in cytoplasmic ribosomes and polyribosonies. The earlier stimulation of labeling of nuclear RNA when not correctcd for acid-soluble radioactivity is presumably duc to a more rapid action of the hormone on the uptake or pool size of uridinc. The abrupt downward trend of the curves is due to the dilution with nucleotides released by the regression of tail, gut, gills, etc. Data compiled from Tatn (1965, 1967a).

acceleration of the rate of nuclear and cytoplasmic RNA synthesis in bullfrog tadpole livcr following induction of metamorphosis with triiodothyronine. Somewhat similar results in the same species of tadpole induced with thyroxine have also been observed in Cohen’s laboratory (Nakagawa and Cohen, 1967). The curves for rates of RNA synthesis in Fig. 2 have been derived from values of specific activity of RNA obtained after correction for changes in uptake of the radioactive precursor which occur earlier after triiodothyronine, as has also been bound by Eaton and Frieden (1969). An early perturbation in the pool size or uptake of precursors of RNA and protein in target tissues has been observed with hormones known to affect protein synthesis (Tata, 1968a; Manchester, 1970; Young, 1970). The question of pool sizes in radioactive labeling of constituents of nonregressing tissues is even more complicated when regression occurs in other tissues in the same organism

3.

METAMORPHOSIS

89

during metamorphosis. Thus, the abrupt downward trend in the specific activity of nuclear and cytoplasmic RNA as was also noticed for amino acid incorporation (Fig. 1), is not due to a sudden reversal of the accelerated rate of synthesis but merely reflects a dilution of radioactive precursors caused by the autolysis in regressing tissues, such as the tail, gut, and gills. It would be tempting to suggest that the additional RNA made following hormone administration includes messengers for proteins like urea cycle enzymes, serum albumin, etc., especially as a sustained RNA synthesis is important for metamorphosis to occur (Weber, 1965; Nakagawa et al., 1967). Furthermore, an alteration in template activity of liver chromatin from thyroxine-treated bullfrog tadpoles has been observed (Kim and Cohen, 1966). I n our laboratory, however, we failed to demonstrate, by base analysis, sucrose density gradient fractionation and DNA-RNA hybridization that a significant part of the additional nuclear RNA synthesized in vivo a t the onset of metamorphosis was messenger- or even DNA-like RNA (Tata, 1967a; Wyatt and Tata, 1968). Our DNA-RNA hybridization studies have shown that whereas there did occur a net increase in the amount of readily hybridizable nuclear RNA, there was also a drop in the overall “hybridization efficiency” of the RNA formed during metamorphosis. This paradoxical finding really suggests that very small increases in DNA-like RNA following hormone administration are accompanied by relatively massive increases in the rate of synthesis of ribosomal RNA. This is not to say that no changes occurred in the nature of genes transcribed under hormonal influence but that one does not yet have sensitive enough methods to detect a small change in a wide spectrum of nuclear RNA molecules with eventual messenger function. Much of the DNA-like RNA in the nucleus, which is also rapidly hybridizable, is not found in the cytoplasm, and although it may have a function in differentiation (Britten and Davidson, 1969) it is not thought to be the direct precursor of cytoplasmic messenger RNA (Billing and Barbiroli, 1970; Penman et al., 1970). 3. Cytoplasmic R N A

A consequence of the burst of nuclear RNA synthesis is the appearance of additional cytoplasmic RNA, mainly in the heavier polyribosomal aggregates with a concomitant decrease in the relative amount of monomeric ribosomes or ribosomal subunits (Tata, 1967a). This build-up of polyribosomes illustrated in Fig. 3 occurs a t a time just preceding the appearance of new proteins, as can be judged from a

90

J. R. TATA

i

o’6 0.4

600 a

400

0.2 I

0

10

5

15

20

25

f iE

30

Fraction number

-

\ In

1.2 0

r

11200

b

-$

wN 1.0 0.a

0.6

0.4 0.2

0 Fraction number

FIQ.3. Distribution of nascent protein synthesized in t h o on hepatic polysomes obtained from (a) premetamorphic bullfrog tadpoles and (b) animals in which metamorphosis had been induced 7.0 days before the preparations were made. Proteins were labeled by injecting 4 pCi of a mixture of %-labeled amino acids (algal-protein hydrolyeate) 15 minutes before the animals were killed. Polysome profiles were determined on mitochondria-free supernatant treated with 0.15% sodium deoxycholate. All other details were the same as in Fig. 7. -, ESo; - - - -, radioactivity. From Tata (1967a).

comparison of Figs. 1 and 2. An increased rate of accumulation of newly synthesized ribosomes and polyribosomes coinciding with new or additional protein synthesis seems to be a common feature of regulation of protein synthesis during growth and development (see Tata, 1967b, 1968a, 1970c; Hamilton, 1968). However, unlike the situation of hormone-induced growth in many mammalian tissues, there is no appreciable accumulation of ribosomes per cell during the initial phase of induction of metamorphosis (6-10 days). This conclusion was borne out by the double isotope labeling studies shown in Table 11, which suggests that there is an accelerated turnover of cytoplasmic RNA and ribosomes as additional ribosomes appear after hormone administration. It seems that there is some mechanism within the developing cell that selectively breaks down “old” ribosomes and allows the preferential accumulation of “new” ribosomes formed after the metamorphic stimulus was applied.

3.

91

METAMORPHOSIS

TABLE I1 DOUBLE-LABELINQ OF RIBOSOMAL RNA DEMONSTRATINQ THATBOTHBREAKDOWN AND SYNTHESIS OF RNA ARE ACCELERATED AT THE ONSETOF METAMORPHOSIS~,~ Day on which orotic acid-14C Metamorphosis administered induced 0

2 4 6

-

++++

Specific activity (cpm/mg of RNA) *H

1 4 c

aH: 14C ratio

17,350 14,900 15 ,765 8,010 15,500 6 ,360 12,870 5,085

4 ,600 5 ,230 5,680 7,825 3,650 11,300 4,100 9,420

3.77 2.85 2.78 1.02 4.24 0.56 3.13 0.54

From Tata (1967a). A batch of 64 tadpoles was divided into eight groups of eight each. Each tadpole received 12 pCi of uridine-aH 8 days before metamorphosis w&sinduced (day 0). Two groups (one control, one induced) were then given 2 pCi of orotic acid-g-‘C on days 0, 2, 4, and 6 and killed 24 hours later. Ribosomes were prepared from mitochondriafree supernatant treated with 0.4 % sodium deoxycholate. 5

4. Distribution of Ribosomes in the Cytoplasm and Proliferation of Endoplasmic Reticulum Membranes What is perhaps of utmost importance in studying the sequential events that occur between the initial burst of RNA synthesis following triiodothyronine administration and the appearance of newly synthesized proteins in a redistribution of ribosomes attached to membranes of the endoplasmic reticulum (Tata, 1967a). That such a redistribution may have occurred was first suggested in experiments in which liver mitochondria-free supernatant fractions from premetamorphic and induced tadpoles were titrated against Na deoxycholate (see Table 111). When the fraction of ribosomes remaining attached to microsomal membranes was measured as more deoxycholate was added, thmose from metamorphosing tadpoles were found to be more tenaciously bound to endoplasmic reticulum. The next question was to find out whether there was a redistribution of all “old” and “new” ribosomes on preexisting and stable membranes or whether there was also some alteration in the formation or proliferation of membranes. It is known that cellular membranes, even in resting cells, are not metabolically stable but turning over quite rapidly (Siekevitz et al., 1967; Arias et al., 1969). To study this problem we compared accumulation of newly formed ribosomes,

TABLE I11 EFFECTOF INDUCING METAMORPHOSIS O N THE RELATIVE DISTRIBUTION OF NEWLYSYNTHESIZED RNA IN MEMBRANE-BOUND A N D FREERIBOSOMAL FRACTIONS FROM LIVERMITOCHONDRIA-FREE SUPERNATANTS TREATED WITH DIFFERENT AMOUNTS OF SODIUM DEOXYCHOLATE~ Specific activity (cpm/EZsounit) and distribution of RNAb in fractions (% in parentheses) Tadpoles Control

Conc. of sodium deoxycholate (%)

Membrane-bound ribosomes

Polysomes

Dimers monomersC

<50 S

0.08 0.16 0.30

295 (4.7) 418 (2.0) - (0)

227 (21.5) 254 (22.8) 238 (18.7)

107 (56.6) 92 (53.0) 122 (62.5)

58 (12.0) 41 (8.8) 73 (14.5)

0.08 0.16 0.30

875 (20.5) 990 (12.8) 1185 (1.5)

430 (33.5) 518 (35.1) 550 (25.8)

492 (26.0) 631 (32.9) 675 (29.0)

196 (10.5) 281 (15.1) 255 (11.9)

+

‘ 3 k

6 Days after triiodothyronine

a

From Tata (1967a).

* Microsomal RNA was labeled by injecting

11.5 pCi of uridine-5H per tadpole 22.5 hours before they were killed. Each group comprised 10 tadpoles, and one group was treated with 0.7 pg of triiodothyronine 6 days before the experiment. Equal amounts of mitochondria-free supernatants were treated with different quantities of sodium deoxycholate, indicated by their final concentration, but the sucrose-density-gradient centrifugation was carried out in a Spinco No. 30 rotor as described by Tata (1965). Fractions (1 ml) were then pooled, after determination of E 2 6 0 , into the four fractions. The “membrane-bound” ribosomal fraction is that which did not, sediment through the 2.0 M sucrose interface. Also includes ribosomal precursor particles smaller than 78 S monomers.

?-

3.

93

METAMORPHOSIS

their distribution on endoplasmic reticulum and the labeling of membrane phospholipids as an index of membrane proliferation. As shown in Fig. 4 there occurred an almost simultaneous increase in the synthesis of ribosomes and membranes of the endoplasmic reticulum, especially those to which ribosomes are bound (rough endoplasmic reticulum). There were no obvious qualitative changes in the types of

I

50

0

50

I

I

1

100

150

200

100

150

200

Time after T,(hours)

FIG.4. Coordinated proliferation of endoplasmic reticulum and the increase in hepatic protein synthesis in vivo in the heavy rough membranes (densely packed ribosomes attached to membranes) during triiodothyronine-induced metamorphosis of young bullfrog tadpoles. (a) T h e appearance of newly synthesized (“P-labeled) heavy rough membrane phospholipids (0) and R N A ( 0 ) ;(b) the recovery of labeled nasrent proteins formed iu vzuii after a short pulse of radioactive amino light) rough ,membranes ( A ) , and free acids in heavy rough membranes (I polysomes ( 0 ) For . details, see Tata (1967a).

membrane phospholipids synthesized, but it is interesting to note that the simultaneous increase in the formation of the two rough endoplasmic reticulum components is accompanied by an enhanced rate of protein synthesis in vivo (Fig. 4b). Titration of mitochondria-free extracts with sodium deoxycholate showed that the newly formed ribosomes (as judged from the differences in specific activity of RNA in Table 111) in induced animals were more tightly bound to membranes of the rough endoplasmic

94

J. R. TATA

reticulum, i.e., they required a higher amount of detergent for their release. I n Fig, 4 it can be seen that the “heavy rough membrane” fraction, i.e., the one in which ribosomes are more densely packed on membranes, was more active in protein synthesis in vivo than were free ribosomes or the “light rough membrane” fraction. Although the evidence is too tenuous to warrant associating a distribution pattern in broken cell preparations with differential protein synthetic rates in the intact cell, it is tempting to suggest that the protein biosynthetic response to the hormone may be contained in the newly formed ribosomes. The above ribosomal redistribution changes are so marked that it is easy to corroborate them by electron microscopy, as shown in Fig. 5. These studies revealed that induction of metamorphosis caused a shift in the distribution of ribosomes from around the simple vesicular membrane structures of the immature larvae to the more complex double lamellar structures more commonly seen in mature tissues (Tata, 1967a,b). At the initial stages of metamorphosis, these new structures are predominantly located in the perinuclear region, an event that could be easily discerned by light microscopy of toluidine blue stained preparations. It is interesting that this shift in structural organization of polyribosomes on the endoplasmic reticulum coincided with the biochemical observations, mentioned earlier, on the formation of RNA and phospholipid and the synthesis of proteins by polyribosomes. Rather similar proliferation of components of the endoplasmic reticulum accompanying changes in protein synthesis has also been described more recently in thyroxine-induced and natural metamorphosis of the bullfrog tadpole by Cohen (1970) and Bennett (Bennett et al., 1970; Bennett and Glenn, 1970). Such a coordination between formation and distribution of ribosomes on the one hand and their protein synthetic activity on the other is not a feature that is exclusive to amphibian metamorphosis but is found in a wide variety of systems of rapid growth or functional maturation (see Tata, 1968a, 1970b,c). Ultrastructural changes of the kind described above raise the more general question of the role of attachment of ribosomes to membranes of the endoplasmic reticulum. The major role assigned so far to the attachment is that of facilitation of secretion of proteins for export (Palade, 1966; P. N. Campbell, 1970; Hendler, 1968). However, the proliferation of endoplasmic reticulum or a preferential shift from free to membrane-bound ribosomes is a characteristic of growth and development. This is true for both nonsecretory and secretory cells in which relatively more protein for intracellular use is made during rapid growth (Tata, 1970b; 1971b). It is likely that the attachment of ribosomes to membranes serves some other role during growth and development.

3.

METAMORPHOSIS

FIG.5 . a b and c. Seep. 98 for legend.

95

96

J. R. TATA

FIG.5. d and e . See p. 98 for legend.

3.

METAMORPHOSIS

FIG.5f. See p. 98 for legend.

97

98

J. R. TATA

The significance of the coupling of structural alterations and biosynthetic function that is noticed in tadpole liver during metamorphosis, mentioned above, may be that: (a) a response to growth and developmental stimulus is contained in newly formed ribosomes, and (b) that there is a topographical segregation on membranes of the endoplasmic reticulum of differently precoded polyribosomes carrying out the synthesis of different groups of proteins. There is, of course, no direct experimental evidence to verify either of these suggestions, but some recent work from our laboratory on the additive effects of growth and developmental hormones in mammalian organs provides an indirect support (Tata, 1968a, 1970b). I n these experiments, when growth hormone, triiodothyronine and testosterone were administered in different combinations, the increased rates of formation of ribosomes and the membranes of the rough endoplasmic reticulum were coordinated with those of synthesis of different classes of proteins in the liver and seminal vesicles. Since these hormones have quite different latent periods of action, the simultaneous administration of any two caused stepwise increases in the rates of proliferation of ribosomes and rough endoplasmic membranes, each burst corresponding in its time-course and magnitude t o that induced by each individual hormone. It is assumed in the above suggestions that an exchange between polysomes and membranes proceeds a t a low rate relative to the lifetime of these structures. An indirect approach to testing the idea of a topographical segregation would be to identify induced proteins on proliferating rough endoplasmic reticuFIG. 5 . Explanation of plates. (a) and (b) Sections of bullfrog tadpole liver as seen by light microscopy, showing typical patterns of cytoplasmic basophilic material (stained with toluidine blue) in a premetamorphic control animal (a) and 6 days after induction of metamorphosis with triiodothyronine (b). I n treated animals the longer clumps of basophilic material (RNA) have a striated appearance. (c) and (d) Electron micrographs showing either scattered distribution of ribosomes (R) or ribosomes associated with simple vesicular membranes (V) of the endoplasmic reticulum (ER) in the hepatocyte of a premetamorphic bullfrog tadpole. A well defined annular polysome complex is shown at P. M, mitochondria; Lip, lipid; N, nucleus; CM, cell membrane. (e) and (f) Electron micrographs of part of hepatic cell from a bullfrog tadpole 6 days after induction of metamorphosis. The cytoplasm has a higher density of ribosomes than in the premetamorphic controls and many of the ribosomes are now found associated with membranes of the endoplasmic reticulum (ER) of the double lamellar type. A large concentration of polysomes (P) is found in areas where the section grazes the rough endoplasmic reticulum. N, nucleus, M, mitochondrion. The inset of (f) is an electron micrograph a t higher magnification of hepatic cell cytoplasm in a metamorphosing tadpole, showing well developed polysome complexes. The author is indebted to Dr. J. A. Armstrong for the electron micrographs. Plates reproduced from Tata (c) ~ 6 0 , 0 0 0 ;(d,e) ~ 2 8 , 0 0 0 ;(f) ~ 3 2 , 0 0 0 ;inset, ~ 6 0 , 0 0 0 . (1967a,b). (a,b) ~ 1 9 0 0 ;

3.

METAMORPHOSIS

99

lum during development. We are now attempting by histochemical methods the detection and localization of ornithine and aspartate carbamyl transferases as they increase during induced metamorphosis. The recent work of Pitot et al. (1969) on the immunochemical location of serine dehydratase has shown that this inducible enzyme is almost exclusively synthesized on membrane-bound and not free ribosomes of the liver. In a different system, Leduc et al. (1968) have shown that antibody synthesized at the initial stages of stimulation of the lymphocyte by a given antigen is localized in the perinuclear rough endoplasmic reticulum structures. Even in reticulocytes it seems that different classes of proteins are synthesized on membrane-bound and free ribosomes (Bulova and Burka, 1970). The suggestion of a seggregation by membrane attachment of different populations of ribosomes carrying out the synthesis of different classes of proteins may not be as far-fetched as it may seem (see Tata, 1971b).

V. The Role of DNA Synthesis The role of DNA synthesis is quite obvious in those tissues that are rapidly formed during metamorphosis, such as the limbs and lungs. However, the possible role of a restricted DNA synthesis is not clear in those tissues that do not grow rapidly but undergo a functional maturation, i.e., the tadpole liver and skin. That a relatively small burst of DNA synthesis during metamorphic maturation may be important and needs careful investigation is emphasized by the work of Topper’s group on a limited DNA synthesis or cell division as a prerequisite for milk protein synthesis in prolactin-induced development of mammary gland in culture (Lockwood et al., 1967). McGarry and Vanable (1969) have indeed found that cell division is a prerequisite for thyroxine-induced maturational changes in skin gland from Xenopus in organ culture. Ingram’s (Moss and Ingram, 1968) work on the switch from fetal to adult hemoglobin during bullfrog metamorphosis suggests that DNA synthesis is important in that a different population of nucleated erythrocytes now takes over the synthesis of hemoglobin for terrestrial life. It is thought that the switching itself takes place in tadpole liver. A different kind of involvement of DNA synthesis may underlie the observations of A. M. Campbell et al. (1969) that triiodothyronine stimulated mitochondrial DNA polymerase in bullfrog tadpole liver. It is, however, difficult to say whether an early alteration in cytoplasmic DNA metabolism is part of some general developmental process or that it merely is a first stage of an increase in mitochondrial size and number that is known to occur during metamorphosis (see Tata, 1967a; Cohen, 1970; Bennett et al., 1970; Bennett and Glynn, 1970).

100

J. R. TATA

VI. Requirement of RNA and Protein Synthesis for Tissue Resorption

Tissue regression or cell death is an important and integral part of many embryonic developmental processes (Saunders, 1966) . During amphibian metamorphosis, regression of organs like the tail, gut, and gills accompany the maturation or formation of organs like the liver, limbs, and eyes. The biochemical basis of amphibian tail resorption has been studied in detail in Frieden’s (1967; Frieden and Just, 1970) and Weber’s (1963, 1969) laboratories, and these workers have established that the increase in activities of many hydrolases as the basis for the regression. A question that had not been answered until recently was whether or not regression or cell death was also based on a hormonal regulation of biosynthetic processes as we have seen above for maturation of the liver. We studied this question in our laboratory by using the technique of thyroid hormone-induced regression of the isolated tadpole tail maintained in organ culture (Tata, 1966). The technique of tail organ culture was already successfully used by Weber (1963) and the induction of regression in vitro corresponds in magnitude and speed to that seen in the intact tadpole (see Fig. 6 ) . Decreasc in the size of the amputated tail in vitro was accompanied by an increase in the activity of enzymes involved in regression such as cathepsin, phosphatases, and deoxyribonuclease (Fig. 7 ) . Earlier work on the comparison of the properties of deoxyribonuclease and cathcpsin in regressing and nonregressing tadpole tails had suggested that a part of the additional enzyme activity following hormone treatment may differ from the basal enzyme present in nonregressing tails (Coleman, 1962; Weber, 1963). It is therefore interesting to note in Fig. 7 that there is a burst of both RNA and protein synthesis just when regression sets in in vitro. Recently, Tonoue and Frieden (1970) have found that in vivo, administration of triiodothyronine to bullfrog tadpoles rapidly (within 1-3 hours) lead to a decrease in the incorporation of radioactive leucine in proteins of the tail and other regressing tissues. This apparently opposite result from that found in organ cultures may be a manifestation of different properties of thyroid hormone. Besides the fact that an enhanced rate of protein synthesis takes 1-2 days to be manifested, the rapid decrease in labeling of proteins may be due to changes in the precursor pool sizes although these authors seem to rule it out on the basis of indirect observations. Not only thyroid hormones (see Tata, 1968a, 1970c) but all hormones, whether they affect growth and development or cause a rapid change in metabolic activity, are known to cause rapid alterations in pool size or uptake of sugars, amino acids and nucleotides in their target tissues (Riggs, 1964; Manchester, 1970; Young; 1970).

3.

1\IIETAMORPHOSIS

101

FIG.6. Triiodothyronine (Td-induced regrcssion of amputated tails of Rana temporaria in organ culture. (a) Control samples on the first day of culture; (b) control, after 8 days of culture; (c) on the fourth day of culture in medium containing 1 pg T3/ml; (d) on the eighth day of culture with TI; (e) the same as (c) but with 2.5 pg actinomycin D/ml; (f) as (d) but with actinomycin. See Tata (1966) for details.

102

J. R. TATA

Days of culture

FIG.7. Accumulation of cathepsin ( 0 ) and deoxyribonuclease (A)and burst of additional RNA and protein synthesis during regression of tadpole tails induced in organ culture with triiodothyronine added to the medium. ( 0-01,Incorporation of uridine-’H into RNA (12.5 hours after the addition of 10 $3); A-A, incorporation of “C-labeled amino acids into protein (8.8 hours after the addition of 1.6 pCi of a mixture of labeled algal amino acids). Incorporation on day 0 is the average value for controls; all other points refer to samples to which T, wm added. ....., Tail length of controls over the duration of the experiment; 0 - - 0,length of triiodothyronine-treated tails showing a marked onset of regression between the second and third day after culture. Curves comuiled from data of Tata (1966).

--

T o determine whether any of the additional RNA and protein synthesized a t the onset of regression induced in cultured tails was essential for the process itself, we turned to the use of inhibitors of RNA and protein synthesis. From these experiments, summarized in Tables I V and V, it was quite clear that inhibition of RNA synthesis with actinomycin D or of protein synthesis with puromycin or cycloheximide completely abolished the regression in tail organ cultures that was induced by triiodothyronine. It was also found that inhibition of protein synthesis abolished the hormone-induced increase in hydrolytic enzyme activities. Eeckhout (1966) has also noted, in a different species of the tadpole, that protein synthesis inhibitors (such as puromycin and cycloheximide) arrested tail regression in organ cultures, and Weber (1965) observed that in Xenopus tail regression was more sensitive than was limb generation to the administration of actinomycin D to the intact tadpole. It seems that tail regression, and perhaps even that of other tissues, is

TABLE I V

THEINCORPORATION OF URIDINE-~H INTO RNA AND THE

Treatment

None

BY NONREGRESSING ISOLATED TAILS EFFECT OF ACTINOMYCIN D asb

Time uridine-aH added (days after amputation)

Length of tail when ~ r i d i n e - ~added H (mm)

Total RNA-3H per tail (CPm)

Specific activity (cpm/mg RNA)

1 2

15.6 15.3 15.3 15.1

2,250 4,800 6,450 5,400

3,840 8,230 11,100 10,500

15.2 14.5 12.7 10.1

2,600 4,000 13,680 11,350

4,170 7,580 33,250 32,600

,.

4

4

Actinomycin D (5 pg/ml)

2 3

15.4 15.3

960 380

2,060 952

TI and actinomycin D

2 3

15.6 15.4

1,130 80

2,230 184

From Tata (1966).

* Regression was induced by TIadded to the medium a t the beginning of culture. Tails were washed with nonradioactive uridine, 12.5 hours after the addition of 10 pCi 3H-labeled uridine and then homogenized. Each value is an average of three determinations in duplicate with variations of +5-15%.

w

TABLE V

THEINCORPORATION

Treatment None

Puromycin (100 pg/ml)

TI and puromycin Cycloheximide (30 pg/ml)

Ta and cycloheximide

OF

0 I&

14C-LABELED AMINOACIDSI N T O PROTEIN B Y ISOL.4TED TAILS,WITH OR WITHOUT T J - I ~ ~ u c e ~ A N D THE EFFECT OF PUROMYCIN AND CYCLOHEXIMIDE~,~ REGRESSION Time of addition of 14C-labeled amino acids (days after amputation)

Length of tail when 14C-labeled amino acids added (mm)

Total 14Cprotein per tail (CPm)

Specific activity (cpm/mg protein)

1 2 3 4 1 2 3 4 2 3 2 3 2 3 2 3

15.0 14.9 14.8 14.7 15.2 12.4 10.2 8.4 15.5 15.5 15.4 15.2 14.7 14.7 14.8 14.6

2,150 6,000 5,700 2,150 7,800 32,100 15,270 10,350 1,100 3,020 1,290 2,870 3,800 2,150 2,490 2,200

1,349 5,080 4,860 2,040 4,750 19,900 13,520 12,800 600 1,640 787 2,010 2,390 1,410 1,140 1,500

From Tata (1966). Triiodothyronine (T3) and the inhibitors were present, where indicated, from the beginning of the experiment. 14C-labeledamino acids (1.6 pCi) were added at different times during the culture, and the tails were withdrawn and washed with nonradioactive casein hydrolyzate 8.8 hours after the addition of the isotope. Each value is an average of 3 determinations with variations in the range of 10-20%. a

b

*

?

F c3

3.

METAMORPHOSIS

105

not brought about by a direct hormonal activation of lysosomal enzymes but by the formation of a new population of hydrolase molecules. Thus, not only cell growth and maturation, but also cell death during development may require a genetically determined synthesis of specific proteins. VII. Conclusions and Future Problems

The above account reemphasizes the advantages of studying biochemical processes in amphibian metamorphosis as a model for studying regulation in late embryonic development. Not only are the embryos free-living but the artificial induction of metamorphosis by thyroid hormones a t developmental stages well before spontaneous metamorphosis have made it possible to establish a sequence of events preceding the acquisition of adult functions and structures. Among the earliest responses of target cells is a readjustment of permeability barriers t o a variety of nutrients and precursors of macromolecules. However, the eventual specificity of developmental changes is most likely to reside in the nature of new or additional species of RNA and protein molecules formed. Due to a combination of the complexity of nuclear RNA, the high degree of nuclear restriction of the rapidly turning over nuclear RNA and the inadequacies of the currently available analytical techniques it has not been possible to relate the burst of RNA synthesis with selective gene activation a t the onset of metamorphosis. I n the liver of the bullfrog tadpole a relatively long period of time elapses between an accelerated synthesis of RNA in the nucleus and the appearance of new proteins that characterize metamorphic change (compare Figs. 1 and 2 ) . During this period there occurs a substantial increase in the turnover of ribosomes in the cytoplasm accompanied by their redistribution on membranes of the endoplasmic reticulum. At the same time there is a tight coordination in the rate of formation of ribosomes and the proliferation of membranes of the endoplasmic reticulum to which they are bound. This phenomenon of structural reorganization of the protein synthesizing machinery is common to many late embryonic developmental systems, and i k significance may reside in a topographical segregation of precoded polysomes engaged in the synthesis of different classes of proteins (see Tata, 1971b). A prominent feature of amphibian metamorphosis is the convenience of studying tissue regression or cell death. It seems that regression is not merely due to activation of existing lysosomes but that an activation of RNA and protein synthesis underlies the process of regression. Experiments with inhibitors of RNA and protein synthesis in the resorption of tadpole tails in culture have shown that cell death during meta-

106

J. R. TATA

morphosis requires the formation of new proteins just as it is necessary for those cells that are programmed for further growth and development. Thus it is now possible to describe sequential phenomena concerning the synthesis of RNA and protein a t the onset of metamorphic maturation or growth. But there are several questions still to be resolved. For example, the nature of RNA made, its transfer into the cytoplasm, and its role as a messenger will have to be clarified-a problem now facing almost all work on differentiation. Are new genes really transcribed or is there a switch in the selection and transfer to the cytoplasm of the types of RNA molecules which are constantly made since an early stage in differentiation? In the cytoplasm, it will be most worthwhile exploring the possibility, by a combination of biochemical, histochemical, and immunochemical techniques, of identifying a separate class of ribosomes preferentially satisfying the demand for metamorphic proteins. The fact that cell growth and cell death proceed simultaneously within the same organism creates problems of a flux of precursor pools which is impossible to control. It is obvious that more emphasis will now have to be devoted t o studying late embryonic developmental process in culture. Organ cultures of tadpole tail, skin, and liver have already been studied, and it would be useful to extend these studies to dispersed cell cultures, providing that cultured cells retain the full developmental competence to respond to the hormonal stimulus. Finally, the question of developmental competence is related to that of the initial site of thyroid hormone action, i.e. the hormone receptors. Some recent experiments from our laboratory have shown that in Xenopus larvae metamorphic competence was acquired very early (Nieuwkoop-Faber stages 38-45) in development (Tata, 196813). Metamorphic competence was assessed biochemically by a change in the overall rate of synthesis of DNA, RNA, phospholipids and protein, water loss, and altered permeability to anions like phosphate. These changes indicate that the cells (undetermined) which were initially unresponsive have suddenly become sensitive to thyroid hormones. It should be realized, however, that in normal development there is a variation in the magnitude of sensitivity of different tissues as a function of the age of the tadpole. One interpretation of such experiments was that the acquisition of sensitivity to the hormone was a consequence of the first appearance of hormone receptors. Thus when the capacity of Xenopus larvae to bind radioactive thyroid hormones was measured, there was an excellent correlation between high affinity (average Kd = 10-0 to l0-lo) thyroxine-binding sites and the rate a t which the organism becomes hormone-sensitive (see Fig. 8 ; Tata, 1970d). It has yet to be shown that one is dealing not with nonspecific binding, but

3.

METAMORPHOSIS

107

;

Days after ferlilizalion

FIG. 8. Correlation between the appearance of temperature-sensitive binding capacity for triiodothyronine ( 0 ) and the acquisition of a metamorphic response to thyroid hormone by developing Xeriopus larvae. The temperature-sensitive binding cornponcnt(s) was calculated by subtracting the binding capacity for triiodothyronine at 25" from that expressed at 5 " . The metamorphic response to triiodothyronine is illustrated for the increase in the rate of RNA synthesis ( A ) or the diminution of uptake of PO?- ions (0) when Xenopus larvae a t different stages of development are exposed to lo-' M triiodothyronine. Data from Tata (196813, 1970d).

with true receptors whose interaction with the hormone would lead to the normal chain of events responsible for metamorphic changes. A selective rather than random distribution of hormone-binding sites has also to be determined. I n the meantime, however, the study of coordinated acquisition of metamorphic competence and hormone-binding offers a developmental approach to the problem of hormone receptors which is now a key issue in understanding the biochemical basis of hormonally regulated processes. REFERENCES Allen, B. M. (1916). Science 44, 755. Andrews, T. M., and Tata, J. R. (1968). Biochem. Biophys. Res. Comm. 32, 1050. Andrews, T. M., and Tata, J. R. (1971). Biochem. J. 121, Arias, I. M., Doyle, D., and Schimke, R. T. (1969). J. Biol. Chem. 244, 3303. Balinsky, J. B., Shambaugh, G. E., 111, and Cohen, P. P. (1970). J. Biol. Chem. 245, 128. Bennett, T. P., and Frieden, E. (1962). Comp. Biochem. 4, 483. Bennett, T. P., and Glenn, J. C. (1970). Develop. Biol. 22, 535.

108

J. R. TATA

Bennett, T. P., Glenn, J. C., and Sheldon, H. (11970). Develop. Biol. 22, 232. Berman, R., Bern, H. A., Nicoll, C. S., and Strohman, R. C. (1964). J. Exp. 2001. 156, 353. Bern, H. A., and Nicoll, C. S. (1969). Recent Progr. Horm. Res. 24, 681. Bern, H. A., Nicoll, C. S., and Strohman, R. C. (1967). Proc. SOC. Exp. B w l . Med. 126, 518. Billing, J., and Barbiroli, B. (1970). Biochim. Biophys. Acta 217,434. Blatt, L. M., Kim, K.-H., and Cohen, P. P. (1969). J. Biol. Chem. 244, 4801. Britten, R. J., and Davidson, E. H. (1969). Science 165, 349. Bulova, S. I., and Burka, E. R. (1970). J. Biol. Chem. 245, 4907. Campbell, A. M., Corrance, M. H., Davidson, J. N., and Keir, H. M. (1969). Proc. R o y . SOC.Edinburgh, Sect. B 70, 295. Campbell, P. N. (1970). FEBS L e t t . 7 , 1 . Chefurka, W., Yapo, A., and Nisman, B. (1970). Can. J. Biochem. 48, 893. Cohen, P. P. (1966). Harvey Lect. 60, 119. Cohen, P. P. (1970). Science 168, 533. Coleman, J. R. (1962). Develop. Biol. 5, 232. Eaton, J. E., and Frieden, E. (1969). Gen. Comp. Endocrinol. 5, 43. Eeckhout, Y. (1966). R e v . Quest. Sci. 3,377. Etkin, W. (1964). In “Physiology of the Amphibia” (J. A. Moore, ed.), pp. 427. Academic Press, New York. Etkin, W. (1968). In “Metamorphosis” (W. Etkin and L. I. Gilbert, eds.), pp. 313-348. Appleton, New York. Etkin, W., and Gona, A. G. (1967). J. Exp. 2001.165, 249. Frieden, E. (1967). Recent Progr. Horm. Res. 23, 139. Frieden, E., and Just, J. J. (1970). I n “Biochemical Actions of Hormones” (G. Litwack, ed.), Vol. 1, p. 1. Academic Press, New York. Georgiev, G. P. (1966). Progr. Nucl. Acid Res. Mol. Biol. 6,259. Gross, J. (1964). Medicine (Baltimore) 43, 291. Gudernatsch, J. F. (1912). Arch. Entwicklungs mech. Organismen 35, 457. Hamilton, T. H. (1968). Science 161, 649. Harris, H. (1964). Nature (London) 202, 249. Hendler, R. W. (1968). “Protein Biosynthesis and Membrane Biochemistry.” Wiley, New York. Henshaw, E. C. (1968). J. Mol. Biol. 36, 401. Karlson, P. (1963). Angew. Chem. 2, 175. Kijima, S., and Wilt, F. H. (1969). J. Mol. Biol. 40, 235. Kim, K. H., and Cohen, P. P. (1966). Proc. Nut. Acad. Sci. U.S. 55, 1251. Kim, K. H., and Cohen, P. P. (1968). Biochim. Biophys. Acta 166, 574. Kollros, J. J. (1942). J. Exp. Zool. 89, 37. Kollros, J. J. (1968). Growth N e w . Syst., Ciba Found. Symp. pp. 179-192. Korner, A. (1970). I n “Control Processes in Multicellular Organisms” (G. E. W. Wolstenholme and J. Knight, eds.), pp. 86-99. Churchill, London. Leduc, E. H., Avrameas, S., and Bouteille, M. (1968). J. Exp. Med. 127, 109. Lockwood, D. H., Stockdale, F. E., and Topper, Y. J. (1967). Science 156, 945. McGarry, M. P., and Vanable, J. W., Jr. (1969). Develop. Biol. 20, 291. Manchester, K. L. (1970). Biochem. J. 117, 457. Metaenberg, R. L., Marshall, M., Paik, W. K., and Cohen, P. P. (1961). J. Biol. Chem. 236, 162. Moss, B., and Ingram, V. M. (1968). J. Mol. Biol. 32, 493.

3.

METAMORPHOSIS

109

Nakagawa, H., and Cohen, P. P. (1967). J. Biol. Chem. 242, 642. Nakagawa, H., Kim, K. H., and Cohen, P . P. (1967). J. Biol. Chem. 242, 635. Ohtsu, K., Naito, K., and Wilt, F. H. (1964). DeveEop. Biol. 10, 216. Olsnes, S. (1970). Eur. J . Biochem. 15, 464. Paik, W. K., and Cohen, P. P. (1960). J. Gen. qhysiol. 43, 683. Palade, G. E. (1966). J. Amer. M e d . Ass. 198, 815. Penman, S., Rosbasli, M., and Penman, M. (1970). Proc. Nnt. Acrid. Sci. U S . 67, 1878. Pitot, H. C., Sladek, N., Ragland, W., Murray, R. K., Moyer, G., Soling, H. D., and Jost, J.-P. (1969). In “Microsomes and Drug Oxidations” (J. R. Gillette et al., eds.), p. 59. Academic Press, New York. Priestley, G. C., Pruyn, M. L., and Malt, R. A. (1968). Biochim. Biophys. Acta 190, 154. Riggs, T . R. (196.1). I n “Actions of Hormones on Molecular Processes” (G. Litwack and D. Kritchevsky,eds.), p. 1. Wiley, New York. Samarina, 0. P., Lukanidin, E. M., Milnar, J., and Georgiev, G. P. (1968). J. Mol. Biol. 33, 251. San Pietro, A., Lamborg, M. R., and Kenney, F. T., eds. (1968). “Regulatory Mechanisms for Protein Synthesis in Mammalian Cells” (see various authors). Academic Press, New York. Saunders, J . W., J r. (1966). Science 154, 604. SaxBn, L., SaxCn, E., Toivonen, S., and Salimaki, K. (1957). Endocrinology 61, 35. Scherrer, K., and Marcaud, L. (1969). J. Cell Physiol., Suppl. 1, 181. Schlessinger, D. (1963). J. Mol. Biol. 7, 569. Schneiderman, H. A., and Gilbert, L. I. (1964). Science 143, 325. Shambaugh, G. E., 111, Balinsky, J. B., and Cohcn, P. P. (1969). J. Biol. Chem. 244, 5295. Shearer, R . W., and McCarthy, B. J. (1967). Biochemistry 6, 283. Siekevits, P., Palade, G. E., Dallner, G., Ohnd. I., nnd Omura, T. (1967). I n “Organizational Biosynthesis” (H. J. Vogel, J. 0. Lampcn, and V. Bryson, eds.), p. 331, Academic Press, New York. Spirin, A. S. (1969). Eur. J . Biochem. 10, 20. Spohr, G., Granboulan, N., Morel, C., and Scherrer, K. (1970). Eur. J. Biochem. 17, 296. Tata, J . R. (1965). Nature (London) 207, 378. Tata, J. R. (1966). Develop. Biol. 13, 77. Tata, J. R. (1967a). Biochem. J. 105, 783. Tata, J. R. (196713). Biochem. J . 104, 1. Tata, J. R. (1968a). Nature (Londor~)219, 331. Tata, J. R. (196813). Develop. Biol. 18, 415. Tata, J. R. (1970a). In “Control Processes in Multicellular Organisms” (G. E. w. Wolstenholme and J. Knight, eds.), pp. 131-150. Churchill, London. Tata, J. R. (1970b). Biochem. J . 116, 617. Tata, J. R. (1970~).I n “Biochemical Actions of Hormones” (G. Litwack, ed.), Vol. 1, p. 89. Academic Press, New York. Tata, J. R. (1970d). Nature (London) 227, 686. Tata, J . R . (1971a). Symp. Soc. Exp. Biol. 25, 163. Tata, J. R. (1971b). Subcellular Biochem. 1, 83. Tomkins, G. M., Gelehlter, T. D., Granner, D. M., Jr., Samuels, H. H.,

110

J. R. TATA

and Thompson, E. B. (1969). Science 166, 1474. Tonoue, T., and Frieden, E. (1970). J . Biol. Chem. 245,2359. Tonoue, T., Eaton, J. E., and Frieden, E. (1969). Biochem. Biophys. Res. Commun. 37, 81. Unsworth, B. R., and Cohen, P. P. (1968). Biochemistry 7, 2581. Weber, R. (1963). Lysosomes Ciba Found. Symp., 1963 pp. 282-300. Weber, R. (1965). Experientia 21, 665. Weber, R. (1967). I n “The Biochemistry of Animal Development” (R. Weber, ed.), Vol. 2, pp. 227-301. Academic Press, New York. Weber, R. (1969). I n “Lysosomes in Biology and Pathology” (J. T. Dingle and H. B. Fell, eds.), pp. 437-461. North-Holland Publ., Amsterdam. Weiss, P., and Rossetti, F. (1951). Proc. Nat. Acad. Sci. US.37, 540. Williams, C. M. (1961). Science 133, 1370. Wilt, F. H. (1959). Develop. Biol. 1, 199. Wolstenholme, G. E. W., and J. Knight, eds. (1970). “Control Processes in Multicellular Organisms” (see various authors). Churchill, London. Wool, I. G., Stirwalt, W. S., Kurihara, K., Low, R. B., Bailey, P., and Oyer, D. (1968). Recent Progr. Horm. Res. 24, 139. Wyatt, G. R. (1962). I n “Insect Physiology,” p. 23. Oregon State Univ. Press, Wyatt, G. R., and Tata, J. R. (1968). Biochem. J . 109, 253. Young, D. A. (1970). J . Biol. Chem. 245, 2747.

CHAPTER 4

HORMONAL CONTROL OF A SECRETORY TISSUE* H . Yomo and J. E. Varner MSU/AEC

PLANT RESEARCH LABORATORY AND DEPARTMENT OF BIOCHEMISTRY

MICHIGAN STATE UNIVERSITY EAST LANSING, MICHIGAN

The aleurone layer of the seeds of Gramineae is a secretory tissue surrounding the endosperm reserves of the seed (Figs. 1-3). During germination, gibberellins from the embryo induce the aleurone layer cells to synthesize and secrete several hydrolases, including amylases and proteases, which hydrolyze the reserve starch (Fig. 4) t o maltose and glucose and the reserve proteins to amino acids. These end products of hydrolysis are transported to the growing regions of the embryo (Fig. 5 ) by way of the scutellum, the absorptive organ of the embryo (Fig. 5 ) . ’ With the exception of the role of the gibberellins, the secretory function of the aleurone layers was understood by Haberlandt as early as 1890 (Table I ) , confirmed a fcw years later (Brown and Escombe, 1898), and reconfirmed many years later by several investigators (Gruss, 1928 ; Schander, 1934; MacLeod and Millar, 1962; Briggs, 1963; Paleg, 1964; Yomo and Iinuma, 1964; Varner and Chandra, 1964). I n 1960 it was shown that the factor from the embryo (Table I) (Haberlandt, 1890; Kirsop and Pollock, 1958; Yomo, 1958) necessary for activation of the secretory function of the aleurone layers was gibberellin-like (Table 11) (Yomo, 1960a,b) and that this factor could in fact be replaced by gibberellic acid (Table 111) (Yomo, 1960c; Paleg, 1960). Furthermore, the aleurone layer stripped free of the starchy endosperm (Fig. 2) responds to added gibberellins by producing and secreting those same amylases and proteases (Briggs, 1963; Paleg, 1964; Yomo and Iinuma, 1964; Varncr, 1964). I n the intact germinating seed the embryo-both the scutellum (Fig. 5) (Radley, 1967, 1969) and the nodal region (Fig. 5 ) (MacLeod and Palmer, 1966) -synthesizes (Yomo and Iinuma, 1966) the gibberellins,

* This work was supported in part by the United States Atomic Energy Commission under Contract AT(11-1)-1338 and in part by a grant from the National Science Foundation (GB-8774). 111

112

H . YOMO AND J. E. VARNER

FIG.1. Left: Barley (Hordeum vulgare, cv. Himalaya) grain, Right: Endosperm half or half-seed. Photo by K. Raschke and J. E. Varner. TABLE I DIASTASEPRODUCTION BY ALEURONE LAYERSOF Secale cerealea 1. Liquefaction of the starchy endosperm starts next to the scutellum and the entire aleurone layer and proceeds inward. 2. Pieces of seedcoat with aleurone layer taken from germinating seeds contain diastase, which corrodes starch grains. 3. The diastase of the aleurone layers does not originate in the scutellum. 4. .4 viable and growing embryo is necessary for the ,production of diastase by the aleurone layers. a

From Haberlandt (1890).

probably GA, and GA, (Fig. 6 ) , that control the aleurone response. This is appropriate and logical, because it is the embryo that gains from an ordered mobilization of the endosperm reserves. Thus the cereal grain provides a convenient system for the study of the plant’s use of a hormone to integrate the various functions of the different tissues and organs.

4. HORMONAL

CONTROL O F A SECRETORY TISSUE

113

FIG.2. Alenrone layer partially peeled away from the starchy endosperm of a fully imbibed half-seed. Photo by K. Raschke and J. E. Varner.

FIG.3. Longitudinal section of n dry barley grain. A, Embryo; B, aleurone layers; C, starchy endosperm. Fixation and sectioning by Linda Franaen; photo by Me1 Dickerson.

114

H. YOMO AND J. E. VARNER

FIG.4. Half-seed 3 days after imbitition of 1 /AM gibberellic acid (GAA.

TABLE I1 PROPERTIES OF AMYLASEACTIVATING SUBSTANCE^ Not present in ungerminated embryo Produced by germinating embryo Dialyzable Destroyed by acid and by alkali Absorbed on charcoal Absorbed on anion exchange resin Soluble in organic solvents Has one -OH (infrared spectrum) Has two -C=O (infrared spectrum) R, values same as for gibberellins Is gibberellin or gibberellin-like 0

From Yomo (1958; 196Oa,b,c).

4.

HORMONAL CONTROL O F A SECRETORY T I S S U E

115

FIG.5. Barley embryo vertical section. N, nodal region; S, scutellum; A, acrospire; R, rootlet; C, coleorhiza; M, micropyle. From MacLeod and Palmer (1966). TABLE I11 INFLUENCE OF GIBBERELLIN CONCENTRATION’ Amylases

0 0.001 0.01 0.1 1 5 50

D. P.”

S.P.c

17 200

730 920 1050 1070 1010 1220 1250

640

900 790 760 730

a

From Yomo (1960~).

c

Saccharifying Power.

* Dkstatic Power.

116

H. YOMO AND J. E. VARNER

HO

@

H'

,

COOH

CH,

H2

k

FIG.6. Structural formulas for GA, and GA3.

250

2

half - seeds

200

Q

Hours

FIG.7. Time course for the development of a-amylase activity in half-seeds and in isolated aleurone layers. The fresh weight of ten half-aleurone layers is 168 mg. GA3 (1 p M ) was added at zero time. From Varner and Chandra (1964).

Hours of incubation

FIG. 8. The time course of release of a-amylase and protease by aleurone layers in the presence of 1 p M GA3. From Jacobsen and Varner (1967).

4.

HORMONAL CONTROL O F A SECRETORY TISSUE

-E

3.0

250

117

Lo

a A

L

-

%

0

200 5

e e

150

>

._

-z

100: .

C

1.0

3

v

%

a

50

c

2 a

E

a 0

log Concentration of gibberellic acid

FIG.9. The release of &-amylase and protease by aleurone layers in response to various concentrations of GA,. From Jacobsen and Varner (1967).

TABLE I V EFFECTOF CALCIUM AND OTHERDIVALENT IONS ON THE PRODUCTION OF PAMYLASE BY ALEURONE LAYERS~~~ a-Amylase per 10 aleurone layers Treatment Control 0 . 1mM 1 mM 10 mM 20 m M 100 m M 10 m M 10 m M 10 mM 10 m M

Calf Ca2+ CaZ+ Ca*+ Ca*+ Sr2+

Mgz+ Ba2+ CdZ+

(PP)

69 74 198 293 328 335 223 42 70 2

From Chrispeels and Varner (1967a). Ten half aleurone layers were incubated with buffer and 1 p M GAI and the appropriate concentration of the salt. a-Amylase was assayed in the medium and the tissue extract after 24 hours of incubation.

118

H. YOMO AND J. E. VARNER

A few hours after normal germination the embryo begins to synthesize gibberellins (Yomo and Iinuma, 1966). These move by way of the scutellum to the aleurone layers-a target tissue-which are composed entirely of a single cell type. Because the aleurone layers are readily stripped away from the starchy endosperm (Fig. 2) they are suitable for study of the action of gibberellic acid a t the cellular and subcellular levels.

Fraction number

FIG.10. Upper: DEAE chromatograms of the medium from 10 half-seeds incubated with 10 pCi of ~-phenylalanine-C'~in 1 F M GAS. Lower: DEAE chromatograms of the extract from the same 10 hnlf-seeds. The filled circles indicate counts per minute per fraction. The open circles indicate 0-amylase values (in arbitrary units). From Varner (1964). The response of aleurone layers to added gibberellins can be studied by incubating half-seeds (Fig. 1) or aleurone layers (Fig. 2) in a suitable buffer solution containing the hormone. Synthesis and secretion of a-amylase (Figs. 7 and 8) and protease (Fig. 8) begins 8-10 hours after the addition of the hormone. I n the absence of added hormone little or no amylase and protease are produced (Fig. 9 ) . Over a 24-hour period, the total amylase and protease produced is proportional to the logarithm of the concentration of gibberellic acid applied from a concentration

4. HORMONAL

119

CONTROL O F A SECRETORY TISSUE

M (Fig. 9 ) . For maximum accumulation of a-amylase, of M to calcium ions must be added to the incubation medium to protect the amylase from attack by the protease (Table I V ) . Maximum production of amylase in response to added gibberellic acid also requires that the half-seeds or aleurone layers be well aerated during the incubation (Table V) . Because production of the hydrolases is inhibited by dinitrophenol, p-fluorophenylalanine, and cycloheximide (Tables V and VI) , TABLE V FACTORS AFFECTINGPAMYLASE FORMATION^^^

Treatment Control Control (anaerobic) - Gibberellic acid Dinitrophenol 10-8 M 10-4 M 10-6 M p-Fluorophenylalanine

+

+

+

M 10-6 M 10-6 M Chloramphenicol (1 mg/ml)

Medium (PP)

Extract (PP)

Total

80 5 0 0 3 24 28 28 82 10

65 8 4 8 12 24 26 47 80 46

145 13 4 8 15 48 54 75 162 56

(PP)

From Varner (1964). The control incubation medium contained 1 m M sodium acetate (pH 4.8) and 1 pM GA3. b

TABLE VI L-LEUCINE-14c INCORPORATION AND AMYLASE FORMATION BY CYCLOHEXIMIDE AND p-FLUOROPHENYLALANINEn INHIBITION O F

Incorporation (CPd Conditionsb Control GA3 GA, p-fluorophenylalanine GA, cycloheximide

+ +

I n extract 5200 3200 4320 560

I n medium 600 1880 360 100

a-Am ylase (pg) 18 126 27 12

From Varner et al. (1965). half-seeds (10) were further incubated for 24 hours with GAa, 46 pCi of ~leucine-14Cin 5 x 10-3 M carrier Lleucine, and 4 X M p-fluorophenylalanine or 5 pg/ml of cycloheximide where indicated. The numbers shown are the averages of duplicates.

* Preincubated

120

H. YOMO AND J. E. VARNER

it is clear that phosphorylative energy and protein synthesis are required for the appearance of the hydrolase activities. I n fact, a-amylase becomes radioactively labeled when i t is produced by layers incubated in I4C-labeled amino acids (Fig. l o ) , and i t becomes density labeled when

4

I00 E

a 0

50

I00

$\

B

44

E

0

50 b

-P FIG.11. ( A ) Equilibrium distributions of radioactivity (-@-) and a-amylase activity (-O-) after centrifugation of a mixture of 2 pg of purified aamyla~e-'~O~H and ca. 25 pg of crude a-amylase induced in H21B0in 3.0 ml of CsCl solution, p = 1.30, at 39,000 rpm in a Spinco SW 39 rotor for 65 hours a t 4°C. Density increases to the left. (B) Equilibrium distribution of radioactivity (-@-) and a-amylase enzyme activity (-O--) after centrifugation of a mixture of 2 pg of purified a-amylase-160SH and ca. 25 pg of crude a-amylase induced in Hz'*O. Centrifugation conditions are the same as for ( A ) . From Filner and Varner (1967).

the layers are incubated in the presence of H,lSO (Fig. 11). The density label could be introduced into amylase only by the path: Reserve proteins

Ha180

180-amino acids -i180-amylase

Therefore all the observed a-amylase activity is synthesized de novo in response to added gibberellic acid, as is the protease (Fig. 12). The aleurone layers synthesize a t least four a-amylase isozymes (Fig. 13). The synthesis of a-amylase by the aleurone layers can be stopped by removal of the gibberellic acid from the incubation medium and started again by adding it back to the medium (Fig. 14, left). Amylase

4. HORJIONAL

CONTROL OF A SECRETORY TISSUE

121

I00 ," 80 ._ 5

60 a -5

40

4-

0 -

20

I00 + 80 ._ 5 ._

2 60 aY

.z c 40

-0 B

20

Fraction number

FIG.12. The distribution of protcnsc released by aleurone layers in the presence of H,'"O and HPO after equilibrium centrifugation on cesium chloride gradients. TritiiEtcd amylase was used as the rcfcrencc. Other conditions of centrifugation were as in Fig. 11. From Jncobsen and Varncr (1967). TABLE VII INHIBITION OF (Y-AMYLASE PRODUCTION BY ACTINOMYCIN

Treatment Control Actinomycin D 20 /.lg/ml 50 a / m l 100 ~ u e :in1

Micrograms per 10 aleurone layers

% Inhibition of incorporation % Inhibition of uridineJ4C

359

-

-

332 241 152

7 33 58

40

63 75

From Chrispeels and Varner (1967a). Samples of 10 aleurone layers were incubated with buffer, 1 ~ c r MGA3 and 20 m M CaC12. Appropriate amounts of actinomycin D were added Et the same time as GAa. The enzyme was assayed after 24 hours of incubation. I n a parallel experiment, aleurone layers were incubated without GA3 and allowed to incorporate uridine-14C (1 pCi per flask specific activity 26 mCi/mmole) between 4 and 8 hours after the addition of actinomycin D. RNA was extracted by the method of Kirby. a

122

H. YOMO AND J. E. VARNER

FIG. 13. Incorporation of tritiated leucine into the &-amylase isozymes originating in the aleurone layers of barley in response to GA,. Lower: Agar gel electrophoretic separation of amylases produced in the presence of leucine-3H. Upper : Distribution of tritium in thin sections of the agar gel. From Jacobsen et al. (1970). TABLE VIII WAMYLASE SYNTHESIS A N D u R I D I N E - l 4 c INCORPORATION BY ACTINOMYCIN D ADDED4 A N D 8 HOURSAFTER GA35sb

INHIBITION O F

Treatment G-43 GA3 GAS

+ actinomycin D after 4 hours + actinomycin D after 8 hours

a-Amylase per 10 Uridine-14C aleurone layers incorporated (rg) (CPd 359 212 325

1380 466

From Chrispeels and Varner (196710). Ten aleurone layers were incubated in 0.1 p.M GA3, and after 4 or 8 hours actinomycin D (100 pglml) was added. Enzyme synthesis was measured at the end of the 24-hour incubation period. UridineJ4C (1 pCi per flask) was added 4 hours after actinomycin D and incorporation was allowed to proceed over a 4-hour period. a

4. HORMONAL CONTROL OF

123

A SECRETORY TISSUE

synthesis can also be stopped a t any time by the addition of cycloheximide (Fig. 14, right) or by the addition of the plant hormone abscisic acid (Figs. 14-16). Thus the synthesis of a-amylase can be closely controlled by the levels of these two hormones. The physiological role of abscisic acid, if any, during normal germinaltion of barely seed is not known. ~

in

~~~

300

-0

E"

7U

Control

200

.4-

0

* I00

rn

Cycloheximide 0

8

16

24

8

16

24

Hours of incubation

FIQ. 14. Left. Effect of removing GA, at the end of the lag period. Aleurone layers were incubated for 7 hours in 0.5 p M GG. G G was then removed by 4 consecutive 0.5-hour rinses. The aleurone layers were further incubated either with GAS, without GA,, or with GA, added again at 15 hours. Total a-amylase synthesis was measured 15 and 23 hours after the start of incubation. Right: Mid-course inhibition of a-amylase synthesis by abscisic acid (abscisin) and cycloheximide. Aleurone layers were incubated in 0.1 p M GA, for 11 hours. At this time abscisic acid ( 5 p M ) or cycloheximide (10 pg/ml) was added and a-amylase synthesis W B S measured 2.5, 5, and 10 hours later. From Chrispeels and Varner (196713).

FIG.15. S-Abscisic acid (formerly abscisin).

At what level-transcription, translation, or other-do the two hormones exert their control of the aleurone cells? Actinomycin D is effective in preventing a-amylase synthesis only if added a t high concentrations (Table VII) early in the course of the response (Table VIII) of the aleurone cells to gibberellic acid. At lower concentrations, or when added several hours after the addition of gibberellic acid, actinomycin D in-

124

H. YOMO AND J. E. VARNER

150-

-t

El00 -

a I

U +

' 0

50-

L,

'

II

15

19

Hours of incubation

FIQ. 16. Mid-course inhibition of a-amylase synthesis by abscisic acid, 6-methylpurine, and 8-azaguanine. Alcurone layers were incubated in 0.05 I.LMGAS for 11 hours. At this time the medium was removed, the aleurone layers were rinsed, or with GAS and 10 and they were further incubated wit,h 0.05 pM GAS (.-a), 5 mM 6-mechylpurine (O-n), 0.5 mM 6-methylpurine pM abscisic acid (0-O), (M-M), or 5 d 8-asaguanine (A-A), From Chrispeels and Varner (196713).

TABLE IX INHIBITION OF WAMYLASE SECRETION BY ACTINOMYCIN Da-* a-Amylase per 10 aleurone layers

Treatment Control Actinomycin D, 25 pg/ml Actinomycin D, 50 rg/ml

Medium

Extract

(rd

(PP)

Total (rg)

324 176 108

79 197 194

403 373 302

From Chrispeels and Varner (1967a). Details the same as in Table VII. Ten aleurone layers were incubated in 0.1 pM GA, and 25 or 50 r g of actinomycin D per milliliter. Enzyme synthesis and secretion were measured a t the end of the 24-hour incubation period. 0

4.

HORMONAL CONTROL OF A SECRETORY TISSUE

125

TABLE X INHIBITION OF WAMYLASE SYNTHESIS BY

6-METHYLPURINE"'b a-Amylase per 10 aleurone layers

Treatment, time of addition of 6-methylpurine (hours) 0 4

a

Control

0.1 mM of 1.0 m M of 6-methylpurine 6-methylpurine (rg)

(pg)

38 115 208 384

9 55 140 426

From Chrispeels and Varner (1967b). Ten aleurone layers were incubated in buffer, 20 mM, CaCll and 0.5 p M GA,. 6-Methylpurine was added a t the same time as GA3or 4 or 8 hours later and total a-amylase production was measured after 24 hours of incubation. 0

FIG. 17. Light micrograph of imbibed aleurone tissue showing distribution of organelles within the cells. W, Cell wall; SC, seed coat; N, nucleus; AG, aleurone . Jones (1969a). grain; G, phytin globoid. ~ 1 7 0 0From

126

H. YOMO AND J. E. VARNER

FIG. 18. Electron micrograph of barley aleurone cell from half-seed imbibed on moist sand for 2 days. Note large alcurone grains with densely stained phytin globoid inclusions, spherosomes (small unstained bodies), proplastids, mitochondria, and occasional endoplasmic reticulum. KMnOl fixation. X5000. From Jones (1969b).

hibits secretion of a-amylase more than it inhibits synthesis of amylase (Table IX), and there is an accumulation of amylase inside the cells. However 6-methylpurine effectively inhibits amylase synthesis when added as late as 11 hours after the addition of gibberellic acid (Fig. 16, Table X ) . The incorporation of uridine-"C by the aleurone layers is inhibited 70% by 1 mM 6-methylpurine within 4 hours of the addition of the 6-methylpurine (Chrispeels and Varner, 1967b). The incorporation of leucine-14C by the aleurone layers into the cellular proteins is inhibited

4.

HORMONAL CONTROL O F A SECRETORY TISSUE

127

FIG. 19. Section of aleurone cell after treatment of the aleurone layer with 1 p M GAI for 24 hours. x5300. From Jones and Price (1970).

by only 30% 4 hours after the addition of 6-methylpurine, while a-amylase is completely inhibited (Chrispeels and Varner, 196713). If we were sure (but we are not) that the only effect of 6-methylpurine is to inhibit RNA synthesis, we might conclude that a-amylase synthesis is dependent on the continued synthesis of a relatively unstable RNA. At present

128

H. YOMO AND J. E. VARNER

FIG. 20. Low magnification view of an aleurone cell from behind the scutellum after 36 hours of germination. Lipid bodies (LB) and aleurone grains (AG) abound in this cell and occupy most of the cell volume, leaving little room for mitochondria ( M ) , microbodies ( M b ) , and plastids. However, segments and laminated arrays of rough endoplasmic reticulum (RER) are numerous, the latter being confined to the periphery of the cell. While polysomes (Ps) and isolated segments of R E R appear near the nuclear envelope, large patches of polysomes are best seen closer to the cell (CW). Glutaraldehyde/Osmium. ~3500.Vigil and Ruddat (1971a,b).

we can neither rule in nor rule out the possibility that gibberellic acid induces the transcription of certain kinds of RNA specific for the synthesis of a-amylase and for other hydrolases. Massive synthesis and secretion of hydrolases begins 8-10 hours after the cells of the isolated aleurone layers have been exposed to gibberellic acid. What is happcning in the cells during this “lag period”? During normal germination the internal structures of the aleurone

4.

HORJIONAL CONTROL OF A SECRETORY TISSUE

129

FIG.21. Enlargement of several stacks of rough endoplasmic reticulum showing the uniform parallel arrangement of individual cisternae of endoplasmic reticulum. Glutaraldeh~de/osniium. x 16,000 (Vigil and Ruddat, 1971a,h).

cells (Figs. 17 and 18) undergo marked changes including the development of rough endoplasmic reticulum (Van der Eb and Nieuwdorp, 1967; Jones 1969c) (Figs. 19-22), the disappearance of the protein matrix and the phytin globoids of the aleurone grains, segmentation of the rough endoplasmic reticulum (Fig. 23), gradual diminishment of the lipid-containing spherosomes, enlargement and fusion of the aleurone vacuoles (Fig. 19), and development of the cristae of the mitochondria and the enveloping membranes (Jones, 1969c; Vigil and Ruddat, 1971a,b).

130

H. YOMO AND J. E. VARNER

FIG.22. Surface view of rough endoplasmic reticulum illustrating the presence of numerous polysomes on the membrane. Glutaraldehyde/osmium. x35,OOO. Vigil and Ruddat (1971a,b). TABLE XI LOCATION OF ACID-PRECIPITABLE TRYPTOPHAN-3H I N BARLEYA L E U R O N E ~ . ~ Homogenate Treatment

Cpm

Tryptophan : tyrosine ratio

48,000 32 ,400

0.87 0.64

Medium Tryptophan: Cpm tyrosine ratio

~~~

+GAs -GAa

9600 5200

4.78 1.35

From Evins (1970). Triplicate samples of 10 aleurone layers were incubated a t 25°C in the presence and in the absence of 1 p M GA3. Between 8 and 10 hours, 5 pCi of tryptophan-aH and 1 pCi of tyrosine-**Cwere added. The trichloroacetic acid precipitates of aliquots of the homogenates and media were collected on Millipore filters and washed with 5% trichloroacetic acid containing carrier amino acids. The samples were counted in a Beckman 3-channel liquid scintillation counter and the tryptophan: tyrosine ratios were calculated using specially prepared SH and 14Cstandards. 5

4.

HORMONAL CONTROL O F A SECRETORY TISSUE

131

Gibberellic acid added to isolated aleurone layers evokes cytological changes similar to those occurring in the aleurone layers during germination. The appearance of the cells of fully imbibed aleurone layers (Fig. 24A) changes little during further incubation for 24 hours in the absence of added gibberellins (Fig. 24B). During incubation with gibberellins, extensive vacuolation of the cells occurs (Figs. 19 and 24C), reserve

FIG.23. Portion of an aleurone cell from the same region as Fig. 20, but taken a t 48 hours of germination. The rough endoplasmic reticulum is no longer in stacks, but is segmented into small vesicles. These cells are highly vacuolate owing to depletions of stored lipid and protein previously present in lipid bodies and aleurone grains, Glutaraldehyde/osmium. x 14,000. Vigil and Ruddat (1971a,b).

proteins of the aleurone grains disappear (Fig. 24C), there is extensive development of rough endoplasmic reticulum (Jones, 1969c), and there are increased numbers of polysomes and some development of mitochondria and of microbodies (Jones, 1969d; Vigil and Ruddat, 1971a,b). Biochemical data consistent with the observed cytological changes include an increased yield of polysomes (Figs. 25 and 26) from cell homogenates and an increased proportion of tryptophan in the puromycin peptides released by puromycin from these polysomes (Tables XI and

Fro. 24. Light micrographs of sections of aleurone layers and seed coats. (A) Layers from fully imbibed half-seeds. (B) Layers after incubation in -GA3 medium for 24 hours. ( C ) Layen after incubation in 1 p M GA, medium for 24 hours. Fixation, embedding, and sectioning by T. J. O’Brien and Linda Franeen. Photo by Me1 Dickerson. ~ 2 0 0 . 132

Volume

FIG.25. The effect of GA, on polyribosome formation. Polysomes were isolated from 40 barley nleuronc layers that were incubated for 18 hours in the presence of 1 p M GA, ( + G A ) , or in the absence of the hormone (-GA) a t 25" on a Dubnoff metabolic shaker. Polysome profiles were determined in 0.3 to 1.0 M isokinetic sucrose gradients. From Evins (1970).

3000

t

t GA

4 2000

.

L

0

0

5

10

15

Hours

FIG.26. The cffcct of GA, on polvsomc formation. The absolute amount of polysomes per 100 bnrlcy alcuronc hycrs wvns dcterminrd by centrifuging ribosomes in 0 3 to 1.0 M isokinctic S L I C ~ O ~gradicnts. C Forty nleurone 1:iyers were incubated a t 25" for various times in 1 mM acetate buffer, pH 4.8, with 20 mM CaCl? and either +GA = 1 JLM giliberellic acid (GA,) or -GA = without GG. The polysome area wns mcasurcd with a. planinirtrr Area measurements arc in relative units. Each point represents the average of duplicate samples. From Evins (1970). 133

134

H. YOMO AND J. E. VARNER

TABLE X I 1 LOCATION A N D PUROMYCIN RELEASEOF TRYPTOPHAN-RICH NASCENTPOLYPEPTIDES4,b Tryptophan: tyrosine ratio Treatment

Label ratio

Pellet

Supernatant

+GA +GA

+ puromycin

3H : 14C

1.35 f 0.34 0.47 k 0.24

1.55 & 0.58 2.19 f 0.59

- GA -GA

+ puromycin

3H:14C

0.317 0.315 0.17 & 0.16

2.34 f 0.72 1.53 & 0.20

~~~~~~

From Evins (1970). Forty aleurone layers were incubated during the last 16 minutes of a 10-hour incubation period with 25 pCi of t r y p t , ~ p h a n - ~and H 5 pCi of tyrosine-14C in the presence and in the absence of 1 p M GAI. The polysoma1 pellets were treated with 7.5 X M puromycin and the released nascent peptidylpuromycin was separated from the ribosomes by centrifugation through a discontinuous sucrose gradient. The samples were counted on a Beckman 3-channel liquid scintillation counter (the pellet was precipitated with 10% trichloroacetic acid, collected on a Millipore filter, and washed with 5% trichloracetic acid containing carrier amino acids, and the supernatant was counted directly with Bray's scintillation fluid) and the tryptophan: tyrosine ratios were calculated using specially prepared 3H and 14Cstandards. a

TABLE XI11 CHOLINE-14C INCORPORATION (ACID I N S O L U B L E ) VARIOUSCELLFRACTIONS~J'

Fraction Microsomal fraction Supernatant First pellet Second pellet

IN

Specific activity Relative specific activity (% of microsomal (cpm/mg protein) fraction) 2506 234 65 305

100.0 9.5 2.6 12.2 ~

~

From Evins and Varner (1971). Forty aleurone layers were labeled for 30 minutes with cholinemethyl-14C following an 8-hour incubation period a t 25°C. The results are the averages of triplicate samples.

135

4. HORMONAL CONTROL OF A SECRETORY TISSUE TABLE XIV

EFFECTOF GA3 A N D ABSCISICACID (ABA) O N THE RATE OF ENDOPLASMIC RETICULUM SYNTHESISQ.~

Treatment - GA +GA +GA +GA a

+ ABA (last 2 hours) + ABA (6 hours)

Choline incorporated (CPd

Choline incorporated (CPm) vs GA (%)

Increase uptake us GA (%)

3160 10200 3200 2800

31 100 31 27

21 0 2 2

+

+

From Evins and Varner (1971).

* Thirty aleurone layers were incubated for 6 hours with

M GA a t 25°C and 2.5 X 10-7 M ABA and choline-methyl-14C. The results shown are the averages of duplicate samples. Similar results were obtained in three experiments.

TABLE XV TIMECOURSEOF GAS ENHANCEMENT OF 32P-LABELED Pi INCORPORATION INTO PHOSPHOLIPIDW~ Counts per minute

Hours

- GA3

+GAa

1600 1660 1420 2100

1400 1420 2700 9720

From Koehler and Varner (1971). Ten Aleurone layers, were stripped from halfseeds which had been on moist sand for 3 days and incubated for the times shown. They were then labeled with 125 NCi of Pi-32P for 30 minutes and ground in buffered sucrose; the 10,000 g supernatant was extracted with ch1oroform:methanol (2: l ) , and an aliquot of this extract was counted. Because the rate of Pi-32P incorporation into phospholipids is linear up t o a 2-hour incorporation period, this technique measures the rate of phospholipid synthesis. b

136

H. YOMO AND J. E. VARNER

XII) [amylase, and perhaps some of the other hydrolases secreted by the aleurone cells, contain about three times as much tryptophan (mole %) as the average barley protein (Varner, 1964)l. I n addition, gibberellic acid induces an increased incorporation of choline-I4C into a TCA-precipitable chloroform-methanol soluble component (phospholipid) of the microsomal fraction of the cell homogenate (Tables XI11 and XIV), and an increased incorporation of Pi-32Pinto the phospholipids of the cells (Table XV). There is also a gibberellic acid enhanced increase in the activity of a t least two of the enzymes involved in the biosynthcsis of phospholipids (Johnson and Kende, 1971). Abscisic acid added several hours after the addition of gibberellic acid prevents any further increase in the rate of phospholipid synthesis (Table XVI). TABLE XVI EFFECTSOF ABA A N D GA3 ON P,-32P INCORPORATION INTO THE PHOSPHOLIPIDS OF BARLEY ALEURONE LAYERS~~~

+GA3 +GA3 +ABA +GA3 +ABA -GA3

Time (hours)

Incorporation (CPd

7 7 Last 2 7 7 7

2940 1800 400 790

From Koehler and Varner (1971). Aleurone layers were incubated 7 hours with or without GA (10-6 M). At 0 time or a t hour 5, abscisic acid (2 X 10-6 M ) was added. At 6.5 hours, 100 Ci of Pi-32P was added for 30 minutes. Layers were ground in buffered sucrose and centrifuged a t 10,000 g; the supernatant was extracted with ch1oroform:methanol (2: 1). a

b

Levels of actinomycin D that inhibit secretion of a-amylase without inhibiting its synthesis (Chrispeels and Varner, 1967a) do not inhibit the gibberellic acid-enhanced incorporation of Pi-32Pinto phospholipids (Koehler and Varner, 1971) and do not prevent the formation of rough endoplasmic reticulum (Figs. 27 and 28), although the rough endoplasmic reticulum has an unusual appearance. I n the later stages of the response of aleurone cells to gibberellic acid, the rough endoplasmic reticulum becomes distended (Fig. 23), vesicles bleb off from the rough endoplasmic reticulum and appear to accumulate a t the periphery of the cells (Fig. 29), and fragments of membranes arc visible just outside the plasmalemma (Fig. 29).

4.

HORJIONAL CONTROL OF A SECRETORY TISSUE

137

FIG.27. Addition of actinomycin D (100 pglml) to the 1.35 p M solution of GAS effects aggregation of rough endoplasmic reticulum in aleurone cells after 22 hours of incubation. There is little evidcnce suggesting that actinomycin D causes additional alteration of cellular structures in thcse cells, notably nuclear pores (NP) , mitochondria ( M ) , and plestids (P). Glutaraldehyde plus formaldehyde and CaCL/osmium. Aqueous uranyl acetate en bloc staining prior to dehydration. x 11,000.Vigil and Ruddat (1971a,b).

H. YOMO AND J. E. VARNER

FIG.28. Section of an aleurone cell treated as in Fig. 27, illustrating the large accumulation of rough endoplasmic reticulum (RER) alongside the nucleus that is characteristic of these cells. Very few, if any, polysomes appear by the nuclear pores, but they are present on small segments of endoplasmic reticulum (arrow). Close examination of the RER area reveals the presence of small lipid bodies surrounded by fibrous material, a few ribosomes, and pieces of RER. x14,OOO. Vigil and Ruddat (1971a,b).

4. HORMONAL

CONTROL OF A SECRETORY TISSUE

139

FIG. 29. Vesiculate rough endoplasmic reticulum (RER) in an aleurone cell from an aleurone layer on a half-seed incubated in 1.35 p M GAI for 41 hours. Several vesicles along the plasmalemma (PI) appear smooth surfaced. The overlapping of the plasmalemma in this region suggests that secretion of a-amylase through the R E R vesicles is more rapid than membrane reabsorption. Similar vesicles are prevalent also in aleurone cells from aleurone layers incubated for 22 hours in GA. The large amount of a-amylase in the ambient medium is most likely the result of direct secretion via the RER since there is no evidence of physical contact between the dictyosome vesicles (DV) and R E R vesicles, as is true for acinar cells of the pancreas. Glutaraldehyde/osmium. Uranyl acetate en bloc staining in 70% alcohol. x 14,000. Vigil and Ruddat (197la,b).

140

H. YOMO AND J. E. VARNER

Secretion of a-amylase-the movement of the completed amylase molecules to the outside of the plasmalemma-is inhibited by dinitrophenol (Figs. 30 and 31), carbon monoxide, pentachlorophenol, and ioxynil and by anaerobiosis (Varner and Mense, 1971), but secretion

FIQ.30. Flow apparatus for measuring release of a-amylase from the aleurone layer from one half-seed. From left to right : buffer reservoir, capillary restriction to control flow rate, chamber for aleurone layer (immersed in a water bath), and test tube in standard fraction collector. One milliliter fractions are collected every 6 minutes and assayed for a-amylase. Varner and Mense (1971).

TABLE XVII RELEASEOF PROTEIN B Y ALEURONE LAYERSIN ABSENCE OR PRESENCE OF GIBBERELLIC ACID^,^ Protein (mg/lO aleurone layers)

Time of incubation (hours)

- GA3

SGA3

1.5 24

0.480 0.665

0.510 2.455

From Melcher, (1970). Incubation medium contained 1 mM sodium acetate, 10 m M calcium chloride, and 1 pM GA3, where added. 5

b

is not inhibited directly by cycloheximide or actinomycin D and does not seem to be under the direct control of gibberellins or abscisic acid (Varner and Mense, 1971). What is the primary site of action of gibberellic acid in the aleurone cells? Does gibberellic acid control transcription in the nucleus, or translation of those proteins being synthesized on the rough endoplasmic

4.

HORMONAL CONTROL OF A SECRETORY TISSUE

141

4 Buffer

-

0

0

0.2 ~

0

o

o

0 Oo

o 0

o 0

0

oooo

0 0 0

O

-

Q

.

n

0.4

4 Yer DNP

4 Buffer

3.

0

0

0 0

0

0.2

0

0 0

0 0 0 1

00

a

0 0 1

0

Minutes

FIG.31. Release of a-amylase from an aleurone layer. Upper: HCl (1 mM) inactivates a-amylase in the cell walls in passage from the plasmalemma to the medium. Buffer (2 mM sodium acetate, 10 mM calcium chloride, and I p M GAd allows secretion and release to resume. Lower, Dinitrophenol (100 p M ) prevents secretion of a-amylase (control experiments show that dinitrophenol does not prevent diffusion of oc-amylase across the cell walls). Varner and Mense (1971).

reticulum or perhaps simply the release from the reserve compartments (spherosomes, aleurone grains, and phytin globoids) of the materials needed for the synthesis of membranes and hydrolases? The release-but not the synthesis-of p-glucanase (Jones, 1971) and of reserve proteins (Table XVII) is controlled by added gibberellic acid. Casein hydrolyzate in the incubation medium does not obviate the requirement of isolated alcurone layers for gibberellic acid in the synthesis of amylase and phospholipids (Koehler et al., 1971), nor does casein hydrolyzate prevent the inhibitory effect of abscisic acid on phospholipid

142

H. YO310 AND J. E. VARNER

FIG.32. Barley aleurone cell after centrifugation for 2 hours a t 90,000 g of an aleurone layer stripped from a fully imbibed half-seed. KMn04 fixation. X3000. Inset shows light micrograph of another cell from the same centrifuged aleurone . Jones (1969b). layer. ~ 8 0 0 From

synthesis and amylase synthesis. It is possible, however, by the use of bromate to inhibit proteolysis of the reserve proteins, to make continued amylase synthesis dependent upon added casein hydrolyzate (Table XVIII) . These experiments allow us to conclude that gibberellic acid and abscisic acid probably do not act by controlling directly the release of reserve materials from the aleurone grains.

4.

HORMONAL CONTROL OF A SECRETORY TISSUE

143

TABLE XVIII EFFECT OF BROMATE A N D AMINOACIDSO N THE PRODUCTION OF AMYLASE BY BARLEYALEURONE LAYERS~~~ Amylase (units/lO aleurone layers) Without amino acids Medium Extract GA3 alone GA3, 1 m M KBr03 GA3, 5 m M K B r 0 3 GA3, 10 mM, K B r 0 3 -GA3

39 26 4 1 2

15 17 9 3 2

Total

With amino acids Medium Extract

53 43 13 4 4

33 12 10 3 2

22 26 20 8 2

Total 55 38 30 11 5

From Melcher (1970). Incubation medium was the same as in Table XVII. c Neutralized casein hydrolyeate powder, 20 mg. a

Further thinking about the primary site of action of gibberellic acid and abscisic acid would be easier if we knew the intracellular localization of these hormones. It appears that such localization of tritiated hormones by autoradiography could be greatly aided by the use of centrifugation to stratify the aleurone cell contents (Fig. 32). REFERENCES Briggs, D. E. (1963). J. Znst. Brew., London 69, 13. Brown, H. T., and Escombe, F. (1898). Proc. Roy. SOC.63, 3. Chrispeels, M. J., and Varner, J. E. (1967a). Plant Physiol. 42, 398. Chrispeels, M. J., and Varner, J. E. (196713). Plant Physiol. 42, 1008. Evins, W. H. (1970). Ph.D. Dissertation, Michigan State University. Evins, W. H., and Varner, J. E. (1971). Proc. Nut. Acad. Sci. U.S. 68, 1631. Filner, P., and Varner, J . E. (1967). Proc. Nut. Acad. Sci. U.S. 58, 1520. Gruss, J. (1928). Wochenschr. Brauerei 45, 539. Haberlandt, G. (1890). Ber. Deut. Bat. Ges. 8, 40; also see Haberlandt, G. (1914). In "Physiological Plant Anatomy" (English translation from 4th German ed.), pp. 505-507. Jayyed Press, Delhi. Jacobsen, J. V., and Varner, J. E. (1967). Plant Physiol. 42, 1596. Jacobsen, J. V., Scandalios, J. G., and Varner, J. E. (1970). Plant Physiol. 45, 367. Johnson, K. D., and Kende, H. (1971). Proc. Nut. Acad. Sci. U.S. (in press). Jones, R. L. (1969a). Planta 85, 359. Jones, R. L. (196913). Plant Physiol. 44, 1428. Jones, R. L. (1969~).Planta 87, 119. Jones, R. L. (1969d). Planta 88, 73.

144

H. YOMO AND J. E. VARNER

Jones, R. L. (1971). Plant Physiol. 47, 412. Jones, R, L., and Price, J. M. (1970). Planta 94, 191. Kirsop, B. H., and Pollock, A,, Jr. (1958). J . Inst. Brew., London 64, 227. Koehler, D. E., and Varner, J . E. (1971). Unpublished data. Koehler, D. E., Mense, R. M., and Varner, J . E. (1971). Unpublished data. MacLeod, A. M., and Millar, A. S. (3962). J. Inst. Brew., London 68, 322. MacLeod, A. M., and Palmer, G. H. (1966). J . Inst. Brew.,London 72, 580. Melcher, U. (1970). Ph.D. Dissertation, Michigan State University. Paleg, L. (1960). Plant Physiol. 35, 293. Paleg, L, (1964). I n “RCgulateurs naturels de la croissance v6gCtale” (J. P. Nitsch, ed.), pp. 303-317. CNRS, Paris. Radley, M. (1967). Planta 75, 164. Radley, M. (1969). Planta 86, 218. Schander, H. (1934). 2.Bot. 27,433. Van der Eb, A. A., and Nieuwdorp, P. J. (1967). Acta Bot. Neer. 15, 690. Varner, J. E. (1964). Unpublished data. Varner, J. E. (1964). Plant Physiol. 39, 413. Not. Acad. Sci. U.S. 52, 100. Varner, J. E., and Chandra, G. R. (1964). PTOC. Varner, J. E., and Mense, R. M. (1971). Plant Physiol. (in press). Varner, J . E., Chandra, G. R., and Chrispeels, M. J. (1965). J . Cell. Comp. Physiol, 66, Suppl. 1, 55. Vigil, E., and Ruddat, M. (1971a). Private communication. Vigil, E., and Ruddat, M. (1971b). Plant Physiol. (in press). Yomo, H. (1958). Hakko Kyokai Shi 16,444. Yomo, H. (1960a). Hnkko Kyokai Shi 18,494. Yomo, H. (1960b). Hakko Kyokai Shi 18, 603. Yomo, H. (1960~).Hakko Kyokai Shi 18, 600. Yomo, H., and Iinuma, H. (1964). Proc. Amer. SOC.Brew. Chem. 97, p. 97. Yomo, H., and Iinuma, H. (1966). Planta 71, 113.

CHAPTER 5

GENE REGULATION NETWORKS: A THEORY FOR THEIR GLOBAL STRUCTURE AND BEHAVIORS Stuart Kaufman DEPARTMENT OF THEORETICAL BIOLOGY, A N D DEPARTMENT O F MEDICINE, THE UNIVERSITY OF CHICAGO, CHICAGO, ILLINOIS

I. Introduction.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145 11. Global Behaviors of Gene Control Systems . . . . . . . . . . . . 146 111. Homeostasis: Constrained Dynamic Behavior, . . . . . . . . . . . . . . . . 148 IV. Model Systems.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 149 V. One-Input Control Systems. . . . . . . VI. Multiple-Input Control Systems.. . VII. Forcing Structures in Switching Nets. VIII. The Size of Forcing Structures as a F Inputs per Element in Model Genetic I X . Behavior as a Function of the Size of a System, and the Number of Control Inputs per Model Gene.. . . 156 X. Biological Implications. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 160 A. Constrained Pat 161 B. Cell Cycle Time C. Number of Cell D. The Flow Matrix: Homeostasis and Restricted Pathways of Differentiation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 166 E. Capacity to Evolve ....... . . . . . . . . . . . . . . . . . 168 XI. Expected Character of Forcing Structures as a Function of the 170 Number of Forcing Connections. . . . . . . . . . . . . . . . . . . . . . . . . . . . XII. Control Advantages of Forcing Structures. . . . . . . . XIII. Molecular Mechanisms. . . . . . XIV. Additional Evidence for the Theory A. Forcible Operons.. . . . . . . B. Number of Inputs per Gene . . . . . . . . . . . . 175 C. Extended Forcing Structures. . . . . . . . . . . . . . . . . D. Developmental Genes,, . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 177 E. Macroscopic Tests . . . . . . . . . . . . . 178 XV. Alternative Theories.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 179 XVI. Conclusions and Summary.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 180 References. . . . . . . . ....... . . . . . . . . . . . . 181

1. Introduction Several years ago the late mathematician John Von Neuman remarked that the study of complex systems composed of many parts could profitably be decomposed into two complementary tasks: the eluci145

146

STUART KAUFFMAN

dation of the mechanisms and laws of the individual parts; and the no less important and difficult task of analysis of the organization of the parts into a functioning whole. There can be no doubt that recent advances in molecular and cellular biology in understanding, not only the chemical structure of DNA, RNA, and protein, but also in elucidating some of the mechanisms controlling transcription, translation, and enzyme activity, must soon bring to the foreground questions of how these processes are integrated. Unfortunately, our theories here are less richly developed than those concerned with the chemical mechanisms of the parts. The purpose of this article is to try to state explicitly a t least some of the tasks the integrated gene control system seems to accomplish; to develop a theory which allows us to see some ways that control system might work; to discuss implications of the theory for the kinds of control functions which molecular mechanisms that regulate transcription, translation, and enzyme activation might follow, that would yield systems whose global behavior is as orderly as cells’; to evaluate several predictions from the theory; and to attempt to find various empirical approaches to test them. The emphasis of the article is on the urgent need for theories about the ways in which integrated genetic control systems might function.

II. Global Behaviors of Gene Control Systems It is now clear that both prokaryotes and eukaryotes are capable of controlling the onset and cessation of DNA synthesis, transcription, translation, and enzyme activity-in many cases of quite specific species of molecules (Jacob and Monod, 1963; Shires et al., 1971; Tomkins, 1968). The molecular mechanisms accomplishing these tasks may not be identical in prokaryotes and eukaryotes ; for example, no “classical” operon has been found in a eukaryote, whose transcription regulation may involve acetylation of chromosome bound histones (Allfrey, 1968) or other mechanisms. We will refer to the system of controls concerning DNA replication, transcription of particular genes, translation of particular mRNA, and enzyme activity as the integrated gene control system. However these processes are controlled in metazoans, a central tenet of current biology is that cells in an organism differ predominantly as a result of differential biosynthetic activity, not usually of loss of genetic material. Attempts to frame theories about the integration of these still only partially known components might well begin with a clear statement of those global behaviors of cellular gene control systems which might be hoped to reflect something of the control system’s organization.

5.

GENE REGULATION NETWORKS: A THEORY

147

Several characteristics of the behavior of metazoan gene control SYStems are so ubiquitous that they are rarely mentioned; nevertheless, precisely because of their ubiquity, they are likely to be of fundamental importance. Perhaps the most obvious of these is the apparently constrained dynamic behavior of metazoan gene control systems. Consider that a metazoan cell may have perhaps 100,000 or more different genes. Suppose each gene is capable only of being active or inactive, and ignore for the moment activities of mRNA and protein. Then a metazoan cell zs 1030*000 potential different states with only 100,000 genm has 21009000 of gene activity. At known rates of alteration of gene activity, a cell could not explore that dynamic space in billions of times the history of the universe. Just how minuscule a subset of patterns of gene activity a given cell type is restricted to is entirely unknown, but presumably it is very small, since we manage to recognize the same cell type over time and cell divisions. Even if an organism has, say, 100 cell types, these jointly would seem to be restricted to a very small subset of the enormous potential variation in gene activity. Whatever the criteria by which we recognize distinct cell types, usually by gross histology and cytohistology, different organisms have different numbers of cell types. The numbers may be expected to be correlated with the numbers of distinct genes of an organism, or its DNA content. I n fact, for 13 organisms ranging from Escherichia coZi through sponges, yeast, round worms, and man, the log log correlation is nearly linear with a 0.5 slope (Kauffman, 1969a) suggesting that the number of cell types of an organism is crudely a square root type function of the quantity of its DNA. We must ask whether such a correlation, if it holds for more types of organisms, is likely to be an accident of selection or whether it reflects something basic about the number of ways a gene control system can behave as a function of the number of components of the system. When the zygote of higher metazoans begins the process of differentiation, it differentiates into intermediate cell types which themselves branch further into different cell types. One can conceive of a system in which the initial blastula cells differentiated directly into as many cell types as the adult contains. That, in higher metazoans, each cell type seems to differentiate directly into rather few other cell types may be expected to be a fundamental character of metazoan gene control systems. We will refer to this apparent property of metazoan gene control systems as RESTRICTED LOCAL ACCESSIBILITY. Whatever the conditions which direct differentiation in specific ways, the outcome is reliable to a rather high degree.

148

STUART KAUFFMAN

Finally, gene control systems must be able sucessfully to evolve. Although we are becoming familiar with how enzymes might evolve to higher specificity and greater efficiency, rather less thought has been directed to how an interlocked system of genes, mRNA, and proteins might successfully evolve. When we note that the cell types in a human and a shark are much the same, it is apparent that the gene control system must be so designed that cell types can be left roughly unaltered, while the proportions and places of their occurrence in an organism change in evolution. 111. Homeostasis: Constrained Dynamic Behavior We return to the problem of achieving restricted patterns of gene activity in organisms with perhaps thousands of genes. As is well substantiated (Jacob and Monod, 1961; Gilbert and Muller-Hill, 1966, 1967; Ptashne, 1967a,b), the rate of transcription in bacteria is controlled by inputs to an operon of specific products of other genes (e.g., the repressor) and, for the operons studied, a small metabolite. We need now to ask whether genes tend to be able to assume finely graded levels of steady activity or whether they tend to be either very active or very inactive. Both theory and experiment seem to suggest the latter to be true. It should be noted first, however, that if binding of control inputs to a gene are readily reversible, it is a t least conceivable that finely graded intermediate levels of gene activity could occur. Theoretical grounds to suppose that genes tend to be either nearly fully active or nearly fully inactive stem from a t least two sources. First, the number of copies of many genes per cell is small, usually one or two. Graded levels of activity by having varying proportions of many genes active is impossible, and a single gene a t any moment is either actively transcribing or is not. Intermediate levels of gene activity could be had, however, by time averages over periods long with respect to the time of transcription. A second theoretical reason to think genes tend to be either highly active or nearly inactive stems from the behavior of allosteric enzymes. All known allosteric enzymes are multimeric (Monod et al., 1965), a property which confers upon them the capacity for cooperative behavior (Monod et al., 1965). Cooperative behavior evidences itself in sigmoid response curves to levels of substrate and allosteric inhibitors (Monod et al., 1965) [although there is some difficulty with allosteric activators (Monod et al., 1965) 1. Perhaps the most important property of catalytic components realizing sigmoid response functions on either substrate or control inputs is that the sigmoid function can behave like a threshold

5.

GENE REGULATION NETWORKS: A THEORY

149

device. Levels of input on the lower alssymptote have little effect. If the sigmoid function is steep, small changes in the input level quickly shift activity to near maximum (or minimum), and further increases in input level have little effect. If the sigmoid is shallow, Walters et al. (1967) have shown that a sequence of sigmoids in series can behave as a set, like a single steep sigmoid, thus providing a threshold-like device. Sigmoid devices behave crudely like switches. Since the lac repressor is known to be a tetramer protein (Riggs et al., 1970), i t is not unreasonable to think that the bound repressor would yield a sigmoid response curve on its inputs. Experimental evidence also suggests that genes tend to be either strongly on or nearly off, rather than to assume finely graded steady responses. Ashburner (1970) has shown in Drosophila melanogaster that the response of the first polytene puff in the ecdysone-induced puffing sequence shows a sigmoid response to the levels of ecdysone administered. Nevertheless, data showing that metazoan gene transcription is nearly full on or full off is not yet convincing. More direct experimental evidence for the thesis that genes tend to be nearly fully on or off in prokaryotes comes from work on bacterial operons. In vivo, the lac operon induction curve is distinctly sigmoidal (Herzenberg, 1959; Boezi and Cowie, 1961 ; Bourgeois, 1966). Coupled with the evidence that the lac repressor is a tetramer which may be capable of cooperative behavior, the data are suggestive. We will therefore suppose that genes tend to be either quite active or nearly inactive; and we consider the gene control system to be the integrated system of genes, mRNA, proteins, and metabolites, by which one gene’s product can influence the rate of activity of other genes. We will focus attention on transcription control, and ask how genes which are nearly full on or full off can be coupled to one another so that the entire system exhibits constrained homeostatic behavior, that is, restricted patterns of gene activation, and strong tendencies to return to those patterns after many different pertubations, and how they might be coupled to achieve restricted accessibility during differentiation, the capacity to evolve, and their other global control tasks, as described in Section 11. IV. Model Systems

To facilitate the discussion, we introduce several idealizations. First, the gene will he considered a binary switch, capable only of being fully on or fully off. Time will bc considered to occur in discrete, clocked moments. Thc pathway by which the output of a gene comes to influence another gene (say by translation of a specific mRNA to a repressor

150

STUART KAUFFiMAN

molecule, or to an enzyme which catalyzes the formation of a metabolite which serves as an input to the target gene) will be ignored. I n another paper (Kauffman, 1971b) we discuss relaxation of the idealizations to include genes whose activities can vary continuously with their inputs, consider continuous time, and consider the pathways of interconnection between genes. A model gene control system is a set of N binary “genes” coupled together such that the outputs of genes serve as inputs which control the activity of other genes, together with a rule (Boolean function) for each gene specifying how it, will behave a t the next integer time moment for any current set of values of its inputs. The structure of such model gene control systems will be classified by the number of components, AT, and the mean number of inputs per component, I - . We distinguish three broad ways in which a n input control molecule to a gene, say a repressor, might bind to the target gene. (1) The repressor might bind weakly and reversibly, so that maintenance of repression of the gene requires a concentration of repressor molecules sufficient to guarantee that a new repressor binds nearly as soon as an old repressor comes off the locus. (2) The input molecule might bind firmly, and only be removed by a specific other molecule. (3) Input molecules might bind irreversibly to the target gene, such that the only way an unbound copy of the gene could be obtained was by replication. I n this paper we consider only the first of these; the two latter are discussed by Kauffman (1971b). They do not alter the conclusions we will reach. W. One-Input Control Systems Gene control systems of cells are almost certainly not one-input systems. Indeed, in the cases which are best known, bacterial operons, the operator locus has a t least two specific inputs. For example, for a n inducible gene, the inputs are the repressor molecule and the inducer metabolite. Nevertheless, it is useful to consider the characteristics systems in which each component has only one input. Perhaps the first structure which comes to mind when one begins to consider coupled nets of genes is a hierarchical, acyclic, one-input system derived from a single highest member of the hierarchy. The notion was first mentioned for gene control systems by Waddington (1962), who coined the phrase “cascade depression” to describe the behavior of the system when the first gene is activated and subsequently activates its immediate descendents in the hierarchy, and so on. The notion of a cascade sequence branching downstream was taken over by Britten

5.

G E N E REGULATION

NETWORKS:

A THEORY

151

and Davidson (1969) in their concept of a hierarchy of batteries of genes. Its attractive feature is that i t allows a single input to the initial gene of the hierarchy to affect the activities of many genes, so providing a way a single steroid molecule, for example, might have wide genotropic effects. There are two major disadvantages to one-input systems such as this. The most obvious is the lack of reliability of the behavior of the system when faced with component failure due either to mistakes while running or to mutation of the genes involved. There is no redundancy in the command structure. A single mutation can disconnect all members of the hierarchy descendent from the mutated gene. The second disadvantage occurs in one-input systems which have structural loops. As we shall see in detail below, such systems do not behave in very restricted ways, and they exhibit rather little homeostatic tendency to return to a mode of behavior once perturbed. VI. Multiple-Input Control Systems

Cellular gene control systems appear to have more than one input per gene, and, noting the feedback loops in repression and enzyme inhibition, are not acyclic structures. The possibility of many inputs per element allows the possibility of redundancy, and thus more reliable behavior. Furthermore, in general, the greater the number of inputs to elements of a system, the more subtle and complex can be the system’s behavior. However, these advantages are bought for a price. As we shall show below, systems with many inputs per element do not usually show highly restricted patterns of activity, nor strongly homeostatic properties. To obtain constrained, homeostatic behavior requires increasingly subtle construction as the number of inputs per element increases. One of the most obvious ways to obtain homeostatic behavior is to build multi-input systems which have many Forcing Structures in them. Indeed, as we shall see, i t now begins to appear that perhaps the only way to build large nets of switching elements that exhibit homeostasis and the other biologically “good” global behaviors noted in Section 11, without highly orderly construction of the interconnections between genes, is to build systems either rich in extended forcing structures or a t least rich in components which are forcible on one or more input lines. VII. Forcing Structures in Switching Nets

Elements in a switching net realize Boolean functions on their inputs.

A Boolean function is a rule which prescribes for an element with K inputs, what its value (0 or 1) shall be at time T 1, for each of

+

152

STUART KAUFFMAN

the 2 possible sets of values of its K inputs a t time T (see Fig. 1 ) . Consider an element which receives two inputs, and switches on at T 1 if, and only if, either the first, or the second, or both input lines carried a 1 value at time T . The element realizes the OR function. An element will be said to be forcible on a given input line, if one of the two possible values of the input line causes the element to assume one of its two values a t the next time moment, regardless ,of the values on any of the remaining input lines. We restrict the definition to elements with more than one input. For example, an element realizing the OR function is forcible on both its input lines, since a 1 value on either input line forces the element to assume the value 1 a t the next time moment regardless of the value on the other input line. The FORCED VALUE of an element is that value to which i t is forcible. An element realizing an OR function has a forced value of 1. An element realizing

+

or

and

exclusive

or

if and only if

FIG.1. Four Boolean functions showing the binary value of 2 a t time T for all possible values of inputs A and B at time 2'.

+

+ 1,

the function EXCLUSIVE OR switches on a t T 1 if and only if, at T , either its first input was on and the second was off, or the second was on and the first was off. If both were simultaneously on, or off, the element switches to 0 a t T 1. An element realizing the EXCLUSIVE OR function is not forcible on either input line, for no value on either input line guarantees that the element will be 0, or 1, the next time moment, regardless of the value of the other input line. If an element A is an input to element B, A will be said to force B if and only if: ( 1 ) A is itself forcible on a t least one of its own input lines, (2) B is forcible on the input line from A ; (3) the FORCED VALUE of A is the value of A which forces B to its forced value. For example, suppose A has two inputs and realizes an OR function on them; and let B have two inputs, of which one is A , and realize an OR function on them. Then A forces B because A is itself forcible on a t least one input; B is forcible on the input line from A ; and the forced value of A , 1, is the value of the input line from A to B which forces B to its fo'rced value. However, if A had realized the AND func-

+

5.

G E N E REGULATION NETWORKS: A THEORY

153

tion, then its forced value would have been 0, not 1, which is not the value of the input linc from A to B that forces B, and A would not have forced B . So defined, the relation between elements of FORCING is a transitive, and nonsymmetric, relation. If A forces B and B forces C, then A forces C with a delay of 2 time moments. T h a t A forces B does not imply that B forces A . However, since A can be an input to itself, and force itself, the relationship of Forcing can be reflexive. In order to apply the concept of forcing to switching nets, we need two simple theorems and some concepts from graph theory.

THEOREM 1B. If a n element, A , i s forcible on more than one input line, then its FORCED VALUE is identical for all the lines on which it is forcible." THEOREM 1A. For an element with M inputs, the maximum number of forcing inputs is M.* The minimum is 0. A directed graph is a set of points, a set of arrows, and two rules. The first rule assigns the tail of each arrow to a particular point; the second rule assigns the head of each arrow to some particular point. A SUBGRAPH is a subset of the points and arrows of the initial graph, connected as they were in the initial graph. We may consider each gene in a model genetic switching net as a point, and the input lines as incoming arrows, hence representing the net by a graph. Making usc of our definition of forcing function, we may examine the Boolean function realized by each gene on its inputs, and create a subgraph of the switching net by keeping only those connections which are forcing. This FORCING GRAPH of the switching net is a subgraph embedded in the entire switching net; in it, a n arrow from A to B means that A forces B in the initial switching net's behavior (Fig. 2 ) . We define the FORCING STRUCTURE OF AN ELEMENT to be (1) the set of all elements of the forcing graph which that element can reach by directed paths (i.e., by following arrows tail to head sequentially) ; plus (2) the set of all elements which can reach that element by directed paths. The forcing structure of an element is its descendent tree plus its antecedent tree, those elements which directly or indirectly it forces, or force it. An element may be a member of its own forcing structure; if so, then the element lies on a FORCING LOOP or FORCING CYCLE such that *All mathematical derivations except the formula for the number of cycles in onpinput nets arc in Appendices 1, 2, or 3 of Kauffman (1971a). The one-input net state cycle analysis is in Kauffman (1971~).

154

STUART KAUFFMAN

FIG.2. Heavy arrows represent forcing inputs, light arrows are nonforcing inputs, in a switching net.

it is its own predecessor and successor. We now investigate the properties of forcing loops. Consider a forcing loop in a switching net of many components, in which A forces B , B forces C, and C forces A , and where each receives other nonforcing inputs (see Fig. 3 ) . By the definition of forcing, if A is currently in its forced value, B must assume its forced value one moment later, regardless of what B’s other inputs are doing. Thus, if

FIG.3. A three-element forcing loop, A, B, C, with descendent forcing tree.

5.

GENE REGULATION NETWORKS: A THEORY

155

A is currently in its forced value, then a forced value propagates around the three element forcing loop in three time moments. No values of any inputs in the switching net arriving at A , B , or C, can dislodge the forced value from propagating around the loop. On the other hand, new forced values can enter the loop a t loci not currently forced. When A is not currently in its forced value, then B’s behavior the next time moment depends not only on A’s value, but on its remaining inputs. If these values happen to cause B to assume B’s forced value, a new forced value enters the forcing loop and cannot thereafter be dislodged. Eventually, every available locus in the forced loop will be expected to become filled with a forced value, and thereafter, the behavior of the loop will be fixed a t a steady state in which each element remains a t its forced value. Every gene in the forcing loop may send forcing arrows to genes not on the loop. This entire descendent forcing structure will also eventually be forced to the appropriate forced values and remain fixed in one state. Obviously, forcing loops constrain the behavior of a switching system powerfully, and exhibit a strong homeostatic tendency to return t o the state with all elements forced, if ever perturbed. Three interrelated factors tend to constrain dynamic behavior in forcing structures which have no loops-that is, unilaterally connected forcing structures. Consider a straight chain, in which A forces B , B forces C , etc., and let this forcing structure from a model genetic switching net have nonforcing inputs to its members from elsewhere in the genetic net. As in the forcing loop, by definition of forcing, once a forced value enters this straight forcing chain a t any point, i t cannot leave except by propagating until it reaches the last member of the forcing chain, and passes off. But forcing values can enter a t any element on the chain whose forcing input is not currently in its forced value. This creates a strong tendency for the later members of the chain to be in their forced value most or all of the time. This effect is enhanced if, instead of a straight chain, the ancestor tree to any element is well branched, for the chance that an element is currently forced is the appropriate sum of the chances that each of its predecessors was forced a t the right moment previously. Finally, consider an acyclic forcing structure in which A reaches B by several directed paths. Let there be a path length 1 from A to B , a path length 2, another length 3, and so on up to a path length K. Then if A is currently in its forced value, B will be forced for K consecutive moments. For a rather extensive acyclic forcing structure rich in such directed semicycles, later members of the structure will be forced most of the time.

156

STUART KAUFFMAN

VIII. The Size of Forcing Structures as a Function of the Number of Inputs per Element in Model Genetic Control Systems

The extent to which behavior in a switching net is constrained by forcing structures, depends upon the size and extent of those structures spreading throughout the net. The number of possible Boolean functions a switching element with K inputs might realize is 22K.The fraction of these which are forcible on one or more input line is maximum when K = 2 ; 14 of the 16 functions are forcible. As K grows large, the subset of forcible Boolean functions grows much smaller. If Boolean functions are assigned to elements from the entire 22Kpossible functions, we show (Kauffman, 1971b) that the expected number of actual forcing connections, R in a net of N elements is less than: K -2

2K+’(K

R<-

2

+ I2= O (22’ - 1)(K - I ) ( K - I - 1)) 29

I’

which grows very small for more than 3 inputs per element. Forcing structures will be largest in nets with 2 inputs per element, if all possible Boolean functions may be used. However, nets with many inputs per element rich in forcing structures may be built by utilizing only forcible Boolean functions from the increasingly improbable subset of such functions as K increases. We now consider systems constructed without careful selection from restricted subsets of Boolean functions, in order to evaluate the effect of the average number of inputs per element, and the number of elements in the net, on the global properties of the system’s behavior (Kauffman, 1969a).

IX. Behavior as a Function of the Size of a Model Genetic Control System, and the Number of Control Inputs per Model Gene These model genetic nets are constructed by choosing a value of N the number of elements in the net, and of K , the number of input lines to any gene. Each gene receives exactly K inputs, one from each of K model genes among the N . Nets are randomly constructed in two distinct senses, the K inputs to each gene are chosen randomly; to each gene one of the 22K Boolean functions of K inputs is assigned randomly. After construction, the structure of the nets is fixed. We assume all genes compute one step in one clocked time moment. The behavior of such systems was explored by computer simulation. A STATE of the net is a list of the present value of each gene. Since each element can be on or off, there are 2N possible states.

5.

157

GENE REGULATION NETWORKS: A THEORY

+

If the syetem is placed in some state at time T, a t T 1, each gene scans the present value of each of its inputs, consults its Boolean function and assumes the value specified by the function for that input configuration. The net passes from a state to only one subsequent state. There are a finite number of states. As the system passes along a sequence of states from any arbitrarily chosen initial state, it must evantually reenter a state previously passed. Thereafter, the system cycles continuously through the reentered sequence of states, called a STATE CYCLE, whose length is the number of states on the cycle (Fig. 4). T \EC

T+I AEC

D00l000 3 0 1 110 3 1 0 101 3 1 1 I l l 100 0 1 1 101 I l l

llo

I l l

09

1-i 2-5

3/4

A/

I l l

(01

(dl

FIG.4a. Switching net in which each element has inputs from the other two, and realizes an OR function on its inputs. FIG.4b. Functions realized by A, B, and C on their inputs, as derived from Fig. 4a. Each row is a state of the net. States are numbered &7. FIG. 4c. Kinetograph showing state transitions from 4b. There are two state cycles, each one state long. FIG.4d. Flow matrix for the two state cycles for all possible reversals t o one element’s activity. All three perturbations cause state cycle 0 to go t o state cycle 7. No perturbation causes state cycle 7 to go to state cycle 0. FIG.4e. Flow among cycles without, and with, slight pcrturbations.

A model genetic net must have a t least one state cycle, but may have more. The number of state cycles is the number of distinct different ways the system can behave. PERTURBATION. As a net passes around a state cycle, one uinit of perturbation may be introduccd by arbitrarily changing the value of a single gene, then releasing the system. After perturbation, the system may return to the state cycle from which it was perturbed, or run to another cycle. By perturbing all states on each state cycle in all possible ways one gene a t a time, a matrix may be obtained listing the total number of times the system returned to the state cycle perturbed, or

155

STUART KAUFFMAN

ran to each other cycle. Dividing the value in each cell of this matrix by its row total yields the corresponding matrix of transition probabilities between state cycles under the drive of random, one unit, perturbation. This transition probability matrix will be called the FLOW MATRIX among the state cycles.

TOTALLY CONNECTED NETS, K = N When N is large, and K = N , so that each gene has an input from every gene, the density of forcible functions is very low, so R is very small, and nets have almost no forcing structures. Behavior would not be expected to be localized, nor to show homeostatic return to a localized subset of states after small perturbation. I n these nets, the expected state cycle length is 2N/z,the square root of the number of states. With only 2000 genes, such a system would cycle through 2looox l O 3 O o states. Although state cycles are very long, K = N nets exhibit a very important characteristic: the expected number of distinct cycles is only l / e ( N ), a linear function of N . This strong constraint will assume greater importance later on. K = N nets show very little or no homeostatic tendency to return to the cycle from which they were perturbed. Further, a single unit perturbation can take the system from nearly any of its l / e ( N ) cycles to any other one. That is, these systems also do not show restricted local reachability.

K

=

1

NETS

We saw that real genetic control systems were almost certainly not one-input systems. One-input nets have no forcing structures, by the definition of forcing. Such a net falls apart into separate cyclic structures, and state cycles are roughly the lowest common multiple of the structural loop lengths. Cycle lengths are very long, the number of state cycles is huge-about In N* 2Nlln N [2(N/ln N ) ] One-input systems show almost no homeostatic return to a state cycle which is perturbed, and each cycle can reach approximately N other state cycles by a single perturbation, and can reach all possible state cycles by a sequence of single perturbations. That is, one-input systems do not show restricted local reachability. K = 2 NETS The density of elements forcible on one or more input lines, or actual extended forcing structures is high in K = 2 nets. The behavior of typical nets is highly localized, exhibits marked homeostatic return to a per-

5.

GENE REGULATION NETWORKS: A THEORY

159

turbed cycle, and marked restricted local reachability. State cycle lengths = are typically ‘only square root N . A net with 10,000 genes and 2101000 103000states, cycles repeatedly through a minuscule 100 states. The number of state cycles is also very few, only square root N . This result is particularly important. Further, the system exhibits strong homeostatic properties, returning to the cycle perturbed for about 0.9 to 0.95 of the possible unit perturbations. Finally, K = 2 nets also show marked restricted local reachability, for even if the number of state cycles is, say, 50, perturbation can cause a single cycle to flow to only about 5 or 6 other cycles. That is, the FLOW MATRIX of a K = 2 net has mostly 0 entries.

K= 2

NETS BUILT WITH NEITHER FORCIBLE ELEMENTS, NOR EXTENDED

FORCING STRUCTURES

Among the 16 Boolean functions of 2 inputs, two are not forcible on either input. These functions are EXCLUSIVE OR, and IF AND ONLY IF. Nets built using exclusively these two functions have neither forcible elements, nor forcing structures, and have enormously long state cycles (Walker and Ashby, 1965). The expected number of cycles in such nets is not known.

K

= 2 NETS BUILT WITH MANY FORCIBLE ELEMENTS, BUT WITHOUT EXTENDED FORCING STRUCTURES

An element realizing the AND function is forcible on both inputs to 0, an element realizing an OR function is forcible on both inputs to 1. However if an element realizing AND on its inputs is an input to an element realizing an OR function, the “AND” element does not actually force the OR elcment, since the forced value of the “AND” element, 0, is not the value which forces the “OR” to its forced value 1. Preliminary results suggest that two input nets built with many elements that are forcible on both inputs, but with the actual number of forcing connections, R , small (from 0.2N to 0 . 5 N ) , have short state cycles roughly like randomly built K = 2 net rich in extended forcing structures. When R is 0.5N or less, we show in Section XI that the net will contain few extended forcing structures. This result, coupled with studies on nets with extensive forcing structures, suggests that extensive forcing structures may be a sufficient, but not necessary, condition for the occurrence of short state cycles. Further, the presence of very many forcible components by itself seems to be sufficient to guarantee reasonably short state cycles. These nets also seem to have about square root N state cycles. The character of their flow matrices is unknown.

160

STUART KAUFFMAN

Nets with 3 inputs per element, where the Boolean function of each element is chosen a t random from the 256 possible functions, have state cycles which become too long to be found by computer simulation search through 10,000 states when N is greater than 100. However, if the restricted set of Boolean functions of 3 inputs which are forcible on one, two, or all three inputs, are used to generate model nets, such nets behave as do nets with 2 inputs per element. These computer simulations suggest a very strong conclusion. When K = N , the number of state cycles is only ( l / e ) N . When K = 2 or 3, the number of state cycles in square root N . Thus the number of state cycles, as a function of K , only increases from a square root function to a linear function of N as K goes from 2 to N . These results suggest that when the number of inputs K is small, almost any switching net will have about square root N modes of behavior; and certainly almost any net rich in forcing structures will have about NM modes of behavior unless it is constructed in rather particular ways. Furthermore, any net rich in forcing structures should show strong homeostatic properties, and strongly restricted local reachability, unless constructed in an odd way. The properties of forcing structures developed in Section VII, led us to think that most members of an acyclic forcing structure would be fixed in their forced values, and that all members of forcing loops and their descendent trees would be fixed in their forced values. Results with K = 2 nets suggest that this is true. On any state cycle in such a net, usually about 0.7-0.8 of elements are fixed at a constant value, the remainder oscillate between being active and inactive. Generally, about 0.6-0.7 of the elements are fixed a t the same value, 0 or 1, on all the state cycles of which the net is capable. I n K = 2 nets rich in forcible elements, but constructed without extended forcing structures, although state cycles are short, only about 0.15N to 0.25N of the elements are fixed in value on one-state cycle. Even fewer are fixed on all state cycles.

X. Biological Implications I n Section I1 we mentioned several global characteristics of the behavior of real genetic control systems which must be explained: Patterns of gene activity are highly constrained relative the potential range of variability in gene activity, cells exhibit marked homeostatic properties, any metazoan cell type differentiates directly into only rather few other cell types (restricted local accessibility), there is a strong correlation

5.

G E N E R.EGULATION

NETWORKS:

A THEORY

161

between the amount of DNA in the cells of an organism and the number of its cell types, and the gene control system must be able to evolve successfully. We now show that model genetic nets rich in forcing structures behave in ways that could account for these properties.

A. CONSTRAINED PATTERNS OF GENEACTIVITY If cell types correspond to extremely restricted patterns of gene activity, and if genes can be well modeled by assuming the gene is either off or on, then switching nets rich either in forcible elements alone or in forcible elements connected into extended forcing structures exhibit the extremely restricted patterns of activity required. Randomly built nets with two inputs per element, which are automatically rich in forcible elements and structures, are restricted to oscillate through state cycles only fl long out of 2~ states, for example, to 100 out of 2 1 ~ 9states ~~0 when iV = 10,000. The remarkable additional fact is that achievement of such constrained behavior does not even require a precise choice of wiring interconnection between the genes, nor choice of the Boolean function realized by each gene. Rather, it is a n automatic consequence of the fact that most Boolean functions of two inputs are forcible. With increasing numbers of inputs, the fraction of Boolean functions which are forcible drop off, and therefore cycle lengths increase. “Good” global behaviors require the use of forcible Boolean functions. We show below that many genes seem to have two major control inputs, and that genes may plausibly be supposed to realize forcible functions on somewhat more numerous inputs. Since precise construction is not required of the control system, a consequence of the theory is that random mutations, would find ‘Lworkable” control systems more easily than might have been expected. B. CELL CYCLETIME There is a strong correlation between the amount of DNA per cell, and cell cycle time. Some minimum cell cycle time is requisite, as exemplified in bacteria, by the time necessary to replicate the bacterial 20 minutes. In higher metazoan cells, the DNA is not DNA-about replicated sequentially, as in bacteria. If it were, a mammalian cell cycle would have to be 1000-fold longer than the bacterial cycle, since mammalian cells contain about 1000 times the bacterial quantity of DNA. Instead, many segments of a metazoan’s DNA replicate simultaneously. Each segment, of unclear length, is called a replicon, and probably has about a bacterial equivalent of DNA (Berendes and Beermann, 1969; Okada, 1968). Painter et al. (1966) estimate that a human cell has 10,000 replicons. If all replicons replicated simultaneously, a mammalian

162

STUART KAUFFMAN

cell cycle might conceivable be roughly the duration of a bacterial cell cycle, but the typical mammalian cell cycle is about 24 hours. (Cleaver, 1967). I n fact, the mean cell cycle time in minutes is about a square root function of the DNA per cell of the organism (see Fig. 5 ) . Model nets with two inputs per element, or those with three or four inputs which utilize only forcible functions, have state cycles whose length increases a t about a square root function of the number of elements of the net. Thus, the time required for a switching net to traverse its state cycle is about a square root function of the number of elements in the net.

Number of elements In the random net and estimated number of genes per cell

FIG.5. Correlation between quantity of DNA per cell and cell cycle time. Also plotted is state cycle length as a function of N in K = 2 nets either using all 16 Boolean functions of two inputs, or excluding tautology and contradiction.

Presuming that the number of genes, or replicons is roughly a linear function of the amount of DNA in a cell, the results are suggestive. There is, by now, considerable evidence that replicons replicate in rather specific temporal sequences during S (Lima-de-Faria, 1969) that mRNA and protein synthesis of substances other than DNA polymerase and other obvious enzymes is needed to initiate replication on yet unreplicated replicons (Kim e t nZ., 1968), and that replication of “early S” replicons, in some species, may bc a necessary condition for the synthesis of “trigger” molecules which initiate the replication of “late S” replicons (Cummins and Rusch, 1966). In complex dynamic systems with many inhibitory connections, the occurrence of an inhibition creates a pause, or temporal gating, while the inhibitory material decays away. However, the signals that control the temporal sequence of replicon activity are integrated into the entire genetic control system, one would expect that

5.

GENE REGULATION NETWORKS: A THEORY

163

the periodicity of the behavior of the set of replicons would reflect the overall periodicity of the entire genetic control system in which it was embedded. With so little information a t this stage on how the order and and duration of replication in S is controlled, these observations a t best suggest that the global periodicity of the cell cycle is roughly a square root function of the quantity of its DNA, a result which parallels the prediction which might be made from the behavior of typical switching nets with about two inputs per element, and rich in forcing structures. The observed cell cycle length might therefore be due to this kind of temporal gating in a system with many inhibitory influences.

C. NUMBEROF CELL TYPES A strong result of computer studies of switching nets is that almost any model genetic control system with large forcing structures, particularly with two or three inputs per element, will have about square root N distinct state cycles, each a different permanent way the model genetic

6%

6% 5E z:

Number of elements m the random net andestimated number of genes per cell

FIG.6. Log number of cell types plotted against log estimated number of genes. The theoretical curve shows log number state cycles against log N .

system can behave. Let us, for the moment, suppose that each state cycle of a model genetic control system corresponds to a distinct cell type of which the model control system is capable. Then the theory makes the strong prediction that the number of cell types in an organism ought to be roughly a square root function of its number of genes. I n Fig. 6 we plot the logarithm of the number of cell types of various organisms as classified by gross histology and sometimes cytohistological techniques, against the logarithm of the estimated number of genes in the organism. The number of genes was estimated by assuming the E. coli has about 2000 genes, and taking linear proportions for cells with more DNA than E. coli. The data lie roughly on a straight line for 13 organisms. Its slope, like that of the number of state cycles as a

164

STUART KAUFFMAN

function of N is 0.5, a square root type slope in a log-log plot. Thus, a t face value, the model seems rather strikingly to give some information rn to the rate a t which the number of cell types in organisms increases as a function of the quantity of DNA per cell. Is the parallelism to be taken seriously? Before discussing several severe criticisms of the hypothesis, it is worth considering three reasons why the hypothesis is important. First, we must raise the question of whether the observed, and observable, correlation between DNA content of cells in an organism and the number of its grossly recognizable cell types is an accident of selection, or whether it offers some fundamental information about the way the genetic control system is organized and could have been organized during evolution. This is obviously an enormously important question. I s there some reason, due to the kind of control structure which might most plausibly be constructed by evolution of DNA and protein control systems, that such systems are very likely to have the observed number of modes of behavior-that is, cell types-as a function of the number of, say, distinct genes? Or, might genetic control systems with almost any number of stable modes of behavior have easily been fashioned by evolution from DNA and protein building blocks, and only selection has constrained it to the observed relation? The second reason to take the hypothesis seriously is that it appears to be a very reliable property of switching systems with modes of behavior. few inputs per element that they have about N1/2 We have seen that it is precisely systems with few inputs per element and therefore with many forcible elements and forcing structures which might be expected to exhibit the sort of homeostatic, constrained dynamic behavior we think gene control systems require. It seems to be just these systems which also may be expected to have square root A: modes of behavior-a result which parallels to an intriguing degree the observed rate of increase of the number of cell types as a function of the quantity of DNA per cell. While we will see many reasons to be critical of this hypothesis, such a clue ought not to be easily thrown away. Third, the results indicate that it is switching systems with few-usually two or three-inputs per element which are most likely to exhibit homeostatic behavior and have N'/? modes of behavior. There is, indeed, a good reason to think that genetic control systems fashioned of DNA and proteins are likely to give rise to control systems of the sort described. What we mean by macromolecular specificity is just that any molecule affects the behavior of only a very few other molecules. Creating systems with few inputs per component is precisely what genes and proteins are good at, and have obviously done. We turn now to several major criticisms of the hypothesis.

5.

GENE REGULATION NETWORKS: A THEORY

165

A first criticism is that almost any data look roughly linear in a log-log plot, so the apparent linearity in the data on the number of cell types says little. The slope of the line, however, does yield sensitive information, and it is the similarity of slope in the log-log plot to which we wish to allude. Second, estimates of the number of cell types, and the number of genes as a function of the quantity of D N A per cell, are both difficult a t best and arbitrary a t worst, particularly considering the evidence for redundant, nonsense D N A (Britten and Davidson, 1968). Certainly the number of cell types we will say an organism has by gross morphological, histological criteria will be different from our classifications as techniques to describe cells become more refined. However, if it should be the case that the effective number of genes in a cell’s D N A is some fairly constant fraction of the total DNA, and if more refined techniques for description of ccll types will raise thc number of cell types of each organism roughly proportionally to the number we now classify it as having, then the slope of the curve relating the number of cell types to the number of genes in a log-log plot will remain about the same. The plot of the number of cell types is parallel to, but not identical with the predicted square root relation between number of genes and number of cell types. Taking the data at face value, the hypothesis suggests itself that genes work in clusters in such a way t h a t the effective number of genes is lowered, to bring the experimental curve in line with the theoretical one. One of thc most obvious “clustering” of genes occurs by their grouping onto replicons. Most bacteria have one replicon ; Painter et al. (1966) suggested that man may have about 10,000, Substituting estimated number of replieons for estimated number of genes does, in fact, bring the observed numbers of cell types into almost exact agreement with the prediction from model switching net, a coincidence which might bc a cluc. Intercstingly, tlierc is evidence (Beerman and Berendes, 1969) that onc chroniomcrc, apparently one functional puffing unit of transcription in dipteran chromosomes, may constitute a single replicon. Drosophiln has (Beerman and Berendes, 1969) about 3000 chromomeres. If we may guess that Drosophila has a number of cell types roughly cqual to that of annelid worms, and substitute 3000 “genes” as the estimate of a fly’s number of genes, then the data point is again shifted over almost exactly to the theoretical N112 curve. On the other hand, it is equally plausible to note that our estimates of the number of cell typcs on morphological grounds may underestimate the true number of cell types; multiplication by roughly a constant would move the observed curve up toward the theoretical one. Finally we mention a set of criticisms which are discussed in

166

STUART KAUFFMAN

Kauffman (1971b). The hypothesis treats the cell type as a direct 1:1 expression of the pattern of gene activity in the cell. As such, it ignores the fact that different spatial arrangcments of components, consistent with the same pattern of genetic activity, could markedly alter the properties of a cell, For example, the local structure of cell membranes might alter even when constructed of the same unit molecules. The model, as developed, assumes that each “gene” is free to evaluate the state of its inputs at each successive tirnc moment. This corresponds to the assumption that control inputs bind weakly and reversibly to target genes, coming off with sufficient frequency that the gene can reevaluate and respond to the current states of its control inputs. We have not shown that the same qualitative results about localized behavior, and the expected number of modes of behavior remain valid when it is assumed that an input molecule stays attached to a target gene until specifically removed by a different input molecule, or actually binds irreversibly with the target gene. Further, genes manufacture mRNA with different half-lives (Spirin, 1966), some very short-lived, others very stable. The model ignores this important complication. The numbers of cell types in higher plants is rather less than for animals with comparable amounts of DNA per cell. Finally, the model deals with binary, discontinuous switching devices, not with components whose activity can vary continuously with the level of input. We must ask whether there is reason to think the results on switching nets give any information about systems of continuous components. Continuous systems are also discussed in Kauffman (196913) and Newman and Rice (1971). We now examine the biological implications of the homeostatic properties and marked restricted local reacliability evidenced by switching model gene control systems with few inputs per element.

D. THEFLOW MATRIX:HOMEDSTASIS AND RESTRICTED PATHWAYS OF DIFFERENTIATION If the state cycles, the stable modes of behavior of a model gene net, correspond to cell types, then flows between state cycles model differentiation. As we noted in Section 11, there is considerable evidence in higher metazoans that each cell type differentiates directly to at most rather few other cell types, which may themsclvcs branch to further cell types. Furthermore, there is evidence suggesting that a given cell type, perturbed with a wide variety of stimuli, can still only differentiate directly into one of a few cell types. The strongest evidence for this lies in the long story of neural induction, in which it has become clear that very many agents can induce the presumptive neural tissue to differentiate into neural tissue. This might be due to the fact that all the diverse

5.

GENE REGULATION

NETWORKS:

A THEORY

167

agents act on the same point in the system, causing the same result, but it is surely more plausible to think that the presumptive neural tissue is poised, as i t were, and nearly any stimuli, acting on diverse points in the system, will push it into being neural tissue. As we noted, switching nets with 2 or 3 inputs per component exhibit precisely these properties under the drive of random, small perturbations. Any mode of behavior is returned to after reversal of the value of a randomly chosen gcne with a probability of about 0.9. If perturbations can cause the system to move away from one state cycle, they can only cause it to move to a few, say five or six, other states cycles. Thus, the systems exhibit both marked homeostasis and marked restricted local reachability. It may be noted that, in general, the transition probabilities between two state cycles are not symmetric; usually it is easier to go from one to the other than back. In addition, these that is, a t least nets often exhibit RESTRICTED GLOBAL ACCESSIBILITY; some state cycles are unable to reach all the other state cycles (through intermediate state cycles) by sequences of perturbations to one gene’s activity at a time. Although I know of no data on the transition probabilities between cell types of an organism under the drive of small perturbations, Hadorn’s remarkable work (1967) on transdetermination of anatomic structures in Drosophila melanogaster shows striking parallels to the behavior of these model genetic systems. Hadorn and others have studied the “stability” of the “determined” state of imaginal discs of Drosophila larva by culturing specific disks, for example, the wing disk, in the abdomen of an adult fly for some number of generations of transfer from adult abdomen to adult abdomen, then transplanted the progeny of the disk tissue back into a larva undergoing metamorphosis to see whether the tissue would give rise to that structure for which i t was initially determined. Generally such homotypic development occurs. However, occasionally, tissue from one disk becomes “transdetermined” and gives rise to a structure deriving normally from a different imaginal disk. Several characteristics of the flow pathways of such transdeterminations are of interest to the present discussion (Fig. 7). Hadorn et al. have now found several structures which transdetermine one to another. I n general, the time of occurrence of a given transdetermination seems to be random, but to occur with roughly constant probability per transfer generation. Usually the probability of transdetermination from one structure to a second is not identical with the reverse probability. Each structure can only transdetermine into two or three others directly, and there are some structures which cannot reach all other structures by transdetermination. For example, all structures can directly or indirectly trans-

STUART KAUFFMAN

(a)

1 \-1-

[TGq

( b1

Tysigrnent

Occiput

-

-Wing (Spread, Costa )

Proboscis

t

Polpus + Antenna

Eye

-

Tarsus

Mesothorox

? + -

Proboscis

FIG. 7. Schemes of the transdetermination sequences. (a) Scheme based on clonal analysis. 7 1 autotypic elements. (b) Comprehensive scheme summarizing the observations of several authors. Short arrows indicate rare transdeterminations. The dotted arrow indicates that transdetermination to genital cells was observed, but it is not known from which cell type these are derived. From Gehring (1968).

determine into mesothorax, but mesothorax has never been found to transdetermine into any other structure. The pathways of transdetermination are not the normal paths of differentiation of these structures. It is not unreasonable to conclude that these abnormal flow pathways reflect something about the “neighbor” relation of the cell types or structures involved. In a particularly interesting study, Gehring (1968) has examined transdetermination in the aristapedia mutant of Drosophila melanogaster, which grows a leg where it should grow an antenna. Strikingly, antenna imaginal disks from this mutant transdetermine immediately into leg structures. It might be asked whether the genome of this mutant had lost the capacity to make antennae. The answer is no, for Gehring found that, left long enough, antenna1 imaginal disk which had transdetermined to give rise to leg structures on reimplantation, would transdetermine back to give rise again to antennae-a behavior that never occurs in the normal course of this mutant’s ontogeny. Somehow, the capacity to make antennae has persisted, but the stimuli or conditions to bring i t about have either changed, or do not occur.

E. CAPACITY TO EVOLVE Gehring’s startling result exemplifies the fact that a great deal of evolution seems to have kept cell types much the same, but altered the circumstances of their occurrence, the structures built of them, and the proportions of different cell types utilized in the different organisms. I n short, the construction of the genetic control net must allow alterations of genes which largely leave cell types unaltered, but which change the conditions under which those cell types will occur. It is an obvious, but important, point that not every conceivable

5.

G E N E REGULATION

NETWORKS:

A THEORY

169

gene control system can operate this way. For example, if a gene control system were a net of switches each 7f which had inputs from very many other genes and the Boolean functions were assigned completely randomly, from among all possible functions of those numerous inputs, then mutation of any gene to produce a useless product would be expected to alter every state cycle of the system, that is, t o alter every “cell type.” On the other hand, systems with few inputs per element can be altered by mutation in such a way that all state cycles are left unchanged, and only the stimuli which cause the system to flow between the state cycles are altered. Nets with two inputs per element have short state cycles because many elements of the forcing structures fall to their forced value and remain fixed there. Suppose therc is a gene, i, which is inactive on all state cycles of the system. If that element is removed from the net-or switched permanently off by a mutation which renders its output useless-that removal will not alter any state cycle of the system. Symmetrically, if an element j is active on all state cycles, its mutation to a “constitutive” state will not alter any state cycle of the system. Such mutations, however, do alter which states run into which state cycles. The consequcncc is that, although the state cycles of the nonmutant system are left unaltered, the transition probabilities of flow among the cycles by perturbation may be changed. If modes are cell types, such a mutation alters the flow among cell types, but not the cell types. The same effect may be obtained by adding a new gene, m, to the system and giving it a forcing repressing input from a member of the net which is always active on all state cycles. Then m will be inactive on all those state cycles, and the cycles will be unaltered. However, if its repressive input is transiently inhibited, m can be active and lead to new behavior, either to a flow to another old state cycle, or to a newly created state cycle. In short, old modes of behavior may be kept while new ones are added and flows are altered. Clearly, the capacity to do these things rests on the occurrence of elements which are constantly active or constantly inactive on all state cycles, i.e., on all cell types. Systems with forcing structures seem most likely to have these properties. Such a system is obviously highly desirable, for it partially uncouples the cell type from the conditions of its occurrence. If favorable cell types are found in evolution, they can be kept, and utilized in diverse circumstances. We note that nets rich in forcible elements, but without extended forcing structures have only about 0.2N elements fixed on all state cycles. It would be harder for such systems to evolve without altering state

170

STUART KAUFFMAN

cycles, than nets rich in forcing structures where about 0.5N to 0.06N elements are fixed on all state cycles. The effects of mutation of a gene normally fixed off, or on to one irreversibly off or on, has been studied by computer simulations on nets with 50 elements and 2 inputs per element. If a net has L state cycles, then each state cycle can conceivably flow by perturbation to L state cycles. The FLOW MATRIX is an L x L matrix. Let M be the number of elements in a net which are either fixed off on all state cycles, or fixed on. The effects of irreversible “mutation” of an element to its behaviorally fixed state was examined by trying all M mutations, one a t a time, and examining the corresponding FLOW matrix for each mutant net. Two general results emerge: (1) Any single mutation of this type tends to affect from none to about one-third of the L2 flows among the L state cycles. (2) Some flows are far more resistant to alteration by mutation of any one of the M than are others. The first would be important in the capacity to evolve in specific ways, the second might play a role in the canalization of behavior of a genetic control system (Rendel, 1967; Fraser, 1970). The discussion of this section led us to distinguish between what might be called DEVELOPMENTAL genes, and CELL TYPE genes, in organisms. That is, there are likely to exist in organisms, genes which are either normally inactive on all cell types, or active on all cell types. Mutation of these to and from constitutively inactive, or active, states would be expected to alter differentiation among cell types, but not the cell types themselves. Such developmental genes would be expected to be members of forcing structures when not mutated to a constitutive form. I n particular, the easiest way to add new genes to a genetic control system in such a way that all old cell types are left unaltered, is to add them in such a way that they are held inactive by a repressing forcing input from an element which is itself held always active bjj being a forced member of a forcing structure. Therefore, the newly added developmental genes would themselves be members of a forcing structure, and their forced value would be 0. Since this is the easiest way to add new developmental genes yet keep old cell types, this hypothesis leads to the suggestion that the genetic control systems of higher metazoans are likely to have a large number of (developmental) genes which are normally inactive on all stable cell types. XI. Expected Character of Forcing Structures as a Function of the Number of Forcing Connections We have seen that homeostatic behavior and other global behaviors which seem to fit many global features of cellular gene control systems

5.

G E N E REGULATION NETWORKS: A THEORY

171

occur in switching nets with few inputs per component, and rich in forcing structures which constrain their behavior. Let R be the number of forcing connections in a net of N components. We now examine the kinds of forcing structure to be expected as a function of R relative to N under random construction rules. Selection may or may not have wielded possible real forcing connections in real genetic systems into highly improbable structures. Our examination will give a background for further evaluation. We define a WEAK FORCING STRUCTURE to be a set of genes each of which is connected to at least one other member of the set by a forcing arrow, but with no regard for the direction of the arrow. From Erdos and Renyi’s (1960) work on random graphs, when R is very small relative to N, the switching net contains very many very small weak forcing trees. When R = N/2, a threshold has been reached and most of the small weak forcing trees have grown together into one giant weak forcing structure, with few small separate forcing trees. Further, R = N appears to be a second threshold a t which directed forcing loops become plentiful. We saw above that such forcing loops would be expected to fall to a steady state with all elements at their forced value, and subsequently all members of the loop’s descendent forcing tree would fall to fixed state. For K = 2 nets, R x 0 . 9 N ; while for K = 3, R x 0.5N and for K = 4, R 0.004N. Thus only randomly built K = 2 nets have large forcing structures. With more inputs per element, forcible Boolean functions must be chosen from among the many nonforcible functions, in order to obtain large forcing structures. Thus in switching nets with few inputs per component and rich in forcing connections, we may expect one huge weak forcing structure extending like a reticulum throughout the net, and several small separate weak forcing structures. Embedded a’s subgraphs within the different weak forcing structures will be distinct directed forcing loops with their descendent directed forcing trees, and different acyclic directed forcing structures. Scattered throughout will be pockets of nonforced elements. On this model of real genetic control systems, the constant pattern of activity provided by many directed forcing structures in their fixed forced states would provide a basic aspect of homeostasis in the system. Remembering that a forcing loop can be maintained in its NONforced state by ensuring that no element of the loop is ever induced t o its forced value, distinct cell types would consist of diverse sets of forcing structures in their fixed forced states, plus additional patterns of oscillation of the pockets of nonforced elements.

-

172

STUART KAUFFMAN

XII. Control Advantages of Forcing Structures I n Section V we discussed one-input control systems. The advantage, with a hierarchically ordered descendent branching tree, is that a single input, like a hormone, to the highest gene in the tree can control the activities of the many genes in the cascade below it. The two major disadvantages were: (1) the lack of control redundancy, such that any single mutation could disconnect all genes below the mutation from higher control; (2) also, one input control systems with structural loops behave on long state cycles, have many state cycles, and show neither much homeostasis nor restricted local reachability. By contrast, an extended forcing structure maintains the control advantages of a one-input system, but overcomes the disadvantages. I n a large forcing structure, any component, placed in its force value, determines uniquely the subsequent states of the members of its descendent forcing structure. Hence, input to such a gene by, say, a steroid, which put that gene in its forced value, could uniquely determine the activities of many genes. Unlike one-input systems, only one value, 1 or 0, of a member of a forcing structure propagates uniquely. Since a gene which is forcible on more than one input line is always forcible to the same forced value on all those lines, forcing structures allow, and commonly have redundancy of control. A forcing value on any input line forces the element t o the same binary value. Loss of one input line by mutation need not greiatly alter the system's behavior. While one input structures with structural loops show little homeostatic return when perturbed, forcing structures, particularly those with structural loops, exhibit marked homeostatic behavior. Directed forcing structures are particularly well suited to play a role in propagating a differentiated state in a proliferating cell population. Maintenance of a given differentiated state presumably requires establishment of the same pattern of gene activities in both daughter cells. It is obviously advantageous if the control system to do this has

/'J\

FIQ.8. A strong forcing structure of three elements, S. Each element in S can reach all elements in A in exactly four steps: that is, S' is complete. If any element is currently forced, all are simultaneously forced 4 moments later, and remain forced thereafter. Strong forcing structures do not require an element which forces itself, but only that there be an integer R , for which SR is complete.

5.

G E NE REGULATION

NETWORKS:

A THEORY

173

sufficient homeostasis that erroneous activities of a few genes in a daughter cell can be corrected. Forcing loops, and particularly strong forcing structures (Fig. 8 ) , which have the property that if any single member of the structure is in its forced value, then all elements of that structure will be fixed in their forced values a short time later, could provide just the required homeostasis. XIII. Molecular Mechanisms

If molecular mechanisms which could realize forcible functions, and so generate forcing structures, were hard for organisms to build, it would be correspondingly difficult to imagine that gene control systems are rich in extended forcing structures. It seems to be relatively easy, however, to build molecules which realize functions forcible on one or more control input, and might be rather harder to build molecular mechanisms which at the same time can be reliably controlled to be nearly off or nearly on, and which realize nonforcible functions. Consider a hypothetical allosteric enzyme (or repressor molecule) with one kind of allosteric site. Suppose that site binds two input metabolites, A and B, and both cause the enzyme (repressor) to be active. Then it is easy to suppose that if A alone causes the enzyme to be active, and so does B alone, then in the presence of both A and B, whichever binds to the allosteric sites, the enzyme will be active. I n this simple case, the enzyme realizes the OR function on its inputs, and can be forced by either input to be active regardless of the presence of the other input. On the other hand, suppose the enzyme is to realize the nonforcible function EXCLUSIVE OR, where it is activated by A alone, or B alone, but both. It is hard to imagine this occurring; it requires that either A or B when alone activates the enzyme by binding to the allosteric sites, but that both jointly present fail to do so. Three ways this might happen are if: (1) A and B formed a product which is unable to bind to enzyme-but we note that there is no evidence that metabolite inputs to operons form such products; (2) conceivably A and B would distort any single copy of the allosteric site in different ways and if the enzyme were a multimer, it might be unable to bind further inputs once one A and one B were bound-which seems implausible; (3) finally, if the allosteric enzyme had two distinct kinds of allosteric sites, one for A and one for B, then the EXCLUSIVE OR function could be realized readily. This point shows that one of the most obvious ways to realize a nonforcible function of two inputs is to have two kinds of sites; with only one kind of site, i t is easier to make forcible functions. In general, with M kinds of sites on a molecule, all 2'" func-

174

STUART KAUFFMAN

tions can be realized. If M is much smaller than the number of input species competing for those sites-say M is l-and if all inputs to one kind of site have effects with the same sign, repressing or enhancing, then it is easy to make functions forcible on all input lines, but hard to make nonforcible functions. If different inputs to the same (say single) kind of allosteric site have effects of different sign, then so long as there is a fairly wide diversity in the binding constants for the different inputs, that input with the strongest binding constant will tend to be a single forcing input. If, instead of an allosteric enzyme, we suppose the protein which is modified by control input is, say, the repressor protein for a given operon, we see it seems not likely to be hard to make molecular devices which give forcible functions for transcription control. A current speculative model for transcription control (Georgiev, 1969) pictures a promoter region, then a linear string of separate operator loci before the structural gene. A polymerase is supposed to bind to the promoter and move past the operator loci, if they are all empty, to the structural genes. If any operator locus is occupied with a repressor, the polymerase is blocked, and transcription of the structural genes does not occur. I n this scheme, every operator locus is a forcible input line with veto power over transcription of the structural gene. It is rather difficult to imagine how such a device could realize an arbitrary nonforcible function on the set of K operator loci, such that, for example, with every other operator occupied it would transcribe, but with every third one it would not. Thus, the linear character of DNA itself markedly favors realization of forcible functions and renders nonforcible functions unlikely. XIV. Additional Evidence for the Theory We note a t the outset that the major strength of the theory is also its major weakness. The theory is not concerned with particular dynamic systems, but with the behavior of entire classes of dynamic systems. The strength of such a theory is that it offers hope of explaining some of the ubiquitous global properties of cellular gene control systems in terms of membership of all those control systems in the same “good” class of dynamic systems. However, the theory makes no statements at all about the detailed pattern of behavior of any hypothetical member of the class of systems studied, so it cannot be directly tested by attempting to describe in detail, for example, the modes of biochemical oscillations of any given cell type. How, then, can such a theory be tested? There are two obvious directions to look: to the structure of the control net for forcible elements and forcing structures; and to known or new global behaviors which are purportedly predicted and explained.

5.

GENE REGULATION

NETWORKS:

A THEORY

175

A. FORCIBLE OPERONS The by now classical bacterial operons supply the most direct evidence that genes actually realize forcible functions on their control inputs in real cellular gene control systems. Consider a typical inducible enzyme such as ,f3-galactosidase. A repressor protein (Riggs et al., 1970) binds, apparently fairly firmly, to the operator locus. A metabolic input, lactose, specifically removes the repressor, and allows transcription. The operon is forcible to the active state by both inputs, for if the regulator gene making the repressor is inactive, then the operon is active; if the metabolite is present, the operon is active. As noted for forcible functions multiple forcing inputs must force i t to the same forced value. Here both forcible input lines force the operon to the same forced valueactive, neither input alone can guarantee that the operon is inactive. The operon is only forcible to one value, not both, as we showed is characteristic of forcible functions. The same considerations apply to a repressible operon with its aporepressor and corepressor. Each, by being absent, can force the operon to be active, neither can alone force it to be inactive. If catabolite repression is considered, then the lac operon has two additional inputs: cyclic AMP and the associated protein are jointly necessary for polymerase attachment to the promotor. This system, realizing the AND function, is forcible off by absence of either input. The lac, considered to have four inputs, is forcible off on two input lines. If both cyclic AMP and protein are present, lac exhibits “nested” forcibility, that is, it is forcible on by either remaining input. There appears to be considerable evidence that allosteric enzymes can realize forcible functions on their inputs. I n a t least three places where several important end products are branches in a reaction sequence of intermediate metabolism, the joint action of two or more of the end products is necessary to inhibit the activity of enzymes concerned with the common path. Such multivalent repression (Cohen, 1968) of aspartokinase I and homoserine dehydrogenase I, each by the pair threonine isoleucine, has been well established. Further, valine, leucine, isoleucine, and patothenate are well all required to repress the enzymes in the biosynthetic pathways leading to valine and isoleucine (Cohen, 1968). I n all these cases, the enzyme realizes a forcible function, for any input can force the enzyme to be active by being itself absent.

+

B. NUMBEROF INPUTS PER GENE As we have noted, bacterial operons seem to have two highly specific inputs, protein and metabolite, although other metabolites may have some influence on the protein in physiological circumstances. Equivalent

176

STUART KAUFFMAN

data for metazoans is not available, but i t seems unlikely that any gene has very many direct inputs.

C. EXTENDED FORCING STRUCTURES We stress that while several operons appear to realize functions which are forcible by one or more single control input, this does not yet constitute evidence that any gene i forces gene j. The binary relation “element i forces element j” requires that i iteslf be forcible, and that its forced value force j to j’s forced value. The input to each bacterial operon may well be its single constitutive regulator gene which may itself (1) have no control input, and (2) in being constitutively active, may put out its nonforcing value to the operon. Many such small separate control structures in a cell may be best adapted for a bacteria’s need to respond to a highly varying environment by making one, or a few enzymes. The evidence is presented to show genes do realize forcible functions. As we showed before, nets with many forcible elements, even without extended forcing structures, have short and few state cycles. Extended forcing structures, however, appear necessary to control reliably the activity of many genes by input to one gene. We suggest that the occurrence of extended forcing structures is most likely to play a role in higher metazoans, whose cells face a stable environment, and where cell differentiation demands the stable occurrence of quite different patterns of many genes’ activities, controlled by hormone, or hormonelike inputs to one or a few genes. There is no direct evidence supporting the predicted existence of extended connected forcing structures. However, discovery of such structures is experimentally feasible. Such a structure would be characterized by a set of genetic loci each of which had more than one control input, yet which behaved in such a way that a given activity (or inactivity) of some component would propagate a specific pattern of activity and inactivity to several other loci of the set, regardless of the behavior of the remaining inputs to those other loci. The reverse activity of the first component would not propagate in a repeatable way to the remaining loci regardless of the behavior of other inputs. Specific sequences of puffing in dipteran giant chromosomes might be examples of extended forcing structures, but such a conclusion is far from established. Clever (19661, Ashburner ( 1969a,b,c) , Berendes (1967, 1968), and others have shown that a specific sequence of puffs in induced following stimulation of Drosophila melanogaster, Drosophila hydei, and Chironomus tentans larval salivary glands with ecdysone. Although it is tempting to conclude that early puffs cause the occurrence of later puffs in the sequence, this is not clearly established. The number of control inputs to the diverse

5.

G E N E REGULATION NETWORKS: A THEORY

177

puffs is surely not established, and therefore it is far from established that the normal pattern of puffing propagates regardless of the activities of those other control inputs which are not members of the hypothetical forcing structure. Sporulation and germination of bacteria may involve extended forcing structures. Schaeffer e t al. (1964), in Bacillus subtilis, have noted six successive morphologic stages of sporulation. H a l v o m n (1965) concluded that “sporulation can bc viewed as a process in which, following the inactivation of the vegetative genome, a large number of spore specific components are synthesized sequentially.” Sporulation is all or none, once initiated. Srinivasan and Halvorson (1963) isolated a molecular factor that induces sporulation in vegetative log phase bacteria. This factor may be a trigger, and the sequence of genes transcribed might constitute a forcing structure. Similarly, when spores germinate, proteins of the vegetative state quickly occur in an ordered sequence in which each is synthesized in a brief burst. Since the sequence precedes the initiation of DNA synthesis, it cannot be a sequence due to obligate transcription during the linear sequence of replication. This ordered transcription sequence might also be a forcing structure. The data on asynchronous DNA replication, discussed in Section X, coupled with evidence in many species of an increasing rate of DNA synthesis in the first half of S (Cleaver, 1967), are consistent with the hypothesis that replicons, when replicating, synthesize trigger molecules (Cummins and Rusch, 1966) which stimulate subsequent replicons to replicate, and that the triggering structure is a descendent branching tree providing thc accelerating rate of DNA synthesis in early S. Since some deletion mutants survive, it is plausible to suppose that replicons in such a hypothetical replicon net would have more than one triggering input from other replicons each able to trigger the replicon, thereby providing redundancy to resist the lethal consequence of deletion of any member of the triggering structure. Such a replicon net would be an extended forcing structure.

D. DEVELOPMENTAL GENES When discussing the capacity of a gene control system to evolve, we raised the possibility that the easiest ways to alter flows among cell types, or add new cell types, while leaving old cell types unaltered was to fix genes in their forced values irreversibly, or add genes to be held inactive by another active gene. This possibility rested on the fact that in switching nets with few inputs per element, about 0.5N0.7N elements are fixed in the same state of activity, or inactivity

178

STUART KAUFFMAN

on all state cycles. The latter leads to the hypothesis that cellular DNA contains many genes which are inactive on all stable cell types, but active transiently in guiding differentiation from one to another stable cell type. There are, in fact, data suggesting that a large proportion of a metazoan’s DNA complement is inactive; unfortunately, the data are not good. Beerman (1952) and many others (Berendes, 1967, 1968) have noted that only about 15% of the observable bands in Drosophila polytene chromosomes are ever observed to puff, where puffing is presumed a sign of gene activity. However, these data are derived from the restricted number of cell types which have polytene chromosomes, and do not necessarily support the claim that a large fraction of DNA is inactive on all cell types. DNA-RNA rehybridization studies by Davidson (1969) and many others, suggests that only about 2-5% of the DNA is active in transcription in any cell type. The techniques used, however, respond only to highly redundant, not single copy, genes, or DNA sequences, and therefore fails to demonstrate either that any cell transcribes so little of its DNA, or that much DNA is never transcribed in any (say stable adult) cell type. Newer hybridization techniques (Davidson, 1969) are opening the possibility of establishing the activity of genes present in single copies; if so, the existence of gene sequences inactive on all adult cell types might be able to be studied directly. The occurrence of homeotic mutants such as aristapedia are good evidence for genes which control flow among cell types and might be active predominantly during development. It would be of great interest to see whether such genes mapped to loci which could be shown to be inactive in transcription on adult cell types.

E. MACROSCOPIC TESTS I n addition to microscopic data related to the kinds of functions genes realize on their inputs, and the existence of forcing structures, macroscopic behavior of gene control systems should fit the predictions of the theory. We have already noted the roughly square root relation between DNA content and cell cycle time in many organisms, which parallels expected state cycle length as a function of N . Homeostasis and a restriction on the number of modes of behavior any state cycle can reach are typical properties of these nets, claimed to underlie homeostasis and thc restricted capacity of cells to differentiate directly into only a few other cell types. Hadorn’s data are suggestive, but not as helpful as they might be, since they deal with transdeterminations among structures, not among cell types. The model predicts that the same sort of restricted reacha-

5.

GENE REGULATION NETWORKS: A THEORY

179

bility would occur among cell types perturbed to differentiate oddly (Burnett, 1968). The theory predicts that patterns of gene activity in different cell types will differ in only about 10% of the loci,. Berendes (1966) , observing 110 loci in Drosophila hydei in Malphigian, salivary, and stomach cells, found puffing patterns differed in about 10% of the loci between different cell types. New hybridization techniques should allow study of other organisms. The model makes the further prediction that mutants should exist which alter the transdetermination probability between different cell types, i.e., the flow probability, and that some flows should be much more sensitive to alteration by mutation than others. The prediction that an organism has a number of cell types about equal to a square root type function of its quantity of DNA has been discussed and criticized. XV. Alternative Theories We wish t o emphasize two alternative, or a t least complementary, ways an integrated gene control system might achieve its orderly global behaviors. We defined the concept of forcible elements and of a forcing structure earlier, and showed systems rich in such components and structures behaved with biologic-like global order. If the gene can usefully be modeled as a binary switching device, we may ask whether forcible elements are necessary to obtain “good” global behavior. One can build large switching nets without forcible elements or extended forcing structures which have quite localized behavior, but, not really orderly enough, I think, for biological systems. Consider a binary gene with many, K , inputs. There are 2“ possible states of its inputs. We might assign the element a Boolean function which assumed the value 1 for 75% of those 2K input states, and 0 for only 25%. Consider a net built of such biased elements. There will be a marked constraint of sequences of states to converge toward the region of state space where each element is in value 1. Where K = N , and Boolean functions are chosen a t random without such bias, state cycle lengths are the square root of the number of states, and the number of distinct state cycles is ( l / e ) N . I n those unbiased nets, there is little convergence in state space. With strongly biased components, 75 :25, we would expect far fewer distinct state cycles and much shorter state cycles. Nevertheless, Boolean functions of many inputs which yield 1 on 75% of the input states may easily be chosen that are not forcible on any input line a t all. Such nets would show constrained behavior, but contain no forcing structures, or forcible elements. The number of state cycles in such nets is unknown, but probably

180

STUART KAUFFMAN

approaches square root N . However, state cycles, while very short with respect to the state space, are far too long to be biologically meaningful. Monte Carlo study of such nets reveals that cycle lengths a’re 0.0005(N4), which becomes enormously long for N =1000. Increasing the bias of the functions appears unlikely to help as N grows reasonably large. In addition to having long cycles, such systems would not be likely to show restricted local accessibility among cycles under perturbation. Further, lacking forcing structures, no single gene’s behavior could uniquely determine the subsequent activity of a downstream tree of other genes. It is difficult to see how hormone or other signals would work reliably. We conclude that, a t this point, it has begun to appear that the way truly large nets of switching elements are likely to behave with “good” biological global properties, is to be rich in forcing structures. If genes really are usually nearly off or nearly full on, extended forcing structures seem likely to exist in cellular control circuits. On the other hand, genes may be capable of finely graded levels of activity. Obtaining steady-state behavior in normal simple open chemical systems with reversible reactions does not require forcing structures. The extent to which cells achieve their global behaviors through such properties of some continuous dynamic systems is unclear. Perhaps the most reasonable guess a t this point, is that genes may rely on forcing structures to obtain homeostasis and other globally useful behavior, while the intermediate metabolic system may rely much more on nonextreme steady state behavior of its readily reversible reactions.

XVI. Conclusions and Summary The picture of a metazoan cell’s genetic net which emerges is that of a system rich in forcible genes and with extensive forcing structures weaving through it rather like a reticulum, additional isolated forcing structures, and pockets of nonforced genes functionally isolated from one another by the unalterable behavior of the forcing structures, once in their forced states. The forcible elements and forcing reticulum would provide the basic localization, and homeostasis in the behavior of the system. Different cell types would correspond to different sets of isolated forcing structures in their stable forced states or held in their metastable nonforced states, plus different patterns of activity, either steady state or cyclic, of the isolated, nonforced pockets. We began by asking whether several global behaviors ubiquitously present in the gene control systems of organisms offered clues to the construction of those control systems by reflecting some basic ways genetic nets might most readily be built, or whether the ubiquitous char-

6.

GENE REGULATION NETWORKS: A THEORY

181

acteristics were present as highly selective consequences of evolution. The building blocks of the genetic control circuitry are DNA, RNA, and protein. These molecules are particularly good a t exhibiting high molecular specificity, thereby creating dynamic systems with few inputs per component, and rich in forcible elements and forcing structures. Just these s y s t e m almost automatically exhibit the ubiquitous properties mentioned : homeostasis, limited local reachability, a capacity to evolve while keeping old useful cell types, roughly the right number of cell types and right length periodicities, etc. Whether the model is right in detail, and it surely is not, it is nevertheless a real intimation that such a theory can be had, and that there is reason to hope for some common discoverable plan to the organization of genetic control systems. ACKNOWLEDMENTS The author wishes to thank Drs. John Maynard Smith, Stuart Newman, and Leon Glass for many helpful discussions. He is particularly grateful to the Department of Biology, University of Cincinnati, for their generous offer of computer time. This research was supported in part by the Sloan Foundation. REFERENCES Allfrey, V. G. (1968). In “Regulatory Mechanisms for Protein Synthesis in Mammalian Cclls” (A. San Pirtro, M. R. Lamborg, and F. T. Kenney, eds.), p. 65. Bcademic Press, Ncw York. Ashburner, M. (1970). Personal communication. Ashburner, M. (1969a). Chiornosoma 27, 47. Ashburner, M. (196913). Chromosoma 27, 156. Ashburner, M. ( 1 9 6 9 ~ ) .In “Chromosomes Today” (C. D. Darlington and K. R. Lewis, eds.), Vol. 2 . Oliver & Boyd, Edinburgh. Beermann, W. (1952). Chromosoma 5, 139. Berendes, H. (1966). J . E x p . Zool. 162, 209. Berendes, H. (1967). Chromosoma 22, 274. Berendes, H. (1968). Chromosoma 24, 418. Berendes, H., and Beerman, W. (1969). I n “Handbook of Molecular Cytology” (A. Lima-de-Faria, rd.), p. 500. North-Holland Publ., Amsterdam. Boezi, J. A., and Cowie, D. B. (1961). Biophys. J . 1, 639. Bonner, J., and Huang. R. C. (1967). Pioc. Nal. Acud. Sci. U S . 48, 1216. Bourgeois, S. (1966). Thcsis, Faculty of Sciences, University of Paris. Britten, R. J., and Davidson, E. H. (1969). Science 165, 349. Burnett, A. (1968). In “The Stability of the Differentiated State” (H. Ursprung, ed.), p. 109. Springer Publ., New York. Cleaver, J. E. (1967). “Thymidine Metabolism and Cell Kinetics.” North-Holland Publ., Amsterdam. Clever, U. (1966). Ann. Zool. 7, 33. Cohen, G. N. (1968). “Thc Regulation of Cell Metabolism.” Holt, New York. Cummins, J. E., and Rusch, H. P. (1966). J . Cell Biol. 31, 577. Davidson, E. H. (1969). “Gene Activity in Early Development.” Academic Press, New York. Davidson, E. H., and Hough, 13. R. (1969). Proc. Nut. Acacl. Sci. U S . 63, 342.

182

STUART KAUFFMAN

Erdos, P., and Renyi, A. (1959). “On the Random Graphs 1,”Vol. 6, Nos. 3-4, Inst. Math. Univ. De Breceniensis, Debrecar, Hungary. Erdos, P., and Renyi, A. (1960). “On the Evolution of Random Graphs,” Publ. No. 5. Math. Inst. Hung. Acad. Sci. Fraser, A. S. (1970). In “Towards a Theoretical Biology” (C. H. Waddington, ed.), Vol. 3, p. 56. Aldine Publ. Co., Chicago, Illinois. Gehring, W. (1968). I n “The Stability of the Differentiated State” (H. Ursprung, ed.), p. 136. Springer Publ., New York. Georgiev, G. P. (1969). J . Theor. Biol. 25, 473. Gilbert, W., and Muller-Hill, B., (1966). Proc. Nnt. Acnd. Sci. U S . 56, 1891. Gilbert, W., and Muller-Hill, B., (1967). Proc. Nut. Acad. Sci. U.S. 58, 2415. Hadorn, E. (1967). I n “Major Problems in Developmental Biology” (J. M. Locke, ed.), p. 85. Academic Press, New York. Halvorson, H. 0. (1965). Slymp. SOC.Gen. Microbial. 15, 343. Herzenberg, L. A. (1959). Biochim. Biophys. Acta 31, 525. Jacob, F., and Monod, J. (1961). Cold Spring Hnibor Sump. Quant. Biol. 26, 193. Jacob, F., and Monod, J. (1963). In “Cytodifferentiation and Macromolecular Synthesis” (M. Locke, ed.),p. 30. Academic Press, New York. Kauffman, S. (1969a). J . Theor. Biol. 22, 437. Kauffman, S. (196913). Nature (London) 224, 177-178. Kauffman, S. (1971a). “Lectures on Mathematics in the Life Sciences,” Vol. 3. Amer. Math. SOC.(in press). Kauffman, S. (1971b). In preparation. Kauffman, S. (1971~).Cybernet. 1, No. 1, p. 71-96. Kim, J. H., Gelbard, A. S., and Perez, A. G. (1968). Ezp. Cell Res. 53, 478-487. Lima-de-Faria, A., ed. (1969). “Handbook of Molecular Cytology.” North-Holland Publ., Amsterdam. Monod, J., Changeaux, J., and Wyman, J. P. (1965). J . Mol. Biol. 88, 12. Newman, S., and Rice, S. (1971). Proc. Nat. Acad. Sci. U.S. 68, 92. Okada, S. (1968). Biophys. J . 8, 650. Painter, R. B. et al. (1966). J . Mol. Biol. 17, 47. Ptashne, M. (1967%).Proc. Nut. Acad. Sci. U.S. 57, 306. Ptashne, M. (196713). Nature (London) 214, 232. Rendel, J. M. (1967). “Canalization and Gene Control.” Logos Press, London. Riggs, A. D. et al. (1970). J . Mol. Biol. 51, 303. Schaeffer, P., Ionesco, H., Rytes, A,, and Balassa, G. (1964). Colloq. Int. Cent. Nat. Rech. Sci. 124, 553. Shires, T. K., Kauffman, S., and Pitot, H. C. (1971). “The Membron-A Functional Hypothesis for the Translational Regulation of Genetic Expression, in Biomembranes,” Vol. 2. Dekker, New York (in press). Spirin, A . S. (1966). Curr. Top. Develop. Biol. 1, 1. Srinivansan, V. R., and Halvorson, H. 0. (1963). Nature (London) 197, 100. Tomkins, G. (1968). I n “Regulatory Mechanisms for Protein Synthesis in Mammalian Cells” (A. San Pietro, M. R. Lamborg, and F. T. Kenney, eds.), Academic Press, New York. Waddington, C. H. (1962). “New Patterns in Genetics and Development.” Columbia Univ. Press, New York. Walker, C., and Ashby, W. R. (1965). Zcybernetics 3, 100. Walters, C. et al. (1967). J. Theor. Biol. 15, 208.

CHAPTER 6

POSITIONAL INFORMATION AND PATTERN FORMATION Lewis Wolpert DEPARTMENT O F BIOLOGY A S APPLIED TO MEDICINE, THE MIDDLESEX HOSPITAL MEDICAL SCHOOL, LONDON, ENGLAND

I. Introduction .............................................. 183 11. Pattern and Form.. ....................................... 184 111. French Flag Problem.. .................................... 186 IV. Pattern Regulation.. ...................................... 188 A. Morphallaxis and Epimorphosis. ......................... 188 B. Interaction between Parts of the Pattern.. . . . . . . . . . . . . . . . . 191 V. Universality and Prepatterns. .............................. 192 VI. Model Systems and Mechanisms.. .......................... 196 A. Diffusion.. ..................................... 198 B. Insect Inte and the Homeostatic Model.. . . . . . . . . . . 198 C. Hydra and the Follow-up Servo Model.. .................. 200 D. Changes with Time ... . . . . . . . . . . . . . . . . . . 201 E. Active Transport a ermeability.. . . . . . . . . . . . . . . 204 F. Periodic Signaling and the Phase-Shift Model 205 G. Validation of Models. .................................. H. Boundary Regions. ..................................... 205 VII. Polarity ............................... 207 VIII. Intercell ................................ 209 IX. Interpretation. . . . . . . . . . . . . . . . . . . ..................... 211 X. Precision ........................ ..................... 212 XI. Cell Movement.. ......................................... 215 XII. Mosaic Development. . . . . . . . . . . . . . . . . . . . . . . . . . . 217 XIII. Growth and Cell Division.. ........................ . . . 218 XIV. Spacing Patterns.. ................................ . . . 219 XV. Pattern Formation in Plants.. .............................. 220 XVI. Conclusions.. ............................................. 220 References. . ....................... 221

1. Introduction The expression of genetic information in terms of pattern and form is a central problem, not only for developmental biologists, but for biology as a whole. The translation of genetic information into shapes and patterns is what links genetics to morphology-genotype to phenotype-and must have crucial consequences for a variety of central biological issues from evolution to learning. For example, it is becoming 183

184

LEWIS WOLPERT

increasingly clear that if we understood the mechanism whereby nerve connections were formed (Gaze, 1970) we might have a much better chance of understanding the neural mechanisms involved in learning. Does the genetic material provide a description of the adult? What, one may ask, are the genes for leg formation in tetrapods, and how do they make a leg? Or, what are the genes for gastrulation? A current fashion in molecular biology is to suggest, either explicitly or implicitly, that the answers t o such problems will come from deeper and deeper molecular probings. Characterize the RNA and proteins and the form will look after itself, or at least be immediately explicable (Lederberg, 1967). Such a view suggests that if we understood cytodifferentiation or molecular differentiation then pattern would be explicable. I wish to take a rather different view and would suggest that the development of form and pattern, while related to molecular differentiation, can be viewed in their own right. Moreover, the rules, laws, or principles for the expression of genetic information in terms of pattern and form will be as general, universal, elegant and simple as those that now apply to molecular genetics (Wolpert, 1969, 1970). Again in contrast to the prevailing ethos, I would suggest that the solution to the problems must initially he sought not a t the molecular level but a t the cellular level, since until the cellular basis is understood, the correct molecular questions cannot be posed. In this paper I wish to consider some general aspects of pattern formation in terms of positional information. Inductive processes will not be discussed, and I regard the misuse of concepts of induction as a major feature preventing progress in understanding pattern formation. II. Pattern and Form

A useful but by no means absolute distinction between form and pattern (cf. Waddington, 1962), is that form involves cell movement and changes in shape, its genesis requiring an understanding of the forces involved, whereas pattern does not involve changes in shape or cell movement, but rather the specification of spatial differences. Thus, the distinction rests in the nature of the cellular processes involved. Pattern formation will usually precede cell movement since it is necessary to specify which cells will move, and where. In very general terms, the pattern problem may be stated as follows: given an ensemble of more or less identical cells, how can states be assigned to these cells such that when they undergo molecular differentiation the cells will form a well defined spatial pattern. The difference between molecular or cytodifferentiation and spatial specification is a crucial one. Molecular differentiation can be viewed in terms of the control of synthesis of specific macromolecules,

6.

POSITIONAL INFORMATION AND PATTERN FORMATION

185

particularly luxury molecules to use Holtzer’s (1971) evocative terminology. The process of pattern formation, on the other hand, must be viewed in terms of the spatial organization of such activities. For example, cytodifferentiation in a tetrapod arm and leg is probably very similar with respect to eary muscle and cartilage development : the difference between an arm and leg lies in the spatial arrangement of these processes. A suggested solution to the pattern problem is that the cells are assigned positional information which effectively gives them their position in a coordinate system, and this positional information is used to determine the cell’s molecular or cytodiff erentiation (Wolpert, 1969). An example of the relationship between pattern and form comes from sea urchin gastrulation which involves pseudopodal activity by mesenchymal cells providing the cellular forces for invagination (Gustafson and Wolpert, 1963, 1967). The form resulting is due to the deformation of tlie cell sheet. The pattern aspect of this problem is the assignation to a particular group of cclls of the cellular property of pseudopodal activity. Our studies 011 tlie carly development of form in the sea urchin embryo (Gustafson and Wolpcrt, 1963, 1967) and other studies (see Trinkaus, 1969) suggest that the cellular forces responsible for cell movement and the generation of form are relatively few in number and are repeatedly used in a variety of animals and processes. I n particular, localized cellular contractions whether in pseudopods or near the cell surface play a central role. This type of cellular force has been invoked for cleavage, blastula formation, gastrulation, formation of the coelom and primary pore canal in tlie sca urchin (Gustafson and Wolpert, 1963, 1967) ; contraction of the ascidian tail a t metamorphosis (Cloney, 1966) ; the initial phase of gastrulation and ncurulation in Amphibia (Baker, 1965; Schroedcr, 1970; Gingell, 1970) ; and the development of convolutions in the salivary gland (Spooner and Wessells, 1970). These studies immediatcly provide the possible link between gene action and form since it presents no conccptual problem to discuss contractile processes near the cell surface in terms of gene action, however ignorant we remain of the mechanisms (Wolpert, 1971). It is worth recalling that virally transformed cclls often involve changes in cell surface properties and cell motility, and it is possible that from such studies one might obtain the best clucs as to how gcnes affect surfaces and motile processes (e.g., Dulbecco, 1970). One important aspcct of such mechanisms for generating form by cellular forces is that the specification of how to bring about the change in form is almost always very much easier than describing the changes

186

LEWIS WOLPERT

that result. It is, for example, easy to specify a mechanism for gastrulation in sea urchin embryo, but rather difficult to describe the final result. We have used this point to emphasize that the genetic information in the egg does not contain a description of the adult but a set of instructions on how to make it and that this set of instructions is much simpler than any description (Apter and Wolpert, 1965). It should be a morphogenetic maxim to avoid inferring developmental mechanisms from final form or pattern. 111. French Flag Problem

The formulation of the pattern problem given above is perhaps misleading, starting as it does with the cell ensemble, since most real biological problems deal with situations in which the ensemble is derived by cell division from a single cell, the egg. In many situations pattern is specified during or following subdivision of the egg. Viewed in these terms one can distinguish between two extreme situations. I n the one there is very little interaction between the cells and the differences that arise between the cells reflects differences already existing within the egg. I n the other, the differences arise due to intercellular signaling. The experimental basis of this difference is that in the former, the socalled mosaic case, removal of a small piece of the system results in a defect, whereas in the latter, or regulative systems, it does not. I n general, most systems show both aspects, but a t different times (Weiss, 1939; Davidson 1969). Another way of looking a t the problem is to consider it in terms of presumptive regions and potentialities. Presumptive regions are classically defined as indicating which cells will normally develop into a particular structure such as eye or limb. Following Driesch (1908) and Spiegelman (1945) one might formulate a DrieschSpiegelman law along the following lines: the amount of material capable of developing into a particular region is always greater than the presumptive region. Since the potentiality is always greater than the expression, some mechanism for restriction of the expression is required. It will be important to know whether the law is universally true, particularly in mosaic eggs (Section XII) . The situations in which cell-to-cell communication is clearly dominant are those situations in which regulation occurs, either during development as in sea urchin or amphibian development, or in regeneration as in hydra or amphibian limbs. These can be illustrated by the French Flag Problem (Wolpert, 1968) (Fig. 1 ) . This problem is concerned with the necessary properties of cells in a line, and the intercellular communication between them, such that if each cell has three possibilities for molecular differentiation-blue, white, and red-the system always

6.

POSITIONAL INFORMATION AND PATTERN FORMATION

187

FIG.1. The French Flag Problem. A line of cells (a) or a shorter line (b) undergo molecular differentiation so that one third are blue (B), white ( W ) , and red (R), as seen in (c) and (d). This may be achieved by providing the cells with positional information in the form of a linear gradient whose boundary values a and 01’ are fixed, and having rules for interpretation based on thresholds an and aw. Another way of providing positional information is to have two gradients as in (g) and (h), and basing the rules for interpretation on the ratio between them. Note that in this case the boundary values and the slopes of the gradients remain constant with changes in length of the line. forms the French Flag irrespective of the number of units or which parts are removed. I n terms of positional information, the solution is that each cell is assigned, by appropriate signals, a positional value which specifies its position with respect to the two ends. The cells interpret this positional value by turning on the genes for blue, white, or red, and so form the French Flag. The positional signal may involve, for example, a diffusing substance or might be the time difference between two periodic signals (Section V I ) . I n both cases they would specify a positional value which may be looked upon as a cellular property which was graded from one end of the line to the other. It could be a gradiant in one, or two, substances or some more complex cellular property. The crucial feature is that the positional value provides the cell with its position within the system and that this value is used together with the cell’s genome to specify its molecular differentiation. This process is defined as the interpretation of the positional information. The set of cells which have their position specified with respect to the same set of points is defined as a field. It will be seen that an essential

188

LEWIS WOLPERT

feature of regulation of fields is the reestablishment of boundary values and the reassignment of positional information. The idea that position is important in determining how a cell will differentiate is a very old one going back a t least to Driesch in the 1890’s (see Wilson, 1925) and classical gradient theory (Child, 1941) has long implied that something like position is important. These ideas, however much they may be part of the embryologists preconscious thoughts have always been rather vague and have been little analyzed or used in recent years. The concept of positional information attempts to define a much sharper and more useful framework and draw attention to several implications that immediately arise once i t is accepted that a cell is having its position specified, and that this information is interpreted in terms of cytodiff erentiation. IV. Pattern Regulation

A. MORPHALLAXIS AND EPIMORPHOSIS Although these are rather ghastly words, they have a long history and very usefully distinguish between two types of regulation. It was recognized a long time ago that regeneration was of two main types and we can here make an extension to regulation in general. Morphallactic regulation involves remodeling of the missing parts from the residue whereas in epimorphosis there is growth of tissue which then forms the regenerated region. While many cases of regulation might involve both types it is helpful to consider two ideal cases. Consider the regulation of a simple linear system, the French Flag (Fig. 2 ) . If the blue region together with part of the white is removed, then in morphallactic regulation the remaining white and red region regulate, without growth, to give the Flag again. I n terms of positional information the cut surface becomes the new boundary, with the appropriate boundary value, and all the cells are assigned new positional values with respect to it, and are able to interpret their new positional value. Morphallactic regulation probably applies to systems like hydra regeneration and sea urchin development. By contrast, in epimorphic regulation there is proliferation of cells at the cut to give rise to a blastema which then forms the missing parts. I n terms of positional information the positional value a t the cut is unchanged and the end of the blastema takes on the boundary value. Positional values are assigned within the blastema. This type of regulation probably occurs in vertebrate limb regeneration and regulation of insect imaginal discs. These examples should serve to illustrate the importance of the mechanism whereby boundary values are established. I n morphallaxis

6.

POSITIONAL INFORMATION AND PATTERN FORMATION

I

B

I

I

W

189

R

XI

FIO.2. Regulation of the Frcnch Flag by epimorphosis and morphallaxis. The pattern is cut a t level X-X corrcsponding the ktli cell, and the left-hand piece is removed. In morphallaxis the valuc of the positionnl information a t the level of the cut bccomes a, and a new gradient is established. With epimorphosis there is growth from a blastema a t the lcvel of the cut to replacc the lost part of the pattern.

the problem is to assign the houndary value to the cells a t the cut surface ; a possible mcchanism for doing this, with specific reference to hydra, will be given in Section V1,C. With epimorphosis, the problem a t this stage is different and rather more complex, for the cells a t the cut surface maintain their value and new positional information is assigned within tlie blnstcma. The problem hcrc is to establish the appropriate boundary values within the blastema, the extreme boundary value a t the distal end, and a, proximal boundary valuc appropriate to the level of cut. This latter houndnry value ensures integration of the blastema witli tlie stump. JYIiilc the distal boundary value might be established by a mcchani~msimilar to that in hydra, special attention must he given to the proximal boundary value. A key question is whether the establishment of this proximal boundary involves signaling from stump to blastema, a view which is implicit in the literature (see GOSS, 1961). I would like to suggest that there is no signaling from stump to blastcma, but that the proximal boundary value is determined by an intrinsic property of the cells that form the blastema. The suggested

190

LEWIS WOLPERT

rule is that cells cannot have positional values more proximal than the level from which they were derived but that they can have positional values corresponding to more distal regions. This immediately provides an explanation of the rule of distal transformation for limbs (Rose, 1962), which states that in normal regeneration the structures regenerated are always those distal to the cut surface. This occurs even if i t is the proximal face of the cut surface that is exposed, in which case a mirror image regenerate is formed (Fig. 3 ) . Further evidence

B

I

B

w I

R

I

I

W

R

I

B

lw

R

FIG. 3. Regeneration and duplication in epimorphosis. The right-hand piece regenerates the pattern as in Fig. 2, but the left-hand piece duplicates itself since the cells forming the blastema cannot take on positional values corresponding to positions to their right; that is, positional values corresponding to those between the kth and Nth cell.

in support of this view comes from culturing early blastemas removed from the stump. Stocum (1968) and de Both (1970) have shown the blastema to have remarkable powers of self-differentiation. I n most cases the cultured blastema forms only structures distal to the stump, but de Both (1970) has claimed that by combining blastemas more proximal structures formed. If this is confirmed the suggested mechanism will clearly require modification. Another test of the theory will be provided by reciprocal grafting of blastema from proximal to more distal regions. For example a blastema from an amputation surface near the shoulder when placed on a stump from a cut near the wrist, should result in an additional set of mid-arm bones, radius-ulna, being formed.

6.

POSITIONAL

INFORMATION

AND PATTERN FORMATION

191

It is of considerable interest, and suggestive of similar mechanisms, that phenomena relating to the rule of distal transformation have been reported in other systems. Preliminary results on the regulation of halfeyes during development of Xenopus suggest that temporal halves regulate normally, but nasal halves give reduplication (Gaze, 1970). Schubiger (1971) using in vitro techniques has shown for the imaginal leg discs of Drosophila that regenerative ability was restricted to the upper medial quarter of the disc, whereas other regions show a high tendency to give reduplicated structures. Bryant (1971) operating in situ has obtained substantially similar results. These results could be explained by regulation from a blastema-like region originating a t the cut surface as suggested from the amphibian limb. This explanation may even be extended to account for what Nothiger and Schubiger (1966) have termed proliferative regulation and in which only parts already present are duplicated (see Lawrence, 1970, for review of this and related phenomena). The crucial experimental observation was that a half genital disc would normally regenerate its symmetrical mirror image, but if part of that half genital disc was irradiated with ultraviolet light so as to eliminate the presumptive clasper region, then claspers would be absent from both sides after regeneration. I would suggest that this might result from UV damage to some of the cells that would give rise to the blastema, and this would lead to absence of the structures with those positional values. The system behaves in some ways like an anterior half of a limb, which only regenerates an anterior half (Goss, 1961). B. INTERACTION BETWEEN PARTS OF THE PATTERN I n terms of positional information, pattern formation is a two-step process : first the specification of positional information within the field and then the interpretation of this information by the cells which results in the formation of the pattern. With this formulation there is no interaction in a field between parts of the pattern as distinct parts. This is a very important point and is quite different, for example, from Rose’s concept (1970) of pattern formation, which is based on specific inhibition passing from one region to another, or any theories which rely on sequential spatial induction. For example, in the French Flag there would be no interaction between the blue, white, or red regions as separate regions. On this view, hydra, for example, would comprise a single field and the hypostome and tentacles would be regarded as the morphological expression of the cell’s interpretation of positional information (Wolpert et al., 1971). Cellular interactions and communication are of course required for specifying positional information.

192

LEWIS WOLPERT

There is the possibility that interactions between differentiated regions might play some role in pattern formation although not in forming the primary pattern. They might be involved in sharpening up boundaries through localized inhibition (see Section XIV) . V. Universality and Prepatterns One of the most interesting features of positional information in relation to pattern is that there is no unique relationship between the form of the gradient in positional value and the pattern which results from the cells’ interpretation of it. Using the identical gradient in positional value one can get completely different patterns, simply by changing the rules for interpretation. For example, in a two-dimensional case, different rules for interpretation with the same coordinate system will give in the one case the French Flag, another the Union Jack, and another the Stars and Stripes (and even the cytodifferentiation is the same in the three cases). There is thus the possibility that the mechanism for specifying positional information might be universal. This view of pattern formation must be contrasted with those views which explicitly or implicitly claim that in order to make a pattern it is necessary to generate a spatial variation in something which resembles in some way the pattern. For example, the French Flag would require the generation of a steplike function. This view of pattern formation is characterized by the work of Turing (1952), Stern (1968)’ Maynard-Smith and Sondhi (1961), Gmitro and Scriven (1966), and Waddington (1962, 1971) and is t,he antithesis of the positional information approach. All these workers have considered the appearance of an overt pattern to be the expression of an underlying prepattern. I n general, this approach makes use of singularities, that is, local maxima or minima in continuous curves to specify special features of the pattern. An example based on the work of Stern (1968) is given in Fig. 4, in which a discrete region is specified along a line of cells-for example, a bristle. I n terms of prepattern concepts this requires a local maximum or minimum a t this region whereas in terms of positional information this is not necessary. The mechanism for changing the site of the region requires in the prepattern case the shift of the peak whereas in terms of positional information it is the cell’s interpretation that alters. This approach logically requires a different prepattern for different patterns and poses severe problems as to how they could be generated. The concept of a universal mechanism for positional information gets over all these problems and does not require different patterns to be generated by generating a prepattern first. The extreme universalist view, to which I a t present adhere, is that the mechanism for specifying

6.

POSITIONAL INFORMATION A N D PATTERN FORMATION

193

I

FIG. 4. The differences between mechanisms of pattern formation based on prepatterns and positional information. I n (a) one cell in a line undergoes a specific cytodiffercntiation, and in prepattern terms this requires a singularity in some property at that region (11). If, as in (c) the site of cytodifferentiation changes, then so will the singularity, as in ( d ) . In terms of positional information, there are no spatial singularities (c) and the patterns arise from the interpretation of the cells. For (a) the cells respond t o a positional value a. whereas for the second pattern (c) the cells eytodifferentiate only for positional values ac.

positional information and interpreting it is the same in all multicellular biological systems, not only at the cellular, but at the molecular, level. This implies a universal coordinate system. It means that while boundary conditions and sizes of fields may vary, the positional signals and positional values are always the same. Thus all fields would be similar and, from a cell’s point of view, indistinguishable and cells would behave according to position and genome. Onc could imagine that if presumptive amphibian neural tissue were placed a t the proximal end of an insect imaginal leg disc it would become forebrain, while at the distal end it would form hindbrain. Such experiments would be hindred, not least, by surface incompatibilities preventing intercellular channels for signaling being formed. The general principles are illustrated with respect to Flags in Fig. 5 and predicts autonomy of interpretation with respect to position. While extreme experiments of the frog/insect type have not

194

LEWIS WOLPERT

(b)

(f)

FIQ. 5 . Some examples to show some possible implications of the universality of positional information. Consider a rectangular field and two different genotypes. Genotype fr results in the interpretation of the positional information so that a French Flag is formed (a) while genotype us results in the Stars and Stripes (b). If, at an early stage, two pieces are interchanged as in (c), and if positional information in the two fields is the same, then the results shown in (d) and (e) will follow: that is the cells behave according to their genotype and position and are indifferent to the nature of the surrounding tissue. Similarly, if two halves of different genotypes are joined as in (f) a mosaic as in ( g ) will form (B is blue, W is white, R is red). From Wolpert (1969).

been performed there is a considerable body of evidence to support the idea of a universal coordinate system as illustrated in Fig. 5 , and this is summarized in Table I. The experiments have been done either by grafting or by making genetic mosaics, and in almost every case the cells have behaved according to the predictions of a universal coordinate system: that is they develop according to their position, genome, and developmental history. The most detailed analysis along these lines has been made by Stern (1968) using genetic mosaics. He set out to test whether pattern mutants in insects were due to a change in the prepattern or to the cells' com-

6.

POSITIONAL INFORMATION AND PATTERN FORMATION

195

TABLE I

POSITIONAL FIELDINVARIANCE AND AUTONOMY OF INTERPRETATION Insect (Drosophila) Bristle formation or absence with genetic mosaics of wild type and achaete (ac), scute (Sc), hairy wing (Hw), and Tufted Insect (Drosophila) Sex comb on leg: gynandric mosaics and mosaics with wild type and mutants sexcombless (Sz) and engrailed (en) Insect (Drosophila) Bithorax mosaics Insect (Drosophila) Genetic mosaic of antenna and legs using aristapedia and wild type Insect (Drosophila) Mosaic of antenna and mesothoracic leg from antennapedia Regional character of cuticle when Insect (Galleria) grafted to different segments Contiguous grafting of forelimb and Insect (Ephestia) hindwing hinges Insect (Vanessa) Grafts between forelimb and hindlimb Sea urchin Amphibian Amphibian Birds

Chick Chick

Arnheim (1967), Stern (1968) Stern (1968)

Lewis (1963) Roberts (1964) Postlethwait and Schneiderman (1969) Stumpf (1968) Kroeger (1960)

Bodenstein (1935) ; (see Lawrence, 1970) Grafts of mesenchyme one species with Horstadius (1936a) ectoderm another gives skeleton of mesenchymal type Holtfreter and HamAnuran suckers on ventral side head burger (1955) urodele, and horny teeth in mouth region, by grafting Exchange of melanoblasts between Twitty (1949) species: pigmentation pattern is of melanoblast genotype Rawles 1948, 1955) Exchange of melanoblasts between varieties of domestic fowl, pheasants, guinea fowl and some common song birds: pigmentation pattern and color is of melanoblast genotype Kieny ( 964) Contiguous grafts between wing and leg buds Saunders et al. (1957) Graft of presumptive thigh to tip of wing gives toe

petence to respond to the prepattern. A theory based on universal positional information must predict that no mutants affecting one prepattern would be found, since it should not be possible to locally alter the positional information in one field, without affecting all other fields. I n every case except one, he and his co-workers have concluded that the prepattern is unchanged in the mutant, the cells behaving genome auton-

196

LEWIS WOLPERT

omously and according to position. The single case where it has been argued that a prepattern change is involved is that of the effects of the mutant eyeless-dominant (ey”) on Drosophila legs (Stern and Tokunaga, 1967), since wild-type tissue surrounded by the mutant formed extra structures not found in wild-type tissue. A single case demonstrating that a mutant locally alters a field would greatly undermine the concept of a Universal coordinate system. It is thus necessary to point out that in the case of eyD mutants the width of the segment is increased, and the segmentation is disturbed. Both of these could cause an alteration to the boundaries of the field and thus the gradient in positional values, which might produce the observed results (see Wolpert, 1969; Lawrence, 1970). It is a situation which requires further detailed investigation. While most of the data in Table I have been discussed elsewhere (Wolpert, 1969 ; Lawrence, 1970) the experiments with pigment patterns in birds require special comment. A large body of results has been obtained in which melanoblasts have been introduced into feather primordia and in which the host and donor patterns are very different. The consistent result has been that the melanophores interact with foreign feather germs by producing a typical donor color and pattern (Rawles, 1948, 1955). Thus the genetic constitution determines the melanophore response, and this is consistent with the positional information type model. Even more striking is the evidence that the local pattern in any part of the same animal is determined by the position of the feather germ. Thus melanoblasts of different local origin but similar genotype behave differently in different positions in the body. VI. Model Systems and Mechanisms

Mechanisms for specifying positional information may be thought of in terms of the requirements for establishing a linear gradient with fixed boundary values since the absolute value a t any point will provide information as to the relative position a t that point with respect to the ends. A similar result can be achieved by setting up two gradients with fixed boundary values a t opposite ends, the ratio of the values a t any point providing positional information (Fig. 1 ) . I n addition to having to set up the gradient with fixed boundary values, it is necessary to consider how the system could regulate and be made size invariant over a certain range of dimensions. This is essentially a problem of reestablishing the boundary conditions. It is also necessary to consider how the polarity of the system is determined and could be reversed. When thinking about models, one requires information on the size

6.

POSITIONAL INFORMATION AND PATTERN FORMATION

197

of the field, the time required to set it up and regulate it, and the precision with which i t must be specified. I originally suggested (Wolpert, 1969) that what one required was a mechanism that could specify up to 50 cells with an error of +2 in about 10 hours, and this requirement is still probably about right. The estimated size of fields is given in Table 11, and it can be seen that their size is about 50 cells, or 1 mm. The time requirements and precision will be returned to below. TABLE I1

MAXIMUM LINEARSIZE

OF PO6ITIONAL

FIELDS Approximate size

System Axial length of Hydra littoralis ectoderm Early amphibian gastrula-animal pole to dorsal lip Early sea urchin gastrula-animal pole to vegetal pole Early starfish gastrula-animal pole to vegetal pole Larval insect segment-epidermal cells from front edge to back Diameter of amphibian retina a t stage 29 Mesenchyme of chick limb from trunk to apical ridge a t stage 24 Width of amphibian medullary plate Imaginal disc of leg of DTosophila occurs (Bryant and Schneiderman, 1969; Poodry and Schneiderman, 1970) Blastema of regenerating limb of newt Blastema of regenerating limb of larval axolotl Apical meristem of shoots Culminating slime mold

Cell number 60 30 30 50 50-100

30 100 40 100

pM 2000

1000 200

300 500

200 500 200 -

30

1000 500 500

50

1000

100 50

Crick (1971) has drawn a distinction between two basic types of intercellular communication that could be opcrative in establishing positional information. The one hc terms a random walk type, such as diffusion, where each cell affects all its neighbors and where the velocity with which any changc is propagated falls off with distance. I n general a localized change influenws nearby regions very quickly and distant regions very slowly. The other type of communication can be regarded as genuine signaling. In such cases there is a directional quality such that the signal affects the next cell, but not all its neighbors, and transmission of the impulsc along a ncrve provides a useful analogy. Such signaling may not involve a reduction in velocity with distance, and a well worked out model of this type is the phase-shift model of Goodwin and Cohen (1969).

198

LEWIS WOLPERT

A. DIFFUSION

It has been argued by Crick (1970) that very serious attention should be given to models based on diffusion because of the simplicity of the mechanisms involved. The simplest type of mechanism providing a gradient is one comprising a source a t one end and a sink a t the other (Stumpf, 1967; Wolpert, 1968; Crick, 1970). If the concentration of the substance be kept constant a t source and sink ends a linear gradient will be established. Crick (1970) and Munro and Crick (1971) have investigated how long such a gradient would take to be set up from a variety of initial conditions. For the simple case where the concentrations of source and sink are fixed, the time required is about four times longer if the concentration of the substance is initially zero everywhere than if there is a uniform concentration equivalent to half the maximum value or, surprisingly, even if the initial condition is a reversed gradient. The times required were not grossly altered when more complex models, involving a more realistic behavior of the source in relation to fluxes, were used. They expressed the times in a dimensionless form [ D t / L 2 ] where D is the diffusion constant, L the total length, and t the time. Typical values ranged from 0.1 to 0.5. Crick (1970) has argued that a plausible value for an effective diffusion constant for substance of mocm/sec. Them for distances lecular weight about 500 is about 0.3 X of about 1 mm, the times involved are a few hours, which is plausible in terms of the experimental data (see below).

B. INSECT INTEGUMENT AND THE HOMEOSTATIC MODEL A simple source/sink model with diffusion has not been found adequate in the two situations in which it has been looked at, namely the insect integument and regulation in hydra. Lawrence (1971) has been investigating the possibility that the ripple pattern in Rhodnius could be determined by a diffusing substance, the ripples forming a t right angles to the slope of the gradient in the substance. They have compared models using simple diffusion with the results obtained from 90° rotation of pieces of the segment (Fig. 6 ) . While models based on pure diffusion alone gave a broadly similar pattern to that observed in the cuticle, a quantitative investigation of the angle of inclination of the ripples in the center of the pattern after 90° rotation showed them to be quite different from the diffusion model. However, the socalled homeostatic model (Crick, 1970) gave very similar results to those observed. I n this model the cells can be either sources or sinks, depending on their new environment: they become sources and make the substance

6.

POSITIONAL INFORMATION AND PATTERN FORMATION

199

Frc. 6. The relation between polarity and the gradient are illustrated in (a) and (b), in which the gradient represents the concentration of a diffusing substance. The polarity is determined by the sign of the slope of the gradient as indicated by the horizontal arrows. If a piece between X and Y is rotated through 180°, then the gradient will be as in (b) the dotted lines indicating how diffusion will tend to restore the gradient to its original form. In (b) the polarity of the rotated piece is reversed, as shown by the reversal of the slope. The actual effect obtained by Lawrence (1971) by rotating a piece of Rhodnius epidermis through 90” is shown in (c) together with a computer simulation based on the homeostatic model. The ripple lines are thought to form at right angles to the maximum slope of the gradient. (Kindly supplied by Dr. P. Lawrence.)

if its concentration is reduced, and become sinks and destroy it if its concentration is increased. The cells thus try to maintain the original Concentration. This model represents an important departure from simple diffusion, there being now not only both synthesis and destruction of the diffusing substance but what is effectively a second gradient: the “remembered” or preset value of the concentration which the cell tries to maintain. A particularly interesting and important feature of this model may be that the preset level can be altered m l y at cell division. While this model has been used mainly to explain changes in polarity either the absolute value of the concentration or the “preset” value could give positional information.

200

LEWIS WOLPERT

Lawrence (1970, 1971) has reinterpreted some of Marcus’ experiments (1962, 1963) on transplantation of pieces of insect segment in terms of the concentration providing positional information. A key feature of his analysis is that the introduction of a low point into the gradient resulted in a local dip in the gradient which altered both the polarity and the developmental fate of the cells. It is also important that the anterior and posterior edges of the segment seem special and suggest their role as sources and sinks. All these results, taken together with Stumpf’s studies (1967, 1968) provide an impressive case for a linear gradient in a substance whose form is partly determined by diffusion, being the basis of both polarity and positional information in the insect epidermis.

C. HYDRA AND THE FOLLOW-UP SERVO MODEL From our own studies on polarity and positional information in hydra, we have come to a rather similar conclusion (Wolpert, Hicklin, and Hornbruch, 1971). Whereas in the insect integument studies local markers are available for both polarity and position, in hydra only the ends can be reliably assayed. In addition, most attention in hydra has been given to the re-establishment of boundary values, a feature barely touched on in the insect studies. Studies from grafting have shown that the presence of a head effectively inhibits head formation lower down the animal. For example, representing the regions in hydra by H1234 B56F (Fig. 7), a graft of H12 on 12 . . F never results in heads a t the junction, whereas a graft of 12 on 12 . . . F usually results in heads and/or feet forming there. Such grafts provide a biological assay for the various regions and we have extensively explored the dynamics of the syst,em and have concluded that there are two gradients, one of which, the positional value, changes much more slowly than the ot.her. Our results suggest that the main way in which positional value changes is by synthesis and breakdown of substances. Changes in posi-

.

H

1

2

3

4

8

5

6

F

FIQ.7. Diagram of hydra to indicate the various regions. H is the head; 1234, the gastric region; B, the budding region; 56, the peduncle; and F, the foot.

6.

POSITIONAL INFORJIATION AND PATTERN FORMATION

201

tional value are localized and rapid when a boundary is being formed, whereas changes in positional value are much slower away from the boundary. I n one formulation (Wolpert et al., 1971) we suggested the second gradient is of an inhibitor which controls bouiidary formation. Since then we have bcen exploring a more explicit model which is, in some ways, more similar to the homeostatic model of Crick (1970). In the new model, the second gradient, while effectively acting as a boundary inhibitor, now provides the signal for positional information. The relationship between this diffusible signaling substance and positional value involves a follow-up scrvomechanisrn rather than a homeostatic rule. Consider two gradienbs, one the positional signal S which is produced by a source from the head end and is diffusible, and the other the positional value P, which is not diffusible. (To make the situation more concrete, one may very cautiously think of S as a diffusible inducer on an enzyme P.) P changes in two ways depending on whether or not it is a t a boundary. If it is not a t a boundary, P is either synthesized or broken down in order to make the difference between S and P minimal: P follows S according to the relationship in Fig. 8. The establishment of a boundary occurs if S falls to a critical value below P, in which case P is synthcsizcd rapidly to the boundary value and a t this valuc hecomes a source of S. S is produced only a t the boundary. Fig. 8 shows the change in the two gradients when the head is removed. This model can probably account for a variety of grafting experiments relating to head formation (Fig. 9 ) , but the model can be tested reliably only by computer simulation (Clarke and Wolpert, 1971). I n its present form it refers only to head formation and will require additional features to account for foot formation, but we have no reason to believe that different principles will be involved.

D. CHANGES WITH TIME Grafting experimcnts with hydra provide some data on the time required for changes in the positional value and positional signal. We have shown, for example, that region 3 of a 34 . . , F will become region 1 within about 10 hours but in a H/34 . . . F combination it takes about 5 times as long. Also it seems that hcad determination occurs some time before full inhibitory properties are restored. While local changes in positional valuc can occur within about 1 hour the times of greatest interest are those in which a change is propagated over distances of the order of the size of the field. I n general, changes of this type require about 10 hours. However we have recently been investigating a case in which information is signaled over half the length of hydra in about

202

LEWIS WOLPERT

I

FIQ.8. The follow-up servo model for hydra. The suggested relationship between S and P is shown in (a). The steady resting state of S and P are as in (b). Following removal of the head, which is the source of S, its concentration falls as in (c), and when it reaches a critical level below P, P is synthesized until its original boundary value is reached (d). At this boundary value of P, a source of S is again established and results in the restoration of the S gradient (el. P now follows up to the S value according to the relationship defined in (a).

6.

POSITIONAL INFORMATION AND PATTERN FORMATION

203

FIQ. 9. Some examples of axial grafting in hydra. A graft of H12/12 . . . F (a) does not result in heads forming a t the junction, but a graft of H1234/12 . . . F does, since S now falls the critical amount below P (b). A graft 12/56F results in a head forming from the 2 at the junction (c) since that is where the critical fall in S in relation to P has occurred. A head also forms from the 1. I n (d) a head is grafted onto a 34 . . . F. The 3 region increases its positional value rather slowly, according to the relation in Fig. 7a.

1 hour. This occurs in relation to the inhibition of foot determination by the head. The 4 region of a H1234 takes about 3 hours to become determined as a foot as measured by grafting. I n the absence of the head, that is, the 4 in a 1234, the time for determination is 1 hour. This means that the removal of the H is made known to the 4 region within a n hour. This would seem to be near the limit of any mechanism based on diffusion (Clarke and Wolpert, 1971). Such experiments are important since, if, for example, an effect could be obtained over such distances within say 10 minutes, diffusion could be excluded and a signaling mechanism would be required. The times required in hydra for the changes in positional value are of the order of several hours, and it is encouraging that the times required for reversal of polarity of the distal tip of the chick wing (Saunders and Gasseling, 1963) and the regeneration of the apical cap in the amphibian limbs (Thornton, 196af are both about 12 hours.

204

LEWIS WOLPERT

E. ACTIVETRANSPORT AND SPECIAL PERMEABILITY The models considered thus far for establishing positional value are based on diffusion, synthesis, and destruction. Further features can be added, such as active transport and specific permeability relationships. Active transport as a mechanism for maintaining a concentration gradient is extremely attractive and has been proposed by Lawrence (1966) for the insect epidermis and Wilby and Webster (1970b) for hydra. This type of mechanism is particularly attractive for determining polarity. Cohen (1971b) has put forward a somewhat more detailed model which effectively requires specialized boundary regions acting as sources and sinks. It will be of the greatest value if the necessity of an active transport model could be demonstrated, my personal prejudice against it being based on the total absence of evidence for intercellular active transport in physiological systems. Another type of modification could be due to the nature of the passive permeability between the cells. One might imagine, for example, that intercellular flow of the substance could occur only if the concentration difference between the cells were greater than a critical value. There would be a threshold phenomenon, and Stein (1971) has pointed out that this is not incompatible with facilitated transfer mechanisms. Such a mechanism would result in the substance having a minimum slope in the gradient, and also reducing the flow at equilibrium.

F. PDRIODIC SIGNALING AND

THE

PHASE-SHIFT MODEL

A very different type of model based on cell-to-cell signaling is provided by the phase-shift mechanism of Goodwin and Cohen (1969). I n this highly original model, which has been worked out in some detail, the gradient in positional value is established by the propagation of two waves with different velocities from a boundary pacemaker cell, the phase difference between the two waves providing a gradient in positional information. This mechanism is sometimes affectionately referred to as the clap of lightning model, because of the analogy with the delay between seeing the lightning and hearing the thunder providing a measure of the distance of the observer from the lightning. I n the phase-shift model the establishment of the boundary values requires the presence of a gradient in autonomous frequency, so that during regulation the cell with the highest autonomous frequency becomes the pacemaker boundary cell, and entrains the others. I n some ways the pacemaker cell is like the source cell in source/sink diffusion models. It must be emphasized that the phase-shift model is but one of a class of models based on signaling using periodic events. The basic mechanism

6.

POSITIONAL INFORMATION AND PATTERN FORMATION

205

of signaling embodied in the phase-shift model is a rich one and has many implications and variations (Cohen, 1971a,b). Other models based on a single event, using one signal, may be formulated, and one can conceive of models based on amplitude decay. It is also necessary to consider hybrid models using signaling, diffusion, active transport, synthesis, and destruction (Cohen, 1971a). One of the most interesting aspects of the phase-shift model is the insight it gives, and the image it provides, of two-dimensional fields (Goodwin and Cohen, 1969). Cooke and Goodwin (1971) have carried out a rather detailed analysis of the early amphibian embryo. This analysis shows the necessity of deciding where the boundary regions are, and they assign the pacemaker region to the animal pole with a secondary center a t the dorsal lip: the resultant two-dimensional positional field is analyzed. A similar type of analysis has been carried out by Goodwin (1971) for the amphibian retinal-tectal projection. An analysis of Gaze’s experiments (1970) on compound eyes correctly predicted that a doublc ventral eye would give a particular distortion to the pattern. A crucial feature of this analysis was an interaction between the positional gradients along the nasotemporal and dorsoventral axes. I n Goodwin’s modcl this interaction was provided by having the temporal edge as the pacemaker and the dorsal edge as the secondary center. OF MODELS G. VALIDATION

Up to the present it has not been possible to eliminate any but the simplest and crudest models by experimental test. This is not surprising since the experimental assays now available are still rather crude, and models can usually be modified to account for uncomfortable results. Simulating results by model systems will probably not resolve the problem: the only way to do that will be by direct demonstration, preferably a t the molecular level, of the main features of a particular model. This could be, for example, a gradient in a diffusible substance that correlates with positional value or periodicity in activity of an enzyme. The models, are invaluable, however, in telling us what sort of biochemical experiments might be reasonable. For example, the phase-shift model strongly warns against a naive approach of looking for biochemical gradients.

H. BOUNDARY REGIONS A striking feature of the current models is that they all require rather special properties to be present a t boundary regions-such as sources, sinks, or pacemakers. It would be of great interest if a model could be made that did not require the establishment of special boundary

206

LEWIS WOLPERT

conditions, our own attempts along the lines of Turing (1952) have been unsuccessful, and Cohen (1971b) has shown why this is so. The biological data, however, do support the idea that boundary regions are rather special, and this is illustrated in Table 111, which shows that, in systems from a variety of phyla, boundaries have rather similar TABLE I11 BOUNDARY REGIONS System

Properties

Reference

Hydra-hypostome

Induction of secondary axis and Webster and Wolpert reversal of polarity; (1966); Wilby and inhibition Webster (1970b) Amphibian-dorsal lip Induction of secondary axis Saxen and Toivonen (1962) Sea urchin-micromeres Induction of secondary axis and Horstadius (1939) reversal of polarity Chick limb-posterior Induction of secondary axis and Saunders and Gasnecrotic zone reversal of polarity seling (1968) Amphibian-epidermal cap Aggregation of mesenchymal Thornton (1960) of regenerating limb cells and regenerative outgrowth Planaria-head Reversal of polarity Ansevin (1969) Insects-intersegmental Reversals of polarity Lawrence (1970) membrane Some land plants-single Totipotent: gives rise to all the Clowes (1961) apical cell in meristem shoot tissues Quiescent region of root Clowes (1961) meristems Slime mold-tip Affects polarity; removal delays Robertson (1971) morphogenesis

and special properties. This is in a sense a restatement of an old idea of Child (1941) which assigned great importance to the so-called dominant region. The similarity of the effects and importance of boundary regions in such diverse systems is again a strong case for universality. The mechanism whereby the boundary region is established is a crucial feature in systems that regulate. For example, most of the studies on hydra referred to in this paper deal with the formation of new boundary regions when the existing boundary regions are removed. Since any part of the system can become a boundary region, part of the problem is to understand how the boundary region becomes localized. Some sort of inhibitory type of mechanism seems to be operating in hydra (Webster, 1971) such that boundary regions inhibit the formati,oa of similar regions, and this could be viewed as an a priori requirement.

6.

POSITIONAL INFORMATION AND PATTERN FORMATION

207

The presence in eggs of gradients which determine the primary axes and the polarity of the embryo is an old idea, but the relationship between these gradients and those that might be involved in specifying positional information is far from clear. I n the case of early sea urchin development, my analysis (Wolpert, 1969) was based on the idea that the gradients in the egg did not specify positional information directly, but rather the axes and boundary values with respect to which positional information was specified. VII. Polarity

Polarity is a rather neglected aspect of development but is central for pattern formation. It is not an easily defined term in the context of biological forms and patterns, but it is generally used in the sense of implying both an axis and differences along it. It can be regarded as an arrow with a head and tail, or in more mathematical terms it defines vectorial properties. The definitions from the Oxford English Dictionary are illuminating: (i) the possession of two points called poles having contrary qualities or tendencies ; (ii) tendency to develop in two opposite directions in space, time, serial arrangement, etc. I n relation to development I think that the term polarity has been used in two rather different senses. In the first, related to the dictionary definitions, it refers to those situations in which an initially little-structured system becomes increasingly different a t what, is effectively two ends of an axis. For example, the egg of Fucus (Jaffe, 1968) appears to have no polarity, but one can be induced by an asymmetric environment: this results in the basic polarity of the egg becoming established such that one end becomes the root and the other the plant, the two ends becoming increasingly different. Again, in early sea urchin development a n inside/outside polarity is established early and the differences become progressively greater (Wolpert and Mercer, 1963). I n the second sense, the magnetic analogy is stronger and a vectorial field is implied; that is, a value and a direction can be assigned a t each point in the system, and i t is in this sense that it is relevant to positional information. It is a field property, and the polarity at each point is defined. This type of polarity can usually be made manifest only by biological tests. For example, the polarity of hydra is defined with respect to which end will form the head and which end will form the foot. It thus refers to the spatial order of the structures which develop. I n some systems the polarity has a more local and direct visual representation such as the integument of some insects where the bristles point in a particular direction, or in the lamphibian ectoderm where the beat of the cilia reveals the polarity. Evidence is now available to suggest that field polarity may partly

208

LEWIS WOLPERT

be represented by the sign of the slope of a diffusible substance. This most important suggestion comes from Lawrence (1966) and Stumpf (1967) from their studies on the insect integument, which suggested that a reversal of the slope of the gradient resulted in a reversal of polarity. I originally (Wolpert, 1969) defined polarity as the direction in which positional information was specified, making an analogy with a coordinate system and suggested that these may be separate by interdependent systems for determining polarity (polarity potential) and positional information as there is in the phase-shift model. However, in the above models, such as the follow-up servo mechanism for hydra, this is not necessary, and the polarity arises from the dynamic interaction between the two gradients specifying the boundary conditions. Even in the homeostatic model for the insect segment the polarity involves two gradients though the cells may respond to only one of them when polarity bcomes manifest by ripple patterns or hair direction. Thus Lawrence’s model (1966) for polarity determination in terms of a gradient must be viewed in a more dynamic manner. Nevertheless in gross terms it seems to be able to explain reversal of polarity in a variety of systems, particularly if the boundary cells are the source of the gradient (see Table 111) (for further examples and discussion, see Wolpert, 1969; Lawrence, 1970; Wilby and Webster, 1970b ; Webster, 1971). Crick (1971) has drawn attention to the confusion that can arise because the word gradient in relation to pattern formation usually means a decreasing spatial concentration of a substance, whereas mathematically it is the slope of the curve defining the concentration a t a particular point. As he points out, one can always define a vector in a scalar field: for example, the slope at any point in such a variation in concentration. Thus if we have a decreasing concentration of a substance along a line of cells, we at once have a polarity. The converse is not t r u e - o n e can have polarity without a scalar field. For example, if a line of cells can pump a substance in a particular direction there is a well defined polarity but there is not necessarily any variation along the line nor is there any difference between the ends. This in fact illustrates two important different ways in which polarity may be established (Wolpert, 1968). In the first polarity is a global property defined by variation in a scalar. I n the second, polarity is a local property of the cells, is vectorial, and there is no scalar variation. It is of great importance to know which of these mechanisms is involved in the establishment of polarity, and this is reflected on the current controversy relating to polarity in hydra. Wilby and Webster (1970b) suggest that polarity is a local property of the cells possibly involving active transport, whereas

6.

POSITIONAL INFORMATION AND PATTERN FORMATION

209

we have argued that it may be explained in terms of dynamics of the gradients (Wolpert ef al., 1971). It is worth pointing out that all mechanisms based on a bioelectric mechanism for establishing polarity (Rose, 1970) should be treated with great caution. I know of no good experimental evidence for significant potential differences between parts of a field and our own studies on the slime mold and the early amphibian embryo (Garrod et al., 1970; Slack and Wolpert, 1971) show that there are no differences greater than about 1 mV. Jaffe (1968) has argued, however, that there is a field of lO-VV/cm in the early Fucus embryo. There are a number of studies on the early development of polarity in such systems as the amphibian nervous system and eye (reviewed in Jacobson, 1970), limbs (Harrison, 1936), and ear (Yntema, 1955) which show that the polarity of the anterior-posterior axis becomes determined before that of the dorsoventral axis. The polarity of the axes of feathers also seem t o be determined a t different times (Saunders and Gasseling, 1957). Again in the sea urchin, the animal vegetal polarity is very difficult t o alter compared to the dorsoventral one, and the latter appears to be determined later. These time differences are very interesting and raise the question whether positional information is specified independently along different axes in a two-dimensional field. The analysis of Cooke and Goodwin (1971) and Goodwin (1971) suggests that this is not the case. I n this connection it is necessary to emphasize that specification of the polarity of an axis is not the same as specifying positional information since Jacobson (1970) has somewhat confused the issue by discussing his important experiments on polarity determination in the early retilna in terms of specification of central connections. In view of the growth of the retina from this stage by marginal accretion (Gaze, 1970; Straznicky and Gaze, 1971), this distinction is particularly important and raises some very subtle problems for pattern development. An interesting, but little known system exhibiting an obvious polarity is the ectoderm of larval amphibia whose cilia usually beat in an anterior-posterior direction. Studies by Tung e t al. (1948) suggest that this polarity reflects a gradient in a chemical substance, and, remarkably, they claim that polarity reversal can be induced by reversal of the underlying tissue, even through an agar barrier. This is particularly interesting since Piepho (1970) has claimed a similar induction of polarity reversal in the insect epidermis. VIII. Intercellular Communication It is commonly believed ( e g , Goodwin and Cohen, 1969) that the cell-to-cell interactions necessary for pattern formation occur via the low-resistance junctions that have been found between embryonic cells

210

LEWIS WOLPERT

in a variety of systems (Furshpan and Potter, 1968; Palmer and Slack, 1970; Bennett and Trinkaus, 1970). This idea is very attractive since one might expect that the molecules for communication would be retained within the cells and yet pass between them, and it is exactly the sort of channel the low resistance junctions provide. Nevertheless, the demonstration that these junctions are in any way involved in communication in development is entirely lacking. There are, for example, no experiments showing that uncoupling of the cells in any way interferes with development, although Wilby and Webster (1970a) have some very indirect evidence that this is the case. I n more general terms one may ask whether cell-to-cell contact is necessary for communication in a positional field and the answer is by no means unequivocal (cf. Wolpert and Gingell, 1969). My impression is that cell-to-cell contact is necessary and in hydra we have some evidence for this (Wolpert et al., 1971). However, in a most important experiment Saunders and Gasseling (1963) have shown, using Millipore filters that cell-to-cell contact is not necessary for repolarization of the distal region of the chick limb. Rose (1970) too has presented evidence that substances controlling regeneration in Tubularia can move under the force of an electric field, from one tissue to another, even when these are not in contact. Such experiments, as well as those of Tung e t al. (1948) referred to above, are very puzzling since the extracellular movement of important substances controlling either pattern or polarity is surprising. Such experiments urgently need confirmation. Returning to low-resistance junctions on the assumption that they are developmentally significant one can ask to what extent they might provide a mechanism for restricting channels of communication. Might, for example, the boundaries of fields be defined by an absence of low resistance junctions at the boundary, and could absence of low resistance coupling between different cell types be developmentally significant? At present, the answer to these questions seems to be in the negative. The available data suggests that in early development there is no significant spatial restriction of functional coupling, although uncoupling between regions such as the neural tube and ectoderm in amphibians has been found by Warner (1970). I n the sea urchin Tupper e t al. (1970) have found that while there is no coupling between cells prior to 16-cell stage, coupling does develop after this. There are as yet no reports of different cell types from early embryos being unable to form low-resistance junctions with each other and the studies of Hyde e t al. (1969) show that heart cells and fibroblasts in culture will form such low resistance connections. That there is at least some restriction to coupling is shown by Loewenstein (1967), who found that coupling did not occur between the cells of two species of sponge.

6.

POSITIONAL INFORMATION AND PATTERN FORMATION

21 1

One may also consider what sort and size of molecules could pass between coupled cells. The electrophysiological data show only that there is a reduced resistance to small ions, and this type of data has been gradually supplemented by investigation of the possible passage of labeled molecules (Furshpan and Potter, 1968). I n several coupled systems, dyes, such as fluorescein, appear to pass relatively easily. There are very few data for developing systems, and in the early amphibian embryo the junctions are not permeable to fluorescein, (Slack and Palmer, 1969). There is at present no evidence for molecules greater than a molecular weight of 1000 passing between cells through lowresistance junctions. It is of great interest that, using the quite different system of metabolic cooperation between cells in culture, Pitts (1971) (and personal communication) has concluded that the molecules going from one cell to another across cells contacts have a maximum molecular weight of about 500. IX. Interpretation The interpretation of positional information is the key process iln the formation of the pattern and is the raison d’dtre for positional fields. It is by this process that the pattern is made manifest. It is not possible to overemphasize its importance since the concept of universality of positional fields places the main burden for pattern formation on the interpretative process. Unfortunately, very little if anything is known about it, and this is due largely to our ignorance of the molecular bases of positional value, but interpretation will possibly involve specific gene activation along the lines currently being considered for cytodiff erentiation (e.g., Britten and Davidson, 1969). Certainly advances in the molecular basis of differentiation will greatly help our understanding of interpretation. It is worth remembering that in any positional field the interpretative decisions a cell may take may be relatively simple in the sense that the decision is usually among only a few possibilities. For example, in the developing chick limb the mesenchyme’s decision is mainly between cartilage and muscle. I doubt that any interpretative decision often involves a choice involving more than three polssibilities, but this requires careful investigation. For example, in Marcus’ system (1962, 1963) it seems that interpretation may involve seven different types of cuticle. The number of interpretative decisions in any cell lineage is probably not greater than ten. Current thinking on the mechanism of interpretation is not in terms of gene activation but rather in terms of concentrations and thresholds: how a cell could, in principle, give different responses to the variations in concentration of one or more substances which specify its positional value. This is probably required by both diffusion and phase-shift

212

LEWIS WOLPERT

models. Crick (1971) has briefly discussed this problem and has suggested how enzymes might be able to recognize the absolute concentration of molecules other than its substrate. I n this connection it may be an advantage for the positional field to use two gradients, as in Fig. 1, since it may be easier for the cell to detect the change in ratio between two substances than the absolute concentration of just one, particularly near the boundaries. It is also becoming clear that there is a close relationship between determination, as classically defined, and interpretation of positional values. I n some cases, determination implies that the interpretation will be different: for example, one can think of the determination of wing and leg in the chick in terms of difference in interpretation to a similar positional field (Wolpert, 1969) . One might also view transdetermination (Hadorn, 1966) in insects in similar terms ; that is, transdetermination involves not a change in the field but in the cells’ interpretation of it. A similar view may be taken of homeotic mutants which result in very similar conversions (Fristrom, 1970). Such studies might provide a clue to the processes of interpretation, and it is significant that transdetermination can occur within a group of adjacent cells and that the rate of transdetermination increases with an increased rate of cell division. The ability, apparently, of a single gene in homeotic mutants, to completely alter the interpretation is something that must be taken into account when the genetic circuitry of interpretation begins to be modeled. One can see the problem more clearly with leg and wing. What does a difference in determination mean, in terms of intracellular circuitry, for the turning on of, say, the genes for muscle cells. Since muscle cells differentiate in different positions in leg and wing, are the conditions for the activation of the same set of genes altered, or are there two sets of muscle genes with different thresholds, one for wing and another for leg. More generally, how much genetic information is needed for specifying the difference between leg and wing: this is particularly interesting since the luxury molecules are very similar. An answer might come from molecular hybridization, which could tell us whether arm is to leg is as kidney is to liver in terms of common genome sequence expressed.

X. Precision It has proved very difficult to obtain data on the precision with which positional information is either specified. or interpreted. While Maynard-Smith (1960) has provided an original discussion of continuous and quantized variation in adult structures, his data are not directly

6.

POSITIONAL INFORMATION AND PATTERN FORMATION

213

relevant to the precision of developmental processes involving pattern formation in field. Nevertheless, it is very interesting that the coefficient of variation of a variety of continuous varying characters in a variety of animals-from insect wing length to the human skull b r e a d t h w a s from about 2% to about 6%. While some of this is due to genetic variation even in genetically homogeneous populations under standardized conditions, no coefficient of variation of less than 1% is reported. One may ask whether any developmental process is ever so accurate as to provide this degree of precision. The situation may be made more specific by considering a line of cells 50 cells long. How accurately could any one of these be specified? or if a boundary were to be made half way, how precisely could this be done? The only developmental situation that I know of where this type of data is available is the specification of the number of primary mesenchyme in sea urchin development. There are about 100 cells in the late blastula of which about 50 become primary mesenchyme. Taking data from Peters (1909) and Horstadius (1936b), the range of mesenchyme is, for example, 3&55 with a mean of 40 and with a coefficient of variation of about 10%. This is rather imprecise and confirms our overwhelming impression that in sea urchin morphogenesis processes such as primary mesenchyme formation and movement, gastrulation, and skeleton formation are not very precise (Gustafson and Wolpert, 1963). In hydra, too, tentacle number is variable (Lashley, 1915). Data from other systems arc hard to obtain. Mosaic eggs give the impression of a very precise cleavage pattern and program of differentiation during early stages, although E. B. Wilson (1904) mentions that the number of primary trochoblasts in the mollusc Patella a t 30 hours varies from 19 to 21. It is of interest to consider how precisely neuronal connections are established, since one might think that it is here that precision might be most manifcst. Gaze (1970) has considered the precision with which the retinotectal precision is restored after various insults and concluded that his technique of recording from the optic tectum, while the most reliable available, gave only about 10 points along a 3-mm line which would contain about 350 optic nerve fiber terminations. Thus, a t the cellular level, the technique provides a very poor sample. Using an anatomical approach, Horridge and Meinertzhazen ( 1970) have provided the most detailed study of precision using the insect eye. They have examined a system with a repeated pattern of connections lying in parallel, which is manifest in the first synaptic layer of the fly’s optic lobe (Braitenberg, 1967). In order to avoid anatomical details of the system, I will give the results in a simplified and formalized form, al-

214

LEWIS WOLPERT

though the system will repay learning as it will be increasingly studied. I n the compound eye of the fly there are about 3200 ommatidia, each of which contains a group of eight photoreceptor cells which are arranged in a well defined trapezoid-like pattern. The axons of 6 of these photoreceptor cells interweave in a complex pattern and terminate in the first synaptic layer in discrete cablelike structures called optic cartridges. There is one optical cartridge below every ommatidium or group of photoreceptor cells, the remarkable feature of the system being that none of the short axons from six of the receptors form synapses with optic cartridge beneath. If we restrict ourselves to a line along the dorsoventral axis and number the ommatidia and oartridges along it from the ventral to the dorsal edge 1 to iV, and number the photoreceptor cells 1 to 8, then the rules for connectivity in the dorsal half for ommatidium n are: receptor cell 1 joins cartridge n 1 , 2 joins cartridge n - 1 , 3 joins cartridge n - 3, and 4 joins cartridge n - 2; 5 and 6 will not be considered because they join lateral cartridges, and 7 and 8 are somewhat different. This pattern in the dorsal half of the eye has a mirror image in the ventral half. Here, the trapezoidal arrangement of the photoreceptors is in precise mirror image and the rules for connectivity can be obtained by reversing the sign between n and the integer. For example 1 now joins n - 1. Horridge and Mcinertehagn (1970) undertook the extremely tedious task of tracing 650 short axons from the receptor cells to their cartridge. The remarkable finding was that none of the axons terminated in the wrong cartridge. It is not worthwhile to speculate too much in the absence of knowledge of how the eye and its connections develop, particularly if it develops by a series of repeated events starting a t the equator. What, we may ask is whether each ommatidium is uniquely specified. If it were, it would be by far the most accurate system known. Whatever the mechanism for the insect eye, the results strongly suggest that 8 adjacent cells can be uniquely and reliably specified with respect to their connectedness. Are any bigger or better examples known? It should be added that the receptor axons take up specific positions in the cartridge, and here occasional errors are found ; Horridge and Meinertzhagen suggested that the terminals settle in the segment of their cartridge where they first arrive. One possible approach to precision might be via genetic analysis. For example, i t would be very interesting to know how fine a genetic control is exerted in the insect eye; could a mutant be found that could affect just one of the 3200 ommatidia, or even a row? There must be a great deal of information available on the fineness of genetic control of patterns which could provide data on precision. On the available data, my impression is that the half-way boundary

+

6.

POSITIONAL INFORMATION AND PATTERN FORMATION

215

of our 50 cell line would be specified between about 23 and 26. However, what does seem to be very well specified is the serial order of structures in all patterns, and I know of no case where an error is made. Forebrain is always a t the front, and toes a t the end. XI. Cell Movement Most current views on the mechanism of directed cell movement or movement of cells to form a pattern is in terms of differential adhesiveness (Steinberg, 1970; Gustafson and Wolpert, 1967; Trinkaus, 1969). For example, in our studies of the mechanism whereby the primary mesenchyme cells of the sea urchin embryo moved from their point of entry into the blastocoel to form a relatively well defined ring, we concluded that it could be accounted for by random pseudopodal activity and a variation in adhesive contact in the substratum over which the cells were moving. The cells would move to those regions where their pseudopods made the most stable contact. This explanation assigns a variation in cellular properties to the substratum which reflects the distribution of the mesenchyme cells, this pattern being effectively a template. I n these terms one might expect, and might even look for, a region of increased adhesiveness a t the surface of the cells where the mesenchyme form the ring (Gustafson and Wolpert, 1963, 1967). Recently, however, I have begun to question this view and to reconsider it in terms of positional information and Goodwin and Cohen’s suggestions (1969) for how this could guide cell movement. It is now clear that a different explanation can be offered. Instead of there being a ring of increased adhesiveness in the ectoderm there would be no pattern as such, but only the assignation of positional information. Thus moving mesenchyme cells would “test” the positional information of the cells over which they were moving and would stop at that value for which they were programmed. The difference between the two mechanisms is illustrated in Fig. 10. In (a) the cells collect on the line because of a localized adhesiveness in the substratum whereas in (b) there is no discontinuity and the cells stop a t a particular positional value because this is a property of the moving cells. In this second case the pattern is generated by cellular response within a coordinate system. Thus in the case of the sea urchin embryo, the mesenchyme cells might be responding not to a template but just to the coordinate system provided by positional information, and if this were true it would be absurd to look for sticky patches. There is unfortunately no good evidence a t the moment to resolve this issue but the apparent ability of melanoblasts, which migrate very actively, to sense the positional informrtion in the

216

LEWIS WOLPERT

FIG. 10. If cells moving over a cell sheet by pseudopods come to stop in a particular region as in (a), then one possibility is that there is an increased adhesiveness between the pseudopods and the cell sheet as indicated in (b). A different view is that there is no variation in adhesiveness but that the moving cells make use of the positional information in the sheet as in ( c ) and tend to stay where the positional value is a*.

substratum over which they have moved (see Section V) is very suggestive. Clearly the situation remains confused. But at least it is necessary to begin to take into account that possibility that moving cells make use of positional information. Goodwin and Cohen (1969) put forward quite a detailed mechanism to account for neuronal cell migration in establishing retinotectal connections and suggested that periodic signals from the tectal cells could exert their influence on retinal fiber cells by sensitizing the membranes to produce or inhibit pseudopods. In general it does not seem too difficult to construct appropriate models, particularly if a signaling channel in the form of low-resistance junctions was established quite quickly between a variety of moving cells. This channel could be used to signal or inhibit pseudopodal activity (cf. Wolpert and Gingell, 1969). Such views force one to consider the importance and relevance of differential adhesiveness for morphogenesis in situations other than sorting out of aggregates (Steinberg, 1970) which is highly artificial. It is also significant that Gaze’s studies (1970) and analysis of the development of retinotectal specifications lead him repeatedly to concepts of patterns of neuronal connections and relative position rather than unique molecular specificity and matching of neurones of the type suggested by Sperry (1963).

6.

POSITIONAL INFORMATION AND PATTERN FORMATION

217

XII. Mosaic Development One of the main features of positional information as originally formulated was the idea that intercellular signaling was involved, and it is an essential requirement of regulative systems. It is thus essential to examine the so-called mosaic or nonregulative systems within the same conceptual framework, since i t seems that here one may well be dealing with a different type of system. The nature of cytoplasmic localization in early development has recently been reviewed by Davidson (1969) in relation to the gene activity in early development. While few systems are completely either mosaic or regulative some groups of animals such as annelids and molluscs, seem to show mosaicism to a high degree. Others, such as the insects, contain both highly mosaic and highly regulative eggs, and it is this latter fact, mosaicism and regulation displayed to varying extents within a single group, that encourages me to continue to look for common features. It is also worth remembering that regulative eggs such as the sea urchin have cytoplasmic localization which may be used to specify boundaries and polarity (Wolpert, 1969). From the point of view of pattern formation, Davidson’s analysis, like the classic work of Wilson (1925), is inadequate since both are more concerned with cytoplasmic localization than pattern formation. For pattern formation the apparent absence of intercellular communication is striking. For example, Horstadius (1937) showed that after the %cell stage the blastomeres of the primitive worm Cerebratulus can be isolated or recombined without any way altering their development. Each fragment develops just as it would have in the intact embryo. The problem of mosaic eggs might be stated as follows: to what extent does the specification of the spatial pattern of differentiation depend on the cytoplasmic differences within the egg, not on intercellular communication? and what is the mechanism whereby cytoplasmic differences are used? An idealized situation is illustrated in Fig. 11 in which the system becomes divided up into N cells which are different from each other because of the cytoplasm (or membrane) i t contains. One might say that positional information has been assigned to the cells by cytoplasmic localization. Such a mechanism raises many problems. For example, how accurately must the cytoplasmic localization and the planes of cleavage be specified. This will depend on whether the cytoplasmic differences are graded quantitatively or are qualitatively discrete. The latter would probably make specification much easier since a cell would respond according to the presence or absence of a discrete set of substances. If localization were graded, local intercellular communication a t the time of, or just following, cleavage could amplify differences between the

218

LEWIS WOLPERT

**xxxr* xx)rYxYx

x x

Y

Y

x

x

x x x

X X X

x

n**

x

Y

x

x

o

o

x

x

*

o

o

X

O

-

+ *

-

o - - t +

t

t X X Y

x

Y

x

*

*

O

xnlrx X x X K Y x x x ~ x x xY x (a)

I

Q

t

t u u

x x

o o

o - o - -

*

+ + +

(b)

FIQ.11. The development of a mosaic-type system, based on cytoplasmic localization. Two possible ways in which this might be done are (a) by graded cytoplasmic differences, and (b) by discrete cytoplasmic differences. Both types raise quite difficult problems with respect to the accuracy of the cleavage planes and the necessary precision of the cytoplasmic differences.

two cells. Might one think of a universal cytoplasmic localization system in eggs: one should certainly be hesitant to think of the localization of specific organ-forming substances. The presence of mosaic and regulative features in the same system is illustrated in the squid (Arnold, 1968), where early damage to a localized region near the egg surface leads to the absence of a n eye; however, a t later stages the presumptive eye region can be made to give rise to two lenses. What, one wonders are the relative merits of signaled positional information and cytoplasmic mosaicism for pattern formation, and why should some eggs be highly mosaic and others highly regulative. Is it partly a matter of distance over which cell-to-cell communication occurs? It is also curious that some mosaic eggs develop into animals with remarkable regenerative powers. The mosaic problem is not just confined to eggs and early development: it is made manifest during the further growth and development of all systems involving positional information (Wolpert, 1970). For example, in chick limb development gross regulation does not seem to occur after about stage 25 and one thinks of the system as a mosaic. How does further pattern become manifest as the system grows? Does the system break down into smaller fields or are the mechanisms operative dependent on cell division and polarity? Do cell lineage and clones of the type studied by Mints (1971) now predominate? The growth of the amphibian retina (Gaze, 1970) may well be an example otf such problems.

XIII. Growth and Cell Division It is an attractive possibility that one type of interpretation of positional information would be entry into the cell cycle since this could provide a means of growth control based upon absolute rather than relative size (Wolpert, 1969). The pattern of cell division in the crypt of Lieberkuhn (Lamerton and Steel, 1968) is certainly suggestive of

6.

POSITIONAL INFORMATION AND PATTERN FORMATION

219

this since dividing cells are restricted to the base of an axis running from the base of the crypt to the villus. However, much more persuasive evidence has come from the relationship between segmental gradients and growth in insects. Lawrence (1970) has pointed out that it is as though the tissues had an inherent steepness of gradient and growth continues until that steepness is reached. For example, Bohn (1970) has found that extirpating middle pieces out of the legs of the African Roach, resulted in intercalary regeneration replacing the lost p a r k and this, combined with various grafting experiments, suggests that the steepness of the gradient is a determining factor. Lawrence (1970) also found that the growth of segments in insects seemed based on similar principles. Such examples lead one to consider the growth of tissues in other developing systems such as the blastema of regenerating limbs or the growth of the chick limb bud (Hornbruch and Wolpert, 1970), in terms of positional information. XIV. Spacing Patterns There seems to be a class of patterns which may be conveniently grouped together under the name of spacing patterns, a term suggested to me by Dick Gordon. Typical examples of spacing patterns are the more or less evenly spaced primary hair follicles in the skin of sheep (Claxton, 1964), the spaced bristles and hairs on the surface of insects (Lawrence, 1970) and stomata on leaves (Bunning, 1965). The development of such patterns is thought to involve a considerable random element and an inhibitory mechanism preventing similar elements developing too close to each other. I n a detailed analysis of the development of hairs and bristles in Oncopeltus, Lawrence (1970) has provided an alternative to Wigglesworth’s competition model (1940) in terms of a local inhibitory field around the developing organelle. I n this new model it is not the developing organelle that is responsible for the inhibition, but the adjacent epidermal cells. As he points out, this brings the process more into line with other field phenomena. This important conclusion means both that the inhibitory process in the development of spacing patterns may involve a mechanism identical to that for specifying positional information, and also that all inhibitory fields may be the same. As with positional information the inhibitory field may vary in extent but the main effect will depend on the properties of the responding cells. It must not be thought, for example, that because hair formation is inhibited, a specific hair inhibitor substance is involved. Nickerson (1944) has suggested a mechanism for the barring pattern in feathers based on the black bands inhibiting pigment formation in the adjacent region, and it is thus rather similar to a spacing pattern.

220

LEWIS WOLPERT

We are currently considering the possibility that some mammalian coat patterns such as those seen in the cheetah, giraffe, and zebra may be based on a similar developmental mechanism. It is probably necessary to consider that both the growth and inhibitory rules may vary in the animal from one region to another, possibly making use of positional information. We are also investigating the extent t o which two rather reliable patterns, the development of the segmented pattern shown by somites and the feather follicle pattern in birds, are to be regarded in terms of positional fields or in terms af spacing patterns. XV. Pattern Formation in Plants

There seem to be two main types of pattern formation processes in plants. The one takes place in embryonic or apical meristematic regions in which the pattern of organs is specified within the primordia when these are less than about 1 mm; e.g., leaf primordia usually form about 500 pm from the tip of apical shoot meristem. This type, which might involve positional fields, should be contrasted with the control of large-scale branching patterns in which the interactions may typically occur over large distances, even up to a meter. This second type of pattern may be thought of as involving “competition” between different potential growth points as is manifest in the phenomenon of apical dominance. The meristem may, from one point of view, be considered as the region which defines the boundaries and position of secondary primordia, such as leaves and “flower members,” without itself undergoing overt morphogenesis. It thus, for example, is responsible for phyllotaxis, which has many features in common with spacing patterns (see Section XIV) (see Wardlaw, 1968). A feature of many meristems is their capacity to regulate such that when cut in half, each half will regulate and form a whole meristem. It is thus clear that the meristem corresponds in some ways to a positionla1 field, and the presence in most meristems of a relatively well defined region such as the apical cells or quiescent zone suggests that these are the boundary regions and strengthens the analogy. It is not clear whether the behavior of the cells within a meristem is determined by positional information. XVI. Conclusions In this essay I have tried to show that within the conceptual framework of positional information a new, and simple, way of looking a t pattern formation may be obtained. In pressing the possibility of universality I am deliberately taking an extreme stand, but a t least it serves to counterbalance the special-substance-inductiveview of pattern

6.

POSITIONAL INFORMATION AND PATTERN FORMATION

221

formation. Also in order to show its possible relevance to pattern formation, and even cell movement, I have often taken a somewhat Procrustean view of the data. One of the virtues of the positional information mechanism of pattern formation is that with the same system for positional information one can generate an enormous number of different patterns by changing the cell’s rules for interpretation. Since interpretation will be gene determined there is little difficulty in seeing how this can be achieved. I n fact, the concept of positional information makes excellent use of a central feature of development, that all cells carry the same genetic information. A further corollary is that it is as easy to make an apparently complex asymmetrical pattern as it is to make a simple one. Pattee (1971) has pointed out the necessary complexity for making the Fench Flag using positional information considering that the French Flag is so “simple” a pattern. This is, however, to miss the point, since while the mechanism is not very simple, the simplicity lies in the fact that it can be used equally easily for apparently complex patterns and the same mechanism may be used for an enormous variety of patterns. ACKNOWLEDGMENTS This work is supported by the Nuffield Foundation. I am grateful to Professor D. Cohen for discussions on plant development, and for the comments of Dr. D. Garrod. REFERENCES Ansevin, K. D. (1969). J . Ezp. 2001.171, 235. Apter, M. J., and Wolpert, L. (1965). J . Theor. Biol. 8, 244. Arnheim, N. (1967). Genetics 56, 253. Arnold, J. M. (1968). Develop. Biol. 18, 180. Baker, P. C. (1965). J . Cell Biol. 24, 95. Bennett, M. V. L., and Trinkaus, J. P. (1970). J . Cell Biol. 44, 592. Bodenstein, D. (1935). Wilhelm Roux’ Arch. Entwicklungsmech. Organismen 133, 156.

Bohn, H. (1970). Wilhelm Rozix’ Arch. Entwicklungsmech. OrganGmen 165, 303. Braitenberg, V. (1967). Exp. Brain Res. 3, 27. Britten, R. J., and Davidson, E. H. (1969). Science 165, 349. Bryant, P. (1971). Personal communication. Bryant, P., and Schneiderman, H. A. (1969). Devel. Biol. 20,263. Bunning, E. (1965). In “Handbuch der Pflanzenphysiologie” (W. Ruhland, ed.), Vol. 15, pp. 146-188. Springer-Verlag, Berlin and New York. Child, C. N. (1941). “Patterns and Problems of Development.” University of Chicago Press, Chicago, Illinois. Clarke, M., and Wolpert, L. (1971.) In preparation. Claxton, J. H. (1964). J. Theor. Biol. 7, 302. Cloney, R. A. (1966). J . Ultrastruct. Res. 14, 300.

222

LEWIS WOLPERT

Clowes, F. A. L. (1961). “Apical Meristems.” Blackwell, Oxford. Cohen, M. H. (1971a). In “Some Mathematical Problems in Biology” (M. Gerstenhaber, ed.). Amer. Math. SOC.(in press). Cohen, M. H. (1971b). Symp. SOC.Exp. Biol. 25. “Control Mechanism of Growth and Differentiation,” pp. 455-476. Cambridge Univ. Press, London and New York. Cooke, J., and Goodwin, B. C. (1971). In “Some Mathematical Problems in Biology” (M. Gerstenhaber, ed.), Vol. 3. Amer. Math. SOC.(in press). Crick, F. H. C. (1970). Nature (London) 225, 420. Crick, F. H. C. (1971). Symp. SOC.Exp. Biol. 25. Control Mechanism of Growth and Differentiation,” pp. 429438. Cambridgc LTniv. Press, London and New York. Davidson, E. H. (1969). “Gene Activity in Early Development.” Academic Press, New York. de Both, N. J. (1970). Wilhelm Roux’ Arch. Entwicklungsmech. Organismen 165, 242. Driesch, H. (1908). (‘The Science and Philosophy of the Organism.” Black, London. Dulbecco, R. (1970). Proc. Nut. Acad. Sci. U.S. 67, 1214. Fristrom, J. W. (1970). Annu. Rev. Genet. 4, 325. Furshpan, E. J., and Potter, D. D. (1968). Curr. Top. Develop. Biol. 3, 95. Garrod, D. R., Palmer, J. F., and Wolpert, L. (1970). J. Embryol. Exp. Morphol. 23, 311. Gaze, R. M. (1970). “Formation of Nerve Connections.” Academic Press, New York. Gingell, D. (1970). J. Embryol. Exp. Morphol. 23, 583. Gmitro, J. I., and Scriven, L. S. (1966). Symp. Int. SOC.Cell Riol. 5, 221-255. Goodwin, B. C. (1971). In “Some Mathematical Problems in Biology” (M. Gerstenhaber, ed.), Vol. 3. Amer. Math. SOC.(in press). Goodwin, B. C., and Cohen, M. H. (1969). J. Theor. Biol. 25,49. Goss, R. J. (1961). Advan. Morphog. 1, 103. Gustafson, T., and Wolpert, L. (1963). Int. Rev. Cylol. 15, 139. Gustafson, T., and Wolpert, L. (1967). Biol. Rev. 42, 442. Hadorn, E. (1966). I n “Major Problems in Developmental Biology” (M. Locke, ed.), pp. 85-104. Academic Press, New York. Harrison, R. G. (1936). Proc. Nat. Acad. Sci. U.S. 2, 238. Holtfreter, J., and Hamburger, V. (1955). In “Analysis of Development” (R. H. Willier, P. A. Weiss, and V. Hamburger, eds.), pp. 230-296. Saunders, Philadelphia, Pennsylvania. Holtzer, H. (1971). In “Control Mechanism in the Expression of Cellular Phenotypes” (H. A. Radykula, ed.), pp. 69-88. Academic Press, New York. Hornbruch, A., and Wolpert, L. (1970). Nature (London) 226, 764. Horridge, G. A., and Meinertzhagen, I. A. (1970). Proc. Roy. SOC.,Ser. B 175, 69. Horstadius, S. (1936a). Mem. Mus. Hist. Natur. Belg. [21 3, 801. Horstadius, S., (193613). Wilhelrn ROW’ Arch. Entwicklungsmech. Organismen 135, 40.

Horstadius, S. (1937). Biol. Bull. 73, 317. Horstadius, S. (1939). Biol. Rev. 14, 132. Hyde, A., Blondel, B., Matter, A., Cheneval, J. P., Felloux, B., and Giradier, L. (1969). Progr. Brain Res. 31, 283. Jacobson, M. (1970). “Developmental Neurobiology.” Holt, New York. Jaffe, L. F., (1968). Advan. Morphog. 7, 295. Kieny, M. (1964). Develop. Biol. 9, 197. Kroeger, H. (1960). Naturwissenschaften 47, 148.

6.

POSITIONAL INFORMATION AND PATTERN FORMATION

223

Lamerton, L. F., and Steel, E. G. G. (1968). Progr. Biophys. Mol. Biol. 11, 247. Lashley, K. S. (1915). Proc. Nut. Acad. Sci. U.S. 1, 298. Lawrence, P. A. (1966). J. Exp. Biol. 44, 607. Lawrence, P. A. (1970). Advan. Insect Physiol. 7, 197. Lawrence, P. A . (1971). Symp. Xoc. Exp. Biol. 25. “Control Mechanisms of Growth and Diffrrentiation,” pp. 379-390. Cambridge Univ. Press, London and New York. Lederberg, J. (1967). Curr. Top. Develop. Biol. 1, ix. Lewis, E. B. (1963). Amer. Zool. 3, 33. Loewenstein, W. R. (1967). Develop. Biol. 15, 503. Marcus, W. (1962). Wilhelm Roux’ Arch. Entwicklungsmech. Organ. 154, 56. Marcus, W. (1963). Zool. Anz., Suppl. 26, 168. Maynard-Smith, J. (1960). Proc. Roy. SOC., Ser. B 152,397. Maynard-Smith, J., and Sondhi, K. C. (1961). J. Embryol. Exp. Morphol. 9,661. Mints, B. (1971). I n “Control Mechanisms in the Expression of Cellular Phenotypes” (H. A. Radykula, ed.), pp. 15-42. Academic Press, New York. Exp. Biol. 25 (in press). Munro, M., and Crick, F. H. C. (1971). Symp. SOC. Nickerson, M. (1944). J . Exp. Zool. 95, 361. Nothiger, R., and Schubiger, G. (1966). J . Embryol. Exp. Morphol. 16, 355. Palmer, G . F., and Slack, C. (1970). J. Embryol. Exp. Morphol. 24, 535. Pattee, H. H. (1969). I n “Communication in Development” (A. Lang, ed.), pp. 1-16. Academic Press, New York. Peters, K. (1909). Arch. Entwicklungsmech. Organ. 27, 164. Piepho, H. (1970). Biol. Zentralbl. 89, 573. Pitts, J. D. (1971). In: Growth Control Cell Cult., Ciba Found. Symp. (G.E. W. Wolstenholme and G. Knight, eds.). Churchill, London. Poodry, C. A,, and Schneiderman, H. A. (1970). Wilhelm Roux’ Arch. Entwicklungsmech. Organismen. p. 166. Postlethwait, J. H., and Schneiderman, H. A. (1969). Proc. Nut. Acad. Sci. U S . 64, 176. Rawles, M. E. (1948). Physiol. Rev. 28, 383. Rawles, M. E. (1955). In “Analysis of Development” (R. H. Willier, P. A. Weiss, and V. Hamburger, eds.), pp. 499-519. Saunders, Philadelphia, Pennsylvania. Roberts, P. (1964). Genetics 49, 593. Robertson, A. (1971). In “Some Mathematical Problems in Biology” (M. Gerstenhaber, ed.), Vol. 3. Amer. Math. SOC.(in press). Rose, S. M. (1962). I n “Regeneration” (D. Rudnick, ed.), pp. 153-176. Ronald Press, New York. Rose, S. M. (1970). Amer. Zool. 10, 91. Saunders, J. W., Jr., and Gasseling, M. T. (1957). J. Exp. Zool. 135, 503. Saunders, J. W., Jr., and Gasseling, M T. (1963). Develop. Biol. 7, 64. Saunders, J. W., Jr., and Gasseling, M. T. (1968). I n “Epithelial-Mesenchymal Interactions” (R. Fleischmajer and R. E. Billingham, eds.), pp. 78-97. Williams & Wilkins, Baltimore, Maryland. Saunders, J. W., Jr., Cairns, J. M., and Gasseling, M. T. (1957). J. Morphol. 101, 57. Saxkn, L., and Toivonen, S. (1962). “Primary Embryonic Induction.” Logos Press, London. Schroeder, T. E. (1970). J. Embryol. Exp. Morphol. 23, 427. Schubiger, G. (1971). Develop. Biol. (in press). Slack, C., and Palmer, J. F. (1969). Exp. Cell Res. 55, 416. Slack, C., and Wolpert, L. (1971). I n preparation. Sperry, R. W. (1963). Proc. Nut. Acad. Sci. U . S. 50, 703.

224

LEWIS WOLPERT

Spiegelman, S. (1945).Quart. Rev. Biol. 20, 121. Spooner, B. S.,and Wessells, N. K. (1970).Proc. Nut. Acad. Sci. U.S. 66,360. Stein, W. D. (1971).Personal communication. Steinberg, M. S. (1970).J. Exp. Zool. 173, 395. Stern, C. (1968). “Genetic Mosaics,” and other essays. Harvmd Univ. Press, Cambridge, Massachusetts. Stern, C., and Tokunaga, C. (1967).Proc. Nut. Acad. Sci. U.S 57,658. Stocum, D.L. (1968).Develop. Biol. 18,457. Straznicky, K.,and Gaze, R. M. (1971).J. Embryol. Exp. Morphol. 26,67. Stumpf, H. (1967).Wilhelm Roux’ Arch. Entwicklungsmech. Organismen 158, 315. Stumpf, H. (1968).J. Exp. Biol. 49, 49. Thornton, C. S.(1960).Develop. Biol. 2, 551. Trinkaus, J. P. (1969). “Cells into Organs.” Prentice-Hall, Englewood Cliffs, New Jersey. Tung, T. C., Tung, Y. F. Y., and Chang, C. H. (1948). Proc. Zool. Soc. London 118, 1134. Tupper, J., Saunders, J. W., Jr., and Edwards, C. (1970).J . Cell Bid. 46, 187. Turing, A. M. (1952).Phil. Trans. Roy SOC.London Ser. B. 237,37. Twitty, V. C. (1949).Growth Symp. 9, 133. Waddington, C. H. (1962).“New Patterns in Genetics and Development.” Columbia Univ. Press, New York. Waddington, C. H. (1971). In “Developmental Systems-Insect” (S. J. Bounce, ed.), Academic Press, New York (in press). Wardlaw, C. (1968).“Morphogenesis in Plants.” Methuen, London. Warner, A. E. (1970).J . Physiol. (London) 210, 150. Webster, G. (19Zl).Biol. Rev. 46, 1. Webster, G., and Wolpert, L. (1966). J. Embryol. Exp. Morphol. 16, 105. Weiss, P. (1939).“Principles of Development.” Holt, New York. Wigglesworth, V. B. (1940).J. Exp. Biol. 17, 180. Wilby, 0. K.,and Webster, B. (1970a). J. Embryol. Exp. Morphol. 24, 583. Wilby, 0.K., and Webster, G. (1970b). J. Embryol. Exp. Morphol. 24, 595-613. Wilson, E.B. (1904).J . Exp. Zool. 1, 197. Wilson, E. G. (1925). “The Cell in Development and Heredity.” Macmillan, New York. Wolpert, L. (1968). In “Towards a Theoretical Biology” (C. H. Waddington, ed.), Vol. 1, pp. 125-133. Edinburgh Univ. Press, Edinburgh. Wolpert, L. (1969).J. Theor. Biol. 25, 1. Wolpert, L. (1970). In “Towards a Theoretical Biology” (C. H. Waddington, ed.), Vol. 3, p. 198. Edinburgh Univ. Press, Edinburgh. Wolpert, L. (1971).Annu. Rev. Sci. Basis Med. 29 (in press). Wolpert, L., and Gingell, D. (1969). In “Ciba Symposium on Homeostatic Regulators” (G. E. W. Wolstenholme and J. Knight, eds.), p. 241. Churchill, London. Wolpert, L., and Mercer, E. H. (1963).Exp. Cell Res. 30, 280. Wolpert, L., Hicklin, J., and Hornbruch, A. (1971). Symp. SOC.Exp. Biol. (in press). Yntema, C. L. (1955). In “Analysis of Development,” (R. H. Willier, P. A. Weisa, and V. Hamburger, eds.), pp. 415-428. Saunders, Philadelphia, Pennsylvania.

AUTHOR INDEX Numbers in italics refer to the pages on which the complete references are listed. A

Abbott, J., 195 Abraniova, N. B., 46, 58, 60, 72,75, 76 Ajtkhozin, M. A., 62, 77 Akhalkazi, R., 53, 76 Allen, B. M., 80,107 Allfrey, V. G., 146, 181 Anderson, 0. R., 11, 43 Andrew, T. M., 85, 107 Ansevin, K. D., 206,221 Apter, M. J., 186, 221 Arias, I. M., 91, 107 Arnheim, N., 195, 221 Arnold, J. M., 218, 221 Aronson, A. I., 55, 75 Ashburner, M., 149, 176, 181 Ashby, W. R., 159,182 Avanesov, A. C., 62, 7'6 Avrameas, S., 99, 108

Bernstiel, M., 46, 76 Billing, J., 89, 108 Blakeman, J. P., 28, 43 Blatt, L. M., 85, 108 Blondel, B., 210, 222 Bodenstein, D., 195, 221 Boezi, J. A., 149,181 Bohn, H., 219, 221 Bonner, J., 181 Botvinnik, N. M., 72, 76 Bourgeois, S., 149, 181 Bouteille, M., 99, 108 Bracker, C. E., 20, 21,43, 44 Braitenberg, V., 213, 221 Briggs, D. E., 111, 143 Britten, R. J., 84, 89, 108, 150, 165, 181, 211, 221

Brown, D. D., 46, 75 Brown, H. T., 111,143 Bryant, P., 191, 221 Bulova, S. I., 99, 108 Bunning, E., 219, I 1 Burakova, T. A., 51, 76 Burka, E. R., 99,108 Burnett, A., 179, 181

B

Bachvarova, R., 48, 51, 75 Bagrova, A. M., 58, 60, '76 Bailey, P., 84, 110 Baker, P. C., 185, 221 Balassa, G., 177, 182 Balinsky, J. B., 87, 107,109 Barbiroli, B., 89, 108 Bartnicki-Garcia, S., 21, 43 Beerman, W., 161, 165, 178, 181 Belitsina, N. V., 62, 77 Benne!t, M. Q. L., 210,821 Bennett, T. P., 80, 94, 99, 107, 108 Berendes, H., 161, 165, 176, 178, 181 Beritashvili, D. R., 53, 76 Berman, R., 81, 108 Bern, H. A., 81, 108

C

Cairns, J. M., 195, 209, 223 Camargo, E. P., 20,21, 43 Campbell, A. M., 99, 108 Campbell, P. N., 85, 94, 108 Campbell, R. N., 21, 41, 44 Cantino, E. C., 2, 3, 5, 6, 7 , 9, 10, 11, 13, 15, 17, 20, 23, 24, 25, 27, 28, 31, 43,

44

Chamberlain, J. P., 49, 76 Chandra, G. R., 111, 116, 119, 144 Chang, C. H., 209, 210, 224 225

226

AUTHOR INDEX

Changeaux, J., 148,186 Chefurka, W., 85, 108 Cheneval, J. P., 210, 628 Cheng, T.-Y., 76 Child, C. N . , 188, 206, 261 Chrispeels, M. J., 117, 119, 121, 122, 123, 124, 125, 126, 136, 143, 144 Clarke, M., 201, 203,221 Claxton, J. H., 219, 221 Cleaver, J. E., 162, 177,181 Clever, U., 176, 181 Cloney, R. A., 185, 621 Clowes, F. A. L., 206,222 Cognett, G., 64, 68, 7 6 Cohen, G. N., 175, 181 Cohen, M. H., 197, 204, 205, 206, 209, 215, 216, 222 Cohen, P. P., 79, 80, 81, 85, 86, 87, 89, 94, 99, 107, 108,109,110 Coleman, J. R., 100, 108 Comb, D. A., 49, 76 Connolly, T. N., 26, 44 Cooke, J., 205, 209, 662 Corrance, M. H., 99,108 Cotter, D. A., 43 Cousineau, G. H., 64, 76 Cowie, D. B., 149, 181 Craig, S. P., 49, 76 Crick, F. H. C., 197, 198, 201, 208, 212, 222, 223 Crippa, M., 53, 54, 76 Cummins, J. E., 162, 177, 181 Czihak, G., 48,49, 76 D

Dallner, G., 85, 91, 109 Das, C. C., 71, 76 Davidson, E. H., 48, 51, 52, 76, 84, 89, 108, 150, 151, 165, 178, 181, 186, 211, 217, 231 Davidson, J. N., 99, 108 Dawid, I. B., 46, 76 Deaven, L., 59, 77 de Both, N. J., 190, 226 Deering, R. A., 24,& Dietrich, C. P., 20, 43 Donzova, G. V., 48,53,76 Doyle, D., 91, 107

Driesch, H., 186, 226 Dulbecco, R., 185, 2i?2 E

Eaton, J. E., 87, 88,108, 110 Edwards, C., 210, 624 Eeckhout, Y., 102, 108 Erdos, P., 171, 186 Escombe, F., 111, f 4 3 Etkin, W., 80, 81,108 Evans, D., 46, 76 Evins, W. H., 130, 133, 134, 135, 143 F

Fawcett, D. W., 3, 44 Felicetti, L., 67, 76 Felloux, B., 210, 222 Filner, P., 120, 143 Fletcher, J., 28, 43 Fraser, A. S., 170, 182 Freed, J. J., 76 Frieden, E., 80, 85, 87, 88, 100, 107, 108, 110

Frietrom, J. W., 212, 222 Fuchs, W. H., 21, 44 Fuller, M. S., 3, 4, 5, 9, 10, 41, Furshpan, E. J., 210, 211, 262

44

0

Gambino, R., 67, 76 Garrett, M. K., 28, 44 Garrod, D. R., 209, 222 Gasarian, K. G., 53, 76 Gasseling, M. T., 195, 203, 206, 209, 210, 223

Gay, H., 71, 76 Gaze, R. M., 184, 191, 205, 209, 213, 216, 218, 222, 664 Gehring, W., 168, 182 Gelbard, A. S., 162, 182 Gelehlter, T. D., 83,84, 87,109 Georgiev, G. P., 58, 61, 62, 76, 84, 108, 109, 182 Gilbert, L. I., 80, 81, 109 Gilbert, T. W., 148, 182 Gingell, D., 185, 210, 216, 66.2,624 Giradier, L., 210, 626

227

AUTHOR INDEX

Giudice, G., 53, 76 Glenn, J. C.,94,99,107, 108 Gligin, M.V.,49,76 Gligin, V. R., 49, 76 Gmitro, J. I., 192,222 Goldstein, A., 29, 31, 33, 36, 44 Goldstein, L.,59, 76 Gona, A. G., 81,108 Good, N. E.,26,44 Goodwin, B. C.,197, 204, 205, 209. 215, 216, 222 Goss, R. J., 189, 191,222 Granboulan, N.,109 Granner, D.M., Jr., 83,84,87,109 Greenberg, J. R.,76 Gross, J., 83, 108 Gross, P. R.,49, 55, 64, 68, 76, 76 Grove, S. N., 20, 44 Griiss, J., 111, 143 Gudernatsch, J. F.,80,108 Gustafson, T.,185, 213, 215, 222 H

Haberlnndt, G., 111, 112,143 Hadorn, E.,167, 182, 212, 222 Halvorson, H. O., 177,182 Hamburger, V.,195,222 Hamilton, T.H., 90,108 Harris, H., 84, 108 Harrison, R. G., 209,222 Hartman, I. F.,49,76 Hendler, R. W.,85,94,108 Henning, W., 71,76 Henshaw, E.C., 84,108 Herzenberg, L.A.,149,182 Hess, O.,71, 76 Hicklin, J., 191, 200, 201,209,210, 224 Hickman, C.J., 42,44 Ho, H. H., 42,44 Horstadius, S., 195, 206, 213, 217, 222 Holtfreter, J., 195, 222 Holtzer, H., 185,222 Holwell, M.E. J., 3,4.4 Horenstein, E. A., 5, 17, 20, 43 Hornbruch, A., 191, 200, 201, 209, 210, 219, 222, 224 Horridge, G. A., 213, 214, 222 Hough, B. R., 181 Huang, R.C.,181

Hunter, A.L., 64,68,76 Hyatt, M.T., 2, 43 Hyde, A.,210, 222 Hynes, R.O., 49,77 I

Ignatieva, G. M., 48, 76 Iinuma, H., 111, 118,144 Il'in, M.J., 64,68,69,76 Infante, A. A., 76 Ingram, V. M., 99,108 Ionesco, H., 177, 182 Ivanchik, T.A., 52,76 Izawa, S.,26, 44 J

Jacob, F., 146, 148,182 Jacobsen, J. V., 116, 117, 121, 122,143 Jacobson, M.,222 Jaffe, L. F., 207,209,122 Johnson, K. D.,136,143 Jones, R.L.,125, 126, 127, 129, 131, 141. 142, 143, 144 Jost, J.-P., 99, 109 Just, J. J., 80, 85, 100,108 K

Kafiani, C. A.,48,53,56,69,76, 76 Karlson, P., 80, 108 Kauffman, S., 146, 147, 150, 153, 156, 166, 182 Kaufmann, B. P., 71,76 Kavanau, J. L., 19,44 Kedes, L. H., 49, 55, 64, 68, 76 Keir, H. M., 99, 108 Kelley, D. E.,76 Kenney, F. T., 83,109 Kende, H., 136,143 Kieny, M.,195,222 Kijima, S., 52, 55, 76, 84, 108 Kim, J . H., 162, 182 Kim, K . H., 85, 86, 87, 89, 108 Kirsop, B. H., 111, 144 Kleevecz, R.,59,77 Knight, J., 83, 110 Koch, W.J., 24,25, 44 Koehler, D.E.,135,141,143, 1 4

228

AUTHOR INDEX

Kollros, J. J., 81, 83,108 Korner, A., 84, 108 Kostomarova, A. A., 48, 50, 51, 54, 58, 76, 77 Krigsgaber, M. R., 52, 62, 64, 68, 69, 70, 76 Kroeger, H., 53, 76, 195, 222 Kukhanova, M. K., 63,76 Kvavilashvili, I. S., 53, 76 1

Lamborg, M. R., 83,109 Lamerton, L. F., 218,223 Lashley, K. S., 213, 223 Lawrence, P. A., 191, 195, 196, 198, 199, 200,204, 206, 208,219, 223 Leak, L. V., 6,43 Lederberg, J., 184, 223 Leduc, E. H., 99, 108 Leseman, D. E., 21, 44 Lessie, P. E., 3, 4, 6, 20, 44 Lewis, E. B., 195, 223 Lezzi, M., 53, 76 Lima-de-Faria, A., 162, 182 Lockwood, D. H., 99,108 Loewenstein, W. R., 210, 225 Lovett, J. S., 2, 3, 4, 6, 9, 15, 17, 20, 22,23,40,41,43,44 Low, R. B., 84,110 Lukanidin, E. M., 84, 109 Lythgoe, J., 6, 43

M McCarthy, B. J., 84, 109 McCurdy, H. D., Jr., 23,44 McGarry, M. P., 99, 108 Mack, J. P., 6, 43 MacLeod, A. M., 111, 115, 144 Malt, R. A., 85, 109 Mantieva, V. L., 58, 76 Manton, I., 21, 44 Marcaud, L., 84, 109 Marcus, W., 200, 211, 223 Markman, B., 51, 76 Marshall, M., 86, 108 Matsumae, A,, 6, 17, 44 Matter, A., 210, 222 Maynard-Smith, J., 212,225

Meinertzhagen, I. A., 213, 214, 22,9 Meir, H., 9, 44 Melcher, U., 140, 143, 144 Melnikova, N. L., 48, 69, 76 Mense, R. M., 140, 141,143,144 Mercer, E. H., 207, 224 Metafora, S., 67, 76 Metzenberg, R. L., 86,108 Meuser, R., 21, 43 Miles, C. A,, 3, 44 Millar, A. S., 111, 144 Milman, L. S., 72, 76 Milnax, J., 84, 109 Mintz, B., 223 Monesi, V., 71, 76 Monod, J., 146, 148, 182 Monroy, A., 49,67,76,76 Morel, C., 109 MorrB, D. J., 20, 44 Moss, B., 99, 108 Muller-Hill, B., 148, 182 Munro, M., 198, 223 Murray, R. K., 89, 109 Mutolo, V., 53, 75 Myens, R. B., 6, 17,20, 44

N

Naito, K., 81, 109 Nakagawa, H., 85, 86, 87, 88, 89, 109 Nechaeva, N. V., 50, 76 Nemer, M., 49, 76 Newman, S., 166, 188 Neyfakh, A. A., 46, 47, 48, 52, 53, 58, 60, 62, 63, 64, 66, 68, 69, 70, 72, 76, Y6 Nickerson, M., 219, 223 Nicoll, C. S., 81, 108 Niessing, J., 61, 76 Nieuwdorp, P. J., 129,144 Nisman, B., 85,108 Nothiger, R., 191, 223 0

Ohad, I., 85, 91, 109 Ohtsu, K., 81,109 Okada, S., 161, 182 Olsnes, S., 84, 109 Olson, L. W.,,4, 10, 44

229

AUTHOR INDEX

Omura, T., 85, 91, 109 Ovchinnikov, L. P., 62, 76 Oyer, D., 84, 110

P Paik, W. K., 85, 86, 91, 94, 108, 109 Painter, R. B., 161, 165, 182 Paleg, L., 111, 144 Palmer, G. H., 111, 115, 144 Palmer, J. F., 209, 210, 211, 222 Pattee, H. H., 223 Penman, M., 89, 109 Penman, S.,YY, 89,109 Perez, A. G., 162,182 Perry, R. P., YG Peters, K., 213, 223 Piepho, H., 209, 223 Pik6, L., 46, YG Pitot, H. C.,99,109, 146,182 P i t h, J . D., 211, 223 Pollock, A., Jr., 111,144 Poodry, C. A., 223 Potter, D. D., 210,211,222 Prescott, D. M., 59, 76 Price, J . M., 144 Priestly, G. C., 85, 109 Pruyn, M. L., 85, 109 Ptashne, M., 148,182

R Rachkus, J. A., 48, 56, 76, YG Radley, M., 111, 144 Radziel-skaya, V. V., 70, YG Ragland, W., 99, 109 Rao, M. V. N., 59, Y6 Raper, K. B.,43 Rawles, M. E., 195, 196, 223 Reichle, R. E., 3, 4, 5, 9, 41, 44 Rendel, J. M., 170, 182 Renyi, A., 171, 182 Rice, S.,166, 182 Riggs, A. D., 149, 175, 182 Riggs, T. R., 100,109 Rinaldi, A. M., 49, YG Roberts, P., 195, 223 Robertson, A., 206, 223 Robinson, P. M., 28,44

Roels, 0. A., 11, 43 Rosbash, M., 89,109 Rose, S. M., 190, 191, 209, 210, 223 Rossetti, F., 83, 110 Rott, N. N., 47, 48, 49, 50, 51, 54, 76, YG, Y7 Ruddat, M., 128, 129, 130, 131, 137, 138, 139, 144 Rusch, H. P., 162, 177, 181 Rytes, A., 177, 182 5

Salimaki, K., 80, 109 Samarina, 0. P., 61, 62, 76, 84, 109 Samuels, H. H., 83, 84, 87, 109 San Pietro, A., 83, 109 Saunders, J. W., Jr., 100, 109, 195, 203, 206,209,210,223,224 SaxCn, E., 80, 109 SaxBn, L., 80, 109, 206,223 Scandalios, J. G., 122, 143 Schaeffer, P., 177, 182 Schander, H., 111, 144 Scherrer, K., 84, 109 Schimke, R. T., 91, 107 Schlessinger, D., 84, 109 Schmoyer, I. R., 2, 23, 40, 41, 44 Schneiderman, H. A,, 80, 81, 109, 195, 223 Schroeder, T. E., 185, 223 Schubiger, G., 191, 223 Scriven, L. S., 192, 222 Sekeris, C. E., 61, Y6 Shambaugh, G. E.,111, 87, IOY, 109 Shaw, D. S., 3, 6, 7, 9, 10, 23, 24, 28, 43, 44 Shearer, R. W., 84, 109 Sheldon, H., 94,99,108 Slieveleva, G. A., 51, Y7 Shires, T. K., 146, 182 Shiokawa, K., 54, 77 46, 77 Shmerling, Zh. G., Siekevitz, P., 85, 91, 109 Singh, M. M., 26, 44 Singh, U. N., 55, Y7 Slack, C., 209, 210,211,223 Sladek, N., 99, 109 Slater, D. W., 49, Y7 99, 109 Soling, H. D.,

230

AUTHOR INDEX

Soll, D. R., 2, 3, 4, 7, 9, 10, 15, 17, 20, 22, 23, 27, 28, 29, 34, 35, 37, 39, 40,

44 Solovjeva, J. A., 49, 77 Sondhi, K. C., 223 Sonneborn, D., 10, 20, 21, 23, 27, 29, 34, 35, 43, 44 Sosinskay, J. J., 49, 77 Sperry, R. W., 216, 223 Spohr, G., 109 Spiegelman, S., 49, 77, 186, 224 Spirin, A. S., 61, 62, 76, 77, 84, 109, 166, 182 Spoonor, B. S., 185, 224 Srinivansan, V. R., 177,182 Steel, E. G. G., 218,223 Stein, W. D., 204, 224 Steinberg, M. S., 215, 216,224 Stern, C., 192, 194, 195, 196, 224 Stinvalt, W. S., 84, 110 Stockdale, F. E., 99, 108 Stocum, D. L., 190, 224 Straznicky, K., 209, 224 Strohman, R. C., 81,108 Strominger, J. L., 20, 43 Stufflefield, E., 59, 77 Stumpf, H., 195, 198, 200, 208, 224 Suberkropp, K. F., 2, 23, 24, 25, 27, 31, 43, 44 Sussman, A. S., 2, 28, 44 Svetajlo, N. A., 70, 76 T

Tartof, K. D., 76 Tata, J. R., 80, 85, 86, 88, 89, 90, 91, 92, 93, 94, 98, 99, 100, 105, 106, 107, 109, 110 Temmink, J. H. M., 21, 41, 44 Terehova, T. A., 76 Thompson, E. B., 83, 84, 87, 110 Thorton, C. S., 203, 206, $34 Timofeeva, M. J., 48, 49, 56, 69, 70, 76, 77 Toivonen, S., 80, 109, 206, 223 Tokunaga, C., 196,224 Tolstorukov, I. I., 48, 53,76 Tomkins, G. M., 83, 84,87,109 Tompkins, G., 146, 182

Tonoue, T., 87, 100,110 Topper, Y. J., 99,108 Trinkaus, J. P., 185,210, 215,221, 224 Truesdell, L. C., 2, 3, 5, 6, 7, 9, 10, 11, 13, 15, 17, 20, 23, 24, 25, 27, 28, 31, 43, 44 Tung, T . C., 209, 210, $24 Tung, Y. F. Y., 209, 210,224 Tupper, J., 210, 224 Turian, G., 4, 44 Twitty, V. C., 195, 224 Tyler, A., 46, 76 U

Unsworth, B. R., 87, 110 V

Vanable, J. W., Jr., 99, 108 Van der Eb, A. A., 129,144 Van Etten, J. L., 23, 44 Varner, J. E., 111, 116, 117, 118, 119, 120, 121, 122, 123, 124, 125, 126, 134, 135, 136, 140, 141, 143, 144 Vigil, E., 128, 129, 130, 131, 137, 138, 139, 144 Vinograd, J., 46, 7G W

Waddington, C. H., 150, 182, 184, 192, 224 Walker, C., 159, I S 2 Wallers, C., 149, 181 Wardlaw, C., 220, 224 Warner, A. E., 210, 224 Weber, R., 79,80,89,100,102,109,110 Webster, B., 208, 210, 224 Webster, G., 206, 208, 224 Webster, J., 9, 44 Weinberg, R. A., 77 Weiss, P., 83, 110, 186, 224 Wigglesworth, V. B., 219, 224 Wilby, 0. K., 2M, 210,224 Williams, C. M., 81, 110 Wilson, E. B., 206, 213,224 Wilson, E. G., 188,224

231

AUTHOR INDEX

Wilt, F. H., 49, 52, 55, Y6, YO, 77, 81, 84, 109, 110 Winget, G. D., 26, 44 Winter, W., 26, 44 Wolpert, L., 184, 185, 186, 191, 194, 196, 197, 198, 200, 201, 203, 206, 207, 208, 209, 210, 212, 213, 215, 216, 217, 218, 219, 221, 222, 223, 224 Wolstenholme, G. E. W., 83, 110 Wool, I. G., 84, 110

Wyatt, G. R., 80, 89,110 Wyman, J. P., 148,152 Y

Yamana, K., 54, YY Yapo, A., 85, 105 Yntema, C. L., 209, 224 Yomo, H., 111, 118, 144 Young, D. A,, 88, 100,110 Yurowitzki, Yu. G., 72, Y6

flow matrix, 166-168 gene activity patterns, 161 control advantages of forcing structures, 172-173 evidence for theory of, 174-179 developmental genes, 177-178 extended forcing structures, 176-177 forcible operons, 175 macroscopic tests, 178-179 number of inputs per gene, 17b176 expected character of forcing structures, 170-171 forcing structures in switching nets,

A

Amphibian metamorphosis DNA synthesis in, 99 hormone role in, 80-81 protein synthesis during, 79-110 proteins involved in, 81-83 regulation of, 83-99 in tadpole hepatocyte, 85-99 tissue resorption in, requirement of RNA and protein synthesis in, 100-105 B

151-155

global behaviors of, 146-148 homeostasis and, 148-149 molecular mechanisms, 173-174 multiple-input type, 151 one-input type, 150-151 theoretical model of, 145-182 model systems, 149-150 Genetic information realization in early development, 45-77 regulation, 52-54 RNA synthesis in nuclei, 47-48 rate, 50-52 in regulation of enzymatic activity,

Blastocladiella emersonii, zoospore germination in, 1 4 4 D

Diffusion, as model system for pattern formation, 198 DNA, synthesis in amphibian metamorphosis, 99 E

Early development, protein synthesis in, 62-84

72-74

RNA transport to cytoplasm, 5 4 6 1 transcription of, 46-54 changes in template quantity, 46-47 translation from RNA, 61-72 informosomes in, 61-62 Gibberellins, in control of secretory tksue, 111-144

F French flag problem, 186 G

Gene control systems alternative theories of, 179-180 biological implications of, 160-170 capacity to evolve, 168-170 cell cycle time, 161-163 cell types, 163-166

H

Hormones control of secretory tissue by, 111-144 232

233

SUBJECT INDEX

role in amphibian metamorphosis, 80-81 Hydra, as model system for pattern formation, 200-201

R

RNA onset of synthesis in nuclei, 4 7 4 8 transport to cytoplasm, 54-61

I

Informosomes, in translation of genetic information, 61-62 Insect integument, as model system for pattern formation, 198-200

Secretory tissue, hormonal control of, 111-144 Sulfonic acid azo dyes, effects on encystment of zoospore, 35

P

T

Plants, pattern formation in, 220 Positional information and pattern formation, 183-224 cell movement and, 215-216 growth and cell division, 218-219 intercellular communication, 209-211 interpretation of, 211-212 model systems and mechanisms, 196-207 active transport and special permeability, 204 boundary regions, 205-207 changes with time, 201-203 diffusion, 198 hydra, 200-201 insect integument, 108-200 periodic signaling, 204-205 validation of models, 205 morphallaxis and epimorphosis, 188-192 mosaic development, 217-218 pattern and form, 184-186 pattern regulation, 186-188 in plants, 220 polarity, 207-209 precision of, 212-215 spacing patterns, 219-220 universality and prepatterns, 192-196 Protein synthesis in amphibian metamorphosis, 79-110 in eady development, 62-64 dependence on RNA, 64-67 nonnuclear control a t translational level, 67-72 regulation of, 83-85

Thyroid hormone effect on amphibian metamorphosis, 82 protein synthesis, 86-87 Tissue resorption, requirement of RNA and protein synthesis in, 100-105

Z Zoospore of Blastocladiella emersonii cytoplasmic inclusions of, 6 encystment of, 2-3 fine-structural changes during, 10-17 formahion of myelinlike figures, 11 induction of, 34-35, 38-39 kinetics of, 35-43 low-temperature effects on, 25-27 resume of changes during, 6 4 spore behavior during, 6-8 structural changes during, 11-14 environmental influences on, 23-25 low temperatures, 23-27 self-inhibition in spore populations, 27-34 flagellar retraction and rotation of nuclear apparatus, 6-7 mechanics of, 8-10 structural changes after, 14-17 structural changes during, 14 gamma particles of, 5-6 role in cell wall formation, 19-21 germination of, 1-44 backing membrane changes, 18 macromolecular synthesis during, 21-23

234

SUBJECT INDEX

structural changes during, 17-21 volume changes during, 7-8 kinetosome, flagellum, and banded rootlet of, 4-5 lipid bodies, SB matrix, and backing membrane of, 5

nongerminating type structure of, 3-6 mitochondrion of, 5 nucleus and nuclear cap of, 4 breakdown of, 18-19 vacuole formation in, 7