Molecular Genetics and Metabolism 71, 1–9 (2000) doi:10.1006/mgme.2000.3048, available online at http://www.idealibrary.com on MINIREVIEW Genetic and Epigenetic Regulation Schemes: Need for an Alternative Paradigm Magali Roux-Rouquie Institut Pasteur, 25–28, rue du Docteur Roux, 75724 Paris Cedex 15, France Received July 31, 2000 ments (as it was formally introduced by Leibnitz in 1666). This thought is still prevalent and forms a basis for the ambiguities currently in force about the concept of systems and results in the approach of systems through the mathematical set theory. Whereas a set is defined by the basic concept of membership of its individual elements, a system is delineated by the new properties gained from the arrangement and the interactions of components (1). They are: (i) the property of emergence, which refers to the new and additional characters undergone by the system compared to the qualities of the components considered separately or arranged separately in another system (e.g., emerging functions); (ii) the property of selectivity, which means that any organizational relation exerts restrictions and constraints on the components so that only a limited part of the properties of these components are being recruited (e.g., specific gene expression in differentiated cells). In this respect, the immune system, the nervous system, and embryogenesis, which are highly integrated systems, could be designed as supersystems (2). Embryogenesis illustrates the central role of interactions in the self-organization of living systems. The initial steps of early development in Drosophila are well known, in particular those leading to the segmental patterning along the anterio-posterior (A–P) axis. This development program is under the control of BX–C complex of which genes are regulated in cis and trans in addition to the rule of colinearity. This rule states that the order of the BX–C loci in the chromosome Self-organization of living cells results from the tangle of positive and negative feedback developed to ensure their homeostasis and/or their differentiation. There are three major means cellular regulation operates: the genetic, the epigenetic, and the metabolic ones. The regulation type in each of them has been overviewed. Further examination of relations between complexity and developmental stability points out sui generis properties of feedback loops, which are redundancy and pleiotropy. Prototypical schemes for positive and negative regulation with redundant and pleiotropic (including multifunctional) proteins are presented. They stress a theoretical shift from the analytical to the systemic framework. The systemic paradigm appears to be of increasing interest and importance in the study of concepts for the representation of genetic and epigenetic regulations. © 2000 Academic Press Key Words: regulation; feedback; self-organization; complexity; systemics; genetic; epigenetic; paradigm. SELF-ORGANIZATION IN LIVING SYSTEMS Living cells are usually seen as self-organized complex systems that exchange matter, energy, and information with their environment. As systems, they are nonreducible global units made up of various molecular structures in relation, which fit together into macromolecular structures themselves in relation. Regarded as a whole, cells are homogeneous entities; according to their components, they are heterogeneous. In this respect, we must distinguish a system from a whole of ele1 1096-7192/00 $35.00 Copyright © 2000 by Academic Press All rights of reproduction in any form reserved. 2 MAGALI ROUX-ROUQUIE parallels the order in which the units are expressed according to the A–P axis in the embryo. During the segmentation, three groups of genes (gap, pair-rule, and segment-polarity) are involved: the gap genes activate the pair-rule genes, which control the segment-polarity genes. Last, the segmental specificity conferred by homeotic genes is under the control of the gap gene. This linear cascade of gene expression corresponds to the first definition of the BX–C syntagm (3). There are three types of syntagms: (i) Time syntagm. With respect to segmentation, it consists of a selector gene (S) (one of the BX–C genes) activated by one activator (A) (gap and pairrule activating the selector of the relevant region) to target one realisator (targets of BX–C genes) (4). (ii) Space syntagm. It makes it possible to translate quantitative information (e.g., gradient of gene expression) to qualitative information (boundaries, cell population) (4). (iii) Form syntagm. This was found notably with gain-of-function mutation of rolledSevenmaker (rlSem) gene, which activates more efficiently development pathways than the rolled (rl) gene (5). In addition to this classification with respect to space, time, and form, syntagms can also be distinguished according to whether they can be broken down into elementary syntagms. This nonreducibility of syntagms consists of an emerging property in the sense of the systemic paradigm (see below). Consistently, elementary syntagms can be part of more complex ones. Self-organization of supramolecular structures has been reproduced in vitro with a limited number of models. It was reported using microtubules and molecular motors known to participate in the formation of many cellular structures. Interestingly, it was found that in a constrained environment, the same final structure was obtained through different assembly pathways (6). Otherwise, Ca 21 -dependent aggregation of lipid membranes by functional annexins was obtained through self-organization of the proteins into twodimensional plaques (7). In any case, self-organization results from a dynamic interplay between composition and regulatory processes. It provides a pool of possible components in which the regulatory system can select the most appropriate to express a given cellular function. TABLE 1 Major Features in Negative and Positive Feedback Feature Negative feedback Positive feedback Organization Behavior Form Action Constant Homeostasis Morphostasis Repetitive Changeable Differentiation Morphogenesis Dispersive REGULATION IS CONSUBSTANTIAL WITH ORGANIZATION Regulation is the process by which an entity is maintained constant with respect to a given law whatever the occurring disturbances. From a general point of view, the concept of regulation is related to those of structural and functional stability: the regulating system interprets the signals emitted by the system to be controlled in order to ensure the conservation of both the structures and the functions. These two subsystems can be clearly identified: when they are gathered in only one system, it is called self-regulation. One can distinguish two types of regulation on the basis of their effects on the organization of the controlled system (Table 1): (i) regulation of a synchronic type that results from a compensation brought by a component of the system whose properties neutralize the disturbance without impact on the organization (homeostasis); (ii) regulation of a diachronic type forcing the system toward other states when some parameters reach a limiting value and thus somewhat modifying the organization of the system (differentiation). These regulations utilize either negative feedback processes (repressive action of the elements of the feedback loop on the evolution of the system), or positive ones (activator role of the regulation on the evolution). These two types of regulation have basically different effects since the negative feedback sustains homeostasis phenomena whereas the positive feedback leads to differentiation processes. The processes of negative regulation are extremely frequent and make it possible to maintain a state close to its threshold value. The positive regulations are autocatalytic reactions in which the product of the reaction is necessary to its own production, whether the stimulus persists or not. E. Morin, the French epistemologist, coined the tangle of relations between positive and negative feedback loops in the mid-seventies (Fig. 1). To quote Morin: “the self- GENETIC AND EPIGENETIC REGULATION SCHEMES 3 CELLULAR MEANS OF REGULATION FIG. 1. Morin’s feedback loop (8). expansion of living systems is achieved from selfregulated systems and this self-regulation is achieved on the basis of disordered proliferation” (8). Since then, numerous studies have demonstrated positive and negative regulations in a variety of models (9 –17). In a recent article, Demongeot et al. described a minimal regulatory module consisting of a negative and a positive circuits and termed logical regulon. This regulon was found to display numerous dynamic properties allowing memory storage and mnesic reminder (18). In cellular differentiation, the introduction of multiple positive circuits was suggested to account for the variety of the steady states. Interestingly, as few as eight genes exercising direct or indirect positive control onto their own transcription can produce up to 256 (2 8) steady states (19). Conversely, negative circuits would be involved in maintaining oscillation processes (19). Oscillatory regimes have been associated with a wide range of biological processes. Oscillations in the concentration of cytosolic Ca 21 (20) have been shown more effective than a single prolonged increase in Ca 21activated transcription factors NF-AT and NF-kB. Specific gene transcription was obtained according to oscillation frequency (21,22). This cross-talk between positive and negative regulation led to the concept of average connectivity (interactions leaving a regulator gene/interactions arriving a regulated gene), which supplies information on the number of interactions per gene. In E. coli, it was found that the connectivity of regulated genes was approximately three. Positive and negative circuits were distinguished through matrix representation of the feedback interactions (all genes controlled by the same gene are located in the same column, whereas all regulators of a given gene are located on the same line). Thus, 55 of 96 transcriptional regulators were involved in interregulation, leading to 87% negative, 6.5% positive, and 6.5% dual circuits (23). Accordingly, the instantiation of regulations lies in a tangle of interaction networks. This is particularly well-documented in the development of Drosophila, which shows a spatio-temporal pattern of gene expression controlled through a combination of independent and interaction-dependent regulation pathways. There are three means of regulation that ensure correct cell functioning: on one hand, genetic and epigenetic regulations and, on the other, metabolic regulations. The metabolic regulations, notably circuits involving post-translational regulation or various metabolic events (e.g., proteolysis, phosphorylation), are out of the scope of this article, which is only devoted to the regulations involving genetic material; however, the lack of a precise distinction between all three regulation mechanisms legitimates the examination of the characteristics of metabolic regulation. 1. Metabolic regulation refers mostly to homeostasis. Two models have been proposed to explain the stability paradox. The first model suggests that cell behavior is similar to that expected for a bag of enzymes; thus, the mechanism is achieved depending on the type of enzyme involved: (i) enzymes obeying Michaelis–Menten kinetics function at near-equilibrium conditions and display high catalytic capacities ensuring strong responses to small changes in [substrate]/[product] ratios; (ii) with allosteric enzymes functioning far from the equilibrium, strong changes in rate are obtained with small changes in key effectors; (iii) with enzyme activity regulated notably by covalent modifications (phosphorylation– dephosphorylation, acetylation– deacetylation, etc.), the ratio of catalytically active-to-inactive enzyme accounts for large effects despite minimal change in substrate concentration. With respect to this so-called traditional model, homeostasis is achieved through (i) simple feedback and mass action controls, (ii) allosteric controls, (iii) regulation of the concentration of catalytic sites (covalent modification, activation through translocation, etc.). The second model assumes that most metabolic systems function within a dynamic space where order, structure, and circulation constrain the intracellular behavior of molecules. In this model, intracellular convection would be the critical means for regulating rates of enzyme–substrate encounter (reviewed in 24). 2. Genetic regulation controls when and how often a given gene is transcribed by molecular signals. In genetically controlled processes, the protein encoded by one gene often regulates expression of other genes. The regulation of expression occurs during initiation and early elongation. This involves the assembly of a preinitiation complex at the promoter. This complex formation is regulated by positively and negatively acting factors that bind to specific 4 MAGALI ROUX-ROUQUIE promoter elements: a myriad of activation and repression regions have been identified (25,26). Repression regions are found to be alanine-rich or proline-rich domains. In addition, high-charge regions seem a common feature among other repression motifs. Once repression domains have been carefully mapped, a remarkably small region exhibits the repressor activity. In contrast, activation domains tend to map larger and even multiple regions, similar to the modular large repression region of nuclear receptors. In a recent review, we compiled most of the human activator and repressor complexes as well as the associated diseases (27). The range of mechanisms used by different types of protein–protein interactions is just beginning to be elucidated. It involves protein–protein binding with targets consisting of basal transcription factors, and coactivator and corepressors proteins. The identification of additional mechanisms will continue to develop this emerging field. In particular, the assembly of stereospecific DNA–protein complexes seems an important mechanism to explain the enhancer function in genetic regulation (28). Recently, there have been significant advances in understanding the molecular mechanisms by which, in vivo, transcription is celland tissue-regulated. Studies of gene regulation in Drosophila muscle cell lineage have revealed combinatorial control involving the dispensable MEF2 factor and other MADS-box transcription factors as well as additional lineage-restricted activators that lead to the activation of specific subprograms (syntagms) even within a single muscle cell type (reviewed in 29). Similarly, the mechanisms of transcriptional controls have been dissected in adaptative thermogenesis to assess the function of mitochondrial gene regulation in obesity (30). The genetic regulation of many other systems, including aging and apoptosis, are under investigation (31,32). 3. Epigenetic regulation: The logic of gene regulation in prokaryotes and eukaryotes is fundamentally different due to the packaging of eukaryotic DNA into transcriptionally restrictive chromatin, whereas there is no restriction to gain access to prokaryotic DNA (33). Due to the restrictive ground state of eukaryotic DNA, epigenetic regulation refers to different states of phenotypic expression caused by differential effects of chromosome or chromatin packaging rather than by a different DNA sequence, thus resulting in a stable differential expression of identical information in cell population over many generations (34). These phenotypic changes keep the memory of their genetic determi- nation over time. They are generally considered to be cleared on passage through germ cells so that only genetic characters are inherited; nevertheless, some states can be propagated through meiosis in violation of Mendel’s laws of inheritance (35–37). One exciting question to answer now is how these epigenetic events are transmitted during chromosomal duplication. To date, the major types of epigenetic regulation are dosage compensation, imprinting, and position effect variegation. All involve alterations in chromatin structure, while the mechanisms of each are quite different. (i) Dosage compensation: In XX females, and XY or XO males, a number of regulations compensate for sex-chromosome dosage. In mammals, the inactivation is directed by a single X-linked locus Xic (Xinactivating center). Although the Xic functions are contained in a discrete region of the X chromosome, they are genetically separable, indicating that the locus contains several regions. The initial steps of X-chromosome silencing are not known but several features distinguish Xi from active Xa; notably histone H4 is hypoacetylated on Xi compared with Xa or with autosomes. Xi chromosome is also late replicating and is sequestered near the nuclear membrane, a common feature of heterochromatin. All of these features are important to silencing but none of them is currently known to be the primary cause of silencing (reviewed in 38). (ii) Imprinting: the term “imprinting” denotes parent-of-origin-dependent silencing of autosomal genes. To date, more than 20 imprinted genes have been discovered in mice and humans. Imprinting of some of these genes is tissue-specific. Cis-regulatory imprinting control elements (ICEs) have been defined and are absolutely required for correct epigenetic inheritance. One example is the imprinting center for the human chromosome 15q11-q13 of which deletions are found in Angelman and PraderWilli syndromes. Differential methylation is widely held to be the primary cause of genomic imprinting and appears to be exceptionally stable in the vicinity of ICEs (reviewed in 34). (iii) Position effect variegation: In Drosophila, when a gene white (w1) conferring red eye color is translocated to the vicinity of centromeric heterochromatin, the inactivating effect of heterochromatin is thought to spread into the gene w1, conferring the white-mottled, clonally inherited variegated phenotype called position-effect variegation (PEV). Many suppressors have been found in Drosophila GENETIC AND EPIGENETIC REGULATION SCHEMES and S. pombe that promotes PEV. The sequence of these transacting factors containing a chromodomain suggests that silencing occurs through the organization of a heterochromatin-like structure (35). COMPLEXITY AND DEVELOPMENTAL STABILITY Complexity is the property of systems likely to express behaviors which are not all predeterminable (necessary) although potentially anticipable (possible) (39). Accordingly, one presumed necessary is only one possible, even if it is the most frequently observed. A matricial model would help to explain these behaviors (Le Moigne, personal communication). Concretely, the term located at the intersection of line C (which represents the state C of the system) and column T j (which refers to the instant j), means that the system is in the state C j at the time T j; nevertheless, because of its essential unpredictability, it could have “chosen” the states B j, D j, or E j as well. In this respect, the heterogeneity emerging from isogenic or clonal cell cultures after serial passages (even in carefully controlled conditions) has been related to the complex nature of living systems. Thus, cells would generate phenotypic diversity corresponding to states B j, C j, D j, etc., of the former matrix. Similarly, pathogenic bacteria and viruses use phenotypic variations to modify surface antigens in order to escape the host cell. The molecular origin of this randomness is poorly understood. It has been suggested that the transcription of individual genes occurs randomly and infrequently (40). A stochastic mode for gene expression has been shown, with proteins being produced in short bursts of variable numbers of molecule occurring at random time intervals (41,42). These fluctuations (noise) in gene expression could lead to qualitative differences in cell phenotypes if the expressed gene acts as input to downstream regulatory thresholds. Otherwise, cells could use transcriptional noise to develop epigenetic diversity and suit environment changes. To some extent, cells “select” one or several states (boxes) within the matrix of possible. Despite these genetic variations, how is biological stability achieved? Many means of regulation have highly predictable outputs and some pathways take place so precisely that one could use them to adjust a clock. In addition to the feedback loops that sus- 5 tain cell organization, sui generis properties of these circuits would account for their deterministic behavior. They are: 1. Redundancy (injectivity). Redundant genes are genes able to substitute for a loss of function in one member of the group. It is well-researched in muscle cell differentiation: specification into myoblasts of the cells of the vertebrate lineage requires the expression of two myogenic master genes (MyoD and Myf5) whose encoded transcription factors activate the genes of two other muscle-specific transcription factors (MRF-4 and Myogenin). These four myogenic gene-encoded products are part of redundant and complex cross-activation network in which any one of the four master proteins accumulating to a threshold level leads to a self-maintaining myoblast (43). 2. Pleiotropy (surjectivity). Pleiotropy is the property of a single mutant gene that influences multiple phenotypic traits (44). It refers to multiple functions embedded in the same object. Different biological concepts are hidden behind this term and seven types of pleiotropy have been listed (45). Among them, combinatorial pleiotropy is the most frequent. It concerns a gene product that interacts with a variety of different partners in different cell types. Mutations affecting such proteins have broad effects. Most of transcription factors as well as cofactors fall into this category (review in 27). Another representative illustration of this functional versatility is shown by cytokines and growth factors that play pivotal roles in cell growth, differentiation, and survival (46). Otherwise, unifying pleiotropy describes multifunctional proteins that carry multiple domains involved in distinct functions. This is notably the case for the multifunctional phosphoglucose isomerase, which is both an autocrine motility factor and a neuroleukin (47). But compared to combinatorial pleiotropy, this is of a different nature. Thus, the dynamic stability in the regulatory circuits results principally from the exploitation of redundancy and feedback. Using several cycles of positive and negative redundant circuitry, the system can respond in a robust way despite substantial changes in local conditions (damage, mutation, etc.). Digital organisms have shown that the more complex genetic networks are, the more robust they are with respect to the average effects of single mutations (48). 6 MAGALI ROUX-ROUQUIE FIG. 2. Genetic and epigenetic regulation schemes. Numbers in parentheses (right) refer to gene level (1), gene product level (RNA or protein) (2), functional level (3) illustrated with syntagms (i.e., a subprograms allowing special functions to develop). Gray arrows (left) represent the external fluctuations due to the environment; the perturbations of endogenous origin are not represented. GENETIC AND EPIGENETIC REGULATION SCHEMES The steady state of a living system can be characterized as a competition between stabilizing system dynamics and destabilizing random fluctuations. As mentioned above, Morin first linked positive- and negative-feedback concepts into a unique loop (Fig. 1). This is far from the traditional design that conceived positive and negative feedback only in a disjunctive way. This joined representation of positive/negative feedback refers to coordination, antagonism, and competition (8). Many examples can be found to illustrate the biological consistency of functional links between positive and negative feedback. To bring out the intelligibility of genetic and epigenetic regulations, I attempted a unifying scheme by taking advantage of the Morin’s feedback loop and subsuming all concepts examined above (Fig. 2). Epigenetic regulations that are ultimately led by gene activation and/or repression are included this representation. In the upper part of Fig. 2a, genes b and c are under the control of the trans-acting factor A. Links determine the relations involving the products B (purple), C (green), and D (gray) of the gene b, c, and d, respectively. These relations are: (i) regulation loops: with direct positive feedback (link 2), indirect cis-acting positive feedback (link 4) and indirect cis-acting negative feedback (link 3). In addition, the product D negatively self-regulates gene d. (ii) stabilization sequences using pleiotropic and redundant properties (link 5 to syntagm I and link 6 to syntagm II): product C shows pleiotropy (links to syntagms I and II) and product D is redundant with products B and C. In the upper part of Fig. 2b, homologous domains and/or motifs in genes b, c, and d are represented as well as the corresponding multifunctional proteins (colored barrels). In protein B, one motif (black bar- 7 GENETIC AND EPIGENETIC REGULATION SCHEMES rel) is involved in indirect negative feedback of gene d, whereas another domain (red barrel) accounts for pleiotropic role of protein B in syntagm II. Since a regulatory protein may act in combination with other signals to control many other genes, complex branching networks of interactions are possible, as previously shown (27). A prototypical scheme of gene regulation is of particular interest in pointing out the concepts with which we are dealing. In addition to the notions clarified above, this representation emphasizes the (i) action performed by the objects (gene and/or gene products) within an (ii) environment that fluctuates in order to realize a particular (iii) project (syntagms and functions). NEED FOR AN ALTERNATIVE PARADIGM: FROM AN ANALYTICAL APPROACH TO A SYSTEMIC MODELING OF GENETIC REGULATION The regulation schemes presented above stress the problems that must be dealt with in achieving and coordinating cellular functions, i.e., homeostasis and/or differentiation under the constraints of exogenous and/or endogenous fluctuations. Because of the regulations, living cells are complex systems of self-organizing components showing properties of emergence and selectivity. This is completely different from a bag containing a set of static molecular and macromolecular objects combined. These statements raise the problem of implementing a theoretical framework to clarify these new aspects of biology. Over the past decades, considerable advances have been made using analytical approaches developed in the fields of molecular biology, biochemistry, and genetics. Major findings concern cell-specific gene transcription, DNA arrangements, chromatin structure, combination of regulatory factors, intraand extracellular signaling, cell competence, apoptosis, etc. Such progress has been made by increasingly sophisticated molecular dissection of prokaryotic and eukaryotic cells using disjunctive logic. This approach implements the analytical modeling paradigm inherited from “Le Discours de la Méthode” of Descartes (1637), and aims to answer the question, What is it made of? According to Descartes’ second principle, the analytical framework isolates objects under study to divide them into simple elements taking in account neither their organization nor their context. The object is viewed as a static struc- TABLE 2 Analytical and Systemic Paradigms: Major Concepts Inquiry System type Logic Element Component Context Organization Behavior Analytical paradigm Systemic paradigm What is it made of ? Complicated Disjunctive Simple Object None Synchronic Deterministic What does it do? Complex Conjunctive Implex Process Included Diachronic Teleological-adaptive (self-organized and self-organizing) ture. This representation could generate mistaken conclusions, because it makes the phenomena static, notably omitting the interactivity between objects and environmental effects (49). In fact, the integration of data realized to represent the regulation schemes in Fig. 2 required a shift in the theoretical framework. The cybernetic modeling paradigm (50) was indeed the first attempt to take into account the interactions between system components by proposing the first regulation formalism of comportment on the “goal-seeking” type. This paradigm has often proved to oversimplifying the multifunctional complexity of living systems. Consequently, many researchers have tried to develop a less limiting concept of the modeling of systems seen as complex. The works of Yovit and Cameron (51) and, in particular, Von Foerster (52) should be credited for the renewal of interest in modeling paradigms of complex systems. The first presentation of such can be found in Le Moigne’s Systemics (1,53). The systemic paradigm aims to answer the questions: What does it do? and Why, When, and Where, suggesting some answers to the How question, which can be of the “black box” (or more generally of the functional) type. The systemic paradigm appears to be of great interest in the study of concepts for the representation of genetic regulations. A presentation of the Le Moigne’s Systemics has been made recently (54). The key epistemological features of the Systemic Modeling Paradigm are summarized in Table 2. The major concepts focus on modeling of the effective processes instead of the static objects or state. Processes are seen as active, evolving, and transforming themselves and their environment along time, in an organized and organizing manner. Those multifunctional processes are also seen as being able to even- 8 MAGALI ROUX-ROUQUIE tually elaborate their own, dependent (teleologicaladaptive) strategies, and are not exclusively reduced to some unique, stable exogenous program. This recursive and irreversible behavior can be more easily described today through the use of computer simulation. The systemic paradigm is of great help in exploiting the current flood of descriptive knowledge (55). tastasis: an imbalance of positive and negative regulation. Cancer Res 51:5054s–5059s, 1991. 15. Cato AC, Wade E. Molecular mechanisms of anti-inflammatory action of glucocorticoids. Bioessays 18:371–378, 1996. 16. Loewen PC, Hu B, Strutinsky J, Sparling R. Regulation in the rpoS regulation of Escherichia coli. Can J Microbiol 44:707–717, 1998. 17. Miyazono K. Positive and negative regulation of TGF-beta signaling. J Cell Sci 113:1101–1109, 2000. 18. Demongeot J, Kaufman M, Thomas R. Positive feedback circuits and memory. C R Acad Sci 323:69 –79, 2000. 19. Thomas R. Laws for the dynamics of regulatory networks. Int J Dev Biol 42:479 – 485, 1998. 20. Hajnoczky G, Thomas AP. Minimal requirement for calcium oscillations driven by the IP3 receptor. EMBO J 16:3533– 3543, 1997. 21. Dolmetsch RE, Xu K, Lewis RS. Calcium oscillations increase the efficiency and specificity of gene expression. Nature 392:933–936, 1998. 22. Li W, Llopis J, Whitney M, Zlokarnik G, Tsien RY. Cellpermeant caged InsP3 ester shows that spike frequency can optimize gene expression. Nature 392:936 –941, 1998. 23. Thieffry D, Huerta AM, Pérez-Rueda E, Collado-Vides J. From specific gene regulation to genomic networks: A global analysis of transcriptional regulation in Escherichia coli. Bioessays 20:433– 440, 1998. 24. Hochachka PW. The metabolic implications of intracellular circulations. Proc Natl Acad Sci USA 96:12233–12239, 1999. ACKNOWLEDGMENTS I thank Jean-Louis Le Moigne for helpful comments and writing that provoked some of the ideas for this review; André Rouquié and François Kourilsky for continuous support; Christiane Nivet and Béatrice Thomasset for carefully reading the manuscript on a sunny Sunday. This work is dedicated to Evelyne Roux. REFERENCES 1. Le Moigne JL. Théorie du Systéme Général, Théorie de la Modélisation. Paris: PUF, 1977, 1995. 2. Tada T. The immune system as a supersystem. Annu Rev Immunol 15:1–13, 1997. 3. Garcia-Bellido A. The Bithorax-syntagma. In Advances in Genetics, Development and Evolution in Drosophila (Lakovaara S, Ed.). New York: Plenum, pp. 135–148, 1981. 4. Huang F. Syntagms in development and evolution. Int J Dev Biol 42:487– 494, 1998. 25. 5. Oellers N, Hafen E. Biochemical characterization of rolledSem, an activated form of Drosophila mitogen-activated protein kinase. J Biol Chem 271:24939 –24944, 1996. Tjian R, Maniatis T. Transcription activation: A complex puzzle with few easy pieces. Cell 77:5– 8, 1994. 26. 6. Nedelec FJ, Surrey AC, Maggs AC, Leibler S. Self-organization of microtubules and motors. Nature 389:305–308, 1997. Hanna-Rose W, Hansen U. Active repression mechanism of eukaryotic transcription repressors. Trends Genet 12:229 – 234, 1999. 27. 7. Lambert O, Gerke V, Bader MF, Porte F, Brisson A. Structural analysis of junctions formed between lipid membranes and several annexins by cryo-electron microscopy. J Mol Biol 272:42–55, 1997. Roux-Rouquie M, Chauvet ML, Munnich A, Frezal J. Human genes involved in chromatin remodeling in transcription initiation and associated diseases: An overview using the GENATLAS database. Mol Genet Metab 67:261–277, 1999. 8. Morin E. Method, Vol. 1 Peter Lang Publishing, Inc., New York. American University Studies. Series V: Philosophy, pp 219 –223, 1992. 28. Roux-Rouquie M, Marilley M. Modeling of DNA local parameters predicts encrypted architectural motifs un Xenopus laevis ribosomal gene promoter. Nucleic Acids Res, in press, 2000. 9. Rowley DA, Kohler H, Cowan JD. An immunologic network. Contemp Top Immunobiol 9:205–230, 1980. 29. 10. Maniatis T, Goodbourn S, Fischer JA. Regulation of inducible and tissue-specific gene expression. Science 236:1237– 1245, 1987. Firulli AB, Olson EN. Modular regulation of muscle gene transcription: A mechanism for muscle cell diversity. Trends Genet 13:364 –369, 1997. 30. Lowell BB, Spiegelman BM. Towards a molecular understanding of adaptive thermogenesis. Nature 404:652– 660, 2000. 11. Raymondjean M, Vaulont S, Cognet M, Decaux JF, Puzenat N, Bergot MO, Kahn A. Positive and negative regulation of gene expression by insulin and glucagon: The model of Ltype pyruvate kinase gene. Biochimie 3:41– 45, 1991. 31. Osiewacz HD. Genetic regulation of aging. J Mol Med 75: 715–727, 1997. 32. 12. Lowy DR, Zhang K, DeClue JE, Willumsen BM. Regulation of p21ras activity. Trends Genet 7:346 –351, 1991. Wyllie AH. The genetic regulation of apoptosis. Curr Opin Genet Dev 5:97–104, 1995. 33. 13. Roeder RG. The complexities of eukaryotic transcription initiation: Regulation of preinitiation complex assembly. Trends Biochem Sci 16:402– 408, 1991. Struhl K. Fundamentally different logic of gene regulation in eukaryotes and prokaryotes. Cell 98:1– 4, 1999. 34. Lyko F, Paro R. Chromosomal elements conferring epigenetic inheritance. Bioessays 21:824 – 832, 1999. Liotta LA, Stetler-Stevenson WG. Tumor invasion and me- 35. Klar AJ. Propagating epigenetic states through meiosis: 14. GENETIC AND EPIGENETIC REGULATION SCHEMES 36. 37. 38. 39. 40. 41. 42. 43. 44. 45. 46. Where Mendel’s gene is more than a DNA moiety. Trends Genet 14:299 –310, 1998. Cavalli G, Paro R. The Drosophila Fab-7 chromosomal elements conveys epigenetc inheritance mitosis and meiosis. Cell 93:505–518, 1998. Morgan HD, Sutherland HG, Martin DI, Whitelaw E. Epigenetic inheritance at the agouti locus in the mouse. Nature Genet 23:314 –318, 1999. Meller VH. Dosage compensation: Making 1X equal 2X. Trends Cell Biol 10:54 –59, 2000. Le Moigne J-L. Le Constructivisme tome 1: des fondements. Paris: ESF, p 177, 1994. Ko MSH. Induction mechanism of a single gene molecule: Stochastic or deterministic? Bioessays 14:341–346, 1992. McAdams HH, Arkin A. Stochastic mechanisms in gene expression. Proc Natl Acad Sci USA 94:814 – 819, 1997. McAdams HH, Arkin A. It’s a noisy business! Genetic regulation at the nanomolar scale. Trends Genet 15:85– 89, 1999. Weintraub H. The MyoD family and myogenesis: Redundancy, networks and thresholds. Cell 75:1241–1244, 1993. Waxman D, Peck JR. Pleiotropy and the preservation of perfection. Science 279:1210 –1213, 1998. Hodgkin J. Seven types of pleı̈otropy. Int J Dev Biol 42:501– 505, 1998. Hirano T. Molecular basis underlying functional pleiotropy 9 of cytokines and growth factors. Biochem Biophys Res Commun 260:303–308, 1999. 47. Sun YJ, Chou CC, Chen WS, Wu RT, Meng M, Hsiao CD. The crystal structure of a multifunctional protein: Phosphoglucose isomerase/autocrine motility factor/neuroleukin. Proc Natl Acad Sci USA 96:5412–5417, 1999. 48. Lenski RE, Ofria C, Collier TC, Adami C. Genome complexity, robustness and genetic interactions in digital organisms. Nature 400:661– 664, 1999. 49. Van Regenmortel MHV. Molecular recognition in the postreductionism era. J Mol Recognit 12:1–2, 1999. 50. Klir J, Valach M. Cybernetic Modelling. London: Iliffe, 1967. 51. Yovit M, Cameron S. Proceedings of the 59th Conference on Self-Organizing Systems New York: Pergamon, 1960. 52. Von Foerster H. Observing Systems. Seaside: Intersystems, 1984. 53. Le Moigne JL. la Modelisation des Systémes Complexes. Paris: Dunod, 1990. 54. Eriksson DM. A principal exposition of Jean-Louis Le Moigne’s Systemic theory. Cybernetics Hum Knowing 4:35– 77, 1997. 55. Roux-Rouquie M, Bravais E, Capponi C. Systemic modeling of biological functions application to the development of the GENINTER software dedicated to the compilation of interrelationships between genes and/or gene products. In Advanced Studies in Plant Biology: Integrating Plant Systems (Greppin, Ed.). University of Geneva, in press, 2000.