Perspectives

steady-state spatial patterns could also arise T i m e l i n e from such processes in living systems21. The full formalization of the nature of Self-organization in cell biology: self-organization processes came from the work of Prigogine on instabilities and the a brief history emergence of organization in ‘dissipative systems’ in the 1960s22–24, and from Haken who worked on similar issues under the Eric Karsenti name of synergetics11 (Timeline). Abstract | Over the past two decades, molecular and cell biologists have made It was clear from the outset that the emergence of dynamical organization important progress in characterizing the components and compartments of the observed in physical and chemical systems cell. New visualization methods have also revealed cellular dynamics. This has should be of importance to biology, and raised complex issues about the organization principles that underlie the scientists who are interested in the periodic emergence of coherent dynamical cell shapes and functions. Self-organization manifestations of life and developmental concepts that were first developed in chemistry and physics and then applied to biology have been actively working in this field19,25–29. From a more general point various morphogenetic problems in biology over the past century are now of view, Kauffman built on the ideas of beginning to be applied to the organization of the living cell. Prigogine and Haken in an attempt to explain the origin of order in biology30–32. One of the most fundamental problems in this complex state of living matter as a self- Self-organization was also invoked to biology concerns the origin of forms and organized end8–10. This led him to question explain the formation of regular patterns in their associated functions. This has been a the validity of using the causality principle the fur of animals and the collective behav- long-lasting question in developmental of classical physics to explain life, and to iour of organisms in ant colonies, termite biology, but similar questions must also suggest that a new kind of science would be nest building, schools of fish and flocks of be addressed at the cellular level. Since required to study how purpose and means birds33,34. The importance of self-organiza- the discovery of the structure of DNA, the are intricately connected8. tion processes in molecular cell biology genome has often been thought of as the The new science he was talking about began to be recognized in the 1980s and overriding architect: a given combination did emerge much later, from observations 1990s1,35–40, but only really started to gain of genes that determines the phenotype and studies made by chemists and physicists momentum recently6,41–43 (Timeline). through a linear chain of causal events. who discovered new, more complex forms In the following article, I do not deal The problem is that embryogenesis and of causality than what Kant had foreseen11,12. with developmental biology issues but dynamic cell forms and functions emerge Ironically, although Kant attempted to specifically focus on how self-organization from multiple molecular interactions and characterize life as a self-organization principles and mechanisms (BOX 1) can help interconnected regulatory feedback loops1–4. process in opposition to non-living mat- to understand subcellular and whole-cell Moreover, many parameters, such as physi- ter, the first well-defined concepts and morphogenesis. I first summarize the cal constraints and collective behaviours, are observations of self-organized processes essence of the theory of self-organization in not under the direct control of the genome. came from theoretical considerations by physico-chemical systems in simple terms. Therefore, we cannot hope to explain cell Lotka13,14, from chemistry by Bray15 and I then show how this concretely applies to morphogenesis, for example, by invoking from the Belousov–Zhabotinsky reaction16–19 some examples of cell organization and simple linear chains of causal events that link (Timeline). Chemical oscillations emerged function. genes to phenotypes5–7. from reaction–diffusion processes that It seems that the philosopher Kant was were formalized in mathematical terms Self-organization concepts the first to define life as a “self-organized, by Kolmogorov et al.20 in the 1930s and by The initial definition of self-organization self-reproducing” process (Timeline). Turing in the 1950s, who predicted that by Kant as a characteristic of living systems Through pure reasoning, he defined life as implied the existence of a loop between the emergence of functions by self-organiza- organization and function. A simpler tion. He said that in an organism, every It was clear from the outset definition used by modern scientists is that part owes its existence and origin to that that the emergence of dynamical dynamic organization emerges from the of the other parts, with the functions organization observed in physical collective behaviour of ‘agents’, the individual that are attributed to a complete living organ properties of which cannot account for or organism emerging from the properties and chemical systems should be the properties of the final dynamic pattern. of the parts and of the whole. He defined of importance to biology... This definition is more general and has the

P e r s p e c t i v e s

Timeline | Key events in the application of self-organization concepts in cell biology

(1972–1977) Biological pattern formation30,31. (1999–2005) Kant and the self- First oscillating Exploratory Reaction–diffusion organized nature chemical reaction Turing (1972–1977) Oscillations behaviours in cell Self-organized and intracellular (2005–2007) The self- of life8–10. in solution15. patterns21. in glycolysis25–27. morphogenesis36. microtubule patterns37,38. morphogenesis3,4. organized cell41,43,80,88,89.

1790 1900 1921 1951 1952 1967 1972 1977 1986 1990 1991 1997 1999 2003 2005

Discovery of the Belousov–Zhabotinsky Dissipative systems11,22. (1977–1984) Experimental Turing (1997–2001) Emergent Self-organized cell polarity91–94. Bénard rolls44. oscillatory reaction16–19. Multicellular29,33 patterns46,47. cytoskeleton patterns and mitotic spindle from collective self-organization39. Self-organized animal behaviours57,58. populations33,34.

advantage of being applicable to systems that Origins in thermodynamics. At first sight, energy that flows through it can be used to do not necessarily acquire a function, even the spontaneous emergence of order in the decrease its entropy (that is, generate order). though they become dynamically organ- universe contradicts the second law of thermo Molecules can suddenly organize themselves ized. The understanding of the emergence dynamics: a thermodynamically closed sys- in dynamic patterns. Bénard rolls44,45 (long of function can be studied separately. The tem settles in the most disordered state (that itudinal cylinders of liquid molecules that other advantage of this definition is that it is, the state with the highest entropy, when form precise and stable dynamic patterns) establishes how self-organized dynamical molecules occupy all of the space randomly; represent such an example and were called systems should be studied: the goal of the FIG. 1a). Ordered states can and do emerge at ‘dissipative structures’ by Prigogine22 (FIG. 1b). science of self-organization is to identify thermodynamic equilibrium (for example, the principles and mechanisms by which an crystals, lipid bilayers, molecular complex Collective behaviour and the Bénard rolls. ensemble of agents in interaction evolves formation), but they are static. Bénard rolls form when a liquid is heated towards a particular dynamical temporal or In a thermodynamically open system from below, which generates a temperature spatial pattern. that receives energy from the outside, the gradient. Molecules at the bottom of the container are more agitated than at the top, creating a lighter layer of liquid than that Box 1 | Self-organization concepts and mechanisms at the top. Roll formation results from local Self-organization occurs when elements interact dynamically with each other to generate a system instabilities that lead the system to break its that acquires emergent properties that cannot be directly predicted from the individual properties symmetry when molecules start to behave of the elements. This only happens when the system dissipates energy. collectively. This happens at different critical temperatures for different fluids, but always Principles Mechanisms Examples in the cell occurs when all of the parameters balance Thermodynamics: non- Thermal, chemical or other ATP consumption coupled to each other so that they satisfy a universal equilibrium thermodynamics. energy dissipation that is dynamic pattern formation. number, called the Rayleigh number, which associated with dynamic 33 pattern formation. equals 1708 under specific conditions . The system can generate rolls that, at Symmetry breaking: occurs Gravity, temperature or Intrinsic asymmetry of a given position in the container, move when a system switches chemical gradients, local agents, nonlinear reactions, from one symmetry level to fluctuations. stereospecific localization clockwise or counter-clockwise with alternate another. of enzymes, pre-existing orientation. When the temperature is raised structures. to the critical value, the system bifurcates Emergence: a new property Collective effects and Cytoskeleton behaviour, between two alternative steady states (FIG. 1b). that arises from the collective reactions that lead to systems enzymatic oscillators, The bifurcation is based on local fluctuations behaviour of agents. properties. functional networks. that occur at the critical temperature and Robustness: the system Feedback loops, physical or Reaction networks with is almost irreversible. This comes from the evolves towards a steady collective constraints. two-state systems, such as fact that the molecules in the liquid begin to state that constrains its kinase–phosphatase or small behave collectively, all moving up together agents to remain within this G‑protein systems, combined steady state. with feedback loops. Physical, on one side of the roll and downwards on the chemical and collective other (FIG. 1b). A long-range correlation has constraints. been established between the molecules of the Bifurcation: the system Local instabilities at Toggle switches between system: the whole pattern of rolls in the cont moves from one steady state critical parameter values, network states, switches ainer emerges from the collective properties of to another when a specific nonlinearities. between collective dynamics the molecules in the fluid and the geometry parameter varies around a states. of the container, and cannot be predicted critical value. from the properties of any of its parts.

P e r s p e c t i v e s

Feedback loops and spatio-temporal pat- a terns. In the example of Bénard rolls, order emerges from collective physical inter actions. In Belousov–Zhabotinsky reactions and Turing patterns, order emerges from the nonlinear kinetic properties of chemical b reactions: that is, from a combination of Up diffusion and feedback loops in the reaction system. Turing21 showed that a system of 0 Bifurcation reactants that are initially homogeneously distributed in a solution can generate Down products that segregate in spatial patterns 0 0 if certain conditions are met. The product tc Temperature of a reaction should act as a short-range Figure 1 | Bénard rolls. a | In a liquid layer, molecules are agitated by thermal motion. b | The molecules positive activator of its own production in the liquid layer are heated from below (red zone) and self-organizeNatur einto Revie rollsws (drawn| Molecular in cross-section) Cell Biology when the temperature reaches a critical value (t ). At this value, the molecules start to move collectively while activating the production of an c 46,47 either up or down at point 0, which determines the alternative orientation of the rotation of the rolls inhibitor that diffuses much faster . This throughout the layer. The orientation of the rotation choice is unpredictable and determined by local forms a local positive feedback loop and fluctuations at t . a long-range negative feedback loop. The c symmetry of the solution is broken by diffusion-driven instabilities that get amplified by the feedback loops. Waves or steady-state and extremely varied dynamical inter In 1991, Verde and colleagues showed patterns can emerge48,49. Again, the system actions between molecules that require that randomly nucleated microtubules in can bifurcate between different states. In energy dissipation (self-organization). frog egg extracts organized themselves biology, such reaction–diffusion processes Here, I give some recent examples in which with the minus-end motor dynein to form have obvious morphogenetic potential50,51. the problem of cellular self-organization steady-state asters38. This was further The problem is to identify them and under- has been addressed — from well-studied developed in vitro57 and then in silico58. stand how biological molecules interact to systems to more complex and less Computer simulation analyses showed that form the appropriate reaction networks. understood ones. various patterns can emerge from interac- Biologists have been trying to find reaction– tions between similar components as a diffusion mechanisms that fit the Turing ...in the living cell, self- function of their physical properties, com- 58,59 model exactly, especially in developmental organization processes also plexity and concentration . Asters, vort biology29,28. Because biological molecules ices, antiparallel microtubule bundles and are complex, various types of reaction– occur through the dissipation of spindle networks were observed (FIG. 2a). In diffusion processes that involve enzymatic ATP or GTP. these systems, symmetry breaking comes reactions may occur — these might have from the asymmetric shape of the parts: the powerful morphogenetic properties without tubulin molecule is asymmetric60 and the having the exact characteristics of the Patterns and oscillators from collective motors move towards one microtubule end Turing equations3,4. In addition, combina- behaviours. In 1990, Tabony and Job37 or the other61. Therefore, the system has a tions of positive and negative feedback loops published a paper in which they showed that built-in symmetry-breaking mechanism or other properties, such as ultrasensitivity pure microtubule solutions self-organized (FIG. 2a). Self-organization comes from the and cooperativity, provide powerful sources into stripes when they were incubated for collective behaviour of the motors and of nonlinearities that can lead to the build- several hours at 33oC, and noticed the ana microtubules, with the energy being dis- ing of enzymatic oscillators or steady-state logy with patterns formed in the Belousov– sipated by the motors as they move along patterns27,52. Therefore, when trying to Zhabotinsky reaction. Computer simulations microtubules. understand the self-organization properties suggest that the patterns may arise through Recently, these observations have been of cells or embryos, it is more important to a reaction–diffusion mechanism53. extended to the acto-myosin system62. simply treat them as nonlinear dynamical However, different interpretations of these Just as for microtubules, myosin II systems than to try to find exactly what observations have been proposed and have crosslinks actin filaments and the system Turing had predicted. suggested that, instead, the patterns arise self-organizes into various patterns includ- through collective effects that result from ing rings. Physicists call these systems Self-organization and the cell dynamic instability coupled to microtubule active gels (as opposed to gels that are A large part of cell organization depends buckling54. In fact, the stripes are apparently made of polymers that do not contain on self-assembly processes that do not made of aligned microtubules and, interest- motors) and have developed a theory that involve energy dissipation (thermodynamic ingly, the patterns formed are affected by predicts the behaviour of these active gels63. equilibrium). But in the living cell, self- boundary conditions55. These properties For an example of structures that are gen- organization processes also occur through may contribute to complex patterns that are erated by self-assembly at thermodynamic the dissipation of ATP or GTP. In fact, the formed in vivo56, but microtubule equilibrium, see the study by Haviv et al.64, dynamic order of the cell results from a pattern formation in cells often requires who obtained self-assembled actin stars combination of complex stereospecific both the regulation of microtubule dynamics that formed in a passive gel of actin and interactions (deterministic self-assembly) and the contribution of motors. actin regulators.

P e r s p e c t i v e s

a cycle in eukaryotes as an example because it is probably the best-characterized system. Minus-end motor (Further examples in biology in general and Microtubules in cells more specifically can be found in Plus-end motor REFS 27, 29, 42, 66.) + In eukaryotic cells, an oscillator coordi- + nates DNA replication with chromosome – segregation. The cyclin-dependent kinase (CDK) oscillator67,68 (FIG. 3a) self-organizes – – – through the permanent synthesis of one + protein, cyclin, which triggers intertwined + positive and negative feedback loops68–74. + The CDK oscillator is a truly self-organized temporal pattern because it is entirely autonomous. The principle of its mechanism + is analogous to a Belousov–Zhabotinsky reaction, except that it is built of much more + complex molecules. It has two stable states — bifurcation between the two states is – – – – triggered when cyclin reaches a threshold concentration and when it is degraded. It + works because there is a long time-delay built into the negative feedback loop that 72 b WT tea mutant orb mutant leads to cyclin degradation . The function of this oscillator is to change Cell shape Microtubule abruptly the cytoplasmic state of the cell. pattern When CDK1 is inactive, a nucleus assembles and DNA replicates, and when it is active, a mitotic spindle assembles. In both cyto- Nucleus position plasmic states, chromatin breaks the spatial + – + symmetry of the cytoplasm by the local accumulation of regulatory factors (through stereospecific targeting) that set off a series Figure 2 | Examples of self-organized microtubule patterns and cell shapes. a | Self-organization of (enzymatic) reaction–diffusion processes, Nature Reviews | Molecular Cell Biology of mixtures of microtubules and motors. A minus-end motor can, under certain conditions, crosslink which leads to the emergence of a nucleus microtubules and focus the minus ends to form asters (top). A mixture of minus-end and plus-end in interphase and a spindle in metaphase3,75 motors can form various patterns. An antiparallel pattern with overlapping plus ends59 is shown (bot- (FIG. 3b). Therefore, this reaction–diffusion tom). b | Self-organization of microtubule patterns and cell shape in Schizosaccharomyces pombe. process contributes to the formation of spa- The elongated shape of S. pombe forces microtubules to align because microtubules depolymerize tial patterns inside the cytoplasm. The whole when they reach the tips of the cells. Motors (together with crosslinking molecules) force microtu- cell cycle in eukaryotes can be seen as being 87 bules to form antiparallel bundles with the minus ends at the cell centre . In tip-elongation-aberrant based on the principle of self-organization (tea) mutants, microtubules keep growing when they reach the cell tips. Because tip-promoting fac- by reaction–diffusion, both temporally and tors move towards microtubule plus ends and because, in this case, microtubules curl along cell edges, additional growth tips can form, thereby generating T‑shaped cells84. Cells that are mutated spatially. I say reaction–diffusion, although in the Ser–Thr protein kinase Orb6 (orb mutants) have a round shape and microtubules cannot organ- it is important to realize that none of these ize into long bundles83. This shows that self-organization of microtubules and the cell cortex feed processes are true Turing patterns. Indeed, back on each other to generate a self-organized dynamic cell shape. The circularity of this process is the symmetry is not broken by spontaneous shown on the right of the figure. instabilities, but rather by deterministic effects (cyclin synthesis for the oscillator, and stereospecific targeting of a small G‑protein An important aspect of self-organization rebinding, which leads to dynamic instability exchange factor to chromatin for nuclear and in living systems concerns the formation of of the force–velocity relationship and collec- spindle assembly). oscillators. They arise from the collective tive periodic binding and unbinding. This behaviour of cytoskeletal systems or from demonstrates how nonlinear collective effects The emergence of functions. Above, I have enzymatic networks (see below). Kruse and can lead to periodic temporal patterns (for described purified cytoskeletal molecules Jülicher have reviewed this field recently42, many other examples, see REF. 42). that can self-organize into patterns as a result and the first theoretical treatment of the of their collective behaviour. In isolation, emergence of spontaneous oscillations of Patterns and oscillators from enzymatic these patterns have no function because they collective molecular motors was published in feedback loops. Substantial work on oscilla- have nothing to act on. In the cell, however, 1997 (Ref. 65). The principle is relatively sim- tors has already been mentioned above, but this might be different. During mitosis, for ple: a collection of motors can lose its grip on relatively little has been said about pattern example, chromosomes induce the assembly the filaments in a cooperative manner before formation inside cells. Here, I take the cell of a spindle that acts on the chromosomes

P e r s p e c t i v e s

a b RanGDP RanGDP

RanGTP RanGTP Interphase Metaphase Interphase Metaphase Interphase ER

– – RanGDP RanGDP RanGTP RanGTP + +

Time Interphase Metaphase CDK1 concentration c Cyclin B concentration CDK1 activity CDK1 activity Microtubule

Nucleus Metaphase state State transition Interphase state

Chromosome

RanGTP concentration Time Figure 3 | Examples of self-organized cell-cycle processes. a | In eukary- assembly in interphase and spindle assembly in metaphase. The local otes, the timing of the cell cycle is determined by an oscillator that is production of RanGTP is determinedNa bytur ae reaction–diffusionReviews | Molecular mechanism.Cell Biology driven by the accumulation of cyclin B, which binds to cyclin-dependent c | The transition between metaphase and interphase corresponds to a kinase-1 (CDK1). When the concentration of the cyclin B–CDK1 complex bifurcation between two steady states; that is, nuclear components are in reaches a threshold, it triggers a positive feedback loop that leads to the two discrete dynamic interaction states that respond to the presence of abrupt activation of CDK1. When CDK1 activity reaches a threshold level, DNA by self-organizing into either a nucleus or a spindle. When CDK1 is it triggers the delayed degradation of cyclin B (negative feedback), which inactivated at the end of metaphase, the cytoplasm moves through a trans results in the onset of anaphase. b | Chromatin generates a gradient of ient state. This is when chromosomes are segregated. ER, endoplasmic RanGTP that, through a series of complex reactions, triggers nuclear reticulum.

themselves. Therefore, the chromosomes The functional organization of the function of the self-organized cytoskeletal trigger the self-organization of a pattern nucleus in interphase is another interesting structure is triggered by the cell domain on (the spindle) that acquires the function of example of self-organization in which func- which it should act. segregating them76. This is exactly the loop tion and structure are interdependent, as The second example concerns the spon- between organization and function that Kant described by Misteli, Glick and Cook41,43,77–80, taneous beating of axonemes that results was looking for and there is nothing mysteri- although the exact mechanisms involved from the self-organization of bending waves, ous about it — so, we can now understand the are still under investigation. which are generated by the collective effects structure of this logic because we understand Let’s take just two more examples of the of motor activity and bending elasticity of many of the underlying mechanisms. emergence of function by self-organiza- microtubules42,82. Axoneme beating (the There are three interlocked, self-organized tion. We have seen above that actin and function) emerges from the self-organiza- systems that constitute the whole cell cycle: myosin can form self-organized contractile tion properties of the system, which itself the oscillator (temporal self-organization), rings in vitro62. Again, these rings have results from the interaction between dynein the spindle and the nucleus (spatial self- no function. However, in vivo, similar and microtubules. organization), and each subsystem has its rings self-organize, triggered by the small own specific function. However, it is interest- G protein RhoA that is locally activated Self-organized cells. The loops that link ing to note that chromosome segregation at the plasma membrane at the mid-zone self-organization and function are also does not occur while the spindle is at steady of the dividing cell. This contractile ring found when we start to look at the self- state, but during anaphase; that is, while functions to cut the cell in two81. In fact, organization of large systems such as whole the cell bifurcates back towards interphase, this system works exactly like chromo- cells. In Schizosaccharomyces pombe, for when a new steady state is established and a some-induced spindle assembly, both example, the self-organization of micro- nucleus has reassembled (FIG. 3c). logically and mechanistically: the final tubule bundles is required to maintain the

P e r s p e c t i v e s

a b c d I-CAC I-CAC Cytoplasmic [RanGTP]↑ Phosphatase Cytoplasmic Phosphatase [PLK4]↑ [RanGTP]↑ A-CAC A-CAC [PLK4]↑ Local instabilities Local instabilities

[PLK4] [PLK4] Figure 4 | Self-organized versus templated pattern transmission. centrioles, PLK4 needs to be present in high concentrations in the cytoplasm a | A model of centriole duplication or spontaneous formation. The Polo-like to counterbalance the phosphatase enzymeNature Reof vieCAwsC. |I nMol thisecular case, Ce localll Biolog instay- kinase PLK4 is present in very low amounts in the cytoplasm but binds bilities can lead to localized assembly of A‑CAC that could trigger the strongly to existing centrioles (turquoise) at the time of centriole duplica- autocatalytic assembly of centrioles. This is based on the same idea as that tion102. Because its substrate (centriole assembly complex; CAC) and the of Turing: a local positive feedback loop coupled to long-range inhibition opposing phosphatase are in solution, inactive CAC (I-CAC) becomes phos- (the phosphatase). c | The previous scenario is analogous to the mechanism phorylated and activated (A-CAC) just around centrioles, which leads to the of RanGTP-induced local spindle assembly in the presence of a localized assembly of new centrioles that bind more PLK4. This, in turn, leads to the GTP exchange factor on chromatin, or d | global spontaneous spindle autocatalytic assembly of a new centriole. b | In the absence of pre-existing assembly if RanGTP is overexpressed throughout the cytoplasm.

simple cylindrical morphology. But the periodically de novo through instability- self-organization76 (FIG. 4c,d). Indeed, in this reverse is also true: cell shape is important to driven self-organization by virtue of the latter case, when chromosomes are present determine the morphology of microtubule physical properties of their components? they also induce the self-organization of bundles that self-organize through dynamic A good example is the centriole. There spindles around them. However, on over instability, the action of motors and asso have been many discussions about the expression of the signalling molecules that are ciated molecules, and interaction with the templated duplication of centrioles because normally activated by chromosomes, spindles cell periphery83–87. Again, we find a loop of their beautiful geometry and the mecha- self-organize randomly in the cytoplasm in between the whole and the parts88,89 (FIG. 2b). nism of duplication at an exact angle to the absence of chromosomes103 (FIG. 4d). In S. pombe, we really begin to under- each other96–99. However, centrioles do arise The case of the Golgi apparatus is also stand how the nucleus, the microtubule de novo in many cases100, and the recent interesting. Again because of its unique system and cell shape form a circular inte- discovery of Polo-like kinase-4 (PLK4), structure, it has been proposed that the Golgi grated self-organized system in which none which is essential for centriole duplica- duplicates during cell division through a of the elements comes first (FIG. 2b). This is tion, indicates how this de novo formation templated process104. However, further work also a beautiful example of modular self- may occur101. Indeed, overexpression of suggested that, in fact, new Golgi buds out organization in which one versatile system PLK4 triggers the formation of ectopic free from the endoplasmic reticulum (ER) at sites (the microtubule system) interacts in vari- centrioles, which suggests that under some that are apparently specified by the basal ous ways with other parts of the cell that are conditions, pre-existing centrioles act as a body105,106, and it now seems possible that themselves self-organized, to form a whole. concentration spot for PLK4 (that is, it binds the Golgi apparatus self-organizes from the Chemotaxis and cell polarization are also to centrioles)102. If the opposing activity of a ER in response to its very function41,107–109. interesting in this context. Cell polarization phosphatase is present throughout the cyto- can occur in the absence of any cue through plasm, PLK4 may locally phosphorylate key The essence of life and evolution spontaneous symmetry breaking90,91, substrates around pre-existing centrioles, There is a tendency for engineers who involving instabilities in small G proteins, leading to local self-organization of cen enter the field of biology to speak of design. phosphorylation and cytoskeletal networks. triolar components through a reaction– Design does not exist in living matter This self-organized process can become diffusion mechanism (FIG. 4a). In the absence (unless we believe in creationism). Nobody functional for the cell by responding in a of pre-existing centrioles, an excess of PLK4 ‘thought’ through the advantage of positive directional way to signalling gradients90–95. could phosphorylate the same substrate and negative feedback loops to build the by opposing the phosphatase globally and, cell-cycle oscillator in Xenopus laevis. It just De novo versus templated. A discussion of through local instabilities, trigger spatially springs from a mixture of gene products self-organization brings us naturally to the restricted positive feedback loops that lead that interact dynamically with each other, old problem of templated versus de novo to local assembly of the nucleating agents as in Belousov–Zhabotinsky reactions. formation of organelles. It is clear that struc- that are required to initiate centriole self- Something incredibly important for the tural heredity exists: the structure and shape organization (FIG. 4b). This would be similar understanding of the origin of life and evo- of membranes, the Golgi apparatus, centri- in principle to chromatin-induced spindle lution is emerging here: self-organization oles and even whole cytoskeleton patterns principles tell us that if there is an ensemble pass from one generation to the next. But ...it now seems possible of products that can interact dynamically how are these structures transmitted? Does that the Golgi apparatus to reach a functional steady state, they will this occur through some sort of template do so robustly at least under certain condi- that duplicates and on which new structures self-organizes from the ER in tions31. Suddenly, life becomes much less grow, or do these structures assemble response to its very function. improbable, as Kauffman suggested30.

P e r s p e c t i v e s

The principles that are associated with 16. Belousov, B. [A periodic reaction and its mechanism]. 47. Ouyang, Q. & Swinney, H. Transition from a uniform Compilation of Abstracts on Radiation Medicine 147, state to hexagonal and striped Turing patterns. self-organization processes tend to indicate 145 (1959) (in Russian). Nature 352, 610–612 (1991). that the driving force behind the diversity of 17. Zhabotinsky, A. [Periodic processes of malonic acid 48. Shoji, H., Yamada, K., Ueyama, D. & Ohta, T. oxidation in a liquid phase.]. Biofizika 9, 306–311 Turing patterns in three dimensions. Phys. Rev. E Stat. life and its evolution is not mainly selection. (1964) (in Russian). Nonlin. Soft Matter Phys. 75, 046212 (2007). Instead, it may derive largely from the intrin- 18. Zhabotinsky, A. M. & Zaikin, A. N. Autowave 49. Yang, L. & Epstein, I. Oscillatory Turing patterns in processes in a distributed chemical system. J. Theor. reaction–diffusion systems with two coupled layers. sic properties of living matter and the com- Biol. 40, 45–61 (1973). Physic. Rev. Lett. 90, 178303 (2003). bination of various self-organized functional 19. Tabony, J. Historical and conceptual background of 50. Gierer, A. & Meinhardt, H. A theory of biological self-organization by reactive processes. Biol. Cell 98, pattern formation. Kybernetik 12, 30–39 (1972). modules. The maximum diversity of life is 589–602 (2006). 51. Meinhardt, H. & Gierer, A. Pattern formation by local probably represented by the parameter space 20. Kolmogorov, A., Petrovsky, L. & Piskunov, N. An self-activation and lateral inhibition. Bioessays 22, investigation of the diffusion equation combined with 753–760 (2000). within which dynamic interactions between an increase in mass and its application to a biological 52. Tyson, J. J., Chen, K. C. & Novak, B. Sniffers, buzzers, the whole and the parts robustly produce problem. Bull. Uni. Moscow Ser. Int. A1 6, 1–26 toggles and blinkers: dynamics of regulatory and (1937). signaling pathways in the cell. Curr. Opin. Cell Biol. 15, a cell or an organism that can survive in a 21. Turing, A. M. The chemical basis of morphogenesis. 221–231 (2003). given environment. In a sense we are moving Philos. Trans. R. Soc. Lond. B Biol. Sci. 237, 37–72 53. Glade, N., Demongeot, J. & Tabony, J. Comparison of (1952). reaction–diffusion simulations with experiment in back to the views of D’Arcy Thompson, who 22. Prigogine, I. & Nicolis, G. On symmetry-breaking self-organized microtubule solutions. CR Biol. 325, thought that biologists overemphasize the instabilities in dissipative systems. J. Chem. Phys. 46, 283–294 (2002). 3542–3550 (1967). 54. Guo, Y., Liu, Y., Tang, J. X. & Valles, J. M. role of evolution over that of physical laws in 23. Prigogine, I., Nicolis, G. & Babloyantz, A. Polymerization force driven buckling of microtubule the origin of growth and form5. Nonequilibrium problems in biological phenomena. bundles determines the wavelength of patterns Ann. NY Acad. Sci. 231, 99–105 (1974). formed in tubulin solutions. Phys. Rev. Lett. 98, 24. Nicolis, G. & Prigogine, I. Self-Organization in 198103‑1‑4 (2007). Conclusion Nonequilibrium Systems: From Dissipative Structures 55. Cortes, S., Glade, N., Chartier, I. & Tabony, J. to Order Through Fluctuations (Wiley, New York, 1977). Microtubule self-organisation by reaction–diffusion Studying self-organization processes in 25. Goldbeter, A. & Lefever, R. Dissipative structures for processes in miniature cell-sized containers and cell biology forces us to focus on principles an allosteric model. Application to glycolytic phospholipid vesicles. Biophys. Chem. 120, 168–177 oscillations. Biophys. J. 12, 1302–1315 (1972). (2006). and collective behaviours rather than on 26. Boiteux, A., Hess, B. & Plesser, T. Oscillatory 56. Maly, I. V. & Borisy, G. G. Self-organization of single molecules. This makes it necessary phenomena in biological systems. FEBS Lett. 75, 1–4 treadmilling microtubules into a polar array. Trends (1977). Cell Biol. 12, 462–465 (2002). to use mathematics and physics as well as 27. Goldbeter, A. Biochemical Oscillations and Cellular 57. Nédélec, F. J., Surrey, T., Maggs, A. C. & Leibler, S. computer simulations to analyse the often Rhythms: The Molecular Bases of Periodic and Self-organization of microtubules and motors. Chaotic Behaviour (Cambridge Univ. Press, 1996). Nature 389, 305–308 (1997). counter-intuitive properties of dynamical 28. Maini, P. K., Baker, R. E. & Chuong, C. M. 58. Surrey, T., Nédélec, F., Leibler, S. & Karsenti, E. systems. This also forces us to clarify Developmental biology. The Turing model comes of Physical properties determining self-organization of molecular age. Science 314, 1397–1398 (2006). motors and microtubules. Science 292, 1167–1171 the difference between principles and 29. Murray, J. (ed.) Mathematical Biology (Springer, New (2001). mechanisms (BOX 1). York, 2007). 59. Nédélec, F. Computer simulations reveal motor 30. Kauffman, S. At Home in the Universe (Oxford Univ. properties generating stable antiparallel microtubule Eric Karsenti is at the Cell Biology and Biophysics Unit, Press, 1995). interactions. J. Cell Biol. 158, 1005–1015 (2002). European Molecular Biology Laboratory, 31. Goodwin, B. C., Kauffman, S. & Murray, J. D. Is 60. Nogales, E., Whittaker, M., Milligan, R. A. & morphogenesis an intrinsically robust process? Downing, K. H. High-resolution model of the Meyerhofstrasse 1, 69117 Heidelberg, Germany. J. Theor. Biol. 163, 135–144 (1993). microtubule. Cell 96, 79–88 (1999). e-mail: [email protected] 32. Kauffman, S. The Origins of Order: Self-Organization 61. Vallee, R. B. & Stehman, S. A. How dynein helps the doi: 10.1038/nrm2357 and Selection in Evolution (Oxford Univ. Press, 1993). cell find its center: a servomechanical model. Trends 33. Ball, P. The Self-Made Tapestry (Oxford Univ. Press, Cell Biol. 15, 288–294 (2005). 1. Babloyantz, A. Molecules, Dynamics, and Life: An 1999). 62. Backouche, F., Haviv, L., Groswasser, D. & Bernheim- Introduction to Self-Organization of Matter (Wiley, 34. Camazine, S. et al. Self-Organization in Biological Groswasser, A. Active gels: dynamics of patterning New York, 1986). Systems (Princeton Univ. Press, 2001). and self-organization. Phys. Biol. 3, 264–273 2. Murray, J. Discussion: Turing’s theory of 35. Karsenti, E., Newport, J., Hubble, R. & Kirschner, M. (2006). morphogenesis — its influence on modelling biological Interconversion of metaphase and interphase 63. Kruse, K., Joanny, J., Julicher, F., Prost, J. & patterns and form. Bull. Math. Biol. 52, 119–152 microtubule arrays, as studied by the injection of Sekimoto, K. Asters, vortices, and rotating spirals in (1990). centrosomes and nuclei into Xenopus eggs. J. Cell Biol. active gels of polar filaments. Phys. Rev. Lett. 92, 3. Bastiaens, P., Caudron, M., Niethammer, P. & 98, 1730–1745 (1984). 078101–078104 (2004). Karsenti, E. Gradients in the self-organization of the 36. Kirschner, M. & Mitchison, T. Beyond self-assembly: 64. Haviv, L. et al. Reconstitution of the transition from mitotic spindle. Trends Cell Biol. 16, 125–134 (2006). from microtubules to morphogenesis. Cell 45, lamellipodium to filopodium in a membrane-free 4. Kholodenko, B. N. Cell-signalling dynamics in time and 329–342. (1986). system. Proc. Natl Acad. Sci. USA 103, 4906–4911 space. Nature Rev. Mol. Cell Biol. 7, 165–176 (2006). 37. Tabony, J. & Job, D. Spatial structures in microtubular (2006). 5. Thompson, D. W. On Growth and Form (Cambridge solutions requiring a sustained energy source. Nature 65. Jülicher, F. & Prost, J. Spontaneous oscillations of Univ. Press, 1942). 346, 448–451 (1990). collective molecular motors. Phys. Rev. Lett. 78, 6. Kirschner, M., Gerhart, J. & Mitchison, T. Molecular 38. Verde, F., Berrez, J. M., Antony, C. & Karsenti, E. 4510–4513 (1997). “vitalism”. Cell 100, 79–88 (2000). Taxol induced microtubule asters in mitotic extracts of 66. Goldbeter, A. Oscillations and waves of cyclic AMP in 7. Kurakin, A. Self-organization versus watchmaker: Xenopus eggs: requirement for phosphorylated factors Dictyostelium: a prototype for spatio-temporal ambiguity of molecular recognition and design charts and cytoplasmic dynein. J. Cell Biol. 112, 1177–1187 organization and pulsatile intercellular communication. of cellular circuitry. J. Mol. Recognit. 20, 205–214 (1991). Bull. Math. Biol. 68, 1095–1109 (2006). (2007). 39. Karsenti, E. Mitotic spindle morphogenesis in animal 67. Novak, B. & Tyson, J. J. Modelling the controls of the 8. Kant, E. Critique de la Faculté de Juger (Gallimard, cells. Semin. Cell Biol. 2, 251–260 (1991). eukaryotic cell cycle. Biochem. Soc. Trans. 31, Paris, 1985) (in French). 40. Mitchison, T. J. Self-organization of polymer-motor 1526–1529 (2003). 9. Van de Vijver, G. Self-Organization and Emergence in systems in the cytoskeleton. Philos. Trans. R. Soc. 68. Sha, W. et al. Hysteresis drives cell-cycle transitions in Life Sciences (Springer, Dordrecht, 2006). Lond. B Biol. Sci. 336, 99–106 (1992). Xenopus laevis egg extracts. Proc. Natl Acad. Sci. USA 10. Fox Keller, E. Contenders for life at the dawn of the 41. Misteli, T. The concept of self-organization in cellular 100, 975–980 (2003). twenty-first century: approaches from physics, biology architecture. J. Cell Biol. 155, 181–185 (2001). 69. Murray, A. W., Solomon, M. J. & Kirschner, M. W. and engineering. Interdiscip. Sci. Rev. 32, 113–122 42. Kruse, K. & Jülicher, F. Oscillations in cell biology. The role of cyclin synthesis and degradation in the (2007). Curr. Opin. Cell Biol. 17, 20–26 (2005). control of maturation promoting factor activity. 11. Haken, H. Nonequilibrium phase transitions and self- 43. Glick, B. S. Let there be order. Nature Cell Biol. 9, Nature 339, 280–286 (1989). organisation in physics, chemistry, and biology. In 130–132 (2007). 70. Murray, A. W. & Kirschner, M. W. Dominoes and Synergetics: An Introduction (Springer, Berlin, 1977). 44. Bénard, H. Les tourbillons cellulaires dans une nappe clocks: the union of two views of cell cycle regulation. 12. Prigogine, I. & Stengers, I. Order Out of Chaos liquide. Rev. Gen. Sci. Pure Appl. 11, 1261–1271 Science 246, 614–621 (1989). (Bantam, Toronto, 1984). (1900) (in French). 71. Félix, M.‑A., Pines, J., Hunt, T. & Karsenti, E. Temporal 13. Lotka, A. J. Contributions to the theory of periodic 45. Rayleigh, L. On convective currents in a horizontal regulation of Cdc2 mitotic kinase activity and cyclin reactions. J. Phys. Chem. 14, 271–274 (1910). layer of fluid when the higher temperature is on the degradation in cell-free extracts of Xenopus eggs. 14. Lotka, A. Elements of Physical Biology (Williams and under side. Philos. Mag. 32 (1916). J. Cell Sci. 12, 99–116 (1989). Wilkins, Baltimore, 1925). 46. Castets, V. V., Dulos, E., Boissonade, J. & De 72. Félix, M. A., Labbé, J. C., Dorée, M., Hunt, T. & 15. Bray, W. A periodic reaction in homogeneous solution Kepper, P. Experimental evidence of a sustained Karsenti, E. Triggering of cyclin degradation in and its relation to catalysis. J. Am. Chem. Soc. 43, standing Turing-type nonequilibrium chemical pattern. interphase extracts of amphibian eggs by Cdc2 kinase. 1262–1267 (1921). Phys. Rev. Lett. 64, 2953–2956 (1990). Nature 346, 379–382 (1990).

P e r s p e c t i v e s

73. Yang, L., MacLellan, W. R., Han, Z., Weiss, J. N. & 87. Janson, M. E. et al. Crosslinkers and motors organize role of the mother centriole in centriole biogenesis. Qu, Z. Multisite phosphorylation and network dynamic microtubules to form stable bipolar arrays in Science 316, 1046–1050 (2007). dynamics of cyclin-dependent kinase signaling in the fission yeast. Cell 128, 357–368 (2007). 102. Azimzadeh, J. & Bornens, M. Structure and eukaryotic cell cycle. Biophys. J. 86, 3432–3443 88. Carazo-Salas, R. & Nurse, P. Sorting out interphase duplication of the centrosome. J. Cell Sci. 120, (2004). microtubules. Mol. Syst. Biol. 3, 95 (2007). 2139–2142 (2007). 74. Zwolak, J. W., Tyson, J. J. & Watson, L. T. Parameter 89. Castagnetti, S., Novak, B. & Nurse, P. Microtubules 103. Heald, R. & Weis, K. Spindles get the ran around. estimation for a mathematical model of the cell cycle offset growth site from the cell centre in fission yeast. Trends Cell Biol. 10, 1–4 (2000). in frog eggs. J. Comput. Biol. 12, 48–63 (2005). J. Cell Sci. 120, 2205–2213 (2007). 104. Pelletier, L. et al. ��Golgi biogenesis in Toxoplasma 75. Hetzer, M., Gruss, O. J. & Mattaj, I. W. The Ran 90. Devreotes, P. N. & Zigmond, S. H. Chemotaxis in gondii. Nature 418, 548–552 (2002). GTPase as a marker of chromosome position in eukaryotic cells: a focus on leukocytes and 105. He, C. Y., Pypaert, M. & Warren, G. Golgi duplication spindle formation and nuclear envelope assembly. Dictyostelium. Annu. Rev. Cell Biol. 4, 649–686 in Trypanosoma brucei requires Centrin2. Science Nature Cell Biol. 4, E177–E184 (2002). (1988). 310, 1196–1198 (2005). 76. Karsenti, E. & Vernos, I. The mitotic spindle: a 91. Devreotes, P. & Janetopoulos, C. Eukaryotic 106. He, C. Y. et al. Golgi duplication in Trypanosoma self-made machine. Science 294, 543–547 (2001). chemotaxis: distinctions between directional sensing brucei. J. Cell Biol. 165, 313–321 (2004). 77. Cook, P. R. Predicting three-dimensional genome and polarization. J. Biol. Chem. 278, 20445–20448 107. Thiele, C. & Huttner, W. B. Protein and lipid sorting structure from transcriptional activity. Nature Genet. (2003). from the trans-Golgi network to secretory granules — 32, 347–352 (2002). 92. Wedlich-Soldner, R., Wai, S. C., Schmidt, T. & Li, R. recent developments. Semin. Cell Dev. Biol. 9, 78. Iborra, F. J. & Cook, P. R. The interdependence of Robust cell polarity is a dynamic state established by 511–516 (1998). nuclear structure and function. Curr. Opin. Cell Biol. coupling transport and GTPase signaling. J. Cell Biol. 108. Glick, B. S. Can the Golgi form de novo? Nature Rev. 14, 780–785 (2002). 166, 889–900 (2004). Mol. Cell Biol. 3, 615–619 (2002). 79. Meaburn, K. J., Misteli, T. & Soutoglou, E. Spatial 93. Xu, J. et al. Divergent signals and cytoskeletal 109. Simpson, J. C., Nilsson, T. & Pepperkok, R. Biogenesis genome organization in the formation of chromosomal assemblies regulate self-organizing polarity in of tubular ER‑to‑Golgi transport intermediates. translocations. Semin. Cancer Biol. 17, 80–90 neutrophils. Cell 114, 201–214 (2003). Mol. Biol. Cell 17, 723–737 (2006). (2007). 94. Maly, I. V., Wiley, H. S. & Lauffenburger, D. A. Self- 80. Misteli, T. Beyond the sequence: cellular organization organization of polarized cell signaling via autocrine Acknowledgements of genome function. Cell 128, 787–800 (2007). circuits: computational model analysis. Biophys. J. 86, I thank colleagues of the Cell Biology and Biophysics Unit at 81. Piekny, A., Werner, M. & Glotzer, M. Cytokinesis: 10–22 (2004). the European Molecular Biology Laboratory for all of the dis- welcome to the Rho zone. Trends Cell Biol. 15, 95. Wedlich-Soldner, R. & Li, R. Spontaneous cell cussions and collective work. I also thank the three anony- 651–658 (2005). polarization: undermining determinism. Nature Cell mous reviewers who helped bring this text to something that 82. Camalet, S. & Jülicher, F. Generic aspects of axonemal Biol. 5, 267–270 (2003). looks, I hope, like a balanced and coherent whole. I apologize beating. New J. Phys. 2, 24.1–24.23 (2000). 96. Nigg, E. A. Centrosome duplication: of rules and to all those who have not been cited owing to lack of space or 83. Verde, F., Mata, J. & Nurse, P. Fission yeast cell licenses. Trends Cell Biol. 17, 215–221 (2007). because of my ignorance. morphogenesis: identification of new genes and 97. Dutcher, S. K. Finding treasures in frozen cells: new analysis of their role during the cell cycle. J. Cell Biol. centriole intermediates. Bioessays 29, 630–634 131, 1529–1538 (1995). (2007). DATABASES 84. Brunner, D. & Nurse, P. New concepts in fission yeast 98. Vladar, E. K. & Stearns, T. Molecular characterization UniProtKB: http://beta.uniprot.org/uniprot morphogenesis. Philos. Trans. R. Soc. Lond. B Biol. Sci. of centriole assembly in ciliated epithelial cells. J. Cell CDK1 | Orb6 | PLK4 | Ran | RhoA 355, 873–877 (2000). Biol. 178, 31–42 (2007). 85. Carazo-Salas, R. E. & Nurse, P. Self-organization of 99. Pelletier, L. Centrioles: duplicating precariously. FURTHER INFORMATION interphase microtubule arrays in fission yeast. Nature Curr. Biol. 17, R770–R773 (2007). Eric Karsenti’s homepage: Cell Biol. 8, 1102–1107 (2006). 100. Bornens, M. & Karsenti, E. In Membrane Structure http://www-db.embl.de/jss/EmblGroupsOrg/g_40.html 86. Daga, R. R., Lee, K. G., Bratman, S., Salas-Pino, S. & and Function (ed. Bittar, E. E.) 99–171 (Wiley, New Rayleigh number: http://scienceworld.wolfram.com/ Chang, F. Self-organization of microtubule bundles in York, 1984). physics/RayleighNumber.html anucleate fission yeast cells. Nature Cell Biol. 8, 101. Rodrigues-Martins, A., Riparbelli, M., Callaini, G., All links are active in the online pdf 1108–1113 (2006). Glover, D. M. & Bettencourt-Dias, M. Revisiting the