arXiv:1011.4125v1 [q-bio.PE] 18 Nov 2010 et n h orc ubro ee ssalr closer smaller, is genes commented: of Gould number 21,000[2]). correct to roundworm the tiny and the rect, of many genome as the half elegans in genes, genes esti- those 40,000 the initial as and hand, again an 30,000 between yielding in of counted, genome mate the entire immediately at the be Crick with could lauda- and For, from Watson far by tory. was tone DNA Gould’s of Laboratory, Cavendish elucidation structure the with the earlier in revo- years of biology was 50 molecular nearly occasion the begun the of lution, Although apotheosis the Times. sense York some piece evolu- New Op-Ed noted the extraordinary the in an penned Gould, biologist, Jay tionary Stephen Genome[1], man lg.Temjrt fclua iei irba,adthese and microbial, is life cellular bi- of in majority phenomena The collective ology. of example isolated an means appropriately. in treated collective be strongly to fundamental are need us the and nature, shaped genes: have of that mechanics processes statistical a need hrl fe h ulcto ftedato h Hu- the of draft the of publication the after Shortly h opeiiso h ua eoeaeb no by are genome human the of complexities The we biology, our understand to are we if words, other In eepanda h ee fterapaac,frthey parts for underlying appearance, separate the their alone.”[3] from of predicted level be must the cannot (as jargon) at interactions technical explained these the of be use many to and properties, by emergent generated code of interactions genes, units more and fewer not combinations is complexity we more to that key but the The system causal marks complex ... biology totality, the of call for elaborate direction reductionism to one of codes failure and of basic protein, doctrine from one the flow as for of genes gene collapse many the one ....The as under roundworm again tiny develop half the only bodies of our influence words, directing other In genes... Hm ain osse ewe 000ad40,000 and 30,000 between possesses sapiens “Homo w o nwta h nta siaewsincor- was estimate initial the that know now (we iei hsc:eouina olciepeoeo a fr far phenomenon collective a as evolution physics: is Life 1 eateto hsc,Cne o h hsc fLvn Cell Living of Physics the for Center Physics, of Department efrfrnilntr feouinr dynamics. evolutionary growt of open-ended evol nature the in self-referential synthesis, evolutionary perspective modern useful the a surve of provide genomic might modern concepts by revealed physics non-equilibrium been plethor has in the by evolution problem evidenced shaping as a collective, are as modes evolution dynamical deve medicine, on the to focuses retarded applications review its arguably for has consequences process disastrous evolutionary wides of the lack The of, formulated. and addressed satisfactorily eut h e suso o ail vlto cus n i and occurs, i evolution thus rapidly and how genetics, of population issues of key subset the a result, as treated widely vlto stefnaetlpyia rcs htgvsri gives that process physical fundamental the is Evolution nvriyo liosa raaCapin 10Ws Gree West 1110 Urbana-Champaign, at Illinois of University INTRODUCTION 2 eateto irbooyadIsiuefrGnmcBiolo Genomic for Institute and Microbiology of Department 0 ot odi vne raa L681 USA 61801, IL Urbana, Avenue, Goodwin South 601 ie Goldenfeld Nigel C. 1 n alWoese Carl and n oprtv ees gis antagonists[8]. against defense differentiation cooperative cellular and labor, of division coordinated colonies with and multicellular 7], biofilms—spatially-extended motility)[6, form (swarming surfaces translocate over sensing)[5], collectively are (quorum cells Microbes com- between transfer)[4], municate gene (horizontal interacting. genes exchange strongly to able are also organisms r ilg,hglgtn h oeo olcieeet and effects collective of evolution- role of the survey highlighting biology, selective ary necessarily a providing by life? to molecules bring to deeper work are at or principles of reactions, set chemical rich a complicated biol- reverse-engineering of extremely in study exercise the an Is physical merely ogy purely formation. pattern in as absent the such seem processes extend principles that may revealing self-organization physics, biology of non-equilibrium that Equally of notion frontier the biology. is in fascinating from illuminating arising be behavior collective interactions—can extent perspective—with of physics the principles unifying matter assess its condensed to a timely which is to it in change[24–26], ecology[14–18], phenomena global climate even in and collective microbiology[21–23] also 20], of immunology[19, but recog- importance especially[10–13], growing develop- the evolution the the With of during biology[9]. nition role solved molecular peripheral a it of a ment as selection”, to off “natural relegated catch-phrase written was the itself; under evolution problem of process the molecular ology. bi- in computational advances and biophysics technological single-molecule biology, tremendous outgrowth an the biological is of today the departments being physics research by in biological performed the neglected of relatively most community; physics been have they h ups fti ril st ics hs questions these discuss to is article this of purpose The is justification, less even with but neglected, Also but biology, in abound processes collective Clearly, clg n h lblevrnet This environment. global the and ecology ra prcainfr n understanding and for, appreciation pread fmbl eei lmnswoerl in role whose elements genetic mobile of a toaybooy h ocpulfailings conceptual the biology, utionary fcmlxt,adtequintessentially the and complexity, of h omn fbooya cec,with science, a as biology of lopment sculn oeooyhv o been not have ecology to coupling ts s edsushwcnesdmatter condensed how discuss We ys. et ilgclpeoea e tis it Yet phenomena. biological to se ,adIsiuefrGnmcBiology, Genomic for Institute and s, ssoei rical iie.A a As limited. artificially is scope ts ttsia ehnc,weetekey the where mechanics, statistical t,Ubn,I 10,USA 61801, IL Urbana, St., n 1 , 2 gy, mequilibrium om 2 thereby the possible connections with condensed matter ology in the second half of the twentieth century, and it physics. Thus, no attempt is made here at a complete is now firmly established that cells contain a multitude review of evolution. Instead, we have tried to focus on of molecular components to perform the functions of life. points of principle, where the shortcomings of present- These include such examples as: phospholipid bilayers day understanding seem most acute to us. It would be to bound the cell; cytosol containing water and organic impossible to review the entire literature of evolution or molecules that form the fluid matrix of the cell; nucleic even the sub-set of this literature that is potentially en- acids to record genetic information; proteins comprised gaging for physicists. For a useful and self-contained in- of strings of amino acids that inter alia structure the cell, troduction to the conventional, eukaryote-centric frame- pump molecules in and out of the cell, catalyze metabolic work of evolutionary theory, accessible to physicists with processes, perform mechanical functions, participate in little biological background, we refer the reader to the the cell cycle, transcribe and translate genetic informa- standard text by Maynard Smith[27]; an introduction tion, and allow for signaling between cells; and small that also covers physics-related topics especially has been molecules such as adenosine triphosphate that transport given by Drossel[28]. Most of life is microbial, and a mod- energy within a cell. The study of life must proceed by ern microbe-centric view of evolution can be found in the the study of the machinery within the cell, with addi- book by Sapp[29]. tional consideration given to multi-cellular phenomena We envisage a readership with a wide range of knowl- where applicable. Moreover, any discussion of the ori- edge and interest in evolutionary biology. To the biologist gin of life from early geochemistry should focus on the interested in practical issues, we ask that you do not dis- chemical environments from which life might have arisen, miss the seemingly useless and na¨ıve issues that we neces- and describe the biochemical pathways through which sarily raise. On the one hand, a fundamental understand- metabolic and genetic information could arise sponta- ing of evolution may not seem to offer immediate benefits neously from such a milieu. Indeed, as van Helmont con- in terms of finding the next wonder drug; on the other cluded in 1648, and as is even today the rallying cry at hand, the lack of appreciation for the rapidity and per- conferences on the origin and evolution of life, it seems vasiveness of evolution has, within a lifetime, destroyed quite clear that “all life is chemistry”[33]. the effectiveness of numerous antibiotics[30], and proba- Or is it? bly is responsible for the limited success of the treatment of cancer[31]. The biomedical-industrial complex cannot afford to ignore the need to create a fundamental science Life is Physics of biology. To the physicist, who might be repelled by the seeming lack of structure represented by biology, we ask From a condensed matter physicist’s perspective, we that you look beyond the currently incomplete state of have been led to a rather strange conclusion. To see biological understanding, and be open to the possibility why this is the case, and why it is important to dwell that evolution is both a physical phenomenon and the on this point, let us reflect on the modern way in which natural framework in which biology is embedded. The many readers of Annual Reviews of Condensed Matter lack of structure in the way that biology is traditionally Physics view their field. We will contrast that with presented reflects the field’s unavoidable focus on a single an explicitly reductionist perspective, one which typi- sample path; however, the underlying evolutionary pro- fies most (but certainly not all) reasoning in biology. To cess itself is surely one with deep mathematical structure, begin, suppose that we were to apply this “biological” capable of expression and elucidation in unifying terms perspective to study the phenomenon of superconductiv- based on emergent physical laws. This is a true frontier ity. The first step might be the construction of a cat- of physics, but one that will require a great deal of what alogue of the known superconductors, a list that might has been termed (in another context of non-equilibrium include elements such as Niobium or Tin, the cuprate ox- physics) “open-minded spadework”[32] to unearth. ides such as La2−xSrxCuO4, the alkali-doped fullerenes such as Cs2RbC60, and the heavy-fermion materials such as URu2Si2. From consideration of the structure of the EVOLUTION AS A PROBLEM IN CONDENSED electronic band structure of these materials, the biologist MATTER PHYSICS might try to argue, quite reasonably, that the transfer of electrons between outer atomic orbitals is somehow the Life is Chemistry cause of the interesting transport, thermodynamic and electrodynamic response of these materials. Even though Most living organisms are composed of cells. Setting we would not be able to actually construct a predictive aside, for the moment, the question of what constitutes theory of superconductivity, we might still conclude that “life”, one might attempt to gain further insight into the “all superconductivity is chemistry”. Although we have nature of life by examining carefully the contents of cells. referred to this hypothetical line of reasoning as a “bi- This enterprise culminated in the advent of molecular bi- ological perspective”, it was actually the approach tried 3 originally by physicists, an approach that failed. In fact, lowable process can occur, and therefore it behooves one such a line of reasoning was attempted by A. Einstein to try and identify the universality classes (i.e. cate- in 1922[34] (see also [35] for historical context) prior to gories) of interesting phenomena, and then to try and the development of quantum theory in its mature form. identify the likely realizations of them. This pattern of From the manifest failure of his effort, Einstein concluded discovery is a relatively recent one, arising during the (correctly) that not only was superconductivity of quan- emergence and maturation of condensed matter physics tum mechanical origin, but that it must involve elec- as a scientific discipline during the last 25 years of the trons transported between closed chains of orbitals, an twentieth century, and exemplified, for example, by the uncanny precursor of Feynman’s theory for macroscopic new and important condensed matter sub-field of topo- quantum order in superfluid helium[36]. We learn from logical insulators, where the theoretical understanding all this that reductionism is a natural, intuitive step in led to a concerted and successful search for experimental the construction of theory, and that its failure mode can realizations. This example is by no means an isolated be an instructive pointer to the necessary ingredients of one: other examples of major significance include the a successful theory. Aharonov-Bohm effect, the quantum spin hall effect, lo- The fallacy of “superconductivity is chemistry” is more calization, and the renaissance in atomic, molecular and evident than the fallacy in the logic leading to the con- optical physics provided by the experimental realization clusion that “life is chemistry”. We know now, starting of atomic Bose-Einstein condensates. with the work of Ginzburg and Landau, that supercon- ductivity has, in fact, little to do with the quantum chem- istry of the atoms in the material. Instead, supercon- The Need for a Physics of Living Systems ductivity is best understood as arising from the break- ing of the global U(1) gauge symmetry in the effective The point of this lengthy discourse should now be field theory that describes the interaction between off- clear. It is as limiting to view life as chemistry as it is diagonal long-range order and the electromagnetic field. to view condensed matter collective states as being pri- Any microscopic Hamiltonian whose effective field theory marily about the atoms. The physics of living systems is Ginzburg-Landau theory will give rise to the phenom- would have at its core a generalized description of evo- ena associated with superconductivity. There is noth- lutionary processes, reflecting allowable dynamical sym- ing fundamental about the atoms or molecules. Indeed, metries, not merely static configurational ones. Such a our putative biologist would be astonished to learn of description would, in a very crude sense, be a counterpart color superconductivity in deconfined quark matter[37], to the description of emergent states of matter embod- likely to be realized in the neutron star remnants of core- ied by effective theories such as the Ginzburg-Landau collapse supernovae. Such high temperature supercon- theory alluded to above. A genuine physics of living sys- ductivity would be inconceivable from a narrow perspec- tems would encompass different limits of the evolutionary tive based purely upon atomic chemistry; moreover, the process, each recognized as, and described by different ef- conceptual relationship between the astrophysical and fective theories. condensed matter versions of superconductivity would A biological example of a universality class and the cor- appear obscure. responding effective theory is classical population genet- To summarize, the phenomenon of superconductivity ics, itself sub-divided into two universality classes, one for as a process is captured by the universal, symmetry- sexual organisms and one for asexual organisms. Taken based Ginzburg-Landau theory, but that process can at face value, population genetics contains phenomeno- have many different realizations or instantiations from logical parameters, such as effective population size, fit- matter at a variety of energy and length scales. This level ness and growth rate, and attempts to model generic of understanding is one that crucially informs the modus aspects of populations and their genes. It certainly is operandi of condensed matter physics. Typically, we re- not a microscopic theory, because it lacks a biochemical gard a phenomenon as essentially understood when two level of description, and it is certainly an effective theory, conditions have been met. The first is that the reasons valid only when there is a separation of scales between for the very existence of the phenomenon are known, usu- ecosystem dynamics and gene mutation dynamics. Con- ally from some form of symmetry or topological consid- densed matter physics was liberated from the hegemony eration. The second is that we have a way to determine of Fermi liquid theory by the recognition that there are under what circumstances a particular system (atomic, many other universality classes describing the behavior molecular, nuclear, etc.) represents an instantiation or of electrons in solids, representing different types of col- realization of the phenomenon. Thus, the mantra of con- lective behavior and interactions between (e.g.) spin and densed matter physics, sometimes attributed to Murray charge degrees of freedom. Similarly, we anticipate that Gell-Mann but surely anticipated in spirit by Lev Lan- evolutionary biology can be liberated from the hegemony dau, becomes a recipe for discovery: “That which is not of classical population genetics through the recognition forbidden is mandatory”. By this is meant that any al- that other universality classes must exist, and will be 4 manifest under the appropriate conditions. We will say science, akin to trying to design integrated circuits with- more below about the ways we see that the discoveries of out a fundamental knowledge of quantum electronics and modern biology are positioning it to take the next step semiconductor physics. beyond the paradigm of classical population genetics. Perhaps the primary shortcoming of the biological en- This perspective redefines what we demand of biologi- terprise is the manifest failure to account for the phe- cal understanding in two additional ways. First, the very nomenon of the existence of life. Without doubt, this existence of the phenomenon of life needs to be under- failure reflects not only on biologists, but also on physi- stood. Second, the realization or instantiation of it, on cists. We say this because the majority of biologists Earth, for example, needs to be understood. For the most would probably regard their primary role as being in one part, it is fair to say that the discipline of biology has ne- sense or another to reverse engineer the myriad specific glected the first condition, and in pursuit of the second, realizations of organic life on Earth — the reductionist has confused understanding of the realization with un- exercise that has been notably successful within its own derstanding of the phenomenon. This has had a number terms of reference. On the other hand, the existence of of unfortunate consequences, which arguably have hin- the phenomenon of life, if it can be understood at all in dered both the conceptual development of biology and generic terms, is surely an emergent phenomenon, arising the proper application of foundational understanding to somehow as an inevitable consequence of the laws of non- societal applications. equilibrium statistical physics. How is it that matter self- Let’s begin with the conceptual difficulties. A unified organizes into hierarchies that are capable of generating view of a phenomenon, such as we have alluded to in feedback loops which connect multiple levels of organi- superconductivity, has the benefit that further instances zation and are evolvable? When life emerged from early of it do not come as a surprise, and do not require fur- geochemistry the process must have been driven by irre- ther ad hoc explanation. With a proper understanding versible thermodynamics, but the extension of that pro- of superconductivity as a symmetry-breaking process, for cess into the emergence of evolvable structures remains example, one does not find it surprising to learn of super- mysterious to us. The physical laws that govern far from conductivity in nuclear or astrophysical contexts. With a equilibrium dynamics are still not known. proper understanding of the phenomenon of life as a dy- namical process, for example, one would not find it sur- prising to learn of life in so-called extreme environments Are there new physical laws in Biology? (such as deep beneath the ocean floor) or even on other planets. Amongst the surprises that biology has encoun- In 1949, Delbr¨uck, inspired by Bohr, famously ex- tered recently are several of major significance, including: pressed the view that biology might exhibit phenomena the discovery of horizontal gene transfer in multicellular that are beyond quantum mechanics: eukaryotes[38–42] and the discovery of a fully anoxic mul- “Just as we find features of the atom, its stability, for ticellular life form[43]. In short, a unified view prevents instance, which are not reducible to mechanics, we may the unnecessary multiplication of hypotheses that is the find features of the living cell which are not reducible sure sign of a lack of fundamental understanding (think to atomic physics, but whose appearance stands in a epicycles!). complementary relationship to those of atomic physics.” — M. Delbr¨uck, A Physicist Looks at Biology (1949). The second consequence of a lack of fundamental un- [49] derstanding in biology is the failure to recognize that biology is a manifestation of evolution — not the other Today, few seriously expect that such features or phys- way round. Interventions in biological systems inevitably ical laws will be found. However, in the same essay, in a provoke an evolutionary response which is rapidly emerg- less-celebrated passage, Delbr¨uck drew attention to the ing and spatially-distributed. Examples of the ability problem of spontaneous generation of life. Surprisingly, of biological systems to defeat human attempts at mit- to him the interesting issue was how statistical fluctua- igation include: (1) the world-wide spread of antibi- tions in the kinetics of the emergence of life would lead to otic resistance genes across distantly-related bacteria, a possible lack of determinacy in the biochemistry of liv- crossing species and phylum boundaries and physical ing organisms: the organism might not precisely reflect locations[30]; (2) the rapid evolution of cancer tumors the geochemistry from which it arose. Delbr¨uck was writ- in the face of chemical attack[31, 44, 45]; (3) the abil- ing at a time when phase transitions were only beginning ity of HIV to out-adapt treatment[46, 47]; (4) the ability to be understood, and the notions of emergence and spon- of life to adapt to the massive poisoning of the Precam- taneous symmetry breaking were in their infancy. Thus, brian atmosphere by cyanobacteria-released oxygen 2.4 he probably had no precise notion of a “law” of physics billion years ago (changing from a reducing to an oxidiz- being a description of an effective theory, one that is valid ing atmosphere[48]) with a remarkable subsequent flow- on an intermediate asymptotic scale of length, time or ering of life. With the exception of example (4), these energy, and systematically related to a deeper level of examples indicate a fundamental limitation to medical description. Whereas Delbr¨uck looked to biology to ex- 5 tend quantum mechanics, we look to it as a source of variety of possible niches in the physical environment en- insight into non-equilibrium statistical mechanics of the able the multiplication of species, which then interact to evolutionary process. create new niches. Thus, in this picture, evolution is es- Thus, the study of biology should be more than simply sentially synonymous with population genetics. Genes cataloguing the wonders of biological organization. We are assumed to be the only dynamical variables that are see no reason why the mantra attributed to Gell-Mann tracked, and are associated with a fitness benefit that should not apply with equal force and predictive power is difficult to define or measure precisely, but is quanti- in biology, and become part of its methodology. Today fied by a fitness landscape that describes how the pop- a condensed matter physicist envisages a new class of ulation fitness depends on the genotype[56–59]. Traits collective processes and finds realizations in the world are simply associated with genes, and gene interactions of materials, or for that matter, optical lattices. What are often ignored, or at best handled through the fitness would biology look like as a science if we sought to antic- landscape[59, 60]. ipate the types of evolutionary processes available to the Implicit in this approach is the assumption that the suite of genetic operators now known? What realizations evolutionary time-scale is different from the time-scale of such processes could we find, if we simply looked? of the ecosystem. The crucial question of the timescale Maybe this sounds far-fetched, but in fact this method- of the evolutionary process, even taking at face value the ology has already been used to good effect in biology. perspective of the modern synthesis (which we do not) re- One example we have in mind is the groundbreaking dis- mains a thorny issue[61], and indeed it is fair to say that covery of the class of molecules known as topoisomerases, the theory’s conceptual framework is so poorly quantified whose existence was first recognized theoretically, leading that one cannot confidently make sensible and realistic to their eventual discovery[50, 51]. The topoisomerases estimates of timescales (for an excellent pedagogical dis- are a class of enzymes that are able to perform the mir- cussion of this point, see [62]). acle of passing one strand of DNA through another, by breaking and then reforming them. The end result of this process is that DNA can be uncoiled through a process Epistasis of successive topological changes, allowing transcription and replication to occur. Without the topoisomerases, Even within the framework of the Modern Synthesis, this would be essentially impossible on the timescale rel- one can begin to make model calculations of how evo- evant to cellular processes. lution rates vary with population size and the nature of the fitness landscape[59, 63–66], and probe the role of collective effects—epistasis—between genes. For exam- BEYOND THE MODERN SYNTHESIS ple, phenotypic variance is generally thought to be the result of many-gene interactions[67], as documented in a The classical and widely-accepted framework of evo- tour de force analysis of the yeast metabolic network, for lution is the so-called “Modern Synthesis” or “Neo- example[68]. In yeast, a single-cell eukaryote, any posi- Darwinism”, based on the fusion of Mendelian genetics tive selection that accrues from many interacting genes with Wallace[52] and Darwin’s ideas[53] about “natural with small but positive contributions to fitness must con- selection” (or “survival of the fittest”, the terminology tend with recombination that mixes up and randomizes preferred by Wallace)[27, 54]. This theory and addi- the genotype. This competition can lead to collective tions to it, primarily those due to Kimura[55], account states, more or less by analogy to phase transitions: a for very simple genetic processes, such as point muta- high recombination rate (analogous to high temperature) tion and sexual recombination, leading to random single where genes are weakly correlated and genotypes are nucleotide polymorphisms. A characteristic of these clas- short-lived, and a low recombination rate (analogous to sical theories of evolution is that their genome dynamics low temperature) regime where favorable genotypes are is linear, diffusive in nature, and the population sizes of stable and compete essentially clonally[60]. In bacteria, communities are typically sufficiently large that fixation which reproduce clonally, recombination is a stabilizing times are long. This union of evolution and genetics that factor and can compete in a similar way with point mu- developed in the 1930’s and 1940’s presumed that evo- tations, leading to two phases: one that contains a nar- lution proceeded through the simple mechanism of her- row distribution of genotypes and one that is genomically itable mutation and survival of the fittest. Organisms diverse, with global genome sequence divergence arising have offspring that survive and propagate based on the through the propagation along the genome in evolution- quality of the inevitably mutated genome they inherited ary time of diversification fronts triggered by horizontal from their parents (vertical gene transfer). New positive gene transfer events[69], or perhaps even by indels[70]. traits spread through the population because the individ- How these dynamics play out when the spatial struc- uals with those traits were more successful in surviving ture of microbial populations is included is a fascinating and breeding than other members of the population. A question, bringing together genome dynamics, ecosystem 6 dynamics and population dynamics in a way that has not is now incontrovertible that the inheritance of acquired yet been explored. characteristics, long discredited, but violating no known law of nature, can sometimes occur, not only through horizontal gene transfer (e.g. in microbes), but through Mobile genetic elements so-called epigenetic mechanisms that by-pass the usual modes of inheritance[87, 88] (especially in ciliates[89]). The difficulties in making detailed and quantitative Not only is the Modern Synthesis afflicted by strong theories of the rate of evolution become vastly more interactions, but its very foundation is questionable. The acute as a result of discoveries from the emerging sci- evident tautology embodied by “survival of the fittest” ence of genomics[13]. Building on the seminal work of serves to highlight the backwards-looking character of the Barbara McClintock[71], the past decade or so has wit- fitness landscape: not only is it unmeasurable a priori, nessed the discovery of a plethora of what one might term but it carries with it no means of expressing the growth of “classically-forbidden” processes that radically trans- open-ended complexity[90] and the generation of genetic form our understanding of dynamics at the level of the novelty. Thus, the Modern Synthesis is, at best, a partial genome[72]. Of particular importance is the discovery representation of population genetics, but this on its own of mobile genetic elements in many forms, ranging from is a limited subset of the evolutionary process itself, and transposons to horizontal gene transfer agents[4, 73–78], arguably the least interesting one. whose levels of activity and evolutionary impact have al- most surely been severely underestimated[79]. Horizontal transfer means that genes or other genetic materials are Coupling between evolution and ecology transmitted through a variety of non-hereditary mecha- nisms from one organism to another, and subsequently It is not only the microscopic basis for current evo- expressed, thus altering the behavior (phenotype) of the lutionary theory that has been challenged by recent ad- recipient. Long known to be present in Bacteria and vances in biology. There is now a substantial and grow- Archaea, horizontal gene transfer is now known to be ing literature that documents a surprisingly rapid rate present in multicellular Eukaryotes as well, as a result of of evolution in numerous systems[91–98], ranging from genome-wide surveys published in the last year or so[38– cancer tumors and the immune system to ecosystem- 42, 80]. While the horizontal transfer of genes is widely driven adaptations in all three domains of life. More- recognized to be a major evolutionary force in Archaea over, detailed observations document the coupling be- and Bacteria, it is still too early to be precise about its tween evolutionary and ecological timescales[92, 99–105]. role in Eukaryotes. In a predator-prey system realized in rotifer-algae inter- However, it is not only genes that can be transferred: actions, the rapid evolutionary dynamics is responsible in Eukaryotes, genes are a small fraction of the total for the unusual phase-lag characteristics of the observed genome, with non-coding DNA and transposable ele- population oscillations[106]. ments making up the majority of the genome in some Evolutionary and ecological timescales can also become cases. Transposable elements, sometimes inaccurately coupled if the ecological timescale becomes very long: but colorfully known as “jumping genes”, can easily move an important example of this is provided by the ecol- around and between chromosomes, and through the dis- ogy of the translation apparatus in the cell. The ge- ruption that they potentially inflict upon a genome, cause netic code—the map between triplet codons and amino deleterious mutations and illness. In humans, it is esti- acids is degenerate: there can be several synonomous mated that about 45% of the genome is composed of codons that code for the same amino acid. It is well- transposable elements[1]. Horizontal transfer of trans- established that the synonomous codons are not used posable elements can be a major driver of eukaryotic with equal frequency, and this codon usage bias reflects genome evolution and a source of genetic innovation[81– selection for speed or accuracy of translation of highly- 83]. Indeed, the textbook picture of a static genome expressed genes[107, 108]. The translation process is a composed of genes and junk DNA has now been super- highly-complex one, but relies essentially on the avail- seded by recent findings, and it is arguably more appro- ability of resources, in particular tRNA molecules. Evo- priate to think of the genome as a set of one-dimensional lution of the genome can of course also lead to a co- ecosystems, coupled together by horizontal transfer, and evolution of the abundance of tRNA in the cell, leading containing numerous genetic elements interacting with to a nonlinear dynamics of the genome and its tRNA each other, creating niches for themselves, and evolving abundance distribution[109]. By going beyond the clas- stochastically to create a community ecology[84]. sical mutation-drift-selection framework into the regime These genetic mechanisms permit the spread of genetic of nonlinear evolutionary dynamics, the theory predicts novelty much faster than vertical or hereditary transmis- multi-stability of the genome and an explanation of the sion of genes, essentially amounting to a Lamarckian[85] pattern of observed microbial genome biases, not only in dynamic of evolution[12, 86]. If this was not enough, it translation, but also in transcription and replication. 7

The coupling between environmental and ecological ably to maintain the host as a functioning phage factory, timescales has also been argued to lead to another generic thereby increasing the production of phages during the feature of biology: the prevalence of modularity[110– lysis process as the host cell is destroyed. From analysing 113]. Modularity refers to the relative independence of a the molecular sequences of these genes, and reconstruct- biological component or network—relative, because ev- ing their evolutionary history, Chisholm’s group at MIT erything is connected, of course, but the intramodule have documented the transfer of photosystem II genes connections are more important than intermodule con- back and forth between these marine cyanobacteria and nections. Modularity carries with it the connotation of their phages[120]. Moreover, these genes underwent evo- reuse of motifs, simple building blocks from which com- lution and sequence shuffling while resident in the phages. plex systems can be built[110]. One might think that Thus, rather than supporting the traditional view of such networks could be generically obtained from sim- the relationship between microbes and viruses as being ulations of the evolutionary process, for example using a predator-prey relationship, the new findings suggest genetic algorithms and digital life simulations[114], but that there are collective interactions between microbes remarkably this is not the case[112]. The reason is that and viruses through gene exchange, with the creation typically such simulations emulate the assumed process of an effective global reservoir of genetic diversity that of “natural selection”: networks are evolved by muta- profoundly influences the dynamics of the major marine tions, recombination and other genetic operators, and ecosystems. These findings had been anticipated many only those which perform a defined goal well enough years earlier by numerous investigators[73, 75, 121–123], are permitted to enter the next generation. The key who appreciated and rediscovered the collective outcome to modularity seems to be the coupling to the environ- of horizontal gene transfer. ment, as evidenced by two rather different calculations. During the last few years, an even more astonishing In the first, the network was evolved in an environment of example has come to light, prompted in part by the at- goals that changed in a modular fashion[112]; moreover, tempt to find the cause of colony collapse disorder—the a follow-up study[113] showed clearly that environmen- dramatic reduction in honey bee population (in the US, tal fluctuations do indeed accelerate the rate of evolu- losses of adult workers were 23% during 2006-7, 36% dur- tion. In the second study[115, 116], modularity emerged ing 2007-8)[124]. One of the potential pathogenic causes, spontaneously as an outcome of horizontal gene trans- the Israeli acute paralysis virus (IAPV), was found to fer in the presence of environmental fluctuations. This be able to integrate harmlessly its genome into that of result, admittedly obtained in a rather specific model, the bee host, and thus confer immunity on the host nevertheless highlights the importance of collective inter- to further infection. The surprise is that this virus is actions and the interplay between environmental fluctu- not a retro-virus: it does not need to integrate itself ations and evolution, neglected in the modern synthesis. into the host genome in order to replicate, and so it How does horizontal gene transfer influence the ar- lacks the genetic machinery for reverse transcription of chitecture of actual biological networks? Comparative its RNA into the host DNA. It is not currently known, genomics, coupled with the flux balance analysis of the therefore, how IAPV was able to work its way into metabolic network of E. coli has demonstrated that the the host genome. This is not an isolated example: it network grows by acquiring genes individually and in is now known that a similar process has occurred in groups (typically operons governing coupled reactions), at least 19 vertebrate species, the relevant viruses that which are attached preferentially at the edges of the ex- have conferred immunity being the lethal Bornavirus and isting network[117]. This form of network organization Ebolavirus[125, 126]. It seems that this mechanism is a does not necessarily imply modularity, but it suggests eukaryotic analogue to lysogeny in microbes. These find- one way that modularity can arise. Although there is ings support the notion that there are collective interac- no unique measure of modularity, the available analyses tions between viruses and their hosts. indicate that modularity is an increasing function of the Evolution and ecology couple not only through time, variability of the environment[118] and that modularity but also through space. Wallace was the first to empha- also reflects the number of niches available and is associ- size that speciation is a phenomenon localized in both ated with horizontal gene transfer[119]. space and time[127]: evolution proceeds through a pro- An example of the non-trivial coupling between evo- cess of front propagation in space that couples to pop- lution and ecology has been obtained by recent metage- ulation genetics in ways that are conceptually simple nomic surveys of marine microbial environments, sam- but only now beginning to be understood in a quan- pling and analysing environmental DNA collected from titative way[128–131]. As fronts expand, the pioneer the Sargasso Sea and the Red Sea. Thirty percent organisms at the leading edge experience large demo- of global carbon fixation occurs through the photosyn- graphic fluctuations that are known to play a significant thetic pathways of two cyanobacteria, Prochlorococcus role in temporal oscillations[132] and spatial patterns in and Synechococcus. Remarkably, the phages of these ecosystems[133]. It is important to stress that horizontal organisms also contain photosystem II genes, presum- gene transfer is also strongly influenced by spatial struc- 8 ture. For example, it was recently established that the environment. At the same time, this environment ac- frequency of conjugation events between bacteria is de- tually shapes genomes, through gene transfer processes pendent on the local density, being essentially one/per and phenotype selection. Thus, we encounter a situation generation in closely-packed biofilms, and an order of where the dynamics must be self-referential: the update magnitude smaller in planktonic culture[77]. How the rules change during the time evolution of the system, and interplay between evolutionary dynamics, ecosystem dy- the way in which they change is a function of the state namics and species distribution is reflected in patterns of and thus the history of the system. To a physicist, this species abundance distributions, diversity measures, and sounds strange and mysterious: What is the origin of this community structure is a frontier topic in ecology, and feature that sets biological systems apart from physical relevant to the emerging conceptual framework of niche ones? Aren’t biological systems ultimately physical ones construction[134, 135]. anyway, so why is self-reference an exclusive feature of biological systems (whatever they are!)? The simple answer seems to be that self-reference arises THE DYNAMICS OF EVOLUTION because the biological components of interest are emer- gent, and we are seeking a description of biological phe- Most existing approaches to formulating evolutionary nomena in terms of these biological components only. Ul- dynamics mathematically, as Drossel explicitly points timately, if we used a level of description that was purely out[28], share the limitation that the space in which evo- atomistic, for example, this self-referential aspect of biol- lution takes place is fixed. For example, some models ogy would not arise. This argument does not distinguish consider the dynamics of genomes of fixed length in a biology from condensed matter physics, where many of specified fitness landscape, others consider the interplay the degrees of freedom are emergent also. between agents who are using specified strategies of be- Thus, it is interesting to ask if there are analogues of havior in their repeated encounters with other agents. this phenomenon in condensed matter physics, generi- Such approaches to evolution miss what is to us the cally arising from coarse-grained description of systems central aspect of evolution: it is a process that contin- with order parameter dynamics. To answer this question, ually expands the space in which it operates through a recall, for example, the two types of description that we dynamic that is essentially self-referential. Self-reference have of superconductivity. On one hand there is the BCS should be an integral part of a proper understanding of theory of superconductivity, which works at the level of evolution, but it is rarely considered explicitly. This description of fermions coupled through a pairing interac- point is so important, because it is at the root of why tion (whose microscopic origin need not be specified, but evolution represents a non-trivial extension of the sorts is due to phonons in classic superconductors and perhaps of dynamical processes we encounter in condensed mat- spin or other interactions in high temperature supercon- ter physics. In condensed matter physics, there is a clear ductors). On the other hand, there is the coarse-grained separation between the rules that govern the time evo- order parameter description due to Ginzburg and Lan- lution of the system and the state of the system itself. dau. Frequently, physicists use the latter as a convenient For example, in studying fluid dynamics, a firm basis for model that is easy to calculate with, in systems of ar- theory is provided by the Navier-Stokes equations, and bitrary geometry or with spatial variation. Moreover, regardless of whether the flow is at low Reynolds num- this level of description is frequently used to study the ber, dominated by viscous effects, or at high Reynolds time-dependent phenomena in superconductors, given by number and dominated by inertial effects, the underly- the time-dependent Ginzburg-Landau equations. Near ing equations are capable of capturing all the phenomena. the superconducting critical point, this description is a The mathematical reason for this is that the governing generic consequence of critical dynamics. However, away equation does not depend on the solution of the equation. from the critical region this equation cannot be systemat- The evolution operator is independent of the state of the ically derived. Similarly, in superfluids and Bose-Einstein system. In biology, however, the situation is different. condensates, the zero temperature dynamics can be well- The rules that govern the time evolution of the system described by the Gross-Pitaevskii equations, and near the are encoded in abstractions, the most obvious of which critical point, a more complicated dynamic universality is the genome itself. As the system evolves in time, the class is believed to be appropriate. Away from these two genome itself can be altered, and so the governing rules regimes, however, there is no universally-agreed upon de- are themselves changed. From a computer science per- scription that is systematically accurate. The reason that spective, one might say that the physical world can be there is confusion surrounding the dynamics at interme- thought of as being modeled by two distinct components: diate temperatures has in fact been well-understood, but the program and the data. But in the biological world, not widely appreciated, for many years: the assumption the program is the data, and vice versa. For example, that there exists a description local in both space and the genome encodes the information which governs the time is false. This can be seen from the derivation of response of an organism to its physical and biological the Ginzburg-Landau description from the more micro- 9 scopic BCS theory, which in general involves a memory ogy are the open-ended growth of complexity that we see function that becomes local in space and time only in in the biosphere, the large population fluctuations, and special limits. The breakdown of locality that accom- the widespread occurrence of the “law of unintended con- panies effective descriptions of dynamical phenomena is sequences” when trying to manipulate ecosystems[137]. well-known beyond superconductivity, of course, and is Thus, although complexity is hard to define precisely and a feature of several approaches to non-equilibrium sta- usefully, we regard the defining characteristic of complex- tistical mechanics, including mode-coupling theory and ity as the breakdown of causality[138]. Simply put, com- renormalization group approaches to effective actions in plex systems are ones for which observed effects do not field theory. have uniquely definable causes, due to the huge nature Why self-reference is a specific feature of biological of the phase space and the multiplicity of paths. systems and not physical systems should now be evi- Ecosystems are never static but are continually chang- dent: self-referential dynamics is an inherent and prob- ing and adapting, and their response involves all levels ably defining feature of evolutionary dynamics and thus down to the genome and even smaller (because viruses biological systems. Thus, the question really is how self- are an integral part of an ecosystem). In an attempt to referential dynamics arises as a universality class from capture such complex dynamics, researchers have made the basic laws of microscopic physics, as an expression extensive use of digital life simulations. of nonequilibrium physics. Although there is a recogni- Digital life simulations[28, 114, 139–144] use inter- tion of this sort of question in some of the literature on acting synthetic organisms, with predefined rules for philosophy, artificial life and the evolution of language, replication, evolution and interaction with each other nothing approaching a serious calculation has been done and their digital environment. Experiments on digital to our knowledge. organisms are an accurate and informative methodol- The fact that evolution is a process that transcends its ogy for understanding the process of evolution because realization means that evolution is able to act on its own the entire phylogenetic history of a population can be basic mechanisms. This nonlinearity of the evolutionary tracked, something that is much more difficult—but not process is sometimes referred to as evolvability. It has im- impossible[145]—to do with natural organisms[146]. Ex- portant generic ramifications that have been explored in periments on digital organisms can be performed over a model calculation of protein evolution[115], in particu- time scales relevant for evolution, and can capture uni- lar that evolvability is preferentially selected during peri- versal aspects of evolutionary processes, including those ods of increased rate of environmental change. Whether relevant to long-term adaptation [147, 148], ecological or not evolvability can be selected for, in the conven- specialization[149, 150] and the evolution of complex tional parlance, is a controversial topic, partly because of traits[151]. Despite this progress, the way in which evo- the unfashionable nonlinearity of the process, and partly lution leads to ever increasing complexity of organisms because evolvability seems to undermine the robustness remains poorly understood and difficult to capture in of biological organization. Presumably, a detailed quan- simulations and models to date. Is this because these titative understanding of evolution would flesh out the calculations are not sufficiently realistic, extensive, or de- balance between evolvability and robustness[136]. tailed, or has something fundamental been left out? As early as 1971, Woese speculated on the emergence of genetic organization[152]. He was concerned with the Evolution and Complexity evolution of complexity, but approached from a molec- ular viewpoint, namely that of the origin of quaternary Complex systems are characterized by the presence structure in proteins. Woese portrayed evolution as a of strong fluctuations, unpredictable and nonlinear dy- cyclic process, in which gene products evolved first to a namics, multiple scales of space and time, and frequently dimerized state, followed by gene duplication, and sep- some form of emergent structure. The individual compo- arate evolution of the gene products so that the dimer nents of complex systems are so tightly coupled that they finally consists of related but not identical sub-units. If cannot usefully be analyzed in isolation, rendering irrel- the two genes that code for the two distinct halves of the evant traditional reductionist approaches to science, ob- dimer subsequently fuse, a new composite molecule has scuring causal relationships, and distinguishing complex- arisen that he called a “co-dimer”. In this co-dimer, the ity from mere complication. Biological complexity, is an two previously-related components can evolve separately extreme example of complexity, and arises from the inclu- but in a complementary fashion, as long as the biochemi- sion of active components, nested feedback loops, compo- cal properties of the co-dimer as a whole are not adversely nent multifunctionality, and multiple layers of system dy- affected. In this way, one half of the co-dimer could, for namics, and is relevant to numerous aspects of the biolog- example, evolve into a control site for the enzyme that is ical, medical and earth sciences, including the dynamics the other member of the co-dimer. The point of this ar- of ecosystems, societal interactions, and the functioning gument is that it provides a molecular realization of the of organisms. Perhaps the most striking features of biol- process of how evolution can cross wide fitness valleys[66]. 10

He argued that the dynamics of co-dimerization, repeated chy, an inevitable reflection of the fact that the dynamic ad infinitum, led to the growth of complexity of all macro- we are seeking is inherently self-referential. The way molecular components of the cell, including proteins and that the conundrum can be resolved is to begin with the translational machinery itself. If codimerization were an infinite-dimensional dynamical system which sponta- an important part of the evolutionary process, it would neously undergoes a sequence of symmetry-breaking or perforce entail there being a high abundance of homod- bifurcation events into lower-dimensional systems. Such imeric proteins in the cell, and moreover that the num- transitions can be thought of as abstraction events: suc- ber of subunits would be an even number. At the time cessive lower dimensional systems contain a representa- Woese’s paper appeared, the paucity of available data tion in them of upper levels of the original system. A did not permit a test of these predictions. Early pro- precise mathematical prototype of such a construction tein structures were limited to small molecules because can be constructed from consideration of the dynamics of the difficulty of crystallizing larger molecules, so that of functions on a closed one-dimensional interval[160], ad- when the Protein Data Bank was surveyed in 1995, 66% mittedly with very little direct biological interpretation, were found to be monomeric[153]. By 2000, the situa- but with the positive outcome of generating a hierarchi- tion had dramatically changed, with 19% of the proteins cally entangled dynamical network. Such networks are surveyed being monomeric. Of the remainder, 59% were not simple tree structures, and this means that the nodes either dimers or tetramers, with a clear preference for and links of the network drive each other in a way that even numbers of sub-units[154, 155]. More recent anal- a biologist would interpret as co-evolutionary. Another yses confirm these findings and are even able to probe way to interpret such dynamical systems is that the el- the evolutionary dynamics that has led to the observed ements are multifunctional: their input-output map de- structure of protein complexes[156]. Moreover, not only pends on the state of the system, rather than being a con- has the detailed structural evidence consistent with the stant in time. Such systems can have no static fixed point co-dimerization model been fully elaborated, but also it to their dynamics: in the language of earlier work[90], appears that the assembly of proteins follows the evolu- they must exhibit a complexity cascade, just as Woese tionary development of their sub-unit structure[157]. had earlier argued in the context of co-dimerization. Co-dimerization and gene duplication are examples of how biological systems exploit redundancy as one of the prime mechanisms for evolution. This insight re-emerges Coevolution and game theory in recent simple models of evolutionary dynamics, which show that open-ended complexity is only possible as An alternative to treating organisms as evolv- an outcome of complexity-scale-invariant genomic oper- ing in a fixed environment or fitness landscape is ators, such as gene duplication[90]. That is, if there are coevolution[161]. In coevolution, organisms interact and genetic operators that bias organisms to have a specific their interactions drive each to evolve, leading to a con- complexity, then the complexity of the system will not tinuing process of phenotype evolution, although not nec- increase without bound. This invariance is similar in essarily an increase in complexity, as occurs in the com- spirit to that which lies at the heart of the Richardson plexity cascade[90]. Examples include mutualism and an- cascade in turbulence[158, 159]. Guttenberg and Gold- tagonistic coevolution. In mutualism, the interaction is enfeld showed in an explicit model of digital life[90] how basically symbiotic, and frequently occurs between plants different genetic operations behaved with regard to this and animals[162], microbes and plants[163] and, gaining invariance criterion, and thus were able to devise ecosys- increasing attention in the last few years, humans and tem models that evolved open-ended complexity. Despite their microbiomes[164]). Antagonistic coevolution de- its popularity, a static “fitness landscape” [56–58] picture scribes how opposing organisms, such as predators and of evolution does not satisfy the proposed invariance cri- prey, develop an arms race[165, 166] as a result of their terion, and is indeed conceptually insufficient to account competition, an effect generally referred to as the Red for the open-ended growth of complexity. Thus, digital Queen effect[165]. life simulations do not generally evolve qualitatively new There is a growing empirical literature[167–173] on the responses or modes of behaviour; they cannot “think out- way in which antagonistic coevolution can accelerate evo- side the box”. lution and dominate a systems’s response to changing en- The emphasis on redundancy as a motif suggests that vironmental conditions. However, this topic has received a component of evolution is multifunctionalism. To see much less theoretical attention to date. To investigate why, consider how a system is modeled, perhaps as a set the interaction of genomic evolution and population dy- of differential equations or lattice update rules. These namics, we require that the matrix describing the inter- rules themselves need to evolve: but how? We need actions between predator and prey evolve with genomic an additional set of rules describing the evolution of the fitness[174, 175]. In these approaches, the interaction ma- original rules. But this upper level of rules itself needs trix is a linear function of either the fitness of the preda- to evolve. And so we end up with an infinite hierar- tors or the relative fitness of a pair, which is appropriate 11 to describe the stabilization of complex foodwebs with a a priori, and is not able to evolve along with the sys- large number of species, but probably not adequate to tem itself. Thus, the effect of each agents’ actions on describe systems with a Red Queen dynamics, where the changing the behavior of other players and the dynam- genomic fitness is a distribution for a species instead of ics of the environment is neglected. A bona fide game a single number. In this case, the evolution of the inter- theory approach to evolution would allow the game rules action matrix is nonlinear, and the dynamics needs to be themselves to change as a function of the state of the treated as a stochastic individual-level model[176]. players and their intrinsic dynamics. This is important Coevolutionary dynamics is also an important arena for the following reason: in the usual formulation of game for game theory dynamics to play out in evolution and theory[178], there is an equilibrium state that can arise ecology, because the level of cooperation between organ- known as an evolutionary stable strategy (ESS). Loosely isms can be analyzed on an encounter-by-encounter basis speaking, this can arise when a population of individuals as the repetition of a cooperative game, such as the Pris- play in a cooperative game in which the payoff represents oner’s Dilemma[177–181]. In the Prisoner’s Dilemma, the the fitness, and selection is assumed to be operating. An two players have two states: they can either “cooperate” ESS is a strategy that if used by a large enough number or “defect”. If they cooperate, they receive a reward R, if of individuals in a population is resistant to invasion by they both defect, they receive a punishment P , and if one alternative strategies, and as such represents some sort defects while the other cooperates, the former receives a of equilibrium (in fact a Nash equilibrium[186]). This temptation T while the latter receives the sucker’s re- equilibrium state is in some sense analogous to thermal ward S. If T>R>P>S, then the following dilemma equilibrium and reflects the static nature of the game it- arises: a rational player would defect, because it yields self. If the game was allowed instead to be dynamic, with the highest reward independent of the state of the other the rules of the game able to change due to the states of player. So in a contest with two rational players, each the players, then in addition to static equilibria there will end up with the punishment P , which is a shame: could also be dynamic equilibria, analogous to a non- because if they had both cooperated, they would have equilibrium steady state. Such a game could describe received the reward R. This game illustrates the para- a steadily evolving system, one with a stationary com- doxical nature of two-body interactions in cooperative plexity cascade (itself is analogous to a turbulent non- dynamical systems, but can sometimes provide an accu- equilibrium state). rate idealization of actual biological interactions, if the Akiyama and Kaneko[187, 188] seem to be the only biology can be meaningfully mapped into simple game researchers who have explicitly argued along these lines, theory interactions terms. making a first step towards a theory for truly-evolving In a remarkable experiment, Turner and Chao[182] games using dynamical systems theory to analyze a co- studied the evolution of fitness (measured in terms of operative game with an evolving pay-off matrix. Other population growth) of phages that can multiply infect game theoretic models that include some sort of dynami- the same bacterial cell. Viruses can cooperate by shar- cal or learning behavior, such as players using their scores ing intracellular enzymes needed for reproduction and to adjust the frequency of a finite set of a priori given re- can defect by sequestering the enzymes. Turner and sponses, have been developed, describing the system tra- Chao engineered two strains to behave as cooperators jectories, showing that chaotic dynamics can arise, and and defectors, and found that the strain with high rate shedding doubt about the applicability of Nash equilib- of co-infection initially increased in fitness over time, but ria in the real world of dynamical games[189]—and by eventually evolved lower fitness, a counter-example to the extension to evolution. usual cavalier assumption that “fitness always increases”. A related development has been the dynamics of In this case, Turner and Chao were able to show that spatially-extended game theoretic models that study the this decline in fitness arose from collective effects: the evolution of cooperation, in which there is coevolution of fitness of the virus strains conformed to the Prisoner’s the network of connections between players[190, 191] and Dilemma, whose pay-off matrix they were able to mea- the players’ strategies as the game progresses. Typically sure. Game theory is not the only way to interpret this players update their strategy based upon their interac- finding[183], but the key point is that collective effects tions with their network neighbours during the game, provide important and sometimes counter-intuitive in- representing perhaps the simplest game theory models fluences in ecological and evolutionary interactions. The where there is feedback between the environment and huge literature on this area is beyond the scope of this the agents[192]. Finally, there has been progress in the review, but other ecological interactions are measurable most difficult aspect of the relationship between game and interpretable in game theory terms[184, 185]. theory and evolutionary biology: understanding how ef- We have emphasized in this review that the essence fective game theoretic description can arise from the of evolution is self-reference, but it is apparent that this macroscopic dynamics of agents interacting with their is not captured by the game theory models described environment[193]. We conclude this section by recalling above. The problem is that the pay-off matrix is given that evolutionary dynamics is more general than biology 12 itself, and thus it should come as no surprise that there CONCLUSION are applications of evolutionary game theory to evolu- tionary finance[194, 195], and indeed meaningful analo- In the natural development of the sciences, issues of gies between ecosystem dynamics and finance[196]. complexity are sensibly postponed until they can no longer be avoided. Physics was able to delay serious con- sideration of collective effects for nearly three hundred years, and only in the last thirty years or so has it con- fronted complex collective phenomena involving multiple scales of space and time, unpredictable dynamics and IS EVOLUTION RANDOM? large fluctuations. Its track record of success is mixed. Biology was not so lucky: at its outset, complex phe- nomena were encountered, but tools were lacking to cope We would be remiss in ending this article if we did with the difficulty. Rather than abiding by ignorance, not briefly mention the fascinating question: is evolu- a language-culture was developed to explain away the tion random? More precisely, does variation precede conceptual difficulties using guesswork solutions such as but not cause adaptation—the central tenet of the mod- “natural selection”. As Schr¨odinger wrote, ern synthesis—or do environmental changes alter the stochastic nature of the evolutionary process? Any in- “Instead of filling a gap by guesswork, genuine sci- ence prefers to put up with it; and this, not so much dication that organisms can chose which mutations arise from conscientious scruples about telling lies, as from after an environmental stress has been applied would be the consideration that, however irksome the gap may anathema to the central tenet of the modern synthesis, be, its obliteration by a fake removes the urge to seek and would require a re-evaluation of how evolution is after a tenable answer.”—E. Schr¨odinger, Nature and widely understood. the Greeks, pp7-8.[202] Today, with the “urge” removed, the development In a classic experiment involving the exposure of of sophisticated technology has allowed biology to take a strain of E. coli to bacteriophages, Luria and refuge in single-molecule biophysics, genomics and molec- Delbr¨uck[197] showed that the probability distribution ular biology. But the stultifying language-culture still for the number of mutants exhibited the characteristics remains. This sanctuary is an illusionary respite: the expected only if random mutations had been present be- core problems of biology remain irksome to some, and fore exposure to the phage, and apparently ruling out are inextricably interwoven with evolution. Indeed, the the hypothesis that mutations occurred as a result of very existence of biological phenomena is an expression the phage. This might seem to put the matter to rest, of physical laws that represent a new asymptotic realm but because there is a priori no theoretical reason why in nonequilibrium statistical physics. Ulam famously a cell could not sense environmental stress and respond quipped[203] “Ask not what physics can do for biology; in a non-random way, researchers have persisted in ex- ask what biology can do for physics.” Our answer is clear. ploring this issue experimentally. Although early exper- iments are generally recognized as not being properly analyzed[198], a plethora of mechanisms have now been reported to give rise to an adaptive response to stress, ACKNOWLEDGEMENTS including regulation of mutation rates (non-random in time) and localized variation along the genome (non- We thank the following for their helpful comments on random in genome space)[88, 91, 199]. earlier versions of this manuscript: Elbert Branscombe, There is also compelling evidence that not only may Lynn Caporale, Jim Davis, Michael Deem, Barbara mutations be non-random but horizontal gene transfer Drossel, Freeman Dyson, George Dyson, Larry Gold, Leo too need not be random. Enterococcus faecalis, a gut- Kadanoff, Alan McKane, Tim Newman, Jim Shapiro, dwelling bacterium, can be resistant to certain antibiotics and especially James Langer and Kalin Vetsigian. This if it contains the plasmid (an extrachromosmal loop of work was partially supported by the US National Science DNA) pCF10. This plasmid can be horizontally trans- Foundation under Grants NSF-EF-0526747 and NSF- ferred from a donor with the plasmid to a recipient ini- PFC-082265, through the Center for the Physics of Living tially without it, through the process of conjugation (bac- Cells. terial sex). The remarkable feature of this organism, how- ever, is that the transfer is controlled by and initiated by signals sent from the recipient[200]. The vancomycin re- sistant strain V583 of this organism is now one of the [1] E. Lander, L. Linton, B. Birren, C. Nusbaum, leading causes of hospital-acquired infection, spreading M. Zody, J. Baldwin, K. Devon, K. Dewar, M. Doyle, rapidly through horizontal gene transfer[201]. W. FitzHugh, et al., “Initial sequencing and analysis of 13

the human genome,” Nature, vol. 409, no. 6822, pp. 860– tive self-organization of microorganisms,” Advances in 921, 2001. Physics, vol. 49, no. 4, pp. 395–554, 2000. [2] M. Clamp, B. Fry, M. Kamal, X. Xie, J. Cuff, M. F. Lin, [22] N. Wingreen and S. Levin, “Cooperation among mi- M. Kellis, K. Lindblad-Toh, and E. S. Lander, “Distin- croorganisms,” PLoS Biology, vol. 4, no. 9, pp. 1486– guishing protein-coding and noncoding genes in the hu- 1488, 2006. man genome,” Proceedings of the National Academy of [23] G. Velicer and M. Vos, “Sociobiology of the myxobac- Sciences, vol. 104, no. 49, pp. 19428–19433, 2007. teria,” Annual review of microbiology, vol. 63, pp. 599– [3] S. J. Gould, “Humbled by the genome’s mysteries.” New 623, 2009. York Times, Feb 21 2001. [24] T. Lenton, H. Held, E. Kriegler, J. Hall, W. Lucht, [4] H. Ochman, J. Lawrence, E. Groisman, et al., “Lateral S. Rahmstorf, and H. Schellnhuber, “Tipping elements gene transfer and the nature of bacterial innovation,” in the Earth’s climate system,” Proceedings of the Na- Nature, vol. 405, no. 6784, pp. 299–304, 2000. tional Academy of Sciences, vol. 105, no. 6, pp. 1786– [5] C. Waters and B. Bassler, “Quorum Sensing: Cell-to- 1793, 2008. Cell Communication in Bacteria,” Annu. Rev. Cell Dev. [25] R. Bardgett, C. Freeman, and N. Ostle, “Microbial con- Biol, vol. 21, pp. 319–46, 2005. tributions to climate change through carbon cycle feed- [6] R. Harshey, “Bacterial Motility On A Surface: Many backs,” The ISME Journal, vol. 2, no. 8, pp. 805–814, Ways To A Common Goal,” Annual Reviews in Micro- 2008. biology, vol. 57, no. 1, pp. 249–273, 2003. [26] K. Cuddington, W. Wilson, and A. Hastings, “Ecosys- [7] M. Copeland and D. Weibel, “Bacterial swarming: a tem Engineers: Feedback and Population Dynamics,” model system for studying dynamic self-assembly,” Soft The American Naturalist, vol. 173, no. 4, pp. 488–498, Matter, vol. 5, no. 6, pp. 1174–1187, 2009. 2009. [8] J. Shapiro, “Thinking about bacterial populations as [27] J. Maynard Smith, “The theory of evolution,” CUP, multicellular organisms,” Annual Reviews in Microbiol- Cambridge, 1993. ogy, vol. 52, no. 1, pp. 81–104, 1998. [28] B. Drossel, “Biological evolution and statistical [9] C. Woese and N. Goldenfeld, “How the microbial world physics,” Advances in Physics, vol. 50, no. 2, pp. 209– saved evolution from the scylla of molecular biology and 295, 2001. the charybdis of the modern synthesis,” Microbiology [29] J. Sapp, The new foundations of evolution: on the tree and Molecular Biology Reviews, vol. 73, no. 1, pp. 14– of life. Oxford University Press, USA, 2009. 21, 2009. [30] A. Salyers and C. Amabile-Cuevas, “Why are antibiotic [10] A. Herz, “Collective phenomena in spatially extended resistance genes so resistant to elimination?,” Antimi- evolutionary games,” Journal of theoretical biology, crobial agents and chemotherapy, vol. 41, pp. 2321–2325, vol. 169, no. 1, pp. 65–87, 1994. 1997. [11] A. Nowak Martin, Evolutionary Dynamics: Exploring [31] R. Gatenby, “A change of strategy in the war on can- the Equations of Life. Cambridge University Press, cer,” Nature, vol. 459, no. 7246, pp. 508–509, 2009. 2006. [32] J. Langer, “Instabilities and pattern formation in crys- [12] N. Goldenfeld and C. Woese, “Biology’s next revolu- tal growth,” Reviews of Modern Physics, vol. 52, no. 1, tion,” Nature, vol. 445, p. 369, 2007. pp. 1–28, 1980. [13] J. Shapiro, “Mobile DNA and evolution in the 21st cen- [33] J. B. V. Helmont, Ortus Medicinae. Amsterdam, 1648. tury.,” Mobile DNA, vol. 1, no. 4, pp. 1–14, 2010. English translation in Great Experiments in Biology, p. [14] S. Levin, “The problem of pattern and scale in ecol- 155, Gabriel, M.L. and Fogel, S. (eds.) (Prentice Hall, ogy: the Robert H. MacArthur award lecture,” Ecology, NJ, 1955). vol. 73, no. 6, pp. 1943–1967, 1992. [34] A. Einstein, Theoretische Bemerkungen zur [15] J. Toner, Y. Tu, and S. Ramaswamy, “Hydrodynamics Supraleitung der Metalle. Eduardo Ijdo, Leiden, 1922. and phases of flocks,” Annals of Physics, vol. 318, no. 1, [English translation by Bjoern Schmekel as Theoretical pp. 170–244, 2005. remark on the superconductivity of metals, available [16] D. Sumpter, “The principles of collective animal be- online at http://arxiv.org/abs/physics/0510251]. haviour,” Philosophical Transactions of the Royal Soci- [35] T. Sauer, “Einstein and the Early Theory of Supercon- ety B: Biological Sciences, vol. 361, no. 1465, pp. 5–22, ductivity, 1919–1922,” Archive for history of exact sci- 2006. ences, vol. 61, no. 2, pp. 159–211, 2007. [17] F. Schweitzer and J. Farmer, Brownian Agents and Ac- [36] R. Feynman, “Atomic theory of the λ transition in He- tive Particles: Collective Dynamics in the Natural and lium,” Physical Review, vol. 91, no. 6, pp. 1291–1301, Social Sciences. Springer Verlag, 2007. 1953. [18] J. Haloin and S. Strauss, “Interplay between Ecological [37] M. Alford, A. Schmitt, K. Rajagopal, and T. Sch Communities and Evolution,” Annals of the New York ”afer, “Color superconductivity in dense quark matter,” Academy of Sciences, vol. 1133, no. The Year in Evolu- Reviews of Modern Physics, vol. 80, no. 4, pp. 1455– tionary Biology 2008, pp. 87–125, 2008. 1515, 2008. [19] I. Cohen and D. Harel, “Explaining a complex living sys- [38] E. Gladyshev, M. Meselson, and I. Arkhipova, “Massive tem: dynamics, multi-scaling and emergence,” Journal horizontal gene transfer in bdelloid rotifers,” Science, of the Royal Society Interface, vol. 4, no. 13, pp. 175– vol. 320, no. 5880, p. 1210, 2008. 182, 2007. [39] P. Keeling and J. Palmer, “Horizontal gene transfer in [20] I. Cohen, “Real and artificial immune systems: comput- eukaryotic evolution,” Nature Reviews Genetics, vol. 9, ing the state of the body,” Nature Reviews Immunology, no. 8, pp. 605–618, 2008. vol. 7, no. 7, pp. 569–574, 2007. [40] B. Palenik, Q. Ren, V. Tai, and I. Paulsen, “Coastal [21] E. Ben-Jacob, I. Cohen, and H. Levine, “Coopera- Synechococcus metagenome reveals major roles for hor- 14

izontal gene transfer and plasmids in population di- [59] S. Park, D. Simon, and J. Krug, “The speed of evolu- versity,” Environmental Microbiology, vol. 11, no. 2, tion in large asexual populations,” Journal of Statistical pp. 349–359, 2009. Physics, vol. 138, no. 1, pp. 381–410, 2010. [41] A. Monier, A. Pagarete, C. de Vargas, M. Allen, [60] R. Neher and B. Shraiman, “Competition between re- B. Read, J. Claverie, and H. Ogata, “Horizontal gene combination and epistasis can cause a transition from transfer of an entire metabolic pathway between a eu- allele to genotype selection,” Proceedings of the National karyotic alga and its DNA virus,” Genome Research, Academy of Sciences, vol. 106, no. 16, pp. 6866–6871, vol. 19, no. 8, p. 1441, 2009. 2009. [42] J. Pace, C. Gilbert, M. Clark, and C. Feschotte, “Re- [61] P. Gingerich, “Rates of Evolution,” Annual Review of peated horizontal transfer of a DNA transposon in mam- Ecology, Evolution, and Systematics, vol. 40, pp. 657– mals and other tetrapods,” Proceedings of the National 675, 2009. Academy of Sciences, vol. 105, no. 44, p. 17023, 2008. [62] D. Fisher, “Course 11: Evolutionary dynamics,” in Les [43] R. Danovaro, A. Dell’Anno, A. Pusceddu, C. Gambi, Houches Summer School Proceedings, vol. 85, pp. 395– I. Heiner, and R. Mobjerg Kristensen, “The first meta- 446, 2007. zoa living in permanently anoxic conditions,” BMC Bi- [63] M. Desai, D. Fisher, and A. Murray, “The speed of evo- ology, vol. 8, no. 1, p. 30, 2010. lution and maintenance of variation in asexual popu- [44] M. Kimmel, “Evolution and cancer: a mathematical bi- lations,” Current Biology, vol. 17, no. 5, pp. 385–394, ology approach,” Biology Direct, vol. 5, no. 1, p. 29, 2007. 2010. [64] E. Brunet, I. Rouzine, and C. Wilke, “The stochastic [45] C. Attolini and F. Michor, “Evolutionary theory of can- edge in adaptive evolution.,” Genetics, vol. 179, no. 1, cer,” Annals of the New York Academy of Sciences, pp. 603–620, 2008. vol. 1168, no. 1, pp. 23–51, 2009. [65] C. Fogle, J. Nagle, and M. Desai, “Clonal Interference, [46] S. Duffy, L. Shackelton, and E. Holmes, “Rates of evolu- Multiple Mutations and Adaptation in Large Asexual tionary change in viruses: patterns and determinants,” Populations,” Genetics, vol. 180, no. 4, pp. 2163–2173, Nature Reviews Genetics, vol. 9, no. 4, pp. 267–276, 2008. 2008. [66] D. Weissman, M. Desai, D. Fisher, and M. Feldman, [47] R. Neher and T. Leitner, “Recombination rate and se- “The rate at which asexual populations cross fitness lection strength in HIV intra-patient evolution,” PLoS valleys,” Theoretical population biology, vol. 75, no. 4, Computational Biology, vol. 6, no. 1, pp. e10006600/1– pp. 286–300, 2009. 7, 2010. [67] J. Moore, “A global view of epistasis,” Nature genetics, [48] P. G. Falkowski and L. V. Godfrey, “Electrons, life and vol. 37, no. 1, pp. 13–14, 2005. the evolution of Earth’s oxygen cycle,” Philosophical [68] D. Segre, A. DeLuna, G. Church, and R. Kishony, Transactions of the Royal Society B: Biological Sciences, “Modular epistasis in yeast metabolism,” Nature genet- vol. 363, no. 1504, pp. 2705–2716, 2008. ics, vol. 37, no. 1, pp. 77–83, 2005. [49] M. Delbr¨uck, “A physicist looks at biology,” Trans. [69] K. Vetsigian and N. Goldenfeld, “Global divergence of Conn. Acad. Arts and Sciences., vol. 38, pp. 173–190, microbial genome sequences mediated by propagating 1949. fronts,” Proceedings of the National Academy of Sci- [50] J. Wang, “Moving one DNA double helix through an- ences, vol. 102, no. 20, pp. 7332–7337, 2005. other by a type II DNA topoisomerase: the story of [70] J. Chen, Y. Wu, H. Yang, J. Bergelson, and M. Kreit- a simple molecular machine,” Quarterly reviews of bio- man, “Variation in the Ratio of Nucleotide Substitution physics, vol. 31, no. 02, pp. 107–144, 1998. and Indel Rates across Genomes in Mammals and Bac- [51] J. Wang, “A Journey in the World of DNA Rings and teria,” Molecular Biology and Evolution, vol. 26, no. 7, Beyond,” Annual review of biochemistry, vol. 78, pp. 31– pp. 1523–1531, 2009. 54, 2009. See p.40,second column. [71] B. McClintock, “The Significance of Responses of the [52] A. Wallace, “On the tendency of varieties to depart in- Genome to Challenge,” Science, vol. 226, pp. 792–801, definitely from the original type,” Journal of the Lin- 1984. nean Society of London, Zoology, vol. 3, pp. 53–62, 1858. [72] J. Shapiro, “Revisiting the central dogma in the 21st [53] C. Darwin, On the origin of species by means of natural Century,” Annals of the New York Academy of Sciences, selection, or the preservation of favoured races in the vol. 1178, pp. 6–28, 2009. struggle for life (1859). J. Murray, London, UK, 1859. [73] M. Syvanen, “Horizontal gene transfer: evidence and [54] J. Huxley, Evolution: the modern synthesis. MIT Press, possible consequences,” Annual Review of Genetics, 1942. vol. 28, no. 1, pp. 237–261, 1994. [55] M. Kimura, The neutral theory of molecular evolution. [74] J. Filee, P. Forterre, and J. Laurent, “The role played Cambridge University Press, 1985. by viruses in the evolution of their hosts: a view based [56] S. Wright, “The roles of mutation, inbreeding, cross- on informational protein phylogenies,” Res. Microbiol, breeding and selection in evolution,” Proceedings of the vol. 154, pp. 237–243, 2003. Sixth International Congress on Genetics, vol. 1, no. 6, [75] M. Weinbauer and F. Rassoulzadegan, “Are viruses pp. 356–366, 1932. driving microbial diversification and diversity?,” Envi- [57] S. Gavrilets, Fitness landscapes and the origin of ronmental Microbiology, vol. 6, no. 1, pp. 1–11, 2004. species. Princeton University Press, Princeton, US, [76] L. Frost, R. Leplae, A. Summers, and A. Toussaint, 2004. “Mobile genetic elements: the agents of open source [58] H. Orr, “The genetic theory of adaptation: a brief his- evolution,” Nature Reviews Microbiology, vol. 3, no. 9, tory,” Nature Reviews Genetics, vol. 6, no. 2, pp. 119– pp. 722–732, 2005. 127, 2005. [77] A. Babic, A. Lindner, M. Vulic, E. Stewart, and M. Rad- 15

man, “Direct visualization of horizontal gene transfer,” [97] Z. Zhang, G. Weinstock, and M. Gerstein, “Rapid evo- Science, vol. 319, no. 5869, pp. 1533–1536, 2008. lution by positive Darwinian selection in T-cell anti- [78] C. Pal, M. Macia, A. Oliver, I. Schachar, and A. Buck- gen CD4 in primates,” Journal of Molecular Evolution, ling, “Coevolution with viruses drives the evolution of vol. 66, no. 5, pp. 446–456, 2008. bacterial mutation rates,” Nature, vol. 450, pp. 1079– [98] K. Summers and B. Crespi, “Molecular evolution of the 1081, 2007. prostate cancer susceptibility locus RNASEL: evidence [79] L. D. McDaniel, E. Young, J. Delaney, F. Ruhnau, K. B. for positive selection,” Infection, Genetics and Evolu- Ritchie, and J. H. Paul, “High frequency of horizontal tion, vol. 8, no. 3, pp. 297–301, 2008. gene transfer in the oceans,” Science, vol. 330, p. 50, [99] J. Meyer, S. Ellner, N. Hairston, L. Jones, and 2010. T. Yoshida, “Prey evolution on the time scale [80] J. Hotopp, M. Clark, D. Oliveira, J. Foster, P. Fischer, of predator–prey dynamics revealed by allele-specific M. Torres, J. Giebel, N. Kumar, N. Ishmael, S. Wang, quantitative PCR,” Proceedings of the National et al., “Widespread lateral gene transfer from intra- Academy of Sciences, vol. 103, no. 28, pp. 10690–10695, cellular bacteria to multicellular eukaryotes,” Science, 2006. vol. 317, no. 5845, pp. 1753–1756, 2007. [100] K. Hillesland, G. Velicer, and R. Lenski, “Experimental [81] C. Bi´emont and C. Vieira, “Genetics: junk DNA as an evolution of a microbial predator’s ability to find prey,” evolutionary force,” Nature, vol. 443, no. 7111, pp. 521– Proceedings of the Royal Society B: Biological Sciences, 524, 2006. vol. 276, no. 1656, pp. 459–467, 2009. [82] S. Schaack, C. Gilbert, and C. Feschotte, “Promiscu- [101] E. Palkovacs and D. Post, “Experimental evidence that ous DNA: horizontal transfer of transposable elements phenotypic divergence in predators drives community and why it matters for eukaryotic evolution,” Trends in divergence in prey,” Ecology, vol. 90, no. 2, pp. 300– Ecology & Evolution, 2010. in press. 305, 2009. [83] D. Rankin, E. Rocha, and S. Brown, “What traits are [102] J. Bailey, A. Hendry, M. Kinnison, D. Post, E. Palko- carried on mobile genetic elements, and why&quest,” vacs, F. Pelletier, L. Harmon, and J. Schweitzer, “From Heredity, vol. 1, pp. 1–10, 2010. in press. genes to ecosystems: an emerging synthesis of eco- [84] S. Venner, C. Feschotte, and C. Bi´emont, “Dynamics of evolutionary dynamics,” New Phytologist, vol. 184, transposable elements: towards a community ecology of no. 4, pp. 746–749, 2009. the genome,” Trends in Genetics, vol. 25, no. 7, pp. 317– [103] F. Pelletier, D. Garant, and A. Hendry, “Eco- 323, 2009. evolutionary dynamics,” Philosophical Transactions of [85] J. de Lamarck, Philosophie zoologique. C. Martins, 1809. the Royal Society B: Biological Sciences, vol. 364, [86] E. Koonin and Y. Wolf, “Is evolution Darwinian or/and no. 1523, pp. 1483–1489, 2009. Lamarckian?,” Biology Direct, vol. 4, no. 42, pp. 1–14, [104] L. Jones, L. Becks, S. Ellner, N. Hairston Jr, T. Yoshida, 2009. and G. Fussmann, “Rapid contemporary evolution and [87] E. Jablonka and M. Lamb, “Epigenetic inheritance in clonal food web dynamics,” Philosophical Transactions evolution,” Journal of Evolutionary Biology, vol. 11, of the Royal Society of London, Series B: Biological Sci- no. 2, pp. 159–183, 1998. ences, vol. 364, no. 1523, pp. 1579–1591, 2009. [88] O. Rando and K. Verstrepen, “Timescales of genetic and [105] C. terHorst, T. Miller, and D. Levitan, “Evolution of epigenetic inheritance,” Cell, vol. 128, no. 4, pp. 655– prey in ecological time reduces the effect size of preda- 668, 2007. tors in experimental microcosms.,” Ecology, vol. 91, [89] M. Nowacki and L. Landweber, “Epigenetic inheritance no. 3, pp. 629–636, 2010. in ciliates,” Current opinion in microbiology, vol. 12, [106] T. Yoshida, S. Ellner, L. Jones, B. Bohannan, R. Lenski, no. 6, pp. 638–643, 2009. and N. Hairston Jr, “Cryptic population dynamics: [90] N. Guttenberg and N. Goldenfeld, “Cascade of Com- rapid evolution masks trophic interactions,” PLoS Biol, plexity in Evolving Predator-Prey Dynamics,” Physical vol. 5, no. 9, p. e235, 2007. Review Letters, vol. 100, no. 5, p. 58102, 2008. [107] M. Bulmer, “Coevolution of codon usage and transfer [91] L. Caporale, Darwin in the genome: molecular strate- RNA abundance,” Nature, vol. 325, no. 6106, pp. 728– gies in biological evolution. McGraw-Hill Professional, 730, 1987. 2003. [108] M. Bulmer, “The selection-mutation-drift theory of [92] N. G. Hairston, S. P. Ellner, M. A. Geber, T. Yoshida, synonymous codon usage,” Genetics, vol. 129, no. 3, and F. J. A, “Rapid evolution and the convergence pp. 897–907, 1991. of ecological and evolutionary time,” Ecology Letters, [109] K. Vetsigian and N. Goldenfeld, “Genome rhetoric and vol. 8, pp. 1114–1127, 2005. the emergence of compositional bias,” Proceedings of the [93] L. Cowen and S. Lindquist, “Hsp90 potentiates the National Academy of Sciences, vol. 106, no. 1, pp. 215– rapid evolution of new traits: drug resistance in diverse 220, 2009. fungi,” Science, vol. 309, no. 5744, p. 2185, 2005. [110] L. Hartwell, J. Hopfield, S. Leibler, and A. Mur- [94] J. Thompson, “Rapid evolution as an ecological pro- ray, “From molecular to modular cell biology,” Nature, cess,” Trends in Ecology & Evolution, vol. 13, no. 8, vol. 402, pp. C47–C52, 1999. pp. 329–332, 1998. [111] A. Barab´asi and Z. Oltvai, “Network biology: under- [95] S. Carroll, A. Hendry, D. Reznick, and C. Fox, “Evolu- standing the cell’s functional organization,” Nature Re- tion on ecological time-scales,” Ecology, vol. 21, pp. 387– views Genetics, vol. 5, no. 2, pp. 101–113, 2004. 393, 2007. [112] N. Kashtan and U. Alon, “Spontaneous evolution of [96] D. Steinhauer and J. Holland, “Rapid evolution of RNA modularity and network motifs,” Proceedings of the Na- viruses,” Annual Reviews in Microbiology, vol. 41, no. 1, tional Academy of Sciences, vol. 102, no. 39, pp. 13773– pp. 409–431, 1987. 13778, 2005. 16

[113] N. Kashtan, E. Noor, and U. Alon, “Varying environ- [131] K. Korolev, M. Avlund, O. Hallatschek, and D. Nel- ments can speed up evolution,” Proc. Natl. Acad. Sci. son, “Genetic demixing and evolution in linear step- USA, vol. 104, pp. 13711–13716, 2006. ping stone models,” Reviews of Modern Physics, vol. 82, [114] J. Holland, Adaptation in natural and artificial systems. pp. 1691–1718, 2010. MIT press Cambridge, MA, 1992. [132] A. McKane and T. Newman, “Stochastic models in pop- [115] D. J. Earl and M. W. Deem, “Evolvability is a selectable ulation biology and their deterministic analogs,” Phys- trait,” Proc. Natl. Acad. Sci. USA, vol. 110, no. 32, ical Review E, vol. 70, no. 4, p. 41902, 2004. pp. 11531–11536, 2004. [133] T. Butler and N. Goldenfeld, “Robust ecological pattern [116] J. He, J. Sun, and M. Deem, “Spontaneous emergence of formation induced by demographic noise,” Physical Re- modularity in a model of evolving individuals and in real view E, vol. 80, no. 3, p. 30902, 2009. networks,” Physical Review E, vol. 79, no. 3, p. 31907, [134] F. Odling-Smee, K. Laland, and M. Feldman, Niche 2009. construction: the neglected process in evolution. Prince- [117] C. P´al, B. Papp, and M. Lercher, “Adaptive evolution of ton Univ Press, 2003. bacterial metabolic networks by horizontal gene trans- [135] R. Day, K. Laland, and J. Odling-Smee, “Rethink- fer,” Nature genetics, vol. 37, no. 12, pp. 1372–1375, ing Adaptation,” Perspectives in Biology and Medicine, 2005. vol. 46, no. 1, pp. 80–95, 2003. [118] M. Parter, N. Kashtan, and U. Alon, “Environmental [136] R. Lenski, J. Barrick, and C. Ofria, “Balancing ro- variability and modularity of bacterial metabolic net- bustness and evolvability,” PLoS Biol, vol. 4, no. e428, works,” BMC Evolutionary Biology, vol. 7, pp. 169/1–8, pp. 2190–2192, 2006. 2007. [137] S. R. Palumbi, “Better evolution through chemistry: [119] A. Kreimer, E. Borenstein, U. Gophna, and E. Rup- evolution driven by human changes to the chemical en- pin, “The evolution of modularity in bacterial metabolic vironment,” in Chemical evolution II: from the origin of networks,” Proceedings of the National Academy of Sci- life to modern society (L. Zaikowski et al., eds.), Amer- ences, vol. 105, no. 19, pp. 6976–6981, 2008. ican Chemical Society, 2009. [120] M. Sullivan, D. Lindell, J. Lee, L. Thompson, [138] N. Goldenfeld, “No man is an island: quoted in editorial J. Bielawski, and S. Chisholm, “Prevalence and Evolu- on complex systems,” Nature Physics, vol. 5, p. 1, 2009. tion of Core Photosystem II Genes in Marine Cyanobac- [139] T. S. Ray, “Evolution, ecology, and optimization of dig- terial Viruses and Their Hosts,” PLoS Biol, vol. 4, no. 8, ital organisms,” in Scientific Excellence in Supercom- p. e234, 2006. puting: The IBM 1990 Contest Prize Papers (K. R. [121] E. S. Anderson, “Possible importance of transfer factors Billingsley, E. Derohanes, and I. H Brown, eds.), in bacterial evolution,” Nature, vol. 209, pp. 637–638, pp. 489–531, The University of Georgia: The Baldwin 1966. Press, 1991. [122] N. G. Anderson, “Evolutionary significance of virus in- [140] C. Adami and C. T. Brown, “Evolutionary learning in fection,” Nature, vol. 227, pp. 1346–1347, 1970. the 2d artificial life systems: Avida,” in Proc. Artificial [123] S. Sonea, “A bacterial way of life,” Nature, vol. 331, Life IV (R. Brooks and P. Maes, eds.), (US), pp. 377– no. 6153, p. 216, 1988. 381, MIT Press, 1994. [124] R. Johnson, J. Evans, G. Robinson, and M. Berenbaum, [141] G. Caldarelli, P. Higgs, and A. McKane, “Modelling “Changes in transcript abundance relating to colony Coevolution in Multispecies Communities,” Journal of collapse disorder in honey bees (Apis mellifera),” Pro- Theoretical Biology, vol. 193, no. 2, pp. 345–358, 1998. ceedings of the National Academy of Sciences, vol. 106, [142] C. Adami, “What is complexity?,” BioEssays, vol. 24, no. 35, pp. 14790–14795, 2009. no. 12, pp. 1085–1094, 2002. [125] M. Horie, T. Honda, Y. Suzuki, Y. Kobayashi, T. Daito, [143] A. McKane, “Evolving complex food webs,” The Eu- T. Oshida, K. Ikuta, P. Jern, T. Gojobori, J. Coffin, ropean Physical Journal B-Condensed Matter, vol. 38, et al., “Endogenous non-retroviral RNA virus elements no. 2, pp. 287–295, 2004. in mammalian genomes,” Nature, vol. 463, no. 7277, [144] M. Maron and C. T. Fernando, “Food webs and the pp. 84–87, 2010. evolution of organism complexity,” in Complexity Work- [126] V. Belyi, A. Levine, and A. Skalka, “Unexpected In- shop, Artificial Life X, 2006. heritance: Multiple Integrations of Ancient Bornavirus [145] S. Elena and R. Lenski, “Evolution experiments with and Ebolavirus/Marburgvirus Sequences in Vertebrate microorganisms: The dynamics and genetic bases of Genomes,” PLoS Pathog, vol. 6, no. 7, pp. e1001030/1– adaptation,” Nature Reviews: Genetics, vol. 4, no. 6, 12, 2010. pp. 457–469, 2003. [127] A. Wallace, “On the law which has regulated the intro- [146] C. Adami, “Digital genetics: unravelling the genetic duction of new species,” Annals and Magazine of Nat- basis of evolution,” Nature Reviews Genetics, vol. 7, ural History, vol. 16, no. 2, pp. 184–196, 1855. pp. 109–118, 2006. [128] K. Ibrahim, R. Nichols, and G. Hewitt, “Spatial pat- [147] C. Wilke and C. Adami, “The biology of digital organ- terns of genetic variation generated by different forms isms,” Trends in Ecology & Evolution, vol. 17, no. 11, of dispersal during range expansion,” Heredity, vol. 77, pp. 528–532, 2002. no. 3, pp. 282–291, 1996. [148] R. E. Lenski, C. Ofria, T. C. Collier, and C. Adami, [129] O. Hallatschek and D. Nelson, “Gene surfing in ex- “Genome complexity, robustness and genetic interac- panding populations.,” Theoretical Population Biology, tions in digital organisms,” Nature, vol. 400, pp. 661– vol. 73, no. 1, pp. 158–170, 2008. 664, 1999. [130] O. Hallatschek and K. Korolev, “Fisher waves in the [149] C. Adami, C. T. Brown, and M. R. Haggerty, strong noise limit,” Physical Review Letters, vol. 103, “Abundance-distributions in artificial life and stochas- no. 10, p. 108103, 2009. tic models: “age and area” revisited,” in Proceedings of 17

the Third European Conference on Advances in Artifi- Biological Sciences, vol. 271, no. 1534, p. 107, 2004. cial Life, (London, UK), pp. 503–514, Springer-Verlag, [169] G. Fussmann, M. Loreau, and P. Abrams, “Eco- 1995. evolutionary dynamics of communities and ecosystems,” [150] E. Ostrowski, C. Ofria, and R. Lenski, “Ecological spe- Ecology, vol. 21, pp. 465–477, 2007. cialization and adaptive decay in digital organisms.,” [170] S. Gandon, A. Buckling, E. Decaestecker, and T. Day, Am. Nat, vol. 169, pp. E1–E20, 2007. “Host-parasite coevolution and patterns of adaptation [151] R. Lenski, C. Ofria, R. T. Pennock, and C. Adami, across time and space,” Journal of Evolutionary Biol- “The evolutionary origin of complex features,” Nature, ogy, vol. 21, no. 6, pp. 1861–1866, 2008. vol. 423, pp. 139–144, 2003. [171] M. Brockhurst, “Using Microbial Microcosms to Study [152] C. Woese, “Evolution of macromolecular complexity,” Host–parasite Coevolution,” Evolution: Education and Journal of Theoretical Biology, vol. 33, no. 1, pp. 29– Outreach, pp. 1–5, 2010. 34, 1971. [172] S. Paterson, T. Vogwill, A. Buckling, R. Benmayor, [153] S. Jones and J. Thornton, “Protein-protein interactions: A. Spiers, N. Thomson, M. Quail, F. Smith, D. Walker, a review of protein dimer structures.,” Progress in bio- B. Libberton, et al., “Antagonistic coevolution acceler- physics and molecular biology, vol. 63, no. 1, pp. 31–65, ates molecular evolution,” Nature, 2010. 1995. [173] K. Hillesland and D. Stahl, “Rapid evolution of stability [154] D. Goodsell and A. Olson, “Structural symmetry and and productivity at the origin of a microbial mutual- protein function,” Annu. Rev. Biophys. Biomol. Struct, ism,” Proceedings of the National Academy of Sciences, vol. 29, pp. 105–153, 2000. vol. 107, no. 5, p. 2124, 2010. [155] K. Plaxco and M. Gross, “Protein Complexes: The Evo- [174] M. Kondoh, “Foraging adaptation and the relationship lution of Symmetry,” Current Biology, vol. 19, no. 1, between food-web complexity and stability,” 2003. pp. R25–R26, 2009. [175] G. Ackland and I. Gallagher, “Stabilization of large gen- [156] J. Pereira-Leal, E. Levy, C. Kamp, and S. Teichmann, eralized Lotka-Volterra foodwebs by evolutionary feed- “Evolution of protein complexes by duplication of ho- back,” Physical review letters, vol. 93, no. 15, p. 158701, momeric interactions,” Genome biology, vol. 8, no. 4, 2004. pp. R51/1–12, 2007. [176] Z. Wang and N. Goldenfeld, “Stochastic model of an- [157] E. Levy, E. Erba, C. Robinson, and S. Teichmann, “As- tagonistic coevolution in a predator-prey model.” un- sembly reflects evolution of protein complexes,” Nature, published. vol. 453, no. 7199, pp. 1262–1265, 2008. [177] R. Axelrod and W. Hamilton, “The evolution of coop- [158] L. Richardson, Weather Prediction by Numerical Pro- eration.,” Science, vol. 211, no. 4489, pp. 1390–1396, cess. Cambridge University Press, Cambridge, UK, 1981. 1922. [178] J. Smith, Evolution and the Theory of Games. Cam- [159] A. N. Kolmogorov, “Local structure of turbulence in an bridge Univ Pr, 1982. incompressible fluid at very high Reynolds numbers,” [179] M. A. Nowak and K. Sigmund, “Evolutionary Dynam- Dokl Acad Nauk USSR, vol. 30, pp. 299–303, 1941. ics of Biological Games,” Science, vol. 303, no. 5659, [160] N. Kataoka and K. Kaneko, “Dynamical networks in pp. 793–799, 2004. function dynamics,” Physica D: Nonlinear Phenomena, [180] A. Nowak Martin, Evolutionary Dynamics: Exploring vol. 181, no. 3-4, pp. 235–251, 2003. the Equations of Life. Cambridge, 2006. [161] R. Anderson and R. May, “Coevolution of hosts and [181] B. McGill and J. Brown, “Evolutionary Game Theory parasites,” Parasitology, vol. 85, no. 02, pp. 411–426, and Adaptive Dynamics of Continuous Traits,” Annu. 1982. Rev. Ecol. Evol. Syst, vol. 38, pp. 403–35, 2007. [162] J. Bascompte and P. Jordano, “Plant-Animal Mutualis- [182] P. Turner and L. Chao, “Prisoner’s dilemma in an RNA tic Networks: The Architecture of Biodiversity,” Annu. virus.,” Nature, vol. 398, no. 6726, pp. 441–443, 1999. Rev. Ecol. Evol. Syst, vol. 38, pp. 567–93, 2007. [183] S. Brown, “Collective action in an RNA virus,” Jour- [163] E. Kiers and R. Denison, “Sanctions, cooperation, and nal of Evolutionary Biology, vol. 14, no. 5, pp. 821–828, the stability of plant-rhizosphere mutualisms,” Annual 2001. review of ecology, evolution, and systematics, vol. 39, [184] B. Kirkup and M. Riley, “Antibiotic-mediated antago- pp. 215–236, 2008. nism leads to a bacterial game of rock–paper–scissors in [164] L. Dethlefsen, M. Mcfall-Nagai, and D. A. Relman, vivo,” Nature, vol. 428, no. 6981, pp. 412–414, 2004. “An ecological and evolutionary perspective on human- [185] J. Gore, H. Youk, and A. Van Oudenaarden, “Snow- microbe mutualism and disease,” Nature, vol. 449, drift game dynamics and facultative cheating in yeast,” no. 7164, pp. 811–818, 2007. Nature, vol. 459, no. 7244, pp. 253–256, 2009. [165] L. Van Valen, “A new evolutionary law,” Evolutionary [186] J. Nash, “Equilibrium points in n-person games,” Pro- Theory, vol. 1, no. 1, pp. 1–30, 1973. ceedings of the National Academy of Sciences of the [166] J. Weitz, H. Hartman, and S. Levin, “Coevolutionary United States of America, pp. 48–49, 1950. arms races between bacteria and bacteriophage,” Pro- [187] E. Akiyama and K. Kaneko, “Dynamical systems game ceedings of the National Academy of Sciences, vol. 102, theory and dynamics of games,” Physica D: Nonlinear no. 27, pp. 9535–9540, 2005. Phenomena, vol. 147, no. 3-4, pp. 221–258, 2000. [167] M. Brockhurst, A. Morgan, P. Rainey, and A. Buck- [188] E. Akiyama and K. Kaneko, “Dynamical systems game ling, “Population mixing accelerates coevolution,” Ecol- theory II:: A new approach to the problem of the social ogy Letters, vol. 6, no. 11, pp. 975–979, 2003. dilemma,” Physica D: Nonlinear Phenomena, vol. 167, [168] M. Brockhurst, P. Rainey, and A. Buckling, “The effect no. 1-2, pp. 36–71, 2002. of spatial heterogeneity and parasites on the evolution [189] Y. Sato, E. Akiyama, and J. Doyne Farmer, “Chaos of host diversity.,” Proceedings of the Royal Society B: in learning a simple two-person game,” in Proceedings 18

of the National Academy of Science, vol. 99, pp. 4748– no. 6, pp. 491–511, 1943. 4751, 2002. [198] P. Sniegowski and R. Lenski, “Mutation and adapta- [190] M. Zimmermann, V. Eguıluz, and M. Miguel, “Cooper- tion: the directed mutation controversy in evolutionary ation, adaptation and the emergence of leadership Eco- perspective,” Annual Review of Ecology and Systemat- nomics with Heterogeneous Interacting Agents vol 503,” ics, vol. 26, no. 1, pp. 553–578, 1995. Lecture Notes in Economics and Mathematical Systems, [199] R. Galhardo, P. Hastings, and S. Rosenberg, “Mutation pp. 73–86, 2001. as a stress response and the regulation of evolvability,” [191] H. Ebel and S. Bornholdt, “Coevolutionary games on Critical Reviews in Biochemistry and Molecular Biology, networks,” Phys. Rev. E, vol. 66, no. 5, pp. 056118/1–8, vol. 42, no. 5, pp. 399–435, 2007. 2002. [200] G. Dunny, “The peptide pheromone-inducible conjuga- [192] M. Perc and A. Szolnoki, “Coevolutionary games–A tion system of Enterococcus faecalis plasmid pCF10: mini review,” BioSystems, vol. 99, pp. 109–125, 2009. cell–cell signalling, gene transfer, complexity and evo- [193] Y. Sato, E. Akiyama, and J. Crutchfield, “Stability and lution,” Philosophical Transactions of the Royal Soci- diversity in collective adaptation,” Physica D: Nonlin- ety B: Biological Sciences, vol. 362, no. 1483, pp. 1185– ear Phenomena, vol. 210, no. 1-2, pp. 21–57, 2005. 1193, 2007. [194] J. Farmer and A. Lo, “Frontiers of finance: Evolu- [201] J. Manson, L. Hancock, and M. Gilmore, “Mecha- tion and efficient markets,” Proceedings of the National nism of chromosomal transfer of Enterococcus faecalis Academy of Sciences of the United States of America, pathogenicity island, capsule, antimicrobial resistance, vol. 96, no. 18, pp. 9991–9992, 1999. and other traits,” Proceedings of the National Academy [195] T. Hens and K. Schenk-Hopp´e, “Evolutionary finance: of Sciences, vol. 107, no. 27, pp. 12269–12274, 2010. introduction to the special issue,” Journal of mathemat- [202] E. Schr¨odinger, Nature and the Greeks. Cambridge Uni- ical economics, vol. 41, no. 1-2, pp. 1–5, 2005. versity Press, 1954. [196] R. May, S. Levin, and G. Sugihara, “Ecology for [203] H. Frauenfelder, P. Wolynes, and R. Austin, “Biologi- bankers,” Nature, vol. 451, no. 21, pp. 893–895, 2008. cal physics,” Reviews of Modern Physics, vol. 71, no. 2, [197] S. Luria and M. Delbr¨uck, “Mutations of bacteria from pp. 419–430, 1999. virus sensitivity to virus resistance,” Genetics, vol. 28,