- Open Access
Interaction-based evolution: how natural selection and nonrandom mutation work together
Biology Direct volume 8, Article number: 24 (2013)
The modern evolutionary synthesis leaves unresolved some of the mostfundamental, long-standing questions in evolutionary biology: What is therole of sex in evolution? How does complex adaptation evolve? How canselection operate effectively on genetic interactions? More recently, themolecular biology and genomics revolutions have raised a host of criticalnew questions, through empirical findings that the modern synthesis fails toexplain: for example, the discovery of de novo genes; the immenseconstructive role of transposable elements in evolution; genetic varianceand biochemical activity that go far beyond what traditional naturalselection can maintain; perplexing cases of molecular parallelism; andmore.
Presentation of the hypothesis
Here I address these questions from a unified perspective, by means of a newmechanistic view of evolution that offers a novel connection betweenselection on the phenotype and genetic evolutionary change (while relying,like the traditional theory, on natural selection as the only source offeedback on the fit between an organism and its environment). I hypothesizethat the mutation that is of relevance for the evolution of complexadaptation—while not Lamarckian, or “directed” to increasefitness—is not random, but is instead the outcome of a complex andcontinually evolving biological process that combines information frommultiple loci into one. This allows selection on a fleeting combination ofinteracting alleles at different loci to have a hereditary effect accordingto the combination’s fitness.
Testing and implications of the hypothesis
This proposed mechanism addresses the problem of how beneficial geneticinteractions can evolve under selection, and also offers an intuitiveexplanation for the role of sex in evolution, which focuses on sex as thegenerator of genetic combinations. Importantly, it also implies that geneticvariation that has appeared neutral through the lens of traditional theorycan actually experience selection on interactions and thus has a muchgreater adaptive potential than previously considered. Empirical evidencefor the proposed mechanism from both molecular evolution and evolution atthe organismal level is discussed, and multiple predictions are offered bywhich it may be tested.
This article was reviewed by Nigel Goldenfeld (nominated by Eugene V.Koonin), Jürgen Brosius and W. Ford Doolittle.
To explain adaptive evolution, we still use today ideas from the foundations of themodern evolutionary synthesis formed in the 1920s and 1930s. Yet there has been asea of change in the empirical realities since then. The molecular biology andgenomics revolutions have occurred and brought with them fundamental new empiricalfindings. Some of these findings were simply unexpected from traditional theory andare unengaged by it, including the discovery in the 1960s of far more geneticvariance than could be subject to selection according to traditional theory [1, 2], and ENCODE’s very recent finding that the majority of the humangenome is biochemically active . From the perspective of traditional theory, we are now forced to predictthat much of this activity is just “biochemical noise” and not reallypart of the organism, again because traditional natural selection cannot act on somuch evolving matter and for other important reasons [4–8]. Other empirical findings have been more directly challenging. Considerfor example de novo genes (e.g., [9–13])—genes that presumably have arisen “out of thin air” bya sequence of random mutations that came together into a new functioning gene,including signals for transcription and translation and even alternative splicing . This de novo formation takes place even though traditionalnatural selection could not have acted on this sequence of mutations until the genewas already complete (substantial enough to be active), in clear contradiction withwhat Jacob justifiably predicted to be impossible . Also challenging to traditional theory are findings of such fundamentalsignificance for our understanding of evolution as the evolutionary organizing ofmore than 1500 genes into a new genetic network underlying a novel, complexadaptation by transposable elements . Whether for these or other reasons, a sense of curiosity about the newempirical reality has been conveyed by such luminaries as Doolittle , Graur and Li , Wagner , Fedoroff , West-Eberhard  and others.
In light of these findings, it is commonly assumed that traditional natural selectionoperates to the extent that it can, and that originally neutral mutations accountfor anything that selection does not account for. But this modern approach leads toa deep inconsistency. The original idea of natural selection and random mutation,implicit in Fisher’s work , was to minimize the amount of “work” done by chance in theevolution of adaptation and let natural selection do the job of evolving anadaptation by pulling out from the noise the supposed slightly beneficial mutationsand causing them to accumulate inexorably toward the evolution of adaptation. It isinconsistent to invoke this idea, which attempted to minimize the amount ofevolutionary work done by fortuitous chance, while at the same time allowing for anunspecified number of originally neutral mutations to play an inherent role in theevolution of adaptation, as is currently done for example in the case of denovo genes. Indeed, there is no quantification of the amount of chance thatwe call upon to explain the evolution of adaptation (namely the chance that isinvolved in the arising of accidental mutations and in random genetic drift, to theextent that the latter is invoked)—a deep problem not yet addressed at all bythe whole body of population genetics.
This paper holds that the key to solving the fundamental problems brought about bythe molecular biology and genomics revolutions is to go back and revisit somefundamental old problems in evolutionary theory that have been open since beforeeven the rise of molecular biology itself. Attending to these old open problems, wemay be able to offer a deep change to the core of the theory of natural selectionthat will reconnect the theory better to the evidence available today. I will beginby discussing two fundamental unresolved problems, namely the role of sex inevolution and how selection on interactions between alleles of different genes canplay an evolutionarily constructive role. I will show that, in fact, these twoproblems are different aspects of one and the same thing.
My general approach will be as follows. I will continue to assume that selection isthe only source of feedback on the fit between an organism and its environment.However, I will revisit the question of the nature of the mutation that drivesevolution. Here, I will continue to assume that mutation is not Lamarckian, and thata given mutation is not more likely to occur in an environment where it increasesfitness than in an environment where it does not [17, 22, 23]. However, I will show that there is another alternative, which has notbeen attended to yet, which is neither accidental mutation nor mutation thatviolates our core assumptions. Revisiting the question of the nature of mutation, Iwill construct a new theory of how adaptive evolution happens, based on selection,but also on a new connection between selection on the phenotype and geneticevolutionary change. I will show that this approach addresses the unresolvedproblems of sex and interactions from a unifying perspective, and at the same timebegins to propose a mechanism at the point where traditional theory relies only onpure chance. Empirical evidence for and predictions derived from this new mechanismwill be discussed for a variety of topics at both the organismal and molecularlevels (from plant mating systems and canalization, to molecular parallelism and thenature of mutation, to genetic mechanisms in the sperm cells), with relevance thatultimately goes beyond science to medicine.
The theory will be proposed verbally, and not mathematically, because it is not clearthat traditional mathematical tools are immediately suitable for itsmathematization. The price of accepting the benefit of unification—where theproblems of sex, interactions and the lack of quantification of chance intraditional theory are addressed in one—will be to accept that what we knowregarding how evolution happens is merely the tip of the iceberg. An outline of themain points is given in the Summary section.
Fundamental problems in traditional evolutionary theory: sex and interactions
The most obvious effect of sex is that it creates an exponentially large number ofdifferent potential combinations of alleles at different loci—indeed it makesindividuals unique. When biologists are asked what the role of sex is in evolution,they often say that from a given number of alleles at different loci it creates thisalmost endless number of different genetic combinations; and since natural selectionoperates on genetic variation, this “increased variance” facilitatesevolution. But the insufficiency of this explanation is well known to investigatorsof the evolution of sex and recombination . What is the point of creating, by the shuffling of genes, a variety ofgenetic combinations that will be tested by natural selection? One may wish to saythat, among the many combinations, particularly good ones will be found that wouldnot have existed otherwise. But in saying this, a basic point is forgotten: thesecombinations of alleles at different loci are not heritable. Just as sex brings themtogether, so too it breaks them down.
Consistent with this point, the core of the Fisherian theory of adaptive evolution,which forms an essential part of the modern synthesis of evolution , is structured in a way that makes these combinations of alleles ascomplex wholes inessential: following it, population geneticists have often assumedthat each allele can have a selective value in and of itself—it can be a“good” or a “bad” mutation (“beneficial” or“deleterious”) with little consideration of the genetic context . This way an allele is “blind” to the particular combinationsit goes through. Selection operates statistically on each allele independently ofother alleles, because any given allele makes essentially the same additivecontribution in different individuals toward the numerical sums that are thoseindividuals’ “fitness values”. Alleles pass each other like shipsin the night as they move through the population , and the population is treated as a collection of allele frequencies,each for an independent, essentially non-interacting locus . While Fisher did discuss interactions both within and between loci, evenin the context of recombination , those were not part of his core process of adaptive evolution, which wasinstead based on independent (or “additive”) effects of separateloci.
However, this way of thinking has left the role of sex a mystery. Notice that thesame beneficial or deleterious mutations could have arisen and been favored ordisfavored in a sexual as well as an asexual population. By providing a basicmechanism for evolution that works with or without sex, the Fisherian theory hascreated a view of evolution where sex is not really essential. Since then,investigators only proposed subsidiary and circumscribed benefits that sex may bringon top of an evolutionary mechanism that can work essentially without it (e.g., [26–30]). But all such “bonuses” proposed so far require ratherspecific conditions , and, even considering all of these bonuses together, it is not clearthat their collection forms an appealing way of explaining the near-ubiquity of sex [32, 33].
Wright never accepted Fisher’s conceptualization of evolution. Wright believedthat genes interacted in complex networks and that likewise alleles at differentloci must interact with each other to generate any notable evolutionary change [34–36]. The notion of selection acting on each allele in and of itself seemed tohim fundamentally insufficient for explaining the evolution of complex adaptation . Note, however, that an interaction between alleles at different locicannot be persistently selected on, according to the traditional view, preciselybecause sex disassembles such combinations of alleles, as discussed. Instead ofselection, Wright proposed in his shifting balance theory that the basis for anadaptive complex of genes will first arise by chance (after the constituent allelesat different loci have not only arisen by chance, but have also spread by randomgenetic drift in a given subpopulation), and then natural selection will come tobear on the process by simple (non-interactive) improvements and by helping tospread the constituent alleles from the given subpopulation to other subpopulationsthrough migrants [34, 35]. This theory required stringent conditions on the population structure [37, 38], attempted to obtain the basis for a new complex adaptation by purechance, and has not been uniformly accepted [38, 39]. Thus, in distinction from selection on separate genetic effects and thesupposed chance formation of the basis for beneficial genetic interactions by randomgenetic drift, we still do not have a theory for how selection on geneticinteractions can be at the core of the adaptive evolutionary process.
There are multiple ways to derive the theory presented here, but the one describedbelow begins with the problem of sex and interactions just mentioned. In accordancewith the long-standing intuition of biologists, I will argue that the essentialthing about sex is that it generates combinations of alleles at different loci;indeed I will argue more: that these combinations are a matter of necessity forevolution. From the traditional theory, this cannot be, because these transientcombinations cannot be inherited. But we will soon realize that they can, though notin the traditional way. This will take a sweeping change of outlook, which at firstappears to be itself impossible: I will posit that mutation is nonrandom, and showthat this solves the problem from the traditional theory that combinations ofalleles cannot have a lasting effect. This appears impossible at first because weare correctly trained to avoid Lamarckian transmission  and Lamarckian “directed mutation” as possible explanationsfor evolution at the general level [22, 23]. But the nonrandom mutation discussed here will not be of these kinds.The “nonrandomness” I will refer to is emphatically not the one wheremutation is more likely to occur in an environment where it increases fitness, andis therefore not the one disallowed by traditional theory [17, 22, 23].
Selection on interactions can drive evolution when mutation is nonrandom
Let us develop the concept of nonrandom mutation here carefully from square one. By“nonrandom mutation” we will mean that the mutation that drivesevolution is not accidental—it is not an unintended disruption of the geneticcode, caused for example by external agents or by oxidative stress (althoughmutations of such kinds do happen and can lead to disease). We will take this tomean that the mutation that drives evolution is the result of an organic processthat belongs to the organism.
If so, then like all other biological processes that belong to the organism, thisprocess must be specified by the genes. These genes interact, as genes alwaysinteract in the determination of a trait, except that, while a classical trait issomething that serves in the survival and reproduction of the organism, here we aretalking about a trait whose end result is genetic change. While genes interact andlead to a classical trait like the ear, here genes interact and cause geneticchange.
Given that genes interact in the determination of genetic change, and keeping theassumption that their alleles interact, this means that the mutation that drivesevolution is a process that combines information from alleles at multiple loci andwrites the result of the combination operation into one locus—the locus beingchanged by mutation (Figure 1a). (Also if multiple loci arechanged at once, information is combined from multiple loci to enact these multiplechanges.) By combining information from alleles at multiple loci into one locus,this operation creates from the combination of alleles a piece of information thatis not broken by the sexual shuffling of the genes, and is therefore heritable(Figure 1b). (It creates an allele, and this is an elementaryunit for the shuffling; the shuffling breaks only combinations of alleles). Thismeans that combinations of alleles at different loci do have an effect that laststhrough the shuffling: they transmit information to future generations through themutations that are derived from them.
This general-level point is as simple as it is crucial: if mutation is nonrandom,then selection on interactions has a hereditary effect. While selection oncombinations means that successful combinations survive and reproducepreferentially, the writing of mutation takes these successful combinations andmakes heritable mutations from them that will be transmitted to the next generation.Thus, natural selection on genetic combinations and nonrandom mutation worktogether.
Interestingly, there have always been only two main ways of thinking about adaptiveevolution (though more if we consider smaller variants and less influentialstreams): One has been the Darwinian theory of natural selection, which was turnedinto the neo-Darwinian theory of natural selection and random mutation (ns/rm) inthe 1920s and 1930s. In this theory, differential survival and reproduction is thesource of feedback that allows the fit between an organism and its environment. Theother has been the Lamarckian-transmissionist one, which holds that the organism issomehow able to sense what is needed for improvement in terms of the fit to theenvironment and then is able to change the hereditary material in a way thatimproves this fit, thus transmitting the improvement to the next generation. ThisLamarckian-transmissionist option is not only impossible as a general-levelexplanation for evolution , but, interestingly, if it were possible, its action would have renderedselection redundant . Therefore, the Lamarckian kind of nonrandom mutation on the one hand,and natural selection on the other hand, are rival hypotheses. We can now see thatthe theory presented here is a third alternative, distinct from the above two. Thenonrandom mutation considered here and natural selection are complementary, indiametric opposition to the above rivalry. Differential survival and reproduction isthe source of feedback on organismal fit to the environment. Nonrandom mutationcollects this feedback in a manner that allows natural selection to act on geneticinteractions. Thus, selection on the organism as a unified whole is possible.
The theory just proposed connects empirical facts at a deep level. It explains sexwhile making a substantial statement about the empirical nature of mutation: themutation that drives evolution is nonrandoma—it is anorganic process that belongs to the organism. Evidence and predictions regardingthis statement will be discussed later (see the section “Evidence from andpredictions for molecular evolution”), after further theory is developed thatwill make them clearer.
In the following sections I will discuss the prevalence, origin and maintenance ofsex, the nature of the evolution of complex adaptation at the phenotypic level, andhow they connect to the above. The reader who is primarily interested in themolecular side of this theory may skip to the section “A more detailed lookinto the new theory”.
Sex as a matter of necessity for evolution
Having described the core of the theory we can now expand on our empirical view.I use Barton and Charlesworth’s  evolutionary definition of sex as the shuffling of genes amongindividuals that leads to the creation of offspring that are geneticallydifferent from their parents. According to this most basicevolutionary-biological definition, sex is nearly universal : it occurs in plants and animals by syngamy, in fungi via the fusionof hyphae and in bacteria by conjugation and other means [33, 42]. Many species are capable of reproducing both sexually and asexually,but because their bouts of sexual reproduction keep their genes shuffled, theywill be considered sexual here. We will consider “asexual” thosespecies in which the shuffling of genes does not occur. Those are the obligateasexuals.
Several important facts can now be pointed out. First, obligate asexuals are veryrare. For example, Vrijenhoek  estimated that about 1 in 1000 animal species is an obligate asexual.Second, they appear to be headed toward ultimate extinction without leavingdescendant species behind. This point has been inferred from their phylogeneticdistribution: they inhabit small, recent, sparsely distributed twigs on the treeof life, which is consistent with the idea that they occasionally arise asterminal offshoots from sexual species (sexuals are the source and asexuals arethe sink) [44–48]. Indeed, their structure shows that they are recent derivations fromsexual ancestors: selfing plants still have reproductive structures that haveserved them in sexual reproduction in recent evolutionary times . Given this evidence (see further discussion in the next section), wecan infer that the immortal part of the tree of life is sexual.
Interestingly, and consistent with the above, Stebbins concluded from extensivestudies of plant morphology that asexuals are incapable of true evolutionaryinnovation . In accord with Stebbins , they have often been called “evolutionary dead ends”. Wemust also ponder the great extent of adaptive structure and effort devoted toimplementing the shuffling of genes throughout the biological world. Fromflowers to butterflies to human behavior, we do not need science to tell us thatsex forms an important part of the biological world. Indeed, it is intertwinedwith biological structure and function down to the molecular level, wheremeiosis involves extremely complex molecular machinery that implements theshuffling of genes.
With these facts in mind, we can now obtain a high-level insight on sex bycomparing it to its “peer” biological phenomena. What otherphenomena are ubiquitous across the immortal part of the tree of life? Sex canbarely be matched in terms of this ubiquity and importance. In this part of thetree of life, it can only be matched by such things as reproduction per se,metabolism in general, and the existence of the genetic code itself.Importantly, these phenomena are parts of the fundamental framework of life.They are not there because their “benefits outweigh their costs”;they are simply necessary. They are part of the definition of the process, as wedo not contemplate biological evolution without some kind of a conveyor ofhereditary information, without reproduction or metabolism. In accordance withthe evidence, these are the “peer phenomena” of sex; and in keepingwith a parsimonious picture, I hold that like its peer phenomena, sex is also amatter of necessity for evolution, a part of the infrastructural group.
Now note that the principle that sex is a matter of necessity for evolution,based on empirical facts, is consistent with the new theory of evolution justproposed, but is inconsistent with traditional theory. It is consistent with thenew theory because this theory argues that genetic combinations are a matter ofnecessity for evolution (selection operates on them), and sex creates thesecombinations. It is inconsistent with traditional theory as alreadydiscussed—the Fisherian theory offers a way of understanding evolutionthat takes sex conceptually out of the level of the essentials.
A prediction following work from Meselson’s lab
The contrast just mentioned renders particularly important the empirical questionsurrounding the putative ancient asexuals. It has been thought for a while thatsome asexuals may have evolved and diversified substantially, giving rise toasexual clades, the most famous example being the bdelloid rotifers. Since noone has observed these minute organisms in the act, they have been thought tohave evolved and diversified asexually for more than 35 million years, givingrise to 4 orders, 18 genera and 363 “species” according to onereport . The possibility that there are such exceptions to the rule ofasexuals as dead-ends has not been a fundamental problem for the traditionaltheory. Under the traditional theory, sex is not part of the evolutionaryinfrastructure but a bonus for which various separate reasons have beenproposed, each with its own specific conditions required. Thus, if an ancientand diversified asexual clade is observed, it can always be argued that it doesnot satisfy any of the requirements for sex without violating the core oftraditional theory (see Judson and Normark  for a discussion of this topic). Indeed, the problem lies more in theother direction: one may ask why there are not many more putative ancientasexuals, as no clarity is given from traditional theory over why the specificconditions required for the various bonuses proposed would sum up to covernearly all of nature.
However, for the theory presented here, the existence of an ancient, diversifiedasexual clade would be a fatal problem; because it would show that trueevolution can happen without sex, thus refuting the new theory. This raises aprediction. According to the theory presented here, all the putative diversifiedasexual clades are false examples in the following sense: if their members haveundergone substantial adaptive evolution and diversification, they have done soin a sexual state. Two possibilities that are in accord with this prediction arethat most of their members are still sexual today, or that a sexual core exists  from which asexuals are continually spun off due to hybridization orother reasons. According to both of these possibilities, even if we have not yetobserved mechanisms of sexual shuffling of genes in these organisms, they areout there to be found, and so if we look for them we will find them, accordingto this theory.
It is of interest, therefore, that Meselson recently reported [52, 53] that, having set to prove once and for all that the bdelloid rotifersare asexual, his lab seemed to have found the opposite: genetic analysis showshomologous gene shuffling in bdelloid rotifers. However, we still do not knowhow they do it—by what mechanisms they exchange genes or what triggerstheir elusive bouts of gene exchange. Assuming this result, not yet published atthe time of writing, holds, one prediction of the theory presented here isalready underway to being confirmed. Beyond this case, there are a couple ofdozen other cases of putative ancient asexual clades , which provide opportunity to test, and refute, this theory.
Sex predates asex
As soon as one proposes the principle that sex is a matter of necessity forevolution, a question comes up: If evolution started in an asexual state, withsex emerging at a later point, then evolution was already taking place beforesex. This in turn would mean that sex is not necessary for evolution.
Indeed, discussions of the origin of sex have often been couched implicitly orexplicitly in terms of “why did sex arise?” (presumably from asex)and “what benefit did it bring that gave it the advantage and led to itsprevalence?” [24, 46]. This discourse shows that there has been a tacit assumption that sexarose from asex, and that it outcompeted asex because its“advantages” outweighed its “costs”. If it arose fromasex, this implies that it is not a matter of necessity, as just mentioned; andif it succeeded because its “benefits outweighed its costs”, then itis not a matter of necessity—it is not a member of the infrastructuralgroup, to which this balance of costs and benefits is not applied.
But why have we been making this tacit assumption? One reason might be that sexappears to be more complex than asex, so it seems as though asex should havecome first, and sex should have been derived from it. But the fact that thesexual mechanisms of today appear complex does not mean that they have alwaysbeen so. In fact, if we ask ourselves what sex is at the most basic level, wewill find that it is merely the mixing of genetic material. So even if we pushthe conversation all the way back to the so-called “primordialsoup”—an era of utter speculation—we will find no reason toinsist that this primordial soup must have been asexual. The free mixing ofcompounds that the image of the “primordial soup” entails could justas well have been a “sexual beginning”. Indeed, all that we see frompresent evidence is that asexuals arise from sexuals; and that the asexuals areless complex than the sexuals because they are “brokensexuals”—sexuals with a missing piece in them. The hidden assumptionthat life started asexually must be exposed, because it has no empirical basis.Instead, the theory presented here is supportive of pioneering theories of Woese , Brosius [55, 56] and Vetsigian et al.  on rampant gene exchange in early life, and of Williams’s andothers’ views that sex is original .
An adaptation evolves by convergence on the population level
I will now describe the second point of the theory: if evolution is based oninteractions, then a trait arises not by sequential addition of one change at atime, each serviceable on its own, but by gradual stabilization of the trait asa complex whole—by a process of convergence on the population level, asdefined below.
For a trait to be part of the long-term process of adaptive evolution—insuch a manner that it is not transient, but rather further adaptive evolutioncan be based on it—we expect it to ultimately belong to all theindividuals in the population or species of interest (even if we are interestedonly in the population of individuals of a particular morph or sex). How does anew trait come to be shared by all individuals in a given population? TheFisherian theory has a ready mechanism for it: A new allele arises by randommutation that has a phenotypic meaning and a fitness value in and of itself. Itmakes the same change in the phenotype regardless of the particular individualgenetic combination it is in. If this allele is “beneficial”, itwill spread by traditional natural selection from the one individual in which itarose to the many, bringing along with it the change that it causes in thephenotype to the whole population. Thus the population comes to share thischange. Then, another beneficial allele will arise in some individual, spreadingand bringing its own change to all, and so on and so forth. It is very easy tosee here how the population comes to share a new trait.
However, if evolution is based on interactions, and interactions are notheritable in the same way that a Fisherian allele is, how does the populationcome to share a new trait?
Let us define a trait on a population level as something that belongs to allindividuals in the population or species of interest and thus does not changemuch as we move from one generation to the next through the sexual shuffling.(Note that this definition defines the trait on a population-level. One canstill talk about a “trait” that belongs to an individual, or anindividual variant. But an evolved adaptation is shared among individuals, andis captured by our definition of a trait). Now, consider the genetic differencesbetween individuals in a population at some arbitrary generation, generationt0. Over the generations, some of the t0 alleles become fixed, others become extinct, and thus the geneticdifferences of t0 gradually disappear (even as they give rise to new differences inthe meantime in accord with the theory presented here, as will be seen in thenext section). This means that the effect of the sexual shuffling on thephenotype that is due to the interactions between these genetic differences oft0 gradually becomes smaller. This means that the parents ofgeneration t x and their offspring in generation tx+1 gradually become more similar to each other as far as thephenotypic differences caused by the genetic variance of t0 are concerned, as x is increased. The differences oft0 have been removed, and something has become stable in thegenes.
We must conclude from the above that the evolution of an adaptation occurs byconvergence on the level of the population as a whole. It is a process ofstabilization. (I use the word “convergence” here not in itsevolutionary jargon meaning but in its dictionary meaning of “movingtoward union or uniformity”  or, in the verb form “converge”, “gradually changeso as to become similar or develop something in common”, or “cometogether from different directions so as eventually to meet” ). Interestingly, this gradual stabilization on the population levelin the long term of what previously used to vary fits much better withDarwin’s own observations on variance  than the so-called “neo-Darwinian” (the traditional)theory does.
Thus, as to the question of how a new trait comes to be shared, we see that, inthe present theory, alleles still spread in the population. At the end of a timeperiod, many alleles would have reached from the individuals in which they aroseto the entire population. These alleles represent a certain amount ofinformation that has come to be shared by all, and thus a new trait can beshared. The difference from Fisher’s additive-effect–based theory isthat each allele does not have its own phenotypic meaning and the trait does notarise in a one-at-a-time fashion by the additive accumulation of independentsteps. Instead, the meaning comes from the whole of those interacting geneticchanges taken together.
The writing of mutation provides the physical basis of convergence
Note that we have just derived the fact that evolution happens by convergence onthe level of the population as a whole from the fact that evolution is based oninteractions (that combinations of alleles, and sex, are matters of necessityfor evolution). Interestingly, though the writing of mutations was derivedindependently from the same fact, we can now see that it helps us understand thephysical basis of convergence. Thus, the two independent derivations cometogether.
I argued that the writing of mutations combines multiple pieces of informationfrom alleles at multiple loci as it puts them into one mutation—into onelocus. Now, many such writing acts take place across the genome and over thegenerations, and a new allele that is the outcome of the writing in onegeneration is part of the input into the writing in another (it is at the tip ofthe writing “funnel” in one generation and part of thefunnel’s base in another). Thus, if we take the many writing acts acrossthe genome and over the generations together, we can see that each allele in alate generation traces its origin to many alleles at different loci in asufficiently remote early generation (much like an individual in a sexualpopulation traces its origin to many ancestors in a sufficiently remote earlygeneration) (Figure 2). This means that, the farther weget in time from the early generation, the more the basis of information in theearly generation comes to be shared by individuals. In other words, thepopulation is converging, and the writing, by actually putting information fromdifferent individual combinations (and from different loci) together, providesthe basis for this convergence.
Note that the writing acts are connected in a network: they represent a flow ofinformation over the generations from many loci into one and from one to themany. This flow converts information from a state where it is unstable under theshuffling of genes to a state where it is stable under this shuffling, and theresult is the writing of a genetic network.
Obligate asexuality evolves by “breakage”
The empirical evidence fitting with the principle that sex is a matter ofnecessity for evolution provides empirical support for my theory, as discussed.Additional evidence from the topic of sex comes from the question of itsmaintenance. The reduction principle [32, 61–63]—one of the most robust findings of theoretical populationgenetics in the 20th century—shows that, in a world consistentwith the modern synthetic view, it would be hard for the sexual recombinationrate to be maintained rather than be reduced. This has been an important,negative result showing a difficulty in explaining sex in a straightforwardmanner from a traditional perspective.
However, if sex is necessary for the evolution of complex adaptation, and thisevolution happens by convergence, then there is a barrier to evolving obligateasexuality, because the closer the population gets to obligate asexuality, theless it is able to further evolve adaptively in this direction (or in anydirection). This leads to the interesting prediction that the process ofadaptive evolution toward asexuality will slowly grind to a halt and will notreach the pure asexual state. Therefore, if there are obligate asexual species,it is not long-term adaptive evolution that led to them, but some kind of“breakage” of the sexual mechanism.
This constitutes a very different approach than available so far to the questionof how sex is “maintained” despite its “costs” . I have claimed earlier that it is incorrect to discuss sex assomething whose “costs” and “benefits” determine itsexistence, because it is a matter of necessity. And here I claim that it is notactively “maintained”, but rather no substantial adaptive evolutionoccurs without it, and so obligate asexuality cannot gradually and adaptivelyevolve. It can only arise non-adaptively. The rest of this section will considerevidence and predictions regarding this point.
Note that, among vertebrates, all known unisexual lineages according to Avise  have arisen from hybridizations, which is a sudden, breakage event.Indeed, it is thought that hybridization probably disrupted the meioticoperations by reducing chromosomal homology enough to disrupt synapsis [64–67]. While this fact is in accord with the new theory as just stated, inretrospect, one may try to argue that it is consistent with the traditional onetoo, because if sex is already well established in two separate sexes, then itis hard to see how it will evolve into asex except by breakage. Let us thereforetake the battle to the flowering plants: there, most species are capable of bothselfing and outcrossing (selfing being akin to asexuality). According totraditional theory, the entire range from pure selfing to pure outcrossing isopen to them, and adaptive evolution should be able to push species all the wayto pure selfing or pure outcrossing . Indeed, from that theory, based on inbreeding considerations, thepaper that initiated the modern interest in this field predicted that pureselfing and pure outcrossing are the only stable equilibria under adaptiveevolution . But that approach gives no clear reason why there are overwhelminglymore species at the outcrossing end of the spectrum than at the selfing end [69–74]. This empirical fact supports the theory proposed here while standinguncomfortably with the traditional one. That is, according to my theory,evolution in a mixed selfing-outcrossing system is possible, but pure selfingcan only be reached by breakage. Pure selfing is rare because it requiresbreakage, which can occur only under very specific conditions. (In the unisexualvertebrates, for example, it has been argued that the hybridizing species needto be genetically close enough to produce a viable hybrid but far enough todisrupt meiosis [43, 75] and/or satisfy more specific restrictions .)
To be sure, other explanations have been offered in the plant-mating literaturefor the lack of pure asexuals (e.g., [77–84]), but the explanation proposed here is both simpler and more general.Indeed, it predicts residual outcrossing in regular biparental inbreeding animalspecies, which goes beyond hermaphrodites.
Because the new theory holds that obligate asexuality is arrived at by breakage,it predicts the lack of fine-tuned adaptations ensuring obligate asexuality. Incontrast, from traditional theory, one would expect adaptations for pure asex inlike manner as for pure sex. This suddenly renders of particular importance theempirical question of what obligate selfers are like. Pannell  mentions two notable examples of obligate selfers. One exampleinvolves the loss of males in populations or species of androdioecious animals.In these animals, such as the mangrove killifish (Kryptolebiasmarmoratus) or Caenorhabditis elegans , individuals are either male or hermaphrodites that can mate withmales but not with each other [86–88]. Loss of males in such a situation leads to obligate asexuality. Butnotice that this loss of males is not a long-term adaptive evolutionary process,but a situational event. It can occur, for example, due to the absence of malesfrom a founding population. Even if it is assumed, hypothetically, that the lossof males is due to selection favoring hermaphrodites and leading to the loss ofan allele for male determination , this is still a short-term, population dynamical process where noevolution of new adaptations or structures occurs. Only the simple loss ofpreexisting parts of the sexual machinery occurs, which does not contradict thetheory proposed here. This case can be classified as a breakage event, broadlyconstrued, and does not provide an example of gradual adaptive evolution of newstructures.
The other example concerns the cleistogamous plants, and provides a test-case forthe theory proposed here. In these plants, some flowers never open, and onlyselfing can occur within them. Most cleistogamous species have both closed andopen flowers [90–92], and it has been suggested that the closed ones provide a cheapsupply of seeds and reproductive assurance under unfavorable conditions (see  and references therein). The closed flowers have adaptivemodifications to facilitate selfing [90, 91] and, according to the present theory, these adaptive modificationscan evolve in a mixed mating state, i.e., while the species has both open andclosed flowers and reproduction both by selfing and by outcrossing occurs. Ofinterest are the ∼10% of cleistogamous species that have only closedflowers at present . If their complete lack of open flowers arose by adaptive evolution,it would refute the present theory; if it arose by breakage it would supportit.
It is conceivably possible to try to distinguish empirically whether completecleistogamous species evolved by adaptive evolution or by breakage. We know thatthe flowers of cleistogamous plants are generally sensitive to the environmentalconditions, such that they often remain closed under unfavorable environmentsand open under favorable environments [90–92]. Thus, it is possible that the loss of an environmental conditioncauses a transition from partial to complete cleistogamy without adaptiveevolution occurring. Furthermore, genetic deterioration, perhaps even due toinsufficient pollination in a mixed-mating state, could be another reason forthe failure of flowers to open. It is interesting that the failure of flowers toopen could be the reason for the switch to complete cleistogamy, because itmeans that the biological nature of this prominent case of obligate asexualitymakes an allowance for the present theory, whereas from traditional theory thereis no reason why the nature of the situation would be as it is. (The point isthe mechanistic nature of an adaptation and the breaking of it: the breaking ofthe process of the opening of a flower in a partially cleistogamous species willmake the flower not open—it will make cleistogamy complete. Compare thesituation to that of the true, complex and fine-tuned adaptation that isself-incompatibility [93, 94]. There, breakage would only lead to less, not more, of what thatadaptation provides.)
Detailed studies of the history and nature of adaptation of completecleistogamous species and other pure asexuals can serve as empirical test casesof the theory presented here. At most, this theory may be consistent with verylimited evolutionary modifications in a pure asexual state, perhaps due toresidual writing activity inherited from sexual ancestors, and those limitedmodifications may tend to show simplification and destruction of parts. But itdoes not allow for the evolution of a novel, complex adaptation in a purelyasexual state. In contrast, observations of breakage of various kinds in theevolution of pure asexuality would support the theory presented here. Note thatthe fact that the examples considered either explicitly fit or potentially fitwith breakage is consistent with the present theory and has no generalexplanation from the traditional theory.
To clarify, my theory does not argue that no degree of selfing can evolveadaptively [95, 96]. It only argues that pure asex cannot evolve in this manner. One canthink of it as follows: according to my theory, if there is an“objective” that an evolving adaptation maximizes, it is the extentof participation in the sexual population; not simply the expected number ofsurviving offspring without regard to their sexuality. Under the rightconditions, a high rate of selfing may maximize the bottom-line participation ofa lineage in the sexual population; but pure selfing fails in this objective.The conditions that have been found to be empirically associated with increasedselfing may be interpreted from this perspective without further change. Thesituation is analogous to the choice between saving versus spending in economicmodels , or the choice between investing in survival vs. reproduction in lifehistory theory . Given the right conditions, saving more can lead to an overallgreater consumption over time, and investing more in survival may lead to agreater number of surviving offspring at the bottom line. But it is not asolution to spend nothing; it is not a solution to not reproduce at all; and byanalogy, pure asex is not an outcome of adaptive evolution, according to thetheory presented here.
In sum, it can be concluded that the entire traditional conceptualization of sexneeds to be changed:
Sex is necessary for evolution, it is not a “bonus”.
Sex cannot evolve from asex and never did.
Sex (as opposed to pure asex) is not actively maintained under some cost–benefit balance as previously discussed. Rather, pure asex arises only through breakage and never through gradual adaptive evolution.
Empirically, the process of the evolution of adaptation looks likeconvergence
The process of convergence described here fits better than traditional theorywith what the evolution of adaptation looks like empirically. A single examplewill be given here, from one of the best studied cases of the phenotypicevolution of complex adaptation .
Sand wasps (Bembicinae; previously Nyssoninae) dig a long, narrow tunnel into theground at the end of which they construct a cell or a complex of cells wherethey lay their eggs and provision their larvae. Their parasites, certain groupsof flies and parasitoid wasps, aggressively seek their nests to lay their owneggs in them, for example by flying over the ground, constantly tapping the soilwith their antennae. In many sand wasp species, a behavior has evolved where thesand wasp digs one or more false burrows that extend from nearby the real nestentrance into the ground. They leave these decoys’ entrances open, and thereal entrance closedb.
Comparative ethological studies [99–101] show a range of species from primitive to advanced in this behaviorof constructing false burrows. In species more primitive in this behavior, thefalse burrows are short and unstable, and can be easily destroyed by theelements. In species more advanced in this behavior, the false burrows are longand pronounced and, in some of these species, they are actively maintained (thatis, restored if disturbed). Importantly, in the species that are more primitivein this behavior, the construction of the decoy burrows is highly variable amongindividuals—it is disorganized: it varies in terms of whether or not afalse burrow appears, how pronounced it is, where it appears spatially, when itappears in the course of nest construction, and the digging that causes it canbe scattered over time—it is unfocused. In brief, the whole operation iscrude, or “fuzzy” (but it is there as a whole). Whereas, in theadvanced species, individual variation in the behavior is far reduced, and theoverall pattern of construction is much more stable. The tunnels appearregularly and are pronounced, and they generally have a time and a place offocused construction. In a word, the operation is sharp, like clockwork; and itis far more similar among individuals. Since it is standard to infer from atransitional series of contemporary variants to the evolutionary process of onevariantc, this evidence suggests that the process ofthe evolution of this complex adaptation has been a process of convergence onthe population level—a process of stabilization—where the trait as awhole evolved from a state of high variance to a state of low variance.
This stabilization and sharpening of the trait as a whole clearly fits with theprocess of convergence predicted by the theory presented here, but it is notinherent to traditional theory. Investigators have tried to explainstabilization without invoking nontraditional theory by invoking two separatetraditional selective forces: one selecting for traits themselves, and oneselecting for the stabilization of traits, the latter being called“stabilizing” or “canalizing” selection [20, 102–104]. The view from my theory is simpler: it holds that there is but oneprocess—that of convergence and stabilization on the population level.There is no need for a separate force of traditional selection forstabilization. Stabilization is an automatic concomitant of the processdescribed here.
A more detailed look into the new theory
The writing phenotype evolves, and the writing and performing phenotypesshare alleles
I argued that the writing of mutations is an organic process that belongs to theorganism. Let us call it henceforth the “writing phenotype”. Whiletraditional theory has had only one kind of phenotype, which we will call herethe “performing phenotype”, here we have two: the writing and theperforming phenotypes. Let us now derive further theoretical points about howthey work.
First, if the writing phenotype is like the performing phenotype, being coded bygenes and alleles, then, like the performing phenotype, it must also beevolving.
Second, the writing and the performing phenotypes are obviously different. Oneimplements genetic change, and the other is responsible for survival andreproduction. But although they are different, we can quickly see that they mustbe sharing alleles, as will now be explained.
As just noted, the writing evolves, and we can now add that it needs to evolveunder the influence of natural selection. Otherwise, how could it ever getfeedback from the outside world, and how could it be different from randommutation indefinitely, when the performing phenotype clearly changes vastlythrough the eons in accord with the environment? Without a source of feedback onthis outside world and the organism ever-changing with it, mutation musteventually become accidental to the organism. (With no flow of information fromB to A, and no predetermination of both B and A by C, A must be random toB).
Now, by definition, the effect of selection is registered in the frequencies ofperforming alleles. Therefore, if the writing phenotype evolves under theinfluence of selection, it means that performing alleles influence the writingphenotype. If they influence the writing phenotype they participate in thewriting. Therefore, performing alleles are also writing alleles.
There is another way to derive the same point. The writing solves the problemthat combinations under selection must have an effect. To solve this problem, itmust be that a combination of performing alleles at different loci is taken andan allele is derived from it. This means that performing alleles are inputs intothe writing—they affect the writing operation. But if they affect thewriting operation, they are writing alleles too.
Thus from both directions we see that the writing phenotype and the performingphenotype share alleles. But the alleles do not mean the same thing to them. Thealleles’ full meaning is generated by the way they modify thetaxonomically-shared part of the writing phenotype and the taxonomically-sharedpart of the performing phenotype respectively, which are different.
We obtain the following picture: Alleles participate in the writing of alleles,and alleles are selected. The writing performs an operation, whose inputs arealleles and whose output is an allele. The writing itself always evolves.
This concise statement is what we are led to, and it deserves much reflection. Aconcrete example will help to explain the idea that the same alleles, andtherefore the same genes, can participate pleiotropically in both the writingand performing phenotypes. According to the theory proposed here, theTRIM5 and CypA genes, which participate in the performingphenotype, also participated in complex genetic activity in the germline thateventually led to their fusion [105, 106], indeed to their independent fusion in different monkey lineages [105, 107–112].
It necessarily follows from the points above that the writing always evolvesalong with the evolution of the adaptation. It accumulates information fromselection, and the alleles that it generates are specific to the evolutionarytimes. But it is never “ahead of” selection—it never takesupon itself the forbidden role of producing something known in advance toincrease fitness—it never replaces natural selection in its role.
We can further illuminate the nature of the writing phenotype by contrasting itwith “cranes” . Those are hypothesized phenotypes that are pre-evolved, generic andrepetitive devices that supposedly speed up evolution based on a traditional,ns/rm core. An example of a crane is presumably given by the hypothesis that theSOS response system in bacteria induces temporary general hypermutability inresponse to stress, and that this general hypermutability speeds up ns/rm-basedevolution and thus hastens the arrival of a solution at a time of need (reviewedand criticized in )d. “Cranes” such as the hypothesizedtemporary general hypermutability system would be of long-term evolutionarybenefit but are not themselves evolving along with any specific adaptationevolving at present, and therefore are not tailored to any particularadaptation, and still rely on traditional ns/rm to do the “work” ofevolving an adaptation. They are thus “add-ons” to the traditionalperspective, and they are not easy to justify from that perspective, because thebenefit they bring is a long-term, evolutionary one. In contrast, the writingphenotype is not generic. It evolves along with the adaptation. It is thereforespecific to the adaptation and the evolutionary times, and only thanks to itsevolution the adaptation can evolve. Thus, the theory presented hereemphatically agrees with Koonin’s conclusion that evolvability can evolve , however, it proposes that evolvability reflects the evolution of thewriting phenotype. This is a far more direct explanation for evolvability thanhigh-level selection.
Interestingly, Wagner has already noted that one of the most interesting thingsthat transposable elements demonstrate vividly is that the options available forgenetic evolutionary change are specific to the evolutionary times . In a sense, I am generalizing Wagner’s deep insight here fromtransposable elements to the entire writing phenotype.
We can conclude that the writing phenotype and the performing phenotype evolvetogether. Indeed, their coevolution explains how they relate to each othersyntactically—they never “lose track” of each other. Nonrandommutation is neither a Lamarckian-transmissionist “seer” that usurpsthe role of natural selection nor an “add-on” on top of traditionalselection. It is a continually evolving system that sits at the heart of theadaptive evolutionary process.
The new theory predicts that genetic activity implementing the writing ofmutations exists in the germline
Several easy predictions now follow from the above.
First, for the writing of mutations to have an evolutionary effect, it obviouslyneeds to take place in the germline. This means that there must be biochemicalactivity in the germline responsible for the writing of mutation. To continuethe example from the previous section, it has been noted that CypA ishighly expressed in the germline, and that this may have contributed to theindependent arising of the TRIM5–CypA gene fusion in at least twodifferent monkey lineages [106, 116]. While from a traditional perspective we could stop the intellectualinquiry here, and assume that this germline activity is simply an accidentalsituation, the theory proposed here considers this situation to be the result ofa long-term evolution of the writing phenotype, essential for the long-termevolution of the performing phenotype (they coevolve, as stated). In otherwords, we are dealing here not with accidental boundary conditions, but withevolved writing activity.
Second, according to my theory, alleles from different loci must interact in thedetermination of mutation. Thus, mutation determination is complex—genesmust interact in the germline in the determination of mutation, enabling thefact that alleles interact. The determination of mutations cannot be exclusivelysimple, single-locus based.
Third, because the performing and the writing phenotypes are different, but theyshare alleles (meaning, the same genetic difference that plays a role in theperforming phenotype also plays a role in the writing phenotype, though thisgenetic difference has different phenotypic meanings in these two phenotypes),the same alleles will participate in biochemical activities in both germline andsoma, but those activities will be different. Hence, genetic activity observedin the germline should not be immediately assumed to be serving the performingphenotype of the germline—it could be writing activity. Furthermore, thisactivity may involve somatic performance genes.
Genetic evolutionary trends exist on all timescales
The writing phenotype can be understood better by analogy to the performingphenotype. Four-legged animals use their legs for locomotion by pressing themagainst the ground. In this general sense, quadrupeds are all similar. But thisgeneral description is filled with detail as we move to finer taxonomic levels:horses gallop, rabbits hop. The details continue to be filled as we get to theindividual level. Individuals can have shorter or longer limbs, differentproportions of fore and hind limbs, different details of their muscularactivation, etc. These individual-level details, though small in comparison tothe general mode of locomotion, are very important—they are theindividual-level variation that is the basis of natural selection. Thus, notethat there is a spectrum of contributions to the performing phenotype, includinga basis that is persistent and slowly changing, and is generally defined, aswell as ever increasing detail that distinguishes between ever finer taxonomicentities and evolves on ever shorter timescales.
Now, I argued that the writing phenotype is an evolving phenotype, and thereforehas the same structure as the performing phenotype. In light of the above, thismeans that there are contributions to the writing phenotype from all taxonomiclevels. The more widely shared these contributions are, the more generally theyare defined, the slower they change, and the longer the timescale on which theypersistently act. Accordingly, at the deep end of this spectrum we find that allorganisms have a genetic code, whose characteristics begin to define the rangeof possible mutations in a very general sense. Further along the spectrum wefind that different taxonomic groups have somewhat different methods of geneduplication and different transposable elements, for example, further delimitingthe range of possible mutations. And at the far end of this spectrum, writingevents in a particular individual are defined in a perfectly concretemanner—these are the particular mutations occurring in the individual.According to the new theory, the details on the individual level are important:they are nonrandom (because mutation is nonrandom), and they enableinteraction-based evolution by natural selection.
Note that, whether we take the traditional standpoint or the new standpoint, wemust accept that there are ever finer specifications of the range of possiblemutations. But while the traditional theory must draw a line at some point andsay that “up to this point the machinery defines the range of mutations,and beyond this point mutation is random”, the theory proposed hererefuses to draw such a line, and completes the spectrum by saying that mutationis determined by the writing phenotype all the way up to the individual level,and is individual-specific, just like the performing phenotype is. We may callthis “individually determined mutation”.
Note also that the line drawn by traditional theory is arbitrary. From atraditional standpoint, we start by assuming that there is a genetic code. Thenwe add that there is replication or other error, hence point mutation. Toaccount for new genes, needed for the evolution of complexity, it was added thatwhole gene duplication exists . But now we must assume that we are lucky enough that the geneticsystem is constructed in such way that gene duplication exists, but that thisextraordinarily important machinery of gene duplication  must be applied here and there by chance. There is theoreticalarbitrariness in saying that, up to here the range of mutation is constrained bythe system, and beyond here it is not constrained at all, when no reason isgiven for why such a dividing line should be placed at one point rather thananother. Indeed, the more we study the situation empirically, the more we seefiner determination of the range of mutations. Gene duplication is stronglyinfluenced by the location of segmental duplications/low copy repeats (see thesection “Evidence from and predictions for molecular evolution”);the location of segmental duplications/low copy repeats is strongly influencedby the location of transposable elements (see the section “Evidence fromand predictions for molecular evolution”); and the location oftransposable elements is strongly influenced by various sequencecharacteristics. The dividing line between “mechanistic” and“random” keeps being pushed back. Here I argue that there is no suchline. Any line would be arbitrary. The removal of this arbitrary line is anindependent point of entry into the new theory, because by removing it, weimmediately get to individually determined mutation.
Now consider the existence of the genetic code; the fact that the “errorrate” in replication supposed under the random mutation view is not toohigh and not too low, so that it allowed evolution; the fact that the geneticsystem is structured such that whole gene duplication, necessary for long-termevolution, is possible, etc. From the traditional perspective, we are lucky thatall these things exist, so that evolution as we know it is possible. Theexistence of these phenomena cannot be easily explained under the traditionaltheory, because from that theory we normally take them as given and do not beginto think about evolution before we imagine them in place (we do not normallythink of them as evolving) (see  for an opposing, nontraditional view, consistent with the presentwork). We cannot say that they are explained by the benefit they bring toevolution in the long term, because traditional theory can only explain theevolution of traits based on short-term, individual-level advantage [6, 16]. Indeed, these phenomena are rather parts of the evolutionary“infrastructure”. Since we cannot explain their existence by thetraditional process, from the traditional view we can only say that theyappeared by chance or by an unknown process outside of the theory. This leavesus with a number of fundamental biological phenomena which enable evolution butare not explained by the traditional evolutionary process.
One possibility is to apply high-level selection to this gross problem [8, 119]. However, the whole situation is seen differently from theperspective of the theory presented here. Even though the theory presented here,like the traditional one, cannot explain in detail how these phenomena arose andtheir current form, the theory presented here inherently includes a mechanismthat supports their existence and evolution. Namely, mutations are effected by awriting phenotype. Since this phenotype obeys the same rules of biologicalstructure as the performing phenotype, as explained above, it has long-termenabling effects on evolution (in addition to short-term ones). This succinctlyprovides a framework for understanding these phenomena’s long-term effecton evolution, which the traditional theory does not. That is, these phenomenadefine the range of mutations, and are part of the writing phenotype. Thisframework is entirely different from both sides of the levels-of-selectiondebate.
An additional, important prediction can now be made. I argued that the morewidely-shared aspects of the writing phenotype are more generally defined andmore slowly changing, and therefore act more persistently on a longer timescale.If a general writing trait has been in existence for a long period of time, onlyslowly changing, then it has been guiding the writing activity during thatperiod of time in a somewhat persistent manner, giving rise to some degree of“directionality” in genetic evolution. I predict that thisdirectionality will be observed in the form of hitherto unexplained long-termgenetic evolutionary trends. These trends do not define the evolutionary changescompletely. They are rather filled with detail at finer taxonomic scales. Andalthough they constitute a certain amount of internal guiding to geneticevolution, this internal guiding does not work by itself, but only together withnatural selection, and is in fact itself the result of past selection andwriting.
Context-dependent selection participates in the formation of the phenotypicmeaning of an allele
When selection operates on interactions—meaning it is contextdependent—then the change in the frequency of an allele is inconsistent inits direction, because this change depends on the context of other alleles,which is itself changing at the same time. The dynamics of allele frequenciesare nonlinear.
Context-dependent selection has two interesting consequences. The first concernsthe phenotypic meaning of an allele.
In the traditional mindset, we think of effective selection as acting mostly onindependent alleles. To be precise, random mutation arises that interacts withthe fixed genetic background but not with concomitant alleles at other loci, andin that interaction with the fixed genetic background it has its own phenotypicmeaning that is complete at the moment of the arising of this mutation and thatis unchanging throughout the period of its selection. All that remains fornatural selection to do is to check whether this mutation is “good”or “bad” in and of itself. Thus, in the random mutation case,selection is an external judge of a phenotypic meaning formed at random beforeselection takes place.
In stark contrast, under context-dependent selection, the phenotypic meaning of aspreading allele (an allele whose frequency is increasing, albeitinconsistently) depends on which other alleles are spreading. But which otheralleles are spreading is affected by selection on interactions. Therefore,natural selection affects the phenotypic meaning of an allele—itparticipates in forming this meaning. Thus, according to my theory,selection is not an external judge of a pre-made phenotypic meaning, but is anactive participant in the formation of it. This alone means that the phenotypicmeaning of a mutation is not random to natural selection, because informationfrom natural selection is already in it. Selection is inside, not outside, theprocess of formation of the phenotypic meaning of an allele.
At the beginning of this paper we found that the need for selection oninteractions to have an effect is answered by the writing of mutations—bygenetic change having a mechanistic and organic basis, and in that sense beingnonrandom. Now we have just derived from selection on interactions thatselection participates in the formation of the phenotypic meaning of an allele,which shows that the phenotypic meaning of genetic change is not random.Interestingly, these two points naturally come together, defining nonrandommutation from above and below.
What appears neutral under the assumption of additive alleles can actuallyexperience selection on interactions
The second point of interest that follows from context-dependent selectionconcerns the neutral theory. Haldane’s  calculation of the “cost of natural selection” was animportant reason behind the advent of the neutral theory . This calculation had put a severe limit on the rate of substitutionthat could be driven by traditional natural selection, and the actual rate ofsubstitution  as well as the amount of present genetic variation [1, 2] later discovered vastly exceeded this expectation . Hence Kimura proposed that the vast majority of mutations are simplynot under selection and just drift to either fixation or extinction .
However, the theory presented here holds that selection operates on interactions;and since Haldane’s calculation was based on traditional assumptions, hereit simply does not apply. Moreover, when selection acts on interactions, allelesexhibit inconsistent change in frequency, which may appear to us as drift. Inother words, alleles that appear to be drifting may actually be experiencingselection on interactions. What looks neutral through the lens of thetraditional, additive-effect–based theory may not be neutral from aselection-on-interactions view. This does not mean that traditional drift cannotexist in addition to selection on interactions, however, it does suggest thatso-called “neutral” matter can be subject to selection and thus hasa vast adaptive potential.
Evidence from and predictions for molecular evolution
We may categorize mutation into two high level categories: rearrangement mutation andpoint mutation. I will discuss them below in turn.
Rearrangement mutation is nonrandom
It is now clear that the genome is highly dynamic, involving a great deal ofrearrangement—where sequences are duplicated, deleted, inserted, invertedor translocated . This ongoing rearrangement is a new reality in molecularbiology—exposed by modern technologies and unknown at the foundation ofthe evolutionary synthesis. This rearrangement was first thought to be random,but it is now clear that it is locus-specific, that it is effected by biologicalmechanisms, and that these mechanisms are guided to their places of action byDNA sequences [123–125].
Four main categories of mechanisms of rearrangement are: non-allelic homologousrecombination (NAHR), non-homologous end-joining (NHEJ), replication-basedmechanisms (RBMs) and transposition.
NAHR  occurs when sufficiently long non-allelic sequences of high homologyalign and cross over. When this crossing over is between homologous chromosomesor sister chromatids, the result is a duplication and/or deletion of thesequences between the non-allelic homologous sequences and of one of thenon-allelic sequences. If it is between repeats on the same strand that align asthe strand coils, there are two options: if the repeats are in directorientation, the result is a deletion (transposable elements are often preciselyexcised in this way ); if they are inverted, the result is an inversion. And if thiscrossing over occurs between repeats on non-homologous chromosomes, the resultis a translocation. Notably, duplications, deletions, inversions andtranslocations of whole genes would not have been possible without a mechanismto enact them, and there is elegance in the mechanism of homology andrecombination that is able to produce quite different outcomes based ondifferent parameters of the situation, and that indeed is also the basis ofsexual recombination.
The regions of sufficient homology are usually provided by low copy repeats(LCRs) or segmental duplications (SDs)—terms that are usedinterchangeably, though defined originally independently using somewhatdifferent parameters (SDs were defined as segments ≥ 1 kb in size and≥ 90% sequence identity , and LCRs were defined as intrachromosomal duplications ≥ 10 kbin size and ≥ 97% in sequence identity ). It is thought that ∼5% of the human genome consists ofLCRs/SDs, and they are particularly prevalent in pericentromeric andsubtelomeric regions (reviewed in ). NAHR can also occur between tandem duplications, and more rarelybetween repetitive sequences, which are shorter and much more numerous incomparison to LCRs/SDs (transposable elements constitute about half of the humangenome). In this case, the repetitive sequences are expected to be closer toeach other as compared to the LCRs/SDs that cause NAHR, and the rearrangementstend to be smaller [123, 124].
NAHR is not random. Not only does it require the biological mechanisms ofcrossing over to be implemented, the LCRs/SDs specify the locations where itusually takes place, and it is often recurrent . Indeed, the breakpoints are further specified within the LCRs/SDs,where they are clustered in narrow hotspots, often nearby DNA sequences such asdirect and inverted repeats, which form hairpins, cruciforms and other non-B DNAstructures, known to induce double-strand breaks (DSBs) involving enzymaticprocesses [123, 130–132]. Their precise locations can be very close to meiotic recombinationhotspots , implying the sharing of features with meiotic recombination hotspots  (reviewed in ), which are known to be associated with consensus sequences and more(reviewed in ; to be discussed later). Furthermore, the non-allelic sequencescausing NAHR have functional relatedness: they share long sequence homology, andwe know that sequence defines function; and the recombining LCRs/SDs need to besufficiently close to each other (the more so the smaller they are) (reviewed in [123–125]), either by simply being nearby on the chromosome or because thethree-dimensional structure of the DNA brings them together from regions thatare remote in two dimensions, and we know that closeness in two dimensions aswell as in three dimensions is to some degree related to function .
Non-homologous end-joining (NHEJ) is able to recognize two ends of DNA (doublestranded), modify them and join them together. If the two ends come from twodistant points rather than one, a deletion or inversion occurs ; and if they come from one point, but double-strand break homologousrepair is performed before the end joining, it can lead to duplication [137, 138]. NHEJ is nonrandom: it occurs in hotspots (e.g., ), though they do not cluster as sharply as in NAHR . These hotspots are often within repetitive elements such as LINE andAlu and near sequence motifs that can curve DNA and cause DSBs, andone of the breakpoints in a rearrangement event is often found within a LCR(though the LCR is not necessary for homology in this case) (see  and references therein). Thus, local genome architecture influencesthe occurrence of these events.
Complex rearrangement events by mutational mechanisms are also possible, andreplication-based mechanisms (RBMs) have been proposed that may cause suchevents. In general, RBMs include replication slippage (RS; ), serial replication slippage (SRS; ), fork stalling and template switching (FoSTeS; ) and microhomology-mediated break-induced replication (MMBIR; ). In replication slippage , microhomology between short repeats allows the nascent strand tomove a few base-pairs forward or backward on the template strand and continuereplication from there, which causes a short duplication if it moves backward ora short deletion if it moves forward. In serial replication slippage , multiple forward and backward movements within a replication forkcan occur, leading to a small but complex rearrangement event. Invasion of a newtemplate due to microhomology between inverted repeats can also lead tosynthesis of an inverted segment. The FoSTeS  and MMBIR  models propose that the lagging strand from one replication fork candisengage and invade another fork that is probably close to it in 3-D spacebased on microhomology and continue replication there, leading to deletions,duplications, inversions and/or translocations based on the parameters of thesituation. Serial disengagements and invasions can lead to complex rearrangementevents, as in the single-fork case, but this time they involve larger sizes ofsegments and larger distances between segments. These mechanisms do not actrandomly, as they involve microhomology and non-B DNA structures [123, 130, 143, 144]. Clear groupings of breakpoints have been observed in some cases thathave been attributed to FoSTeS/MMBIR [141, 145]. And closeness in 3-D between forks may suggest related function asmentioned . Although the mechanistic aspects of NHEJ and RBMs have beenilluminated by studies of genomic disorders, these mechanisms may account forthe majority of non-pathological copy number variation [142, 146, 147].
Rearrangement by transposition is the fourth main category of rearrangementchanges. It occurs when transposable elements (TEs) move themselves as well asother pieces of genetic material incorporated in them—DNA transposons andinsertion sequences with the help of the enzyme transposase (with or withoutmaintaining a copy at the source), and retroelements with the help of the enzymereverse transcriptase and integration into the DNA (always maintaining a copy atthe source) [17, 148]. As will be discussed later, one school of thought, associated withthe traditional framework, has held that TEs are “selfishelements”—parasites of the genome—and that occasionally theyare coopted by chance for other functions [5, 6, 149]. But we will see later how TEs can have the appearance of selfishelements yet be an inherent part of the mutational mechanisms that serve theevolution of the organism. Indeed, they donate every kind of functional element,including promoters, enhancers, splice sites, coding sequences and sequencemotifs, and have an extraordinarily wide and deep range of evolutionaryinfluences [15, 17, 148, 150, 151]. TE movement is not random. They have a wide range of preferences fortarget sites, some showing affinity to certain chromosomes, others to locidistinguished by certain sequences, others to loci of a particular nucleotidecomposition, etc. [17, 148]. It is also thought that TEs are involved in the formation ofSDs/LCRs discussed before. Alus have been observed at the end-points ofnearly 30% of the LCRs/SDs in humans [152, 153], implying Alu-based homology is involved in theirproliferation.
Korbel et al.  and Kidd et al.  systematically analyzed structural (rearrangement) variationbreakpoints in the human genome, and have found that almost all breakpointsanalyzed have signatures of one of the four mechanisms above. As previousauthors already noted [123, 124], this means that the vast majority of rearrangements in humans aredue to biological mechanisms whose action is directed by DNA sequence andstructure and are therefore not random. We need only to add that this sequenceand structure is itself evolving.
Point mutation is nonrandom
We discussed rearrangement mutation above. The other general category of mutationis point mutation, nowadays referring to a single nucleotide change from one ofthe four kinds of nucleotide to another. Naturally, we used to think that thesechanges are random, but cutting edge research in molecular biology is showingthat, as in the case of rearrangement mutation, a great deal of point mutationis nonrandom.
Point mutations are not uniformly distributed at random across the genome, butinstead the mutation rate per locus varies across the genome on all scales, fromthe single-base resolution through the gene scale and mega-base scale to thechromosome scale .
Many point mutations in humans are due to a change in the cytosine of CpGdinucleotides (dinucleotides where cytosine is adjacent to guanine in the5’-CpG-3’ orientation)  that are spread out over the genome outside of the relatively narrowCpG-rich regions (themselves not experiencing this high rate of mutation; [157–159]). This change is due to methylation of the cytosine, which, in thisCpG context, is the predominant target of DNA methylation in vertebrates [159, 160]. The methylation is enzymatic and controlled by evolved machinery,and following deamination it leads to a C→T mutation at a very high rate(reviewed in ). This high rate of transition is either because of chemicalinstability of the methylated cytosine, or due to an enzymatic process yet to bediscovered . However, we already know that this kind of mutation is nonrandombecause of the biological marking of the cytosine, which causes the mutation oneway or the other. Notably, ∼24% of all point mutations in humans are dueto this mutational process .
In addition, ∼18% of the human genome is within 10 bp of a CpG, and an∼50% increase in single nucleotide polymorphisms (SNPs) has been observedwithin this distance in methylated regions . It has been proposed that deamination of the methylated cytosine isfollowed by “error-prone repair” which not only establishes theC→T mutation but also gives rise to point mutations in nearby bases at thesame time [162–164] (but of course, “error-prone repair” may also be called a“change-inducing mechanism”).
Other short sequences also exist that have a substantial association withmutation rate. The sequences ATTG and ATAG have a mutation rate of T→C inthe second position that is 3.5- and 3.3-fold higher, respectively, than thegenome-wide average T→C mutation rate, and ACAA has a mutation rate ofA→C in the first position that is 3.4-fold higher than the averageA→C mutation rate, in humans  (compare to a 5.1-fold excess of C→T mutations in CpGs in thesedata ). The average mutation rates of other short sequences also differsignificantly amongst each other, and farther nucleotides also have asignificant but ever smaller effect (reviewed in ).
In addition, there are loci at the single-base resolution that undergo pointmutation preferentially even though no simple sequences have been found yet inthese loci [166–168]. We know of these loci from studies of coincident SNPs (cSNPs), whereSNPs are observed in the same locations in related species [166–169] (understandably, they also tend to exist in the same loci wheresingle nucleotide substitutions are observed in between-species comparisons; [166–168, 170–172]). It has been said that traditional natural selection does not appearto explain these coincidences in the location of variance , and so we know that these mutations are guided, though we do notknow how. According to Hodgkinson and Eyre-Walker , this part of the variance in the human mutation rate across locithat is accounted for by cSNPs unexplained by simple context, called“cryptic variance” , is as large as that of CpG mutations (the latter alone involving∼24% of all point mutations in humans, as said). Thus, we already see thata large percentage of the total variance in the per-locus mutation rate inhumans is accounted for by cSNPs and CpG mutations, two obviously nonrandomprocesses.
It is also worthwhile mentioning that there is a strong association betweenmeiotic recombination hotspots and mutation hotspots [173, 174]. Meiotic recombination hotspots move rather quickly duringevolution—i.e., they are not conserved between humans andchimpanzees—but they remain within a certain region for longer periods oftime (in other words, they move quickly on the single-base scale but more slowlyon the Mb scale; [175, 176]). Within these regions, substitution rates are elevated [173, 176]. Importantly, meiotic recombination hotspots are clearly nonrandom:their locations involve DNA sequence motifs and, according to Wahls and Davidson , are determined by the combinatorial effect of the binding ofmultiple transcription factors at multiple transcription factor binding sites.This complex determination of meiotic recombination locations, interesting inand of itself, will be discussed later, but in the present context it impliesthat the point mutations co-localized with recombination hotspots are alsononrandom, as their location is biologically determined (even without furtherdirect evidence speaking to these mutations, we know that their rates could notbe randomly elevated particularly at those places where recombination isnonrandomly placed).
Finally, as discussed in the previous section, rearrangement mutations arenonrandom, and point mutations and rearrangement mutations are in generalrelated. The rate of point mutation is substantially increased near insertionsand deletions (reviewed in ).
Taking together the predictive power of simple contexts, of cryptic variance, ofthe recombination–point mutation association, and of the associationbetween the locations of rearrangement mutations and point mutations, we alreadyknow that much of point mutation is nonrandom and under biological control.
The traditional theory leads to paradoxes when facing new knowledge frommolecular biology
Traditionally, we had been thinking that mutation was random and caused byexternal agents such as UV radiation or toxic chemicals, or by “copyingerrors”. But we now see that a great extent of genetic evolutionary changeis under biological control. Applying traditional thinking to this observation,we still say that all of this mutational activity must ultimately be accidentalto the organism: that the biological mechanisms cause it by making errors asthey try to restore the previous genetic state or by failing to recognize thatstate following an accidental disruption. But this view leads to paradoxes.
One such paradox is that mutation hotspots are particularly concentrated in zonesof adaptive evolution. This is indicated in several ways. First, genes whoseproducts interact rather directly at the molecular level with the externalenvironment, like chemo-sensory perception genes, immune and host-defense genes,and metabolism and detoxification genes, display a high concentration ofmutation hotspots [129, 177–179], and to some degree we have independent evolutionary-ecologicalreasons to expect to see much adaptive evolution in those genes . Second, a high dN/dS ratio (a high ratio of non-synonymoussubstitutions per non-synonymous site to synonymous substitutions per synonymoussite) has been observed in such genes [181, 182], an observation commonly used as an indicator that genes are underpressure for change. Thus, mutation hotspots are concentrated in zones that, forboth reasons just mentioned, are expected to be under pressure for change.Indeed, these mutation hotspots are not just there and disassociated from theadaptive evolution of these genes, but rather appear to play an active role inthis adaptive evolution, as demonstrated, for example, by the defensin geneclusters . Third, evidence arising from detailed studies of particular cases,such as evidence of hypermutability in toxin-encoding genes in snails of thegenus Conus[183, 184] and evidence of hypermutability of HoxA13a in zebrafish andrelated taxa (Cypriniformes) , is consistent with a connection between mutation hotspot locationsand adaptive evolution.
But how did mutation hotspots come to be concentrated where they are needed? Thetraditional view cannot explain this association well, because this viewrequires an immediate benefit for the spread of a mutation based on an advantagethat it supposedly brings in and of itself, whereas the “benefit”from the presence of these hotspots is due to changes that they bring in theevolutionary long-term, and which are part of the evolution of the population asa whole. What the association rather means is that the biological control ofmutations is not incidental to adaptive evolution.
The association between zones of adaptive evolution and genetic disease canbe understood as evolutionary friction points between the writing andselection
It is not the intention of the new theory to suggest that, since mutationhotspots are placed in zones of adaptive evolution, they can“outguess” natural selection. It rather suggests that mutationhotspots are positioned in a long-term evolutionary process that is constantlyreceiving feedback from natural selection, and they never take on the role ofnatural selection. This point is underscored by the fact that mutation inmutation hotspots often leads to recurrent genetic disease [186, 187].
The importance of recurrent genetic disease has been becoming clear in the lastdecade, and is a bit of a curiosity from the traditional perspective. In fact,there appears to be a triple association between mutation hotspots, zones ofadaptive evolution, and genetic disease (e.g., [129, 131, 180, 181, 188, 189]). The new theory offers the following view on this situation.Recurrent genetic disease represents evolutionary friction points, where thepressure for change that comes from the writing phenotype and its mutationhotspots—which, according to the new theory, belong to writing mechanismsthat have been evolving in the long-term—clashes with the pressure forimmediate performance of the focal loci in the context of the current state ofthe organism. That is, improvement in a complex system is hard to achieve, andit takes an evolutionary “negotiation” process between writing andselection pressures, until either the focal writing trend readjusts and pushesin a new direction, or other loci change and remove the block imposed by naturalselection on this writing trend and allow it to persist in its direction. Thisgives us a way of understanding the triple association just mentioned. Incontrast, from the traditional perspective, the long-term persistence inparticular places of mutation hotspots that are enabling of adaptive evolutionin the long-term yet are costly in terms of recurrent genetic disease in theshort-term has no equally intuitive explanation.
We can conclude from the molecular biological evidence so far that the biologicalcontrol of mutation is plainly fitting with the theory presented here, and infact connects the two grand phenomena of sex and nonrandom mutation; but itleads to paradoxes from the traditional one.
The new theory predicts that the determination of mutation is complex, andthis prediction is confirmed
Not all of the variance in the mutation rate across loci is predicted by simplecontext—i.e., by a simple consensus sequence that is present in everylocus where the mutation happens. And the presence of one of the simpleconsensus sequences in some locus does not in and of itself guarantee a mutationin that locus, it only increases the likelihood that we will see a mutationthere. As mentioned, a substantial amount of the variance in the mutation rateacross loci is cryptic .
This is in accordance with the theory proposed here. If all point mutations werecompletely determined by simple local context, this would not allow alleles fromdifferent loci to be involved in the determination of mutations, because eachlocus in this case would specify its own mutation by itself; whereas, crypticvariance means that mutation is nonrandom, yet local allelic information doesnot completely determine it, implying that allelic information from other lociparticipates in its determination, exactly as predicted by the theory proposedhere.
Interestingly, Wahls and Davidson  argued that simple consensus sequences are not sufficient for thecomplete determination of meiotic recombination hotspots. Rather, the meioticrecombinational activity is determined combinatorially by the binding ofmultiple transcription factors that interact with each other . In addition, we know that meiotic recombination hotspots are alsomutation hotspots, as said. Combining these two facts, we see that, at least inthe case of this type of hotspot, the location of mutation is determinedcombinatorially by the binding of multiple interacting factors, much like thelocation of transcription is determined in the performing phenotype (writingfunction is determined much like performing function). This enables alleles frommultiple loci to interact in the writing. Thus, recombination–mutationhotspots as described by Wahls and Davidson are a living example of theindividually determined mutation predicted by the new theory—the writingphenotype.
Evidence of “divergent parallelism” is in accord with the newtheory
Cryptic variance relates to another interesting point. In The Origin ofSpecies , an observation of high generality is emphasized, according to whichtraits that have been experiencing adaptive evolution in recent evolutionarytime are also the ones that continue to vary substantially between individualsat present. It is interesting that this very general observation, so importantto Darwin, has not had an obvious place within the traditional (neo-Darwinian)theory: according to traditional theory, which is based on random mutation,variation is supposed to hit where it hits, selection is supposed to act whereit acts, and there should be no relation between the two.
Now, molecular evolutionary studies, including those investigating crypticvariance, produce evidence that precisely mirror Darwin’s observation atthe molecular level: loci of substitutions between related species (recentevolution) are associated with SNPs (present variance) [166–168, 170–172] and with regions of adaptive evolution .
From Darwin’s observation alone one could have inferred that there is somelong-term persistence in what evolves and therefore some directionality inevolutionary change. Furthermore, since there is persistence in what varies,there is parallelism in what varies, and this parallelism cannot be explained bysimilarity of selection pressures experienced by related speciesalone—there must be also similarity in the guiding of the variance.Furthermore, since separate species cannot forever evolve in exactly the sameway, but must gradually become more and more different, Darwin’s pointimplies that evolution proceeds by what might be called “divergentparallelism”: something guides the variance but it gradually evolves. Allof this is exactly in accord with the theory proposed here: the writingphenotype evolves.
Indeed, while presenting their results on cryptic variation, Seplyarskiy et al.  noted that while a substitution in a certain locus in gorillaincreases the chance of an SNP in the corresponding locus in humans by about30%, the substitutions that have occurred on the path connecting two species oflemurs show practically no correlation with the locations of SNPs in humans.Consistent with this example, they suggested that “[p]erhaps the patternsof the cryptic variation of the mutation rate are subject to evolution and,thus, become more and more different in more and more distant genomes”.This suggestion of divergent parallelism is precisely in accord with the newtheory.
A new interpretation for recent findings
In light of the new theory proposed here, new interpretations become possible forsome recently discovered puzzling phenomena. I will discuss under this headingde novo genes, epistatic capture, the interpretation of TEs and ofjunk DNA, transcriptional promiscuity and the unusual genetic activity in spermcells.
De novo gene evolution may be subject to indirect natural selectionthrough the writing phenotype
All would agree that random, accidental mutation cannot be expected tosuddenly produce out of thin air a large and complex beneficial change.Therefore, the Fisherian theory of evolution , which has been so important in our understanding of adaptiveevolution, has a basic idea behind it: it is to minimize the amount of“useful work” that random mutation can supposedly do in any onemutational step. The idea, then, is to let natural selection check eachmutation and let through only the useful ones, and thus gradually accumulatethe small, additive effects of many such mutations into a substantialphenotypic change .
One question that arises, then, is how a new gene emerges. A gene is acomplex entity that cannot arise out of thin air. It includes hundreds orthousands of bases of DNA, including both regulatory signals and RNA- orprotein-coding sequences, and it cannot be active and subject to traditionalnatural selection until many of those bases are in place. For this reason,it was rightly suggested already in the 1930s that new genes originate bywhole gene duplication : First, a previously complete and active gene is duplicated by asingle “duplication mutation” all at once along with itsregulatory and coding sequences. Then, point mutations may graduallyaccumulate in one or both of the copies, eventually making themsubstantially different from each other and thus leading to the arising of a“new gene” . In line with this, Jacob argued in 1977 that it is impossible toget a gene out of nothing —a gene always starts by drawing on the pre-existing. Theword “alchemy”  may be attached to this impossibility of complexity out of thinair.
One deep philosophical problem with gene duplication from the traditionaltheory has already been discussed: it is that we are lucky to have themechanisms that enable duplication mutations, indeed the mechanismsdiscussed earlier, because they are necessary for long-term evolution, buttheir existence is not easily explained by the traditional theory. But thereis another problem, raised by recent evidence.
Since 2006, results have accumulated showing the existence of a completesequence of an active gene in one or a few closely-related species, and theexistence of substantially similar (syntenic) sequences in multiple relatedspecies that are incomplete and are missing some of the regulatory signalsand coding sequences that make the gene what it is in the species where itis active [9–13, 193–201]. Because of the nature of the phylogenies involved, it has beeninferred (and on this all agree) that the common ancestor of those sequenceswas nonfunctional (because if it were functional, a larger number ofindependent evolutionary events of repeated dysfunctionalization in multiplespecies would need to be assumed); and thus, in the course ofevolution—in fact in the course of millions of years ofevolution—signals and coding sequence elements have been graduallyadded until the sequence has become an active gene in one or a few closelyrelated focal species. But this means that, in these cases, referred to ascases of “de novo” genes, a gene has been created notfrom copying and gradual change of a previously complete gene, but in a waythat appears, from traditional theory, out of nothing—out of“junk” . That is, multiple random mutations supposedly had accumulatedbefore the gene was activated and thus before they could have experiencedtraditional natural selection, and these mutations created a whole,functioning gene. It was therefore simply inferred that Jacob was wrong , and that random mutations unchecked by natural selection canaccumulate and create a whole new gene after all.
I side with Jacob, however, that this should not be possible. But this meansthat the facts are not fitting with traditionaltheorye.
The results from de novo gene studies are so tantalizing that theyshould have received more attention than they did, and that Siepel noted in2009  that much care should be taken to ensure that they are not due tosome methodological fluke.
Assuming that these results hold, the theory presented here can offer anexplanation to them, and is the only theory currently offering anexplanation: according to it, the writing phenotype can bring informationinto the evolving de novo locus from elsewhere in the genome overthe long-term.
According to the theory presented here, the writing of mutation bringstogether information from multiple loci into the single locus where themutation is written. So, bringing information into an evolving locus fromelsewhere is already part of this theory. Indeed, looking at the data ofLevine et al. , one can see large insertions in the sequence of the evolvingde novo gene. Presumably, these insertions have come fromelsewhere. In some other studies, Alu elements have been observedto contribute to sequence evolution of de novo genes (e.g., [12, 200]). The involvement in the evolution of de novo genes ofAlus—the same elements so thoroughly intertwined with themechanisms of nonrandom mutation discussed earlier—throws much lighton the topic, because the involvement of counter-traditional elements joinsthe impossibility of gene-out-of-nothing in placing the evolution of denovo genes far from the reach of traditional theory. In short, Iargue that de novo gene evolution demonstrates long-term movementof information by writing mechanisms. This is in accord with Jacob’sassertion that a new gene always draws on the preexisting. Indeed, in mytheory, every mutation draws on the preexisting.
Importantly, arguing that the long-term action of writing mechanisms givesrise to a new gene does not imply that evolution has “foresight”of the kind long rejected. The writing mechanisms have evolved and keepevolving under the influence of natural selection. They never“guess” what would be beneficial under natural selection. Theydo not create information out of nowhere but rather process information thatis present. Indeed, the long-term trend that culminates in the emergence ofa new gene in the de novo locus does not work on its own. Rather,it embeds a new gene in the larger genetic network, while changes in otherloci make room for this new gene in the network (evolution according to thetheory presented here is based on interactions—on network evolution).And, this long-term writing trend itself evolves in the long-term under theinfluence of natural selection. Thus, I propose that the de novogene, even prior to its transcription and translation, always evolves underthe influence of natural selection, but this influence is nontraditional: itaccumulates in the long-term through the evolution of the writing mechanism,and is indirect. The crucial difference from traditional theory is this:traditional theory, by lacking the writing phenotype, has no indirect routeby which natural selection can influence the evolving de novolocus, and thus reaches the paradoxical conclusion that a whole newfunctional gene evolves absent natural selection.
Armed with this new theoretical framework, we can take a closer look at thede novo gene data. Consider, for example, the case of thePoldi gene analyzed by Heinen et al. . In the house mouse (Mus musculus) and closely relatedspecies, this gene is transcribed in postmeiotic cells of the testis andshows evidence of functionality (reduced sperm motility and testis weight inknockout mice). In Figure 3, the signals forPoldi transcription and splicing are shown for mammalianspecies of increasing distance from Mus musculus. Notice how inhumans (the most distant species from Mus musculus in the sample),only 2 out of 6 signals are present. In Rattus norvegicus, 4 out of6 signals are present. In the basal Mus species Mus caroliand Mus famulus, as well as in Mus spicilegus, 5 out of 6signals are present. And in the remaining, focal species of Mus,all 6 signals are present. By parsimony, it is assumed that the gene wasmissing at least one signal at the root of this phylogeny, and thattherefore at least one if not more signals were added in time. Looking atthis phylogenetic tree without preconceptions, we see the possibility of aslow and tentative construction of a gene over the long-term and thereforein multiple lineages, where in the Mus genus it reaches the pointof transcription first.
Notice that this evolutionary trend seen in the data takes place on thetimescale of millions of years. This is an additional problem for thetraditional theory, beyond the problem of gene out of nothing (i.e., beyondthe problem of constructive evolution before transcription/translation),because we do not see from the traditional view what would spread thisactivity out over such a long timescale. But it is fitting with the theoryadvanced here, which predicts long-term trends in the writing.
Indeed, a bit more can be said about the fit between the theory presentedhere and events unfolding on the long timescale. The theory presented hereis a theory of the evolution of interactions. A new gene does not arise inand of itself as a separate event of traditional separate benefit. It ispart of a massive network-level evolution. We may expect that the sequenceof changes constructing a new gene will take much time to accumulate,because in parallel to it, vast amounts of changes in the genome are madethat allow for and accommodate this gene. Thus, much time is taken by thiscomplex evolutionary work.
This timescales issue is an important general point. New bona fide genes,sufficiently different from other genes, generally arise on the timescale ofmillions of years, regardless of their mechanism of arising. Thesemechanisms include not only whole gene duplication and gradual divergence,which may be seen as the gradual arising of a new gene from a traditionalperspective, but also chimeric genes and other genes that appear fromtraditional theory to have arisen by sudden events. Thus, from a traditionalperspective, rare events are interspersed with continuous evolution that arenot really part of this continuous evolution, and we are lucky to have suchevents at all because they are crucial for long-term evolution, yetapparently, according to the traditional view, evolution does very wellwithout them in the “in between” periods. The situation is seendifferently from the theory presented here, which has not two separateevolutions, one for the short term and one for the long term, but geneticevolutionary trends across the timescales that are complementary and worktogether in the gradual construction of complex genetic networks.
Two more specific predictions can now be raised. First, if long-term writingmechanisms participate in the creation of de novo genes, asstipulated by the present theory, then to some degree there may be molecularparallelism in the establishment of de novo genes even before thetime that they first become transcribed or translated. Such parallelism, iffound, could not be explained from the traditional theory. According totraditional theory, parallelism is due to similar selection pressures acrossthe populations or species concerned. That is, if the same random mutationoccurs in each population or species independently by chance, it could befixed in all due to the similar selection pressure. In de novogenes prior to transcription or translation, we have a situation wheretraditional natural selection cannot yet take place, and parallelism here,if found, would be consistent with the theory proposed here and not with thetraditional one.
With regards to parallelism in general, note that even though the theorypresented here is thoroughly in agreement with the widely accepted notionthat the number of independent evolutionary events should be considered as acost in the construction of phylogenetic trees, it allows much moremolecular parallelism than the traditional theory. This is because,according to the theory proposed here, similar selection pressures as wellas similar writing phenotypes in related species support parallelism.However, hypothesized parallelism still needs to go “with thedirection of the phylogenetic tree” rather than against it. In otherwords, independent evolutionary events do not all have the same cost inphylogeny construction but are more likely to occur the closer the speciesunder consideration are, because both the writing phenotype and theperforming phenotype (and hence natural selection) are more similar there.Thus, we may expect some molecular parallelism in the construction of denovo genes, even though, like traditional theory, we do not expectparallel dysfunctionalization of an originally functional gene.
Second, if the writing of mutations can write regulatory information into theevolving de novo locus while using information from elsewhere andthus drive evolution toward a functioning product before transcriptionand/or translation take place, then we also need to consider the possibilitythat it can move coding sequence information into the de novo locusfrom elsewhere before transcription and/or translation. The PIPSLgene (Akiva et al. 2006)—a new gene transcribed in the testis ofhumans and chimpanzees —provides an example of the kind of empirical evidence thatwould be relevant here. Zhang et al.  found evidence of strong positive selection (evolution of theamino acid sequence) in PIPSL in the lineage leading to humans andchimpanzees, even though the gene appears not to be translated in eitherspecies. They furthermore argued against the idea that the gene has beendysfunctionalized and that such dysfunctionalization is the reason itappears not to be translated . But because, from the traditional perspective, one does notexpect to see a high dN/dS ratio with no translation, the authorsproposed that the protein is there and that we just have not found ityet—perhaps it is translated during some brief time-window that has sofar escaped observation. While they may be right in saying that there isboth a signal of selection and a protein, the theory presented here bringsup an additional possibility: that there is a signal of selection yet thereis no protein. If the protein is searched for thoroughly and is not found,it would be an intriguing negative finding, because it would beunderstandable from the new theory but not from the traditional one. Thatis, although it does not necessarily follow from the new theory that thereshould be sequences currently undergoing positive evolution that are not yetexpressed, because for all we know, the transfer of coding information doesnot lend itself to a dN/dS >1 read, the possibility that it doesmay be pursued as a potential distinguishing factor between the new theoryand the traditional one, in the PIPSL gene and in other examplesf.
Epistatic capture amplifies the issue of de novogenes
De novo genes exemplify a problem that goes beyond the genes thathave been given this name and that in fact suggests that they do not form aseparate category. In recent work, Lynch et al.  have shown that a network of more than 1500 genes has beencoopted for decidualization of the endometrial stromal cells in placentalmammals—a key step in the establishment of pregnancy. Lynch et al.  and Lynch et al.  have furthermore shown that TEs have made a large contribution tothe organization of this network, activating genes that had been previouslysilent. Furthermore, according to Emera and Wagner , it appears that for many genes, with examples both in theendometrial decidualization network and elsewhere [204, 205], the insertion of a promoter-carrying TE was not sufficient forthe activation of the gene, but rather multiple further modifications wererequired after the insertion. This point was studied in detail in the caseof the decidual prolactin gene, dPrl, where modifications afterinsertion provided multiple transcription factor binding sites that bindfactors that interact with each other . Emera and Wagner  have called this process “epistatic capture”, a namethat underscores the importance of multiple changes acting as a whole andnot in a piecemeal fashion.
We can now see the essential similarity between the above and denovo genes evolution: in both cases we see that multiple changesare needed before a gene is transcribed and can be subject to traditionalnatural selection. Indeed, the fact that, in this example, before the geneis transcribed, first it is silenced, then a TE is inserted, and thenfurther modifications occur, such as insertions of additional TEs and pointmutations, shows that a whole lot must happen before it can become subjectedto traditional natural selection in its new context. This multiplicity ofchanges is essentially like that of the evolution of de novo genes,and expands the point from the de novo genes section, because nowthe multiplicity of changes is thought to happen in each of many genes thatare part of a network.
The fact that TEs are involved both in epistatic capture and networkorganization and in de novo gene evolution is of further interest.In accord with the new theory, TEs can participate in bringing informationfrom one locus to another, and, since their movement is affected not only bythemselves but also by sequences at or other characteristics of their pointof insertion, TE movement in fact combines information from multiple loci asit generates mutation.
Note that in the case of the evolution of endometrium decidualization, as inthe case of de novo genes, the multistep process takes millions ofyears. We have already discussed how this spread of activity over thelong-term fits with the new theory but is unexplainable from traditionaltheory.
Lynch et al.  have argued that their results demonstrate that novel, complexadaptations evolve not by the traditional process of independently selectedsteps but rather by network-level evolution. Though they did not propose anynew theoretical development in that regard, their statement is thoroughlyfitting with the theory presented here, because network-level evolution isinteraction-based evolution.
Addressing the conflict in the interpretation of TEs
There have been two ways of interpreting TEs. One emphasizes that they areserviceable to the organism [15, 19], [150, 207–210]. The other sees them as “selfishelements”—parasitic material—“junk” [5, 6, 149]. According to the latter, TEs are the remnants of viruses thatreplicate themselves “for their own selfish benefit” at theexpense of the “host”, though occasionally, by chance, bits andpieces of them are coopted by the host for host use. This “selfishgenes” school of thought has much to commend itself by, because if wefocus on the short-term, it looks indeed as though TEs just “replicatethemselves” and are not really needed for the organism’sperformance; if anything, they can cause disease. The problem is that, whilea “little bit” of fortuitous cooption of TEs for“host” use may seem reasonable to assume, the immensecontribution of TEs in the long-term to the evolution of organisms is a bithard to assimilate under traditional principles. Would supposedly fortuitousmovements of genomic parasites organize more than 1500 genes into a novelgenetic network underlying the important, complex adaptation that is thedecidualization of the endometrium [15, 204]? The situation from a traditional perspective is a stalemate:both sides have important arguments to support themselves, yet they areconflicting. Doolittle [8, 119] offered a way of resolving this conflict, by proposing thatclade-level selection helps to explain the existence of a system hospitableto TEs which is useful in the long term. But the debate over whetherselection at levels above the gene and individual is strong enough to affectsuch things is far from resolved. Hence we must admit that the question isopen.
The theory presented here sees TEs as a part of the system—as a part ofthe writing phenotype. Their contribution is systematic, and does not ariseas a fluke. This system, however, is not a “homunculus” thatoutguesses selection. Instead, it is a complex system—an ecology ofwriting activities. In such a decentralized system, TEs may well appear tobe “pressing” to self-replicate and integrate where they can,even while other writing forces remove them, change them, or silence them;and it is through this tension of writing activities that evolution happens,according to the new theory. The same view of “negotiation”between contradictory forces has been applied earlier here to theunderstanding of the triple association between zones of adaptive evolution,mutation hotspots, and genetic diseaseg.
What of the relation of TEs to viruses? At first sight, it might seem tosupport the selfish elements view: if TEs evolved from viruses (indeed“viruses” as we see them today—parasites of their host,unnecessary to the host and its evolution) then at least originally, theirincorporation in the evolution of the host genome has been a fortuitouscooption of parasite parts; and if this happened originally, we might aswell assume that it keeps happening. But we do not know that TEs evolvedfrom viruses originally. The perspective given here, which sees TEs as asystematic part of the organism’s genome, encourages us to considerthe possibility that at the origin of viruses were elements much moreintertwined with the functioning of the organism.
Another point raised by the present theory regarding viruses concerns thequestion of how they evolve. They seem to be too small to include much ifany writing mechanisms. However, much like their performing phenotype is not“their own”, but due to an interaction between them and thehost, so too could their writing phenotype be not their own, but due to suchinteraction. This leads to a certain prediction: that viruses will showspecific characteristics of molecular evolution (idiosyncrasies of specificwriting mechanisms) that parallel those of their present and past hosts.
So-called “junk DNA” may participate in evolution in anontraditional manner
The sequences from which de novo genes arose are called“junk” from a traditional perspective , but I argued that writing activity takes place there. Repetitiveelements have been said to constitute “junk” , but I argued that they are part of the writing machinery. Thus,according to the theory proposed here, such so-called “junk DNA”may participate in evolution in a nontraditional manner.
Recently, the ENCODE consortium announced, based on empirical evidence, thatas much as ∼80% of the human genome is biochemically active andtherefore “functional”. This statement was criticized ontheoretical grounds in papers that were invaluable for bringing evolutionarytheory to bear on the results [7, 8]. However, since these theoretical grounds trace themselves backto traditional theoretical assumptions, we must go over the assumptionsunderlying the criticism and see how they may change once we take the pointof view proposed here.
First, it has been thought that junk DNA is junk because it is not conserved.Evolution in it appears to be neutral, and if it were functional accordingto the traditional theory, it would have represented far too much geneticload [4, 7]. However, I have argued that, in accord with the new theory,things can evolve that have a neutral appearance yet are under the influenceof natural selection on interactions. Evolution of interactions isexplicitly nontraditional, and traditional load calculations do not apply toit. Thus, according to the new theory, these traditional reasons to believethat the majority of the human genome is “junk” do notapply.
Second, it has been thought that junk DNA is junk because of the C-valueparadox: organisms may vary greatly in genome size without relation toapparent organismal complexity, which has been taken to suggest that much ofthe genome is not needed . But first of all, in what sense is it “not needed?”The traditional theory considers only the performing phenotype. Once weadmit the existence of the writing phenotype, there is room for much writingactivity across the genome. As Eddy wrote, it is one question how much DNAit takes to design a human; it is a whole other question how much DNA ittakes to evolve a human  (and one may add: the amount of DNA participating in mutationalwriting in a given organism at a given point in time may not necessarilyfollow organismal complexity closely).
If junk DNA is part of the evolving writing ecology, then we see that itcould vary much and yet be an inherent part of the evolving organism.Indeed, TE bursts and whole genome duplication are obviously of majorimportance for evolution in the long term, and they cause quick changes ingenome size. Those who support the junk concept do not contest this lastfact that junk can become of use in the long-term [7, 8]; indeed, some have criticized the use of the word“junk” because of this . The issue is rather how we conceptualize the evolutionaryprocess. The traditional view has a difficulty in seeing long-term activityas systematic. TEs serve as an example: their short-term costs have beenpart of seeing them as “selfish elements”, and their long-termbenefit has been seen as a fluke. Thus, the traditional view sees activityin the “junk” as random noise out of which adaptive materialarises fortuitously. However, my theory allows for biological activity oflong-term constructive consequence even with short-term costs. Moregenerally, junk DNA may play an important, nonrandom role in evolution inthe long-term. What we are observing in the non-conserved DNA may be anunconcentrated mass of interacting material out of which more concentrated,“genic” material gradually arises in the long-term by anontraditional evolutionary process.
Thus, while findings of the ENCODE consortium have already shown that themajority of the human genome is biochemically active , the new theory allows for the possibility that this activity ispart of the evolving organism and is important; whereas traditional theoryseems to predict that only a small amount of it could be functionallyimportant, and the rest must somehow be random noise. The only reason tobelieve that so much activity in the organism does not really belong to theorganism and is just “noise” is the fact that the traditionaltheory is short-term performance focused, single-allele focused, andrandom-mutation based. If this reason is removed, our understanding of junkDNA is changed.
Many writing mechanisms may exist in the sperm cells
There are several fundamental differences between the genetic system of thesperm and that of the soma. These include, but are not limited to,transcriptional promiscuity (reviewed in ), alternative splicing promiscuity [213, 214], and a specialized RNA interference and chromatin organizationsystem in the sperm cells of mammals (based on PIWI-domain proteins) .
Transcriptional promiscuity (TP) occurs during the development of the spermcells—in cells at the meiotic stage (spermatocytes) and early haploidsperm cells (round spermatids) . These cells are much more transcriptionally active than somaticcells by several measures: expression is extremely diverse, showing geneproducts that are different than the usual ones, including partial products,and many RNAs are expressed at much higher levels than usual . TP is a highly involved mechanism, requiring the orchestrationof a complex machinery to both create and compensate for the pattern ofexpression , and its evolutionary origin is a mystery.
Interestingly, TP has characteristics that make it useful for the writing ofmutations predicted by the theory proposed here . This writing requires bringing together information frommultiple loci, which means that at least some genes that affect the writingmust be transcribed, so that their products can reach elsewhere and allowinteraction between loci. TP can allow many genes to interact and be part ofthe writing activity, including genes with established somatic functions aswell as genes that are in the process of formation and have only minimalpromoters [212, 216]. The TP stages of the developing sperm cell overlap with themeiotic stage.
With regards to the sperm-specialized RNAi system in mammals, this system isthought to be involved in the control of TEs in the germline . We have already seen the counter-traditional nature ofTEs’ role in evolution: they are important in mutational mechanismsand in such examples of evolution as the organization of genetic networksand the long-term writing of de novo genes—activities whichmust be taking place in the germline. It has also been proposed that thegermline provides special opportunities for the activity of TEs, especiallythe state of hypomethylation thought to be involved also in TP [216, 217]. It has also been proposed that in mammals there is controlledDNA “repair” by transposons , which of course may also be DNA “change”. Theexistence of such a deeply evolved system that does not abolish TEs but infact regulates their activity [19, 209, 210, 215] is consistent with the view that TEs are not incidental remnantsof viruses that the organism just defends against and that occasionallycontribute to its evolution as a fluke, but rather are a systematic andinherent part of the writing machinery.
From a traditional standpoint, one could have raised the question of why allthese phenomena with a deep potential to affect the mechanics of evolutionoccur in the germline. Of course, the germline is where mutations areheritable, and so only there is the evolutionary potential of these highlyinvolved phenomena fulfilled. But from the traditional theory, which isbased on random mutation and immediate advantages, one cannot easily explainthe evolution of the abovementioned phenomena based on their effect onevolution, and they cannot be easily put together into a systematic picture.The theory proposed here, on the other hand, predicts that the writing ofmutations exists, and operates on all timescales. If so, we should expect tosee it in the germline. This offers an explanation for the multiplicity offundamental differences between the genetics of germline and soma.
However, this does not yet clarify all aspects of the situation. My theorypredicts that there are writing mechanisms both in the sperm and in the egg(or the cells that lead to them), because both male and female traits mustbe under the influence of natural selection (and according to my theory,nonrandom mutation collects the effect of natural selection on geneticcombinations). Why the asymmetry between sperm and egg in many organismsstudied so far? It may be that writing mechanisms that operate in thelong-term and evolve in the long-term and that are complemented byindividually determined mutation mechanisms, but need not co-occur withthem, are free to be either in the sperm or in the egg or in the cells thatgive rise to them (whereas individually determined mutation must be inboth). Thus, if the abovementioned phenomena are writing mechanisms, itmeans that, for some reason, at least in the organisms studied so far, thedeveloping sperm cells have been specialized as the locus of much long-termwriting activity. In this case, why it is the sperm and not the egg cellsthat have been specialized for this part of the writing is not specified bythe theory, but one may wonder if the excessive number of sperm compared tothe number of eggs enables this excessive writing activity, for example byproviding enough cells so that some will be functional despite intensiveevolutionary work causing a high error rate. We may call this “theworking sperm” hypothesis.
Indeed, it is interesting that the results of ENCODE, which found so muchbiochemical activity across the genome, were based mostly on pluripotentstem cells and cancer cells , which are known to have unusual genetic activity ; according to some, the latter behave in several ways like germcells, and especially like sperm [218, 219]. ENCODE may have helped to bring to light the exceptional geneticactivity of the germline.
It is also interesting that new genes have a strong bias of being expressedin the testis, whereas older genes have stronger and broader expressionpatterns [193, 216, 220, 221]. To explain this bias, it has been suggested that new genes arisefirst in the sperm to serve sperm functions, and that later these genes aresomehow coopted for somatic functions [216, 221, 222]. This has been called the “out of testis” hypothesis [216, 221, 222]. Two reasons have been given for why this happens in the spermspecifically: One is the existence of TP there. Interestingly, TP is used inthis hypothesis in a manner not far from the above—it is seen there asa phenomenon that facilitates genetic evolutionary change—though itsorigin in the first place is not easily explained from this view. Second, itwas assumed that the sperm are under much stronger pressure to evolverapidly than other cells in the body, and are thus in “highdemand” of new genes [212, 216].
However, a main reason to think that these new genes really serve the spermin its performing phenotype is that knockout of them disrupts thedevelopment of sperm. According to the traditional theory, which has onlythe performing phenotype, this evidence of functionality from knockoutindeed means that these genes serve the sperm. But the theory proposed herepredicts the existence of the writing phenotype, which raises thepossibility that many (though not all) of what we think of as “newsperm genes” serving the sperm’s performing phenotype are notnecessarily traditional “sperm genes” but are either genes withan evolving somatic function that are the locus of much writing activity orgenes that belong to the body of the writing phenotype, and that disruptionof them by knockout derails the writing system and makes it cause damage inthe sperm cell (indeed, the sensitivity of sperm cells is well known). Thus,the observation that has led to the assumption that the sperm are underpressure for rapid evolutionary change, which has underlain theout-of-testis hypothesis in the first place, may not be only due to rapidevolution of the sperm performing phenotype. It could be that, to a notabledegree, the sperm appears to be so rapidly evolving because of the writingactivity in it and the evolution of this activity.
Indeed, we may now note that, of the examples of adaptive evolution detectedby dN/dS >1 mentioned in , the first two (and in that sense prominent) categories ofexamples mentioned involve molecular environment interaction genes and rapidevolution of sperm, and both of these take on entirely new meaning accordingto the theory proposed here.
Though much more data are needed on the material discussed in this section,one thing we need to notice is that the “working sperm”hypothesis has relevance beyond science. In 2001, Old  (see also ) suggested that cancer cells imitate germ cells and trophoblastsin several respects which appear to be part of the malignancy of thedisease, including global hypomethylation, expression of cancer-testis (CT)antigens, expression of chorionic gonadotropin, downregulation of the majorhistocompatibility complex and immune evasion, and more. Indeed, cancercells are exceptionally genetically active. We should ask, therefore,whether activation of the evolutionary writing activity is part of theetiology of cancer. (It is understandable, then, that in some sense HeLacells appear “nonhuman” [7, 224].)
A quintessential example of ns/rm may be an example of mutational writing
Since the works of Pauling et al.  and Ingram , the evolution of malaria resistance and sickle cell anemia has beentaken as a quintessential example of the ns/rm process. It has been thought thatrandom mutation caused a change at nucleotide 2 in codon 6 in theβ-globin gene from A to T (Glu→Val), and that naturalselection caused a substantial increase in frequency of this new allele. In theheterozygote form, this allele (henceforth HbS) provides notable protectionagainst the malaria parasite Plasmodium falciparum, and in thehomozygote form it causes sickle cell anemia [227, 228]. According to Haldane’s model of heterozygote advantage , selection in malaria-stricken areas maintains the HbS allele in thepopulation despite its cost.
How do we deal with this “queen of examples” of ns/rm from theperspective of the present theory? There are at least two options. First, onemay admit that this indeed is a case of evolution by ns/rm. If so, then, if thepresent theory is correct, one would have to say that this is the exceptionrather than the rule. In this case, one would say that, occasionally, a randommutation can cause a simple adaptive change, and that the HbS mutation is such achange, but that such mutations cannot build up toward a complex adaptation andtherefore cannot be the main drivers of evolution. Essentially this approach wastaken by Behe while criticizing the traditional theory .
However, there is another possibility. It is that the HbS mutation was not randomafter all. Consider more background first. HbS is one of severalhemoglobinopathies that provide some degree or another of protection againstmalaria and that are due to genetic changes in the α- andβ-globin genes. The most common of these are the regulatorychanges (including whole gene deletion) causing α-thalassemia andβ-thalassemia and the following point mutations causingstructural change in the hemoglobin protein: HbS, HbC (which, like HbS, involvesa point mutation in codon 6 of β-globin, this time causing a Glu→Lys change) and HbE (codon 26 of β-globin, Glu→Lys) . Details of the mechanism of protection are still debated , but, at the most general level, notice that all these changes affectthe internal composition and behavior of red blood cells, which is where themalaria parasite grows. We know that natural selection is involved in theirprevalence, mainly from the fact that there is a strong geographical correlationbetween their prevalence and the incidence of malaria (reviewed in [231, 233]). This is consistent with both the new theory and the traditionaltheory, because both rely on natural selection for the fit between the organismand its environment. It remains to be asked whether the mutations arise randomlyor not.
Now, notice that there is substantial mutation and recombination hotspot activityin the α- and β-globin gene clusters [234, 236]. Indeed, some malaria-stricken populations are so riddled withmutations affecting red blood cells that in most individuals the cells areabnormal . From the traditional standpoint, why would these mutation hotspotsbe there? One might say that mutation hotspots are randomly positionedthroughout the genome, and that it so happens that genes with a potential to beinvolved in malaria resistance also have them. But we have already seen thatmutation hotpots are in general frequent in regions undergoing adaptiveevolution and are also in general associated with disease. The case ofhemoglobinopathies is not a special case but follows this rule.
Indeed, we already know that deletions in the α-globin gene, whichprovide some protection against malaria and cause α-thalassemia,occur by NAHR and are recurrent . HbD-Punjab and HbO-Arab, which are the 4th and5th most prevalent point mutations causing structural change inthe hemoglobin protein respectively, are both in the same position—thefirst nucleotide of codon 121 of β-globin, which is right at thesharp bend of a 4-nucleotide hairpin, a DNA structure thought to facilitatemutations . These cases are already consistent with the writing of mutationsaccording to my theory, because they are evidently guided by internal factors.It is only reasonable to ask whether the HbS mutation was also guided by amechanism yet to be determined.
Indeed, the HbS mutation appears in Africa on four different genetic backgroundsunrelated to each other by simple meiotic recombination events, whereas the HbEmutation appears in south-east Asia on at least two such backgrounds (reviewedin ). But HbS does not appear in south-east Asia, and HbE does not appearin Africa . It is usually concluded from these data that HbS arose four timesindependently in Africa, and HbE arose twice independently in south-east Asia . Now, since there is no reason to expect that these mutations wouldwork in one place but not the other, if these were random mutations, and therate of random mutation was high enough that HbS arose independently four timesin Africa, then why does this mutation not appear in south-east Asia? And if itwas high enough so that HbE arose in south-east Asia twice independently, whydoes that mutation not appear in Africa? To address this question, Flint et al. [231, 233] proposed that the HbS mutation arose only once in Africa, and wastransferred to other genetic backgrounds by means of gene conversion of shortstretches assisted by a meiotic recombination hotspot (which, from a traditionalperspective, just happens to be there). But there are multiple other cases ofglobin gene mutations that show the same kind of pattern, and it is unlikelythat this hypothesis applies to all of them (indeed, Flint et al. do not attempta general explanation, and others do not think that conversion really applies toHbS ). For example, according to Flint et al. , the α-thalassemia deletion mutation -α4.2 appears to have arisen six times independently in Melanesiaspecifically, the -α3.7III mutation has been found in other parts of Oceania exclusively,and the -α3.7I arose three times independently and is most common in theMediterranean and in Africa. According to Kazazian and Boehm , the G→C mutation at intervening sequence 1 nucleotide 5 of theβ-globin gene (IVS-1 nt 5) appears on different backgrounds inAsia and Melanesia, whereas the G→A mutation at intervening sequence 2nucleotide 1 of the β-globin gene (IVS-2 nt 1) appears ondifferent backgrounds in people of African and Mediterranean origins. Accordingto Thein et al.  and Kazazian et al. , multiple rare dominant β-thalassemia mutations in exon3 of the β-globin gene have been observed outside of the malariabelt in Northern Europeans, notable among them is the third and last possiblemutation at the first position of codon 121, G→T , which has arisen recurrently there , and is different than the abovementioned HbD-Punjab and HbO-Arab,which are mutations from G to C and from G to A respectively, common in otherpopulations. Details may change based on newer information, but the principle isunlikely to disappear. In general, it appears that malaria resistance solutionshave a strong tendency to evolve recurrently, and furthermore they are moresimilar within human populations than between them (this well-established ethniceffect occurs even between separate but geographically neighboring populations;reviewed in ).
This ethnic effect of malaria resistance mutations, whose paradoxical nature froma traditional standpoint was well articulated by Flint et al. , becomes much more understandable from the theory presented here.According to this theory, the writing phenotype evolves, and therefore differentpopulations will reach somewhat different solutions to the same problem,appearing as recurrent mutation. It is nothing but an example of divergentparallelism, which is how evolution in general unfolds according to thistheory.
It would be ironic if HbS—the first mutation to be characterized at themolecular level, and a prime example of traditional ns/rm—turns out to bean example of nonrandom mutation after all. Regardless, in accord with bothevidence of other malaria resistance mutations and previous molecular biologicalevidence discussed, we are left to conclude that, empirically, mutation isaffected by internal biological factors, and that these factors themselvesevolve. This grand empirical fact is clearly in accord with the theory presentedhere, but plays no role in the traditional theory and is not understandable in astraightforward fashion from the traditional perspective.
Final remarks and summary
The origin of life
It is not my purpose here to make speculations on the origin of life. But withouthaving clarified my position on the origin of sex, it may have been difficult tosee how sex could be a matter of necessity for evolution. Similarly, withoutclarifying my position on the question of the origin of life, it may bedifficult to accept the fact that nonrandom mutation is at the basis ofevolution.
Let us see, first, that the two origin questions are related. I argued that sexis a matter of necessity. This raised a question of origins: if life beganasexually, then biological evolution in the beginning occurred without sex, butthis would seem to suggest that sex is not a matter of necessity for evolution.I then argued that, for all we know, life did not begin asexually. But if lifedid not begin asexually, then it did not start with a single organismeither.
People often imagine that life began with a chance event giving rise to a“self-replicating” molecule [149, 242, 243]. This self-replicating molecule was the “first organism”,so to speak. By self-replicating, it created a population of such molecules,which were then able to compete. Under the assumption of errors inself-replication, neo-Darwinian evolution started and gave rise to all of life,supposedly.
This single molecule/“naked gene”  scenario is an extension of the neo-Darwinian idea to the“beginning” of life. Since we must begin with this molecule and endwith humans, we then assume that organization at different scales has been addedas layers one by one: first came self-replicating molecules, then cells, thenmulticellular organisms, and then societies.
This view is not the only view on the origin of life (see, for example, ), but it serves as a contrast to the theory presented here. If lifebegan with a molecule that arose by chance and self-replicated with error, thenlife began with a period where there was no writing of mutations—there wasno “higher level phenotype” to speak of to enact any change to thepresumed chance-arising “genome”. To presume that life began with anasexual, random mutation evolutionary process, and at some point switched to athoroughly different process based on sex and nonrandom mutation introduces anarbitrary line into history and into theory of the kind that we should be happyto remove and that shows a disagreement between this hypothetical originscenario and the theory presented here.
But life did not have to start with a single, chance event at some particularpoint in space and time. Instead, consider the possibility that there has been asmooth transition from a “chemical” primordial soup to a“biological” one. In this case, life “began” with a“proto ecology”—a complex world of chemical reactions. In factit is not correct to use the word “beginning” here, because it didnot start from a particular point in space and time. What would later become thegenome and what would later become the phenotype, including the writing ofmutations and every other fundamental aspect of life, have descended from thisproto-ecology together, coevolving. Genotype and phenotype coevolved. Thisenables gradual evolution of the framework of life as we see it today, whileavoiding sharp transitions from “chemical” to“biological”, from “asexual” to “sexual”,from “single organism” to many, etc. This view is consistent inimportant respects with those of Woese , Brosius [55, 56], Vetsigian et al.  and others.
One may ask: But did life not have to begin with a self-replicating mechanism,the arising of which has to be explained by a chance event, because without it,there could be no population of individuals to undergo selection? The answer isno. Things can be similar not because their ancestor arose in a point and led tothem by asexual reproduction, but because of the application of the same lawsand materials everywhere . Complex organismal entities that are similar to each other couldhave gradually arisen not because of an asexual spread of a chance gene from onepoint to the rest, but by a process of convergence like the one discussed inthis paper. In other words, things can come to share characteristics bymultidimensional exchange of information (such an exchange is the shuffling ofgenes of today), and not just by sequential, one-way spread of non-interactingpieces of information [54–57].
Indeed, “self-replication” is a misleading term. Strictly speaking,there is no such thing as “self-replication”. Do we mean by it thatnothing other than the object itself takes part in the replication of theobject? This is logically impossible and empirically evident not to exist. Anindividual can only be replicated in the right environment, not to mention thatits replication involves material and energy coming from the environment. Sincethe right environment is indispensable, the responsibility for replication isnot only within the “self”. Furthermore, under sexual reproduction,the individual is not really “replicated” at all.
We tend to put “self” in “self-replication” because theindividuals that we see struggle so much to ensure their reproduction, and theirteleological behavior makes us focus on them as actors, and because they carryin them information needed for their reproduction. But all of this could havegradually evolved from a “proto-ecology” of chemical interactions,involving the ancestral matter of both writing and performing. Life did not haveto start, anthropomorphically, with a chance-like event of the sudden origin ofa mechanism capable of “replicating itself”. Life did not have tostart with an “Adam” molecule.
Placing the theory presented here in context of previous thought
When Darwin began thinking about the mechanism of evolution he started byspeculating that sex was the driver of it , showing how important this phenomenon is in the eyes of theuninitiated. When he saw Malthus’s paper and came up with the idea ofnatural selection, he largely put the question of sex aside, though he kept inmind a fuzzy notion of “blending inheritance” that is due to sex.When his theory of natural selection became known in 1859, it was notimmediately accepted by the biological community (only the fact that evolutionhappened was), but rather continued to be debated for 70 years, becauseinheritance was critical to the theory, yet Darwin had only a vague notion ofit. Especially, Galton posited that the ever-present individual variation thatDarwin relied on could not be the source for evolution, because under blendinginheritance—which is due to sex—the individual makeup could notpersist; and thus a special, sudden mutation was needed, whose character wouldnot be lost in the passage from one generation to the next [246, 247]. We can easily see that, from the beginning, the problem ofinheritance was thoroughly intertwined with the problem of sex. It is only sincethe modern synthesis that we have forgotten that these problems are one and thesame. The theory presented here, however, treats them as one and the same:information from allele combinations is inherited through nonrandom mutation,which solves the original problem of sex.
It was not until the works of Fisher, Wright and Haldane in the 1920s and 1930sthat Darwin’s theory of natural selection was finally accepted, but onlyafter having been crucially changed. The simplest and clearest wasFisher’s theory of adaptive evolution, and its critical assumption wasthat each allele is thought to make a small, separate, additive contribution tofitness, or to some quantitative phenotype, and is largely independent of allother alleles. This way, each allele maintains its meaning despite the sexualshuffling, and there is no longer a problem of sexual inheritance (while Fisheracknowledged the presence of interactions, they were not part of the core of histheory of adaptive evolution). It has often been said that the key of themodern-synthetic revolution was that it married Darwin’s theory of naturalselection with Mendelian genetics, or in other words with the sexual shufflingof genes. But in fact it proposed a concept of selection that worked despitethis form of inheritance, not with it, as until today it is easier to understandthis theory without sex than with it.
There is, however, a flaw in Fisher’s theoretical assumption that additiveeffects are at the core of adaptive evolution. All would agree that an alleledoes not work on its own as though it exists in a vacuum. The allele mustinteract with the genetic background. According to the assumption justmentioned, it interacts largely with the fixed genetic background. That is,alleles are not allowed to interact substantially with each other ingeneral, because such interaction is “noise” that drownsthe supposed signal of single-allele–based selection. But this assumptionintroduces an arbitrary line into the foundations of the traditional theory: Howcan we expect that alleles will systematically be able to“distinguish” between genetic partners that are fixed in thepopulation and partners that are not fixed, so that interactions with the fixedbackground are systematically strong and interactions with concomitant varianceare systematically weak? It is interesting that we have not asked this questionfor so long, but it is time to ask it. Removing this arbitrary line weimmediately get to the framework of the theory presented here. This is a third,independent entry point into my theory (independent because removal ofarbitrariness from theory is good independent of anything else).
Interestingly, Mayr has already made his criticism of the Fisherian, generallynon-interactive approach loud and clear . He considered it of vast importance that selection in the presenceof sex is context-dependent, and that, with it, the genome evolves as a unified,cohesive whole. However, he did not propose what makes this context-dependentselection possible—what allows selection on interactions to driveevolution. Here, I have done so.
Finally, Shapiro deserves much credit for being the first to promote heavily anidea of nonrandom mutation. Indeed, he has summarized his view in a highlyinspiring and informative book . However, he has not specified the mechanistic underpinnings of thejoint operation of natural selection and nonrandom mutation. Also noteworthy anddirectly relevant to my work are works by Caporale, who has expressed viewsrelated to a part of the present paper, namely an association between mutationhotspots and regions of adaptive evolution, which she calls a “mutationphenotype” [177, 178]; and works by Stoltzfus and colleagues, who have written extensivelyon the importance of mutation in evolution from nontraditional views, includingon mutation bias [249–252].
This paper holds that the mutation that drives evolution is not a result ofrandom accident but an outcome of a mutational writing phenotype. This phenotypeitself evolves, like anything else inherent to the organism. It absorbs theinformation that comes from selection and guides selection further by providingfurther variation.
I have presented evidence—from Darwin’s organismal-level observationsof the evolution of variation, to modern research on cSNPs, to observationscollected here on the evolution of malaria resistance genes—that issubstantially fitting with the above. Evidence shows that the generation ofvariance is more similar between more related species or populations, supportingthe idea of “divergent parallelism”, which follows from an evolvingmutational writing phenotype.
To clarify, this process is not Lamarckian. It is not subsumed by previousdiscussions of “directed” or “adaptive” mutation (e.g., [22, 253–255]) and it would be incorrect to equate it with those (thoughinteresting connections may exist). The nonrandom mutation proposed here doesnot usurp the role of selection, it rather absorbs information from selection oninteractions. This mutation is dependent on the genetic state of the organism,which itself depends on past selection (and past mutations). Importantly, whilethis mutation is distinctly different evolutionarily from accidental mutation,this does not imply that a given mutation is more likely to arise in anenvironment where it increases fitness than in an environment where it does not,nor that mutation is more likely to increase rather than decrease fitness.
This process is also not equivalent to “cranes” such as thehypothesized SOS system in bacteria, which are hypothesized generic“tricks” that are presumed to facilitate evolution based on apresumed ns/rm core and that are presumed to have evolved by ns/rm. Quite incontrast, the mutational writing phenotype discussed here evolves along with theevolution of adaptation and is therefore specific to the evolutionary times, andis part of the core of the evolutionary process, not an “add-on” ontop of it. In designing experimental evolution approaches in accord with theprocess described here, one should be mindful of the fact that this process isexpected to be generally a long-term one.
The process described here is “Kantian” in that it shows thatevolution is driven not only by external forces. It is not random accident thatgenerates the variance that selection operates on. Rather, a phenotype causingsyntactic internal change is absorbing information from the outsideworld—from natural selection—and changes itself in the process.
This solves the problem of sex in a manner very different from before. No longerdo we treat sex as a phenomenon of potential subsidiary benefit, but rather wetreat it as a fundamental part of evolution by natural selection. The theoryproposed here does so by tying sex to the question of interactions.Investigators have suspected from the beginning that interactions must somehowbe formed to allow for the evolution of complex adaptation. But how are theyformed under selection? Some resorted to chance to explain interactions; butthis approach was not followed here. In the presence of transmission ofhereditary information through syntactic mutational writing, selection oninteractions influences future generations.
To reiterate, mutation combines information from multiple loci as it changes acertain locus. While natural selection operates on the genetic combinationscreated by sex, the writing of mutations combines information from multiple lociinto new mutations, which are not themselves broken by the sexual shuffling, andthus allows the combinations to have hereditary effects according to theirfitness. Thus, natural selection and nonrandom mutation work together, wherenonrandom mutation allows selection to operate on the organism as a complex,interacting whole. Here, a new mode of information transmission wasproposed.
Interestingly, this opens up neutral evolution to a whole new interpretation.With selection on interactions, we no longer expect that each allele will beeither neutral or have a tendency to move straight toward fixation or straighttoward extinction, but rather the frequencies of alleles will change in anunpredictable manner, owing to the nonlinear dynamics of selection oninteractions, and this movement is not necessarily easily distinguishable fromdrift. Therefore, what has been seen before through a traditional lens asneutral matter can be experiencing selection on interactions and thus can play anon-fortuitous role in adaptive evolution.
Cutting-edge evidence from molecular evolution supports the proposition thatmutation is nonrandom. More specifically, evidence on cryptic variance and thecomplex determination of mutation-recombination hotspots supports theproposition that mutation combines information from multiple loci into one. Manyother cases that speak to this latter point may be lurking in the literature andstill others may have yet to be empirically discovered. Epigenetic inheritancemay follow this pattern of combining information from multiple loci into one,and the whole connection between epigenetic inheritance and long-term geneticchanges is a massive area that needs to be explored from the perspective of thepresent theory.
Another point of interest is how an adaptation comes to be shared among themembers of a species. A new trait comes into being not by the sequential spreadof mutations that supposedly bring separate phenotypic changes from theindividuals in which they arose to the whole population. Instead, while allelesspread, they interact, and the new trait arises at the level of the populationas a whole from these interactions. We saw that this is necessarily a process ofconvergence, where gradually the trait becomes less influenced by the sexualshuffling of genes and thus more uniform across individuals. It is therefore aprocess of stabilization, one that is an automatic concomitant of the adaptiveevolutionary process described here, and does not require an extra traditionalselective force specifically for stabilization, as assumed in theories ofstabilization or canalization. The writing of mutations enables this process ofconvergence by combining information from different individuals (and fromdifferent loci) over the generations. Interestingly, this convergence processconnects molecular evolution to phenotypic-level evolution better than before,because empirically, the evolution of complex adaptation looks like a process ofconvergence at the population level.
Below is an outline of the main points made in this paper:
Mutation is the outcome of a nonrandom, biological process.
It follows that mutation combines information from multiple loci into one.
By combining information from multiple loci into one, mutation allows selection on genetic interactions to have a hereditary effect according to fitness.
This revises the connection between selection on the phenotype and evolution of the genotype proposed in the 1920s–30s in a way that connects the theory of evolution better to modern evidence. Mutation has a complex genetic component, and the causes of variance and the nature of inheritance are not separate issues.
This view is a third way of thinking about evolution: it is neither neo-Darwinian nor Lamarckian.
Given that selection can operate directly on genetic interactions, sex becomes an element of fundamental importance for evolution, not one of subsidiary, circumscribed benefits, since it is the generator of genetic combinations.
It follows from the above that: a) sex is original—it did not evolve from asex; b) sex (or a mix of sexual and asexual reproduction in a species) is not actively maintained against obligate asex, but rather, and more simply, long-term adaptive evolution is not available for obligate asex to arise by.
It is therefore predicted that obligate asex arises by breakage and that no fine-tuned adaptations ensuring obligate asexuality exist. This prediction offers a new look into a key open question in plant mating systems, namely why pure asexuals are exceptionally rare. It is confirmed in vertebrate unisexual animals and in androdioecious animals, and remains to be tested in complete cleistogamous species and/or other cases.
It is also predicted that putative ancient asexuals have not substantially evolved and diversified in an asexual state. This prediction is confirmed in the case of the bdelloid rotifers according to statements by Meselson, and remains to be tested more thoroughly in these organisms and others.
It follows from the theory that an adaptation arises by a process
Stabilization arises automatically; it does not require an extra
Evidence shows that rearrangement mutation and point mutation are
It is noted that interpreting mutation as ultimately accidental
Since the theory proposed here holds that mutation is nonrandom
The above points show that the theory proposed here ties together
It is proposed that genetic disease can be seen as the result of
The so-far quintessential example of evolution by random mutation
A more advanced consideration of the new theory shows that alleles
It is predicted that there must be biochemical activity in the TRIM5–CypA gene fusion was proposed as an example. These predictions provide a more parsimonious view than that of traditional theory on the fundamental differences between the genetic activity of germline and soma.
The existence of mechanisms of mutational writing may inform our
The writing phenotype view holds that there is no dividing line
The last point leads to the prediction of long-term genetic
Selection is not an external judge of phenotypic meaning as in the
What appears as neutral from traditional theory actually can be
De novo genes exemplify the paradox of “explaining” adaptive evolution by chance. The writing phenotype offers an indirect route by which selection can exert itself on the evolving de novo locus.
The questions have been raised of whether there exist sequences PIPSL, and/or in other cases), and whether molecular parallelism could be found in sequences that cannot be subject to traditional natural selection.
Epistatic capture amplifies the point above on de novo genes and ties it to TEs.
TEs may appear as “selfish elements” but nonetheless
The “working sperm” hypothesis has been proposed,
ENCODE’s results may have indirectly helped to expose the
Reviewer’s report I: Nigel Goldenfeld, University of Illinois atUrbana-Champaign
The main ideas in this long paper are:
Mutation is not purely a random process but contains a deterministic contribution arising from the interaction of genes and from the way in which the organism interacts with its environment.
The mutation at a given locus is a function of the alleles at all other loci so is inherently an interacting many-body process. Mutations are not to be thought of as arising from processes acting on individual alleles. This is the sense in which (3) is claimed to be true, thus explicitly avoiding teleology.
Selection acts in the canonical way, on combinations of alleles, but these combinations are not disrupted by shuffling due to recombination. Fitness is a collective property of many alleles: a single allele does not have a phenotype and fitness ascribed to it.
Sex (here defined in a general way that includes bacterial conjugation, for example) is the rule: obligate asexuality is breakage of sex, the exception not the rule and indicative of a dead end phylogenetically. The problem of the evolution of sex is actually the other way round: sex is the natural result of the evolutionary process as proposed in this paper, and one really needs to ask about the evolution of asex.
The writing of mutations from multiple allele interactions into a single locus is itself a phenotype, and interacts with other phenotypes. Consideration of this process corresponds to the notion in classical population genetics of fixation.
A number of qualitative predictions are made from the author’s perspective, related in particular to the cost of natural selection, transposable elements, nonrandomness of genome rearrangements and point mutations, de novo genes and the origin of life.
If Biology Direct is able to publish very long essays, then I think thispaper could be published in some form, because its viewpoint is provocative andstimulating.
Author response: I would like to thank this eminent referee for his review. It is excitingto hear that he finds this viewpoint provocative and stimulating.
I see the above as a good summary of some of the main points of the paper(though naturally, each author prefers his or her own wording), and I havenow added my own, more comprehensive point-by-point outline to the summarysection of the final version of the paper.
I personally found the article too long, and felt that it could be made moresuccinct with writing discipline. I am sure many readers will find its title,abstract and introduction too grandiose, so the author should think carefullyabout that. I hope it will engender discussion and perhaps a vigorous debatewith proponents of a more orthodox way of thinking about evolution.
Author response: I also hope it will engender vigorous debate. The previous title and earlyparts of the paper have been replaced with much better ones, thanks toProfessor Goldenfeld’s comments, and the rest has been editedthroughout.
The paper is certainly long; however, I would like it to be able to serve asa reference for future works, as well as show the breadth of evidencesupporting the ideas presented here, in order to draw investigators frommultiple fields into the discussion.
Many of the examples in the paper are in fields of biology that are beyond myexpertise or ability to evaluate critically. I hope that another referee hasthat level of knowledge.
I think there are some ways to improve this paper. Claims (1–3) above arein some sense consistent with my own prejudices about the evolutionary process,and probably those of some other workers, so I was hoping that the paper wouldhave more quantitative analysis of the issues it raises, to really force aconfrontation between theory and experiment, or even new theory versus thestatus quo. However, that is not the case: this paper is purely descriptive andqualitative. In particular, I would like to know if the author’sperspective makes novel predictions about such phenomena as the rate ofevolution, the stress-dependence of hypermutation, or the prevalence oftransposable (and other mobile genetic) elements as a function of stress orenvironmental conditions. These and other phenomena are beginning to be exploredquantitatively in the laboratory, so the value of a new perspective will be whatit enables us to compute and predict. I would like to suggest that the authortry to wrap up this paper with a section that gives a concise and explicit setof predictions, even if only qualitative, and discusses how this perspectivecould be advanced to the level of a legitimate theory and potentiallyfalsified.
Author response: A quantitative analysis based on the ideas proposed here will indeed beworthwhile, and I believe that it will greatly facilitate the beginning of anew way of thinking about the evolutionary process. While I will be workingon mathematizing the ideas proposed here and invite others to do the same,my goal in this paper is different—it is to propose, for the firsttime, the conceptual foundation that will make such quantitative analysispossible as well as elicit empirical work directly.
Naturally, those who develop mathematical models tend to focus on the math.But while mathematical modeling is clearly an important tool in biology, theview according to which all that is important is in the math would be toolimiting. Darwin’s own theory of natural selection consisted ofconcepts and empirical observations, showing the power of words. Verbaltheory has also been used to great effect by Lorenz, Tinbergen, Doolittle,Brosius and others, and if we do not use it today we stand to losesomething.
My theory shows, for the first time, how sex, selection on interactions andnonrandom mutation come together as three aspects of one and the sameevolutionary process. Insodoing, it addresses fundamental problems that havebeen open so far, and raises new predictions and new avenues for research.This theory not only is refutable, it makes strong empirical predictions.Take for example the prediction that sex, defined as the shuffling ofhereditary material between individuals by any means, is necessary for theevolution of complex adaptation. One could try to refute it by showing thatany one of the putative ancient asexuals has really substantially adaptivelyevolved or diversified in a purely asexual state. Or take the predictionthat there can be no evolution of a complex adaptation ensuring obligateasexuality, but only breakage events leading to obligate asex. One could tryto refute it by finding a single true adaptation (as opposed to a breakageevent) that ensures obligate asexuality, and opportunities for doing so werediscussed. The general-level prediction that mutational writing mechanismsexist in the germ cells can guide further research, indeed in a directionthat has not been seen from the perspective of traditional theory, andmultiple other questions amenable to empirical investigation have beenraised. Quantitative predictions regarding some of the topics that thereviewer mentions may also be drawn from the ideas advanced here, but Ibelieve that they deserve their own, separate treatment.
In response to the reviewer’s comments above, I have added totheSummarysection an outline of the main points of this paper,including empirically testable predictions and directions for futureresearch.
Some specific comments I have are as follows:
1. Page 4. An important part of the author’s perspective is that information is conserved under allele shuffling. The argument seems to be that information from multiple alleles is combined into one allele, and so is not destroyed by shuffling or even the disappearance of the contributing alleles. How is this different from the well-known concept of epistasis?
Author response: As the reviewer writes, the point of interest is the information-transferprocess itself. This paper explains why, because of nonrandom mutation,information is transferred to future generations from combinations ofinteracting alleles at different loci, despite the fact that the allelescomprising those combinations are continually shuffled. Previous discussionsof epistasis do not mention this point, which plays a central role in thispaper.
That being said, the term “epistasis” is very closely related tothe phrase “interaction between alleles at different loci” as Imean it in this paper, with three differences that are worth noting. First,traditional theory often conceives of epistasis as a small deviation from asupposed, larger, additive effect. In contrast, this paper does not assumean additive basis for adaptive evolution. Second, traditional theory isconcerned with low-order epistasis terms and not high-order ones. Incontrast, this paper leaves room for interactions that are highly complex.Third, and critical to the novel point abovementioned, we are used todiscussing epistasis in the context of its effect on survival andreproduction only. In contrast, here interactions are discussed not only insuch terms but also in terms of their effect on mutation.
2. Page 5. Lamarckism is stated as being impossible but weak forms of it, such as epigenetics, are widely acknowledged to occur. How does DNA methylation for example affect the interactions between alleles in the author’s theory? And what about Landweber’s recent work on the role of RNA in ciliates, which shows evidence for Lamarckian modes of evolution at the molecular level?
Author response: Professor Landweber’s recent work [256, 257], like her other work, is fascinating, cutting-edge research. I think thatthe results from her work, as well as other work on DNA methylation andepigenetics, are in no way contradictory to my paper, and actually would bevery interesting to examine from the perspective of the theory proposedhere.
What I wish to reaffirm prior to proposing a new theory of adaptiveevolution is that traditional Lamarckism is not an answer to the question ofhow multicellular organisms evolve. The reason, as has been articulated wellby Haig , Koonin and Wolf and others, is that we do not expect it to be possible for a mechanism toexist that could sense what the organism needs for improvement at thephenotypic level and then translate it into and implement the needed geneticchange that would cause the desired phenotypic improvement in the course ofthe complex process of development.
However, this theoretical block imposed on Lamarckian transmission does notpreclude actions at the molecular level from having a heritable effect, andin fact my paper here relies on actions of this sort. The point is that, inmy paper, these actions do not “reverse engineer” what is needed for improvement at the phenotypic level, but instead are theresult of a continually evolving mutational writing phenotype that enablesthe absorption of information from selection on genetic combinations.Therefore, I find empirical results like those of Landweber andcollaborators on actions of heritable effect to be very interesting in thecontext of the theory proposed in this paper. But I think that the term“Lamarckism” does not apply to my interpretation of them.
3. Page 7. The section entitled “A prediction following work from Meselson’s lab” needs to be rewritten. Meselson’s work on bdelloid rotifers is described but as a “so-far unofficial result” and indeed references 37 and 38 are to a website and a talk, not to papers. This is strange, because Meselson’s work was published in Science, see: Gladyshev et al., Science 320, 1210–1213 (2008). Since this result predates the present manuscript, this is at best a “post-diction”, certainly not a prediction.
Author response: Meselson’s interesting 2008 paper in Science, to which the reviewerrefers, reports on non-homologous horizontal gene transfer concentrated intelomeric regions , whereas, in contrast, here I am referring to Meselson’s recent (andstill unpublished at the time of this writing) statements that, after yearsof having thought the opposite, they found that bdelloid rotifers undergo“homologous gene transfer” , taken by Meselson as a proof that bdelloid rotifers are sexual. [
Another group has just published in this area , and their publication serves to demonstrate an important point about thedefinition of sexuality. Their evidence suggests lack of conventionalmeiosis in the individual bdelloid rotifer that they sequenced . However, the definition of sex that is of interest for us here is theshuffling of hereditary information at the population level, by any means.Indeed, this group concludes that their results do not exclude“parasexuality”; and that:
“The high number of horizontally acquired genes, including some seeminglyrecent ones, suggests that HGTs may also be occurring from rotifer to rotifer.It is plausible that the repeated cycles of desiccation and rehydrationexperienced by A. [Adineta] vaga in its natural habitats have had amajor role in shaping its genome: desiccation presumably causes DNAdouble-strand breaks, and these breaks that allow integration of horizontallytransferred genetic material also promote gene conversion when they arerepaired. Hence, the homogenizing and diversifying roles of sex may have beenreplaced in bdelloids by gene conversion and horizontal gene transfer, in anunexpected convergence of evolutionary strategy with prokaryotes.” 
Thus, both the statements by Meselson and the statement just quoted callinto question the notion that bdelloid rotifers have evolved without sex(without genetic shuffling).
Whether this should be considered a “prediction” or a“retrodiction” at this point may not be the crucial question.Not only are we still far from having a detailed picture of what hasoccurred in the evolutionary past across the bdelloid rotifers, there is anumber of other examples of putatively ancient asexual clades, and in eachcase my theory predicts that these organisms did not substantiallyadaptively evolve and diversify without some form of shuffling of hereditarymaterial. This should be testable. Traditional theory, in contrast, does notmake a sufficiently serious prediction about the existence of hereditaryshuffling as to be refutable by such tests.
4. The “writing phenotype” plays a major role in this article. I did not feel that I came away with a clear understanding of what that is at the molecular level. Given the exquisite knowledge we have now about genome dynamics (e.g. as summarized in ref. ) it should be possible to be much more explicit about this, in particular to get to the question posed near the end of the article: how does writing [know how to] process information? (I am deliberately removing the anthropomorphic language here).
Author response: The theoretical and empirical exploration of the mechanisms of themutational writing phenotype will be exceptionally exciting. But in contrastto the referee, I do not think that analyzing the workings of the writingphenotype is a simple task, despite our current knowledge of genomedynamics. My goal in the current manuscript is to raise the possibility thata new mechanism for evolution exists. Therefore, I am satisfied withpositing for the first time that the mutational writing phenotype exists,and with tying it to the problem of sex and the nature of the evolution ofcomplex adaptation, and I leave for future research the vast question of itsinternal workings.
5. I would recommend that the title of the paper and indeed the abstract be toned down, and summarise what is actually proposed here rather than the claim to make a new theory. That is, be more specific rather than the rather grandiose but uninformative statement made in the “results” section of the abstract, for example.
Author response: The draft version of the paper that Professor Goldenfeld mentions has beenmuch improved with the help of his comments, though the result is far fromperfect. Nonetheless, the reader should know that, with all thequalifications, I am proposing here no more and no less than a new way ofthinking about how adaptive evolution works. In that sense, it is a newtheory.
I would like to thank Professor Goldenfeld for his very insightful commentsand suggestions for improvement, which I have taken to heart and which Ibelieve have improved the paper substantially.
Reviewer’s report II: Jürgen Brosius, University ofMünster
This is an interesting and thought provoking read containing many“eye-openers” and emphasizing yet unsolved questions concerning theevolutionary significance of sexual reproduction and the proposal of a newtheory in harmony with sex.
Author response: It is a great honor to be told by this pioneering thinker that the paper isthought provoking and has many “eye-openers”.
I would like to thank Professor Brosius for his thoughtful and detailedreview. In his comments outlined below, he will attempt to raisedifficulties of various kinds with the theory proposed here, including thequestion of whether the mutational writing process I propose is more like“writing” or “scribbling”; whether there is a“direction” to the mutational process I postulate; how allelescould influence mutation; the role of transposable elements; and more. Inthe following section, I will answer each question in turn, and explain whytraditional theory does not provide a sufficient explanation for thephenomena discussed here.
However, in my opinion, this attempt falls short for a number of reasons outlinedbelow.
Foremost, I would make a distinction between the introduced term“writing” and a possible alternative, namely“scribbling”. Most, if not all aspects the author has covered seemmore like “scribbling” rather than “writing” (see below)and despite all the efforts to present presumed examples, I am not convinced ofa “writing” process in genomes. It should be noted that thisskepticism comes from someone who does not outright reject ‘genomewriting’. In contrast, I was among the early voices who considered ourrecently acquired capabilities to actively write into genomes, including our ownin a directed manner as a very significant evolutionary transition:
“...Homo sapiens, by being able to influence its own genes stands at thebrink of a significant transition. We will soon have the ability to use genetherapy to correct genetic disease, clone individuals from somatic cells,introduce desired traits or remove undesirable ones, design genes from scratchand introduce additional chromosomes. Lamarckism is raising its head, after all,albeit without violating the Darwinian principles” (reviewer’s ref.1).
“Presently, we are about to witness yet another major evolutionarytransition. Through our advances in biology we are now able to transmitknowledge and experimental experience into the germ line of virtually all livingspecies including our own. We will be able to correct the genetic causes ofhereditary diseases and implant desired traits into future generations. In 3.5billion years of evolution, life was perhaps never so close to some form ofLamarckian mechanism as now (...); whether this is a desirable development is,of course, yet another question” (reviewer’s ref. 2).
Prior to the 1970s/80s, all we did was wait for mutations to occur naturally andselect for desired traits. There was however an intermediate period lastcentury, when we scribbled by increasing random mutation rates aided bychemicals, UV radiation, X-rays and radioactivity in conjunction with the powerof selection in applications such as plant/animal breeding and modification ofmicroorganisms.
Author response: The reader who has started with the reviews before reading the paper shouldnote that the quote above from Professor Brosius’s earlier work, whileinteresting in its own sake, is on a topic other than the one that is infocus in this paper. Professor Brosius is referring to the process ofartificial induction of mutation, whereas the current paper proposes a newtheory about how the mutations that occur naturally drive evolution.
Therefore, the point in his comments above that is of direct relevance tothis manuscript is that he is not convinced of a mutational writing processsuch as I describe here. But it is not a requirement to be convinced at theoutset when a new theory takes on such a grand topic as how adaptiveevolution works. While I hope that he will eventually be convinced of themechanism proposed here, the pertinent question for the moment is notwhether he is personally convinced, but whether the theory proposed hereanswers in a parsimonious way questions that have been left open by previoustheory and raises new predictions.
Consequently, I can subscribe to the “cranes” concept that includeshypermutability by point mutations or retroposition. Without selective advantageunder non-stress situations, lineages that fortuitously kept such mechanisms hada long-term advantage (with hindsight). I have more problems with direction ofthese processes. Too often, such a “unicorn in the garden” turnedout to be a single-horned goat after rigorous experimentation, data analysis,and interpretation (reviewer’s refs. 3–4). Furthermore, during allconsiderations of directed mutations one has to remember that occurrence of the‘right’ mutation is one side of the coin—the other ispersistence of the mutation. I do not know, but would assume that at least somestudies have examined the ratio of neutral mutations nearby (e.g., third aminoacid codon positions) versus the ‘right’ mutations.
Author response: My paper does not propose that a mutation is more likely to occur in anenvironment where it increases fitness than in an environment where it doesnot, and should not be confused with such proposals. I have clarified thisissue in the final draft.
Interestingly, the concept of “cranes” gives no sufficient orwidely-accepted explanation for evolvability. This concept is defined as a“subprocess or special feature of a design process that can bedemonstrated to permit the local speeding up of the basic, slow process ofnatural selection, and that can be demonstrated to be itself predictable (orretrospectively explicable) product of the basic process” . As such, cranes are not well-explained by individual-level selection, andthis is well-known. The inventor of this concept addresses this problem inhis discussion of sex by implying that cranes evolve for a reason that has nothing to do with theevolutionary heavy-lifting that they do—namely, with speeding upevolution. Instead, he implies that they are fortuitously adept at thisheavy lifting. But how often can we excuse by fortuitousness fundamentalbiological phenomena such as the effect of sex on evolution?
A more intuitive approach is to argue that cranes arise by high-levelselection and that this explains why they are so adept at speeding upevolution. However, it is inconsistent for evolutionary theory to propose,on the one hand, that high-level selection is weak [23, 48], and on the other hand invoke it to explain phenomena so important forlong-term adaptive evolution as sex, the constructive contribution of TEs,and more. This contradiction in evolutionary theory, where biologicalphenomena of central interest are either explained by fortuitousness orexplained by a theory that is considered weak by many, reveals a fundamentalproblem unsolved by traditional theory.
My proposed solution to this problem (see the section “Geneticevolutionary trends exist on all timescales”) warrants attention,because it is distinctly different from both sides of thelevels-of-selection debate.
A key concept described on page 4 and illustrated in Figure 1a of the interaction between alleles at multiple loci being“written” into a further single locus that is being inherited is toovague and hard to understand. Vague, because merely presenting the term“interaction” and statements that alleles from different loci mustinteract in the determination of mutation fails to give the reader any clue tothe molecular genetic mechanisms of allele interactions (alleles merely beingvariants of genes and not genes or retroposons etc. per se) and how they couldmodify an additional locus in a heritable mode. It is hard to imagine how oneallele combination would “write” (even scribble) differently thananother. The reader should be enlightened by examples or at least suggestions ofmore detailed molecular genetic mechanisms.
Author response: Consider the ethnic effect of malaria resistance mutations, which Idiscussed in the section “A quintessential example of ns/rm maybe an example of mutational writing”. One could try to argue that eachmutation arises at the same rate in all populations and that different onesare fixed repeatedly in different populations, but that would leave manyfacts of the situation unexplained. It rather appears that different malariaresistance mutations tend to arise in different human populations, whilewithin a population the same or similar mutations tend to arise repeatedly.My theory is the first to explain this evidence in principle. Now, thisevidence suggests that different alleles lead to different mutations. Whileone is free to say that it is hard to imagine how different alleles (asopposed to genes per se) could lead to different mutations, saying it doesnot address the empirical data, which show in fact that they do.
Also, we would not claim that one combination of alleles at different locicould not have a different effect on survival and reproduction thananother—that is the concept of epistasis. Why should we thinkdifferently of the way that DNA sequence and structure and gene productsaffect or bias mutation? There is no fact in our entire understanding ofmolecular and cellular biology that suggests that different geneticcombinations could affect survival and reproduction differently but notaffect mutation differently.
Regarding detailed molecular mechanisms, it would be very exciting to havethem and sooner or later we may have them. However, the goal of the presentpaper is to set the stage for the exploration of these mechanisms by showingindirectly that they exist. It would take many papers to not only show thatthey exist but also lay them out in full molecular detail. Furthermore, thelack of molecular detail should not be confused with the lack of a concreteand important high-level outline of the mechanisms, which in fact has beenproposed here. Scientists now have the option of going ahead and exploringin detail the mechanisms that have been predicted here at a generallevel—from the fact that alleles must interact in the biologicaldetermination of mutation, to the fact that these interactions occur in thegermline, to the possibility that the mutational writing may inform ourunderstanding of the etiology of cancer.
Therefore, I expect the molecular detail to slowly accumulate in the courseof future work. Having stated this opinion, I would now like to explain at ahigher level why incompleteness of the kind seen here is not a problem forthe process of science. While this part of the reply will be lengthy, Ibelieve that it will be informative on the nature of this work as well asthe context of biological thought in which it fits.
The very idea of a mutational writing phenotype as articulated here is new.It provides a unifying and parsimonious framework through which to viewseveral heretofore unexplained and important biological phenomena. It showshow sex, selection on interactions, and the nature of mutation come togetheras different aspects of one process, while raising multiple predictions anddirections for future research. At the same time, the mechanisms ofmutational writing have not been fully articulated as of yet. Thisincompleteness may be a block to some readers, but it is also a part of thescientific process.
Traditional evolutionary theory seemed to give us an explanation forevolution that is complete at the level of the essentials. It holds thataccidental mutation, random genetic drift, and natural selection togetheraccount for evolution. We like to admit that open questions remain, such asthe mechanisms of speciation or the role of sex in evolution as seen from atraditional perspective; but those questions are left at some distance fromthe fundamentals, and in that sense, traditional theory can be said to becomplete. But this completeness may be a bit misleading. Which of theavailable mathematically precise theories truly explains phenomena such asthe arising of de novo genes, chimeric genes and the evolution of complexgenetic networks organized evolutionarily to a large degree by transposableelements? So far, we have had no explanation for these phenomena but purechance and fortuitousness as key factors that complete a theory that wassupposed to be based on natural selection. But here is the crux of thematter. Saying that pure chance explains the initial arising of de novogenes is a completely well-defined thing to say; it admits no lack ofessential knowledge: what occurs by chance no longer needs to be studied,and thus the evolutionary process is presumably completed by chancemolecular events. But we must ask ourselves whether this explanation isrealistic and satisfying. The vagueness that arises from my theory is due tothe fact that, instead of invoking pure chance in addition to an unknownamount of traditional natural selection, I have proposed the beginning of amechanism, incorporating selection and nonrandom mutation in one unifiedprocess. In such a proposal there is much to be asked and to bestudied.
Indeed, incompleteness has played a constructive role in the history ofevolutionary thought. Darwin’s own theory of natural selection wasvague on the mechanisms of inheritance. For this reason it was debated formany decades. It was Fisher and Wright who, borrowing what they did fromDarwin, proposed mathematical theories of natural selection. But althoughtheir theories have been immensely useful, and though they have given theconcept of natural selection a semblance of preciseness, they have leftimportant questions in our understanding of evolution unanswered. Inparticular, interactions are notoriously difficult to treat mathematically.Furthermore, it is much easier to construct mathematical models under theassumption of random accidental mutation as opposed to nonrandom mutation,because then one does not need to add the structure to the mathematicalmodels that would have been needed in order to describe nonrandom mutation.The current paper holds that selection on interactions and nonrandommutation are critical for adaptive evolution. Thus, according to this paper,by putting selection on interactions outside of our core understanding ofadaptive evolution, and by not examining the possibility of nonrandommutation, the mathematical models have missed an important part ofreality.
To understand how the history of the field has led us to our contemporaryway of thinking, we need to remind ourselves of the problem that Fisher andWright tried to solve. As alluded to in this paper, Galton , Jenkin and others never accepted Darwin’s theory of natural selectionbecause of the problem of blending inheritance, which arises in the contextof sex. They thought that traits could not persist in the face of sexualreproduction as to be subject to effective selection. In response, Galtonproposed the notion of saltation—a big, heritable phenotypic changethat on a rare occasion would happen in a single individual and thereafterbe untouched by blending [246, 247]. This avoided use of the ever-present individual variation that Darwinrelied on, on account that selection on it could not have a strong heritableeffect, and instead relied on rare, big changes—a concept that waslater mirrored in Goldschmidt’s “hopeful monsters” and that was ultimately rejected. It was Fisher and others who, in theearly part of the 20th century, tried to bring back Darwin’s relianceon genetic variation across the population in a manner consistent withMendel’s laws, or in other words, with inheritance through sexualreproduction; but interestingly, the way this was done shared an importantelement with Galton’s approach in order to solve the same problem thatGalton tried to solve. That is: Fisher’s mutations also do not blend;they also bring about phenotypic changes that are supposed to occur insingle individuals by chance one day and continue undisturbed from then ondespite the sexual shuffling of genes. Like Gaton’s saltations, theyare independent changes, but “small” ones that can sum up. Thissmallness was supposed to make them more likely to happen by accident,though at the heart of the matter, we still have no quantification of thesupposed Fisherian “smallness” of chance, not to mention that weare no longer using it in an internally-consistent fashion (see introductionto this paper and later discussion in this review). As discussed in thispaper, Fisher’s mathematical framework of additive effects did notinvolve interactions as the drivers of evolution, and it did not actuallyexplain the sexual shuffling of genes, but rather neutralized it; that is,from the Fisherian view, the shuffling of the hereditary material is notnecessary.
In this paper, we have gone back to these deep roots of the modernsynthesis—to the basic assumptions on which the modern theoreticalapproach stands. I have proposed that the shuffling of the hereditarymaterial is of necessity for the evolution of complex adaptation; that itcreates a wide variety of combinations of interacting alleles at differentloci that are then selected; and that a complex adaptation arises by aprocess of “convergence” as described in the paper, whereinformation from combinations of alleles at different loci is put togetherby the writing of mutations. In other words, inheritance involves morebiological detail than we had thought: it involves the heredity of mutationsthat combine information from transient genetic combinations. This amountsto a new connection between selection on the phenotype and geneticevolutionary change.
We can now see both how the theory proposed here does better than Fisher andWright’s theories in some ways and why it is incomplete at the sametime. The additive-effects–based connection between selection on thephenotype and genetic evolutionary change that Fisher invented, while beingperfectly crisp and immediately amenable to mathematization, did away withindividuals as complex wholes. It encouraged instead the very crispperspective that adaptive evolution is based on supposedly accidentalmutations that are normally beneficial as single units. In contrast, thepresent theory offers a connection between selection on the phenotype andgenetic evolutionary change that allows for the first time selection oninteractions to have a direct hereditary effect and thus drive evolution. Itallows selection on complex wholes. This connection is not only moreconsistent with the long standing intuition of biologists that interactionsmust be critical for the evolution of complex adaptations, but also resolvessome of the current mysteries brought about by the genomics revolution (seepaper). Now, this development leaves open the question of the detailednature of the mutational writing mechanisms, and thus makes the theory vagueon an important point. But while the theory as a whole is currently morevague than Fisher’s theory, importantly, it is less vague thanDarwin’s (since Darwin’s theory had no mechanism ofinheritance), and it solves the problem at the basis of the modernevolutionary synthesis in a deep way that is more in line withDarwin’s own observations than the way of Fisher and Wright (see, forexample, the connection between “divergent parallelism”discussed here and Chapter V of the Origin of Species).
In fact, Darwin’s theory was vague not only on the mechanism ofinheritance. It was also vague on the central question of the causes ofvariation. In the beginning of Chapter V of the Origin of Species Darwinwrote: “I have hitherto sometimes spoken as if the variations... weredue to chance. This, of course, is a wholly incorrect expression, but itserves to acknowledge plainly our ignorance of the cause of each particularvariation” . Until now, this incompleteness has left a big hole in our thinking aboutevolution.
Interestingly, in this paper, Darwin’s two great vague and incompleteareas—that of the mechanism of inheritance, and that of the causes ofvariation—(and in fact other mysteries left open by him, such as thequestion of why sex exists) are put together into one: nonrandom mutation ispart of the nature of inheritance and allows selection on complex wholes;the cause of variance and the nature of inheritance are not separate things.But in putting them together, this work opens a vast new vaguearea—the mechanisms of mutational writing.
It is surprising that the author did not even mention the term“epigenetic(s)” once in the entire manuscript (except in about 3references).
Author response: To examine epigenetics from the perspective of the present theory would beof great interest, and I agree that the relevance is clear. Having hadalready tended to so many topics, I had reluctantly decided to leave thisone for future research in the writing of the first draft. I have now addeda very brief mention of it in the summary, thanks to ProfessorBrosius’s comment. This brevity stands in no relation to theimportance of the topic and should not be seen as disinterest or belittlingof it.
Perhaps an important explanation for the role of sex in evolution might be thefact that TEs would not be successful without sex (reviewer’s ref. 5).About a decade later, after the occasional beneficial effects of TEs on genomesbegan to emerge, the same author wrote: “It has been shown that molecularsymbionts (such as transposons and plasmids) derive a major selective advantagefrom conjugation and sexual outbreeding” (reviewer’s ref. 6); seealso (reviewer’s ref. 7). This is impressively documented by the rapidspread of P-elements in the Drosophila genus (reviewer’s refs.8–10). While in the short run, asexual species might have a selectiveadvantage, in the long run only those lineages survived that happened tomaintain sexual reproduction—sex being a relict from as early as the RNAworld [54, 55]. As an aside, bdelloid rotifers might have been sexless for over 35million years, but instead, they have been involved in rampant horizontal genetransfer, which, if not sex, is an efficient substitute . The interesting concept of Michael Ghiselin that a species should beconsidered an individual should also be discussed in the context of theevolutionary significance of sex (reviewer’s refs. 11 and 12).
Author response: I disagree with the traditional approach according to which sex can beexplained by some particular secondary benefit or a conglomeration of those.In the above and in the below parts of this review, the reviewer suggeststhat an important explanation for the role of sex in evolution could be thatTEs are only successful with it; that the reason why many plants are at theoutcrossing end of the spectrum but few are at the selfing end might be dueto better geographical dispersal of alleles or allele combinations (seebelow); and that de novo genes fortuitously arise from the transcriptionalnoise generated by transcriptional promiscuity (see below).
As legitimate as these hypotheses may be, the strength of my theory is inproviding a unifying approach: it explains the role of sex in evolution, howselection on interactions can drive evolution, why the evolution of complexadaptation appears as it does at the phenotypic level, why there isdivergent parallelism at both the molecular level and phenotypic levels, whytranscriptional promiscuity exists in the first place, why there are fewspecies at the outcrossing end, and many other things. A unifyingperspective engages the data better than a perspective that treats problemsin isolation from each other, and it opens up new directions for researchthat cannot be seen from the latter.
Using the term “convergence” in its dictionary meaning of“moving toward union or uniformity” might lead to confusion. A fewpages down, the reader might slip back to the established meaning inevolutionary biology. Perhaps a term “comulgation” from the Spanishlanguage “comulgar”, meaning “to share, to communicate”could be introduced. Unfortunately, it also has a religious use:http://www.wordreference.com/es/en/translation.asp?spen=comulgarhttp://dictionary.reverso.net/spanish-english/comulgar From my ownexperience, though, I have to point out that it is difficult to introduce novelterms, however useful.
Author response: Following much consideration, I have decided to leave the terminology as itis, while fully recognizing the importance of Professor Brosius’spoint.
Concerning the statement: “...the farther we get in time from the earlygeneration, the more the basis of information in the early generation comes tobe shared by individuals” the author should consider that new alleles areconstantly being formed as well.
Author response: Agreed, and this has already been taken into consideration—see Figure 2.
The reason why in plants many species are at the pure outcrossing end, and yetvery few at the pure selfing end might simply be due to better geographicaldispersal of alleles or allele combinations (see above).
Author response: See response above to the comment on the role of sex in evolution andTEs.
The locus of retroposition is, apart from a preference for ubiquitous A/T-richtarget sequences only determined by complementarity of the retroposed RNA3’-end and a ragged-ended DNA strand for priming. Since there are a numberof tailless SINE elements, priming could, in theory, occur at any sequence inthe genome (reviewer’s ref. 13).
The introduction to the first sentence of the second paragraph, page 24 is notquite correct. Of course, there is rearrangement between TEs and thecut-and-paste mechanism of DNA transposons is some sort of rearrangement.However, retroposition is more like a duplication of—if notgenes—but of genetic elements and it is quite random.
Author response: I consider “duplication” to be included under the notion of“rearrangement”. It is easy to call “quite random”things we do not understand, and which could be the result of neither pureaccident nor an omniscient process, but rather the result of a decentralizedwriting system—an ecology of writing activities.
Also, it should be kept in mind that the “donation of every kind offunctional element” mostly requires additional and fortuitous mutationalsteps that may take tens of millions of years to occur and if they occur, theymight not persist, because they still might be neutral or only slightlyadvantageous (reviewer’s ref. 14).
Concerning de novo gene evolution (pp. 30/31), the author states:“First, a previously complete and active gene is duplicated by a single‘duplication mutation’ all at once along with its regulatory andcoding sequences”. This is only partially true: retroposition matchesexisting mRNA reverse transcripts with novel regulatory elements(reviewer’s ref. 15).
Author response: The point that I am making is not that only the whole gene duplicationroute is available to new genes, but that only this route is consistent withthe traditional idea of the gradual accumulation of small-effect chancemutations under traditional natural selection, and that it may be imprudentto dismiss the other routes, which are inconsistent with this traditionalidea, as the result of pure chance without much thought.
Author response: References added.
Much more common is the exaptation of novel gene parts from retroposons oractually any neutral sequence (reviewer’s refs. 16–20).
Author response: I agree with the importance of movements of genomic pieces and ask whethermechanisms (indeed, evolvability) or only pure chance are involved in theirmovements.
Re-wiring of the gene regulatory landscape of endometrial stromal cells (ESCs) ofthe placenta, if true, only can be a random process. If 1,500 MER20 elementswere recruited into this regulatory network, what about the remainder of the15,000 MER20 elements in the human genome? I highly recommend the criticalreader to look at the chapter (actually the entire manuscript is excellent)entitled “Transcription factor binding does not equal function” byDan Graur and colleagues . Furthermore, although Lynch et al.  could show with reporter constructs ex vivo that MER20 elementsrespond to progesterone/cAMP in ESCs, it is only part of the confirmation of aregulatory network. The problem with these and similar studies is, that currentscience politics might grant us the time to prove a working hypothesis but notto falsify it (reviewer’s ref. 21). Not many laboratories can afford theleisure to test the influence of TEs on gene expression by costly andtime-consuming targeted deletions in mouse or other animal models.
Author response: What we are concerned with here [15, 205, 206]is the evolutionary organizing by transposable elements of more than 1500genes into a new genetic network underlying a novel, complex adaptation thatis the decidualization of the endometrium. It is not that we had not knownbefore that TEs play a constructive role in evolution; it is rather themassiveness of this example that is intriguing. This work comes out ofGünter Wagner’s lab, who has been a leader in evolutionarybiology, pushing the envelope on our understanding of evolution throughouthis career. These results provide strong p-values for the nonrandomassociation of MER20s with this network, and in my opinion they are quitechallenging as they are.
The problem that these results raise is as follows. If one were to explainfrom traditional theory the evolution of a network of this sort, the mainway of doing so would be to say that it is due to some mix of selection andneutral evolution. But how much fortuitous chance would be involved in sucha mix? How many neutral movements of TEs and neutral mutations in them hadto take place before something was established that could be subject totraditional natural selection and explain the arising of a new network, ifwe operate under the assumption that it is accidental mutation and naturalselection that explain things, and does this explanation make sense?
The reviewer argues that, assuming that 10% of MER20s are involved in tyingtogether this network, and that the rest fall elsewhere, it must have been arandom process that gave rise to this network. Perhaps this argument wouldhave been true if the only alternative to accidental mutation were anomniscient process that frugally used each type of TE for one purpose; butaccording to this paper, this is not the only alternative.
In fact, saying that the arising of this network must have been a randomprocess is problematic. At its core, the traditional view of adaptiveevolution holds that small chance-events occur, and selection pulls out ofthe noise beneficial changes, which can thus accumulate and create anadaptation. When faced with evidence not fitting with this view, such as denovo genes, this view forces us to argue that it is still just a smallgenetic change that arises by a sequence of fortuitous chance events absentselection (a small whole gene, in this example), and that it could happen byaccident after all (we will see if this approach is valid below). But in theexample discussed here we have the evolution of a network of more than 1500genes that come together to underlie a complex, novel adaptation. Therefore,the question that this example helps us highlight is this: Where do we drawthe line? When is the amount of accidental chance that we invoke forexplaining the evolution of complex adaptation from the traditional view toomuch, and when is it not too much? The answer that we are currently using isobvious: the line is defined post hoc so that it always includes everyempirically discovered case of evolution as one that could in principle beexplained by traditional accidental mutation, random genetic drift and someunquantified amount of selection. These post hoc explanations harbor thedouble standard of saying that we have an explanation for the evolution ofadaptation, in the form of a process where natural selection inexorablytests many mutations, among them beneficial ones, each of a slight enoughbenefit that could presumably have arisen by accidental chance, while at thesame time invoking just as much additional chance as the situation requires,in the form of neutral evolution that also happens to play an inherent rolein the evolving adaptation (this invocation of additional chance undoes thewhole point of the frugality of the reliance on chance, a frugality whichwas supposed to be the anchor of the scientific explanation). This doublestandard is a severe problem with the traditional view of adaptiveevolution.
We are therefore led to ask whether there are mechanisms involved in thepresent example beyond those considered by traditional theory. I would agreewith the view that the proliferation of TEs of the same kind alreadyequipped with cryptic binding sites [205, 206]could subject many genes to activation by the same transcription factorsand is therefore a very good way of tying those genes together [15, 208]. But note that this view makes TEs inherently useful for evolution in amechanistic way, and this prepotency is more congruent with the theory thatI have proposed here than with traditional theory.
Two to three decades ago, it was extremely difficult to convince the scientificcommunity about occasional exaptations of a TE or part thereof into a novelfunction. More recently, we have the opposite problem. Namely, that there aremany attempts to sweepingly assign functions to the majority of TEs in thegenome. While it is clear that we still have a lot to learn about the grammar ofgenomes, trying to read too much into their structures is somehow reminiscent oftea-leaf reading akin to the “bible code” (reviewer’s ref.22).
Author response: My paper does not suggest that all or most TEs play a role in survival andreproduction. What I have suggested is in line with Fedoroff’s viewthat TEs are inherently useful for evolution . It does not follow that the majority of TEs have functions in theperforming phenotype at any given point in time. The usefulness of TEs forcurrent survival and reproduction and their usefulness for evolution are twodifferent things .
A similar problem concerns the detection of RNA transcripts from >60%of the human genome by ultra-deep RNA sequencing which leads to absurdities ofequating function to any aberrant RNA snippet and hence claiming, as ENCODErecently did, that most “junk” DNA is functional indeed. The authorappears to fall into a similar trap: “Transcriptional promiscuity is ahighly involved mechanism, requiring the orchestration of a complex machinery toboth create and compensate for the pattern of expression , and its evolutionary origin is a mystery”. Others would callit basal levels of transcription or insufficient transcriptional silencing orread-through transcription or spurious transcription initiation and elongationetc. ( and reviewer’s ref. 23). It simply is an imperfection akin topoint mutation where replication is not completely error-free. However, I agreewith the notion that such transcripts have the potential to fortuitously lead tonovel genes encoding functional RNAs, or even protein coding mRNAs out of thesespurious mostly low-level transcripts (reviewer’s ref. 23. and ).
Author response: The reviewer touches here on an important difference between his view andmine. His view is that TP is error, and that mutation is error. My view isthat TP is hard to explain biologically as an error, because it requiresevolved adaptations to compensate for it . I hold that this fact, along with its ability to allow interactions in thewriting of mutations as predicted by my theory, makes TP an intriguingphenomenon.
From the perspective of the theory proposed here, many differentobservations fall into place as pieces of one puzzle, includingtranscriptional promiscuity, molecular parallelism, the nonrandomness ofmutation that comes out very clearly from the empirical evidence, and muchmore, as already stated in the above comments and in the paper. From thetraditional theory, these things are dismissed one by one: mutation is arandom accident (despite all the evidence to the contrary—traditionaltheory does not provide an answer to the paradoxes that I have elaborated onin this paper); mutation hotspots just fortuitously happen to be in lociundergoing concentrated adaptive evolution; transcriptional promiscuity isjust an error that happens to take place, of all places, in the cells whereit can affect evolution, and, fortuitously, it is not disruptive; TEsfortuitously acquire functions, so much so that a notable percentage of TEsof a particular kind have played a substantial role in the evolutionaryorganizing of a complex network of more than 1500 genes; the incredibleevolutionary usefulness of TEs is then explained [8, 119]as being partly the result of extremely high-level selection, even whilethe effectiveness of high level selection is far from being widely accepted,for basic reasons [23, 48]; extensive molecular parallelism just happens to happen, and it is thensimply assumed that the mutation rate is high enough to allow it to happen,even while cases of parallelism such as the independent arising of theadaptive TRIM5–CypA gene fusion in different monkey lineages [105, 107–112]show that the assumption of random mutation is faltering (not to mention the curious connection of the high expression level of theCypA gene in the germline , which is precisely in accord with the theory presented here), and indeedno calculation is provided by traditional theory showing that accidentalmutation can account for such cases; mutational mechanisms are reported andlabeled as “deeply perplexing” , are discussed here at length in connection with organismal-levelobservation crucial to Darwin, but again these observations have been putaside by traditional theory for the lack of a suitable theoreticalframework; and so on and so forth. This approach dismisses many criticalobservations and explains away others in an ad hoc fashion, whereas, incontrast, the present paper provides a unifying framework that seriouslyengages the findings and that opens up new avenues for future research.
Comment concerning the statement “To address de novo genes from atraditional viewpoint, it is said that Jacob did not know that there is so muchtranscriptional ‘noise’...”
Above all, Jacob did not know that genomes of multicellular organisms contain somuch non-functional DNA as cradle or cauldron for de novo genes.
“This example shows that there is a severe problem of lack ofquantification of the amount of random chance that we call upon, not to mentionsuch facts as that the Poldi de novo gene arose with an existingalternative splicing pattern .”
This is no surprise, as a novel gene is not expected to arise with perfect splicesites; hence alternative splicing patterns are common.
Author response: I argue that there is no calculation that shows that it is reasonable toexpect genes to arise de novo, even accounting for the large size of thepool of transcripts to draw from, and the quotations above are taken fromthe part of my paper that makes this argument.
Some would dismiss de novo genes as no surprise, on account that de novogenes are small enough, and that the pool of transcripts to draw from islarge enough, so that it is reasonable to assume that once in a while one ofthe transcripts will find use as a new gene by chance. But the Poldi gene is853 nucleotides long with exon 2, and 785 without. There are 4785 random genetic sequences of this smaller length. This numberdwarfs the number of atoms in the visible universe, and thus also dwarfs thenumber of transcripts available due to TP. The question therefore is not whetherde novo genes are “sufficiently small” or whether TP provides“so many transcripts”—these statements do not address thechallenge that the evidence has brought forth. The question is what fraction ofrandom sequences of such sizes will be useful in any given organism. Theliterature does not provide an answer to this question. There is no hint of acalculation or empirical evidence showing that a de novo gene can arisefortuitously without involvement of selection. Our lack of ability to answerthis question from traditional theory should be acknowledged as a problem. Incontrast, my theory begins to address this issue, by saying that there aremechanisms in place that enable the evolutionary route taken by de novo genes,mediating between them and selection. Interestingly, my theory argues that TP isone of these mechanisms.
In trying to explain de novo genes in a way other than just by saying thatthey arise by pure chance, one might argue that there must be smallerintermediates on the route to a de novo gene, and that those intermediateswere somehow subject to natural selection. I would agree with this line ofreasoning, but add that if one wanted to explain the arising of anintermediate in a de novo manner, the same question would apply again. Thenumber of random genetic sequences only 50 nucleotides long still dwarfs thenumber of transcripts available due to TP. Furthermore, de novo pieces are aproblem all the way down, since at some point the many de novo pieces alsoneed to be connected together, and that would require again anunquantifiable amount of pure chance according to the traditional view.
I would like to thank the reviewer again for sharing his highly informativeviews and his expert knowledge, which greatly helped to explore points ofdifficulty and made a very important contribution to this work, as well asfor other helpful comments of his not included in the above, all of which Ihave taken into consideration.
Reviewer’s report II: reference list
Brosius J: From Eden to a hell of uniformity? Directed evolution in humans. Bioessays 2003, 25(8):815–821.
Brosius J: Disparity, adaptation, exaptation, bookkeeping, and contingency at the genome level. Paleobiology 2005, 31(2):S1–S16.
Stahl FW: Unicorns revisited. Genetics 1992, 132(4):865–867.
Smith KC: Spontaneous mutagenesis: experimental, genetic and other factors. Mutat Res 1992, 277(2):139–162.
Hickey DA: Selfish DNA: a sexually-transmitted nuclear parasite. Genetics 1982, 101(3–4):519–531.
Hickey D: Molecular symbionts and the evolution of sex. J Hered 1993, 84(5):410–414.
Wright S, Finnegan D: Genome evolution: sex and the transposable element. Curr Biol 2001, 11(8):R296–R299.
Good AG, Meister GA, Brock HW, Grigliatti T, Hickey D: Rapid spread of transposable P elements in experimental populations of Drosophila melanogaster . Genetics 1989, 122(2):387–396.
Engels WR: The origin of P elements in Drosophila melanogaster . BioEssays 1992, 14(10):681–686.
Silva JC, Kidwell MG: Evolution of P elements in natural populations of Drosophila willistoni and D. sturtevanti . Genetics 2004, 168(3):1323–1335.
Ghiselin MT: A radical solution to the species problem. Syst Zool 1974, 23(4):536–544.
Ghiselin MT: The Economy of Nature and the Evolution of Sex. New York: University of California Press; 1974.
Schmitz J, Churakov G, Zischler H, Brosius J: A novel class of mammalian-specific tailless retropseudogenes. Genome Res 2004, 14(10A):1911–1915.
Krull M, Petrusma M, Makalowski W, Brosius J, Schmitz J: Functional persistence of exonized mammalian-wide interspersed repeat elements (MIRs). Genome Res 2007, 17(8):1139–1145.
Brosius J: Retroposons–seeds of evolution. Science 1991, 251(4995):753.
Lev-Maor G, Sorek R, Shomron N, Ast G: The birth of an alternatively spliced exon: 3’ splice-site selection in Alu exons. Science 2003, 300(5623):1288–1291.
Singer SS, Männel DN, Hehlgans T, Brosius J, Schmitz J: From “junk” to gene: Curriculum vitae of a primate receptor isoform gene. J Mol Biol 2004, 341(4):883–886.
Krull M, Brosius J, Schmitz J: Alu-SINE exonization: en route to protein-coding function. Mol Biol Evol 2005, 22(8):1702–1711.
Möller-Krull M, Zemann A, Roos C, Brosius J, Schmitz J: Beyond DNA: RNA editing and steps toward Alu exonization in primates. J Mol Biol 2008, 382(3):601–609.
Baertsch R, Diekhans M, Kent WJ, Haussler D, Brosius J: Retrocopy contributions to the evolution of the human genome. BMC genomics 2008, 9:466.
Popper K: Unended Quest: An Intellectual Autobiography, Karl Popper. London: Routledge; 1993.
Drosnin M: The Bible Code. New York: Simon & Schuster; 1997.
Brosius J: Waste not, want not–transcript excess in multicellular eukaryotes. Trends Genet 2005, 21(5):287–288.
Reviewer’s report III: W. Ford Doolittle, Dalhousie University
I confess that I found this a very irritating essay, and several times nearlygave up reading it. It is clear that Prof. Livnat has thought and read much anddeeply about evolution and does seem to be offering hope for a new conceptualframework within which to rationalize observations that many claim to findpuzzling. He admirably summarizes a vast number of phenomena which neoDarwinistshave to stretch themselves to rationalize, and argues that these are“consistent with the present [that is, his] theory.”
But I’ll be damned if I can tell you just what exactly that theory is.Livnat packages a variety of accepted observations about epistatic interactionsas these affect not only gene expression but also mutation under the notion of“writing”, but these seem to me not much different than the sorts ofthings people are referring to when they write about “evolvability”and claim (reasonably enough) that it too evolves. Nobody thinks that mutationsare not complexly caused, nor that evolution does not impinge upon mutationalmechanisms and their specificity. Nor does any sensible contemporaryneoDarwinian deny that the complexity with which function is determined in agenome such as ours has important effects on the direction and speed with whichit evolves.
Most interesting, and possibly the core of the “present theory”, isthe notion that although synergistic interactions between multiple alleles atunlinked loci brought together by sex and recombination are transient (becauseof sex and recombination), they may by their joint mutational effect on somethird locus result in a novel and potentially useful new allele which in itssingleness can be a permanent contribution (not broken up by recombination). Sosex is important, indeed foundational for evolution, because it createsadvantageous new genes that are immune to it. Maybe this is an interesting newway of looking at things: time will tell.
Author response: Addressing the problem of sex is no small matter. The reviewer acknowledgesthe novelty of the hypothesis that lies at the core of the theory that Ihave proposed here and writes that it would make sex “important,indeed foundational for evolution,” and that maybe it is “aninteresting new way of looking at things”.
The reader familiar with previous hypotheses on the role of sex in evolutionshould note that this hypothesis is very different from the previous ones,because it connects sex and mutation while implying that the nonrandomaspect of mutation is critical: it allows selection to act on combinationsof alleles at different loci as interacting wholes while having a heritableeffect.
We now need to pursue the consequences of this hypothesis to see how theyaddress the reviewer’s questions.
Traditional theory only defines random mutation with respect to its effectson immediate fitness. As evolutionary biologists, we are quick to admitthat, in other senses, mutation need not be random, and that variouscomplicating factors may cause the mutation rates to be higher in someplaces rather than others, or even affect what mutational change will takeplace. But the question that my paper raises is whether these complicatingfactors are or are not of profound consequence for our understanding of theprocess of adaptive evolution. My paper holds: “yes, theyare”.
The reviewer writes that no one would dispute that evolution impinges uponmutational mechanisms, and indeed some of the most inspiring papers inpopulation genetics have been written on the evolution of modifiersaffecting the mutation rate or the recombination rate [61–63, 263–268]. But the ways in which evolution is thought to impinge upon mutationalmechanisms are not a systematic part of our traditional explanation of howadaptive evolution happens at its core. The core is that of ns/rm plusdrift, and on top of this basis various effects have been modeled. Thistraditional core is very different from what I am proposing here. To makethis explicitly clear, the traditional idea of the mutation that drivesevolution is that mutation is the result of accident. From that perspective,the complicating factors that Professor Doolittle claims no one disputesare, to our deep understanding of evolution, complicating factors. They arenot front and center. In contrast, my theory states that the mutation thatis of relevance for the evolution of complex adaptation is not the result ofaccident, but the outcome of an evolved and continually evolving biologicalphenotype. What were previously considered “complicatingfactors” are actually the basis of things. The possibilities forongoing mutation are defined through genetic interactions, and this fact isat the heart of the evolutionary process.
Thus, while the reviewer writes that nobody would dispute that mutation is“complexly caused,” the question at hand is whether thesecomplex genetic influences on mutation are a fundamental part of theadaptive evolutionary process. If sex becomes foundational according to thepossibility that the reviewer recognizes, then at the same time theseepistatic effects on mutation become foundational, because they are the onesthat underlie this hypothesis on sex in the first place.
Note that, according to this view, there is no selection acting onaccidental variation that, by acting indirectly on modifiers, or by favoringsome higher level entities (such as species or clades) over others, evolvesevolvability, because adaptive evolution is not based on traditionalaccidental variation in the first place.
The key to understanding this paper with regards to evolvability isunderstanding the full implication of the assumption that mutational writingcan be seen as a phenotype, and that there is no adaptive evolution but thejoint evolution of the writing and performing phenotypes. As explained inthe section “A more detailed look into the new theory”, if mutational writing is a phenotype, then this immediatelyimplies that it would include both taxonomically shared phenomena thatdefine the possibilities for genetic change at a general level, such as sexand recombination, as well as more specific influences on the possibilitiesfor change, up to and including the individually varying epistaticinfluences on mutation that figure into the hypothesis on sex that thereviewer recognizes. This means that we do not treat evolvability as asecondary issue: inference of the writing phenotype from the many pieces ofevidence discussed in this paper implies evolvability directly. Importantly,there is no longer a question from this perspective of how the writingphenotype (and thus evolvability) evolved independently of other biologicalstructure, as though we are still looking for an explanation of origins inan ns/rm core. The mutation that drives evolution has always been theoutcome of biological actions, and this biological activity from the“beginning” has evolved along with the performing phenotype toits present state.
This of course ties to the view that sex is original.
Given this theory, our understanding of evolvability is improved. Whentrying to explain, from the traditional theory, the evolvability provided bysuch phenomena as sex, recombination and an evolutionarily productive rateof mutation, there is a problem. Evolvability, by definition, is somethingthat facilitates population-level evolution. It is not a property of anindividual, because an individual does not evolve; populations do. Theindividual does not benefit in terms of its own fitness as compared to thefitness of other individuals in the population based on how evolvable thepopulation that it belongs to is. Therefore, how can traditional naturalselection acting on individuals lead to the evolution of evolvability?
Working from the traditional theory, one possibility is to propose thatevolvability evolves not based on individual-level selection, but based onselection at the level of groups, species, or clades (e.g., ). However, high-level selection is considered by many theoreticians to beweak [48, 263]. Therefore, to be forced to explain complex biological phenomena that areof much importance for evolvability by applying high-level selection is tobe in a weak position. I will discuss this in detail later.
Another possibility is to address evolvability through modifier theory . While this theory avoids high-level selection entirely, it also recognizesthe problem that evolvability is not necessarily favored in the process ofindividual-level selection and is a priori agnostic on what outcome toexpect . In this approach it is assumed, for example, that one locus controls themutation rate at another locus or the recombination rate between two otherloci, and a model is constructed to examine the evolution of allelefrequencies in that gene [61, 62, 263]. Note that these models do not require epistasis in the determination ofthe modifiers’ action (whereas my theory requires it for the corehypothesis on sex, highlighting a difference in mechanism). Moreimportantly, these models presumably have been interpreted as though themutational cause of the alleles in the modifier locus and in the other lociis accidental.
This important modeling framework has actually exposed a difficulty inevolving evolvability within the traditional framework, namely the reductionprinciple [61–63]. This principle shows that traditional natural selection indirectlyaffecting modifier loci is actually often expected to decrease mutation andrecombination rates (to shut down evolvability) [32, 61–63, 263–266]. Conditions can be found where the opposite happens, and the behavior ofthe system is complex (e.g., [267, 268]). But ultimately, this modeling framework does not provide a systematicsolution to the evolution and maintenance of evolvability based on ns/rm [32, 266]. Quite the contrary, I would argue that its greatest contribution was inshowing the lack of such a solution from the traditional theory , thus highlighting the problem of evolvability for traditional theory.
According to my theory, we do not need to look for circuitous explanationsfor evolvability from an ns/rm core. Inference of the writing phenotype fromthe many empirical observations discussed in this paper immediately impliesevolvability, and the writing phenotype did not evolve from ns/rm, butrather, just like the performing phenotype, has been around as long as lifehas been around, and both have become more elaborate. This explainsevolvability in the sense that it now becomes yet one more of many piecesthat are connected in a parsimonious framework.
Why I didn’t give up reading was because Livnat considered but thendismissed the sort of levels-of-selection explanation for the evolvability ofevolvability that I myself endorse (his references [8, 16] and ), thus giving me the opportunity to put in a word for it in thisreview. He writes that “TEs [transposable elements] can have theappearance of selfish elements yet be an inherent part of the mutationalmechanisms that serve the evolution of organisms”. I would say thattransposable elements are selfish and spread at the genomic level because ofthat, though they are disadvantageous at the level of individual organismswithin species, where they are selected against. But because possession of thesemutational agents may speed speciation or delay extinction at a still higherlevel (species?), they can be seen as advantageous and selected for at such alevel.
Livnat equivocates on this view, indeed rather cops out. He writes that“Doolittle offered a way of resolving this conflict, by proposing thatclade-level selection is responsible for the existence of a system hospitable toTEs due to their long-term usefulness. But the debate over whether selection atlevels above the gene and individual is strong enough to affect such things isfar from resolved. Hence we admit that the question is open”. Mostbiologists know that group selection as espoused by Wynne-Edwards is in very badodor and shy away from invoking it. But surely if species or higher taxa differin their evolvability because of TE accumulation they will be differentlysuccessful in the long run even if the TEs spread because of selfishness. Maybemore theory needs to be developed here, divorced from issues of altruism andcomplex population behaviors (the usual battleground). So it is true that the“question is open”, but that phrase should not be read asdismissive.
Author response: I do not mean to equivocate on this issue, and I will make furtherclarifications here. In his writings [16, 119], Professor Doolittle acknowledges the problem of high level selection. AsWilliams argued, selection at levels higher than the individual is much weaker thanselection at the level of individuals for the following reason. There aremany generations of individuals, and many individuals in each generation.Selection has much to choose from, and is therefore strong. By comparison,how many groups are there, and how often do they give rise to“offspring groups?” Likewise, the turnover rate of species isincredibly small relative to the turnover rate of individuals. Selection hasfar fewer opportunities to act, and thus the potential for the buildup of anadaptation by high-level selection is far smaller than that which we expectfrom individual-level selection. Make no mistake: by reminding the reader ofWilliams’s argument, I do not mean to imply that I am a supporter ofWilliams’s world-view, which I am not. Rather, what I mean to state isthat, as long as we work from the framework of traditional theory, thatargument is still relevant, and assigning importance to the explanatoryvalue of high-level selection proposals is not a straightforwardmatter.
Now, I do not disagree with Professor Doolittle that TEs can be described asthough they are selfish elements on the molecular level, with costs at theindividual level and long-term usefulness for evolution. However, I amconcerned with the question of what explains the fact that they are usefulin the first place, evolutionarily. Using only the past arsenal of ideas,one possibility is to propose that the system of TEs and their regulationare fortuitously useful in the long-term, and that once they are there,high-level selection plays some role in their prevalence [8, 16, 119]. Another possibility is to suggest that high-level selection has graduallybuilt up the system of TEs and their regulation and thus explains the originof their long-term usefulness. One could also combine these possibilities,proposing an initially fortuitous usefulness and further gradual elaborationof it by high-level selection. Doolittle’s arguments in [16, 119], emphasized in the context of the evolutionary usefulness of intronsthrough exon shuffling, but also discussed as applicable to TEs, raised thefirst possibility, and a parenthetical note in his 2013 paper (page 5298) may also be read as consistent with the third. Whileconsidering these options, Professor Doolittle consistently admits therelative weakness of high-level selection, implying the untenability ofmaintaining a trait by high-level selection in the face of noticeableindividual-level costs, and he carefully constructs his arguments in amanner that recognizes this problem. In the case of TEs, he relies on TEsbeing selfish elements at the gene level to provide their own proliferationand maintenance, thus countering from below the individual-level selectionpressure against them (in other words, TEs are favored at the gene level(strongly), selected against at the individual level (strongly), andselected for again at the species or clade levels (weakly)). However, asmore empirical results are obtained, the more we see that the contributionof TEs is immense, and that their regulatory system is complex [15, 19]. To me, these findings make it increasingly unappealing to explain theorigin of the usefulness of TEs by fortuitousness. They also make thequestion more pressing of whether high-level selection can gradually buildup a complex system with notable long-term benefits despite its relativeweakness. In the explanations above, only fortuitousness and high-levelselection are combined, and so the less we use one, the more we use theother (and I have just criticized both).
This problem can be sharpened by discussions of introns and of sex, becauseboth are phenomena that, like TEs, and like an evolutionarily productivemutation rate, provide evolvability and thus need an explanation . In his chapter of 1990 , Professor Doolittle provokingly writes that the existence of introns andthe entire apparatus that allows for exon shuffling is in a sense moreinteresting than the entire part of the evolutionary process thattraditional theory attempts to address. Rather than small quantitativechanges, this apparatus allows for “quantum leaps” through thecreation of new genes and enzymes. He then attempts to address this issuewith high-level selection, but again admits that all that this selection canbe expected to do is favor species that for one reason or another have lostfewer of their introns or have their introns positioned better in terms oftheir long-term usefulness through exon shuffling. However, the origin ofthe usefulness is not thus explained, and seems to be left tofortuitousness; and once we admit that the more important part of evolutionis enabled fortuitously by the existence of a complex biological system, itis not clear how much of evolution is really explained or is explainable bythe traditional theory anyhow.
Consideration of sex as a phenomenon that provides evolvability and thatneeds an evolutionary explanation also helps to sharpen the problem above.In the preface to the 1996 edition of his book—the book where he hadargued that there are no high-level adaptations—Williams conceded thatperhaps his greatest mistake regarded his discussion of sex . Previously he had interpreted sex as a complex adaptation elaborated byindividual selection. Now he admitted that he had underestimated theindividual-level costs of sex; that it had long-term benefits; and thathigh-level selection most likely plays a role in explaining it. He now seemsto treat it as an exception, aligning himself with common wisdom. But thereis a point that I believe he missed: If the rule is that high-leveladaptations do not exist because high-level selection is much weaker thanindividual-level selection, then if a certain evolved adaptation stands asan exception, appearing to be a high-level one, would we not expect it to besimple rather than complex, and of little rather than substantialindividual-level costs, so that it would not strain the difference ineffectiveness between the different levels of selection? Is it not a bitstrange that the one case that evolutionary biologists tend to make anexception for is more weighty than all of the other traits that have beendiscussed in the context of the levels-of-selection debate, one that is sohighly complex and advanced in its biological mechanisms of implementation,and that affects the structure and function of the organism across thescales of organization so thoroughly—indeed that defines the processof selection and inheritance (see the section “Fundamentalproblems in traditional evolutionary theory: sex andinteractions”)?
Given the paragraph above, and given the relatedness of the phenomena abovein terms of them being different manifestations of the problem ofevolvability, I actually agree with an earlier quote from ProfessorDoolittle’s work—from his famous 1980 paper with Sapienza.Discussing the possibility of explaining TEs by high-level selection, theywrite:
“The selective advantage represented by evolutionary adaptability seems fartoo remote to ensure the maintenance, let alone to direct the formation, of theDNA sequences and/or enzymatic machinery involved. A formally identicaltheoretical difficulty plagues our understanding of the origin of sexualreproduction, even though this process may now clearly be evolutionarilyadvantageous.”
I argue that the evolvability that ties together the question of sex andTEs, and that Doolittle and Sapienza concluded could not have arisen byhigh-level selection, is no more satisfactorily “explained” byfortuitousness.
In this paper, I have provided an alternative: the mutational writingphenotype implies evolvability directly, and this ties to a newunderstanding of sex and of the complex factors affecting mutation. Both sexand these complex influences on mutation become central to the process:while the shuffling of the genes creates new genetic combinations, thewriting of mutations combines information from different loci and thusallows selection on individuals as complex wholes to have a hereditaryeffect in accord with fitness—which is actually not allowed bytraditional theory in sexual populations.
As a further means of clarification in light of this reviewer’squestions, notice further differences between my theory and previous work.Modifier theory, in addition to not relying on epistasis in the control ofmutation, is split into theory of selected modifiers, which do not concernthe mutation and recombination rates, and theory of “neutral”modifiers, which do (but do not themselves affect survival andreproduction). In contrast, my theory holds that the control of mutation isepistatic, and that genes participate pleiotropically both in the performingand in the writing phenotype. For example, my theory predicts that thefusion between TRIM5 and CypA [105, 107–112]involved genes that, on the one hand, participated in the performingphenotype, and on the other hand participated in genetic activity in thegermline that ultimately led to their fusion . This kind of epistatic activity affecting mutation and performancepleiotropically has not been modeled, and it demonstrates that the controlof mutation through genetic interactions is at the heart of the evolutionaryprocess. Indeed, I would argue that the “quantum leaps” of thegeneration of new genes that Professor Doolittle refers to in his work are not random, and that a gradual process of evolution involving both thewriting and performing phenotypes predisposes the genetic system to producethem. Thus, this paper offers a new look on germline genetics, suggestingavenues for research not conceived of from traditional theory, withpotentially intriguing implications (see, e.g., the subsections “ Many writing mechanisms may exist in the sperm cells ” and “ De novo gene evolution may be subject to indirect natural selection throughthe writing phenotype ”).
As I have argued above, there is no longer a question of how the writingphenotype itself evolved, as though this question can be separated fromothers. Evolution occurs by the joint evolution of the writing andperforming phenotypes. Biological action has always been central tomutation, and one should not look for an origin in an ns/rm core. Instead,the question becomes: How does the joint evolution of the writing andperforming phenotypes exactly happen? In this paper I have merely begun todescribe this process (see points on sex, on convergence, and on selectionbeing not a passive judge of phenotypic meaning, but an active participantin its formation). To make an analogy, the “learning apparatus”of evolution is not accidental mutation running through a sieve; it is alearning apparatus that absorbs information from selection based on itsabilities evolved so far and that grows along with the information that itabsorbs over evolutionary time.
Despite differences between my viewpoint and Professor Doolittle’s onthe topic of high-level selection, I would like to mention that I have foundhis publications on evolvability in the contexts abovementioned inspiring.While others have commonly either ignored the question altogether or havenot even recognized that it exists, this luminary has been unique indiscussing it prominently and openly and in articulating the interest forevolution that lies in it. This has been both a great inspiration for me andhas also had a profound effect on my own thinking and the development of myideas on these topics.
I would also like to mention that, in addition to his points of criticismwhich I have addressed above, Professor Doolittle writes that my hypothesison sex would make it fundamental for evolution; that I have summarized a“vast number of phenomena which neoDarwinists have to stretchthemselves to rationalize”; and that I do seem to be offering hope fora new conceptual framework within which to address these phenomena.
I would like to thank him greatly for his time and effort in reading andcommenting on the first and less clear draft of this manuscript. In aneffort to clarify the paper, I have revised key parts of the textsubstantially and have added a point-by-point outline of the materialdiscussed (see Summary section).
a By defining the mutation that drives evolution as described in this paper, Ido not mean that this mutation is nonrandom in a global sense. Events affectingmating as well as other events affecting the outcome of recombination and thewriting of mutation are not predictable as a whole and provide sources of“randomness” in the sense in which this word is normally perceived. (Inother words, mutation depends on the individual genotype, and the composition of thelatter is, to an important degree, random.) Thus, in a global sense, the mutationthat drives evolution as described in this paper is still random. However, thenature of this mutation as described here is unambiguously different from the oneheld by the traditional theory of evolution and nonrandom in that this mutation iscaused by an organic process that is part of the evolving organism. In fact, it isthe outcome of evolved and continually evolving biological system. This new conceptof the mutation that drives evolution (not inclusive of all the mutations that causedisease) is further developed in the section “A more detailed look into thenew theorypredictions for molecular evolution”.
b These are interpreted as decoys because multiple investigators haveoccasionally observed parasites in them, and the cost of being located by a parasiteis large and obvious .
d Interestingly, and in accord with the present view, as willbecome clearer later, the SOS response in terms of the increase in mutation rate isin fact not equal across the genome but modulated by hotspots .
e To address de novo genes from a traditional viewpoint,it is said that Jacob did not know that there is so much transcriptional“noise”, that de novo genes are “sufficientlysmall”, and that they only need “minimal functionality” to“get started” ; and that with so much transcriptional noise, and minimal requirements,occasionally some pieces of neutrally-evolving junk fortuitously acquire use in theorganism. But how can we be sure that the de novo genes are small enough,that the amount of transcriptional noise is large enough, and that opportunities forfunctionality, even minimal, are common enough? Say that a person has left Jerusalemby foot, and that we believe that he performs a random walk, stepping in entirelyrandom directions forever after; and say that we all agree that he would never reachParis in this manner, because it is too far for a random walk. Now say we observehim in Istanbul. Would we say that it makes sense that he got there, becauseIstanbul is closer to Jerusalem than Paris is? Or should we perhaps reexamine ourassumptions about the person? This example shows that there is a severe problem oflack of quantification of the amount of chance that we call upon, not to mentionsuch facts as that the Poldi de novo gene arose already with an alternative splicing pattern . The reason that we say that de novo genes must be “smallenough” is not because we know that this “explanation” works, butbecause we have not had any other explanation up to now. In fact, de novogenes are small because they are new genes, and there is a trend whereby the lengthof a gene increases with age. This trend would be expected from the present theory,which upholds a mechanistic, gradual origin of de novo genes. Last but notleast, according to the present theory, the fact of transcriptional promiscuity isindeed eminently relevant to the situation, but in a mechanistic way, as will bediscussed later in this paper.
f To clarify, like Jacob, I do not think that a protein can beconstructed without natural selection. But if protein parts are brought fromelsewhere, in a long-term evolutionary process enacted by the writing phenotype,which itself gets continuous feedback from natural selection, then it is possiblethat we will see here things that are contradictory to traditional notions.
g Note that the word “selfish” in “selfishelements” comes from an analogy to human behavior, where in fact it is a keyprinciple of economics that actions in the actors’ self-interest contribute tothe economy as a whole.
AL graduated from Princeton University, Ph.D. in Ecology and Evolutionary Biology,and is now an Assistant Professor in the Department of Biological Sciences atVirginia Tech. He has done work in theoretical population genetics, the evolution ofsex and recombination, and work at the interface of biology and theoretical computerscience. He conceived this project while being a Postdoctoral Research Fellow at theMiller Institute for Basic Research in Science and the Computer Science Division, UCBerkeley.
The Encyclopedia of DNA Elements consortium
Fork stalling and template switching
Low copy repeats
Microhomology-mediated break-induced replication
The natural selection andrandom mutation process
Single nucleotide polymorphism
Serial replication slippage
Lewontin RC, Hubby JL: A molecular approach to the study of genic heterozygosity in naturalpopulations. II. Amount of variation and degree of heterozygosity in naturalpopulations of Drosophila pseudoobscura. Genetics. 1966, 54: 595-609.
Harris H: Enzyme polymorphisms in man. Proc R Soc Lond B. 1966, 164: 298-310. 10.1098/rspb.1966.0032.
The ENCODE Project Consortium: An integrated encyclopedia of DNA elements in the human genome. Nature. 2012, 489: 57-74. 10.1038/nature11247.
Ohno S: So much “junk” DNA in our genome. Brookhaven Symp Biol. 1972, 23: 366-370.
Orgel LE, Crick FH: Selfish DNA: the ultimate parasite. Nature. 1980, 284: 604-607. 10.1038/284604a0.
Doolittle WF, Sapienza C: Selfish genes, the phenotype paradigm and genome evolution. Nature. 1980, 284: 601-603. 10.1038/284601a0.
Graur D, Zheng Y, Price N, Azevedo RBR, Zufall RA, Elhaik E: On the immortality of television sets: “function” in the humangenome according to the evolution-free gospel of ENCODE. Genome Biol Evol. 2013, 5: 578-590. 10.1093/gbe/evt028.
Doolittle W: Is junk DNA bunk? A critique of ENCODE. P Natl Acad Sci USA. 2013, 110: 5294-5300. 10.1073/pnas.1221376110.
Cai J, Zhao R, Jiang H, Wang W: De novo origination of a new protein-coding gene in Saccharomycescerevisiae. Genetics. 2008, 179: 487-496. 10.1534/genetics.107.084491.
Knowles DG, McLysaght A: Recent de novo origin of human protein-coding genes. Genome Res. 2009, 19: 1752-1759. 10.1101/gr.095026.109.
Heinen TJAJ, Staubach F, Häming D, Tautz D: Emergence of a new gene from an intergenic region. Curr Biol. 2009, 19: 1527-1531. 10.1016/j.cub.2009.07.049.
Li CY, Zhang Y, Wang Z, Zhang Y, Cao C, Zhang PW, Lu SJ, Li XM, Yu Q, Zheng X, Du Q, Uhl GR, Liu QR, Wei L: A human-specific de novo protein-coding gene associated with human brainfunctions. PLoS Comput Biol. 2010, 6: e1000734-10.1371/journal.pcbi.1000734.
Li D, Dong Y, Jiang Y, Jiang H, Cai J, Wang W: A de novo originated gene depresses budding yeast mating pathway and isrepressed by the protein encoded by its antisense strand. Cell Res. 2010, 20: 408-420. 10.1038/cr.2010.31.
Jacob F: Evolution and tinkering. Science. 1977, 196: 1161-1166. 10.1126/science.860134.
Lynch VJ, Leclerc RD, May G, Wagner GP: Transposon-mediated rewiring of gene regulatory networks contributed to theevolution of pregnancy in mammals. Nat Genet. 2011, 43: 1154-1159. 10.1038/ng.917.
Doolittle WF: Understanding introns: origins and functions. Intervening Sequences in Evolution and Development. Edited by: Stone EM, Schwartz RJ. 1990, New York: Oxford University Press, 43-62.
Graur D, Li WH: Fundamentals of Molecular Evolution. 2000, Sunderland: Sinauer Associates
Heard E, Tishkoff S, Todd JA, Vidal M, Wagner GP, Wang J, Weigel D, Young R: Ten years of genetics and genomics: what have we achieved and where are weheading?. Nat Rev Genet. 2010, 11: 723-733. 10.1038/nrg2878.
Fedoroff NV: Presidential address. Transposable elements, epigenetics, and genomeevolution. Science. 2012, 338: 758-767. 10.1126/science.338.6108.758.
West-Eberhard MJ: Developmental Plasticity and Evolution. 2003, New York: Oxford University Press
Fisher RA: The Genetical Theory of Natural Selection. 1930, Oxford: The Clarendon Press
Sniegowski PD: The origin of adaptive mutants: Random or nonrandom?. J Mol Evol. 1995, 40: 94-101. 10.1007/BF00166600.
Futuyma DJ: Evolution, 2nd edition. 2009, Sunderland: Sinauer Associates
Barton NH, Charlesworth B: Why sex and recombination?. Science. 1998, 281: 1986-1990.
Wade MJ, Goodnight CJ: Perspective: The theories of Fisher and Wright in the context ofmetapopulations: when nature does many small experiments. Evolution. 1998, 52: 1537-1553. 10.2307/2411328.
Muller HJ: Some genetic aspects of sex. Am Nat. 1932, 66: 118-138. 10.1086/280418.
Muller HJ: The relation of recombination to mutational advance. Mutation Res. 1964, 1: 2-9. 10.1016/0027-5107(64)90047-8.
Levin DA: Pest pressure and recombination systems in plants. Am Nat. 1975, 109: 437-451. 10.1086/283012.
Jaenike J: A hypothesis to account for the maintenance of sex within populations. Evol Theory. 1978, 3: 191-194.
Kondrashov A: Selection against harmful mutations in large sexual and asexualpopulations. Genet Res. 1982, 40: 325-332. 10.1017/S0016672300019194.
West SA, Lively CM, Read AF: A pluralist approach to sex and recombination. J Evol Biol. 1999, 12: 1003-1012. 10.1046/j.1420-9101.1999.00119.x.
Feldman MW, Otto SP, Christiansen FB: Population genetic perspectives on the evolution of recombination. Annu Rev Genet. 1997, 30: 261-295.
Stearns SC, Hoekstra RF: Evolution: An Introduction. 2005, New York: Oxford University Press
Wright S: Evolution in Mendelian populations. Genetics. 1931, 16: 97-159.
Wright S: The roles of mutation, inbreeding, crossbreeding and selection inevolution. Proc 6th Int Cong Genet. 1932, 1: 356-366.
Provine W: Sewall Wright and Evolutionary Biology. 1986, Chicago: The University of Chicago Press
Barton N, Rouhani S: Adaptation and the ‘shifting balance’. Genet Res. 1993, 61: 57-74. 10.1017/S0016672300031098.
Coyne JA, Barton NH, Turelli M: Perspective: A critique of Sewall Wright’s shifting balance theory ofevolution. Evolution. 1997, 51: 643-671. 10.2307/2411143.
Leigh Jr EG: Introduction. The Causes of Evolution/by J.S.B. Haldane; with a new Preface and Afterwordby E. G. Leigh. 1990, Princeton: Princeton University Press,
Haig D: Weismann rules! OK? Epigenetics and the Lamarckian temptation. Biol Philos. 2007, 22: 415-428. 10.1007/s10539-006-9033-y.
Berg LS: Nomogenesis; or, Evolution Determined by Law. 1926, London: Constable & Company
Dougherty EC: Comparative evolution and the origin of sexuality. Syst Zool. 1955, 4: 145-190. 10.2307/2411667.
Vrijenhoek RC: Genetic and ecological constraints on the origins and establishment ofunisexual vertebrates. Evolution and Ecology of Unisexual Vertebrates. Edited by: Dawley RM, Bogart JP. 1989, Albany: New York State Museum, 24-31.
Stebbins GL: Variation and Evolution in Plants. 1950, New York: Columbia University Press
Van Valen L: Group selection, sex, and fossils. Evolution. 1975, 29: 87-94. 10.2307/2407143.
Maynard-Smith J: The Evolution of Sex. 1978, Cambridge: Cambridge University Press
Bell G: The Masterpiece of Nature: The Evolution and Genetics of Sexuality. 1982, Berkeley: University of California Press
Williams GC: Adaptation and Natural Selection. 1966, Princeton: Princeton University Press, (8th edition 1996)
Stebbins GL: Self fertilization and population variability in the higher plants. Am Nat. 1957, 91: 337-354. 10.1086/281999.
Judson OP, Normark BB: Ancient asexual scandals. Trends Ecol Evol. 1996, 11: 41-46. 10.1016/0169-5347(96)81040-8.
Hurst LD, Hamilton WD, Ladle RJ: Covert sex. Trends Ecol Evol. 1992, 7: 144-145. 10.1016/0169-5347(92)90205-P.
Meselson M: Molecular and cellular biology, faculty profiles. https://www.mcb.harvard.edu/mcb/faculty/profile/matthew-s-meselson/Accessed 1/7/2013,
Meselson M: “Sex and death in bdelloid rotifers.” The Second Annual Arthur W.Galston Memorial Lecture given for the Yale Interdisciplinary Center forBioethics on April 16, 2010. http://archive.org/details/MathewMeselsonSexandDeathinBdelloidRotifersAccessed 1/7/2013,
Woese CR: On the evolution of cells. P Natl Acad Sci USA. 2002, 99: 8742-8747. 10.1073/pnas.132266999.
Brosius J: Gene duplication and other evolutionary strategies: from the RNA world to thefuture. J Struct Funct Genomics. 2003, 3: 1-17. 10.1023/A:1022627311114.
Brosius J: Echoes from the past – are we still in an RNP world?. Cytogenet Genome Res. 2005, 110: 8-24. 10.1159/000084934.
Vetsigian K, Woese C, Goldenfeld N: Collective evolution and the genetic code. P Natl Acad Sci USA. 2006, 103: 10696-10701. 10.1073/pnas.0603780103.
Merriam-Webster Online: Convergence. http://www.merriam-webster.com/dictionary/convergence Accessed1/7/2013,
Oxford English Dictionary Online: Converge. http://oxforddictionaries.com/us/definition/american_english/converge?q=convergeAccessed 1/7/2013,
Darwin C: The Origin of Species by Means of Natural Selection, Or The Preservation ofFavoured Races in the Struggle for Life, 6th edition, Chapter V. 1876, London: John Murray,
Nei M: Modification of linkage intensity by natural selection. Genetics. 1967, 57: 625-641.
Feldman MW: Selection for linkage modification. I. Random mating populations. Theor Popul Biol. 1972, 3: 324-346. 10.1016/0040-5809(72)90007-X.
Feldman MW, Christiansen FB, Brooks LD: Evolution of recombination in a constant environment. P Natl Acad Sci USA. 1980, 77: 4838-4841. 10.1073/pnas.77.8.4838.
Avise JC: Clonality: The Genetics, Ecology, and Evolution of Sexual Abstinence inVertebrate Animals. 2008, New York: Oxford University Press,
Neaves WB: Tetraploidy in a hybrid lizard of the genus Cnemidophorus (Teiidae). Breviora. 1971, 381: 1-25.
Cole CJ: Evolution of parthenogenetic species of reptiles. Intersexuality in the Animal Kingdom. Edited by: Reinboth R. 1975, Berlin: Springer-Verlag, 340-355.
Darevsky IS: Evolution and ecology of parthenogenesis in reptiles. Soc Study Amphib Reptiles Contr Herpetol. 1992, 9: 21-39.
Lande R, Schemske DW: The evolution of self-fertilization and inbreeding depression in plants. I.Genetic models. Evolution. 1985, 39: 24-40. 10.2307/2408514.
Goodwillie C, Kalisz S, Eckert CG: The evolutionary enigma of mixed mating systems in plants: Occurrence,theoretical explanations, and empirical evidence. Annu Rev Ecol Evol Syst. 2005, 36: 47-79. 10.1146/annurev.ecolsys.36.091704.175539.
Jarne P, Charlesworth D: The evolution of the selfing rate in functionally hermaphrodite plants andanimals. Annu Rev Ecol Syst. 1993, 24: 441-466. 10.1146/annurev.es.24.110193.002301.
Tian-Bi YNT, N’Goran EK, N’Guetta SP, Matthys B, Sangare A, Jarne P: Prior selfing and the selfing syndrome in animals: an experimental approachin the freshwater snail Biomphalaria pfeifferi. Genet Res. 2008, 90: 61-72.
Winn AA, Moriuchi KS: The maintenance of mixed mating by cleistogamy in the perennial violet Violaseptemloba (Violaceae). Am J Bot. 2009, 96: 2074-2079. 10.3732/ajb.0900048.
Barrett SCH, Eckert CG: Variation and evolution of mating systems in seed plants. Biological Approaches and Evolutionary Trends in Plants. Edited by: Kawano S. 1990, Tokyo: Academic Press, 229-254.
Jarne P, Auld JR: Animals mix it up too: The distribution of self-fertilization amonghermaphroditic animals. Evolution. 2006, 60: 1816-1824.
Moritz C, Brown WM, Densmore LD, Wright JW, Vyas D, Donnellan S: Genetic diversity and the dynamics of hybrid parthenogenesis in Cnemidophorus(Teiidae) and Heteronotia (Gekkonidae). Evolution and Ecology of Unisexual Vertebrates. Edited by: Dawley RM, Bogart JP. 1989, Albany: New York State Museum: , 87-112.
Darevsky IS, Kupriyanova LA, Uzzell T: Parthenogenesis in reptiles. Biol Reptilia, Volume 15. Edited by: Gans C, Billett F. 1985, New York: Wiley, 411-526.
Holsinger KE: Mass-action models of plant mating systems: The evolutionary stability ofmixed mating systems. Am Nat. 1991, 138: 606-622. 10.1086/285237.
Porcher E, Lande R: The evolution of self-fertilization and inbreeding depression under pollendiscounting and pollen limitation. J Evol Biol. 2005, 18: 497-508. 10.1111/j.1420-9101.2005.00905.x.
Johnston MO: Evolution of intermediate selfing rates in plants: Pollination ecology versusdeleterious mutations. Genetica. 1998, 102/103: 267-278.
Harder LD, Wilson WG: A clarification of pollen discounting and its joint effects with inbreedingdepression on mating system evolution. Am Nat. 1998, 152: 684-695. 10.1086/286199.
Vallejo-Marin M, Uyenoyama MK: On the evolutionary costs of self-incompatibility: Incomplete reproductivecompensation due to pollen limitation. Evolution. 2004, 58: 1924-1935.
Morgan MT, Wilson WG: Self-fertilization and the escape from pollen limitation in variablepollination environments. Evolution. 2005, 59: 1143-1148.
Sakai S, Ishii HS: Why be completely outcrossing? Evolutionary stable outcrossing strategies inan environment where outcross-pollen availability is unpredictable. Evol Ecol Res. 1999, 1: 211-222.
Cheptou PO: Allee effect and self-fertilization in hermaphrodites: reproductive assurancein demographically stable populations. Evolution. 2004, 58: 2613-2621.
Pannell JR: On the problems of a closed marriage: celebrating Darwin 200. Biol Lett. 2009, 5: 332-335. 10.1098/rsbl.2009.0142.
Kiontke K, Gavin NP, Raynes Y, Roehrig C, Piano F, Fitch DHA: Caenorhabditis phylogeny predicts convergence of hermaphroditism andextensive intron loss. P Natl Acad Sci USA. 2004, 101: 9003-9008. 10.1073/pnas.0403094101.
Sassaman C, Weeks SC: The genetic mechanism of sex determination in the conchostracan shrimpEulimnadia texana. Am Nat. 1993, 141: 314-328. 10.1086/285475.
Mackiewicz M, Tatarenkov A, Taylor DS, Turner BJ, Avise JC: Extensive outcrossing and androdioecy in a vertebrate species that otherwisereproduces as a self-fertilizing hermaphrodite. P Natl Acad Sci USA. 2006, 103: 9924-9928. 10.1073/pnas.0603847103.
Otto SP, Sassaman C, Feldman MW: Evolution of sex determination in the conchostracan shrimp Eulimnadiatexana. Am Nat. 1993, 141: 329-337. 10.1086/285476.
Darwin C: The Different Forms of Flowers on Plants of the Same Species. 1877, New York: Appleton and Co.
Lord EM: Cleistogamy: a tool for the study of floral morphogenesis, function andevolution. Bot Rev. 1981, 47: 421-449. 10.1007/BF02860538.
Culley TM, Klooster MR: The cleistogamous breeding system: a review of its frequency, evolution, andecology in angiosperms. Bot Rev. 2007, 73: 1-30. 10.1663/0006-8101(2007)73[1:TCBSAR]2.0.CO;2.
de Nettancourt D: Incompatibility and Incongruity in Wild and Cultivated Plants. 2001, New York: Springer Verlag
Franklin-Tong VE: Self-incompatibility in Flowering Plants: Evolution, Diversity, andMechanisms. 2008, Berlin: Springer Verlag,
Lloyd DG: Self- and cross-fertilization in plants. II. The selection ofself-fertilization. Int J Plant Sci. 1992, 153: 370-380. 10.1086/297041.
Feldman MW, Christiansen FB: Population genetic theory of the cost of inbreeding. Am Nat. 1984, 123: 642-653. 10.1086/284229.
Ramsey FP: A mathematical theory of saving. Econ J. 1928, 38: 543-559. 10.2307/2224098.
Stearns SC: The Evolution of Life Histories. 1992, Oxford: Oxford University Press
Evans HE: The Comparative Ethology and Evolution of the Sand Wasps. 1966, Cambridge: Harvard University Press
Tsuneki K: Comparative studies on the nesting biology of the genus Sphex (s.l.) in EastAsia (Hymenoptera, Sphecidae). Mem Fac Lib Arts, Fukui Univ Ser II. 1963, 13: 13-78.
Evans HE: The accessory burrows of digger wasps. Science. 1966, 152: 465-471. 10.1126/science.152.3721.465.
Schmalhausen II: Factors of Evolution. 1947, Philadelphia: Blakiston
Waddington CH: The Strategy of the Genes. 1957, London: George Allen and Unwin Ltd. Publishers
Wagner GP, Booth G, Bagheri-Chaichian H: A population genetic theory of canalization. Evolution. 1997, 51: 329-347. 10.2307/2411105.
Virgen CA, Kratovac Z, Bieniasz PD, Hatziioannou T: Independent genesis of chimeric TRIM5-cyclophilin proteins in two primatespecies. P Natl Acad Sci USA. 2008, 105: 3563-3568. 10.1073/pnas.0709258105.
Johnson WE, Sawyer SL: Molecular evolution of the antiretroviral TRIM5 gene. Immunogenetics. 2009, 61: 163-176. 10.1007/s00251-009-0358-y.
Nisole S, Lynch C, Stoye JP, Yap MW: A Trim5-cyclophilin A fusion protein found in owl monkey kidney cells canrestrict HIV-1. P Natl Acad Sci USA. 2004, 101: 13324-13328. 10.1073/pnas.0404640101.
Sayah DM, Sokolskaja E, Berthoux L, Luban J: Cyclophilin A retrotransposition into TRIM5 explains owl monkey resistance toHIV-1. Nature. 2004, 430: 569-573. 10.1038/nature02777.
Liao CH, Kuang YQ, Liu HL, Zheng YT, Su B: A novel fusion gene, TRIM5-Cyclophilin A in the pig-tailed macaque determinesits susceptibility to HIV-1 infection. Aids. 2007, 21 (Suppl 8): S19—S26-
Brennan G, Kozyrev Y, Hu SL: TRIMCyp expression in Old World primates Macaca nemestrina and Macacafascicularis. P Natl Acad Sci USA. 2008, 105: 3569-3574. 10.1073/pnas.0709511105.
Wilson SJ, Webb BL, Ylinen LM, Verschoor E, Heeney JL, Towers GJ: Independent evolution of an antiviral TRIMCyp in rhesus macaques. P Natl Acad Sci USA. 2008, 105: 3557-3562. 10.1073/pnas.0709003105.
Newman RM, Hall L, Kirmaier A, Pozzi LA, Farzan M, O’Neil SP, Johnson W, Pery E: Evolution of a TRIM5-CypA splice isoform in old world monkeys. PLoS Pathog. 2008, 4: e1000003-10.1371/journal.ppat.1000003.
Dennett DC: Darwin’s Dangerous Idea: Evolution and the Meaning of Life. 1996, New York: Simon & Schuster
Roth JR, Kofoid E, Roth FP, Berg OG, Seger J, Andersson DI: Regulating general mutation rates: examination of the hypermutable statemodel for Cairnsian adaptive mutation. Genetics. 2003, 163: 1483-1496.
Koonin EV: The Logic of Chance: The Nature and Origin of Biological Evolution. 2011, Upper Saddle River: FT Press
Kaessmann H, Vinckenbosch N, Long M: RNA-based gene duplication: mechanistic and evolutionary insights. Nat Rev Genet. 2009, 10: 19-31.
Muller HJ: Bar duplication. Science. 1936, 83: 528-530.
Nei M: Mutation-Driven Evolution. 2013, Oxford: Oxford University Press
Doolittle WF: What introns have to tell us: hierarchy in genome evolution. Cold Spring Harb Symp Quant Biol. 1987, 52: 907-913. 10.1101/SQB.1987.052.01.099.
Haldane JBS: The cost of natural selection. J Genet. 1957, 55: 511-524. 10.1007/BF02984069.
Kimura M: Evolutionary rate at the molecular level. Nature. 1968, 217: 624-626. 10.1038/217624a0.
Bryson V, Vogel HJ: (Eds): Evolving Genes and Proteins: A Symposium Held at the Institute ofMicrobiology of Rutgers, with Support from the National ScienceFoundation. 1965, New York: Academic Press,
Gu W, Zhang F, Lupski JR: Mechanisms for human genomic rearrangements. PathoGenetics. 2008, 1: 4-10.1186/1755-8417-1-4.
Zhang F, Gu W, Hurles ME, Lupski JR: Copy number variation in human health, disease, and evolution. Annu Rev Genomics Hum Genet. 2009, 10: 451-481. 10.1146/annurev.genom.9.081307.164217.
Mani RS, Chinnaiyan AM: Triggers for genomic rearrangements: insights into genomic, cellular andenvironmental influences. Nat Rev Genet. 2010, 11: 819-829.
Lupski JR: Genomic disorders: structural features of the genome can lead to DNArearrangements and human disease traits. Trends Genet. 1998, 14: 417-422. 10.1016/S0168-9525(98)01555-8.
Bailey JA, Gu Z, Clark RA, Reinert K, Samonte RV, Schwartz S, Adams MD, Myers EW, Li PW, Eichler EE: Recent segmental duplications in the human genome. Science. 2002, 297: 1003-1007. 10.1126/science.1072047.
Stankiewicz P, Lupski JR: Genome architecture, rearrangements and genomic disorders. Trends Genet. 2002, 18: 74-82. 10.1016/S0168-9525(02)02592-1.
Sharp AJ, Cheng Z, Eichler EE: Structural variation of the human genome. Annu Rev Genomics Hum Genet. 2006, 7: 407-442. 10.1146/annurev.genom.7.080505.115618.
Wells RD: Non-B DNA conformations, mutagenesis and disease. Trends Biochem Sci. 2007, 32: 271-278. 10.1016/j.tibs.2007.04.003.
Zhao J, Bacolla A, Wang G, Vasquez KM: Non-B DNA structure-induced genetic instability and evolution. Cell Mol Life Sci. 2010, 67: 43-62. 10.1007/s00018-009-0131-2.
Pfeiffer P, Goedecke W, Obe G: Mechanisms of DNA double-strand break repair and their potential to inducechromosomal aberrations. Mutagenesis. 2000, 15: 289-302. 10.1093/mutage/15.4.289.
De Raedt T, Stephens M, Heyns I, Brems H, Thijs D, Messiaen L, Stephens K, Lazaro C, Wimmer K, Kehrer-Sawatzki H, Vidaud D, Kluwe L, Marynen P, Legius E: Conservation of hotspots for recombination in low-copy repeats associatedwith the NF1 microdeletion. Nat Genet. 2006, 38: 1419-1423. 10.1038/ng1920.
Lindsay SJ, Khajavi M, Lupski JR, Hurles ME: A chromosomal rearrangement hotspot can be identified from population geneticvariation and is coincident with a hotspot for allelic recombination. Am J Hum Genet. 2006, 79: 890-902. 10.1086/508709.
Wahls WP, Davidson MK: Discrete DNA sites regulate global distribution of meiotic recombination. Trends Genet. 2010, 26: 202-208. 10.1016/j.tig.2010.02.003.
Rass E, Grabarz A, Plo I, Gautier J, Bertrand P, Lopez BS: Role of Mre11 in chromosomal nonhomologous end joining in mammalian cells. Nat Struct Mol Biol. 2009, 16: 819-824. 10.1038/nsmb.1641.
Woodward KJ, Cundall M, Sperle K, Sistermans EA, Ross M, Howell G, Gribble SM, Burford DC, Carter NP, Hobson DL, Garbern JY, Kamholz J, Heng H, Hodes ME, Malcolm S, Hobson GM: Heterogeneous duplications in patients with Pelizaeus-Merzbacher diseasesuggest a mechanism of coupled homologous and nonhomologousrecombination. Am J Hum Genet. 2005, 77: 966-987. 10.1086/498048.
Lee JA, Inoue K, Cheung SW, Shaw CA, Stankiewicz P, Lupski JR: Role of genomic architecture in PLP1 duplication causing Pelizaeus-Merzbacherdisease. Hum Mol Genet. 2006, 15: 2250-2265. 10.1093/hmg/ddl150.
Streisinger G, Okada Y, Emrich J, Newton J, Tsugita A, Terzaghi E, Inouye M: Frameshift mutations and the genetic code. Cold Spring Harb Symp Quant Biol. 1966, 31: 77-84. 10.1101/SQB.1966.031.01.014.
Chen JM, Chuzhanova N, Stenson PD, Férec C, Cooper DN: Complex gene rearrangements caused by serial replication slippage. Hum Mutat. 2005, 26: 125-134. 10.1002/humu.20202.
Lee JA, Carvalho CMB, Lupski JR: A DNA replication mechanism for generating nonrecurrent rearrangementsassociated with genomic disorders. Cell. 2007, 131: 1235-1247. 10.1016/j.cell.2007.11.037.
Hastings PJ, Ira G, Lupski JR: A microhomology-mediated break-induced replication model for the origin ofhuman copy number variation. PLoS Genet. 2009, 5: e1000327-10.1371/journal.pgen.1000327.
Zhang F, Carvalho CMB, Lupski JR: Complex human chromosomal and genomic rearrangements. Trends Genet. 2009, 25: 298-307. 10.1016/j.tig.2009.05.005.
Voineagu I, Narayanan V, Lobachev KS, Mirkin SM: Replication stalling at unstable inverted repeats: interplay between DNAhairpins and fork stabilizing proteins. P Natl Acad Sci. 2008, 105: 9936-9941. 10.1073/pnas.0804510105.
Carvalho CM, Zhang F, Liu P, Patel A, Sahoo T, Bacino CA, Shaw C, Peacock S, Pursley A, Tavyev YJ, Ramocki MB, Nawara M, Obersztyn E, Vianna-Morgante AM, Stankiewicz P, Zoghbi HY, Cheung SW, Lupski JR: Complex rearrangements in patients with duplications of MECP2 can occur byfork stalling and template switching. Human Mol Genet. 2009, 18: 2188-2203. 10.1093/hmg/ddp151.
Korbel JO, Urban AE, Affourtit JP, Godwin B, Grubert F, Simons JF, Kim PM, Palejev D, Carriero NJ, Du L, Taillon BE, Chen Z, Tanzer A, Saunders ACE, Chi J, Yang F, Carter NP, Hurles ME, Weissman SM, Harkins TT, Gerstein MB, Egholm M, Snyder M: Paired-end mapping reveals extensive structural variation in the humangenome. Science. 2007, 318: 420-426. 10.1126/science.1149504.
Lupski JR: Genome structural variation and sporadic disease traits. Nat Genet. 2006, 38: 974-976. 10.1038/ng0906-974.
Shapiro JA: Evolution: A View from the 21st Century. 2011, Upper Saddle River: FT Press
Dawkins R: The Selfish Gene. 1976, Oxford: Oxford University Press
Brosius J, Gould SJ: On “genomenclature”: A comprehensive (and respectful) taxonomyfor pseudogenes and other “junk DNA”. P Natl Acad Sci USA. 1992, 89: 10706-10710. 10.1073/pnas.89.22.10706.
Brosius J: RNAs from all categories generate retrosequences that may be exapted as novelgenes or regulatory elements. Gene. 1999, 238: 115-134. 10.1016/S0378-1119(99)00227-9.
Bailey JA, Liu G, Eichler EE: An Alu transposition model for the origin and expansion of human segmentalduplications. Am J Hum Genet. 2003, 73: 823-834. 10.1086/378594.
Kim PM, Lam HYK, Urban AE, Korbel JO, Affourtit J, Grubert F, Chen X, Weissman S, Snyder M, Gerstein MB: Analysis of copy number variants and segmental duplications in the humangenome: Evidence for a change in the process of formation in recentevolutionary history. Genome Res. 2008, 18: 1865-1874. 10.1101/gr.081422.108.
Kidd JM, Cooper GM, Donahue WF, Hayden HS, Sampas N, Graves T, Hansen N, Teague B, Alkan C, Antonacci F, Haugen E, Zerr T, Yamada NA, Tsang P, Newman TL, Tuzun E, Cheng Z, Ebling HM, Tusneem N, David R, Gillett W, Phelps KA, Weaver M, Saranga D, Brand A, Tao W, Gustafson E, McKernan K, Chen L, Malig M, et al: Mapping and sequencing of structural variation from eight human genomes. Nature. 2008, 453: 56-64. 10.1038/nature06862.
Hodgkinson A, Eyre-Walker A: Variation in the mutation rate across mammalian genomes. Nat Rev Genet. 2011, 12: 756-766. 10.1038/nrg3098.
Fryxell KJ, Moon WJ: CpG mutation rates in the human genome are highly dependent on local GCcontent. Mol Biol Evol. 2005, 22: 650-658.
Bird AP: DNA methylation and the frequency of CpG in animal DNA. Nucleic Acids Res. 1980, 8: 1499-1504. 10.1093/nar/8.7.1499.
Cohen NM, Kenigsberg E, Tanay A: Primate CpG islands are maintained by heterogeneous evolutionary regimesinvolving minimal selection. Cell. 2011, 145: 773-786. 10.1016/j.cell.2011.04.024.
Deaton AM, Bird A: CpG islands and the regulation of transcription. Genes Dev. 2011, 25: 1010-1022. 10.1101/gad.2037511.
Suzuki MM, Bird A: DNA methylation landscapes: provocative insights from epigenomics. Nat Rev Genet. 2008, 9: 465-476.
Arnheim N, Calabrese P: Understanding what determines the frequency and pattern of human germlinemutations. Nat Rev Genet. 2009, 10: 478-488.
Qu W, Hashimoto S, Shimada A, Nakatani Y, Ichikawa K, Saito TL, Ogoshi K, Matsushima K, Suzuki Y, Sugano S, Takeda H, Morishita S: Genome-wide genetic variations are highly correlated with proximal DNAmethylation patterns. Genome Res. 2012, 22: 1419-1425. 10.1101/gr.140236.112.
Walser JC, Ponger L, Furano AV: CpG dinucleotides and the mutation rate of non-CpG DNA. Genome Res. 2008, 18: 1403-1414. 10.1101/gr.076455.108.
Walser JC, Furano AV: The mutational spectrum of non-CpG DNA varies with CpG content. Genome Res. 2010, 20: 875-882. 10.1101/gr.103283.109.
Panchin AY, Mitrofanov SI, Alexeevski AV, Spirin SA, Panchin YV: New words in human mutagenesis. BMC Bioinformatics. 2011, 12: 268-274. 10.1186/1471-2105-12-268.
Hodgkinson A, Ladoukakis E, Eyre-Walker A: Cryptic variation in the human mutation rate. PLoS Biol. 2009, 7: 226-232.
Johnson PL, Hellmann I: Mutation rate distribution inferred from coincident SNPs and coincidentsubstitutions. Genome Biol Evol. 2011, 3: 842-850. 10.1093/gbe/evr044.
Seplyarskiy VB, Kharchenko P, Kondrashov AS, Bazykin GA: Heterogeneity of the transition/transversion ratio in Drosophila andHominidae genomes. Mol Biol Evol. 2012, 29: 1943-1955. 10.1093/molbev/mss071.
Hodgkinson A, Eyre-Walker A: The genomic distribution and local context of coincident SNPs in human andchimpanzee. Genome Biol Evol. 2010, 2: 547-557. 10.1093/gbe/evq039.
Stoneking M: Hypervariable sites in the mtDNA control region are mutational hotspots. Am J Hum Genet. 2000, 67: 1029-1032. 10.1086/303092.
Bazykin GA, Kondrashov FA, Brudno M, Poliakov A, Dubchak I, Kondrashov AS: Extensive parallelism in protein evolution. Biol Direct. 2007, 2: 20-10.1186/1745-6150-2-20.
Hodgkinson A, Eyre-Walker A: Human triallelic sites: Evidence for a new mutational mechanism?. Genetics. 2010, 184: 233-241. 10.1534/genetics.109.110510.
Lercher MJ, Hurst LD: Human SNP variability and mutation rate are higher in regions of highrecombination. Trends Genet. 2002, 18: 337-340. 10.1016/S0168-9525(02)02669-0.
Webster MT, Hurst LD: Direct and indirect consequences of meiotic recombination: implications forgenome evolution. Trends Genet. 2012, 28: 101-109. 10.1016/j.tig.2011.11.002.
Myers S, Bottolo L, Freeman C, McVean G, Donnelly P: A fine-scale map of recombination rates and hotspots across the humangenome. Science. 2005, 310: 321-324. 10.1126/science.1117196.
Duret L, Arndt PF: The impact of recombination on nucleotide substitutions in the humangenome. PLoS Genet. 2008, 4: e1000071-10.1371/journal.pgen.1000071.
Caporale LH: Darwin in the Genome: Molecular Strategies in Biological Evolution. 2003, New York: McGraw-Hill
Caporale LH: Natural selection and the emergence of a mutation phenotype: An update of theevolutionary synthesis considering mechanisms that affect genomevariation. Annu Rev Microbiol. 2003, 57: 467-485. 10.1146/annurev.micro.57.030502.090855.
Chuang JH, Li H: Functional bias and spatial organization of genes in mutational hot and coldregions in the human genome. PLoS Biol. 2004, 2: 253-263.
Nguyen DQ, Webber C, Ponting C: Bias of selection on human copy-number variants. PLoS Genet. 2006, 2: 198-207. 10.1371/journal.pgen.0020198.
Clark AG, Glanowski S, Nielsen R, Thomas PD, Kejariwal A, Todd MA, Tanenbaum DM, Civello D, Lu F, Murphy B, Ferriera S, Wang G, Zheng X, White TJ, Sninsky JJ, Adams MD, Cargill M: Inferring nonneutral evolution from human-chimp-mouse orthologous genetrios. Science. 2003, 302: 1960-1963. 10.1126/science.1088821.
Nielsen R, Bustamante C, Clark AG, Glanowski S, Sackton TB, Hubisz MJ, Fledel-Alon A, Tanenbaum DM, Civello D, White TJ, Sninsky JJ, Adams MD, Cargill M: A scan for positively selected genes in the genomes of humans andchimpanzees. PLoS Biol. 2005, 3: 976-985.
Woodward SR, Cruz LJ, Olivera BM, Hillyard DR: Constant and hypervariable regions in conotoxin propeptides. EMBO J. 1990, 9: 1015-1020.
Olivera BM, Walker C, Cartier GE, Hooper D, Santos AD, Schoenfeld R, Shetty R, Watkins M, Bandyopadhyay P, Hillyard DR: Speciation of cone snails and interspecific hyperdivergence of their venompeptides: potential evolutionary significance of introns. Ann NY Acad Sci. 1999, 870: 223-237. 10.1111/j.1749-6632.1999.tb08883.x.
Crow KD, Amemiya CT, Roth J, Wagner GP: Hypermutability of HoxA13a and functional divergence from its paralog areassociated with the origin of a novel developmental feature in zebrafish andrelated taxa (cypriniformes). Evolution. 2009, 63: 1574-1592. 10.1111/j.1558-5646.2009.00657.x.
Inoue K, Lupski J: Molecular mechanisms for genomic disorders. Annu Rev Genom Hum G. 2002, 3: 199-242. 10.1146/annurev.genom.3.032802.120023.
Veltman JA, Brunner HG: De novo mutations in human genetic disease. Nat Rev Genet. 2012, 13: 565-575. 10.1038/nrg3241.
Voight BF, Kudaravalli S, Wen X, Pritchard JK: A map of recent positive selection in the human genome. PLoS Biol. 2006, 4: 446-458. 10.1371/journal.pbio.0040446.
Crespi BJ, Summers K: Positive selection in the evolution of cancer. Biol Rev. 2006, 81: 407-424.
Dawkins R: The Blind Watchmaker: Why the Evidence of Evolution Reveals a UniverseWithout Design. 1986, New York: WW Norton & Company,
Ohno S: Evolution by Gene Duplication. 1970, Heidelberg: Springer-Verlag
Siepel A: Darwinian alchemy: Human genes from noncoding DNA. Genome Res. 2009, 19: 1693-1695. 10.1101/gr.098376.109.
Levine MT, Jones CD, Kern AD, Lindfors HA, Begun DJ: Novel genes derived from noncoding DNA in Drosophila melanogaster arefrequently X-linked and exhibit testis-biased expression. P Natl Acad Sci USA. 2006, 103: 9935-9939. 10.1073/pnas.0509809103.
Begun DJ, Lindfors HA, Kern AD, Jones CD: Evidence for de novo evolution of testis-expressed genes in the Drosophilayakuba/Drosophila erecta clade. Genetics. 2007, 176: 1131-1137.
Chen ST, Cheng HC, Barbash DA, Yang HP: Evolution of hydra, a recently evolved testis-expressed gene with ninealternative first exons in Drosophila melanogaster. PLoS Genet. 2007, 3: 1131-1143.
Zhou Q, Zhang G, Zhang Y, Xu S, Zhao R, Zhan Z, Li X, Ding Y, Yang S, Wang W: On the origin of new genes in Drosophila. Genome Res. 2008, 18: 1446-1455. 10.1101/gr.076588.108.
Toll-Riera M, Bosch N, Bellora N, Castelo R, Armengol L, Estivill X, Alba MM: Origin of primate orphan genes: A comparative genomics approach. Mol Biol Evol. 2009, 26: 603-612.
Wu DD, Irwin DM, Zhang YP: De novo origin of human protein-coding genes. PLoS Genet. 2011, 7: e1002379-10.1371/journal.pgen.1002379.
Tautz D, Domazet-Lošo T: The evolutionary origin of orphan genes. Nat Rev Genet. 2011, 12: 692-702.
Zhang YE, Liu CJ, Li Y, Zhang R, Wei L, Li CY, Xie C: Hominoid-specific de novo protein-coding genes originating from longnon-coding RNAs. PLoS Genet. 2012, 8: e1002942-10.1371/journal.pgen.1002942.
Neme R, Tautz D: Phylogenetic patterns of emergence of new genes support a model of frequentde novo evolution. BMC Genomics. 2013, 14. doi:10.1186/1471–2164–14–117,
Babushok DV, Ohshima K, Ostertag EM, Chen X, Wang Y, Mandal PK, Okada N, Abrams CS, Kazazian Jr HH: A novel testis ubiquitin-binding protein gene arose by exon shuffling inhominoids. Genome Res. 2007, 17: 1129-1138. 10.1101/gr.6252107.
Zhang Y, Lu S, Zhao S, Zheng X, Long M, Wei L: Positive selection for the male functionality of a co-retroposed gene in thehominoids. BMC Evol Biol. 2009, 9: 252-10.1186/1471-2148-9-252.
Lynch VJ, Nnamani M, Brayer KJ, Emera D, Wertheim JO, Kosakovsky-Pond SL, Grutzner F, Bauersachs S, Graf A, Kapusta A, Feschotte C, Wagner GP: Lineage-specific transposons drove massive gene expression recruitmentsduring the evolution of pregnancy in mammals. arXiv preprint. 2012, arXiv, 1208.4639-
Emera D, Wagner GP: Transposable element recruitments in the mammalian placenta: impacts andmechanisms. Brief Funct Genomics. 2012, 11: 267-276. 10.1093/bfgp/els013.
Emera D, Wagner GP: Transformation of a transposon into a derived prolactin promoter withfunction during human pregnancy. P Natl Acad Sci USA. 2012, 109: 11246-11251. 10.1073/pnas.1118566109.
McClintock B: Components of action of the regulators Spm and Ac. Carnegie Inst Wash Yearbook. 1965, 64: 527-534.
Britten RJ, Davidson EH: Gene regulation for higher cells: A theory. Science. 1969, 165: 349-357. 10.1126/science.165.3891.349.
Georgiev GP: Mobile genetic elements in animal cells and their biological significance. Eur J Biochem. 1984, 145: 203-220. 10.1111/j.1432-1033.1984.tb08541.x.
Brosius J: Genomes were forged by massive bombardments with retroelements andretrosequences. Genetica. 1999, 107: 209-238. 10.1023/A:1004018519722.
Eddy SR: The ENCODE project: Missteps overshadowing a success. Curr Biol. 2013, 23: R259—R261-
Kleene KC: Sexual selection, genetic conflict, selfish genes, and the atypical patternsof gene expression in spermatogenic cells. Dev Biol. 2005, 277: 16-26. 10.1016/j.ydbio.2004.09.031.
Kan Z, Garrett-Engele PW, Johnson JM, Castle JC: Evolutionarily conserved and diverged alternative splicing events showdifferent expression and functional profiles. Nucleic Acids Res. 2005, 33: 5659-5666. 10.1093/nar/gki834.
Elliott DJ, Grellscheid SN: Alternative RNA splicing regulation in the testis. Reproduction. 2006, 132: 811-819. 10.1530/REP-06-0147.
Thomson T, Lin H: The biogenesis and function of PIWI proteins and piRNAs: Progress andprospect. Annu Rev Cell Dev Biol. 2009, 25: 355-376. 10.1146/annurev.cellbio.24.110707.175327.
Kaessmann H: Origins, evolution, and phenotypic impact of new genes. Genome Res. 2010, 20: 1313-1326. 10.1101/gr.101386.109.
Miller D, Brinkworth M, Iles D: The testis as a conduit for genomic plasticity: an advanced interdisciplinaryworkshop. Biochem Soc Trans. 2007, 35: 605-608. 10.1042/BST0350605.
Old LJ: Cancer/Testis (CT) antigens—a new link between gametogenesis andcancer. Cancer Immunity. 2001, 1: 1-
Simpson A, Caballero O, Jungbluth A, Chen YT, Old L: Cancer/testis antigens, gametogenesis and cancer. Nat Rev Cancer. 2005, 5: 615-625. 10.1038/nrc1669.
She X, Horvath JE, Jiang Z, Liu G, Furey TS, Christ L, Clark R, Graves T, Gulden CL, Alkan C, Bailey JA, Sahinalp C, Rocchi M, Haussler D, Wilson RK, Miller W, Schwartz S, Eichler EE: The structure and evolution of centromeric transition regions within thehuman genome. Nature. 2004, 430: 857-864. 10.1038/nature02806.
Vinckenbosch N, Dupanloup I, Kaessmann H: Evolutionary fate of retroposed gene copies in the human genome. P Natl Acad Sci USA. 2006, 103: 3220-3225. 10.1073/pnas.0511307103.
Marques AC, Dupanloup I, Vinckenbosch N, Reymond A, Kaessmann H: Emergence of young human genes after a burst of retroposition in primates. PLoS Biol. 2005, 3: e357-10.1371/journal.pbio.0030357.
Nei M: Selectionism and neutralism in molecular evolution. Mol Biol Evol. 2005, 22: 2318-2342. 10.1093/molbev/msi242.
Landry J, Pyl PT, Rausch T, Zichner T, Tekkedil MM, Stütz AM, Jauch A, Aiyar RS, Pau G, Delhomme N, Gagneur J, Korbel JO, Huber W, Steinmetz LM: The genomic and transcriptomic landscape of a HeLa cell line. G3. 2013, 3: 1213-1224. 2013.
Pauling L, Itano HA, Singer SJ, Wells IC: Sickle-cell anemia, a molecular disease. Science. 1949, 110: 543-548. 10.1126/science.110.2865.543.
Ingram VM: How do genes act?. Sci Am. 1958, 198: 68-76.
Allison AC: Polymorphisms and natural selection in human populations. Cold Spring Harb Symp Quant Biol. 1964, 29: 137-149. 10.1101/SQB.1964.029.01.018.
Hill AV, Allsopp CE, Kwiatkowski D, Anstey NM, Twumasi P, Rowe PA, Bennett S, Brewster D, McMichael AJ, Greenwood BM: Common West African HLA antigens are associated with protection from severemalaria. Nature. 1991, 352: 595-600. 10.1038/352595a0.
Haldane JBS: Disease and evolution. La Ricera Scientifica Suppl A. 1949, 19: 68-76.
Behe MJ: The Edge of Evolution: The Search for the Limits of Darwinism. 2007, New York: Free Press
Flint J, Harding RM, Boyce AJ, Clegg JB: The population genetics of the haemoglobinopathies. Baillière’s Clin Haem. 1998, 11: 1-51. 10.1016/S0950-3536(98)80069-3.
Kwiatkowski DP: How malaria has affected the human genome and what human genetics can teachus about malaria. Am J Hum Genet. 2005, 77: 171-192. 10.1086/432519.
Flint J, Harding RM, Clegg JB, Boyce AJ: Why are some genetic diseases common?. Hum Genet. 1993, 91: 91-117.
Borg J, Georgitsi M, Aleporou-Marinou V, Kollia P, Patrinos GP: Genetic recombination as a major cause of mutagenesis in the human globingene clusters. Clin Biochem. 2009, 42: 1839-1850. 10.1016/j.clinbiochem.2009.07.014.
Giordano PC, Harteveld CL, Michiels JJ, Terpstra W, Schelfhout LJDM, Appel IM, Batelaan D, van Delft, Plug RJ, Bernini LF: Phenotype variability of the dominant β-thalassemia induced in fourDutch families by the rare cd121 (G→ T) mutation. Ann Hematol. 1998, 77: 249-255. 10.1007/s002770050453.
Holloway K, Lawson VE, Jeffreys AJ: Allelic recombination and de novo deletions in sperm in the humanβ-globin gene region. Human Mol Genet. 2006, 15: 1099-1111. 10.1093/hmg/ddl025.
Sicard D, Lieurzou Y, Lapoumeroulie C, Labie D: High genetic polymorphism of hemoglobin disorders in Laos. Hum Genet. 1979, 50: 327-336. 10.1007/BF00399399.
Kazazian Jr HH, Boehm CD: Molecular basis and prenatal diagnosis of β-thalassemia. Blood. 1988, 72: 1107-1116.
Thein SL, Hesketh C, Taylor P, Temperley IJ, Hutchinson RM, Old JM, Wood WG, Clegg JB, Weatherall DJ: Molecular basis for dominantly inherited inclusion bodyβ-thalassemia. P Natl Acad Sci USA. 1990, 87: 3924-3928. 10.1073/pnas.87.10.3924.
Kazazian Jr HH, Dowling CE, Hurwitz RL, Coleman M, Adams JGI: Thalassemia mutations in exon 3 of the β-globin gene often cause adominant form of thalassemia and show no predilection for malarial-endemicregions. Am J Hum Genet. 1989, 45: A242-
Kazazian Jr HH, Orkin SH, Boehm CD, Goff SC, Wong C, Dowling CE, Newburger PE, Knowlton RG, Brown V, Donis-Keller H: Characterization of a spontaneous mutation to a β-thalassemia allele. Am J Hum Genet. 1986, 38: 860-867.
Troland LT: The chemical origin and regulation of life. The Monist. 1914, 24: 92-133. 10.5840/monist191424113.
Muller HJ: The gene as the basis of life. Proc. 1st Int Congr Plant Sci, Ithaca. 1926, 1: 897-921.
Dyson FJ: Origins of Life. 1985, Cambridge: Cambridge University Press
Ospovat D: The Development of Darwin’s Theory: Natural History, Natural Theology,and Natural Selection, 1838–1859. 1995, Cambridge: Cambridge University Press,
Galton F: Natural Inheritance. 1889, London: Macmillan and Co.
Gayon J: Darwinism’s Struggle for Survival: Heredity and the Hypothesis ofNatural Selection. 1998, Cambridge: Cambridge University Press,
Mayr E: Animal Species and Evolution. 1963, Cambridge: Belknap Press
Stoltzfus A: On the possibility of constructive neutral evolution. J Mol Evol. 1999, 49: 169-181. 10.1007/PL00006540.
Yampolsky LY, Stoltzfus A: Bias in the introduction of variation as an orienting factor in evolution. Evol Dev. 2001, 3: 73-83. 10.1046/j.1525-142x.2001.003002073.x.
Stoltzfus A: Mutationism and the dual causation of evolutionary change. Evol Dev. 2006, 8: 304-317. 10.1111/j.1525-142X.2006.00101.x.
Stoltzfus A, Yampolsky L: Climbing mount probable: Mutation as a cause of nonrandomness inevolution. J Hered. 2009, 100: 637-647. 10.1093/jhered/esp048.
Lenski RE, Mittler JE: The directed mutation controversy and neo-Darwinism. Science. 1993, 259: 188-194. 10.1126/science.7678468.
Hall BG: On the specificity of adaptive mutations. Genetics. 1997, 145: 39-44.
Rosenberg SM: Evolving responsively: Adaptive mutation. Nat Rev Genet. 2001, 2: 504-515.