METAPTATION:
A descriptive category for evolutionarily versatile
Department of Anatomy
and Department of Zoology
Southern Illinois University, Carbondale,
IL 62901, U.S.A.
(Abstract published in 1985, Evolutionary Theory 7:222)
ABSTRACT
"Metaptation" (from meta to change + aptation fitness) is offered as a name for evolved patterns of biological organization which promote evolutionary versatility by causing and constraining mutation and by ontogenetically accommodating to the consequences of mutation. Many mechanisms for actively encouraging genetic variation are already known to exist. Recognizing that such mechanisms have evolved for this role, selected on the basis of past contribution to evolutionary survival, offers a paradigm for investigating the nature of genetic and developmental function. Evolutionarily useful transformations may be explicitly coded by variable genes or genetic regulatory sequences. Organizational patterns which confer long term evolutionary versatility may be as sophisticated and as fundamental as the patterns which provide for immediate adaptive morphogenesis.
INTRODUCTION
Ever since Darwin, we have understood that life has evolved through the natural
selection of heritable variation. This view of life -- with evolution
seen as a three-fold process of variation, selection, and heredity -- is now
firmly grounded in ecology, molecular biology and population genetics. Inheritance
is based on DNA replication. Variation in genomic DNA sequences arises
from several effects which impinge on the replication process. Interactions
between phenotype and environment yield differential survival and reproduction
among these various patterns of genetic information. The result is a gradual
transformation in the information contained within the genomes of all the organisms
which comprise each population. Although these processes can be summarized
quite simply, their consequences are extremely complex. Emerging from
evolution is adaptation, the astonishing "fit" between organism and environment
which has enabled the continuing passage from generation to generation of an
intricate and ever-changing pattern of biological organization.
Intimately associated with adaptation is a fundamental question for biology: How is genetic variation translated into those variations in phenotypic form and function upon which selection operates? Most adaptations are not independent traits, each regulated by its own discrete gene; rather each adaptation arises from a complex genetic program whose successful evolutionary modification is constrained by the need for developmental and functional integration among many interrelated traits. Although basic mechanisms for translating genes into proteins have become well known, for higher levels of biological organization we still share the concern expressed by Darwin in 1859 in a letter to T. H. Huxley: "...what the devil determines each particular variation? What makes a tuft of feathers come on a cock's head, or moss on a moss-rose?" (F. Darwin, 1892). In more modern jargon, what manner of genomic alteration can produce the specific changes in functional morphology which lead to adaptive evolution?
This paper adopts the view of Gould (1982a, b; 1985), Arnold and Fristrup (1982) and Vrba and Eldredge (1984) that evolution is hierarchical. Variation and selection apply to individuals at each of several levels of biological organization ranging from gene to clade. Developing such an expanded evolutionary theory (Gould 1982a) as a tool for analyzing biological processes at many levels will require the formulation of concepts appropriate to each level, and will benefit from clear language that differentiates among corresponding processes in hierarchically distinct levels. Pursuing a hierarchical viewpoint may enable the realization of unexpected conclusions, the appreciation of emergent processes and effects which had not been readily apparent from a single, exclusively low-level perspective.
Gould (1985) has begun this task by labelling three levels or "tiers" of evolution: "first-tier" selection is conventional Darwinian selection of individual organisms; "second-tier" selection emerges from differential speciation and extinction among lineages; "third-tier" selection reflects the infrequent catastrophes which may eliminate many forms of life without respect for their varying adaptive or competitive advantages. This essay shall discuss some implications of second-tier selection. A new term -- metaptation -- shall be introduced to identify those traits which are both the basis for and the product of such selection. Features which facilitate the evolutionary continuation and branching of lineages shall be distinguished from those features -- usually called adaptations -- which contribute primarily to the survival and reproduction of individual organisms. That species differ in their evolutionary plasticity, in their potential for speciating and for accommodating to diverse and shifting environments, is well known. Less well understood is how traits which promote variation might be responsible for such differential versatility, and might themselves be selected for, elaborated upon and improved through long-term selection for lineage survival. The concept of metaptation shall be offered then not only as a name for features which affect second-tier selection but also as a paradigm for exploring the nature of evolutionary variation.
THE PROBLEM
Mutations, defined as any "alterations of the information carried by the genetic
macromolecules" (Futuyma 1979, p. 22), must provide the ultimate source for
evolutionary variation. But beyond the central requirement that
variation must be adaptively undirected -- for otherwise variation itself, not
selection, would be the primary directing and creative force of evolution (Gould
1982a) -- evolutionary theory has not addressed the causes of mutation. Mutations
are seen as no more than the inevitable result of imperfect reproduction; they
are fortuitous accidents.
But what if much evolutionarily significant variation were not accidental? What if such variation were the product of internally organized mechanisms for promoting mutations -- for causing variation which is not directed toward particular adaptive ends but is nevertheless constrained within viable domains of genetic organization? In fact, much genetic variation is not "accidental" in any fundamental sense but is both caused and constrained by active molecular systems which are themselves the product of evolutionary selection.
"Gene-processing enzymes...probably are the source of most of the variability important for evolution. This is significant because enzymes are notable for the specificity of the chemical transformations that they catalyze, while the hallmark of classical mutation is its presumed randomness... Even those mutational changes not actually catalyzed by enzymes are picked over and winnowed by multiple systems of sophisticated error-correcting enzyme pathways to leave a small pool of highly edited mutant alleles for selection" (Campbell, 1985).
Far from being a passive bearer of information, the genome has turned out to be an active agent of mutation, highly organized to influence its own restructuring. If an understanding of evolutionary patterns can reveal necessary constraints upon variation, thereby exposing genetic principles which govern the occurrence of mutation and the relationship between genome and organism, then the historical science of evolutionary biology may become coequal with the experimental sciences as a tool for unlocking the mysteries of genetics and development.
The emerging view of mutation as an orderly product of organized systems seems, at least superficially, to depart from the Darwinian expectation that evolutionary variation should be undirected. Confusion over this apparent inconsistency has provoked some extreme misinterpretations of molecular organization. Indeed, one researcher with a predominantly molecular perspective has gone so far as to state that "variants ... are neither random by a strict definition, nor random by the definition of undirectedness. The reality is rather than adaptation is directed by nonrandomness, which is orthogenetic and deterministic" (Fox 1984). Mechanisms underlying genetic variation have even been named "evolutionary directors" and "evolutionary drivers" which "drive evolution in the same sense that a bus driver drives his vehicle" (Campbell 1985), in clear contradiction of classical evolutionary theory.
Such misinterpretations have arisen to fill a conceptual gap left by the modern synthesis. Far from offering guidance for analytical research in genetic organization and developmental mechanism, classical evolutionary theory has simply assumed that mutations must occur in sufficient quantity and appropriate quality to supply the raw material for adaptive evolution. The adequacy of mutation is evident from the reality of evolutionary change. But in light of current knowledge this adequacy of mutation is no longer just a necessary assumption; rather it has become a curious fact which must itself be explained within the framework of evolutionary theory. The classical view that natural selection, not any internal control system, provides the adaptive direction for evolution does indeed require that genetic variation be adaptively undirected. But this principle does not demand that such variation be accidental. Recent discoveries in molecular genetics call for alternative hypotheses; the assumption of accidental mutation is no longer an adequate explanation for the source of evolutionary variation. Nevertheless, the opposing view that internally constrained mutation must orthogenetically direct evolution is also unacceptable. It is past time to reconcile the existence of active internally organized mutation-generating mechanisms with the primacy of natural selection as a directing force for evolution.
Advocates of evolutionary orthodoxy have routinely argued that natural selection should not, indeed could not, favor the active facilitation of mutation. "In other words, natural selection of mutation rates has only one possible direction, that of reducing the frequency of mutation to zero" (Williams 1966, p. 139). Although this argument appears valid at the level of gene selection, its truth depends upon selection acting only at this level in the hierarchy of evolutionary process. As long as selection is defined at this level alone, then amenability of an allele (defined as a particular nucleotide sequence) to mutation must by definition confer selective disadvantage. But the genome represents many levels of genetic organization beyond that of peptide-coding alleles, and natural selection by differential reproduction acts upon these higher levels as well as on genes. At higher levels the important genetic messages are not merely those of accurate structural duplication but also those of survivable functional effectiveness. Accurate copying of the genetic message remains vitally important at all levels, but the message to be copied differs at different levels. At higher levels, effective replication of a functional message may require some orderly variation in its lower-level structural representation. For such higher-level messages -- those for which efficient enzyme catalysis is more important than particular amino-acid sequence and viable offspring are more valuable than precisely cloned DNA -- orderly but undirected variation can yield a greater degree of reproductive success for the higher-level plan than would overly precise replication.
Presumably natural selection would act against any mechanism which promoted deleterious mutations. But a mechanism which would preferentially yield adaptively neutral variations could survive selection, since neutral mutations by definition will not be selected against. If any such "neutral mutation" mechanism happened to produce evolutionarily useful variation (and hypothetical examples are easy to imagine without invoking either foresight or determinism; see below), then such a mechanism would be strongly favored over systems lacking such capacity, by selection acting at the second tier.
The principle that reproduction with variation can offer substantial benefits, and that the benefits are sufficient for the selection of elaborate and expensive mechanisms for promoting such variation, is widely accepted as the principal explanation for sexual reproduction (e.g., Bell 1982). But sex is just one application of this principle. Systematic, genetically based variation in the surface antigens of parasitic organisms or in the antibody production systems of mammals are also familiar, and these are only special cases. The underlying concept of selective advantage accruing to active, orderly, but undirected variation is quite general. However, the broad implications of this principle for our understanding of biological form and function seldom seem to have been fully appreciated (but see Conrad 1983).
Biologists have long recognized that certain fortuitous traits can provide the foundation for subsequent adaptive evolutionary change. Such traits have been called "preadaptations" but are now better named "exaptations" or "preaptations" depending on their present function or lack thereof (Gould and Vrba 1982). By definition, such traits did not originate in selection for their current or future contribution to fitness. Such traits are not the intended target for the term metaptation. Rather, metaptations are evolved patterns of organization which have been shaped by natural selection for their evolutionary function, for their capacity to promote evolutionary versatility. Metaptation can be subtle. Metaptations need never make direct contribution to fitness. For instance, a duplicate copy of a gene for a useful protein may be a preaptation, leading by genetic drift to a valuable exaptation (Goulds and Vrba 1982). But some way to duplicate or otherwise add genetic material is necessary just for the possibility of evolutionary increase in genomic complexity. Any genetic mechanisms which could allow gene duplication in the first place, and which could cause or permit mutations within the copies, would be important metaptations whose adaptive role, if any, is indirect. Metaptations may originate through selection for adaptation (e.g., enhanced gene function in the case of duplicated genes), but neither presence nor absence of direct adaptive function is required by the definition of metaptation. Metaptive function transcends adaptation and should be clearly distinguished. Metaptation responds to selection at a different level from adaptation. The only adaptive first-tier constraint upon metaptation is that it not be seriously maladaptive; otherwise the potentially metaptive trait would be too quickly eliminated by first tier selection.
The reality of metaptations, of mechanisms which permit and promote variation, can be appreciated at several levels of biological organization. At the molecular level, metaptations are traits which allow the occurrence of mutation, especially mutation which is constrained within viable domains of genetic organization. Molecular metaptations include the enzymes and controlling elements underlying gene duplication, transposition, and recombination. At the level of organisms, metaptations are features of structure or of development which enable complex and highly integrated systems to accommodate mutational modification. An obvious morphological example is repetitive structure, such as segmentation or redundant neural pathways, which can encompass considerable variation in number or individual form of subunits without detrimental effect on organismic function. Developmental examples might include any form of regulation which allows genetic alteration in one system to be met by concordant shifts in other interrelated systems (e.g., "ontogenetic buffer mechanisms," Katz and Lasek 1978). Developmental processes which permit the genetic dissociability of basic ontogenetic mechanisms are of fundamental importance to evolution (Needham 1933). They create the possibility for non-lethal variation within a highly integrated developmental program. Organism-level metaptations such as developmental regulation can offer immediate survival value by compensating for lost or damaged parts and by buffering against disturbances during growth. But the compensatory capacity of many redundant structures and regulatory systems often greatly exceeds that needed for accommodating any reasonable injury which might be survived outside the laboratory. Such traits do convey some conceivably adaptive protection against implausible environmental insults, but they are also powerfully metaptive. They permit a viable pattern of phenotypic organization to persist through and benefit from extensive genomic variation.
But does the concept of metaptation need a special term? Features that confer evolutionary plasticity have long been recognized. Why then introduce a new name for such features? The absence of a specific word for traits which encourage variation reinforces a conventional view that the evolutionary function for such traits is merely fortuitous, that evolutionary versatility is an incidental byproduct of certain adaptive patterns of organization, albeit one with important consequences. Interpreting metaptive function as fortuitous inhibits the realization that such function may be far more common and far more effective than an accidental origin would imply. Similarly, the related view that mutation is no more than an inherent imperfection in molecular copying obscures the practical reality that the rates for various classes of mutation can be subject to exquisite genetic control. Efficient and complex functional organization does not arise fortuitously. Functional complexity, including that represented by elaborate molecular mechanisms for genetic rearrangement and for developmental regulation, is presumptive evidence for the active, directing agency of natural selection in its origin (e.g., Ghiselin 1984).
"Adaptation" is the name usually given to the evolved products of natural selection (e.g., Gould and Vrba 1982; Brandon 1985). But adaptation, like fitness, is usually understood in terms of selection acting at the first tier, through the survival and reproduction of individual organisms. Evolutionary plasticity, on the other hands, is not an attribute of individual organisms but of populations or of genomic information systems, because only populations can evolve, not individual organisms. Therefore, by conventional usage, "evolutionary plasticity is not an adaptation that can be produced by natural selection" (Williams 1966, p. 139, emphasis added.) However, this conclusion is not a natural law but a tautology. It arises from a commonly accepted definition of adaptation as a feature conferring immediate benefit upon individual organisms. This does not imply that evolutionary plasticity cannot be the product of natural selection, only that such plasticity should not be equated with adaptation of individual organisms. Unfortunately, adaptation is the only name we have for traits which have been shaped by selection. Therefore, although a new name is not justified for traits which fortuitously contribute to future evolution, new concepts and new terminology are necessary to identify the origin of specialized mechanisms which persist because they cause and constrain mutation while accommodating to its consequences. "Metaptation" identifies traits whose present forms have been selected at the second tier because they have facilitated evolution.
This concept of metaptation is similar to that of "biotic adaptation" which Williams (1966) has argued against so effectively. Unfortunately Williams chose brazenly teleological language to define biotic adaptations. These, he says, have been "designed to perpetuate a population or more inclusive group," while conventional adaptations merely "function to maximize the genetic survival of individuals" (Williams 1966, p. viii). Of course biotic adaptation, so defined, is not a valid concept. Nothing, not even the most marvelous adaptation, has been "designed" for any future function: What survives, what is selected, is only what has worked in the past. But the process of selection can nevertheless operate at several distinct, hierarchical levels.
Williams is honest enough to predict when his view must be expanded:
"We must take the theory of natural selection to its simplest and most austere form...and use it in an uncompromising fashion whenever a problem of adaptation arises...When the best such explanation is complex and not very plausible, the way is paved for a better theory." (Williams 1966, p. 270).
Sexual reproduction has long been recognized as an "exceptional adaptation that was 'evolved for the specific rather than for the individual advantage'" (Williams 1966, p. 157, quoting from R. A. Fisher 1930). With arduous logic, sex can be explained as a (somewhat) conventional adaptation selected at the first tier. But sex need not be explained entirely at one level of selection. "Perhaps sex predominates because two levels interact positively and are not suppressed any any higher level" (Gould 1982a). In other words, sex is both an adaptation and a metaptation. But even if the question of sex were to be adequately answered with reference to "simple and austere" first tier selection, recent discoveries of molecular genetics are not so easily resolved. To account for the presence of each active molecular system that can modify the genome, conventional "explanation" must postulate the existence of some unknown adaptive advantage to offset the supposed genetic load of mutation. But perhaps the time has come to accept Williams' challenge, to begin formulating an expanded evolutionary theory which does not limit selection only to those traits which confer simple short-term adaptive value. Traits whose primary effect is the creation of distant genetic diversity may confer no immediate value, may even be mildly disadvantageous over the short term, and yet be strongly favored by selection which operates not solely through immediate reproductive success but also through every descendent generation. The proposed term "metaptation" avoids the semantic pitfalls of naming as adaptations such traits whose evolved function is long-term genetic versatility.
Metaptations evolve not because they anticipate a future need for evolutionary change but because past success has been based on their contribution to evolutionary versatility. Because environments do change, especially environments which include other evolving organisms, any lineage with limited capacity to undergo extensive evolutionary transformation will stagnate and must eventually risk extinction. Lineages are more likely to persist if they can track a changing habitat and diversify into many niches, thus increasing the probability that at least some descendent branches will survive the vicissitudes of an uncertain environment. Selection at the second tier, at the level of anagenesis and speciation, will eliminate lineages that lack versatility. Second-tier selection will favor the perpetuation and improvement of any genomic or phenotypic patterns which, however baroque they may appear from an adaptive perspective, increase the general capacity of a lineage for adapting to new or altered conditions. The greater this capacity, the greater will be the probability for long term survival of at least some derivatives of the fundamental genetic system. Whatever the original source for traits which increase this capacity -- whether they arise directly as primary mutations fixed by drift or indirectly as adaptations fixed by first-tier selection -- once metaptive traits appear they can be preserved by second-tier selection, then augmented and elaborated by the selective accumulation of additional improvements. Metaptations are certainly fortuitous, just as every adaptation is the fortuitous result of a historical series of chemical reactions. But once numerous fortuitous variations have been preserved and modified by selection into a complex and functional system, then that system should no longer be viewed as merely fortuitous, whether its function is adaptive or metaptive.
The process of selection for metaptation, for traits which confer evolutionary versatility, has been presented by Conrad (1983). Most simply, the adaptive evolutionary success of descendent lineages results from and therefore perpetuates the metaptive patterns which made such success possible. Conrad summarizes this as the "bootstrap principle of evolutionary adaptability": "Evolutionary amenability in effect pulls itself up by its own bootstraps by hitchhiking along with the advantageous traits whose emergence it facilitates" (Conrad 1983, p. 196). Different traits may be adaptively equivalent, yet differ substantially with respect to selection for versatility. Lineages which embody more versatile traits, which are better metapted, will be favored by selection at the second tier.
Even traits with significant adaptive (first tier) costs can prosper if every alternative is eventually eliminated at the second tier. Sexual reproduction presents a useful example. Among complex organisms, parthenogenesis can under certain circumstances confer considerable adaptive advantage through increased reproductive efficiency. But, on the basis of the phylogenetic record, parthenogenetic lineages are generally short-lived with limited potential for diversification. Sex persists, not because it is always or even often the most efficient method for reproduction, but because alternatives lack the necessary versatility for continued survival at the second tier. Any traits which "lock in" sex, which hinder its adaptive evolutionary loss by making reproduction absolutely, irrevocably dependent on sex, will be favored by second-tier selection and will help assure that sex will predominate even under conditions where parthenogenesis might be more immediately adaptive.
One might argue that second tier selection could not be an effective agency for evolutionary change, because species are numerically so many fewer than the organisms upon which first-tier selection acts. But the raw material for second tier selection is quantitatively equivalent to that available for conventional Darwinian selection. Every individual organism is potentially the bearer of a metaptively significant mutation. As long as metaptively beneficial variation coincides with adaptively advantageous traits so as not to be selected against (and the presence of metaptation should itself encourage this coincidence), then novel metaptive variants will become established within a population by the bootstrap principle, as an indirect consequence of first-tier selection. The long term value for metaptation, the selection based upon its functional effect, occurs most strongly at the second tier. But the genetic variation upon which this selection acts originates within individual organisms.
Significance of Metaptation.
The concept of fitness can be stretched to include the survival and reproductive
success of progeny. Indeed, this must be done to rationalize the adaptive
advantage of sex. Similarly, the concept of adaptation could be expanded
to include traits which indirectly increase the probability of survival for
descendants more remote than immediate offspring, and would thus grow to embrace
metaptation. From this perspective, an emphasis on second-tier selection
simply accentuates the singular importance of those genetic patterns whose most
potent selective value or cost emerges over very many generations and encompasses
speciation events. (See Arnold and Fristrup 1982 for further clarification
of this concept.) But such semantic stretching results in practical difficulties.
By way of analogy, life may be nothing but chemistry, but biologists do
not confine their studies to processes which can be adequately described by
stoichiometric equations. Rather, we have found it helpful, even necessary,
to label phenomena such as mitosis, reproduction, and evolution with words which
can facilitate our understanding of higher-level organization before we know
anything about the chemical reactions involved. Appreciating the causes
and consequences of selection at the second tier will also benefit from a hierarchical
terminology.
The value of metaptation as a concept distinct from adaptation lies with our inability to comprehend the details of any process whose adaptive function is extremely indirect. That indirect adaptive function can be difficult to recognize is evident from our historical difficulty in explaining even so obvious a process as sexual reproduction, whose selective advantage arises indirectly from the phenotypic consequences in succeeding generations of mechanisms for meiosis and fertilization which were of no direct benefit to the parents. Indirect advantage outweighs considerable immediate risk.
The concept of metaptation circumvents such difficulty by addressing second-tier advantage directly. Our concepts of adaptation are currently limiting the ease with which we can assimilate new discoveries in molecular genetics and developmental biology. Consider for a moment whether sex would ever have been understood, if its basic function had demanded justification in conventional adaptive terms before its intricate functional design could even have been acknowledged. How obscure might seem the "real" adaptive function for copulatory organs if our theoretical framework insisted that their contribution to sexual reproduction must be simply fortuitous, because natural selection could hardly favor any process which caused random disruption of parental gene combinations while effecting a fifty percent loss of reproductive efficiency.
Yet, this is exactly the problem we are now encountering with molecular and developmental genetics, where unexpectedly complex mechanisms for active genetic rearrangement and regulation are being discovered with surprising frequency. Most of these mechanisms have profound evolutionary implications (e.g., Echols 198l, Britten 1982, Davidson 1982, Dover 1982, Campbell 1983, Shapiro 1983a, Temin and Engels 1984, Campbell 1985). But, while the origin or selective value for such traits is assumed to depend on unknown adaptive function, their evolutionary value is commonly dismissed as "fortuitous." With this dismissal, evolutionary theory can offer no guidance to further research in molecular genetics. However, with the concepts of second-tier selection and resulting metaptation, we become free not only to perceive or propose metaptive functions for patterns which seem otherwise inexplicable (such as the phylogenetic increase in non-protein-coding "junk" DNA) but to accept second-tier selection as a plausible explanation for the evolutionary origin of such patterns. We can even begin to predict the existence of unsuspected new mechanisms. For example, any occurrence of heterochrony suggests the existence of a gene that controls a particular developmental rate; the importance of heterochrony in evolution suggests that regulatory gene-complexes may be selected for metaptive properties of gradational variation and rearrangement (see below). Every instance of effective evolutionary transformation, every instance of extensive adaptive radiation, points not only toward immediate (first tier) selection pressure but also toward the probable existence of an underlying metaptive pattern of genomic organization and developmental process.
Metaptive function.
By definition, any evolved trait which permits or promotes viable genetic variation
is a metaptation. Some very basic attributes for genetic organization
offer obvious metaptive value. One example would be a pattern of genetic
regulation which could permit mutation at a single locus to cause coherent evolutionary
variation in all systems associated with one adaptively significant trait. The
existence of high-level growth regulation, for instance, permit simple genetic
change to affect body size as a single parameter. In contrast, a hypothetical
non-metapted pattern might regulate each organ or system independently so that
genetic variation at one locus could not possibly be advantageous unless it
were improbably coincidental with concordant variation at many other loci. Another
important attribute for metaptation is gradational change. Genetic patterns
organized so that probable mutations will cause preferentially small effects
on phenotype will offer greater evolutionary versatility than would patterns
in which most mutations cause discontinuous and capricious phenotypic alterations.
"The degree of gradualism with which the phenotype changes with independent
changes in genotype is adjustable and adjusted for effective evolutionary behavior
in the course of evolution" (Conrad 1983, p. 221) through selection at both
first and second tiers.
Also valuable would be complex metaptations which could influence the number and dimension of degrees of freedom available for variation. As formulated by Vermeij (1973), the concept of evolutionary versatility requires many degrees of freedom: "Provided the parameters can vary independently with respect to one another, an increase in the number and range of the parameters results in the concomitant increase in versatility..." But degrees of freedom are not available without cost. Too many degrees of freedom can drastically reduce the probability that any random change will maintain a viable level of functional integration. Most genes have pleiotropic effects and most adaptively significant traits are influenced by several genes. Biological integration thus involves the interdependence of many different genes, so that uncorrelated variations in low-level genetic structures are likely to disrupt high-level organismic functions. Traits which might seem genetically independent at a low level of description may not in fact be independent variables at an adaptively relevant level of functional organization. As a consequence, "...integration might lead to a reduction rather than to an increase in potential versatility" (Vermeij 1973).
Any increase in the number of parameters which can vary independently might thus seen to imply a reduction in the potential for correlated, integrated changes to occur across many functionally interrelated parameters. However, both extensive integration and great versatility can be available if complex organization were to embody many overlapping regulatory functions, each with control over a different set of dependent variables. Thus changes in any one high-level regulator would cause one pattern of integrated variation in all dependent traits, while changes in other regulators would induce other patterns of dependent change. Variation in the patterns of overlapping sets of traits could permit a relatively few regulators to control a potentially unlimited set of dimensions for change. For example, many integrated aspects of bodily proportion may be developmentally (allometrically) dependent on total body size. A simple change in body size will thus cause correlated changes in many dependent parameters. But with variation in the interactions (heterochrony) among mid-level growth regulators can come shifts in allometry, so that any one of many different changes in shape could accompany a simple increase or decrease in body size. The concept is not limited to body size and shape, of course, but can be applied to any ontogenetically interrelated processes (see Gould 1977, Raff and Kaufman 1983). Primary variation in the pattern of interdependence among various parameters could remain adaptively invisible until disclosed by an evolutionary change in the ruling parameter (e.g., body size). Thus a population of individuals which were phenotypically similar but nonetheless embodied covert variation in patterns of regulatory interdependence could respond along many diverse dimensions to any adaptive selection for a change in body size. This in turn would increase the probability that at least some descendents of the population might become exapted in new and advantageous ways.
Thus, metaptations for producing genetic variation in the developmental interdependence among various parameters at different levels in a regulatory hierarchy could create tremendous evolutionary potential by establishing within a population many degrees of freedom for adaptive change while yet maintaining extensive integration within each individual genome. The process can be conveniently illustrated by referring to "Galton's polyhedron." Gould (1980) has elaborated this analogy to make the point that "organisms are not billiard balls, struck in deterministic fashion by the cue of natural selection and rolling to optimal positions on life's table. They influence their own destiny in interesting, complex and comprehensible ways." A species is visualized as a metaphorical polyhedron, resting stably (adaptively) on one facet:
"Change cannot occur in all directions, or with any increment...When the polyhedron tumbles, selection may usually be the propelling force. But if adjacent facets are few in number and wide in spacing, then we cannot identify selection as the only, or even the primary control upon evolution. For selection is channeled by the form of the polyhedron it pushes, and these constraints may exert a more powerful influence upon evolutionary directions than the external push itself" (Gould 1980).
Gould apparently assumed that the polyhedron's shape as well as its position would be fixed by first tier selection. Adjacent facets would then represent exaptations or other incidental correlates of primary adaptation. Metaptation offers an additional interpretation: The polyhedron may be malleable, with the locations and shapes of its facets amenable to adaptively neutral remodeling based on active internal mechanisms for genetic reorganization. While the species resides adaptively on one facet, metaptive variation could randomly alter the positions of adjacent facets and the sharpness of intervening edges until, under direction of selection pressure to which the species had been previously unresponsive, it rolls smoothly onto a newly prepared facet. The facet upon which the polyhedron rests is determined by current phenotype; adjacent facets represent alternative phenotypes which are available through variation in the fundamental genomic pattern. The sharpness of the edges which separate the facets represents the degree of gradualness with which genetic change can cause a shift between phenotypes. Sharp edges represent stability in the face of selection, while rounded edges represent easy transitions from one form to another. Thus metaptations which permit variation in the rules by which specific genetic modifications can affect particular aspects of morphogenesis offer a possibility for covert preparation of new adaptive modes free from constraint upon immediate adaptive value. A metaptively versatile species could include many phenotypically-equivalent genotypes which differed in dimensions for potential variation. Subpopulations could be adaptively identical yet represent polyhedra of several distinct shapes. As a consequence, such a versatile species might be able to respond to selection pressure almost as freely as a metaphorical sphere, but with diverse metaptive variants underlying its ability to escape from narrow evolutionary channels while yet maintaining adaptive integration within the complex genome of each individual organism.
Hofstadter (1985) offers another powerful metaphor for evolutionary variation: "twiddling a knob on a concept." "Making variations on a theme is really the crux of creativity." Although Hofstadter was discussing intellectual rather than evolutionary creativity, the metaphor is just as apt. A metaptation can be envisioned as a biological, genetic knob which permits the gradual and continuous variation of one parameter. To paraphrase Hofstadter's discussion of a computer for designing typefaces: A single master knob can control a feature common to a group of related subsystems. Although there may be thousands of knobs when you count all the genes which regulate the synthesis of all the proteins in the body, there will be a far smaller number of master knobs, and they will have a deeper and more pervasive influence on the whole organism. What happens, in effect, is that by twiddling the master knobs alone, you have a way of drifting smoothly through morphological space (Hofstadter 1985, p. 240). A genome with only low-level knobs might be extremely versatile, but finding any workable combination of settings for an integrated organism from among the infinite set of possibilities could be practically impossible. Practical change requires the presence of knobs for adaptively meaningful traits. "Making variations is not just twiddling a knob before you; part of the act is to manufacture the knob yourself" (Hofstadter 1985, p. 251). The existence of a knob corresponds to the presence of a rounded edge on Galton's polyhedron. Making a new master knob amounts to molding a new facet, with the addition that the "knob" metaphor recognizes potential for an unlimited hierarchy ("tangled hierarchy" is even more descriptive; see Hofstadter 1979) of interacting dimensions for regulatory variation.
Returning to biology, these metaphors suggest a hypothesis: Specific genes or genetic regulatory sequences may exist as explicit representations of evolutionary useful transformations in organism form and function. Such genes would display a relatively high rate of small-scale variation. Although such a gene need not code directly for a particular structure, variations in the gene would correspond with variations in the regulated trait. The gene could be intimately involved in developmental regulation, but the principal value as a discrete variable sequence would reflect selection for evolutionary versatility of the genome rather than immediate survival of each individual organism. The normal existence of variable genes with metaptive function could easily escape detection by normal genetic techniques. Experimental deletion could cause phenotypic effects ranging from the inconspicuous to the lethal (depending on the extent to which normal functioning of the intact gene had displaced the regulated trait from its unregulated value), but in any case, deletion of the variable gene would reveal little about the normal range of variation it had controlled. Non-destructive experimental mutation would by initial hypothesis by indistinguishable from and obscured by spontaneous variation of the same gene, while spontaneous variation could mimic variation arising from recombination among multiple alleles of a polygenic trait. Conclusive demonstration of metaptive gene function may therefore be somewhat difficult, requiring comparative studies which combine DNA sequencing with phylogenetic analysis of adaptively significant transformations. But the possible presence of a metaptive gene with variable regulatory function may be suggested by any consistent pattern of evolutionary transformation among related species.
The metaptive functions of gradational transformation and of covert preparation of novel degrees of freedom -- edge-rounding, facet-molding, knob-building -- can be accommodated by many plausible molecular mechanisms. Mechanisms for creating new proteins and for gradually modifying old ones are already familiar (e.g., Ohno 1970, Conrad 1983; Ninio 1983; Gilbert 1985). Higher-level mechanisms for establishing novel degrees of freedom for graded variation in adaptively integrated morphology are largely unknown. One conceivable mechanism which could not only accomplish this function efficiently but could also account for the presence of "meaningless" non-protein-coding DNA would be for repetitive, transposable sequences to code for non-specific regulatory function. Repetitive sequences are highly mutable, with easy transitions to longer or shorter sequences by insertion or deletion of extra repeats. Insertions and deletions in repetitive sequences are not readily recognized or corrected by intrinsic mechanisms for DNA repair (Drake, Glickman and Ripley 1983). If the length or quantity of a repetitive sequence controlled the expression of another gene, the variable-length repetitive sequence would represent a "knob" for adaptively adjusting the regulated trait. Any population with such a knob would evolve under first-tier selection to an adaptively advantageous mean setting of the variable regulator while continually generating small-scale variation on either side of the mean. The variation could itself be advantageous in a tangled-bank environment, if the rate of variation were adjusted so that only an acceptably small proportion of extreme variants represented any genetic load. Yet the evolutionary potential for such a "knob" would be limited not by the variation currently available within the population but only by the extreme limits of the adjustable regulator. An evolving population could track environmental change smoothly and continuously over a range far exceeding the potential implied by alleles extant at any particular moment, by actively and continuously generating new variation about any current mean. (Incidently, were such a mechanism to exist, one might anticipate a rather high level of viable, small-scale genetic variation even in a highly inbred population. Such effects may have been observed -- see Fitch and Atchley, 1985, for new data and review.)
Duplication and transposition of a general-purpose mutationally-adjustable regulator, one whose target would be determined by its location rather than by its specific function, would enable similar adaptive "tuning" of other genes and gene-combinations. Such non-specific regulators could also be "stacked" so that one original sequence could eventually come to control not only other genes but also lower-level versions of its own initial regulatory function. Genetic transposition of multiple copies of standard regulatory sequences would thus permit the formation of new master knobs at any level in a complex hierarchy of genomic organization. Each new pattern of genetic regulation could be adaptively invisible (no direct phenotypic effect) while creating new dimensions for future gradational adaptive variation.
The hypothetical molecular mechanism outlined above is based primarily on considerations of abstract metaptational effectiveness, of genetic properties which could contribute to evolutionary versatility and thereby be favored by second-tier selection. Nevertheless, this mechanism bears obvious resemblance to well-established patterns of movable controlling elements within the genomes of many different organisms (see Shapiro 1983b for review). Whether or not the resemblance reflects the developmental/evolutionary function of transposable genes, this discussion has hopefully illustrated how the idea of metaptation can guide hypotheses about genetic regulation. The relationship between evolutionary change and developmental regulation has been widely discussed (e.g., Davidson 1982; Raff and Kaufman 1983), but the possibility that selection acting on evolutionary versatility has shaped the mechanisms of developmental regulation is still unfamiliar. Although the neoDarwinian beads-on-a-string metaphor for genetic organization is clearly outdated, to most biologists the "new" genome still appears unexpectedly complex. Yet many features of a complex and dynamic genome may turn out to be unsurprising (even predictable!) from a second-tier perspective.
Metaptation as mutational grammar.
Language has provided a plentiful supply of metaphors for interpreting genetic
information. Thus four nucleotide "letters" yield sixty-four possible
"words" or codons, which in turn can combine into an unlimited number of protein-coding
"sentences". The concept of "generative grammar" has recently been applied
to the rules of genetic variation within the immune system (Jerne 1985). Rules
of grammar impose orderly constraints upon the sequences of letters and words
which are used for communication. Grammar does not force any particular
meanings upon the permitted sequences; grammar merely creates the possibility
for meaning by excluding an infinitely vast array of nonsense sequences. Grammar
creates freedom from chaos yet permits the expression of unlimited meaning.
Within the immune system, a generative grammar for antibody formation
can produce a tremendous number of proteins to recognize virtually any antigen
by employing a small number of rules for somatic mutation and recombination.
There is no foresight, no direction toward antibodies that can recognize
particular pathogens, yet the system is remarkably efficient at preventing disease.
Extending the metaphor of generative grammar to other forms of genetic
variation suggests that metaptations might be conceptualized as molecular or
developmental mechanisms which impose grammatical rules on mutational change,
rules which can themselves be altered by mutation. (In fact, the immune
system may be only one specialized adaptive expression of a more general molecular
grammar for cellular specification, genetic regulation and mutational variation.)
By constraining the domains in which genetic variation can occur, and
by promoting variation within domains of high probability for potential viability,
metaptation can permit the possibility of evolutionary change within systems
so complex and highly integrated that truly random mutation could never be beneficial,
at least not within the practical limits of geological time. Like grammatical
rules, metaptations may impose order on mutation. But as with antibody
formation, rules of order need not be deterministic and do not imply direction.
In a sufficiently complex system, constraints upon variation are not limits
or narrow channels but rather broad avenues toward virtually unlimited latitude
for change while avoiding chaos. Orderly but undirected variation can
be permitted and encouraged within selectively evolved constraints. Mechanisms
of disorderly, "accidental" change will be selected against.
The concept of metaptation as an evolving mutational grammar promises to reconcile the paradox of nonrandom but undirected evolutionary variation, while correcting two fallacies which recur frequently in the literature on evolution.
First is the fallacy of accidental mutation, the idea that all mutations are "mistakes in the coping of the genetic message" (Futuyma 1979, p. 233). Mutations which occur under constraint of elaborate molecular mechanisms are not satisfactorily classified as mistakes. Certainly copying errors do occur, but these are not necessarily the stuff of significant evolutionary variation. Accidental mutations are conspicuous as a direct result of the damage they cause. But the visibility of deleterious mutations, the fact that these mutations are familiar subjects for laboratory experiment, should not prejudice our attitude toward all classes of genetic modification. The difference between genetic mistakes and grammatical genetic change is the difference between accidental chromosomal abnormalities (gross aneuploidy such as trisomy-21, which is predictably deleterious) and the orderly assortment of chromosomes which occurs "randomly" within the evolved constraints of meiosis and fertilization. Orderly assortment is not directed toward any particularly adaptive phenotypes, but nevertheless maintains a very high probability for viable gene combinations while producing genotypes which never before existed.
The second fallacy assumes that if genetic change is internally caused and constrained, then the resulting phenotypic variation must be directed by the constraint. Consequently, according to this fallacy, evolutionary change must be "orthogenetic and deterministic" (Fox 1984). Again, independent assortment and recombination during sexual recombination provides a model. Within the constraints of crossing-over, segregation and pairing among homologous chromosomes, the number of possible genotypes is practically unlimited; there is no "direction." Yet a vast number of non-viable, chaotic combinations of chromosomes are effortlessly avoided by the orderly grammar of meiosis and fertilization. Constraints need not be limits, and active internally controlled mutation need not imply directed evolution, any more than non-random, grammatical sequences of letters and words imply that meaning is restricted and directed by syntax. A molecular grammar for mutation can create the orderly dimensions for variation that are necessary for evolution. But selection, and selection alone, gives direction to adaptive evolutionary change.
CONCLUSION
Mutations need not be accidents; genetic systems have an evolved ability to
generate orderly variation. But evolutionary change is only facilitated,
not directed, by orderly and internally constrained mutation. At a molecular
level, chemical reactions occur according to random thermal interactions. At
this level, the processes of life do not differ from the processes of inorganic
chemistry. But at the level of organisms, chemical reactions are controlled
by elaborate enzyme systems to serve particular adaptive functions. Although
the reactions are still stochastic, an appearance of purpose has arisen from
first-tier selection, selection for fitness. Similarly, genetic variation
occurs "at random," undirected by specific adaptive needs. But as a result
of selection at the second tier such random variation occurs within evolved
constraints and serves the metaptive function of facilitating evolutionary versatility,
just as adaptively evolved enzymes react stochastically while yet facilitating
organism survival. Phenotypic variation is thereby constrained yet undirected.
Genotypic variation is stochastic within evolved constraints, but is not accidental.
This thesis contributes to a long but unorthodox tradition which has persistently maintained that more than accidental mutation is involved in the origin of evolutionary variation. When an old idea has been repeatedly rebutted but refuses to die, when it emerges again and again in new guises supported by new data, then perhaps the idea still contains a germ of truth which has yet to find its proper expression. For most modern biologists including this author, overwhelming evidence compels belief that evolution arises entirely from mechanisms of undirected variation and natural selection. The neoDarwinian synthesis does indeed offer a correct account of evolutionary process. But evidence that biological organization also embodies complex patterns which are not yet satisfactorily explained by neoDarwinian mechanisms, at least not when the mechanisms are interpreted too simplistically, has become too strong to ignore. Recent discoveries in molecular genetics have forced the paradox of nonrandom but undirected mutation to surface once again. Fortunately, this time a growing concept of evolution as a complex hierarchical process is available (Gould 1982a,b, 1985; Arnold and Fristrup 1982; Vrba and Eldredge 1984) and may bring about a reconciliation of views which have previously seemed to be mutually contradictory. As a result, evolutionary theory may be entering a new and exciting period of growth.
The same forces which have shaped adaptation can also create metaptation. Elaborately constructed forms of mutational constraint, which encourage yet do not direct evolution, have been produced by the laws of selection acting around us. With this realization, "a grand and almost untrodden field of inquiry will be opened, on the causes and laws of variation" (Darwin 1859), a field which may soon begin to yield deeper insights into the macroevolutionary processes of ontogeny and phylogeny.
ACKNOWLEDGEMENTS
Comments and questions: dgking@siu.edu
Department of Zoology e-mail: zoology@zoology.siu.edu
Comments and questions related to web server: webmaster@science.siu.edu