The Structure of Evolutionary Theory

by Stephen Jay Gould

  Their stated preference for the second alternative (Shubin et al., 1997, p. 647) illustrates the intermediary character of this example, as recording too broad a constraint to permit the identification of any particular verte­brate limb as “the same” structure as any given arthropod appendage, but also as based upon a sufficiently detailed and complex set of genetic homologies to set a common developmental channel with more specificity than a Pharaonic brick, albeit not at the level of constraint of a full Corinthian col­umn: “This ancestral structure need not have been homologous to arthropod or vertebrate limbs; the regulatory system could have originally patterned any one of a number of outgrowths of the body wall in a primitive bilaterian for example . . . The key step in animal limb evolution was the establishment of an integrated genetic system to promote and pattern the development of certain [Page 1142] outgrowths. Once established, this system provided the genetic and de­velopmental foundation for the evolution of structures as diverse as wings, fins, antennae and lobopodia.”

  Movement three, Scherzo: Does evolutionary change often proceed

  by saltation down channels of historical constraint?

  As I documented in Chapters 4 and 5, internally channeled evolution (ortho­genesis) has been intimately linked with discontinuous change (saltationism) in the history of structuralist thought (with the model of “Galton's polyhe­dron” serving as the classical image for the connection). The linkage isn't physically necessary or logically impelled, for some orthogeneticists have fa­vored gradualism (C. O. Whitman, pp. 383-395), whereas some saltationists have rejected internal directionality (Hugo de Vries, pp. 415-451). But in ex­panding the causes of evolutionary change beyond the incremental gradual­ism of externally directed Darwinian selection, and in regarding internal channels of developmental constraint as important mediators of phyletic trending, most advocates of formalist or structuralist explanation (Bateson, D'Arcy Thompson, and Goldschmidt, for example) have supported some linkage of channeled directionality with at least the possibility of saltational movement down the channels — if only because the potential phyletic analog of such ontogenetic phenomena as metamorphosis seems intriguing and worth exploration.

  Thus, with the reintroduction of internal channeling by historical con­straint (based on genetic homology) into our explanatory schemes, we must ask whether saltational themes (that had been even more firmly rejected by the Darwinian Synthesis) can also advance a strong case for a rehearing. My own conclusions are primarily negative (hence my parsing of this theme as a scherzo, and as the shortest movement of my analysis), but the subject clearly merits some airing (and undoubtedly holds limited validity), if only as a sign of respect for the intuition of so many fine evolutionists, throughout the his­tory of our subject, that structural channeling — now clearly affirmed as a theme of central importance — implies a serious consideration of saltational mechanics.

  As discussed extensively in Chapter 9, in the context of debate over punc­tuated equilibrium, notions of “rapidity” depend strongly upon the time scales of their context. Invocations of suddenness raise quite different evolu­tionary issues at each level of consideration. In this section, I shall discuss true saltation (discontinuous changes, potentially across a single generation, and usually mediated by small genetic alterations with major developmental ef­fects), and not punctuational patterns at larger scales of time (continuous changes that would be regarded as slow and gradual across human lifetimes, but that appear instantaneous when scaled against the millions of years in stasis for a resulting species or developmental Bauplan).

  Nonetheless, I note in passing the relevance of developmental themes to punctuational patterns at these larger, and very different, scales of explana­tion. For example, several authors have argued that our emerging concepts of deep homology might help to elucidate such macroevolutionary “classics” of [Page 1143] large-scale rapidity as the Cambrian explosion. Under Lewis's (1978) original model of evolution from ancestral homonomy (multiple, identical segments) by accretion of duplicated Hox genes to achieve differentiation of specialized parts along the body axis, the Bauplan of the major animal phyla must origi­nate separately and gradually, as each added developmental component per­mits further differentiation. How, then, could so many basic designs make such a coordinated first appearance in five to ten million years, unless some genetic glitch or unknown environmental trigger initiated a rampant episode of duplication in many lineages simultaneously, or unless the pattern only represents an artifact of preservation, rather than an actual macroevolutionary event?

  But the first fruits of evo-devo (see next movement, pp. 1147–1178 for full discussion) have reversed this scenario by documenting a full complement of Hox genes in the most homonomously segmented invertebrate bilaterian phyla, thus suggesting the opposite process of loss and divergence for the differentiation of numerous complex and specialized patterns from initial homonomy (De Rosa et al., 1999). The punctuational character of the Cambrian explosion seems far easier to understand if the basic regulatory structure already existed in ancestral homonomous taxa, and the subsequent diversification of Bauplan therefore marks the specialization and regionalization of potentials already present, rather than a dedicated and individual­ized addition for each major novelty of each new Bauplan. The Cambrian ex­plosion still requires a trigger (see Knoll and Carroll, 1999, for a discussion of possible environmental mediators, including the classical idea of an achieved threshold in atmospheric oxygen), but our understanding of the geological rapidity of this most puzzling and portentous event in the evolution of ani­mals will certainly be facilitated if the developmental prerequisites already ex­isted in an ancestral taxon.

  Knoll and Carroll (1999, p. 2134) stress this point in a section of their arti­cle entitled “Cambrian diversification: So many arthropods, so little time.” They add (loc. cit., see also Grenier et al., 1997, p. 551):

  The entire onychophoran-arthropod clade possesses essentially the same set of Hox genes that pattern the main body axis. Thus, Cambrian and recent diversity evolved around an ancient and conserved set of Hox genes . . . Increase in segment diversity is correlated with changes in the relative domains of Hox gene expression along the main body axis . . . Most body plan evolution arose in the context of very similar sets of Hox genes, and thus was not driven by Hox gene duplication . . . Bilaterian body plan diversification has occurred primarily through changes in developmental regulatory networks rather than the genes themselves, which evolved much earlier.

  But returning to the very different, central subject of this section — the pos­sibility and meaning of evolutionary saltation at the organismic level of dis­continuity across generations — we may at least assert a case for plausibility, so that, at the very least, this perennially contentious subject will not be dis­missed a priori. First of all, we cannot deny either the existence of such large [Page 1144] and discontinuous phenotypic shifts in mutant organisms, or the conven­tional basis assigned to them: small genetic alterations with major develop­mental consequences. For example, a single base substitution in bicoid, the maternal gene product that sets the AP axis by supplying positional informa­tion within the Drosophila larva, can reverse the axes of symmetry (Frohnhofer and Nusslein-Volhard, 1986; Struhl et al, 1989). Of this and other cases, Akam et al., in the introduction to their 1994 book on The Evolution of Developmental Mechanics write (1994, p. ii): “It is a commonplace of de­velopmental genetics that minimal genetic change can lead to the most dra­matic morphological effect.”

  Second, we can also posit believable mechanisms that avoid the classical problems of specifying how genes with such disruptive effects could ever be integrated into the intricate and finely tuned development of a complex metazoan ontogeny and, even if viable, how such saltations could spread through populations following their mutational origin in single individuals, given the almost inevitable fitness depression that must accompany any inter­breeding with modal individuals. Schwartz (1999), for example, pre
sents a modern version of the old argument for origin in a nonlethal recessive state, followed by accumulation without expression in heterozygotes until the achievement of a critical frequency permits an effectively simultaneous overt appearance of the phenotype in numerous homozygotes. Addressing the second aspect of this problem, and basing their case on a viable homeotic mutant in a homonomous species of centipedes, Kettle et al. (1999, p. 393) argue that most workable mutations of such large effect may arise in homonomous ancestors of more specialized groups: “Perhaps the severe fitness depression accompanying homeotic transformation would have been less pronounced, even absent, in a primitive arthropod with many similar seg­ments.”

  But mere plausibility doesn't imply likelihood, and two strong arguments would seem to indicate a minimal role for the evolutionary efficacy of such developmental saltations: first, a negative statement based on the fallacy of usual sources of inference; and, second, a positive argument about more plau­sible alternatives for the same sources of inference.

  For the negative statement, both common arguments for inferring saltational origins from modern developmental patterns falter upon the general fallacy (discussed in detail in Chapter 11) of invoking current circumstances to make unwarranted inferences about historical origins:

  1. The fact that major phenotypic effects accompany the repression or al­teration of key developmental switches in modern organisms does not imply a saltational origin either for the switch itself, or for any extensive conse­quences of its mutational variations. For example, the fact that Hox genes now repress the expression of Dll in the insect abdomen, thus suppressing the development of appendages — and that a single mutation (albeit ultimately le­thal) in one Hox gene can reverse this effect and emplace leg rudiments on each larval segment (Lewis, 1978) — does not permit the inference that insects lost their abdominal legs in one phylogenetic saltation.

  2. Reasonable inferences about saltational losses cannot be theoretically inverted [Page 1145] into hypotheses about sudden origins for the developmental cascades thus repressed. (We note here a developmental version of the error so com­monly committed as a thoughtless consequence of naming the normal func­tion of genes for the results of their discombobulation. If the mutational si­lencing of a gene precludes the development of a child's ability to read, we have not thereby identified a “reading gene,” although such taxonomies and inferences remain all too frequent in our literature, particularly on human cognitive abilities.) For example, the fact (Swalla and Jeffery, 1996) that a loss-of-function mutation in the Manx gene of the tunicate Mogula can re­press chordate features and lead to a tailless (anural) larva — and that this function can be restored in interspecific hybrids with other Mogula species that develop with a tailed (urodele) larval form — does not imply that the tuni­cate larval tail and notocord (once a popular theme in theories about the origin of vertebrates, as in the classic paper of Garstang, 1928) arose by saltational introduction of Manx activity. (Swalla and Jeffery make no such inference, of course, and I hate to see their fascinating discovery so misused, as in many press reports.)

  As a general structural principle, applicable across a full range of natural phenomena, from cosmology to human social organization, complex systems can usually collapse catastrophically, whereas the construction of such func­tional intricacy can only occur by sequential accumulation — a pattern that I have called “the great asymmetry” (Gould, 1998a).

  For the positive argument, more plausible continuationist scenarios can ex­plain the modern phenomena that most often tempt us to invoke hypothetical saltation to resolve their origin. In an important article, for example, Akam (1998) discusses the property of Hox action that has often led to saltational inferences: “The Hox genes might justifiably be considered master control genes (Gehring, 1996) for segment identity. For most segments of the insect trunk, they provide the only conduit for channeling axial information from the early embryo to cells at the later stages of development.” Akam then ex­poses the same fallacy that I discussed above as negative point 2: “It is tempt­ing to shift this process into reverse, and to assume that segment diversifica­tion has been achieved by a series of overt homeotic mutations generating novel complexity.”

  Akam then develops a much more plausible evolutionary model for the incremental origin of developmental patterns mediated by so-called “selector genes,” which have generally been viewed “as stable binary switches that di­rect lineages of cells to adopt alternative developmental fates” (Akam, 1998, p. 445). Akam proposes an alternative concept “for the regulation of Hox genes within compartments” by “enhancer modules,” conceptualized as “lo­cal signals, hormone receptors or any of the other stimuli that commonly me­diate gene regulation. In this regard, it makes the Hox genes like any other genes. It predicts that small changes, particularly in the structure of their promotor modules, will change the phenotype of segments” (p. 448). “By ac­cepting a role for the regulation of Hox genes within compartments,” Akam adds (p. 448), “we demote them from their privileged status as stable binary switches.” [Page 1146]

  Akam envisages the gradual evolution of different enhancers (or different levels and mixtures of the same enhancers) in various arthropod segments, leading to a phyletically diverging regulation of the same Hox gene down the body axis, but with continued expression in a segment-specific manner. Ubx for example is regulated by the 'abx' enhancers in parasegment 5, which inte­grate patterning information in one way, but by the 'bxd' enhancers in para­segment 6, which specify a different within-segment pattern.” As these alter­ations in expression evolved gradually within different segments, “the change would not necessarily be recognized as a 'homeotic mutation'” (p. 448). These and other models reinforce the important principle that extensive and discontinuous phenotypic effects in the development of modern organisms do not imply the saltational origin of these features in phylogeny.

  Other models, however, permit more space for an important frequency of saltational shifts in evolution. Duboule and Wilkins (1998), for example, tie an increasing propensity for saltational change to functional recruitment of genes for multiple tasks: “Transitions from gradual to discontinuous rates of evolutionary change are an inevitable consequence of the multiple use of genes through evolutionary tinkering, given appropriate selective pressures” (1998, p. 54). Interestingly, in the context of this chapter and the history of this subject, their model explicitly links this increasing frequency of salta­tion to the “hardening” of internal constraints that arise as a consequence of incorporating key genes into multiple networks of regulation. They write (p. 58): “The greater the number of networks that a gene product is involved in, the smaller the scope for new variations to be offered to natural selection. The idea of internal constraints leading to restrictions in the production of evolutionary novelties is not new. However, we would like to argue that inter­nal constraints result, indirectly but inevitably, from the increasing work load imposed by successive recruitment of genes to new functions.” In networks of such complexity, they conclude, any “novel equilibrium will have to be estab­lished as a one-step event and not through the accumulation, in time and space, of many mutations of small effect, or gradualistic change” (p. 58).

  In any case, and however important such saltational changes may be in establishing fundamental evolutionary novelties (my own betting money goes on a minor and infrequent role), phyletic discontinuity at lower taxonomic levels, based on small genetic changes with large regulatory effects, has been documented in several cases. In a fascinating example, Ford and Gottlieb (1992) found that about 20-30 percent of several hundred Clarkia concinna plants growing in a single locality at Point Reyes, California displayed the bicalyx mutant, a homeotic variation that replaces the usual circlet of four bright pink petals with a second circlet of sepals.

  By Mendelian analysis of ratios in cross breeding between normal and bicalyx plants, Ford and Gottlieb established that a single point mutation produces the bica
lyx phenotype. Moreover, and in contrast to many home­otic mutations, the bicalyx plants show no developmental abnormalities and no apparent fitness depressions; insect pollinators continue to visit bicalyx flowers with no apparent reduction in frequency. Ford and Gottlieb note (1992, p. 673) that “homeotic mutants have been found and propagated in [Page 1147] gardens, but have almost never been reported as natural populations” — thus giving special interest to this case of viability in nature. (They also men­tion the intriguing historical footnote that Linnaeus himself found a peloric Linaria (with normal upper petals replaced by spurred lower petals), appar­ently growing in abundance within its habitat, but probably seed sterile, and therefore arising as a vegetatively propagated clone.)

  Ford and Gottlieb outline a reasonable scenario for the promotion of this fecund, and naturally growing, homeotic form to specific status (1992, p. 673):

  The absence of deleterious pleiotropy or fitness-reducing epistatic inter­actions in bicalyx suggests that mutations with extensive morphological consequences can be successfully accommodated by plant developmental systems. If such mutants were to become associated with chromosomal rearrangements reducing the fertility of hybrids between them and their progenitors, a process that has occurred repeatedly in Clarkia, the new population would probably be accorded species status (p. 673). Bicalyx demonstrates that a large morphological difference governed by a simple genetic change can become established in a natural plant population (p. 671).