The Structure of Evolutionary Theory

by Stephen Jay Gould

  In this particular case, for example, Deutsch and Le Guyader (1998) have suggested a historically prior function for Hox (and other zootype) genes in designing “an appropriate neuronal network in bilaterian animals” (p. 713). Recognizing the relevance of this idea to the issue of contingency and the Cambrian explosion (1998, p. 716), they write: “Hence, the presence, before the Cambrian explosion, of a large number of Hox genes, whose domains of activity extend from the post-oral head to the abdomen, cannot be accounted for by a function in driving morphological diversity. Another role has to be assumed for the ancestral function of the Hox genes. We postulate that the zootype genes primitively specified neural identity.”

  Second, we must ask if the realized variants that congealed so quickly as specialized and differentiated body plans (the major bilaterian Bauplane) — permitting no further origin of novel anatomies sufficiently distinct to war­rant taxonomic recognition as phyla — represent a predictable set of “best so­lutions” within the broad possibilities of historical constraint permitted by shared developmental rules? Or do they constitute a subset of workable, but [Page 1161] basically fortuitous, survivals among a much larger set that could have func­tioned just as well, but either never arose, or lost their opportunities, by his­torical happenstance? I admit my partisanship for the latter position (Gould, 1989c) and freely acknowledge that my judgments have won some support, but no consensus to say the least (Conway Morris, 1998). I would only point out that even the strongest opponents of contingency admit that arthropod disparity (the measured range of anatomical designs, not the number of spe­cies) had reached a fully modern range in the Burgess Shale faunas (Middle Cambrian, about 10 million years after the explosion) — and that more than 500 million years of additional arthropod evolution has not expanded the scope of anatomical disparity at all (Briggs et al., 1992; Foote and Gould, 1992, present evidence for the counter view that Cambrian disparity ex­ceeded modern levels, despite much lower species diversity). I would also urge my colleagues to spend more time studying Cambrian “oddballs” that do not easily fit into recognized higher taxa, including Xidazoon among “or­phan” taxa (Shu et al., 1999), or Fuxianhuia among arthropods that do not belong to any recognized class (Chen et al., 1995), and not to focus so strongly, as most studies have done in recent years, upon cladistic attempts to place all Cambrian forms at least into the stem regions of major phyla, if shared derived markers of crown groupings bar their entry — a strategy that leads researchers to ignore the autapomorphies of these peculiar taxa, and to coax other features into plesiomorphy with modern taxa.

  CHANNELING THE SUBSEQUENT DIRECTIONS OF BILATERIAN HISTORY FROM THE INSIDE. If the bilaterian ancestor possessed a full complement of Hox genes, and if all major variants upon this initial system had already con­gealed by the end of the Cambrian explosion, then subsequent bilaterian evo­lution must unfold within the secondary strictures of these realized specializa­tions upon an underlying plan already channeled by primary constraints of the common ancestral pattern. But lest we begin to suspect that rigid limita­tion must represent the major evolutionary implication of such constraint, I must reemphasize the positive aspect of constraint as fruitful channeling along lines of favorable variation that can accelerate or enhance the work of natural selection. Moreover, the evolutionary flexibility of developmental channels achieves its most impressive range — as Chapter 11 will discuss as its primary subject — through the crucial principle of cooptation, or the extensive and inherent capacity of genes evolved for one particular function to operate, through evolutionary redeployment, in strikingly different adaptive ways.

  Among “higher” triploblast phyla of markedly divergent design, echinoderms represent the obvious test case for studying the flexibility of homolo­gous developmental genes. With their remarkable autapomorphies of radial symmetry, calcitic endoskeleton, and a water vascular system for circulation, how could these creatures evolve within the confines of a genetic regulatory system that builds bilaterial, axially specialized organisms with blood vascu­lar systems in both their immediate sister phylum (the vertebrates) and in plesiomorphic taxa of more distant common ancestry (the protostome phyla [Page 1162] on both major branches). Did echinoderms delete their ancestral determi­nants to evolve such an aberrant morphology or did they acquire entirely new regulatory genes and developmental rules?

  Few data now exist to address this important issue, but preliminary results suggest that echinoderms have retained their genetic homologies with other bilaterian phyla, while coopting several of these genes (with stable function in other phyla) for different roles in their own unique development. In a pio­neering study, Lowe and Wray (1997) documented the expression in echino­derms of orthologs of three important regulatory genes that encode transcrip­tion factors with a homeodomain, and that generally function in the same broad way in both vertebrates and arthropods (and must therefore be plesiomorphic to any derived condition in echinoderms): distal-less for proximo-distal patterning in outgrowth of limbs, engrailed for neurogenesis along the axis of the CNS, and orthodenticle for the differentiation of anterior structures.

  Lowe and Wray documented a full spectrum of results, ranging from reten­tion to cooption for markedly different echinoderms roles. At an extreme of retention, the brittle star Amphipholis squamata expresses engrailed in neuronal cell bodies along the five radial nerves. Lowe and Wray note (1997, pp. 719-720): “This expression is superficially similar to that in bilaterial an­imals, in which engrailed is expressed within a serially repeated subset of ganglionic neurons along the antero-posterior axis. It is possible that a neurogenic role for engrailed is widely conserved among triploblastic animals.”

  In the intermediary state of a retained general role transferred to novel or­gans, sea urchins express distal-less at the distal ends of the five primary po­dia (tube feet) soon after their formation — thus preserving the standard func­tion of regulating outgrowths from a body axis by expression at their distal tips, but now applied to an autapomorphic outgrowth with no homolog in any other bilaterian phylum! Finally, at the extreme of full cooptation (for new functions in new organs), brittle stars express orthodenticle in ectoderm overlying the terminal ossicles at the ends of the arms — a position with only tenuous and hypothetical connection to the anterior end of the AP axis in bilaterian phyla. Moreover, at least for engrailed and orthodenticle (the copy-number of distal-less remains undetermined in echinoderms), only one ortholog exists in any echinoderm studied so far — so new functions cannot be ascribed to the cooptation of duplicated copies.

  Lowe and Wray's final statement (1997, p. 721) emphasizes the important conclusion that genetic and developmental homologies of triploblast animals still permit enormous flexibility in evolutionary diversification — primarily by the principle of cooptation: “The highly derived body architecture of echino­derms evolved at least in part through extensive modifications in the roles and expression domains of regulatory genes inherited from their bilaterial an­cestors. Even the limited number of genes and species we examined demon­strates a remarkable evolutionary flexibility in genes that have previously been considered interesting mainly for their conserved roles in arthropods and chordates.” [Page 1163]

  If we now turn our attention to these “conserved roles in arthropods and chordates,” at least three sources of evidence underscore the central con­clusion of this section: that the evolution of differentiated and specialized Bauplane from a presumably homonomous common ancestor proceeds — paradoxically, and contrary to the scenario of Lewis's (1978) original hypoth­esis about Hox genes — by reduction and restriction, rather than by addition of genes or expansion of their domains of activity.

  1. Specialization of anatomy sometimes correlates with deletion or unem­ployment of Hox genes, or their redeployment to other functions. For exam­ple, Zen, the insect ortholog (in both structure and position) to vertebrate Hox3, exhibits no Hox function and is not expressed along the AP axis of the developing larva, but
plays some role instead in the formation of extra-embryonic membranes. In a paper that wins, by acclamation, the Steinbeckian prize for title parodies (“Of mites and Zen”), Telford and Thomas (1998) cloned the homolog of Drosophila Zen in the orbatid mite Archegozetes. They found expression of this chelicerate homolog “in a discrete antero-posterior region of the body with an anterior boundary coinciding with that of the chelicerate homolog of the Drosophila Hox gene proboscipedia” (p. 591). This fascinating result suggests that zen may have lost its Hox func­tion in Drosophila as a consequence of functional redundancy due to overlap with another Hox gene.

  Taking the argument further, Telford and Thomas present evidence that the Drosophila pair-rule gene fushi tarazu may also be “a divergent Hox gene that has adopted a new role” (p. 594; see also Dawes et al., 1994, on a locust homolog of fushi tarazu that shows no pair-rule function). These observa­tions on the original roles of Drosophila zen and fushi tarazu suggest “that the original complement of arthropod Hox genes must be revised from eight to ten” (Telford and Thomas, 1998, p. 594), thus emphasizing the role of gene loss in the specialization of body plans.

  2. Stasis or slow change in Hox genes indicates their conserved role in evolution. Akam et al. (1994), in a section of their paper entitled “Hox genes that got away,” contrast the conservation of Hox genes in insects (as docu­mented by high levels of sequence similarity among taxa) with much higher rates of divergence in homeobox genes that do not now function within the Drosophila Hox series, but may have belonged to the Hox cluster of an ar­thropod common ancestor: the maternally expressed bicoid (encoding the morphogen that produces a crucial AP gradient in the early syncytial em­bryo), the pair-rule gene fushi tarazu, and the two zen genes. The putative orthologs of fushi tarazu in other insects “are almost unrecognizable outside of their homeodomains, and have accumulated approximately 10 times as many changes in their homeodomains as have homeotic [i.e., Hox] genes in the same comparisons” (Akam et al., 1994, p. 209). The authors then gener­alize about these non-Hox homeobox genes (p. 214): “We think that these genes may be derived . . . from Hox genes which, in the lineage leading to Drosophila, have escaped from the conservative selection that characterizes homeotic genes.” [Page 1164]

  3. Overexpression, in both position and amount, of vertebrate Hox genes has generated atavisms in several experiments, thus suggesting that derived specializations evolve by tighter regionalization and restriction of expression in individual Hox genes. Pollock et al. (1995) studied the influence of over-expression for Hoxb-8 and Hoxc-8 upon the skeletal development of mice, concluding that “many of the morphological consequences of expanding the mesodermal domain and magnitude of expression of either gene were atavis­tic” (p. 4492). For example, the earliest Paleozoic vertebrates grew “free ribs” (independent from and articulating with the vertebrae) along the entire body axis, from the base of the skull to the tail. Many subsequent tetrapod lineages, particularly among mammals, reduced the number of free ribs dra­matically. But vestiges of the ancestral free ribs sometimes remain as small units fused with the vertebrae. In particular, the lumbar pleurapophyses of posterior mammalian vertebrae “most likely represent an ancestral rib that has fused with the lateral portion of the vertebrae and now serves as a point of attachment for muscle groups of the back” (p. 4495). Pollock et al. (1995) documented “the reappearance of free ribs at the expense of lumbar pleurapophyses” in Hoxb-8 transgenic mice — “a clear example of atavism” (p. 4495). In another experiment, mice developing with overexpression of both Hoxb-8 and Hoxc-8 grew costal tubercles on their lower thoracic ribs. Costal tubercles represent a vestige of the second head of the articulating boss in free ribs. Normal mice develop no costal tubercles on these ribs at all.

  In a similar experiment, Lufkin et al. (1992) ectopically expressed a Hoxd gene “more rostrally than its normal mesoderm anterior boundary of expres­sion” (p. 835) at the level of the first cervical somites. This anomalous ante­rior expression generated “a homeotic transformation of the occipital bones towards a more posterior phenotype into structures that resemble cervical vertebrae” (p. 835). One should not read too much evolutionary meaning into one experimental manipulation, but since the same ectopic expression also induced other changes of a potentially atavistic nature (particularly “the presence of clearly segmented neural arches arising from the most anterior so­mites,” p. 840), and since the vertebrate skull and forebrain probably arose as novel features at the anterior end of a more homonomous ancestor, any transformation of skull parts towards the phenotype of more homonomous posterior vertebrae can hardly fail to elicit thoughts about the phylogeny of the vertebrate skull — especially when these potential atavisms arise by revers­ing the presumed phyletic restriction and posterior localization of Hox ac­tion. Lufkin et al. (1992), in what I can only regard as an expression of chutz­pah (but still worth pondering), even ask: “Would ectopic expression of additional Hox genes be required to convert fully the neurocranium into ver­tebrae?” (p. 840).

  Pollock et al. (1995, pp. 4495-4496) also reach a bold conclusion that may go too far, but that merits careful consideration:

  The observation that expansion of the functional domain of a Hox gene can result in the transformation of a modern costal structure to a more [Page 1165] ancient form suggests that regional repression of Hox gene expression could have played a role in the evolution of the vertebral column . . . We propose that, in antecedent vertebrates, Hox genes involved in pat­terning the axial skeleton were expressed in relatively broad regions of paraxial mesoderm. The resulting less complex Hox code would have es­tablished the similar vertebral identities observed in broad regions of early vertebrate skeletons. Regionalization of the vertebral column sub­sequently evolved in concert with the evolution of restricted patterns of Hox gene expression in paraxial mesoderm.

  When we turn to specific examples in the evolution of differentiated and complexified axial structures in arthropods and vertebrates, we commonly find correlations between morphological inventions and the restriction and regionalization of Hox gene expression. I have already discussed several cases in previous sections of this chapter — particularly the complex and elegantly documented story of multiple parallel evolution of crustacean maxillipeds from ancestral limbs that were more homonomous with the rest of the poste­rior thoracic series. This complexification originates by suppression of Ubx and abdA in just those anterior thoracic segments that develop maxillipeds in each case.

  Enlarging upon this example, Averof and Akam (1995) found that three middle Hox genes — Antp, Ubx and abdA — “are expressed in largely overlap­ping domains in the uniform thoracic region” of branchiopod crustaceans (1995, p. 420, and see Fig. 10-27). In the more highly differentiated insects, on the other hand, the same three genes show more restricted expression in discrete domains, where “they specify distinct segment types within the tho­rax and abdomen” (p. 420). In particular, Antp turns on in thoracic segments (that develop legs and wings) but not in abdominal segments. Similarly, both Vbx and abdA are expressed in all abdominal segments, where they repress distal-less and therefore presumably regulate the repression of legs on all ab­dominal segments.

  In an elegant affirmation that has already become a classic of the evo-devo literature, Warren et al. (1994) demonstrated that lepidopteran larvae de­velop prolegs on their abdominal segments by localized deletion of Ubx and abdA expression, followed by subsequent derepression of distal-less in the small bilateral patches on each abdominal segment, from which the prolegs then grow. They conclude (1994, p. 458) that “abdominal limb formation in butterflies has been made possible by the evolution of a regulatory mecha­nism for shutting off these two BX-C [Hox in modern terminology] genes in selected cell populations, which then permits Dll and Antp to be expressed.”

  Warren et al. then raise an obvious question that exemplifies a related prin­ciple in the evolution of differentiated complexity from homonomous ances­try: Why didn't butterflies evolve
their abdominal prolegs by the “easier” route of fixing a Dll mutation to release repression by Ubx and abdA? Why follow the more complex scenario of first repressing the two Hox genes in lo­cal patches, and then permitting the ordinary action of Dll? Warren et al. offer [Page 1166] the reasonable explanation that derepression of Dll will not be sufficient, by itself, to build a full proleg. A much more extensive cascade of down­stream genes generates the proleg (presumably with a crucial boost from Dll) — and upstream Hox repression must be released in order to potentiate the full downstream cascade. Thus, differentiation from homonomy may pro­ceed either by regionalizing and restricting the Hox genes themselves, or by altering and specializing the downstream cascades regulated by Hox genes.

  As a further example of this second process, Warren et al. (1994) also documented the absence of Ubx from Drosophila wing imaginal discs, whereas Ubx occurs at high levels in the section of the dorsal metathoracic disc that generates the halteres (markedly reduced wings that function as balancing or­gans) on the third thoracic segment. This observation led to the conjecture that, just as Hox genes regulate the appearance or repression of legs on par­ticular segments, Hox genes might also determine the presence or absence of wings in a similarly direct manner. Thus, if Ubx prevents full wing develop­ment in Drosophila T3, perhaps a suppression of Ubx permits the generation of a large and complete second pair of wings on the homologous T3 segment of Lepidoptera.