Despite these caveats, we can only conclude that the first fruits of evo-devo have revealed some remarkable, and extensive, homology in both genetic structure and action among animal phyla (particularly between arthropods and chordates, the former prototypes of separation in our traditions and literature), and that these data have confirmed some important aspects of the most ridiculed formalist theory of constraint in the history of morphological and evolutionary thought: Geoffroy's claim for homology between vertebrate and arthropod segments, with the idealized segment itself regarded as the basic unit of generation.
The history of our dawning realization — as expressed in an acknowledgment that this heresy of homology between phyla must be reclothed in modern genetic language as a partial reality — also followed an interesting pathway of strong reluctance gradually yielding to bemused acceptance of ever widening scope. From the initial Mayrian stance of near theoretical impossibility for recognizable genetic homology, we first admitted (with the discovery of vertebrate Hox genes) that footprints of common ancestry could be preserved during more than 500 million years and such substantial anatomical divergence. But we still doubted that such genetic resemblances could continue to encode phenotypic homologies in these disparate phyla. Then, in a second step, we acknowledged the potential homology of Hox action in general spatial organization along the A-P axis, but still declined to accept any common basis for segmentation in arthropods and vertebrates (the key to Geoffroy's hypothesis).
In a third stage, the similarity of Hox action in patterning vertebrate rhombomeres and arthropod segments demonstrated at least some homology between modes of differentiation in arthropod segments and in the compartmentalized organization of the vertebrate hindbrain and its extensive derivatives. In a fourth stage, researchers then discovered some partial and limited homologies in earlier determinants of segmentation itself (pair-rule and segment-polarity genes) between arthropods and some aspects of segmental development in the main vertebrate body axis posterior to the rhombomeres.
In short, and in a story to be, no doubt, extensively continued (expanded, contracted, changed, reinterpreted, etc.), Geoffroy's theory of complete and overarching homology based on a common vertebral archetype surely will not prevail in anything like its original form. But this most ridiculed of all heresies, so contrary in principle to strict Darwinian expectations of the Modern Synthesis, and so widely dismissed as a romantic delusion until just a few years ago, has now resurfaced, in appropriately revised terms, as a primary, and initially surprising example of the unanticipated durability of ancient genetic pathways, and of their continuing power to constrain the subsequent [Page 1117] phylogeny of life along broad and fruitful (but still limited) routes of wonderfully diverse, but historically rule-bound, adaptive designs.
GEOFFROY'S SECOND ARCHETYPAL THEORY OF DORSO-VENTRAL INVERSION OF THE COMMON BILATERIAN GROUNDPLAN. The ridicule heaped upon Geoffroy's second archetypal theory for homologizing arthropods and vertebrates did not descend entirely from his genuine and original argument, but from an explicitly phyletic version championed by later evolutionists, but never intended by Geoffroy himself. (Geoffroy maintained a generally supportive attitude towards an evolutionary world view, and even gave his name to a theory of causation that he did advocate in passing, but never really developed in extenso — the idea that soft inheritance could operate by immediate impress of external conditions upon parts of organisms, yielding inheritable changes directly, rather than by the more indirect route of Lamarckian organic response to “felt needs.” In fact, late 19th century discussions of evolutionary mechanisms often listed three primary contenders: Darwinism for the theory of natural selection, Lamarckism for soft inheritance by organic response, and so-called “Geoffroyism” for soft inheritance following direct imposition. In this light of Geoffroy's positive attitude to evolution, his failure to cite, as support for transmutation, his archetypal theory of dorsoventral inversion between insects and vertebrates provides strong evidence that he did not intend this anatomical comparison to be read in a phylogenetic context.)
After resolving (to his satisfaction) the common structure of vertebrate and arthropod segmentation, and dealing with the inconvenient fact, for a hypothesis of homology, that vertebrates grow internal hard parts and arthropods an exoskeleton (see pp. 304-306 for Geoffroy's ingenious, and gloriously wrong, resolution), Geoffroy moved to a second archetypal theory for the outstanding remaining difference in basic anatomy between the phyla: their apparently reversed dorsoventral orientations — for the two main nerve cords of arthropods run along the ventral surface below the central gut, while the single nerve tube of vertebrates runs along the dorsal surface, above the gut. As we all learned in Biology 1 (but usually not with proper respect or understanding for Geoffroy's interesting conjecture), Geoffroy resolved this striking difference by suggesting that the same groundplan underlay the development of both phyla, but that this common design appeared in reversed orientation, with arthropods interpreted as, essentially, vertebrates turned on their backs (see Fig. 10-19).
We also learned, quite correctly — for Geoffroy's supporters never resolved this issue with any plausibility — that such a reversal scarcely brings the two phyla into complete correspondence. In the knottiest remaining difficulty, the front end of the arthropod gut passes between the nerve cords and emerges on the lower surface as a ventral mouth. In an overturned position, the vertebrate mouth should therefore pass by the dorsal nerve cord and emerge on the upper surface. But vertebrate mouths are also ventral. So Geoffroy weakly argued that the mouths of the two phyla are not homologous — and [Page 1118] that the original mouth, which would have opened dorsally in vertebrates, simply closed up, while a separate ventral mouth originated as a neomorph.
This history, discussed extensively in Chapter 4, becomes crucially relevant to the modern validation of Geoffroy's own centerpiece for his theory of dorsoventral inversion. I must also address this question in the largely non-historical second half of this book because later exegetes have seriously misrepresented Geoffroy's intent by substituting a later phyletic version that modern research in evo-devo rightly rejects — and that, if conflated with Geoffroy's actual theory, will lead to continued and unjustified dismissal of his second, and remarkably ingenious, archetypal formulation.
In short, Geoffroy never advanced (and, I suspect, never even conceptualized) a historical argument about direct evolutionary transformation: the claim that vertebrates evolved from arthropods when an ancestral trilobite or merostome literally flipped upside down during its phyletic ascent. However, the American morphologist William Patten did popularize such an evolutionary account in arguing that the first prominent group of putative fossil vertebrates, the jawless “ostracoderms,” did not belong to a completed vertebrate line, but represented an intermediate stage between arthropods and fishes along the “great highway of organic evolution” (his words, see Patten, 1912, 1920), with the transition literally achieved by anatomical inversion, as an arthropod that swam on its back settled to the bottom, thus converting its original dorsal side to a new belly. (The ostracoderms, with their external plates,
10-19. An unintentionally amusing illustration from Gaskell, 1908, showing the inverted topology of vertebrates and arthropods — with major nerve cord above the gut in vertebrates and below in arthropods.
[Page 1119]
do resemble, but by convergence, the eurypterid arthropods in several features of external form and function.)
For Geoffroy, however, the inversion of axes between vertebrates and arthropods does not denote an evolutionary transition in either direction, but represents instead (as the archetypal mode of thinking would imply) two opposite specializations upon a shared abstract groundplan that generated both great phyla along predictable pathways of internally specified laws of form and their permissible transformations.
As a structuralist thinker, committed to a formal, rather than a functional
, approach to the explanation of organic design and variation, Geoffroy argued that the apparently fundamental difference in disposition of organs between arthropods and vertebrates should be reconceptualized as both secondary and superficial — a consequence of opposite ecological orientations for the same archetypal structure. The shared and constraining pattern specifies a central gut and a peripheral major channel for the nervous system, both oriented parallel to the body's A-P axis. Vertebrates, so to speak (and befitting their higher status and dignity), have oriented their main nerve tract upwards toward the sun and surface, while the humbler arthropods have directed the same peripheral aspect of archetypal form downward towards the earth and ocean bottom.
A functional theory, like Darwinian natural selection, would tend to interpret this ecological correlation as a primary impetus for the later evolutionary fixation of these two opposite arrangements. But to a structuralist thinker like Geoffroy, the same ecological situation becomes both derivative and temporally consequential (not to mention ideologically inconsequential as well). The established differences represent a realized subset of possible transformations for an archetypal form under structurally determined rules of geometric constraint and possibility. The happenstance of opposite ecological orientation for a common archetypal design only records a later adaptive overlay — a diversity of form arising for structural reasons and then finding both an appropriate suite of functions and the right environments for their realization. Thus, in Geoffroy's view, the inversion of dorsoventral axes in arthropods vs. vertebrates does not validate a direct flip of evolutionary transformation, but rather represents two separately developed expressions of a common archetypal structure, one oriented up towards the sun, the other down towards the earth, in a secondary ecological specialization that can only obscure an underlying, and truly ruling, unity of constrained design.
The modern version of Geoffroy's vision — so different in genetic and evolutionary (as opposed to formal and archetypal) evidence, yet so eerily similar in philosophical style (as a structuralist account based on internal channels of transformational constraint) — originated in the mid 1990's based on unanticipated discoveries of genetic homology in genes that operate in patterning dorsal and ventral surfaces and structures in Drosophila and Xenopus. (See Sasai, et al., 1994; Holley et al., 1995; and De Robertis and Sasai, 1996, for the pioneering work of De Robertis's lab at UCLA, and Francois et al., 1994; and Francois and Bier, 1995, for studies of similar import from Bier's lab at [Page 1120] the University of California, San Diego. See also the general commentary of Hogan, 1995; De Robertis, 1997; and Gould, 1997c.)
The chordin (chd) gene of Xenopus codes for a protein that operates in patterning the dorsal side of the developing embryo, and also plays an important role in formation of the dorsal nerve cord. But sog, the homolog of chd in Drosophila, is expressed on the ventral side of the developing larva, where it acts to induce the formation of ventral nerve cords. Thus, the same gene by evolutionary ancestry acts in the development of both the dorsal nerve tube in vertebrates and the ventral nerve cords in Drosophila — in conformity with Geoffroy's old claim that the two phyla can be brought into structural correspondence by inversion.
Two further discoveries then promoted this intriguing hint into a strong case. First, major gene acting in development and specification of the dorsal surface in flies (decapentaplegic, or dpp) has a vertebrate homolog (Bmp-4) that patterns the ventral side of Xenopus. Moreover, the entire system seems to operate in a similar manner — but inverted — in the two phyla. That is, dpp, diffusing from the top to the bottom, can antagonize sog and suppress the formation of the ventral nerve cords in Drosophila — while Bmp-4 (the homolog of dpp) diffusing from the bottom to the top, can antagonize chordin (the homologue of sog) and suppress the formation of the dorsal nerve cord in vertebrates (see Fig. 10-20).
Second, the fly genes work in vertebrates, and vice versa. Vertebrate chordin can induce the formation of central nerve tissue in flies, while fly sog can induce dorsal nerve tissue in vertebrates. These three discoveries, taken
10-20. A highly idealized and schematic illustration of the common developmental pattern in the inverted topologies for gut and nerve cords in vertebrates and arthropods. Chd in Xenopus is the homolog of Sog in Drosophila — and both help to regulate the development of the nerve cords, dorsally in vertebrates and ventrally in arthropods.
[Page 1121]
together, offer strong support for a modern recasting of Geoffroy's old theory of inversion.
These studies have aroused considerable excitement and controversy, and a substantial set of alternative interpretations has been proposed. But, in my judgment, some of these objections attack the wrong target (Patten's hypothesis of direct evolutionary transition, not Geoffroy's argument for common structural design), while others raise legitimate questions of an interesting and fundamental nature, although not yet resolvable by information now in hand. To cite a most cogent example in each category:
Jacobs et al. (1998) compare the neural organization of arthropods and vertebrates to a platyhelminth outgroup bearing a potentially plesiomorphic design: “The flatworm nervous system is often conceived of as having an anterior nerve ring with four major nerves emanating posteriorly from it” (1998, p. 348). They then point out, citing Bier (1997), that “the default condition of the ectoderm is neurectoderm” (p. 349), and that the development of non-neural ectoderm therefore requires additional and apomorphic down-regulation, now largely accomplished, in vertebrates and arthropods, by the chd/sog and dpp/BMP-4 systems described above. Jacobs et al. (1998) therefore interpret the relatively inverted systems of neural development in arthropods and vertebrates as two different specializations from the plesiomorphic (flatworm) condition of four major nerves extending posteriorly in radial symmetry around the anterior ring. Thus, arthropods retain the two ventral cords (and suppress dorsal neurectoderm by dpp action described above), whereas vertebrates keep the plesiomorphic state in a dorsal position and suppress ventral neurectoderm by the action of BMP-4. Jacobs et al. (1998, pp. 349-350) conclude: “The bilaterian central nervous system would be the product of concentrating the nervous organization in part of the ectoderm, by eliminating it from other regions ... If this were the case, then the ventral nervous system in protostomes could derive from the ventral pair of nerves in the orthogon [the fourfold system of flatworms] and the dorsal system in vertebrates from the dorsal pair.”
So far so good, and so reasonable. But Jacobs et al. (1998, p. 350) then make a false inference about Geoffroy's views: “The above scenario explains the available data without invoking an instantaneous dorsoventral inversion as envisioned in the transcendental scheme of Geoffroy.” But Jacobs et al., while correctly criticizing Patten's theory of flipover in direct phyletic transition, misattribute this view to Geoffroy. In fact, the scenario of Jacobs et al. fits splendidly with Geoffroy's actual hypothesis of separate and different transformation, constrained by structural rules of growth, from a common archetype — in this particular case, suppression of either the two dorsal, or the two ventral, nerve cords in an originally radially symmetrical circlet of four. (Gerhart, 2000, questions the inversion hypothesis with an alternative strikingly similar to the proposal of Jacobs et al., but equally, and truly, consonant with Geoffroy's actual claim.)
In a different potential criticism, Bang et al. (2000) accept the description [Page 1122] of homologous systems of neural generators and suppressors operating at opposite poles of the dorsoventral axis in arthropods and vertebrates. But they question the status of this system as an ancient and conserved primary marker and definition of the body axis throughout the history of bilaterian animals, from the time of the ancestral “urbilaterian” (De Robertis and Sasai, 1996) through the differentiation of arthropods, vertebrates, and all other phyla deriving from this common node.
Perhaps, they argue, the dpp/sog and BMP-4/chd interaction expresses a much more general (
and perhaps more ancient) signaling pathway “that has been conserved in evolution but coopted for patterning very different aspects of the body” (Bang et al., 2000, p. 23). In potential support, they note (see Yu et al., 1996) that, in Drosophila, “dpp is expressed in vein precursor cells in the pupa, whereas sog is expressed in the intervein-cells and suppresses the formation of veins.” The separate cooption, in arthropods and vertebrates, but in reversed orientation, of such a general signaling pathway would represent a parallelism based on so broad and abstract a homology of underlying genetic routes of development that an evolutionary interpretation in terms of constraint would become uninformative because the “hold of history” would then become so loose and unspecific. (I shall devote the entire second part of the next section — pp. 1134–1142 — to this central issue, by elucidating the contrast between the genuine but uninteresting homology of Pharaonic bricks and the important historical mark and constraint of Corinthian columns. I will therefore let this example stand as a prelude to this forthcoming discussion, while also adding an incisive comment from Wray and Lowe (2000, p. 48): “The existence of developmental modules that are reapplied in functionally similar contexts in nonhomologous structures poses a very real problem for testing hypotheses of homology among morphological structures.”)
For now — and so much more shall be discovered in the first years of our new millennium — we may recapitulate the stunning novelty of this first theme by contrasting Mayr's conventional 1963 statement that genetic homology between phyla may be dismissed a priori and in principle, based on our general understanding of the power of natural selection, with a 1996 statement by Kimmel (p. 329), not at all intended as a “gotcha” or an ironic commentary on Mayr's misplaced confidence, but certainly appropriate as an opening sentence for a 1996 article on a new view of life: “We have come to find it more remarkable to learn that a homolog of our favorite regulatory gene in a mouse is not, in fact, present in Drosophila than if it is, given the large degree of evolutionary conservation in developmentally acting genes.”
The Structure of Evolutionary Theory Page 178