Movement two, Elaboration: parallelism of underlying generators:
Deep homology builds positive channels of constraint
PARALLELISM ALL THE WAY DOWN: SHINING A LIGHT AND FEEDING THE WALK. The deep homologies discussed in the last section operate as shared starting points and subsequent conduits for historically constrained change. But, in an even more positive role for the historical shaping of evolution from within, homologous developmental pathways can also be employed [Page 1123] (and deployed) as active facilitators of homoplastic adaptations that might otherwise be very difficult, if not impossible, to construct in such a strikingly similar form from such different starting points across such immense phyletic gaps. In short, this fascinating evolutionary phenomenon, long discussed under the rubric of convergence in our literature, now stands ripe for reinterpretation, in several key cases, as the positively constrained outcome of remarkable homologies in underlying pathways of genetic and developmental construction.
This general shift in viewpoint — from a preference for atomistic adaptationism (favoring the explanation of each part as an independent and relatively unconstrained event of crafting by natural selection for current utility) to a recognition that homologous developmental pathways (retained from a deep and different past, whatever the original adaptive context) strongly shape current possibilities “from the inside” — has permeated phylogenetic studies at all levels, from similarities among the most disparate phyla to diversity among species within small monophyletic segments of life's tree. No case has received more attention, generated more surprise, rested upon firmer data, or so altered previous “certainties,” than the discovery of an important and clearly homologous developmental pathway underlying the ubiquitous and venerable paradigm of convergence in our textbooks: the independent evolution of image-forming lens eyes in several phyla, with the stunning anatomical similarities of single-lens eyes in cephalopods and vertebrates as the most salient illustration. As Tomarev et al. (1997, p. 2421) write: “The complex eyes of cephalopod molluscs and vertebrates have been considered a classical example of convergent evolution.” (Assertions of anatomical convergence remain valid in the restricted domain of final products, whereas a phenomenon of opposite theoretical import now holds sway for the pathway of construction itself.)
PARALLELISM IN THE LARGE: PAX-6 AND THE HOMOLOGY OF DEVELOPMENTAL PATHWAYS IN HOMOPLASTIC EYES OF SEVERAL PHYLA
Data and discovery. Salvini-Plawen and Mayr (1977), in a classical article nearly always cited in this context, argued that photoreceptors of some form have evolved independently some 40 to 60 times among animals, with six phyla developing complex image-forming eyes, ranging from cubomedusoids among the Cnidaria, through annelids, onychophores, arthropods and mollusks to vertebrates along the conventional chain of life. In the early 1990's, using Drosophila probes, researchers cloned a family of mammalian Pax genes, most notably Pax-6, which includes both a paired box and a homeobox (Walther and Gruss, 1991). Soon thereafter, several recognized mutations in the form and function of eyes were traced to alterations in Pax-6. For example, mice heterozygous for Small eye (Sey) have reduced eyes, whereas lethal homozygotes, before their death, develop neither eyes nor nose. Similarly, human Aniridia (An) causes reduced eyes, sometimes lacking the iris, in heterozygote form, while the lethal homozygotes also develop no eyes at all. Further studies then demonstrated expression of Pax-6 in the spinal [Page 1124] cord, several parts of the brain, and especially in the morphogenesis of vertebrate eyes, “first in the optic sulcus, then in the optic vesicle, the pigmented and the neural retina, the iris, in the lens and finally in the cornea” (Gehring, 1996, p. 12).
Although the existence of a Drosophila homolog could certainly have been anticipated — the Pax genes, after all, were found with fly probes — few researchers expected that a Drosophila version would also function in the same basic way. But the Drosophila Pax-6 homolog mapped to the eyeless (ey) locus (Quiring, et al., 1994), named for a mutation discovered early in the 20th century, and producing, in homozygous state, flies with strongly reduced eyes, or lacking eyes entirely. Moreover, the conservation between mammalian and insect Pax-6 sequences is impressively high, with 94 percent amino acid identity in the paired box and 90 percent in the homeobox (Gehring, 1996).
The similar function of these Pax-6 homologs in different phyla was then dramatically affirmed by expressing the mouse gene in Drosophila (Haider et al., 1995), and finding that the mammalian version could still induce the formation of normal fly eyes. Noting that Pax-6 acts as an upstream regulator of a large set of more specific genes, Gehring (1996, p. 14) makes the obvious, but important, point: “Of course, the eyes that are induced by the mouse gene are Drosophila compound eyes, since the mouse gene is only the switch gene and another 2500 genes from Drosophila are required to assemble an eye.”
In the boldest of all experiments, leading to results that attracted substantial and well-deserved public attention, Gehring and colleagues then found that ectopic expression of either the murine or Drosophila version of Pax-6 could induce supernumerary eyes on the antennae, legs and wings of flies (Fig. 10-21), thus supporting Gehring's designation of Pax-6 as a “master control gene” for the development of eyes. Gehring (1996, p. 13) wrote that these “ectopic eyes are morphologically normal with normal photoreceptors,
10-21. From Gehring, 1966. One of the most remarkable discoveries from the early days of evo-devo. An ectopic eye (smaller and to the left of the normal eye in A; enlarged in B) can be induced in Drosophila by targeted expression of the mouse homolog of Pax-6 in flies.
[Page 1125]
lens, cone and pigment cells and an electroretinogram as it is typical for photoreceptor cells can be recorded, when the ectopic eyes are exposed to light.” Nonetheless, as these ectopic eyes are not neurologically “wired up,” the fly presumably cannot use them for vision. (To give some sense of the excitement and weirdness of these results upon their initial discovery — for we have, a mere five years later, already become accustomed to such findings — I include as Fig. 10-22 the “Post-it” note that Gehring penciled when he sent me his first reprint announcing this achievement.)
But this conserved developmental pathway for insect and vertebrate eyes, however surprising in the light of previous assumptions about the impossibility of such genetic homology between phyla, did not yet directly address the theoretical issue of convergence in evolution. After all, the single-lens eye of vertebrates bears little anatomical similarity to the multifaceted fly eye, and no claims for convergence had been staked upon this case. But the discovery
10-22. A personal touch expressing the excitement of this discovery: when Gehring sent me his 1996 reprint on inducing ectopic fly eyes with mouse genes, he inserted this Post-it just above his finding that the ectopic eyes are morphologically normal and react to light in the normal way. I added the marginal notation — not to pour water on a great discovery — that these eyes are not wired to the brain and therefore will not function.
[Page 1126]
of a homologous developmental pathway for the disparate eyes of two such different phyla raised an obvious question about the generality of Pax-6 as a “master control gene” (using Gehring's phrase again) for all complex eyes, including the “paradigm for convergent evolution” (Gehring, 1996, p. 14): the remarkable similarities in function and structure between the single-lens eyes of vertebrates and cephalopods.
This case has persisted as a classic, ever since the formulation of convergence as a concept, because the two eyes look so much alike, and work so similarly, despite their separate origins from different tissues: the vertebrate eye as an evagination of the brain, and the cephalopod eye by invagination of the epidermis. The squid eye forms from a monolayer of epidermis that becomes thickened, multilayered and internalized on the dorsal side of the head lobe. The outer ectodermal layer forms the iris and the outer lens portion, while the inner half of the lens arises from the inner ectoderm
al layer. Thus, the adult lens contains two parts, divided by a septum. Meanwhile, the cornea, also of ectodermal origin, derives from a quite different source on the edge of the arms, as they grow forward. In vertebrates, by contrast, the optic vesicle arises as an evagination of the diencephalon, whereas the lens then develops from overlying ectoderm. As the most interesting consequence of these differences — well known, perhaps, because the vertebrate eye seems more “jury-rigged” than the eye of the conventionally “inferior” squid on this basis — the polarity of photoreceptors becomes inverted in vertebrates, but remains everted (an apparently superior design) in cephalopods.
In a keenly anticipated result, Tomarev et al. (1997) found a homolog of vertebrate and arthropod Pax-6 in the squid Loligo opalescens. This gene is expressed in the development of the embryonic eyes, olfactory organs, brain and arms. (This common expression in visual and olfactory systems bears further study, especially given the common ectodermal origin of both organs in vertebrates and their embryonic interaction with adjacent regions of the
10-23. The squid version of Pax-6 also induces the development of ectopic eyes in Drosophila. From Tomarev et al., 1997.
[Page 1127]
developing brain.) In the most satisfying result (see Fig. 10-23), ectopic expression of squid Pax-6 also induced supernumerary eyes in Drosophila. Tomarev et al. write (1997, p. 2424): “Squid Pax-6 is able to induce ectopic Drosophila eyes on wings, antennae, and legs, as was previously demonstrated for Drosophila eyeless and mouse Pax-6. All Drosophila eye-specific structures including cornea, pigment cells, cone cells and photoreceptors with rhabdomeres were formed in the ectopic eyes induced by squid Pax-6 DNA.”
Theoretical Issues. Two related questions have dominated the emerging discussion of these Pax-6 homologies: the putative status of this gene as a “master control” for eyes, and the impact of Pax-6 upon the claim for independent evolution of eyes by convergence. Gehring (1996, 1998) bases his terminology of “master control” upon three properties of Pax-6: its status as upstream regulator of a substantial cascade of more specific eye-forming gene products; its interchangeability among phyla, while always acting as a trigger to the downstream production of the “right” eyes for any given animal; and its general ability to trigger the formation of supernumerary eyes in odd places.
We may legitimately quibble, as Jacobs et al. (1998) and many others have done, that upstream position in a cascade should not be equated with either causal or temporal primacy, for novel regulatory elements can be introduced by evolution into any position of a developing sequence. Nonetheless, I would not begrudge a researcher the right to bestow an incisive name upon such an important discovery. Pax-6 may be no more important than hundreds of other genes in the sense that usable eyes will not form, absent its normal operation. But the designation, as “master control,” of such early and such general action (including the key property of interchangeability among phyla) does no violence to ordinary linguistic usage.
However, this very generality raises the crucial issue (see forthcoming pages for a fuller discussion) of whether the action of Pax-6 must then be regarded as too broad and too universal to sustain any argument for meaningful constraint upon the evolution of eyes in disparate phyla. After all, if activating Pax-6 represents little more than flicking on a master switch at the power plant (with animal development than analogized to the operation of any electrical device thereafter), then its admittedly necessary action fails to identify any channel of development specific enough to warrant designation as a dedicated impetus for the evolution of one adaptive solution over other attainable possibilities. However, in this case (but not in others, as I shall argue on pp. 1034–1042), the actions of Pax-6 are sufficiently specific and precise to set a definite channel among conceivable alternatives, and not just to open a floodgate through which subsequent cascades might flow in any direction (see discussion of this point and listing of criteria for specificity by Jacobs et al., 1998, p. 334, who conclude that this “documentation of eye homology was quite a coup”).
This acknowledgment of sufficient specificity for Pax-6 then raises a final question about the extent of revision thus required in evolutionary concepts of convergence vs. constraint. Some enthusiasts have claimed that genetic [Page 1128] homology in such a crucial and early-acting developmental pathway requires a wholesale reinterpretation of this classic convergence as a pure case of parallel evolution based upon underlying constraint. For example, Tomarev et al. (1997, p. 2426) end their important paper on squid Pax-6 by stating: “Our data support the idea that morphologically distinct eyes of different species have arisen through elaboration of a common conserved Pax-6-dependent mechanism that is operative at early stages of eye development and that the anatomical differences among eyes arose later in evolution. Consequently, we believe that eyes in cephalopods and vertebrates have a common evolutionary origin and are products of parallel rather than convergent evolution.”
This question would be unresolvable, and would become a source of endless terminological wrangling if a single and exclusive answer — either independent adaptations by convergence or similar solutions by constraints of parallelism — had to emerge as the explanation for a unitary phenomenon. (Several participants in the developing debate have operated upon just such a contentious assumption, hence the need for explicit treatment of this eminently resolvable question.) But the issue of how evolution can generate such similar and highly complex eyes in disparate phyla requires an invocation of both phenomena at different levels of analysis. The conventional view of convergence cannot be denied for the final products of adult anatomy, as documented in my previous discussion of the fascinating differences in form and developmental origin for the strikingly similar eyes of cephalopods and vertebrates. But the traditional claim for exclusive convergence at all scales implies a purely functional explanation, positing an independent evolution of eyes along entirely separate and internally unconstrained sequences of natural selection, with no aid from any common starting point or channel of development.
However, the Pax-6 story has now furnished an important homological basis in underlying developmental pathways for generating complex eyes in cephalopods and vertebrates. Thus, a channel of inherited internal constraint has strongly facilitated the resulting, nearly identical solution in two phyla, and evolutionists can no longer argue that such similar eyes originated along entirely separate routes, directed only by natural selection, and without benefit of any common channel of shared developmental architecture. But just as the advocates of pure convergence erred in claiming exclusive rights of explanation, the discovery of Pax-6 homologies does not permit a complete flip to exclusive explanation by constraint.
As so often happens in our world of biological hierarchy, convergence prevails at one level, and constraint at another. The similarities in adult anatomy are primarily convergent, but Pax-6 establishes an important homology in underlying pathways of generation. We thus encounter a case of homoplasy in final results based upon significant homology in underlying developmental architecture. As discussed extensively on pages 1061–1089, and as presented in tabular form on page 1078, this common circumstance, however muddled by a century of confusion in our literature, nevertheless enjoys a clear and simple solution in proper formulation of the concept of parallelism, or homoplasy [Page 1129] of results based on homology of underlying generators. This recasting of the paradigm case for pure convergence as an outcome of substantial parallelism in a key developmental channel has now highlighted the neglected role of constraint as a strongly positive force in organismic adaptation. For we must now grant strong probability to the proposition that, absent an “internal” direction supplied by the preexisting Pax-6 developmental channel, natural selection could not have crafted such exquisitely similar, and beautifully adapted, final products from scratch, and purely “from the outside.”
Moreover, two studies published after my initial composi
tion of this section strongly reinforce the increasing emphasis on constraint and parallelism, rather than independent adaptation and convergence, in the evolution of complex eyes in widely separated phyla of animals. First, Pineda et al. (2000) report homologs of both Pax-6 and sine oculis in the planarian Girardia tigrina. These genes operate in the same cascade, with Pax-6 directly regulating sine oculis, as in phyla with complex lens eyes. But the much simpler visual system of Girardia includes no lens. Pineda et al. (2000, p. 4525) write: “The eye spots of planarians are one of the most ancestral and simple types of visual systems, close to the prototypic eye proposed by Charles Darwin. The planarian eye spots consist of two cell types: a bipolar nerve cell with a rhabdomere as a photoreceptive structure and a cup-shaped structure composed of pigment cells.”
Thus, the basic genetic cascade had already originated, and already regulated visual systems, before the evolution of complex lens eyes, indicating the preexistence of the developmental pathway as a positive constraint of parallelism. Pineda et al. show that repression of the sine oculis homolog completely suppresses the development of eyes in regenerating planarians, thus demonstrating commonality of function as well as structure in the developmental genetics of some of the simplest and most complex eyes among disparate animal phyla.
Second, and from the other end of the logic of the general argument, further aspects of underlying developmental homology have been found in the general construction of anatomically divergent lens eyes of arthropods and vertebrates — so the evidentiary basis of parallelism now extends well beyond the Pax-6 system itself. Neumann and Nusslein-Volhard (2000) show that the retinas of both Drosophila and zebra fish are patterned by a morphogenetic wave of strikingly similar form and timing — driven by Hedgehog in Drosophila and by its homolog Sonic Hedgehog in zebra fish — both inducing a cascade of neurogenesis across the retina. The strikingly unexpected finding of this additional homology in patterning for such anatomically different products led the authors to conclude (2000, p. 2139):
The Structure of Evolutionary Theory Page 179