E. B. Lewis used such homeotic mutations to develop his model for the evolution and operation of the bithorax complex in Drosophila, the breakthrough that effectively began the modern study of evo-devo and that won a most deserved Nobel Prize for its pioneer. (The Nobel awards include no category for evolutionary studies. Only twice has a prize been given for work in evolutionary biology, each time by nuancing the definition of medicine to include work with legitimate consequences for health, but scarcely in the mainstream of medical research — first to Lorenz, Tinbergen, and von Frisch, for foundational studies in ethology, and second to my dear colleagues Ed Lewis, Christiane Nusslein-Volhard, and Eric Wieschaus for unlocking the genetic basis of fundamental architectures in animal development.)
In a simple and brilliant model, Lewis (1978) inferred that the bithorax complex evolved by gene duplication, with all members (up to eight) remaining aligned in a tandem array on the third chromosome. Since these BX-C genes regulated developmental positions in the posterior part of the thorax [Page 1097] and throughout the abdomen (see Fig. 10-15), Lewis assumed that the functional basis of duplication lay in the need for more genes to achieve evolutionary differentiation from the ancestral homonomy of repeated and similar (if not identical) appendages on each body segment — in this case, and for dipterans in general, to suppress the development of legs on abdominal segments and to convert the second pair of wings (on the third thoracic segment) into the small pair of balancing halteres.
The model then implied an elegant mechanism for gene functioning in morphological differentiation. Lewis argued that the first gene in the array turned on in the second thoracic segment and in all posterior segments, with each subsequent gene having its anterior boundary of expression one or two segments further back, but then turning on from there to the posterior end of the fly. Clearly, such a system would build a simple and linear gradient with least gene product at the anterior end of expression for the entire array (where only the first gene turns on), and most products at the posterior end of the animal (where all genes are active).
The further beauty of this model then lies in the simple testability of the implied
10-15. E. B. Lewis's original, brilliant, but not entirely correct model for developmental action of genes in the bithorax complex of Drosophila. Lewis assumed that differentiation of complexity from original homonomy, particularly the conversion of the second pair of wings to halteres, and the suppression of legs on the abdominal segments, required a duplication of further genes in the set. He proposed a tandem array of up to eight genes, each turning on in sequence, but with expression beginning in successively more posterior parts of the developing larva — thus establishing a gradient with ever more gene product accumulating towards the rear of the animal. Therefore — and this part of the model remains basically correct — loss-of-function mutations should weaken the gradient and cause anterior structures to develop in a more posterior position; while gain-of-function mutations should intensify the gradient and cause posterior structures to develop in more anterior sites.
[Page 1098]
mechanism for initiating appropriate structures in each segment: the more gene product, the more posterior the appearance (given a linear gradient with greatest concentration at the rear end). Thus, any loss-of-function mutation, leading to a weakening of the gradient, should cause anterior structures to develop in a more posterior position. In a corresponding manner, gain-of-function mutations, or ectopically induced overexpressions, should intensify the gradient and cause posterior structures to grow in more anterior positions. Shifts in both these directions would produce homeotic effects under Bateson's original definition — and the BX-C complex had originally been recognized by a set of arresting homeotic mutations.
Lewis's model neatly explained the most famous and puzzling homeotic transformations, both based on loss-of-function mutations. Bithorax, the celebrated four-winged fly, does not represent an atavistic reversion to the ancestral state, but arose by a weakening of the gradient that caused the third thoracic segment (usually bearing the much reduced second set of wings in the derived form of balancing halteres) to develop instead as a supernumerary second thoracic. Since second thoracics bear ordinary wings, a fly with two-second thoracics will grow two pairs of wings. Similarly, the equally peculiar eight-legged, or bithoraxoid, fly developed by another loss-of-function mutation under the same rules of Lewis's gradient. The gradient became sufficiently weakened in the first abdominal segment to cause this normally legless module to develop instead as a supernumerary third thoracic. Since each thoracic segment bears a pair of legs (giving insects their defining six for the animal's three thoracic segments), a fly with (effectively) four thoracic segments would grow eight legs.
As a virtually definitional consequence of truly great theories developed in a previous terra incognita, several aspects of an original formulation invariably turn out to be wrong, while central concepts persist in greatly improved form. The most interesting development since the classical formulation (Lewis, 1978), has reversed Lewis's argument that the duplications arose to provide positional cues needed to potentiate the evolution of the distinctive insect body plan (in particular, to suppress legs on the abdomen and convert wings to halteres on the last thoracic segment). In formulating his original hypothesis, Lewis (1978) made the conventional assumption of both Darwinian and ordinary vernacular reasoning: that greater specialization of the phenotype would correlate with increase in the number of generating units. But the idea that morphological novelties must “await” the provision of new genetic material by duplication (or some other process) has been disproven by the fascinating discovery — with central implications for my general argument about constraint, to be developed in the concluding fourth “movement” of this “symphony” (pp. 1147–1178) — that all major arthropod Hox genes had already appeared before the separation of arthropod classes and, for that matter, of protostome phyla as well.
Homologs for all 8 insect Hox genes have been found in other arthropod classes, including the maximally homonomous (identically segmented) [Page 1099] Myriapoda and, for that matter, in the equally homonomous sister phylum of onychophorans (Grenier et al., 1997). De Rosa et al. (1999) conclude that the full complement must be even more ancient, as phyletic analysis indicates a minimum of 7 Hox genes for the bilaterian ancestor, and at least 8 for the common ancestor of protostome phyla.
Thus, the differentiation of distinctive bilaterian body plans has occurred not by the duplication or recruitment of additional Hox genes, but by changes in their regulation and their downstream targets. Presumably, Hox genes “read” positional information to set the location of differentiating structures, thereby triggering the cascade of downsteam architects, but not building the varied structures themselves. As Warren et al. (1994, p. 461) write: Hox genes “provided a pre-existing groundplan upon which insect segmental diversity evolved.” Carroll (1995, p. 483) therefore restated the Lewis hypothesis as follows: “What has evolved in the course of insect and fly evolution are not new genes but new regulatory interactions between BX-C proteins and genes involved in limb formation and wing morphogenesis.”
The discovery of the homeobox — a 180 base pair unit coding for a 60 amino acid homeodomain with important regulatory action as a DNA binding protein — as a common constituent of Hox genes (and others as well) opened the floodgates of this amazingly fruitful research in the early 1980's. By probing for homeoboxes, Hox genes could quickly be located and characterized, and (even more crucially for evolutionary analysis) their homology to genes of other organisms (even in other phyla) established. The two homeotic complexes of Drosophila — Antennapedia (ANT-C) and Bithorax (BX-C) — were quickly revealed as separated subunits, controlling the positioning of anterior and posterior structures along the A-P axis respectively, of a single Hox cluster that maintains its integrity in the beetle Tribolium, and in other nondipteran insects. Powers et al. (2000) show
that the mosquito Anopheles gambiae also retains a single and undivided Hox cluster, so the Drosophila subdivision does not characterize Diptera in general.
The established rules of “hoxology”* vindicated the central principles of morphogenesis in Lewis's model, though under an interestingly different genetic regime. (The BX-C component of the Drosophila Hox sequence contains only three genes, and if they arose, one from the other, by tandem duplication, these events probably preceded the separation of protostome and deuterostome phyla.) But Lewis could not have been more prescient in recognizing the essential sequence and form of Hox action, and in specifying the implied consequences and tests. Lewis's principle established the basis for discovering homologous genes (and homologous actions) in distant groups, thus potentiating evo-devo's greatest and most surprising discovery of “deep homology” among animal phyla — the key to the reevaluation of [Page 1100] historical constraint as an essential component of evolutionary theory and pattern.
Manak and Scott's (1994, p. 63) epitome of “hoxology” illustrates the centrality of Lewis's original conceptions in a different guise:
Several rules governing homeotic gene function have been fairly well conserved. (1) Genes are ordered along the chromosome in the same order as their expression and function along the anterior-posterior axis of the animal. (2) More genes are usually expressed in more posterior regions. (3) Loss of gene function leads to loss of structures or to development of anterior structures where more posterior structures should have formed. (4) Activation of genes where they should be off, i.e. gain-of-function mutations, leads to posterior structures developing where more anterior structures would normally be found. To these generalizations we may add some molecular data. (5) Each homeotic gene contains a single homeobox and encodes a sequence-specific DNA-binding protein, which acts as a transcription factor. (6) Most of the homeotic genes are transcribed in the same direction, with the 5' ends of transcription units oriented toward the posterior end of the Hox cluster.
The perfect colinearity of spatial order along the chromosome with the sequence of morphological differentiation along the developing animal's anteroposterior axis summarizes the most stunning conclusion of this research, and also generates most other hoxological regularities. This central property of colinearity supplies a rationale for Lewis's original concept of a gradient generated by tandem duplicates turning on in spatial order along the chromosome. (The spatial sequence usually reflects a temporal order as well, as morphologically anterior and genetically 3' units generally operate first in ontogeny, with differentiation then proceeding temporally towards the posterior. Some models of Hox evolution regard the temporal factor as primary (see Duboule, 1992; Dolle et al., 1993; Deutsch and Le Guyader, 1998), and I shall discuss this issue further in the last part of this section.)
The other morphological rules also follow from this central precept of colinearity (Lewis could not have known about items 5 and 6 in the above list when he devised his model). The rules for loss and gain-of-function mutations express this key property in a particularly convincing manner. I have already discussed the classic cases of four-winged and eight-legged flies as anteriorizations of posterior segments caused by loss-of-function mutations. The ultimate loss, a fly developing with no Hox gene function at all, leads to lethality, with the dead embryo as a grim and fascinating manifestation of expected rules: a misfit bearing antennae on each of its segments (Shubin et al., 1997, p. 644). (The antennae, or most anterior appendages, normally develop with no Hox activity at all.) The famous Antennapedia mutant introduces Hox activity into this anterior region and thus grows a leg in the antennal position. Other gain-of-function mutations also cause posterior structures to move forward, as expected. The first discovered gain mutation in the Hox genes, Contrabithorax (Cbx), causes the second thoracic segment to differentiate [Page 1101] as another third thoracic — and the fly therefore grows two pairs of halteres and no wings (Lewis, 1992, p. 1530)!
Raff (1996, p. 307) has expressed the surprise of colinearity, and its evolutionary implications for constraint, in the opening words of his section on “frozen controls?”:
Constraint in gene organization is a clouded topic at best, but disturbing observations loom up like logging trucks on a foggy mountain road. The Hox genes have presented the most puzzling instance of deeply conserved gene order. In all phyla so far examined (arthropods, nematodes, and vertebrates), the Antennapedia and Bithorax homeotic gene homologues are clustered, they have the same transcriptional orientation and order of activation, and their transcription is colinear with the body axis. The conservation of a set of clustered genes over half a billion years is difficult enough to accept, but colinearity with body axis defies credibility. Yet it's true.
Vertebrate homologs in structure and action. So far, the formalist or archetypal content of this discussion has been largely limited to the Goethian theme of common bases for the generation of differentiated serial homologs in a single organism — in other words, to internal constraints and channels in the evolutionary history of particular forms and lineages. But the more radical archetypal theories — including both of Geoffroy's derided arguments about vertebral foundations and dorsoventral inversions — postulate the maintenance of such constraints in phyla of distant taxonomic separation and immensely long periods of independent evolution. Such theories of constraining homologies among groups focus our attention upon the quite different and larger issue of inhomogeneities in the morphospace of animal designs. Does the markedly nonrandom clumping of organisms within this morphospace record historical constraint (where organisms have been, and where, in consequence, they then cannot go), and not only the power of selection (where organisms do best, with all workable positions accessible)?
The discovery of homeoboxes, and the development of simple probes for their identification, provoked a grand “fishing expedition” (or “gold rush” for a more positive metaphor) throughout the taxonomic pool of organisms. When such procedures become easy, efficient and inexpensive enough, scientists will be tempted to try experiments that would otherwise be deemed foolish.
As an obvious candidate for crazy experiments, especially in the persistently dim light of Geoffroy's archetypal hypothesis for arthropods and vertebrates, a search for vertebrate homologs of arthropod Hox genes could hardly have remained unthought or undone, although I doubt that anyone dared to anticipate success (again, see Mayr's canonical quotation on p. 1066). As we all now know and utilize the stunning successes of these experiments, a reminder of the initial astonishment, and of the tentative nature of first conclusions, dramatically illustrates how far this research has proceeded [Page 1102] in 15 years (and will no doubt extend, thereby rendering these pages obsolete, in just a few additional years).
Not only do Hox genes exist in vertebrates, but also homologs for all Drosophila Hox genes have been found, arranged in the same linear order on chromosomes, and acting with the same colinearity in development along the A-P axis of the vertebrate body. Moreover, vertebrate Hox genes have undergone fourfold replication and exist as four paralogous sequences on four different chromosomes. (The vertebrate sister taxon, amphioxus, has but a single Hox cluster, so we can make good inferences about the timing of amplification in our lineage. The agnathan lamprey probably has only three Hox sequences. Interestingly, and uniquely among deuterostomes, or any other animal, the single Hox cluster of amphioxus has an “extra” or 14th Hox gene at the 5' end — see Ferrier et al., 2000.) The vertebrate Hox genes can be arranged into 13 paralogy groups. (No vertebrate genome includes all 13 genes in any single cluster. The mouse, for example, has 39 of the 52 possible genes — Ferrier et al., 2000. The single sequence of amphioxus, however, does include a copy of each Hox gene. The increase in potential number within each group occurred largely by duplications of the posteriormost (5') homologs of Drosophila Hox genes.)
Lewis (1992, p. 1529) captured the excitement of this work in a single opening
adverb: “Astonishingly, mice and humans not only have cognates of the BX-C and ANT-C genes in a single HOM-C, but the complexes occur in four sets, each in a different chromosome.” Slack et al. (1997, p. 867) echoes a consensus in designating this discovery of deep homology as “the most spectacular achievement of molecular developmental biology.” Yet initial expectations certainly did not forecast emerging realities. In a 1990 review, De Robertis described the decision to undertake an experiment leading to the discovery of the first vertebrate homeobox gene in Xenopus laevis (Carrasco, McGinnis, Gehring, and De Robertis, 1984 — a good Orwellian year). I was a bit saddened (but mostly amused) by the closing observation on the counter-intuitively negative correlation that often emerges (or gets imposed by the realities of laboratory culture) between youth and willingness to think the unthinkable. To any graduate student reading this book, I can only say: Verbum sapientiae . . . “We decided to try what seemed, at the time, a crazy experiment: to isolate a gene similar to Antennapedia from frog DNA with McGinnis and Gehring's fruit fly homeobox probes. There was little reason to believe that the frog DNA contained such a gene or that the genes of such unrelated species would be significantly similar. Still, we felt it was worth the attempt. Some of our colleagues were skeptical that such an experiment could ever work, and two of our students declined to help on those grounds.”
The initial discovery of homology in genetic structure for arthropod and vertebrate Hox did not seal the case for evolutionary meaning, since no one yet knew how vertebrate Hox genes operated. Carrasco et al. (1984, p. 409) wrote of their original discovery: “If the frog gene cloned here eventually turns out to have functions similar to those of the fruit fly genes, it would represent the first development-controlling gene identified in vertebrates.” Evidence [Page 1103] for similarity of action soon followed, thus securing the argument for meaningful morphogenetic conservation across at least 530 million years, and almost maximal bilaterian separation.
The Structure of Evolutionary Theory Page 175