2. MORE EXTENSIVE HOMOLOGIES THROUGHOUT THE DEVELOPING SOMITES. If homologies based on the Hox code place vertebrate rhombomeres into phylogenetic union with arthropod metameres, must we conclude that the far more prominent somites of the gnathostome vertebral column bear no relationship of homology with arthropod segments? Such a conclusion need not follow, for the obvious reason that development and specification of arthropod segments requires the operation of several genetic systems prior to and beyond the activation of Hox genes. The Hox genes, after all, do not regulate the formation, number and timing of segments, and
10-16. Schematic diagram from Raff, 1996, showing that each rhombomere in the developing embryonic hindbrain of vertebrates correlates directly with the pharyngeal arches developing just alongside, with each arch corresponding to two rhombomeres.
[Page 1110]
they do not code for the actual structures built within each segment. The Hox genes turn on after the segments have been generated by other systems. They then act to regulate the appropriate (and different) downstream cascades that actually build the specialized structures of each segment. Thus, we may also search for homologies between vertebrates and arthropods in the prior systems that specify numbers and positions of segments before Hox genes begin their work in regulating specific fates.
Although the long germ-band style of segmentation in Drosophila (all segments forming simultaneously as divisions of an embryo with a fully established A-P axis) represents a highly derived condition with respect to the plesiomorphic state of most insects (short germ-band development, with new segments added in a temporal sequence, one by one at the posterior end), we almost inevitably turn to Drosophila as an arthropod model of segmentation because our knowledge of this fly so exceeds our understanding of any other arthropod's development. The identities and differentiation of Drosophila's segments occur in a programmed cascade of linked and ever-finer specifications that always draws my mind to the basic model of Genesis I (by which I intend no statement about creation, needless to say, but refer only to the geometric style of building complexity by successive division and differentiation out of primal homogeneity, rather than by addition). In this primal tale of Western culture, the cosmos begins “without form and void,” and its products then originate by compartmentalization and increasing specification of units: light from darkness on day one; earthly from heavenly waters on day two; earthly land from earthly water on day three; and division of heavenly light into sun and moon on day four.
Drosophila's first specification even begins in a prior generation, for protein products of maternal genes like bicoid and nanos appear in the egg cytoplasm to designate the anterior and posterior embryonic poles. These maternal genes activate gap genes like hunchback that specify broad regions along the A-P axis. Gap genes then regulate the expression of pair-rule genes, whose bands of activity establish the embryo's parasegment boundaries. These pair-rule genes express themselves in every other segment, but also regulate the next level of differentiation in the genetic cascade: segment-polarity genes like engrailed and wingless. The action of segment-polarity genes finally establishes the anterior and posterior domains of each segment. Now that segment boundaries have been set, and the spatial domains of each segment determined, Hox genes can finally establish segment identities by regulating downstream cascades of appropriate architects.
Interestingly, although evidence remains limited to a few taxa and effects as I write this section in January 2000, some apparent vertebrate homologs of these segmental cascades have been detected with reasonable confidence.
(1) Pair-rule genes and somite formation in zebra fish and chicks. Miiller et al. (1996) studied the expression of herl, a homolog of the Drosophila pair-rule gene hairy, in the zebra fish Danio rerio. Expression of herl occurs in transient stipes within the presomitic mesoderm. Although more than 10 [Page 1111] bands eventually form, no more than three are expressed at any one time, because older (anterior) bands fade as new bands appear at the tail bud (see description in Kimmel, 1996). Since the her1 bands form and fade before the appearance of somites, Miiller et al. (1996) traced the fate of cells from the her1 bands in later embryos. In a particularly gratifying result, cells of the first herl band formed somite 5, while cells of the second band generated somite 7, thus confirming homology of action for pair-rule genes (expression in every other somite) as well as homology of genetic sequence.
Pennisi (1997) then described the work of Pourquie and colleagues on chairy (for chick hairy), the chick homolog of the same Drosophila pair-rule gene hairy. This study added important data on the timing of gene action, again linking the spatial order along a major body axis to a temporal sequence that can easily implicate heterochrony, the classical rubric for elucidating relationships between ontogeny and phylogeny, as a pathway (and preferred channel) for evolutionary change. The early chick embryo grows an elongated region where about 50 somites will originate, one at a time, starting at the anterior end, and taking about 90 minutes for each to form. Pourquie and colleagues found that chairy first becomes active in the rear 70% or so of the entire elongated region. The band of expression then narrows and shifts forward towards the head, finally becoming concentrated in a thin stripe at the rear edge of the next somite to form. After this stripe appears, the gene turns on again over the same broad region, beginning the cycle anew and ending in a sharp stripe at the next posterior segment in the developing array.
(2) A segment polarity gene in amphioxus. Although vertebrate homologs of arthropod segment polarity genes do not seem to function in establishing segmentation (for their expression begins only after somitic boundaries have formed), AmphiEn, the only amphioxus homolog of the Drosophila segment polarity gene engrailed, appears in stripes at the posterior border of the first eight somites to develop (Holland et al., 1997). (Drosophila engrailed appears in a similar position at the anterior borders of developing parasegments, which become the posterior borders of adult segments, since each final segment forms from the junction of the posterior half of one parasegment with the anterior half of the next parasegment in the A-P array.) Holland et al. (1997, p. 1723) draw a strong inference about segmental homology: “The segmental expression of AmphiEn in forming somites suggests that the functions of engrailed homologs in establishing and maintaining a metameric body plan may have arisen only once during animal evolution. If so, the protostomes and deuterostomes probably shared a common segmented ancestor. “
(3) Does resegmentation occur in developing vertebrae, and could such a process be homologous with the conversion of embryonic insect parasegments to adult segments? As De Robertis (1997) reminds us, anatomical data known for more than a century indicate that a subset of cells in each somite (called the sclerotome) forms a vertebra. But each adult vertebra arises by [Page 1112] “resegmentation” as the posterior half of one sclerotome fuses to the anterior half of the next sclerotome along the A-P axis. “The end result is a phase shift of the vertebra with respect to the muscle, so that the segmental muscles can span, and move, adjoining vertebrae” (De Robertis, 1997).
These anatomical data, never satisfactorily verified, have now been confirmed by cell lineage studies in birds. De Robertis argues that such vertebral resegmentation may be homologous, and not merely analogous, with the similar construction of insect segments from conjoined halves of adjacent parasegments. De Robertis concludes (1997): “It seems improbable that such a complicated way of making individual metameres would have arisen independently twice in evolution.”
3. Some caveats and tentative conclusions. I need hardly remind my fellow evolutionary biologists that these results, no matter how fascinating and surprising, show only limited and partial homology, in the strict sense needed to affirm Geoffroy's archetypal notions, between arthropod metameres and vertebrate somites. To cite the two most important caveats: First, even the most impressive finding, the mapping of Hox activity to rhombomeres of the developing vertebra
te hindbrain, does not establish full homology between particular arthropod and vertebrate segments. We may, I think, legitimately speak of homology in the basic function, and in the spatiotemporal operation of the Hox genes themselves, and therefore in the fundamental patterning of the A-P axis. But the segments along this axis have already been established by this point in development, and the action of Hox genes (as discussed previously on p. 1107) does not build the segments, but rather turns on downstream cascades that differentiate the “right” structures in the appropriate places.
At this point, we have no evidence for, and some substantial (albeit negative) evidence against, the building of rhombomeres along genetic pathways homologous with those that determine arthropod segments. No data suggest that gap genes, pair-rule genes, and segment polarity genes — the temporal cascades responsible for the development of arthropod segments — also build vertebrate rhombomeres. Thus, in the overall case for homology between vertebrate and arthropod segments, the rhombomeres can only claim an architectural status as “preformed” compartments in which a homologous set of genes then operates to regulate the further differentiation of appropriate structures within each segment. But we cannot claim homology in the pathways of genetic construction for the compartments themselves.
Second, although some impressive homologies may now be asserted for structures along the main A-P axis of arthropods and vertebrates (despite their major differences in adult appearance and function), two important comparisons in Geoffroy's hypothesis cannot, for different reasons, be defended as support for strongly constraining homology: The relationship of insect appendages with vertebrate limbs, and the interpretation of the vertebrate head as an amalgam of several vertebrae (which might then be viewed as potentially homologous with the arthropod head, construed as a tagma of several segments). [Page 1113]
As for limbs, I will argue in the next section (pp. 1134–1142) that the putative homology of some genetic pathways resides in such generalized rules of morphological organization (for the initiation of any “outpouching” orthogonal to a major axis, for example) that little support for particular historical constraints can be drawn from the claim for genetic retention. (After all, properties so pervasive and general as the structure of DNA, or so broad as the necessary physical geometry of elongation and outpouching, do not manifest the specificity required to identify limitations or channels arising from definite historical positions on life's phyletic tree.)
As for the vertebrate head, current knowledge favors a status even less congenial to claims for homology across phyla, and to strong historical constraint: interpretation of this definitive vertebrate structure as a true novelty and neomorph, and not as a highly modified organ constructed from parts homologous to units of the arthropod Bauplan. The foundation of this argument rests upon distinctive features of the vertebrate neural crest and its astonishing range of developmental derivatives and influences (Gans and Northcutt, 1983). Thus, despite important homologies in products of the developing hindbrain and its rhombomeres, the vertebrate mid and forebrain seems to represent a largely “suradded” structure, unique to the vertebrate (or at least to the chordate) lineage.
I do not challenge this general argument, but some aspects of the vertebrate fore and midbrain may exhibit developmental homology with anterior segmentation in protostome phyla. In particular (see Simeone et al., 1992; Holland et al., 1992; and Raff, 1996, pp. 199-200), the Drosophila gap genes orthodenticle (otd) and empty spiracles (ems) operate in the establishment of head segmentation at the fly's front end, anterior to the domain of expression for Hox genes. Two homologs of each of these homeobox genes (Otx1 and 2, and Emx1 and 2) have now been identified in mice, and their domain of action also maps to the forebrain and midbrain, anterior to the expression of Hox genes in the rhombomeres of the hindbrain (see Fig. 10-17, taken from Holland et al., 1992, p. 627). But we do not yet know if these genes encode common modes of action (in addition to their similarity in genetic structure and locus of operation).
On this chapter's central subject of degrees of constraint, Holland et al. (1992) offer the interesting suggestion that these gap gene homologies might enforce less channeling upon patterns of development than the Hox genes impose, and that the greater independence, flexibility and subsequent novelty and variety of the vertebrate head might flow, in part, from the absence of more constraining Hox action in the mid and forebrain regions. (In particular, as Figure 10-17 illustrates, “the four Otx and Emx genes show a nested series of posterior expression boundaries, in contrast to clearly nested anterior expression boundaries in the Hox genes” (Holland et al., 1992, p. 627.) Moreover, whereas the anatomical expression of Hox genes strictly parallels their spatial order on the chromosome, Otx and Emx show no evidence for similar clustering in the genome). Holland et al. (1992, p. 628) therefore hypothesize: [Page 1114]
If roles for Hox, Otx and Emx genes in body regionalization evolved early in metazoan radiation, the fundamental molecular dichotomy within the vertebrate neural tube is a legacy from events preceding the evolution of the vertebrate head; nonetheless, there are likely to be adaptive consequences. If the different homeobox gene families are under different modes of regulation (for example if the tight clustering of Hox genes restricts mutational change) then subsequent variation and adaptive radiation will have been constrained to different extents anterior and posterior of the midbrain/hindbrain boundary. We suggest this could be a molecular basis for the comparative evolutionary plasticity of the vertebrate forebrain and midbrain, but conservation of hindbrain morphology, during vertebrate evolution.
Lumsden and Krumlauf (1996) discuss another prospect for potential homology in genetic action at the anterior end of arthropods and vertebrates (although such examples, as for the previous case of Otx and Emx, bear limited application to Geoffroy's particular theory about the segmental basis of
10-17. Note, in the rhombomeres of the developing mouse brain (part B of the figure), the nested anterior expression boundaries of Hox genes, as opposed to the posterior nesting of expression boundaries in Otx and Emx. From Holland et al., 1992.
[Page 1115]
anatomical homology, because any similar action occurs within segments at an arthropod's frontal end, but operates within the apparently unsegmented mid and forebrain of developing vertebrates). In the chick midbrain, rostral to the anterior limit of Hox action, a long-range signaling region, located at the isthmic constriction between the posterior end of the midbrain and the rhombomeres behind, regulates AP patterning within the unsegmented field of the developing midbrain. Signals from this isthmus regulate the action of En-1 and En-2, two engrailed genes homologous with the prominent segment-polarity engrailed gene of Drosophila. In chicks, the engrailed gradient (see Fig. 10-18) spreads from the isthmus in both directions, decreasing anteriorly through the mesencephalic vesicle and also posteriorly through the first rhombomere (Lumsden and Krumlauf, 1996, p. 1112). Moreover, these authors add (p. 1112), “En expression is the earliest known marker for mesencephalic polarity.”
Finally, although this argument only applies to the relationship of vertebrates with other chordates, and not to any protostome group, the forebrain, and even the neural crest, may not be so confined to true vertebrates as previous views generally assumed. In overt appearance, the anterior end of amphioxus does not include any organs comparable with the vertebrate mid or forebrain. But Holland and Holland (1998) report that amphioxus homologs of two genes with important action in the vertebrate fore and midbrain also operate in generating the anteriormost cerebral structures of amphioxus. (AmphiOtx, the homolog of the vertebrate Otx that operates in both fore and midbrain, is expressed at the anterior end, and in the ventral and lateral walls, of the cerebral vesicle in amphioxus. AmphiDll, the homolog of vertebrate Dlx that operates in the forebrain, is expressed at the extreme anterior end of the cerebral vesicle, and also in the dorsal wall.)
Holland and Holla
nd conclude (1998, p. 651) “the expression patterns of these amphioxus genes suggest that the cerebral vesicle is largely homologous to the vertebrate forebrain, but cannot rule out a midbrain homo-log.” Moreover, the expression pattern of AmphiDll during neurulation implies “that the epidermal cells bordering the neural plate may represent a
10-18. Possible homology in genetic action at the anterior end of both arthropods and vertebrates. In chicks, a gradient of Engrailed expression spreads from the isthmus of the developing brain in both directions, decreasing anteriorly through the mesencephalic vesicle, and also decreasing posteriorly through the first rhombomere. From Lumsden and Krumlauf, 1996.
[Page 1116]
phylogenetic precursor of the vertebrate neural crest” (p. 648). Nonetheless, emphasizing the novelty of vertebrate usage, whatever the homology of development with amphioxus, “the migrating cells of amphioxus do not differentiate into the wide variety of cell types known to originate from the vertebrate neural crest, but eventually remain part of the epidermis” (Holland and Holland, 1998, p. 654).
The Structure of Evolutionary Theory Page 177