The vertebrate Hox genes also exhibit the crucial colinearity between sequential order on the chromosome and site of action along the body's A-P axis. Moreover, and most impressively, several early studies confirmed that the familiar arthropod rules for loss-of-function (anterior structures move back) and gain-of-function (posterior structures more forward) generally apply to vertebrate development as well (although unique and non-homeotic effects have also been demonstrated, as in Pollock et al., 1995). For example, in loss-of-function experiments, Le Mouellic et al. (1992) deactivated the mouse Hoxc-8 gene (previously, as in this 1992 paper, called Hox-3.1) and noted anteriorization of vertebral form throughout a substantial region of the body axis extending from the 14th to the 21st vertebra (T7 to LI). In the most striking effect, a supernumerary pair of ribs (characteristic of thoracic vertebrae) grew on the first lumbar vertebra. In general, “vertebrae and ribs displayed more or less pronounced transformations, turning them into structures resembling those characteristic of the adjacent anterior segment” (1992, p. 251).
Rancourt et al. (1995) also observed anteriorization towards the adjacent segment in mice with disrupted expression of Hoxb-5 and Hoxb-6. The first thoracic segments often lost their rib heads and grew altered lateral processes “making them indistinguishable from C7” (1995, p. 112). Since, with the rarest exceptions of 6 to 9 in sloths and 6 in manatees, all mammals possess 7 cervical vertebrae (yes, including giraffes, who grow very long cervicals but don't augment their number!), this homeotic transformation of the first thoracic to the form of a supernumerary (or eighth) cervical seems as curiously in violation of basic taxonomic signatures as the more famous four-winged and eight-legged Drosophila.
In an interesting temporal analog, illustrating the common coincidence of spatial and temporal ordering in the expression of Hox sequences, Dolle et al. (1993) disrupted the most 5' (and therefore last acting) Hoxd-13 gene in mice, and noted a variety of effects upon the limbs, all interpretable as neotenic changes expressing developmental delays evoked by deactivating the last stages of a normal temporal sequence in ontogeny. (I particularly appreciate Dolle et al.'s conscious linkage of these genetic results to the classical data on heterochrony (see Gould, 1977b) as a morphological approach to questions about the regulation of development.) Dolle et al. (1993, p. 438) note an interesting relationship between these genetic results and common pathways of evolutionary change in heterochronic phenotypes, thus invoking this chapter's central theme of positive constraints based on internal channels:
In such evolutionary modifications, the first skeletal elements to be lost are usually those that are formed last during the establishment of the chondrogenic pattern. In Hoxd-13 mutants, the missing skeletal elements are precisely those that appear last during the development of the [Page 1104] autopods. There is therefore a correlation between the extreme 5' location of the Hoxd-13 gene within its complex, its last position in the temporal sequence of activation and its involvement in the patterning of the last-appearing structures. The Hoxd-13 phenotype may thus be considered as resulting from a block in a developmental sequence. This arrest occurs at the end of the process and corresponds to the time at which this gene is supposed to become active. Consequently, only those structures appearing at the end of the process, or parts of those structures still developing at this stage, will be altered.
In a corresponding manner, gain-of-function mutations often yield the expected effects of posteriorization. Kessel et al. (1990) induced overexpression of the mouse Hoxa-7 gene (previously called Hox-1.1) by inserting a promoter sequence of chicken DNA. Two results indicate a forward movement of posterior structures: (i) the first two vertebrae, the atlas and axis, became simplified, assuming a “structure characteristic of more posterior vertebrae” (1990, p. 302); (ii) the last cervical vertebra of one animal developed a pair of ribs and assumed the form of the next posterior series of thoracic vertebrae.
Kessel and Gruss (1991) then induced overexpression by application of retinoic acid. “Posterior transformations occurred along the complete body axis after RA administration on day 7 of gestation and were accompanied by anterior shifts of Hox gene expression domains in embryos” (1991, p. 89). In a particularly interesting result, Lufkin et al. (1992) ectopically expressed Hoxd-4 (previously Hox-4.2) in regions of the developing head anterior to its usual boundary of expression in somites of the cervical vertebrae. “This ectopic expression results in a homeotic transformation of the occipital bones towards a more posterior phenotype into structures that resemble cervical vertebrae” (p. 835). Phyletic inference is treacherous, and absurd claims have been made in misanalogies between phyletic history and developmental anomaly. But a transformation of skull bones towards the identity of vertebrae does induce thoughts of a presumably more homonomous ancestral vertebrate.
Interestingly, the A-P axis of the vertebrate limb also seems to follow the same rules of colinearity. Morgan and Tabin (1994) demonstrated the importance of the Hoxd series in differentiation of the chick limb bud. They observed expression of successive 5' genes in progressively more posterior regions. Overexpression of Hoxd-11 in regions anterior to its normal domain led to the growth of an additional phalanx in digit 1 (which normally has one, while subsequent digits have 2, 3, and 4 respectively, excluding the terminal claw) — “leading to a morphology similar to that of digit 2” (p. 183), a posteriorization anticipated in gain-of-function regimes. Ectopic expression of Hoxd-11 in anterior regions of the chick wing that normally grow no skeletal elements at all induced the growth of a supernumerary digit (resembling digit 2 in morphology) at the wing's anterior edge.
Tickle (1992) noted the similarity of Hoxd expression in the chick wing to [Page 1105] Lewis's gradient model for establishing domains of differentiation. Of the genes at the 5' end of the complex, she wrote (1992, p. 188): “Cells in the posterior part of the bud that will give rise to posterior structures such as a 'little finger' express all the genes, whereas anterior cells that will give rise to the anterior 'thumb' express only Hox-4.4” (Hoxd-9 in modern terminology). These rules, apparently pervasive (at least in bilaterally symmetrical Bilateria with an A-P axis), also explain several well-known empirical regularities in the classical literature on experimental embryology. Citing the correlation of spatial order and temporal sequence, Tickle (1992, p. 188) notes: “Because activation can proceed in only one direction along the complex, this explained why manipulations can convert anterior structures into posterior ones, but never posterior into anterior.”
The high degree of sequence similarity often found between homologous arthropod and vertebrate Hox genes (amounting to near identity of homeodomains in some cases) leads to the remarkable, but (by now) scarcely surprising, interphylum substitutability revealed by so many experiments (and further discussed as evidence for parallelism in the evolution of eyes on pages 1123–1132). Fly Hox genes, expressed in vertebrates, usually broker the same developmental sequences as their vertebrate homologs — and vice versa. Needless to say, such experiments yield the “correct” morphologies for each phylum, thus reinforcing the well-established conclusion that Hox genes specify proper positions and regulate downstream cascades, but do not build anatomical structures themselves. If Hox genes worked as architects as well as specifiers, then the frights of Hollywood horror movies might become realities, and the fly with a human head might really scream, “please help me” from the despair of his spider-web prison.
As one example among so many, the Drosophila Hox gene Antennapedia promotes leg identity, presumably by repressing previously unknown antennal genes. Casares and Mann (1998) have now identified two antennal determiners, including homothorax (hth). As one line of evidence, they cloned Meisl, the mouse homolog of Drosophila hth, and expressed it ectopically in the fly's anal primodium, which normally develops without expressing any Hox genes. The anal plates of these flies then grew as antennae. (Most Hox genes suppress antennae, so ectopic expression of
Meisl in Hox domains does not generate antennae in odd places, but induces other malformations, including markedly truncated legs on the thoracic segments.)
As a person with literary pretensions, I am always fascinated by the sure signal of scientific progress conveyed by the evolution of a rationalized and simplified terminology. The original Hox terminologies were eclectic and specific. Students of Drosophila first identified two clusters of homeotic genes, but could not recognize them as separated parts of a single ancestral sequence. So they awarded different names: Antennapedia complex (ANT-C) for genes regulating anterior structures, and Bithorax complex (BX-C) for genes operating in the fly's rear half. When homologs of both were detected as a single sequence in beetles, terminology began to coalesce, and the entire series [Page 1106] assumed the name of HOM-C. When researchers discovered vertebrate homologs, they did not want to use the same names at first, for they had not yet affirmed the corresponding similarities of colinearity and action (and were probably still reeling from the basic shock of the discovery itself). So the vertebrate homologs became Hox genes. This potentially anarchic situation deteriorated further when, after finding four Hox complexes in vertebrates, researchers started naming the genes in each complex by their order of discovery, and not by their invariant spatial positions along the chromosome. (Perhaps they did not yet believe that colinearity could prevail here as well.) Thus, Hox-1.1 denoted the first discovered, not the most 3', gene of the first Hox series.
Happily, these discrepancies and illogicalities have now been sorted out and — like the standardization of railroad gauges, or the choice of an internal combustion engine for all cars (thus abandoning a host of other early and workable devices) — a common and integrated terminology has developed, not by the official fiat of any particular meeting or official commission, but by obvious advantages in daily use. The four vertebrate complexes have been renamed Hoxa to Hoxd and the genes within each have been numbered from 1 to 13 in their proper A-P, or 3' to 5', order. Meanwhile, acknowledging the proven homologies of gene structure, position and action, the fly folks have dropped their different name for the complex, and now also denote their sequence as Hox, rather than HOM-C. This congelation of a simple and unified taxonomy, replacing the previous promiscuity of different and uncoordinated names for each gene, marks the coherence and maturation of an important field from an initiating chaos of uncoordinated empirical promise.
SEGMENTAL HOMOLOGIES OF ARTHROPODS AND VERTEBRATES: Geoffroy's vindication. The discovery of these deep homologies in genetic structure and action among phyla (particularly between vertebrates and arthropods) brings us back to Geoffroy's daring theory of the vertebral archetype. Researchers have documented homology in key regulatory genes of development, and have also shown the conservation of basic developmental patterns between the two phyla, particularly in differentiation of structures along the A-P axis under the influence of homologous Hox genes and their principles of colinearity. But Geoffroy's formalist theory rests upon an additional and crucial premise — one that continued to strike most researchers as unlikely, even after the first discovery of these broad commonalities in development. For Geoffroy postulated that the segment (the vertebra in Geoffroy's terminology) represents a fundamental — and truly homological — unit of construction in both phyla. Therefore, to validate the basic premise of Geoffroy's theory, the vertebrate somite must also be homologous with the insect metamere (similar patterns of differentiation along the A-P axis cannot suffice), and such a close comparison seemed exceedingly unlikely, if not anathematic, to most biologists. In the classic pre evo-devo book on the origin of the coelom and segmentation, Clark (1964) described the independent origin of arthropod and vertebrate segments as “universally accepted.” And [Page 1107] Moore and Willmer (1997, p. 34) although writing after most of the genetic discoveries discussed in this section, affirmed the independent evolution of segmentation as virtually beyond dispute, and therefore an exemplar and “type case” for good pedagogy in phyletic inference: “As an object lesson to begin with, it is evident . . . that the character we score as 'segmentation' has to have arisen at least twice, since it occurs in the protostome annelid/arthropod grouping and again in the very distant deuterostome chordates, but not in any of their possible common ancestors.” (Their confidence, presumably, would only be increased by the subsequent discovery of a fundamental split among the protostome phyla, with arthropods on one branch and annelids on the other — thus implying a third independent origin of segmentation.)
But now, at a dawning millennium in human calendrics, two sequential sets of discoveries have provoked a rethinking even of this most “settled” issue, and some genuine segmental homology between arthropods and vertebrates now seems almost inescapable. No simple one-for-one correspondence of somite with metamere can be specified down the A-P axes of these phyla, and no archetypal form like Geoffroy's “vertebra” can be reconstructed as an ancestral prototype for all segments. Moreover, vertebrate somites do not seem to be constructed by the arthropod cascade of gap, pair rule, segment polarity genes, etc. — see p. 1110 for more detail on these differences. But anatomical homologies between these two-segmented phyla on maximally divergent boughs of the bilaterian tree extend well beyond mere positioning and pattern of A-P differentiation, and also include important aspects of segmentation as well. If the common ancestor of arthropods and vertebrates did not already possess a segmented body, this “urbilaterian” (in the terminology of De Robertis, 1997) had probably established the fundamental genetic pathways behind segmentation and the differentiation and specialization of segments — a system maintained ever since in both phyla, and based in large part on the Hox sequences and their colinearity.
1. REDISCOVERING THE VERTEBRATE RHOMBOMERES. Initial data on the mode of action of vertebrate Hox genes seemed, at first, to support the traditional conclusion that no segmental homology existed between the two phyla. The primary sites of Hox action generally correlate with the anterior expression boundary of each gene — and these boundaries extended past the developing vertebral column into anterior regions of the embryo. Some enterprising geneticists then rediscovered an important fact, established in the 19th century by the great German school of descriptive anatomists, and then forgotten by several subsequent generations who dismissed such work as the dullest form of cataloguing done at the least causally relevant scale by the most hidebound methodology of holistic observation. For these 19th century anatomists had discovered that the vertebrate hindbrain eventually develops into a unitary structure, but begins as a linear series of 7 or 8 segments called rhombomeres. Moreover, specific rhombomeres seem to control (or at least correlate with) the development of important aspects of anterior anatomy, including the deployment of the cranial nerves. Finally, as the spur [Page 1108] to renewed respect for such “trivial” data of gross anatomy, the anterior expression boundaries of several Hox genes map consistently to specific rhombomeres.*
The striking similarity between the action of vertebrate Hox in rhombomeres and insect Hox in metameres generates strong suspicions of homology. For example, some vertebrate Hox sequences follow the common insect pattern that the anterior expression boundary of each successive 5' gene “skips” a segment, appearing two segments towards the animal's posterior. In mice, Hoxb-2 turns on in the third rhombomere, Hoxb-3 in the fifth, and Hoxb-4 in the seventh. Moreover, cell populations of the rhombomeres seem to follow the same “compartment” rules of insect parasegments — i.e., cells originating before the formation of rhombomere boundaries may place progeny in several rhombomeres, but the clones of all cells formed after the development of a rhombomere boundary do not transgress into adjacent rhombomeres.
These observations may lead a skeptic to admit that some segmental homology exists, but only between the bulk of an arthropod's body and a relatively insignificant portion of a vertebrate's anterior end (and not even to the crucial face or forebrain). At this point, however, a key paleontolog
ical fact should convert skepticism into strong interest. The rhombomeres of the embryonic hindbrain correlate directly with the pharyngeal arches developing just alongside (Fig. 10-16). In fact, each pharyngeal arch corresponds with two rhombomeres (Raff, 1996, p. 343). As we should remember from our elementary courses, all early vertebrate embryos develop pharyngeal arches, or gill slits. Tetrapods lose these structures in later embryology, but their positions determine important aspects of embryological topology (including migratory paths of neural crest cells and the subsequent locations of cranial nerves, as mentioned above), while some of their parts transform into important organs of gnathostome vertebrates. (Most famously, the jaw arises from the first gill arch, while an element of the second arch becomes, in jawed fishes, the hyomandibula (suspending the upper jaw to the braincase) and later, in tetrapods, the stapes, or hearing bone.)
But, more importantly for acknowledging a meaningful segmental homology between arthropods and vertebrates, the rhombomeres and their underlying Hox codes do not only generate some important features of later tetrapod anatomy. [Page 1109] They also constitute, in the earliest agnathan vertebrates, the major functional aspect and structural extent of the organism's segmental anatomy — and not just a small portion of the anterior end. The region of the agnathan gill slits occupied more than half the body's length in many early forms. Moreover, the pharyngeal clefts functioned not only in breathing, but also, as the branchial basket, in gathering and filtering food. In fact, these earliest vertebrates may have fed in the manner of many arthropods, by passing food along a series of segments and their appendages, from posterior to anterior towards the jawless mouth (rather than in the reverse direction that we know so well from our own experience!). For many of the earliest agnathan vertebrates, and without gross exaggeration, one might be tempted to regard the posterior vertebral column (behind the branchial basket) as an add-on and afterthought. In this historical sense, if insect metameres are homologs of rhombomeres in the developing hindbrain of vertebrates, then segmental homology between the two phyla governs the major primordial system of vertebrate segmentation, even if most later gnathostome clades deemphasized this anterior system and strengthened the somites of the subsequent and posterior vertebral column.
The Structure of Evolutionary Theory Page 176