Hox Expression in the American Alligator and Evolution of Archosaurian Axial Patterning

Transcription

1 RESEARCH ARTICLE Hox Expression in the American Alligator and Evolution of Archosaurian Axial Patterning JENNIFER H. MANSFIELD 1 AND ARHAT ABZHANOV 2 1 Department of Biological Sciences, Barnard College, Columbia University, New York, New York 2 Department of Organismic and Evolutionary Biology, Harvard University, Cambridge, Massachusetts ABSTRACT J. Exp. Zool. (Mol. Dev. Evol.) 314B, 2010 The avian body plan has undergone many modifications, most associated with adaptation to flight and bipedal walking. Some of these modifications may be owing to avian-specific changes in the embryonic Hox expression code. Here, we have examined Hox expression in alligator, the closest living relative of birds, and an archosaur with a more conservative body plan. Two differences in Hox expression between chick, alligator, and other tetrapods correlate with aspects of alligator or bird-specific skeletal morphology. First, absence of a thoracic subdomain of Hoxc-8 expression in alligator correlates with morphological adaptations in crocodilian thoracic segments. Second, Hoxa-5, a gene required to pattern the cervical thoracic transition, shows unique patterns of expression in chick, alligator, and mouse, correlating with species-specific morphological patterning of this region. Given that cervical vertebral morphologies evolved independently in the bird and mammalian lineages, the underlying developmental mechanisms, including refinement of Hox expression domains, may be distinct. 314B, & 2010 Wiley-Liss, Inc. How to cite this article: Mansfield JH, Abzhanov A Hox expression in the American alligator and evolution of archosaurian axial patterning. 314B:[page range]. Variation in the axial skeleton, comprising the vertebrae and ribs, supports different modes of locomotion, feeding, and other behaviors in different tetrapod groups. This variation consists of both differences in the morphologies of segments within each region of the vertebral column (cervical, thoracic lumbar, sacral, caudal), and also of differences in the number of segments participating in each region, termed the axial formula. Both the axial formula and the morphology of individual vertebral segments are governed, at least in part, by Hox transcription factors. Early in development, these proteins there are nearly 40 in tetrapods are expressed in restricted and overlapping domains along the anterior posterior axis of the body, including in the paraxial mesoderm that gives rise to most of the axial skeleton, musculature, and connective tissue. Mutations in Hox genes lead to homeotic transformations in segmental identities (Krumlauf, 94). It has been proposed that the identity of each segment is conferred by the combination of Hox genes expressed in that segment, also called the Hox code (Kessel and Gruss, 91). Transplantation studies have shown that the identity of somite-derived axial structures is determined by Hox expression established early, in the pre-somitic mesoderm, but it is also clear that Hox proteins function later, within differentiating somites, to regulate the morphology of somitederived tissues. (Kieny et al., 72; Yueh et al., 98; Nowicki and Burke, 2000; Carapuco et al., 2005). Correspondingly, expression of Hox genes is dynamic and varied across and within the different compartments of individual somites. Although direct transcriptional targets of Hox proteins are for the most part unknown, it is evident that during skeletal development they influence morphogenetic processes, including cell proliferation, adhesion, and differentiation (Rijli et al., 93; Yokouchi et al., 95; Goff and Tabin, 97; Kanzler et al., 98; Yueh et al., 98; Stadler et al., 2001; Aubin et al., 2002; Suzuki and Kuroiwa, 2002; Boulet and Capecchi, 2004; Massip et al., 2007). Hox proteins thus play Additional Supporting Information may be found in the online version of this article. Correspondence to: Jennifer H. Mansfield, Department of Biological Sciences, Barnard College, Columbia University, 3009 Broadway, New York, NY jmansfield@barnard.edu Received 28 December 2009; Revised 28 May 2010; Accepted 2 June 2010 Published online in Wiley InterScience ( DOI: /jez.b & 2010 WILEY-LISS, INC.

2 2 MANSFIELD AND ABZHANOV many roles in regulating the morphology of the developing axial skeleton. Changes in the Hox expression code have contributed to the diversification of vertebrate body plans. Conserved combinations of Hox expression mark morphological transitions between vertebral types across vertebrate groups; evolutionary shifts in the location of such transitions are associated with corresponding shifts in the Hox code (Gaunt, 94; Burke et al., 95; Ohya et al., 2005). Birds have undergone extensive modification to their axial skeletons compared with basal archosaurs, a group whose only extant members are birds and crocodilians. For example, birds are distinguished by long, flexible necks, which contain an expanded number of cervical segments. Cervical vertebrae further show modifications, including the loss of ribs (via fusion of cervical rib precursors with the vertebral centra; Goodrich, 30), and the development of heterocoelus centra, both of which contribute to the flexibility of the neck. In contrast, thoracic, lumbar, and caudal regions contain reduced numbers of segments and show extensive vertebral fusions: fusions between thoracic vertebrae form a fulcrum for the wings, fusion of thoracic, lumbar, and sacral segments with the pelvic girdle form the synsacrum, and posterior caudal segments fuse to form the pygostyle (Liem et al., 2001). Although the Hox expression code in chicks has been extensively studied, it is not known whether or how Hox code evolution contributed to these and other derived features of avian axial skeletons. Crocodilians are the closest living relatives of birds; their last common ancestors lived approximately 250 million years ago. The crocodilian body plan is highly conservative and, notably, lacks the avian specializations described above. Axial formulae are nearly invariable among the crocodilians and are, in fact, quite similar to those in the basal archosaurs and even basal amniotes (Hoffstetter and Gasc, 69; Cruickshank, 72; Gow, 75; Gauthier et al., 88; Berman et al., 2004; and see discussion). Thus, they would not be expected to share any Hox expression changes associated with bird-specific specializations in the axial skeleton. Here, we have cloned and examined the expression of Hox mrnas in embryos of the American alligator, Alligator mississippiensis. We focused on Hox group 4 8 paralogs, which collectively participate in patterning cervical and thoracic segments. Not surprisingly, we find that expression boundaries are generally similar in alligator and chick embryos. However, several differences were observed, in anterior posterior boundaries or within somitic subdomains, two of which were correlated with alligator or bird-specific skeletal morphology. A loss of one subdomain of Hoxc-8 in alligator thoracic somites correlates with larger neural spines (and epaxial musculature), which Hoxc-8 patterns in mice. In addition, a key regulator of cervical segmental identity in mice, Hoxa-5, shows unique patterns of expression in birds, mammals, and alligator that correlate with different cervical morphologies. This suggests a potentially important role for Hoxa-5 in the independent evolution of a flexible and ribless neck in both mammals and birds. RESULTS Axial Skeleton of Alligator mississippiensis The axial formula of crocodilians, including A. mississippiensis, has been previously described (Reese, 15; Mook, 21), and shown in Figure 1A, B. In amniotes, the five anterior-most somites contribute to the occipital region of the skull (Burke et al., 95). Posterior to the skull, alligators have 9 cervical vertebrae, 10 thoracic, 5 lumbar, 2 sacral, and a variable number of caudal segments nearing 40. Cervical vertebrae bear short ribs that increase in size near the cervical thoracic transition, such that the transition is graded rather than sharp, as it is in birds (Reese, 15; Fig. 1B, asterisk). The first eight thoracic ribs articulate with the sternum and the posterior two are free. Lumbar vertebrae are morphologically similar to posterior thoracic segments, but with wide, flat transverse processes; however, short ribs occasionally develop on L1 L2 as well (Fig. 1B). Overall, the alligator axial skeleton is marked by the more gradual morphological transitions between regions compared with birds, and by greater similarities in the morphology of vertebrae across different regions, especially in the cervical, thoracic, and lumbar segments. Identification of Alligator Hox Sequences Partial cdna sequences for alligator Hox genes were obtained by RT-PCR. Briefly, degenerate primers were used to amplify a region of variable sequence between that encoding the conserved YPWM motif (involved in Pbx cofactor binding) and the WFQNRR motif in the third helix of the homeodomain, following a previously described strategy (Abzhanov and Kaufman, 2000). For some genes, additional 5 0 sequence between the start codon and YPWM motif was cloned. In total, sequences corresponding to 10 out of the predicted 15 Hox group 4 8 orthologs were isolated. Each could be unambiguously identified by comparison of predicted amino acid sequences with Hox sequences from chick, mouse, and human (Fig. 2). More extensive alignments with orthologs from additional vertebrates are shown in Supplementary Figure 1. These reveal variation within and across vertebrate groups, including representative mammals, archosaurs, amphibians, and sarcopterygian and teleost fish, which is also consistent with known phylogenetic relationships. It is notable that some proteins, such as Hoxc-8, show very high sequence conservation while others, such as Hoxb-4, show much more variation across groups. Such differences might reflect varied levels of functional redundancy among Hox proteins (which could itself vary across lineages), permitting either lineage-specific functional specializations or simply an accumulation of neutral or deleterious changes. Nevertheless, given high conservation in biochemical activity (sometimes from insects to mammals; reviewed in Svingen and Tonissen, 2006) and that most

3 HOX EXPRESSION IN ARCHOSAURIAN AXIAL PATTERNING 3 Figure 1. Alcian blue staining shows cartilage morphology at embryonic day 14 (E14, A)andE28(B)inA. mississippiensis embryos. E14 (A), is the stage at which Hox expression was examined. Alcian blue positive condensations are apparent in the vertebral bodies and neural arches of axial segments, as well as in the skull, pharyngeal arch, limbs, and heart valves. Vertebral identities are indicated. At E28 (B) development of vertebral segments is more advanced. Morphological transitions between vertebral types are indicated. Note that short cervical ribs are present on anterior cervical segments. These grade into longer cervical ribs near the cervical thoracic transition. The most posterior cervical (C9) rib is marked with a yellow asterisk. The first eight thoracic ribs articulate with the sternum and the last two are free. Short rib projections are also present on the first two lumbar segments in this embryo. Diffuse staining is also present in the location of the gastralia, connecting the sternum and pubis. functions of Hox proteins are likely to be conserved at least across vertebrates, gene expression changes might be hypothesized to account for most Hox-dependent changes in axial patterning. Alligator Hox Expression Code Expression of alligator Hox genes was examined between embryonic days (E10 E14). These stages correspond to Hamburger Hamilton (HH) stages in chicks, a time when somitic Hox expression is robust and anterior posterior boundaries are stable. We focused on determining expression boundaries in the somitic mesoderm, although in some cases boundaries are described in neural tube and lateral plate (LP) mesoderm. LP mesoderm gives rise to the sternum and gastralia (ventral rib-like structures) (Chevallier, 75; Vickaryous and Hall, 2008), and LP and somitic cells both contribute to a few axial tissues (Kieny et al., 72; Nowicki and Burke, 2000; Nowicki et al., 2003; and see below). Importantly, Hox expression boundaries are different in these tissues and are regulated, at least in part, by different mechanisms. In addition to mapping expression borders along the AP axis, transverse sections were examined to characterize expression within somitic compartments. Chick somites have been extensively fate mapped and the progenitors for many axial structures can be localized based on position and/or gene expression domains (reviewed in Brent and Tabin, 2002; Christ et al., 2004; Scaal and Christ, 2004). A diagram of a somite cross-section is shown in Figure 3M. When possible, we attempted to describe the intrasomitic localization of Hox expression in chicks and alligators, and with respect to the chick fate map. Hox Group 4 Hox group 4 paralogs pattern cervical segments and the cervical thoracic transition. We found that anterior expression limits for Hoxb-4, Hoxc-4, and Hoxd-4 were either identical or similar in alligator compared with chick. Hoxb-4 has an anterior limit at prevertebra (pv)3 (C3; somite (so) 7/8) in both species (Fig. 3A, B; Burke et al., 95). Hoxc-4 is expressed from pv4 (C4; so 8/9) in alligator and pv6 (C6; so 10/11) in chicks (Fig. 3C, D; Burke et al., 95). Hoxd-4 expression is identical to Hoxb-4: in both species, it has an anterior limit at pv3 (C3; somite (so) 7/8) (Fig. 3E, F; Burke et al., 95). All three transcripts are most highly expressed in the anterior part of their domain in both chicks and alligators; this is most pronounced for Hoxb-4 (Fig. 3A, B; between arrowheads). Hox group 4 expression was also observed in transverse sections through cervical segments. Like the anterior boundaries, regionalized expression within somites was mostly similar in the two species, although there were differences, most notably for Hoxb-4. In chicks,hoxb-4 is expressed evenly throughout the sclerotome (Fig. 3G; open arrowhead) and is most concentrated in the dorsal myotome, which contains epaxial muscle precursors at this stage (Fig. 3G; closed arrowhead). In alligator, Hoxb-4 expression is uniform in the sclerotome as it is in chick, but unlike chick, was not detected in the myotome (Fig. 3H; open and

4 4 MANSFIELD AND ABZHANOV Figure 2. Neighbor-joining trees with bootstrap values for 1,000 replicates are shown for predicted amino acid sequences from alligator (Ami), chick (Gga), human (Hsa), and mouse (Mmu) for paralog groups 4 8. Ten out of the 15 group 4 8 Hox genes were isolated from alligator. Clustal W alignments and neighbor-joining trees were generated with Mega4.0. Positions containing gaps and missing data were eliminated from the dataset, and distances were computed using the Poisson correction method. Scale bars indicate number of amino acid substitutions per site. Genbank accession numbers for chick, human and mouse sequences are given in Supplementary Table 1. closed arrowheads, respectively). Although it is not obvious from this expression pattern or from mouse mutant phenotypes (Ramirez-Solis et al., 93) whether this variation correlates with a specific morphological change, it could regulate some aspect of epaxial muscle patterning or differentiation. Furthermore, Hoxb-4 expression is different in each species where it has been

5 HOX EXPRESSION IN ARCHOSAURIAN AXIAL PATTERNING 5 described, and is thus quite labile: in mouse embryos, Hoxb-4 mrna and protein are localized exclusively to the myotome and not observed in sclerotome at all (Brend et al., 2003). In chick, Hoxc-4 and Hoxd-4 are also expressed in the dorsal (epaxial) myotome and are present in the lateral, but not central, sclerotome (Fig. 3I, K). Lateral sclerotome gives rise to neural arches and lateral vertebral structures, including transverse processes and distal ribs; this region also contains syndetome, composed of axial tendon progenitors, at the AP margins of somites (reviewed in Christ et al., 2004). Although staining was weaker and, therefore, difficult to localize in alligator sections, it is apparent in both the myotome and sclerotome for both genes (Fig. 3K, L). A species difference in expression was observed in forelimbs: in chicks, Hoxc-4 expression is restricted to the proximal limb bud (Fig. 3C, arrow; Nelson et al., 96), whereas in alligators, expression extends distally along most of the anterior limb bud (Fig. 3D, arrow). Hoxc-4 is associated with forelimb identity in chick embryos (Nelson et al., 96), although it is not known to be required for forelimb patterning in mice, so the significance of this difference is not clear. Hox Group 5 Hoxa-5 and Hoxb-5 pattern segments around the cervical thoracic transition. Although Hoxb-5 expression is nearly identical between species, we observed striking differences in Hoxa-5 expression that correlate with differences in vertebral morphology. In both chicks and alligators, Hoxb-5 is expressed from pv2 (C2, so 6/7) and extends caudally to the tail (Fig. 4A; Ohya et al., 2005). Neural tube expression is present from the hindbrain to the tail. In chick transverse sections, expression is strong and uniform in the myotome and throughout the sclerotome (Fig. 4B, C). In cervical segments, expression is mostly restricted to somites (Fig. 4B; above arrowhead), but in thoracic segments it is also observed ventrally, in the LP-mesoderm (Fig. 4C; arrow). In alligators, a nearly identical pattern is observed in cervical and thoracic segments (Fig. 4D, E). Indeed, Hoxb-5 expression is broadly conserved: similar boundaries of somitic expression have been reported in mouse and turtle (Gaunt et al., 90; Ohya et al., 2005). In contrast, Hoxa-5 expression differs across species, both with respect to its anterior posterior expression boundaries and its pattern of expression within somites. In chicks, Hoxa-5 is most strongly expressed in the dorsal somites, with sharp anterior and posterior borders at pv8 and 14 (C8 C14; so 12/13 so 18/19), respectively. This posterior border corresponds to the cervical thoracic transition (Gaunt, 2000; Fig. 4F; arrowheads). A second, weaker expression domain is observed in ventral somites from pv12 (C12; so 16/17) posteriorly, grading out near the hind limb level (Fig. 4F). In alligators, somitic Hoxa-5 expression is not concentrated in cervical segments, but rather is strongest between pv9 and 12 (C9 T3; so 13/14 16/17), and is also present more weakly in the posterior thoracic region, between pv13 and 16 (T4 T7; so 17/18 20/21) (Fig. 4G). In specimens where in situ hybridization signal is allowed to develop longer, expression can also be detected weakly up to approximately C2 (Fig. 4H; arrowhead). Also in contrast to chick, no clear distinction between dorsal and ventral somitic expression is apparent in whole mount (compare Fig. 4F, G, H). This difference is further highlighted in transverse sections. In chick cervical somites, Hoxa-5 mrna is absent from the myotome, but present in sclerotome, where it is further localized to two regions: first, in the dorsal lateral-most sclerotome cells (Fig. 4I; arrowhead) and second, in a ventral cap of sclerotome that surrounds the ventral lateral lip of the myotome (Fig. 4I, arrow). Ventral sclerotome expression is also present in thoracic segments (Fig. 4J; arrow), but the dorsal domain is specific to the cervical region, as was also observed in whole-mount staining. As found for Hoxb-5, LP expression is present in thoracic but not cervical segments. In alligators, in contrast, Hoxa-5 expression is detected in both the myotome and the sclerotome. The ventral lateral cap of strongest sclerotome expression is present, similar to chick, in both cervical and thoracic segments (Fig. 4K, L, arrows), suggesting that this domain arose before the last common ancestor of birds and alligators. In contrast, the dorsal lateral sclerotome expression is unique to chicks and could have arisen within the avian lineage. Interestingly, this difference in Hoxa-5 expression correlates with variation in vertebral morphology. Unlike alligators, chicks have a sharp morphological boundary between cervical and thoracic segments, most obvious in the development of thoracic but not cervical ribs. Cervical rib loss in both birds and mammals occurs through fusion of embryonic rib progenitors with developing vertebral centra; therefore, these progenitors are specified at all axial levels, but undergo different development in cervical compared with thoracic segments (Goodrich, 30). Fate mapping has shown that the lateral sclerotome, where high Hoxa-5 expression is observed in chick, contains rib precursors (Aoyama et al., 2005). This could suggest a role for Hoxa-5 in regulating chondrogenesis of these precursors. Consistent with such a role in mice, Hoxa-5 mutants develop ectopic ribs on the last cervical segment (Jeannotte et al., 93). We next reexamined Hoxa-5 expression in mouse somites because cervical rib loss occurred independently in birds and mammals (Benton, 2005) and, so, could have partially distinct developmental mechanisms. These could include independent recruitment of Hoxa-5 for rib repression, through changes in its expression domain along the AP axis or within segments. In mice, as described earlier, Hoxa-5 is expressed from pv3 to 9 (C3 T2; so 7/8 13/14) and each of these segments is improperly patterned in null mutants (Jeannotte et al., 93; Fig. 4M). In transverse sections through cervical (Fig. 4N) or anterior thoracic segments (not shown), Hoxa-5 expression is present throughout the sclerotome, except in the most medial region. No concentrated expression in the dorsal or ventral lateral sclerotome is

6 6 MANSFIELD AND ABZHANOV Figure 3. Expression of Hoxb-4, Hoxc-4, andhoxd-4 in alligator and chick. (A) In chick (Gg), Hoxb-4 has an anterior limit at pv3 and is strongest from pv3 to 10 (C3-C10; so 7/8-so 14/15; arrowheads) and is less abundant in more posterior segments. (B) In alligator (Am), Hoxb-4 expression has an anterior limit at pv3 (C3; so 7/8) and is expressed most strongly between pv3 and 6 (C3-C6; so 7/8-so 10/11; arrowheads). (C) Chick Hoxc-4 expression has an anterior limit in somitic mesoderm at pv6 (C6; so 10/11; arrowhead). Expression visible anterior to this border is present in the dorsal root ganglia, not somitic mesoderm. In chick, as in mouse, Hoxc-4 expression is confined to the proximal anterior limb bud (arrow). (D) Alligator Hoxc-4 has an anterior expression limit at pv4 (C4; so 8/9; arrowhead). Somitic expression is strongest in the anterior-most 7 8 segments of the domain; posterior segmental expression is largely confined to dorsal root ganglia. Expression in the forelimb is different from chick, extending distally along most of the anterior limb bud (arrow). (E) Chick Hoxd-4 has an anterior expression limit at pv3 (C3; so 7/8). (F) Similarly, alligator Hoxd-4 shows an anterior expression limit at pv3 (C3; so 7/8). (G) In transverse section, chick Hoxb-4 is observed throughout the sclerotome (open arrowhead) and at higher levels in the dorsal (epaxial) myotome (closed arrowhead). (H) In alligator, Hoxb-4 is expressed evenly throughout the sclerotome, as in chick (open arrowhead). In contrast to chick, none or very low expression is observed in the myotome (closed arrowhead). (I) In transverse sections, chick Hoxc-4 is expressed in the lateral sclerotome and

7 HOX EXPRESSION IN ARCHOSAURIAN AXIAL PATTERNING 7 observed, however, suggesting that this domain could be unique to archosaurs. Alternatively, it could reflect species differences in sclerotome morphology. Nevertheless, strong cervical expression of Hoxa-5 is associated with a sharp cervical thoracic transition in both chicks and mice, and variation in Hoxa-5 expression, both along the AP axis and within segments, in mice, chicks, and alligators could suggest a partially derived role for Hoxa-5 in patterning avian cervical segments. Hox Groups 6 7 Hox groups 6 7 pattern thoracic segments. We examined the expression of Hoxa-6, Hoxc-6,andHoxa-7. No somitic expression of Hoxa-6 could be detected (not shown) and, correspondingly, Hoxa-6 expression is relatively low in chicks (not shown; Ohya et al., 2005). Both Hoxc-6 and Hoxa-7 expression patterns were largely conserved between alligators and chicks. Hoxc-6 has an anterior expression border at the cervical thoracic transition in all species examined, including mouse, chick, goose, frog, and lizard (Gaunt, 94; Burke et al., 95; Woltering et al., 2009; and see Fig. 5A, B). The same is true in alligator. Expression grades out gradually over approximately 9 somites and is highest between pv11 and 18 (T1 T8; so 15/ 16 22/23) (Fig. 5C, D). As described earlier, chick Hoxc-6 is expressed in the lateral sclerotome, dorsal dermis, and LP mesoderm, but only at low levels in the myotome (Ohya et al., 2005; Fig. 5E; arrowhead). In alligators, expression is relatively uniform throughout the sclerotome, dermis, and lateral plate, and seems most concentrated in myotome, unlike chick (Fig. 5F). No role for Hoxc-6 in muscle-specific patterning or differentiation is known, and thus the potential significance of this difference is not obvious; it is likely to be specific to alligators, because turtles and mice also have none or very little myotome expression (Ohya et al., 2005). The Hoxc-6 neural tube boundary is in the mid-cervical region in both species: in alligators it is at pv5 (C5, so 9/10) (Fig. 5D). In both species, transcripts are also detected in the proximal forelimb and hind limb buds, and in the flank surrounding hind limb buds, which contains LP mesoderm-derived precursors of the pelvic girdle (Fig. 5A, C, asterisks; Nelson et al., 96; Malashichev et al., 2005, 2008). Hoxa-7 is expressed in interlimb (flank) somites in both alligator and chick. In alligators, it is strongly expressed from pv11 to 17 (T2 T8; so 15/16 21/22) and grades out posteriorly over several segments (Fig. 5G). Similarly, in chicks it is expressed between pv16 and 21 (T2 T7; so 20/21 25/26) (Gaunt et al., 99). In alligator, the neural tube expression boundary is at pv5 (C5, so 9/10). In both species, Hoxa-7 transcripts are also detected in the posterior and distal limb buds and in the flank surrounding hind limb buds, which contains LP mesodermderived precursors of the pelvic girdle (Malashichev et al., 2005, 2008) (Fig. 5G). In alligator sections, Hoxa-7 expression seems similar to that of Hoxc-6: it is most concentrated in the myotome, and is also expressed in the dermis and both lateral and central sclerotome (Fig. 5H). In chick thoracic somites, Hoxa-7 is strongly expressed in the sclerotome and dermis, but unlike alligator myotome, expression in chick is restricted to just the dorsal and ventral lips (Ohya et al., 2005). Hox Group 8 Hox group 8 paralogs pattern thoracic segments. We were unable to detect somitic expression of Hoxb-8 in alligator (not shown). However, Hoxc-8 has an expression pattern unique to alligators. In chicks, the anterior border of Hoxc-8 is at pv19 22 (T5; so 23/24), with the most concentrated expression in the anterior-most 4 5 segments of its domain. (Burke et al., 95; Ohya et al., 2005; Fig. 5I, J; arrowheads). Weaker somitic expression extends into the caudal segments. A nearly identical pattern is observed in mice and turtles (Burke et al., 95; Ohya et al., 2005). In alligators, Hoxc-8 has an anterior border at pv10 (T1; so14/15) and expression continues posteriorly into the lumbar region (Fig. 5K, L). The position of the anterior border is thus shifted anteriorly relative to other species, so that it lies at the cervical thoracic transition rather than within the thoracic region. Additionally, no concentrated expression is observed in Figure 3. Continued. myotome (arrowhead), where it is most concentrated dorsally. (J) In alligator transverse sections, a similar pattern of Hoxc-4 expression is observed, with weaker expression in the sclerotome and more concentrated in myotome. (K) In chick transverse sections, Hoxd-4 expression resembles Hoxc-4: it is most highly expressed in myotome (arrowhead) and is detected weakly in lateral sclerotome. (L) In alligator transverse sections, Hoxd-4 is detected weakly in the sclerotome and more strongly in the myotome (arrowhead). (M) Schematic diagram of a somite cross section, with different regions of the somite labeled. At the stages observed, epithelial myotomes contain epaxial muscle precursors dorsally and hypaxial muscle precursors ventrally. Localization within sclerotome can be described along the medial lateral and dorsal ventral axes (ML, DV) and more broadly in lateral vs. central sclerotome, which have different derivatives (see text). (Differences along the anterior posterior axis are not shown in this plane of section). Hox expression in this and subsequent Figures was described whenever possible with reference to these regions and the chick fate map (see text). For all panels, whole-mount embryos shown are between HH24 and 25 (chick) or E10 (alligator); sections are HH24 (chick) and HH14 (alligator). Gg, Gallus gallus; Am,Alligator mississippiensis; a,aorta;d, dorsal root ganglion; m, myotome; nt, neural tube; nc, notochord; s, sclerotome; so, somite.

8 8 MANSFIELD AND ABZHANOV Figure 4. Expression of Hoxb-5 and Hoxa-5.(A) Hoxb-5 expression, shown in a whole-mount E14 alligator, has an anterior border at pv2 (C2, so 6/7; arrowhead) and extends posteriorly into caudal segments. This pattern is identical to that described in chick. (B) In a chick cervical segment, Hoxb-5 is expressed throughout the myotome and sclerotome, but is largely absent from LP (arrowhead marks the approximate border between somitic and LP tissue). (C) In chick anterior thoracic somites, Hoxb-5 expression is identical to cervical segments, except that it is also present in thelp(arrow).(d) In an alligator cervical section, Hoxb-5 is similar to chick: it is detected throughout the myotome and sclerotome (arrowhead). (E) Hoxb- 5 expression in alligator anterior thoracic segments is also similar to chick present in somitic derivatives and additionally present in LP (arrow). (F) In chicks, Hoxa-5 expression is predominantly cervical, with dorsal somitic expression from pv8 to 14 (C7-14; so 12/13-16/17; between arrowheads). Ventral somitic expression and PNS expression extends throughout the thoracic and lumbar segments, posteriorly to the hind limb level. (G) In alligator embryos, in contrast to chick, Hoxa-5 is expressed most strongly in thoracic somites between pv9 12 (C9-T3; so13/14-16/17; arrowheads). Weak staining is apparent in several segments anterior and posterior to this domain, particularly in embryos where in situ hybridization detection proceeded longer, (H) where Hoxa-5 can be detected weakly up to C2 (white arrowhead; black arrowheads bracket the same segments indicated in G). (I) In a chick cervical section, Hoxa-5 expression is mostly in the sclerotome, where it is most concentrated in two locations: first, in a ventral cap of sclerotome cells surrounding the ventral edges of the myotome (arrow). An additional chick-specific domain is present in the dorso-lateral sclerotome (arrowhead). In contrast to other species, expression is weak or absent in myotome. (J) In chick anterior thoracic segments, Hoxa-5 transcripts are weak or absent in the dorso-lateral sclerotome, but the ventral lateral expression domain is present and slightly expanded compared with cervical somites (arrow). Additional expression is present in the medial body wall and thoracic organs. (K) In an alligator cervical segment, Hoxa-5 is expressed in myotome, unlike chick, but like chick is present in the lateral sclerotome. The highest accumulation is observed in ventral lateral sclerotome, surrounding the ventral edges of the myotome (arrow), a domain shared with chick; in contrast to chick, no dorsal sclerotome accumulation of Hoxa-5 is observed. (L) In an alligator anterior thoracic segment, Hoxa-5 expression is identical to cervical segments. (M) In mouse, Hoxa-5 is expressed in the somites between pv3 and 9 (C3-T2; so 7/8-13/14; arrowheads). Weak expression continues in more posterior somites. (N) In a mouse cervical section, Hoxa-5 expression is detected in the myotome and lateral sclerotome, with no apparent ventral or dorsal localization, in contrast to chick and alligator. In all panels, alligator embryos (section and whole-mount) are at E14. Chick embryos are HH24 and mouse embryos are E11. Gg, Gallus gallus; Am, Alligator mississippiensis; Mm, Mus musculus; a, aorta; d, dorsal root ganglion; m, myotome; nt, neural tube; nc, notochord; s, sclerotome; so, somite.

9 HOX EXPRESSION IN ARCHOSAURIAN AXIAL PATTERNING 9 Figure 5. Expression of Hoxc-6, Hoxa-7, andhoxc-8. (A) In chicks, Hoxc-6 expression is observed up to an anterior limit at pv15 (T1; so 19/20) (shown in lateral and dorsal view (B), arrowheads). Expression is also present in anterior proximal regions of the limbs (asterisks). (C) In alligators, Hoxc-6 transcripts are detected in thoracic segments with an anterior limit at pv10 (T1; so 14/15), as observed in other species and grading out posteriorly around the level of pv18 (T8; so 22/23; arrowheads). Neural tube expression is weak, but in specimens where in situ hybridization signal was processed for longer, (D), it can be detected with an anterior limit at pv5 (C5; so9/10) (arrow). Expression can also be detected in the anterior proximal region of the limbs (asterisks). (E) In chick thoracic sections, Hoxc-6 is expressed in dermis and lateral sclerotome but only weakly in the myotome (arrowhead). It is also expressed in the LP mesoderm. (F) In alligator thoracic sections, Hoxc-6 transcripts are present in all somitic derivatives and LP mesoderm. Unlike in chick, expression seems most concentrated in the myotome. (G) Inalligator,Hoxa-7 is expressed in somites from pv11 to 17 (T2-T8; so 15/16-21/22; arrowheads). Neural tube expression extends to the level of pv5 (C5; so 9/10). (H) In an alligator thoracic section, Hoxa-7 expression is highest in myotome and is also present at lower levels in sclerotome and dermis. (I) and(j) show lateral and dorsal views, respectively, of Hoxc-8 expression in chick. Hoxc-8 has an anterior somitic boundary at pv19 (T5; so 23/24) and is most concentrated in the four anterior-most segments of its domain (between arrowheads), a pattern also observed in mice and turtles. Weaker expression is present posteriorly into the caudal segments. (K) and (L) show lateral and dorsal views, respectively, of Hoxc-8 expression in alligator. The anterior limit of somitic expression is at pv10 (T1; so 14/15), but is strongest from T2 posteriorly (arrowhead). Concentrated Hoxc-8 expression in the anterior-most segments of its domain, observed in chick and mouse, is absent in alligator. In the lateral plate mesoderm, Hoxc-8 expression has an approximate anterior limit at pv18 (T9; so 22/23; arrow), a border corresponding to the transition between sternal and free ribs as in mice. (M) In chick transverse thoracic sections, Hoxc-8 is expressed in the sclerotome and dermis and is largely excluded from the myotome, similar to Hoxc-6. Expression is strong in the dermis (arrowhead). (N) Ina thoracic-level alligator section Hoxc-8 is detected in lateral sclerotome (arrowhead) and myotome but seems lowest in dermis. Chick sections are from a HH24 embryo; whole-mount embryos are HH27 (A, B) or HH28 (I, J). Alligator embryos are at E14 in all panels. Gg, Gallus gallus; Am, Alligator mississippiensis; a, aorta; d, dorsal root ganglion; m, myotome; nt, neural tube; nc, notochord; s, sclerotome; so, somite.

10 10 MANSFIELD AND ABZHANOV the anterior-most 4 5 somites of its domain as it is in chicks, mice, and turtles. We also examined younger (E6) and older (E28) alligator embryos, but did not find evidence for this expression domain (not shown). Owing to its conservation, it was most likely present in basal amniotes (and archosaurs), but lost within the crocodilian lineage. Although it is possible that failure to detect this domain is related to the probe strength, we were able to detect strong Hoxc-8 expression in the neural tube, PNS and LP mesoderm. Interestingly, this difference correlates with specializations in alligator thoracic vertebrae, which have enlarged thoracic neural spines compared with other species. Mice bearing both loss and gain of function Hoxc-8 alleles show altered patterning in thoracic neural spines and arches, and Hoxc-8 can negatively regulate their growth in a dose-dependent manner (Le Mouellic et al., 92; Pollock et al., 95; Yueh et al., 98; van den Akker et al., 2001; Juan and Ruddle, 2003). Therefore, this expression difference is potentially of functional significance, and is discussed below. In alligators, we also observed expression of Hoxc-8 in the LP mesoderm extending from the level of pv18 (T9; so 22/23) posteriorly (Fig. 5K; arrow). The anterior border of this domain corresponds to the transition from sternal to free ribs. The same LP expression, corresponding to the sternal/non-sternal rib border, has been described in mice (Ohya et al., 2005), but not turtles, which do not form a sternum at all. Interestingly, Hoxc-8 represses formation of sternebrae in mice and could, therefore, play a similar role in alligator (Le Mouellic et al., 92). In sections through chick thoracic somites, Hoxc-8 expression is largely excluded from myotome, but is present in the dorsal dermis and expressed evenly throughout the lateral sclerotome (Fig. 5M). Dermal expression appears highest near the border between the somites and LP mesoderm (Fig. 5M; arrowhead; Ohya et al., 2005). In alligator, although expression in thoracic somites is low, it seems most concentrated in myotome and lateral sclerotome, but dermal expression is low (Fig. 5N). DISCUSSION Hox proteins play a key role in patterning the AP body axis in most animal embryos and regulate the patterning and subsequent differentiation of vertebrate axial tissues. Changes in the Hox expression code have contributed to the diversification of vertebrate body plans, and mapping these changes across all crown groups of vertebrates will likely continue to produce insights into how the vertebrate body plan has evolved. The avian axial skeleton has undergone many modifications, most associated with adaptation to flight and bipedal walking (Benton, 2005). Most of these changes, including the expansion of the neck and repression of cervical ribs, fusion of thoracic vertebrae, and formation of the synsacrum and pygostyle through fusion of lumbar-sacral and caudal vertebrae, respectively, are specific to one axial level. As such, they may have been influenced by changes in the Hox expression code, which is required for anterior posterior patterning of most tissues. Among amniotes, chick and mouse Hox expression has been best characterized. Given that the ancestors of birds and mammals diverged early in amniote history, features of Hox expression shared between them are likely to represent features of a basal amniote Hox code, although convergence is also possible. It is difficult to determine, however, whether each of the many differences in gene expression is owing to changes in the avian code, mammalian code, or both. Furthermore, although some differences are functionally important, others likely are not. By comparing chick Hox expression to that of a more closely related species, alligator, we aimed to uncover avian-specific Hox expression patterns associated with unique features of axial patterning in birds. In terms of the axial formula, the crocodilian body plan lacks the avian specific modifications described above. Indeed, crocodilians have particularly conservative axial formulae. Within the crocodilian lineage, there has been little variation, with 8 9 cervical, 13 dorsal (thoracic and lumbar), and 2 sacral vertebrae in most described species (Hoffstetter and Gasc, 69). Furthermore, this body plan closely resembles that of basal archosaurs, sister groups to archosaurs, and even basal amniotes, which generally had 8 cervical vertebrae with long unfused (free) ribs, approximately rib-bearing dorsal vertebrae, and 2 sacral vertebrae. Lumbar vertebrae (with short fused ribs rather than long thoracic ribs) appear in Diadectomorpha, the closest outgroup to amniotes and are thus ancestral, although they were lost in many lineages (Ewer, 65; Cruickshank, 72; Gow, 75; Gauthier et al., 88; Berman et al., 2004). The schematic in Figure 6 compares the axial formulae of O. pabstii, a Diadectomorph resembling basal amniotes, Euparkeria, a close sister group to archosaurs, and our representative crown-group archosaurs, alligator and chick. The mouse axial formula is also shown and this is also highly derived compared with early mammals, which likewise had similar body plans to basal amniotes (Stovall et al., 66). A comparison of the AP boundaries of somitic Hox expression in alligator, chick, and mouse is also summarized in Figure 6. Expression boundaries for chick and mouse have been published earlier (Gaunt et al., 90, 99; Puschel et al., 91; Gaunt, 94, 2000; Burke et al., 95; Akasaka et al., 96; Ohya et al., 2005). We also examined and compared intrasomitic expression for each gene, although this is not represented in Figure 6. Not surprisingly, somitic Hox expression is largely conserved between alligators and chicks. However, several differences were observed and two of these, in Hoxc-8 and Hoxa-5, correlate with skeletal morphologies particular to each lineage. These are discussed further in the sections below. Other differences were less obviously linked to a specific morphological variation; for example, the expression of Hoxb-4 in chick cervical dorsal myotomes but not in alligator and, conversely, expression of

11 HOX EXPRESSION IN ARCHOSAURIAN AXIAL PATTERNING 11 Figure 6. Schematic representation of axial formulae and Hox expression boundaries in the somitic mesoderm. The pre-caudal vertebrae in different regions are represented by boxes (cervical, light blue; thoracic, teal; lumbar, dark blue; sacral, purple); non-sternal ribs are drawn as short horizontal lines and ribs articulating with the sternum as long horizontal lines. The alligator axial formula is quite conservative, revealed by comparison to Euparkeria, a sister group to stem archosaurs, and Orobates pabstii, a member of the sister taxon to stem amniotes, Diadectomorpha. In contrast to crocodilians, bird (and mammalian) axial formulae are highly modified. Hox expression is summarized next to the diagrams of alligator, chick, and mouse. Lines represent the anterior posterior extent of somitic Hox expression in the somites. Where a boundary of strong posterior expression was observed that boundary is shown; in some cases, weaker expression continues posteriorly. All alligator expression patterns are described in this work; chick and mouse AP boundaries indicated have previously been reported, as cited in the text. Hoxc-6 and Hoxa-7 throughout thoracic myotomes of alligator but not other species. Whether these differences contribute to lineage-specific differences muscle patterning could be tested in functional studies. Out of the 15 predicted Hox group 4 8 genes, we cloned 10 and detected somitic expression for 8. Owing to sequence similarity, particularly within paralog groups, cross-reactivity of hybridization probes is a potential concern, even though probes were designed to include variable regions. Although we cannot rule it out, we believe that probes were specific because each revealed a unique expression pattern and, furthermore, for most genes examined, expression patterns corresponded in most respects with those described in other tetrapods. Also, in cases where expression was different in alligator, it did not correspond to that of a closely related Hox gene, which might have suggested cross-reactivity. We have focused on somitic Hox expression, and indeed most of the axial skeleton and musculature are derived from and develop surrounded by somitic derivatives. However, LP mesoderm gives rise to the appendicular skeleton and sternum. In addition, some ventral axial structures, which form following migration of somitic cells into the lateral plate, bear contributions from both populations. For example, in mice, the first rib contains somite-derived chondrocytes but its surrounding perichondrium/periosteum, a tissue that regulates cartilage differentiation and morphology, is mostly of LP origin (Durland et al., 2008). The same is true for multiple sternal ribs in chicks (Nowicki et al., 2003). Likewise, some axial muscles that span dorsal and ventral regions contain contributions from both populations. Heterotopic transplantations have shown that structures derived exclusively from somites (primaxial tissues) are patterned by autonomous Hox expression established before segmentation, but that those axial structures comprised of somitic cells that migrate into the lateral plate and bear some LP-derived connective tissue (abaxial tissues), change fate, and Hox expression following transplantation (Kieny et al., 72; Nowicki and Burke, 2000). Thus, Hox patterning in abaxial tissues relies on the LP Hox code.

12 12 MANSFIELD AND ABZHANOV Coordinate patterning of developing axial and appendicular musculoskeletal systems, and coordination between somitic and LP mesoderm, must necessitate extensive communication between these tissues. Thus, it will be important to consider gene expression changes including but by no means limited to Hox genes in both somites and LP, and to consider these changes in the context of species-specific differences in how these cell populations interact and contribute to axial structures (reviewed in Winslow et al., 2007; Shearman and Burke, 2009). Consideration must also be made of surrounding tissues, including neural tube, notochord, and ectoderm, which provide signals required to pattern somites and maintain and regulate the differentiation of somitic derivatives and are, therefore, essential regulators of the morphology of axial tissues (reviewed in Brent and Tabin, 2002). Changes in signaling, downstream responses to signaling, or relative positions of these tissues could also underlie variation in the patterning and morphology of axial tissues. Hoxc-8 and Alligator Thoracic Morphology In chicks, mice, and turtles, Hoxc-8 has a somitic anterior expression limit in the mid-thoracic region, with strongest expression in the sclerotome and dermis of the anterior-most 4 5 segments of its domain (Burke et al., 95; Ohya et al., 2005; Fig. 5I, J). In mice, Hoxc-8 is required to pattern both the thoracic lumbar and lumbar sacral transitions and both are shifted posteriorly in null mutants. In addition, the thoracic neural arches and spines are improperly patterned in both gain and loss of function mutants. Additional defects include formation of ectopic sternebrae and malformation of the xiphoid process (Le Mouellic et al., 92). Alligators share with mice an anterior border of LP Hoxc-8 expression at the transition between ribs that articulate with the sternum and those that are free. (Ohya et al., 2005; Fig. 5K). In contrast, in turtles, which lack a sternum, Hoxc-8 has a much more anterior LP border. This correlation in LP expression of Hoxc-8 with sternebrae repression, functionally demonstrated in mice, suggests that this role for Hoxc-8 is likely conserved across amniotes. It is likely that the expression of Hoxc-8 in both LP and somites is important for patterning sternebrae, although this awaits functional testing. In addition to early expression of Hoxc-8 in both tissues, the mouse Hoxc-8 enhancer is active in both costal cartilage and sternum during late embryogenesis, and sternebrae form as cartilaginous fusions between distal rib tips and the sternum (Pollock et al., 95; Yueh et al., 98). Although the sternum is completely LP derived, distal ribs show varying composition: in mice, all ribs except T1 are primaxial, whereas in chicks (and perhaps alligators, although it is not known) the most distal (sternal) rib segments bear LP-derived periosteum and are therefore abaxial (Nowicki et al., 2003; Durland et al., 2008). As postulated earlier, sternebrae formation would depend on both the patterning and maturation of cartilage in these two tissues and on the proper alignment between them (Pollock et al., 95), which could be negatively regulated by Hoxc-8 expression in either or in both tissues. Unlike mouse, chick and turtle, alligators lack concentrated Hoxc-8 expression in the anterior-most 4 5 somites of its domain, and the anterior boundary is shifted forward, to the cervical thoracic transition (compare Fig. 5I, J, K, L). Interestingly, crocodilian thoracic segments have proportionally larger dorsal neural spines, supporting larger epaxial musculature, compared with these other species (Organ, 2006). This is required in part for alligators mode of feeding, which involves underwater rolling and shaking prey. Hoxc-8 is required to pattern thoracic neural arches and dorsal neural spines in mice; therefore, loss of this domain could help promote larger spines in alligators. In addition to homeotic transformations associated with Hoxc-8 mutations, which include a shift in the position of the transitional vertebra (marked by a switch in neural arch articulation pattern and spine morphology), Hoxc-8 alsoregulates the morphology, including the overall size, thoracic neural arches and spines. Indeed, overexpression of Hoxc-8 within its normal domain leads to a dose-dependent reduction in the size of these structures (Pollock et al., 95; Yueh et al., 98). It is, therefore, possible that loss of strong somitic Hoxc-8 in the alligator lineage enhances neural spine development. A different mechanism must be invoked to explain the origin of enlarged cervical and thoracic neural spines, as these were present in basal archosaurs as well (Gauthier et al., 88). However, loss of the alligator-specific dorsal domain could promote further neural spine enlargement within this lineage. Future functional studies, such as loss of function in chick or gain of function in alligators, could establish a role for Hoxc-8 in negatively regulating thoracic neural spine development. Hoxa-5 Variation Correlates with Cervical Segment Morphology in Chicks, Alligators, and Mice A striking difference in Hox expression between the alligator, chick, and mouse is in Hoxa-5. In both chicks and mice, Hoxa-5 expression is highest in the cervical somites;alligators, in contrast, show mostly thoracic expression (Fig. 4F, G, H, M). Thus, strong cervical expression of Hoxa-5 correlates with ribless flexible cervical segments in chicks and mice, whereas weaker, more posterior expression in alligator correlates with development of cervical ribs. Hoxa-5 has rib-repressing activity in mice: null embryos show posterior transformations, including development of ribs on the last cervical vertebra (Jeannotte et al., 93; Tabaries et al., 2007). In mice, as described above, the first rib bears contributions from both somitic and LP mesoderm, so loss of expression in either or both tissues may contribute to this phenotype (Nowicki et al., 2003; Durland et al., 2008). Whether chick Hoxa-5 also represses cervical rib development awaits functional characterization, but its expression in somitic rib precursors and near identity to the mouse ortholog (which