The evolutionary origins of the chordate central nervous system remain uncertain. Conclusions drawn from classical morphological comparisons and from a broad range of metazoan phyla conflict with the new molecular genetic information from developmental model systems characterized by central nervous systems. This has led to debate as to the nature of the ancestral deuterostome nervous system. Hemichordates as basal deuterostomes occupy a phylogenetically critical position for addressing hypotheses on the evolution of the chordate nervous system. Characterizing the molecular basis of the development of their diffuse nervous system offers insights into the role of conserved body patterning genes in the evolution of specific neural anatomies. We present a description of hox gene expression during the development of the direct-developing hemichordate Saccoglossus kowalevskii. The nested ectodermal expression of these genes in a hemichordate suggest that they play a general patterning role in the anterior/posterior regionalization of the ectoderm of bilaterians rather than being uniquely associated with the development and evolution of central nervous systems.
Both the evolution of bilaterian brains and the origins of the chordate central nervous system have been hotly debated over the past decade. Data have emerged from molecular genetics on the remarkable similarities between the molecular patterning involved in morphogenesis of chordate and arthropod central nervous systems (Arendt and Nubler-Jung 1996; De Robertis and Sasai 1996; Nielsen 1999; Reichert and Simeone 1999; Holland 2003; Lowe and others 2003; Lichtneckert and Reichert 2005). This new dataset conflicts with the classical view that nervous system complexity has gradually increased, with early bilaterian animals being characterized by more diffuse and simple neural organization. These early forms evolved independently into the complex central nervous systems associated with the later diversification of the protostomes and deuterostomes (Holland 2003). A growing list of genes has been proposed to play a conserved role in the patterning of the central nervous system of vertebrates and of Drosophila. Of genes involved in anterior specification, otx (Acampora and others 1998; Leuzinger and others 1998) and pax6 (Halder and others 1995; Gehring and Ikeo 1999) have been studied most comprehensively, but the list continues to grow (Lowe and others 2003). Initial comparisons were restricted to an anterior otx domain and a posterior hox domain, but this has been refined to include numerous other molecular genetic similarities between the anteroposterior axis of the developing tripartite brain of vertebrates and arthropods (Hirth and others 2003), and further expanded to the dorsoventral axis (Arendt and Nubler-Jung 1996; De Robertis and Sasai 1996; Ferguson 1996; Holland 2003). This has led to the hypothesis that the central nervous systems and tripartite brains of arthropods and chordates are homologous (Hirth and others 2003; Lichtneckert and Reichert 2005). Alternatively, central nervous systems and tripartite brains could have evolved independently in protostomes and deuterostomes: their conserved molecular similarities could reflect a regulatory conservation conveying basic axial positional information, rather than a tight coupling with any particular neuroanatomical structure. Resolution of these ambiguities might be reached through the analysis of the molecular characteristics of nervous systems of yet other bilaterian groups (Holland 2003).
Hox genes, which are part of the conserved suite of genes involved in neuraxial patterning in arthropods and chordates, have been examined in a range of bilateria. They play critical roles in the patterning of the hindbrain and spinal cord in vertebrates (Krumlauf and others 1993; Lumsden and Krumlauf 1996; Pearson and others 2005), and in the development of the central nervous system of Drosophila. The phylogenetic sampling of hox genes and their expression during development has improved with data from groups such as annelids (Irvine and Martindale 2000) and mollusks (Hinman and others 2003), both showing an association with the development of the ventral nervous system. There is however little information on the role of hox genes in the development of animals with a more diffuse organization of the nervous system. This raises the following question: are hox genes associated with the formation of all bilaterian nervous systems, diffuse or centralized, and as such unreliable markers for testing hypotheses of specific neuroanatomies, or are they uniquely associated with central nervous systems?
Hemichordates are representatives of the deuterostome lineage, along with their sister group the echinoderms, chordates (Cameron and others 2000; Blair and Hedges 2005), and a poorly described new group, Xenoturbella, currently only represented by 1 species (Bourlat and others 2003). In contrast to chordates, the major organizational feature of the hemichordate nervous system is that it is basiepithelial. There is no central nervous system and cell bodies are scattered throughout the epithelium. A mat of axons is spread out along the basement membrane, which is thickened in certain areas, such as at the base of the proboscis (prosome), along the anterodorsal region of the body in the mesosome (collar), and in both the dorsal and ventral midlines of the metasome (trunk) (Bullock 1946; Knight-Jones 1952; Bullock and Horridge 1965). During development, the entire ectoderm is neurogenic, as shown by the broad expression of pan-neural markers at early developmental stages of the direct-developing enteropneust hemichordate Saccoglossus kowalevskii (Lowe and others 2003). The diffuse nervous system of this group, and their key phylogenetic position as an outgroup to the chordates, make hemichordates a particularly interesting phylum for testing hypotheses on the origins of the chordate nervous system. They also provide an opportunity to examine the reliability of conserved developmental genes for testing hypotheses of neuroanatomical homology between distantly related groups with central nervous systems. Our work addresses whether conserved developmental genes are involved in the generation of specific neuroanatomical features, or if they were more generally involved in the A/P patterning of all bilaterian nervous systems, irrespective of their specific organizational complexity or level of integration. Previously, we reported the expression of a suite of such genes with conserved roles in the development of vertebrate and Drosophila central nervous systems, including 5 hox genes (Lowe and others 2003). In this study, we extend the analysis and report the cloning and expression of 6 additional hox genes. This almost completes the entire hox complement in hemichordates. Hox genes have been proposed to have an ancestral role in patterning the nervous system of chordates (Schilling and Knight 2001), and we examine and compare their expression during the early development of S. kowalevskii.
Materials and methods
Animal collection, spawning and embryo culture were carried out as described previously (Lowe and others 2004). Cloning of hox genes and in situ hybridization was carried out as described previously (Lowe and others 2003; Lowe and others 2004). Briefly, genes were cloned either from a collection of ESTs from 3 different developmental stages (gastrula, neurula and juvenile) or primers were designed to conserved domains using Codehop (Rose and others 2003) for PCR assay of orthologous sequences. The hox sequences from S. kowalevskii were added to a previously published alignment (Peterson 2004) using ClustalX (Jeanmougin and others 1998). Neighbor-joining trees were constructed using PAUP* (Swofford 1999). Nonparametric bootstrap proportions for clades were assessed with 1000 pseudoreplicates.
Previously, we established the clear orthology between several of the members of the Hox complex, namely hox1, hox3 and hox4 (Lowe and others 2003). At that time we assigned 2 genes as hox7/8 and hox11/13 since our phylogenetic analysis was too ambiguous to assign definitive orthology. Subsequent to this study, an analysis of hox genes has been carried out in another species of hemichordate Ptychodera flava (Peterson 2004) and in the sea urchin Strongylocentrotus purpuratus (Cameron and others 2006). We have cloned 6 additional hox genes and carried out a phylogenetic analysis using the homeodomain aligned with hox sequences from a range of metazoans. The addition of a complete echinoderm hox complement, and additional hemichordate sequences has improved the resolution of the tree, making more definitive orthology assignments possible. The analysis is shown in Figure 1. Our original orthology assignments still stand for hox1, hox3 and hox4, but the previously assigned hox7/8 is now reclassified as hox7 (GenBank accession no. DQ985450) as it grouped with strong bootstrap support (96) with the 2 sea urchin hox7 sequences. Hox11/13 is now reclassified as hox11/13c (GenBank accession no. DQ985454) and grouped closely with P. flava Hox11/13c (77). Of the new sequences added to the analysis we established clear orthology for hox2 (Genbank accession no. DQ985445) (82) and hox5 (GenBank accession no. DQ985448) (75) with good bootstrap support. The position of another center-class gene was difficult to resolve, but grouped closely with hox6 (GenBank accession no. DQ985449) from P. flava and is thus assigned orthology group 6. We were not able to clone an ortholog of hox8 from our EST collection. Our complement of posterior class genes includes hox9/10 (GenBank accession no. DQ985451), hox11/13a, (GenBank accession no. DQ985452), and 11/13b (GenBank accession no. DQ985453), as has been described in P. flava and the sea urchin S. purpuratus. Our collection now includes all the hox members originally distinguished by a PCR screen of homeobox genes in S. kowalevskii (Pendleton and others 1993) on the basis of 82 base homeobox fragments. The results strongly suggest that, like other basal deuterostomes, S. kowalevskii has only one Hox cluster.
Expression of hox genes during development of S. kowalevskii
Hox1. Expression of hox1 was originally reported by Lowe and colleagues (2003). The details of expression remain as reported (Fig. 2A–C), but we add additional stages at Days 2 and 3 of development (Fig. 2A and B); expression is first detected in the early gastrula in a strong band of ectodermal expression beginning just anterior to the ciliated band in the metasome and continuing posterior to this band. No expression is detected within the epithelium of the ciliated band (Lowe and others 2003). Strong expression continues anterior to the ciliated band in subsequent stages of development (Fig. 2A–C) as the embryos elongate. At Day 4 and later, expression extends further anterior along both ventral and dorsal midlines, but also dips more posteriorly in the lateral regions following the posterior limit of the first gill slits (Fig. 2C). Expression is also detected in the endoderm at later stages in a domain directly underlying the strong ectodermal band of expression in the anterior metasome (Fig. 2C). The endodermal domain of hox1 gene expression below the gill slit marks the posterior boundary of the pharyngeal endoderm. A similar hox1/labial domain is found in the anterior gut of both Drosophila and chordates including amphioxus (Schubert and others 2005).
Hox2. Expression of hox2 is much weaker than other genes in this study. Color development of the whole mount in situ hybridization was extended in order to detect signal. Consequently, background staining was much more pronounced in the coelomic cavities and endoderm, particularly in later developmental stages. Expression is first detected in the late gastrula in a domain quite similar to hox1 except that the anterior limit is found more posteriorly (Fig. 2D–F). There is an ectodermal ring just anterior to the ciliated band, and around the blastopore (Fig. 2D). At Day 2 of development, although extensive background staining is detected in the endoderm, the ectodermal expression maintains a consistent domain similar to that in the gastrula (Fig. 2E). At Day 3 of development, expression on the dorsal side of the embryo in the ectoderm becomes weaker, and expression is stronger on the ventral side (Fig. 2F). We could not detect expression in later stages close to hatching (data not shown).
Hox3 and hox4.Hox3 and hox4 were previously described by Lowe and colleagues (2003) and exhibit similar domains of expression that resemble that of hox5, in later developmental stages.
Hox5. We first detect hox5 during gastrulation around the blastopore (Fig. 2G). After embryo elongation at the end of Day 2, expression is detected in a thin line immediately anterior to the ciliated band and in the ectoderm posterior to this structure (Fig. 2H). Expression of this gene is stronger in the most posterior ectoderm than in other more anterior hox genes surveyed above. In late stage embryos at Day 4 of development, as the first gill slit is forming, expression is detected throughout the posterior ectoderm. At this stage the ciliated band is displaced; extending from a posterior ventral position up to a more anterior dorsal position (Fig. 2I).
Hox6 and hox7.Hox6 and hox7 generally show quite similar expression profiles to hox5. At gastrula and Day 2 of development (Fig. 2J, K, M and N) expression is detected just anterior to the ciliated band as in the case for hox2 and hox5, but hox6 is expressed evenly around the blastopore, whereas hox7 is expressed more on the ventral side of the blastopore. At later stages the expression domains of the 2 genes diverge and hox6 resembles hox5 quite closely (Fig. 2L and I), whereas hox7 is expressed more ventrally than dorsally, and the staining extends into the ciliated band (Fig. 2O).
Hox9/10. Of the posterior class members, hox9/10 shows the furthest anterior limit of expression, and is expressed quite strongly. We first detected expression early during gastrulation around the blastopore (data not shown). By the end of gastrulation, expression extends throughout all the posterior ectoderm and into the ciliated band (Fig. 2P). By Day 2 of development, this pattern of expression remains consistent (Fig. 2Q). At later stages close to hatching, when gill slits have formed, the strongest expression is detected in the most posterior ectoderm and is no longer detected in the epithelium of the ciliated band (Fig. 2L).
Hox11/13a.Hox11/13a expression is unusual in first being detected exclusively in the ciliated epithelium of the developing ciliated band from early gastrula onwards (Fig. 2S). This association holds throughout all stages (Fig. 2S–U). By Day 3 of development, after the extension of the dorsal side of the embryo, expression expands into the surrounding ventral and most posterior ectoderm close to the ciliated band. At the same time, it is downregulated in the lateral and dorsal parts of the ciliated band. The latest developmental stage examined for expression of hox11/13a was a late stage juvenile at Day 13 of development (Fig. 2β). These animals had developed a long, ventral postanal tail, and expression was detected at the very posterior tip of this structure (Fig. 2β).
Hox11/13b.Hox11/13b is first detected after gastrulation as the embryo elongates at Day 2 of development. Expression initially occurs in the posterior ventral ectoderm and the underlying posterior endoderm (Fig. 2V and W). Later stages also show a similar topology of expression (Fig. 2X).
Hox11/13c. Expression of this most posterior member is almost identical to that of hox11/13b but begins earlier in gastrulation. This expression was originally reported in Lowe and colleagues (2003), as hox11/13. Expression persists throughout all stages examined up until hatching in the posterior ectoderm and the very posterior endoderm (Fig. 2Y, Z and α). At Day 13 of development, after hatching and well into juvenile growth, expression of hox11/13c (as in hox11/13a) is restricted to the ectoderm in the postanal tail (Fig. 2γ).
The hemichordate hox gene complement
We present an additional 6 hox genes to supplement the 5 we cloned in a previous study (Lowe and others 2003). Hemichordates have a complement of hox genes similar to that found in the sea urchin S. purpuratus, a member of their sister group, the echinoderms (Fig. 3). However, in S. purpuratus, hox4 has been lost due to cluster rearrangements (Cameron and others 2006), but is present in hemichordates and another echinoderm class, the asteroids (Long and others 2003), suggesting that the loss of hox4 occurred subsequent to the divergence of asteroids from other echinoderms, and that it was present in the ancestral deuterostome Hox cluster. Assigning definitive orthology to the central class genes is more difficult, but based on the assignments in sea urchins from a combined phylogenetic study and Hox cluster linkage analysis, S. kowalevskii has at least 2 members of this group: hox6 and 7. We were not able to clone an ortholog of hox8 from our EST collection, but negative data do not necessarily imply its absence from the hemichordate cluster. Genomic sequence and cluster analysis will be required to definitively determine the presence or absence of this member of the cluster. The entire complement of posterior class genes cloned from S. purpuratus and other echinoderms is present in both species of hemichordate. Previously it was reported that amphioxus posterior class genes are evolving more rapidly than are other members of the cluster, leading to the proposal that perhaps deuterostome posterior hox genes are evolving more rapidly than are other cluster members (Ferrier and others 2000). However, the grouping of echinoderm and hemichordate posterior hox members suggest that this may not be a general feature of deuterostome posterior hox evolution (Peterson 2004).
While we have now cloned almost all the cluster members, there are currently no data on the genomic arrangement of hox genes in hemichordates. Recent data suggest that this group will be informative for addressing the relationship between colinearity and regulation of the Hox complex as the cluster has been reorganized independently in several of the major lineages in the deuterostomes. In the sea urchin S. purpuratus, the Hox cluster has undergone a series of reorganizations resulting in the loss of hox4 and inversion of the cluster from hox5 to hox11/13c (Cameron and others 2006) (Fig. 3). Within the urochordates, in ascidians and larvaceans (appendicularians), the Hox cluster has been dispersed to varying extents. In the case of larvaceans there is no evidence of clustering (Seo and others 2004), and in ascidians colinearity within the cluster has been partially dispersed with the hox genes arranged on 2 chromosomes, and with some changes in gene order (Ikuta and others 2004; Spagnuolo and others 2003) (Fig. 3). Interestingly, in larvaceans, despite the dispersion of the Hox complex, the expression pattern of the cluster members is still axially ordered. Hox cluster information from hemichordates will help clarify the picture of early Hox cluster evolution in the deuterostomes.
Colinearity of hox expression
With the possible exception of the hox8, this study completes the cloning and expression profile of the hox complement from S. kowalevskii. The entire complex of hox genes is expressed almost exclusively in the developing surface ectoderm, which contains finely intermixed neural and epidermal cells. As in other bilaterian phyla during the development of the adult body plan, an extensive neural hox domain is involved in the development of the region of the animal posterior to the otx domain (Hirth and others 2003). In a variety of chordates, arthropods and annelids, the genes in the most 3′ region of the cluster are expressed more anteriorly, while increasingly more 5′ members of the cluster are expressed more posteriorly in the A/P axis, resulting in a colinear arrangement of cluster position and expression domain. Figure 4B shows a model of a generalized vertebrate central nervous system, representing a consensus of the A/P progression in expression domains of different Hox clusters.
Although its Hox cluster organization has not yet been determined, S. kowalevskii exhibits a similar pattern with the anterior class hox genes, such as hox 1, expressed anteriorly around the developing gill slits and the posterior class members expressed increasingly posteriorly (Fig. 4A). While this general trend of posterior progression is obvious over the entire hox complement, many of the genes are initially quite similar in their expression at early stages of development. For example, hox2, hox5, hox6 and hox7 are all expressed just anterior to the ciliated band of early stages, and it is not until later developmental time points that differences in expression can be detected among these genes (Fig. 2D–O, summarized in Fig. 4A). Expression analysis in the hatched juvenile stage, when the metasome (trunk) is elongated, is difficult because signal is obscured by extensive background staining due to probe trapping. It is however possible that colinearity of expression may be much more pronounced in these later developmental stages as the trunk continues to elongate.
Another feature of hox gene regulation that has been described in many but not all bilaterian groups is temporal colinearity, which correlates with spatial colinearity. Anterior hox genes are activated before the onset of more posterior cluster members (Izpisua-Belmonte and others 1991). In S. kowalevskii, however, there was no obvious trend of earlier activation of anterior class hox expression. We neither tightly bracketed early developmental stages nor did we examine the expression of most genes before gastrulation. Consequently, we may have missed the temporal colinearity of hox gene expression during development. Alternatively, S. kowalevskii may be a bilaterian in which, like Drosophila, anteroposterior timing differences are not a feature of hox gene expression.
Hox and ectodermal patterning
The role of the Hox cluster is to establish positional information along an axis (Pearson and others 2005). It has been recruited and deployed in many disparate developmental contexts during animal evolution, but the Hox cluster is associated with the development of the central nervous system in many different taxa (Lumsden and Krumlauf 1996; Irvine and Martindale 2000; Schilling and Knight 2001; Hinman and others 2003). In vertebrates, the patterns of hox gene expression are linked with the generation of rhombomeres and neural crest in the hindbrain, and make up part of a combinatorial code for the specification of rhombomere identity (Lumsden and Krumlauf 1996; Schilling and Knight 2001). Work from urochordates and amphioxus has demonstrated nested hox expression domains in the developing nervous system (Gionti and others 1998; Wada and others 1999; Schilling and Knight 2001; Seo and others 2004; Holland 2005; Keys and others 2005). It has been proposed that the ancestral role of hox genes in chordates is in the A/P regionalization of the central nervous system (Wada and others 1999; Schilling and Knight 2001). However, in both urochordates and amphioxus, colinear expression of hox genes are not only restricted to central nervous system but also to the epidermis (Seo and others 2004; Holland 2005; Keys and others 2005; Schubert and others 2005). For example, in amphioxus, hox1, 3, 4, and 6 are all expressed in nested domains in the central nervous system, and also in corresponding domains in the general ectoderm (Holland 2005). In the ascidian Ciona intestinalis, a screen for cis regulatory elements of hox genes uncovered several regulatory elements that drive epidermal expression of hox genes in a colinear fashion (Keys and others 2005), and expression of hox1 and hox11/12 was detected in the ectoderm by in situ hybridization. Even in vertebrates early ectodermal expression of hox genes is not always restricted to neurectoderm (Kolm and Sive 1995), although the role of hox genes in early vertebrate ectodermal patterning has received little attention (Holland 2005). Keys and colleagues (2005) discussed the ancestral role of hox genes in the ectoderm was in patterning both epidermis and neurectoderm, similar to the role of hox genes in arthropod development. The hox expression data from S. kowalevskii in circumferential rings, in a finely mixed ectodermal population of both neural and epidermal cells, is consistent with the hypothesis that the ancestral role of hox genes in deuterostomes, and perhaps even in bilaterians, was in patterning ectoderm rather than neurectoderm in particular.
The echinoderms, as the sister group to hemichordates, are a critical node for testing hypotheses of the roles of hox genes in deuterostomes. However, the hox expression data are limited, and no complete expression study has been carried out during the development of the adult or the larval body plan. The most comprehensive data are from 2 different sea urchins with contrasting life history strategies; the indirect-developing sea urchin S. purpuratus and the direct-developing sea urchin Holopneustes purpurescens. The results from these 2 species are quite different: in the indirect-developing species most of the hox genes are not expressed during embryogenesis and larval stages, but hox7, hox8, hox9/10, hox11/13a,b are initiated in a colinear pattern in the posterior, mesodermally derived coeloms during the establishment of the adult radial body plan. A limited number of hox genes are detected in the ectoderm, where neural patterning would be expected to occur, but never in a colinear fashion during the development of the adult or the larva (Arenas-Mena and others 2000; Peterson and others 2000; Popodi and Raff 2001). These data contrast with what has been found in the direct-developing species: 2 hox genes are expressed in the ectoderm of the juvenile urchin along the oral/aboral axis during the development of the adult body plan (Morris and Byrne 2005), but not in the mesoderm as found in S. purpuratus. Otx, which is normally expressed anterior to hox genes in most bilaterians, is expressed more orally in the radial nerves and circumoral nerve ring, and the 2 hox genes in the study, hox5 and hox11/13, are expressed more caudally, away from the mouth. Although Morris and Byrne (2005) agree with the conclusions of Arenas-Mena and colleagues (2000), it is difficult to resolve the differences in expression between the 2 datasets. Clearly, more information is needed from echinoderms to establish the basal role of hox genes in the echinoderms and also to help resolve the ancestral role of Hox proteins in the deuterotomes.
Hox genes and proposed homology of central nervous systems
Generally, the correlated expression domains in the antero/posterior of many key developmental transcription factors in S. kowalevskii are very similar to those in vertebrates and Drosophila (Lowe and others 2003). The expression of the hox genes in ectodermal rings during the development of the hemichordate basiepithelial nervous system has implications for our understanding of the evolution of deuterostome nervous systems. These data also raise questions as to whether body patterning genes are useful characters for reconstructing ancestral neural anatomies of distantly related groups of animals. Do gene expression domains allow us to test hypotheses of morphological homology between hemichordates and other bilaterians? Molecular similarities, such as hox gene expression conservation, have been used to support the hypothesis that the ancestor of protostomes and deuterostomes possessed a complex and centralized nervous system (Hirth and others 2003; Lichtneckert and Reichert 2005). However, the correlated expression domains of hox genes, and the more anteriorly localized suite of genes (such as otx) (Lowe and others 2003), in broad domains of the ectoderm of hemichordates, raise the possibility that this group of transcription factors acts as a conserved regulatory suite that patterns all bilaterian nervous systems, irrespective of the specific morphologies. The similarities of hox genes between arthropods and chordates may therefore be insufficient to prove homology of the arthropod ventral nerve cord and the chordate dorsal nerve cord. These expression studies from S. kowalevskii establish that conserved patterning genes, such as the hox genes, were not uniquely associated with central nervous system development during the course of evolution. While it also possible that hemichordates lost centralization from a deuterostome ancestor with a central nervous system, correlated suites of developmental genes expressed in similar topologies, between distantly related groups are clearly deployed in the patterning of diverse types of neural architectures. Caution should therefore be used when arguing for hypotheses of morphological homology of specific neuroanatomies based on the deployment of this suite of genes.
Clearly, the similarities in gene expression between vertebrates and hemichordates do not indicate any sort of morphological homology between regions of the vertebrate brain and the basiepithelial plexus of the hemichordates (Lowe and others 2003). Despite the similarities of gene expression, we have no grounds for speaking of the forebrain, midbrain or hindbrain of the hemichordate. This leads to the issue of what exactly these data allow us to compare between groups, and what they tell us about the early evolution of the chordate nervous system. We argue that these data provide, for the first time, a transcriptional map from which to directly compare the body plans of vertebrates and hemichordates. This map provides some critical insights into what molecular patterning elements were used to regionalize the deuterostome ancestor that gave rise to the chordates. While the details of the ancestral morphology of the nervous system (centralized or diffuse) are difficult to reconstruct from these kinds of data, the ancestor clearly used this map of genes to pattern its A/P axis, and likely its nervous system.
Functional analyses will be necessary in S. kowalevskii to fully test the role of hox genes in ectodermal patterning, and to establish whether hox genes are predominantly involved in neural patterning or general ectodermal patterning. Further comparative work within the deuterostomes will help clarify the role of hox genes in axial patterning within this group. Progress will come from both genomics and a greater phylogenetic sampling of deuterostome taxa for developmental studies. A better understanding of the role of hox genes in echinoderm development will be particularly informative.
We thank John Gerhart and Marc Kirschner for continued support and encouragement; Eric Lander for EST sequencing; Chris Gruber of Express Genomics for the production of cDNA libraries; the staff of Marine Biological Laboratory for help in collection and support; Bob Freeman for Bioinformatics support, Mike Wu and Ellyn Farelly for providing valuable technical help, and Imogen Hurley and John Gerhart for providing comments on the manuscript. We would also like to thank one of the anonymous reviewers for constructive and valuable comments on the manuscript. This work was supported by a Colwin Fellowship (to CJL) from the Woods Hole Marine Biological Laboratory, a NSERC predoctoral grant (to JA) and by USPHS grants HD37277 to M. Kirschner and HD42724 to J. Gerhart.