Comparison of the Acropora and Nematostella Sox complements
Each of the six Acropora Sox genes reported has a clear counterpart in Nematostella, and these pairs of genes are probable orthologs. A total of 14 Sox sequences has been reported from Nematostella , so presumably more Acropora genes will be discovered as a result of continuing transcript characterisation using 454 sequencing. However, most of the Nematostella genes without Acropora counterparts (NvSoxA, NvSoxE2, NvSoxF2, NvSoxJ, NvSox4, NvSox5) were predicted from genomic sequence data so their existence as functional transcripts remains unproven. In our analyses (Fig. 1), three of the Nematostella Sox genes (NvSox1, NvSox2, NvSox3) that were unassigned in the Magie et al. study  fell clearly within the SoxB clade. Factors contributing to these differences, and the significantly higher resolution in our analysis, include the deliberate exclusion of highly divergent bilaterian sequences and the use of only complete HMG domain sequences in our case.
Anthozoan Sox genes and the evolutionary divergence of the Sox family
Sox genes have previously been reported from both cnidarians and sponges [6–9] but, as a number of these surveys have been based on PCR screens or scanning genomic resources, pseudogenes may have been included and in some cases assignments have been based on incomplete HMG domain sequences. The analyses presented here are based on complete HMG domain data derived from cDNAs, overcoming previous limitations. In terms of the representation of the ten (A-J) recognized classes of Sox genes, our studies confirm the presence of group B, C, E and F Sox types in cnidarians, and are consistent with the absence of group D, G, H, and I as reported by Magie et al. . However, in our analyses Nematostella genes previously assigned to groups A and J fell into group B (Additional File 9) in agreement with . The strong bootstrap support (≥ 90%) for monophyly of Sox groups B, E and F in our analyses (Fig. 1) indicates that each of these (and probably also the SoxC type) were distinct in the common ancestor of cnidarians and bilaterians, and the assignment of sponge genes to the Sox B, F, and C classes in our analyses implies that these classes had already diverged in Urmetazoa (the common ancestor of all animals), consistent with . Hence our analysis is broadly consistent with that in a recent study of the Sox gene complement of the sponge Amphimedon .
According to our analyses none of the cnidarian Sox sequences fall into subgroup B2, the distinctness of which is well supported (79%) by bootstrap probability, and the relationship of the non-bilaterian sequences to subgroup B1 is not simple (Fig. 1). It appears that Sox group B2 of the Bilateria may have arisen from a B1-like precursor after the Cnidaria/Bilateria divergence. However, the extent to which the phylogeny might be biased by the presence of an extra amino acid residue in two of the Acropora HMG domains (and their Nematostella counterparts) is not yet clear. Therefore, although the tree presented here (Fig. 1) contradicts the conclusions of Larroux et al.  with respect to divergence within the Sox B class, relationships between the SoxB1 and SoxB2 types remain unresolved. The additional amino acid residue in the HMG domains of two Acropora Sox proteins and their likely Nematostella orthologs, a lysine residue in the case of AmSoxBa and NvSoxB2 and an arginine residue in AmSoxBb and NvSox3, is thus far unique to these group B genes, and its origin may post-date the Cnidaria/Bilateria split.
In terms of HMG domain sequences, sizes and protein domain structures, the Acropora Sox proteins are strikingly similar to the vertebrate members of each of these Sox groups, whereas more differences are apparent in vertebrate/Drosophila and vertebrate/Caenorhabditis comparisons. This implies that, as is the case for some other genes , vertebrate and cnidarian Sox genes may more closely reflect ancestral characteristics than do their fly and worm counterparts. However, in some respects, the structures of the sponge Sox proteins  differ substantially from their cnidarian and bilaterian counterparts. The sponge sequences are more divergent in the HMG domains, lack other obvious conserved domains, and differ in overall size and position of the HMG domain. It is unclear, however, whether these differences reflect divergence within the sponge lineage or innovations within the cnidarian/bilaterian lineage potentially underpinning the transition to tissue level organization.
Expression patterns of Acropora Sox genes – glimpses of ancestral functions?
The restriction of the Acropora SoxBb and SoxB1 mRNAs to the presumptive ectoderm and SoxE1 to the presumptive endoderm during gastrulation is suggestive of roles for these genes in germ-layer specification. Although often referred to as neural markers , members of Sox group B are also important for germ layer formation and gastrulation in both vertebrates and invertebrates. Sox3 (a group B1 gene) regulates gastrulation and germ layer formation in both Xenopus and zebrafish . In the sea urchin, SoxB1 and B2 are expressed in the presumptive ectoderm during gastrulation and are necessary for gastrulation and vegetal development . In the hemichordate Saccoglossus kowalevskii, Sox1/2/3 (a SoxB gene) is expressed in the entire ectoderm of the gastrula embryo . These similarities suggest an ancestral function of group B Sox genes in germ layer specification. The significance of the AmSoxE1 expression pattern is more difficult to assess, as few early SoxE expression patterns have been reported. Roles for Sox9 (group E) genes in neural crest development [34–36] presumably reflect co-option.
Zygotic AmSoxE1 expression starts from the late prawnchip or early donut stage, and this gene marks the presumptive endoderm during gastrulation (Fig.3). Although this is suggestive of a role for AmSoxE1 in endoderm determination, there are no clear precedents for this. In the sea urchin, SoxE transcripts are localized in small micromere descendents at the tip of the archenteron during gastrulation , but this gene is not expressed in the blastula stage. However, few comparative expression patterns have been reported, and more general roles for SoxE genes in early tissue patterning cannot yet be ruled out. Although AmSoxF is not expressed until after gastrulation (Fig. 8), expression is limited to the endoderm, so it is possible that AmSoxF plays a role in the maintenance of endodermal identity.
Heterogeneity in early Sox expression patterns within the Anthozoa
Although Acropora and Nematostella are both members of the same subclass (Hexacorallia, or Zoantharia) within the cnidarian Class Anthozoa there are apparent differences in the expression patterns of presumably orthologous genes. Some of these differences may just reflect the more complete series of early developmental time points reported for Acropora, while others presumably result from fundamental differences in the overall developmental biology of the anemone and the coral (see [38, 39]).
The early development of Nematostella is now well documented up to blastopore closure, although some details of the mechanism of gastrulation remain equivocal [9, 40, 41, 42]. Later development is less well understood with the most complete description still being that of Hand and Uhlinger . Acropora development has been characterised much less thoroughly although it is clear that there are some dramatic differences between the two species [44, 45]. Firstly, it appears that the Acropora egg contains much more yolk than that of Nematostella, consistent with the frequently longer planktonic life of the former species. Secondly, the prawn-chip stage is more exaggerated in Acropora, in that many more cells are present at this stage, so that the embryo takes on the appearance of a warped dinner plate, rather than a small bowl. The morphology of the post-gastrulation larva in the two species is also quite different due to the large amount of yolk present in Acropora. Thus the Nematostella planula has a far more developed endoderm and structures such as the septa are apparent from shortly after blastopore closure. In contrast, the early Acropora planula has a poorly developed endoderm consisting of a thin layer of cells lying beneath the mesogloea plus small cells scattered among the large yolk cells that pack the cylindrical central axis of the larva. It is only late in planula development that the tightly packed core of yolk begins to thin at the oral end and septal development becomes apparent. These morphological differences mean that even when a gene is functioning in a similar manner its pattern of expression may appear somewhat different.
Magie et al.  comment that NvSox3 is the only Nematostella Sox gene that is highly expressed maternally. Its Acropora ortholog, AmSoxBb (Fig. 3G–L) is also expressed in the egg and in early stages of embryonic cell division. A second Acropora gene, AmSoxB1, is also maternal, as evidenced by detection of the mRNA from the earliest stages of development (Fig. 3A), and has a very similar pattern of expression to AmSoxBb. The earliest stage at which expression of NvSoxB1 (the ortholog of AmSoxB1) is shown by Magie et al.  is the gastrula, when its mRNA is restricted to the aboral ectoderm and a discrete area around the blastopore which will give rise to the pharynx. Sox E may be an example of the phenomenon noted above, where orthologs may function similarly in spite of initial expression patterns that appear quite different. Thus, early expression in Nematostella appears to be ectodermal while in Acropora it is clear that the expression will clearly become endodermal much earlier. However, by the end of gastrulation the expression is endodermal in both organisms so the way this expression is arrived at may not be functionally significant.
The cell-specific expression pattern of NvSoxC, the Nematostella ortholog of AmSoxC, in the ectoderm is broadly similar to that of the Acropora gene. However, there are obvious differences. Firstly, expression begins in Acropora well before the blastopore has closed while expression of NvSoxC begins later – expression of the anemone gene could not be detected during gastrulation. A second difference is that expression associated with the oral pore appears just after the blastopore has closed in Nematostella (Fig. 6), while in Acropora this is not seen prior to the planula stage. Finally, while NvSoxC is clearly expressed in developing tentacles in Nematostella (Fig. 6J–L), no such expression is seen in Acropora at comparable (i.e. later in development) stages of tentacular development (not shown).
The expression pattern of NvSoxC reported here (Fig. 6) has some striking similarities with that previously reported for another Nematostella Sox gene, the B type gene NvSox2 . Like NvSoxC, NvSox2 is expressed in a subset of ectodermal cells that are distributed in a scattered pattern early in post-gastrulation development but become restricted to the ends of the developing tentacles later; the similarity is most striking at the stage shown as Fig. 6H – compare this pattern for NvSoxC with Fig. 3ll in Magie et al.  for NvSox2. Unfortunately, direct comparisons are limited by the fact that no NvSox2 in situ data are available for the period corresponding to Fig. 6D–G. Note that the two Nematostella genes are on separate scaffolds in genome assembly v1.0 (JGI) so it is unlikely that they are tightly linked.
Direct comparisons of expression patterns later in development are complicated by the more extensive differences between Nematostella and Acropora (e.g. the incomplete metamorphosis of the anemone). Whilst AmSoxBa and NvSoxB2 have similar cell-specific expression patterns throughout gastrulation, in Acropora this ectodermal pattern persists through to the planula stage (Fig. 7), whereas NvSoxB2 is expressed in a cell-specific manner in both endoderm and ectoderm . Immediately prior to settlement, expression of AmSoxBa becomes restricted to the aboral half of the ectoderm (Fig. 7), but nothing like this axial restriction is seen during Nematostella development before tentacle formation.