The innate immune repertoire in Cnidaria - ancestral complexity and stochastic gene loss

Analysis of genomic resources available for cnidarians revealed that several key components of the vertebrate innate immune repertoire are present in representatives of the basal cnidarian class Anthozoa, but are missing in Hydra, a member of the class Hydrozoa, indicating ancient origins for many components of the innate immune system.


Background
The innate immune system is the first line of defense against pathogens, and in non-chordates is assumed to be the sole means by which any non-self cells are detected and either killed or contained [1]. Innate immunity in vertebrates is essentially a two-tier system consisting on one hand of phagocyte activation by the interaction of specialized surface receptors with pathogens or pathogen-derived components, and on the other of the direct opsonization and lysis of pathogens via the complement cascade. Whilst the vertebrate innate immune system has been the subject of intense investigation and is relatively well understood, studies of invertebrate immunity, which have focused primarily on the arthropods Drosophila and various horseshoe crab species [2][3][4], have revealed some striking similarities. For example, in both Drosophila and vertebrates, the Toll/Toll-like receptor (TLR) mediates the activation of appropriate response genes to microbial challenge [5,6].
Toll and the TLRs are transmembrane proteins with a characteristic domain structure consisting of an extracellular amino-terminal domain containing leucine-rich repeats (LRRs) responsible for pattern recognition and an intracellular Toll interleukin receptor (TIR) domain that mediates signal transmission. Although the Toll and TLR families of arthropods and mammals are thought to have independently diversified [7,8], all Tolls and TLRs signal via a common pathway that is conserved between Drosophila and mammals. The ultimate step in this pathway is translocation of nuclear factor (NF)-κB or its fly counterpart (the Dif/Rel heterodimer) into the nucleus, where it stimulates transcription of appropriate response genes. The immune repertoire of the horseshoe crab Carcinoscorpius includes a complex complement pathway that has both opsonic and lytic effector functions [9]. Horseshoe crab complement C3 is functionally homologous with mammalian C3, mediating phagocytosis of bacteria (by hemocytes) in a strikingly similar manner.
Whilst these specific studies imply that at least some innate immune mechanisms have been conserved, broader comparative studies highlight the extent of gene loss and divergence in various metazoan lineages. For example, although Carcinoscorpius clearly uses a vertebrate-like complement system, none of the central components of the cascade (C2, C3, C4, C5) are encoded by the genomes of the ecdysozoans Drosophila, Caenorhabditis or Anopheles. Moreover, the sole Toll/TLR in Caenorhabditis elegans and C. brigssae is not known to function in the context of immunity, nor does that reported in the horseshoe crab Tachypleus tridentatus [10]. There are also important differences between the Toll/TLR systems of Drosophila and mammals. For example, some mammalian TLRs themselves act as pattern recognition receptors (PRRs) upon microbial challenge, whereas in fly this is not the case [11]. Moreover, whereas most of the ten or so vertebrate TLRs function primarily in immunity, only one of the nine fly (and ten mosquito) Tolls functions in this con-text. The others play a role in development [10], most famously in controlling differentiation in the dorsal/ventral axis.
The significance of gene loss in animal evolution has recently been brought into focus by preliminary expressed sequence tag (EST) and genomic analyses of some 'basal' animals ( Figure 1), particularly the anthozoan cnidarians Acropora millepora and Nematostella vectensis [12,13] and the planarian Dugesia japonica [14]. Paradoxically, the genomes of these morphologically simple animals contain many genes previously thought to have evolved much later in the context of vertebrate complexity, and most of the complexity of signaling pathways and transcription factors associated with higher animals is represented in the anthozoan datasets [13,[15][16][17]. In contrast to Drosophila and Caenorhabditis, which have undergone substantial gene loss, for at least some groups of genes Acropora and Nematostella appear to have preserved much of the genetic complexity of the common metazoan ancestor. For example, whereas fly and worm have each lost approximately half of the ancestral Wnt complement, all but one of the 12 known Wnt subfamilies is represented in Nematostella [15]. The emerging cnidarian EST and genomic datasets are, therefore, potentially highly informative with respect to the ancestral immunological repertoire. In addition to this basic evolutionary significance, understanding the bases of cnidarian immunity is of major applied significance in light of the dramatic decline in coral health that has occurred on a global scale over the past 20 years. Increasing human activity in coastal zones throughout the world has led to declines in water quality with increased sediment, nutrient and heavy metal concentrations. These have all had detrimental effects on corals with an associated increase in the prevalence of disease, and perhaps led to some new diseases, although this is less certain, as coral diseases are very poorly understood. Most are named for their symptoms, for example, 'white band disease', 'black band disease', and 'rapid wasting syndrome' and causative agents are frequently unknown. In the face of this uncertainty it is important that coral defense mechanisms should be better understood.
Although cnidarians have no specialized immune cells, at least some display highly specific allorecognition characteristics. Allorecognition, xenorecognition, and killing mechanisms have been demonstrated in several hydrozoans [18][19][20] and anthozoans [21][22][23]. Allorecognition is thought to protect colonial cnidarians from fusion with genetically different individuals and to prevent germ line parasitism. The effector mechanisms range from contact avoidance involving chemical sensing, to barrier formation, or usage of nematocysts. For example, the sea anemone Anthopleura xanthogrammmica will 'tolerate' adjacent clonal individuals, but will attempt to 'reject' heterogenic clones with which it comes into contact [24,25]. In the anthozoans, Stylophora pistillata and Montipora verrucosa branches within one colony will readily fuse while branches of genetically different individuals never undergo fusion [22,26,27]. Fusion of two conspecific individuals is occasionally referred to as 'natural transplantation'. Observations in sea anemones (Anthopeura elegantissima, Phymactis clematis) and gorgonians (Eunicella stricta) indicate that individual colonies possess unique sets of histocompatibility elements, which are recognized as nonself by all other conspecific colonies [23,28]. In Hydrozoa, the same phenomenon was reported for Millepora dichotoma [29] and studied in great detail in the colonial marine hydroid Hydractinia echinata [30]. Hydractinia in fact was among the first invertebrates shown to display a genetically based system of intolerance against allogenic tissue: for more than 50 years it has been known that allorecognition and the inability to fuse stolons of different colonies is under the control of one polymorphic gene [31,32]. Recent efforts using defined genetic lines of the hydroid Hydractinia symbiolongicarpus have confirmed this and shown that the single chromosomal region contains at least two loci [33].
The availability of whole genome sequences of two basal cnidarians at respectably high levels of redundancy -the hydrozoan Hydra magnipapillata (>6-fold coverage) and the anthozoan Nematostella vectensis (>7.6-fold coverage [17])together with large-scale EST datasets for these and for the coral Acropora millepora potentially offer new perspectives on the origins of mammalian immune functions. Here we report the results of a screen of the available genomic and EST resources for the cnidarian counterparts of key components of the vertebrate innate immune repertoire.

The toll receptor and other proteins containing the TIR domain
Searching the Hydra predicted protein collection using the available hidden Markov models (HMMs) identified only four TIR domain-containing proteins, two of which are clearly related to MyD88, which functions downstream of TLRs (Table 1). Consistent with their assignment as MyD88 family members, both of these Hydra proteins also contain the characteristic DEATH domain. The two other Hydra TIR proteins are atypical transmembrane proteins in having relatively short extracellular domains that are devoid of the LRR domains that characterize Toll and the TLRs (Figure 2). cDNAs encoding these proteins have previously been isolated by the Bosch laboratory (unpublished data) and their functions are presently under investigation; these proteins are known as HyTRR-1 and HyTRR-2. Surprisingly, extensive searching of the Hydra genome and all available EST/cDNA resources failed to identify any proteins having the canonical Toll/TLR structure, characterized by possession of both LRR and TIR domains.
Whereas only four TIR proteins are present in Hydra, substantially more could be identified amongst the predicted proteins from Nematostella using HMM-based search methods. Five of them were sufficiently complete to be included in the analyses presented here. These include a single MyD88 homolog (NvMyD88) and a protein (NvTLR-1) clearly related to members of the Toll/TLR family ( Figure 2). Whereas the mammalian TLRs, and some members of the fly Toll/TLR family, have only a carboxy-terminal cysteine-rich motif flanking the LRRs proximal to the membrane, Nematostella NvTLR-1 is predicted to contain both carboxy-and aminoterminal-flanking cysteine-rich motifs in the extracellular part of the protein ( Figure 2). This suggests that fly and anemone Toll more closely reflect the ancestral domain structure than do the mammalian TLRs. Moreover, a phylogenetic analysis ( Figure 3) groups the TIR in Nematostella NvTLR-1 with its fly and human counterparts, with strong bootstrap support. Surprisingly, three more of the predicted Nematostella TIR proteins also contain multiple immunoglobulin (Ig) domains (Figure 2), and thus reflect the domain structure of mammalian interleukin 1 receptors (IL-1Rs). NvIL-1R1 and NvIL-1R2 each contain three Ig domains, and NvIL-1R3 contains two predicted Ig domains ( Figure 2) but may be incomplete. In the phylogenetic analysis based on TIR domains the Nematostella IL-1R-like proteins form a clade distinct from both the MyD88 and Toll/TLR types (Figure 3), although these cnidarian TIRs appear to be distinct from those in the vertebrate IL-1Rs (data not shown). Several other TIR proteins were detected amongst the sequences of Nematostella (Additional data file 1), but were not subjected to further analysis as the TIR domains were incomplete or the sequences were judged likely to be artifactual. Two complete TIRs were identified by searching the available coral datasets. The trace archive yielded one TIR from Acropora palmata (ApGe-  Other TLR related proteins

Overview of innate immunity components present or absent in selected Cnidaria
nomic) and a second was encoded by an A. millepora EST (AmTIR-1). These two coral TIRs are most similar to those in the Nematostella IL-1R-like proteins ( Figure 3), but no linked domains have yet been identified in these cases.
The Müller group recently reported the identification of MyD88 in a demosponge, Suberites domuncula [34]. However, whilst phylogenetic analyses clearly grouped the TIR in this sponge sequence with those present in unambiguous MyD88 orthologs (Figure 3), domain searching indicates that the predicted sponge protein may not have a functional DEATH domain.

The Toll/TLR pathway is ancestral but some components are missing or highly divergent in Hydra
Most of the intracellular mediators of Toll/TLR signaling could be identified in Nematostella and Acropora, but some key components appear to have either been lost or diverged beyond recognition in Hydra ( Table 1). The absence of a Toll/ TLR protein sensu stricto from Hydra is discussed above, but in addition only a single highly derived Rel domain could be found in Hydra whereas unambiguous NF-κB homologs are present in both Nematostella and Acropora (Table 1). In addition to the pathway leading to nuclear localization of NF-κB, Toll/TLR signaling can activate the Jun N-terminal kinase (JNK) and p38 mitogen-activated protein kinase (MAPK) pathways, leading to transcription of a range of target genes via the AP1 (activating protein 1)/ATF (activating transcriptionfactor 1) factors. Toll/TLR signaling via JNK/ MAPK requires the participation of the ECSIT (evolutionarily conserved signaling intermediate in Toll pathways) adaptor protein [35], which also provides a link between the Toll/TLR and TGF-b (transforming growth factor-beta)/BMP (bone morphogenic protein) pathways [36]. The presence of ECSIT as well as the key components of the JNK/MAPK pathway in the cnidarian datasets (Table 1, Figure 4) indicates an early origin for this variant of Toll/TLR signaling. The discovery of a conserved predicted ECSIT coding sequence in the fresh water sponge Ephydatia fluviatilis (N Funayama, personal observation) additionally supports this view.
Wiens et al. [34] have suggested that a sponge-specific cell surface protein known as SLIP (sponge LPS-interacting protein) functions as a pattern recognition receptor and effectively substitutes for Toll/TLR in antimicrobial defence. The sponge MyD88-related protein nominally functions downstream of SLIP in the proposed pathway [34]. In support of the idea that sponges lack Toll/TLRs, the authors cite a lack of TLRs in a screen of 15,000 ESTs. As sponge 'MyD88' and SLIP are co-immunoprecipitated by the reciprocal antibodies [34], they clearly can interact in vitro. Some TLR pathway components could also be identified in the available sponge data, but the equivocal status of the sponge MyD88 related protein means that it is unclear at this time whether sponges have a canonical Toll/TLR pathway.

Cnidarian complement C3 and related proteins
The complement component C3 has recently been reported in another anthozoan cnidarian, the octocoral Swiftia [37], and the corresponding gene has recently been cloned from Acropora (Hayward, unpublished data). The Acropora C3 (C3-Am) gene is first expressed strongly in the endoderm of the planula as it elongates following gastrulation (Figure 5a). The endodermal expression is not uniform, being most intense in a subset of dark staining cells that have not yet been characterized. As the planula elongates expression becomes somewhat weaker, with the strongest expression localized to the aboral endoderm (Figure 5b). Post-settlement (Figure 5c-e) expression is limited to the endoderm and is particularly strong in the endoderm of the polyp as it rises from the calcifying platform at its base (for example, Figure 5d).
C3 has a complex domain structure. Whilst anthozoan C3s resemble their deuterostome counterparts both in domain structure ( Figure 5f) and sequence, not only could no corresponding gene be identified in Hydra, but also some of the domains characteristic of C3 (ANATO, C345C; Figure 5f) could not be detected in any Hydra protein. Although lacking a canonical C3, Hydra contains a gene encoding A2M related domains. Interestingly, in situ hybridization in Hydra using a probe covering these typical A2M-related domains (Figure 5f; A2M-comp/A2M-recep) showed expression restricted to the endodermal epithelium (Figure 5g), as was the case with Acropora C3.

MAC/PF domain containing proteins in Cnidaria
Searching for other components of the complement cascade, we identified proteins containing a membrane attack complex/perforin domain (MAC/PF) similar to that present in complement component C6 and related proteins. HMM searching identified just two MAC/PF domain-containing proteins in Hydra (Table 1), whereas four proteins were identified in Nematostella. Two MAC/PF proteins were also identified amongst the Acropora ESTs. Database searches and analyses of predicted domain structures revealed that most of the cnidarian MAC/PF sequences are likely to fall into three Phylogenetic analysis of cnidarian TIR sequences in comparison to a selection of TIR domains from other species   Figure 5h).
TBlastN-based searches of the Nematostella genome identified a gene matching strongly to the human macrophage expressed protein 1 (MPEG1; GenBank:XP_166227) and its abalone homolog abMPEG1 (GenBank:AAR82936) [38]. A clearly related gene in S. domuncula has recently been implicated as an effector in a hypothetical sponge innate immune defence pathway [34]. Recombinant Suberites MPEG has anti-bacterial activity against Gram-negative bacteria, and is up-regulated after lipopolysaccharide (LPS) treatment [34]. The MPEG1 family clearly has an ancient evolutionary history (the sponge and human sequences have 28% identity and 46% similarity) [34] but only in Suberites has any functional characterization been done. Despite the presence of MPEG1 in the sponge and an anthozoan, no corresponding gene could be identified in Hydra.
The nematocyst venom of at least some anthozoans contains the protein TX-60A [39], and two of the Nematostella MAC/ PF proteins and one of the Acropora ESTs clearly correspond to this protein type (Table 1). TX-60A has an epidermal growth factor (EGF) domain immediately carboxy-terminal of the MAC/PF domain. In Hydra, this domain structure can be found in Hy-MAC, one of the two Hydra MAC/PF proteins (Figure 5h, Table 1). However, it is unclear whether the Hydra and anthozoan sequences are orthologous, as overall sequence identity is low. In situ hybridization analysis shows that expression of Hy-MAC is restricted to gland cells that are interspersed throughout the endoderm of Hydra (Figure 5i). Since endodermal gland cells and nematocysts are terminally differentiated [40], this pattern of expression is not easy to reconcile with a common function for the venom TX-60A and Hy-MAC.

Apextrin, a gene lost from Nematostella
The third class of cnidarian MAC/PF proteins represented in the Hydra and Acropora ESTs (Figure 5h) contains no identifiable domains other than MAC/PF. These proteins have moderate overall similarity to the echinoderm apextrins [41,42] and to the apicomplexan protein family to which the Plasmodium membrane attack ookinete protein (MAOP) [43] belongs. MAOP is responsible for rupture of epithelial cells in the insect host by the ookinete stage of the parasite. Surprisingly, apextrin seems to be a case of gene loss from Nematostella as, despite clearly related genes being present in Hydra and Acropora, extensive searching of both the predicted protein collection and the anemone genome using a variety of tools failed to identify an apextrin-related gene ( Table 1).
To explore the significance of this case of apparent gene loss, the expression pattern of an apextrin gene was examined by in situ hybridization in Acropora (Figure 5k-o). Apextrin-Am expression first appears in scattered ectodermal cells at the oral (blastopore) end of the embryo as it begins to elongate following blastopore closure (Figure 5k). As the embryo continues to elongate expression increases in intensity and the zone of expression spreads toward the aboral end, initially still in scattered cells (Figure 5l), but as elongation continues apparently in all ectodermal cells (Figure 5m), as is clear in transverse section (Figure 5n). As the planula settles, expression becomes less obvious at the oral end, eventually becoming limited to a belt marking the transition between the tissue that was formerly aboral (bottom), and the end bearing the oral pore, which subsequently forms the mouth of the polyp (Figure 5o). These expression data suggest that the primary function of apextrin-Am is in the ectoderm leading up to metamorphosis; hence, secondary loss of the corresponding gene from Nematostella may be explicable in terms of the very different modes of development of these two animals (see Discussion).
In adult Hydra, whole mount in situ hybridization showed an expression of the apextrin-like gene in groups of ectodermal cells arising from the interstitial cell lineage (Figure 5j), which may reflect a possible functional shift in an organism where metamorphosis is absent.
In addition to the above, the Nematostella dataset yielded MAC/PF domain-containing proteins having high similarity to the neural cell adhesion molecule spondin-1 (Additional data file 1).

Discussion
These preliminary analyses of the newly available genomic and EST datasets indicate that a surprising number of key components of the innate immune system, including the Toll/ TLR pathway and some complement cascade components, were in place at the base of the Eumetazoa. Also represented in the datasets are proteins strongly matching both the tumor necrosis factor (TNF) and Nod-like receptors and with the same domain structures (data not shown).
The analyses presented here are consistent with the idea of a genetically complex common metazoan ancestor [12,13]; clearly the Toll/TLR, MyD88 and IL-1R protein families were distinct prior to the divergence of the Cnidaria from the Bilateria, and a Toll/TLR pathway may predate even the Porifera/ Eumetazoa split. The discovery of proteins with the same domain structure as the IL-1R in Nematostella indicates that this receptor type predates chordate origins and that its original ligands may not have been interleukins. However, as the TIR domains in the cnidarian IL-1R-like and mammalian IL-1R proteins are divergent, separate evolutionary origins cannot yet be ruled out despite the similarities at the level of domain structure.
It is also likely that a prototypic complement effector pathway involving C3 and multiple MAC/PF proteins was in place in Signaling pathways downstream of the Toll/TLRs Figure 4 Signaling pathways downstream of the Toll/TLRs. Pattern recognition, either indirectly or directly, by Toll/TLRs results in activation of NF-κB (vertebrates) or the Dif/Rel heterodimer (Drosophila) and thus transcription of appropriate immune response genes. At TRAF6, the classic Toll/TIR pathway (shown in the right branch) is linked to the JNK/p38 pathway (shown in the left branch) by the ECSIT protein, which acts as a regulator of MEKK-1 processing [35]. Components of both pathways downstream of Toll/TLRs are represented in the cnidarian datasets (Table 1). ECSIT may also act as a link between these and the TGF-b signaling pathway, since it forms complexes with BMP-pathway restricted Smads and is essential for regulation of the BMP-target gene Tlx2 [36]. All of the components of the TGF-b signaling pathway are also known from anthozoan cnidarians [13].  Ureumetazoa. Comparisons of the TIR and MAC/PF complements of Hydra and Nematostella highlight the likely extent of gene loss and sequence divergence in the former -not only does Hydra appear to have lost the Toll-receptor, but NF-κB has also either been lost or has diverged beyond recognition. Moreover, only a few TIR proteins could be identified in Hydra compared to the rich representation of these in the anthozoans. In addition, Hydra appears to have lost a number of MAC/PF proteins and lacks an equivalent of the ancestral complement C3 protein, implying that although anthozoans most likely have a prototype complement effector pathway, this has undergone degeneration in Hydra. Hydra is an efficient producer of a number of potent antimicrobial peptides (Bosch, unpublished data), indicating that at least some of the functions of the complement effector system may have been subsumed by pathways involved in synthesis of these peptides.
The observation that genes of two of the three classes of cnidarian immune-repertoire genes examined here are expressed in the endoderm ( Figure 2) was unexpected, but is consistent with the primary sites of expression of such genes in many animals. In Drosophila, the fat body (an endodermal derivative, and the functional homolog of the mammalian liver) is the predominant source of the antimicrobial peptides whose synthesis is under the control of the Toll and Imd pathways [44] and in Caenorhabditis the gut is the primary source of antimicrobials [45,46]. Although mammalian skin contains immune sensory cells (Langerhans cells, dendritic cells and so on), these are also endodermal derivatives, hence there is a deep evolutionary link between the immune system and the endoderm.
Other specific cases of gene loss from Hydra have been documented [47] (Seneca, unpublished data), hence there are precedents for the apparent loss of immune repertoire components reported here. More surprising is the implication that gene losses may also have occurred in Nematostella, as evidenced by the apparent absence of the apextrin gene that is present in both Hydra and Acropora. This loss may be associated with differences between Nematostella and Acropora during the planula to polyp transition. Metamorphosis from the motile planula to the sessile polyp involves dramatic tissue remodeling [48] and gene expression changes in Acropora (Grasso, unpublished data), and the massive up-regulation of apextrin leading up to and during metamorphosis (Figure 5k-o) is consistent with a role in this process. It is worthy of note that apextrin is expressed in the tissue that is least remodeled at the time of metamorphosis. By contrast, it appears that there are no dramatic events comparable to Acropora metamorphosis occurring during this same period of Nematostella development; rather, Nematostella undergoes a simple and continuous transition from planula to polyp and appears to show some neotenous features, continuing to glide over the bottom aboral end first for some time after metamorphosis [49], and never forming a pedal disc [50]. Thus, loss of this gene may be related to the different modes of development of these two animals. This comparison between two anthozoans with different modes of development is clearly reminiscent of the original identification of apextrin as a gene specifically expressed in the 'direct' developing seaurchin Heliocidaris erythrogramma during metamorphosis, but which is not expressed during the development of H. tuberculata, a congeneric 'indirect' developer [41,42]. Because indirect development is considered to be ancestral in sea urchins, these authors hypothesized that apextrin had been coopted to function in metamorphosis in H. erythrogramma. The specific ectodermal expression of homologous proteins during metamorphosis in a cnidarian and an echinoderm is either a remarkable example of convergence or reflects conservation of function at some level. This latter possibility carries with it the implication that the 'simple' continuous planula to polyp transition seen in Nematostella might represent a derived state, and that the more dramatic metamorphosis seen in Acropora might more closely reflect the ancestral condition.

Conclusion
An important general implication that flows from these data is that gene loss may occur stochastically. If the genes in a pathway function only in that pathway, then one might expect that entire pathways would disappear following loss of one key component. However, the Hydra data appear to contradict this; most of the intracellular intermediates of Toll/TLR signaling are present despite loss of all of the major receptor types (both Toll/TLR and the IL-1R types known from more basal cnidarians are apparently absent from Hydra). Moreover, simple comparisons between related animals (for example, Nematostella versus Acropora) are unlikely to be informative in terms of understanding the origins of genes [50] -loss of a gene from Nematostella and loss of a specific function in the indirect developing urchin seems more likely than the independent co-option of the same protein to related roles in Acropora and a direct developing urchin. Stochastic loss underlying the distribution patterns of genes across the Metazoa may account for some cases of assumed lateral gene transfer -chance retention of an ancestral gene in one or a few animal lineages might easily be confused with lateral transfer.
Reconstructing the genome of the common animal ancestor will not, therefore, be a simple undertaking; it will require whole genome data for a wide range of 'lower' as well as 'higher' animals.

Datasets
Genomic and EST sequence data were downloaded from public databases at NCBI (dbEST, Trace archive) or originate from private investigators (D Miller, K Agata) and were stored on a local comparative genomics analysis platform [51].  36,820 ESTs from the fresh water sponge E. fluviatilis is maintained in the lab of K Agata. For the corals A. millepora, A. palmata and P. lobata only pilot genomic sequencing data are available to date. Genomic trace data were maintained in the original raw format, whereas ESTs were clustered and assembled using TGICL tools from TIGR [52]. The ESTscan software [53] was used to infer unigene and predicted peptide sequences from the assembled ESTs.

Database search, sequence analysis and phylogenetic methods
For database searches a local Blast-platform [51] and the public Blast platform at NCBI [54] were used. Domain searches were performed using SMART [55] and a local install of HMMer [56]. Profile HMMs for the investigated domains were obtained from PFAM [57] and Superfamily [58] databases. Seqtools [59] and BioEdit [60] were used for general sequence analysis. Protein sequence alignments were created using ClustalW [61] and the HMMalign script included in the HMMer package. Maximum likelihood phylogenetic analyses were undertaken using MolPhy version 2.3 [62] using the Dayhoff substitution matrix and local rearrangement search mode.

In situ hybridization
Whole mount in situ hybridization on Hydra polyps was performed as previously described by Grens et al. [63]. Acropora embryos were fixed and processed for in situ hybridization as described in de Jong et al. [64] following the detailed in situ protocol given in Hayward et al. [65]. For all genes shown, sense controls gave no hybridization signals.

Additional data files
The following additional data are available with the online version of this paper. Additional data file lists accession numbers for additional sequences identified within the database searches that were not further characterized in the present study.
Additional data file 1 Accession numbers for additional sequences identified within the database searches that were not further characterized in the present study Accession numbers for additional sequences identified within the database searches that were not further characterized in the present study Click here for file