We have previously characterized one qua, one hog-only, ten wrt, 17 grd, and 32 grl ORFs from C. elegans, three of which are pseudogenes 1938. Furthermore, we have identified 49 hh-related genes in the related nematode Caenorhabditis briggsae 38. We correct this number to 48 hh-related genes here, because C. briggsae wrt-8 is the same locus as wrt-4. To retrieve sequences from other species we used selected Hh, WRT, QUA, GRD and GRL protein sequences as queries for tblastn and blastp searches at Stellabase, the DOE Joint Genome Institute, The Genome Sequencing Center at the Washington University School of Medicine, The Institute for Genomic Research (TIGR), and NCBI (see Methods). The recovered sequences were aligned to sequences that we had assembled previously 181938. When obvious discrepancies in conserved regions were found in the newly retrieved ORFs, genomic sequences were inspected for additional or extraneous exons or alternative splice sites, and ESTs were examined for frameshifts. ORFs were corrected to optimize matches to existing motifs, and extraneous N-terminal residues were truncated when methionine residues followed by good N-terminal signal peptides for secretion were found. One caveat is that our ORF predictions from genomic sequences are still limited due to partial nature of the various contig assemblies. In some instances an ORF runs into an unsequenced region (e.g., B. malayi wrt-4). In the case of ESTs it was often possible to assemble several ESTs into contigs, but in most instances ORFs derived from ESTs lack either N-terminus and/or C-terminus. Considering also that the various genome projects are are in different states of completion, the nomenclature given here to the ORFs should be considered preliminary. After correction of the ORFs multiple sequence alignments of the different protein domains were made and used for phylogenetic analyses using Neighbor Joining and Maximum Likelihood. We also prepared protein sequence logos of the Hog domains of Hh and nematode Hh-related genes to aid with the analysis of more divergent Hog domains (Figure 1, Additional files 1, 2, 3). We extended the motif nomenclature of inteins by introducing motifs J, K, and L (Figure 1, Additional file 2). Motif J corresponds to motif G in inteins 34353637, however, because this motif is so distinct in Hog domains, we have ventured to give it its own name here. Motifs K and L are located in the SRR and are primarily found in Hog domains of Hh proteins (Figure 2, 3, 4, Additional file 1). In the Hog domains of nematode Hh-related proteins, these two regions show a number of differences compared to the Hh proteins (Figure 2, 3, 4, Additional files 1, 2, 3), and, as will be shown below, motifs K and L provide useful diagnostic functions for evaluating Hog domains.

The nematode B. malayi is a parasitic nematode that is one of the more distantly related members to C. elegans within the order of Rhabditiada 39. We have previously described a quahog gene, qua-1, in B. malayi 19, and obtained partial sequences from ESTs for two wrt, one grd and two grl genes 17. Here we retrieved a total of four wrt genes, one with a Hog domain (Bm wrt-6), two without a Hog domain (Bm wrt-10, Bm wrt-5/3) and one whose C-terminus is presently unknown (Bm wrt-4) (Table 1, Fig. 2, 3, 4, Additional file 4). Based on phylogenetic analyses of both the Hog domain and the Wart domain (Figure 5, 6, 7, Additional files 5, 6, 7), Bm wrt-10 is a clear orthologue of Ce wrt-10, wrt-5/3 is a co-orthologue of Ce wrt-5 and wrt-3, and – based primarily on the Hog domain – Bm wrt-6 is an orthologue of Ce wrt-6. The wrt-6 ORF encodes a full Wart domain, however the previously identified wrt-6 EST 17 lacks the C-terminal half of the Wart domain. Comparison with the genomic sequence revealed that this EST spans exons 1, 2, the first 10 nucleotides of exon 3 and continues then into exon 10, 11, and 12, which contain the Hog domain (data not shown). The point of discrepancy in exon 3 is not at a splice site, therefore this unusual EST might represent a cloning artifact. Bm wrt-4 cannot be assigned unequivocally as orthologue of Cb wrt-2 or Cb wrt-4, but it appears to group with them. Ce wrt-10 lies next to Ce wrt-1 on the chromosome 17, however the Bm wrt-10 contig is too small to determine, whether another wrt gene resides next to it.

One Ground domain gene was recovered from B. malayi, Bm grd-5, that is co-orthologous to Ce grd-5 and grd-10 (Figure 8, Additional file 8). We have previously identified a few grl genes from B. malayi in searches of ESTs 17. Here, thirteen grl genes were recovered from B. malayi, however, only a few could be identified as orthologues of C. elegans genes, i.e. grl-4, grl-16, and perhaps grl-7 and grl-17 (Figure 8, Additional file 8). Other grl genes have clearly duplicated within the B. malayi branch, e.g. Bm grl-x1, gr l-x2 and grl-3, which are more similar to each other than to other genes.

A few ESTs were retrieved from other Chromadorea nematodes: In Meloidogyne incognita we found a gene with similarity to wrt-6 (Figure 2, 3, 4, 5, 6), and one grl gene, Msp3, which is expressed in the esophagal gland cells 40 (Additional file 8). In Parastrongyloides trichosuri a gene with similarity to wrt genes was found (Additional file 7).

C. elegans and B. malayi belong to the class of Chromadorea. Our database searches revealed now also Hog-containing genes from the distantly related class of Enoplea nematodes, i.e. Xiphinema index CSEQDL01, and T. spiralis, both members of the Dorylaimia 39. From Xiphinema we retrieved ESTs for nine distinct genes, and from T. spiralis five (Table 1), one of which (Ts Xhog1) is also supported by ESTs. All five T. spirals ORFs have a signal peptide sequence for secretion, and although many of the Xiphinema ESTs are incomplete, in several instances methionine residues followed by good signal peptides could be found at the 5' of the ESTs (XC Thog, Shog1, Shog2) (Additional file 9). One gene from Xiphinema (XC hh) and one gene from T. spiralis (Ts hh) are clearly hh genes (Figure 2, 3, 4, 9, Additional file 10), since they both have a Hedge domain and a Hog domain. One gene from T. spiralis has a QUA domain upstream of the Hog domain (Ts qua-1). While its Qua domain is rather divergent, the cysteine residues are all conserved (Additional file 11). Ts Xhog3 has a rather short region upstream of the Hog domain, which cannot be extended, because it is delimited by an upstream cyclin gene, for which ESTs are available (data not shown). After cleavage of the signal peptide and subsequent autoprocessing through the Hog domain the predicted N-terminal peptide of Ts Xhog3 would only be 34 residues long. In Xiphinema the three ORFs with a signal peptide (XC Shog1, Shog2, and Thog) have rather short predicted N-terminal sequences as well. In the case of XC Shog1 it is only 15 residues long, in the case of XC Shog2 it is 25 residues, and in the case of XC Thog it is 79 residues long with an unusual stretch of about 70 residues almost entirely composed of threonine and serine residues. Two T. spiralis genes reside next to each other on the chromosome (Ts Xhog1 and Xhog2). They share sequence similarity upstream of the Hog domain with six conserved cysteine residues. In addition, XC Xhog5 also has sequence similarity to the upstream regions of Ts Xhog1 and Xhog 2 (Figure 10).

Apart from the similarities in the regions N-terminal to the Hog domain indicated above, the remaining N-terminal sequences show no obvious similarities between each other or to any other proteins. Only the threonine-rich stretch is reminiscent of the 200 residue long threonine stretch in the N-terminal region of the choanoflagellate Hoglet protein 22. However, this may be a case of convergent evolution. No Wart, Ground, or Ground-like domains could be detected in the genome of T. spiralis or in EST database searches restricted to Enoplea.

Based on phylogenetic analyses of the Hog domains, XC Xhog1, Xhog2, Xhog3, and Ts Qua-1 form a clade with the Quahog proteins (Figure 5, 6, Additional files 5, 6, 7). While N-terminal sequences for XC Xhog1, 2 and 3 are lacking they could be bona-fide Quahog proteins. A second, distinct clade is formed by Ts Xhog1, Xhog2, and XC Shog1, Xhog4, Xhog5 and Thog, indicating that they are derived from a common ancestor (Figure 5, 6, Additional files 5, 6, 7). In three cases, Ts Xhog1, Xhog2 and XC Xhog5), a common upstream sequence (Enop) has been identified (Figure 10), which seems to be specific to Enoplea nematodes, suggesting that at least in the cases of XC Shog1 and Thog the N-terminal regions have diverged relatively recently.

Almost all nematode Hh-related proteins form a distinct clade, the only exception being the Hh proteins, and Ts Xhog3 and XC S2hog, which are both very divergent and do not fall into the Hh clade of genes either (Figure 5, 6, Additional files 5, 6, 7). Two features distinguish the Hog domains of the nematode Hh-related proteins from those of the Hh proteins (Figure 1, 2, 3, 4, Additional files 1, 2, 3). 1) The regions corresponding to motifs K and L have characteristic differences in their conserved residues in nematode Hh-related proteins. 2) Two conserved cysteine residues are found in the central region of the Hog domain. When these two residues are mapped onto the X-ray structure of the C-terminal autoprocessing domain of Drosophila Hh 12, it emerges that they lie adjacent to each other and therefore could form a disulfide bond. This feature might stabilize this type of Hog domain in an extracellular environment, and this extra stability might possibly provide some new functionality. It is however not unique to nematode Hog domains. Zebrafish ihha and ihhb and fugu dhh (fhh) also have this extra cysteine pair, which must represent convergent evolution. It is worth pointing out that Ts Hh lacks the two cysteine residues and has motifs K and L as expected from a bona-fide Hh molecule. However, the quite divergent Ts Xhog3 protein, which lacks a Hedge domain, also lacks the cysteine residues and has motifs K and L.

tlastn searches of the N. vectensis predicted ORFs returned 10 hits. Several turned out to be differently predicted ORF variants most likely derived from the same locus, since corresponding genomic sequences for some of these loci displayed >99% identity. In the end six distinct ORFs were retrieved that all had good signal sequence for secretion. For four of the ORFs ESTs were found that at least partially support the predictions (Table 1, Additional file 9). In the case of Nv 239508 the corresponding genomic region seems to have undergone a recent duplication as two virtually identical Hog domains are present there (Additional file 12). In addition to the N. vectensis sequences, ESTs for two genes from Acropora millepora and one gene from Hydra magnipapillata were identified (Additional file 9). The EST from H. magnipapillata could be extended using the blastn of the NCBI trace archives, which also revealed a second, closely related paralogous gene (Additional file 9). Two genes from N. vectensis and one from A. millepora are bona-fide hh genes, because they both encode a Hedge domain and a Hog domain (Figure 2, 3, 4, Additional file 10). Two other ORFs, Nv 120428 and Acm DY579185, share conserved sequences upstream of the Hog domain with at least 3 conserved cysteine residues (Figure 11). The N-termini of Nv 140260 and 239508 do not show any similarity with known motifs, and the processed N-terminal peptide of Nv 140260 is only 86 residues long. Similarly, the upstream region of Hm CO905822 and its close paralog do not shown any similarity to the upstream regions of other cnidarian Hog proteins.

Last but not least, Nv 200640 is predicted to be 3592 amino acids long and is highly unusual. It is similar to the Hh proteins through the N-terminal Hedge domain (blast expected probability: 1e-18 to Ciona Hh), but no Hog domain follows (Additional files 10, 13). The Hedge domain is encoded by two exons, and after an intron of 600 bp many additional exons continue the ORF of the JGI prediction, but nowhere in this genomic region resides a Hog domain. Analysis of the ORF using the SMART server revealed that these extra exons encode multiple motifs with significant sequence similarity to other proteins (Additional file 13). The first motif, encoded by exons 3 and 4, contains a von Willebrand factor (vWF) type A domain (VWA). For example, the VWA domain of chicken collagen, type XIV, alpha 1 (undulin) is retrieved with a blastp probability of 8e-28. After the VWA domain, 21 CA (Cadherin repeat) domains follow, they occur as repeats in extracellular regions and are thought to mediate cell-cell contact when bound to calcium. Further follow two Immunoglobulin C-2 Type domains, two EGF repeats, a transmembrane region, and finally an SH2 domain.

The phylogenetic analysis of the cnidarian Hog domains reveals that they cluster primarily with the Hh Hog domains (Figure 5, 6, Additional files 5, 6, 7), albeit mostly with insignificant bootstrap values. The Hog domain of Nv 241466 Hh has the best similarity to the Hh Hog domains, and clusters with the deuterostome Hh proteins. Nv 140260 and Nv 239508 are most similar to each other, suggesting a likely duplication event within the cnidarian lineage. Nv 120428 and Acm DY579185 may also be related to these two proteins via their Hog domain (Figure 5, 6, Additional files 5, 6, 7), but the bootstrap values are not significant. The Nv 95413 Hh protein is rather divergent, and the Hydra sequence Hm CO905822 is also very divergent and does not from a clade with any of the N. vectensis sequences. Therefore, it is not possible to determine, whether all the cnidarian Hog genes originated all from a single ancestral gene in the cnidarian lineage, or whether hh and other Hog genes were already present before the split of Cnidaria and Bilateria. The Hedge domains of the three N. vectensis ORFs are more divergent than the bilaterian Hedge domains (Figure 9, Additional file 10). The Hedge domain of Nv 241466 Hh is most similar to bilaterian Hh proteins, with a best blast probability of 2e-52 to a fish Hedge domain. Nv 95413 Hh is more divergent, with a best blast probability of 5e-36, and Nv 200640 is the most divergent Hedge domain, with a probability of 1e-18 to a Ciona Hh.

In order to detect Hh sequences from lower eukaryotes, tblastn searches were performed using the organism restriction "eukaryotes NOT bilateria". This recovered a number of genomic and EST matches from lower animals, fungi, plants and protists (Figure 2, 3, 4, Additonal files 9, 14). One EST was recovered from the sponge Oscarella carmela, which was previously described 27. Analysis of this sequence shows that, while it does have a Hedge domain, the downstream sequence does not contain the start of a Hog domain in any frame (Additional file 10). No sequence similarity to a VWA domain is detected in that fragment either. Nevertheless, it indicates that as in the case of Nv 200640, this gene may not contain a Hog domain.

A match was detected to the gene GmGIN1 from the fungus Glomus mosseae, which belongs to the Glomeromycota, a sister group of ascomycetes and basidiomycetes and had already been described as having similarity to Hh 29. The Hog domain has a blast probability of 7e-18 to the best matching Hh Hog domain, which is much better than the blast probability of choanoflagellate Hoglet to the best matching Hh Hog domains (4e-10). Furthermore, good matches to motifs J and K, as well as a region with similarity to motif L. Therefore, GmGIN1 contains a bona-fide Hog domain (Figure 2, 3, 4, Additional file 14). The upstream domain of GmGIN1 shares similarity with Ras-like GTPases, e.g. the Arabidopsis protein AIG1 (avrRpt2-induced gene 1) and the animal The IAN/IMAP subfamily 29. However, this ORF lacks a signal peptide and may therefore not be secreted.

A number of matches were found in Alveolata, i.e. in the dinoflagellates Alexandrium tamarense, Amphidinium carterae, and Karlodinium micrum (blast expected probability of aKm Hog: 9e-17 to the best matching Hh Hog domain; note: blast probabilities below also refer to Hh Hog domains) and the apicomplexans Cryptosporidium muris and Cryptosporidium parvum (blast prob. of aCp Hog: 4e-08). Their Hog domains contain motifs J and K, although in a few cases the cysteine has been replaced with a serine in motif J (Figure 2, 3, 4, Additional file 14). The aCm and aCp sequences are most likely full length, they have signal peptide sequences for secretion and share a conserved upstream region of about 100 residues in length that contains two conserved cysteine residues (Figure 12), but no sequence similarity of this motif to other known domains was found.

Further Hog sequences were found in red algae and mosses (Figure 2, 3, 4, Additional file 14): In the mosses Selaginella moellendorffii (blast prob.: 1e-09) and Physcomitrella patens one sequence each with a Hog domain; in the red algae Chondrus crispus two sequences (blast prob. of rCc Hog: 1e-10); in Griffithsia japonica one sequence (blast prob.: 3e-08); in Porphyra haitanensis (blast prob. of rPh Hog: 4e-07) and Porphyra yezoensis two Hog domain ORFs each; and in Gracilaria changii six ORF fragments (blast prob. of rGc Hog1: 3e-12). Those moss and red algae Hog domains that are not truncated have motifs J and K, although the cysteine residue in motif J has been changed to serine, threonine, or aspartate. Alignment of rPy and rPh Hog2 revealed conserved sequences upstream of the Hog domain, however, these two sequences are relatively closely related so this conservation is not surprising (Figure 13). Blast searches with this upstream region did not reveal matches in any other organisms. Similarity, alignment of the moss sequences revealed also a conserved upstream region that was not found in other organisms (Figure 14). The P. patens sequence is presumably full length, since it was predicated from genomic sequence, and it has a good signal peptide. One EST sequence supposedly stems from rice (XX 104K18), however, it has a much better match to Hh Hog domains (blast prob.: 3e-24) than other non-metazoan Hog domains, and we could not find any match to rice genomic sequences. Therefore, this sequence may come from a contaminating organism and is designated as species XX here.

Additional Hog-like sequences were recovered from the cercozoan Bigelowiella natans (blast prob.: 9e-10), from the cryptophyte Guillardia theta (blast prob. of crGt Hog1: 6e-11 to Hog of Mo hoglet), and from the jakobid Jakoba libera (blast prob. jJl Hog1: 2e-05) (Additional file 14). These sequences have motif J, although the cysteine has been replaced, and in those cases, where the C-terminal region is complete, it is clear that motif K is not conserved. Sequence alignment of the J. libera Hog sequences revealed conserved upstream sequences with some conserved cysteine residues (Figure 15). Numerous ESTs cover jJl Hog1 and therefore its ORF could be complete. If this is the case, the putative start methionine has a good signal sequence for secretion, and therefore jJl Hog1 has the same global structural features as the animal Hh and Hh-related proteins, i.e. a secreted N-terminal domain followed by a Hog domain. Finally, three sequences were recovered from the haptophyte Pleurochrysis haptonemofera. Sequence alignment revealed sequence conservation upstream of the Hog domain with conserved cysteine residues. However, it is noteworthy that the Hog domain is much better conserved than the upstream region, indicating that the upstream region can evolve more rapidly (Figure 16).

Overall, these results show that Hog domains occur in many different branches of the major groups of eukaryotes. However, multiple losses seem to have occurred, since in many branches we did not detect Hog domains, for example, in Arabidopsis thaliana and other higher plants, or in the currently sequenced ascomycetes and basidiomycetes, or in other sequenced organisms such as Dictyostelium.

C. elegans not only lacks a bona-fide Hh molecule, but several other components of the Hh signaling pathway have been lost as well. In particular orthologs of the Hh signaling pathway in recipient cells, i.e. smoothened, fused, suppressor of fused (sufu), and costa are missing 18. On the other hand, a homolog of the transcription factor Cubitus interruptus (Ci/Gli) is present, albeit it has been adapted for sex determination. And multiple homologs for the receptor of Hh, i.e. Patched, have been found in C. elegans 18, as well as the related molecule Dispatched, required for secretion of Hh. Patched, Dispatched, Smoothened, Ci/Gli and Hip have already been found in N. vectensis 2526. We were particularly interested to find components lacking in C. elegans in the relatively well sequence genomes of N. vectensis and T. spiralis. Using reciprocal blast searches, we have attempted to identify these components of the pathway in Nematostella and Enoplea (Table 2). In Xiphinema we only detected an EST for patched, but this is not surprising giving the limitations of the current dataset. In T. spiralis we detected patched, dispatched, dally-like and Ci/Gli, but found no evidence for Ihog, smoothed, costa, fused, and sufu. This is actually identical to the situation in C. elegans. Presently about 56.8 Mb of an estimated genome size of about 65 Mb has been sequenced for T. spiralis 41. If we assume that only about 80% has been sequenced, the probability of finding only the genes listed in Table 2, but missing the others is 0.013%. If the sequence coverage is higher, this probability would even be lower. Therefore, we have to assume that in T. spiralis, even though it has a bona-fide hh gene, the Hh signaling pathway is compromised in a similar way as in C. elegans.

In N. vectensis we found good orthologues for dispatched, dally-like, patched, smoothened, fused, sufu and Ci (Table 2). No obvious homolog was found for Ihog. In the case of Drosophila costa, good matches to its human homologs were found, and Drosophila costa is rather divergent. Recently it has been shown that the mammalian homologues of fused and costa do not play the same key role in the pathway as in flies, instead sufu plays a major role 4243. Overall, it looks like most of the key players of the Hh pathway are present in N. vectensis so that it is clear that the pathway was already well established before the split of Cnidaria and Bilateria.

During the tblastn searches ESTs and ORFs from non-Hog genes such as inteins were discovered, usually in the non-significant zone at the bottom of the results lists. One group of genes attracted our attention, because upon closer inspection it became apparent that these genes had an amino-terminal domain comprised of a VWA domain followed by a region that has good similarity to the first part of the Hint domain, i.e. in particular motifs A and B (Figure 17, 18, 19, Additional file 15). This observation was intriguing given that in Nematostella Nv 200640 a Hedge domain is followed by a VWA domain. Further blast searches revealed the presence of these VWA-Hint proteins in Tetrahymena, several fungal species, the Heterolobosea Naegleria gruberi, the parabasilid Tritrichomonas foetus, dinoflagellates, the slime mold Physarum polycephalum, rice and the chanoflagellate Monosiga brevicollis. Additional matches in other species, for example, pine tree, were found in the EST database, but not included here, because the fragmentary nature of the sequence information made it impossible to determine, whether the VWA domain resides in the same transcript as the Hint domain (data not shown). No match could be found for the cDNA sequence from rice (pOs AK110392) in the genomic sequence, but ESTs recovered from other plants support the notion that VWA-Hint proteins exist in plants.

The VWA-Hint proteins do not seem to have a signal peptide for secretion. The VWA domain is located at the N-terminus of the proteins, although in four cases a Ubox precedes the VWA domain (Figure 17, 18, 19). A region of around 300 residues separates the VWA domain from the Hint domain. This region has several small patches of conservation and one large region, that we propose to call Vwaint domain. At the C-terminus a Hint-like domain follows, which is of similar size as a Hog domain. However, the best conserved features are only motifs A and B, i.e. the N-terminal region of the Hint-like domain. One region shares a little similarity with motif F of inteins and BIL-Bs, but motifs J, K and L are lacking (Figure 17, 18, 19). The Hint-like domain is also rather different from inteins or Hog domains, the best blast matches of aTt 00471620 are to honeybee Hh with a probability 0.013. Therefore, these sequences cannot be classified as intein, Hog or Bil domains, and we refer to these genes as Vint genes. Vint genes are apparently so wide spread in eukaryotes that we have to assume that a common ancestor was present in early eukaryotes. However, Vint genes seem to be lacking in Arabidopsis, many fungi (for example, Saccharomyces cerevisiae), and in Metazoa. Multiple independent losses in different lineages seem the most likely explanation for this absence.

Our searches also revealed a group of proteins from bacteria that had a Hint-like domain at their C-terminus and shared some weak sequence similarity in their N-terminal region (Additional files 16, 17). At least some of these proteins are predicated to have signal peptides for secretion, and the upstream region has two cysteine residues conserved between all sequences. The Hint-like domains of these bacterial proteins are also quite divergent from inteins, Hog and BIL domains, and represent yet another subgroup. This subgroup has previously also been detected by Dassa and Pietrokovski 21. The new members we retrieved here support the notion that this is yet another new type of Hint protein.

Hh genes are present in deuterostomes as well as in several different protostome phyla such as molluscs, annelids, and arthropods (Figure 20). However, in nematodes the situation is more complex. In C. elegans, C. briggsae and B. malayi we find no hh gene but instead many hh-related genes. We recovered 19 hh-related genes from the nematode B. malayi. Based on empirical evidence from other gene families we estimate that the genome of B. malayi is around 80% complete (K. Mukherjee and T. B., unpublished). Therefore, some additional hh-related genes might still be forthcoming. But the present survey shows that members of the qua, wrt, grd and grl gene families are all present in B. malayi. Only a representative of a grd gene with a Ground domain has so far not been found. The phylogenetic analyses show that while there are some instances of direct orthology between B. malayi and Caenorhabditis genes, in many instances, in particular for the grl genes, the relationship is not clear and in fact suggests that independent diversification occurred in these two Chromadorea branches. This shows that these gene families have been actively evolving in nematodes.

In the more distantly related Enoplea nematodes Xiphinema sp. and T. spiralis a strikingly different pictures emerges. In both species we find both a hh gene as well as several hh-related genes. In T. spiralis we also find a quahog gene, and – based on the phylogenetic analyses – some of the Xiphinema genes could also be quahog genes. Two T. spiralis and one Xiphinema protein share a new motif (Enop motif) upstream of the Hog domain that appears to be specific to Enoplea nematodes. However, there are also a number of instances of N-terminal sequences that are very short. Several of these proteins cluster with the "Enop" proteins in the phylogenetic analyses, suggesting that they diverged from a common ancestor. However, two proteins with short N-terminal regions (Ts Xhog3 and XC Shog2) are rather divergent and do not reliably fall within the clade of nematode-specific Hog proteins ("Nema-Hog" proteins) in phylogenetic analyses. In particular Ts Xhog3 lacks the conserved cysteine pair usually found in Nema-Hog domains, and it shares motifs K and L with Hh Hog domains, indicating it could be derived from a Hh protein. Therefore, while these genes could have diverged from hh or Nema-Hog genes, it may also be possible that the represent ancestral genes that were lost in Chromadorea. In conclusion, we think that there were probably at least three different types of Hog genes in the common ancestor of Enoplea and Chromadorea, one hh gene, one quahog gene and one gene which give rise to the wrt/grd branch in Chromadorea and the Ts Xhog1/2 branch in Enoplea. But possibly up to five Hog genes could have existed in the common ancestor. The proliferation into further distinct groups such as wrt, ground and ground-like appears to have happened later in a branch specific manner.

Many different N-termini now exist in Nema-Hog proteins. Two possible mechanisms can explain this diversity: Either acquisition of new N-terminal domains, or divergent evolution of existing N-terminal domains. A relatively good case can be made that all Wart, Ground and Ground-like domains arose from a single common ancestor based on weak sequence similarities between the motifs 17. This relationship is also supported by the phylogenetic analyses of the Hog domains. Therefore, multiple loss of the Hog domain must have occurred secondarily within the wrt and ground families. The presence of the rather short N-termini in Enoplea suggests that these regions have evolved and diverged through mutations, rather than by acquisition of a new domain. The threonine-rich stretch in XC Thog is very likely the result of polymerase slippage, though it is striking that this feature has evolved separately also in the choanoflagellate Hoglet protein 22. It is also worth mentioning that some of the Caenorhabditis N-terminal domains have repetitive regions outside of the conserved Ground and Ground-like domains, mainly proline, glycine and serine. For example, Ce grl-23 has a 176 residue long stretch upstream of the Ground-like domain containing 125 glycine residues. In conclusion, most of the observed variability in the N-terminal domains of nematode Hh-related proteins is probably the result of sequence divergence from a progenitor, rather than acquisition of new domains. Loss of N-terminal domains in the case of C. elegans Hog-1, as well as loss of Hog domains did occur however.

A surprising observation is the fact that T. spiralis has a hh gene, but apparently lacks several components of the Hh pathway, such as Smoothened. Particularly noteworthy is that the components that appear to be missing are the same as in C. elegans. This would suggest that the signaling pathway was modified by loss already before the split of Enoplea and Chromadorea, even though hh was maintained in Enoplea. While one could imagine that Hh could be maintained in an animal parasite such as T. spiralis to affect host cells, this is very unlikely in the case of the plant nematode Xiphinema. It implies that Hh has an important function even in the absence of Smoothened, and it refutes the hypothesis that the Nema-Hog genes evolved directly from hh concomitantly with the other changes in the Hh pathway.

In Cnidaria we also encounter a complex situation with both Hh and Hh-related proteins. Both in N. vectensis and A. millepora we find bona-fide hh genes that have a Hedge and a Hog domain. Another gene is well conserved between N. vectensis and A. millepora and has a distinct, novel secreted N-terminal domain. Two further Hh-related proteins in N. vectensis have yet other, distinct N-termini. The upstream region of the two closely related genes retrieved from Hydra do not share any similarity with those in Nematostella, indicating divergent evolution. No sequence similarity of these new N-terminal motifs has been found outside Cnidaria.

The hh-related genes from Cnidaria are however distinct from those in nematodes, since the phylogenetic analyses of the Hog domains does not show them to be closely related. Therefore, we would like to suggest that – as in the case of the nematode hh-related genes – the Cnidarian N-terminal domains have evolved from common ancestors by divergent evolution rather than by domain acquisition.

The case of Nv 200640 is perhaps a special exception. In this protein we find an N-terminal Hedge domain fused to a large extracellular protein that contains a VWA domain as well as CA and EGF repeats, but it clearly lacks a Hog domain. The VWA domain is a 200 residue long domain first identified in von Willebrand Factor 4445. VWA domains are found both in extracellular and intracellular proteins, such as non-fibrillar collagens, plasma proteins such as complement factors and integrins, and they mediate adhesion via metal ion-dependent adhesion sites. Likewise, the CA repeats also mediate adhesion in a Ca2+ dependent fashion. Therefore, the Nv 200640 protein is probably involved in cell adhesion. This shows that the Hedge domain can also evolve in a modular fashion and separate from the Hog domain. The EST recovered from sponges also has a Hedge domain that lacks the immediately following Hog domain, and may perhaps represent also a protein lacking a Hog domain.

We have recovered a substantial number of Hog domain proteins from many diverse groups of eukaryotes, mostly protists, such as red algae, moss, alveolates (ciliates, dinoflagellates, apicomplexans), cryptophytes, jakobids, haptophytes, cercozoa and Glomeromycota fungi. While some of these Hog sequences are quite divergent, they are invariably most closely related to Hog domain proteins from animals, and not to inteins, such as those found in fungi, or to BIL or Vint domains. Given the widespread occurrence in many of the major groups of eukaryotes (46, we must conclude that Hog proteins were present already in the earliest eukaryotes. We find diverse N-termini associated with the Hog domain that are only conserved to limited extends within groups (case in point are the various conserved N-termini in nematodes). Many of these limited conserved N-termini have conserved cysteine residues, and in cases, where one can be quite confident of the start methionine, they start with a good signal peptide for secretion. Only in the case of the fungal protein GmGIN1 and the choanoflagellate Hoglet are distinct other N-terminal domains fused to the Hog domain. Therefore, we postulate that an ancestral Ur-Hog gene existed, with a secreted N-terminal domain and an autoprocessing Hog domain, that may have added a sterol or similar moiety to its secreted N-terminus. This gene evolved in concert with eukaryote evolution and was lost in several branches. In animals, the question arises about the origin of the Hedge domain. Both in sponge and in Nematostella we find a Hedge gene that lacks a Hog domain. Perhaps such a gene merged with a Hog domain in early metazoans. However, the reverse process is also possible: the Hedge domain evolved as an N-terminal variant of a Hog protein in early metazoans, and in the two Hedge genes in sponge and Nematostella the Hog domain was lost later. Both in Cnidaria and nematodes we find both hh and hh-related genes. Did the hh-related genes evolve twice independently from a hh precursor in each lineage? This is certainly the most parsimonious hypothesis. Nonetheless, in an alternative scenario, a hh and a hh-related gene could have been present in the common ancestor of eumetazoa, and the hh-related gene would have given rise to the cnidarian and nematode hh-related genes. For this hypothesis to be true, we would have to postulate three separate losses of hh-related genes: in deuterostomes, in lophotrochozoa, and in arthropods. While this seems rather unlikely, we do observe many losses of Hog genes in various branches of eukaryotes, as well as loss of the Hog domain only in a number of nematode genes so that such a series of losses may not be totally impossible.

Our searches revealed new genes with Hint motifs merged to VWA domains. Given that a Hedge domain was found fused to a VWA domain in Nematostella we investigated this further and recovered a novel gene family. The well-conserved gene structure consists of a VWA followed by a new domain, termed Vwaint, followed by the "Vint"-type Hint domain. Unlike the Hog proteins, these proteins are most likely not secreted and instead are processed inside the cell. The Vint genes are present in many eukaryotic groups, but must have been lost multiple times, in particular in multicellular eukaryotes. Multiple loss seems to be a common theme also in Hog proteins and especially inteins 2132. Inteins may be subject to special selective pressure for loss 2132, and this pressure may also extend to Hog and Vint proteins. However, gene loss is not uncommon. The N. vectensis genome contains a remarkable complexity of highly conserved gene families 25, and several instances of later gene loss in the protostome or deuterostome lineage, for example in the homeobox gene family, have been found 4748, indicating gene loss later in evolution is feasible.