Introduction

Hox genes are found in all metazoan phyla except poriferans (Larroux et al. 2007). They are active in distinct domains along the main body axes and direct the morphogenesis of segment-specific structures via the activation of downstream target genes (reviewed by Gehring 2007). These genes are organized into a so-called “Hox cluster” (Bürglin 1994). A Hox cluster consists of anterior, central, and posterior genes; the central genes have been identified only in bilaterians, but not in cnidarians, which are a basal metazoan phylum that arose before the divergence of bilaterians (Chourrout et al. 2006). The central Hox genes are diverse, and their diversity is considered the basis of morphological complexity in bilateral development (Ogishima and Tanaka 2007). Understanding the evolution of the central Hox genes should thus lead to understanding the evolution of bilaterian body plans. The reconstruction of Hox cluster evolution in the Metazoa should provide valuable data for understanding the evolution of bilaterian body plans and the relationship between genetic and morphological complexity (reviewed by Ferrier 2007).

It has been proposed that lophotrochozans have an ancestral Hox complement composing 10–11 genes (de Rosa et al. 1999; Balavoine et al. 2002; Kulakova et al. 2007): Hox1, Hox2, Hox3, Hox4, Hox5, Lox5 (which is probably orthologous to ftz from ecdysozoans; Telford 2000b), Hox7 (named after the Lineus sanguineus LsHox7, and similar to ecdysozoan Antp; we have followed this nomenclature after Kulakova et al. 2007), Lox2, Lox4, Post-1, and Post-2. Genes clearly orthologous to Hox1-5 are found in both protostomes and deuterostomes. The genes Lox5, Lox2, Lox4, Post-1, and Post-2 are, on the other hand, characteristic of lophotrochozoans (Balavoine et al. 2002). Lox5 genes are characterized by the presence of the conserved “Lox5” parapeptide, which is N-terminal to the homeodomain. Lox2 and Lox4 genes share with ecdysozoan Ubx and Abd-A the “Ubd-A” parapeptide. It is not clear whether Lox2 and Lox4 originated as an independent duplication of an ancestral “Ubd-A” gene in lophotrochozoans, or whether the last common ancestor of protostomes already had two “Ubd-A” genes.

Metazoans also possess an ancestral Parahox cluster (Garcia-Fernàndez 2005). Hox and Parahox clusters probably originated from the duplication of an ancestral ProtoHox cluster; therefore, Hox genes are not monophyletic, with anterior Hox genes being most related to the Gsx Parahox genes, Hox3 genes with Xlox Parahox genes, and posterior Hox genes to Cdx Parahox genes. All three Parahox genes have been found in several lophotrochozoans (Ferrier and Holland 2001; Kulakova et al. 2008).

Platyhelminthes constitute a group of organisms that display a range of diverse life histories from free-living to parasitic. As they lack a coelom, segmentation, elaborated organs, and anus, they were considered to be among the basal groups of bilaterians. However, analyses based on 18S ribosomal and homeobox sequences placed Platyhelminthes as a derived bilaterian phylum, among the lophotrochozoans (Aguinaldo et al. 1997; Balavoine 1997). Also, many recent morphological analyses coincide in placing the platyhelminthes not as basal to all bilaterians but as derived protostomes, together with other spiralian phyla (Ax 1996; Nielsen 2001). The exact position of platyhelminthes among lophotrochozoans is not clear. Indeed, they have been placed at a wide range of positions, from being basal to all other lophotrochozans (Helmkampf et al. 2008) to a sister relationship with Annelida (Lartillot and Philippe 2008).

Information about Hox genes in plathyhelminthes is mostly about planarians (Tricladida) (Bartels et al. 1993; Tarabykin et al. 1995; Bayascas et al. 1997; Orii et al. 1999; Saló et al. 2001; Nogi and Watanabe 2001). Orthology relationships of Hox genes of triclads have been well established (Saló et al. 2001), in groups PlHox1–PlHox9. These groups have been proposed to be orthologous to Hox1-5, Lox5, Lox4, and Post-2. Few Hox genes have been isolated from parasitic flatworms (Neodermata). Studies included Schistosoma mansoni (Pierce et al. 2005; Webster and Mansour 1992), in which Hox1, Hox4, Lox5, and Lox4 orthologs were identified, and the cestode Taenia asiatica (Kim et al. 2007). Recently, Olson (2008) extensively reviewed the Hox genes from the platyhelminthes, performing a phylogenetic analysis together with other selected lophotrochozoan sequences. This analysis included, also, unpublished sequences from Hymenolepis microstoma. Orthology assignment between Hox genes of triclads and neodermatans, or of platyhelminthes and other lophotrochozoans, was based exclusively on phylogenetic relationships of the homeodomain, and only some orthology groups were clearly recovered. Information about Parahox genes in platyhelminthes is scarcer; although they have been searched for extensively (Saló et al. 2001; Olson 2008), only Xlox and Cdx orthologs have been found in the polyclad Discocelis tigrina (Saló et al. 2001).

Our study was directed at obtaining Hox gene sequences from neodermatans, using two different strategies. We used the Schistosoma mansoni and Echinococcus multilocularis genomic assembled contigs from the Sanger Institute to search for Hox and Parahox genes; this approach should be insensitive to the amplification bias of degenerate PCR strategies. Furthermore, the 7× coverage of the S. mansoni genome makes unlikely the possibility that existing Hox genes might be missed. We also searched for Hox genes by degenerate PCR in the cyclophyllidean cestode Mesocestoides corti. We have inferred orthology relationships among Hox genes of neodermatans, triclads, and other lophotrochozoans by means of phylogenetic analysis, presence of characteristic parapeptides, and unusual intron positions. Our results suggest that the last common ancestor of triclads and neodermatans had a reduced Hox complement compared to other lophotrochozoans, probably due to gene loss.

Materials and Methods

Mice infected with M. corti tetrathyridia were kindly donated by Laura Dominguez and Jenny Saldaña of Facultad de Quimica, Universidad de la Republica, Uruguay. Parasite removal and culture were made following Britos et al. (2000). Tetrathyridia were cultured in vitro to obtain segmented worms, showing an elongate body and numerous proglottids, and separated manually. DNA was extracted from tetrathyridia following McManus et al. (1985). RNA was extracted from tetrathyridia and segmented worms using Trizol (Gibco). cDNA was synthesized from 5 μg tetrathyridia or adult worm RNA with poly dT primer and Superscript II reverse transcriptase (Invitrogen).

Isolation of Hox cDNA Fragments

The strategy employed (Tarabykin et al. 1995) was to amplify Hox homeobox genes from tetrathyridia cDNA with degenerated primers directed to the coding region of the first and third helices of the homeodomain. Primer sequences are S01, GARYTNGARAARGARTT, and S02, CKNCKRTTYTGRAACAA. Cycling conditions were as described by Tarabykin et al. (1995). PCR bands were excised from agarose gels and cloned into pGEM-T-Easy vector (Promega). The 200 recombinant plasmids were sequenced by CTAG Service of Facultad de Ciencias (Uruguay) using a Perkin-Elmer ABI Prism 377 automated DNA sequencer.

Southern Blot Assay

Southern blot assay was performed following Sambrook et al. (1989). A 0.8% agarose gel was loaded with 10 μg M. corti or 20 μg mouse DNA digested with EcoRI and SalI. The radioactive probes were synthesized by EcoRI digestion of the recombinant plasmids containing MvHox1 or MvHox7 fragments. The excised inserts were labeled with 32P-αATP using the Prime-a-Gene kit (Promega).

Search for Hox and Parahox genes in S. mansoni and E. multilocularis genomic assemblies

We searched for Hox and Parahox genes in genomic contigs from S. mansoni genome version 3.1 and E. multilocularis. These sequence data were produced by the Schistosoma and Echinococcus Sequencing Groups at the Sanger Institute and can be obtained from ftp://ftp.sanger.ac.uk/pub/pathogens/Schistosoma/mansoni/ and ftp://ftp.sanger.ac.uk/pub/pathogens/Echinococcus/. Searching for Hox genes in genomic contigs was done by Blastn in the Sanger Blast Server (http://www.sanger.ac.uk/cgi-bin/blast/submitblast/s_mansoni and http://www.sanger.ac.uk/cgi-bin/blast/submitblast/Echinococcus), using Mus musculus HoxA1; Euprymna scolopes EsAntp, EsPost-1, and EsPost2; Nereis virens Gsx and Cdx; and Capitella sp. Xlox protein sequences as queries (accession numbers are provided in Fig. 2). The expected cut-off value was 10 in order to avoid missing homeoboxes with several introns. A list containing all the contigs that were hit was analyzed both manually and by Blastp (after joining the conceptual translation of exons) against the GenBank nr database. Approximately 100 Blast hits were analyzed for each species, resulting in the retrieval of more than 50 homeodomains, not only from the ANTP class but also from others such as POU, paired, LIM, SINE, and ZF (this suggests the search for Hox and Hox-like genes was exhaustive). Homeoboxes and flanking regions from genes identified as Hox or Hox-like were then recovered, and their exons joined manually when introns were present, based on amino acid similarity to other Hox genes and the presence of canonical splice sites. In some cases (i.e., EmHox1), similarity did not allow the confident recovery of the complete homeobox; in these cases the sequence was retrieved only to the nearest possible canonical splice site to the Blast hit. We also searched for Hox genes in unassembled reads of these organisms, in the S. mansoni Genome Database (GeneDB; http://www.genedb.org/genedb/smansoni/), and in ESTs from all platyhelminthes in GenBank.

Phylogenetic Analyses

Conceptually translated amino acidic sequences were aligned using ClustalW software (Thompson et al. 1994). Alignments included numerous sequences from all Hox and Parahox orthologous groups belonging to other lophotrochozoans, together with the deuterostome Mus musculus and the ecdysozoan Drosophila melanogaster. Only some sequences from each planarian orthology group, PlHox1-PlHox9, were included, because genes within these groups are very similar and have been extensively characterized before (Saló et al. 2001; Olson 2008). Unrooted phylogenetic analyses were performed, using only homeodomain sequences. Similar results were obtained when performing rooted analyses using Evx and Mox genes from Mus musculus and Drosophila melanogaster as outgroups, except that bootstrap support values for some clades were lower (data not shown). Maximum parsimony and neighbor-joining (NJ) phylogenetic trees were constructed using Mega (Kumar et al. 2004). Poisson correction was used as the substitution model for NJ; similar results were obtained using other models. Bootstrap support values were estimated using 1000 replicates. Maximum likelihood analysis was performed using the ProML application from BioEdit (Hall 1999), using the Jones-Taylor-Thornton model. All analyses gave very similar results, except in nodes with very low support; therefore, only the NJ analysis is reported.

Results

We found two Hox genes in M. corti by degenerate PCR, seven in the E. multilocularis genome contigs, and nine in the S. mansoni genome contigs and unassembled reads (including the five genes previously reported by Pierce et al. 2005). No Parahox genes were found, although Evx and Mox orthologs, which are basal to Hox/Parahox genes (Minguillón and Garcia-Fernàndez 2003), were clearly identified in both S. mansoni and E. multilocularis (data not shown). We have classified these sequences, together with those of triclads, according to homeodomain sequence similarity and parapeptides (Fig. 1), phylogenetic analysis (Fig. 2), and intron positions. Below, we describe these sequences and classify them according to the orthology groups to which they probably belong.

Fig. 1
figure 1figure 1

Alignments of homeodomain and flanking sequences from M. corti, E. multilocularis, and S. mansoni with related sequences from other lophotrochozoans, the ecdysozoan D. melanogaster and the deuterostome Mus musculus. Numbers above correspond to standard positions for the homeodomain. Alignments of Hox5 and Hox7 sequences are included for comparison. Within each alignment, positions that are absolutely conserved are in white letters with black shading. Residues marked with gray shading within the homeodomain are referred to in the main text. Residues marked with gray shading in the Hox4/Dfd, Lox5, and Lox4 alignments correspond to the Dfd, Lox5, and Ubd-A parapeptides, respectively. Sequences obtained in this study are marked with an asterisk. Abbreviations and accession numbers are given in Fig. 2

Full size image
Fig. 2
figure 2

Neighbor-joining tree of Hox and Parahox sequences from platyhelminthes and other bilaterians. Bootstrap support values are given next to nodes in percentages. Nodes with less than 50% support have been collapsed. Genbank accession numbers (and/or contig numbers for S. mansoni and E. multilocularis) are provided to the right. Sequences obtained in this study are marked with an asterisk. Orthology groupings are indicated by the color codes defined below the tree. Members of the PlHox5 group (Gt DtHoxD and Dj Plox-4) are not color-coded. Species abbreviations: Cap, Capitella sp. (Annelida); Dist, Discocelis tigrina (Polyclada); Dm, Drosophila melanogaster (Arthropda); Dj, Dugesia japonia (Tricladida); Em, Echinococcus multilocularis (Cestoda); Es, Euprymna scolopes (Mollusca); Gt, Girardia tigrina (Tricladida); Hme, Hirudo medicinalis (Annelida); Lan, Lingula anatina (Brachiopoda); Ls, Lineus sanguineus (Nemertea); Mm, Mus musculus (Vertebrata); Nvi, Nereis virens (Annelida); Pni, Polycelys nigra (Tricladida); Pvu, Patella vulgata (Mollusca); Sm, Schistosoma mansoni (Trematoda); Tas, Taenia asiatica (Cestoda)

Full size image

Hox1 Orthologs

Hox1 orthologs from the platyhelminthes have been recovered before in the polyclad Discocelis tigrina (Distox-A, Saló et al. 2001), in several triclads (PlHox1 group; Saló et al. 2001), in S. mansoni (SmHox1; Pierce et al. 2005), in Echinostoma trivolvis (ETOX-A, L19170; see Olson 2008), and in Taenia asiatica (in which two paralogous genes were found, TasHox1a and TasHox1b; Kim et al. 2007).

We found a Hox1 ortholog in E. multilocularis (EmHox1, Contig_0006768) and in M. corti (MvHox1, AY187806). In the case of MvHox1, one of the primers annealed 5′ of the expected site. This circumstance allowed us to have a more complete homeodomain sequence for this gene. To confirm that the isolated sequence is indeed from M. corti and to estimate the gene copy number, we performed Southern blot analysis. When working with parasites the contamination with host material is a concern. For this reason, we included mouse genomic DNA lanes in the Southern blots. The membranes hybridized with MvHox1 probe exhibit one single band for M. corti DNA digested with SalI and EcoRI (Fig. 3). This suggests that MvHox1 is a single copy gene. Mouse DNA did not reveal any band. We have also confirmed by RT-PCR with gene-specific primers that MvHox1 is expressed in both the tetrathyridium stage and in the adult segmented worms (data not shown).

Fig. 3
figure 3

Southern blot analysis of MvHox1 and MvHox7. Mesocestoides corti DNA (10 μg) or Mus musculus (20 μg) was digested with the restriction enzyme indicated above and hybridized with a radioactive probe generated with the sequence available for MvHox1 and MvHox7. The size of the molecular weight marker fragments is indicated on the left

Full size image

In phylogenetic analysis, all putative Hox1 genes from the platyhelminthes cluster with moderate support with Hox1 orthologs from other bilaterian phyla (Fig. 2); furthermore, all Hox1 genes from cestodes are clustered too. Alignment of these sequences reveals that although some characteristic residues of Hox1 genes are conserved in some cyclophyllidean cestodes (such as threonine in position 43), these sequences are rather divergent. In position number 23, most homeodomains have an asparagine residue, but Hox1 representatives of the cestodes M. corti, E. multilocularis, and T. asiatica show serine/threonine. In addition, homeodomains of Hox1 orthologs have arginine or lysine in position 24, but MvHox1, EmHox1, and TasHox1a display the amino acid histidine. In position 29 all Hox1 proteins have an alanine, except cestode ones.

Olson (2008) labeled several other genes from Echinococcus spp. (Hbx1 and Hbx2) as Hox-1-like genes; however, close inspection clearly demonstrates that these genes are actually from the NKL subclass.

Hox2 Orthologs

In S. mansoni, we have found a putative Hox2 ortholog, SmHox2 (Smp_contig022848). It is located in scaffold Smp_scaff000314, in which SmHox4 is also located. It has a phase-0 intron between codons 46 and 47 of the homeobox. This homeodomain is present in Smp_166150 in the GeneDB; it seems that this CDS is incorrectly assembled, fusing the homeodomain to a serine/threonine kinase.

Phylogenetic analysis groups this sequence with Hox2 genes from other bilaterians with moderate support. The amino acid sequence shares several conserved residues characteristic of Hox2 (such as T11) or both Hox2 and Hox3 genes (C27, P29). Before this work, the only putative ortholog of Hox2 genes recovered from any platyhelminth was a small fragment from Girardia tigrina, DtHoxB (Bayascas et al. 1997; Saló et al. 2001). In our analysis, DtHoxB is clustered with low support with Hox3 genes (this node is collapsed in Fig. 2), but given its similarity to SmHox2 and the limited information in this short sequence, this might be an artifact.

Hox3 Orthologs

In triclads, the PlHox3 group has been linked to Hox3 genes due to overall sequence similarity and low phylogenetic support (Balavoine and Telford 1995; Bayascas et al. 1997; Saló et al. 2001), although it is highly divergent. In neodermatans, putative PlHox3 orthologs have been found in T. asiatica (two paralogs, TasHox3a and TasHox3b) and a short sequence similar to TasHox3b was found in Echinostoma trivolvis (ETOX-E, L19216). We have found similar sequences in both S. mansoni (SmHox3, Smp_contig022107; Smp_164610 in the GeneDB) and E. multilocularis (EmHox3, Contig_0003912). All these sequences are clustered together with good bootstrap support, constituting a clade that is basal to both Hox2 and Hox3 genes (this node is collapsed in Fig. 2). Therefore, the assignment of these sequences to the Hox3 group remains tentative.

This clade is clearly resolved in phylogenetic analyses into two groups, with very good support; the differences between them are apparent in the alignment (Fig. 1). One branch (Hox3a group) contains genes from triclads and TasHox3a; the other (Hox3b group) has genes from the neodermatans E. multilocularis, S. mansoni, and the TasHox3b gene. This latter group would also include ETOX-E, which is very similar to these sequences and grouped with them in the analysis by Olson (2008).

One explanation for this topology could be that the last common ancestor of triclads and neodermatans possessed two Hox3 genes, and that these were either selectively lost or artifactually not recovered in searches for Hox genes in different lineages (Hox3b in triclads, Hox3a in most neodermatans). Alternatively, TasHox3a could be a T. asiatica-specific paralog that converged with Hox3 sequences of triclads.

Central Genes

Among central genes, resolution is usually very low in phylogenetic analysis using the homeodomain sequence (Balavoine et al. 2002), and this is the case in our analysis. We were able, however, to establish orthology relationships using parapeptide sequences.

Hox4 Orthologs

Hox4 genes have been described from several platyhelminthes (Distox-D from D. tigrina, PlHox4 group from triclads, SmHox4 from S. mansoni, and the short sequence TasHox4 from T. asiatica). Olson (2008), based on phylogenetic analyses, labeled members of the PlHox4 group as belonging to the Hox1-like genes; however, the presence of the Dfd parapetide clearly demonstrates the affinity of these sequences to Hox4 genes, as suggested by Saló et al. (2001).

One putative Hox4 ortholog from E. multilocularis (EmHox4; Contig_0004814) was found in our search. In the phylogenetic analysis, the three neodermatan sequences clustered together with good support, but Hox4 genes formed part of a large polytomy of central genes. We were able to unite them, however, thanks to the presence of the Dfd parapeptide (Balavoine et al. 2002) in all of these sequences (Fig. 1). This parapeptide is very well conserved in all sequences from the platyhelminthes. This orthology group would also include Fhhbx2 (X66824) from Fasciola hepatica and ETOX-B (L19171), which are very similar to SmHox4.

Lox5 Orthologs

Lophotrochozoan Lox5 genes can be clearly distinguished by the presence of the conserved Lox5 parapeptide (Balavoine et al. 2002). In triclads, the Lox5 gene has duplicated, and therefore the PlHox6 group has two genes per species. In S. mansoni, Pierce et al. (2005) also found two Lox5 paralogs: Smox1 and the more divergent SmLox5, in which the Lox5 parapeptide is less well conserved. Smox1 is placed in scaffold Smp_scaff000004, together with the SmPost-2b gene. SmLox5, on the other hand, is located in an unplaced read in S. mansoni assembly 3.1, (schisto_3266f03.p1k).

We have found a Lox5 ortholog in the E. multilocularis genome, EmLox5 (Contig_0009134), with a well-conserved Lox5 parapeptide (Fig. 1). A very similar sequence from Taenia solium was identified in several ESTs (Genbank acc. nos. EL756693, EL759423, EL762691). This is the first identification of Lox5 orthologs in cestodes.

The PlHox6 genes from Girardia tigrina (DtHoxC, DtHoxE; Bayascas et al. 1997), Smox1, and EmLox5 all possess two very unusual introns within the homeodomain; the first is a phase-2 intron in codon 24, and the second is a phase-0 intron between codons 51 and 52. The more divergent SmLox5 shares the first intron position but not the second one. This gives further support to the hypothesis that all these genes are orthologous.

Planarian PlHox5 genes have been tentatively proposed to be orthologous to Hox5, based on overall similarity (Bayascas et al. 1997; Saló et al. 2001). It is interesting that the PlHox5 genes from the tricladida also share these intron positions. The Girardia tigrina DtHoxD gene (Bayascas et al. 1997) is known to have a phase-2 intron in codon 24. We also compared by Blastn the recently described cDNA sequence of the HoxD-like gene from Schmidtea mediterranea (EU082824; Iglesias et al. 2008) with the S. mediterranea draft genome assembly (AAWT00000000) and confirmed that it shares both intron positions. Furthermore, PlHox5 genes have Q in position 6 of the homeodomain, which is characteristic of Lox5, Hox7, Lox2, and Lox4 genes, instead of T6, as found in Hox5 genes. All these results suggest that PlHox5 genes are probably Lox5 orthologs (and therefore paralogs of PlHox6) that lost the Lox5 parapeptide after the duplication, and that Hox5 genes have not been found in the triclads. Similarly, we have been unable to find Hox5 genes in the S. mansoni and E. multilocularis genomes.

Lox2/Lox4 Orthologs

Several genes most similar to Lox2 and Lox4 have been recovered from the platyhelminthes before, although in many cases these sequences are too small to allow observation of the Ubd-A parapeptide. As pointed out by Olson (2008), these sequences are very similar to each other and are therefore easily identifiable. These include Distox-F from D. tigrina, the triclad PlHox7/8 group, the gene ETOX-C from Echinostoma trivolvis (L19172; see Olson 2008), SmHox8 from S. mansoni, and TasHox6/8 from T. asiatica.

In M. corti, we obtained a short sequence, MvHox7 (AY187808), that is 100% identical at the amino acid level to TasHox6/8. Southern blot analysis demonstrates that there is a single copy of this gene in M. corti (Fig. 3). In E. multilocularis, we identified the gene EmLox4 (Contig_0005027), which has a well-conserved Ubd-A parapeptide and several characteristic residues within the homeodomain (Fig. 1).

All “Ubd-A” genes from the platyhelminthes are most similar to the Lox4 genes of other lophotrochozoans, and they cluster together in our phylogenetic analysis (albeit with very low support; this node is collapsed in Fig. 2). They present two synapomorphic residues (Q21, [R/K]60) and two plesiomorphic residues (H24 and A35) described for Lox4 (Telford 2000a). No genes with Lox2-specific residues have been found in the platyhelminthes.

Post-2 Orthologs

Because of their divergent nature, few posterior genes have been recovered from flatworms by degenerate-PCR surveys. The only published posterior gene sequences belong to the triclad PlHox9 group, very simlar to Post-2 genes, which would include an unpublished G. tigrina gene (GtAbdB-b, Saló et al. 2001), Dugesia japonica Abd-Ba and Abd-Bb (Nogi and Watanabe 2001), S. mediterranea Abd-Ba (EST EG409633, and others) and AbdB-b (EST EG404975, and others), and a Dugesia ryukyuensis AbdB-b EST (BW635170). We have found two posterior genes in S. mansoni: SmPost-2a (two exons, in Smp_contig024481 and Smp_contig024482, which are contiguous in scaffold Smp_scaff000397; the first exon is present in Smp_087070 in GDBase, although it is incorrectly assembled) and SmPost-2b (two exons, in Smp_contig000845 and Smp_contig000846, which are contiguous in scaffold Smp_scaff000004; it is not present in GeneDB). An EST similar to part of the SmPost-2b gene is present in Schistosoma japonicum (AA143933). Two posterior genes were also identified in E. multilocularis: EmPost-2a (Contig_0005028) and EmPost2-b (Contig_0009147). All these genes contain a phase-0 intron between codons 44 and 45 of the homeodomain.

All identified posterior genes from flatworms are clustered with good support with Post-2 genes from other lophotrochozoans. Indeed, the alignment clearly shows that these genes have several Post-2-specific residues, such as M14, [I/V]15, C36, and K37, in addition to P7, which is characteristic of all posterior genes. The phylogenetic relationship among them is ambiguous, however, and very sensitive to parameter changes (data not shown). Thus, it is not possible to determine whether two Post-2 genes were present in the common ancestor of triclads and neodermatans, or whether independent duplications have occurred.

Three unpublished posterior genes were reported by Olson (2008) in the cestode Hymenolepis microstoma. One of these clearly clustered with Post-2 genes; the other two (Post-1a and Post-1b) were referred to as divergent Post-1-like genes, with uncertain phylogenetic affinities. We have not been able to identify any Post-1-like genes in S. mansoni and E. multilocularis.

Problematic Sequences

In previous degenerate-PCR surveys of Echinococcus granulosus and M. corti, we have found several short homeobox fragments (Genbank acc. nos. AF095860-62, AY187809-13). These sequences have been included in analyses by other groups (Kim et al. 2007; Olson 2008). Strikingly, there are no sequences similar to these in E. multilocularis. Furthermore, these sequences are most similar to vertebrate Hox genes, although they are not identical to any sequence in the GenBank nr database. Specifically, the EgHox10 and MvHox10 sequences are most similar to vertebrate Hox10 paralogs, but the amino acid identity with the best Blastp match (NP_032289) is only 63%.

Southern blot analyses under high stringency conditions failed to detect these sequences in either M. corti or E. granulosus, except in the case of EgHox1 where multiple bands were observed (data not shown). We have provisionally ruled out the possibility that they originated by contamination from the hosts Bos taurus or Mus musculus, because they are not present in the sequenced genomes, but they could have originated from some other contamination source.

It is possible that some of these sequences are true Hox or Hox-like genes from E. granulosus and M. corti. Therefore, we have chosen not to discontinue these sequences from the GenBank database yet, although we question their origin, and they should be used with caution.

Similarly, three Hox/Parahox sequences were found in E. multilocularis reads with very high similarity to vertebrate Hox/Parahox genes, even at the nucleotide level in introns and UTRs, when compared to Mus musculus and Rattus norvegicus (reads emu-907i19.p1k, emu-923f09.p1k, emu-956k24.p1k; data not shown). These reads are not present in the assembled contigs. We have interpreted them as possibly originating from host contamination.

In our analysis, we have been unable to find Hox7 orthologs in platyhelminthes. The only published sequence that shows similarity to this gene group is the very short sequence L19173, from Echinostoma trivolvis. This sequence, however, is also very similar to other orthology groups, and strangely, it is 100% identical at the nucleotide level to the CTs-Dfd homeobox (S76416) from the annelid Ctenodrilus serratus.

Finally, we have reanalyzed some sequences from the planaria Polycelis nigra and Phagocata woodworthi that have been considered orphans (Orii et al. 1999; Olson 2008). Pnox6 (Balavoine and Telford 1995; AAB17625) from Polycelis nigra shows the highest similarity in Blast searches to ind, the D. melanogaster ortholog of the Gsx Parahox gene. In all our phylogenetic analyses, it clusters with moderate support with Gsx genes (Fig. 2). Its sequence is, however, rather divergent. The short sequences PwoxF and PwoxG (L19174 and L19179) from Phagocata woodworthi are very similar to Pnox6. The Pnox5 homeodomain fragment (Balavoine and Telford 1995) shows the highest similarity in Blast searches to Mnx/exex genes, especially to the Mnx gene from the placozoan Trichoplax adherens (DQ355807).

Discussion

In this study we have found Hox genes in three neodermatan species and attempted to determine relationships of orthology among them and with other lophotrochozoans. According to our results and our reinterpretation of the PlHox5 group, all Hox groups previously identified in triclads (PlHox1-9) are present in the neodermatans: Hox1 (PlHox1), Hox2 (PlHox2), Hox3 (PlHox3), Hox4 (PlHox4), Lox5 (PlHox5, PlHox6), Lox4 (PlHox7/8), and Post-2 (PlHox9). This implies that four genes that have been proposed to be present in the ancestral lophotrochozoan are missing in neoophoran platyhelminthes: Hox5, Hox7, Lox2, and Post-1. (The Post-1-like genes from H. microstoma reported by Olson in 2008 have not been published, preventing us from including them in our analysis. They did not group, however, with other Post-1 genes in that work.) It is possible that one of these groups has been missed in our search and those of others; however, we consider this unlikely, especially in the case of S. mansoni, in which there is good genomic coverage and an independent degenerate-PCR strategy, and in triclads, where extensive research has been done. Furthermore, analysis of the S. mediterranea draft genome sequences did not produce Hox genes from any of the orthology groups missing in this work (data not shown).

We have also been unable to find Parahox genes in neodermatans, which parallels the results of PCR surveys. Parahox-like genes were found in Discocelis tigrina (Saló et al. 2001), and Pnox6 from Polycelis nigra might be a Gsx ortholog, as described before. Therefore, Parahox genes could be missing specifically in neodermatans.

Therefore, we propose that the last common ancestor of triclads and neodermatans had a Hox gene complement of at least seven or eight genes (depending on whether Post-2 duplication occurred only once or independently in several lineages). If Platyhelminthes is the most basal lophotrochozoan phylum studied yet, the absence of some of these genes (Lox2, and perhaps Post-1 and Hox7) could be interpreted as plesiomorphous (Fig. 4). Indeed, the presence of a single Lox2/Lox4 gene in triclads was interpreted as evidence of a basal position of platyhelminthes among lophotrochozoans (Saló et al. 2001). If the absence of Hox7, Post-1, and Lox2 is plesiomorphous, then the last common lophotrochozoan ancestor would have had a Hox gene complement of only eight genes [Hox1-5, a Lox5/Hox7 ancestor, a single Ubd-A gene (Lox2/Lox4 ancestor), and a single Post-1/Post-2 ancestor].

Fig. 4
figure 4

Two possible evolutionary scenarios for central and posterior Hox genes in neoophoran platyhelminthes. a Scenario inferred by Saló et al. 2001. Supposing playthelminthes were basal lophotrochozoans, some of the gene absences could be plesiomorphous (as the absence of Lox2 is interpreted here); also, the PlHox5 group is interpreted as orthologous to Hox5, making the intron positions in PlHox5 and PlHox6 convergent. b In this scenario, which maximizes the number of possible gene losses, the last common ancestor of all lophotrochozoans is hypothesized to have had 11 genes, four of which were lost in the lineage leading to neoophoran platyhelminthes. Color coding as in Fig. 2

Full size image

If, however, platyhelminthes are not basal to all other studied lophotrochozoan species (including members of Annelida, Mollusca, Brachiopoda, and Bryozoa; Balavoine et al. 2002; Kulakova et al. 2007), the clear implication is that Hox5, Hox7, Lox2, and Post-1 have been lost in the lineage leading to neoophoran platyhelminthes (the second scenario in Fig. 4). If Platyhelminthes is the most basal lophotrochozoan phylum, this scenario would still be equally valid.

On the basis of several grounds, we favor the gene-loss scenario. First, there is no consensus on the position of platyhelminthes among lophotrochozoans, and they cannot thus be assumed to be basal. Second, all Hox genes from the platyhelminthes show characteristics that link them to the corresponding specific genes of other lophotrochozoans (they do not show primitive characteristics). For example, Post-2 genes from the platyhelminthes share several synapomorphic residues with those of other lophotrochozoans. Assuming that they are basal to both Post-1 and Post-2 genes from other lophotrochozoans, these synapomorphies would have been subsequently lost in Post-1 genes. Similar arguments can be made for the position of Lox4 genes from flatworms. Finally, Lox5 genes from the platyhelminthes possess a well-conserved Lox5 parapeptide; if these genes are basal to Lox5 and Hox7 from other lophotrochozoans, this parapeptide would have had to be secondarily lost in Hox7 genes. Furthemore, Telford (2000b) has proposed that Lox5 is orthologous to ecdysozoan ftz genes, and that Hox7 could be orthologous to Antp. If this is correct, then both genes would have been present in the last common protostome ancestor and, therefore, in the last common lophotrochozoan ancestor.