Identical genomic organization of two hemichordate Hox clusters

Summary

Genomic comparisons of chordates, hemichordates, and echinoderms can inform hypotheses for the evolution of these strikingly different phyla from the last common deuterostome ancestor [1–5]. Since hox genes play pivotal developmental roles in bilaterian animals [6–8], we analyzed the Hox complexes of two hemichordate genomes. We find that Saccoglossus kowalevskii and Ptychodera flava both possess 12-gene clusters, with mir10 between hox4 and hox5, in 550kb and 452kb intervals, respectively. Genes hox1-9 of the clusters are in the same genomic order and transcriptional orientation as their orthologs in chordates, with hox1 at the 3′ end of the cluster. At the 5′ end, each cluster contains three posterior genes specific to Ambulacraria (the hemichordate-echinoderm clade), two forming an inverted terminal pair. In contrast, the echinoderm Strongylocentrotus purpuratus contains a 588kb cluster [9] of 11 orthologs of the hemichordate genes, ordered differently, plausibly reflecting rearrangements of an ancestral hemichordate-like ambulacrarian cluster. Hox clusters of vertebrates and the basal chordate amphioxus [10] have similar organization to the hemichordate cluster, but with different posterior genes. These results provide genomic evidence for a well-ordered complex in the deuterostome ancestor for the hox1-9/10 region, with the number and kind of posterior genes still to be elucidated.

Results and Discussion

Here we characterize the order, transcriptional orientation, and clustering of the Hox genes of the genomes of two widely studied model hemichordates, S. kowalevskii and P. flava [4, 11, 12] that represent two major evolutionary branches of enteropneust hemichordates diverged by an estimated 400 MYa [13]: Saccoglossus, of the direct developing harimaniids, and Ptychodera, of the indirect developing ptychoderids [14]. Both are vermiform marine animals with gill slits, a non-segmented body, and a muscular proboscis for burrowing (so-called “acorn worms”). A third branch, not included here, is that of pterobranch hemichordates, which may be an offshoot of the harimaniids or basal to both branches [14]. Together, the phyla of hemichordates and echinoderms (diverged an estimated 540 MYa [13]) comprise the Ambulacraria, sharing a last common ambulacrarian ancestor. Similarly, the Ambulacraria and the chordate phylum comprise the Deuterostome superclade, sharing a last common deuterostome ancestor. The similarities of Hox genes and their clusters across these groups can, if found, provide information about the Hox complement of the ambulacrarian and deuterostome ancestors that gave rise to these three diverse phyla [3–5].

Characterization of the Hox genes of S. kowalevskii and P. flava

Following the isolation of ANTP-class Hox gene sequences from the sea urchin S. purpuratus [15], an echinoderm, cDNA clones were obtained from hemichordates for 11 of 12 Hox genes in S. kowalevskii (excepting hox8) [16, 17], and 8 of 12 in P. flava (excepting hox2, 3, 7, and 8) [18]. At the time, there was no information on the genomic linkage of Hox genes in hemichordates. For hemichordate genes hox1-8, numerical identities were preliminarily assigned based on the best match of the homeodomain sequence to paralogous groups of vertebrate Hox genes and to the sea urchin genes. Hox9/10 and hox11/13a, b, c, which were first named in S. purpuratus [15] because of their broad similarity to vertebrate posterior genes, were used as gene names for subsequently discovered ambulacrarian orthologs. The hox1-11/13c (except for hox8) genes of S. kowalevskii are expressed along the antero-posterior axis of embryonic ectoderm and juvenile epidermis [16, 17] in the same order as their assigned numerical gene names, and similar to the ordered expression of the orthologous Hox genes in the vertebrate central nervous system and in the ectoderm of protostomes such as Drosophila [6, 7].

In the course of our genomic analysis, we obtained the previously uncharacterized Hox genes of S. kowalevskii and P. flava (see Supplementary Material). The total complements of 12 genes each for these species are summarized in the phylogenetic tree of Figure 1 and the homeodomain sequence alignments of Figures S1 and S2. Membership of each gene in a vertebrate-defined paralogous group was further assessed by submitting the homeodomain sequences to HoxPred [19]. From these analyses, hox1-8 of the two hemichordates are found to strongly resemble each other as well as hox genes of echinoderms and chordates. The hemichordate sets contain: 1) two anterior genes (hox1, 2) probably descended from PG1 and PG2 orthologs dating back to the bilaterian ancestor, 2) a group3 gene (hox3) probably descended from a PG3 ortholog of that ancestor, and 3) five central genes (hox4, 5, 6, 7, 8) that can be matched with chordate counterparts.

Figure 1. Molecular phylogenetics of the hemichordate Hox genes.

A maximum-liklihood phylogram was constructed from the 60-amino-acid homeodomain sequences from 64 Hox genes, and using two NK2.1 genes as an outgroup. The Hox genes of the hemichordates P. flava and S. kowalevskii are shown in blue. Red circles indicate genes characterized anew in this study. PhyML was used (www.phylogeny.fr/version2_cgi/one_task.cgi?task_type=phyml see Supplementary Material). The aLRT branch support values are shown as percentage in red. Branches with support values under 50% were collapsed. Branch length is indicated by the bar in the lower right. Pf, Ptychodera flava; Sk, Saccoglossus kowalevskii; Sp, Strongylocentrotus purpuratus; Mr, Metacrinus rotundus (for hox4, which is absent in S. purpuratus); Am, Branchiostoma floridae (amphioxus); and Mm, Mus musculus.

The four posterior genes (hox9/10, 11/13a,b,c) of the hemichordate sets, while clearly similar to each other and to the echinoderm posterior genes and while having PG9 character, have only poorly supported phylogenetic relationships to specific chordate posterior genes, as others have noted [15, 20]. From phylogenetic analysis of the 60aa homeodomains (Figure 1) as well as of 75aa blocks encompassing the homeodomain and conserved among posterior hox genes (Figure S3), we find that the hemichordate and echinoderm hox9/10 genes, in keeping with their name, group with vertebrate hox9 and somewhat less well with vertebrate hox10, and Amphioxus hox9 and 10, whereas the other genes (hox11/13a, b, c of ambulacraria) form a phylogenetic group separate from those of vertebrate hox11-14 and amphioxus hox13-15, and appear to have diverged within the ambulacrarian lineage (Figures S2 and S3). Lacking evidence for orthology to vertebrate hox11, 12, 13, and with evidence for their close relatedness to the echinoderm genes, we suggest the 11/13a,b,c genes be renamed as ambPa,b,c (ambulacrarian Posterior a,b,c). The extensive diversification of posterior gene sets in the different lineages of deuterostomes, as compared to their anterior and central gene sets, has been attributed to “deuterostome posterior flexibility” [20] and to multiple independent duplications [19], as discussed later.

Assembly of the Hox complexes

For both hemichordates, the genomic analysis entailed the isolation of overlapping BAC sequences carrying subsets of hox genes, complemented by the assembly of whole genome shotgun sequences to produce large scaffolds or supercontigs containing the entire clusters (see Supplementary Material). For S. kowalevskii a single scaffold (Scaffold_166 of 951kb; Figure 2A) contains the 12-gene cluster. Nine genes (hox1 to 9/10) are arranged in the same order and transcriptional orientation as their vertebrate orthologs, with hox1 at the 3′end, and with three posterior genes at the 5′ end, namely, hox11/13a in tandem with hox1-9/10 but hox11/13c and 11/13b inverted as a terminal pair. The entire cluster, from exon2 of hox1 to exon2 of the inversely oriented hox11/13b, spans a genomic distance of 550kb. A hox8-like homeodomain sequence was recovered in the interval between hox7 and hox9/10 (deposited as a GNOMON model [Genbank gi|291221533|]). Furthermore, a mir10 sequence was found between hox4 and hox5.

Figure 2. Clustering of Hox genes in hemichordate genomes.

Genomic regions approximately 500 kb-long are shown for (A) Saccoglussus kowalevskii and (B) Ptychodera flava. Both contain the twelve genes Hox1 to Hox11/13b (blue arrows in the direction of transcription). In both, the ten genes Hox1-Hox11/13a are aligned in the same direction, while two genes, Hox11/13c and Hox11/13b, are in the opposite direction. Red bars indicate the position of mir10 genes, and the orange bar in (B) indicates a gap. BAC clones are shown as green lines, scaffolds or contigs obtained from whole-genome shotgun reads as purple lines, and PCR-amplified fragments as brown lines.

For P. flava, four BAC clones and three genome shotgun-derived contigs were linked to form a supercontig of 455.7kb, which included the 12 P. flava Hox genes within a 451.7kb-long genomic region (Figure 2B). Remarkably, the organization of the cluster is the same as that of S. kowalevskii, namely, hox1 to 9/10 are arranged in the same order and transcriptional orientation as their vertebrate orthologs, with hox1 at the 3′end, and three posterior genes are arranged at the 5′ end, with hox11/13a in tandem with hox1-9/10 but hox11/13c and 11/13b inverted as a terminal pair. A mir10 sequence resides between hox4 and hox5.

From the assembled complexes we estimated the exon-intron structure, size, and intergenic distances of the 12 genes of each (Table S1A and B). For S. kowalevskii, all genes have two exons except hox9/10, which has four, and all homeobox sequences are contained within single exons. The average primary transcript size and mRNA size for the 12 genes are 5,698 and 2,651 nucleotides, respectively (based on cDNA sequences placed against the Scaffold_166 sequence; Supplementary Materials). An intergenic distance of 127kb separates hox11/13a and 11/13c; that is, it separates the group of ten tandemly ordered genes (hox1-hox11/13a) from the inverted pair (hox11/13c and 11/13b). When this interval is omitted, the average intergenic distance is 36kb for the remaining genes of the cluster.

For the P. flava complex, all genes have two exons except hox5 and hox7, which have three, and all homeobox sequences are contained in single exons (The estimates of different numbers of exons for hox5, 7, and 9 between the two species are not definitive because they are based on cDNA analysis in some cases and on gene prediction in others; Table S1A and B). The average mRNA size is 1691 nucleotides in the P. flava complex. The average intergenic distance is 34kb for all genes of the cluster. Like S. kowalevskii, the largest (66kb) intergenic space occurs between the group of ten tandem genes (hox1-11/13a) and the inverted pair (hox11/13c and b).

In both complexes, no other protein-coding gene was identified within the cluster by routine sequence searches (Figure 2A and B). However, for S. kowalevskii, two sense and two antisense non-coding transcripts of unknown function have been detected in and near the cluster (Table S4, Supplementary Materials). Outside the S. kowalevskii cluster, an astacin-like protease sequence and a thioredoxin4-like sequence reside within 11kb and 47kb, respectively, of the terminal hox11/13b gene. However, within the 110 kb interval from hox11/13b to the end of Scaffold_166, no evx sequence was found, whereas evx is located adjacent to posterior genes of vertebrate clusters A and D, the amphioxus cluster, and the sea urchin cluster. The S. kowalevskii evx gene is on a separate Scaffold_145 (954kb), 120kb from one end; we have no evidence linking this scaffold to that of the Hox cluster.

In summary, these two hemichordate species, separated by an estimated 400 My [14], possess 12-gene Hox clusters of identical organization, with nine genes (hox1-9/10) arranged in the same genomic order and transcriptional orientation as their orthologs in chordates, with hox1 at the 3′ end of the cluster. At the 5′ end, both clusters contain three posterior genes specific to Ambulacraria, two forming an inverted terminal pair. The cluster sizes are 550 and 452kb respectively for S. kowalevskii and P. flava.

Comparison of deuterostome Hox clusters

Figure 3 summarizes the genomic Hox clusters for the two hemichordates reported here and for the sole echinoderm example, the 11-gene cluster of S. purpuratus, which has a different order and orientation of genes from the hemichordate clusters. These differences can be interpreted as arising by way of rearrangements of an ancestral ambulacrarian cluster having hemichordate-like organization (Figure 3, hypothetical cluster 1). That is, five genes in the middle of the echinoderm cluster (hox6, 7, 8, 9/10, 11/13a) have the same order and orientation as their orthologs in the hemichordate clusters (Figure 3; dotted lines). If within the lineage to sea urchins, the ancestral ambulacrarian cluster underwent an inversion of the [hox1, 2, 3, 4, mir10, 5] group of genes at the 3′ end of the cluster, and then a translocation of the inverted genes [hox1, 2, 3, 4, mir10] but not hox5 to the 5′ end of the cluster, followed by loss of hox4 but retention of mir10, plus an inversion and translocation of hox11/13c, the outcome would be the present-day echinoderm S. purpuratus cluster of mixed orders and orientations. The loss of hox4 presumably occurred after the echinoid lineage split from crinoids and asteroids, which have hox4 [21, 22]. This and other rearrangement schemes have been previously discussed [9, 15].

Figure 3.

Summary of Hox clusters of extant two hemichordates and one echinoderm (Ambulacraria) and two hypothetical reconstructions of the cluster of the ambulacrarian ancestor, arranged in a phylogenetic tree [1]. Hox gene location and transcriptional orientation are indicated by triangles, the colors of which represent the different categories of Hox genes: anterior (purple), group3 (yellow), central (green) and posterior (red). The mir10 locus is indicated by name, between hox4 and 5. Note the identical organization of the six gene segment (hox6,7,8,9/10,11/13a, 11/13c) of the echinoderm (S.purpuratus, sea urchin) and hemichordate clusters (bounded by dashed lines). In the hypothetical cluster 1 (upper left), the minimal ancestral cluster is proposed to be the same as the extant hemichordate clusters, including the inverted hox11/13b,c pair of genes, Then, within the lineage to sea urchins, the cluster underwent rearrangements and gene loss (hox4) as well as an inversion and translocation of hox11/13c. In the hypothetical cluster 2 (lower left), the minimal ancestral cluster is proposed to be like that of hemichordates except that hox11/13b and c are in tandem with the other genes, not inverted as a pair. Then within the echinoderm lineage, the cluster underwent rearrangements of hox1-5 and gene loss (hox4), as well as an inversion, but not translocation of hox11/13b, while hox11/13b and c inverted as a pair in the hemichordate lineage. The proposal of the kinds of hox genes in the ancestral cluster is supported by the hox sequence repertoire across extant Ambulacraria [6,18, 20, 21]. Whereas hox1-9 (called 9/10 in Ambulacraria) are orthologous to chordate genes 1–9, the ambulacrarian posterior genes 11/13a,b,c differ from chordate genes 11–13, and are identified here as AmbPa,b,c to indicate their ambulacrarian-specific ancestry.

From the comparisons of cluster organization and from Hox sequences shared across ambulacraria (including the ptychoderid hemichordate Balanoglossus simodensis [23] and the echinoderms Metacrinus rotundus and Heliocidaris tuberculata, which possess the same ambulacrarian-specific posterior Hox genes [Figure S3]), we suggest in Figure 3 (hypothetical cluster 1) that the ambulacrarian ancestor possessed minimally a 12-gene Hox cluster similar to the hemichordate cluster, with at least nine genes (hox1-9/10) arranged in a genomic order and transcriptional orientation the same as their chordate orthologs. Of the three remaining posterior genes, one (11/13a;ambPa) was probably in tandem order and orientation with the hox1-9/10 genes, and two (hox11/13c and b; ambPb and c) were inverted at the 5′ end of the ancestral cluster, like the posterior genes of extant hemichordate clusters (and like hox11/13a and hox11/13c of S. purpuratus). Alternatively, as diagramed in Figure 3 (hypothetical cluster 2), the ambulacrarian ancestor may have had the 11/13b and c genes, not inverted, but in tandem with all other genes of the cluster, and then the 11/13b,c genes inverted as a pair in the hemichordate lineage while only 11/13 b inverted in the echinoderm lineage (with the same rearrangements and gene loss for hox1-11/13a in the echinoderm lineage as described above). Furthermore, in light of the large number of chordate posterior Hox genes, the ambulacrarian ancestor may have possessed more than four, with parallel gene loss in the hemichordate and echinoderm lineages or in their common stem lineage before the ancestor (see Figure S4). Further elucidation of the ancestral cluster requires more information about hox sequences and cluster organization from additional ambulacraria.

From shared features of the chordate and ambulacrarian clusters, we suggest, as have others [15, 20], that the last common deuterostome ancestor had a well ordered cluster of at least nine genes including a hox9-like gene (hox9/10 in ambulacraria). The number and kind of additional posterior genes of this ancestor remain difficult to estimate due to the differing numbers of such genes in different extant lineages (7 in Amphioxus, 6 in vertebrates, 4 in Ambulacraria, and 1 or 2 in protostome outgoups) and their differing homeodomain sequences, yielding only weak supported orthologies by phylogenetic analysis. Sequence differences of the posterior genes in deuterostome evolution have been attributed to multiple gene duplications [19], gene loss, and deutersotome posterior flexibility (extensive homeobox sequence change [20]).

We join others [6, 19, 20, 24] in drawing attention to the extensive evolutionary modifications of posterior Hox genes in deuterostomes, far more than has occurred for their anterior and central genes. This greater gene evolution at the 5′ end of the cluster raises questions of whether significant Hox-related morphological evolution occurred at the posterior end of the body axis of early deuterostomes, where these genes were presumably first expressed (though co-option of them may have occurred later for other uses). In extant chordates, they are expressed in the posterior body including the post-anal tail, a phylotypic trait of chordates (fin and limb expression arising later in fish and tetrapod vertebrates), and in S. kowalevskii they are expressed in the posterior body and the contractile, adhesive post-anal extension (sometimes called “tail-like” or “stalk-like”) of the juvenile animal [16, 17]. (This post-anal region is, however, absent in extant ptychoderids such as P. flava). Among other deuterostomes (not yet analyzed genomically), pterobranch hemichordates possess a contractile stalk (separated from the anus), crinoid echinoderms have stalks, and echinoderm-related fossils (e.g., solutes) reveal muscular stalks/holdfasts [25]. We favor the possibility that the deuterostome ancestor, and then early stem Ambulacraria and chordates, engaged in posterior axial innovations such as tails, stalks, and holdfasts, some with dorsoventral asymmetry to exclude the gut (hence, post-anal), and that the evolution of these diverse modifications involved the expansion, loss, and sequence diversification of posterior Hox genes in the various lineages (Figure S4).

In summary, we find that two hemichordates, separated by an estimated 400 MYa [14], have the same genomic organization of their 12-member Hox clusters, and that apart from differences at the posterior end of this cluster, the hemichordate arrangement matches that of amphioxus and vertebrates. Thus, these hemichordates should provide us with useful new systems in which to address the function and regulation of Hox genes and gain greater insight into the ancestral deuterostome condition.

Highlights.

• *The 12-gene Hox clusters of two hemichordates have identical organization.
• *hox1-9/10 are arranged like their orthologs in the vertebrate cluster, hox1 at the 3′ end.
• *Three posterior genes are at the 5′ end of the cluster, two inverted as a terminal pair.
• *The ancestral ambulacrarian cluster was likely similar to the hemichordate cluster.