Evolution of Hoxgene clusters in deuterostomes

Xenoturbellida and aceolomorpha: to be or not to be … deuterostomes

Xenoturbellids and acoelomorphs (acoels plus nemertodermatids) were classically classified as platyhelminthes. However, in the last decade and with the advance of molecular phylogenomics, their position has changed dramatically, depending on the methods and datasets used. Whereas acoelomorphs were classified as basal bilaterians [18–21] and Xenoturbellida formed a new phylum within deuterostomes [22, 23], the most recent analyses have grouped Xenoturbella and acoelomorphs in a monophyletic group, the so-called Xenacoelomorpha [24], either at the base of the Bilateria [25] or as a sister group of Ambulacraria within the deuterostomes [24]. Although the definite positions of xenoturbellids and acoelomorphs remain uncertain, here we will discuss the implications of the different possibilities.

Xenoturbellid Hox genes have been studied solely by a polymerase chain reaction (PCR) amplification survey in the species Xenoturbella bocki[26]. This study identified only an anterior Hox1, three central Hox genes (HoxM1, HoxM2 and HoxM3) and a posterior HoxP gene [26]. Acoels, on the other hand, have only three Hox genes: one anterior, one central and one posterior [27, 28]. Although acoel Hox genes are not clustered [28], they retain a spatially colinear expression pattern (we use the term colinearity with regards to PG order, regardless of gene clustering) [28, 29]. Regarding Nemertodermatida, only one study has been reported in which two central and one posterior Hox genes were identified by means of degenerate PCR [30].

Establishing a plausible evolutionary scenario for the Hox content of the LCA of Xenacoelomorpha remains difficult, because most of the sequences of the above-mentioned PCR surveys are very short and thus the datasets are poor. The simplest landscape is that of a single cluster consisting of one anterior, one central and one posterior Hox gene. If the Xenacoelomorpha is indeed confirmed as a group of the Deuterostomata, this would imply—compared with ambulacrarians and chordates—that the lineage suffered a massive loss of Hox genes and a complete disintegration of the cluster, at least in acoels. This would be in line with their simplified body plans from a more complex, Hox-rich, deuterostome ancestor. If eventually they are classified as the sister group of Nephrozoa (protostomes + deuterostomes), then their Hox content would nicely fit with that expected for their earlier bilaterian ancestor.

Ambulacrarian Hoxgenes

Within ambulacrarian groups, echinoderms are the most well studied, and several PCR surveys have been used to partially determine the Hox gene inventories of different species (see [31] and references therein), although the genomic organizations remain unknown in most cases (Figure 1). The exception within echinoderms is the well-known sea urchin Strongylocentrotus purpuratus, whose draft genome has been published [32]. S. purpuratus possesses a single cluster of about 600 kb in length that contains 11 Hox genes (Figures 1 and 2). However, the order of the S. purpuratus Hox genes in the cluster is altered in comparison with other deuterostome Hox clusters, so that some anterior genes are near to posterior ones and have different transcriptional orientations (Figures 1 and 2; [33]).

Regarding the hemichordates, the presence of a single Hox cluster has been identified recently in two different enteropneust species: Saccoglossus kowalevskii and Ptychodera flava[34]. Their Hox clusters show identical organization, with 12 Hox genes arrayed in ~500 kb, all with the same transcriptional orientation except for the terminal pair of Ambulacrarian-specific posterior Hox genes AmbPb and AmbPc (previously named Hox11/13b and Hox11/13c, respectively [34]). Overall, the hemichordate Hox cluster reflects a more prototypical organization than its sea urchin counterpart and its comparison with that of S. purpuratus allowed Freeman and colleagues [34] to infer the changes accounting for the scrambled condition of the latter. For example, S. purpuratus lacks a Hox member of the PG4 when compared with hemichordates. Therefore, sea urchins would have lost Hox4 at some point in echinoderm evolution, probably arising from the genomic rearrangements that provoked first the inversion of Hox1–5 genes and then the translocation of Hox1–4 to the 5′ end of the cluster, eventually causing the loss of Hox4[34]. Besides, crinoids (Metacrinus rotundus) and asteroids (Asterina minor and Patiriella exigua) have a Hox4 gene [35–37]; thus, the loss of Hox4 must have occurred at least after the split of echinoids, holothuroids, and ophiuroids from the rest of the echinoderms (Figure 1). Nonetheless, the genomic sequences of more ambulacrarian lineages, mainly those within the group of echinoderms, must be investigated to depict a precise evolutionary scenario for the ancestral ambulacrarian Hox cluster.

Hoxgenes of cyclostomes: the main vertebrate gap

Cyclostomes (the only extant group of agnathans or jawless vertebrates) are the sister group of gnathostomes (jawed vertebrates) and are much less known than the latter, also in terms of Hox gene content. Cyclostomes are composed of two different groups: lampreys and hagfishes. In the case of the hagfish, the most recent report is that of Stadler et al. [56], who obtained up to 33 Hox genes using degenerate PCR, a lower number than that of a general gnathostome, but still compatible with multiple clusters. The authors [56] concluded from their analyses that cyclostomes had split at least after the first WGD and that subsequent independent gene/cluster duplications expanded their Hox inventory. However, their analysis was based on very short sequences, so their conclusions remained very speculative. A conclusive repertoire of Hox genes and clusters of the hagfish is far from definitive, since the genomic organization of the hagfish Hox genes still remains a mystery. New advances in hagfish research [57, 58] together with the new and powerful sequencing techniques available will surely help to fill this gap in the future.

Regarding lampreys, several PCR surveys have been done using different lamprey species: Petromyzon marinus[59–63], Lampetra planeri[64] and the Japanese lamprey Lethenteron japonicum[44, 65, 66]. Moreover, the draft genome of the sea lamprey (P. marinus) has been published recently [67] and the Hox genes were investigated. A total of 25 Hox genes were found but only two Hox clusters were recognizable: clusters Pm1Hox (with Hox2, 3, 4, 5, 6, 7, 8, 9 and 11) and Pm2Hox (Hox1, 4, 5, 7, 8, 9, 10 and 11). Besides these clusters, eight additional genes were found lacking genomic information, except for two of them that are linked, indicating the presence of a putative third cluster [67].

The number of Hox clusters of the lamprey can be estimated to be four or five: there are four cognates for at least PG1, 8 and 10 (see Table 1); strikingly, there are five distinct genes for PG4, 9 and 11 and although there are four Hox1 genes, the cluster Pm1Hox does not seem to have a Hox1 member [67]. Conversely, there might be fewer clusters, depending on allelic polymorphisms or independent tandem duplication events within a cluster [63]. Nonetheless, none of the above-mentioned scenarios is conclusive and further investigation is needed. At present, it is not clear whether lampreys hold representatives of the four gnathostome Hox clusters or whether some originated by independent duplication events (for example, if five clusters are confirmed) [68]. Finally, phylogenetic analysis of dozens of gene families [69] and the recent analysis of the P. marinus genome [67] pointed to a post 2R-WGD origin of cyclostomes, indicating that the LCA of vertebrates most probably possessed four Hox clusters. Accordingly, the synteny of non-Hox genes linked to the Hox clusters has been generally conserved between cyclostomes and gnathostomes, although with a differential retention of paralogues [67].

Condrichthyans: a Hoxcluster loss

The gnathostomes consist of two main groups: condrichthyans (cartilaginous fishes) and osteichthyans (bony vertebrates); and all of them possess members of the four Hox clusters (A, B, C and D), indicating that they diverged after the 2R-WGDs [70]. The condrichthyans are divided into two subclasses: Elasmobranchii (sharks, skates and rays) and Holocephali (chimaeras).

Apart from some partial reports on the horn shark Heterodontus francisci (elasmobranch) [71–73], the first complete condrichthyan Hox repertoire was reported in a chimaera, the elephant shark Callorhinchus milii (holocephalian). C. milii possesses a total of 45 Hox genes, retaining the HoxD14 gene, plus two Hox14 pseudogenes (A and B, see Figure 2). The C. milii Hox repertoire substantially increased the putative Hox contents of the gnathostome ancestor [74]. Surprisingly, elasmobranchs seem to have lost the HoxC cluster completely. Studies with both the lesser-spotted cat shark Scyliorhinus canicula and the little skate Leucoraja erinacea did not find any of the HoxC cluster members, including sequences encoding Hox cluster-associated microRNAs [75, 76]. These are the first reports on the loss of a complete Hox cluster type (A, B, C or D) within jawed vertebrates, a loss that dates back to 250 Mya [77].

Osteichthyans

The Hox clusters of osteichthyans are the most well-known within vertebrates and deuterostomes. Here we include recent reports on newly investigated species that have changed the overall scenario of the previously inferred ancestral conditions [5].

Osteichthyans comprise two major groups: sarcopterygians (coelacanths, lungfish and tetrapods) and actinopterygians (ray-finned fishes, including teleosts). Within the former, the complete repertoire of many tetrapods (chicken, mouse and human genomes among others) is known [5, 78]. Comparison of the Hox inventories of different tetrapods allows to infer a tetrapod ancestral condition of up to 41 genes [78], one more than previously thought [5], and an amniote ancestral condition of 40 Hox genes (after the loss of HoxC1; Figure 2), of which only the green anole (Anolis carolinensis) retains all of them [78, 79], while mammals and the chicken have lost the HoxC3 gene independently. Liang et al. [78] also did not identify HoxC3 in crocodiles and turtles by degenerate PCR and it is absent from turtle genomes [80, 81]. Therefore, HoxC3 was probably lost before the evolutionary split of archosaurians and turtles. Amphibian repertoires vary: Xenopus tropicalis has 38 Hox genes, three fewer than in the ancestral tetrapod. One of these lost genes is HoxC1, also lost from amniotes. Because HoxC1 is still present in the amphibian caecilian Ichthyophis bannanicus[78], it means that it was lost independently in amniotes and X. tropicalis (Figure 2). Taking into account data from I. bannanicus and the salamander Batrachuperus tibetanus[78], the amphibian ancestor probably had 40 genes, after losing HoxD12 from its tetrapod ancestor (Figure 2).

The Hox clusters of the coelacanths Latimeria menadoensis[82] and Latimeria chalumnae[83], the closest living relative to the sarcopterygian ancestor [83], include a HoxA14 gene that was later lost in tetrapods. Moreover, a PCR survey of the lungfish Protopterus annectens also found a HoxA14 gene ([78]; data from the lungfish are not included in Figure 2 because they lack clustering information, but they were taken into account to infer the ancestral condition). This enabled Liang and colleagues [78] to reconstruct a more complete ancestral sarcopterygian Hox inventory, with a total of 43 Hox genes (Figure 2).

The case of actinopterygians is rather more complicated, because after the teleost-specific 3R-WGD, more genes—and in some cases a complete cluster—were lost differentially (Figure 2). Teleosts are the most numerous group of vertebrates with more than 27,000 extant species [84]. It has been already 15 years since more than four Hox clusters were found in the cyprinid zebrafish Danio rerio[60, 85], which instead has seven Hox clusters: HoxAa, HoxAb, HoxBa, HoxBb, HoxCa, HoxCb and HoxDa. The cluster HoxDb has been reduced to only the Hox-related microRNA miR-10d[86]. Another cyprinid, Megalobrama amblycephala, might also lack the HoxDb cluster [87], implying that it was probably lost in the cyprinid ancestor. Other teleosts with known Hox complement include pufferfishes, cichlids, medakas, sticklebacks, salmonids and some early lineages such as the eel and the bichir (although data are not complete for the last species). The Hox complement of several pufferfishes has been already reported: Spheroides nephelus[88], Takifugu rubripes[88, 89] and Tetraodon nigroviridis[90] showing differential species-specific losses (see Figure 2). The case of T. rubripes is curious. Amores and colleagues described a possible duplication of the HoxAa cluster in T. rubripes, which they called HoxAc[88]. This cluster duplication was considered doubtful by other authors [4, 5]. Later it was found that the sequence of this HoxAc cluster, present in version 2.0 of the genome of T. rubripes, corresponded to a bacterial artificial chromosome (BAC) clone of the Nile tilapia and was not present in later versions of the T. rubripes genome [91]. Comparisons of the Hox complement of the above-mentioned pufferfishes with those of the medaka Oryzia latipes[92], cichlids such as the Nile tilapia Oreochromis niloticus[93] (not present in Figure 2 because the clusters are largely not sequenced), of Astatotilapia burtoni[94, 95] and the three-spine stickleback, Gasterosteus aculeatus[94], show a common loss of the HoxCb cluster in their LCA (the ancestral Neoteleosteii; see Figure 2). Moreover, independent losses of Hox genes are not rare: for example HoxBa7 and HoxCa1[94, 95] have been lost several times (Figure 2). HoxBa7 is present in O. niloticus and thus was independently lost in the medaka and A. burtoni.

Strikingly, the Atlantic salmon Salmo salar and the rainbow trout Oncorhynchus mykiss contain 13 Hox clusters [55, 96, 97], arising from a salmonid-specific 4R-WGD [54], the former with a total of 118 Hox genes plus eight pseudogenes: the largest Hox repertoire to date [55]. S. salar lost the HoxDb cluster before the 4R-WGD (as did zebrafish independently). S. salar still retains HoxD1a copies (α and β) within the HoxDa cluster. Therefore, HoxD1a was indeed present in the teleost ancestor, in contrast to previous ideas [5]. The complete repertoire of Hox clusters of the European eel Anguilla anguilla, representing the earliest branch of teleosts [98], helped in reconstructing a more reliable teleost ancestral inventory [99]. A. anguilla also retains HoxD1a, like S. salar, meaning that it was independently lost later in zebrafish and neoteleosts (e.g., medakas, pufferfishes, cichlids and sticklebacks; Figure 2). A. anguilla is the best representative of the teleost ancestor, having lost only four Hox genes (three of them are pseudogenes) and retaining all eight Hox clusters after the 3R-WGD ([99]; Figure 2). The osteoglossomorph Hiodon alosoides (the goldeye), a representative of a basal branch diverging after the eels [98], also seems to have retained the original eight clusters [100]. In summary, the teleost ancestor probably had at least 74 Hox genes, with eight Hox clusters (after the 3R-WGD), quickly followed by differential gene and cluster losses in the different lineages ([99]; Figure 2).

There are few reports on Hox genes of fishes outside the teleosts. Several studies report either few Hox genes or partial information about Hox clusters for the Senegal bichir Polypterus senegalus. The bichir is a representative of the earliest branch of actinopterygians that diverged prior to the 3R-WGD. Only the HoxA cluster sequence out of a total of four clusters has been reported thus far [101, 102]. There is also a report of cDNA screening in another bichir species, Polypterus palmas[103]. Although again, while the data presented are fragmented and not complete, these studies allow to infer the presence of HoxD2 in the LCA of osteichthyans (and actinopterygians), a gene that was lost in all the osteichthyans species studied so far except in the bichir ([102]; Figure 2). In another gnathostome species, the catshark S. canicula, HoxD2 is expressed in very important tissues such as the pharyngeal arches and the rhombencephalon [104]. However, we predict that the loss of this gene may have little importance in the general patterning of such tissues, because the expression of Hox genes is quite redundant; once a gene is lost, other Hox genes patterning the pharynx and the hindbrain could compensate for its function. The information for other early branches of actinopterygians (the paddlefish, sturgeon, gar and Amia calva), diverging after the bichir, is also scattered and partial [6, 105], although it is important to note that some of them, for instance sturgeon and paddlefish, probably have more than four Hox clusters caused by lineage-specific WGDs [105]. In fact, the American paddlefish Polyodon spathula has at least two HoxA and two HoxD clusters (each termed α and β, to denote their different duplicative origins from the teleost 3R-WGD). Surprisingly, P. spathula retains a HoxD14β paralogue [105], indicating that, unlike as previously inferred, HoxD14 was present in the LCA of actinopterygians (and thus in the osteichthyan LCA) and was secondarily lost in the rest of the lineages of the group (shown in Figure 2).

Future perspectives

We have presented here a catalogue and a current view on the evolution of Hox gene families in deuterostomes, showing that we are still far from picturing a conclusive scenario for the ancestral conditions. In fact, many early branches within vertebrates remain to be examined, as do more invertebrate deuterostomes (for example, more ambulacrarians); hence we cannot exclude the existence of more Hox genes in the ancestral states until the number of sequenced genomes from non-model animals has increased considerably. With the new genome projects being carried out currently in a wide range of animals, mainly in vertebrates (e.g., the 10 K Genome Project [109]), this landscape will change in the very near future. On the other hand, the problematic origin of the posterior Hox genes will not be solved just by reporting more Hox genes, but also will require improved methods, both phylogenetic and non-phylogenetic.