Hox gene clusters of early vertebrates: do they serve as reliable markers for genome evolution?

Hox genes, responsible for regional specification along the anteroposterior axis in embryogenesis, are found as clusters in most eumetazoan genomes sequenced to date. Invertebrates possess a single Hox gene cluster with some exceptions of secondary cluster breakages, while osteichthyans (bony vertebrates) have multiple Hox clusters. In tetrapods, four Hox clusters, derived from the so-called two-round whole genome duplications (2R-WGDs), are observed. Overall, the number of Hox gene clusters has been regarded as a reliable marker of ploidy levels in animal genomes. In fact, this scheme also fits the situations in teleost fishes that experienced an additional WGD. In this review, I focus on cyclostomes and cartilaginous fishes as lineages that would fill the gap between invertebrates and osteichthyans. A recent study highlighted a possible loss of the HoxC cluster in the galeomorph shark lineage, while other aspects of cartilaginous fish Hox clusters usually mark their conserved nature. In contrast, existing resources suggest that the cyclostomes exhibit a different mode of Hox cluster organization. For this group of species, whose genomes could have differently responded to the 2R-WGDs from jawed vertebrates, therefore the number of Hox clusters may not serve as a good indicator of their ploidy level.

Introduction

Hox genes in eumetazoan genomes are generally organized into clusters (. During development, the orders of genes comprising the Hox clusters are converted into information governing specification of different body compartments along the anteroposterior axis. In vertebrates, this scheme was originally presented for the patterning of the hindbrain and pharyngeal arches (“Hox code”) (. Later, similar patterns were observed in the development of limb buds ( and gut endoderm (. With some exceptions, Hox genes closer to the 3′-end of Hox clusters are expressed earlier in development (temporal collinearity) and also more anteriorly in embryos (spatial collinearity) (. In all vertebrate genomes sequenced to date, Hox genes are found in multiple clusters, and the number of Hox clusters exactly (in tetrapods) or roughly (in teleost fishes) corresponds to their ploidy levels, indicating how many times those species experienced whole genome duplications (WGDs) (.

Hox gene regulation is often introduced as one of the most striking examples of conserved molecular programs underlying conserved morphological architecture of animal body plans. However, even among vertebrate model systems serving as traditional laboratory animals, Hox cluster composition varies to some extent as a result of secondary gene losses and cluster doublings (. Most importantly, our knowledge about vertebrate Hox cluster organization and gene regulation has concentrated mostly on traditional laboratory model systems, all of which belong to a subgroup of vertebrates, Osteichthyes (bony vertebrates). With all technical factors regarding access to animal resources, recent development in genomic and embryonic studies on cyclostomes and cartilaginous fishes is providing a new frontier of vertebrate Hox studies. In this review, focusing on these early-branching evolutionary lineages, I summarize the progress broadening our scope, which previously depended exclusively on osteichthyans. In addition, insights into evolutionary processes of Hox gene cluster organization are also presented.

Hox Genes in Cartilaginous Fishes and Cyclostomes

Hox gene repertoire in cartilaginous fishes

Among cartilaginous fishes, the first report of Hox cluster organization came from the horn shark Heterodontus francisci (. This species is categorized in the order Heterodontiformes, and this order is regarded as the most basal among extant members of Galeomorphi. In this species, only two Hox clusters, designated HoxM and HoxN, were sequenced. Presence and absence of particular Hox paralog groups, as well as lengths of intergenic regions and molecular phylogeny, consistently supported the homology of HoxM and HoxN clusters to HoxA and HoxD, respectively 7, 8.

Later, in the elephant shark (also called ghost shark or elephant fish), Callorhinchus milii, genomic sequencing of Hox clusters was completed based on the 1.4× whole-genome shotgun reads and targeted bacterial artificial chromosomes (BAC) clone screening 9, 10. Importantly, this species retains four Hox clusters, containing 45 Hox genes in total (compared to 39 Hox genes in human). Also, 1-to-1 homology of these four clusters to osteichthyan HoxA-D was firmly supported. In C. milii, seven paralog groups (PGs) were detected with four members on clusters A-D, namely PG-1, -3, -4, -5, -9, -10 and -13, while there are only three in human, namely PG-4, -9 and -13. This allowed a more reliable analysis to investigate the process of cluster duplications. The molecular phylogenetic analysis supported the scenario with the 1-2-4 pattern 6, 11, namely the tree topology [(HoxA, HoxB), (HoxC, HoxD)] 9, 12.

More recently, the organization of Hox genes and clusters of the lesser spotted dogfish Scyliorhinus canicula was revealed by transcriptome sequencing and BAC clone screening (. This species has been one of the most promising systems for developmental biology because of its relatively small adult body size and oviparity (. The most striking result in this study was the absence of HoxC members in both genomic and transcriptomic sequencing (Figure 1), while the other three clusters were revealed to have retained a comparable number of Hox genes to that in C. milii (.

Figure 1

Timings of whole genome duplications and multiplicity of Hox clusters. Numbers of Hox clusters identified so far are shown in the box in the middle adjacent to taxon names. Exact numbers of Hox clusters are given for taxa represented by species with genomic data, while the symbol “?” is included for other taxa. Note that two urochordates were shown to possess single, but atomized Hox clusters 1, 46, 47. The possible loss of the HoxC cluster in the chondrichthyan lineage might have occurred even before the split between sharks and rays/skates. Based on the current understanding that cyclostomes may have a smaller number of Hox clusters than four, two possible evolutionary scenarios were presented on both sides. The “post-2R cyclostome” hypothesis (left), compatible with a large-scale phylogenetic analysis employing non-Hox genes (, postulates a secondary reduction of Hox clusters. Based on the “mid-2R cyclostomes” hypothesis (right) 48, 49, the possibly smaller number of Hox clusters in cyclostomes is explained by their divergence before the second round of the whole genome duplication. Intensity of green color indicates the number of Hox clusters in individual evolutionary steps and taxa.

Overall, except for the possible absence of the HoxC cluster in S. canicula, cartilaginous fishes have retained more ancestral members of Hox genes than osteichthyans. More precisely, HoxD2, HoxD5 and HoxD14 have been identified only in cartilaginous fishes, and are thought to have been lost from the HoxD cluster in the basal osteichthyan lineage. Retention of ancestral features by cartilaginous fishes has also been shown in conservation of intergenic sequences in Hox clusters (. The HoxA cluster was also recently sequenced for the little skate Leucoraja erinacea, and compared with that of H. francisci (. This ray–shark comparison equates to a timespan of more than 250 million years 16, 17 and highlighted a higher level of conservation of non-coding sequences inside the Hox cluster than that in the osteichthyan lineage of a comparable evolutionary distance (.

Hox gene expression in cartilaginous fishes

In cartilaginous fishes, investigations on roles of Hox genes have been driven by interests in the evolution of developmental programs responsible for limb/digit formation (. Analyses on cartilaginous fish Hox genes have so far concentrated on species in two oviparous groups, namely those in the genus Scyliorhinus (S. canicula and S. torazame) and L. erinacea. In S. canicula, 5′ Hox genes (HoxD9-13, HoxA11 and HoxA13) were reported to be expressed in a nested pattern, consistent with both spatial and temporal collinearity 19, 20, 21, as seen in osteichthyans 22, 23. Expression of S. torazame HoxD14 was shown in a subpopulation of cells surrounding the hindgut, but not in tissues that would normally express Hox genes, such as the neural tube, somites and fin folds, suggesting the decoupling of this gene from the Hox code (. Hindgut-associated expression is also documented for HoxA13 and HoxD13 of L. erinacea, and these genes are thought to be involved in early hindgut patterning as in amniotes (.

No expression patterns of Hox genes in PG1-8 had been reported for cartilaginous fishes until very recently. As elasmobranchs have been studied as a model for craniofacial development since the 19th century (, this avenue of research needed to be urgently pursued. Very recently, Oulion et al. reported embryonic expression patterns of 34 Hox genes in S. canicula and concluded that their nested expression of Hox genes (Hox code) in branchial arches, hindbrain and somites was maintained in this species and is a ground plan of embryonic architecture, which underwent only small amount of changes during jawed vertebrate evolution (.

Hox gene repertoire in cyclostomes

Cyclostomes are divided into hagfishes (Myxiniformes) and lampreys (Petromyzontiformes). These two lineages separated more than 400 million years ago (. Each of Myxiniformes and Petromyzontiformes consists of only up to 50 species that diversified within 200 million years—a long time after the separation between Myxiniformes and Petromyzontiformes (. Molecular sequence data are reported mostly for species in the northern hemisphere. In Petromyzontiformes, northern hemisphere species form a distinct family Petromyzontidae that diversified only within 50 million years (.

Hox cluster organization in cyclostomes is still controversial because of some difficulties unique to this group of species (. Firstly, the so-called two-round whole genome duplications (2R-WGDs) ( occurred close to the split between cyclostome and gnathostome lineages, and therefore multiple alternative scenarios for ploidy levels of this group have been postulated (. Secondly, initial attempts to isolate cDNA and genomic DNA fragments were made using several different northern hemisphere lamprey species [Petromyzon marinus (; Lampetra planeri (; Lethenteron japonicum 34, 35], and this made the counting of gene numbers difficult. It has been shown by a molecular phylogenetic analysis involving 55 non-Hox gene families that the 2R-WGDs probably occurred before the cyclostome–gnathostome split (. Based on this scenario, if Hox cluster organization truly reflects the genomic ploidy level, cyclostome species are supposed to retain four Hox clusters. However, genomic sequencing of Hox-containing regions on P. marinus resulted in fragments containing up to only five Hox genes inside 37, 38. This situation does not allow a reliable conclusion on the number of clusters, although it is highly likely that the lamprey has multiple Hox clusters (Figure 1).

In lampreys, members of all paralog groups except for PG12 have been identified. Because of the ambiguous timing of the 2R-WGDs, the 1-to-1 orthology of a lamprey (and also hagfish) gene to either of gnathostome HoxA-D is normally not reliably shown in molecular phylogenetic analyses (. They sometimes rather support exclusive clustering of lamprey sequences, suggesting independent gene duplications in the lamprey lineage (. The same feature was also observed for hagfish Hox genes (. These issues should be scrutinized more intensively by obtaining complete cyclostome genomes, ideally representing both Myxiniformes and Petromyzontiformes. In the genome assembly of P. marinus (version 3.0; http://genome.wustl.edu/), none of the available supercontigs harbors multiple Hox genes that are thought to belong to the same Hox cluster. It is notable that the secondary cluster breakage and gene loss, suggested by the currently available data of the lamprey, were proposed also for ParaHox gene organization in a hagfish (.

Hox gene expression in cyclostomes

So far, no Hox expression has been described for hagfishes since their embryos are very difficult to access (. Cyclostome Hox expression studies are concentrated on the Japanese lamprey L. japonicum 34, 35 and the sea lamprey P. marinus (. The former species is used to elucidate conservation of Hox gene expression in neural crest derivatives. In this context, members of PG1-8 were characterized with in situ hybridization, which resulted in evidence of spatial collinearity in the neural tube and pharyngeal arches 34, 35. In contrast, temporal collinearity did not seem to be organized in this early-branching vertebrate (.

Members of PG9-11 (namely LjHox9r, LjHoxW10a, LjHox10s and LjHox11t) as well as Hox13α were shown to be expressed in the tailbud of L. japonicum 24, 35. Moreover, L. japonicum LjHox14α, LjHox13α and LjHox13β were shown to be expressed in the hindgut (. This is reminiscent of the hindgut-associated expression of HoxD14 in the dogfish, and HoxA13 and HoxD13 in the little skate mentioned above, suggesting that their role in hindgut patterning was already established before the radiation of all extant vertebrates.

Lessons from Basal Vertebrate Lineages

Inclusion of basal vertebrates in the discussion of vertebrate Hox evolution broadens our appreciation of the variety of states in which vertebrate Hox gene clusters can exist. This includes not only the possible loss of an entire cluster, but also molecular phylogenetic patterns of retention and loss of particular genes. For example, in PG2, H. francisci, S. canicula and C. milii all retain HoxD2 genes, while this PG2 member is absent in all osteichthyan species examined to date. In lampreys, only one L. japonicum sequence with sufficient length, LjHox2, has been identified.

Figure 2 shows how these early vertebrate members of PG2 are related to osteichthyan homologs. In HoxA2 and HoxB2, cartilaginous fish sequences exhibit shorter branches, suggesting their less derived nature, compared with osteichthyan sequences. The HoxD2 group, absent in osteichthyan species as explained above, consists only of cartilaginous fish members. In this maximum likelihood (ML) tree, the lamprey sequence is placed outside gnathostome sequences (Figure 2A), which is incompatible with the “post-2R cyclostome” scenario proposed based on non-Hox genes (. However, the low bootstrap support for this gnathostome group of 66 suggests that there are other possible tree topologies in which the lamprey sequence could in fact be located within the gnathostome group. Notably, one of them is the tree topology combining the lamprey LjHox2 and the chondrichthyan HoxD2 (Figure 2B). Based on this tree topology, the lamprey LjHox2 is interpreted as an ortholog of HoxD2, or as a relict member of HoxC2 whose ortholog was lost in the basal gnathostome lineage (Figure 2C).

Figure 2

Molecular phylogenetic tree of vertebrate Hox2 genes. A. ML tree based on 104 amino acid sites. Phylogenetic relationships within black circles were constrained according to the generally accepted species phylogeny. The dataset consists of invertebrate outgroups, cartilaginous fishes (HoxA2, HoxB2 and HoxD2), osteichthyans (HoxA2 and HoxB2) and a lamprey sequence, LjHox2 (AY497314). Other cyclostome Hox2 sequences in the public database [P. marinus HoxE2 (AF410908), L. planeri LpHox2A (AF044800), and Eptatretus stoutii Hox2_Z (AY445839)] were not included because of their short lengths. Cartilaginous fish sequences were shown in purple, while the lamprey sequence was shown in red. Values at nodes are bootstrap probabilities based on resampling of estimated log-likelihood (. B. One of the alternative tree topologies supported by the ML method (ΔlogL=2.50±3.90). This tree topology places the lamprey LjHox2 sequence the most closely to HoxD2, suggesting that they are orthologous to each other. C. An alternative hypothesis explaining the tree topology in Panel B. Genes lost or unidentified yet are shown in gray (Gna., gnathostomes; Lamp., lamprey). In this hypothesis, the lamprey LjHox2 gene is the only relict member of HoxC2. Note that the tree topologies in Panels B and C are consistent with the order of cluster duplications recently proposed (, namely [(HoxA, HoxB), (HoxC, HoxD)], whereas the tree topology in Panel A is not.

In fact, both Figure 2A and 2B show tree topologies compatible with the hypothesis that the 2R-WGD occurred before the cyclostome–gnathostome split [pan-vertebrate quadruplication (PV4) hypothesis (]. The tree topology in Figure 2C is particularly striking because this suggests that different subsets of duplicates have been retained between lamprey and gnathostome lineages—paralog C have been kept in the lamprey, while paralogs A, B and D have been kept in jawed vertebrates. This is a typical situation referred to as “hidden paralogy”, and more cases have been introduced for other cyclostome genes 44, 45. If the two lineages, namely cyclostome and jawed vertebrate lineages, evolved differently subsequent to the WGD event, it would not be surprising to detect the differential patterns of gene retention between these groups.

Conclusion

Cartilaginous fishes have well conserved ancestral Hox cluster organization except for the possible absence of a HoxC cluster in S. canicula. The 4-cluster state, observed at least in C. milii, resembles that in non-teleost osteichthyans (for example, mouse, chicken and coelacanth). In contrast, the present uncertainty surrounding our understanding of Hox cluster organization in cyclostomes prevents us from reliably deducing the number of Hox clusters and using it as a proxy for ploidy. This ambiguity could also be interpreted as a reflection of the phylogenetic landscape of non-Hox gene families in which the orthology of cyclostome genes to gnathostome genes cannot unambiguously be established. If we are based on the recently proposed PV4 hypothesis assuming “post-2R cyclostomes” (, cyclostomes would be expected to have four Hox clusters, or at least had four in their ancestry. If a smaller number of Hox clusters are to be found in them, this implies additional events, such as loss of an entire cluster as a significant step in cyclostome evolution (Figure 1). This would violate the notion that Hox cluster organization serves as a reliable marker of ploidy levels. This in turn might have important implications for the evolution of the cyclostome body plan and may be consistent with its hypothesized evolution via a certain degree of simplification or degeneration.