The functional evolution of hbn, rx and otp
Homeobrain was originally identified in Drosophila, and mapped to a region of chromosome 2 that contained two additional PRD-class homeobox genes, Orthopedia and rx[44, 45]. Homeobrain is expressed in the fly embryonic brain and ventral nerve cord . Recently the first lophotrochozan homeobrain-like gene was reported , and found to be expressed in a restricted region of the anterior brain. Hbn expression has also been reported in the sea urchin, where it is expressed in oral ganglia of the animal pole .
Otp is associated with neural development in a phylogenetically diverse collection of animals, and this may represent an ancestral role for this homeodomain family. Orthopedia-related genes in mouse are involved in brain patterning and development [47–52]. In the hemichordate Saccoglossus, otp has punctate, ectodermal expression in neural domains of the prosome (proboscis) and mesocome (collar) . In planaria, orthopedia is implicated in patterning the branch structure of the brain [54, 55]. In the limpet, otp is involved in the development of the larval apical sensory organ , and in flies it is involved in the developing CNS and hindgut and anal pads . However, a connection to neural development is not obvious in the sea urchin, where otp appears to be involved in larval skeletal morphogenesis and the establishment of oral ectodermal cell fate [57–61].
Rx (retinal associated homeobox) genes have been extensively studied in many deuterostome taxa including human, mouse, Xenopus, chicken, zebrafish, medaka and tunicate [45, 62–70]. In all of these organisms, rx is involved in brain development and the formation of retinal territories and associated neural structures. In the hemichordate, rx is expressed throughout the prosome ectoderm, the most anterior region of this organism, but absent from the most apical pole . Rx is expressed more broadly during sea urchin development, with punctate expression in the animal pole and developing gut . A fruit fly rx gene has also been identified, and studies suggest that it is necessary for proper brain development, but it is not required for eye development [45, 71].
Alternative splice variants in orthopedia and rx
In addition to a homeodomain and a PRD domain, two additional motifs have been identified in some PRD-like genes. The octapeptide motif [72, 73], located towards the N-terminus, is involved in transcriptional repression . The OAR domain, found in otp, aristaless and rx[40, 51, 67, 75] is typically located at the C-terminus and is known to function as a transactivator in Otp[40, 51].
We have identified alternative transcripts for Nematostella orthopedia and rx. For rx, some transcripts encode a highly conserved OAR domain and one transcript does not. For orthopedia, none of our RACE products nor any ESTs in publicly available databases included the OAR domain. However, a well-conserved OAR domain was identified downstream and in-frame of these sequences in predicted genomic gene models, suggesting that this domain is probably expressed in an as yet unidentified splice variant. The OAR domain has been described as an intramolecular switch, which acts to reduce the affinity of the homeodomain transcription factor for its binding site. In the mouse, ectopically expressed mutant forms of the Alx3 and Cart1 proteins lacking the OAR domain exhibit increased binding to their DNA targets .
Alternative splice variants involving the presence or absence of an OAR domain have also been identified the mouse prx1 gene, a member of the PMX family of PRD-class homeobox genes . The carboxy terminus of the Prx1a protein includes an activation domain and an OAR domain, whereas the carboxy terminus of Prx1b encodes a repression domain and lacks an OAR domain [77, 78]. The tissue distribution of both transcripts appears to be similar in mice and humans, but different tissues exhibit pronounced differences in the relative ratios of prx1a and prx1b[79, 80]. It has been hypothesized that the presence of the OAR domain in prx1a could render it sensitive to modulation via an unidentified partner protein that interacts with the OAR domain itself. In the absence of this cofactor, the OAR domain masks the activation domain and reduces the affinity of prx1a for DNA binding sites. When bound by its co-factor, the activation domain becomes unmasked and the DNA binding affinity increases [76, 77, 81]. In the case of Nematostella rx, the situation may be somewhat simpler than in the mouse prx1 gene, because except for the presence or absence of the OAR domain itself, the alternative splice variants encode essentially identical proteins.
Functional inferences about Nematostella otp, rx and hbn based on expression data
In the most famous example of a conserved homeobox cluster, the Hox genes, the spatial ordering of Hox expression territories along the body's main axis and the timing of their onset mirrors the physical ordering of linked Hox genes (although not in all taxa). This correspondence is termed colinearity. However, in animals with dispersed Hox clusters, such as the urochordate Oikopleura dioica, some spatial colinearity remains whereas temporal colinearity is absent . This suggests that it is temporal rather than spatial colinearity that is driving the maintenance of these clusters. Although work has been performed in studying temporal colinearity in Hox, ParaHox and NK clusters, the homeobrain cluster could prove to be another supporting example.
In Nematostella, hbn, rx and otp appear to be expressed in a temporally colinear pattern. Hbn is expressed first, in the blastula stages, followed by rx at mid gastrulation and finally otp in the planula. However, in the case of the Drosophila homeobrain cluster, there is no clear evidence of temporal colinearity. Homeobrain is expressed first, in the syncytial blastoderm, then otp is expressed slightly before rx[31, 51, 65]. Future studies in other animals will help determine whether temporal colinearity is widely conserved among homeobrain clusters.
Although the expression of Nematostella hbn, rx and otp is consistent with temporal colinearity, it is not consistent with spatial colinearity. The three genes are expressed in non-overlapping domains along the oral-aboral axis; however, these domains do not appear to be related to their position in the cluster. NvRx is expressed at the most aboral domain, although expression is not present at the aboral pole, where the apical tuft will form. NvHbn is more broadly expressed in oral ectoderm during early embryogenesis and becomes confined to the most oral ectoderm, mainly around the base of the tentacles. Finally, NvOtp is also expressed in the oral ectoderm, in the domain that will invaginate and form the pharynx. NvOtp and NvHbn are both expressed in close proximity to three paralogs of Otx (NvOtxA, B, C) in the pharyngeal ectoderm and tentacles surrounding the oral pole .
In addition to the broad non-overlapping domains, all these genes are also expressed in individual cells throughout the body column. Based on their cell morphology, it appears that these cells may be neurons. Additionally, the expression of NvOtp and NvOtx(A,B,C) in the oral ectoderm coincides with formation of the oral nerve ring [30, 42]. Considering the function of these genes in other animals and the expression patterns seen here, it is likely that hbn, rx and otp play some role in neural development in Nematostella. However, it is interesting that there is no expression in the apical tuft, another strongly neurogenic region in Nematostella.
The evolutionary history of the homeobrain cluster
The literature on homeobox clusters would suggest that the history of the ANTP class has been qualitatively different from that of the histories of other classes of homeobox genes. The most intensively studied and widely conserved homeobox clusters are all composed of ANTP-class homeobox genes (the Antennapedia complex, the Bithorax complex, the Hox cluster, the ParaHox cluster, the NK cluster and the EGH box cluster). Ultimately, all of these ANTP-class clusters may derive from a single ancestral cluster. For example, the ANTP-C and BX-C of Drosophila are clearly derived from a single ancestral Hox cluster that is widely shared by protostomes and deuterostomes, and in simpler form, by cnidarians. The Hox cluster, in turn, appears to have broken off from an ancestral 'mega-Hox' or 'extended Hox' cluster that at one time may have encompassed the Hox cluster, the ParaHox cluster, the EGH cluster and the NK cluster. Over time, the hypothetical ancestral cluster appears to have fragmented, and different remnants of this ancestral cluster may be more highly conserved in different animal lineages .
This study is the first to provide evidence that a non-ANTP-class homeobox cluster was conserved over hundreds of millions of years of animal evolution. Clearly, the hbn-rx-otp cluster has been fairly well conserved over the evolutionary history of holometabolous insects. A cluster with the same constituent genes, in the same orientation, spanning a comparable distance is found in representatives of three different orders of insects (Coleoptera, Diptera and Hymenoptera; Figure 4). The cluster also appears to date to the ancestral Protostome, as evidenced by conservation of the three-gene cluster in the limpet Lottia, although the order of the genes has changed. In addition, the cluster was also likely to have been present in the cnidarian-bilaterian common ancestor, some 600 million years ago. The cluster in Nematostella involves the same closely linked genes in the same order as the inferred ancestral cluster of holometabolous insects, but the otp locus has been inverted. Similarly, in four other key taxa (cephalochordate, sea urchin, annelid and placozoan) we find evidence of a partial cluster, in which two of the three genes are linked. Further sequencing and assemblies of the echinoderm and hemichordate genomes will reveal the extent of the hbn-rx-otp cluster in the deuterostome ancestor. The placozoan genomic data suggests that a portion of this cluster (hbn and otp) dates back even further to the ancestral eumetazoan. However, it remains to be seen whether rx was present in this ancestor and lost in the placozoan lineage, or whether rx evolved after the split.
Mechanistically, it is easier to envision how a cluster of three closely related genes might remain linked than to envision how three closely related genes, if already dispersed, could independently become so closely juxtaposed in multiple taxa. If the genes arose by tandem duplication, the cluster would have originated as a result of the gene duplication; that is, the starting point would have been a cluster. Subsequently, the cluster could have been maintained over hundreds of millions of years of evolution in multiple animal lineages by stabilizing selection.
Closely linked genes will tend to reside in the same chromosomal territories, and they may come under the influence of shared regulatory elements. For this reason, the proper regulation of linked genes may be related to their physical proximity in the genome. This is the general explanation for why Hox genes have remained clustered for hundreds of millions of years in many animals that have been examined. In the case of Hox genes, the spatial ordering of Hox expression territories along the body's main axis mirrors the physical ordering of linked Hox genes along the chromosome. This correspondence is termed colinearity. In the anterior CNS, rx, hbn and otp are expressed in nested territories, which is somewhat reminiscent of Hox genes. Future bioinformatics studies could test for conserved regulatory elements within the hbn-rx-otp clusters of fly, honeybee and flour beetle. The functionality of these putative enhancer-binding sites could then be studied experimentally. In addition, the effect of cluster disruption can be examined experimentally in Drosophila and in mice, as it has been with Hox genes [84, 85]. It may also prove very informative to use evolutionary comparisons of taxa with intact and disrupted clusters to investigate the consequences of cluster disruption, as has recently been carried out for the ParaHox cluster .
The loss of homeobrain in chordates
The homeobrain gene appears to have been lost in chordates. We could not identify it in the sequenced genomes of human, mouse, chicken, clawed frog, zebrafish, lancelet or tunicate (Figure 4). However, we identified clear orthologs in two other deuterostomes (sea urchin and hemichordate) and throughout the protostomes and 'basal' metazoans. Thus, this gene may have been truly lost from the genome or it may have become so highly modified that it is no longer recognizable as a homeobrain ortholog. The phylogenetic analysis performed here, based on homeodomain sequences, does not strongly suggest another PRD-class gene as a possible ortholog. Unfortunately, we cannot rely on regions outside the 60-amino acid homeodomain to obtain additional phylogenetic signal because there was not sufficient sequence conservation to permit alignments across all genes represented in the phylogeny. The highly conserved 128-amino acid Paired domain, present in many PRD-class homeobox genes, is absent from the homeobrain, orthopedia and rx families, among others [29, 39, 40]. With the present data, our analyses support the hypothesis that homeobrain was lost early in the chordate lineage.
Coding polymorphisms in Nematostella
Despite potential stabilizing selection to maintain the cluster, we observed many polymorphisms in the Nematostella genes comprising the cluster, including nonsynonymous substitutions in the homeodomain of rx that have a patchy distribution in natural populations. The functional role of these polymorphisms awaits future experimental characterization. The presence of few homozygotes for the rare amino acid for each position in our sampling of natural populations is potential evidence for a functional difference between alleles.