Broken colinearity of the amphioxus Hox cluster
© Pascual-Anaya et al.; licensee BioMed Central Ltd. 2012
Received: 25 June 2012
Accepted: 4 October 2012
Published: 3 December 2012
Skip to main content
© Pascual-Anaya et al.; licensee BioMed Central Ltd. 2012
Received: 25 June 2012
Accepted: 4 October 2012
Published: 3 December 2012
In most eumetazoans studied so far, Hox genes determine the identity of structures along the main body axis. They are usually linked in genomic clusters and, in the case of the vertebrate embryo, are expressed with spatial and temporal colinearity. Outside vertebrates, temporal colinearity has been reported in the cephalochordate amphioxus (the least derived living relative of the chordate ancestor) but only for anterior and central genes, namely Hox1 to Hox4 and Hox6. However, most of the Hox gene expression patterns in amphioxus have not been reported. To gain global insights into the evolution of Hox clusters in chordates, we investigated a more extended expression profile of amphioxus Hox genes.
Here we report an extended expression profile of the European amphioxus Branchiostoma lanceolatum Hox genes and describe that all Hox genes, except Hox13, are expressed during development. Interestingly, we report the breaking of both spatial and temporal colinearity for at least Hox6 and Hox14, which thus have escaped from the classical Hox code concept. We show a previously unidentified Hox6 expression pattern and a faint expression for posterior Hox genes in structures such as the posterior mesoderm, notochord, and hindgut. Unexpectedly, we found that amphioxus Hox14 had the most divergent expression pattern. This gene is expressed in the anterior cerebral vesicle and pharyngeal endoderm. Amphioxus Hox14 expression represents the first report of Hox gene expression in the most anterior part of the central nervous system. Nevertheless, despite these divergent expression patterns, amphioxus Hox6 and Hox14 seem to be still regulated by retinoic acid.
Escape from colinearity by Hox genes is not unusual in either vertebrates or amphioxus and we suggest that those genes escaping from it are probably associated with the patterning of lineage-specific morphological traits, requiring the loss of those developmental constraints that kept them colinear.
Chordates include the group olfactores (vertebrates and urochordates) and the cephalochordates (Figure1). However, urochordates, as a reflection of their highly reorganized genome and extensive gene losses, do not retain the typical clustered organization of Hox genes with only some genes linked, as in the ascidian Ciona intestinalis[13, 14] (Figure1), or as an atomized cluster, as is the case of the larvacean Oikopleura dioica (Figure1). Nonetheless, the cephalochordate amphioxus, representing the most basal branch of chordates, has a rather prototypical genome and possesses a single cluster of 15 Hox genes, where all of them are transcribed in the same orientation, as in vertebrates. Thus, amphioxus represents the best model to compare with vertebrates for illuminating the basal condition of chordates for both Hox content and regulation. However, the expression of amphioxus Hox genes is scarcely reported and studies have focused mainly on the anterior ones. The genes of the Floridian amphioxus Branchiostoma floridae Hox1 4 and Hox6 have been reported to be expressed during development in a colinear manner in the central nervous system (CNS), highlighting that, although it is not morphologically segmented as in vertebrates, the CNS in both groups to some extent conserves the same nested Hox system. In addition, the expression patterns of Hox1, Hox3, and Hox4 have been reported in the epidermis and have been associated with the determination of different sensory neurons along the A-P axis. Moreover, Hox1 mRNA is expressed in the middle part of the gut.
The acidic form of vitamin A, retinoic acid (RA), has an important role in the regionalization of morphological structures along the A-P axis of vertebrates, acting as the main posteriorizing factor during neural determination. Its function is carried out in a gradient-dependent manner, with higher concentrations in posterior parts. In the case of amphioxus, an excess of RA during development also causes changes of anterior into posterior identities, and the mouth and gill slits do not form. Conversely, treatments with RA antagonists result in a caudal extension of the pharynx[23, 24]. These effects caused by altered concentrations of RA during development are equivalent to those observed in vertebrate embryos and highlight that determination mechanisms of the structures along the A-P axis are somehow conserved in chordates. RA carries out its function by binding heterodimers of the RA receptor (RAR) and retinoid X receptor (RXR), which regulate the transcription of their target genes by binding to RA response elements (RAREs) in the regulatory regions of the genome. RAREs consist of two direct repeats (DRs) separated by a variable number of nucleotides. In the case of RAR/RXR heterodimers, they have been shown to bind DRs separated by one (DR1), two (DR2), or five (DR5) nucleotides. RA has an important role in controlling Hox genes[26, 27] via RAR/RXR binding to RAREs[28–30]. Besides, the anterior Hox genes in amphioxus are regulated by RA[18–20], and morpholino knock-down of Hox1 produces the same phenotype as treatment with an RA antagonist, indicating that Hox1 mediates the function of RA to establish the posterior limit of the pharynx. Therefore, the regulation of anterior Hox genes by RA seems to be conserved between vertebrates and amphioxus, as suggested by heterologous reporter assays using regulatory regions of amphioxus Hox genes in both the mouse and chicken[31, 32]. Again, most of these studies were focused on the anterior part of the cluster; hence, a general scenario for Hox cluster regulatory evolution is not yet available.
In this study, we report previously undescribed expression patterns of Hox genes of the European amphioxus Branchiostoma lanceolatum, and, surprisingly, find that some of them are not expressed in a colinear manner either in space or time, thus breaking the paradigm of Hox colinearity in amphioxus. We identified a different expression for B. lanceolatum Hox6 than that previously reported for the Floridian amphioxus[18, 33] and detected Hox14 expression in the pharyngeal endoderm at the level of the endostyle, the mid-hindgut, and notochord. Strikingly, it was detected in the cerebral vesicle, a part of the CNS where no Hox expression has been detected so far. Thus, this gene has escaped from the Hox coding pattern, as has been reported for vertebrate Hox14 genes[34, 35]. We then investigated the regulation of these genes by RA and found that RA regulated the expression of Hox6 and the expressions of Hox14 in the gut and notochord. It also affected Hox14 expression in the cerebral vesicle, whereas the regulation of Hox14 in the pharyngeal endoderm seemed to be RA independent. The presence of RAREs near these genes, conserved between both the Floridian and European amphioxus species, makes these genes colinearity-breakers but still likely targets of RA regulation.
Sexually mature amphioxus adults (B. lanceolatum) were collected in Argelés-sur-Mer (France) during the summer of 2009. Spawning was induced in the laboratory by heat shock. After fertilization, embryos were reared in filtered seawater at 17°C. Treatments with RA (in DMSO), the RA antagonist BMS009 (in DMSO), or equivalent amounts of DMSO (as control) were carried out at the late blastula stage at a final concentration of 1 × 10-6 M as described[23, 24]. At the early neurula stage, embryos were transferred to untreated filtered seawater, washed a few times, and kept in normal conditions. The control DMSO treatment did not affect development. Embryos and larvae were fixed at frequent intervals with 4% paraformaldehyde overnight at 4°C in a buffer containing 0.1 M MOPS, 0.5 M NaCl, 2 mM MgSO4, 1 mM EGTA, pH 7.4 for in situ hybridization.
Because the coding sequences of B. floridae and B. lanceolatum are extremely conserved, we decided to use primers based on the B. floridae Hox11 exon 2 to amplify the first intron of B. lanceolatum's one. Using a forward primer from the B. lanceolatum Hox11 exon 1 (5′-ATGGACGGTTACTGGCTGC-3′,) and a reverse primer designed on the B. floridae Hox11 exon 2 sequence (5′-CTGCCTATCCGTGAGGTTG-3′,), we amplified a band of approximately 2.5 Kbp using B. lanceolatum genomic DNA as a template. We cloned it into pGEM-T Easy Vector (Promega) and sequenced it. The sequence corresponded to the first exon, first intron, and second exon of B. lanceolatum Hox11. The genomic sequence with the new annotation has been uploaded to the NCBI GenBank Database (http://www.ncbi.nlm.nih.gov/genbank/) under accession no. JX508623.
Sequences of lambda phages containing either B. lanceolatum Hox6 and Hox5 (lambda phage no. λ4131+λ4184 described in) or Hox14 (lambda phage no. λ4100 in) and their equivalent sequences from B. floridae were analyzed by nuclear hormone receptor (NHR) scan using default parameter values. We compared the results for sequences in both species and only those elements conserved in both B. floridae and B. lanceolatum were considered.
Apart from Hox4 and Hox6, no other Hox gene expressions have been reported for the central group in cephalochordates. We detected a very weak expression of Hox7 in the CNS, mesoderm, and tail bud (Figure3). Due to its weak expression, we cannot rule out the possibility that B. lanceolatum Hox7 was expressed in other tissues. At the late neurula stage, the anterior limit of Hox7 was at a level equivalent to that of Hox4, but the anterior expression was so blurred that establishing a clear boundary was difficult; thus, we could not assess the colinearity relationship between Hox4 and Hox7 at this stage. However, from the pre-mouth stage the expression is more posteriorly restricted than for Hox4. On the other hand, it began to be expressed after more anterior Hox genes, from the late neurula stage, thus retaining temporal colinearity.
No expression of Hox genes from posterior groups has been characterized in cephalochordates so far. Here we investigated the expression patterns of B. lanceolatum Hox10 and Hox14. We found that Hox10 expression followed a similar pattern to that for Hox7, with very weak expression in the CNS and mesoderm (Figure3). Again, the weak and blurred expression of Hox10 makes difficult to exclude the possibility that is actually expressed in other tissues. As for Hox4 and Hox7, the anterior limit of expression of Hox10 was very diffuse, and although it seemed to be more rostral than the expressions of Hox4 and Hox7, and thus breaking spatial colinearity, the diffuse anterior limit found makes evaluating the colinearity difficult.
The most unexpected expression pattern was that of Hox14. B. lanceolatum Hox14 is expressed from the pre-mouth larval stage. As for most abovementioned genes, a probe for the coding sequence produced high background and unspecific signals, so we decided again to clone the 3′-UTR. We found that it was split into two exons, with a small intron of 45 bp. Amphioxus Hox14 was expressed in the mid-hindgut, in the posterior part of the notochord, and in the tail bud (Figure4B′, C′). Strikingly, Hox14 was also detected in anterior structures such as the cerebral vesicle and the left side of the pharyngeal endoderm at the level of the endostyle (Figure4B′ and inset).
Apart from the genes whose expression patterns we have been able to identify, other Hox genes were detected by means of RT-PCR during B. lanceolatum development. Among these genes are B. lanceolatum Hox2, Hox5, Hox8, Hox9, Hox11, Hox12, and Hox15. Hox13 was not identified either by RT-PCR or in the recently published embryonic transcriptome of B. lanceolatum. We obtained the CDSs for B. lanceolatum Hox2, Hox5, Hox8, Hox9, Hox11 (first and second exons but not the third), Hox12, and the recently discovered Hox15[17, 37], and 3′-UTRs by using 3′-RACE RT–PCR for B. lanceolatum Hox2, Hox5, Hox8, and Hox12. We also amplified the 5′-UTR by means of ′-RACE RT–PCR of Hox2, Hox9, and Hox11. Nonetheless, we were not able to obtain any signal by WISH (Figure2B).
The case of amphioxus Hox2 is quite similar to that of Hox6. So far, two reports about its expression have shown two different expression patterns. The first one reported the breaking of colinearity for Hox2, while the second reported a colinear expression with respect to Hox1 and Hox3. As with Hox6, we wanted to test which Hox2 expression pattern could be the correct one using B. lanceolatum. However, although we performed WISH with different probes based on the sequences of the CDS or the 3′-UTR, we were not able to obtain specific signals (Figure2B).
We found that the 5′-UTR of Hox9 had two different versions. One was shorter, with a canonical 5′-UTR next to the start codon, and a second larger one was divided into four exons when aligned against the B. floridae Hox cluster (the complete sequence of the B. lanceolatum Hox cluster is still not available): three of them were far upstream from Hox9. The first exon was placed approximately 5.4 Kbp downstream of Hox11, the second was approximately 9 Kbp downstream of Hox10, and the third was approximately 25 Kbp upstream of Hox9. The fourth corresponded to the canonical 5′-UTR (Figure2A).
For B. lanceolatum Hox11, only the first exon has been annotated. Here, we have extended the previously described genomic sequence up to exon 2 (see Methods), which allowed us to find Hox11 by 5′-RACE RT-PCR. The 5′-UTR was shorter than expected, which means that the previously automatically annotated exon 1[37, 38] was not correct and the actual exon 1 is shorter. As with the other Hox genes, a probe based on the coding sequences of exons 1 and 2 gave no signal in WISH (Figure2B).
The RA-Hox system controls the patterning along the A-P axis during development of chordates (for a review see), and such control has been reported widely for anterior Hox genes in amphioxus[18–20, 23, 24, 44]. Because we detected a different expression pattern for the B. lanceolatum Hox6 gene than that reported previously and given that amphioxus Hox14 has been shown to have a non-canonical expression pattern, we treated embryos with RA, the RA antagonist BMS009, or with DMSO as an inert negative control, and carried out WISH experiments.
In RA-treated embryos, the anterior limit of Hox6 moved rostrally up to the level between somites 3 and 4 (compare Figure4A with Figure4A′), whereas the posterior limit was unaltered. When treated with the antagonist, Hox6 expression disappeared. This can be explained by the anterior limit shifting posteriorly to the extreme of its fixed posterior extent, thus making Hox6 expression disappear (compare Figure4A′ with Figure4A′′). Then, the level of the anterior limit would be that changed when taking the somites as a reference point, as in other more anterior Hox genes demonstrating that the changes in expression were regulated by RA (directly or indirectly) and not because of a general shift of internal structures.
In RA-treated larvae, the anterior limit of B. lanceolatum Hox14 expression (at least in the gut) was shifted anteriorly in a significant manner compared with the control, taking as a reference point the mid pigment spot of the CNS. However, it was not so clear for the expression in the notochord (in Figure4, compare B′ and C′ with B and C, respectively). In contrast, when treated with BMS009 the expression of Hox14 in both the notochord and the gut shifted strongly to the posterior (in Figure4, compare B′ and C′ with B′′ and C′′, respecively). Surprisingly, expression of Hox14 in the pharyngeal endoderm did not disappear completely in either RA- or BMS009-treated embryos (black arrowheads in Figure4). Although formation of the pharynx is strongly reduced in RA-treated larvae, with mouth and gill slits failing to form, we detected faint expression in the pharynx of Hox14 in the pre-mouth-stage larvae (Figure4B, black arrowhead), while Hox14 was detected clearly in the case of the 2-day-old RA-treated larvae (Figure4C, black arrowhead). This suggests that regulation of Hox14 expression in the endostyle is RA independent. The amphioxus cerebral vesicle is a structure that is also reduced in RA-treated larvae but does not disappear, as has been shown using cerebral vesicle markers. Interestingly, the expression of Hox14 in the cerebral vesicle was extremely reduced with both RA and BMS009 treatments (Figures4B,4B′′,4C and4C′′, white arrowheads), suggesting that the cerebral vesicle expression domain is somehow very sensitive to variations in RA level.
RA regulates the expression of its target genes via heterodimers of RAR/RXR that bind RAREs. These heterodimers can bind DR1-, DR2-, and DR5-type RAREs. Using an NHR scan, which has been shown to be effective in the prediction of DR2 and DR5 surrounding amphioxus ParaHox genes, we looked for RAREs near to amphioxus Hox6 and Hox14 genes, using the same genomic regions analyzed previously for non-conserved regions (see Methods). We also screened the corresponding B. floridae genomic sequences used in our previous comparative regulatory analysis, to exclude predictions that have not been conserved between both amphioxus species, because they are probably not real and functional elements. We found that most of the predicted RAREs were not conserved between both species (see Additional file1: Tables S2 to S5). Thus, we regarded these as false-positives and they were discarded. Using these criteria, we detected one DR2 and three DR5 elements near to Hox6, and two DR1 elements within the Hox14 locus: one within the second intron and the other located in the second exon of the 3′-UTR (Figure4D).
The expression patterns of Hox1, Hox3, and Hox4 in amphioxus epidermal neurons have been reported for B. floridae. However, we did not detect Hox4 expression in B. lanceolatum epidermis. Therefore, our data are not consistent with the hypothesis of a ‘skin brain’ (similar to the diffuse net of neurons in hemichordates) in amphioxus[19, 46].
As for Hox6, unlike the two patterns described previously in the Floridian amphioxus, we have found B. lanceolatum Hox6 only at the mid-neurula stage in a restricted stretch of the neural plate (Figure4A′). The anterior limit of B. lanceolatum Hox6 is one somite level more rostral than that described for B. floridae by Schubert and colleagues (between somites 7 and 8,) and much more caudal than the anterior limit found by Cohn.
One question arises from past and current data: what can explain such different expression patterns in three different experiments? One possibility is that one of the expression patterns of B. floridae (in the case of Hox6, most likely that reported by Schubert et al. rather than that by Cohn, because the signal presented by the former seems more reliable than the faint one of the latter report) and those ones presented here for B. lanceolatum actually reflect a real species-specific difference. If so, it means that the expression of B. floridae Hox6 in the CNS in a colinear manner with the other Hox genes is not conserved in B. lanceolatum CNS patterning. On the other hand, it is possible that experimental consideration such as probe design may explain the differences. For example, since the nucleotide sequences of the homeobox regions of all central Hox genes are highly similar (see Additional file2: Figure S2), a probe spanning this sequence might cause cross-hybridizations and thus partial miss-assignments of expression patterns. In fact, when we used a probe based in the CDS for most of the genes, we obtained either no signal or unspecific staining of the B. lanceolatum embryos (red lines in Figure2B). Therefore, we decided to use 3′-UTR-based probes, which are unable to cross-hybridize with other Hox genes. The 3′-RACE RT-PCR using gene specific primers designed in the first exon gave only a single band in both Hox4 and Hox6 (see Additional file2: Figure S3), indicating that alternative splicing does not lie behind the difference and that we were detecting only the expression of Hox4 and Hox6 transcripts. However, we cannot conclusively discard the presence of alternative transcripts that could account for the different expression patterns obtained upon the use of different probes.
We believe that a revisit of expression patterns in both B. floridae and B. lanceolatum and, essentially, in the Asian species Branchiostoma belcheri, will help to elucidate if the discrepancies reported come from truly species-specific differences or have an experimental nature.
In B. lanceolatum, Hox6 was expressed slightly more rostrally than Hox4 and thus did not maintain spatial colinearity. It was also expressed in an earlier stage (mid-neurula) than that of the onset of Hox4 expression (late neurula), therefore, Hox6 also deviated from temporal colinearity. The function of Hox6 in amphioxus is not known, but it is likely involved in the patterning and regionalization of the CNS in a very specific domain of the neural plate at a very specific time in development. In vertebrates, Hox6 is expressed in the spinal cord behind the rhombencephalon to the caudal end of the spinal cord and also in the mesoderm. Therefore, the expression of Hox6 of amphioxus and vertebrates is not conserved. Given that the vertebrate Hox6 genes maintain both spatial and temporal colinearity, we believe that they represent the ancestral condition, while the expression of amphioxus Hox6 is probably more divergent. In addition, we have shown that Hox6 is still regulated (directly or indirectly) by RA, as are the more anterior Hox genes, suggesting that it is derived from the ancestral state of canonical nested expression together with its mode of regulation.
The expression of amphioxus Hox6 and the effects of RA and the RA antagonist are very similar to those of the amphioxus ParaHox gene Gsx. Amphioxus Gsx is expressed in a few cells in the neuroectoderm, at the level of somite 5, just anterior to the Hox6 domain. Amphioxus Hox6 and Gsx likely participate in the A-P patterning of limited parts of the neuroectoderm in a similar manner, probably in combination with other Hox genes that overlap with them. RA treatments enlarge and shift the Gsx rostral limit of expression anteriorly whereas RA antagonist treatments make the Gsx domain disappear, as in the case of Hox6. As with the posterior limit of Hox6, which is unaffected by RA treatment, the posterior limit of Gsx did not change dramatically with RA treatment. Therefore, as Osborne et al. have suggested for Gsx, the anterior limit of Hox6 would be regulated by RA, but the posterior limit would not. Thus, in both Hox6 and Gsx, the loss of the domain following BMS009 treatment can be explained by a caudal shift of the anterior limit until it reaches its posterior one, making the expression to disappear. In vertebrates, it is not known whether Hox6 paralogs are direct targets of RA regulation. However, other central Hox genes such as HoxA7 and HoxC8 shift their anterior limit rostrally in the paraxial mesoderm of mouse embryos after RA treatment (Hox 1.1 and Hox 3.1, respectively, in), and different Hox4 paralogs have been shown to be regulated directly by RA[47–49]. However, in other cases such as in the chicken neural tube, the expression levels of genes from HoxB6 to HoxB9 have been shown to be refractory to RA treatment. Thus, further investigation is needed in both cephalochordates and vertebrates to understand the ancestral mode of regulation of the central Hox genes by RA.
We have also detected the expression of three other Hox genes not studied so far: Hox7, Hox10, and Hox14. While the anterior limit of Hox7 expression is similar to that of Hox4 at the late neurula stage, it was more caudal from the pre-mouth larval stage onwards, thus keeping its spatial colinearity. However, the anterior limit of Hox10 expression in the CNS and mesoderm seemed to be more anterior than that of Hox4 and Hox7. Nonetheless, it is necessary to point out that the anterior limits of amphioxus Hox7 and Hox10 were very diffuse, unlike their vertebrate counterparts, which usually display sharp rostral limits, and their colinearity nature is then far from conclusive. Although this difference in the anterior limit between amphioxus and vertebrates might reflect different modes of regulation, we believe that this is the result of the clearly segmental nature of the vertebrate CNS (for example, rhombomeres) compared with the amphioxus CNS. Thus, the real anterior level up to where these genes are expressed in amphioxus might be different to that detected by WISH and thus their colinearity might differ, as discussed here. In vertebrates, Hox6 and Hox10 paralogs retain their colinearity and they have important opposite roles: Hox6 genes encode rib-promoting factors whereas Hox10 genes are rib-inhibiting. Thus, the colinear expression of Hox6 and Hox10 genes in vertebrates is under strong developmental constraints that are not present in amphioxus, allowing these genes to escape from colinearity.
The most striking case of colinearity breakage is that of Hox14. Posterior Hox genes are involved in the appearance of morphological innovations and are also related to changes in the evolution of the vertebrate bauplan, such as the type of vertebrae or morphological variability within squamates. Interestingly, lamprey and shark Hox14 genes have non-canonical expression patterns. They are expressed only in the posterior part of the endoderm in the lamprey and in a very specific posteroventral area in the shark surrounding the cloaca. Amphioxus Hox14 is expressed in anterior structures such as the cerebral vesicle. This is the first Hox gene to be detected in such an important organ. In vertebrates, no Hox genes are expressed in the midbrain or forebrain and all are excluded from Otx and Pax expression territories. However, what was thought to be a universal rule is broken in amphioxus. Likewise, the expression of Hox14 in the pharyngeal endoderm is an exceptional case, because no amphioxus Hox gene has been detected earlier in the pharynx. RA regulates the expression of Hox14 in the notochord and mid-posterior gut in the same manner as anterior genes (enlarged expression anteriorly when embryos are treated with RA, or posterior shift of the anterior limit when treated with an RA antagonist; arrows in Figure4B-B'', C-C''). On the other hand, Hox14 seems not to be regulated by RA in the pharynx, because RA treatment did not make the expression disappear, and nor did treatment with the antagonist expand it posteriorly (Figure4B-B'', C-C'', black arrowheads) as would be expected. Surprisingly, Hox14 regulation in the cerebral vesicle appeared to be very sensitive to RA, because both excess and a drop in RA level affected its expression strongly (Figure4B-B'', C-C'', white arrowheads). Thus, the regulatory regions of Hox14 must be modular. Some expression domains, namely the gut and the notochord, would depend upon RA. Although we cannot discern from the data presented here whether this control is direct or indirect, the presence of DR1 elements within amphioxus Hox14 locus could give clues for future experiments. Thus, an RA-independent module might regulate the pharyngeal endoderm domain. By contrast, a module sensitive to RA concentration might control the cerebral vesicle domain, perhaps from an indirect effect of RA treatment on some transcription factors that are directly regulated by RA. For vertebrates, a putative regulation of Hox14 by RA has not been studied. However, other posterior Hox genes respond in the opposite way to the anterior genes. For example, HoxB9 is refractory to RA treatment in the neural tube of chicken embryos, as is HoxB6, and in the mouse a putative function of RA seems to be to prevent the expression of posterior HoxD genes in the anterior domain[54, 55].
The uncoupling of vertebrate and amphioxus Hox14 genes from a canonical Hox code is one sign of relaxation of the posterior part of the cluster, but not the only one. Posterior Hox genes of cephalochordates, urochordates, echinoderms, and hemichordates do not have clear orthologous relationships to the posterior paralogy groups of vertebrates, probably because of the higher evolutionary rate of this class of genes, a phenomenon named deuterostome posterior flexibility. If the posterior Hox genes of amphioxus and vertebrates are true orthologs, our data imply that decoupling from the Hox code of Hox14 genes occurred in the last common ancestor of chordates. However, although not conclusively, phylogenetic analyses of deuterostome posterior Hox genes, including the recently reported amphioxus Hox15, suggest that the posterior Hox genes of amphioxus and vertebrates likely originated from independent duplications[17, 34, 35]. In line with the posterior flexibility hypothesis, the intergenic regions of the posterior Hox cluster are less conserved than the anterior ones in both amphioxus[37, 38] and gnathostomes. This trend for the posterior Hox cluster might be explained if the posterior genes had originated by specific expansions, for example, via tandem duplication, to give 14 Hox genes in the last common ancestor of vertebrates and 15 in Branchiostoma. Santini et al. suggested that this lack of constraint among the posterior Hox cluster would have allowed these genes to be involved in the patterning of secondary axes in vertebrates, such as fins and limbs. In amphioxus, the same reasoning would apply to Hox14 and its unusual expression territories. If the origin of the posterior Hox cluster is truly independent, the decoupling of the Hox14 genes from the classical Hox code must have happened independently in the amphioxus and vertebrate lineages.
The escape of Hox genes from canonically nested expression is not unusual. For instance, the Hox1 and Hox2 paralogs of vertebrates also do not follow spatial colinearity, so that the anterior limit of Hox2 is more rostral than that of Hox1, which is expressed only in rhombomere 4. This delimited expression of Hox1 without extension to more caudal regions is similar to that of amphioxus Hox1 in the CNS. However, it is worth noting that this similarity cannot account for any homology between the Hox1 domain in amphioxus and rhombomere 4 in vertebrates. Furthermore, in lampreys, temporally colinear expression of Hox genes has not been detected. Thus, different Hox genes, in different animals escape in one way or another from the colinearity ‘rule’. We believe that these escapes might be associated with the patterning of lineage-specific morphological traits that first requires a loss of the constraint that kept them colinear.
We thank the anonymous reviewers for constructive criticisms that improved the manuscript. We are indebted to Hector Escriva and the ASSEMBLE FP7 EU programme for providing space and support during amphioxus sampling in Laboratoire Aragó, Banyuls-sur-mer, France. We thank Ina Arnone and Rossella Annunziata of the Stazione Zoologica Anton Dohrn of Naples (Italy) for their kind help in RA experiments and Ángel R. de Lera for providing the RA antagonist BMS009. We also thank Beatriz Albuixech-Crespo for her kind help with preparing the figures. The authors thank Senda Jiménez-Delgado, Ignacio Maeso, Manuel Irimia, Beatriz Albuixech-Crespo, and Ildikò M. L. Somorjai for their fruitful suggestions and discussions. We also want to thank Ignacio Maeso and Manuel Irimia for critical reading of the manuscript. This work was supported by the Ministerio de Educación y Ciencia (Spain) BMC2008-03776 and BFU2011-23921 and the ICREA prize to JGF. JP-A held a ‘FI’ PhD fellowship of the Generalitat de Catalunya (Spain) and SDA a ‘Juan de la Cierva’ postdoctoral contract of the Ministerio de Educación y Ciencia (Spain).
This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.