Genetic topography of the vertebrate dorsal mesoderm evolved through A/P expression domain shift of amphioxus mesodermal genes
In the dorsal mesoderm of amphioxus embryos, somites and notochord are regionalized during gastrulation, and an equivalent structure of the vertebrate unsegmented head mesoderm is thought to be absent in amphioxus [7] (Fig. 1). The dorsal mesoderm of vertebrates such as Xenopus is formed by massive mesodermal involution during the early-gastrula stage, which is not present in amphioxus embryos and is regionalized into head mesoderm, notochord, and somites [27, 37] (Fig. 1), suggesting that genetic programs govern differences between amphioxus and vertebrate mesodermal formation. We thus examined the molecular topography of the dorsal mesoderm in amphioxus and Xenopus. We first compared the mid-gastrula stage of amphioxus embryos and stage 10.5 Xenopus embryos because the orientation of dorsal mesodermal tissue is approximately parallel to the A/P body axis in both species, and thus comparable at these stages [33] (Additional file 1: Figure S1). The key regional genes include gsc (head mesoderm), bra (notochord), delta (somite), wnt8 (somite) and dkk1 (head mesoderm, somite).
By the mid-gastrula stage, all genes examined were expressed around the blastopore and showed similar patterns in amphioxus and vertebrates (Fig. 2a and b). However, by the late-gastrula stage of Xenopus, the expression domains of the regional marker genes became separated anteroposteriorly, with the gsc expression domain barely overlapping with those of delta, wnt8, and bra as the head and trunk mesodermal identities became distinct (Fig. 2c–f). These dynamic shifts were also observed in basal vertebrates, such as the lamprey (L. japonicum) and shark (S. torazame) (Additional file 1: Figures S2, S3 and Table S1). During the late-gastrula stage, the presomitic mesodermal region became distinct around the blastopore, and dkk1/2/4, delta, bra, and wnt8 were co-expressed in Xenopus presomitic mesoderm (Fig. 2d). In amphioxus, somites were found to form directly from the tail bud, and the expression of dkk1/2/4, delta, bra, and wnt8 largely overlapped in the prospective tail bud region (Fig. 2c). These results suggest that during gastrula stages, the mesodermal genes segregate anteroposteriorly only in vertebrates, whereas in amphioxus, these genes overlap considerably (Fig. 2f).
Interference of mesodermal involution promotes an amphioxus developmental mode in Xenopus embryos
Just after the stage in which mesodermal gene expression arrangements were similar in amphioxus and Xenopus, the dorsal mesoderm in Xenopus spread anteriorly to the blastocoel to form the head mesoderm (Fig. 3a and b). However, in amphioxus, the relative increase in mesodermal size was much lower than that in vertebrates (Fig. 3c). This suggests that, in vertebrates, the dynamic mesodermal gene shift is achieved by increasing the mesoderm, which is primarily dictated by mesodermal cell movements (e.g. involution, convergence and extension) (Fig. 3c). During gastrulation, mesodermal involution is controlled by convergence and extension of the dorsal axial mesoderm in vertebrates [38, 39]. To determine whether mesodermal involution is essential for the mesodermal gene shift, we suppressed convergent extension by inhibiting the Wnt planar cell polarity (PCP) signalling pathway, (a key signal transduction pathway in convergence and extension) [40, 41]. For the loss of function study of Wnt/PCP signalling, we injected Xldd1 (a dominant-negative form of Xldsh; [42, 43]) mRNA into Xenopus embryos [26].
In the dye only-injected control embryos, labelled cells migrated anteriorly and expressed gsc but not delta-2 or bra (Fig. 3d, e, g and i). In Xldd1 mRNA-injected embryos, however, labelled cells did not migrate anteriorly, but remained close to the blastopore, and gsc, delta-2, and bra were not separated anteroposteriorly as observed in the control embryos during the late-gastrula stage (Fig. 3f, h and j). Additionally, the size of the dorsal mesoderm was much smaller in the Xldd1 mRNA–injected embryos compared with control embryos (Fig. 3e–j), indicating that the developmental sequences of the dorsal mesoderm were somewhat similar to those in amphioxus. Microinjection of XldshdelDEP mRNA, a mutant dsh that specifically inhibits the Wnt/PCP signalling pathway [44], indicated that the effect of Xldd1 injection resulted from suppression of the Wnt/PCP signalling pathway (Additional file 1: Figure S4A–F). Overlapping head and trunk marker gene expression was also detected based on the results of the dorsal marginal zone assay, suggesting that the effect of Xldd1 injection was not attributable to differences in the amount of yolk, but more likely to the loss of mesodermal cell movement (Additional file 1: Figure S4I–L).
These results are consistent with morphological changes observed at the tail bud stage in Xldd1-injected Xenopus embryos assimilated to an amphioxus-like condition. Specifically, the anterior-most somite, normally appearing just posterior to the anterior end of the notochord, was extended anteriorly into the prechordal domain (Fig. 3k–p). In these embryos, the somite marker myoD was expressed normally, similar to mrf1 in amphioxus embryos (Fig. 3q–s). Interestingly, ectopic expression of tbx1, a head mesoderm marker in vertebrates [10], was also detected in the Xldd1-injected Xenopus somites, similar to amphioxus tbx1/10, a somite marker in amphioxus [45], expression (Fig. 3t–w and Additional file 1: Figure S4G and H). These results suggest that the vertebrate-specific mesodermal involution during gastrulation is likely responsible for the A/P distinction between head and trunk mesoderm, which does not occur in amphioxus.
A/P patterning in the dorsal mesoderm by different Wnt/β-catenin-signalling input is a vertebrate novelty
Previous functional studies in vertebrates have shown that the dorsal mesoderm is regionalized by a Wnt/β-catenin-signalling gradient along the A/P axis during early embryogenesis [46]. The failure of A/P segregation of mesodermal regional gene expression in Xldd1 mRNA-injected embryos suggests that Wnt/β-catenin-signalling pathway control of these downstream genes is compromised in this context. We examined the nuclear localization of β-catenin, a downstream factor in the Wnt/β-catenin signalling pathway, in Xldd1 mRNA-injected embryos. In the control embryos, nuclear localization of β-catenin was observed in the posterior dorsal mesodermal cells, but not in the anterior region (Fig. 4a–c). In Xldd1 mRNA-injected embryos, however, β-catenin localized to the nucleus in some cells, but there was no clear A/P difference in the degree of nuclear localization (Fig. 4d–f). The lack of obvious A/P difference in β-catenin nuclear localization was also observed in the amphioxus dorsal mesoderm (Fig. 4g–l). Consistent with this result, a previous functional study showed that Wnt/β-catenin signalling had no role in the A/P patterning of the dorsal mesoderm during the gastrula stages [23]. These findings suggest that anterior low and posterior high Wnt/β-catenin-signalling input is important in the segregation of regional marker genes of the dorsal mesoderm along the A/P axis that evolved in vertebrates.
Evolution of head mesoderm in vertebrates
In this study, we propose that the vertebrate dorsal mesoderm evolved as an entirely novel pattern associated with a new mechanism of mesoderm specification. As was first described by Ernst Haeckel [47], amphioxus gastrulation involves a simple invagination similar to that in cnidarians, with little change in the spatial relationships between the ectoderm and mesoendoderm during development. Unlike in vertebrates, the dorsal mesoderm in amphioxus co-expresses both vertebrate head and trunk mesoderm marker genes (Fig. 5). Thus, the amphioxus dorsal mesoderm remains unspecified along the A/P axis due to the absence of a vertebrate-specific developmental program (Fig. 5). Vertebrate mesoderm, however, is uniquely polarized along the A/P axis into the head and trunk mesoderm based on mesodermal patterning mediated by vertebrate-specific cell movement. By the end of gastrulation, the head and trunk mesodermal identities are specified by anteroposteriorly dislocated expression of regional marker genes. This patterning mechanism also controls A/P patterning of the overlying neuroectoderm. In vertebrates, the A/P regional identity of the CNS is organized by vertical signals from the underlying dorsal mesoderm containing A/P pattern information [48]. In amphioxus, based on the topography of regional markers, the CNS is patterned into domains largely comparable to the fore-/mid-/hindbrain and the spinal cord in vertebrate embryos [5]. However, three major signalling centres (anterior neural ridge, zona limitans intrathalamica, and midbrain–hindbrain boundary) are absent in the amphioxus CNS [5, 49]. Given that our current study indicated that the A/P regional identity of the vertebrate dorsal mesoderm is fundamentally different from that of amphioxus (Fig. 5a), the three major signalling centres in the neuroectoderm of vertebrate embryos may have evolved through a reorganization of the dorsal mesoderm in an amphioxus-like chordate ancestor.
Evolutionary reorganization of the entire dorsal mesoderm as described above would imply that individual amphioxus somites are not homologous to any specific region of the vertebrate head mesoderm. Our scenario instead favours the novel nature of the vertebrate mesoderm generated by modification in mesodermal patterning dynamics. From the perspective of vertebrate mesodermal specification, the segmented mesoderm in amphioxus appears as an intermediate between the head mesoderm and trunk somites, and vertebrate somites do not represent primitive traits, but rather derived traits established by removal of head mesoderm-like properties from ancestral somites. This scenario of vertebrate head evolution correlates with the observation that peripheral nerves in amphioxus possess traits of both cranial and spinal nerves [50, 51]. The mesodermal developmental pattern is shared among chordates only in the early gastrulae, in which the mesoderm is not yet anteroposteriorly polarized, possibly representing a plesiomorphic state (Fig. 5). The A/P patterning of mesodermal identities through the different Wnt/β-catenin-signalling input takes place only in vertebrates later in the developmental process; along with unique changes in cell movement, this can be considered a synapomorphic developmental trait for this animal subphylum. As proposed by Haeckel, this indicates that the vertebrate body plan is established by recapitulating an amphioxus-like ancestral pattern (plesiomorphy) during the early-gastrula stage and engaging a novel pattern (synapomorphy) during the later stages. The comparison of mesodermal gene expression of vertebrates, amphioxus, and hemichordates (as an out-group taxon) suggests that the amphioxus mesoderm has an intermediate nature, possibly representing a plesiomorphic state for deuterostomes (http://www.ibiology.org/ibioseminars/evolution-ecology/marc-w-kirschner-part-3.html). This characterization of the evolutionary sequence of developmental dynamics in chordates also provides insight into the potential origins of mesoderm and mesodermal segments in bilaterians.
Possible mechanism of mesodermal involution unique to vertebrates
In this study, we showed that inhibition of mesodermal involution in vertebrate embryos recapitulated amphioxus development (Fig. 3). The lack of mesodermal involution and likely convergent extension in amphioxus gastrulation indicates that the developmental program for mesodermal involution in vertebrates is absent in amphioxus. Disruption of cadherin-mediated cell–cell adhesion is essential during mesodermal involution and convergent extension in vertebrates, and fibronectin leucine-rich-repeat transmembrane 3 (Flrt3) and a small GTPase (Rnd1) control C-cadherin degradation [26, 34, 52, 53]. A BLAST search revealed a homologue of rnd1 in amphioxus, but not of flrt3 (Additional file 1: Figure S5A and B). In amphioxus, rnd1 expression was observed around the blastopore (Additional file 1: Figure S6A–H). However, overexpression of Bfrnd1 mRNA could not rescue the loss of endogenous rnd1 in Xenopus (Additional file 1: Figure S6I–M). These findings suggest that involvement of the cadherin degradation system in mesodermal involution as well as convergence and extension may have emerged in the vertebrate lineage.