NK and NKL gene repertoires in Onychophora and Tardigrada
Our searches and phylogenetic analyses revealed a total of 17 NK and NKL genes in the onychophoran E. rowelli and the tardigrade R. varieornatus, whereas the other analyzed tardigrade species, H. exemplaris, possesses 18 genes. We identified nine NK cluster genes, including single copies of NK1, NK3, NK4, NK5, Msx, Lbx, Tlx and two different transcripts of NK6, in the transcriptome of E. rowelli. The genomes of the two tardigrade species and the transcriptome of H. exemplaris show identical sets of eight NK cluster genes, which comprise single copies of NK1, NK3, NK4, NK5, NK6, Msx, Lbx and Tlx. These results show that, apart from the duplication of NK6 in onychophorans (or the onychophoran subgroup containing E. rowelli), onychophorans and tardigrades possess an identical set of eight NK cluster genes, whereas NK7 is missing in both taxa. However, the presence of NK7 in the genomes of several arthropods [Additional file 1; 23, 58] as well as other bilaterians [Additional file 1; 24, 59] suggests that a complete “protoNK cluster” [7, 23, 60, 61] consisting of nine genes was present in the last common ancestor of Panarthropoda and that NK7 was most likely lost in the onychophoran and tardigrade lineages.
Interestingly, localization of NK cluster genes in the genomes of the two tardigrade species revealed that most of these genes are located on different scaffolds. While the NK cluster genes are encoded on six scaffolds in H. exemplaris (with four scaffolds containing one and two scaffolds including two NK genes each), they are encoded on three scaffolds in R. varieornatus (with each scaffold containing one, two and five NK genes, respectively). The gene pairs NK3/NK4 and NK6/Tlx occur in both tardigrade species, suggesting that these pairs were present in the genome of the last common ancestor of the two species. However, even those NK genes that are located on the same scaffold do not neighbor each other but lie at a distance ranging from several kb to up to ~ 1.5 Mb. Moreover, large numbers of non-NK genes are interspersed between the individual NK genes. For comparison, the NK cluster of the fruit fly D. melanogaster spans only about 180 kb and contains only four non-NK genes [18, 23]. We therefore conclude that the NK cluster might have been fragmented in tardigrades. Similar breakups have been reported for the tardigrade Hox cluster (see supplementary figure S3 in Ref. [62]) as well as for the Hox and ParaHox clusters in other bilaterians, including the fruit fly D. melanogaster, the nematode C. elegans and the tunicate Ciona intestinalis [63,64,65]. Comparative analyses of the genomes of H. exemplaris and R. varieornatus generally revealed a low level of synteny, which might be due to extensive intrachromosomal rearrangements similar to those in C. elegans [66]. It has been hypothesized that such breakups of gene clusters are correlated with rapid modes of embryogenesis, possibly due to the loss of temporal collinearity and corresponding regulatory mechanisms in these taxa [63]. Since tardigrades also show rapid embryonic development (4–5 days in H. exemplaris [67]), this might be indeed due to the disintegrated homeobox clusters in these animals.
In contrast to the two tardigrade species, well-assembled genomic data are currently unavailable for Onychophora. Since neither the location nor the orientation of NK genes are known in E. rowelli, it remains unclear whether or not the ancestral NK cluster has been retained in the onychophoran genome.
In contrast to the almost identical sets of NK cluster genes in onychophorans and tardigrades, the NKL gene complement is variable in the three species studied. While E. rowelli shows eight NKL genes (NK2.1, NK2.2, Nedx, vax, Emx, Bari, BarH and Hhex), we found ten NKL genes in H. exemplaris (Abox, Ro, Nedx, Emx, Hhex, plus two copies of Barx and three copies of Barh) and nine in R. varieornatus (NK2.1, NK2.2, Abox, Ro, Nedx, Emx, Hhex, plus two copies of Barh). These results, together with the NKL complements of various arthropods, indicate that the last common ancestor of Panarthropoda might have possessed single copies of each NKL gene except for Nanog and Ventx, which were missing, and that several gene losses and gene duplication events occurred in different lineages (Fig. 1B; Additional file 1). The high variation of the NKL gene repertoire in tardigrades, onychophorans and arthropods suggests that the NKL genes are more prone to evolutionary changes than the NK cluster genes.
Segment polarity-like expression patterns of NK1, Lbx and Msx in the onychophoran E. rowelli
Our data revealed that NK1, Lbx and Msx are expressed early in development and show a stereotypic pattern in the mesoderm (somites) of the onychophoran embryo. Each somite of E. rowelli exhibits a wide anterior NK1 domain, followed by a diffuse medial Msx domain and a posterior Lbx domain (Fig. 14A). This pattern strongly resembles the expression of the so-called segment polarity genes (SPG), which are also expressed in segmentally reiterated domains early in development that are aligned in defined anteroposterior positions within each segment [55, 68,69,70]. To our knowledge, a similar expression pattern of these three NK genes has not been reported from any arthropod species studied thus far, except for an early Lbx expression in D. melanogaster that does show an SPG-like pattern in both ectoderm and mesoderm [71]. Interestingly, NK1, Lbx and Msx are also expressed in an SPG-like pattern in the annelid Platynereis dumerilii, which shares with E. rowelli the formation of somites during its larval development [19]. Based on the distinct complementary expression of these genes early in development of P. dumerilii, the authors [19] have concluded that the role of the NK genes in anterior–posterior segment patterning might be conserved in protostomes. If so, one would expect these genes to be expressed in a similar anterior–posterior alignment in onychophorans.
However, besides NK1, Lbx and Msx at least two other NK cluster genes, namely NK4 and Tlx, are expressed in an SPG-like pattern in P. dumerilii (Fig. 14B). Furthermore, despite the SPG-like expression patterns of NK1, Lbx and Msx, the relative positions and dimensions of individual domains within each somite clearly differ between E. rowelli and P. dumerilii. For example, while NK1 and Lbx are expressed in the anterior and posterior halves of each somite in E. rowelli, respectively, both genes are co-expressed in the anterior portion of each somite in P. dumerilii (Fig. 14A, B). Likewise, Msx is expressed in the posterior two-thirds of each somite in E. rowelli, while it occupies the anterior portion of each somite in P. dumerilii. Most importantly, the expression of Lbx and Msx is restricted to the mesoderm in E. rowelli, whereas these genes show additional ectodermal domains in P. dumerilii (Fig. 14A, B). These major positional discrepancies suggest considerable differences in segment patterning mechanisms in these two taxa.
As another line of evidence, Saudemont et al. [19] state that the expression of Lbx is co-localized with the expression of wingless in P. dumerilii in a domain anterior to engrailed, which is similar to the situation in D. melanogaster, where Lbx expression is co-localized with and dependent on wingless expression. Based on these similarities, they consider it as unlikely that the complementary expression of Lbx, wingless and engrailed might have been recruited independently in annelids and arthropods [19]. In contrast to this, Lbx is not expressed in the ectoderm of E. rowelli at any developmental stage and, thus, is not co-localized with the exclusively ectodermal wingless expression [55]. Instead, the Lbx domain might partially overlap with the mesodermal engrailed domain in onychophorans, which is devoid of Lbx expression in D. melanogaster and P. dumerilii. Consequently, one would have to assume that this pattern has been modified to a large extent in onychophorans, while it was retained in annelids and arthropods. However, the argumentation of Saudemont et al. [19] is based on the expression of only a single NK gene in two distantly related species. Further data from other protostome taxa, including other annelids [72, 73] and arthropods [74,75,76,77,78], so far have not revealed any evidence for an involvement of NK genes in segment formation, which would indicate several independent losses of this pattern in many protostome taxa.
In summary, if the NK genes were involved in the anterior–posterior regionalization of segments in the protostome ancestor, as proposed by Saudemont et al. [19], one would expect a similar set of NK genes to be expressed in defined anteroposterior positions in most protostome taxa, similar to what has been described from the segment polarity genes engrailed, hedgehog, wingless and cubitus interruptus [55, 57, 68]. However, the considerable differences and the lack of specific similarities in the expression patterns observed in P. dumerilii and E. rowelli, as well as the absence of similar patterns in almost all protostomes studied thus far rather indicate that NK1, Msx and Lbx might have been recruited independently to fulfill similar functions in the regionalization of segments in annelids and onychophorans. On the other hand, even though these genes show a segmentally reiterated pattern in both species, which resembles an SPG-like expression, so far there is no convincing evidence for their potential role in the anterior–posterior regionalization of segments.
Conserved expression patterns of NK cluster genes in the mesoderm of E. rowelli
One of the most intriguing features of the NK cluster genes is their seemingly conserved expression in mesodermal derivatives across bilaterians, including various somatic muscles [6, 7, 18]. For example, NK1, Msx and Lbx are expressed in non-overlapping patterns in different sets of longitudinal and parapodial muscles in the annelid P. dumerilii, suggesting that their transcripts might provide identity information for the differentiation of these muscles [19]. In the fruit fly D. melanogaster, NK1, Msx and the two Lbx orthologs (lbl and lbe) are mainly expressed in subsets of muscle founder cells, including the developing dorsal body wall muscles, lateral and segmental muscles [18, 37, 71, 77, 79]. In vertebrate embryos, including the mouse Mus musculus and the chicken Gallus gallus, the Msx and Lbx orthologs are expressed in specific sets of muscle precursors that will give rise to the limb musculature [38,39,40,41, 80,81,82].
Similarly, our data revealed that NK1, NK3, NK4, NK5, Lbx and Tlx are expressed in distinct, mesodermal domains in the developing limbs in embryos of E. rowelli. NK1, NK4 and Lbx show a similar pattern in each developing leg, which follows an anterior-to-posterior progression in development. Similar Tlx domains appear in all developing limbs, although they differ in size and shape in the developing jaws and slime papillae, which are modified cephalic appendages and show a derived muscle arrangement ([56, 83]; see also Fig. 9A–C in Ref. [84]).
These findings indicate that the mesodermal domains of NK1, NK4, Lbx and Tlx might correspond in position to the individual developing limb muscles, and that these genes might provide positional identity information for the limb muscles in onychophorans. However, since development of the somatic musculature of onychophorans has not been studied and since the exact number and arrangement of individual leg muscles has not yet been clarified [85,86,87,88,89,90], the relation of these expression patterns to the development of specific sets of muscles in the onychophoran limbs remains unclear. Nevertheless, our results indicate that the involvement of NK genes in the development and differentiation of somatic musculature might be conserved in Bilateria [19], or at least Nephrozoa, since comparative NK gene expression data are missing from Xenacoelomorpha, the sister group of Nephrozoa [16].
In contrast to the mostly similar mesodermal expression patterns of NK1, NK4, Lbx and Tlx, the posterior domains of NK3 and NK5 are elongated in the fourth and fifth developing legs with respect to their corresponding domains in the remaining limbs. Compared to the jaws and slime papillae, however, these limbs do not show a derived muscle arrangement that would explain these differences [56]. Interestingly, these patterns resemble the expression of odd-skipped, pox-neuro and pax3/7, which also show enlarged domains in the fourth and fifth legs [11, 68]. This pattern might correspond to the anlagen of specific types of nephridia along the onychophoran body: The small NK3 domains in the first three leg-bearing segments might be localized in the tiny nephridial anlagen of these segments, while the elongated NK3 and NK5 domains in the fourth and fifth leg-baring segments might be associated with the large, specialized “labyrinth organs” developing in these segments ([91,92,93]; see also Fig. 10A in [94]). This striking pattern indicates that these genes might be involved in the development of nephridia and their derivatives in Onychophora. To our knowledge, a comparable role of NK genes in nephridiogenesis has only been reported from the annelid P. dumerilii thus far [19].
Apart from the expression of NK genes in the developing limb muscles, we observed peculiar mesodermal stripes of NK4 expression along the dorsal rim of the lateral blastoderm bands. The position of this expression might correspond to mesenchymal cells of the dorsal coelomic linings that move into the space above the midgut to form the future heart [56, 95,96,97]. This pattern strongly resembles the previously described expression of the T-box gene H15 in the dorsal tube of the onychophoran Euperipatoides kanangrensis [98]. This, in turn, is reminiscent of the overlapping expression of NK4 and H15 in the beetle Tribolium castaneum, the fly D. melanogaster and the spider Cupiennius salei along the dorsal rim of the lateral germ bands in cells that have been identified as heart precursors in these animals, suggesting that this pattern might be conserved in onychophorans and arthropods [18, 28,29,30, 35]. These similarities support the hypothesis that the panarthropod hearts are homologous. In contrast to onychophorans and arthropods, however, a vascular system with a pulsatile organ is absent from the third major panarthropod taxon, the tardigrades. This absence might be a secondary loss due to the miniaturized body of tardigrades [99]. Depending on the phylogenetic position of tardigrades, our data suggest that the last common ancestor of either panarthropods or onychophorans plus arthropods possessed a pulsatile dorsal vessel that expressed NK4 during development (Fig. 15A, B).
Interestingly, studies of NK4 expression in other bilaterians, including annelids and chordates, revealed similar patterns in the pulsatile dorsal vessel or heart, which has led to the hypothesis that a role in the formation of the pulsatile dorsal vessel or heart might be conserved among bilaterians, or at least nephrozoans, since xenacoelomorphs do not possess a circulatory system [19, 100,101,102]. Thus, the “urnephrozoan” might have already possessed a pulsatile dorsal vessel. A possible conservation of the NK4 expression in heart precursor cells and a common origin of the heart in panarthropods or onychophorans plus arthropods might support this hypothesis. However, many protostome taxa, including cycloneuralians and numerous spiralian/lophotrochozoan taxa, do not possess heart-like organs or pulsatile vessels. If a heart-like structure indeed evolved only once in the nephrozoan ancestor, one would have to assume multiple independent losses of pulsatile organs in these taxa. Like tardigrades, many of these taxa show extensive miniaturization [103]. Since it has been shown that miniaturization often results in the reduction or complete loss of organs or entire organ systems, multiple independent losses of the vascular system including heart-like structures seem possible [103]. In contrast to this, NK4 expression data from the cephalopod mollusk Sepia officinalis revealed that NK4 is not involved in the formation of the heart but rather in somatic muscle development in this species [104]. Consequently, if one assumes a single origin of the heart in the nephrozoan ancestor, mollusks would have lost NK4 expression in cardiac tissue but retained the heart itself (Fig. 15A).
Interestingly, the example of S. officinalis shows that NK4 is not essential for proper heart development and that the formation of pulsatile tissue can be mediated by other molecular mechanisms [104]. The expression of NK4 in the somatic musculature in S. officinalis rather indicates an ancestral role of this gene in somatic muscle development. Thus, an alternative scenario of heart evolution in nephrozoans is conceivable (Fig. 15B). According to this scenario, a heart was absent in the last common ancestor of nephrozoans and pulsatile organs evolved four times independently in deuterostomes, panarthropods (or onychophorans plus arthropods), annelids, and mollusks. According to this scenario, NK4 might have been involved in somatic muscle development in the nephrozoan ancestor and was recruited independently to fulfill major regulatory functions during heart development in annelids, panarthropods and deuterostomes but not in mollusks.
Non-regionalized neural expression of NK cluster genes and the NKL gene NK2.2 in E. rowelli supports convergent evolution of bilaterian nerve cords
In addition to mesodermal domains, we observed an expression of NK3, NK5, NK6.1, NK6.2, Lbx, Msx and NK2.2 in the developing nervous system of the onychophoran embryo. While NK3, NK5, Msx and NK2.2 are expressed in the anlagen of both the brain and the ventral nerve cords, the expression of the two NK6 copies is restricted to the ventral nervous system, and Lbx is confined to the developing brain. Interestingly, an involvement of NK5, NK6, Msx, Lbx, Tlx and NK2.2 in neural development has also been reported from other bilaterians, including arthropods, annelids and vertebrates [18, 19, 44, 105]. This supports the assumption that these genes were involved in neural development in the last common bilaterian ancestor [18, 19, 44].
In the trunk of E. rowelli, NK5, NK6.1, NK6.2, Msx, Lbx, Tlx and NK2.2 are expressed in uniform domains along the body. However, NK2.2, Msx and both copies of NK6 show peculiar mediolaterally regionalized patterns (Fig. 16A). While NK2.2 is expressed in parallel, continuous, medial and lateral stripes along the body, Msx and both copies of NK6 show both, continuous bands of expression within each nerve cord as well as segmentally reiterated domains in the medial ectoderm (Fig. 16A). Interestingly, Msx, NK6 and NK2.2 have been reported to be involved in the mediolateral regionalization of the nervous system in vertebrates, the fly D. melanogaster and the annelid P. dumerilii [19, 44, 46, 105]. This regionalization is mediated by the staggered expression of the transcription factors NK6, Msx, NK2.2, Pax6 and Pax3/7 in the neuroectoderm of these animals (Fig. 16B), thus defining a specific mediolateral arrangement of neurogenic domains and neuron types [19, 44, 105]. While the medial NK2.2+/NK6+ domain gives rise to a medial column of serotonergic neurons, the adjacent NK6+/pax6+ area forms an intermediate column of cholinergic motor neurons and the lateral pax6+/pax3/7+ and pax3/7+/msx+ domains establish a lateral column of interneurons and lateral sensory trunk neurons [17, 44]. This seemingly conserved mediolateral patterning has been used as a major argument for proposing an ancestral condensed, mediolaterally patterned ventral nerve cord in the “urbilaterian” [17, 44, 105]. The absence of a mediolateral patterning of these genes in hemichordates, nematodes and planarians has been interpreted as independent losses in these lineages [44].
However, recent studies of representatives of Xenacoelomorpha revealed that NK6, Msx, NK2.2, pax6 and pax3/7 expression is unrelated to the trunk neuroanatomy in these taxa, suggesting that the mediolateral patterning evolved after the Xenacoelomorpha–Nephrozoa split [17]. Interestingly, investigation of these genes in additional nephrozoan taxa, including Rotifera, Nemertea, Brachiopoda and Enteropneusta, revealed that a mediolateral regionalization is also absent or only partially present in these taxa [17, 42]. Moreover, despite the presence of mediolateral regionalization in the annelid P. dumerilii (Errantia, Phyllodocida), there is no evidence for such a pattern in the annelid Owenia fusiformis (Sedentaria, Sabellida), although its trunk neuroanatomy largely resembles that of P. dumerilii, D. melanogaster and vertebrates [17]. Thus, two possible scenarios on the evolution of condensed medial nerve cords among bilaterians have been proposed ([17] but see an opposing view in Ref. [106]). According to the first scenario, the mediolateral patterning of the central nervous system in vertebrates, D. melanogaster and P. dumerilii is homologous, thus reflecting the ancestral bilaterian or nephrozoan state, which would imply multiple independent losses or modifications. Alternatively, the absence of mediolateral regionalization of the central nervous system in xenacoelomorphs, many spiralians and some annelids might indicate a convergent evolution of this patterning system in vertebrates, arthropods and some annelids.
Our results show that NK2.2, NK6.1, NK6.2 and Msx are expressed in dynamic patterns early and late in development of E. rowelli. At the onset of neurogenesis, when neural precursors segregate from the ventral ectoderm [56], expression of these genes is mainly restricted to the ventral ectoderm (Fig. 16A). NK2.2 and Msx are expressed in non-overlapping medial and lateral domains, respectively, that resemble the NK2.2 and Msx domains of other bilaterians. In contrast to this, NK6 is expressed in a lateral domain, overlapping with Msx, but not with NK2.2 (Fig. 16A). This is different from what has been reported from arthropods, annelids and vertebrates, where NK6 is expressed in a medial domain, overlapping with NK2.2 but not with Msx (Fig. 16B).
Later in development, after most neural precursors have been segregated and the developing nerve cords have delaminated from the ventral ectoderm, these genes are expressed in both the ventral ectoderm and the nerve cords where they show different patterns (Fig. 16A). In the ventral ectoderm, NK2.2 and NK6 are expressed in medially restricted, largely overlapping domains, while Msx expression is confined to the central region, partially overlapping with the medial NK2.2 and NK6 domains. NK2.2 shows an additional domain in the lateral ectoderm, which partially overlaps with the Msx domain, but not with the medial NK6 domain. In the nerve cords, NK6 and Msx are expressed in broad, largely overlapping domains that are confined to the medial part, while NK2.2 is restricted to a spot-like expression in the lateral nerve cord, overlapping with the NK6 and Msx domains (Fig. 16A).
Similarly, previous gene expression data on pax6 and pax3/7 did not provide any evidence for regionalized mediolateral patterning in Onychophora [11, 107]. Instead, pax6 is expressed in a broad domain, which covers the entire width of the nerve cord in late developmental stages (see Fig. 9D in [11]), thus overlapping with the NK2.2, NK6 and Msx domains and extending further laterally (Fig. 16A). In contrast, pax3/7 is not expressed in the developing nerve cords at any developmental stage (see Fig. 8A–D in [11] and Fig. 5A–C in [68]). Thus, irrespective of the developmental stage studied, neither our results nor previously published data on pax6 and pax3/7 expression provide any evidence for the presence of adjacent NK2.2+/NK6+, NK6+/pax6+, pax6+/pax3/7+ and pax3/7+/msx+ columns in the onychophoran nerve cords.
These results are in line with selective stainings of specific neurons, including retrograde fills of the leg nerves, and immunolabeling of various neurotransmitters and neuromodulators, which did not provide any evidence for adjacent columns of medial serotonergic neurons, cholinergic motor neurons, lateral interneurons or sensory neurons in the onychophoran nerve cords [108,109,110]. Instead, most of the neuron types are located in largely overlapping areas [108]. For example, the somata of serotonergic neurons are not restricted to a medial area but are rather distributed in a random fashion in the ventromedial and ventral perikaryal layers [108, 109].
While the expression of NK2.2, NK6 and Msx in the ventral ectoderm of early developmental stages as well as in the nerve cords of late developmental stages might be correlated to the specification of neurons, their corresponding patterns in the ventral ectoderm of late developmental stages are not necessarily correlated with any neural structures. Interestingly, the double-paired patterns of NK6 and Msx in the ventromedial ectoderm correspond to the emergence of paired ectodermal thickenings which are the anlagen of the ventral and preventral organs [84, 94, 111]. Although these structures arise from the ventral ectoderm, they do not seem to contribute any cells to the central nervous system but rather persist as attachment sites for segmental limb muscles in adult onychophorans ([56, 84, 112, 113] but see ref [114] for an opposing view on a putative function of ventral organs in neurogenesis). Interestingly, similar patterns have been reported from Delta and Notch transcripts that are expressed in two pairs of bilaterally symmetric domains on the ventrum of each segment in embryos of Euperipatoides kanangrensis and E. rowelli [84, 114, 115]. This characteristic, double-paired pattern appears only after most neurons have been segregated from the neuroectoderm [84, 109, 116]. Thus, it has been concluded, that these Delta and Notch domains specify regions of the ectoderm that give rise to the ventral and preventral organs rather than neurogenic tissue in the onychophoran embryo [56, 84]. Likewise, our results indicate that the double-paired segmental Msx and NK6 domains in the ventral ectoderm might provide identity information for specific regions of the ectoderm that give rise to the ventral and preventral organs rather than neurogenic tissue because the nerve cords have already been segregated from the ectoderm at this developmental stage (cf. Ref. [56]).
Thus, NK6, Msx, NK2.2, pax6 and pax3/7 expression seems to be unrelated to the trunk neuroanatomy in onychophorans, which is similar to what has been reported from Xenacoelomorpha, Rotifera, Nemertea, Brachiopoda and Enteropneusta, and some Annelida [17]. Consequently, if a complex medial nerve cord with a distinct regionalization of specific neuron types was present in the nephrozoan ancestor, one would have to assume either multiple losses of the mediolateral pattern in several lineages (including nematodes and onychophorans; Fig. 17B), or a single loss in the ecdysozoan ancestor followed by a secondary gain of this pattern in the arthropod lineage (Fig. 17C). The latter scenario would further imply a homology of the medially condensed nerve cords in vertebrates and annelids, but also an independent evolution of medially condensed nerve cords in arthropods (Fig. 17C). Alternatively, a mediolateral regionalization of the nerve cords might have never existed in the panarthropod ancestor, which would support the hypothesis [17] of convergent evolution of a condensed, mediolaterally patterned nerve cord in vertebrates, arthropods and some annelids (Fig. 17A).