Quadruple zebrafish mutant reveals different roles of Mesp genes in somite segmentation between mouse and zebrafish

Basic body structures are highly conserved among vertebrates. It is still not clear, however, to what extent the molecular mechanism underlying the generation of such structures is conserved. Somites, which are basic structures giving rise to skeletal muscle, vertebrae and dermis, are periodically generated in an anterior-to-posterior fashion during early embryogenesis. The periodical generation of somites can be generally explained by the ‘clock and wavefront model’, in which the timing and position of somite boundary formation are determined based on the spatiotemporal information generated by periodical activation of the ‘segmentation clock’ and posterior regression of the ‘determination front’, respectively (Cooke and Zeeman, 1976). Integration of such temporal and spatial information achieves the cyclic generation of presumptive somites in the anterior presomitic mesoderm (PSM). This model is supported by intensive studies using a number of different model animals to identify genes involved in the segmentation of somites (Hubaud and Pourquié, 2014; Oates et al., 2012).

However, in contrast to this conserved scheme of segmentation, the molecular network involved in the generation and propagation of the segmentation clock may not necessarily be conserved (Krol et al., 2011). In the mouse, distinct but mutually related molecular clocks have been identified, including the oscillatory activation of the transcriptional repressor Hes7, and Notch, Fgf and Wnt signaling, whose oscillations are generated by negative-feedback regulation (Hubaud and Pourquié, 2014). Her family genes, which are homologs of mouse Hes genes, have been considered as the core of the clock in the zebrafish. Oscillation of these genes is generated and maintained by self-repression and association with other Her proteins (Hubaud and Pourquié, 2014; Oates et al., 2012; Schröter et al., 2012; Trofka et al., 2012).

To understand the biological significance of the diversity of the molecular clock among species, it is important to show how such diversity is reflected in the morphogenesis of somites. Once the molecular clock is established, this temporal periodicity is converted into the spatial pattern of somites at or around the wave front. This conversion process consists of three steps (Dahmann et al., 2011; Saga, 2012a; Zhao et al., 2015). First, the position of a presumptive somite boundary is determined in the anterior PSM (boundary positioning). Next, epithelialization of cells adjacent to the presumptive boundary occurs to generate the physical segmental boundary of the somite (physical boundary formation). In parallel to physical boundary formation, rostro-caudal (R-C) polarity is established within the somite to provide positional information by which various type cells are generated from the somite in later development (R-C polarity formation).

In mouse somitogenesis, Mesp2, a basic helix-loop-helix (bHLH) transcriptional factor, functions as the key regulator in all three steps of this conversion process (Saga, 2012a). Mesp2 activates the expression of Ripply1 and Ripply2, which determine the somite boundary position through repression of Tbx6 protein (Oginuma et al., 2008; Takahashi et al., 2010; Zhao et al., 2015). Mesp2 expression is then restricted to the rostral half of predicted somite and activates the expression of Epha4, which is proposed to be involved in morphogenesis of the physical segmental boundary (Nakajima et al., 2006). In addition, Mesp2-dependent inhibition of Notch signaling, which is essential to induce the caudal identity of somite, is required for the establishment of the rostral identity of a somite (Takahashi et al., 2000; Morimoto et al., 2005; Sasaki et al., 2011). Since Mesp2 expression, which is periodical in the anterior PSM, is directly induced by periodically activated Notch signaling in mouse somitogenesis, Mesp2 is the master regulator connecting the segmentation clock to somite morphogenesis in the mouse (Morimoto et al., 2005; Yasuhiko et al., 2006). However, because of the lack of comparative analysis of the role of Mesp in other vertebrates, it is still uncertain as to what extent the molecular mechanism of this connection is conserved in other animals in which the temporal periodicity is established by a different molecular mechanism to that in the mouse.

Like the mouse, the zebrafish is a typical model animal widely used for genetic studies. The zebrafish has four Mesp genes, mespaa, mespab, mespba and mespbb, all of which are expressed in the anterior PSM during somitogenesis (Sawada et al., 2000; Durbin et al., 2000; Cutty et al., 2012). However, the roles of these Mesp genes in somitogenesis remain unclear. Zebrafish embryos injected with Mesp mRNA are defective in somite boundary and R-C polarity formation in their somites (Sawada et al., 2000; Durbin et al., 2000). In contrast, the injection of antisense morpholino specific for mespba and mepbb causes abnormal myogenesis but no obvious defect in the somite boundary formation (Windner et al., 2015). Thus, to clarify genetically whether these Mesp genes play a role in somitogenesis, we sought to generate a quadruple mutant for all of the zebrafish Mesp genes by using the TALEN-mediated genome editing approach. Analysis of the quadruple mutant embryos highlighted the conserved roles of Mesp genes between different species and also the species-specific mechanism governing positioning of the somite boundary.

It was predicted that the Mesp gene mutations would be null because mutant fish lines used in this study carried frame-shift mutations upstream of the bHLH domain of each Mesp gene (see also Discussion). Homozygous mutant embryos for any of the four Mesp genes, however, showed no obvious defect during embryogenesis, suggesting that the four genes are functionally redundant. Then, we sought to generate quadruple homozygotes for these four Mesp genes. Since mespaa and mespba, as well as mesppab and mespbb, are located close to each other on the same chromosome, the recombination rate between each pair of these genes would be predicted to be quite low (Fig. 1A). Thus, we first generated two independent double heterozygotes by introducing a mespaa mutation into mespba/+ fish and mespab mutation into mespbb/+ fish. The double mutants appeared normal (Fig. 2C-E). By sequential intercrossing between these double mutant lines, we finally obtained the mespaa^kt1017; mespba^kt1004; mespab^kt1002; mespbb^kt1009 quadruple homozygous mutant, referred to as the Mesp quadruple mutant, which lacked the function of all Mesp genes (Fig. 1B).

In contrast to mouse embryos defective in either or both Mesp1 and Mesp2, which show several obvious defects during early embryogenesis (Saga et al., 1997; Saga et al., 1999; Kitajima et al., 2000), the phenotype of zebrafish Mesp quadruple mutants appeared milder. In contrast to the mouse Mesp2 mutant, in which all somite boundaries are disrupted, most of somite boundaries, except the most anterior 3-5 somites, were still generated in the Mesp quadruple mutant embryos (Fig. 2A,B,C,F); although the morphology of each somite appeared to be different from the normal, chevron-shaped, one (Fig. 2C′,F). Furthermore, the defective mesoderm formation seen in the mouse Mesp1/Mesp2 double-mutant embryo and abnormal heart development reported in the mouse Mesp1 mutant were not observed in the Mesp quadruple mutant embryos (Fig. S1). We also noted that development of the lateral plate mesoderm, which had been reported to be impaired in mespaa morphants (Hart et al., 2007), was normal in Mesp quadruple mutant embryos (Fig. S1).

Several possibilities exist to explain this inconsistency between mutant and morphant phenotypes. One is that the TALEN-mediated mutations might not be null. However, in all of the mutations generated in this study, we introduced frame-shift mutations upstream of the bHLH domain, which is essential for the DNA binding of the Mesp protein. We also confirmed by PCR that unpredicted splicing (Fig. S2A), like exon skip, did not occur in the mutant embryos. Another possibility is that a small peptide fragment containing the bHLH domain, which may be alternatively translated from an AUG site downstream from the first AUG site, might have had an effect on the phenotype. Sequencing analysis revealed that such a possibility could not be excluded in the case of mespab and mespbb alleles. Thus, we further generated additional mutant alleles for these two Mesp genes, from which a functional product theoretically cannot be generated. However, the quadruple mutants containing these two alleles did not exhibit any additional defect (Fig. S2B-N). These findings strongly suggest that the TALEN-mediated mutations generated in this study were null.

Furthermore, we examined the possible contribution of maternally deposited Mesp gene transcripts or proteins to the phenotype of the Mesp quadruple mutant by generating maternal and zygotic (MZ) Mesp quadruple mutants. We did so by crossing Mesp quadruple homozygous fish, which could survive to adulthood and were fertile, because RT-PCR analysis revealed that maternally supplied transcripts of mespaa and mespba were detectable in zebrafish embryos. MZ-Mesp quadruple mutants, however, did not show enhanced abnormality in early mesoderm formation and patterning. Taken together, the lack of several phenotypes observed in mouse Mesp mutants and the zebrafish Mesp morphant was not caused by inefficient disruption of gene function in the quadruple mutant. Thus, the Mesp function was less critical during early embryogenesis in the zebrafish compared with that in the mouse or its function might have been compensated by functionally related genes. For instance, since a T-box protein, encoded by tbx16 (previously known as spadetail) and a homeobox protein, encoded by mixl1 (bonny and clyde), are redundantly required for cardiac mesoderm specification during zebrafish embryogenesis (Griffin and Kimelman, 2002), the lack of Mesp function might have also been compensated by these transcriptional factors in the Mesp quadruple mutant. As the MZ-Mesp quadruple mutant seemed to be identical to the zygotic mutant, we used the former embryos for subsequent experiments.

To better understand the role of Mesp genes in zebrafish somitogenesis, we then precisely examined somite development in the Mesp quadruple mutant embryos. As shown above, the homozygous Mesp quadruple mutant lacked several boundaries between its anterior somites. Only the Mesp quadruple mutant embryo showed this defect among the siblings obtained by the crossing of Mesp quadruple heterozygous parents, indicating highly redundant functions of Mesp genes. A similar defect specific to anterior somites is observed in fn1 or itga5 mutants, in which anterior somite boundaries are initially formed but not maintained (Koshida et al., 2005; Jülich et al., 2005). By contrast, the anterior boundaries were not generated at all in the Mesp quadruple mutant, indicating that Mesp function is not closely correlated with that of fn1 and itga5 (Fig. 2A,B).

In contrast to a few anterior somites, somite boundaries were created between most of the posterior somites. Previous studies revealed that the somite boundary could be established through two distinct processes: positioning of the somite boundary and morphogenesis of the actual physical boundary. The position of presumptive boundaries is initially defined at the anterior edge of the Tbx6 protein domain in both the mouse and zebrafish (Oginuma et al., 2008; Wanglar et al., 2014). In mouse Mesp2 mutant embryos, this anterior edge is not properly created; rather the Tbx6 protein domain is anteriorly extended. In the zebrafish Mesp quadruple mutant embryos, however, the anterior edge of the Tbx6 protein domain was still clearly formed as observed in wild-type embryos (Fig. 3A,B,D,E). However, ripply1 and ripply2 double-homozygous embryos, which we also generated by TALEN-mediated mutagenesis, exhibited abnormal extension of the Tbx6 protein domain, indicating that the function of Ripply genes is essential for the positioning of somites, as predicted by the results of morpholino-based studies (Fig. 3C,F and Fig. S3). Thus, the position of the segmental boundary was determined by a Mesp-independent but Ripply-dependent manner in zebrafish.

Since Mesp2 regulates the expression of Ripply1 and Ripply2, which subsequently define the anterior edge of the Tbx6 protein domain, in mouse somitogenesis (Takahashi et al., 2010; Morimoto et al., 2007), we next examined the role of Mesp for Ripply gene expression in zebrafish. In the wild-type embryo, ripply1 was expressed at both the anterior PSM and rostral compartment of each somite; whereas ripply2 was expressed in two to three strips in the anterior PSM (Fig. 3G,I,K,M). In the Mesp quadruple mutant, although the striped expression pattern of ripply1 was severely perturbed in somites, ripply1 and ripply2 expression patterns were not affected in the anterior PSM compared with that in the siblings (Fig. 3H,J,L,N). By contrast, when mespba was overexpressed in somites and in the anterior PSM region by using the tbx6; gfp-gal4 (ggff) driver (Fig. S4), expression of ripply1 and ripply2 was not ectopically induced in the anterior PSM; although somite expression of ripply1 was expanded to the caudal compartment of somites (Fig. 3O-T). Thus, in contrast to mouse, zebrafish ripply1 and ripply2 were expressed in the anterior PSM independent of Mesp function. The unaffected expression of ripply1 and ripply2 in the anterior PSM appears to preserve the normal positioning of somite boundaries in the Mesp quadruple mutant.

Next, we examined whether zebrafish Mesp genes were involved in morphogenesis of the physical boundary. In our preliminary observation of the Mesp quadruple mutant fish under a stereomicroscope, almost all of the somite boundaries, except for those of several anterior somites, were evident. However, more detailed analysis using Phalloidin staining revealed that the boundary morphology was partially disrupted, even in the posterior somites (Fig. 4A,B). During morphogenesis of somite boundaries in the zebrafish, initially established boundaries are gradually reinforced by the accumulation of ECM components such as fibronectin. This accumulation is required for the cells adjacent to the boundary to establish their apico-basal polarity, which is considered to be important for maintenance of the physical boundaries (Koshida et al., 2005). In the Mesp quadruple mutants, Fibronectin did accumulate at the somite boundaries; but the pattern of its assembly was partially disturbed (Fig. 4A,B). Thus, zebrafish Mesp plays a role in morphogenesis after the positioning of the presumptive boundaries; and its requirement is higher in the anterior somites than in the posterior ones.

Transplantation studies with zebrafish and chick embryos revealed that the repulsive interaction between Epha4 and EphrinB2 is sufficient to generate physical boundaries between transplants and host cells (Barrios et al., 2003; Watanabe et al., 2009). Considering the expression patterns of these genes, this repulsive interaction occurs prior to the accumulation of Fibronectin. Since epha4 expression is directly regulated by Mesp2 during somitogenesis in the mouse, we next examined epha4 expression in the Mesp quadruple mutant (Fig. 4C,D,G,H). In normal zebrafish embryos, epha4 is expressed in a striped pattern in the anterior PSM and somites, and in a gradient fashion along the anterior-posterior (A-P) axis in the posterior PSM. In contrast, the striped expression pattern of epha4 was completely absent in the Mesp quadruple mutant; however, the gradient in expression was not affected. Thus, Mesp was specifically required for the striped expression pattern of epha4 in the zebrafish, as is reported to be the case in the mouse (Nakajima et al., 2006).

On the other hand, previous studies suggested that pcdh8 (papc) is also involved in the morphogenesis of somite boundaries in the mouse (Rhee et al., 2003). In the wild-type embryo, striped expression of pcdh8 was detected in the anterior PSM, whereas in the Mesp quadruple mutant, this striped expression pattern was not affected, although the expression level was slightly reduced, suggesting that the expression of pcdh8 is mostly independent of Mesp (Fig. 4E,F,I,J). These results suggest that two parallel pathways – one Mesp dependent and one Mesp independent – regulate the morphogenesis of physical somite boundaries in the zebrafish.

Mesp2 function is also essential for R-C polarity formation within a somite in the mouse. In zebrafish, although overexpression studies revealed that Mesp genes have a potential to induce rostral identity in somites, their physiological role in R-C polarity formation remains unclear (Sawada et al., 2000; Durbin et al., 2000). We found that expression of tbx18 and fgf8 in the rostral compartment of somites was completely missing in the Mesp quadruple mutant embryo (Fig. 5A-D,I,J). In contrast, the caudal expression of uncx4.1 and myod expanded to the entire somite (Fig. 5E-H,K,L). We also confirmed that mespba was sufficient to induce tbx18 expression and repress uncx4.1 expression in the caudal half of each somite by overexpression of mespba in the anterior PSM and somite region by using tbx6:ggff driver (Fig. 5M-P). Thus, Mesp genes are necessary and sufficient to establish rostral identity in the zebrafish somite, as in mouse somitogenesis.

Previous studies reported that the R-C polarity in a somite is relayed to the sclerotome and essential for the pattern formation of vertebrae (Saga, 2012b). In the zebrafish Mesp quadruple mutant, duplication of hemal spines was observed in some vertebrae, implying that the spines of zebrafish vertebral bone were derived from the caudal compartment of somites, as in the case of the mouse vertebral arch, and duplicated by the caudalization of the somite (Fig. 5Q″,R″). In addition, the tips of several neural spines had split or fused abnormally (Fig. 5Q′,R′), suggesting that Mesp-dependent generation of the R-C polarity is required for the proper formation of vertebrae in later development. Although the R-C polarity was severely impaired along the entire body axis (Fig. 5B,F,J,L), skeletal defects were more evident in the anterior vertebrae than in the posterior ones. In contrast, the segmentation of vertebrae was not affected by Mesp mutation, supporting previous studies showing that vertebral columns are segmented independent of somite segmentation (Fleming et al., 2004).

In the zebrafish Mesp quadruple mutant, none of the somites showed the chevron-shape morphology seen in the wild-type embryo (Fig. 2C,F). Such abnormal morphology is frequently caused by lack of the horizontal myoseptum, which is induced by hedgehog signaling and separates the dorsal and ventral compartment of fast muscle fibers (Halpern et al., 1993; Barresi et al., 2000). This horizontal myoseptum develops from adaxial cells, which are a subpopulation of paraxial mesoderm cells neighboring the notochord (Devoto et al., 1996). A population of these adaxial cells, located in the rostral part of a somite, differentiates into muscle pioneers (MPs), which are identified by their expression of prox1 and engrailed (Felsenfeld et al., 1991; Roy et al., 2001). Then, these MPs elongate along the A-P axis and migrate laterally to generate the horizontal myoseptum, which is continuous with the slow muscle monolayer surrounding the lateral surface of the myotome, and divide the fast muscle myotome into its dorsal and ventral compartments.

In the Mesp quadruple mutant, differentiation into MPs, which were double positive for Prox1 and Engrailed, occurred normally, as observed in the wild-type embryo (Fig. 6A-C). However, in the Mesp quadruple mutant embryo, MPs remained close to the notochord, resulting in interruption of slow muscle fibers at the medial surface in the trunk (Fig. 6D,E). We also observed that the expression of cxcl12a (sdf1a), a chemokine gene, at the superficial myoseptum had completely disappeared (Fig. 6F,G). These results indicated that Mesp function was essential for the correct distribution of MPs and their differentiation into a cxcl12a-positive superficial horizontal myoseptum.

In addition to the lack of the horizontal myoseptum, the number of myogenic progenitor cells that expressed pax7 was decreased in the medial surface of somites and formation of fast muscle was abnormal in the quadruple mutant (Fig. 6H,I). These defects were similar to those shown by morpholino-mediated disruption of mespba and mespbb function (Windner et al., 2015). This complexity of defects in myotome development raised the question of which defect had the greatest contribution. To address this question, we performed a transplantation experiment in which the behavior of wild-type cells was monitored in a Mesp-deficient environment (Fig. 6J-L). In embryos whose phenotype was restored, as judged from cxcl12a expression in the medial surface of the trunk region, cxcl12a expression was restricted only in the wild-type donor cells, suggesting that the Mesp genes in adaxial cells, which are the progenitors of the cxcl12a-positive horizontal myoseptum, were required for proper formation of the horizontal myoseptum.

Expression of cxcl12a in the superficial horizontal myoseptum was reported to have important roles in the pattern formation of various tissues and cells in the body surface, including migration of the posterior lateral line (pLL) primordium and distribution of pigment cells at the body surface (David et al., 2002; Svetic et al., 2007). In the wild-type embryo, neuromasts, which are repetitively deposited from the pLL primordium during its migration, were aligned along the A-P body axis in the trunk and tail (Fig. 7A). However, in the Mesp quadruple mutant embryo, few neuromasts were deposited in the anterior trunk region, suggesting that loss of cxcl12a expression in the medial surface of somites resulted in mis-migration of the pLL primordium (Fig. 7B). Disrupted expression of cxcl12a in the superficial horizontal myoseptum was reported to affect neural crest migration along the lateral pathway in meox1 (choker) mutants, resulting in a disturbed pattern of pigment cells (Svetic et al., 2007). Thus, expression of this chemokine ligand in the medial surface of somites is required for proper alignment of pigment cells in medial body surface region. Similarly, the alignment of melanocytes and iridophores was severely disrupted in the Mesp quadruple mutant embryo (Fig. 7C-F). This phenotype suggests that formation of the superficial horizontal myoseptum, which is regulated by Mesp genes, is required for proper pigment pattern formation in later developmental stages.

In mouse somitogenesis, Mesp2 functions as the master regulator to connect the spatiotemporal information, generated by the segmentation clock and determination wavefront, to the multiple processes of somite morphogenesis (Saga, 2012a). However, most of the functions of Mesp genes in zebrafish somitogenesis had remained unclear until now. In this study, we addressed this issue by generating Mesp mutants using TALEN-mediated mutagenesis. Although all Mesp single mutants showed no obvious defect throughout embryogenesis including somitogenesis, the Mesp quadruple mutant, which lacks the function of all four Mesp genes, exhibited defects in R-C polarity formation of its somites. The physical boundaries between somites were generated even in the Mesp quadruple mutant, although their formation was not completely normal, especially with respect to several anterior somites. In contrast, positioning of somite boundaries occurred normally even in the Mesp quadruple mutant. Thus, unlike in the mouse, zebrafish Mesp genes do not necessarily act as a master regulator to direct all of the three processes involved in somite morphogenesis; and the function of Mesp can be divided into conserved and species-specific functions.

The temporal periodicity generated by the segmentation clock is converted into a spatial pattern as the somite boundary to achieve the sequential and periodical generation of somites. Determination of the somite boundary position is the first step of this process. Previous studies revealed that the position of the somite boundary is defined as the anterior edge of the Tbx6 protein-expressing domain, which cyclically regresses posteriorly in a step-wise manner associated with the periodicity generated by the segmentation clock (Oginuma et al., 2008; Wanglar et al., 2014). In mouse somitogenesis, posterior regression of the Tbx6 domain is severely disrupted in the Mesp2 mutant or Ripply1 and Ripply2 double-mutant embryos. A number of recent studies strongly suggest that Mesp2-dependent generation of anterior border of the Tbx6 protein domain is mediated by Ripply1 and Ripply2 in mouse somitogenesis (Takahashi et al., 2010; Zhao et al., 2015). For instance, expression of Ripply1 and Ripply2 is lost in Mesp2-deficient mouse embryos. Furthermore, the anterior border of the Tbx6 protein domain is properly established even in knock-in mice in which the Mesp2 gene is completely replaced with Ripply2 (Takahashi et al., 2010).

Although Ripply-dependent somite boundary positioning is conserved in zebrafish, the roles of Mesp in this process have remained unclear. Our genetic analysis revealed that Mesp genes but not Ripply genes, are dispensable for somite boundary positioning along the entire body axis in zebrafish. Moreover, we demonstrated that the effect of the Mesp mutation on ripply1 expression is limited to the mature somite region and that the expression of ripply1 and ripply2 in the anterior PSM is regulated in a manner that is independent of the function of Mesp genes. Thus, the anterior border of the Tbx6 protein domain, which is consistent with the position of the presumptive somite boundary, is determined by ripply1 and ripply2 expression in the anterior PSM without the function of Mesp genes in the somites.

In contrast to most posterior somites, several anterior somites showed no clear boundaries in the Mesp quadruple mutant embryos. Of note, ripply1 and ripply2 expression in the anterior PSM and the formation of the anterior border of the Tbx6 protein domain were almost normal, even in the early somite stage. These results indicate that Mesp gene function is dispensable for the positioning of these anterior somites. Thus, the disrupted boundaries between several anterior somites observed in the Mesp quadruple mutant is not caused by impaired positioning of the somite boundaries.

In somitogenesis, physical boundaries between somites are formed at the position defined by the anterior edge of the Tbx6 protein domain (Oginuma et al., 2008). The morphology of segmentation boundaries was partially disrupted in the Mesp quadruple mutant, even though the boundary positioning occurred normally. These results indicate that Mesp function is required for the proper formation of morphological boundaries in zebrafish. Consistent with this conclusion, recent analysis of mouse knock-in embryos in which the Mesp2 gene was replaced with Ripply2 showed that the function of Mesp2 is indispensable for generation of morphological boundaries (Zhao et al., 2015).

During morphogenesis of the segmentation boundaries, cells located at the somite boundary started to epithelialize after the initial establishment of the segment boundary. This initial boundary is considered to be generated by interaction between Eph and Ephrin, because transplantation studies using chick and zebrafish demonstrated that ectopic expression of Epha4 is sufficient to generate segmental boundary-like structures in the unsegmented paraxial mesoderm (Barrios et al., 2003; Watanabe et al., 2009). Furthermore, the interaction between Epha4 and Ephrinb2a in zebrafish is sufficient for the initial generation of the segment boundary, but insufficient for the epithelialization, suggesting that additional molecular machinery is at work in the epithelialization (Barrios et al., 2003). Rather, the epithelialization further required the accumulation of the extracellular matrix component Fibronectin, mediated by Integrin alpha 5, which is activated by Epha4 (Jülich et al., 2015). In the Mesp quadruple mutant, the striped expression pattern of epha4a was lost, suggesting that aberrant boundary formation due to reduced expression of Epha4 causes a disruption in the accumulation of Fibronectin. However, the possibility that Mesp differentially regulates the boundary formation and epithelialization cannot be excluded.

In the Mesp quadruple mutant, the phenotype of somite boundary formation was mild in the posterior somites, suggesting involvement of a Mesp-independent molecular mechanism in this process. pcdh8 is one of the candidate molecules involved in this Mesp-independent somite boundary formation, because its striped expression pattern at the anterior PSM was maintained in the Mesp quadruple mutant. At present, there is scant genetic evidence for the molecular mechanisms involved in cell border formation in vertebrate somitogenesis. Further studies using convenient reverse genetics approaches such as TALEN/CRISPR should clarify this issue.

Previous studies with mice revealed that inhibition of Notch signaling by Mesp2, which is localized in the rostral compartment, is required for the establishment of the rostral identity within the somite (Morimoto et al., 2005; Saga, 2012a). Similar to the mouse, we demonstrated that Mesp function was essential and sufficient to establish the rostral identity of somite even in the zebrafish. The molecular mechanism for the Mesp-dependent establishment of rostral identity remains unclear in zebrafish, because there is no genetic evidence supporting the involvement of Notch signaling in the establishment of the caudal identity, which is known to be the case in mouse somitogenesis.

Notably, the somite boundary, even though it was incomplete, was formed in the posterior body in the Mesp quadruple mutant, indicating that R-C polarity formation and somite boundary formation are distinguishable events in zebrafish. Similarly, a Mesp hypomorphic mouse embryo can form somite boundaries without clear R-C polarity, but the relationship between the two events is not completely identical to that in zebrafish (Nomura-Kitabayashi et al., 2002). In a hypomorphic mouse embryo, somite boundaries gradually disappear after their initial establishment, suggesting that proper R-C polarization is required for the maintenance of somite boundaries. In contrast, our study showed that morphological boundaries were maintained in zebrafish quadruple mutant embryos. Therefore, it seems plausible that not only establishment but also maintenance of somite boundary occur without R-C polarity in the zebrafish.

As previously shown by a knockdown study using morpholinos specific for mespba and mespbb (Windner et al., 2015), suppression of myogenic differentiation and induction of the dermomyotome in the rostral compartment of somites was defective in the Mesp quadruple mutant. In addition to the roles of Mesp genes in fast muscle fiber formation, our study also revealed essential roles of Mesp for the formation of the horizontal myoseptum, which consists of MPs, the progenitor cells of which are derived from the rostral compartment of adaxial cells (Nguyen-Chi et al., 2012). In the caudal compartment of adaxial cells, differentiation of MPs is repressed by Fgf signaling, which is specifically activated in the caudal compartment of somites. This process is possibly regulated by the R-C polarity in a manner that is independent of Mesp function, because the differentiation of MPs was not affected by the Mesp mutation. By contrast, the function of Mesp was essential for the lateral migration of MPs to generate the superficial horizontal myoseptum, which expresses cxcl12a. Thus, two different types of R-C polarity, Mesp independent and Mesp dependent, are likely to regulate specification and migration of MPs in the developing somite. The lateral migration of slow-twitch muscle was reported to be regulated by chemokine signaling and M- and N-cadherin (Chong et al., 2007; Cortés et al., 2003). A cell-autonomous function of Mesp possibly regulates the response to environmental cues provided by surrounding cells.

During vertebrate somitogenesis, the temporal information generated by the segmentation clock is converted to the spatial structure of the somites in the anterior PSM. Since in mouse somitogenesis, Mesp2 expression is activated by the cooperative function of Tbx6 and Notch signaling, whose activity oscillates in the posterior PSM as the core component of the segmentation clock itself, Mesp2 can be considered to act as the master key regulator to connect the segmentation clock to the spatial patterning and morphogenesis of somites (Yasuhiko et al., 2006; Saga, 2012a). However, in zebrafish, Özbuduk and Lewis previously reported that the Notch signaling is not essential for boundary formation of somites (Özbudak and Lewis, 2008). Thus, mouse and zebrafish appear to employ different molecular mechanism for this process, but the molecular mechanism to convert the segmentation clock to the somite boundary remains unclear in zebrafish.

In this study using the zebrafish, we demonstrated that the function of Mesp is not essential for Ripply-dependent boundary positioning, which is the earliest event to connect the segmentation clock to boundary formation (Fig. 8). This result indicates that, unlike in the mouse, neither Notch nor Mesp was essential for the clock-dependent positioning of the somite boundary in zebrafish. Thus, diversity of the molecular mechanism exists not only in terms of the segmentation clock but also with respect to the integration of the clock to the spatial structure in vertebrate somitogenesis. Our results suggest that the segmentation clock in the zebrafish regulates expression of ripply1 and ripply2 in a manner that is independent of Mesp function, although the molecular mechanism still remains unclear.

Zebrafish with the TL2 background were used as the wild type, as described previously (Kishimoto et al., 2004). Collected embryos were grown at 28.5°C or 23.5°C and their developmental stages were determined according to morphological criteria (Kimmel et al., 1995). This study was performed in accordance with the Guidelines for Animal Experimentation of National Institutes of Natural Sciences, with approval of the Institutional Animal Care and Use Committee (IACAC) of the National Institutes of Natural Sciences.

In situ hybridization was performed as described previously (Jowett and Yan, 1996). For synthesis of the mRNA probes, we amplified the fragments of cxcl12a, smyhc, s100t and pax7 using the primers described in Table S3. These fragments were then cloned into the pBS-SK+ vector to use as a template for the RNA probe synthesis.

Capped mRNA, synthesized by use of a message mega-machine kit (Ambion) and plasmid or BAC DNA, purified using PureLink Midiprep Kit (Invitrogen), were dissolved in 0.2 M KCl with 0.05% Phenol Red and injected into one-cell stage zebrafish eggs using an IM300 microinjector (Narishige).

The transplantation experiment was performed as described previously (Kawamura et al., 2005). Cells isolated from host embryos that had been injected with 1% Rhodamine-dextran (Molecular Probes) and 1% biotin-dextran were transferred to the margin of Mesp quadruple mutant embryos at the blastula stage. The host cells were identified by using Rhodamine Red X-NeutraAvidin (Molecular Probes).

For immunostaining, zebrafish embryos were fixed with 4% PFA/PBS at 4°C overnight.

After dechorionation, embryos were dehydrated by immersion in methanol at −20°C for more than 30 min. They were then hydrated by incubation in PBS containing 0.1% Triton X-100, 2% BSA and 1% DMSO for 1 h at room temperature. Then, embryos were incubated overnight at 4°C with primary antibody diluted 1:1000, 1:10, 1:2000 and 1:1000 for Tbx6 (Wanglar et al., 2014), 4D9 (DSHB), GFP (ThermoFisher, G10362) and Prox1 (AngiBio, 11-002P), respectively. After washing three times with PBS containing 0.1% Triton X-100 and 1% DMSO, the embryos were incubated with Alexa Fluor 488- or 555-conjugated secondary antibodies diluted 1:800. GFP signal was amplified using the TSA Plus fluorescence system (PerkinElmer). Double staining for Fn1 (1:100; Sigma, F3648) and Phalloidin (1:100; Molecular Probes) was performed without DMSO.

The TALEN to target each gene was designed using TAL Effector Nucleotide Targeter 2.0 (https://tale-nt.cac.cornell.edu/node/add/talen-old; Table S1), assembled by using a modified Golden Gate Assembly Kit (Sakuma et al., 2013), and cloned into the pCS2+TAL3LL, pCS2+TAL3DD or pCS2+TAL3RR (Dahlem et al., 2012). One nanogram of each capped synthesized mRNA was injected into one-cell stage eggs of TL2. The mutation efficiency was assessed by performing a T7 endonuclease assay (Chen et al., 2012). The primers used to amplify the DNA fragment around the target site are described in Table S2. To obtain mespaa; mespba or mespab; mespbb double-mutant heterozygous fish, we performed sequential mutagenesis. First, mespba or mespab heterozygous F1 mutant fish were established by injection of TALEN mRNA and screening by direct sequencing of the mutated site in the F1 fish. Then, we injected TALEN into the eggs obtained by crossing F1 heterozygous fish and wild-type TL2 fish. The mosaic F2 fish were crossed to TL2 fish to obtain double heterozygous F3 fish, which were crossed to the wild-type fish to confirm that the two mutations were located in the same allele of the chromosome.

Transgenic fish were generated and genotyped using standard techniques using primers as detailed in the supplementary Materials and Methods.

We thank Dr Koichi Kawakami, for providing materials and NIBB Core Research Facilities for their technical support, as well as all members of S.T.'s lab for their helpful discussions.