An anterior limit of FGF/Erk signal activity marks the earliest future somite boundary in zebrafish

Proper formation of a well-proportioned body in multicellular organisms requires spatiotemporal control of multiple biological processes during development. For instance, the regularity of repetitive structures in vertebrate bodies is derived from the vertebrate segments called somites, which are generated by periodic segmentation of the uniform presomitic mesoderm (PSM) (Pourquié, 2001). The periodicity of segmentation is regulated by oscillation of the segmentation clock genes, and the position of segmentation is determined by a gradient of fibroblast growth factor (FGF) (Dubrulle et al., 2001; Dubrulle and Pourquié, 2004; Holley, 2007; Pourquié, 2001; Sawada et al., 2001).

The first detectable sign of a future segment boundary, as visualized by the expression of mesoderm posterior homolog-b (mesp-b; mespba - Zebrafish Information Network), appears at a distance of two somite lengths posterior to B0 (B-2, between S-I and S-II; see Fig. 1D for nomenclature) in the anterior PSM in zebrafish (Sawada et al., 2000), although this may not be the earliest event in boundary determination. A brief exposure of zebrafish embryos to heat shock disrupts somite segmentation after four cycles of normal segmentation (Roy et al., 1999), suggesting that the boundaries of at least five somites (from B-1 to B-5, between S-IV and S-V) are predetermined somewhere in the uniform PSM. However, no molecular sign marking the position of the future somite boundary at B-5 has yet been identified. By precise measurement of the FGF active domain and transient manipulation of FGF signal activity in zebrafish embryos, we show in this study that the anterior limit of FGF/Erk activity corresponds to the future somite boundary at B-5.

In chick, an Fgf8 gradient, generated by restricted transcription and mRNA decay of Fgf8, is essential for setting up the position of the segmentation (Dubrulle and Pourquié, 2004). In zebrafish, segmentation defects can be seen in ace/ace embryos (fgf8a mutants) (supplementary material Fig. S1) (see also Reifers et al., 1998), suggesting that zebrafish fgf8a has a similar role as chick Fgf8 in the context of somitogenesis. However, it has been reported that at least three fgfs (fgf4, fgf8a and fgf24) are expressed in the tailbud region (data retrieved from the Zebrafish Information Network, http://zfin.org/, 4 April 2013), and that fgf8a and fgf24 are together required to generate posterior mesoderm (Draper et al., 2003). We thus observed their expression in the tailbud of zebrafish embryos at a particular somite stage and tested whether the fgf gradient is seen in zebrafish embryos. Although expression of fgf4 and fgf24 was detected in the midline structures of the tailbud, a clear gradient of neither fgf4 nor fgf24 was observed in the PSM (supplementary material Fig. S2). By contrast, fgf8a mRNA was most abundant in a region of the posterior PSM between Kupffer’s vesicle and the posterior end of the tailbud, at the 5-somite stage, and declined gradually towards the anterior (Fig. 1A). To determine how this fgf8a gradient moves during somite segmentation, we measured levels of fgf8a expression in each embryo and then compared the position of the anterior limit of the fgf8a gradient among several embryos, which are arranged in an order of time progression by both the somite number and PSM length as a time indicator (for details, see Materials and Methods). Although PSM length varied between embryos at the same 5-somite stage, the length of the fgf8a gradient between the anterior border and the posterior end remained constant (Fig. 2A,C; supplementary material Fig. S10). When a pair of new somites formed, meaning that the embryo had reached the 6-somite stage, the PSM shortened immediately by about one somite length, whereas the length of the fgf8a gradient did not change at the transition between the 5- and 6-somite stages or indeed throughout the 6-somite stage (Fig. 2A,C; supplementary material Fig. S10). These results suggest that the fgf8a gradient moves continuously towards the posterior as the tail elongates.

Fgf8 binds to and activates Fgf receptors (Fgfrs), which then transduces the FGF signal to the Ras/Erk pathway, leading to the transcriptional activation of target genes including positive and negative feedback regulators (Dorey and Amaya, 2010; Tsang and Dawid, 2004). The Ras/Erk pathway in Xenopus has the mechanistic property of generating a non-linear switch-like response, in which individual cells exhibit either ‘ON’ or ‘OFF’ status (Ferrell and Machleder, 1998; Xiong and Ferrell, 2003). In chick embryos, however, graded activation of the Erk pathway is seen in the PSM (Delfini et al., 2005), whereas in mouse embryos Erk activity displays a different dynamic state, namely an oscillatory pattern within the PSM (Niwa et al., 2007; Niwa et al., 2011). These findings suggest that the Erk pathway operates with species specificity in the PSM. To test the dynamic state of Erk activity in zebrafish PSM, we monitored the active (phosphorylated) form of Erk (p-Erk) by immunostaining using an anti-p-Erk antibody. Although the expression of fgf8a displayed a single peak gradient in zebrafish PSM (Fig. 1A,C), the Erk activity pattern did not (Fig. 1B,C); rather, Erk was highly activated in PSM cells located at rostral and caudal sites of Kupffer’s vesicle, showing a bimodal pattern (Fig. 1B; supplementary material Fig. S3A). In the posterior PSM cells, p-Erk signal intensity in a particular PSM cell was similar to that in its neighbours (supplementary material Fig. S3D), whereas p-Erk was not detected at all in the anterior PSM cells (supplementary material Fig. S3B). Importantly, an obvious difference in the p-Erk signal intensity could be seen at the middle of the PSM (supplementary material Fig. S3C). These results suggest that, in zebrafish PSM, the mechanistic property of the Ras/Erk pathway generates either an ‘ON’ (activated) or an ‘OFF’ (inactivated) status for Erk activity, as seen in Xenopus (Ferrell and Machleder, 1998; Xiong and Ferrell, 2003), and thus a boundary of Erk activity is generated within the uniform PSM.

In most developmental processes, a morphogen gradient is converted into a sharp response boundary within a uniform tissue to form proper patterns. We thus compared the fgf8a and p-Erk gradients in different embryos and found that the p-Erk gradient was much steeper than the fgf8a gradient (Fig. 1C). This result is further confirmed by comparisons of the fgf8a and p-Erk gradients in the same embryos (supplementary material Fig. S4A-D). These results suggest that the gentle slope of fgf8a is converted into the steep gradient of Erk activity, leading to generation of the Erk activity boundary in the uniform PSM. In agreement with this notion, we found that the anterior limit of Erk activity displayed stepwise changes within the PSM that correlated with the spatial periodicity of somite patterning (Fig. 2B,D; supplementary material Fig. S4E,F and Figs S11, S12). Within a particular somite stage, the ON region of Erk activity elongated gradually owing to the tail elongation, whereas the length of the OFF region (distance between B-1 and the anterior limit of Erk activity; see also Fig. 1B) remained constant (Fig. 2B,D; supplementary material Fig. S4E,F). Interestingly, when a pair of new somites formed, the ON region shortened immediately by about one somite length so that the length of the OFF region was precisely maintained (Fig. 2B,D; supplementary material Fig. S4E,F). Furthermore, double staining of p-Erk and her1 expression confirmed that the stepwise changes in the anterior limit of Erk activity were consistent with both somite number and phase propagation of her1 oscillation (supplementary material Fig. S5). These results therefore suggest that the anterior limit of Erk activity displays stepwise movements that correlate with somite segmentation and corresponds to the future somite boundary in the uniform PSM.

Expression of mesp-b appears in the rostral parts of S-I and S-II in the anterior PSM, and is the molecular sign of the future somite boundary B-2 between S-I and S-II (Holley, 2007; Sawada et al., 2000). Double staining for p-Erk and mesp-b expression revealed that the anterior limit of Erk activity was not concomitant with the position at S-II where mesp-b is expressed (supplementary material Fig. S6). The anterior limit of Erk activity was located at a considerably more posterior site in the PSM (supplementary material Fig. S6), suggesting the possibility that the anterior limit of Erk activity marks a future somite boundary, such as B-5 between S-IV and S-V.

If this is the case, misplacement of the Erk activity boundary towards the anterior would result in an anterior shift of B-2 future somite boundary position, marked by mesp-b expression, leading to decreased size of the resultant somite. To test this possibility, we applied a strategy similar to that of the FGF signal perturbations reported by Sawada et al. (Sawada et al., 2001); we enhanced Erk activity for 4 minutes in embryos at the 2-somite stage using the dual-specificity phosphatase inhibitor BCI (Molina et al., 2009) (Fig. 3A). p-Erk signals were intensified immediately after BCI treatment (2-somite stage), resulting in an anterior shift (∼14 μm) of the Erk activity boundary that is estimated by the difference of the OFF region length of p-Erk between embryos treated with DMSO (vehicle) and BCI (Fig. 3B,C; Table 1). However, the anterior limit of Erk activity returned to its normal position by the 3-somite stage (one round of segmentation after BCI treatment) (Fig. 3C). These results indicate that the Erk activity boundary transiently shifted towards the anterior immediately after BCI treatment.

If the Erk activity boundary represents the B-5 future somite boundary, the transient anterior shift of this boundary may alter mesp-b expression position towards the anterior only when three rounds of segmentation (5-somite stage) had occurred after BCI treatment. To test this possibility, we investigated the distribution of mesp-b expression in BCI-treated embryos. Although the anterior limit of Erk activity was shifted just after BCI treatment (Fig. 3C), mesp-b expression occurred normally at this time point (Fig. 3D). Only when three rounds of segmentation had occurred after BCI treatment (5-somite stage) was the width of mesp-b stripes shortened (∼13 μm) (Fig. 3D,E; Table 1). Consistently, this manipulation eventually led to decreased size (∼13 μm) of the resultant 7th somite (Fig. 3F,G; Table 1). Similar results could be obtained when 4- or 10-somite stage embryos were treated with BCI (Table 1). We also investigated whether mesp-b knockdown affected the distribution of the anterior limit of Erk activity. Although mesp-b knockdown resulted in mild disruption of the somite boundary, this manipulation did not affect the stepwise movement of the anterior limit of Erk activity seen in cycles of somite segmentation (supplementary material Fig. S7). These results therefore suggest that the anterior limit of Erk activity is the future somite boundary at B-5; this is an earlier sign of the future somite boundary than that of mesp-b expression at B-2.

Time-lapse imaging followed by p-Erk staining revealed that the position of the anterior limit of Erk activity was maintained in several cycles of somite segmentation (supplementary material Fig. S8A). We thus reasoned that targeted cell fluorescence labelling and time-lapse imaging could confirm whether this position represents the future somite boundary at B-5. To test this possibility, we photo-converted cells at the presumptive p-Erk boundary in Tg[her1:KikGR] embryos and traced the trajectory of the cells (supplementary material Fig. S8B and Movie 1). Although the alignment of the photo-converted cells (supplementary material Fig. S8B, red cells) changed during somitogenesis, a proportion of the cells finally resided at the rostral end of the somite when five rounds of segmentation had completed, meaning that the position adjacent to the photo-converted cells (supplementary material Fig. S8B, red arrowhead) corresponds to the future somite boundary at B-5 (see also Fig. 1D).

Because the repetitive structure of somites is generated from the temporal periodicity created by the somite segmentation clock (Oates et al., 2012; Pourquié, 2011), we hypothesized that proper functioning of the segmentation clock is required for the stepwise movement of the future somite boundary. To test this possibility, we disrupted the segmentation clock in zebrafish embryos by knocking down two segmentation clock genes, her1 and her7, and investigated the dynamics of the anterior limit of Erk activity in the manipulated embryos. As shown previously (Henry et al., 2002), double knockdown of her1 and her7 resulted in the loss of metameric structures of somites (supplementary material Fig. S9B). Erk activation occurred normally even in the manipulated embryos (supplementary material Fig. S9C,D). However, the anterior limit of Erk activity lost its stepwise pattern in each cycle of somite segmentation (Fig. 4A,B; supplementary material Fig. S13). These results therefore suggest that the segmentation clock is required for the stepwise movement of the anterior limit of Erk activity in the PSM (Fig. 4C).

In a clock and wavefront model, the clock controls when the somite boundaries form and the wavefront determines where they form (Cooke and Zeeman, 1976). It has been proposed that, in zebrafish, Her1 homodimer- and Her7:Hes6 heterodimer-based negative feedback loops constitute the clock, and FGF gradients constitute the wavefront (Holley, 2007; Pourquié, 2001; Schröter et al., 2012; Trofka et al., 2012). In this study, we provide additional information for the mechanism underlying zebrafish somitogenesis. Because the single peak fgf8a gradient moved continuously towards the posterior (Fig. 2A,C; supplementary material Fig. S4E,F), the fgf8a gradient acts as a steadily regressing wavefront. As the fgf8a gradient is converted into the Erk activity boundary (Fig. 1; supplementary material Figs S3, S4), which corresponds to the future B-5 somite boundary (Fig. 3; supplementary material Fig. S8B and Movie 1), and because the clock regulates the stepwise movements of the future somite boundary (Fig. 4; supplementary material Fig. S9), the anterior limit of Erk activity represents a ‘signal integration spot’ of the clock and the wavefront.

In mouse, the clock (Notch oscillators such as Hes7 and Lfng) controls oscillation of Erk activity, and Erk oscillation periodically allows Notch signals to induce Mesp2 expression at S-I, suggesting that Erk oscillation represents a rippled wavefront (Niwa et al., 2011). These findings suggest that, although the detailed mechanisms are different, there is a common logic of somitogenesis in zebrafish and mouse: Notch oscillators periodically modulate the FGF/Erk wavefront at the signal integration spot to regulate spatiotemporal periodicity. In chick, although Erk activity declines gradually towards the anterior (Delfini et al., 2005), whether the gradient of Erk activity corresponds to a future somite boundary and whether the gradient shows an oscillatory pattern remain unknown. Therefore, it would be of great interest to understand a common and/or specific logic of somitogenesis in vertebrates by comparing regulatory mechanisms of somite formation in such different species.

Pattern formation is one of the most important processes of vertebrate morphogenesis. In the case of somites, pre-patterning over a distance of about five somites is thought to be determined within apparently uniform tissue composed of somite progenitor cells. Using somitogenesis as a model to investigate how patterns are generated during development, we reveal that a morphogen gradient is converted into a sharp response boundary of intracellular signalling, providing positional information as a future segment boundary, and that clock-dependent stepwise movement of this sharp response boundary is required to generate the perfect metameric pattern of the vertebrate body during development.

Wild-type and ace^ti282a (fgf8a mutant) zebrafish were used in this study. As described previously (Matsui et al., 2011), substitution from G to A in the ace allele was detected by sequence analyses of the DNA fragment obtained from each embryo after in situ hybridization or immunohistochemistry.

Whole-mount in situ hybridization and immunohistochemistry were performed as described previously (Matsui et al., 2005; Matsui et al., 2011). cDNA fragments of fgf8a, mesp-b and her1 were used as templates for antisense probes.

The following antisense MO oligonucleotides against her1, her7 and mesp-b and a control MO were obtained from Gene Tools: control-MO: 5′-CCTCTTACCTCAGTTACAATTTATA-3′; her1-MO: 5′-TTCGACTTGCCATTTTTGGAGTAAC-3′ (Henry et al., 2002); her7-MO: 5′-CAGTCTGTGCCAGGATTTTCATTGC-3′ (Henry et al., 2002); mesp-b-MO: 5′-TCGGTTCTTGCTTGAGGTTTGCATG-3′ (Lee et al., 2009). her1-MO plus her7-MO (6.25 ng each), mesp-b-MO (5 ng) or control-MO (12.5 ng or 5 ng) were injected into the yolk of one-cell-stage zebrafish embryos as described previously (Matsui et al., 2011). Injected embryos were collected and fixed at the indicated time points, and then used for experiments including in situ hybridization and immunohistochemistry.

Embryos were obtained by mating wild-type fish and were cultured normally until 10.5-11 hours post-fertilization (hpf). Embryos at the 2-somite stage were treated with 0.1 mg/ml BCI (Dual Specificity Protein Phosphatase 1/6 Inhibitor, Calbiochem) for 4 minutes, and then washed extensively with embryo medium. Embryos were fixed just after treatment (2-somite stage), or after one (3-somite stage), three (5-somite stage) or eight (10-somite stage) rounds of segmentation following BCI treatment.

The embryos were flat-mounted in Vectashield mounting medium, which contains propidium iodide (PI) (Vector Laboratories). Immunofluorescence signals were visualized and photographed using an LSM510-META or LSM710 confocal microscope (Zeiss). Sequential confocal images were stacked. The position of the somite boundary and posterior end of the PSM was estimated by visual observation of PI staining in combination with bright-field microscopy. Embryos were grouped according to somite number, and were arranged in order of PSM length, because the PSM gradually elongates along with tail elongation. Signal intensity was measured by using ImageJ software (National Institutes of Health), and background was subtracted to remove any tissue autofluorescence from the signal. The position of the anterior limit of fgf expression and the Erk activity was determined by the comparison between visual observation and signal intensity plot.

The coefficient of variation (C.V.) was calculated by the ratio of the s.d. to the mean. Differences in variances and means were analysed by one-tailed F test of equality of variance and two-tailed Welch’s t-test, respectively. Results in the F test and t-test were considered significant when 2P<0.05 and P<0.05, respectively.

In order to label PSM cells with a photo-convertible fluorescent protein named KikGR, we generated a transgenic line Tg[her1:KikGR] using the Tol2 system (Kawakami et al., 2004; Urasaki et al., 2006). Tg[her1:KikGR] carries the zebrafish her1 promoter, KikGR full length cDNA (MBL), nuclear localization signal sequence and SV40 poly A sequence. A 405 nm laser was used in photo-conversion of KikGR. Time-lapse image acquisition was performed by using Olympus FV-1000-D confocal microscope and FLUOVIEW software.

We are grateful to Shinji Takada for helpful discussions and critical reading of the manuscript. We also thank Ian Smith, Wong Kah Loon and Fiqri Dizar Khaidizar for help in preparing the manuscript; Maiko Yokouchi, Hiroko Shigesato, Tatsuro Matta, Hiroshi Okamoto and Ryosuke Tatsumi for technical assistance.