Jawed vertebrates (gnathostomes) have paired appendages (fins or limbs) with species-specific sizes and positions (Supp. Fig. S1, which is available online). Recent studies have shed much light on the developmental mechanisms of limbs and fins, but much remains to be learned about the mechanisms that define body proportions, including the positions and size of limbs, in gnathostomes. Despite the importance of the Hox code for regional specification of the trunk in gnathostome vertebrates (Burke et al.,1995; Rancourt et al.,1995; Cohn et al.,1997; Cohn and Tickle,1999; Minguillon et al.,2005), it is still not known how the Hox code is translated into body proportions in the embryo. Although several genes, including Hox genes, affect the size of the limb bud when misexpressed, most cause changes in the limb-bud size only along the proximal–distal axis. At present, Tbx18 is the only known factor that can alter the size of the limb field along the anterior–posterior axis (Tanaka and Tickle,2004).
In the chick embryo, specific regions of the lateral plate mesoderm (LPM) give rise to the limb bud mesenchyme, which in turn induces the overlying ectoderm to form the apical ectodermal ridge (AER) at the dorsoventral boundary of the limb bud, by secreting fibroblast growth factor 10 (FGF10; Ohuchi et al.,1997; Xu et al.,1998; Yonei-Tamura et al.,1999). The induced AER secretes FGF8 and FGF4 to establish a positive feedback loop with the expression of FGF10 from the underlying mesenchyme and/or SHH from the zone of polarizing activity (ZPA). This loop maintains mesenchymal cell proliferation to promote limb protrusion (e.g., Saunders,1948; Rubin and Saunders,1972; Saunders et al.,1976; Laufer et al.,1994; Niswander et al.,1994; Capdevila and Izpisua Belmonte,2001; Fernandez-Teran and Ros,2008). These mechanisms are considered to be shared among gnathostome vertebrates. The flank region between the forelimb and hindlimb is competent to form additional limbs in response to FGFs (Cohn et al.,1995; Crossley et al.,1996; Yonei-Tamura et al.,1999 [chick]; Tanaka et al.,2000 [mouse]; Abe et al.,2007 [zebrafish]; Yonei-Tamura et al.,2008 [cartilaginous fish] and references therein). However, despite their potential to form additional limbs in the flank, a gnathostome species with a third set of paired limbs has never emerged. Therefore, some mechanism must specify the size of the limb fields within the flank region and prevent additional limb formation. Although no tissue has been directly shown to restrict the development of the limb/flank region, the axial mesoderm and paraxial mesoderm are known to be responsible for limb morphogenesis (Stephens and McNulty,1981; Strecker and Stephens,1983; Stephens et al.,1991; Geduspan and Solursh,1992; Crossley et al.,1996; Dealy,1997) and limb identity (Saito et al.,2002,2006), and accumulating data suggest that the paraxial mesoderm plays a key role in the positioning of the limbs (Burke et al.,1995; for review, see Capdevila and Izpisua Belmonte,2001).
In the present study, we investigated the role of the paraxial mesoderm in limb/flank regional specification in chick embryos. We found that the forelimb-level presomitic mesoderm (PSM) was involved in the emergence and growth of the limb bud, but the flank-level PSM induced apoptosis in the flank LPM. We also found that an additional limb bud induced in the flank prevented apoptosis in the flank LPM. These results suggest that the paraxial mesoderm plays pivotal roles in regulating limb size by specifying the limb/flank fields in the LPM, and in preventing the formation of additional limb buds in the flank LPM.
Forelimb-Level Paraxial Mesoderm Is Involved in Limb Size Determination
To examine the function of the forelimb-level paraxial mesoderm in limb-bud formation, part of the PSM at the forelimb level was ablated from stage 10/11 embryos (Fig. 1A–D). The resultant embryos possessed a smaller limb bud on the PSM-ablated side than on the contralateral side (n = 10/12, Fig. 1A,B′, asterisk indicates the ablated side), and the anterior/posterior length of the fgf8-expressing AER was shortened (Fig. 1B′). To exclude the possibility that we had simultaneously ablated some intermediate mesoderm, we examined the expression of Sim1, an intermediate mesoderm/nephric duct marker (Obara-Ishihara et al.,1999; Mauch et al.,2000) in the operated embryos. Sim1 expression was not affected by the microsurgery (Fig. 1C,D), indicating that the intermediate mesoderm was intact. Whereas PSM ablation resulted in a reduced limb size at stage 10/11, no defect was observed in embryos with somite ablation at the corresponding axial level at stage 12 (n = 7/7, data not shown), suggesting that the PSM is required for budding of the forelimb field before stage 12.
Flank-Level Paraxial Mesoderm Restricts the Position of Limb Bud Formation
To investigate the role of the flank-level PSM in limb-field specification, we ablated the PSM corresponding to the level extending from the posterior end of the forelimb to the flank (called flank-level hereafter, Fig. 2). The simple ablation of the flank-level PSM did not affect the anterior–posterior range (around five somites) of the forelimb (Fig. 2B–D). The AER expands from the middle–posterior region to the anterior end, and the fgf8-expressing AER on the manipulated side reached the anterior end of the limb bud earlier than the AER on the control side (n = 3/3, Fig. 2A). In the later stages, the forelimb on the operated side was bigger along the proximal–distal axis compared with the control side, suggesting that the limb bud had undergone accelerated distal growth compared with the control side (n = 9/12, Fig. 2B–D). The expression domain of fgf10 correspondingly expanded (n = 5/6, Fig. 2C,D) along the proximal–distal axis of the operated side compared with the control side. Although the flank-level ablation included the PSM corresponding to the level of the posterior end of the forelimb, Shh expression was not affected (n = 3/3, Fig. 2B).
Because chick embryos lie with their left side down after stage 14, it is possible that the trypsin solution used to isolate the PSM had a greater effect on the limb bud growth on the left side (the underside) than on the right side. Therefore, we also ablated the PSM on the left side of some embryos (Fig. 2D), and found that, as expected, the forelimb on the operated side was bigger than on the control side (n = 3/3, Fig. 2D). No defects were observed in the embryos when somite ablation was performed at stage 12/13 (n = 14/14, data not shown), suggesting that the end of PSM influence on the limb field correlates with somite formation. These results suggest that the flank-level PSM, but not the later-stage somitic mesoderm, has negative effects on the size of the limb bud in the nearby LPM, or that the LPM loses its responsiveness to the suppressive signal from the flank paraxial mesoderm by stage 13.
Different Functions of the Forelimb-Level PSM and the Flank-Level PSM
We next examined the mesenchymal cell proliferation in the PSM-ablated limb bud compared with that in controls, by immunohistochemical analyses using an antibody specific for the mitotic marker phospho-histone H3. Twenty-four hours after PSM ablation at the anterior forelimb level (see also Fig. 1A), the number of mitotic cells had decreased (Fig. 3A), and the number of phospho-histone H3-positive cells in the limb bud mesenchyme was decreased by around 55% compared with the contralateral control limb buds (Fig. 3B). When the anterior flank-level PSM was ablated (see also Fig. 2A), the number of mitotic cells increased significantly in the manipulated limb bud (Fig. 3C,D). Cell death detected by an antibody for active caspase 3, a marker for early apoptosis, did not increase in the limb bud after ablation of the anterior forelimb PSM (Fig. 3E), suggesting that apoptosis did not contribute to the reduction of limb size after PSM ablation.
To confirm the role of the flank-level PSM, we replaced the forelimb-level PSM with flank-level PSM (Fig. 4). The forelimb bud on the operated side was smaller along the anterior–posterior axis than the control (left) forelimb bud (Fig. 4AB). After 48 hr, the manipulated forelimb bud was much shorter along the proximodistal axis than the left forelimb bud, as indicated by the auxiliary lines shown in Figure 4B. Thus, both the width and the length of the limb bud decreased on the operated side (n = 6/7, Fig. 4B). No defects were found when the forelimb level PSM was replaced with PSM taken from the same axial level of another embryo (n = 3/3, Fig. 4C).
We further examined the function of the forelimb level PSM in limb bud formation. We replaced the anterior flank-level PSM with forelimb level PSM (Fig. 5). The limb bud was wider on the operated side than on the control side 24 hr after the operation (n = 8/11, Fig. 5A, A′). The fgf8-expressing AER expanded in the posterior direction only on the operated side (Fig. 5A′,B′), and the anterior–posterior size was greater, but only by as much as one somite. After 48 hr, the manipulated limb bud was still bigger than the control (Fig. 5C,C′). The forelimb level PSM thus seems to be able to expand the limb field even when transplanted into the anterior flank. Taken together with the data in Figures 2 and 4, showing that the flank-level PSM has the ability to reduce the limb bud size, these data suggest that the flank-level PSM and forelimb-level PSM have distinct functions in limb bud formation.
Apoptosis in the Flank LPM
From the onset of limb bud formation, active apoptosis was transiently detected in the prospective flank LPM (Fig. 5, see also Zuzarte-Luis et al.,2007). Whole-mount TUNEL (terminal deoxynucleotidyl transferase–mediated deoxyuridinetriphosphate nick end-labeling) staining revealed massive apoptosis in the flank LPM at stages 17–19 (Fig. 6B–E), but not at stage 16 or stage 21 (Fig. 6A, and data not shown). Strong anti-active caspase 3 immunoreactivity was observed in the flank LPM at stage 17 but not at stage 16 (Fig. 6H,I and not shown). Thus, the onset of flank apoptosis can be traced back to stage 17 at the earliest. Dying cells in the flank were mostly detected, not in the splanchnic layer, but in the somatic layer of the LPM (Fig. 6I,I′; see also Fig. 7B). Because flank-specific apoptosis was also detected in gecko embryos (Fig. 6F,G) and mouse embryos (Fig. 6J,K) at the corresponding stages, transient apoptosis in the flank LPM appears to be a common mechanism among amniote embryos.
This extensive apoptosis could be suppressed by the application of an FGF8-soaked bead to the flank LPM (Fig. 7A,B), which caused an additional limb bud to form in the flank. We then examined the effect of flank-level PSM ablation on the flank LPM apoptosis, and found that the LPM apoptosis significantly decreased in the region neighboring to the ablated PSM (n = 5/5, Fig. 7C,C′). These results indicate that the apoptosis in the flank LPM is induced by the flank PSM.
Effect of FGF Signaling on Apoptosis in the Flank LPM
When an FGF8-soaked bead was transplanted into the flank LPM, a significant reduction of apoptosis was observed (see above). To reveal the requirement of the FGF signal in preventing apoptosis of LPM cells, the FGF inhibitor SU5402 was applied to the limb-forming LPM, and it induced ectopic apoptosis in the limb field (Fig. 8A). Moreover, the limb bud mesenchyme showed active apoptosis after AER removal at stage 19 (Fig. 8B). These results suggest that the FGF-responsive cells underwent apoptosis when they could not receive sufficient FGF. Because fgf8 is known to be expressed throughout the AER, we hypothesized that the flank-level PSM induces apoptosis in the flank LPM, and that the FGF8-responsive cells are removed by apoptosis during flank development (Fig. 8C).
It is notable that FGF7/10, both of which have the ability to induce the AER structure directly in the flank ectoderm (Ohuchi et al.,1997; Yonei-Tamura et al.,1999), do not cause ectopic limb bud formation at the flank region when implanted into stage 13/14 chick embryos. It takes 16–24 hr for FGF7 to induce the AER in the ectoderm (Yonei-Tamura et al.,1999); this may be too late to induce the limb mesenchyme in the flank through the ectopic AER, because the flank LPM has started dying by stage 17/18, as shown in Figure 6. To test this hypothesis and to examine the function of PSM on suppressing limb bud formation in the flank, we ablated the PSM at the flank level and implanted an FGF7-soaked bead; we used FGF7, because it is reported to be more effective for AER induction than FGF10 (Yonei-Tamura et al.,2008) (Fig. 8E). The FGF7-bead implantation induced an ectopic AER with a small protrusion in the flank (n = 7/7, Fig. 8D,D′), which gave rise to an ectopic limb bud (see also Yonei-Tamura et al.,1999). Moreover, an ectopic limb bud with an fgf8-positive AER was induced in the flank when the flank-level PSM was removed simultaneously (n = 3/7, Fig. 8E,E′). These ectopic limb buds would be expected to give rise to an additional limb skeleton in the flank, according to previous reports (Ohuchi et al.,1997; Yonei-Tamura et al.,1999; Tamura et al., 2001). In a typical sample (Fig. 8E), the large ectopic limb bud in the flank was fused with the hindlimb bud, with missing an anterior portion of the AER in the bud. Taking these observations together with the result shown in Figure 7 that the ablation of flank-level PSM suppresses apoptosis in the flank, it is possible that the combination of flank-level PSM ablation and FGF7 application enables FGF8-responsive cells in the flank to survive until the stage of limb bud outgrowth. It is possible that the flank-level PSM might directly promote cell death or to do so by inhibiting FGF10 function.
The paraxial mesoderm has been suspected to be responsible for the regional specification of the trunk, because of its clear Hox gene expression boundaries and correlation between the specific Hox gene expression boundaries and limb positions (Burke et al.,1995). Numerous reports have shown that alterations in Hox gene expression result in anteriorization or posteriorization of the vertebral column and rib cage, and some of these mutants show a shift in limb position (Horan et al.,1995; Rancourt et al.,1995; Cohn et al.,1997; Chen et al.,1998; van den Akker et al.,2001; Wellik and Capecchi,2003; McIntyre et al.,2007; Wellik,2007, and references therein). It is, however, unclear if Hox genes in the paraxial mesoderm are responsible for limb field specification, because they are also expressed in other tissues, including the LPM itself (Burke et al.,1995). Saito et al. (2006) directly manipulated the PSM and showed that the PSM is involved in specification of forelimb and hindlimb, but the PSM's role in regional specification of the trunk was not addressed.
In this study, we investigated the role of the paraxial mesoderm in the regional specification of the trunk. We found that the PSM has different roles in the emergence and growth of the limb bud, depending on its position along the anterior–posterior axis (Fig. 9A). The forelimb-level PSM promotes limb bud growth and expands the AER, even when transplanted at the flank level (Fig. 5). Consistent with this function, the ablation of forelimb-level PSM caused a reduction of cell proliferation (Fig. 3). The flank-level PSM, in contrast, suppressed limb bud growth in the LPM even when transplanted to the forelimb level (Fig. 4), probably by means of the induction of apoptosis. Ablation of the flank-level PSM promoted cell proliferation and resulted in a bigger limb bud (Figs. 2–4). Few apoptotic cells were detected in this limb bud (Fig. 3 and not shown), but the increase in size may have been related to the reduction of apoptosis in the flank (Fig. 7C,C′). In the chick embryo, anterior-to-posterior cell movement is reported to occur in the LPM around the forelimb bud (Wyngaarden et al.,2010). Some population of forelimb cells may flow out into the flank region. Flank-level PSM ablation and the resultant decrease of flank LPM apoptosis might prevent this outflow of the forelimb bud cells, resulting in a bigger limb bud. As mentioned in the Introduction, no gnathostome has developed a third pair of appendages. Because flank-level PSM ablation in combination with FGF7 application resulted in ectopic limb bud formation, apoptosis in the flank LPM may remove cells that are competent to form a limb bud.
In this study, we revealed that the PSM is important for limb field specification, but this does not exclude the possible contribution of other tissues, including an autocrine effect of the LPM. Wnt2b, expressed in the presumptive forelimb field of the LPM before limb bud outgrowth, is sufficient to induce an additional limb in the flank (Kawakami et al.,2001), and this LPM factor cannot be induced by ectopic transplantation of the paraxial mesoderm into the caudal trunk (Saito et al.,2006). Although the upstream regulation of the wnt2b expression in the LPM is unknown, its expression might be regulated independently of the paraxial mesoderm. It is therefore possible that paracrine effects from the PSM and autocrine regulation in the LPM concomitantly and additively specify the limb size, position, and growth.
We detected Sim1 expression after the adjoining PSM had been ablated, indicating that the intermediate mesoderm remained intact, but it is still possible that the intermediate mesoderm participates in limb field specification. The intermediate mesoderm develops into three tissues (pronephros, mesonephros, and metanephros) and differentiates coincidentally with the segmentation of the adjoining paraxial mesoderm (Hiruma and Nakamura,2003). In the chick embryo, the mesonephros is formed posterior to somite 15; this position corresponds to the boundary between the neck and the forelimb bud (Hiruma and Nakamura,2003). The paraxial mesoderm has been shown to be necessary for pronephros formation, and pronephros induction is required for mesonephros development (Mauch et al.,2000). The PSM at the forelimb level might affect the adjoining intermediate mesoderm to induce the limb field in the adjacent LPM. Interestingly, the forelimb-level intermediate mesoderm expresses fgf8 and Wnt2b (Crossley et al.,1996; Kawakami et al.,2001), although the function of the intermediate mesoderm in limb initiation remains controversial (Geduspan and Solursh,1992; Crossley et al.,1996; Fernandez-Teran et al.,1997). Because the paraxial mesoderm and the LPM are separated by the intermediate mesoderm, a secreted factor(s) from the paraxial mesoderm could affect the LPM either directly or indirectly through the intermediate mesoderm. A molecule that might regulate the fate of the LPM is Growth and differentiation factor 11 (Gdf11), a member of the Transforming growth factor beta (Tgf-beta) superfamily that is expressed in the posterior PSM and is a regulator of Hox gene expression (McPherron et al.,1999; Liu et al.,2001). Moreover, Gdf11 knockout mice show posteriorization of the hindlimb bud (McPherron et al.,1999), and the onset of Gdf11 expression in the posterior PSM is slightly earlier than the onset of Tbx5/4 expression in the LPM (Liu et al.,2001; Saito et al.,2002), supporting this gene as a good candidate for the posterior LPM fate regulator. Other members of the Tgf-beta superfamily may be anterior LPM fate regulators.
The timing of the limb field specification in the LPM is also important for understanding the molecular nature of its specification by the PSM. Tbx5, which does not actually have a forelimb-specific expression domain but is also expressed in the anterior flank (Saito et al.,2006), is required for forelimb bud initiation (Ng et al.,2002; Takeuchi et al.,2003; Rallis et al.,2003; Hasson et al.,2007). Tbx5 is not expressed in the LPM before stage 13 in the chick embryo (Saito et al.,2002), and the LPM fate for forelimb-flank-hindlimb regionalization can be altered up to stage 13 (Saito et al.,2002,2006). It is also evident that the positioning of limb buds is specified well before the LPM is split into the somatic and splanchnic mesodermal layers (Yonei-Tamura et al.,2005), which is complete by the 20-somite stage (around stage 13; Funayama et al.,1999). Our results are consistent with the above reports, as the embryos showed no defect when somites were ablated at stage 12 or later (data not shown), suggesting that the inductive effects of the PSM and/or the competence of the LPM may disappear around this stage.
In the present study, we demonstrated that a signal(s) from the PSM is involved in determining the forelimb size, and consequently, trunk regionalization. As recently shown by Keyte and Smith (2010), a wider limb field can give rise to a larger limb. Marsupial neonates have larger forelimbs than do their eutherian counterparts. In opossum embryos, eight somites are associated with the forelimb field, whereas five to six somites are associated with it in mice and other eutherian mammals. Thus, limb size variation can be attributed to differences in the numbers of associated somites and PSM tissue organization in these animals. However, lizards and amphibians show extreme variations in the number of their trunk vertebrae and in the relative size of their limbs. We can estimate the number of somites at each level by observing the spinal nerves in the adult, because the brachial nerves appear to be induced by the limb bud (Ohuchi and Noji,1999; Turney et al.,2003), with one spinal nerve corresponding to one somite. Therefore, the number of spinal nerve roots that enter the limb represents the number of somites associated with the limb field in embryogenesis. In lizards, the number of nerve roots that contribute to the brachial plexus (usually four) shows no correlation with limb size (unpublished results based on observations of Bipes, Gerrhonotus, Chlacides, Paroedura, and Ameiva). Instead, the number of somites associated with the flank in lizards may contribute to the relative proportions of the limbs to the trunk. Indeed, there is a known correlation between limb reduction and trunk elongation in several lizard lineages (e.g., Wiens and Slingluff,2001), and small-limbed lizard species with an elongated trunk tend to have more trunk vertebrae (and, therefore, more trunk somites) than their close relatives (e.g., Presch,1975; Greer,1987; Caputo et al.,1995).
Thus, different proportions of somite numbers in the limb field and flank (i.e., the relative proportions of the PSMs contributing to these fields) may account for the difference between large-limbed and small-limbed species. Although further studies are necessary to test this hypothesis, it will provide a basis for future efforts to elucidate the mechanism through which the diversity of body proportions evolved among gnathostomes.
Surgical Ablation of PSM
Chick embryos were staged according to Hamburger and Hamilton (1992). For the surgery, a drop of 0.15% trypsin/Calcium Magnesium Free Tyrode's saline (Tyrode,1910) was placed on stage 10/11 embryos and allowed to incubate for 5 min. The ectoderm overlaying the paraxial mesoderm was peeled off, and the PSM at the level of the forelimb or from the posterior end of the forelimb to the flank, a region corresponding to 3–5 somites in length, was ablated with a sharpened tungsten needle. The trypsin was then washed off with Tyrode's saline, and the peeled ectoderm was put back in place. After the surgery, the eggs were sealed with Scotch tape and incubated for 10 to 36 hr. The manipulated embryos were then fixed for whole-mount in situ hybridization, immunostaining, or whole-mount TUNEL staining.
Transplantation of PSM
PSM was removed from the forelimb level or from the level between the posterior end of the forelimb and flank of stage 10–11 and 11–12 embryos, respectively, as described above. Grafts consisting of these tissues were then carefully placed on the ablated site of the recipient. After surgery, the eggs were sealed and incubated for 24 to 48 hr for further analyses.
Implantation of Beads
Affigel blue beads or Affi-gel heparin beads (Bio-Rad) were washed at least three times for 30 min with phosphate buffered saline (PBS) and soaked in 0.5 mg/ml FGF7 (Promega)/PBS, FGF8 (Promega)/PBS, or 30 mg/ml SU5402 (Calbiochem), for 1 hr. One soaked bead was then inserted between the LPM and the overlying ectoderm. We used FGF7 rather than FGF10, because it is known to exhibit a similar activity, with a generally stronger effect (Yonei-Tamura et al.,2008).
Whole-Mount In Situ Hybridization
Whole-mount in situ hybridization was performed as described previously (Yonei et al.,1995; Yonei-Tamura et al.,1999). Digoxigenin-labeled RNA probes for fgf8 (Ohuchi et al.,1997), fgf10 (Yonei-Tamura et al.,1999), Shh (Yonei et al.,1995), and Sim1 (Saito et al.,2006) were used for whole-mount in situ hybridization.
Immunohistochemistry and Cell Counting
Specimens were fixed for 2 hr at 4°C in 4% paraformaldehyde, cryoprotected with 30% sucrose, and embedded in O.C.T. compound. Sections were cut at a thickness of 10 μm with a cryostat. The primary antibodies were rabbit anti-phosphorylated Histone H3 antibody (1:500, Upstate) and rabbit anti-active caspase-3 (1:500, BD Biosciences Pharmingen), and the secondary antibody was Alexa Fluor 488 goat anti-rabbit IgG (H+L) (1:500, Invitrogen). Sections were counterstained with DAPI (4′,6-diamidine-2-phenylidole-dihydrochloride) to detect nuclei. Histone H3-positive cells were counted in a certain area (around 700 cells with DAPI signal) of the limb field on a section, and the ratio of the positive cells to total (DAPI-labeled) cells was calculated. The same cell-counting procedure was repeated with three to four sections for each specimen.
Whole-Mount TUNEL Analysis
Chicken embryos or gecko embryos at 2- and 3-day post-oviposition (dpo) (Noro et al.,2009), were fixed in 4% paraformaldehyde/PBS overnight and dehydrated in graded methanol. The embryos were then immersed in 5% H2O2/methanol for 20 min and then transferred into methanol. All the samples were rehydrated in 0.5% Triton X-100/PBS (PBST). Chick and gecko embryos were treated with 5 μg/ml Proteinase K for 30 min and 20 μg/ml Proteinase K for 20 min, respectively. The samples were then treated with 2 mg/ml glycine for 10 min, washed with PBST for 5 min twice, refixed in 4% paraformaldehyde/0.2% glutaraldehyde/PBST, and washed with TdT buffer (Tris-HCl [pH 6.6], 2.4 mg/ml BSA, 0.2% Tween20, 0.2 M sodium cacodylate, 1.5 mM CoCl2) twice for 10 min. An In Situ Cell Death Detection Kit, AP (Roche), was used for the subsequent reactions. The embryos were soaked in a reaction solution (TdTase solution:FITC-dUTP:TdT buffer 1:9:20 vol/vol) at 4°C for 1 hr and incubated at 37°C for 1 hr. The rest of the reactions were conducted according to the supplier's instructions.
The authors thank Dr. Takuya Minokawa for valuable discussions. H.Y. and K.T. were funded by the Ministry of Education, Culture, Sports, Science, and Technology of Japan, K.T. was funded by the Toray Science Foundation and Sankyo Foundation of Life Science, M.N. was funded by the Fujiwara Natural History Foundation of Japan, and T.T. was supported by JSPS Postdoctoral Fellowships and Grant-in-Aid for JSPS Fellows from JSPS.