ISL1/SHH/CXCL12 signaling regulates myogenic cell migration during mouse tongue development

Human congenital tongue malformations that occur during embryogenesis, such as aglossia, microglossia and macroglossia, greatly affect the quality of life of individuals (Cobourne et al., 2019). Tongue development begins with median lingual swelling of the first branchial arch (Kaufman and Bard, 1999). During this stage, the tongue anlage is mainly composed of neural crest cell (NCC)-derived mesenchymal cells. Lateral lingual swellings are then formed on each side of the median tongue bud. Myoblasts from occipital somites migrate along the hypoglossal cord and reach the first arch at about embryonic day (E) 10.5 (Lofqvist and Lindblom, 1994; Parada and Chai, 2015; Parada et al., 2012). Most of the tongue muscle is derived from the occipital myoblasts, whereas the connective tissue and vasculature are derived from NCCs, which originate from the dorsal neural tube (Ziermann et al., 2018). Myoblasts are precisely regulated to migrate to the tongue primordium and proliferate and differentiate, making the tongue a muscular organ. Although epithelial-derived signaling has been found to control tongue morphogenesis, the mechanisms by which epithelial-derived signaling regulates myoblast migration and invasion during tongue development remain unclear (Jeong et al., 2004; Lin et al., 2011).

Multiple signaling pathways and genes have been reported to be involved in the regulation of myoblast migration. Pax3, Gab1 and Met have been shown to be essential for the delamination and migration of myogenic progenitor cells to other sites of myogenesis (Amano et al., 2002; Bober et al., 1994; Tajbakhsh and Buckingham, 2000). Shh, Wnt5a and the CXCL12/CXCR4 axis are also crucial for cell migration and invasion during organogenesis (Jeong et al., 2004; Vasyutina et al., 2005; Yamaguchi et al., 1999). Hedgehog signaling has been shown to be important for the growth and patterning of facial primordium during craniofacial development (Melnick et al., 2005). Wnt5a has been shown to regulate directional cell migration and polarity (He et al., 2008; Schlessinger et al., 2007; Witze et al., 2008). The WNT5A pathway underlies the outgrowth of multiple structures in vertebrate embryos, and Wnt5a deficiency results in abnormal tongue development (Liu et al., 2012; Yamaguchi et al., 1999). The chemokine Cxcl12 has been reported to provide directional cues for the migration of many different cell types, such as germ cells, lymphocytes and muscle progenitors (Belmadani et al., 2005; Doitsidou et al., 2002; Ma et al., 1998; Ratajczak et al., 2003; Zou et al., 1998). Previous reports indicate that CXCL12/CXCR4 regulates the migration of both proliferating and terminally differentiated muscle cells (Griffin et al., 2010). However, it remains unclear how these signaling pathways interact between epithelial cells, mesenchymal cells and myoblasts to coordinate tongue development.

Isl1, a member of the LIM homeodomain transcription factor family, has been proven to play crucial roles during left-right body pattern formation, mandible development, anterior teeth formation and cell-type specification in many developmental processes (Li et al., 2017; Pfaff et al., 1996; Yang et al., 2006). We previously found that Isl1 was exclusively expressed in the distal epithelium of first branchial arch as early as E9.5, and Isl1 ablation (Isl1^ShhCre) resulted in defective mandibular outgrowth (Li et al., 2017). In the present study, we report that the Isl1^ShhCre mice exhibit aglossia due to defects in distal myoblast invasion. We further reveal that the ISL1/SHH/CXCL12 axis regulates distal migration of myoblasts in the first branchial arch during tongue development. Because myoblast migration plays an important role in the development of many organs, the mechanism we identified has important implications for understanding the processes of organogenesis and malformation.

To investigate the potential function of Isl1 in tongue development, we generated mice with an epithelial cell-specific deletion of Isl1 by crossing Shh-Cre with Isl1^fl mice. Scanning electron microscopy (SEM) analysis demonstrated that the tongue did not develop in Isl1^ShhCre embryos (100% penetration) (Fig. 1A, Fig. S1). X-gal staining of Shh^Cre/R26R embryos revealed Cre activity in the mandibular arch epithelium at E9.5, E10.5 and E11.5 (Fig. 1B). Isl1 is expressed in the distal ectoderm of the first branchial arch, but not in the entire mandibular epithelium, as determined by immunofluorescence and the Isl1-lacZ knock-in allele (Fig. 1B). Anti-MHC (myosin heavy chain) staining showed that only a stack of residual muscle cells existed in the tongue-forming region of the Isl1^ShhCre mutant embryos (Fig. 1C). Taste buds were not detected in the Isl1^ShhCre embryos using an antibody against cytokeratin-8 (CK8, Troma-1; also known as keratin 8, KRT8) (Fig. 1D). Together, these results suggest that an intrinsic function of Isl1 is required for the tongue formation.

The absence of the majority of tongue muscle in Isl1^ShhCre mutant embryos suggests that aglossia may be caused by defective myoblast migration into the tongue primordium. It has been well documented that Pax3 and desmin are expressed in early myogenic precursor cells migrating from the somites into the branchial arch around E10 (Mayo et al., 1992; Tajbakhsh and Buckingham, 2000). Immunofluorescence staining revealed that both Pax3- and desmin-labeled myoblasts reached the floor of the first branchial arch, but did not migrate into the tongue primordium of Isl1^ShhCre embryos (Fig. 2A, Fig. S2), indicating that in Isl1^ShhCre mutant mice, myoblasts fail to invade into the tongue mesenchyme after reaching the first branchial arch. The number of Pax3⁺ and desmin⁺ cells in wild type and Isl1^ShhCre was counted at E10.5 and E11.5. The results showed no significant difference in the number of Pax3⁺ and desmin⁺ cells at E10.5 between control and Isl1^ShhCre (Fig. 2B,C). However, the number of Pax3⁺ and desmin⁺ cells at E11.5 was reduced in Isl1^ShhCre compared with wild type (Fig. 2B,C).

We next investigated whether there were differences in myoblast proliferation between Isl1^ShhCre and wild type. Cyclin D1 immunofluorescence showed that the proliferation of Pax3⁺ myoblasts in Isl1^ShhCre did not change significantly at E10.5, but decreased significantly at E11.5 (Fig. 2D,E). Consistent with this, 5-bromodeoxyuridine (BrdU) and Ki67 (Mki67) labeling was also reduced in desmin⁺ and Pax3⁺ myoblasts in Isl1^ShhCre embryos at E11.5 (Fig. 2D,F,G). Terminal deoxynucleotidyl transferase dUTP nick end labeling (TUNEL) analysis revealed no significant changes in apoptosis in Isl1^ShhCre embryos at E11.5 (Fig. 2H). Myoblast differentiation was examined by expression of the myoblast differentiation marker myogenin (Myog) and no apparent differences were detected between wild-type and Isl1^ShhCre embryos (Fig. 2I). These results demonstrate that in Isl1^ShhCre mutant embryos myoblast proliferation is reduced after reaching the first branchial arch, but their differentiation and survival are not affected.

It has been reported that the Hedgehog signaling in NCC-derived mesenchyme may be involved in the transmission of information from epithelial cells to myogenic progenitor cells (Parada and Chai, 2015). We propose that Isl1 may regulate tongue development by affecting Shh expression in epithelial cells. Real-time PCR results showed that the expression of Shh was significantly downregulated in the mandible of Isl1^ShhCre embryos at E10.5 (Fig. 3A). Ptch1, Gli1, Foxd1, Foxd2, Foxf1 and Foxf2 are all reported to be the target genes of Hedgehog signaling in cranial NCCs (Jeong et al., 2004). Quantitative RT-PCR analysis showed that the expression of Ptch1, Gli1, Foxd1, Foxd2, Foxf1 and Foxf2 was significantly decreased in the first branchial arch of Isl1^ShhCre mutants (Fig. 3A, Fig. S3). Whole-mount in situ hybridization results demonstrated that the expression of Shh was much reduced in the distal epithelium of the first branchial arch (Figs 3B and 4). The expression of Ptch1 was also downregulated in the distal region of the mandible of the Isl1^ShhCre mutant (Figs 3B and 4). These results indicate that Isl1 in the distal ectoderm regulates Shh expression in anterior tongue epithelial cells.

Previous studies have shown that the chemokine CXCL12 provides directional cues for migrating CXCR4-expressing muscle progenitors in the limbs (Vasyutina et al., 2005). In the developing first branchial arch, CXCL12 is expressed in the mesenchymal cells, and its receptor CXCR4 is expressed in the muscle progenitors (Vasyutina et al., 2005). The co-expression of CXCR4 and desmin in myoblasts was confirmed by immunofluorescence (Fig. S4). At E10.5, expression of Cxcl12 mRNA but not Cxcr4 mRNA was detected at the base of the tongue primordium, suggesting that Cxcl12-expressing mesenchymal cells populate the tongue anlage prior to myoblast invasion (Fig. 4C-D′). At E11.5, the expression domains of Cxcl12 and Cxcr4 similarly extended into the middle region of the tongue primordia in wild-type embryos (Fig. 4G,H). In contrast, there was no colocalization of Cxcl12 and Cxcr4 expression in Isl1^ShhCre (Fig. 4G′,H′), suggesting that Cxcr4⁺ myoblasts did not reach the Cxcl12 expression domain. In addition, the expression region of Cxcl12 in the first branchial arch of the Isl1^ShhCre mutant did not extend as far, and its expression in the Isl1^ShhCre mutant was much higher than that in the wild type (Fig. 4C,C′,G,G′). Cxcl12 expression was only detected in the proximal mesenchyme below the Shh-expressing epithelium, suggesting that Shh expression in the mandibular epithelium may determine the region of Cxcl12 expression (Fig. 4A-C′,E-G′). Thus, we hypothesize that aberrant expression patterns of Shh and Cxcl12 in the first branchial arch lead to the migration defect of Cxcr4⁺ myoblasts.

To determine whether Hedgehog signaling regulates myoblast migration, we generated mice with an NCC-specific deletion of Smo by crossing Wnt1-Cre with Smo^fl mice. Similar to the phenotype of Isl1^ShhCre, the tongue was absent in Smo^Wnt1Cre embryos as a result of myoblast migration defects (100% penetration) (Fig. 5A,B). There was no expression of Ptch1 in the first branchial arch and adjacent pharyngeal arch of Smo^Wnt1Cre mice, indicating that the activity of the Hedgehog pathway is blocked (Fig. 5C). In situ hybridization revealed that Cxcl12 expression was absent in the first branchial arch at E10.5 and E11.5, but the expression of Cxcl12 in the adjacent pharyngeal arch was not affected (Fig. 5D), indicating that only the expression of Cxcl12 in the first branchial arch was dependent on activation of the canonical Hedgehog pathway in NCCs. The results of immunofluorescence and in situ hybridization showed that desmin⁺, Pax3⁺ and Cxcr4⁺ myoblasts did not migrate into the tongue anlage in Smo^Wnt1Cre embryos (Fig. 5B,D), suggesting that activation of canonical Hedgehog pathway in NCCs is required for Cxcl12 expression in the first branchial arch and for myoblast invasion. There was no significant difference in the number of Cxcl12⁺ mesenchymal cells between wild type and Isl1^ShhCre at E9.5 and E10.5 (Fig. S5A-D). Our real-time PCR results also showed no significant difference in the expression of Cxcl12 mRNA between Isl1^ShhCre and wild type in the first branchial arch at E10.5 (Fig. S5E). We further investigated the effect of Hedgehog signaling on expression of Cxcl12 at the transcriptional level. The 2 kb promoter of Cxcl12 was cloned into the pGL3 vector (pGL3-CXCL12 Pro). Purmorphamine (PUR), an agonist of the Smo receptor, had no effect on Cxcl12 mRNA expression and Cxcl12 promoter activity (Fig. 5E,F). Therefore, we hypothesize that Shh regulates the expression pattern of Cxcl12 by affecting the migration of Cxcl12-expressing mesenchymal cells in the first branchial arch.

Considering that Wnt5a affects cell migration and can alter outgrowth of many structures, including the tongue (He et al., 2008; Liu et al., 2012; Schlessinger et al., 2007; Witze et al., 2008; Yamaguchi et al., 1999), it is possible that the reduction of Shh expression limits the migration of Cxcl12⁺ mesenchymal cells through Wnt5a. Consistent with this idea, Wnt5a expression was detected in the proximal domain below the Shh-expressing epithelium in Isl1^ShhCre embryos (Fig. 6A). The expression of Wnt5a was also significantly reduced in Smo^Wnt1Cre embryos (Fig. 6B). More importantly, Wnt5a expression was strongly induced by SHH-saturated beads (Fig. 6C), or purified recombinant SHH protein in mandible explant culture (Fig. 6D). These results suggest that expression of Wnt5a in the tongue is regulated by Shh. In addition, WNT5A-saturated beads were able to attract Cxcl12⁺ cells to migrate from the adjacent pharyngeal arch in wild-type embryos (Fig. 6F). Defects in Cxcl12⁺ cell migration could also be rescued by WNT5A beads in Isl1^ShhCre embryos (Fig. 6G). Moreover, the expression of Cxcl12 was not altered after WNT5A treatment (Fig. 6E). Overall, these results suggest that the migration of Cxcl12⁺ mesenchymal cells is regulated by the SHH/WNT5A cascade, and the abnormal migration and accumulation of Cxcl12⁺ mesenchymal cells in Isl1^ShhCre is due to changes in the expression patterns of Shh and Wnt5a.

Mandible explant cultures were used to determine whether the chemotaxis of myoblasts changes with the concentration of CXCL12. Low concentration (100 ng/μl) of CXCL12 protein-saturated beads obviously attracted desmin⁺ myoblasts to the protein beads (Fig. 7A). By contrast, desmin⁺ myoblasts aggregated into circular bands at a certain distance from beads saturated with high concentration (500 ng/μl) of CXCL12 protein (Fig. 7A), indicating that a high concentration of CXCL12 protein can attract myoblasts to aggregate nearby, but prevent myoblasts from approaching. CXCL12 protein diffuses from the beads to the surroundings, forming a concentration gradient from high to low. When CXCL12 reaches a certain concentration in the circular band, the attraction and repulsion effect of CXCL12 on myoblasts reaches a balance.

We hypothesized that the migration of Cxcl12⁺ cells is crucial for establishing a concentration gradient of CXCL12, which attracts myoblast migration during tongue development. CXCL12 beads were implanted in Isl1^ShhCre mandibles to investigate whether myoblasts can be repelled or attracted into the tongue primordium despite the lack of Cxcl12⁺ cell migration (Fig. 7B). The results showed that high (500 ng/μl) and low (100 ng/μl) concentrations of CXCL12 protein-saturated beads repelled or attracted desmin⁺ myoblasts migration, respectively (Fig. 7B). This suggests that myoblast migration is regulated by CXCL12 protein concentration and that the migration of Cxcl12⁺ mesenchymal cells secreting CXCL12 protein is crucial for myoblast migration.

Primary cultures of a mixture of mesenchymal cells and myoblasts isolated from the E11 tongue tissue of wild-type embryos were used to evaluate the chemotaxis of CXCL12. The inserts were coated with Matrigel matrix to mimic the in vivo environment. When CXCL12 was added to the lower chamber, the chemotactic attraction activities of CXCL12 on desmin⁺ myoblast cells were observed with a maximum effect at 10 ng/ml (Fig. 7C). At this concentration, the number of desmin⁺ cells that transmigrated to the lower chamber was more than three times that of control. The movement of desmin⁺ myoblasts towards CXCL12 dramatically decreased when the CXCL12 concentration was increased to 1000 ng/ml (Fig. 7C). The number of desmin⁺ cells that transmigrated was less than half that of the control, suggesting that desmin⁺ myoblasts would move away from CXCL12. These results confirmed that CXCL12 attracts Cxcr4⁺ myoblasts at low concentrations but repels them at high concentrations.

A competitive chemotaxis assay was also used to confirm the bidirectional effect of CXCL12 on myoblast invasion. When CXCL12 was added to the upper chamber, the movement of desmin⁺ myoblasts away from CXCL12 reached a maximum at 1000 ng/ml, with about twice as many desmin⁺ cells transmigrating than in the control (Fig. 7D). Conversely, when the CXCL12 concentration was decreased to 10 ng/ml the number of desmin⁺ cells that transmigrated was less than half of the number transmigrating in control. These data illustrate a concentration-dependent bidirectional movement of myoblasts in response to CXCL12 (Fig. 7D). The direction of myoblast invasion depends on the concentration of CXCL12, suggesting that CXCL12 functions as an attractant for Cxcr4⁺ myoblasts in normal developing tongue tissue, but abnormally high levels of CXCL12 in the tongue primordia of Isl1^ShhCre mutants prevent myoblast invasion.

Treatment with the selective CXCR4 inhibitor AMD3100 resulted in inhibition of myoblast movement in response to CXCL12 in both directions (away and towards) (Fig. 7E). It has been reported that CXCL12 can induce activation of tyrosine kinase and inhibition of cAMP (Chalasani et al., 2003; Ganju et al., 1998; Xu et al., 2010). Pre-incubation of myoblasts with the tyrosine kinase inhibitor genistein only inhibited movement towards CXCL12, but had no effect on movement away from CXCL12. In contrast, the membrane-permeable cAMP agonist 8-Br-cAMP blocked movement away from CXCL12, but had no effect on movement towards the chemokine (Fig. 7E). These data indicate that the attraction and repulsion of myoblasts by CXCL12 are mediated by distinct signaling pathways.

To test further the hypothesis that Isl1 regulates tongue development through Hedgehog signaling, we next examined whether Ihh could rescue the defective tongue morphogenesis in Isl1^ShhCre. We took a transgenic gain-of function approach by overexpressing Ihh in the mandibular epithelium using a conditional Ihh transgenic allele (Tg-pmes-Ihh), as Ihh and Shh could activate the same canonical Hedgehog signaling pathway (Jenkins, 2009). As expected, transgenic Ihh did indeed alleviate the defect of aglossia, and the Isl1^ShhCre; Tg-pmes-Ihh allele displayed microglossia (Fig. 8A,B). The invasion of myoblasts (Pax3⁺ and desmin⁺) was rescued after expression of Ihh (Fig. 8C, Fig. S6) as was the proliferation of myoblasts (Fig. S6). Moreover, Ptch1 expression was induced in the mesenchyme of the tongue, indicating activation of Hedgehog signaling (Fig. 8D). The colocalization of Cxcl12⁺ and Cxcr4⁺ cells was restored by expression of transgenic Ihh (Fig. 8D). The expression of Wnt5a was also restored in Isl1^ShhCre; Tg-pmes-Ihh embryos (Fig. 8D). Although expression of Ihh was insufficient to fully rescue the tongue development defect, the genetic integration of Hedgehog signaling was evident. The signaling cascade model controlling myoblasts invasion during tongue morphogenesis is illustrated in Fig. 9.

Although the tongue is an extremely important organ in mammals, the molecular mechanisms of tongue morphogenesis remain poorly understood owing to its complex anatomy and cell origins (Noden and Francis-West, 2006). In this study, we demonstrate that the transcription factor Isl1 is required for tongue development, as deletion of Isl1 in the epithelium of the first branchial arch results in defects in myoblast invasion.

Tongue myogenic progenitors originate from the occipital somites and migrate along hypoglossal cord to the tongue primordium at about E10.5 (Mackenzie et al., 1998). Both desmin and Pax3 mark the early myogenic precursor cells that migrate from the somite to the branchial arch (Mayo et al., 1992; Tajbakhsh and Buckingham, 2000). In this study, desmin⁺ and Pax3⁺ myogenic cells were observed in the floor of first branchial arch at E10.5 in both wild-type and Isl1^ShhCre embryos. However, the myogenic cells did not invade into the tongue primordium after they arrived the first branchial arch of Isl1^ShhCre embryos. It has been reported that NCCs initiate tongue development and populate the tongue primordium prior to the invasion of myogenic progenitors (Chai and Maxson, 2006). Therefore, we hypothesized that ablation of Isl1 in epithelial cells results in a microenvironment in the tongue primordium that is unconducive to muscle cell invasion.

The chemokine receptor CXCR4 and its ligand CXCL12 provide guidance for many different cell types, such as lymphocytes, germ cells, hematopoietic cells and muscle progenitor cells (Belmadani et al., 2005; Doitsidou et al., 2002; Ma et al., 1998; Ratajczak et al., 2003; Zou et al., 1998). All of these different cell types are simultaneously and accurately guided by CXCL12 to different targets, suggesting tight control over the spatial and temporal distribution of CXCL12 (Lewellis and Knaut, 2012). CXCL12 is expressed in the mesenchyme of limb bud and the first branchial arch, and its receptor CXCR4 is expressed in migrating muscle progenitors (Vasyutina et al., 2005). However, it has been reported that CXCR4-positive muscle progenitor cells did not migrate to the center of limb buds with high CXCL12 expression, but instead localized to the surrounding area (Vasyutina et al., 2005). This strongly suggests that the expression level of CXCL12 may regulate the distribution of muscle progenitors. Here, we found abnormally high Cxcl12 mRNA expression in the mandibular arch of Isl1^ShhCre embryos. We propose that this higher CXCL12 expression in the presumptive tongue primordium leads to defects in myoblast invasion. Consistent with this concept, our chemotaxis assays demonstrate that CXCL12 has a concentration-dependent bi-functional effect on myoblasts invasion, attracting Cxcr4⁺ myoblast invasion at low concentrations but repelling at high concentrations. The invasion and distribution of Cxcr4⁺ myoblasts during tongue and limb development may be precisely regulated by the intensity of CXCL12 expression. This is the first study to find that CXCL12 has a bi-functional effect on myoblast invasion. Consistent with our observations, a previous study reported a similar bi-functional effect of CXCL12 in T-cell migration (Poznansky et al., 2000).

How might Isl1 in the distal ectoderm alter the expression pattern of Cxcl12 in the mesenchyme? Hedgehog signaling plays an important role in coordinating reciprocal epithelial-mesenchymal interactions during organogenesis. The expression domain of Shh in the epithelium seems to determine where the tongue develops (Jeong et al., 2004). Here, we demonstrate that Shh is regulated by Isl1 in oral ectoderm-derived lingual epithelium. We propose that downregulation of Shh leads to defects in myoblast migration during tongue development. A strong correlation between Shh and Cxcl12 expression patterns was observed. The loss of Cxcl12 expression in the first branchial arch of Smo^Wnt1Cre embryos confirms the crucial role of the Hedgehog pathway for Cxcl12 expression. However, activation of Hedgehog signaling had no effect on Cxcl12 mRNA expression and Cxcl12 promoter activity. Therefore, we propose that Shh may affect the expression pattern of Cxcl12 by regulating the migration of Cxcl12⁺ mesenchymal cells in the first branchial arch. Our data indicate that the expression of Wnt5a in the tongue is regulated by Shh, which is consistent with reports of hair follicle morphogenesis (Reddy et al., 2001). Several lines of evidence suggest that Wnt5a may mediate SHH signaling in the regulation of Cxcl12⁺ mesenchymal cell migration. First, it is well documented that Wnt5a plays a crucial role in regulating cell migration and polarity (Yamaguchi et al., 1999). Second, expression of Wnt5a is lost in the tongue of Isl1^ShhCre embryos. Conversely, activation of SHH signaling in tissue explants can induce Wnt5a expression. Third, WNT5A attracts Cxcl12⁺ mesenchymal cells to migrate from the adjacent pharyngeal arch. Moreover, the attenuated defect in tongue morphogenesis and adequate restoration of myoblast invasion by expressing transgenic Ihh further strongly supports that Hedgehog signaling is required for tongue development. Based on these findings, we propose that Isl1 regulates Cxcl12⁺ mesenchymal cell migration through the SHH/WNT5A signaling cascade. Canonical Hedgehog signaling activation in NCCs is essential for the migration of Cxcl12⁺ mesenchymal cells and myoblasts. In Isl1^ShhCre mutants, WNT5A in the proximal mesenchyme below Shh-expressing epithelium attracts Cxcl12⁺ mesenchymal cells to cluster in the relatively small proximal region of the mandibular arch, resulting in a higher concentration of CXCL12 in this domain. The distal migration of Cxcl12⁺ mesenchymal cells is essential for establishing a concentration gradient of CXCL12, which attracts myoblast invasion during tongue development.

Binding of CXCL12 to its receptor, CXCR4, triggers multiple intracellular signaling pathways related to chemotaxis (Chalasani et al., 2003; Ganju et al., 1998; Xu et al., 2010). Both tyrosine kinase and cyclic nucleotides have been reported to be essential for cell polarization and migration (Daaka et al., 1997; Guo and Cheng, 2015; Malherbe and Wang, 2012; Song et al., 1998). Our data suggest that the attractive effect of CXCL12 on myoblasts is mediated by the activity of tyrosine kinase, whereas cAMP is responsible for the repulsive effect of CXCL12 on myoblasts. It is possible that different concentrations of CXCL12 activate the distinct signal transduction pathways downstream of its receptor CXCR4, resulting in a ligand concentration-dependent bi-functional effect on myoblast invasion. In addition to regulation of migration processes, the CXCL12/CXCR4 signaling pathway promotes the growth and survival of different cells types (Molyneaux et al., 2003; Rubin et al., 2003; Sutton et al., 2007). Consistent with this, myoblast proliferation defects in myoblasts were detected in Isl1^ShhCre embryos at E11.5. The reduced cell proliferation could be due to the lack of interaction between Cxcl12⁺ mesenchymal cells and Cxcr4⁺ myoblasts in Isl1^ShhCre embryos. However, the differentiation process of the myoblasts was unaffected, suggesting that interaction between these two cells is required for myoblast proliferation rather than differentiation, which is consistent with previous findings that myoblast differentiation is not affected after CXCR4 inhibitor treatment (Griffin et al., 2010).

To the best of our knowledge, this is the first study to report that the ISL1/SHH/CXCL12 axis regulates Cxcr4⁺ myoblast migration during tongue development. The concentration-dependent bidirectional regulation of CXCL12 on Cxcr4⁺ myoblast migration and invasion may also be significant for the development of other organs, such as limbs. Our findings will promote the understanding of the mechanism of myoblast migration and open perspectives for studies of the pathogenesis of human tongue deformities and related syndromes.

All animal procedures were approved by Hangzhou Normal University Animal Care and Use Committee (HNLA-2012-018). Construction of conditionally targeted Isl1^fl mice, Isl1^lacZ and Tg-pmes-Ihh conditional transgenic mice has been described previously (Elshatory et al., 2007; Li et al., 2017). Shh^{tm1(EGFP/cre)}, Wnt1-Cre, Smo^fl/fl and R26R-lacZ were purchased from The Jackson Laboratory and maintained on a C57BL/6J background. The morning of observed vaginal plug was designated as day 0 (E0) of pregnancy.

Embryos were fixed in 2.5% glutaraldehyde in 0.1 M phosphate buffer (pH 7.4) overnight. After rinsing in PBS, the sections were post-fixed with 1% osmium tetroxide solution for 1.5 h. Samples were dehydrated using a series of acetone solutions: 30%, 50%, 70%, 80%, 90%, 100% (three times) for 15 min each. Samples were critically point-dried after a 15 min treatment with 95% isoamyl acetate. Samples were then mounted on conductive paper and sputter-coated with gold. Images were recorded with a scanning electron microscope (Hitachi S-3000N) with a 15-kV accelerating voltage.

X-gal staining of whole-mount samples was performed as described previously (Li et al., 2017). Embryos were fixed in 4% paraformaldehyde (PFA) solution (with 5 mM EGTA and 2 mM MgCl₂) at 4°C for 1 h. The fixed embryos were rinsed three times in washing buffer (2 mM MgCL₂, 0.01% sodium deoxycholate and 0.02% NP-40 in PBS). The embryos were then incubated in staining solution (0.1% X-gal, 5 mM potassium ferricyanide, 5 mM potassium ferrocyanide, 2 mM Tris, pH 7.3, in washing buffer) overnight in the dark at 37°C. Lastly, the stained embryos were washed in PBS and used for whole-mount analysis. The X-gal-stained sections were counterstained with Nuclear Fast Red.

Total RNA from mouse mandibles was extracted from organ culture explants using TRIzol Reagent (Thermo Fisher Scientific). All RNA samples were treated with DNA-free DNA Removal Kit (Thermo Fisher Scientific). The first-strand complementary DNA (cDNA) was then synthesized from 1 μg of total RNA from each sample using a PrimeScript 1st Strand cDNA synthesis kit (Takara Bio). Quantitative RT-PCR was performed using the following primer sequences: Isl1, 5′-ATGATGGTGGTTTACAGGCTAAC and 5′-TCGATGCTACTTCACTGCCAG; Shh, 5′-AAAGCTGACCCCTTTAGCCTA and 5′-TGAGTTCCTTAAATCGTTCGGAG; Ptch1, 5′-AGACTACCCGAATATCCAGCACC and 5′-CCAGTCACTGTCAAATGCATCC; Gli1, 5′-CCAAGCCAACTTTATCAGGG and 5′-AGCCCGCTTCTTTGTTAATTTGA; Foxd1, 5′-AAAATCGCCCTATGCTGC and 5′-CTGGACCTGAGAATCTCTACACC; Foxd2, 5′-ATTTATGAAGAGTCTCCAGACC and 5′-GATGCTCAAACAGAAAAGC; Foxf1, 5′-ACTCCAGTGTCTTTCACCTTGC and 5′-TGAGCCTGAACTACACCAGC; Foxf2, 5′-TCAGTAGGACATTTCTTCC and 5′-CTGTCACAATACTGAGAGC; Wnt5a, 5′-CTAACAAGTGTGACAAGATCC and 5′-CTTGAAAGCAATGTCTAGC; Cxcl12, 5′-GACAGAAGAAGAGAAAGGCTGC and 5′-CCGAAGAGGGAAGAGTTTACC; 18S, 5′-TAGAGGGACAAGTGGCGTTC and 5′-CGCTGAGCCAGTCAGTGT. The specificity for each primer set was confirmed by electrophoresis and sequencing before use. Real-time PCR experiments were performed in triplicate using SsoFast EvaGreen Supermix as previously described (Li et al., 2017). The relative amount of each gene transcript was calculated using the 2^−ΔΔCT method and normalized to the endogenous reference gene 18S (Rn18s). Data were analyzed with CFX Manager software and are represented as mean±s.e.m.

The collected embryos were fixed in 4% PFA solution, dehydrated, embedded in paraffin and sectioned as previously described (Li et al., 2019). Heat-induced epitope retrieval was used to unmask antigen epitopes (Li et al., 2019). For the cultured cells, 4% PFA solution was used for 10 min. After blocking in 1% bovine serum albumin (BSA), the sections were incubated with antibodies against MHC (1:20; Developmental Studies Hybridoma Bank), desmin (1:200; ab32362, Abcam), cyclin D1 (1:100; ab16663, Abcam), Pax3 (1:200; ab15717, Abcam), BrdU (1:100; ab6326, Abcam), CXCR4 (1:200; ab1670, Abcam) or Ki67 (1:200; ab15580, Abcam) at 4°C overnight. Then, horseradish peroxidase-conjugated secondary antibody against the primary antibody (1:200, sc-2004, Santa Cruz Biotechnology) was applied for 1 h at room temperature. For immunofluorescence, secondary antibodies conjugated with Alexa Fluor 488 or 594 (1:1000, Invitrogen) were used. Images were analyzed using a Leica DM4 B microscope equipped with a digital camera.

Whole-mount and section in situ hybridization were performed as previously described (Li et al., 2017). For section in situ hybridization, samples were fixed in 4% PFA solution, dehydrated, embedded in paraffin and sectioned at 12 μm. For whole-mount in situ hybridization, samples were fixed in freshly made 4% PFA/PBS, dehydrated into methanol and bleached with 6% hydrogen peroxide (H₂O₂). Non-radioactive antisense RNA probes were generated by in vitro transcription using the DIG RNA labeling kit (Roche Diagnostics). For histological analysis, sections were stained with Hematoxylin and Eosin according to standard protocols.

Cell proliferation activity was evaluated by BrdU labeling and immunofluorescence staining. Timed pregnant mice were sacrificed 30 min after a single intraperitoneal injection of BrdU (Sigma-Aldrich) at 3 mg/100 g body weight. The collected embryos were fixed in 10% neutral buffered formalin at 4°C overnight and embedded in paraffin. BrdU-labeled cells were detected immunohistochemically on paraffin sections according to the manufacturer's instructions. The cell growth ratio (%) was calculated by dividing the number of BrdU-positive nuclei by the total number of nuclei.

The chemotactic response of mandible myoblasts to CXCL12 was evaluated using a 24-well fitted Transwell inserts with membranes (8 μm pore size, Corning Falcon). The inserts were then coated with Corning Matrigel matrix (Fisher Scientific) at 200 μg/ml. Mouse tongues were carefully dissected from E11 wild-type embryos. To remove the epithelium, the explants were incubated in 2.25% trypsin/0.75% pancreatin on ice for 5 min. One hundred microliters of isolated primary cells (mixture of myoblasts and mesenchymal cells) were added to the upper well of the chamber (2×10⁵ per well). Both upper and lower chambers were loaded with Dulbecco's modified Eagle medium (DMEM) containing 1% fetal bovine serum (FBS). CXCL12 (R&D Systems) was used at concentrations of 0 ng/ml, 10 ng/ml, 100 ng/ml and 1000 ng/ml in the lower chambers of the Transwell. For the inhibitor treatment, cells were pretreated with 0.5 μg/ml AMD3100 (30 min at 37°C; A5602, Sigma-Aldrich), 1 μg/ml genistein (20 min at 37°C; 345834, Sigma-Aldrich) or 100 μM 8-Br-cAMP (15 min at room temperature; B7880, Sigma-Aldrich) before being added to the upper well of the chamber in selected experiments. For the competitive chemotaxis assay, CXCL12 (0, 10, 100 and 1000 ng/ml) was mixed with isolated primary cells (1×10⁵ per well) plated to the upper chamber in media containing 1% (v/v) FBS. Media containing 3% (v/v) FBS was added to the lower chamber. The cells on the upper surface of the membrane were removed using a cotton swab after treatment for 24 h. The membranes were cut from inserts and mounted onto glass slides for immunofluorescence staining. The desmin antibody was used to label myoblasts. All experiments were repeated three to six times, and the number of invaded desmin⁺ myoblasts was calculated.

Mouse tongues were carefully dissected from E11 wild-type embryos. The explants were incubated in 2.25% trypsin/0.75% pancreatin on ice for 5 min to remove the epithelium. One hundred microliters of isolated primary mesenchymal cells were cultured in 24- or 96-well plates. The 2 kb promoter of Cxcl12 was cloned into the pGL3 basic vector (pGL3-CXCL12 Pro) for the luciferase assay. The clone primer sequences were: 5′-CTAGCTCATCGTAGGAGGACC and 5′-CAGAGCTGGACAGCAAGAGG. The pGL3 empty vector (control) or pGL3-CXCL12 Pro vector was transfected with Lipofectamine3000 reagent. The next day, PUR (1 μM) or DMSO (control) was added to the culture media to activate the canonical Hedgehog signaling pathway. After incubation for 24 h in a humidified atmosphere of 5% CO₂ at 37°C, the cells were used for RNA isolation or luciferase activity analysis. Firefly luciferase activities were normalized to Renilla luciferase activity and each experiment was performed in triplicate at least three times.

For bead-implantation experiments, mouse mandibles were carefully dissected from E10.5 wild-type embryos and placed into DMEM (Gibco) supplemented with 10% FCS. The explants were placed on a Nuclepore Track-Etch membrane (0.2-μm pore size) in Trowel-type organ culture dishes. WNT5A-saturated agarose beads (0.5 μg/μl), CXCL12-saturated agarose beads (0.1 and 0.5 μg/μl) or SHH-saturated agarose beads (1 μg/μl) were grafted in the mandible. After being cultured for 16 h in a humidified atmosphere of 5% CO₂ at 37°C, the explants were fixed in 4% PFA overnight at 4°C, washed in PBS and finally stored in methanol at −20°C until analysis by whole-mount in situ hybridization. WNT5A, CXCL12 and SHH were purchased from R&D Systems. For Isl1 overexpression experiments, the Isl1 overexpression vector (pCDH-Isl1) or control vector (pCDH) was transfected with Lipofectamine3000 reagent. After being cultured for 48 h in a humidified atmosphere of 5% CO₂ at 37°C, the explants were used for RNA extraction and real-time PCR analysis.

All data are presented as mean±s.e.m. Unpaired, two-tailed Student's t-test was used to compare data sets. The threshold for statistical significance was P<0.05.

We thank Dr Lin Gan for providing Isl1^fl and Isl1^lacZ mice, and Dr Zunyi Zhang for comments and suggestions. We acknowledge the Center of Electron Microscopy of Zhejiang University for access to the scanning electron microscopes.

The peer review history is available online at https://journals.biologists.com/dev/lookup/doi/10.1242/dev.200788.reviewer-comments.pdf.