The Dlx5-FGF10 signaling cascade controls cranial neural crest and myoblast interaction during oropharyngeal patterning and development

The oropharyngeal region serves as a passage for air, water and food, and also helps to produce articulate speech, which is unique to humans. Consequently, defects in the soft palate adversely affect speech, swallowing, breathing and hearing (Precious and Delaire, 1993). The soft palate forms the posterior one-third of the palate and consists of five muscles in humans: the tensor veli palatini (TVP), levator veli palatini (LVP), palatopharyngeus (PLP), palatoglossus (PLG) and muscle of the uvula (Evans et al., 2010). Cleft of the soft palate leads to misorientation of these muscle fibers and compromises palatal function in the oropharyngeal region. Despite various surgical approaches to reorient the muscles of the soft palate in order to restore its proper function, the outcome is typically less than ideal, owing to insufficient muscle tissue and poor ability to regenerate following the operation (Carvajal Monroy et al., 2012). The lack of a more optimal approach to palatal muscle restoration is partly due to a lack of understanding of the molecular regulatory and cellular mechanisms of soft palate development.

Our recent study has shown that the anatomical structures of the soft palate and cellular contributions to them in mice are similar to those in humans, which renders the mouse an excellent model with which to study soft palate development (Grimaldi et al., 2015). Specifically, cranial neural crest (CNC)-derived cells migrate into the primordium of the soft palate and form a scaffold to guide the migration of myogenic progenitors into the soft palate, where they form the TVP, LVP, PLP and PLG (Grimaldi et al., 2015). The progenitors of these muscles in the soft palate and those in the pharynx/larynx derive from the pharyngeal mesoderm (Michailovici et al., 2015). In comparison with trunk muscle development, the molecular regulatory mechanisms underlying head muscle development are less well characterized. It is clear, however, that head muscle development is governed by both intrinsic and extrinsic programs. For example, Tbx1, Myf5, and Lhx2 form an important intrinsic genetic network to control the fate of myogenic progenitors during head muscle development (Michailovici et al., 2015). As one of the extrinsic regulators, CNC-derived cells control tongue muscle development through CNC-myogenic progenitor interactions (Han et al., 2012; Hosokawa et al., 2010).

Dlx genes play important roles in regulating craniofacial patterning and development. In mammals, six Dlx genes are expressed as convergently transcribed gene pairs, including Dlx1/Dlx2, Dlx5/Dlx6 and Dlx3/Dlx7. These gene pairs have overlapping regulatory systems and expression patterns (Robledo and Lufkin, 2006). Dlx5/Dlx6 is required for establishing mandibular identity, because loss of these proteins leads to the transformation of the mandible into a maxilla (Depew et al., 2002). Hand2 restricts Dlx5/Dlx6 to the proximal domain of the mandibular arch (Barron et al., 2011). Moreover, Dlx5 controls oronasal patterning in the anterior part of the palatal shelf and indirectly regulates Shh signaling in the palatal epithelium during palatogenesis (Han et al., 2009). However, it has remained unknown where Dlx5 is expressed in the posterior part of the palate and whether Dlx5 plays a role in regulating the development of the soft palate and oropharyngeal region.

In this study, we have analyzed the expression and function of Dlx5 during the development of the posterior palate and pharyngeal region. Dlx5 is expressed in CNC cells adjacent to myogenic progenitors that are derived from the 4th pharyngeal arch. Among the soft palate muscles in Dlx5^−/− mice, the TVP appeared indistinguishable from that of control mice, but the LVP, PLP and PLG were absent. Specifically, using cell-lineage tracing analysis, we found that Dlx5-positive CNC cells from the 4th pharyngeal arch migrated into the soft palate primordium to set up a scaffold and were adjacent to muscle progenitor cells in the LVP, PLG and PLP regions. Loss of Dlx5 also affected the proliferation and cell survival of CNC and myogenic cells in the LVP, PLG and PLP. Moreover, loss of Dlx5 resulted in decreased expression of Fgf10. In addition, activation of FGF10 signaling rescued differentiation of muscle progenitor cells and proliferation of CNC cells in Dlx5^−/− mice. Overall, loss of Dlx5-mediated FGF10 signaling led to defects in the development of the LVP, PLG and PLP as well as patterning of the oropharyngeal region.

To investigate the function of Dlx5 during the development of the posterior palate, we analyzed the phenotypes of Dlx5^−/− mice in the craniofacial region. Consistent with previous reports, Dlx5^−/− mice exhibited a shortened soft palate with a uvula-like structure at the end (Fig. 1A,B) (Han et al., 2009). Using microCT analysis, we found that the soft palate failed to connect to the pharyngeal wall and that rugae were prominent in palates of Dlx5^−/− mice (Fig. 1C,D). Next, we focused on the muscles in the soft palate region using histological analysis. In Dlx5^−/− mice, the TVP was indistinguishable from that of control mice (Fig. 1E,F). In contrast, the LVP and PLG were undetectable in Dlx5^−/− mice (Fig. 1G,H, data not shown). Moreover, we found that the midline region of the PLP and superior pharyngeal constrictor (SPC) were missing in Dlx5^−/− mice (Fig. 1I,J; Fig. S1). These results indicate that Dlx5 is crucial for the development of soft palate and pharyngeal muscles.

To analyze where Dlx5 is expressed during development of the soft palate, we used Dlx5-CreER;tdTomato and Wnt1-Cre;ZsGreen mice to compare the distribution of Dlx5-positive cells and CNC-derived cells, respectively. In whole-mount samples, CNC-derived cells were detectable mainly in the pharyngeal arches (PAs) at early stages (E10.5, E11.5) (Fig. 2A,B) and then throughout most of the craniofacial region at later stages (E13.5, E14.5, E17.5) (Fig. 2C-E). Dlx5-positive cells were also detectable in similar, but more restricted, regions in the PAs (Fig. 2F,G), and in the craniofacial region at later stages (Fig. 2H-J). Next, we analyzed endogenous Dlx5 expression in E16.5 Wnt1-Cre;ZsGreen mice and performed double staining of ZsGreen and Dlx5 using RNAscope in situ hybridization analysis. Expression of ZsGreen, indicating the CNC cells, was detectable throughout the soft palate region (Fig. 2K-P). In the TVP, expression of Dlx5 was not detectable and was only detectable in the pterygoid plate (Fig. 2K). In contrast, Dlx5 was expressed adjacent to the muscles of the LVP and PLP (Fig. 2L,M,O,P). Moreover, Dlx5 expression colocalized with ZsGreen expression in the LVP and PLP regions, but not in the TVP region (Fig. 2N-P). These results suggest that Dlx5 is expressed by a subset of CNC-derived cells during key stages of soft palate development.

In order to assess whether Dlx5-positive CNC cells affect muscle development in the soft palate region, we compared Dlx5-positive cells with expression of myoblast determination protein 1 (MyoD) and myosin heavy chain (MHC). MyoD is expressed early during muscle development and acts as a muscle determination gene (Rudnicki et al., 1993). MHC is a marker of mature stages of myogenic differentiation (Chanoine et al., 1990; Parada et al., 2012). Using both of these markers allowed us to assess the possible interaction of Dlx5-positive CNC cells with myogenic cells at various stages of muscle development. In the TVP region of Dlx5-Cre;tdTomato mice, MyoD-positive muscle progenitor cells were detectable at E13.5 and E14.5, but fewer cells were detectable at E17.5 (Fig. 3A-C,S). MHC-positive muscle cells were detectable at E13.5 and in greater numbers at E14.5 and E17.5 (Fig. 3D-F,V). Thus, as MHC-positive muscle cells were increasing in number, MyoD-positive muscle progenitor cells were decreasing, indicating that our assays faithfully represent the myogenic differentiation program in vivo. Next, we compared the Dlx5 expression pattern with that of MyoD- and MHC-positive muscle cells. We found that Dlx5-positive cells were not immediately adjacent to the MyoD- and MHC-positive cells in the TVP region (Fig. 3A-F). Our results suggest that Dlx5-expressing CNC cells may not regulate TVP development.

In the LVP and PLP regions, MyoD-positive muscle progenitor cells were detectable at E14.5 and E15.5 (Fig. 3G,H,M,N). These cells were decreased in number at E17.5 in the LVP region (Fig. 3I,T), but they persisted throughout the PLP from the lateral region to the midline (Fig. 3O,U). MHC-positive muscle cells were detectable at E14.5 and present in greater numbers at E15.5 and E17.5 (Fig. 3J-L,P-R,W,X). These data suggest that the differentiation of the PLP is not yet finished at E17.5 and that the PLP differentiates later than the TVP and LVP. Dlx5-positive CNC cells were immediately adjacent to the MyoD- and MHC-positive cells in the regions of the LVP and PLP (Fig. 3G-R). In addition, more Dlx5-positive cells adjacent to MyoD- and MHC-positive cells were detectable in the LVP and PLP compared with the TVP (Fig. 3Y-B′). These results indicate that Dlx5-positive CNC cells closely associate with muscle progenitor cells in the LVP and PLP, but not the TVP, regions, consistent with a role related to muscle development.

In order to examine the cellular defect of Dlx5^−/− mice in greater detail, we analyzed when the soft palate muscles began to exhibit defects. In both the LVP and PLP regions, there was no significant difference between control and Dlx5^−/− embryos at E13.5 (data not shown). Dlx5^−/− embryos began to exhibit defects in the LVP region by E14.5 (Fig. S2A-F), after palatal shelf elevation, which persisted at E15.5 (Fig. S2G-L). After the PLP began to grow from the lateral region towards the midline, muscle defects were detectable in Dlx5^−/− embryos at E15.5 (Fig. S2M-R). Next, we investigated the potential cellular mechanism for the defects in the LVP and PLP regions of Dlx5^−/− mice by determining whether Dlx5 disruption affects apoptosis and proliferation. First, we compared expression of caspase 3 in control and Dlx5^−/− mice for analysis of apoptosis. In the TVP region, apoptosis was unaffected in Dlx5^−/− mice compared with control mice at E14.5 and E15.5 (Fig. 4A-D,M-P). In contrast, more caspase3-positive cells were detectable in Dlx5^−/− mice compared with control mice at E14.5 and E15.5 in the LVP region (Fig. 4E-H). The apoptotic cells colocalized with MyoD-negative CNC cells at E14.5 and MyoD-positive muscle progenitor cells at E15.5 (Fig. 4G,H,Q-T). In the PLP region, loss of Dlx5 resulted in increased apoptosis in MyoD-negative CNC cells at E15.5 (Fig. 4J,L,V,X), but apoptosis was unaffected in CNC and muscle progenitor cells at E14.5 (Fig. 4I,K,U,W). These results demonstrate that loss of Dlx5 leads to increased apoptosis of CNC cells during early development of the soft palate in the LVP and PLP, but not the TVP. The earlier onset of apoptosis in CNC cells in the LVP at E14.5 versus E15.5 in the PLP is perhaps due to the earlier development of the LVP compared with the PLP (see Fig. 3). Moreover, the increased apoptosis switches from CNC cells to muscle progenitor cells during later development of the LVP.

Next, we compared expression of phospho-histone 3 in control and Dlx5^−/− mice for analysis of proliferation. Robust proliferation was detectable in the LVP and PLP regions of E14.5 and E15.5 control mice, primarily in MyoD-negative CNC cells, whereas we detected decreased cell proliferation in Dlx5^−/− mice (Fig. 5C,D,G,H,K,L,O,P). In contrast, cell proliferation in the TVP of Dlx5^−/− mice was indistinguishable from that of control mice (Fig. 5A,B,E,F,I,J,M,N). These results suggest that Dlx5-mediated signaling in CNC cells may be crucial for the proliferation of CNC cells and the survival of muscle progenitor and CNC cells in the LVP and PLP regions of the soft palate, but not in the TVP region.

In order to analyze the molecular mechanism of muscle development in the LVP and PLP, we performed RNA profile analysis using the samples from soft palate tissues of E14.5 control and Dlx5^−/− embryos. In this comparison, we found 674 genes differentially expressed (≥1.7-fold, FDR P<0.05, n=3), 416 of them downregulated and 258 upregulated in Dlx5^−/− mice.

We focused on FGF signaling, one of the differentially expressed signaling pathways we identified from the RNA profile analysis (Fig. S3). Previous studies have reported that many FGF family members are expressed in the skeletal muscles (Hannon et al., 1996; Hébert et al., 1990). Moreover, our group found that TGFβ-mediated FGF10 signaling in neural crest cells controls the development of muscle progenitor cells during tongue morphogenesis (Hosokawa et al., 2010). Thus, we first analyzed gene expression of Fgf10 in the soft palate region using in situ hybridization analysis. Fgf10 expression was detectable in the TVP of both control and Dlx5^−/− mice (Fig. 6A,D). In the LVP and PLP regions, Fgf10 expression was detectable in control mice but was decreased in Dlx5^−/− mice (Fig. 6B,C,E,F). We confirmed that Fgf10 expression was decreased in the soft palate region using quantitative PCR (Fig. S4). Based on this, we analyzed whether activation of FGF10 signaling could rescue the development of the soft palate muscles in Dlx5^−/− mice. We cultured cells from the soft palate tissues of E13.5 control and Dlx5^−/− mice treated with or without exogenous FGF10 for 1 week and analyzed muscle differentiation by counting MHC-positive cells per DAPI-positive total cells. Consistent with our results above, more MHC-positive cells were detectable in mock-treated control cells than Dlx5^−/− cells (Fig. 7A,C,I). After treatment of Dlx5^−/− cells with FGF10, the number of MHC-positive cells increased and was similar to that of control cells (Fig. 7A-D,I). These results suggest that loss of Dlx5 affects muscle development of the LVP and PLP via regulation of FGF10 signaling. We also analyzed whether activation of FGF10 signaling can rescue proliferative ability in the soft palate of Dlx5^−/− mice. We cultured cells from the soft palate tissues of E13.5 control and Dlx5^−/− mice treated with or without FGF10 for 2 days and analyzed their proliferative ability. More MyoD-negative CNC cells were proliferating in control samples than in Dlx5^−/− samples (Fig. 7E,G,J), and total proliferation was also decreased in Dlx5^−/− cells (Fig. S5). After treatment of Dlx5^−/− cells with FGF10, the number of proliferative CNC cells increased in the Dlx5^−/− samples and was similar to that of control cells (Fig. 7E-H,J). These results suggest that loss of Dlx5 affects proliferative ability of CNC cells via regulation of FGF10 signaling.

In this study, we found that loss of Dlx5 led to defects in oropharyngeal patterning and soft palate muscles derived from the 4th pharyngeal arch (PA). During the formation of the head muscles, the precursors of the pharyngeal mesoderm, which is a subset of the head mesoderm that surrounds the developing pharynx, stream into the neighboring PAs that form the templates of the craniofacial structures. In the soft palate, the TVP is derived from the 1st PA, whereas the LVP, PLP and PLG are derived from the 4th PA (Lieberman, 2011). Muscles derived from different PAs must be physically and functionally integrated. For example, muscles for facial expression that are derived from the 2nd PA and muscles for mastication that are derived from the 1st PA work together. These muscles are highly diverse and reflect evolutionary changes related to sites of jaw articulation and modes of masticatory movement (Noden and Francis-West, 2006). Moreover, many factors contribute to the development of muscles from specific PAs. Tbx1 is expressed in the premyogenic mesoderm of the 1st and 2nd Pas, and affects the maturation of the PA myoblasts and formation of some head muscles (Kelly et al., 2004). Pitx2 is expressed in the 1st PA mesoderm at E9.5 and co-expressed with Tbx1 and MyoD in the 1st and 2nd PAs at E10.5 (Shih et al., 2008). Previous studies have demonstrated that Pitx2 specifies 1st PA-derived muscles through regulating the activation of Tbx1 (Shih et al., 2007). However, the factors that affect muscle development from other PAs remain unclear. In this study, loss of Dlx5 led to defects of the LVP, PLP, PLG and superior pharyngeal constrictor, which are all derived from the 4th PA, whereas the TVP, which is derived from the 1st PA, was unaffected. These results suggest that Dlx5 plays a crucial role in the development of muscles derived from the 4th PA.

Previous studies have shown that there is significant molecular heterogeneity in the regulation of head muscle development (Nathan et al., 2008). In chick embryos, the pharyngeal mesoderm can be subdivided into groups of cells based on their contribution to muscle development associated with different pharyngeal arches (Michailovici et al., 2015). However, there is not yet any definitive information that links specific molecular regulators with the development of muscles within each pharyngeal arch. In this study, the discovery that Dlx5 is specifically associated with muscles derived from the 4th pharyngeal arch in mammals highlights the early establishment of arch identity during head muscle development. It also begins to provide unique tools for us to target pharyngeal arch-specific derivatives in order to investigate the regulatory mechanisms for craniofacial development.

CNC cells play a crucial role in the patterning of the head muscles (Olsson et al., 2001; Schilling and Kimmel, 1997). Previous studies have demonstrated that ablation of CNC cells in chick embryos leads to abnormal migratory pathways and anterior-posterior registration of the mesoderm cells (Rinon et al., 2007). Our previous study showed that CNC-derived cells control muscle development in the tongue through tissue-tissue interactions (Hosokawa et al., 2010). In the soft palate region, CNC-derived cells migrate into the primordium of the soft palate and form a scaffold to guide the migration of myogenic progenitors (Grimaldi et al., 2015). In this study, we demonstrated that Dlx5-positive cells are a subset of CNC-derived cells and that Dlx5-positive CNC cells are adjacent to myogenic cells in the LVP and PLP regions. Earlier in development, Dlx5-positive cells were detectable in the olfactory placode, otic placode and PAs. These patterns are consistent with the results of previous gene expression analyses using in situ hybridization (Depew et al., 1999). Strikingly, Dlx5-positive CNC cells are not adjacent to myogenic cells in the TVP region, which is derived from the 1st PA and appears unaffected in Dlx5^−/− mice. These results suggest that Dlx5-positive CNC cells have a close association with myogenic cells and guide migration of myogenic cells to the terminal differentiation site only in muscles derived from the 4th PA.

CNC cells give rise to all facial bone and some of the calvaria. Head muscles attach to these bones through tendons, which are also CNC-derived tissue (Grenier et al., 2009; Han et al., 2012). Previous studies have shown that CNC cells are progressively involved in patterning and development of head muscles (Tzahor, 2015). As pharyngeal mesodermal progenitors migrate into the pharyngeal arch, they also encounter signals from pharyngeal ectoderm and endoderm. In order to investigate the specific roles of CNC cells in regulating head muscle development, we need to identify cell type-specific regulatory factors and test their instructive role in regulating head muscle development. Clearly, the Dlx5-mediated signaling network is crucial for CNC and myogenic progenitor interactions during the development of 4th pharyngeal arch muscles.

In the PAs, CNC cells that surround the muscle anlagen separate the myoblasts from the overlying surface ectoderm. Subsequently, the muscle progenitors within the PAs differentiate and form each of the muscles (Michailovici et al., 2015). Studies have reported that CNC cells are not essential for early myogenic development, but are necessary for the migration, positioning and differentiation of cranial muscle precursors in vertebrates (Rinon et al., 2007). Previously, our work revealed that Dlx5 controls the fate of CNC cells in the anterior palate region (Han et al., 2009). In this study, we discovered that loss of Dlx5 leads to reduced cell proliferation and increased apoptosis in CNC cells at earlier stages and increased apoptosis in myogenic cells at later stages. In addition to CNC cells, palatal epithelial signals can also affect muscle development through tissue-tissue interactions. For example, epithelial TGFβ signaling can regulate WNT signaling activity in the soft palate mesenchyme and affect muscle development (Iwata et al., 2014). Members of the TGFβ family have been shown to act both upstream and downstream of Dlx5 in regulating the fate of CNC cells during craniofacial development. Our results indicate that Dlx5-mediated signaling affects proliferation and apoptosis of CNC cells. Furthermore, myogenic progenitor cells fail to migrate correctly and myogenic cell survival is affected secondarily in Dlx5 mutant mice. It will be crucial to identify the signaling molecules regulated by Dlx5 that can exert action on myogenic progenitors during soft palate muscle development.

We identified the FGF pathway as altered in Dlx5^−/− mice based on RNA-seq analysis. Many members of the FGF signaling family are expressed in muscle cells (Hannon et al., 1996), and the TGFβ-induced FGF10 signaling pathway plays crucial roles in muscle development of the tongue (Hosokawa et al., 2010). In addition, FGF7 and FGF10 are expressed in the anterior palate region, and Fgfr2b^−/− and Ffg10^−/− mice exhibit cleft palate (Rice et al., 2004). During palatogenesis, Dlx-regulated FGF7 signaling inhibits the expression of Shh, which in turn controls the fate of CNC cells through tissue-tissue interaction (Han et al., 2009). In this study, we found that expression of FGF10 is decreased in Dlx5^−/− mice in soft palate muscles derived from the 4th PA. In addition, activation of FGF10 signaling rescued the differentiation program of myogenic cells and proliferative ability of CNC cells in Dlx5^−/− mice. These results clearly indicate that a Dlx5/FGF signaling interaction plays a key role in patterning and development of the hard and soft palate. Previous studies have shown that key transcription factors expressed in CNC cells may regulate FGF signaling to control craniofacial and cardiac development. For example, neural crest cell ablation in the Hox-negative mid-diencephalon through the rhombomere 2 region led to downregulation of FGF8 expression and abnormal pharyngeal arch development in chick embryos (Creuzet et al., 2004). On the other hand, Hox-positive cardiac neural crest cells appear to inhibit FGF8 signaling (Michailovici et al., 2015). Although more work is need to elucidate the precise mechanism by which neural crest cells regulate FGF signaling, it appears to be a crucial and well-conserved mechanism in mediating tissue-tissue interactions during embryonic development.

In summary, we have found that Dlx5-mediated FGF10 signaling plays a crucial role in oropharyngeal patterning and the development of muscles derived from 4th PA through interactions between CNC and muscle progenitor cells. These findings are important for clinical studies that aim to correct soft palate muscle development defects. Moreover, the mechanisms of soft palate muscle development identified in this study may lead to a better approach to muscle regeneration in individuals with soft palate muscle defects in the future.

To generate Dlx5^−/− mice, we mated Dlx5^+/− with Dlx5^+/− mice. To generate Wnt1-Cre;ZsGreen mice, we mated Wnt1-Cre;ZsGreen^fl/+ males with ZsGreen^fl/fl females. To generate Dlx5-CreER;tdTomato mice, we mated Dlx5-CreER;tdTomato^fl/+ males with tdTomato^fl/fl females. Pregnant females were harvested based on staging determined by counting the days after detection of a vaginal plug and Theiler staging criteria (Theiler, 1989). The animals were euthanized through carbon dioxide overdose followed by cervical dislocation. All studies were performed with the approval of the Institutional Animal Care and Use Committee (IACUC) of the University of Southern California.

For microCT analysis, we used a SCANCO µCT50 device at the University of Southern California Molecular Imaging Center. The images were detected with the X-ray source at 70 kVp and 114 µA. The data were collected at a resolution of 10 µm. 3D reconstruction was performed using AVIZO 7.1 (Visualization Sciences Group).

Tamoxifen (Sigma, T5648) was dissolved in corn oil (Sigma, C8267) at 20 mg/ml and injected intraperitoneally 2 days before sacrificing mice (single injection, 2 mg/10 g body weight).

For Hematoxylin and Eosin staining, samples were fixed in 4% paraformaldehyde (PFA) and decalcified with 10% EDTA. Next, samples were dehydrated with serial ethanol and embedded in paraffin wax. Sections were cut at 7 µm and Hematoxylin and Eosin staining was performed using standard procedures. For immunofluorescence staining, the following primary antibodies were used: myoblast determination protein 1 (MyoD; DAKO, M3512; 1:50), myosin heavy chain (MHC; DSHB, P13538; 1:10), active caspase 3 (Casp3; Abcam, ab2302; 1:100), phosphohistone H3 (PH3; Millipore, 06-570; 1:100) and RFP (tdTomato; Rockland, 600-401-379; 1:200). Alexa Fluor 488 and 568 were used as secondary antibodies. Staining procedure was performed according to standard protocol (Iwata et al., 2012; Sugii et al., 2014). Sections were counterstained with DAPI and imaged by Leica DMI 3000B and Keyence BZ-X710 microscopes.

Soft palate tissue was collected at E15.5 and E16.5, and fixed in fresh 4% PFA. Samples were passed through 15% and 30% sucrose and embedded in OCT compound (Tissue-Tek). Sections were cut at 7 µm and staining was performed with the RNAscope 2.5HD Reagent Kit-RED assay (Advanced Cell Diagnostics, 322350) according to the manufacturer's protocol. The RNAscope probes used in this study were Fgf10 (Advanced Cell Diagnostics, 446371), Dlx5 (Advanced Cell Diagnostics, 478151) and ZsGreen (Advanced Cell Diagnostics, 461251-C2).

Samples from the soft palate tissue of E14.5 Dlx5^−/− and control mice were used to extract total RNA. cDNA library preparation and sequencing were performed at the USC Molecular Genomics Core. A total of 400 million paired-end reads with 75 cycles were performed on Illumina Next Seq 500 equipment for three pairs of samples. Raw reads were trimmed, aligned using TopHat2 with mm 10 genomes and normalized using RPKM. Differential expression was calculated by selecting transcripts that changed with a significance of P<0.05.

Total RNA was isolated from mouse soft palate tissue dissected at E13.5 using QIAshredder and RNeasy Micro extraction kits (Qiagen) as described previously (Iwata et al., 2010). The following PCR primers were used: Fgf10, 5′-TTTGGTGTCTTCGTTCCCTGT-3′ and 5′-TAGCTCCGCACATGCCTTC-3′.

Soft palate tissue was dissected from control and Dlx5^−/− embryos at E13.5, followed by treatment with 2.4 U/ml Dispase I (Roche Applied Science) for 1 h at 37°C. The cells were isolated and cultured as previously described (Hosokawa et al., 2010). Where indicated, mouse recombinant FGF10 (10 ng/ml; R&D) was added to the medium after plating and re-added every other day, followed by cell culture for 2 days or 1 week.

Two-tailed Student's t-tests were applied for statistical analysis. For all graphs, data are mean±s.d. P<0.05 was considered statistically significant.

We are grateful to Dr Julie Mayo and Dr Bridget Samuels for critical reading and editing of the manuscript. We thank Dr John Rubenstein at UCSF for Dlx5 mutant mice, Vasu Punj for analyses of RNA-seq and USC Norris Medical Library Bioinformatics Service for assisting with data analysis. The bioinformatics software and computing resources used in the analysis are funded by USC Office of Research and the Norris Medical Library.