TWIST1 interacts with β/δ-catenins during neural tube development and regulates fate transition in cranial neural crest cells

Craniofacial morphogenesis requires precise interaction between multiple genes and gene environments that are essential for cell fate induction, specification and differentiation (Chai and Maxson, 2006; Nomura and Li, 1998; Padmanabhan and Ahmed, 1997). Disruption of the precisely organized morphogenesis leads to craniofacial disorders, which are the second most common congenital birth defects in humans (Hoyert et al., 2006; Joshi et al., 2014; Murray et al., 2007; Parker et al., 2010). During embryogenesis, signaling molecules from the notochord and mesoderm induce the formation of neuroectoderm. The junctions between non-neural and neural ectoderm are called the neural plate borders, which form distinct domains through inductive interactions with the adjacent cells and the underlying mesoderm (Groves and LaBonne, 2014; Litingtung and Chiang, 2000; Millonig et al., 2000). The CNCCs emerge from the neural fold borders of the midbrain and hindbrain regions, and migrate as unipotent and multipotent mesenchymal cells towards the frontonasal processes (Achilleos and Trainor, 2012; Zhang et al., 2021). The CNCCs give rise to the bone, cartilage, neuron and glial cells of the orofacial skeletal and peripheral nervous systems (Groves and LaBonne, 2014; He et al., 2014; Le Douarin et al., 2004; Noden and Trainor, 2005).

Twist1 encodes a transcription factor that belongs to the basic helix-loop-helix (bHLH) B-family. It was first identified in Drosophila melanogaster where its deletion resulted in disruption of mesodermal specification and ventral furrow formation (Lu et al., 2011; Nüsslein-Volhard et al., 1984; Simpson, 1983). In vertebrates, Twist1 is expressed first at the gastrulation stage in the primitive streak and then in the adjacent mesodermal cells (Füchtbauer, 1995; Stoetzel et al., 1995). At the neurulation stage, mammalian Twist1 is detected in the paraxial and lateral mesoderm, and the migratory CNCCs (Bildsoe et al., 2016; Füchtbauer, 1995; Ota et al., 2004; Soo et al., 2002). Twist1 expression is maintained in the CNCC-derived mesenchyme of the frontonasal and pharyngeal processes (Bildsoe et al., 2009; Füchtbauer, 1995). In humans, mutations in TWIST1 cause Saethre-Chotzen syndrome, which is characterized by craniosynostosis and cleft palate, Sweeney-Cox syndrome and Robinow-Sorauf syndrome (Kunz et al., 1999; Takenouchi et al., 2018; Seto et al., 2007). In mice, Twist1 is crucial for neural tube closure and for CNCC-derived craniofacial bone and cartilage formation (Chen et al., 2007; Bildsoe et al., 2013). TWIST1 has been shown to promote cell survival and proliferation of migratory CNCCs during craniofacial development (Bildsoe et al., 2013). Although both processes involve gaining potency and motility, the in vivo molecular function of TWIST1 as a master regulator of epithelial-to-mesenchymal transition (EMT) in CNCCs was thus far only described in Xenopus (Lander et al., 2013).

TWIST1 is a phosphoprotein with multiple residues that are phosphorylated by several kinases according to numerous cancer studies (Lu et al., 2011; Vichalkovski et al., 2010; Xue and Hemmings, 2012). TWIST1 phosphorylation is crucial for regulating its homo- and heterodimerization with other factors to control multiple cellular activities (Firulli and Conway, 2008; Hong et al., 2011; Lu et al., 2011; Vichalkovski et al., 2010). TWIST1 interacts with other proteins, such as TCF3 and chromatin remodeling factors, to guide CNCC migration during craniofacial development (Firulli et al., 2005; Fan et al., 2021). Misregulation of TWIST1 dimerization with HAND2 is caused by mutations of phosphoresidues in the PKA consensus domain or in the bHLH domain, which have been reported to be associated with Saethre-Chotzen syndrome (Firulli et al., 2005; El Ghouzzi et al., 1997). Mutations within the TWIST-box domain are also associated with the isolated form of craniosynostosis (Seto et al., 2007). TWIST1 is also expressed in cardiomyocytes and is repressed in patients with dilated cardiomyopathy. Knockdown of Twist1 in primary rat cardiomyocytes that are stimulated using phenylephrine/isoprenaline causes significant reduction in cell size (Baumgarten et al., 2013). Overexpression of Twist1^T125D/S127D phosphomimetic alleles (equivalent to T121/S123 in human) in cardiomyocytes resulted in disrupted cell remodeling and heart formation of transient transgenic mouse embryos (Lu et al., 2011).

Our previous work demonstrated that the compound heterozygous mice (Irf6^+/−; Twist1^+/−) have craniofacial abnormalities, i.e. mandibular agnathia, fused maxilla, cleft palate and holoprosencephaly (Fakhouri et al., 2017). Unlike other IRF family members, interferon regulatory factor 6 (Irf6) is a broadly expressed transcription factor in the ectoderm and oral epithelium during embryogenesis. In mice, Irf6 promotes cell differentiation of proliferative epithelial cells, but its function in the neural tube and neural crest cell development is still unknown (Ingraham et al., 2006; Kousa et al., 2019; Richardson et al., 2006). We have recently shown that overexpression of Irf6 in ectoderm by K14-enhancer leads to exencephaly and failure of peritoneal skin development. Furthermore, the compound heterozygous mice for Irf6 and Tfap2a do not show neural tube defects compared to single Tfap2a heterozygous mice (Kousa et al., 2019).

In the current study, we sought to characterize the spatiotemporal expression of Twist1 and Irf6 during neural tube formation and their function in CNCCs. We generated multiple stable Twist1 phospho-incompetent mouse lines to identify the importance of TWIST1 phosphorylation in regulating neural tube and CNCC-derived craniofacial structures. We investigated the molecular role of Irf6 in regulating the integrity of the neural tube and its relation to AP2α expression. Our data show that TWIST1 protein and mRNA are detected at the apical side of dorsal neuroectodermal cells within expression locations similar to β-catenin and tight junction proteins. Further investigation suggests that TWIST1 is expressed in endocytic compartments at the apical surface and partially overlaps with endocytic markers LRP2 and RAB11b. Meanwhile, cytosolic TWIST1 interacts with β/δ-catenin during neural tube formation. We show that Twist1 is crucial to promote cell fate transition of pre-EMT neuroectodermal cells, and loss of Twist1 in neuroectoderm causes ectopic expression of IRF6 and E-cadherin in delaminated CNCCs. We demonstrate that three highly conserved phospho-residues in TWIST1 are crucial for its in vivo function in the neural tube and in CNCC-derived structures. Finally, our data shows that TWIST1 regulates Specc1l expression during CNCC development to possibly control cytoskeletal remodeling in CNCCs.

To determine the role of Twist1 and Irf6 in neural tube formation, we investigated their spatiotemporal expression in wild-type embryos. We performed immunofluorescent staining and in situ hybridization in embryos from E8.5 to E9.5. At E8.5, IRF6 was highly expressed in the basal and apical sides of neural plate, with low expression seen in the middle and weak IRF6 expression detected in some cells beneath the neural plate (Fig. 1A,A′). TWIST1 was unexpectedly detected in vesicle-like structures at the apical surface of neural plate (Fig. 1A,A′, white arrow; Fig. S1A-B). Nuclear TWIST1 was weakly expressed in a few cells at the neural plate borders (Fig. 1B,B′, white arrow). Robust nuclear expression of TWIST1 was observed in the mesodermal cells adjacent to the neural plate (Fig. 1A,B, green arrow). At E9.0, most dorsal and ventral cells in neural folds expressed IRF6, while TWIST1 was mostly expressed at the apical side and dorsal edges of neural folds. TWIST1 expression overlapped with IRF6 at the dorsal edges of the neural folds (Fig. 1E,E′, white arrow; Fig. S1C). To understand the role of TWIST1 apical expression, we investigated whether its expression was associated with adherens and tight junction proteins. TWIST1 apical expression in neuroectodermal cells had similar expression patterns to β-catenin (Fig. 1C,C′; Fig. S1C,D) and claudin 1 at E8.5 (Fig. 1D,D′) and E9.0 (Fig. 1F-G′; Fig. S1E). SOX9 was used as a marker of pre-EMT and migratory CNCCs (Fig. 1B-D, red arrow). TWIST1 apical expression was also observed at the dorsal edges of the neural folds immediately before fusion (Fig. 1H-H′, arrowhead). We tested the specificity of the anti-TWIST1 antibodies in Twist1-null embryos. Histological staining showed the neural tube closure defects in Twist1-null compared with wild-type embryos at E10.5 (Fig. S2A-B′). No signal for TWIST1 protein was detected in the neural tube or mesenchymal cells of a Twist1-null embryo compared with a wild type (Fig. S2C-F′). Notably, Twist1 null embryos had complete neural tube closure defects (Fig. 1J, white arrow) compared with wild-type littermates at E11.5 (Fig. 1I, white arrow). In Twist1 null embryos, the neural plate invaginated and lateral neural folds elevated towards the dorsal midline but the lateral folds failed to close (Fig. 1J, white arrow). To validate the immunofluorescence data, whole-mount in situ hybridization for Twist1 mRNA showed weak expression at the dorsal edges of the neural plate of E8.5 embryos (Fig. 1K-K′; Figs S3 and S4) and of neural folds of E9.5 embryos (Fig. 1L-M′; Fig. S5). Based on a previous publication reporting the expression of LRP2 and RAB11b, a marker of endocytic cellular membrane vesicles, at the apical surface (Kowalczyk et al., 2021), we investigated whether TWIST1 was expressed in endocytic vesicles of RAB11b compartments. The dual immunofluorescent staining data shows that TWIST1 was expressed in endocytic vesicles at the apical surface of neural folds (Fig. 1N-O′). TWIST1 expression partially overlapped with LRP2 at the apical side of neural tube (Fig. 1N,N′, inlet in N′; Fig. S1A,A′) and with RAB11b in endocytic compartments (Fig. 1O,O′, inlet in N′; Fig. S1B,B′). Rab11b mRNA was significantly reduced in hindbrain tissues of Twist1 conditional knockout (CKO), while no change was detected for Lrp2 expression (Fig. S6C,D).

To further determine the significance of TWIST1 apical expression, we performed co-immunoprecipitation (co-IP) for TWIST1 to identify the protein interactors at E8.5-E9.0. The in vivo cellular fractionation co-IP blots show that the β-catenin and δ-catenin in the cytosolic fraction were pulled down with TWIST1 immunoprecipitation using anti-TWIST1 antibodies (Fig. 2A,B). In contrast, nuclear TWIST1 weakly interacted with β-catenin (Fig. 2A). We also performed dual immunostaining for TWIST1 and β-catenin using rabbit anti-β-catenin antibodies. β-Catenin was detected at the apical-lateral side of all the dorsal neuroectodermal cells (Fig. 2C,C′; Fig. S1F,G). Likewise, TWIST1 and δ-catenin were detected at the apical side of neural folds in a similar pattern of consecutive sections (Fig. 1D,D′; Fig. S1H-I′). Fig. 2E shows neural folds and a dotted-line square marks the areas indicated in Fig. 1F,G. β-Catenin apical expression detected in wild type was diffused in the Twist1^cko/− embryo and became mostly cytosolic when detected using mouse anti-β-catenin antibodies (Fig. 2F-G′). To further determine the impact of Twist1 loss in neural tube formation, Hematoxylin and Eosin staining was performed in wild-type and Twist1^cko/− embryonic tissues. Normal neural tube formation was observed in wild type (Fig. 2H,J), while Twist1^cko/− embryos had multiple dorsolateral bend points and expansion of the neural tube (Fig. 2I,K). Histological staining of older wild-type embryos showed a normal development of the cephalic flexure of the midbrain and hindbrain at E12.5 (Fig. 2L,L′; Fig. S6F-F″), while Twist1^cko/− showed abnormal patterning of the cephalic flexure of midbrain and hindbrain at E12.5 (Fig. 2M,M′, black arrows; Fig. S6H-H″).

To determine the role of Twist1 in CNCCs, we used Wnt1-Cre and Wnt1-Cre2 to conditionally knockout (CKO) Twist1 in pre-EMT CNCCs. For cell tracing, we used R26^Tm1 reporter gene and X-gal whole-mount staining. At E9.5, the blue staining of migratory CNCCs was detected in the cephalic flexure of the midbrain, frontonasal process and pharyngeal arches of wild-type embryos (Fig. 3A), and similar expression pattern was detected in wild type at E11.5 but with stronger staining in the rhombic lips (Fig. 3C). At E15.5, migratory CNCCs were observed in the otic pinna and orofacial regions (Fig. 3E). In mutant Twist1^cko/− embryos, weaker blue staining was observed in the frontonasal and pharyngeal processes, while stronger staining was detected in the mid- and hindbrain regions at E9.5, E11.5 and E15.5, suggesting disruption of migratory CNCCs towards frontal processes (Fig. 3B,D,F). We noticed that the first pharyngeal arch of Twist1^cko/− embryos was smaller at E11.5, and the embryos displayed severe brain abnormalities at E15.5 (Fig. 3D,F). The recently generated Wnt1-Cre2 mouse line has been used to validate the CNCC-specific deletion by the Wnt1-Cre line that has a midbrain developmental abnormality (Lewis et al., 2013). The Wnt1-Cre2 line showed a similar staining pattern of migratory CNCCs to Wnt1-Cre (Fig. 3G,I,K) and the craniofacial phenotype associated with Twist1^cko/− (Fig. 3H,J,L). At E12.5 and E14.5, a dorsal image from the whole-mount staining of wild-type embryos showed blue staining in the hindbrain and in the trunk neural tube (Fig. S6E,E′; Fig. 3M, black arrow). Similarly, blue staining was observed in Twist1^cko/− embryos, showing expanded staining in the midbrain and hindbrain compared with the wild type (Fig. S6G,G′; Fig. 3N, black arrow). The embryos with compound alleles of Twist1^cko/− and Irf6^+/− presented with severe hemorrhaging in the frontonasal process and craniofacial abnormalities compared with wild type (Fig. 3O,P; Fig. S6I-K). Stereomicroscope images of a wild-type embryo showed normal craniofacial development at E17.5 (Fig. 3Q), while a Twist1^cko/− embryo showed severe exencephaly and abnormal frontonasal processes (Fig. 3R). Twist1^cko/− embryos lacked cranial, frontonasal and maxillary bones, and had severe mandibular hypoplasia, as depicted by the skeletal staining (Fig. 3T), compared with wild-type littermate (Fig. 3S). Histological staining of a wild-type embryo showed normal development of the neural tube apical and basal layers of the hindbrain and midbrain at E10.5, respectively (Fig. 3U,W), while Twist1^cko/− showed an expansion of the neural tube and partially detached cells at the basal side, and abnormal patterning of the cephalic flexure of the midbrain (Fig. 3V,X).

We further investigated why a loss of Twist1 in pre-EMT neuroectodermal cells led to severely abnormal brain and craniofacial structures, even though a significant number of CNCCs migrated towards the frontonasal and pharyngeal processes. We also asked why the mutant migratory CNCCs did not form craniofacial bone and cartilage (Fig. 3Jj). To answer these questions, we developed a neural tube explant system using a dual reporter transgenic mouse line (ROSA26^{tm4.ACTB-tdTomato-EGFP}), which allows us to capture cell morphogenesis and migration of somatic and CNCCs. The illustrative diagram depicts the procedure for dissecting the neural folds from embryos at E8.5 (Fig. 4A). We used this system to uncover the function of Twist1 in pre-EMT CNCCs, wherein lack of its expression in the neuroectodermal and pre-EMT neural crest cells can explain the phenotype observed in Twist1^cko/−. The cells expressing the Wnt1-Cre transgene excise the membrane-associated mTomato reporter gene, allowing instead the expression of GFP. In wild-type explants, individual migratory CNCCs moved away from the dorsolateral sides of the neural folds (Fig. 4B,B′) and the migratory mesenchymal cells formed long protrusions (inset in Fig. 4B′). In the Twist1^cko/− explant, a considerable amount of Twist1-deficient neuroectodermal cells remained within the neural folds and expanded ventrally (Fig. 4C), and the majority of the detached CNCCs did not transition into mesenchymal cells but instead maintained their cell-cell adhesion and epithelial morphology (Fig. 4C′, inset in C′). Quantitative measurements of the total number of migratory CNCCs were analyzed for each genotype. We detected approximately a 50% reduction in the total number of cells in mutant explants (Fig. 4D), and the average distance of all migratory CNCCs was reduced by 68% (Fig. 4E). We also measured the mean speed of individual CNCCs in wild type and Twist1^cko/−. To determine the mean speed of individual CNCCs, we tracked the CNCCs as they moved away from the neural tube in wild type and Twist1^cko/− (Fig. 4F,F′,G,G′). Micrographs show individual CNCC migration paths during time-lapse imaging (Fig. 4F′,G′, white lines). Wild-type CNCCs mean speed increased by 50% in comparison with Twist1^cko/− (Fig. 4H).

IRF6 is expressed in the non-neural and neural ectoderm during neural tube formation. Its expression is relatively stronger in neural ectoderm than non-neural ectoderm (Fig. 1E,E′; Fig. S1C-E). We hypothesized that a complete loss of Irf6 affects neural tube formation, leading to craniofacial abnormalities. At E10.5, we noticed that the cellular organization of the edges of neural tube and apical membrane were disorganized in Irf6 null compared with wild-type littermates (Fig. 5A-B″, black arrow). Abnormal expansions from the brain cortex were also detected in Irf6-null embryos compared with wild type (Fig. 5C-D′). We checked the expression of SOX9, which is expressed in the neuroepithelial progenitors of glia (Farrell et al., 2011; Vogel and Wegner, 2021), and F-actin in wild-type and Irf6-null embryos. SOX9-positive cells were detected in the dorsal half of the neural tube (Fig. 5E,E′, white arrow). However, we found that the organization of SOX9-positive cells was altered and large intercellular gaps among SOX9 positive cells were detected in Irf6-null neural tube (Fig. 5F,F′, white arrows). A continuous expression of F-actin was noticed in the apical cells of the dorsal edge of wild-type neural tube (Fig. 5E″), while the apical membrane was discontinuous in Irf6-null embryos (Fig. 5F″). We also investigated the expression of non-neural ectodermal marker, AP2α, to uncover whether a loss of Irf6 alters its expression at the neural tube junction. We noticed that the number of AP2α-positive cells at the neural tube dorsal edges increased in Irf6-null embryos compared with wild-type littermates in four different biological replicates of each genotype (Fig. 5G,H). We quantified the number of AP2α-positive cells in the dorsal edges and observed a significant increase in Irf6 null compared with wild type (Fig. 5I).

Based on the CNCC-derived tissue phenotypes observed in Twist1^cko/−, we investigated the expression pattern and level of SOX9 and epithelial genes involved in cell adhesion and differentiation to determine why a loss of Twist1 in delaminated CNCCs maintained their epithelial morphology. In wild-type embryos, IRF6 was mainly detected in the neural tube, whereas TWIST1 was highly expressed in migratory mesenchymal cells that were detached from the neural tube and migrated towards frontonasal and pharyngeal prominences at E9.5-E10.5 (Fig. 6A,A′; Fig. S7A-C). Ectopic IRF6 expression in partially detached mesenchymal cells was detected in a cluster of cells along the neural tube in Twist1^cko/− embryos (Fig. 6B-B′, white arrow). No expression of E-cadherin (E-CAD) was present in mesenchymal cells in wild type (Fig. 6C,C′). However, staining for E-CAD in Twist1^cko/− embryos showed that the partially detached mesenchymal cells express ectopic E-CAD (Fig. 6D,D′). SOX9 expression was used to mark migratory CNCCs along the neural tube and to distinguish CNCCs from mesodermal cells. In a wild-type embryo, an abundant number of migratory CNCCs showed co-expression of SOX9 and TWIST1 (Fig. 6E,E′ white arrows), whereas only a few TWIST1-positive CNCCs can be seen in the Twist1^cko/− embryo (Fig. 6F-F′, white arrow). We also looked at the expression of tight junction proteins occludin and ZO1. Occludin is expressed at high levels at the dorsal junction of neural tube edges of wild-type embryos (Fig. 6G,G′), whereas almost no expression of was detected at the unfused junction of the dorsal edges of Twist1^cko/− neural tube (Fig. 6H,H′). In neural tube explants, an interspaced and broken signal of ZO1 was seen among the migratory mesenchymal cells in wild type (Fig. 6I,I′), whereas a continuous junctional signal of ZO1 was observed among detached mesenchymal cells from the neural tube of Twist1^cko/− (Fig. 6J,J′). We also looked at the expression of a mesenchymal marker, vimentin, and observed a disrupted expression pattern in the mesenchymal cells in Twist1^cko/− embryos and neural tube explants compared with wild type (Fig. S7D-G′). As expected, Twist1 expression was significantly reduced in Twist1^cko/− embryos, and a slight increase was seen in Irf6 mRNA in Twist1^cko/− (Fig. 6K). mRNA encoding E-Cad (Cdh1) N-Cad (Cdh2) and β-catenin (Ctnnb1) were also measured in WT and Twist1^cko/− (Fig. 6L). E-Cad expression was significantly increased in Twist1^cko/− (Fig. 6L), which is consistent with the ectopic expression of E-CAD in partially detached mesenchymal cells (Fig. 6D,D′).

The expression of genes encoding kinases [Akt1, Akt2, members of the Pi3k family, Erk1 (Mapk3), members of the Ck2 family and Src]; epithelial markers [E-Cad, N-Cad and occludin (Ocln)]; neural and non-neural ectoderm markers, and CNCC specification drivers [Snail2 (Snai2), Msx1, Tfap2a, Hand2, Sox2, Sox10 and Hoxd10]; signaling molecules [Yap (Yap1), Wnt1, Wnt3, Jag1 and Mtor]; and actomyosin remodeling regulators (β-catenin, Rhoa, Rhoc and Arhgap29) was measured in dissected tissues of the hindbrain and first pharyngeal arch (Fig. S8). The mRNA expression of Akt2, Snai2, Msx1, Hand2, Sox10, Yap, β-catenin, Rhoa and Arhgap29 was significantly reduced in Twist1^cko/− embryos at E10.5 (Fig. S8A,E,I,G,K). Protein analysis by western blot also showed decreased levels of AKT1, AKT2, AKT3, SNAI2 and WNT3 in Twist1^cko/− embryos (Fig. S8B,F,H). Further details are explained in the legend of Fig. S8.

TWIST1 has six phospho-residues that are highly conserved in vertebrates (Fig. S9A). Our data show that some subpopulations in the dorsal edges of neural folds and in the pre-EMT CNCCs express TWIST1 mostly in the cytosol (Fig. S9B-C′, arrows; Fig. S1A′,D,D′), and it becomes mainly nuclear in detached and migratory CNCCs at E9.5 and E10.5 (Fig. 6A,A′, arrow; Fig. S9D,D′; Fig. S7C). Our in vivo data show that phosphorylated TWIST1 protein can be detected in mesenchymal cells using monoclonal antibodies raised against TWIST1 phospho-S68, and polyclonal antibodies raised against TWIST1 phospho-S42 and phospho-S123 residues at E10.5 (Fig. S9E). TWIST1 unphosphorylated and phosphorylated forms were also detected in the O9-1 cell line (Fig. S9F). Further details are provided in the legend of Fig. S9.

To determine the importance of TWIST1 post-translational modifications in the neural tube and CNCC-derived tissues, we generated four phospho-incompetent mouse lines, including Twist1^S18;20A, Twist1^S42A, Twist1^S68A and Twist1^{T121A; S123A} using CRISPR/CAS9 technology. The founders of Twist1^S42A mice did not give any progeny, and all ten Twist1^{T121A; S123A} founders died before reaching 2 months of maturity. We confirmed the genomic changes in three founders of Twist1^S18I;20A/+ and in four founders of Twist1^S68A/+. The sequence of F6 generations confirmed the changes of serine 18 (S18) to isoleucine and serine 20 (S20) to alanine, and in the second mouse line, serine 68 (S68) to alanine (Fig. 7A,B).

The phenotypic characterization showed that the two phosphorylation sites, S18/20 and S68, are crucial for TWIST1 activity in craniofacial development. The two phospho-incompetent mouse lines have epidermal blebbing, severe edema along the neural tube and significant neural tube defects in Twist1^{S18I;20A/S18I;20A} compared with wild-type littermates (Fig. 7C-F′). The skeletal staining exhibited a remarkable loss of maxillary and mandibular bones in Twist1^{S18I;20A/S18I;20A} and a reduction in skull mineralization in Twist1^S68A/S68A, respectively (Fig. 7G′,H′), compared with wild type (Fig. 7G,H). The histological staining of Twist1^S68A/S68A head structures at E15 showed subepidermal blebbing, separation of fibroblast layers from the meninges and a detachment of the meninges from the surface of the brain cortex (Fig. 7J,J′,L,L) compared with wild-type littermates (Fig. 7I,I′,K,K′). We also noticed lymphocyte infiltration in Twist1^{S18I;20A/S18I;20A} brain tissues and around the frontonasal processes compared with wild-type embryos (Fig. 7M,N). To determine the types of the infiltrated cells, immunofluorescence staining for the lymphocyte marker CD73 and B cell marker CD206 was used on tissues sections at E13.5. No signal was detected in the frontonasal sections of wild type (Fig. 7O,Q); however, enrichment of both markers was noticed in the frontonasal tissues of Twist1^{S18I;20A/S18I;20A} embryos (Fig. 7P,R). The genotype and spectrum of craniofacial phenotype in the homozygous phospho-incompetent embryos are further described in Fig. S10.

Many of the tested cytoskeletal remodeling regulators were significantly reduced at the mRNA level in Twist1^cko/− embryos. We focused on Specc1l because of the phenotypic overlap in craniofacial disorders in humans and mice, and the importance of this gene in regulating the actomyosin cytoskeleton. We tested the hypothesis that TWIST1 regulates Specc1l expression in CNCCs and that both factors interact with adherens junction proteins. Specc1l has 17 exons plus 5′ and 3′ UTR (Fig. 8A). We mapped two putative regulatory elements in intron 1 and 2 based on epigenetic signatures (Fig. S11). We performed ChIP-PCR to determine whether TWIST1 in vivo binds to E1 and E2 putative elements. The results showed enrichment of TWIST1 protein at the putative Specc1l element E1 for both embryonic time points E9.5 and E10.5. The bound TWIST1 to E2 putative enhancer was increased at E10.5 as indicated by the PCR band intensity but not at E9.5 or E11.5 (Fig. 8B). mRNA expression of Specc1l was significantly reduced in tissues extracted from the hindbrain and first pharyngeal arch of Twist1^cko/− compared with wild type (Fig. 8C). We matched the phenotype in Twist1^cko/− and phospho-incompetent embryos with Specc1l loss-of-function mutant mouse embryos to compare the phenotypic overlap in craniofacial tissues and CNCCs. Twist1^cko/− and phospho-incompetent mutant embryos showed brain hemorrhage, subepidermal blebbing and severe edema along the neural tube (Fig. 8D,E). Similarly, the Specc1l^cGT/DC510 compound mutant alleles (described by Hall et al., 2020) show brain hemorrhage, subepidermal blebbing and severe edema along the neural tube compared with wild-type littermate embryos (Fig. 8F). Loss of Twist1 in CNCCs led to disruption in the EMT process and the persistence of cell-cell adhesion in migratory mesenchymal cells (Fig. 8H), and the in vivo accumulation of adherens junction proteins β-catenin and E-CAD in partially detached cells (Fig. 8J,L; Fig. S12B-B‴ and D-D‴) compared with wild type (Fig. 8G,I,K; Fig. S12A-A‴,C-C‴). Similarly, loss of Specc1l function caused the accumulation of membrane-associated β-catenin in migratory CNCCs (Fig. 8N-N″) compared with wild-type CNCCs (Fig. 8M-M″). We tested whether SPECC1L protein interacts with adherens junction proteins using U2OS cell lysate due to high expression in this cell line. SPECC1L and E-CAD were detected in the input of total protein lysate. E-CAD protein was pulled down with SPECC1L in a co-IP assay in a lysate immunoprecipitated by anti-SPECC1L antibodies, and no band was detected in the negative control (Fig. 8O).

The different mouse models of Twist1 loss of expression and function indicate the crucial role of this gene in regulating the development of CNCC-derived structures. To determine the relevance of these animal model findings to humans, we revisited a recently published GWAS meta-analysis of 8246 individuals of recent European ancestry, which identified 203 genome-wide significant signals associated with normal-range facial shape, as previously described by White et al. (2021). When moving from the minor (A) to major (T) allele of the leading SNP, the major shape changes include a less protrusive and wider nose, a shorter and less protrusive upper lip, and a more prominent chin and forehead. These changes are visible in the morphs and in a heatmap that shows the same effect using the normal displacement at each point comprising the facial surface (∼8000 points), with blue representing inward depression and red representing outward protrusion (Fig. 9A). A signal at 7p21.1 near TWIST1 was significantly associated facial shape and the lead SNP (rs212672) is located ∼355 kb downstream of TWIST1 (Fig. 9B). When the face was partitioned into global-to-local anatomical segments, the strongest association with the lead SNP was observed for the whole face (P=1.31E⁻²⁰) and for CNCC-derived facial regions involving the nose and upper lip (see rosette diagram in Fig. 9C).

Uncovering the underlying mechanism of Twist1 and Irf6 function in cell fate regulation is crucial to identify their regulatory pathways and associated genes that can control tissue development and differentiation. Our work tested the hypothesis that Twist1 is involved in neural tube formation and EMT of CNCCs, while Irf6 helps define neural tube dorsal edges and structural integrity. Our immunofluorescence staining data showed that TWIST1 expression has a similar pattern to β/δ-catenin and tight junction proteins at the apical side of neural plate and neural folds. This cellular overlap is validated by the dual immunofluorescence of TWIST1 and β-catenin. Furthermore, the in situ hybridization results showed that Twist1 mRNA is also weakly expressed in the dorsal edges of neural plate and apical side of neural folds. Our findings are consistent with a recent study indicating that a few cells of the neural folds express Twist1 mRNA when detected using single cell RNA-seq technology (Soldatov et al., 2019), although the scRNA-seq data relies on computational analysis to determine the spatial expression and the relative anatomical distribution of this population of cells (Soldatov et al., 2019). Additionally, Twist1 expression was detected in round cells residing in the neuroectoderm of the forebrain during the emergence of CNCCs at E8.5 (Füchtbauer, 1995).

To provide valuable insight on TWIST1 apical expression in the neural folds, we explored whether TWIST1 is co-expressed with LRP2 and RAB11b in endocytic vesicle compartments. We based our investigation on a previous report on LRP2 and RAB11b co-expression in endocytic compartments at the apical surface of neural folds (Kowalczyk et al., 2021). We found that TWIST1 is expressed in endocytic vesicles at the apical surface of neural folds, and its expression partially overlaps with RAB11b and LRP2. Recycling endosomes are important for apical constriction, and loss of Lrp2 leads to neural tube closure defects in mouse embryos (Kowalczyk et al., 2021). Similarly, loss of Twist1, β-catenin and δ-catenin in neuroectodermal cells leads to neural tube closure defects (Kowalczyk et al., 2021; Brault et al., 2001; Pieters et al., 2020). The cellular fractionation co-IP data shows that the cytosolic TWIST1 interacts with β/δ-catenins during neural tube closure, and Twist1 CKO using Wnt1-Cre disrupts the apical expression of β-catenin, which mostly becomes cytosolic. Notably, δ-catenin CKO by Wnt1-Cre also causes a focal loss of N-cadherin at the apical side of neural folds (Pieters et al., 2020). However, we cannot definitively disentangle the cytosolic versus nuclear contribution of Twist1 because it is a dynamic and continuous process, and the disruption of the neural tube formation might impact the delamination of the CNCCs. Therefore, our data suggests that the interaction of cytosolic TWIST1 with β/δ-catenins, and the expression of TWIST1 in endocytic vesicles at the apical surface might facilitate the apical constriction of neuroectodermal cells, whereas lack of Twist1 in neuroectodermal cells leads to an altered cell shape with multiple ectopic dorsolateral hinge points at the apical surface of the neural tube. Our findings are consistent with a recent publication showing that TWIST1 interacts with tight junction protein 1 (TJP1) along the cell membrane of two different cancer cells (Liu et al., 2022). Liu et al. also demonstrated that TWIST1 expression overlaps with TJP1 in vesicles at cellular membrane (Liu et al., 2022).

Twist1-null mice exhibited degeneration of the apical neuroectodermal cells, and increased cell death in migratory CNCCs (Bildsoe et al., 2009; Chen and Behringer, 1995; Soo et al., 2002). Although the adjacent cephalic mesenchyme activity contributes to neural fold elevation, it does not explain why the neural folds in Twist1-null embryos do elevate but fail to bend and close completely. Similarly, in Twist1^cko/− in CNCCs, the neural folds bend as if to join but stop prematurely, leading to improper closure of the neural tube. Although the phenotypes caused by a lack of Twist1 in CNCCs and mesoderm can be explained by an indirect role of Twist1 in regulating neural tube formation (Bildsoe et al., 2013), Twist1 expression in the neural folds provides a complementary mechanism for explaining the neural tube closure defects in Twist1-null embryos, the multiple ectopic bending points and the expansion in the neural tube of Twist1^cko/−. Although a complete loss of Twist1 did not disrupt the anterior-posterior patterning of the neural tube at E9.5, there was expansion in a few components of Shh and Fgf pathways (Soo et al., 2002). A direct role in the expansion of these components has not been previously reported because Twist1 expression was not described in the neural folds (Soo et al., 2002). Twist1 CKO in mesodermal cells disrupts neural fold elevation (Bildsoe et al., 2013), and the rescue experiment with wild-type mesenchymal cells in Twist1-null chimera embryos indicated that cephalic mesenchymal cells are involved in neural fold elevation (Chen and Behringer, 1995). For the function of IRF6 in neural tube, we have shown that overexpression of Irf6 in the basal layer of non-neural ectoderm suppresses Tfap2a in the ectoderm and causes neural tube abnormalities and exencephaly (Kousa et al., 2019). Our current findings show that AP2α expression was expanded into the neural tube in Irf6-null embryos, which could explain the disruption of neural tube edge integrity. In addition, the glial progenitor cells within the neural tube were disorganized in Irf6-null embryos with more noticeable gaps between the progenitor cells.

The Wnt1 ligand is a crucial regulator in the identification of neural plate border domains and in early induction of CNCCs (Parr et al., 1993). The Wnt1-Cre transgene drives the expression of Cre throughout the neural plate at E8.5 and later stages during neural tube development (Chang et al., 2015). Hence, Wnt1-Cre was ideal to delete Twist1 at early stages, if present, during early neural plate formation. Previous studies have suggested that Twist1 CKO in neuroectoderm by Wnt1-Cre does not disrupt CNCC migration towards the frontonasal and pharyngeal processes. However, the X-gal staining in Twist1-deficient CNCCs was reduced in frontonasal and pharyngeal arches, as Twist1 was considered a survival factor in migratory CNCCs (Chen et al., 2007, 2017; Zhang et al., 2012). Apoptosis in Twist1-deficient CNCCs might be one of the disrupted cellular functions behind the loss of craniofacial bone and cartilage. Yet no report has indicated whether mammalian Twist1 is involved in the events before CNCC delamination or whether the migratory Twist1-deficient CNCCs are normal mesenchymal cells. The data in this study from neural tube explants of embryos shows that Twist1 regulates EMT in CNCCs and that a loss of Twist1 in neuroectoderm leads to disruption of cell fate transition and delamination. Our findings also show that TWIST1 suppresses the expression of Irf6 and other epithelial factors during delamination. Irf6 is a terminal differentiation factor of proliferative epithelial cells in the ectoderm and its expression prevents EMT (Fakhouri et al., 2017; Ingraham et al., 2006; Thompson et al., 2019). In agreement with its role in EMT, Twist1 overexpression in the trunk neural tube converts the pre-EMT neuroectodermal cells into cells that resemble CNCCs (Soldatov et al., 2019). Recent studies on the EMT show that a complete conversion from an epithelial to a mesenchymal state is not necessary for cancer cells to migrate and multiple metastable states during cell transition have been described in cancer studies (Cousin, 2017; Nieto et al., 2016; Thiery et al., 2009). However, we did not observed a collective cell migration or epithelial-like cells in wild-type explants or in vivo histological staining. Lineage analysis may be the most appropriate way to detect the effect of Twist1 loss on target genes. Unfortunately, the lack of such analysis limits our interpretation of the impact of Twist1 on target gene expression in migratory CNCCs.

Determining the in vivo role of TWIST1 post-translational modification is vital in delineating its role in neural tube and CNCCs. TWIST1 is a phospho-protein and multiple phospho-residues are necessary for its stability, nuclear translocation and transcriptional regulation, as per previous cancer studies (Bourguignon et al., 2010; Zhao et al., 2017; Xu et al., 2017). Twist1 phosphorylation increases tumor cell motility in squamous cell carcinoma of the head and neck (Alexander et al., 2006; Su et al., 2011). In addition, overexpression of Twist1 phosphomimetic T125D; S127D led to limb abnormalities in transient transgenic mouse embryos (Firulli et al., 2007). We suggest that post-translational phosphorylation of TWIST1 is crucial for regulating its cellular activity in CNCCs. The Twist1^{S18;20A/S18;20A} and Twist1^S68A/S68A mutants embryos showed several craniofacial abnormalities, including epidermal blebbing, severe edema, meningeal detachment and neural tube defects. The skeletal staining of these phospho-incompetent mutant embryos showed bone loss and reduced skull mineralization of the frontonasal and maxillary bone. Yet the in vivo molecular function of the two phospho-sites needs to be delineated, including its importance in protein stability, nuclear translocation, the EMT process, cell survival and differentiation.

The data generated on the differentially expressed genes in Twist1^cko/− embryos demonstrated that many factors involved in intercellular adhesion proteins and cytoskeletal remodeling are significantly altered. Previous studies have shown that mutations in adhesion protein and cytoskeletal regulator genes cause neural tube closure defects and improper delamination of CNCCs, leading to craniofacial disorders (Pieters et al., 2020; Wilson et al., 2016). One of the genes contributing to this delamination defect in mice is Specc1l. Mutations in TWIST1 and SPECC1L lead to craniosynostosis and orofacial clefting in humans (Bai et al., 2021; Bhoj et al., 2019; Goering et al., 2021; Hall et al., 2020). Specc1l is a gene that encodes a novel coiled-coil domain-containing protein, which stabilizes microtubules and actin of the cytoskeleton (Bhoj et al., 2019; Gfrerer et al., 2014). SPECC1L colocalizes with both cellular actin and tubulin (Saadi et al., 2011; Wilson et al., 2016). Thus, without sufficient SPECC1L, actin-cytoskeleton reorganization and cell adhesion are significantly impacted (Saadi et al., 2011). The overlap in cellular function and embryonic phenotype between Twist1 and Specc1l mutant mouse lines suggests that both genes are involved in a similar regulatory pathway to control cytoskeleton reorganization during cell delamination and migration. This study shows that TWIST1 directly binds to putative regulatory elements in the intronic regions of Specc1l and that loss of Twist1 leads to a significant reduction of Specc1l in hindbrain and first pharyngeal arch tissues, suggesting that TWIST1 acts as an activator for Specc1l.

In conclusion, our findings emphasize the direct role of Twist1 and Irf6 in neural tube development, and the conservation of TWIST1 function in regulating the EMT process in CNCCs by controlling the expression of epithelial genes and cytoskeletal regulators. Specc1l and Twist1 play a similar role in the delamination of CNCCs by remodeling the cell-cell adhesion and cytoskeletal reorganization. Thus, previous reports and the current study provide a reasonable explanation for the overlapping phenotypes in mice and humans. In addition, as our results highlight, reanalysis of the GWAS data in humans shows that variants near TWIST1 impact normal-range facial shape, including CNCC-derived structures of the midface (Claes et al., 2018; White et al., 2021). Notably, the most significant SNP near TWIST1 overlaps with epigenetic enhancer signatures in which the elements with the epigenetic marks were tested for enhancer activities to recapitulate TWIST1 endogenous expression (Hirsch et al., 2018). Understanding the transcriptional regulation of TWIST1 and its function in controlling the associated regulatory pathway involved in cell fate determination could lead to the identification of the missing heritability in families that are affected and at high risk of craniofacial birth defects.

All the mice used in this study were generated using the C57BL/6J genetic background. Twist1 heterozygous mice were generated using EIIA-Cre mouse line using the B6;129S7-Twist1^fl/fl [obtained from the Mutant Mouse Resource and Research Center (MMRRC) supported by the National Institutes of Health). The mouse lines 129S4.Cg-E2f1^{Tg(Wnt1-cre2)Sor}/J (022137), BL6.129(Cg)-Gt(ROSA)26Sor^{tm4(ACTB-tdTomato,−EGFP)Luo}/J (007676) and BL6.129S4-Gt(ROSA)26Sor^tm1Sor/J (003474) were obtained from the Jackson Laboratory. The two different cell-tracing strategies R26^{tm4(ACTB-tdTomato,−EGFP)} and R26^tm1.lacZ were used to track in vivo CNCC formation and migration. Twist1^S18;20A/+, Twist1^S42A/+, Twist1^S68A/+ and Twist1^T121;S123A/+ phospho-incompetent founders were generated using CRISPR/Cas9 method at the Baylor College of Medicine. We backcrossed all phospho-mutant founders to C57BL/6J wild-type mice for six generations to avoid non-specific genomic alterations. Genomic DNA from the founders was sequenced to confirm the substitution of S18/20 and S68. Twist1^S18I;20A/+ and Twist1^S68A/+ heterozygous mice were crossed to obtain homozygous Twist1^{S18I;20A/S18I;20A} and homozygous Twist1^S68A/S68A embryos. These embryos were generated to determine the effects on the CNCC-derived craniofacial bone and cartilage, and the earliest time-point of detecting a craniofacial pathology. The animal work was approved by the Center for Laboratory Animal Medicine and Care committee at UT Health Houston under the Approved Animal protocol AWC-19-0045.

Embryos were extracted based on the presence of a copulation plug and the number of days after the last known delivery. Pregnant females were euthanized with CO₂ followed by cervical dislocation. Embryos were genotyped for Twist1^fl/fl, Twist1^+/−, Twist1^−/−, Twist1^fl/−, Wnt1-Cre1, Wnt1-Cre2, Irf6^−/+, Twist1^{S18I;20A/S18I;20A} and Twist1^S68A/S68A alleles via DNA extraction, allele-specific PCR and gel electrophoresis. The ratio of genotype-to-phenotype of Twist1 CKO and phosphor-incompetent mice was within the expected ratio for mouse.

Embryos of wild-type, Twist1^−/−, Twist1^cko/− and phospho-incompetent lines were collected for histological and immunofluorescence staining. Immunofluorescent staining was performed as previously described (Fakhouri et al., 2012). Briefly, mouse tissues were deparaffinized and rehydrated in a series of ethanol dilutions. The slides were boiled for 10 min in 10 mM sodium citrate buffer for antigen retrieval. Sections were blocked with anti-mouse IgG Fab fragment for 30 min, and then with 10% (v/v) normal goat serum and 1% BSA (v/v) in PBS for 1 h, then incubated overnight at 4°C with the following primary antibodies: mouse anti-AKT1/2/3 (1:150, Santa Cruz Biotechnology, sc-81434,), mouse anti-β-catenin (1:200, Santa Cruz Biotechnology, sc-7963), rabbit anti-β-catenin (1:150, Abcam, 16051), mouse anti-δ-catenin (1:150, BD Biosciences, 611537), rabbit anti-LRP2 (1:150, Abcam, ab76969), rabbit anti-RAB11b (1:150, Proteintech, 15903-I-AP), mouse anti-caludin 1 (1:200, Santa Cruz Biotechnology, sc-166338), mouse anti-N-cadherin (1:200, Santa Cruz Biotechnology, sc-59987), mouse anti-occludin (1:150, Santa Cruz Biotechnology, sc-133256), mouse anti-vimentin (1:150, Santa Cruz Biotechnology, sc-6260), mouse anti-E-cadherin (1:150, BD Biosciences, BD-610182), mouse anti-TWIST1 (1:200, Abcam, ab50887), SOX9 (1:200, Abcam, ab185966), mouse anti-ZO-1 (1:200, Santa Cruz Biotechnology, sc-33725), mouse anti-TFAP2α (1:50, DSHB, PCRP-TFAP2A-2C2), rabbit anti-CD73 (1:150, Abcam, ab133582), rabbit anti-CD206 (1:150, Abcam, ab64693), phalloidin (1:1000, Abcam, ab176757) and rabbit anti-IRF6 (1:500: Fakhouri et al., 2012). The secondary antibodies were goat anti-rabbit (1:150, A21429, Molecular Probes) and goat anti-mouse (1:150, A11029, Molecular Probes). We stained nuclei using DAPI (D3571, Invitrogen). An X-Cite Series 120Q laser and a CoolSnap HQ2 photometric camera (Andor Neo/Zyla) installed in a fluorescent microscope (Nikon Eclipse Ni) were used to capture images. For the immunofluorescence staining, sections from more than four embryos of each genotype, wild type and Twist1^cko/−, were used for each embryonic time point tested.

The Twist1 probe for in situ hybridization was designed to recognize a 0.8 kb region in the 3′ UTR region of mouse Twist1, starting at 188 bp and ending at 1062 bp after the Twist1 coding sequence ends. This avoids any conserved regions homologous with other bHLH factors and encompasses the sequence of the shorter probe described by Wolf et al. (1991). Further details are provided in Fig. S3.

R26^Tm1 transgenic mice were used to perform X-gal staining for cell tracing in whole-mount embryos. Whole-mount embryo staining at E9.5, E11.5 and E15.5 was performed as previously described (Metwalli et al., 2018). Stained wild-type and Twist1^cko/− embryos were then visualized, analyzed and compared using a stereomicroscope and NIS Elements AR software. The X-gal staining was performed on more than three biological replicates of each genotype and embryonic stage tested.

Twist1 phospho-incompetent embryos at E16.5 were dissolved in 2% KOH until the skin dissolved, and then stained with Alcian Blue solution overnight for cartilage and Alizarin Red for bone, as previously described (Thompson et al., 2019). The embryos were then submerged in 10% glycerol and 1% KOH to remove excess stain and remaining tissues for 3-5 days.

E8.5-E9.5 embryos from the Twist1^fl/−;Irf6^−/+;Wnt1-Cre2 crosses described previously were extracted from their amniotic sacs and maintained in an organ culture medium. Using microdissection instruments, E8.5-E9.5 embryos were removed from their amniotic sacs. Transverse cuts were made through the optic vesicle and first pharyngeal arch. After the removal of extraneous tissue, the mid and hindbrain regions were sectioned and cultured on collagen type I-coated petri dishes (10 µg/cm²) in alpha-MEM medium (10% fetal bovine serum, 100 U/ml penicillin G, 100 μg/ml streptomycin, and 1% non-essential and essential amino acids) and placed in a 5% CO₂ incubator at 37°C overnight to allow for appropriate adherence.

Time-lapse imaging of explants was performed using the A1R MP Confocal and Multiphoton Microscope (Nikon Instruments). Using overlaid contrast of red (TRITC), green (FITC) and blue (Hoechst) channels, fluorescence color changes from tissue explants were analyzed. Green fluorescence indicated the GFP reporter found only in CNCCs, while red fluorescence indicated all other tissues that express RFP, with Hoechst serving as a nuclei counterstain. Neural tubes from more than three biological replicates of each genotype were imaged in an environmentally controlled chamber maintained at 37°C and 5% CO₂, and the migration of CNCCs was captured using automated acquisition over a 21 h period. The net migration of cell bodies over a 10 h period was calculated using IMARIS software (Bitplane). Parameters included a spot diameter and connected components algorithm used for cell tracking. Cells also were tracked frame by frame using the manual-tracking spot feature of IMARIS. About 16 wild-type and 22 Twist1^cko/− CNCCs were tracked. A two-tailed Mann–Whitney U-test was used with a significance level ****P<0.0001.

We performed immunoblots to measure the quantitative levels of TWIST1 (ab50887), SNAIL2 (sc-166476), AKT1 (sc-81434), E-CAD (sc-8426), N-CAD (sc-59987), β-catenin (sc-7963) and WNT3 (sc-74537). mRNA expression of Twist1 and Irf6 in the hindbrain and first pharyngeal arch was also measured in three pooled biological and four technical replicates at E9.5. We also measured the mRNA level of these genes in addition to Pi3k, Ck2, Erk1, vimentin, occludin, Msx1, Tfap2a, Hand2, Yap, Wnt1, Rhoa, Rhoa, Arhgap29, Sox2, Sox10, Hoxd10, Jag1, Src and Mtor. The levels of phosphorylated and unphosphorylated forms of TWIST1 at S42, S68 and S127 were detected via immunoblot as previous described (Fakhouri et al., 2017; Hong et al., 2011). We received aliquots of rabbit polyclonal antibodies for TWIST1 P-S42 and P-S127 from Dr Brian A. Hemmings (Friedrich Miescher Institute for Biomedical Research, Basel, Switzerland; Vichalkovski et al., 2010). We purchased the monoclonal TWIST1 phospho-S68 from Abcam (ab187008). Monoclonal antibodies for TWIST1 (1:200, Abcam, ab50887) were used for immunofluorescence, immunoblotting and co-IP. We used the same antibodies previously described for E-CAD and β-catenin for immunofluorescence, western and co-IP. For the digestion with phosphatase enzyme, protein phosphatase 2A1 bovine (PP2A; Millipore-Sigma, P6993) was used to digest 100 µl total protein extracted from O9-1 CNCCs. About 10 µl PP2A in 50 mM TRIS-HCl buffer containing 2-mercaptoethanol was used and the solution was incubated at 30°C for 30 min and then analyzed by western blot using TWIST1 monoclonal antibodies (Abcam, ab50887).

Purified genomic DNA from mouse tail snips and purified PCR products were submitted for Sanger sequencing at Genewiz Company. We sequenced exon 1 of the Twist1-null and the two Twist1 phospho-incompetent mouse lines for S18/20 and S68 regions.

We extracted total protein from mid and hindbrain tissues dissected from embryos at E8.5-E9.0. Tissues were ground with a plastic pestle on dry ice and the samples were frozen and thawed twice before centrifugation to remove undissolved materials. Purified total protein was incubated with Protein A/G conjugated to magnetic beads. We used monoclonal antibodies (Abs) for TWIST1 (1:200, Abcam, ab50887), E-CAD (1:150, Santa Cruz Biotechnology, sc-8426), δ-catenin (1:150, BD Biosciences, 611537) and β-catenin (1:150, Santa Cruz Biotechnology, sc-7963). We performed cellular fractionation for total protein extracted from E8.5 and E9.5 hindbrain tissues using a cytoplasmic buffer [50 mM Tris-HCl (pH 8), 137.5 mM NaCl, 0.05% Triton X-100, 10% glycerol, 5 mM EDTA and Protease Inhibitor Cocktail (Thermo Scientific)] and a nuclear buffer [50 mM Tris-HCl (pH 8), 137.5 mM NaCl, 0.1% Triton X-100, 0.5% SDS, 10% glycerol, 5 mM EDTA and Protease Inhibitor Cocktail (Thermo Scientific)]. The cytosolic and nuclear fractions were incubated individually with Protein A/G conjugated to magnetic beads for the co-IP assay.

Chromatin immunoprecipitation (ChIP) was performed in tissues dissected from hindbrain and first pharyngeal arch. We used TWIST1 monoclonal antibodies to detect the binding to two putative enhancer elements within intron 1 and 2 of Specc1l. We also pulled down fragmented chromatins with IgG antibodies as a negative control for non-specific binding. The ChIP-PCR was performed as previously described (Kousa et al., 2019).

The β-actin and ribosomal protein S16 (Rps16) were used for normalization of western blot and RT-qPCR data as internal controls, respectively. A two-tailed Student's t-test analysis was applied for the statistical analysis between the wild-type control and Twist1 CKO quantitative data. The difference in average mean value was considered statistically significant if P<0.05.

We thank April Zhang and Yazan Hasan for their excellent help with the immunofluorescent imaging and histological staining of the mouse embryonic tissues. Special thanks to Derrick Thomas for his help with the revised manuscript. We are very grateful to Dr Brian A. Hemmings at the Friedrich Miescher Institute for Biomedical Research, Basel, who sent us sufficient aliquots of rabbit polyclonal antibodies for TWIST1 P-S42 and P-S127, and to Drs Pierre McCrea at MD Anderson Cancer Research Institute, Noriaki Ono at UT Health School of Dentistry at Houston and David Sheikh-Hamad for providing us with aliquots of the δ-catenin, rabbit β-catenin and LRP2 antibodies, respectively. We thank Dr Brendan Lee and Ms Racel Cela from Baylor College of Medicine for their tremendous help maintaining the Twist1 phospho-incompetent mouse founders at the Baylor mouse facility. We further thank the Genetically Engineered Mouse Core at Baylor College of Medicine for generating the Twist1 phospho-incompetent mouse lines.