Kindlin2 regulates neural crest specification via integrin-independent regulation of the FGF signaling pathway

Kindlin2 is an evolutionarily conserved focal adhesion protein that is involved in integrin activation and linking the extracellular matrix (ECM) and actin cytoskeleton. It contains a FERM domain, where ‘FERM’ represents four proteins that contain this domain: F for 4.1 protein, E for ezrin, R for radixin and M for moesin, all of which serve as membrane cytoskeleton linkers (Chishti et al., 1998). Kindlin2 binds to the cytoplasmic tail of integrin β subunits (Bledzka et al., 2012; Harburger et al., 2009; Raabe et al., 2008), which is crucial to integrin activation, and results in integrin ‘inside-out’ signaling that enhances the affinity of integrins for their ligands (Arnaout et al., 2005; Avraamides et al., 2008). Deletion of Kindlin2 in mice results in early embryonic lethality, and mice carrying only one copy of the kindlin2 (Fermt2) allele have aberrant angiogenesis (Montanez et al., 2008; Pluskota et al., 2011). Furthermore, embryonic stem cells isolated from Kindlin2-deficient mice barely adhere to the feeder layer, indicating that they have severely impaired integrin activation and cell adhesion (Bledzka et al., 2012; Montanez et al., 2008). Similar to the kindlin2-deficient mouse, kindlin2 knockdown Xenopus laevis also show defects in vascular maintenance and angiogenic branching during embryonic development. Furthermore, depletion of maternal Kindlin2 proteins caused arrested embryonic development at early cleavage stages (Rozario et al., 2014). Intriguingly, the expression pattern of kindlin2 in X. laevis embryos showed a strong tissue-specific expression in the neural crest (NC) (Canning et al., 2011). However, the function of Kindlin2 in NC development remains largely unknown.

The NC comprises a transient cell population induced at the neural plate border that is unique to vertebrate embryos. NC cells are multipotent stem cells (Baggiolini et al., 2015; Lignell et al., 2017) that give rise to a large variety of cell types, and the cranial NC cells preferentially contribute to mesenchymal cell types of facial cartilage and bone (Simoes-Costa and Bronner, 2016; Simões-Costa and Bronner, 2015; Soldatov et al., 2019). NC development can be broadly divided into formation, migration and differentiation, all of which are governed by a hierarchically arranged gene-regulatory network (Simões-Costa and Bronner, 2015). At the top of the hierarchy are signaling pathways, including the wingless-related integration site (Wnt), bone morphogenetic protein (BMP) and fibroblast growth factor (FGF) pathways. The combined action of these pathways induces the expression of a set of genes known as neural plate border specifiers, which define the neural plate border territory. Medial neural plate border specifiers, cooperating with signaling pathways that remain active during the time of specification, drive the expression of NC specifier genes such as Foxd3, Ets1, Sox9/10 and Snail1/2 (Mayor and Theveneau, 2013; Shellard and Mayor, 2019). NC specification is followed by the activation of the epithelial-to-mesenchymal transition machinery, which enables NC cells to detach from the epithelium and acquire migratory ability. Guided by environmental cues, the migratory NC cells express differentiation circuits and differentiate into diverse derivatives (Mayor and Theveneau, 2013; Taylor and LaBonne, 2007). Elucidating the multi-step gene-regulatory network that governs NC development is crucial to understanding the molecular mechanisms that underlie human diseases, such as Hirschsprung's disease (a congenital gut-motility disorder) and Treacher Collins syndrome (a facial dysmorphism) (Zhang et al., 2014).

In this study, we investigated the function of kindlin2 during NC formation using Xenopus embryos. We found that depletion of kindlin2 by either morpholino oligonucleotides (MOs) or the clustered regularly interspaced short palindromic repeats (CRISPR)/CRISPR-associated protein 9 (CRISPR/Cas9) technique inhibited the expression of NC specifier genes. Consistent with the observations in whole embryos, kindlin2 MOs inhibited NC formation induced by wnt3a and chordin in the animal cap assay. Furthermore, we showed that the FGF signaling pathway was interrupted by kindlin2 MOs in animal cap explants and that Kindlin2 can enhance the level of FGFR1 protein. In addition, integrin activation-deficient kindlin2 mutants also increased the FGFR1 protein level and rescued the phenotypes induced by kindlin2 MOs in Xenopus embryos. These findings suggest that the function of Kindlin2 during NC formation is independent of its integrin-activating ability. Collectively, our study uncovered a novel function of Kindlin2 in NC formation via regulation of the FGF signaling pathway during embryogenesis.

We examined the expression pattern of kindlin2 throughout embryogenesis in X. laevis embryos and found strong kindlin2 staining in the NC at the mid-neurula stages and late-neurula stages (Fig. S1), which is consistent with the previous studies (Canning et al., 2011). In the later developmental stages, the expression of kindlin2 became restricted to the head, cranial NC streams and somites (Fig. S1). The expression of kindlin2 in somites reinforces the view that Kindlin2 is essential for maintaining muscle integrity, owing to its ability to activate integrins (Dowling et al., 2008; Qi et al., 2015; Rogalski et al., 2000). However, the role of kindlin2 in NC development remains obscure.

To investigate whether Kindlin2 plays functional roles during NC formation, we sought to perform loss-of-function assays using MOs specifically targeting kindlin2 mRNA to block protein translation. Two MOs were designed: MO1, which targeted the 5′ UTR of kindlin2 mRNA; and MO2, which targeted the sequence flanking the start codon of kindlin2 mRNA (Fig. 1A). The knockdown efficiency of two MOs was assessed by western blotting, with proteins extracted from late-blastula stage embryos microinjected with either the mRNA of Myc-tagged kindlin2 (k2-Myc) alone or the k2-Myc mRNA mixed with MO1 or MO2. The results revealed that both low (500 pg/embryo) and high (1000 pg/embryo) doses of k2-Myc mRNA were well expressed in the embryos, but the k2-Myc expression was barely detectable in the presence of either MO1 or MO2 (Fig. 1B; Fig. S2A), indicating that both MOs effectively blocked the translation of k2-Myc mRNA. We then performed knockdown experiments with kindlin2 MOs to investigate the effects of kindlin2 knockdown on NC formation in Xenopus embryos. Following titration of the injection doses of the two MOs, MO1 (10 ng/embryo) and MO2 (20 ng/embryo) were microinjected into one dorsal blastomere of embryos at the four-cell stage, and lacZ mRNA was co-injected as the lineage tracer. The other dorsal blastomere was left uninjected to serve as a control. Whole-mount in situ hybridization revealed that the expression of three NC specifier genes foxd3, sox9 and snail2 was significantly decreased in both the MO1-and MO2-injected embryos (Fig. 1C-E). To confirm the specificity of the kindlin2 MOs, we performed rescue experiments with kindlin2 mRNA containing the coding sequence only (meaning that the MOs were unable to bind to it). When the kindlin2 mRNA was co-injected with MO1, the ratio of embryos exhibiting normal and mildly affected expression of foxd3, sox9 and snail2 was increased to 44%, 57% and 48%, respectively, compared with 0%, 33% and 18%, when MO1 was injected alone (Fig. 1C-E). Similarly, co-injection of kindlin2 mRNA with MO2 significantly increased the ratio of embryos showing normal and mildly affected expression of foxd3, sox9 and snail2 compared with injection with MO2 alone (Fig. 1C-E). Taken together, these results suggest that Kindlin2 is crucial for NC specification.

As NC development is a gradual process that originates from the neural plate border (NPB), we wondered whether the neural plate border specification is affected by MO-mediated knockdown of kindlin2. To test this possibility, we examined the expression of pax3, one of the specifiers of the NPB, in kindlin2 morphants at early neurula (stage 13). The expression of pax3 was decreased in the injected side compared with the uninjected side in all MO1-injected embryos and in 70% of MO2-injected embryos (Fig. 1F; Fig. S3A), suggesting that NPB specification was inhibited by loss of kindlin2. As MO1 also targets X. tropicalis kindlin2 (Fig. S2B), we injected MO1 into X. tropicalis embryos at the two-cell stage and examined pax3 (xtpax3) expression. We found that xtpax3 expression was also consistently suppressed in the injected side in X. tropicalis embryos (Fig. 1F; Fig. S3A). Taken together, these results indicate that loss of kindlin2 inhibits NPB specification.

To substantiate the effect of kindlin2 knockdown on NC specification, we followed the effect of kindlin2 knockdown on the development of NC-derived tissues. As cranial NC cells make a major contribution to the development of craniofacial cartilages, we next examined craniofacial morphogenesis in kindin2 morphants. The kindlin2 morphants that survived to early tadpole stages displayed abnormal head morphology. Alcian Blue staining demonstrated that the sizes of Meckel's, ceratohyal and ceratobranchial cartilages were clearly reduced at the injected side in the MO1 or MO2 morphants (Fig. 1G). These results are consistent with the inhibition of NC development in the early- and mid-neurula morphants shown above (Fig. 1C-F). Taken together, these data demonstrate that Kindlin2 is required in both the induction and specification processes during NC formation, which is fundamental to the normal development of NC-derived craniofacial cartilages.

Depletion of kindlin2 expression causes defects in cardiovascular development (Rozario et al., 2014), and in line with this observation, we found that the expression of nkx2.5, a marker gene for heart anlage, and dab2, a blood-vessel marker gene (Cheong et al., 2006; Shang et al., 2020), were repressed in X. laevis embryos injected with MO1 (Fig. S3B,C). This finding indicates that knockdown of kindlin2 led to defective heart and blood-vessel formation, again supporting the specificity of kindlin2 MOs.

To further validate the effect of kindlin2 deficiency on NC formation, we used the CRISPR/Cas9 technique to disrupt kindlin2 in X. tropicalis. We designed three single-guide RNAs (sgRNA) that target X. tropicalis kindlin2 (Fig. S4A; Fig. 2A). The T7 endonuclease 1 (T7E1) assay showed that all three sgRNAs could effectively disrupt kindlin2 when co-injected with Cas9 mRNA, with sgRNA1 displaying the highest mutagenic efficacy (Fig. S4B). Moreover, when sgRNA1 and Cas9 protein were microinjected into one blastomere at the two-cell stage, the gene disruption ratio was 37% revealed by the T7E1 assay (Fig. 2B). The injected embryos were raised to neurula stage, and 17% (12/70) of them showed reduced expression of foxd3 in the injected side (Fig. 2C). Accordingly, sgRNA1 was used to generate homozygous kindlin2 mutant X. tropicalis frogs by microinjecting sgRNA1 and Cas9 mRNA into X. tropicalis embryos at the one-cell stage. The resulting F0 frogs were crossed with wild-type frogs to obtain heterozygous kindlin2 mutant F1 offspring. In total, 28 (out of 60) F1 frogs carrying five different types of mutations (Fig. 2D) were obtained; five of these frogs carried frame-shift mutations, with three carrying a 10 bp deletion (hereafter referred to as k2 Δ10^+/−), and two carrying a 2 bp deletion. The k2 Δ10^+/− F1 siblings were mated to obtain k2 Δ10^−/−, i.e. homozygous kindlin2 mutant embryos (Fig. 2E), the genotype of which was confirmed by T7E1 assay (Fig. S5A,B). Furthermore, quantitative PCR (qPCR) analysis of embryos at stage 15 (when NC is specified) showed that the level of kindlin2 mRNA was substantially reduced in k2 Δ10^−/− embryos (Fig. 2F), which is likely due to nonsense-mediated mRNA decay (El-Brolosy et al., 2019; Ma et al., 2019). The mRNA level of one paralogue kindlin1 was decreased to a lesser extent (Fig. 2F). Another paralogue kindlin3 could not be detected by qPCR in both wild-type and k2 Δ10^−/− embryos at stage 15 (not shown). We next examined the expression of various NC regulators, including foxd3, sox9, pax3, ets1, tfap2a, msx1, snail1, id3, twist1, sox10, axud1 and c-myc, by qPCR. It revealed that the expression of foxd3, sox9, pax3, tfap2a, twist1, sox10 and c-myc was significantly reduced in k2 Δ10^−/− embryos compared with that in wild-type embryos (Fig. 2G), indicating that Kindlin2 is crucial for NC specification in X. tropicalis.

Having established that Kindlin2 is crucial for NC formation in Xenopus embryos, we next performed an in vitro animal cap assay to determine whether Kindlin2 deficiency affects NC induction in animal cap explants. An animal cap explant can differentiate into various tissues upon different morphogen treatment (Ariizumi et al., 2009), and Wnts mixed with BMP antagonists such as chordin or noggin can induce animal cap explants into NC-type tissues (Maharana and Schlosser, 2018; Sasai et al., 2001; Steventon et al., 2009; Wang et al., 2015; Zhao et al., 2008). We injected both blastomeres of X. laevis embryos at stage 2 with mixture of wnt3a (50 pg/embryo) and chordin (50 pg/embryo), dissected the animal caps at stage 9, and cultured them in vitro until the sibling embryos developed to stage 15, when the NC is specified (Fig. 3A). As expected, the expression of NC marker genes, including sox10, snail2, sox9, foxd3, pax3 and zic1, was efficiently induced in the animal caps dissected from the embryos co-injected with wnt3a and chordin. However, the induction of these NC marker genes was significantly reduced in the animal caps co-injected with kindlin2 MOs (Fig. 3B; Fig. S6A). This finding indicates that knockdown of kindlin2 inhibits NC induction in animal cap explants, and provides further evidence that Kindlin2 plays a role in NC specification.

Next, we sought to investigate the underlying mechanisms of kindlin2 knockdown on NC specification. NC formation depends on signals that arise from its surrounding tissues, including Wnt and BMP signaling in the surface ectoderm (Mayor et al., 1995; Simões-Costa and Bronner, 2015), and Wnt and FGF signaling in the underlying paraxial mesoderm (LaBonne and Bronner-Fraser, 1998; Mayor et al., 1997; Sauka-Spengler and Bronner-Fraser, 2008). As the Wnt and BMP signaling pathways are central to NC formation, we next examined both pathways in the animal caps. To examine whether knockdown of kindlin2 affected Wnt signaling, we performed animal cap assay with the embryos injected with wnt3a or wnt3a mixed with kindlin2 MOs. The expression of siamois and nr3, two target genes of the canonical Wnt signaling pathway, were examined by qPCR after in vitro culture for 3 h. The results revealed that the wnt3a induced expression of siamois and nr3 was increased in the animal caps co-injected with kindlin2 MOs (Fig. 3C), suggesting that knockdown of kindlin2 does not significantly interfere with Wnt signaling activation. As suppression of BMP signaling is required for NC formation, and BMP antagonists such as Chordin are capable of transforming the epidermal cell fate of animal caps into neural cell fate, we therefore tested whether knockdown of kindlin2 interferes with the inhibitory effects of chordin on BMP activity by examining the neural marker genes. To this end, the animal caps were microinjected with either chordin or chordin mixed with kindlin2 MOs. The injected animal caps were then cultured in vitro until their sibling embryos developed to stage 20 when neuralization is completed. The qPCR results showed that expression of sox2 and sox3 was induced by chordin, and slightly increased by co-injection of chordin and kindlin2 MO2 (Fig. 3D). These results suggest that knockdown of kindlin2 does not alter the inhibitory effect of chordin on BMP signaling.

NC formation depends on signals secreted from its surrounding tissues. It is worth noting that not only was the NC suppressed in the kindlin2 morphants, but its surrounding tissues were also disturbed, as indicated by the reduction in the expression of the paraxial mesodermal maker gene myod (Fig. S6B). This observation prompted us to test whether knockdown of kindlin2 affected FGF signaling, another essential signaling pathway for NC specification (Barriga et al., 2015; Mayor et al., 1997). To this end, the animal caps were microinjected with efgf mRNA, which induced the expression of the FGF signaling target genes xbra, egr1 and myod. The expression of these target genes was significantly decreased in the animal caps microinjected with efgf and kindlin2 MOs (Fig. 3E; Fig. S6C). Moreover, the phosphorylation of extracellular signal-regulated kinase (ERK) induced by injection of efgf mRNA into animal caps was also suppressed in the presence of kindlin2 MOs (Fig. 3F). Collectively, these results suggest that knockdown of kindlin2 interrupts FGF signaling during NC specification.

The small guanine triphosphatase Ras functions as a binary molecular switch, and mediates signal transduction between the adaptor complex that binds to FGFRs and the mitogen-activated protein kinase kinase kinase (MAPKKK) (Ornitz and Itoh, 2015). To decipher how kindlin2 might regulate the FGF signaling pathway, we examined whether kindlin2 affects Ras signaling by microinjecting constitutively active HaRas alone or in combination with kindlin2 MOs into animal cap explants. As expected, the expression of the FGF target gene xbra was efficiently induced by the constitutively active HaRas (20 pg/embryo) (Fig. 3G) (Whitman and Melton, 1992; Zhao et al., 2008). In animal caps co-injected with HaRas and kindlin2 MOs, the expression of xbra remained unchanged (Fig. 3G), suggesting that FGF signaling downstream of Ras activation is unaffected by loss of kindln2. Moreover, HaRas rescued the NC defects in kindlin2 morphants (Fig. 3H,I), suggesting that kindlin2 knockdown affects NC formation occurring upstream of Ras activation in the FGF signaling cascade.

We then turned our attention to the FGFRs upstream of FGF signaling, and investigated a possible interaction between Kindlin2 and FGFR. There are four FGFRs (FGFR1-FGFR4) in X. laevis, and FGFR1 is crucial to the development of the NC (Brewer et al., 2016; Monsoro-Burq et al., 2003). We noticed that when FLAG-tagged Xenopus Kindlin2 and Myc-tagged FGFR1 were expressed in HeLa cells, a large fraction of Kindlin2-FLAG signals was found to overlap with the FGFR1-Myc signals, either on or near the cell membrane (Fig. 4A). We next examined whether Kindlin2 physically interacted with FGFR1 using a co-immunoprecipitation (co-IP) assay. Frs2, a known FGFR1-interaction partner, served as the positive control. HEK293T cells were transfected with the plasmids kindlin2-FLAG, FGFR1-Myc or Frs2-FLAG alone, with a combination of kindlin2-FLAG and FGFR1-MYC, or with a combination of FGFR1-Myc and Frs2-FLAG. FGFR1-Myc was co-precipitated with FRS2-FLAG, as previously reported (Fig. 4B) (Burgar et al., 2002). However, FGFR1-Myc was not detected in Kindlin2-FLAG precipitants (Fig. 4B), and FGFR1-FLAG was not detected in Kindlin2-Myc precipitants either (Fig. 4C; Fig. S7A). Thus, our data suggest that Kindlin2 interacts with FGFR1 indirectly, although they are localized closely and cannot be distinguished by standard confocal imaging. Notably, the expression of FGFR1-FLAG was enhanced in HEK293T cells co-transfected with kindlin2-Myc compared with that in HEK293T cells transfected with FGFR1-FLAG alone (Fig. 4C). We next transfected HEK293T cells with FGFR1-FLAG and increased doses of kindlin2-Myc. The plasmid pCS2-Myc was used to adjust the total amount of DNA for transfection. Western blotting showed that exogenous FGFR1-FLAG expression was increased gradually by increasing amounts of Kindlin2-Myc (Fig. 4D). Similarly, endogenous FGFR1 expression was enhanced by transfection of kindlin2-Myc in a dose-dependent manner (Fig. 4E). In line with this observation, the expression of phosphorylated ERK (pERK) was also increased (Fig. 4D,E), suggesting that the activity of the FGF pathway was increased. However, qPCR analysis showed that the mRNA level of FGFR1 remained unchanged (Fig. 4F).

We next tested the effect of kindlin2 overexpression on the FGFR1 protein level in Xenopus embryos. Consistent with the results obtained in HEK293T cells, overexpression of kindlin2 in Xenopus embryos enhanced the protein expression of both exogenous FGFR1-FLAG and endogenous Fgfr1 in a dose-dependent manner (Fig. 4G,H). Conversely, knockdown of kindlin2 reduced the protein expression of both exogenous FGFR1-FLAG and endogenous Fgfr1 in Xenopus embryos (Fig. 4I,J). Taken together, these results suggest that Kindlin2 is involved in regulating the protein level of FGFR1.

To determine whether the enhanced protein level of FGFR1 in the presence of kindlin2 overexpression is due to increased protein stability, we measured FGFR1 protein stability following inhibition of protein synthesis by cycloheximide (CHX). Western blotting revealed that the FGFR1-FLAG protein level was markedly increased when kindlin2-Myc was overexpressed in HEK293T cells treated with CHX (Fig. 4K), suggesting that Kindlin2 plays a role in stabilizing FGFR1. Nedd4 and c-Cbl, two E3 ubiquitin ligases, mediate ubiquitylation of FGFR1 and promote FGFR1 degradation in lysosomes (Haugsten et al., 2008; Persaud et al., 2011; Wong et al., 2002). This led us to test whether Kindlin2 regulates FGFR1 protein stability by interacting with Nedd4 or c-Cbl. However, no interaction was detected between either Kindlin2 and Nedd4 or Kindlin-2 and c-Cbl in co-IP assays (Fig. S7B,C), suggesting that Kindlin2 is not directly involved in FGFR1 ubiquitylation. We then performed a cell-surface biotinylation assay to extract membrane-bound FGFR1 in these cells. Western blotting quantification revealed that membrane-bound FGFR1-Myc was increased in HeLa cells co-transfected with kindlin2 compared with that in HeLa cells transfected with FGFR1-Myc alone (Fig. 4L). In the future, it would be interesting to examine whether Kindlin2 regulates FGFR1 endocytosis and recycling to the cell membrane.

As Kindlin2 is renowned for its ability to activate integrins (Rognoni et al., 2016), we next examined whether the role of Kindlin2 in NC formation is dependent on its integrin-activating ability. Analysis of the human KINDLIN2 protein structure revealed that Gln⁶¹⁴ and Trp⁶¹⁵ in the F3 subdomain are crucial to integrin binding, and that a mutant in which Gln⁶¹⁴ and Trp⁶¹⁵ were substituted by Ala (Q614A/W615A) failed to bind β1 integrin (Shi et al., 2007). Given that the kindlin2 F3 subdomain is highly conserved between human and Xenopus, with 96.4% sequence similarity (Fig. S8B), we introduced these Ala-substitution mutations into Xenopus kindlin2 (hereinafter referred to as kindlin2 QW/AA). We expressed Xenopus kindlin2 wild type or QW/AA mutant in kindlin2 KO HT1080 cells, to assess their abilities to restore kindlin2-mediated cell shape modulation. Typically, kindlin2 KO HT1080 cells have a shrunken morphology (cell I in Fig. 5A1) (Liu et al., 2021). However, the GFP-tagged Xenopus kindlin2 wild type, but not the QW/AA mutant, restored normal cell morphology (Fig. 5A-B1). Moreover, the GFP-tagged Xenopus kindlin2 QW/AA mutant, unlike the GFP-tagged Xenopus kindlin2 wild type, failed to localize to focal adhesions marked by paxillin (Fig. 5A-B1). We also overexpressed Xenopus kindlin2 wild type or kindlin2 QW/AA alone, or in combination with talin-FERM in αIIbβ3 integrin-expressing CHO A5 cells, and performed an integrin activation assay. Similar to mouse Kindlin2 (Li et al., 2017), Xenopus kindlin2 wild type synergized with talin for integrin activation (Fig. 5C). However, the Xenopus kindlin2 QW/AA mutant showed significantly reduced synergistic enhancement of integrin activation (Fig. 5C). Collectively, these results indicate that the Xenopus kindlin2 QW/AA mutant is unable to bind and activate integrins.

We next tested whether the integrin-binding-defective kindlin2 mutant reversed the effect of kindlin2 knockdown on NC formation in Xenopus embryos. To this end, we injected one dorsal blastomere of X. laevis embryos at the four-cell stage with kindlin2 MO1, or kindlin2 MO1 combined with mRNA of either Xenopus kindlin2 QW/AA or mRNA of the mouse kindlin2 L675E mutant, which also fails to localize to focal adhesion and activate integrins (Li et al., 2017). Whole-mount in situ hybridization showed that similar to the kindlin2 wild type (Fig. 1C,E), both Xenopus kindlin2 QW/AA (Fig. 5D-F) and mouse kindlin2 L675E (Fig. 5G,H) partially rescued the kindlin2 MO-induced decreased expression of the NC marker genes foxd3 and snail2. Moreover, the ratio of kindlin2 morphants rescued by kindlin2 QW/AA mRNA was similar to that rescued by kindlin2 wild-type mRNA (Fig. 1C,E). Taken together, these results suggest that the role of Kindlin2 in NC specification is independent of its integrin-activating ability. Furthermore, this function seems to be conserved across species, as the mouse integrin-binding-defective mutant reversed the kindlin2 deficiency-induced NC defects in Xenopus embryos. Furthermore, overexpression of kindlin2 QW/AA, like that of wild-type Kindlin2 (Fig. 4G,H), enhanced the level of both exogenous and endogenous FGFR1 protein in Xenopus embryos (Fig. 5I,J), suggesting that the integrin-activating ability of Kindlin2 is not involved in the Kindlin2-mediated regulation of FGFR1.

While co-injection of wild-type kindlin2, kindlin2 QW/AA mRNA or even mouse kindlin2 L675E mRNA with kindlin2 MOs, reversed the kindlin2 MO-induced downregulation of NC-specific gene expression (Fig. 1C-E; Fig. 5D-H), overexpression of any of these alone in Xenopus embryos had little effect on the expression of NC specifiers (Fig. 6A; Fig. S8A,C,D). These findings suggest that the overexpression of kindlin2 does not substantially affect the NC specification process. Given that NC cells possess several integrins, including integrins α4β1 and α5β1, which mediate NC cell adhesion and migration, and convey survival signals (Bronner-Fraser, 1986; Testaz and Duband, 2001), we next investigated the effect of kindlin2 on the early tailbud stages, when migration has already commenced. The expression of twist, which represses the expression of E-cadherin in delaminating NC cells and is crucial for NC migration (Simões-Costa and Bronner, 2015), was used to indicate NC migration. It revealed that microinjection of wild-type kindlin2 mRNA, but not kindlin2 QW/AA mRNA, inhibited NC cell migration (Fig. 6B,C). This result suggests that a certain level of kindlin2 is essential for the migration of NC cells, which likely occurs via an integrin-mediated mechanism.

Kindlin2 is an evolutionarily conserved protein that is essential in cell-ECM adhesion (Zhan and Zhang, 2018), and its function has been studied in both invertebrate and vertebrate model organisms. Loss of unc112, an ortholog of kindlin2, in C. elegans results in the most severe Pat (paralyzed, arrested elongation at twofold) phenotype (Williams and Waterston, 1994), which is due to impaired cell-ECM adhesion and integrin function (Rogalski et al., 2000). Moreover, studies in vertebrates such as mice, zebrafish and Xenopus have revealed that Kindlin2 is crucial to the development and function of germ layers (Montanez et al., 2008) and parts of organs, including cardiac muscle (Dowling et al., 2008; Qi et al., 2015), and angiogenesis (Pluskota et al., 2011; Rozario et al., 2014). However, Kindlin2-deficient mouse embryos die early (Montanez et al., 2008) and, to date, less is known about the function of Kindlin2 during vertebrate embryogenesis. In this study, we used the Xenopus model system to follow early embryonic development, as local loss- and gain-of-function experiments can be performed by targeted injection of Xenopus embryos, which are not subject to the early lethality observed in Kindlin2-deficient mouse embryos (Montanez et al., 2008). We demonstrated that Kindlin2 is required for NC formation and, mechanistically, we identified that Kindlin2 enhances FGFR1 stability and promotes FGF signaling, independently of Kindlin2 binding to and activating integrins. Our findings identify a novel function of Kindlin2 in NC induction and specification, and provide the first mechanistic explanation of this process.

The NC is a cell population of particular importance in embryos, as it has the potential to give rise to a wide range of cell types, such as cranial facial cartilages, peripheral neurons and melanocytes (Mayor and Theveneau, 2013). The NC originates from the neural plate border, where NC cells are induced between the neural and non-neural ectoderm during gastrulation and neurulation. Emerging evidence shows that various cellular and molecular mechanisms underlie NC formation, and highlight a level of transcriptional regulation and cell signaling interactions (Mayor and Theveneau, 2013; Shellard and Mayor, 2019). Through loss-of-function studies, we demonstrated that Kindlin2 is required for NC induction and specification in X. laevis. Knockdown of kindlin2 with MOs results in inhibition of NC development, as revealed by the decreased expression of the NC-specifier genes foxd3, sox9 and snail2 in the region (Fig. 1C-E), which is rescued by kindlin2 mRNA. Moreover, the expression of pax3, a specifier of the neural plate border, was also decreased in the kindlin2 morphant, suggesting that Kindlin2 may be involved in NC induction. A previous study reported the effect of kindlin2 knockdown with MOs in X. laevis (Rozario et al., 2014), but did not show the effect on NC formation that was observed in our study. We performed localized microinjection to inject one dorsal blastomere of a four-cell stage embryo with kindlin2 MOs. This method allowed us to follow the effect on NC formation and monitor the parallel control in the same embryo, with limited effects on other tissues. In addition, we selected different target sequences of kindlin2 MO from those used in the previous study (Rozario et al., 2014). Two MOs were designed to bind to the sequence located at the 5′-UTR of kindlin2 and the sequence spanning the 5′-UTR (8 bp), and the beginning of the coding sequence (17 bp). Both MOs effectively and specifically knocked down kindlin2 in Xenopus embryos (Fig. 1B-F; Fig. S2). Moreover, MO-mediated knockdown of kindlin2 inhibited the NC formation induced by wnt3a and chordin in the animal cap assay (Fig. 3B; Fig. S6A).

Consistent with the inhibitory effects of kindlin2 deficiency on NC development in early- and mid-neurula kindlin2 morphants, the kindlin2 morphants, which survive and develop to early tadpole stages, displayed severe facial cartilage defects (Fig. 1G). In line with the effects of kindlin2 deficiency in X. laevis embryos, depletion of kindlin2 in X. tropicalis by CRISPR/Cas9 also resulted in a reduction in NC specification in F0 embryos (Fig. 2C). Furthermore, the embryos of homozygous F2 kindlin2 deletion (k2 Δ10^−/−) mutants at stage 15 showed significantly reduced expression of a variety of NC regulators (Fig. 2G). Collectively, these data support the notion that Kindlin2 plays an essential role in NC formation (Fig. 7A).

The animal cap assay results suggest that the effect of Kindlin2 on NC formation occurs via the FGF signaling pathway, which plays an essential role in NC formation in Xenopus and in the mouse. Conditional inactivation of FGFR, ERK, the upstream elements of ERK (B/C-Raf and MEK1/2) or the downstream effector serum response factor in the developing NC cells in mice results in craniofacial anomalies (Newbern et al., 2008; Wang et al., 2013). In kindlin2 morphants, the NC was suppressed, and the paraxial mesoderm, which secrets FGF-signaling triggers, was also disturbed (Fig. S6B) (Sauka-Spengler and Bronner-Fraser, 2008). Furthermore, knockdown of kindlin2 was found to attenuate the expression of FGF target genes induced by overexpression of efgf, such as egr1, xbra and myod (Fig. 3E; Fig. S6C). Concurrently, the efgf-induced phosphorylation of mitogen-activated protein kinase was also reduced in response to kindlin2 knockdown (Fig. 3F). We discovered that the FGFR1 protein level was increased following overexpression of kindlin2 in both human HEK293T cells and Xenopus embryos (Fig. 4D,E,G,H), and was reduced when kindlin2 is knocked down (Fig. 4I,J). Thus, it is attractive to propose that Kindlin2 regulates FGF signaling at least in part by stabilizing FGFR1. Clearly, future studies are required to further investigate the underlying mechanisms of these processes.

Kindlin2 has key roles in integrin activation, consistent with its roles in cell-ECM adhesion (Zhan and Zhang, 2018). However, emerging evidence shows that kindlin2 has novel functions that are independent of its integrin-binding and integrin-activation abilities (Guo et al., 2015; Wei et al., 2013; Yu et al., 2013, 2012). In this study, we demonstrated that two integrin-activation-defective mutants of kindlin2 (Q614A/W615A and L675E), like wild-type kindlin2, reversed the decreased expression of the NC specifier genes that were induced by kindlin2 knockdown (Fig. 5D-H). Moreover, FGFR1 protein expression was enhanced by overexpression of kindlin2 QW/AA (Fig. 5I,J). However, overexpression of kindlin2 did not substantially affect NC specification, but instead inhibited the expression of twist, which is involved in the regulation of NC migration (Fig. 6B,C). This effect appears to be dependent on integrin signaling, as overexpression of integrin-binding-defective mutant kindlin2 fails to do so. As it is well established that integrins play essential roles in cell movement (De Pascalis and Etienne-Manneville, 2017; Hamidi and Ivaska, 2018), we speculate that the overexpression of kindlin2 causes overactivation of integrins, and therefore impairs NC movement, as reflected by decreased twist expression (Fig. 6B,C; Fig. 7B). Although the function of Kindlin2 in integrin activation is well recognized, the dissection of its integrin-independent function is still in the preliminary stages. This represents the first report of a role for Kindlin2 in NC specification, which provides an entry point for Kindlin2-based research of NC development. How Kindlin2 mechanistically regulates each step of NC development is an important scientific question that awaits further investigation.

Morpholinos (MOs) were purchased from GeneTools and had the following sequences: MO1, 5′-TCACGTCACTGCTTCTCAATCTGGT-3′; MO2, 5′-ATACCATCCAAAGCCATGATTCCT T-3′. The following plasmids were used as templates for the synthesis of mRNAs: pCS2⁺-xlkindlin2 and pCS2⁺-xlkindlin2-3′-6Myc, subcloned from pCMV-Sport6-xlkindlin2 (Source BioScience, #IRBHp990C0933D); pCS2⁺-xlkindlin2-QW/AA and pCS2⁺-xlkindlin2-QW/AA-3′-6Myc, generated using PCR-based site-directed mutagenesis (Laible and Boonrod, 2009); and pCS2⁺-hsFGFR1-3′-2FLAG subcloned from pWZL-hsFGFR1-Flag (Addgene, #20486). The plasmid pCS2-mkindlin2-L675E was constructed by PCR, using mkindlin2-L675E-GFP as a template (Li et al., 2017). All of the above constructed plasmids were confirmed by sequencing.

X. laevis and X. tropicalis juveniles or adult frogs were purchased from Nasco. The use of frogs for this study was approved by the Ethics Committee of The Chinese University of Hong Kong and licensed by the Department of Health, Hong Kong. Embryos were obtained by in vitro fertilization or natural mating, according to the published protocols (Sive et al., 2000) and staged according to Nieuwkoop and Faber (1975). Microinjection was performed bilaterally at stage 2 for western blotting and animal cap assays, or unilaterally at stage 3 for whole-mount in situ hybridization and Alcian Blue staining. Animal caps were dissected at stage 9 using forceps, and were cultured in L-15 medium [67% L-15, 7.5 mM Tris-HCl (pH 7.5), 1 mg/ml bovine serum albumin] (Zhao et al., 2008) at room temperature for 4 h or in a 16°C incubator until the stages indicated, before collection.

We titrated a series of injection doses (5, 10 and 20 ng/embryo) of each MO to find the most suitable dose for Xenopus embryos. This minimized the non-specific effects induced by MO or the effects due to the innate immune response to MO injection in embryos (Eisen and Smith, 2008; Gentsch et al., 2018). For MO1, 10 ng/embryo was selected, as the 5 ng/embryo exhibited only slight effects on the expression of NC markers and the 20 ng/embryo had a high ratio of lethality before neurulation. For MO2, the 20 ng/embryo was selected, because its penetrance was lower than that of MO1.

For β-galactosidase staining, the embryos were fixed at desired stages using HEMFA buffer [0.1 M 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid (pH 7.4), 2 mM ethylene glycol-bis(β-aminoethyl ether)-N,N,N′,N′-tetraacetic acid, 1 mM MgSO₄ and 4% formaldehyde] for 1 h at room temperature. Embryos were washed three times with 1× phosphate-buffered saline, and were then stained [40 mg/ml 5-bromo-4-chloro-3-indolyl-β-D-galactopyranoside, 0.5 M K₃Fe(CN)₆, 0.5 M K₄Fe(CN)₆ and 100 mM MgCl₂] until color appeared (Wang et al., 2015). After re-washing and re-fixation, embryos were dehydrated twice with absolute ethanol for 5 min each, and then stored at −20°C in absolute ethanol. Whole-mount in situ hybridization was performed as described previously (Wang et al., 2015, 2019). For Alcian Blue staining, kindlin2 morphants were fixed at approximately stage 46 using HEMFA. After dehydration with absolute ethanol, embryos were stained in Alcian Blue solution (30% acetic acid, 70% ethanol and 20 mg Alcian Blue) for 3 days. After washing with 95% ethanol and clearing in glycerol/KOH, the cranial cartilages were dissected manually with forceps and imaged with a Nikon SMZ1270 stereomicroscope.

HEK293T and HeLa cells were cultured in Dulbecco's modified Eagle's medium with 10% fetal bovine serum in a culture incubator at 37°C under 5% CO₂ as described by Li et al. (2018). Transfection was preformed using Lipofectamine 2000 (Invitrogen), following the manufacturer's instructions. Cells were collected 24-48 h after transfection, as indicated, for protein or RNA extraction.

Proteins were extracted from embryos, as outlined in our previous study (Wang et al., 2015). Dynabeads Protein G (Invitrogen, 10004D) were used to capture the antibody-protein complexes. Membrane-protein extraction was performed using the cell-surface protein biotinylation and isolation kit (Thermo Fisher Scientific, A44390). HeLa cells transfected with FGFR1-Myc alone or mixed with kindlin2 were incubated with biotin. Cell lysates were incubated with Neutravidin resin that can capture the biotinylated membrane proteins. After washing, the proteins were eluted by DTT solutions. Lysates from whole embryos, animal caps, cells and immunoprecipitated samples were resolved by sodium dodecyl sulfate–polyacrylamide gel electrophoresis, transferred to polyvinylidene difluoride membranes and analyzed by immunoblotting. The antibodies used in this study are listed in Table S2.

Total RNAs were extracted from animal caps, cells or whole embryos using TRIzol reagent (Invitrogen, #15596-026). cDNA was generated by PrimeScript RT Reagent Kit (TaKaRa, #RR037A). PCRs were performed using cDNA as a template, on either a general PCR machine (animal cap assay) or on an ABI 7900HT Fast Real-Time PCR instrument (ThermoFisher). The primer pairs used in this study are listed in Table S1.

Kindlin2 KO HT1080 cells stably expressing paxillin-mRuby2 were transfected with either wild-type Xenopus kindlin2-GFP or Xenopus kindlin2 QW/AA-GFP plasmids. Cell images were captured by a Nikon TiE microscope and an Andor iXon DU-897 EMCCD (Andor), as we have described recently (Liu et al., 2021). Confocal microscopy was used to reveal the cellular localization of Xenopus Kindlin2 and FGFR1, and was performed using the same method as outlined previously (Shi et al., 2015). HeLa cells were transfected with kindlin2-FLAG and FGFR1-Myc plasmids. Fluorescent images were captured by a Leica SP8 confocal microscope after immunostaining.

Integrin-activation assays were performed as previously described (Li et al., 2017). Xenopus kindlin2-GFP, kindlin2-QW/AA-GFP, RFP-tagged talin-FERM or a mixture of talin-FERM and individual Xenopus kindlin2 constructs were transfected into CHO-A5 cells harboring stable expression of αIIbβ3. The transfected cells were incubated with anti-PAC-1 (BD) antibody, and then stained with Alexa Fluor 633-conjugated goat anti-mouse IgM (Invitrogen). After staining, the cells were analyzed using a BD FACSCanto Flow Cytometer. Integrin activation was represented by the relative median fluorescence intensities after normalization to the basal PAC1-binding in control cells (CHO-A5).

CRISPR/Cas9-mediated mutant X. tropicalis was generated, and the T7E1 assay was performed as described previously (Liu et al., 2016).

Statistical analysis was performed using either an unpaired two-tailed chi-squared test or an unpaired two-tailed t-test. Data represent mean±s.d. or mean±s.e.m., and P<0.05 was considered statistically significant (*P<0.05, **P<0.01; n.s., no significant difference).

We thank Prof. Igor Dawid, Eddy De Robertis, Richard M. Harland, Christof Niehrs, Tomas Pieler, Chuanyue Wu, Yonglong Chen and Cong Yu for the reagents used in this paper. We appreciate the technical assistance of Mr. Chi Bun Wong in this study.

The peer review history is available online at https://journals.biologists.com/dev/article-lookup/doi/10.1242/dev.199411