Disruption of the nectin-afadin complex recapitulates features of the human cleft lip/palate syndrome CLPED1

Orofacial clefting is a common congenital defect affecting nearly 200,000 infants each year. These frequently-monogenetic developmental malformations usually include cleft lip and/or cleft palate (CP), occurring alone or as part of a syndrome. Cleft lip/palate-ectodermal dysplasia syndrome 1 (CLPED1; OMIM #225060) – also known as orofacial cleft 7 (OFC7), Zlotogora-Ogur syndrome and Margarita Island ectodermal dysplasia (ED4) – was originally identified as a recessive disorder affecting isolated populations in the Middle East, South America and the Caribbean (Bustos et al., 1991; Rodini and Richieri-Costa, 1990; Zlotogora, 1994; Zlotogora and Ogur, 1988). Homozygous mutations in NECTIN1 that truncate the receptor within its extracellular domain at Trp185 were identified in two families with CLPED1 (Suzuki et al., 2000); however, mouse models of Nectin1 loss do not appear to cause CP or ED (Barron et al., 2008; Inagaki et al., 2005; Yoshida et al., 2010).

With its similarity to human craniofacial development and genetic malleability, the mouse is an excellent model for studying palatogenesis (Bush and Jiang, 2012). There are ∼50 genes linked to orofacial clefting in humans and >100 in mice, and many of these are orthologs (Dixon et al., 2011). In murine palatogenesis, palatal shelves (PSs) emerge at E11.5, extend downward between E12 and E13.5, rapidly elevate over the tongue at ∼E14, and elongate horizontally, fusing along the midline by E15.5 (Bush and Jiang, 2012; Greene and Pratt, 1976; Walker and Fraser, 1956). PSs are lined by a bilayered epithelium consisting of basal cells overlaid with protective periderm, which prevents the formation of intraoral adhesions between juxtaposed epithelia, such as the PS and tongue (Hammond et al., 2019; Richardson et al., 2009). As they approach the midline, opposing PS epithelia adopt an adhesive-competent identity as the medial edge epithelia (MEE). They form the midline (medial) epithelial seam (MES) during fusion, with residual epithelial cells removed through a combination of migration, apoptosis and differentiation (Hammond et al., 2019; Kim et al., 2015). Failure at any step can result in CP, perhaps explaining its high incidence and genetic heterogeneity (Dixon et al., 2011; Kousa et al., 2017; Mossey et al., 2009).

Nectins are transmembrane cell-cell adhesion proteins – consisting of four members in mice and humans – that homodimerize in cis and generally heterotetramerize in trans (Momose et al., 2002; Yasumi et al., 2003). Different nectin homodimers interact in trans with variable affinity; e.g. nectin 1 preferentially binds to nectin 3 and nectin 4 (Reymond et al., 2001). This hierarchical binding affords the nectins cell sorting functionality, as demonstrated in the olfactory epithelia and cochlea (Katsunuma et al., 2016; Togashi et al., 2011). The nectin cytoplasmic tail forms an obligate interaction with the actin-binding protein afadin, which links the cytoskeleton to the adherens junction (AJ) (Mandai et al., 1997; Miyahara et al., 2000; Takahashi et al., 1999).

Recent work suggests that an interaction between nectin 1 and nectin 4 may be important for palatogenesis (Mollo et al., 2015; Richardson et al., 2017). Interestingly, mutations in NECTIN4 underlie a related developmental disorder, ectodermal dysplasia syndactyly syndrome (EDSS1; OMIM #613573) (Brancati et al., 2010). However, Nectin4 loss has not been modeled in mice, so its role in palatogenesis remains unknown. Here, using a versatile in utero lentiviral genetic toolkit, we provide the first functional evidence that the nectin-afadin cell-adhesion complex is required for proper palate closure. Furthermore, we demonstrate that Nectin1 and Nectin4 act cooperatively during palatogenesis, and that human CLPED1 mutations induce CP and syndactyly in mice through a dominant-interfering mechanism.

To generate a model of pan-nectin loss, we targeted the obligate nectin binding partner afadin. As germline afadin (Afdn) knockouts are non-viable (Ikeda et al., 1999; Zhadanov et al., 1999), we used a well-characterized in utero, lentiviral-based genetic manipulation technique to generate conditional epithelial loss-of-function embryos (Beronja et al., 2010). This LUGGIGE (lentiviral ultrasound-guided gene inactivation and/or gene expression) approach can achieve ∼50-70% transduction of palatal epithelium (Beronja et al., 2010; Byrd et al., 2016; Lough et al., 2017). Moreover, LUGGIGE delivery of Cre recombinase at E9.5 drives reporter gene expression by E10.5, 2-3 days earlier than a commonly used transgenic Krt14-Cre line (Beronja et al., 2010). A similar approach using adenovirus demonstrated that periderm-specific re-expression of TGFβ3 rescued CP in Tgfb3 knockouts (Wu et al., 2013). Thus, in utero delivery of viral vectors represents a powerful and versatile approach to genetically manipulate oral epithelia during palatogenesis, offering several advantages over traditional transgenics (Lough et al., 2017).

We used LUGGIGE to generate two tissue-specific Afdn loss-of-function models (Fig. 1A): (1) knockdown with a previously validated Afdn²⁷¹¹ shRNA (Lough et al., 2019); (2) knockout using RFP-tagged Cre recombinase injected into Afdn^fl/fl mice (Beaudoin et al., 2012). For the first strategy, we injected E9.5 CD1 mice with lentivirus harboring the Afdn²⁷¹¹ shRNA and an mRFP1 fluorescent reporter fused to histone-2B (H2B-mRFP1) (Fig. 1A). RFP+ and wild-type, uninjected littermates were collected at various timepoints to evaluate palatogenesis and validate shRNA efficiency (Fig. 1B). Afdn²⁷¹¹ embryos displayed near-complete loss of afadin accumulation in palatal epithelia by E14.5 (Fig. S1A). Although CP in wild-type littermates was rare (1/16; 4 litters), most Afdn²⁷¹¹ embryos (13/17; 4 litters) presented with severe CP (Fig. 1C,D). Importantly, neither a Scramble non-targeting shRNA (n=10; 5 litters) nor empty vector H2B-mRFP1 control (n=4) caused CP when evaluated in the same manner (Fig. 1D; Fig. S1B).

We developed a novel contrast-enhanced µCT technique which allows non-invasive visualization of the embryonic palate at 6 µm resolution in three dimensions (see Materials and Methods) (Fig. 1E,F). Using this technique, plus serial cryosectioning and immunofluorescence, we demonstrated that E16.5 Afdn²⁷¹¹ embryos display a variety of defects in both PS elevation and fusion (Fig. 1E,F; Fig. S1C). These diverse phenotypes could be attributable to differences in transduction efficiency or the inherent variability in the progression of palatogenesis between littermates. However, in nearly all cases, the PS remained in direct contact with the tongue, a hallmark of persistent intraoral adhesions. We next examined earlier timepoints and observed that Afdn²⁷¹¹ embryos displayed delays in shelf elevation between E14.5-E15.5 (Fig. 1G,H). Interestingly, E15.5 Afdn²⁷¹¹ knockdown animals with near-complete lentiviral transduction frequently presented with descended PS, which appeared to lack a K8-positive periderm layer between palatal and tongue epithelia (Fig. 1I). Thus, afadin loss could induce CP by delaying palatal shelf elevation, which may cause – or be attributed to – the formation of intraoral adhesions.

We next created conditional knockouts using LUGGIGE to deliver Cre-RFP into homozygous Afdn^fl/fl embryos (hereafter referred to as Afdn lenti-cKO). Like Afdn²⁷¹¹ embryos, nearly all E16.5 Afdn lenti-cKO mutants displayed CP (4/5 embryos; two litters) (Fig. S1D,E). Interestingly, however, epithelial-specific Krt14-Cre; Afdn^fl/fl conditional knockout embryos never displayed CP (0/17 embryos; three litters) (Fig. S1D). We attribute this discrepancy to differences in the timing of Cre activation because Afdn lenti-cKOs display complete loss of afadin immunostaining in E13.5 PS whereas Krt14-Cre; Afdn^fl/fl embryos still show residual protein (Fig. S1F). This highlights potential concerns in using some Krt14-Cre strains because recombination may not occur early enough. However, other Krt14-Cre lines (Andl et al., 2004; Vasioukhin et al., 1999) have been used to generate CP models, suggesting that differences in the timing and/or expression of Cre recombinase must be considered (Lough et al., 2017). Nonetheless, LUGGIGE circumvents this issue because it drives epithelial-specific expression from early stages of palatogenesis.

Despite the well-characterized link between NECTIN1 and NECTIN4 mutations with CP and ED, mouse models have failed to recapitulate these phenotypes. CP has not been reported in any Nectin1, Nectin2 or Nectin3 single mutant or in compound mutants where three Nectin alleles are deleted (Bouchard et al., 2000; Inagaki et al., 2005; Yoshida et al., 2010). In addition to nectin 1 and nectin 2, nectin 4 is also expressed in the MEE (Richardson et al., 2017; Yoshida et al., 2012), but the consequence of Nectin4 loss has not been explored.

Because ‘transcriptional adaptation’ can lead to compensatory upregulation of related genes in germline knockouts (El-Brolosy et al., 2019), we re-evaluated the role of Nectin1 and Nectin4 in palatogenesis using LUGGIGE-mediated knockdown. To do so, we screened panels of shRNAs in cultured keratinocytes via RT-qPCR and chose two for each gene with the highest knockdown efficiencies (see Materials and Methods). As a direct readout of shRNA efficacy, we performed immunofluorescent staining in cultured keratinocytes, which demonstrated robust loss of target protein accumulation (Fig. S2).

Nectin 1 and nectin 4 localize to the interface between palatal basal cells and periderm at E13.5 (Richardson et al., 2017). Whereas Nectin1 is a direct target of the basal-specific transcription factor p63 (Trp63) in both epidermis and palatal epithelium – and is lost in Trp63 knockouts – Nectin4 is regulated by Irf6, and is still expressed in Trp63 knockouts (Ferone et al., 2012; Mollo et al., 2015; Richardson et al., 2017). The fact that Nectin1 and Nectin4 are subject to distinct modes of transcriptional regulation suggests they may be expressed in different cell populations. In support of this, single-cell RNA-Seq data show that Nectin4 is highly enriched in periderm while largely absent from palatal epithelium (Li et al., 2019). Because LUGGIGE achieves mosaic transduction, it can be used to resolve which cell layers express nectin 1 and nectin 4. At E14.5, Nectin1 loss in RFP+ basal cells, but not periderm, was sufficient to ablate nectin 1 staining, while Nectin4 loss in RFP+ periderm achieved nectin 4 protein loss at this interface (Fig. 2A-C). These data suggest that nectin 1 is specifically expressed by basal cells while nectin 4 is predominantly expressed in periderm.

We next examined the requirement of nectin 1 and nectin 4 in palatogenesis using LUGGIGE (Fig. 2D-F). Both Nectin1 and Nectin4 loss resulted in CP, albeit with low penetrance (11% for Nectin1⁴⁴⁰²; 40% for Nectin1²⁷³²; 20% for Nectin4²⁵⁸⁹ and Nectin4²⁶³²). Notably, the phenotype was also less severe than Afdn loss-of-function embryos. Finally, while Nectin1 knockdown caused uniform clefting of the secondary palate region, Nectin4 knockdown frequently resulted in posterior-specific CP (Fig. 2D).

To test whether multiple nectins cooperate to promote palate closure, we first co-injected two lentiviruses targeting Nectin1⁴⁴⁰² and Nectin4²⁵⁸⁹, each with distinct fluorescent reporters (Fig. 3A). This pooled approach resulted in a higher penetrance of CP (3/7 embryos; 3 litters) compared with loss of only one nectin (Fig. 3B,C). Furthermore, two additional embryos presented with a posterior submucosal cleft (SMC). Although this approach increased CP incidence, it could underestimate the effect of combined loss of Nectin1 and Nectin4, as RFP/YFP double-positive cells were rare (Fig. S3A). We therefore generated a new lentiviral construct that allowed for simultaneous expression of multiple shRNAs with a single H2B-mRFP1 reporter by adding an H1 promoter driving Nectin4²⁵⁸⁹ shRNA expression upstream of the original U6 promoter-Nectin1⁴⁴⁰² shRNA cassette (Fig. 3D; hereafter referred to as N1⁴⁴⁰²; N4²⁵⁸⁹). By E14.5, N1⁴⁴⁰²; N4²⁵⁸⁹ RFP+ palatal epithelium displayed efficient loss of both nectin 1 and nectin 4 (Fig. S3B), and at E16.5, nearly all RFP+ embryos displayed CP (11/12 embryos; three litters; Fig. 3E,F). Additionally, in the one N1⁴⁴⁰²; N4²⁵⁸⁹ RFP+ embryo without obvious CP, we observed persistent K14+ epithelial cells within the palatal mesenchyme, indicating a failure to complete MES dissolution (Fig. S3C). Looking earlier, at E14.5, N1⁴⁴⁰²; N4²⁵⁸⁹ embryos were more likely to have both shelves still descended compared to wild-type littermates (Fig. 3G; Fig. S3D). By E15.5, when wild-type littermates had largely completed palatogenesis (9/10 embryos; 2 litters), half of N1⁴⁴⁰²; N4²⁵⁸⁹ embryos displayed an unusual phenotype where only one PS elevated (Fig. 3G,H). Although not as dramatic as in Afdn knockdowns, we observed direct contact between the PS and lateral tongue in E14.5 N1⁴⁴⁰²; N4²⁵⁸⁹ embryos, with similar gaps in K8+ periderm coverage (Fig. S3E). This suggests that Nectin1 and Nectin4, acting through Afdn, promote PS elevation, possibly by preventing aberrant intraoral adhesions.

Mutations in NECTIN1 have been implicated as casual variants for CLPED1 (Suzuki et al., 2000). Most NECTIN1 alterations are nonsense mutations, predicted to truncate the protein following the first Ig-like loop, maintaining the region that mediates trans heterodimerization while eliminating the transmembrane region and afadin binding (Lough et al., 2017; Suzuki et al., 2000; Yoshida et al., 2015). One of the better characterized CLPED1 disease variants is the NECTIN1^W185X nonsense mutation (Sozen et al., 2001; Suzuki et al., 2000). Although it is not known whether truncated NECTIN1 transcripts undergo nonsense-mediated decay, it is conceivable that they could function in a dominant-negative manner by forming unproductive heterotypic interactions that interfere with other nectin binding. As evidence of this, an Fc-fusion of the first Ig-like loop of nectin 1 has been used to bind to and inhibit endogenous interactions between nectins 1, 3 and 4 (Kawakatsu et al., 2002). If this were the case, expression of the NECTIN1^W185X variant would be more detrimental than Nectin1 loss.

To test this, we created a bicistronic lentiviral construct to express human NECTIN1^W185X with a C-terminal V5 epitope tag, together with an H2B-mRFP1 reporter. To mimic homozygosity of the NECTIN1^W185X mutation, we included the Nectin1⁴⁴⁰² shRNA to knock down endogenous murine Nectin1 (Fig. 4A). In E16.5 Nectin1⁴⁴⁰²; NECTIN1^W185X palatal epithelium, both the reporter, labeling transduced cells and the V5 tag (which labels the truncated NECTIN^W185X protein) could be visualized (Fig. 4B). Although all Nectin1⁴⁴⁰²; NECTIN1^W185X PS underwent normal elevation, most failed to approximate (4/7 embryos; 2 litters), whereas the other three displayed a persistent MES with abnormal epithelial organization (Fig. 4C,D). MES dissolution phenotypes have also been reported upon epithelial loss of Tgfbr1/2 (Dudas et al., 2006; Xu et al., 2006) and Myh9 (Kim et al., 2015). However, these mutants display a persistent, bilayered MES, while we observe a highly disorganized epithelium with islands of cells – many of which are RFP+ – within the mesenchyme. As these islands are E-cadherin+, we believe they are likely epithelial in origin, rather than lentiviral transduced mesenchymal cells. Because MES cells normally undergo coordinated displacement towards the oral and nasal epithelial surfaces (Kim et al., 2015), it is tempting to speculate that these residual epithelial islands result from errors in this coordinated migration, possibly owing to the known function of nectins in epithelial cell sorting (Katsunuma et al., 2016; Togashi et al., 2011).

Four lines of evidence support the conclusion that the truncated NECTIN1^W185X protein functions in a dominant-interfering manner. First, the penetrance of palate closure defects was much higher in Nectin1⁴⁴⁰²; NECTIN1^W185X embryos compared with Nectin1⁴⁴⁰² knockdowns (100% vs 11%). Second, Nectin1⁴⁴⁰²; NECTIN1^W185X embryos display MES dissolution defects not observed in either Nectin1 or Nectin4 knockdowns. Third, the subcellular localization of NECTIN1^W185X is abnormal; it fails to accumulate at the basal cell-periderm interface – instead displaying circumferential, cytoplasmic accumulation (Fig. 4E). Furthermore, although the Nectin1⁴⁴⁰² shRNA acts cell autonomously in basal cells, expression of NECTIN1^W185X in periderm was sufficient to disrupt endogenous nectin 1 localization non-cell autonomously in neighboring basal cells (Fig. 4F). Fourth, we observed syndactyly – a clinical feature of CLPED1 where digits are fused – in 2/7 NECTIN1^W185X mutant embryos, but never in Nectin1 or Nectin4 knockdowns (Fig. 4G). It should be noted, however, that we cannot exclude the possibility that the NECTIN1^W185X mutation confers gain-of-function activity, which could be tested by overexpression on a wild-type rather than Nectin1 knockdown background.

The presence of intraoral adhesions and syndactyly in various nectin/afadin perturbations suggests the most likely mechanism of action for the NECTIN1^W185X mutation is disrupting periderm formation or function. Interestingly, there are some parallels in how epithelial cells and periderm coordinate morphogenesis in both palatogenesis and digit separation. For example, during digit separation, epithelial cells form an interdigital epithelial tongue (IET) similar to the MES, with non-adhesive periderm cells functioning to prevent aberrant interdigital adhesions (Kashgari et al., 2020).

In conclusion, our study demonstrates the utility of LUGGIGE in creating allelic series, compound mutants and human disease variants for studying human genetic disorders such as CP. Our results also resolve longstanding uncertainty regarding the role of nectins in mammalian palatogenesis, confirming that, despite the lack of CP in Nectin1 germline knockouts, the nectin-afadin axis is required for normal palate elevation and fusion.

Mice were housed in an AAALAC-accredited (329; June 2017), USDA registered (55-R-0004), NIH welfare-assured [D16-00256 (A3410-01)] animal facility. All procedures were performed under IACUC-approved animal protocols (19-155). CD1 mice (Charles River; 022) were used for all shRNA experiments. Afdn^fl/fl animals (Beaudoin et al., 2012) were maintained on a mixed C57B6/J CD1 background and either bred to the same Krt14-Cre allele (Dassule et al., 2000) or injected with lentiviral Cre-mRFP1 (see below). The procedure for producing, concentrating and injecting lentivirus into amniotic fluid of E9.5 embryos has been previously described (Beronja et al., 2010) and is briefly detailed below.

This protocol is approved via IACUC #19-155. Pregnant CD1 or Afdn^fl/fl females were anesthetized and the uterine horn pulled into a PBS filled dish to expose the E9.5 embryos. Embryos and custom-made glass needles were visualized by ultrasound (Vevo 2100) to guide microinjection of ∼0.7 µl of concentrated lentivirus into the amniotic space. Three to ten embryos were injected depending on litter size. Following injection, the uterine horn(s) were reinserted into the mother's abdominal cavity, which was sutured closed. The incision in the skin was resealed with surgical staples and the mother was provided with subcutaneous analgesics (5 mg/kg meloxicam and 1-4 mg/kg bupivacaine). Once awake and freely moving, the mother was returned to her housing facility for 4-9 days, at which point E13.5-18.5 embryos were harvested and processed accordingly.

E14.5-E16.5 embryos were decapitated and heads were fixed for 90 min in 4% paraformaldehyde at room temperature and rinsed with PBS. Heads were then submerged in a 0.3% (w/v) solution of phosphotungstic acid hydrate (Sigma-Aldrich P4006) and diluted in 70% ethanol for 72 h or longer with agitation. Samples were rinsed in PBS for >20 min prior to image collection using a Scanco Medical microCT-40 machine using a holder with diameter of 12 mm, voxel size of 6 µm and 114 µA/70 kVp/8 W energy exposure.

For Nectin1 and Nectin4 RNAi targeting, we tested ∼10 shRNAs for knockdown efficiency in primary keratinocytes. These sequences were selected from The RNAi Consortium (TRC) Mission shRNA library (Sigma) versions 1.0, 1.5 and 2.0, and cloned using complementary annealed oligonucleotides with AgeI/EcoRI linkers. For Afdn, we used an shRNA that had been previously validated with our lentiviral injection technique (Lough et al., 2019). shRNA clones are identified by the gene name with the nucleotide base (NCBI Accession number) where the 21-nucleotide target sequence begins in superscript (e.g. Afdn²⁷¹¹). Lentivirus was packaged in 293FT or TN cells using the pMD2.G and psPAX2 helper plasmids (Addgene plasmids 12259 and 12260, respectively). For knockdown screening, primary keratinocytes were seeded at a density of ∼150,000 cells per well into six-well plates and grown to ∼80% confluency in E-Low calcium medium and infected with an MOI of ∼1. Approximately 48 h post-infection, keratinocytes were treated with puromycin (2 µg/ml) to generate stable cell lines. After 3-4 days of puromycin selection, cells were lysed and RNA isolated using the RNeasy Mini Kit (Qiagen). cDNA was generated and amplified from 10-200 µg total RNA using either Superscript VILO (Invitrogen) or iScript (Bio-Rad). mRNA knockdown was determined by RT-qPCR (Applied Biosystems 7500 Fast RT-PCR) using two independent primer sets for each transcript with Hprt1 and cyclophilin B (Ppib2) as reference genes and cDNA from stable cell lines expressing Scramble shRNA as a reference control. Knockdown efficiencies (±95% confidence interval) were as follows: Nectin1⁴⁴⁰² (56.2±8.2%), Nectin1²⁷³² (53.6±12.8%), Nectin4²⁵⁸⁹ (82.5±32.9%) and Nectin4²⁶³² (52.5±8.0%). Primer efficiencies were determined using dose-response curves and were required to be greater than 1.8, with relative transcript abundance determined by the ΔΔCT method. RT-qPCR runs were performed in triplicate with the mean knockdown efficiency determined by calculating the geometric mean of the ΔΔCT values for at least two independent technical replicates. The following primer sequences were used: Nectin4 (fwd-1, 5′-CAGCCCCCTCCCAAATACAA-3′; rev-1, 5′-TATGATCACTGAGGCGGACACC-3′; fwd-2, 5′-AGATGTGGGGCCCTGAAGC-3′; rev-2, 5′-GCATTCGTACTCGCCCTCATC-3′) and Nectin1 (fwd-1, 5′-TAACCCGCCAGCCACTGAGT-3′; rev-1, 5′-CTGCGCAGGGCCACTATGA-3′; fwd-2, 5′-CAAACAGAACATGGCCATCTACAAC-3′; rev-2, 5′-TCGCCCTTTAGCCGTGTTTC-3′). The following shRNA targeting sequences were used: Afdn²⁷¹¹ (5′-CCTGATGACATTCCAAATATA-3′), Nectin1⁴⁴⁰² (5′-TAAACGAGAAACCTGTATTAA-3′), Nectin1²⁷³² (5′-GAATGCGAGGCACAGAATTAT-3′), Nectin4²⁵⁸⁹ (5′-TACGTACCTTCTGTAAATTAA-3′) and Nectin4²⁶³¹ (5′-CTGCTTAGACTCCCTTAATAA-3′).

E13.5-16.5 embryonic heads were fixed for 1 h at room temperature in 4% paraformaldehyde (in PBS) and submerged in 15% sucrose overnight (4°C). The next day, heads were submerged in 30% sucrose at room temperature for >6 h (until they sank) then mounted whole in OCT (Tissue Tek) and frozen at −20°C. Infected and uninfected littermate controls were mounted in the same blocks to allow for direct comparisons on the same slide. Frozen samples were sectioned (8-12 µm thick) on a Leica CM1950 cryostat, mounted on SuperFrost Plus slides (ThermoFisher) and stored at −80°C. For staining, sections were thawed at 37°C for 15 min, washed with PBS and blocked for 1 h with gelatin block (5% NDS, 3% BSA, 8% cold-water fish gelatin, 0.05% Triton X-100 in PBS). Primary antibodies were diluted in gelatin block and incubated overnight in a humidity chamber at 4°C. Slides were then washed with PBS and incubated with secondary antibodies diluted in gelatin block at room temperature (∼25°C) for 2 h, counterstained with DAPI (1:2000) for 5 min and mounted in ProLong Gold (Invitrogen). Images were acquired using LAS AF software on a Leica TCS SPE-II 4 laser confocal system on a DM5500 microscope with an ACS Apochromat 20×/0.60 multi-immersion, an ACS Apochromat 40×/1.15 oil-immersion or an ACS Apochromat 63×/1.30 oil-immersion objective.

The following primary antibodies were used: anti-mCherry (rat, Life Technologies M11217, 1:1000-3000), anti-GFP (chicken, Abcam ab13970, 1:1000), anti-E-cadherin (rat, Life Technologies 131900, 1:1000), anti-E-cadherin (goat, R&D Systems AF748, 1:1000), anti-afadin (rabbit, Sigma A0224, 1:500), anti-nectin 1 (rat, MBL International D146-3, 1:250), anti-nectin 4 (rabbit, Millipore HPA010775, 1:100), anti-K14 (guinea pig, OriGene BP5009, 1:1000), anti-K8 (rat, Developmental Studies Hybridoma Bank TROMA-1, 1:500) and anti-V5 (rabbit, Abcam ab9113, 1:500).

The following secondary antibodies were used (all antibodies were highly cross-absorbed and produced in donkey): anti-rabbit AlexaFluor 488 (Life Technologies, 1:1000), anti-rabbit Rhodamine Red-X (Jackson Labs, 1:500), anti-rabbit Cy5 (Jackson Labs, 1:400), anti-rat AlexaFluor 488 (Life Technologies, 1:1000), anti-rat Rhodamine Red-X (Jackson Labs, 1:500), anti-rat Cy5 (Jackson Labs, 1:400), anti-guinea pig AlexaFluor 488 (Life Technologies, 1:1000), anti-guinea pig Rhodamine Red-X (Jackson Labs, 1:500), anti-guinea pig Cy5 (Jackson Labs, 1:400), anti-goat AlexaFluor 488 (Life Technologies, 1:1000) and anti-goat Cy5 (Jackson Labs, 1:400).

Primary mouse keratinocytes were maintained in E medium with 15% chelated FBS and 50 µM CaCl₂ (E low medium). Parental cell lines tested negative for mycoplasma infection using the ATCC 30-1012 K kit. For viral infection, keratinocytes were plated at ∼150,000 cells per well in a six-well plate, incubated with lentivirus in the presence of polybrene (1 µg/ml) and centrifuged at 1100 g for 30 min at 37°C. Stable cell lines were generated/maintained by adding puromycin (2 µg/ml) 48 h after infection and continual antibiotic treatment following. Calcium shifts were performed by seeding ∼45,000 cells per well into 8-well Permanox chamber slides (Lab-Tek 177445) coated with poly-L-lysine, collagen and fibronectin. Once cells reached ∼85% confluency (∼12-16 h) cells were switched to high Ca²⁺ (1.5 mM) medium and grown for 8 h. Cells were fixed with 4% paraformaldehyde in PBS warmed to room temperature. Immunostaining was performed using the same protocol as for slides (see above).

Palatogenesis phenotypes, including CP, residual MES, submucosal cleft and the orientation of the PS, were determined by either stereoscopic imaging (as in Fig. 1C), µCT scanning (as in Fig. 1E), or cryosectioning and immunofluorescent confocal imaging (as in Fig. 1H). For animals E16.5 or older, statistical testing between genotypes was performed using Fisher's exact test, with a binary grouping of phenotypes into either ‘fused’ or ‘abnormal’ groups, the latter of which included embryos displaying submucosal cleft or residual MES, as well as classic CP. For younger timepoints (i.e. E14.5 or E15.5) statistical testing was performed by binning the number of PS that were elevated or descended in each genotype (i.e. for Fig. 1G, wild-type E15.5 animals had 30 PS elevated and 0 descended, while Afdn²⁷¹¹ animals had four PS elevated and four descended) and evaluating significance using Fisher's exact test. All statistical analyses and graphs were generated using GraphPad Prism 8. Figures were assembled using Adobe Photoshop and Illustrator CC 2020.

We thank members of the Williams Lab for their critical feedback throughout the process. We thank Dr Louis Reichardt (UCSF) and Dr Gerard M. Beaudoin III (Trinity University) for sharing the Afdn^fl/fl strain. We thank Dr. Hong Yuan, Jon Frank and the Small Animal Imaging Facility at the UNC Biomedical Imaging Research Center for providing the µCT imaging service. The imaging core is supported in part by an NCI cancer core grant (P30-CA016086-40).

The peer review history is available online at https://dev.biologists.org/lookup/doi/10.1242/dev.189241.reviewer-comments.pdf