The KMT2D Kabuki syndrome histone methylase controls neural crest cell differentiation and facial morphology

Epigenetics play a vital role in facial development, as mutations in chromatin-modifying enzymes have been shown to underlie syndromic craniofacial disorders (Berdasco and Esteller, 2013; Ng et al., 2010; Lederer et al., 2012; Tsurusaki et al., 2012; Santen et al., 2012; Vissers et al., 2004; Petrij et al., 1995; Roelfsema et al., 2005; Sanka et al., 2007). These chromatin regulatory enzymes function by catalyzing the addition or removal of histone post-translational modifications or by remodeling nucleosomes. These activities will alter DNA accessibility and influence the binding of transcriptional activation or repression complexes. Disorders from mutations in chromatin-modifying factors can have overlapping clinical features (Bramswig et al., 2015; Negri et al., 2019; Sakata et al., 2017; Schulz et al., 2014; Verhagen et al., 2014), indicating that they may have co-regulatory functions in facial development. Kabuki syndrome is a craniofacial disorder that manifests from mutations in two distinct histone modifying enzymes and represents a unique disorder that may be dependent on co-regulatory chromatin-modifier function.

Kabuki syndrome is a congenital disorder distinguished by facial dysmorphology, intellectual disability, skeletal and dermatoglyphic abnormalities, developmental delay and postnatal growth deficits (Adam et al., 1993; Kuroki et al., 1981; Niikawa et al., 1981, 1988). Kabuki patients exhibit distinctive facial hypoplasia with a depressed nasal tip, cleft or high arched palate, lower eyelid eversion, high arched eyebrows, elongated palpebral fissures and prominent ears. The craniofacial components of the disorder can be accompanied by a variety of cardiac defects.

Kabuki syndrome is genetically heterogeneous. Mutations were initially, and have been most frequently identified, in KMT2D, a histone H3 lysine 4 (H3K4) methylase (mutated in 39-74% of cases) (Banka et al., 2015; Bogershausen et al., 2016; Cocciadiferro et al., 2018; Hannibal et al., 2011; Makrythanasis et al., 2013; Micale et al., 2011; Miyake et al., 2013; Ng et al., 2010). This predominant molecular subtype has been termed type 1 Kabuki syndrome (KS1). A smaller percentage of Kabuki patients (3-6%) exhibit mutations in UTX (KDM6A), an X-chromosomal histone H3 lysine 27 (H3K27) demethylase and are classified as KS2 (Lederer et al., 2012; Miyake et al., 2013; Banka et al., 2015; Bogershausen et al., 2016; Cocciadiferro et al., 2018). KMT2D and UTX form a biochemical protein complex and thus may function at common sites in the genome (Cho et al., 2007). However, it is not clear whether phenotypic differences exist between KS1 and KS2 patients. Overall, KMT2D mutant patients more frequently demonstrate classical facial dysmorphism compared with patients with other genetic lesions (Adam et al., 2019; Makrythanasis et al., 2013). In large cohort analyses, KMT2D mutant KS1 patients are reported to have more consistent, typical facial dysmorphology and hypotonia, whereas UTX mutant KS2 patients have more prominent short stature and growth deficiencies (Banka et al., 2015; Miyake et al., 2013). In addition, cleft palate is much more prevalent with KMT2D mutations compared with UTX (Adam et al., 2019). In contrast, other reports have described KS2 patients as having classical facial phenotypes with no distinction between UTX or KMT2D mutation (Bogershausen et al., 2016).

Neural crest cells (NCCs) comprise a multipotent stem cell population that gives rise to anterior cranial structures (Jiang et al., 2002; McBratney-Owen et al., 2008). Cranial NCCs are specified in the dorsal neural tube and migrate in streams to anterior facial regions for differentiation towards osteoblast and chondrocyte lineages that will develop bone and cartilage (Dupin et al., 2006; Santagati and Rijli, 2003; Trainor, 2005). Throughout specification, migration and differentiation, NCCs undergo drastic alterations in transcriptional states that can be dependent on epigenetic regulation (Bajpai et al., 2010; Hu et al., 2014; Minoux et al., 2017; Rada-Iglesias et al., 2012; Strobl-Mazzulla et al., 2012). For these reasons, NCC-specific knockouts have been used to model the function of chromatin-modifying enzymes in craniofacial disorders (Chandler and Magnuson, 2016; Li et al., 2013; Sperry et al., 2014).

We have previously examined the function of UTX in mouse craniofacial development through NCC-specific knockout approaches (Shpargel et al., 2017). We used a Wnt1-Cre transgene to drive NCC-specific mouse knockout of a conditional floxed (fl) Utx allele. Homozygous Utx^fl/fl Wnt1-Cre conditional female NCC knockout mice (Utx^cFKO) exhibited severe frontonasal hypoplasia and growth deficiencies, as well as reduced viability before weaning. Hemizygous Utx^fl/y Wnt1-Cre male mice (Utx^cMKO) were viable with mild facial hypoplasia and moderate postnatal growth deficiencies compared with Utx^cFKO due to partial redundancy between UTX and the Y-chromosomal homolog, UTY (Shpargel et al., 2012, 2017). Mouse UTX NCC knockout modeled several dysmorphic facial phenotypes observed in human Kabuki syndrome.

We now establish NCC knockout of KMT2D for comparison of craniofacial anatomical and NCC cellular phenotypes to UTX loss-of-function. We find that heterozygous loss of KMT2D in NCC lineages resulted in mild facial hypoplasia and deficiencies in postnatal growth, whereas homozygous deletion of NCC KMT2D manifested in perinatal lethality. KMT2D NCC knockout phenocopied the frontonasal hypoplasia previously observed with UTX deletion. Although KMT2D mutant NCCs displayed appropriate specification and migration, frontal bone ossification centers displayed altered distributions of NCC osteoblast and chondrocyte differentiation for both KMT2D and UTX loss-of-function. In contrast, fully penetrant secondary cleft palate, mandible hypoplasia and cranial base ossification phenotypes only manifested with KMT2D NCC mutation. NCC KMT2D knockout palatal shelves failed to elevate, with altered expression of extracellular matrix (ECM) components. Within the cranial base, KMT2D knockout produced deficiencies in endochondral ossification of presphenoid and basisphenoid bone due to loss of NCC hypertrophic chondrocyte differentiation. Our results highlight KMT2D function in several aspects of cranial NCC differentiation and identify developmental differences in KMT2D and UTX requirements that may underlie reported KS1 and KS2 phenotypic disparity.

We obtained a Kmt2d conditional allele with floxed exons upstream of the SET methylase domain for conditional knockout approaches (Fig. 1A). Cre-mediated deletion will frameshift the coding sequence to a stop codon, creating a mutant KMT2D protein lacking SET and methylase activity. We utilized a Vasa-Cre transgene to generate a germline knockout of the KMT2D SET domain. KMT2D SET domain knockout embryos were dead and resorbed by embryonic day (E)11.5 (4/4 embryos), similar to published Kmt2d null alleles (Lee et al., 2013). Primary mouse embryonic fibroblasts (MEFs) isolated from a pool of E9.5 embryos carrying the KMT2D SET deletion revealed an absence of KMT2D protein in the expected size range for SET deletion (568 kD, Fig. 1B), consistent with findings that the SET domain is essential for KMT2D protein stability (Jang et al., 2019).

To model the craniofacial dysfunction of KMT2D in Kabuki syndrome, the mouse Kmt2d conditional allele was crossed with the NCC-specific Wnt1-Cre transgene (Danielian et al., 1998). Heterozygous conditional Kmt2d^+/fl Wnt1-Cre mice (Kmt2d^cHet) were viable but exhibited facial dysmorphism (Fig. 1C,D) and postnatal growth deficiency (Fig. 1E). The Kabuki-like facial features such as depressed snout, broad face, dome shaped forehead, ocular phenotypes and abnormal, rounded ears were fully penetrant. These NCC KMT2D haploinsufficient facial deformities phenocopy Utx^cMKO mice (Shpargel et al., 2017).

To understand KMT2D NCC requirements, we examined the phenotypic effects of completely removing KMT2D protein. Kmt2d^fl/fl Wnt1-Cre mice (Kmt2d^cKO) died at birth as no pups were observed at postnatal day (P)2, although eight were expected by Mendelian genetics. At P1, Kmt2d^cKO pups demonstrated severe facial hypoplasia (Fig. 1F,H) relative to wild-type (WT) controls (Kmt2d^fl/fl without Cre). In contrast, Kmt2d^cHet pups had more intermediate dysplasia (Fig. 1G). Quantification of facial characteristics such as a reduction in the nose-to-ear length (Fig. 1I) and frontal angle (Fig. 1J) highlighted the shortened frontonasal structures present in Kmt2d^cHet and Kmt2d^cKO newborn pups and the graded dysmorphism that is dependent on KMT2D dosage.

Whole mount Alizarin Red and Alcian Blue staining for bone/cartilage in Kmt2d^cKO pups (Fig. 2A,B) facilitated quantitation of bone structure (Fig. 2C). Kmt2d^cKO resulted in significant reductions in nasal and frontal bone lengths, but had no effect on parietal bone, which is derived from non-neural crest mesoderm (Jiang et al., 2002). Kmt2d^cHet pups exhibited intermediate reductions in nasal and frontal bone lengths (with frontal bone measurements lacking statistical significance from WT). Although this compilation of craniofacial phenotypes was similar to Utx^cFKO (Shpargel et al., 2017), Kmt2d^cKO exhibited mandible hypoplasia that was not as severely affected in Utx^cFKO at birth (Fig. 2D-F). Kmt2d^cKO mandibles had particularly underdeveloped condylar process with a lack of condylar cartilage (Fig. 2E, asterisk).

We traced NCC development to identify the stage at which Kmt2d^cKO frontal NCCs display an altered phenotype. Mouse cranial NCCs are specified in the dorsal neural tube at E7.5-E8 and complete migration to anterior facial positions at approximately E9 (Serbedzija et al., 1992). Across subsequent post-migratory development (E11.5-E13.5), cranial NCCs experience dramatic changes in cellular properties such as proliferation and differentiation. NCC lineage tracing was performed using a Cre-inducible tomato fluorescent reporter (Madisen et al., 2010). At post-migratory cranial NCC time points (E11.5 and E13.5), Kmt2d^cKO embryos demonstrated similar anterior NCC positioning and fluorescent intensity as WT (Kmt2d^+/+ with Wnt1-Cre) controls (Fig. 3A-F and Fig. S1A,B).

SOX10 is a transcription factor that becomes expressed in all NCCs during migration, but persists in the peripheral nervous system and melanocyte lineages upon migratory completion (Hari et al., 2012; Kuhlbrodt et al., 1998; Southard-Smith et al., 1998). We performed whole-mount in situ hybridization for Sox10 expression to examine migratory NCCs in Kmt2d^cKO embryos (Fig. S1C,D). At E10.5, NCC migration had completed normally in both WT and Kmt2d^cKO embryos, as there were no lagging Sox10-positive cells in the migratory streams expected from cranial NCCs (Serbedzija et al., 1992). Furthermore, Sox10 expression domains in post-migratory peripheral nervous system components such as trigeminal and dorsal root ganglia appeared normal in Kmt2d^cKO embryos. Therefore, cranial NCC specification and migration appeared largely normal in Kmt2d^cKO embryos.

We utilized a second Cre driver with more specificity to remove KMT2D function at the completion of migration. Although SOX10 is expressed in migrating NCCs, a Sox10-Cre transgene has been established (Matsuoka et al., 2005) that does not induce recombination until the completion of migration. Reporter-based analyses have demonstrated that Sox10-Cre activity is delayed compared with endogenous Sox10 expression and is absent from early migrating NCCs (Hari et al., 2012; Jacques-Fricke et al., 2012). In contrast to Wnt1-Cre, Sox10-Cre reporter staining was low in migratory cranial NCCs streams and became activated upon reaching anterior destinations. Upon crossing to this Sox10-Cre transgene, the tomato reporter demonstrated localization restricted to anterior cranial regions characteristic of migrated NCCs (Fig. S1E,F) and lacked the posterior neural tube and midbrain localization characteristic of Wnt1-Cre activity (Fig. 3E). Sox10-Cre had a similar efficiency in differentiating NCCs (Fig. S1G,H) as Wnt1-Cre (Fig. S1A). Sox10-Cre-driven Kmt2d knockout (Kmt2d^cS10KO) produced frontonasal hypoplasia (Fig. 3G-J) similar to Wnt1-driven Kmt2d^cKO mutagenesis. Skeletal staining also demonstrated consistent reductions in frontal, nasal and mandible bone length in Kmt2d^cS10KO pups at birth (Fig. 3K-Q). We conclude that the craniofacial phenotypes in Kmt2d^cKO pups manifest from KMT2D loss-of-function in post-migratory cranial NCCs.

Provided that the cranial NCC distribution appeared normal at E13.5 early differentiation time points (Fig. 3D-F), we examined NCC differentiation events that precede intramembranous ossification of the frontal bone. Frontal bone primordia are specified in supraorbital regions of the head, basolateral to the brain (Ishii et al., 2003; Yoshida et al., 2008). Cranial NCC mesenchymal cells within the supraorbital arch differentiate to osteoblast lineages that are required for the synthesis and mineralization of frontal bone (Ferguson and Atit, 2019; Ishii et al., 2003). In this differentiation pathway (depicted in Fig. 4A), SOX9-expressing ectomesenchymal stem cells initiate RUNX2 expression in an osteochondral progenitor cell that has the potential to differentiate to pre-osteoblasts expressing RUNX2 or pre-chondrocytes expressing SOX9 (Bhatt et al., 2013). Committed differentiation to osteoblast lineages requires RUNX2-induced osterix (OSX; Sp7) expression (Nakashima et al., 2002), whereas SOX9 expressing pre-chondrocytes will induce downstream targets such as type II collagen (COL2). We examined KMT2D function in cranial NCC differentiation towards pre-osteoblast and pre-chondrocyte lineages by immunofluorescence for RUNX2 or COL2 respectively on E13.5 supraorbital coronal sections. In the E13.5 frontal ossification center, WT pre-osteoblast and pre-chondrocyte lineages have diverged with COL2+ pre-chondrocytes specified medially and RUNX2+ pre-osteoblasts occupying a lateral crescent domain (Fig. 4B) similar to previous reports (Goodnough et al., 2012; Ishii et al., 2003; Vivatbutsiri et al., 2008). Kmt2d^cKO embryos have specified these lineages; however, the RUNX2+ osteoblast domain (Fig. 4B,C) and OSX+ committed pre-osteoblasts (Fig. 4D,E) are positioned more laterally compared with WT. The Kmt2d^cKO medial chondrocyte domain has extended laterally, indicating that some osteochondral progenitor differentiation in this region may have instead skewed towards the chondrocyte lineage.

Unlike Kmt2d^cKO, Utx^cFKO frontonasal NCCs experience an increase in apoptosis at the onset of E13.5 differentiation time points (Shpargel et al., 2017). We analyzed Utx^cFKO frontal primordia to identify whether UTX knockout produced an altered pattern of osteochondral differentiation similar to KMT2D. Utx^cFKO cranial NCCs within the supraorbital arch had a laterally shifted pattern of osteoblast and chondrocyte differentiation similar to Kmt2d^cKO, as both RUNX2+ and committed OSX+ osteoblasts were restricted to more lateral domains compared with WT (Fig. 4F-I). In frontal primordia, Kmt2d was expressed throughout both osteoblasts and chondrocytes, but demonstrated elevated osteoblast expression (Fig. S2A) supporting a function in differentiation. UTX expression localized to both chondrocytes and osteoblasts with no tissue specific bias (Fig. S2B,C). The patterns of altered Kmt2d^cKO osteoblast differentiation persisted at later E15.5 stages and led to altered frontal bone shape (Fig. S3A,B). Although the areas of the specified RUNX2+ osteoblast and committed OSX+ osteoblast domains were not impacted by KMT2D NCC knockout (Fig. 4J), the length of the RUNX2+ domain was shorter (Fig. 4K) and the position of the osteoblast domains was consistently altered across Kmt2d^cKO embryos (Fig. 4L). We examined earlier E12.5 stages when frontal primordia are becoming specified. At this stage of development osteochondral progenitor cells have diverged into RUNX2+ pre-osteoblasts and SOX9+ pre-chondrocytes; however, pre-osteoblasts lack OSX expression characteristic of lineage commitment (Fig. S3C-F). At this earlier stage the osteoblast/chondrocyte domains appear similar in WT and Kmt2d^cKO embryos (Fig. S3C,D). A BrdU injection 2 h before dissection indicated normal patterns of proliferation in E12.5 Kmt2d^cKO frontal primordia (Fig. S3G-I), with a lack of apoptosis as indicated by cleaved caspase 3 signal (Fig. S3J,K). Therefore, the Kmt2d^cKO frontal phenotypes arise from altered patterns of osteochondral differentiation as these domains expand at E13.5.

KMT2D mutant KS1 human patients have a higher incidence of cleft palate compared with UTX mutation (Adam et al., 2019). Therefore, we assessed palatogenesis in Kmt2d^cKO embryos. In contrast to Utx^cFKO pups that demonstrate a low penetrance of cleft palate (Shpargel et al., 2017), Kmt2d^cKO embryos had fully penetrant cleft palate that likely contributes to postnatal lethality (Fig. 5A,B). Similarly, Kmt2d^cS10KO embryos (Fig. 3H) demonstrated fully penetrant cleft palate (4/4 embryos).

Secondary palate formation is a highly organized process involving outgrowth of the palatal shelf from the maxillary prominence (E11.5-E12.5), vertical outgrowth and remodeling to elevate the shelves above the tongue to a horizontal position (E12.5-E14.5), and midline fusion of the shelves with degeneration of the epithelial seam (E14.5-E15.5) to form a continuous palate (Bush and Jiang, 2012). We examined the NCC cellular and molecular mechanisms of dysfunctional palatogenesis in Kmt2d^cKO embryos. Compared with WT, E13.5 Kmt2d^cKO embryos lacked vertical outgrowth and extension of the distal tip of the anterior palatal shelf (Fig. 5C,D). Notably, the tongue in Kmt2d^cKO embryos is situated more posterior relative to the anterior palatal domain and may play a role in defective Kmt2d^cKO elevation. As failed outgrowth and elevation may also result from deficient palatal shelf mesenchymal proliferation, we labeled mitotic cells by immunofluorescence for the phospho-histone H3 (Ser10) antibody (pH3S10) (Fig. 5C,D). Overall, there was a similar distribution of mitotic cells in WT and Kmt2d^cKO palatal shelves (Fig. 5G).

Palatal morphogenesis is also dependent on NCC mesenchymal remodeling and condensation in the interior side of the palatal shelf, which is accompanied by appropriate levels of osteogenic differentiation (Fu et al., 2017; Jin et al., 2010; Wu et al., 2008). In WT E13.5 sections, the interior palatal mesenchymal cells express RUNX2, have condensed, and have differentiated to committed OSX+ osteoblasts (Fig. 5E). In contrast, Kmt2d^cKO palatal shelves are deficient in OSX+ differentiation (Fig. 5F,H). To understand the molecular mechanisms underlying KMT2D function in palatogenesis, we dissected E14.25 WT and Kmt2d^cKO palatal shelves before fusion and performed RNA-seq to identify mis-expressed genes and pathways. A total of 252 genes were significantly reduced in expression in Kmt2d^cKO samples, with 161 genes upregulated (Table S1), consistent with KMT2D function in gene activation events. Further refinement of this list for genes that have higher levels of WT palate expression (RPKM>2) with more dramatic Kmt2d^cKO expression changes (logFC<−1) yielded a set of 115 KMT2D affected genes (Table S1: sheet 3). Both gene set enrichment analysis (Broad MSigDB) and Ingenuity Pathway Analysis identified that Kmt2d^cKO downregulated genes were most significantly enriched for factors in ECM organization (Fig. 5I, Table S1: sheet 4). Induction of ECM components are a feature of remodeling mesenchymal cells within the interior vertical palatal shelf (Chiquet et al., 2016; Jin et al., 2010). We find that many matrix remodeling components such as collagens, collagen processing factors and extracellular glycoproteins are consistently reduced in expression with KMT2D loss-of-function (Fig. S4A and Table S1).

We examined the spatial distribution of ECM components using Picrosirius Red histological stain, which has an affinity for matrix collagen. At E14.25, Picrosirius Red ECM stain was enriched at differentiating osteoblasts within the interior of WT palatal shelves as well as mesenchymal supporting cells layered underneath regions of epithelial folding that form the distal palatal tip (Fig. 5J). These ECM enrichments were lacking in Kmt2d^cKO embryos (Fig. 5K) that coincided with an abnormally shaped distal extension. Although the overall area of these extensions was similar (Fig. S4B), WT palatal shelves exhibited a longer extension than Kmt2d^cKO (Fig. 5L). BrdU labeling in E14.25 embryos demonstrated similar proliferation within these extensions (Fig. S4C-E) and cleaved caspase 3 immunofluorescence revealed very few apoptotic cells in either WT or Kmt2d^cKO palatal shelves (Fig. S4F,G). Examination of initial palatal shelf outgrowth from the maxillary prominence at E12.5 revealed inconsistent phenotypes with a lack of Kmt2d^cKO proliferative or apoptotic defects (Fig. S4H-P). The area and length of the Kmt2d^cKO palatal shelf outgrowth was slightly smaller but failed to reach statistical significance (Fig. S4J,K).

Tmem119, one of the most significantly downregulated genes in Kmt2d^cKO palatal shelves (Table S1), is known to function in osteoblast differentiation (Hisa et al., 2011). In WT embryos, TMEM119 was highly expressed throughout the interior palate in regions of differentiating osteoblasts but was also localized more diffusely through distal extensions (Fig. 5M). TMEM119 expression was lost throughout the Kmt2d^cKO palatal shelves (Fig. 5N). Other ECM regulatory genes identified by Kmt2d^cKO RNA-seq were also elevated in osteoblast differentiation (Col8a2 and Pcolce) or in distal palatal extensions (Col8a2) with reduced Kmt2d^cKO expression (Fig. 5O-R), highlighting a role for KMT2D in these processes. Kmt2d expression was elevated in both osteoblast differentiation domains and mesenchymal cells in the vicinity of epithelial folds (Fig. S2F), regions that correlated with transcripts misregulated in Kmt2d^cKO palatal shelves. We hypothesized that perhaps redundancy between UTX and another H3K27 demethylase, KDM6B (JMJD3), is the reason for low cleft palate penetrance in Utx^cFKO embryos (Shpargel et al., 2017). UTX was expressed throughout the palatal mesenchyme, with enrichment in distal and subepithelial regions (Fig. S2G,H). KDM6B expression demonstrated similar subepithelial localization but was also upregulated with osteoblast differentiation in the palate (Fig. S2I,J) as well as the supraorbital ridge (Fig. S2D,E). Therefore, redundancy between H3K27 demethylases may lead to observed Kmt2d^cKO and Utx^cKO phenotypic differences.

The nasal and maxillary components of the viscerocranium (facial skeleton) and calvaria portion (including the frontal bone) of the neurocranium (brain case) are formed directly through osteoblast-dependent intramembranous ossification. In contrast, within the cranial base (basicranium) support of the neurocranium, bone is deposited through endochondral ossification in the presence of cartilage and directed by chondrocyte differentiation. As the anterior cranial base is NCC derived (McBratney-Owen et al., 2008; Szabo-Rogers et al., 2010) and the cranial base provides structural support for the viscerocranium to influence facial morphology (Lieberman et al., 2000; Neaux et al., 2018; Nie, 2005), we examined whether these structures were properly formed in Kmt2d^cKO newborn pups.

We performed Alizarin Red and Alcian Blue staining for bone and cartilage on postnatal WT and Kmt2d^cKO pups at birth for whole-mount imaging of the basicranium (Fig. 6A,B). Whereas the WT cranial base has ossification of both the presphenoid and basisphenoid bones at birth, Kmt2d^cKO pups demonstrated an absence of presphenoid bone deposition and a shortened, abnormally shaped basisphenoid bone. In addition, Kmt2d^cKO pups exhibited underdeveloped pterygoid processes, palatine and tympanic structures. Hyoid bone ossification, also dependent on NCC-based endochondral ossification, was also absent in Kmt2d^cKO pups and only stained for Alcian Blue cartilage. In addition, the mandibular condylar process hypoplasia (Fig. 2E) is due to a lack of endochondral ossification from the condylar cartilage. Postoptic roots of the Kmt2d^cKO sphenoid bone (lesser wings) were ossified (Fig. 6B); however, these structures form from mesoderm-derived hypochiasmatic cartilage (McBratney-Owen et al., 2008). In contrast, Utx^cFKO pups demonstrated proper formation of all basicranial structures, even in instances of cleft palate where the palatine bone has not fused at the midline (Fig. 6C). Kmt2d^cS10KO embryos with loss-of-function in post-migratory NCCs had identical cranial base phenotypes to Wnt1-Cre-based knockout, with the exception that these pups on occasion displayed a small strip of presphenoid bone formation that was greatly reduced compared with WT (Fig. 6D and Fig. S5A). Quantitation of whole-mount bone lengths along the cranial base midline further illustrates the reductions in basisphenoid and presphenoid bone development in both Kmt2d^cHet and Kmt2d^cKO skeletal preparations (Fig. 6E). Notably, the length of the presphenoidal synchondrosis, the intermediary cartilage responsible for formation of basisphenoid and presphenoid bone, was not altered in the Kmt2d^cHet cranial base.

Owing to deficiencies in cranial base endochondral ossification, we performed sagittal sections through the midline of this cranial region to analyze chondrocyte histology. Basicranial bones are formed at ossification centers surrounded by cartilage. The synchondroses (cartilaginous joints) between basicranial ossification centers have bidirectional growth for ossification at opposing ends of the growth plate. The spheno-occipital synchondrosis is the growth center for both the basioccipital and basisphenoid bones (Fig. 7A). The basisphenoid bone will form at the junction of the spheno-occipital and presphenoidal (intersphenoid) synchondroses. The presphenoid bone will form at the junction of the presphenoidal synchondrosis and the septopresphenoidal junction, posterior to the nasal septum.

The synchondroses have zones of resting, proliferative and hypertrophic chondrocyte differentiation (Fig. 7A). Following chondrocyte hypertrophy, osteoblasts will differentiate from the perichondrium surrounding the cartilage anlage, populate the ossification zone and synthesize bone. In contrast to the WT presphenoid bone, which is ossified at birth, Kmt2d^cKO sections have an absence of presphenoid bone synthesis and a reduction in the length of chondrocyte hypertrophic zones at this site (Fig. 7C,G). We quantified the overall length of the hypertrophic zone from both the anterior (septopresphenoidal) and posterior (presphenoidal synchondrosis) growth zones for the presphenoid as well as the proliferative zone from the presphenoidal synchondrosis. Kmt2d^cKO chondrocytes had a normal proliferative zone but had significant reductions in the amount of hypertrophic differentiation and an absence of bone deposition (Fig. 7N). Kmt2d^cHet chondrocytes had an intermediate phenotype (Fig. 7E). Differentiated presphenoid Kmt2d^cKO chondrocytes were not as hypertrophic, with significant decreases in cell area (Fig. 7O-Q). As Utx^cFKO pups demonstrated presphenoid ossification (Fig. 6C), we examined whether UTX is expressed throughout these cell types. UTX was expressed throughout presphenoidal chondrocytes and osteoblasts (Fig. S2K,L) in a similar fashion to KDM6B (Fig. S2M,N), which may be redundant to UTX.

Within the Kmt2d^cKO basisphenoid ossification zone, the anterior growth zone (presphenoidal synchondrosis side) similarly had significant reductions in chondrocyte hypertrophic differentiation and bone deposition (Fig. 7B,D,F,N). However, the effect on the Kmt2d^cKO basisphenoid ossification zone was not nearly as severe as the presphenoid. We performed NCC lineage tracing with the Cre-responsive tomato reporter to examine the contributions of mutant cells to these structures (Fig. 7H-M). In both WT (with Wnt1-Cre) and Kmt2d^cHet, tomato-positive NCC lineages contribute to both the anterior and posterior chondrocyte growth zones surrounding the presphenoid, in addition to the RUNX2 highly-expressing osteoblasts within the bone matrix (Fig. 7I,K). In contrast, only the anterior basisphenoid growth zone and perichondrium are composed of NCC lineages, yet osteoblasts within the basisphenoid matrix are almost entirely NCC derived (Fig. 7H,J). The Kmt2d^cKO basisphenoid bone was instead comprised of a mix of NCC and non-NCC lineages (Fig. 7L). Therefore, the deficiencies in Kmt2d^cKO NCC chondrocyte hypertrophy within the basisphenoid anterior growth zone (Fig. 7N) can be compensated by the posterior spheno-occipital synchondrosis that has been reported to be partially comprised of mesoderm lineages (McBratney-Owen et al., 2008).

Kabuki syndrome is a genetically heterogeneous craniofacial syndrome largely resulting from mutations in KMT2D and UTX, but rare mutations have also been identified in a small subset of other genes (Lintas and Persico, 2018). Clinical diagnosis of Kabuki syndrome is ascribed to a distinct facial gestalt, however the cellular mechanisms underlying altered craniofacial development are largely unknown. The disorder has been modeled in other organisms through whole animal loss-of-function approaches. Knockdown or knockout of KMT2D or UTX in zebrafish produced branchial arch and viscerocranium hypoplasia with altered cartilage structures (Bögershausen et al., 2015; Van Laarhoven et al., 2015). Morpholino knockdown in Xenopus embryos resulted in craniofacial and cartilaginous deformities (Schwenty-Lara et al., 2019). Mouse homozygous Kmt2d knockout leads to early embryonic lethality (Lee et al., 2013) that precludes analysis of Kabuki-dependent phenotypes. Heterozygous Kmt2d mutations have been used to model the neurological, cardiac and skeletal components of Kabuki syndrome (Ang et al., 2016; Bjornsson et al., 2014; Carosso et al., 2019; Fahrner et al., 2019).

We now demonstrate by mouse tissue-specific approaches that KMT2D function intrinsic to NCCs is essential for appropriate craniofacial development. Although early NCC developmental processes such as specification and migration appear grossly normal, we observe NCC dependency on KMT2D for appropriate differentiation in several structural components within anterior cranial tissues. Intramembranous ossification of the frontal bone is dependent on localized osteoblast differentiation from osteochondral progenitors. NCC KMT2D loss-of-function leads to mild reductions and altered distributions of these pre-osteoblast domains in frontal primordia with expansion of pre-chondrocyte differentiation. Alternatively, within the basicranium, KMT2D functions in terminal differentiation of hypertrophic chondrocytes and is essential in formation of bone derived by endochondral ossification. KMT2D is essential in palatogenesis and NCC loss-of-function leads to altered mesenchymal differentiation with altered expression of ECM components. Although we did not observe any significant changes in proliferation or apoptosis of these progenitor domains, it is possible that subtle region-specific changes might contribute to the observed phenotypes. Consistent with our findings, regional mesenchymal condensation as well as expression of ECM collagens, processing factors and glycoproteins are essential in palatal shelf elevation (Chiquet et al., 2016; d'Amaro et al., 2012; Jin et al., 2010; Logan et al., 2019; Zhang et al., 2015b). We find that several of the KMT2D transcriptional targets are expressed in osteoblast differentiation domains and folding subepithelial mesenchyme that may shape the distal palatal shelf (Yu and Ornitz, 2011).

KMT2D has previous ascribed functions in cellular differentiation events. KMT2D cooperated with lineage-determining transcription factors to catalyze H3K4 mono and di-methylation at enhancers during embryonic stem cell, adipocyte and myogenic differentiation (Lee et al., 2013; Wang et al., 2016). Cranial intramembranous ossification is dependent on directed differentiation of a transient osteochondral progenitor cell towards either osteoblast or chondrocyte cell fate (Abzhanov et al., 2007; Akiyama et al., 2005; Bhatt et al., 2013). Osteochondral progenitors express both SOX9 and RUNX2, and the osteoblast or chondrocyte lineage decision will depend on expression levels of these transcription factors and whether SOX9 can effectively repress RUNX2 expression (Zhou et al., 2006). In a similar fashion, chondrocyte terminal differentiation is dependent on the SOX9 to RUNX2 balance. SOX9 is essential for chondrocyte proliferation (Akiyama et al., 2002). RUNX2 expression activates within these proliferating chondrocytes and promotes hypertrophic differentiation and endochondral ossification (Chen et al., 2014; Takarada et al., 2013). Continued SOX9 expression is required to keep RUNX2 expression under control to maintain a proliferative state and prevent hypertrophic chondrocyte and osteoblast differentiation (Dy et al., 2012). As we observe cranial NCC differentiation phenotypes in both intramembranous and endochondral ossification, KMT2D may be involved in regulating expression or activity of these transcription factors. Both KMT2D and UTX have described functions in regulating RUNX2 expression and activity during myoblast differentiation (Munehira et al., 2017; Rojas et al., 2015). We find that RUNX2 can be activated in KMT2D mutant NCCs (Figs 4E, 5F, 7M), so the methylase may act downstream to activate the RUNX2 transcriptional program. KMT2D also restricts chondrocyte differentiation within a teratocarcinoma cell line, through SHOX2-mediated SOX9 repression (Fahrner et al., 2019). In contrast to our findings, endochondral skeletal hypoplasia in Kmt2d heterozygous mice results from expanded growth plates with increased tibia chondrocyte hypertrophy (Fahrner et al., 2019). These differential findings could stem from KMT2D null as opposed to loss-of-function phenotypes or cranial-specific KMT2D functions that may differ from skeletal development. KMT2D knockdown in Xenopus embryos resulted in deficits in NCC neural plate specification and migration (Schwenty-Lara et al., 2019). Although we do not observe a major defect in NCC migration, KMT2D may be partially redundant in mammalian NCC development with the KMT2C homolog as has been observed in other cell types (Hu et al., 2013; Lee et al., 2013). Knockout of both H3K4 methylases may produce severe developmental NCC abnormalities and more complete inhibition of chondrocyte or osteoblast differentiation.

Phenotypic heterogeneity within Kabuki syndrome has led to diagnostic difficulties, and even misdiagnoses (Bramswig et al., 2015; Negri et al., 2019; Sakata et al., 2017; Schulz et al., 2014; Verhagen et al., 2014). KS1 resulting from KMT2D mutation has the most consistent genotype-to-phenotype correlations, whereby patients exhibit more typical facial dysmorphic features and palatal abnormalities (Adam et al., 2019; Banka et al., 2015; Makrythanasis et al., 2013; Miyake et al., 2013). KS2 patients can exhibit varying severity with sexual dimorphism (Banka et al., 2015; Miyake et al., 2013). UTX hemizygous KS2 males had more severe motor delay and cognitive disability (Banka et al., 2015). This variable expressivity was even present in hereditary families, whereby heterozygous UTX females carrying the same mutation as hemizygous UTX males lacked intellectual disability (Kim and Lee, 2017; Lederer et al., 2014). In these cases, heterozygous females have one functional copy of UTX, whereas hemizygous males have the Y-chromosome homolog UTY. The sexual dimorphism may be due to skewed X-chromosome inactivation in heterozygotes (Miyake et al., 2013) or lack of UTY function as a histone demethylase (Hong et al., 2007; Lan et al., 2007; Shpargel et al., 2012). However, UTX function in mouse craniofacial development is redundant with UTY and does not require histone demethylase activity (Shpargel et al., 2017). Allelic heterogeneity also exists for KS1, as KMT2D missense mutations produced less severe facial phenotypes compared with truncating mutations (Miyake et al., 2013). Overall, the UTX causative facial dysmorphism in KS2 lacked some of the more typical Kabuki features that were more consistently present with loss of KMT2D function (Banka et al., 2015; Miyake et al., 2013). However, small cohort sizes, allelic heterogeneity and sexual dimorphism have made cross comparisons difficult.

Modeling craniofacial development in the mouse with NCC-specific knockout approaches has provided us with the opportunity to directly compare null knockout of Kmt2d to Utx to contrast regional specific function of the two chromatin factors (this study and Shpargel et al., 2017). Homozygous NCC knockout of both Kmt2d and Utx produced similar frontonasal hypoplasia. Both chromatin-modifying proteins affected osteochondral progenitor differentiation patterns in frontal primordia. In contrast, Kmt2d NCC knockout produced fully penetrant cleft palate, an observation that was only present in a small percentage of Utx knockout mice. Unlike Utx, Kmt2d NCC knockout produced more severe mandible hypoplasia, and micrognathia is a feature of human Kabuki patients (Porntaveetus et al., 2018). In addition, Kmt2d was uniquely required for cranial NCC-dependent endochondral ossification events within the basicranium, mandibular condyle and hyoid. Alterations to cranial base and condylar process growth can contribute to facial dysmorphology in craniofacial disorders (Lieberman et al., 2000; Neaux et al., 2018; Nie, 2005). We conclude that KMT2D does demonstrate region-specific functions in NCC craniofacial development that may underlie phenotypic heterogeneity in KS1 and KS2 Kabuki patients. UTX either does not function in these cell types or has redundancy with other chromatin-modifying proteins. Notably, a homologous H3K27 demethylase, KDM6B functions in chondrocyte proliferation and hypertrophy in long bone endochondral ossification (Zhang et al., 2015a). Consistent with potential redundancy, we find that UTX and KDM6B expression overlaps in several NCC-derived craniofacial tissues (Fig. S2). In our future research, we will continue to examine the direct role of KMT2D towards induction of cranial NCC differentiation and explore cellular-specific genomic redundancy between chromatin-modifying factors in this process.

All mouse experimental procedures were approved by the University of North Carolina Institutional Animal Care and Use Committee. Mouse strains were crossed for at least four generations onto C57BL/6J backgrounds. The Kmt2d^fl and Utx^fl alleles were previously described (Jang et al., 2019; Shpargel et al., 2014). The Kdm6b^GT allele (Kdm6b^{tm1(KOMP)Wtsi}), targeted in JM8 embryonic stem cells by the Knockout Mouse Project (KOMP), replaced Kdm6b exons 11-20 with a splicing trap and internal ribosome entry site for lacZ expression. Embryonic stem cells were injected into C57BL/6J-Tyr^c-2J blastocysts to generate germline transmission. Wnt1-Cre, Sox10-Cre, and Rosa^Tomato reporter mice were imported from the Jackson Laboratory (Danielian et al., 1998; Madisen et al., 2010; Matsuoka et al., 2005). The Sox10-Cre transgene was active in the male germline similar to the S4F:Cre line which uses a Sox10 enhancer (Stine et al., 2009). Therefore, female mice carrying the Sox10-Cre allele were used for crossing into Kmt2d.

Embryonic fixation and sectioning was performed as previously described (Shpargel et al., 2012). For histology, samples were fixed in 4% paraformaldehyde overnight, embedded in paraffin and sectioned (10 μm) on a microtome (Leica, RM2165) for Hematoxylin and Eosin (H&E) staining, Picrosirius Red staining or in situ hybridization. H&E staining was performed in Harris Hematoxylin, blued with ammonium hydroxide, and counterstained with Eosin. Picrosirius Red staining was performed with 0.1% Direct Red 80 (Sigma-Aldrich, 365548-5G) dissolved in saturated picric acid. RNA in situ hybridization on sections was performed as previously described (Chandler et al., 2007) using T7 transcribed digoxigenin-labeled riboprobes from regions indicated in the primers section. All probes were generated from unique regions with the Kmt2d probe based on a region utilized by the Allen Brain Atlas (https://mouse.brain-map.org) and the Sox10 probe based on 3′ regions from published reports (Kuhlbrodt et al., 1998; Southard-Smith et al., 1998). X-Gal staining was performed as previously described (Chandler et al., 2007).

Whole-mount skeletal preparations were fixed in 95% ethanol and Alizarin Red and Alcian Blue staining was performed as previously described (Lufkin et al., 1992). Whole-mount RNA in situ hybridization was performed as previously described (Chandler and Magnuson, 2016; Jacques-Fricke et al., 2012).

For immunofluorescence, embryos were fixed for 1 h in 4% paraformaldehyde, passed through a sucrose gradient, and cryopreserved/frozen in sucrose/OCT before sectioning (10 μm) on a cryostat (Leica, CM3050 S). Immunofluorescence was performed in PBS/3% bovine serum albumin/10% goat serum/0.1% Triton X-100 using the following antibodies: RUNX2 (1:800, Cell Signaling Technology, 12556S or 1:100, Abcam, ab76956), COL2 (1:100, Thermo Fisher Scientific, MA5-12789), OSX (1:150, Abcam, ab22552), pH3S10 (1:500, Millipore, 06-570), TMEM119 (1:150, Abcam, ab209064), UTX (1:400, Cell Signaling Technology, 33510S), SOX9 (1:400, Millipore, AB5535), BrdU (1:250, Abcam, ab6326), and Cleaved Caspase 3 (1:400, Cell Signaling Technology, 9661S). BrdU labeling was performed by maternal intraperitoneal injection of 50 mg/kg BrdU (Sigma-Aldrich, B5002-250MG) prepared in PBS 2 h before embryonic dissection. BrdU-labeled sections were treated in 2 N HCl/PBS for 30 min at 37° before blocking. Sections were imaged with Zeiss axiovision software. Image stacks were deconvolved and z-projected. Measurements of bone lengths, differentiation domain lengths and area were quantified using ImageJ (https://imagej.nih.gov/ij/).

Statistical analysis for comparison between WT and Kmt2d^cKO samples were performed using an unpaired two-tailed Student's t-test to determine significant differences between groups (P-value<0.05). Asterisks denote significant differences compared to WT (*P<0.05, **P<0.01, ***P<0.001).

E9.5 MEFs were generated as previously described (Shpargel et al., 2012). For western blotting, nuclear lysates were prepared according to Invitrogen's nuclear extraction protocol. Western blotting was performed as previously described (Shpargel et al., 2012) with anti-KMT2D [1:1000, generated in Lee et al. (2013)] and anti-Nucleolin (1:5000, Bethyl Laboratories, A300-711A).

E14.25 palatal shelves were micro-dissected, and RNA was isolated with Trizol as directed (Thermo Fisher Scientific) from three biological replicates of WT or Kmt2d^cKO embryos for cDNA synthesis, ligation of Truseq adapters and library amplification (Kappa KK8580 mRNA HyperPrep Kit). Library samples were multiplexed for Nextseq 500 high output 25 bp paired-end sequencing. The quality of the sequence reads was evaluated using FastQC (http://www.bioinformatics.babraham.ac.uk/projects/fastqc/). Reads were mapped to the MM9 B6 genome using Tophat2 (Kim et al., 2013). We removed any sequencing data on the X and Y chromosomes due to sex differences in male and female embryos. Genic RNA-seq reads were counted with htseq-count (Anders et al., 2015) and analyzed by edgeR using the DESeq2 independent filtering method to determine the cut off for the minimum read counts per sample to identify significant (FDR<0.05) differential expression (Anders and Huber, 2010; Robinson et al., 2010). KMT2D-dependent gene expression was refined for genes that have higher levels of WT palate expression (RPKM>2) with more dramatic Kmt2d^cKO expression changes (logFC<−1). Ingenuity pathway analysis and the Molecular Signatures Database (MSigDB) were used to analyze mis-expressed pathways based on RNA-seq expression data (http://www.ingenuity.com/products/ipa, http://software.broadinstitute.org/gsea/msigdb/index.jsp). RNA-seq datasets were submitted to GEO (GSE149688). RNA-seq was verified by qRT-PCR (Bio-Rad SsoFast EvaGreen, CFX96 real time system). All RT-PCR was normalized to Gapdh expression and graphed relative to control samples.

All genotyping and qRT-PCR primers are listed (Table S2). Regions amplified by RT-PCR for subcloning, in vitro transcription and in situ hybridization are as follows, with T7 promoter underlined: T7-Kmt2d-R, TAATACGACTCACTATAGGGCCCTTGCCAAAGAAAGTATCTG; Kmt2d-F, AAATAACACAGGGTCACAGCCT; T7-Sox10-R, TAATACGACTCACTATAGGGCCAGAAGCATTGATTTTATTACTTGG; Sox10-F, CTGAATGTGGGAACTGGTCG; T7-Col8a2-R, TAATACGACTCACTATAGGGTCCCATCCAAACCTGGTTTG; Col8a2-F, CTCTACCGATGCTGCCAATG; T7-Pcolce-R, TAATACGACTCACTATAGGGACATCAGGCAGCTGTCCTAG; Pcolce-F, TCTCTTCTGAAGGGAACGAG.

The peer review history is available online at https://dev.biologists.org/lookup/doi/10.1242/dev.187997.reviewer-comments.pdf