Kdm6a^tm1d/+ mice exhibit growth deficiency

We utilized a previously described mouse model of KS2, Kdm6a^{tm1d(EUCOMM)Wtsi} [18], generated using the ‘knockout-first’ targeted construct created by the International Knockout Mouse Consortium (IKMC) [23] and referred to as Kdm6a^tm1d/+ henceforth. The tm1d allele results from Cre-mediated excision of Kdm6a exon 3, which leads to a frameshift and premature termination codon (PTC) in exon 6 of 29 (Fig 1A). Female heterozygous Kdm6a^tm1d/+ mice were viable and born at term. However, they experienced an increased rate of perinatal lethality, and survivors had significantly decreased body weight and length at 8 weeks compared to female Kdm6a^+/+ littermates (p = 0.007 and p = 0.001, respectively) (Fig 1B and 1C). Male pups hemizygous for Kdm6a were previously described to succumb perinatally [24], and we likewise observed complete perinatal lethality in Kdm6a^tm1d/y pups. For subsequent experiments, therefore, we focused on heterozygous female mice.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Kdm6a^tm1d/+ mice exhibit generalized growth deficiency.

(A) Schematic of the targeted Kdm6a allele as designed by the International Knockout Mouse Consortium, showing FRT sites as green arrows and loxP sites as orange arrows. Generation of the tm1d knockout allele requires excision of exon 3. Created using BioRender.com. Compared to Kdm6a^+/+ littermates, Kdm6a^tm1d/+ mice have significantly decreased (B) body weight (Kdm6a^+/+ n = 15; Kdm6a^tm1d/+ n = 15) and (C) body length (Kdm6a^+/+ n = 16; Kdm6a^tm1d/+ n = 15). Measurements obtained at 8 weeks of age. Blue circles: Kdm6a^+/+, red triangles: Kdm6a^tm1d/+. **p < 0.01, two-tailed unpaired Student’s t-test. All error bars represent mean ± 1 SD.

https://doi.org/10.1371/journal.pgen.1011310.g001

To assess growth abnormalities in Kdm6a^tm1d/+ mice, we first performed skeletal profiling. Kdm6a^tm1d/+ mice had shorter femurs and tibias than Kdm6a^+/+ littermates (p = 0.0003 and p = 0.001, respectively) at 8 weeks of age (Fig 2A and 2B). Micro-computed tomography (micro-CT) performed concurrently revealed long bone structural differences in both cortical and trabecular parameters (Fig 2C–2E). Kdm6a^tm1d/+ mice had decreased tissue area in the femoral mid-diaphysis (p = 0.005) (Fig 2D and 2F), which was significant even after normalization for femur length (p = 0.033) (S1 Fig). Despite the decreased tissue area, cortical thickness was unaffected in Kdm6a^tm1d/+ mice (Fig 2G), leading to a trend towards a higher bone area-to-tissue area percentage (p = 0.081) (Fig 2H). Trabecular volume was increased in Kdm6a^tm1d/+ mice (p = 0.030) (Fig 2E and 2I), secondary to a trend towards higher trabecular number (p = 0.097) (Fig 2E and 2J) and strikingly increased trabecular thickness in Kdm6a^tm1d/+ mice (p = 0.003) (Fig 2E and 2K). These changes suggest disturbances to long bone development due to the heterozygous loss of Kdm6a.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Kdm6a^tm1d/+ mice have shortened bones and altered skeletal parameters.

Kdm6a^tm1d/+ mice have significantly shorter (A) femurs and (B) tibias compared to Kdm6a^+/+ littermates (Kdm6a^+/+ n = 16; Kdm6a^tm1d/+ n = 15), which is also shown in (C) representative longitudinal micro-CT images of femurs and tibias from Kdm6a^tm1d/+ and Kdm6a^+/+. Representative transverse cross-sectional micro-CT reconstructions are displayed for femoral (D) cortical bone and (E) trabecular bone. Kdm6a^tm1d/+ mice have decreased (F) femoral cross-sectional tissue area. However, there is no significant difference in (G) cortical thickness or (H) bone area/tissue area percentage, despite Kdm6a^tm1d/+ mice displaying a slight upward trend in the latter. (I) Trabecular volume is increased in Kdm6a^tm1d/+ mice. While (J) trabecular number does not differ significantly, (K) trabecular thickness is increased in Kdm6a^tm1d/+ mice. For (F) through (K), Kdm6a^+/+ n = 12; Kdm6a^tm1d/+ n = 12. Blue circles: Kdm6a^+/+, red triangles: Kdm6a^tm1d/+. *p < 0.05, **p < 0.01, ***p < 0.001, two-tailed unpaired Student’s t-test. All error bars represent mean ± 1 SD. n.s., non-significant.

https://doi.org/10.1371/journal.pgen.1011310.g002

Growth plate structure is disrupted in Kdm6a^tm1d/+ mice

We examined whether perturbed bone development may stem from defects in the growth plate by performing hematoxylin and eosin (H&E) staining of proximal tibial growth plates sectioned in the longitudinal plane at 8 weeks of age (Fig 3A). Kdm6a^tm1d/+ mice had an overall decrease in growth plate height (p = 0.042) (Fig 3A and 3B). This was caused by a significant shortening of the hypertrophic zone in Kdm6a^tm1d/+ mice compared to wild-type (p = 0.0008), although the proliferative zone did not differ from wild-type littermates (Fig 3A and 3C). Cell counts in both the proliferative and hypertrophic zones were unchanged between the genotypes (Fig 3D); similarly, there was no difference in cell proliferation by EdU incorporation assay (S2 Fig). Instead, hypertrophic chondrocyte cell size, as measured by cross-sectional cell area in the longitudinal plane, was decreased in Kdm6a^tm1d/+ mice (p = 0.0006) (Fig 3E). Chondrocyte hypertrophy is the main driver of longitudinal bone growth [25]. These changes in the growth plate, particularly in the hypertrophic zone, suggest that an abnormality of chondrocyte function is involved in the pathogenesis of growth deficiency in KS2.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Growth plates have altered structure in Kdm6a^tm1d/+ mice.

(A) Hematoxylin and eosin staining of Kdm6a^tm1d/+ and Kdm6a^+/+ tibial growth plates. Left panels: 10x magnification. Insets: 40x magnification. (B) Growth plate height is decreased in Kdm6a^tm1d/+ mice. This can be attributed to (C) shorter hypertrophic zones (right panel) in Kdm6a^tm1d/+ mice, although proliferative zone heights do not differ (left panel). (D) Cell counts do not differ in either the proliferative (left) or hypertrophic (right) zones, but (E) cross-sectional cell size in the hypertrophic zone is smaller in Kdm6a^tm1d/+ growth plates. Kdm6a^+/+ n = 16; Kdm6a^tm1d/+ n = 13. Blue circles: Kdm6a^+/+, red triangles: Kdm6a^tm1d/+. *p < 0.05, ***p < 0.001, two-tailed unpaired Student’s t-test. All error bars represent mean ± 1 SD. n.s., non-significant. GP, growth plate. PZ, proliferative zone. HZ, hypertrophic zone.

https://doi.org/10.1371/journal.pgen.1011310.g003

Kdm6a^-/- chondrogenic cell lines exhibit precocious differentiation at both phenotypic and transcriptomic levels

To create a tractable in vitro model for chondrocyte differentiation and development in KS2, we chose to use ATDC5, a well-characterized female murine teratocarcinoma cell line that is known to undergo chondrogenesis when induced to differentiate [26–28]. CRISPR-Cas9 editing was used to target Kdm6a exon 6, introducing indels by non-homologous end joining (NHEJ) (Fig 4A). We isolated clones that remained unedited (Kdm6a^+/+) as controls, alongside clones bearing compound heterozygous frameshift mutations (Kdm6a^-/-). All such frameshift mutations led to premature termination codons (PTCs) upstream of the catalytic Jumonji C (JmjC) domain (S1 Table). For each clone, we excluded off-target edits by Sanger sequencing at the top 3 predicted off-target sites, as identified by partial complementarity to the gRNA (S3A–S3C Fig). Kdm6a^-/- lines do not express any KDM6A at the protein level, in contrast to Kdm6a^+/+ controls (Figs 4B and S4A). Intriguingly, Kdm6a^-/- cells did not exhibit higher global levels of H3K27me3 or altered H3K4me1 compared to Kdm6a^+/+ (Figs 4B and S4B–S4D). However, Kdm6a^-/- cells did have significantly higher global H3K4me3 (p = 0.030) (S4E and S4F Fig). We next differentiated the clones to chondrocytes over the course of 28 days and found that Kdm6a^-/- clones displayed enhanced chondrogenesis. This was visualized by increased Alcian blue staining, which labels glycosaminoglycans (GAGs) secreted by differentiated chondrocytes (Fig 4C) [29], and was subsequently quantified by increased absorbance at 605 nm (Fig 4D). Of note, we observed no difference in cell proliferation between genotypes with the MTT assay (S5 Fig); this fits with our findings in Kdm6a^tm1d/+ and Kdm6a^+/+ mouse growth plates (S2 Fig), which showed no difference in proliferative zone height or EdU incorporation between genotypes. Although we had observed a decrease in cross-sectional cell size in Kdm6a^tm1d/+ growth plates (Fig 3E), it was not possible to perform such measurements in vitro due to technical limitations imposed by excessive matrix secretion.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. Kdm6a^-/- cells exhibit dysregulated chondrogenesis.

(A) CRISPR-Cas9 gRNA target site for generating Kdm6a^-/- cell lines. Representative chromatograms from an unedited Kdm6a^+/+ clone and a compound heterozygous Kdm6a^-/- clone, which shows double peaks following the intended Cas9 cut site (blue line). (B) Kdm6a^-/- cells do not express KDM6A protein, and H3K27me3 levels are unaltered from Kdm6a^+/+ control lines. (C) Representative Alcian blue staining images (4x magnification) from a 28-day time course of Kdm6a^-/- and Kdm6a^+/+ cell lines. (D) Kdm6a^-/- cells lines have increased uptake of Alcian blue stain. Each line represents a single technical replicate, comprising the average of 3–4 biological replicates. (E) Principal component analysis of Kdm6a^-/- and Kdm6a^+/+ RNA-seq samples collected at Day 14 of differentiation (Kdm6a^+/+ n = 5; Kdm6a^-/- n = 5). Blue circles: Kdm6a^+/+, red triangles: Kdm6a^-/-. (F) Volcano plot for Kdm6a^-/- versus Kdm6a^+/+ samples, with MGI cartilage development genes highlighted in light blue. False discovery rate (FDR) = 0.1 (red dashed line), which corresponds to a p-value of 0.0055. (G) Wilcoxon rank sum test statistic for MGI cartilage development genes (light blue line, p < 1e-4) and simulated distribution of test statistics (gray). **p < 0.01, ***p < 0.001, two-tailed unpaired Student’s t-test. Cart. dev. genes, cartilage development genes.

https://doi.org/10.1371/journal.pgen.1011310.g004

To identify the transcriptional changes behind the chondrogenic phenotype in Kdm6a^-/- cells, we performed RNA-seq on five Kdm6a^+/+ and five Kdm6a^-/- clones at day 14 of chondrocyte differentiation. We chose this time point immediately prior to the increase in Alcian blue uptake in Kdm6a^-/- because changes in gene expression are expected to precede phenotypic manifestation. Principal component analysis showed that the clones clustered according to genotype, with principal component 1 (PC1) accounting for 44% of inter-sample variance (Fig 4E). A total of 736 genes were differentially expressed at the 10% false-discovery rate (FDR) level (S1 Appendix). Of these, 541 genes were downregulated in Kdm6a^-/- chondrocytes and only 195 genes were upregulated (Fig 4F), suggesting a skew towards repression of gene expression–or possibly, loss of activation. We verified that Kdm6a read counts were significantly lower in Kdm6a^-/- chondrocytes, although transcripts could still be detected, potentially due to a time delay before the initiation of nonsense-mediated decay. Kmt2d was not differentially expressed in Kdm6a^-/- chondrocytes at the transcriptional level; we were unable to assess at the protein level due to technical limitations.

We found perturbations in chondrogenic pathways at the transcriptional level in Kdm6a^-/- cells. From a list of 200 genes annotated by the Mouse Genome Informatics (MGI) database as involved in cartilage development, inclusive of both positive and negative regulators (see Methods) (S2 Appendix), 164 genes were expressed in our RNA-seq dataset, and 21 had significant fold-changes (10 upregulated and 11 downregulated). These genes tended to have greater absolute fold-change in expression level than non-chondrogenic genes, suggesting that cartilage development could be collectively dysregulated in Kdm6a^-/- cells (S6 Fig). To quantify this effect, we showed that the p-value rank-sum of this group of genes was significantly different from the distribution of p-value rank-sums obtained from selecting groups of 164 genes at random (p < 1x10^-4) (Fig 4G). Specific genes of interest include Col10a1, a marker of hypertrophic chondrocytes, which was significantly overexpressed in Kdm6a^-/- cells with a log₂(fold-change) of 2.45 (Fig 4F) [30]. Chst11, a carbohydrate sulfotransferase responsible for the production of chondroitin sulfate, the major proteoglycan component of cartilage [31], was also significantly upregulated. Finally, Sox6, a transcription factor known to regulate chondrogenesis by promoting hypertrophy and organization of the growth plate [32], was also overexpressed. A number of genes annotated in the MGI cartilage development list were also downregulated. Literature investigation of several prominently downregulated genes in our data revealed that these are negative regulators of chondrogenesis (Fig 4F). For example, RSPO2 suppresses chondrocyte differentiation in a human ligament cell model [33]. While Ogn is present at high transcript levels in a distinct murine hypertrophic chondrocyte model, its expression drives the cells towards an osteoblastic fate, and therefore it is downregulated in our chondrogenic model system, as expected [34]. Edn1 overexpression has been found to induce senescence in chondrocytes [35]. Altogether, these RNA-seq findings support that loss of KDM6A causes transcriptional alterations that promote excessive chondrogenesis both directly and indirectly.

We also performed Gene Ontology (GO) molecular function analysis of our RNA-seq results. Several significantly enriched GO classifiers initially looked appropriate, such as collagen binding and extracellular matrix structural constituent (S3 Appendix). Notably, most of the collagen genes within these categories actually had a negative fold-change in our data, such as Col6a4, Col6a6, Col7a1, Col8a1, and Col15a1. Upon further investigation in the literature, we found that these forms of collagen make up the matrix components in non-cartilaginous connective tissues, such as the basement membrane [36,37], vasculature [38], or the cornea [39]. The strong suppression of non-cartilage specific genes further suggests that Kdm6a^-/- cells are poised towards chondrogenesis. This stands in contrast to the aforementioned enhanced expression of Col10a1, a collagen type that is highly specific to hypertrophic chondrocytes in the process of endochondral ossification.

Although we wished to extend our RNA-seq findings in cell lines to our Kdm6a^tm1d/+ mouse model, we found no significant difference between Kdm6a^tm1d/+ and Kdm6a^+/+ mice in protein expression of COL10A1 and EDN1 by immunofluorescence in tibial growth plates (S7A–S7D Fig). However, this may be due to the lack of sensitivity of immunofluorescence as a quantitative technique.

While chondrogenesis was a natural process to study in growth regulation, we were also interested in looking for the transcriptional signatures of other biological processes that could be perturbed by the loss of Kdm6a. Given our finding of decreased cell size in Kdm6a^tm1d/+ growth plates, we performed similar p-value rank-sum analysis for positive and negative regulators of cell growth (S4 Appendix). We did not find significant dysregulation in these groups of genes (S8A and S8B Fig). Because control of cell growth and hypertrophy can vary widely between cell types, some genes annotated by MGI may not be relevant in chondrocytes, thus diluting the power of this kind of collective analysis. However, Igf1, a major driver of cell size, was notably downregulated in Kdm6a^-/- cells (S1 Appendix). Further studies are needed to clarify the contribution of the Igf1 pathway towards the growth phenotype in KS2.

Kmt2d^-/- and Kdm6a^-/- cell lines share a transcriptional profile indicative of excessive chondrogenesis

The similarities of precocious differentiation in both KS1 and KS2 chondrocytes prompted a direct comparison of their transcriptional profiles. We had previously generated Kmt2d^-/- and control Kmt2d^+/+ clones from the same ATDC5 parental cell line by CRISPR-Cas9 gene editing. RNA-seq at Day 7 of chondrogenic differentiation revealed striking differential gene expression in Kmt2d^-/- lines compared to Kmt2d^+/+ [22]. To account for the difference in timepoints and minimize batch effects, we concurrently cultured Kmt2d^-/- and Kdm6a^-/- lines and their respective wild-type control lines. Poly-adenylated RNA samples were sequenced from all samples at both Days 7 and 14 of differentiation. We first validated these results with our prior datasets. We stratified genes on the basis of significance at an FDR cutoff of 10% in our previous data, obtained at Day 7 for KS1 [22]. In our new dataset, we contrasted Day 7 Kmt2d^-/- samples with Day 7 Kmt2d^+/+, and the new experiment indicates a strong enrichment of genes that were also significant in prior data (S9A Fig). Furthermore, the fold-change directionality was concordant for 864 out of 866 genes significant in both datasets (99.8%) (S9B Fig). For KS2, we performed a similar analysis stratifying genes based on prior data collected at Day 14, and again noted significant enrichment for the same genes that were previously identified (S9C Fig), as well as high concordance in the direction of fold-change (403 out of 404 genes significant in both data sets; 99.8%) (S9D Fig). As expected, at both Day 7 and Day 14, significantly fewer reads mapped to Kmt2d in the Kmt2d^-/- lines compared to Kmt2d^+/+, while Kdm6a transcriptional expression was unchanged. Altogether, this confirmed that our new data aligns with prior findings where applicable, and allowed us to directly compare KS1 and KS2.

To investigate how factors such as genotype and differentiation stage affected sample clustering, we performed principal component analysis (PCA). This indicated separation roughly based on genotype, with principal component 1 (PC1) accounting for 32% of variance (Fig 5A). Interestingly, there was substantial variation among wild-type samples. Kmt2d^+/+ lines that were isolated as matched controls for Kmt2d^-/- clones appeared to separate along principal component 2 (PC2, 18% of variance) away from Kdm6a^+/+ controls isolated contemporaneously with Kdm6a^-/- lines, despite the wild-type status of both groups. We have previously confirmed by Sanger sequencing that Kmt2d^+/+ and Kdm6a^+/+ lines do not bear edits in the targeted loci, or in the top predicted CRISPR/Cas9 off-target sites (S3A–S3C Fig). Furthermore, PC1 accounts for nearly twice the variance as PC2 (32% versus 18%). However, in case of divergence between the two sets of wild-type samples over the course of passaging, we elected to make all further comparisons between knockout lines only with their respective matched wild-type controls, henceforth referred to as Kmt2d^+/+ for KS1 and Kdm6a^+/+ for KS2.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 5. Kmt2d^-/- and Kdm6a^-/- chondrocytes bear similar transcriptomic profiles.

(A) Principal component analysis of Day 7 and Day 14 RNA-seq samples from Kmt2d^-/- (KS1; n = 2), Kdm6a^-/- (KS2; n = 3), and wild-type control cell lines Kmt2d^+/+ (KS1; n = 3) and Kdm6a^+/+ (KS2; n = 3), respectively. Blue symbols represent wild-type; red symbols represent knockout lines. Circles represent Day 7; triangles represent Day 14. Open symbols represent KS1 (Kmt2d^-/- and Kmt2d^+/+) and solid symbols represent KS2 (Kdm6a^-/- and Kdm6a^+/+). (B) Conditional p-value histogram displaying p-values from the contrast of Kmt2d^-/- versus Kmt2d^+/+ at Day 14 (KS1 D14), stratified by gene-wise significance at Day 14 in the contrast of Kdm6a^-/- versus Kdm6a^+/+ (KS2 D14; orange line, p-adj < 0.1) or non-significance (black line, p-adj > 0.1). (C) Scatter plot of gene-wise log₂(fold-changes) at Day 14 in the Kdm6a^-/- versus Kdm6a^+/+ contrast (KS2) and the Kmt2d^-/- versus Kmt2d^+/+ contrast (KS1). Orange: gene significant in both KS1 and KS2; gray: gene is significant in either KS1 or KS2, but not both; blue: gene is annotated in MGI as involved in cartilage development. Identity line (x = y) is displayed in dark blue. The (D) conditional p-value histogram is also displayed for Day 7, as is the (E) scatter plot of gene-wise log₂(fold-changes) at Day 7. (F) Scatter plot of gene-wise log₂(fold-changes) between Day 14 and Day 7, for KS1 and KS2. Inset panel shows a magnified area of the plot.

https://doi.org/10.1371/journal.pgen.1011310.g005

Notably, the PCA plot also showed that, for the most part, Kmt2d^-/- samples cluster tightly with Kdm6a^-/- samples regardless of timepoint (Fig 5A). To determine the overlap in gene expression patterns between KS1 and KS2 at Day 14, we compared Kmt2d^-/- and Kdm6a^-/- lines against Kmt2d^+/+ and Kdm6a^+/+ samples, respectively (differentially expressed genes are listed in S5 and S6 Appendices). Gene-wise p-values in KS1 were stratified by their significance in KS2 at an FDR cutoff of 10% (Fig 5B). This revealed that genes with low p-values in KS2 are likewise enriched for low p-values in KS1, suggesting a high degree of similarity between the chondrocyte transcriptomes of the two disorders. The fold-change directionality from wild-type was also extremely concordant between KS1 and KS2; of the 578 genes significant in both conditions, 139 genes were upregulated in both KS1 and KS2 and 362 of genes were downregulated, leaving only 77 genes discordant between the two conditions (Fig 5C). Mutually upregulated genes of interest include: Acan, which encodes aggrecan, a major proteoglycan component in cartilage; Matn1 and Matn3, both encoding matrilin proteins that are part of the cartilage matrix; and several collagen genes including Col10a1 and Col11a1. We also found a high degree of overlap in differentially expressed genes between KS1 and KS2 at Day 7 (Fig 5D; S7 and S8 Appendices), although fold change directionality was generally less concordant at Day 7 than at Day 14 (Fig 5E). Acan, Col10a1, Col11a1, and Col2a1, an early marker for chondrogenesis [40], were significantly upregulated in both conditions at Day 7. These results show that KS1 and KS2 have similar transcriptomic profiles, which support increased chondrocyte differentiation compared to wild-type controls.

We had performed RNA-seq at two time points, on Day 7 and Day 14 of differentiation, in order to examine the progression in gene expression in KS1 and KS2. This allowed us to assess whether the chondrocyte differentiation program differs between KS1 and KS2. We compared Kmt2d^-/- and Kdm6a^-/- samples collected at Day 14 against those collected at Day 7, and there was increased expression of a number of chondrogenic and osteogenic genes at the later time point (Fig 5F). Fgf9, a transcription factor that promotes chondrocyte hypertrophy [40], had 38.8-fold higher expression at Day 14 in KS1 and 79.8-fold increase in KS2, compared to Day 7. Bglap and Bglap2, which encode calcium-chelating protein hormones produced exclusively in bone tissue, were also mutually upregulated–particularly so in KS1. The overall pattern of gene expression changes from Day 7 to Day 14 indicated an increasingly chondrogenic and osteogenic environment in both Kmt2d^-/- and Kdm6a^-/- cells. Interestingly, the similarity in gene-wise fold-changes between KS1 and KS2 is greater when considering the change between time points rather than within a given time point (Fig 5C, 5E, and 5F). For instance, there is a collection of genes that are significantly upregulated in KS1 but downregulated in KS2 at Day 7, and vice versa (Fig 5E). However, with respect to the directionality of fold-change over the course of differentiation from Day 7 to Day 14, almost all significant genes are concordant between KS1 and KS2 (Fig 5F). This suggests that although individual genes may be perturbed in opposing directions at a given time point compared to wild-type, the overall trajectory of gene expression across time in KS1 and KS2 is driven by a prevailing chondrogenic differentiation program that is largely shared between the two disorders.

Kdm6a^tm1d/+ mice display a defect in longitudinal and appositional bone growth

Bone growth is closely coupled with mechanical load. Mice with lower body mass and smaller skeletal frames, as we see with the Kdm6a^tm1d/+ genotype, thus proportionally have lower femoral cross-sectional tissue area. However, we showed that Kdm6a^tm1d/+ mice have a significantly decreased tissue area even after normalization for shorter femur length. As tissue area can only be expanded by osteoblast activity at the periosteal surface, the lack of appropriate tissue area growth suggests that osteoblasts may be another afflicted cell type in KS2. Interestingly, femoral cortical thickness was unchanged in Kdm6a^tm1d/+ mice and trabecular volume was increased. Decreased resorptive capability by osteoclasts at the endosteal and trabecular surfaces may also contribute towards the skeletal phenotype in KS2. Although we focus on chondrocytes here, our in vivo findings suggest that future studies of other cell types involved in bone growth–specifically osteoblasts and osteoclasts–would be important to expand understanding of growth deficiency in KS2.

Recombinant human growth hormone (rhGH) is a potential therapy for KS2-associated growth deficiency

Our ultimate goal is to simultaneously treat the multiple syndromic features of KS2 by addressing the underlying molecular etiology. Specifically regarding growth, however, we believe an FDA-approved therapy, recombinant human growth hormone (rhGH), may also be efficacious in treating KS2-associated growth deficiency. Van Montfort et al. have shown that rhGH treatment improved linear growth in three individuals with KS2 and 13 individuals with KS1 [42]. Our data now provide mechanistic support for the use of rhGH in KS2 by showing that Igf1 expression is reduced in Kdm6a^-/- chondrocytes and that growth plate height is reduced in Kdm6a^tm1d/+ mice due to reduced hypertrophic chondrocyte size. This is notable because growth hormone (GH) increases linear growth both directly and indirectly through IGF1 [43,44]. In rats, GH increases Igf1 transcript levels in the hypertrophic zone of the growth plate [45]. GH also increases growth plate height and longitudinal bone growth in mice that have undergone postnatal ablation of Igf1r, encoding the IGF1 receptor [46]. Therefore, rhGH would be expected to improve bone growth and height if growth deficiency in KS2 can indeed be attributed towards reduced IGF1, as our data suggest.

Overlap between KS1 and KS2 is apparent in gene expression profiles

An extension of the Balance Hypothesis [1] supposes that impaired function of distinct components of the epigenetic machinery may yet converge upon the same phenotype by achieving a shared chromatin state and gene expression. We found that despite the distinct genetic etiologies of KS1 and KS2, downstream gene expression patterns appear to be shared, which has also been noted by others. Luperchio et al. used RNA-seq on B lymphocytes isolated from mouse models of KS1 and KS2 and found a high degree of overlap in differential gene expression [18]. Several groups have developed a genome-wide DNA methylation signature in peripheral blood that can differentiate individuals with KS1 from healthy controls [47–49]. Interestingly, this signature also identifies KS2 samples due to the high degree of overlap, and has even aided in diagnosis [47,50,51]. As gene expression is dependent upon DNA methylation status, this suggests that there is a shared ‘Kabuki transcriptome’ in peripheral blood. Relative to wild-type controls, we showed striking concordance in gene expression between Kmt2d^-/- and Kdm6a^-/- cells at multiple stages of chondrocyte differentiation. The transcriptional signatures of these genotypes were both suggestive of precocious chondrogenesis. This also matched our cellular phenotype of increased Alcian blue staining in Kdm6a^-/- lines, described here, and our previously published findings in Kmt2d^-/- lines [22].

KS1 and KS2 may lead to growth deficiency through distinct pathways

Several pieces of data indicate that the mechanism of growth deficiency in KS1 may still differ from KS2 in subtle ways. Our prior work identified Shox2 as a direct target of KMT2D [22]. Shox2 is expressed in mouse growth plates, and conditional loss of Shox2 in chondrocytes causes shortened limbs [52,53]. Kmt2d^-/- cells exhibited depletion of H3K4me3 at the Shox2 promoter relative to Kmt2d^+/+ lines, concurrent with a 4-fold downregulation of Shox2 expression. The decrease in SHOX2 led to a disinhibition of Sox9 expression in Kmt2d^-/- cells. As SOX9 is a pro-chondrogenic transcription factor [54,55], we proposed that this pathway is implicated in the precocious differentiation of chondrocytes, which we have shown to be a common feature of KS1 and KS2. Yet in Kdm6a^-/- cells, neither Shox2 nor Sox9 is differentially expressed relative to Kdm6a^+/+ controls, at either Day 7 or Day 14 of chondrogenic differentiation. This suggests that the Shox2-Sox9 pathway is not a driving factor in precocious chondrogenesis in KS2. Notably, even in Kmt2d^-/- cells, repletion of Shox2 by lentiviral overexpression was insufficient to restore Alcian blue staining to wild-type levels [22], indicating that additional pathways–possibly shared between KS1 and KS2 –contribute towards this phenotype.

A caveat to these conclusions is that our Kmt2d^+/+ and Kdm6a^+/+ control lines, both previously confirmed to be wild-type, appear to differ at baseline. Whereas this may be due to passaging effects, as the clones were derived on separate instances, or to the slightly different CRISPR methods used to generate the lines, it remains possible that the aforementioned gene expression differences between KS1 and KS2 could be an artifact of our imperfect cell models. However, detailed examination of skeletal profiles in KS1 and KS2 mouse models also reveals differences despite the preserved global picture of growth deficiency. Compared to wild-type littermates, Kmt2d^βgeo/+ and Kdm6a^tm1d/+ mice have significantly lower body weight and body length, and shorter femurs and tibias [22]. Unlike Kdm6a^tm1d/+ mice, however, Kmt2d^βgeo/+ mice did not have any significant difference in tissue area, and trabeculae were thinner compared to wild-type. In the tibial growth plate, Kdm6a^tm1d/+ mice had reduced hypertrophic chondrocyte size, leading to a significant decrease in hypertrophic zone and overall growth plate height. This was the opposite of growth plate findings in Kmt2d^βgeo/+ mice. These data definitively point to chondrocytes as a critical cell type in the pathogenesis of growth deficiency for both KS1 and KS2. However, it remains a question how to reconcile opposing growth plate data with the ultimate shared phenotype of shortened femurs and tibias.

Features of KS1 and KS2 mice recur in mouse models of other genetic diseases

We are not aware of other mouse models with growth plate findings identical to KS1 or KS2 mice that definitively suggest a common mechanism. However, there are reported mouse models that resemble certain aspects of each. In fact, both shortened and expanded growth plates have occurred in association with decreased long bone length. A thanatophoric dysplasia type 1 mouse model containing the hypermorphic Fgfr3^Y367C/+ variant showed shortened long bones and reduced hypertrophic zone height due to reduced cell size [56], similar to our Kdm6a^tm1d/+ mice. However, the growth plates were also severely disorganized, unlike in Kdm6a^tm1d/+ mice. Of note, the genes encoding two FGFR3 ligands, Fgf1 and Fgf9, are upregulated in KS1 and KS2 chondrocytes, supporting a role in activation of this pathway in both disorders. Similarly, we are not aware of a disorder that completely mimics our findings in KS1 mice. Dmp1^-/- mice partially resemble KS1, with shortened long bones and expanded growth plates involving both the proliferative and hypertrophic zones [57]. Distinct from our previous findings in KS1 [22], Dmp1^-/- mice also show reduced apoptosis of hypertrophic chondrocytes [57].

Chondrocyte-to-osteoblast transdifferentiation remains to be explored in KS1 and KS2

Altered cell differentiation is a common disease process in this class of disorders, as described in other MDEMs [22,58,59]. Cell type-specific gene expression is generally controlled through epigenetics, and prior studies indicate that disruptions to the epigenetic machinery impact cell fate and commitment [60]. The classical model of endochondral ossification describes apoptosis of hypertrophic chondrocytes and subsequent invasion of the residual cartilage template by osteoblasts. Accelerated chondrocyte differentiation, as appears to occur in KS1 and KS2, could therefore impact downstream processes of bone formation. Not only might an altered epigenome hasten or delay differentiation, it is also conceivable that the barriers between cell types may become increasingly fluid. Chondrocytes and osteoblasts were originally believed to be mutually exclusive, committed cell fates resulting from differentiation of mesenchymal stem cells [61]. However, lineage tracing experiments using transgenic mice have now led to several hypotheses for chondrocyte-to-osteoblast transdifferentiation [62]. Whereas some groups propose transdifferentiation at the immature chondrocyte stage [63], others suggest dedifferentiation of hypertrophic chondrocytes followed by redifferentiation [64–66], and yet additional data may support direct transdifferentiation from hypertrophic chondrocytes into mature osteoblasts [67–69]. A disruption in this complex interplay of cell fates could plausibly contribute to growth plate differences in KS1 and KS2. We noted an extreme upregulation of Bglap and Bglap2 between Day 7 and Day 14 of chondrocyte differentiation in Kmt2d^-/- cell lines in particular. Alternatively named Osteocalcin, these genes are regarded to be osteoblast-specific [70], suggesting that transdifferentiation may be occurring. Kdm6a^-/- and wild-type cell lines also increased expression of these genes over the course of differentiation, although to a lesser degree. ATDC5 cells are an established in vitro model system for chondrogenesis [26–28]. However, it would be ideal to further investigate in vivo whether KS1 and/or KS2 change the potential for chondrocyte-to-osteoblast transdifferentiation, as an imbalance in these cell types would impact the bone environment and appositional bone growth.

KMT2D and KDM6A co-exist in the COMPASS complex

The Balance Hypothesis is an appealing explanation for the clinical resemblance of KS1 and KS2, but it is important to consider alternative hypotheses as well. In support for the former, Björnsson et al. have shown that Kmt2d^βgeo/+ mice have genome-wide depletion of H3K4me3, which is the expected direct consequence of reduced KMT2D activity [71]. This was reversible upon treatment with a histone deacetylase inhibitor (HDACi), AR-42, which also ameliorated neurologic findings of KS1 in the mice. These data, in combination with the Kabuki syndrome DNA methylation signatures [47–49], point towards an epigenetic etiology for these disorders. Luperchio et al. also found that Kmt2d^βgeo/+ and Kdm6a^tm1d/+ B lymphocytes both tend towards increased promoter accessibility [18]. On the one hand, this is supportive of KS1 and KS2 having a shared chromatin state that contributes to overlapping gene expression, which we saw in our RNA-seq data comparing Kmt2d^-/- and Kdm6a^-/- cells. Yet, loss of either a writer of H3K4me3 or an eraser of H3K27me3 would be expected to have the opposite effect on chromatin accessibility. Two potential possibilities (which are not mutually exclusive) include: 1. An indirect effect that is downstream of KMT2D or KDM6A catalytic activity drives the determination of chromatin state, or 2. The role of KMT2D and/or KDM6A in the pathogenesis of Kabuki syndrome stems from a non-catalytic function in one or both of these proteins.

Our western blot results show that global H3K27me3 levels are unaltered even in Kdm6a^-/- cells with no residual KDM6A protein. This is not entirely unsurprising, as the KDM6 family also includes KDM6B, which shares 70% protein sequence identity with KDM6A and also has H3K27 demethylase activity [12]. However, it raises the possibility that the predominant role of KDM6A in KS2 may be non-catalytic. KMT2D and KDM6A coexist in the COMPASS complex, which broadly serves as an antagonist to gene repression by Polycomb group proteins throughout development [72]. The reduced dosage of either KMT2D or KDM6A could disrupt stoichiometry or destabilize the COMPASS complex–and perhaps the shared features of KS1 and KS2 are rooted in the destabilization of COMPASS rather than the individual enzymatic functions of KMT2D and KDM6A. KDM6A is destabilized in Kmt2d^-/- mouse embryonic stem cells, suggesting interdependency of KDM6A upon KMT2D expression [73]. Due to technical limitations, we were unable to verify the converse at the protein level, although we noted that Kmt2d expression is unaltered at the transcriptional level in KS2 chondrocytes. Interestingly, global H3K4me3 is increased in Kdm6a^-/- cells. Given that KMT2D is a writer of H3K4me3, this result initially appears to contradict the hypothesis of interdependency between KDM6A and KMT2D. However, KMT2D is not the sole methyltransferase to target H3K4 [74], and it remains possible that gene expression changes resulting from loss of KDM6A and/or KMT2D could impact another component of the epigenetic machinery, thus indirectly upsetting the balance of global H3K4me3 levels.

Domain-swapping experiments have shown that KDM6A has a non-catalytic capacity to recruit the histone acetyltransferase p300 to co-localize with COMPASS at enhancer elements [75]. Likewise, catalytic-dead KDM6A is sufficient to rescue differentiation in Kdm6a-knockout mouse embryonic stem cells [24,76], establishing that KDM6A plays important non-enzymatic roles. It would be interesting to consider the formation and genomic localization of the COMPASS complex in the absence or reduced dosage of KMT2D, as well as KDM6A. Should the catalytic role of KMT2D also be implicated in the pathogenesis of KS2, it may unlock the possibility of treatment with pharmacological and/or dietary therapies, which have already been shown to improve neurological outcomes in Kmt2d^βgeo/+ mice [71,77,78].

Summary

In this work, we aimed to further understand the mechanisms involved in growth deficiency for two related MDEMs, KS1 and KS2. We showed that while skeletal and growth plate findings differ in certain aspects between mouse models of KS1 and KS2, chondrocytes bearing Kmt2d or Kdm6a null mutations exhibit a shared phenotype of precocious differentiation. Importantly, the similarity in transcriptional profiles raises the tantalizing possibility of developing a treatment to target both disorders.

Ethics statement

The Johns Hopkins Institutional Animal Care and Use Committee approved all animal studies in this project. Procedures were performed following standards described in the NIH Guide for the Care and Use of Laboratory Animals (National Academies Press, 2011).

Animals

The Kdm6a^{tm1a(EUCOMM)Wtsi} allele (tm1a) was designed by the International Knockout Mouse Consortium (IKMC) and consisted of a FRT-flanked LacZ-neomycin resistance cassette (βgeo), upstream of the loxP-flanked Kdm6a exon 3. Kdm6a^tm1a/+ females were obtained by the Björnsson lab from the European Mutant Mouse Archive (EMMA) and crossed with B6.Cg-Tg(ACTFLPe)9205Dym/J males (Strain No. 005703, The Jackson Laboratory), which express FLP1 driven by the Actb promoter. This excised the βgeo cassette and generated the tm1c allele. Kdm6a^tm1c/tm1c females were next crossed with B6.C-Tg(CMV-cre)1Cgn/J males (Strain No. 006054, The Jackson Laboratory), which carry a human CMV-driven cre transgene integrated on chromosome X. Cre-mediated excision of Kdm6a exon 3 resulted in the tm1d allele. Kdm6a^tm1d/+ females were backcrossed to C57BL/6J males (Strain No. 000664, The Jackson Laboratory) to remove FLP1 and cre transgenes and to maintain the strain subsequently. Due to the perinatal lethality of Kdm6a^tm1d/y males, the tm1d allele can only be maintained in the female line. Therefore, all experimental animals were bred by crossing Kdm6a^tm1d/+ females with wild-type C57BL/6J males. Genotyping was performed by PCR. All mice were in the care of Johns Hopkins Research Animal Resources. Up to 5 mice were housed in each barrier cage with ad libitum access to autoclaved feed (Envigo Teklad 2018SX) and reverse osmosis-filtered, hyperchlorinated water. A cotton square nestlet (Envigo Teklad 6060/6105) and autoclaved corncob bedding (Envigo Teklad 7092/7097) were provided for each cage. Cages were changed every 2 weeks under aseptic conditions. Cage racks were ventilated with HEPA-filtered and humidified air. Animal rooms were maintained on standard light/dark cycles at regulated temperatures. Mice were euthanized by halothane inhalation (Sigma-Aldrich B4388), following the AVMA Guidelines for the Euthanasia of Animals, 2020 edition (https://olaw.nih.gov/policies-laws/avma-guidelines-2020.htm).

High-resolution micro-computed tomography

Mice were euthanized at 8 weeks of age. Femurs and tibias were dissected and fixed with 4% paraformaldehyde in 1x PBS for 24–48 hours at 4°C, then transferred to 70% ethanol. Digital calipers were used to measure bone length. Imaging was performed following the guidelines of the American Society for Bone and Mineral Research [79]. A Bruker Skyscan 1172 desktop microcomputed tomography system was used to scan bones at 65 keV and 152 μA with a 0.5 mm aluminum filter at an isotropic voxel size of 10 μm. Image reconstruction was performed with nRecon (Bruker), and analysis was done with CTAn software (Bruker). A 500 μm region of interest (ROI) centered around the femoral mid-diaphysis was assessed for cortical bone measurements. Trabecular measurements were collected 500 μm proximal to the distal femoral growth plate in a 2 mm ROI. All micro-CT imaging and measurements were carried out under blinded conditions.

Growth plate measurements

Femurs and tibias dissected from 8-week-old mice were fixed with 4% paraformaldehyde in 1x PBS for 24–48 hours at 4°C, then decalcified by continual agitation in a 14% EDTA solution, pH 7.4 (Bio-Rad 1610729) at 4°C. Decalcifying solution was refreshed every 24–48 hours until bones were pliable. Bones were processed by the Johns Hopkins Hospital Reference Histology Laboratory and embedded in paraffin blocks. Hematoxylin and eosin staining (H&E) was performed on mounted 5 μm sections of the proximal tibial growth plates. Images were taken with a Nikon 80i microscope under brightfield illumination at 10, 20, and 40x magnification and analyzed using NIS elements software to obtain height measurements of the proliferative zone (PZ), hypertrophic zone (HZ), and growth plate (GP). Measurements were averaged over 3 sites per section, for 4 sections per sample. Cell counts were obtained in the proliferative and hypertrophic zones and averaged for each mouse. Cell size was assessed for hypertrophic chondrocytes using ImageJ and averaged over 15 cells per section, 4 sections per mouse (60 cells total). All imaging and growth plate measurements were performed blinded to the genotype.

EdU incorporation assay

Kdm6a^tm1d/+ and Kdm6a^+/+ mice were injected with 150 mg/kg of EdU at the age of 8 weeks. Twelve to sixteen hours post-injection, the mice were euthanized. The long bones were dissected and fixed with 4% paraformaldehyde in 1X PBS for 36 hours, before demineralization with 14% EDTA in 1X PBS as described above. Bones were embedded in paraffin blocks and 5 μm sections of proximal tibial growth plates were adhered to glass slides. Sections were deparaffinized in xylene and rehydrated in decreasing concentrations of ethanol. EdU incorporation was detected using the Click-iT EdU Cell Proliferation Kit (ThermoFisher Scientific C10337) and nuclei were counterstained with Hoechst dye.

Growth plate immunofluorescence

Paraffin-embedded longitudinal proximal tibial growth plate sections were deparaffinized and rehydrated. Antigen retrieval was performed with 0.05% citraconic anhydride at pH 7.4 [80]. Slides were then rinsed and permeabilized, blocked with 5% goat serum and then incubated overnight in primary antibody. Following washes, slides were then incubated in secondary antibody. Excess antibody was washed off before slides were stained with DAPI. Quantification was performed by taking the total fluorescence intensity per DAPI-stained cell count in the growth plate in ImageJ. Background subtraction was performed using the rolling ball algorithm with radius size 150 pixels and cell counts were performed in ImageJ. All antibodies are listed in S2 Table. All images and measurements were performed blinded.

ATDC5 cell culture

ATDC5 cells were sourced from the European Collection of Authenticated Cell Cultures (Sigma-Aldrich 99072806). Cells were propagated at 37°C, 5% CO₂, in ATDC5 Complete Media. This consisted of DMEM/F-12 (ThermoFisher Scientific 11320033) supplemented with 5% heat-inactivated fetal bovine serum (ThermoFisher Scientific 16140071), 2 mM L-glutamine (Corning 25-005-CI), 100 U/mL penicillin and 100 μg/mL streptomycin (ThermoFisher Scientific 15140122). Media was changed every 48–72 hours.

Chondrogenic differentiation

24 hours post-seeding, media was exchanged for ATDC5 Complete Media supplemented with 0.05 mg/mL L-ascorbic acid (Sigma-Aldrich A4403), 10 mM β-glycerophosphoric acid (ThermoFisher Scientific 410991000), and 1x Insulin-Transferrin-Selenium (ITS-G) (ThermoFisher Scientific 41400045). Cells were cultured for up to 28 days at 37°C, 5% CO₂, with supplemented media changes every 48–72 hours.

Generation of Kdm6a^-/- cell lines

The crRNA was designed to target exon 6 of Mus musculus Kdm6a with the following guide sequence: 5’-TTCCTTGGCTCGACAAAAGC-3’. Alt-R CRISPR-Cas9 tracrRNA and Alt-R custom Cas9 crRNA (Integrated DNA Technologies) were mixed at an equimolar ratio in Nuclease-Free Duplex buffer (Integrated DNA Technologies 11-01-03-01) and allowed to anneal into duplexes. Alt-R S.p. Cas9 Nuclease V3 (Integrated DNA Technologies 1081058) was introduced for duplex loading. The resulting ribonucleoprotein (RNP) complex was incubated with CRISPRMAX (ThermoFisher Scientific CMAX00001) and Opti-MEM (ThermoFisher Scientific 31985062) for formation of transfection complexes. Meanwhile, ATDC5 cells were rinsed with 1x PBS, dissociated with 0.25% trypsin, 2.21 mM EDTA (Corning 25-053-CL), and collected for counting. Cells were seeded at 2x10⁵ cells per well in 6-well plates concurrently with introduction of the RNP transfection complexes. After 48 hrs at 37°C, 5% CO₂, cells were washed and trypsinized. A sample was lysed with QuickExtract DNA Extraction Solution (Biosearch Technologies QE09050) to assess editing efficiency with the Alt-R Genome Editing Detection Kit (Integrated DNA Technologies 1075932). Remaining cells were seeded either in 96-well plates at 1 cell per 50 μL (limiting dilution cloning), or in 10cm dishes at low densities to obtain single colonies. After 10 days of growth, cells were washed with 1x PBS and underwent mild dissociation in warm trypsin for 1 minute. Single colonies were visually identified, isolated, and expanded to establish clonal lines. Successfully edited lines and unedited control lines were identified using the Alt-R Genome Editing Detection Kit and verified by Sanger sequencing. Forward primer: 5’-AAGATAGAGTGCAGTGGGTTG-3’. Reverse primer: 5’-CAGAAGTCCAAATGCCTTGTAAAT-3’.

Generation of Kmt2d^-/- cell lines

Please refer to Fahrner et al., 2019 [22].

Western blot

Cells were washed with cold 1x PBS twice and lysed with RIPA buffer containing protease inhibitor (ThermoFisher Scientific A32953) to isolate total protein. Histones were extracted using a kit following manufacturer’s protocols (Abcam ab113476). The Pierce BCA Protein Assay kit (ThermoFisher Scientific 23225) was used to measure protein concentrations, and 10 μg of each sample was loaded onto a NuPAGE 4 to 12%, Bis-Tris gel (ThermoFisher Scientific NP0336BOX). Following transfer of proteins, PVDF membranes were blocked with 0.5x Intercept (PBS) Blocking Buffer (LI-COR Biosciences 927–70001) for 1 hour at room temperature. Membranes were stained with primary antibodies overnight at 4°C and repeatedly washed with 1x PBS with 0.1% Tween-20 (PBS-T) the following day. Afterwards, membranes were incubated in secondary antibody for one hour at room temperature. Excess antibody was removed with PBS-T washes. Imaging was performed with the LI-COR Odyssey. All antibodies are listed in S2 Table. Image Studio version 5.2 was used to quantify blots by densitometry.

Alcian blue staining

Cells were seeded at a density of 1x10⁵ cells per well in 6-well plates (Corning 353046). Day 7–28 wells underwent chondrogenic differentiation for the specified number of days; Day 0 wells were collected 24 hours post-seeding. Cells were rinsed with 1x PBS and fixed for 5 minutes in 10% neutral buffered formalin (Epredia 9990244). After 3 washes in 1x PBS to remove residual fixative, cells were stained with 1% Alcian blue solution in 3% acetic acid, pH 2.5 (Sigma-Aldrich B8438) for 1 hour in the dark. Unbound dye was removed with 3 washes in 1x PBS. Whole-well images were captured over a transilluminator. An EVOS FL Auto Imaging System (ThermoFisher Scientific) was used to obtain images at 4X, 10X, and 20X magnification. For quantification of the Alcian blue stain, well contents were lysed in a 10% sodium dodecyl sulfate solution (Sigma-Aldrich L3771) for 20 minutes at minimum. Absorbance was measured at 605 nm with a BioTek Synergy 2 Multi-Mode Microplate Reader.

MTT assay

Cells were seeded at a density of 1x10³ cells per well in 96-well plates (Corning 353072). Day 7–28 wells underwent chondrogenic differentiation for the specified number of days; Day 0 wells were collected 24 hours post-seeding. 50 μL of ATDC5 Complete Media and 10 μL of CellTiter 96 AQueous One Solution (Promega G3580) were added to each well. Plates were incubated at 37°C, 5% CO₂ for 1.5 hours to allow colorimetric conversion to the formazan product. Absorbance was measured at 490 nm with a BioTek Synergy 2 Multi-Mode Microplate Reader.

RNA isolation

Cells were seeded at a density of 1x10⁵ cells per well in 6-well plates (Corning 353046) and underwent chondrogenic differentiation for the specified number of days. Samples were collected with TRIzol reagent (ThermoFisher Scientific 15596026) and chloroform. Following phase separation, the aqueous phase containing RNA was put through the RNA Clean & Concentrator-5 kit (Zymo Research R1013) with DNase I digestion to remove genomic DNA contamination.

RNA-sequencing library preparation

RNA quantity was measured by Qubit RNA Broad Range Assay (ThermoFisher Scientific Q10210). RNA quality was assessed on an Agilent Fragment Analyzer by the Johns Hopkins Single Cell & Transcriptomics Core (JH SCTC) Facility. 1 μg of total RNA per sample was used as input for the NEBNext Ultra II RNA Library Prep kit with Sample Purification Beads (E7775S), in combination with the NEBNext Poly(A) mRNA Magnetic Isolation Module (New England BioLabs E7490L). Libraries were indexed with NEBNext Multiplex Oligos for Illumina (Dual Index Primers Set 1) (New England BioLabs E7600S). Completed libraries underwent quality assessment on the Agilent Fragment Analyzer run by JH SCTC, and were quantified with the NEBNext Library Quant Kit for Illumina (New England BioLabs E7630L). Libraries were pooled to a final concentration of 4 nM and submitted to the Johns Hopkins Genomics Research Core Facility for high-throughput sequencing on the Illumina NovaSeq 6000 platform, using SP flow cells to generate 100-bp paired-end reads.

Analysis of RNA-sequencing data

The GRCm38 transcriptome was indexed using Salmon 1.9.0 [81], using the GRCm38 genome as the decoy sequence (Mus_musculus.GRCm38.cdna.all.fa.gz and Mus_musculus.GRCm38.dna.primary_assembly.fa.gz, from http://nov2020.archive.ensembl.org/Mus_musculus/Info/Index, release 102). FASTQ files of demultiplexed paired-end reads were mapped using Salmon 1.9.0, with selective alignment and GC bias correction. Transcript quantification files were loaded into R 4.1.2, running Bioconductor 3.14, with tximeta 1.12.4 [82], and transcript-level counts were summarized by gene. Gene-level counts were aggregated for any technical replicates and filtered for non- or low-expressed genes (median count over samples < 10). Surrogate variable (SV) analysis was performed using sva 3.42.0 [83] for the dual experiment between Kmt2d^-/- and Kdm6a^-/- cell lines, identifying up to 2 SVs in certain contrasts. Differential expression analysis was performed using DESeq2 1.34.0 [84] with a false-discovery rate threshold < 0.1. biomaRt 2.50.3 was used to annotate genes according to Ensembl Mus musculus version 102 [85]. Differentially expressed genes are listed in S1 and S5–S8 Appendices. For the comparison to prior Kabuki syndrome 1 data, raw reads were downloaded from the Gene Expression Omnibus (GEO), accession number GSE129365, and processed identically. The Mouse Genome Informatics (MGI) database was accessed on July 11, 2023, to download a list of genes falling under the Gene Ontology (GO) category of ‘cartilage development’. Annotations for negative regulators of this process were removed, leaving 200 genes (complete list in S2 Appendix). MGI was again accessed on March 12, 2024, to download lists of genes within the GO categories of ‘positive regulation of cell size’ and ‘negative regulation of cell size’ (S4 Appendix). A two-sample Wilcoxon test statistic was calculated for the p-values of genes included in these lists and compared against the distribution of test statistics obtained from performing this test on random groups of genes for 10,000 iterations.

Gene Ontology analysis

PANTHER overrepresentation test (release 2024-02-26) was performed with the GO Ontology molecular function database (release 2024-01-17) using the list of differentially expressed genes in Kdm6a^-/- cells compared to Kdm6a^+/+ at Day 14. The reference gene list comprised all expressed genes in the dataset (median count over samples > 10).

Statistics

With exception of the RNA-seq data analysis (which is described in Methods, ‘Analysis of RNA-sequencing data’), two-tailed, unpaired Student’s t-tests were used with a cutoff of p < 0.05.