Mesenchyme-specific loss of Dot1L histone methyltransferase leads to skeletal dysplasia phenotype in mice

Abstract

Chromatin modifying enzymes play essential roles in skeletal development and bone maintenance, and deregulation of epigenetic mechanisms can lead to skeletal growth and malformation disorders. Here, we report a novel skeletal dysplasia phenotype in mice with conditional loss of Disruptor of telomeric silencing 1-like (Dot1L) histone methyltransferase in limb mesenchymal progenitors and downstream descendants. Phenotypic characterizations of mice with Dot1L inactivation by Prrx1-Cre (Dot1L-cKO^Prrx1) revealed limb shortening, abnormal bone morphologies, and forelimb dislocations. Our in vivo and in vitro data support a crucial role for Dot1L in regulating growth plate chondrocyte proliferation and differentiation, extracellular matrix production, and secondary ossification center formation. Micro-computed tomography analysis of femurs revealed that partial loss of Dot1L expression is sufficient to impair trabecular bone formation and microarchitecture in young mice. Moreover, RNAseq analysis of Dot1L deficient chondrocytes implicated Dot1L in the regulation of key genes and pathways necessary to promote cell cycle regulation and skeletal growth. Collectively, our data show that early expression of Dot1L in limb mesenchyme provides essential regulatory control of endochondral bone morphology, growth, and stability.

1. Introduction

The majority of the bones develop from a cartilage intermediate through endochondral ossification, a process involving the condensation of the embryonic limb mesenchyme, initiation of chondrogenesis, extracellular matrix production, and maturation of chondrocytes to hypertrophy (1–4). Invasion of blood vessel endothelial cells, hematopoietic cells, osteoclasts and osteoblast progenitors into the cartilage anlagen leads to the resorption of hypertrophic cartilage, formation of marrow cavities, and trabecular bone formation at primary ossification centers (5). During fetal and childhood development, formation of secondary ossification centers (SOC) at the ends of the long bones establishes the structure of the growth plate (GP), and provides osseous support at the epiphyses. A continuous supply of proliferative GP chondrocytes and their controlled hypertrophic differentiation drives postnatal long bone elongation (6–8). This process requires the coordinated activities of stage-specific transcription factors, signaling molecules, as well as nutritional, endocrine, cytokine, paracrine and extracellular factors (9). Misregulation of cartilage and bone formation and growth can lead to various skeletal dysplasias, which manifest in short stature, bone and joint malformations, and painful complications that reduce quality of life (1, 10–12).

The molecular regulation of endochondral ossification and genetic basis of skeletal dysplasia have been intensively studied (13, 14), yet our understanding of the epigenetic modifiers in normal skeletal development and growth is still emerging (14–19). Chromatin conformation can be influenced by the addition or removal of post-translational modification (ie. methylation, acetylation) at lysine residues within the histone tails and globular histone domain by chromatin modifying enzymes. Conformational changes in chromatin alter accessibility of transcription factors and RNA polymerase complexes, and influence the activation or repression of gene expression (14, 20, 21). Specific lysine acetylation and methylation also serve as docking sites for “reader proteins” in complexes involved in transcriptional initiation, elongation and silencing (22).

The evolutionarily conserved Disrupter of telomeric silencing 1-like gene (Dot1L) encodes the only known methyltransferase that catalyzes mono-, di- and tri- methylation (me1, me2, me3) on lysine 79 within the globular domain of histone 3 (H3K79) (23, 24). Dot1L enzymatic activity is associated with euchromatin, as genome-wide chromatin immunoprecipitation sequencing (ChIP) in various cell types show H3K79me2/3 enrichment at enhancers, transcriptional start sites, and within the gene bodies of actively transcribed genes. The highest levels of H3K79me2/3 are typically found just downstream of the transcription start site of actively transcripted genes (25–28). Earlier studies confirmed that Dot1L expression is indispensable for mammalian development since germ-line deletion resulted in mid-gestational lethality in mice (29, 30). Despite its critical role in embryonic development, few studies have explored the role of Dot1L in skeletal development.

In humans, there is significant genetic link between Dot1L variants and the risk of hip and knee joint osteoarthritis (OA) (31–34). In human and murine knee articular cartilage, reduced levels of Dot1L and H3K79me2 are associated with aging and OA progression (35). Important insights into Dot1L function in joint homeostasis have been gained through genetic and pharmacologic inhibition of Dot1L in mouse and human articular chondrocytes, respectively (35, 36). Collectively, these studies concluded that Dot1L function is positively associated with articular cartilage health (35, 36), and preserving its activity upon joint injury or in elderly patients may be an effective OA therapy.

Conditional loss of Dot1L in the chondrogenic lineage during embryogenesis impaired endochondral bone growth in mice, and impeded survival (35, 37). Thus, a thorough understanding of Dot1L-dependent control of skeletal growth and development may be gained through further analysis of the temporal and spatial requirement of Dot1L in endochondral bone formation. In the present study, we crossed mice carrying the conditional floxed allele of Dot1L with mice expressing Prrx1 (paired-related homeobox transcription factor-1) driven Cre recombinase to generate a novel mouse line with earlier developmental loss of Dot1L in uncommitted limb mesenchyme. Through phenotypic, histologic, and molecular analyses, we reveal a unique skeletal dysplasia phenotype in mice with loss of Dot1L in limb mesenchyme, highlighting the importance of Dot1L in the developing long bones.

2. Materials and methods

2.1. Ethics statement.

All animal procedures were approved and conducted in accordance to NIH guidelines, and the Institutional Animal Care and Use Committee (IACUC) of the University of Connecticut Health Center.

2.2. Animals and genotyping.

Dot1L^fl/fl mice carry loxP sites flanking exon 5, which encodes most of the H3K79 methyltransferase domain (Fig. S1) (38). Female Dot1L^fl/fl mice were mated with transgenic male mice expressing Cre recombinase under the control of the Prrx1 enhancer / promoter which marks limb progenitors (Jackson Laboratories, #005584) (39). Resulting Dot1L^fl/+:Prrx1Cre male offspring were then mated with Dot1L^fl/fl females to produce conditional knockout mice (Dot1L^fl/fl:Prrx1Cre, designated as Dot1L-cKO^Prrx1), conditional heterozygous mice (Dot1L^fl/+:Prrx1Cre, designated as Dot1L-cHet^Prrx1) and wild type (Wt) littermate controls which did not express Cre (Dot1L^fl/fl, Dot1L^fl/+) (Fig. S1). To circumvent risk of germ-line deletion of Dot1L, our breeding strategy used only males as the carrier of Cre recombinase. In litters with >6 pups, fostering methods were used to prevent competition for nursing (40). Two or more mothers with litters born within 24 hours of each other were selected for fostering. Dot1L-cKO^Prrx1 pups could be identified within the first 2 days after birth (forelimb shortening, poor locomotion). We collected only Dot1L-cKO^Prrx1 mice and left them in a nest with their mother. Wt pups of comparable number were placed in cages with other mothers to foster. All mice were in the C57BL/6 background and housed in University of Connecticut Health Center animal facility with a 12 hour light-dark cycle and maintained on standard chow. Genotyping was performed using genomic DNA isolated from ear notches. Identification of Cre transgene was determined by Polymerase chain reaction (PCR) amplification using primers Forward transgene 5’-GCGGTCTGGCAGTAAAAACTATC-3’ and Reverse transgene 5’-GTGAAACAGCATTGCTGTCACTT-3’ (100-bp product). Amplification of 324-bp internal positive control DNA fragment used forward control primer 5’-CTAGGCCACAGAATTGAAAGATCT-3’ and reverse control primer 5’-GTAGGTGGAAATTCTAGCATCATCC-3’, using cycling conditions recommended by Jackson Laboratories for detection of Cre. Primer sequences, and cycling conditions used for Dot1L genotyping were described in (38). A 200-bp Dot1L PCR product was detected in Wt conditions, and 300-bp Dot1L PCR product indicated presence of the floxed Dot1L allele. When separated on a 2% agarose gel, Cre-mediated excision of Dot1L exon 5 was confirmed by detection of a ~150 bp fragment (Fig. S1).

2.3. X-ray imaging.

X-ray imaging of live mice, or of dissected forelimbs and hindlimbs from eviscerated mice, was performed at an intensity of 26kV for 3 seconds on a Faxitron X-ray system (Faxitron X-Ray Corp, Wheeling IL).

2.4. Whole mount skeletal staining.

Newborn mice were eviscerated after euthanasia, fixed overnight at room temperature in 95% ethanol, and then subjected to whole mount skeletal staining using Alcian blue and Alizarin Red as previously described (41). Stained skeletons were stored in 100% glycerol, and imaged using Olympus IX71 microscope with CellSens Software (Olympus, Tokyo, JP).

2.5. Histologic analyses, growth plate measurements and immunohistochemistry.

Hindlimb and forelimb bones were fixed in 10% neutral buffered formalin, decalcified in 14% EDTA (pH 7.2), and embedded in paraffin (42). Longitudinal tissue sections (7-8um) were collected through the proximal tibia and distal femur, and stained with hematoxylin and eosin, Alcian blue and Fast Red, or Safranin O and Fast Green as previously described (42, 43). Immunohistochemistry was performed as described in (42), with epitope retrieval by sodium citrate buffer (pH 6) at 90°C for 5 minutes. Sections were incubated with anti-Sox9 antibody (Abcam #ab185230), anti-PCNA antibody (Abcam #ab29), and anti-p57 antibody (Santa Cruz, #ab56341) overnight at 4°C. Negative controls were incubated with normal rabbit IgG. After overnight incubation, sections were washed in PBS, then incubated in horseradish peroxidase (HRP)-linked anti-rabbit or anti-mouse secondary IgG antibodies (Vector Laboratories). Sections were imaged using image analysis software (Olympus IX71 microscope with CellSens Software, Olympus, Tokyo, JP).

Histological measurements of zonal cartilage thicknesses (resting, proliferating, hypertrophic zones) were determined from H&E stained sections encompassing the central (500um) portion of the proximal tibia growth plate in 21 day-old mice. Growth plate measurements were calculated from 6 points within the central portion of each section using at least 2 sections per mice and 4 mice per genotype. Lengths were determined using the CellSens Software package (Olympus, Tokyo, JP). Data presented as mean ± s.e.m.

2.6. Primary cultures.

Limb bud micromass cultures were established using mesenchymal progenitors isolated from the hindlimb buds of wild type C57BL/6 mouse embryos at 11.5 d.p.c. as described in (29). Cells were passed through a 40-micron cell strainer to obtain a single cell suspension, centrifuged (300xg, 5 minutes), and then resuspended in growth media (Dulbecco’s modified Eagle’s medium/F12 media, 10% FBS, penicillin/streptomycin) at 2 × 10⁷ cells/mL. Cells were spotted in 10uL drops per well of 12-well plates, incubated to allow cell attachment, then fed growth media. Media was refreshed daily, with Dot1L inhibitor supplementation from day 0. EPZ-5676 (Selleck Chemicals) was used at 10uM. All experiments were performed using a minimum of triplicate biological replicates and repeated at least twice using separately derived cultures.

Chondrocytes were isolated from epiphyseal cartilage harvested from limbs of 5-6 day old mice as previously described (44). To maximize chondrocyte yield, the epiphyseal portions of the hindlimb and forelimb joints including distal femurs, proximal tibias, and distal humerus dissected from surrounding soft tissue and mineralized bone were pooled. Cell suspensions were filtered through 40-micron cell strainer to discard undigested tissue, then pelleted (300xg, 5 minutes) prior to resuspension in growth media (DMEM high glucose medium, 10% FBS, 100 units/mL penicillin, 100 units/mL streptomycin). Cells were seeded at 2x10⁵ cells/cm² and cultured in growth media in humidified, 5% CO₂ incubator at 37°C. Cultures were maintained for 24 hours to allow recovery and removal of dead cells prior to RNA harvest (passage 0) for downstream RNAseq and qPCR analysis.

2.7. Real time-PCR and western blotting.

Total RNA was extracted from limb bud micromasses or primary chondrocytes (collected one day after plating, passage 0) using TRIzol reagent (Invitrogen). Samples of one microgram were treated with DNaseI (BioRad), then reverse transcribed using iScript Reverse Transcriptase (Qiagen). 20ng of cDNA was used for quantitative PCR analyses. Quantitative PCR was performed using SYBR Green I Mastermix (Roche) and run on a BioRad CFX96 Touch Real-Time PCR Detection System. Relative expression of mRNA was calculated relative to β-actin by comparative cycle threshold analyses. Gene expression values are expressed as 2^ΔΔCt, with ^ΔΔCt defined as the difference in crossing threshold (Ct) values between wild type and mutant samples, using β-actin as an internal standard. Murine oligonucleotide primer sequences were as listed in Table 1.

Table 1.

qPCR Primers

Gene	Forward primer	Reverse primer
Aggrecan	5’-gtggagagtcttctggcattac-3’	5’-cactgagttccacagatcctaac-3’
Alp	5’-tgaccttctctcctccatcc-3’	5’-cttcctgggagtctcatcct-3’
B-actin	5’-agatgtggatcagcaagcag-3’	5’-gcgcaagttaggttttgtca-3
Col2a1	5’-actggtaagtggggcaagac-3’	5’-ccacaccaaattcctgttca-3’
Dot1L	5’-cactacacagcccatgaagc-3’	5’-ggtcgacaaaacaggtcatcc-3’
Igf2	5’-gagcttgttgacacgcttc-3’	5’-acgtttggcctctctgaac-3’
Igfbp2	5’-cagacgctacgctgctatcc-3’	5’-ctccctcagagtggtcgtca-3’
Runx2	5’-aagtgcggtgcaaactttct-3’	5’-tctcggtggctggtagtga-3’
Sox5	5’-acctcagaaggcggaagaag-3’	5’-cttcagggtgtccaccacat-3’
Sox9	5’-aggaagctggcagaccagta-3’	5’-cgttcttcaccgacttcctc-3’

For western blotting, cells were scraped in RIPA lysis buffer supplemented with protease inhibitor cocktail (Sigma-Aldrich). 30 ug of total protein was loaded per well of 12% acrylamide SDS-PAGE gel. Western blotting was performed using anti-histone H3 dimethyl K79 (H3K79me2, Abcam #ab3594, 1:750 dilution), anti-β-actin antibody (Cell Signaling, #3700, 1:2000 dilution). Detection and analyses was performed using Bio-Rad ChemiDoc XRS+ imager and Image Lab software 4.1.

2.8. Microcomputed Tomography (uCT).

Femoral bone lengths and trabecular microarchitecture were determined using a Microcomputed Tomographic Instrument (uCT 40; Scanco Medical AG, Bassersdorf, Switzerland). Right femora were scanned in water at high resolution, energy level of 55 kVp, intensity of 145 uA, with 500 ms integration time. Images were reconstructed and calibrated at isotrophic voxel size of 8 mm³. Trabecular regions were assessed for total volume, bone volume, bone volume fraction (BV/TV), trabecular thickness, trabecular number, trabecular spacing, connectivity density, and structural model index. Analysis of femoral cortical bone excluded the marrow cavity. All analyses were performed in 10 week-old male mice (n=4-6 mice/group).

2.9. RNA sequencing (RNAseq) and analysis.

Chondrocytes were harvested from the hind- and forelimbs of 5 day-old Wt and Dot1L-cKO^Prrx1 mice (n=3 per group, passage 0). Total RNA was isolated using TRIzol reagent, and RNA integrity assessed using Agilent 2100 Bioanalyzer. RNA sequencing libraries were generated using SMARTer v4 chemistry (Clontech). After 1^st strand DNA synthesis, cDNA was purified using Agencourt AMPure beads (Beckman Coulter), and PCR amplified. cDNA yield and fragment size was measured on a 2100 Bioanalyzer (Agilent). cDNA was sheared using a Covaris LE220 system, and fragmented cDNA prepared using the NEBNext DNA library prep kit for Illumina sequencing. Barcoded libraries were multiplexed and sequenced using NextSeq500 2 × 75 base sequencing. Each sample was sequenced to an average depth of 30 million reads. Raw reads were trimmed with Sickle (45), with a quality threshold of 30 and length threshold of 45, and trimmed reads mapped to Mus Musculus genome (GRCm38.97) with HISAT2 (version 2.1.0) (46). SAM files then converted into BAM format using samtools (version 1.9) (47), and the PCR duplicates were removed using PICARD software (http//broadinstitute.github.io/picard/). Counts were generated against the features with htseq-count (48). Differential expression of genes between conditions was evaluated using DESeq2 (49). Genes showing <10 counts across the compared samples were dropped from the analysis. P-values were adjusted for comparison using the false discovery rate (FDR). The criteria for DEG was >0.6 log2 fold change / < −0.6 log2 fold change (absolute fold change 1.5) and adjusted P-values <0.05.

Gene ontology (GO) Enrichment Analysis (http://geneontology.org/page/go-enrichment-analysis) and DAVID annotation (https://david.ncifcrf.gov/) was used for functional annotation, including molecular function, biological processes, and cellular compartments. GO terms with FDR (q < 0.05) were considered significantly enriched within the gene set. Gene set enrichment analysis (GSEA) was performed using multiple annotated gene set databases available from the Molecular Signatures Database v7.1 (MsigDB) (gsea-msigdb.org). For the present study, we used GO gene sets (biological processes, molecular function, cellular compartments), hallmark gene sets and curated gene sets (ie. Reactome) and list of ranked genes based on a score calculated as −log10 off P value multiplied by sign of fold-change. Enrichment of gene sets was performed using the default GSEA Pre-ranked algorithm.

2.10. Statistics.

Statistical analyses were performed using Excel 2007 (Microsoft, Redmond, WA) and GraphPad Prism 8.0 for Mac OS X (GraphPad Software, San Diego, CA). Student’s t-test was used to test differences between two samples. Analysis of variance (ANOVA) with Bonferroni’s post-test was used for multiple comparisons. Significance threshold is P ≤ 0.05. The values are mean ± standard error of mean (s.e.m.)

3. Results

3.1. In vitro pharmacologic inhibition of Dot1L accelerated chondrocyte differentiation in limb bud mesenchyme-derived progenitors.

Expression of Dot1L in murine limb buds was suggestive of a role in early skeletal development (29), yet the functional contribution of Dot1L to chondrogenesis in limb progenitors has not been studied. To address this, uncommitted mesenchymal progenitors isolated from E11.5 hindlimb buds were cultured under chondrogenic micromass conditions. Compared to the undifferentiated (E11.5) progenitors, which we set as baseline, Dot1L transcription increased significantly under chondrgenic micomass conditions (Fig 1A). Increased H3K79me2 protein levels following 10 days of micromass differentiation corroborated our gene expression data (Fig 1B). To examine functional significance, micromasses were treated with or without 10uM EPZ-5676 (EPZ), a clinically relevant Dot1L inhibitor (50, 51). Dot1L inhibition was confirmed by depletion of H3K79me2 proteins in inhibitor-treated micromasses (Fig. 1C). During initial stages of in vitro chondrogenesis, compact condensations were more apparent in EPZ-treated micromasses as compared to vehicle (DMSO)-treated controls (days 2-3). We detected more robust Alcian blue and Col2 staining within condensations in EPZ-treated cultures as compared to vehicle treatment (Fig. 1D,F). During the initial 4 days of differentiation, Dot1L inhibition resulted in significantly higher levels of Col2a1 (5.62-fold, p < 0.00001), and Aggrecan (10.26-fold, p< 0.00001) expression (Fig. 1E). By days 7 and 10 of chondrogenesis, sustained EPZ treatment significantly attenuated the expressions of Col2a1 (0.27-fold, p< 0.00001; and 0.34-fold, p < 0.05) and Aggrecan (0.38-fold, p < 0.001; and 0.53-fold, p < 0.05) relative to controls. Interestingly, decreased expression of markers of immature cartilage matrix in Dot1L inhibited micromasses coincided with increased expressions of genes involved in chondrocyte maturation, including type X collagen (ColX) and alkaline phosphatase (Alp) (Fig. 1G). On day 10 of differentiation, we also observed enhanced Alp activity staining in Dot1L-inhibited micromasses as compared to controls (Fig. 1G). Together, our data indicate that inhibition of Dot1L accommodates onset of chondrogenesis in limb progenitors responsible for endochondral bone formation, and accelerates developmental progression towards chondrocyte hypertrophy.

Figure 1. Pharmacologic inhibition of Dot1L activity in limb bud progenitors accelerated chondrocyte differentiation in vitro.

(A) Increased Dot1L mRNA expression over the time course of chondrogenic differentiation in mesenchymal progenitors from forelimb (FL) versus hindlimb (HL) buds in micromass. Expression relative to day 0 (undifferentiated) progenitors. Significance at * P<0.05, **** P<0.0001. One-way ANOVA using Tukey’s multiple comparison test. Data are mean ± s.d. (B) Western blot analysis of H3K79me2 levels in protein lysates harvested from chondrogenic FL versus HL bud micromass cultures. β-actin served as normalization control. Values represent normalized expression levels determined by densitometry. (C) Western blot analysis of H3K79me2 levels in protein lysates harvested from day 3 micromasses treated with either vehicle (DMSO) or 10 uM EPZ-5676. EPZ-5676 lowered H3K79me2 (17KDa) levels. β-actin served as loading control. (D) Brightfield images (upper panel) of vehicle- versus EPZ-5676-treated micromasses on day 3 of differentiation. Arrows show early condensations in inhibitor-treated cultures. Lower panel shows Alcian blue staining in vehicle- and inhibitor-treated cultures. (E) Gene expression analysis over time course of differentiation showed elevated induction of chondrogenic markers (ie. Col2a1, Aggrecan) in day 4 Dot1L inhibitor treated cultures. n = 4 samples per group * significance at P<0.05. Two tailed Student’s t test. Data are mean ± s.d. (F) Type 2 collagen (Col2) immunocytochemical staining in day 4 micromasses. Scale bar, 200 um (G) q-PCR analyses showed significantly higher expression levels of chondrocyte maturation genes type X collagen (ColX) and alkaline phosphatase (ALP) in micromasses treated with EPZ-5676 for 10 days. N = 3 samples per group ** indicates significance at P < 0.01. Two tailed Student’s t test. Data are mean ± s.d. Side panel shows ALP activity staining was enhanced in day 10 EPZ-5676-treated micromasses as compared to vehicle treatment. Scale bar, 100 um

3.2. Inactivation of Dot1L in Prrx1-expressing embryonic mesenchyme caused limb shortening, secondary patterning defects, and joint defects.

To study in vivo significance of Dot1L, we generated a new mouse line with conditional loss of Dot1L in embryonic limb mesenchyme using Prrx1-(paired-related homeobox transcription factor-1) Cre recombinase driver mice (38, 39). Cre recombinase activity is controlled by a 2.4kb Prrx1 promoter, active in condensed limb mesenchyme from which multiple skeletal lineages are derived (39). Following crosses of male Prrx1Cre mice with Dot1L^fl/fl females, we confirmed efficient Cre-mediated recombination at exon 5 in Dot1L-cKO^Prrx1 cartilage by PCR analyses (Fig. S1.F). Real-time PCR analyses confirmed loss of Dot1L transcript expression in chondrocytes from day 5 Dot1L-cKO^Prrx1 hindlimbs (>98% reduction vs Wt) (Fig. 2A), and Western blot analyses showed global attenuation of H3K79 di-methylated (H3K79me2) histones (Fig. 2B). These data validated functional loss of Dot1L in Prrx1 downstream descendants, and lack of compensatory H3K79 methyltransferase activity.

Figure 2. Validation of Prrx1-Cre mediated loss of Dot1L function in mice.

(A) qPCR analysis confirmed loss of Dot1L mRNA expression in chondrocytes harvested from day 5 Dot1L-cKO^Prrx1 mice (n = 7) as compared to Wt (n = 8) and Dot1L-cHet^Prrx1 (n= 3). Expression values are normalized to β-actin levels. Dots represent data points from individual mice. Significance at ** P<0.01; ***P<0.0007; **** P0.0001. One-way ANOVA. Data are mean ± SD. (B) Western blot analyses confirmed lack of H3K79me2 signal in total protein lysates of chondrocytes harvested from cKO mice as compared to Wt chondrocytes. Dot1L-cHet^Prrx1 showed reduced Dot1L protein levels relative to Wt. β-actin served as loading control.

Male and female homo- and heterozygous Dot1L knockout mice were viable, produced at the expected Mendelian frequency, and appeared grossly similar to Wt and Dot1L-cHet^Prrx1 littermates (Fig. 3A), At birth, there was significant shortening of the long bones (ie. tibia) in Dot1L-cKO^Prrx1 mice as compared to WT, and sustained at 7 days postnatal (Fig. 3B). By age of weaning, mice of either sex from each genotype showed comparable body lengths (Fig. 3C,D). Substantial shortening Dot1L-cKO^Prrx1 hindlimbs at 3 weeks of age was further confirmed by radiographic imaging (Fig. 3E). In additional to shortening, imaging revealed thickening of the fibula (Fig. 3E, white arrowhead), and a diminished region of secondary ossification in Dot1L-cKO^Prrx1 tibias (Fig. 3E, yellow arrow). Limb shortening in Dot1L-cKO^Prrx1 mice was sustained throughout postnatal development (Fig. 3F,G). By 10 weeks, mean femoral length in Dot1L-cKO^Prrx1 mice was 11.64 ± 0.016 mm, compared to 14.93 ± 0.197 mm for Wt, and 14.20 ± 0.985 mm for Dot1L-cHet^Prrx1 mice (Fig. 3G, n= 3-8 male mice per genotype, one-way ANOVA). In additional to shortening, femora from adult Dot1L-cKO^Prrx1 mice were abnormally broad and displayed femoral neck deformity, abnormal morphology of the third trochanter, and osteophytic changes (Fig. 3F).

Figure 3. Dot1L-cKO^Prrx1 mice exhibited significant postnatal hindlimb shortening.

(A) Representative images of newborn Wt, Dot1L-cHet^Prrx1 mice and -cKO^Prrx1 littermates. (B) Tibia lengths in newborn and 7 day-old mice from each genotype. * P < 0.01 versus Wt controls, # P < 0.01 versus cHet. (C) Overall body length was unaffected by loss of Dot1L in Prrx1-expressing mesenchyme. Body lengths (nose to rump length shown in D) were similar across genotypes in 21 day-old mice. Dot1L-cKO^Prrx1 and -cHet^Prrx1 body lengths represented as % of their respective male and female Wt control littermates across multiple litters (E) Representative x-rays from 3 week-old mice showed hindlimb shortening in Dot1L-cKO^Prrx1 mice. Reduced SOC (yellow arrow), and widening the fibula (white arrowhead) was seen in Dot1L-cKO^Prrx1 mice. (F) 3D uCT reconstruction and (G) lengths of femurs from 10 week-old male Wt (n=6), Dot1L-cHet^Prrx1 (n=4) and Dot1L-cKO^Prrx1 (n=5) mice. Dot1L-cKO^Prrx1 femurs were significantly shorter in length, yet broader in comparison to either Wt or Dot1L-cHet^Prrx1. Arrows in (F) indicate osteophytes in Dot1L Dot1L-cKO^Prrx1 femora. Significance at **** P<0.0001. One way ANOVA, using Tukey’s multiple comparison test. Data are mean ± s.d.

Alcian blue and Alizarin staining of whole mount skeletal preparations also revealed structural forelimb anomalies in Dot1L-cKO^Prrx1 neonates (Fig. 4). The radius appeared bowed due to dislocation at the radio-ulnar joint in Dot1L-cKO^Prrx1 (Figs. 4J, 3O). This radial head phenotype was 100% penetrant in male and female mice from multiple breeding pairs, and was not observed in either Wt or Dot1L-cHet^Prrx1 littermates. Histological sections stained with Alcian blue confirmed disruption at the radio-ulnar joint in Dot1L-cKO^Prrx1 mice, revealing a more convex-shaped radial head and aspherical capitellum humeri (Fig. 4P). X-ray imaging of forelimbs demonstrated elbow dislocations and shortening of all forelimb elements in day 7 Dot1L-cKO^Prrx1 mice (Fig. 4M) relative to Wt (Fig. 4N) and Dot1L-cHet^Prrx1 littermates (Fig. 4O).

Figure 4. Novel limb and joint malformations in Dot1L cKO mice.

Whole mount Alcian blue (cartilage) and Alizarin red (mineral) staining was performed in newborn Wt (A), Dot1L-cHet^Prrx1 (E) and Dot1L-cKO^Prrx1 (I) littermates. Scale bar, 20mm. (B, F, and J) Magnified images of radioulnar joint revealing dislocation in newborn Dot1L-cKO^Prrx1 mice (asterisk * in J). (C,G, and K) Magnification of glenohumeral joint showed misshapen humeral head in Dot1L cKO^Prrx1 mice (K). (D, H, and L) Higher magnification image of deltoid tuberosity (dt, red arrows) of humerus showed under-development in Dot1L-cKO^Prrx1 mice (L). Representative radiographs from 7 day-old Wt (M), Dot1L-cHet^Prrx1 (N), and Dot1L-cKO^Prrx1 (O) mice. (P-V) Alcian blue staining of histological sections from forelimbs of 7 day-old Wt and Dot1L-cKO^Prrx1 mice. Staining of the radioulnar joint revealed irregularly shaped capitellum and proximal radius in Dot1L-cKO^Prrx1 mice (S) as compared to Wt mice (P). (Q, T) Alcian blue staining of proximal humerus revealed misshapen proximal humerus shown in Dot1L-cKO^Prrx1 mice (T). (R,U) Deltoid tuberosity (dt, demarcated by dotted line) was under-developed in Dot1L cKO^Prrx1 mice (U). Scale bar 200um. (V) Magnified view of boxed image in (U) showed retention of hypertrophic cartilage at presumptive dt in Dot1L-cKO^Prrx1 mice. Scale bar,100um. Abbreviations; ca, capitellum; Hu, humerus; ra, radius; ul, ulna; dt, deltoid tuberosity.

In addition to elbow dysplasia, the deltoid tuberosity of the humerus appeared under-developed in embryonic (data not shown) and in perinatal Dot1L-cKO^Prrx1 mice (Fig. 4L) when compared to Wt (Fig. 4D) and Dot1L-cHet^Prrx1 mice (Fig. 4H). This bony protrusion on the lateral edge of the midhumeral shaft forms through endochondral ossification, and provides a stable anchoring point for muscles, inserted to the skeleton by tendons (52). Histological sections through this region in 7 day-old Wt mice revealed a mineralized deltoid tuberosity (Fig. 4R), whereas residual cartilage remained within the presumptive tuberosity in Dot1L-cKO^Prrx1 mice (Fig. 4U,V).

3.3. Abnormal growth plate (GP) morphology in Dot1L-cKO^Prrx1 mice.

We examined postnatal GP organization to gain insights into developmental defects attributed to loss of Dot1L in limb mesenchyme. H&E staining of proximal tibia sections from 3 week-old Wt mice revealed a hallmark structure defined by a narrow band of resting zone (RZ) chondrocytes; proliferating zone (PZ) chondrocytes within columns that align parallel to the direction of bone growth; and enlarged chondrocytes within the zone of hypertrophy (HZ) (Fig. 5A). By contrast, the Dot1L-cKO^Prrx1 GP was composed of irregularly shaped PZ chondrocytes that failed to properly align in chondrons (Fig. 5B). The PZ was significantly reduced in Dot1L-cKO^Prrx1 mice, representing only 29.67% ± 2.05% of total GP length as compared to 40.78% ± 1.12% in Wt controls (p = 0.003) (Fig. 5C, Table 2). Zone of chondrocyte hypertrophy and developing trabecular bone were clearly diminished in Dot1L cKO^Prrx1 long bones as compared to Wt controls (Fig. 5A). Chondrocytes within the terminal HZ comprised 45.79% ±1.25% of the total GP length in Wt mice, whereas this zone represented only 30.73% ± 2.59% of total GP length in Dot1L-cKO^Prrx1 (p = 0.002). Interestingly, there was significant expansion of the RZ within the GP of 3 week-old Dot1L-cKO^Prrx1 mice (40.32% ± 4.28%) as compared to Wt mice (9.0% ± 0.83% in Wt mice, p = 0.0004)

Figure 5. Impaired development of growth plate structure in Dot1L-cKO^Prrx1 mice.

(A,B) H&E staining of sections from tibial GP of 21 day-old Wt and Dot1L-cKO^Prrx1 mice. Scale bar, 20um. Higher magnification image in (B) show misaligned chondrocytes with irregular morphology in Dot1L-cKO^Prrx1 mice. Scale bar, 40um. (C) Quantitative morphometric analyses of the proximal tibial GP in Wt and Dot1L-cKO^Prrx1 mice. Zone lengths were calculated as % of overall GP length. n = 4 male mice per genotype. GP, growth plate; RZ, resting zone; PZ, proliferating zone; HZ, hypertrophic zone, Tb, trabecular bone.(D,E) IHC staining for PCNA in tibial GP from Wt and Dot1L-cKO^Prrx1 mice. Scale bar, 50um. Reduced percentage of PCNA positive cells found in Dot1L-cKO^Prrx1 PZ chondrocytes as compared to Wt controls. % of PCNA positive cells was quantified across 3 Wt and 3 Dot1L-cKO^Prrx1 from at least 3 sections per mouse. (F) Alcian blue staining shows diminished proteoglycan content within the proximal tibia GP in Dot1L-cKO^Prrx1 mice as compared to Wt control at 5 weeks of development. Scale bar, 100um. (G). Q-PCR analysis showed significantly lower expression levels of cartilage specific genes SRY-box transcription factor 9 (Sox9), collagen type II (Col2), and Aggrecan (Agc) in Dot1L-cKOPrrx1 primary chondrocytes (n=6) versus wild type chondrocytes (n=5), Significance determined at * P< 0.05; ** P < 0.005; *** P < 0.001.

Table 2. Growth plate (GP) measurements in postnatal growth plates.

Zone-specific lengths were calculated from H&E stained histological sections encompassing the proximal tibias of 21 day-old male Wt and Dot1L-cKO^Prrx1 mice. Lengths (um) were calculated from 6 points within the central portion of each histological section (2 sections per mouse, 4 mice per genotype). Total GP length, and each GP zone was determined using the CellSens Software package (Olympus, Tokyo, JP). RZ, resting zone; PZ, proliferating zone; HZ, hypertrophic zone. Values represent mean ± s.e.m.

	Total length	RZ length (um)	PZ length (um)	HZ length (um)
Wild type	335.51 ± 22.68	28.96 ± 1.59	136.62 ± 9.41	154.22 ± 14.55
Dot1L cKO	319.05 ± 12.14	129.42 ± 13.13	94.19 ± 6.09	98.02 ± 9.96
P-value	0.546	0.0003	0.009	0.019

By immunohistochemistry, we compared the distribution of PZ chondrocytes expressing Proliferating Cell Nuclear Antigen (PCNA) in Wt and Dot1L-cKO^Prrx1 growth plates (Fig. 5D). PCNA-positive chondrocytes represented a significantly lower percentage of total cells in the Dot1L-cKO^Prrx1 GP compared to Wt conditions (38.1% for cKO^Prrx1 vs 81.8% for Wt, p = 0.001) (Fig. 5E). Consistent with these observations, Sox9-expressing cells were less abundant in PZ in Dot1L-cKO^Prrx1 GP as compared to Wt mice (Fig. S3A). Antibody staining for the chondrocyte prehypertrophy / hypertrophy marker p57 showed a more restricted expression domain in Dot1L-cKO^Prrx1 GPs (Fig. S3B). Loss of columnar GP chondrocyte organization is a common feature in chondrodysplasia, and associated with impaired production or structure of the extracellular matrix. A dramatic reduction in proteoglycan staining within the Dot1L-cKO^Prrx1 GPs was apparent in 5 week-old mice (Fig. 5F). Moreover, qPCR analysis of limb chondrocytes (postnatal day 5) showed reduced levels of Sox9, type II collagen (Col2) and aggrecan under conditions of Dot1L genetic loss of function (Fig. 5G). Overall, these data indicate that reduced GP chondrocyte proliferation and hypertrophy, and deficits in extracellular matrix production underlie the long bone growth deficits in Dot1L-cKO^Prrx1 mice.

3.4. Formation of secondary ossification centers (SOC) was defective in mice with mesenchyme-specific deletion of Dot1L.

Formation of a bony epiphysis at the SOC physically separates the articular cartilage from the GP, and protects chondrocytes from mechanical demands associated with weight bearing during juvenile growth (4). In humans, SOCs develop within the long bone epiphysis during late embryogenesis, while SOC formation is initiated shortly after birth in mice (53). Mouse lineage tracing studies determined that periarticular progenitors contribute to SOC initiation and ossification of subchondral bone (54). In Alcian blue-stained sections, perichondrial invaginations at the presumptive SOC were observed in proximal tibias and distal femurs from 7 day-old Wt mice (Figs. 6A and 6C, red arrows). These specialized structures (cartilage canals) bring in blood vessels and loose connective tissue into centrally located, non-mineralized epiphyseal cartilage (54) (Fig. 6E (red arrow). By contrast, evidence of cartilage canals in Dot1L-cKO^Prrx1 long bones was lacking at this stage of development (Figs. 6B and 6D). Enlarged, hypertrophic chondrocytes were readily observed within the presumptive SOC in 7 day-old Wt mice (Fig. 6E), whereas the epiphyseal chondrocytes uniformly remained at a less mature differentiation stage in Dot1L-cKO^Prrx1 mice (Fig. 6F). By 17 days of postnatal development, cartilage was replaced by bone and SOC was well developed at the proximal tibia in Wt mice (Fig. 6G, yellow arrow). In contrast, the Dot1L-cKO^Prrx1 tibial epiphyses still contained a mass of cartilage consisting of hypertrophic chondrocytes and lacked osteoid formation and marrow spaces (Fig. 6H, black arrow). At 3 weeks of development, Safranin O/Fast green staining showed retained hypertrophic cartilage (red) and lack of subchondral bone formation (green) within the epiphyses of Dot1L cKO^Prrx1 mice (Fig. 6I, 6K). Histology revealed deformed epiphyses in older Dot1L-cKO^Prrx1 mice (Fig. 6J, 6L). Nine weeks after birth, the typically domed-shaped tibial epiphyses seen in Wt mice (Fig. 6M) appeared flattened in Dot1L-cKO^Prrx1 mice, and the distance between the GP and articular surface was markedly reduced (Fig. 6N). Together, our data show that Dot1L functions as a critical regulator of SOC initiation and maturation within the long bones.

Figure 6. Deletion of Dot1L by Prrx1Cre impaired the formation of secondary ossification centers within the long bones.

Alcian blue and Fast red staining of proximal tibia (A, B) and distal femur (C, D) in 7 day-old Wt and Dot1L-cKO^Prrx1 mice. Red arrows indicate presence of cartilage canals in Wt proximal tibia (A) and distal femur (C, E), respectively. Dot1L-cKO^Prrx1 mice lacked cartilage canals in both proximal tibia (B) and distal femur (D), and showed diminished hypertrophy (D, F) within epiphyseal cartilage compared to Wt controls. Higher magnification images of boxed area in Wt and Dot1L-cKO^Prrx1 distal femurs shown in (E) and (F) respectively. Scale bar, 50 um. H&E staining of proximal tibia from 17 day-old Wt mice (G) and Dot1L-cKO^Prrx1 mice (H). Formation of subchondral bone (yellow arrow) and marrow cavities detected in SOC of Wt tibia (G), whereas Dot1L-cKO^Prrx1 epiphyses retained hypertrophic chondrocytes (black arrow), and lacked subchondral bone (H). Scale bar, 200um. Safranin O (red) and fast green (green) staining of SOC within proximal tibia of day 21 Wt (I, K) versus Dot1L-cKO^Prrx1 mice (J, L). Images show well-formed SOC (yellow arrow indicates bone) and clearly defined GP in Wt mice (K). Presence of hypertrophic chondrocytes indicated delayed subchondral bone formation in epiphyses in Dot1L-cKO^Prrx1 mice (L). Alcian blue staining of tibial longitudinal sections from 9 week-old mice demonstrated flattening of epiphysis, and diminished proteoglycan staining in Dot1L-cKO^Prrx1 mice (N) compared to Wt (M). Scale bar, 200um

3.5. Conditional knockout of Dot1L in limb mesenchyme impaired trabecular bone formation and microarchitecture.

The effects of Dot1L loss on endochondral bone formation were further studied by quantitative micro-computed tomography (uCT) in femurs from 10 week-old Wt, Dot1L-cHet^Prrx1 and -cKO^Prrx1 male mice. Representative uCT images indicated deficient metaphyseal trabeculation in femora from Dot1L-cKO^Prrx1 and Dot1L-cHet^Prrx1 mice (Figs. 7A, B). Significantly reduced trabecular bone mass in both Dot1L-cHet^Prrx1 and Dot1L-cKO^Prrx1 mice was confirmed by decreased bone mineral density (bone volume/total volume, BV/TV, %) versus Wt controls (5.7% ± 0.442% in Dot1L-cKO^Prrx1; 4.85% ± 1.55% in cHet^Prrx1; and 17.80% ± 0.796% in Wt). Femora from either Dot1L-cKO^Prrx1 or Dot1L-cHet^Prrx1 mice exhibited reduced metaphyseal trabecular thickness (46.50um ± 0.967um in Wt; 29.50um ± 1.5um in Dot1L-cHet^Prrx1; and 34.25um ± 1.315um in Dot1L-cKO^Prrx1) as well as reduced mean trabecular number compared to Wt mice (5.493 ± 0.08 in Wt; 4.265 ± 0.465 in Dot1L-cHet^Prrx1; and 4.613 ± 0.145 for Dot1L-cKO^Prrx1, units 1/mm). In contrast, trabecular spacing in either Dot1L-cKO^Prrx1 (215.5 um ± 7.263 um) or Dot1L-cHet^Prrx1 (236.5 um ± 28.50 um) femurs was significantly increased relative to Wt controls (172.5 um ± 3.428 um) (Fig. 7B–H). Interestingly, uCT analyses also showed significant reduction in cortical bone thickness in Dot1L-cKO^Prrx1 and Dot1L-cHet^Prrx1 as compared to Wt controls (Fig. 7I, J). Collectively, our data formally established the critical requirement for Dot1L in postnatal endochondral bone formation.

Figure 7. Defective trabecular and cortical bone formation in 10 week-old mice with mesenchyme-specific deletion of Dot1L.

(A) 2D uCT scans of intact femora from 10 week old Wt, Dot1L-cHet^Prrx1 and Dot1L-cKO^Prrx1 male mice. Scale bar, 1mm. (B) 3D uCT scans revealed reduced Tb volume and density in the metaphyses of distal femurs from Dot1L-cHet^Prrx1 and Dot1L-cKO^Prrx1 mice as compared to Wt. (C-H) Quantitative analyses demonstrated significant reduction in bone volume fraction (C), Tb thickness (D), Tb number (E), and connectivity density (F) in both Dot1L-cKO^Prrx1 and Dot1L-cHet^Prrx1 as compared to Wt controls. Tb spacing (G) and SMI (H) were significantly increased in Dot1L-cKO^Prrx1 and Dot1L-cHet^Prrx1 femora versus Wt femora. Dot1L-cKO^Prrx1 mice showed widening of medullary cavity and cortical thinning (I, J). Data are Mean ± s.e.m. n = 5 Wt, n = 4 Dot1L-cHet^Prrx1, n = 5 Dot1L-cKO^Prrx1. Significance at *P<0.05; **P<0.01; ***P<0.0005; ****P<0.0001; ns, not significant. Significance determined by one-way ANOVA, using Tukey’s multiple comparison test. BV, bone volume; TV, total volume; Tb, trabecular; Conn, connectivity; SMI, structural model index

3.6. Differential gene expression profiles in Dot1L deficient chondrocytes underlies the dysplastic limb phenotype.

The GP phenotype in Dot1L-cKO^Prrx1 mice led us to focus our molecular analyses on chondrocytes. Unbiased genome-wide RNA sequencing was used to identify genes that were differentially expressed in chondrocytes isolated from the limbs of early postnatal (day 5, passage 0) Dot1L-cKO^Prrx1 mice as compared to Wt mice (n=3 mice/genotype). 1,929 transcripts (6.7% of total) were found be differentially expressed between Wt and Dot1L-cKO^Prrx1 chondrocytes at an FDR adjusted P-value below 0.05, including 1,109 up-regulated transcripts (57.5% of differentially regulated transcripts) and 820 down-regulated transcripts (42.5% of differentially regulated transcripts). Genes that were most significantly up- or down-regulated in Dot1L-cKO^Prrx1 chondrocytes are shown in Fig. S5. Using a threshold of 1.5 absolute fold change, DEGs were uploaded into the GO Enrichment Analysis tool for annotation, visualization and integrated discovery (DAVID) tool (Table S1). We found DEGs involved in biological processes (BP) including cell adhesion (GO: 0007155; FDR=1.2E-7), multicellular organism development (GO: 0007275, FDR=1.0E-6), muscle contraction (GO: 0006936; FDR=3.5E-4), extracellular matrix organization (GO: 0030198; FDR =1.2E-3), skeletal system development (GO: 0001501; FDR=2.1E-3), ossification (GO: 0001503; FDR=9.1E-3), angiogenesis (GO: 0001525; FDR=1.7E-2), and positive regulation of cell proliferation (GO: 0008284; FDR=4.2E-2). The top molecular functions (MF) predicted from significantly enriched GO terms included calcium binding (GO: 0005509; FDR=6.2E-9), receptor binding (GO: 0005102; FDR=8.5E-5), heparin binding (GO: 0008201; FDR=2.2E-3), extracellular matrix structural constituent (GO: 0005201; FDR=2.3E-3), and extracellular matrix binding (GO: 0050840; FDR=4.1E-3). The cellular compartment enrichment for the DEGs was predominantly in the extracellular matrix (extracellular region GO: 0005576, FDR=2.2E-17; extracellular space GO: 0005615, FDR=6.1E-18; proteinaceous extracellular matrix GO: 0031012, FDR = 1.7E-16; extracellular exosome GO: 0070062, FDR = 3.1E-13; and extracellular matrix GO: 0031012, FDR = 7.4E-7).

Gene set enrichment analysis (GSEA) of the differentially expressed genes (DEG) was also used to gain mechanistic insights into key biological processes, molecular functions and cellular compartments altered by loss of Dot1L function in chondrocytes (Fig 8A–C, Table S2). Regarding biological processes, our analyses revealed that loss of Dot1L specifically decreased expression of transcriptional programs associated with regulation of cell cycle process (ie. Pcna, Cdk1, Cdk2, Cdk4, E2f1, E2f8, Igf2). Expression of E2F-target genes linked to cell cycle regulation, DNA synthesis, and proliferation were also decreased in Dot1L-cKO^Prrx1 chondrocytes. Top molecular function gene sets that were negatively associated with loss of Dot1L function included DNA binding and DNA binding transcription factor binding (Table S2). GSEA predicted the enrichment of genes associated with cell-cell adhesion via plasma membrane adhesion molecules in Dot1L-cKO^Prrx1 chondrocytes, including cadherins (Cdh2, Cdh3, Cdh6, Chd11, Cdh13, Cdh17, Cdh19) and protocadherins (Pcdh7, Pcdh10, Pcdh17, Pcdh18, Pcdh19) whose functions are associated with not only cell adhesion and migration, but also linked to major signaling pathways such as Wnt/b-catenin (55)

Figure 8. Gene set enrichment and Panther Pathway anlaysis.

Gene set enrichment analyses (GSEA) of all genes significantly regulated in Dot1L-cKO^Prrx1 chondrocytes at FDR < 0.05. (A, B) GSEA showing enrichment of genes annotated to “positive regulation of cell cycle processes” and “hallmark E2F targets” in genes down-regulated in Dot1L cKO^Prrx1 chondrocytes. (C) “Cell-cell adhesion via plasma membrane adhesion molecules” (C) was among the most significantly up-regulated genes in Dot1L cKO^Prrx1 chondrocytes relative to Wt chondrocytes. FDR, false discovery rate; NES, normalized enrichment score. (D) Enrichment of pathways with deregulation in Dot1L-cKO^Prrx1 chondrocytes was identified by PANTHER Pathway analysis. Pie charts indicate total number and % of genes within each pathway showing significantly differential expression in Dot1L-cKO^Prrx1 chondrocytes as compared to Wt chondrocytes. Data from n = 3 mice per genotype.

Growing evidence suggest that Dot1L does not act as a transcriptional switch, rather it facilitates transcriptional maintenance of a subset of lineage-specific genes (56). We noted that several transcription factors with known relevance to skeletal development were among the genes most significantly down regulated in Dot1L-cKO^Prrx1 chondrocytes. These included Msx1, Runx2, Shox2, Sox5, Sox11, Tbx4, and Tbx5 (Fig. S5A). Sox5 belongs to the trio of Sry-related HMG box-containing transcription factors (Sox5,6,9), which play essential roles at multiple stages of cartilage and endochondral bone development, from formation of the cartilage primordia to establishment of a multi-layered postnatal GP (57). Sox5, together with Sox6 maintains the columnar alignment of proliferating GP chondrocytes, and delays chondrocyte prehypertrophy (58). qPCR analyses shown in Fig. S6 confirmed the decreased expression of Sox5 in Dot1L cKO chondrocytes. There was significant down-regulation of the SoxC family member, Sox11, whose function has been associated with skeletal progenitor survival, skeletal patterning and growth (57). Although differences in Sox9 gene expression did not reach significance by RNAseq analysis, more rigorous qPCR analysis showed significantly reduced levels in Dot1L-cKO^Prrx1 chondrocytes. Type II collagen and Aggrecan, both downstream target of the Sox trio, and key constituent of cartilage ECM were also decreased in Dot1L-cKO^Prrx1 chondrocytes (Fig. 5G). Expression of transcriptional factors associated with anterior/posterior pattern specification and regionalization including Hoxa and Hoxc clustered genes were significantly down regulated in Dot1L-cKO^Prrx1 chondrocytes. Hoxa11, together with Hoxd11, act upstream of Runx2 and Shox2, key regulators of chondrocyte differentiation (59). RNAseq and qPCR analyses showed significantly lower Runx2 levels, a positive regulatory factor in chondrocyte maturation, in Dot1L-cKO^Prrx1 chondrocytes. Hoxa clustered genes and Runx2 are also among the most dysregulated genes upon Dot1L inactivation in leukemia (38). Shox2, a repressor of chondrogenesis at the initial differentiation of chondroprogenitors into early chondrocytes and the maturation of chondrocytes into hypertrophic chondrocytes, was also significantly decreased in Dot1L-cKO^Prrx1 chondrocytes (60). In humans, functional loss of the pseudoautosomal SHOX gene is linked to osteo-chondrodysplasia characterized by short stature (61). Additional genes with known functional relevance to skeletal growth and development included the paternally imprinted insulin growth factor 2 (IGF2) (62, 63). The insulin growth factor (IGF) pathway plays pivotal roles in chondrocyte proliferation, hypertrophy and bone elongation in skeletal growth plates, and is also implicated in SOC formation (63–65). As shown in Table 3, multiple IGF pathway-related genes exhibited differential expression in Dot1L cKO chondrocytes, suggestive of pathway attenuation. Cadherin signaling, endothelin signaling, integrin signaling, Wnt signaling, and inflammation mediated by chemokine and cytokine signaling raised among the top listed fold enriched pathways by Panther analysis (Fig 8C, Table S3).

Table 3. IGF signaling pathway-related genes.

IGF signaling pathway-related genes exhibit significant differential expression in Dot1L-cKO^Prrx1 chondrocytes as compared to Wt chondrocytes. Absolute fold changes in expression in Dot1L-cKO^Prrx1 chondrocytes relative to wild type levels are shown.

Gene Name	Symbol	Fold Change	Adj. P-value
AKT serine/threonine kinase 3	Akt3	−1.7825	3.60E-16
GRB2 associated binding protein 1	Gab1	−1.5522	5.07E-09
Insulin growth factor 2	Igf2	−4.0181	2.52E-25
Insulin growth factor 2 antisense RNA	Igf2as	−2.7300	3.46E-06
Insulin growth factor 2 mRNA binding protein 1	Igf2bp1	−2.2389	9.91E-05
Insulin growth factor 2 mRNA binding protein 2	Igf2bp2	−1.9206	4.99E-27
Insulin growth factor binding protein 2	Igfbp2	4.6772	1.04E-16
Insulin growth factor binding protein 3	Igfbp3	2.1864	1.44E-14
Insulin growth factor binding protein 4	Igfbp4	1.8487	4.11E-16
Insulin growth factor binding protein 5	Igfbp5	−1.5552	5.14E-10
Insulin growth factor binding protein 6	Igfbp6	1.5207	1.89E-4
Phosphodiesterase 3	Pde3b	−1.6920	1.45E-06

Although Dot1L H3K79 methyltransferase activity is strongly correlated with gene activity (euchromatin), the majority of differentially expressed genes in Dot1L-cKOPrrx1 chondrocytes showed elevated expression. Based on gene ontology, genes up-regulated in Dot1L deficient chondrocytes are largely associated with cell adhesion (GO: 0007155, FDR = 1.0E-10), collagen fibril organization (GO: 0030199, FDR = 9.3E-7), extracellular matrix (ECM) organization (GO: 0030198, FDR = 1.9E-6) and ossification (GO: 0001503, FDR = 4.6E-3) (Table S1). Genes associated with fibrous and mesenchymal cells, such as Col3a1 and Col1a2 were among those showing increased expression. The aberrant expression of genes associated with collagen composition and organization is consistent with the reduced cartilage matrix staining and GP abnormalities in Dot1L-cKO^Prrx1 mice (Fig 5). These findings may indicate formation of a more fibrotic ECM, a consequence of decreased levels of key cartilage transcription factors needed for normal cellular organization of the GP, and cartilage matrix production. Together, our bioinformatic analysis revealed genes and pathways that are sensitive to Dot1L expression in murine growth plate chondrocytes, which had not previously been described, and when disrupted in development, may contribute to the skeletal anomalies.

4. Discussion

In this original report, we characterized a unique skeletal dysplasia phenotype in mice with conditional deletion of Dot1L in uncommitted limb mesenchyme and downstream descendants. Homozygous deletion of Dot1L by Prrx1Cre impeded normal skeletal development, as seen by significant limb shortening, impaired formation of secondary ossification centers, and severe forelimb joint abnormalities. Shortening of the Dot1L-cKO^Prrx1 long bones was accompanied by changes in chondrocyte morphology and organization, in addition to diminished proteoglycan content within the postnatal GP. Quantitative uCt analyses of femora demonstrated low trabecular bone volume and aberrant structure in Dot1L-cKO^Prrx1 and Dot1L-cHet^Prrx1 mice, as compared to Wt controls. Transcriptome analyses confirmed that the enriched functions and components of the differentially expressed genes in Dot1L-cKO^Prrx1 chondrocytes are closely related to growth of cartilage cells, development and remodeling of cartilage ECM, and endochondral bone growth. Together, our data highlight the importance of this chromatin modifier in multiple aspects of long bone development, joint morphology, and stability.

Prior in vivo studies exploring the conditional knockout of Dot1L in cartilage (Col2Cre) were limited by embryonic lethality (37), severe dwarfism, or early postnatal death caused by accelerated ossification of the appendicular and axial skeleton (35, 37). In view of the severe growth retardation and lethal phenotype in Dot1L-cKO^Col2 mice, it may seem surprising that earlier inactivation of Dot1L in uncommitted mesenchyme did not impede survival. While the Prrx1Cre transgene targets osteochondral cells as well as soft tissues derived from early limb mesenchyme, its activity remains restricted to the appendicular skeleton (39). Thus, complications attributed to loss of function in the axial skeleton are avoided in Dot1L-cKO^Prrx1 model, as seen in other systems (66). Survival of Dot1L-cKO^Prrx1 mice afforded us the unique opportunity to explore the in vivo consequences of prenatal disruption of Dot1L expression on skeletal growth and development in a broader context. Importantly, the skeletal defects observed in Dot1L-cKO^Prrx1 mice extended beyond the GP abnormalities and limb shortening observed in Dot1L-cKO^Col2 mice. Specifically, loss of Dot1L by Prrx1Cre resulted in delayed secondary ossification, joint abnormalities and secondary patterning defects that were not reported in Dot1L-cKO^Col2 mice. Dot1L function has been linked to a range of cellular processes including transcriptional activation, resumption of transcription after DNA damage, and cell cycle progression (25, 27, 67). Our data demonstrate the importance of Dot1L for maintaining an appropriate supply of proliferative GP chondrocytes. Impaired chondrocyte proliferation was also observed in mice with postnatal deletion of Dot1L in cartilage using tamoxifen-inducible Aggrecan (Acan)-Cre^ERT2 (37), yet these mice did not display limb shortening. Our observations are also in line with studies of Dot1L knockout from the central nervous system, and other organ systems, which showed the importance of Dot1L in balancing progenitor proliferation and differentiation (68–70). Bioinformatic analyses of differentially expressed genes in Dot1L-cKO^Prrx1 chondrocytes revealed significant down-regulation of genes sets associated with cell cycle processes (Fig. 8A, Table S2). Elevated expressions of cyclin dependent kinase (CDK) inhibitors (ie. Cdkn1c, Cdkn2b) in Dot1L-deficient chondrocytes further support the depletion of chondrocytes that are otherwise available for elongation of the long bones.

Interestingly, our RNAseq analysis revealed that over half of the genes differentially expressed in Dot1L-cKO^Prrx1 chondrocytes at an FDR adjusted P-value <0.05 were up-regulated. Mechanisms for elevated expression of select genes in Dot1L-cKO^Prrx1 chondrocytes remain unclear, however, up-regulated genes are likely to be indirectly controlled by Dot1L, as opposed to direct Dot1L targets. Studies on Dot1L perturbation in other cell types under normal and diseased conditions also found that the majority of deregulated genes were up-regulated (38, 71, 72). Whether the effect in Dot1L-cKO^Prrx1 chondrocytes is dependent on chromatin context remains to be determined. De-repression represents a potential mechanism through which some genes become up-regulated with loss of Dot1L. Ezh2, a member of the Polycomb-repressive complex 2 (PRC2), deposits H3K27me3, a mark involved in repression of developmentally regulated genes. Ezh2 was identified as a gene methylated by H3K79 and dependent on Dot1L expression in other cell types including normal peripheral CD8+ T cells, and thymocytes (56, 73). Lowly expressed genes that become up-regulated with loss of Dot1L contained no or very low H3K79me2 under wild type conditions, and were normally enriched for H3K27me3 (56). Thus, Ezh2 and other PRC components may be relevant targets of Dot1L. In depth ChIPseq experiments, at key time points in development, are needed to fully interrogate a functional connection between Dot1L and Ezh2 for proper development and maintenance of the chondrocyte phenotype.

The aberrant bone mass phenotype in both Dot1L-cKO^Prrx1 and Dot1L-cHet^Prrx1 mice underscore Dot1L dosage needed to support trabecular bone formation. Diminished trabecular bone content and aberrant bone microarchitecture in Dot1L-cKO^Prrx1 mice may be explained by impaired chondrocyte hypertrophy and matrix deposition in the postnatal GP. We acknowledge that uCT analysis were performed only in young male mice from each genotype at a single age, and analysis in male and female Dot1L-cKO^Prrx and Dot1L-cHet^Prrx mice over a time course of development may rule out sex- and/or age-related effects of Dot1L loss of function on bone acquisition and remodeling. Spontaneous osteophytic changes and ectopic ossification in Dot1L-cKO^Prrx1 mice (Fig. 3F) mirrored the aging and OA-associated changes induced by the conditional loss of Dot1L in postnatal cartilage (36, 37). While our manuscript was in preparation, the Schwartz lab achieved postnatal loss of Dot1L in cartilage using the inducible AcanCre^ERT2 system in mice, and similarly reported weakening of trabecular bone in mature mice (37). Increased cortical bone thickness - which does not require endochondral ossification - was also reported in tamoxifen-treated Dot1L^fl/fl:Acan^ERT2 mice and was interpreted as an adaptive response to the weakened trabecular bone (37). In contrast, our uCT analyses of young Dot1L-cKO^Prrx1 femurs revealed decreased cortical thickness, possibly due to exhaustion of a progenitor pool. It is notable that Prrx1Cre causes gene deletion of in all cells of the mesenchymal lineage, including osteoprogenitors. Quantitative bone histomorphometric analyses in Dot1L-cKO^Prrx1 mice may further establish the importance of Dot1L function in bone formation, and remodeling.

Bilateral radial head dislocations in Dot1L-cKO^Prrx1 are intriguing. This anomaly manifests uniquely in Dot1L-cKO^Prrx1 mice, and was not reported in mice with cartilage-specific (Col2Cre) deletion of Dot1L (35, 37). Since Prrx1Cre is active in limb mesenchyme days before Col2Cre, the forelimb defects in Dot1L-cKO^Prrx1 mice point to the developmental requirement of Dot1L prior to overt chondrocyte differentiation, or within other lineages derived from limb mesenchyme (osteoblasts, fibrous connective tissues). In humans, congenital radial head dislocation / subluxation (CRHD/S) is the most common elbow abnormality, and also occurs in conditions involving abnormal GP ossification, short stature, and brachydactyly (74–76). While causes of congenital radial head dislocations are largely unknown, dislocations can be triggered by failed development of the capitellum humeri, depriving the radial head of contact pressure necessary for normal development (77). Collagen abnormalities, and altered HoxD expression are associated with CRHD/S pathogenesis (76). Interestingly, several Hox genes involved in anterior/posterior pattern specification and regionalization were among the top down-regulated genes in Dot1L-cKO^Prrx1 chondrocytes. Single cell RNAseq and genome-wide chromatin immunoprecipitation (ChIPseq) analyses will be helpful towards providing an integrated view of Dot1L-dependent epigenetic and gene expression changes, and for uncovering mechanisms underlying congenital bone deformities and joint abnormalities.

Bony protrusions or ridges arise through secondary patterning processes, giving bones their distinct shape as well as stable anchoring points for muscles inserted to the skeleton by tendons (52). Under-development of the deltoid tuberosity in Dot1L-cKO^Prrx1 mice is suggestive of Dot1L-mediated regulation in modular processes that define bone shape, specifically within progenitors cells from which bone superstructures arise. Formation of the deltoid tuberosity is initiated at E11.5 under the control of Tgfb/Bmp signaling (52, 78), and involving distinct embryonic progenitors that co-express Sox9 and Scleraxis (Scx) (78–80). These progenitors coalesce on the tubular substructure of the developing bones to form eminences. Our RNAseq analyses identified Mab21l2, a Tgfb/Bmp downstream target, among the top down-regulated genes in Dot1L-cKO^Prrx1 cells (Fig. S5A). Mice with mutant Mab21l2 lacked deltoid tuberosities, and exhibited long bone shortening and fused joints, recapitulating features of a human skeletal dysplasia syndrome caused by MAB21L2 mutations (81). Whether Dot1L regulates genetic pathways involved in patterning bone superstructures requires more rigorous investigations.

Developmental bone and joint defects elevate the risk of debilitating conditions, including OA, osteoporosis, and non-healing fractures (82). Impaired epiphyseal ossification is associated with multiple epiphyseal dysplasia (MED), a group of human diseases defined by short stature, joint abnormalities, and early onset OA (83). Interestingly, our observations of delayed secondary ossification in Dot1L-cKO^Prrx1 long bones are divergent from the ectopic chondrocyte hypertrophy and premature ossification seen in mice with cartilage-specific Dot1L deletion (Col2Cre) (37). Our findings point to Dot1L function in periarticular mesenchymal cells that initiate SOC formation. RNAseq data revealed aberrant regulation of pathways implicated in SOC formation, including epidermal growth factor receptor (84), Wnt (53), and IGF signaling pathways in Dot1L deficient chondrocytes (64, 85). Understanding their contribution to aberrant skeletal phenotype in Dot1L-cKO^Prrx1 mice will require thorough mechanistic studies.

We acknowledge that primary chondrocyte cultures used in our RNAseq analyses contained a low level of cell type heterogeneity (Fig S4). RNAseq analysis at the single cell resolution will provide a superior approach for delineating genes sensitive to loss of Dot1L in chondrocytes, and in other skeletal cells. Moreover, single cell ATACseq analysis will reveal cell type-specific changes in chromatin that are attributed to Dot1L loss and associated with congenital bone deformities and developmental joint anomalies seen in Dot1L-cKO^Prrx1 mice.

The DOT1L inhibitor EPZ-5676 has entered clinical development for KMT2A-rearranged leukemia where it has proven to be well tolerated (51). The age group of patients with the highest prevalence of KMT2A rearrangements is infants. In addition, outcomes for infants with KMT2A rearrangements are substantially worse than outcomes for older children with the same translocation (86, 87). Infants therefore constitute the patient population with the greatest need for a targeted therapeutic approach to KMT2A-rearranged leukemia. However, given the critical role that DOT1L plays in pre- and postnatal bone growth, our data raises concern that efficient and prolonged DOT1L inhibition in infants and children would lead to severe growth defects.

We provide compelling evidence that disruption of Dot1L expression in limb mesenchyme leads to defects in GP structure, bone elongation, and trabecular bone formation. Elbow dislocations, aberrant bone morphologies, and defective SOC formation in mice with Dot1L loss in limb mesenchyme and descendants further highlights the unique and specialized role(s) of Dot1L in skeletal development. Dot1L-cKO^Prrx1 represents a tractable system useful for interrogating the underlying epigenetic basis of defects that manifest in skeletal dysplasia.

Highlights.

• Pharmacologic inhibition of Dot1L activity in limb bud micromass accelerated chondrocyte differentiation.

• We generated a novel mouse line with conditional genetic loss of Dot1L gene in mesenchymal progenitors of the limb (Dot1L-cKO^Prrx) to study Dot1L role in skeletal development in a broader context.

• Genetic loss of Dot1L function in limb mesenchyme resulted in limb shortening, impaired secondary bone patterning, and joint dislocations.

• Dot1L critically regulates secondary ossification center formation, trabecular bone formation and microarchitecture.

• RNAseq analysis revealed Dot1L-sensitive transcriptional programs closely related to the growth of cartilage cells and extracellular matrix production and remodeling.