Enhancer of zeste homolog 2 (Ezh2) controls bone formation and cell cycle progression during osteogenesis in mice

Epigenetic mechanisms control skeletal development and osteoblast differentiation. Pharmacological inhibition of the histone 3 Lys-27 (H3K27) methyltransferase enhancer of zeste homolog 2 (EZH2) in WT mice enhances osteogenesis and stimulates bone formation. However, conditional genetic loss of Ezh2 early in the mesenchymal lineage (i.e. through excision via Prrx1 promoter–driven Cre) causes skeletal abnormalities due to patterning defects. Here, we addressed the key question of whether Ezh2 controls osteoblastogenesis at later developmental stages beyond patterning. We show that Ezh2 loss in committed pre-osteoblasts by Cre expression via the osterix/Sp7 promoter yields phenotypically normal mice. These Ezh2 conditional knock-out mice (Ezh2 cKO) have normal skull bones, clavicles, and long bones but exhibit increased bone marrow adiposity and reduced male body weight. Remarkably, in vivo Ezh2 loss results in a low trabecular bone phenotype in young mice as measured by micro-computed tomography and histomorphometry. Thus, Ezh2 affects bone formation stage-dependently. We further show that Ezh2 loss in bone marrow–derived mesenchymal cells suppresses osteogenic differentiation and impedes cell cycle progression as reflected by decreased metabolic activity, reduced cell numbers, and changes in cell cycle distribution and in expression of cell cycle markers. RNA-Seq analysis of Ezh2 cKO calvaria revealed that the cyclin-dependent kinase inhibitor Cdkn2a is the most prominent cell cycle target of Ezh2. Hence, genetic loss of Ezh2 in mouse pre-osteoblasts inhibits osteogenesis in part by inducing cell cycle changes. Our results suggest that Ezh2 serves a bifunctional role during bone formation by suppressing osteogenic lineage commitment while simultaneously facilitating proliferative expansion of osteoprogenitor cells.

Mineralization of fetal tissues to form a bony skeleton is restricted to the later stages of gestation, and expression of bone-specific genes remains silenced during the earlier stages of embryogenesis. Silencing of genes and their regulatory regions is mediated by the tightly controlled organization of promoters and enhancers into nucleosomes, the fundamental units of chromatin (1,2). Condensation of chromosomal loci in heterochromatin silences gene expression and occurs concomitant with post-translational modifications of histone N termini. One of the major mechanisms for heterochromatin formation is through trimethylation of histone 3 at lysine 27 (H3K27me3) 2 (3). The polycomb-repressive complex 2 (PRC2), which contains three structural proteins and the catalytic subunit enhancer of zeste homolog 2 (EZH2), catalyzes the mono-, di-, and trimethylation of H3K27. Enhancer of zeste homolog 1 (EZH1) can also serve as the catalytic subunit within the PRC2 complex and may also suppress gene expression, albeit by different mechanisms (3)(4)(5). Methylation of H3K27me3 is reversible by three histone demethylases, JHDM1D, KDM6A, and KDM6B (6). Hence, deposition and retention of H3K27me3 marks reflect the dynamic balance in the activities of opposing enzymes.
With the exception of Ezh1, mice completely lacking expression of any of the PRC2 components are not viable (18). Thus, studies have assessed tissue-restricted genetic loss of Ezh2 in the mesoderm (35), neural crest (36), mesenchymal stem cells (26,35,37), and chondrocytes (38). Loss of Ezh2 in the mesoderm and neural crest causes embryonic lethality, whereas loss of this epigenetic enzyme in the mesenchyme produces viable mice that have several skeletal abnormalities, including craniosynostosis, clinodactyly, shortened limbs, and reduced bone volume. The loss of Ezh2 in cartilage does not result in a significant phenotype, but dual loss of Ezh1 and Ezh2 severely impairs skeletal development (38). A similar phenotype was observed with the loss of EED, a structural component of the PRC2 complex (39), suggesting that both EZH1 and EZH2 require the PRC2 complex for their function.
In this study, we assessed the role of Ezh2 in osteoblast differentiation and function. We show that conditional Ezh2 loss in osteoblast precursors permits normal patterning and development but shows a remarkable trabecular bone phenotype reflected by reduced bone volume in young animals. Mechanistic assessment of this phenotype indicates that loss of bone is due to cytostatic effects on osteoprogenitor cells. The latter finding indicates that beyond the known biological role of Ezh2 in suppressing osteogenic differentiation, it has a novel positive function in early stages of bone formation by supporting preosteoblast proliferation.

Loss of Ezh2 in bone results in phenotypically normal mice
Our laboratory and others have demonstrated that conditional loss of Ezh2 in the mesenchyme (Ezh2 cKO Prrx1 ) results in several developmental defects (e.g. shorted limbs, craniosynostosis, and clinodactyly) (26,37). A subset of these phenotypic changes are easily understood as skeletal patterning defects that may result from the well-known activation of the Prrx1 driver in the limb bud stage (40). However, to gain insight into the specific role of Ezh2 in the osteoblast lineage and bone, Ezh2 flox/flox mice were bred with Osx-Cre mice (41) to generate control (CON, Ezh2 wt/wt : Osx-Cre) and conditional knockout animals (Ezh2 cKO Osx , Ezh2 flox/flox : Osx-Cre). Loss of Ezh2 in osteoblasts results in animals that are indistinguishable from CON mice (Fig. 1a). In addition, X-ray analysis revealed normal skeletal formation in Ezh2 cKO osx animals (Fig. 1b). Weight assessment of female mice at 3, 8, and 12 weeks of age did not reveal any differences between CON and Ezh2 cKO osx animals (Fig. 1c). However, Ezh2 cKO Osx male animals have slightly lower total body mass compared with CON mice at the earlier time points (3 and 8 weeks of age). Micro-computed tomography (CT) analysis of male and female mice did not reveal any changes in the structure of the skulls (Fig. 1, d and e) or abnormalities of the clavicles (Fig. 1f). Thus, with the exception of minor changes in weights of young male mice, the loss of Ezh2 in osteoblasts (in Ezh2 flox/flox : Osx-Cre mice) results in phenotypically normal mice.

Bone-specific loss of Ezh2 results in a low trabecular bone phenotype in young animals
To assess the effects of Ezh2 loss on bone quality, CT analysis was performed near proximal tibial epiphysis (trabecular bone) and tibial mid-shaft (cortical bone) of a fraction of male mice exhibiting similar weights at several stages of life (3, 8, and 12 weeks of age). Assessment of cortical bone demonstrates that there are no major differences between Ezh2 cKO osx and CON mice (Table S1). A minor reduction in cortical porosity is observed at 8 weeks in Ezh2 cKO Osx animals, which is statistically significant with a moderate level of confidence (p Ͻ 0.05). Analysis of trabecular bone reveals several quantitatively modest changes in bone parameters at 3 weeks of age (Table S2 and Fig. 2a). A reduction in trabecular bone volume (BV/TV), trabecular number, and connectivity density is observed in Ezh2 cKO mice. Although the reduction in these bone parameters is numerically modest, it is significant at an adequate confidence level (p Ͻ 0.05). We also observe an apparent increase in trabecular spacing upon bone-specific loss of Ezh2, but this trend is not statistically significant (p Ͼ 0.05). Interestingly, no significant differences in trabecular bone parameters are observed at 8 weeks (Table S2 and Fig. 2b) and 12 weeks (Table S2 and Fig. 2c) of age between CON and Ezh2 cKO Osx animals. Histomorphological analysis was performed on femora derived from 3-week old male mice (Fig. 3). Von Kossa-McNeal (a) and Goldner's trichrome (b) staining of the distal femora demonstrate that the growth plate and cartilage morphologies are similar between CON and Ezh2 cKO animals. However, Goldner's trichrome staining of the femora mid-shafts reveals that cKO animals exhibit enhanced bone marrow adiposity (c). Similar to CT analysis at 3 weeks of age, histomorphometric analysis of distal femora shows a reduction in trabecular bone volume (BV/TV) with the loss of Ezh2 in osteoblasts (p Ͻ 0.05) (d). The change in bone volume is accompanied by a reduction in osteoblasts surface per bone surface, osteoclast surface per bone surface, and number of osteoclasts per bone perimeter (p Ͻ 0.05 in all cases). Although we observed an apparent decrease in the osteoblasts surface per bone surface in adult cKO mice, these numerical changes do not reach significance (p Ͼ 0.05) (Fig.  S1). Yet, reductions in osteoclast surface per bone surface remained lower in Ezh2 cKO mice at 12 weeks of age (p Ͻ 0.05). Thus, whereas the deletion of Ezh2 in osteoblasts does not alter Ezh2 function in proliferating osteoprogenitor cells growth plate and cartilage formation, Ezh2 loss leads to increased bone marrow adiposity and yields a mild low trabecular bone phenotype in young animals. The latter finding indicates that Ezh2 function in osteoblasts is required for formation or maintenance of trabecular bone.

Cre-specific recombination of the Ezh2 conditional allele results in the loss of EZH2 protein in primary calvarial osteoblasts
The SET domain of EZH2 possesses H3K27 methyltransferase activity, and its loss renders EZH2 inactive. To assess the efficiency of the Cre-recombination of the Ezh2 conditional allele (42) in osteoblasts, primary calvarial osteoblasts derived from Ezh2 flox/flox were infected with adenovirus expressing GFP (Ad-GFP) or an adenovirus expressing GFP and Cre recombinase (Ad-GFP-Cre) (Fig. 4a). The administration of Ad-GFP to these cells is expected to generate WT Ezh2, whereas the addition of Ad-GFP-Cre is predicted to result in truncated Ezh2 lacking the SET domain (Fig. 4b). As anticipated, administration of both adenoviruses results in the expression of GFP, whereas Cre recombinase expression is only detected in the cells infected with Ad-GFP-Cre (Fig. 4c). To detect the mutant Ezh2 allele, a primer pair that spans across exons 15 and 20 was used (Fig. 4b). The addition of Ad-GFP does not induce gene rearrangement, but the addition of Ad-GFP-Cre results in the formation of the mutant Ezh2 allele (Fig.  4c). Furthermore, we observe a reduction in the expression of the SET domain and total Ezh2 mRNA levels. The addition of Ad-GFP-Cre results in the loss of EZH2 protein levels as detected by Western blotting 4 days after infection (Fig. 4d). Collectively, these results demonstrate that Cre-specific recombination of the Ezh2 locus results in fusion of exons 15 and 20 at the mRNA level and the loss of EZH2 protein in primary osteoblasts.

Loss of functional Ezh2 inhibits osteoblast differentiation of bone marrow-derived mesenchymal stem cells
To assess a direct effect of Ezh2 loss on progenitor cells, BMSCs were isolated from 6 -8-week-old male and female CON and Ezh2 cKO Osx mice and differentiated into the osteogenic lineage (Fig. 5). RT-qPCR analysis demonstrates that expression of osteogenic genes (e.g. osterix/Sp7, osteocalcin/

Ezh2 function in proliferating osteoprogenitor cells
Bglap, alkaline phosphatase/Alpl, and bone sialoprotein/Ibsp) is suppressed in male and female Ezh2 cKO Osx BMSCs undergoing osteogenic differentiation (Fig. 5a). In support of RT-qPCR analysis, alizarin red staining of BMSCs cultures on day 17 of osteogenic differentiation demonstrates a significant reduction in mineral deposition with Ezh2 loss (Fig. 5, b and c). Similar to alizarin red staining, a reduction in alkaline phosphatase staining is observed with BMSCs derived from Ezh2 cKO osx animals (Fig. S2). Western blot analysis reveals that EZH2 expression is reduced, whereas H3K27me3 is increased during osteogenic commitment of mouse-derived BMSCs (Fig.  5d). A reduction in EZH2 and H3K27me3 is observed (see days 3 and 7) in BMSCs derived from Ezh2 cKO osx animals. As a loading control, the levels of histone H3 remain stable during the differentiation time course. Hoechst staining, which is a proxy for the number of cells in each treatment condition, reveals that Ezh2 loss results in the reduction of DNA content at day 7 (Fig. 5e). These results show that Ezh2 is only actively expressed during early proliferative stages of BMSC differentiation and that H3K27me3 levels are controlled by another enzyme (e.g. EZH1) at later maturation stages of osteogenic differentiation in BMSCs. Furthermore, the loss of osteogenic potential of BMSCs upon genetic Ezh2 loss demonstrates that Ezh2 performs a critical and previously unrecognized positive function during proliferative expansion of BMSCs as they gradually become committed to Osx-positive osteoprogenitor cells within the population.
It is plausible that loss of Ezh2 in Osx-positive osteoblasts in vivo may affect regulatory interplay with BMSCs in situ within the bone environment. Hence, the BMSCs isolated from Osx-Cre-positive mice may differ from naive BMSCs developing in the bone micro-niche of WT mice. Therefore, we assessed BMSCs differentiation after acute deletion of Ezh2 in vitro utilizing Ad-GFP and Ad-GFP-Cre virus particles on BMSCs from Ezh2 flox/flox animals that are otherwise genetically and phenotypically normal (Fig. 6a). Corroborating our data obtained with primary calvarial osteoblasts (Fig. 4), the addition of Ad-GFP-Cre resulted in the expression of Cre recombinase and the truncated version of Ezh2 (Ezh2 mut ) (Fig. 6b). In addition, a reduction in WT mRNAs that encompass the Ezh2 SET domain (Ezh2 SET ) and total Ezh2 mRNA is observed. Acute loss of Ezh2 results in a small, but significant, reduction in cell viability as measured by MTS activity 10 days after the addition of virus particles (Fig. 6c). Although not significant, DNA content as measured by Hoechst staining is also reduced as a result of Ezh2 loss in BMSCs (Fig. 6d). Acute deletion of Ezh2 reduces

Ezh2 function in proliferating osteoprogenitor cells
the expression of several osteogenic markers (e.g. Bglap, Ibsp, Alpl, and osteomodulin/Omd) at day 14 after osteogenic induction (Fig. 6e). In support of mRNA reduction of key osteogenic genes, a significant reduction in alizarin red staining of these primary cell cultures is observed (Fig. 6, f and g). Western blot analysis demonstrates a loss of EZH2 protein and a reduction in H3K27me3 with the addition of Ad-GFP-Cre (Fig. 6h). Consistent with the loss of the exon encoding the SET domain upon Cre excision, a faster migrating EZH2 protein band is observed after the recombination event, reflecting the truncated shortlived residual EZH2 protein that lacks methyl transferase activity. Taken together, our data show that EZH2 protein expression is lost during both in vivo excision (Ezh2 cKO Osx ) and ex vivo excision of Ezh2 (Ad-GFP-Cre with Ezh2 flox/flox BMSCs). Our results indicate that loss of EZH2 protein reduces cellular H3K27me3 levels, which in principle is permissive for osteogenic differentiation but may simultaneously inhibit osteogenic differentiation of primary mouse BMSC populations by reducing the number of osteogenic progenitor cells. This finding corroborates the emerging idea that Ezh2 performs an essential cellular function during osteoprogenitor proliferation while suppressing the bone cell maturation program.

Acute loss of Ezh2 impacts cell cycle progression in mouse BMSCs
Because loss of Ezh2 reduces cell number (as assessed by DNA content) and metabolic activity (as monitored by MTS assays), we examined the contribution of Ezh2 to cell cycle progression of BMSCs derived from WT (C57/BL6) or Ezh2 flox/flox mice (Fig. 7a). The infection of these BMSCs with Ad-GFP-Cre virus results in expression of GFP in both female and male BMSCs (Fig. 7b), which results in the expected recombination of Ezh2 locus in Ezh2 flox/flox BMSCs (data not shown). RT-qPCR analysis shows stable expression of the housekeeping gene eukaryotic elongation factor 1␣1 (Eef1a1) when normalized to that of the housekeeping gene Gapdh at day 6 after infection ( Fig. 7c). However, loss of Ezh2 suppresses the expression of two proliferation markers, Mki67 and Ccnb2, that encode the Ki67 antigen and cyclin B2, respectively. Furthermore, the two principal mRNA products (p16 and p19) of the Cdkn2a locus, which are critical inhibitors of cell cycle progression, are significantly up-regulated with Ezh2 loss. A similar trend is also observed with Cdkn1a (p21), another cell cycle inhibitor. Flow cytometric analysis shows that loss of Ezh2 alters the distribution of cells in the different phases of the cell cycle (Fig. 7d). In both male and female BMSCs, the loss of Ezh2 significantly shifts the cells from S/G 2 phases into the G 0 /G 1 phases of the cell cycle, suggesting that cell cycle progression of BMSCs is impeded in the G 0 or G 1 phase when EZH2 activity is eliminated. Because male and female cells were processed independently, these technical considerations limit direct comparison of cell cycle distributions between the two sexes. In support of these findings, MTS activity assays show reduced metabolic activity on days 6 and 11 with Ezh2 loss (Fig. 7e). No significant changes are observed day after plating (day 0) and 3 days after the addition of the virus to the cell populations. Thus, Ezh2 loss in BMSCs reduces the number of cells by attenuating G 0 or G 1 cell cycle progression in BMSCs.

Ezh2 function in proliferating osteoprogenitor cells Altered expression of cell cycle genes in mouse calvaria lacking Ezh2 expression
To investigate the mechanism by which Ezh2 controls cell cycle progression, we examined in vivo gene expression by RNA-Seq analysis of mRNAs derived from calvaria of CON and Ezh2 cKO pups (Fig. 8). Consistent with genomic truncation of the Ezh2 locus through Cre excision over loxP sites, RNA-Seq analysis of mRNA collected from Ezh2 cKO calvaria exhibits a marked reduction in the number of reads mapped to exons 16 -19 (Fig. 8, a and b). Bio-informatic analysis of genes with robust average expression (Ͼ0.1 reads/kilobase pair/million mapped reads (RPKM)) demonstrates that ϳ500 genes are upregulated and ϳ200 genes are down-regulated upon loss of Ezh2 (fold change Ͼ 1.4 and p value Ͻ 0.05) (Fig. 8c). Similar to observations in primary BMSC cultures, the cell cycleinhibitory gene Cdkn2a is one of the most prominently up-regulated genes in Ezh2 cKO calvaria (Fig. 8, c and d). Of the ϳ700 differentially expressed genes, about 8.5% (n ϭ 61) associate with the gene ontology biological process term "cell cycle" (Fig.  8e). These RNA-Seq results corroborate findings obtained with ex vivo cultures of BMSCs and establish that genetic loss of Ezh2 perturbs in vivo expression of cell cycle-regulatory genes in mouse calvaria.

Discussion
Our group has pursued multiple strategies for understanding the cell-autonomous effects of Ezh2 in mesenchymal stromal cells and committed osteoblasts within bone tissue (reviewed in Ref. 18). We have previously shown that genetic inactivation of Ezh2 in calvarial bones (Ezh2 cKO Prrx1 , Ezh2 flox/flox : Prrx1-Cre) enhances expression of osteogenic genes. These mice exhibited several major phenotypic changes, including a domed head and premature suture fusion (i.e. craniosynostosis) (26), a phenotype that could originate from any development stage when the Prrx1 promoter first begins to express Cre. This permanent genetic deletion of Ezh2 at an early developmental stage limited our interpretation. However, these studies did not permit a direct assessment of the acute cell-autonomous mechanistic effects of Ezh2 deletion in osteoblasts during normal skeletal patterning. Our present study assessed the genetic role of Ezh2 in the osteoblastic lineage and demonstrates that Ezh2 is necessary for early proliferative stages of osteoblastogenesis. This function complements its known activity as a suppressor of osteoblast maturation.
Conditional deletion of Ezh2 in osteoblasts in vivo (using the Osx/Cre driver) results in mice with normal skeletal patterning. Female animals do not exhibit any obvious anatomical changes, whereas male cKO mice weigh less than CON animals at 3 and 8 weeks of age. CT reconstructions indicate that skull structure, calvaria suture formation, and clavicle formation are within the normal range of variation in mice lacking Ezh2 in the osteoblast lineage. Interestingly, CT analysis reveals a mild low trabecular bone mass phenotype at 3 weeks of age that resolves at 8 weeks. These findings are supported by histomorphometric analysis at 3 weeks of age. Interestingly, a reduction in osteoblasts is accompanied by a reduction in osteoclasts at 3 weeks of age, suggesting that bone formation and resorption remain coupled with osteoblastic loss of Ezh2. This coupling effect is also observed at 12 weeks of age, where the number of osteoclasts is significantly different between control and Ezh2 cKO animals. Histologic assessment reveals normal growth plate and cartilage development but shows enhanced formation

Ezh2 function in proliferating osteoprogenitor cells
of cells positive for lipid droplets in the bone marrow cavity of cKO animals. Loss of Ezh2 in mouse BMSCs reduces osteogenic differentiation ex vivo, in part because of negative effects on cell cycle progression that occur concomitant with up-regulation of cyclin-dependent kinase inhibitors (e.g. CDKN2A/p16 INK4A -p19 ARF ) and down-regulation of genes supporting mitotic division (e.g. cyclin B2). Consistent with these results, changes in the expression of cell cycle genes are also observed in Ezh2 cKO calvaria. Of interest, BMSCs from 6 -8-week-old Ezh2 cKO animals exhibit a reduction in osteogenic differentiation, but the bone parameters resolve at this age in vivo. This suggests that Ezh2 is required for the commitment of progenitor cells into the osteogenic lineage, whereas coupling between osteoblast and osteoclast restores net trabecular bone formation in vivo. We note that the contribution of Ezh2 to cell cycle progression has been well documented (43)(44)(45) for nonosseous cell types and that small-molecule inhibitors of EZH2 are used clinically to inhibit cancer cell proliferation (46, 47). Taken

Ezh2 function in proliferating osteoprogenitor cells
together, our study indicates that Ezh2 controls osteoblastic differentiation by both supporting cell cycle progression of osteoblast progenitor cells and suppressing maturation of precursor cells that are committed to the osteoblast lineage.
The gain in our understanding that Ezh2 is a bifunctional regulator of osteoblastogenesis clarifies several other previous results. For example, previous studies have shown that mesenchymal loss of Ezh2 results in several skeletal abnormalities (26,37), and these apparent patterning defects can now be interpreted as resulting from a reduction in the expansion of committed osteoblast progenitors when Ezh2 is inactivated in Ezh2 cKO mice that express the Prrx1-Cre driver. The in vivo importance of Ezh2 in osteoprogenitor cells during early stages of bone formation in skeletally immature mice is corroborated by our observation that Ezh2 cKO that express the Osx-Cre driver have a transient low trabecular bone volume phenotype in young adult mice. Thus, genetic inactivation of Ezh2 in early stages of skeletogenesis generally has negative biological effects. However, Ezh2 inhibition also has clear bone-stimulatory potential from the perspectives of both bone tissue engineering and bone anabolic therapy. For example, transient loss of EZH2 function (knockdown or small-molecule inhibitors) in cultured mesenchymal cells stimulates osteogenic differentiation in vitro, and EZH2 inhibition also enhances in vivo bone formation and prevents bone loss associated with estrogen depletion (26,28,29,33,34).
Our previous work has indicated that Ezh2 loss in the mesenchymal compartment yields a senescence-like phenotype, characterized by enhanced expression of Cdkn2a, a well-characterized cell cycle and senescence marker, as well as enhanced bone marrow lipid droplet formation (26). Deletion of Ezh2 co-regulators, Hdac3 and Zfp521, in osteoblast progenitors produces similar marrow phenotypes (48 -50). Our current study also shows that Ezh2 loss results in bone marrow adiposity and enhanced expression of Cdkn2a. It is plausible that the absence of Ezh2 could potentially activate an adipogenic program that causes accumulation of fat droplets in cells that are precommitted to the osteogenic lineage. Indeed, fate-mapping studies have shown that Osx-Cre-positive cells can contain lipid droplets (51,52). These data suggest that, similar to Hdac3 (53), Ezh2 could potentially regulate lipid storage in pre-osteoblasts. It is noteworthy that the expression of Cdkn2a is more

Ezh2 function in proliferating osteoprogenitor cells
robust in primary calvaria derived from animals that lack Ezh2 expression in the mesenchyme (26) when compared with animals that lack expression of this epigenetic enzyme in pre-osteoblasts (current study). Our current and past findings are also supported by a recent study that shows that the loss of Ezh2 in Nestin-expressing cells, a population of mesenchymal stem cells, induces a senescence phenotype in mice (54).
Recent studies have demonstrated that EZH2 possesses methyltransferase-independent function (55)(56)(57). Therefore, we considered that Ezh2 inactivation by elimination of the SET domain in principle still could result in a truncated EZH2 protein that may retain some of its biochemical functions. We consider it less likely as this mutant protein is rapidly degraded in mesenchymal cells. Permanent Ezh2 loss by genetic inactivation is expected to destabilize the PRC2 complex, resulting in weakening of chromatin structure and inducing gene expression changes indirectly. Long-term loss during development could potentially provoke compensatory mechanisms to ensure skeletal formation that proceeds through less efficient mechanisms.
Our group and others have previously presented solid data in support of the concept that Ezh2 inhibition has bone-stimulatory properties (26, 28, 29, 33, 34). The present work indicates that inhibitory strategies could also provoke negative effects on bone, especially during initial stages of bone growth. Notwithstanding this limitation in these negative biological effects during maturation of the mammalian skeleton, it may still remain realistic to leverage EZH2 inhibition for short-term, local applications in skeletally mature mammalian species and in bone tissue engineering to accelerate maturation of committed osteoblasts. Therapeutic strategies utilizing EZH2 inhibitors could potentially be employed to enhance spinal fusion, stimulate orthopedic implant grafting, and heal nonunion fractures. However, EZH2 inhibition is clearly not beneficial for adolescent or young adult mice; nor does it appear to be suitable for long-term applications (i.e. chronic conditions such as osteoporosis). Our current data indicate that undesired effects of EZH2 suppression, including cell cycle impediment and senescence, may counter the beneficial effects (i.e. enhanced expression of osteogenic genes) in bone-stimulatory strategies.

Bone marrow mesenchymal stem cell cultures
BMSCs were isolated from WT C57Bl/6 mice, Ezh2 flox/flox animals, or Ezh2 CON or Ezh2 cKO Osx mice, as described (59). Briefly, bone marrow was flushed from femurs and tibias of 6 -8-week-old mice, which is an age that yields a robust and consistent number of BMSCs. Cells were immediately seeded for experiments in 6-and 12-well plates (1 million cells/cm 2 ) or cultured in T175 flasks for 4 -6 days before seeding the adherent cell population in 6-and 12-well plates (20,000 cells/cm 2 ). BMSCs were maintained in ␣MEM supplemented with 20% FBS, 1% antibiotic and antimycotic, and 1% nonessential amino acids. BMSCs were selected by their ability to adhere to the culture plates (at least 3 days) before exposure to adenovirus and differentiation medium. Osteogenic differentiation was accomplished by the addition of 50 g/ml ascorbic acid, 10 mM ␤-glycerol phosphate, and 10 Ϫ8 M dexamethasone in the maintenance medium. Media were changed every 3 days.

MTS activity assay
MTS activity was assayed according to the manufacturer's protocol (Promega). Absorbance was measured at 490 nm using a SpectraMAX Plus spectrophotometer (Molecular Devices).

Hoechst staining
Medium was aspirated from wells, and cells were washed one time with PBS. After aspiration of PBS, 500 l of Hoechst 33258 (Sigma, 94403) (0.5 l/ml in PBS) was added to each well. Plates were left to incubate in the dark at room temperature for 20 min. After incubation, fluorescence intensity was measured at a 340-nm excitation wavelength and 460-nm emission wavelength using an F200 Infinite Pro (Tecan) plate reader. Meas-

Ezh2 function in proliferating osteoprogenitor cells
urements were fit to a standard DNA curve to determine relative DNA content.

RT-qPCR
RNA was isolated using the Direct-zol TM RNA kit (Zymo Research). RNA was then reverse-transcribed into cDNA by the SuperScript III first-strand synthesis system (Invitrogen). Transcript levels were measured using the 2 ⌬⌬Ct method and normalized to the housekeeping gene Gapdh (set at 100). Genespecific primers are shown (Table S3).

Alkaline phosphatase and alizarin red staining
For alkaline phosphatase activity, cell cultures were fixed in 10% neutral buffered formalin and stained with 5-bromo-4chloro-3-indolyl-phosphate/nitro blue tetrazolium to monitor the enzymatic activity of alkaline phosphatase (Promega). For alizarin red staining, cells were also fixed in 10% neutral buffered formalin and stained with 2% alizarin red to visualize calcium deposition. Absorption of alizarin red stain was quantified with ImageJ software (60).

Animal welfare
Animal studies were conducted according to guidelines provided by the National Institutes of Health and the Institute of Laboratory Animal Resources, National Research Council. The Mayo Clinic Institutional Animal Care and Use Committee approved all animal studies. Animals were housed in an accredited facility under a 12-h light/dark cycle and provided water and food (PicoLab Rodent Diet 20, LabDiet) ad libitum.

Micro-computed tomography analysis
Right tibias were harvested from 3-, 8-, and 12-week-old CON and Ezh2 cKO osx mice. Soft tissue surrounding the bone was carefully removed, and samples were fixed in 10% neutral buffered formalin for 48 h followed by storage in 70% ethanol. Tibias were scanned with an ex vivo CT system (Skyscan 1174, Skyscan) equipped with a 50-kV, 800-A X-ray tube and a 1.3megapixel charge-coupled device coupled to a scintillator. Samples were maintained in a moist environment during scanning to prevent dehydration and scanned in air using 8-m (3 and 8 weeks old) or 10-m (12 weeks old) isotropic voxels with an integration time of 400 ms. The region of analysis selected was between 13 and 16% of bone length (3-week-old mice) or between 12 and 15% of bone length (8-and 12-week-old mice) relative to the proximal tibial epiphysis for trabecular bone analysis. Mid-shafts of the tibias were assessed for the cortical parameters. Reconstructions were performed in NRecon software (Skyscan) and subjected to morphometric analyses using CTAn software (Skyscan). A minimum gray scale threshold of 60 (3 weeks old) or 70 (8 and 12 weeks old) was used.
Skulls were harvested from 3-week-old CON and Ezh2 cKO osx mice. Samples were fixed in 10% neutral buffered formalin for 48 h followed by storage in 70% ethanol. Skulls were scanned using a Scanco vivaCT40 system (Scanco) at 70 kV, 114 A with an integration time of 221 ms for a 10.5 isometric voxel size. 3D renderings were created using Microview (Parallax).

Histology and histomorphometric analysis
Right femurs were harvested from 3-and 12-week-old CON and Ezh2 cKO osx animals. Soft tissue surrounding the bone was carefully removed, and samples were fixed in 10% neutral buffered formalin for 48 h followed by storage in 70% ethanol. Undecalcified femurs were embedded in methyl methacrylate resin and sectioned using a rotary retracting microtome. Sections of 5-m thickness were stained by Gomori's trichrome method as described previously (61). Calcified bone was visualized in green and osteoid in red. Quantitative histomorphometry was performed utilizing OsteoMeasure TM to estimate the percentage of bone volume (BV/TV, %), osteoblast surface, osteoid surface, and number of osteoblasts and osteoclasts per bone parameter. In addition, trabecular thickness (m) was also derived from trabecular parameters. Six fields of view were measured, scanning a total area of 2.1 mm 2 at a distance of ϳ1 mm away from the growth plate. All measurements were made in the trabecular region. Cortical bone was marked as nontissue and was not included in the measurements.

Flow cytometric analysis
Cells were harvested by trypsin, washed with PBS, and fixed in 70% ethanol overnight at 4°C. Cells were then washed two times with PBS and stained with 0.5 ml of FxCycle propidium iodide/RNase staining solution (Molecular Probes) for 30 min at room temperature in the dark. Cell cycle distribution (Ͼ10,000 events) was collected on the FACSCanto system (BD Biosciences), and the data were analyzed utilizing FACSDiva version 8.0.1 software (BD Biosciences).

High-throughput RNA sequencing and bioinformatic analysis of primary calvaria
Calvaria were dissected out from 2-3-day-old pups, washed in PBS, digested for 20 min in collagenase digestion medium (␣MEM, 2 mg/ml collagenase type II, and 0.005% trypsin) at Ezh2 function in proliferating osteoprogenitor cells 37°C, washed in PBS, and snap-frozen with liquid nitrogen and stored at Ϫ80°C. RNA was then isolated utilizing Ultra Turrax T25 (IKA) and the Direct-zol TM RNA kit (Zymo Research). High-throughput read mapping and bioinformatic analyses for RNA-Seq were performed as reported previously (26,27,29,62). Gene expression is expressed in RPKM. RNA-Seq data were deposited in the Gene Expression Omnibus of the National Institute for Biotechnology Information (GSE111245). Gene tracks were viewed using Interactive Genome Viewer (IGV) version 2.3.98 with the mm9 build loaded (63,64). Cell cycle-related genes (n ϭ 1742) have been obtained from the gene ontology consortium (65,66).

Ex vivo assays
For assays that utilized primary calvarial osteoblasts and BMSCs, cells from several animals (typically 2-5) were pooled and then plated out as three biological replicates. These assays were repeated at least three times. Representative experiments are shown.

Statistics
For ex vivo studies, data are shown as mean Ϯ S.D., and statistical analysis was performed with unpaired Student's t test.
For in vivo studies, results are depicted by scatter plots that contain mean Ϯ S.E. with each dot representing one mouse. To assess for bone parameters (CT and histomorphometry), 6 -10 animals derived from at least three independent litters were used in these studies. Statistical analysis was performed using Student's unpaired t test or Tukey post hoc tests for multiple comparisons using GraphPad Prism version 7 software. Significance is noted in the figures, when applicable (*, p Ͻ 0.05).