Tet1 and Tet2 maintain mesenchymal stem cell homeostasis via demethylation of the P2rX7 promoter

Ten-eleven translocation (Tet) family-mediated DNA oxidation represents an epigenetic modification capable of converting 5-methylcytosine (5-mC) to 5-hydroxymethylcytosine (5-hmC), which regulates various biological processes. However, it is unknown whether Tet family affects mesenchymal stem cells (MSCs) or the skeletal system. Here we show that depletion of Tet1 and Tet2 results in impaired self-renewal and differentiation of bone marrow MSCs (BMMSCs) and a significant osteopenia phenotype. Tet1 and Tet2 deficiency reduces demethylation of the P2rX7 promoter and downregulates exosome release, leading to intracellular accumulation of miR-297a-5p, miR-297b-5p, and miR-297c-5p. These miRNAs inhibit Runx2 signaling to impair BMMSC function. We show that overexpression of P2rX7 rescues the impaired BMMSCs and osteoporotic phenotype in Tet1 and Tet2 double knockout mice. These results indicate that Tet1 and Tet2 play a critical role in maintaining BMMSC and bone homeostasis through demethylation of P2rX7 to control exosome and miRNA release. This Tet/P2rX7/Runx2 cascade may serve as a target for the development of novel therapies for osteopenia disorders. Tet-mediated DNA oxidation converts 5-methylcytosine (5-mC) to 5-hydroxymethylcytosine (5-hmC), which is essential to regulate different biological processes. Here the authors show that Tet1 and Tet2 regulate mesenchymal stem cell and bone homeostasis through demethylation of P2rX7 promoter.

T he ten-eleven translocation (Tet) family is a group of DNA demethylases capable of regulating various epigenetic responses. Tet proteins, including Tet1, Tet2, and Tet3, are able to convert 5-methylcytosine (5-mC) to 5hydroxymethylcytosine (5-hmC) and its oxidative derivatives in Fe(II)-and alpha-ketoglutarate (α-KG)-dependent oxidation reaction to promote DNA demethylation and gene transcription [1][2][3][4] . Previous studies showed that 5-hmC is abundant in both adult cells and embryonic stem cells (ESCs) [5][6][7] . Upon ESC differentiation, the expression levels of Tet1 and Tet2 are downregulated, suggesting that Tet1 and Tet2 may be associated with the maintenance of ESC pluripotency through regulation of lineage-specific genes 1 . It was reported that the expressions of Tet1 and Tet2 were regulated by Oct4/Sox2 complex, and the depletion of Tet1 impairs the self-renewal and differentiation of ESCs 5,8 . In contrast to its role in maintaining ESC pluripotency, Tet proteins have different effects on adult stem cells. Hematopoietic stem cells (HSCs) from Tet2-mutant mice exhibit increased repopulating capacity, augmented HSC expansion, and impaired differentiation toward the myeloid lineage [9][10][11] . Moreover, Tet2 and 5-hmC levels are increased during smooth muscle cell differentiation 7 . Adult Tet1-mutant mice exhibit defective self-renewal of neural progenitor cells and impaired memory extinction 12 . These studies suggest that Tet proteins may possess unique functional roles in epigenetic regulation of stem cell function. Depletion of Tet1 resulted in a reduction of 5-hmC levels and impaired chondrogenic differentiation in a chondroprogenitor cell line 13 . Additionally, 75% of the newborn Tet1depletion mice exhibit a smaller body size at birth but seemed to gain their body weight when growing to 6-9 weeks 14 , suggesting a potential skeletal defect. However, the role of Tet proteins in skeletal development remains largely unknown.
Bone marrow mesenchymal stem cells (BMMSCs) are a population of non-hematopoietic multipotent stem cells with selfrenewal and multipotent differentiation capacities. They play an essential role in maintaining bone/marrow homeostasis 15,16 . BMMSCs can be regulated at both transcriptional and epigenetic levels in response to stimulation from various environmental elements [17][18][19] . BMMSC deficiency may contribute to bone degenerative phenotypes in osteopenia disease models 20,21 .
In this study, we show that Tet1 and Tet2 are required to maintain BMMSC and bone homeostasis. The depletion of Tet1 and Tet2 may lead to hypermethylation of the P2rX7 promoter to block miR-297a-5p, miR-297b-5p, and miR-297C-5p release, leading to downregulation of Runx2 signaling and osteopenia phenotype.

Results
BMMSCs express Tet proteins. Since Tet proteins are expressed in various tissues and play an essential biological role in epigenetic regulation, we hypothesized that Tet proteins may affect BMMSC function. We found that both human and mouse BMMSCs express Tet1, Tet2, and Tet3, as assessed by western blotting and real-time polymerase chain reaction (qPCR; Fig. 1a, b). Double immunostaining confirmed that BMMSCs co-express CD146, a mesenchymal stem cell marker, with Tet1, Tet2, and Tet3 (Fig. 1c). It was reported that different Tet proteins may display distinct roles in developmental processes 9 . To explore the possible roles of Tet family members in maintaining BMMSC and bone homeostasis, we used a BMMSC impairment model (ovariectomized (OVX) mice) to assess whether the expression levels of Tet family members were altered in impaired BMMSCs 22 . Microcomputed tomography (micro-CT) and histological analysis confirmed that bone mineral density (BMD), cortical bone area (Ct.Ar), cortical thickness (Ct.Th), and distal femoral trabecular bone volume of OVX mice were markedly decreased compared with the sham-treated group ( Supplementary Fig. 1a-c). The number of colony-forming unit fibroblasts (CFU-F) was significantly elevated in OVX BMMSCs ( Supplementary Fig. 1d). Bromodeoxyuridine (BrdU)-labeling assay confirmed that OVX BMMSCs had an increased proliferation rate ( Supplementary  Fig. 1e). Moreover, OVX BMMSCs showed impaired osteogenic differentiation, as indicated by reduced mineralized nodule formation assessed by alizarin red staining and reduced expression of the osteogenic genes runt-related transcription factor 2 (Runx2), alkaline phosphatase (ALP), and osteocalcin (OCN) after 14 days of osteogenic induction, assessed by western blotting (Supplementary Fig. 1f,g). Interestingly, we found that the expression levels of Tet1 and Tet2 were significantly decreased in OVX BMMSCs (Supplementary Fig. 1h). However, the Tet3 expression level remained unchanged. We detected a reduced level of 5-hmC in OVX mice bone marrow stem cells using immunostaining and dot blot assay ( Supplementary Fig. 1i-1j), consistent with the fact that Tet proteins convert 5-mC to 5-hmC.
Tet DKO mice show osteopenia phenotype and BMMSC impairment. To explore the role of Tet1 and Tet2 in maintaining BMMSC and bone homeostasis, we compared the bone phenotype of Prx1 cre (control), Tet1 −/− , Prx1 cre Tet2 fl/fl , Tet1 −/− ; Prx1 cre Tet2 fl/fl double knockout (Tet DKO) mice at 8-10 weeks of age. Micro-CT and histological analysis showed that Tet DKO mice, but not in Tet1 −/− mice, had significantly reduced BMD, bone volume/tissue volume (BV/TV), Ct.Ar, Ct.Th, and the distal femoral trabecular bone volume (Fig. 2a-c) compared to littermate controls (Fig. 2a-c). The BMD, BV/TV, Ct.Ar, and the distal femoral trabecular bone volume of Tet DKO mice were significantly lower than Tet1 −/− and Prx1 cre Tet2 fl/fl mice, and the bone volume of Prx1 cre Tet2 fl/fl mice were lower than control group (Fig. 2a-c and Supplementary Fig. 2a). Then, we also performed a calcein-labeling assay to show that the Tet DKO mice had a lower bone turn-over rate, which indicated that their bone formation rate was comparatively decreased (Fig. 2d).
To examine whether Tet1 and Tet2 affect BMMSC function, we isolated BMMSCs from 8-10-week-old Tet1 −/− , Prx1 cre Tet2 fl/fl , Tet DKO mice and littermate controls ( Supplementary Fig. 2a, b). Flow cytometric analysis showed that BMMSCs from both control and Tet DKO mice were positive for stem cell surface markers Sca1, PDGFR, CD105, CD90, and CD73, but were negative for hematopoietic lineage markers CD34 and CD45 ( Supplementary  Fig. 2c) 23 . The number of CFU-F was significantly elevated in Tet DKO but not Tet1 −/− and Prx1 cre Tet2 fl/fl BMMSCs compared to the control group (Fig. 3a). BrdU-labeling assay confirmed that Tet DKO but not Tet1 −/− and Prx1 cre Tet2 fl/fl BMMSCs had an increased proliferation rate (Fig. 3b). In addition, we found that Tet DKO and Prx1 cre Tet2 fl/fl BMMSCs but not Tet1 −/− BMMSCs showed reduced osteogenic differentiation, as indicated by reduced mineralized nodule formation (Fig. 3c) and decreased expression levels of osteogenic-related genes Runx2, ALP, and OCN (Fig. 3d). In addition, the osteogenic differentiation capacity of Prx1 cre Tet2 fl/fl BMMSCs was significantly higher than Tet DKO BMMSCs (Fig. 3c, d). We further confirmed that BMMSCs derived from Tet DKO and Prx1 cre Tet2 fl/fl mice had significantly reduced capacity for new bone formation when subcutaneously implanted into immunocompromised mice (Fig. 3e). It was reported that Tet1 and Tet2 may compensate for each other 24 . Since the osteopenia phenotype of Tet DKO mice and the impairment of Tet DKO BMMSCs were significantly severe than that of Tet1 −/− and Prx1 cre Tet2 fl/fl mice, we thus focus on study of Tet DKO mice. Control and Tet DKO BMMSCs showed similar adipogenic differentiation capacity under adipogenic induction, as indicated by Oil Red O staining and the expression of adipogenicrelated genes lipoprotein (LPL) and peroxisome proliferatoractivated receptor γ2 (PPARγ2) ( Supplementary Fig. 2d, e). These data indicate that Tet1 and Tet2 depletion impaired BMMSC function.
To further confirm the role of Tet1 and Tet2 in maintaining stem cell properties of BMMSCs in vitro, we used small interfering RNAs (siRNAs) to knockdown Tet1 and Tet2 expression in BMMSCs ( Supplementary Fig. 2f). Knockdown of Tet1 and Tet2 in BMMSCs induced a higher proliferation rate ( Supplementary Fig. 2g); reduced mineralized nodule formation ( Supplementary Fig. 2h); decreased expression levels of osteogenic-related genes Runx2, ALP, and OCN ( Supplementary  Fig. 2i); and reduced capacity for new bone formation when subcutaneously implanted into immunocompromised mice ( Supplementary Fig. 2j). The elevated proliferation rate and impairment of osteogenic differentiation were consistent with what was observed in Tet1-and Tet2-depleted BMMSCs. These data confirmed that Tet1 and Tet2 are required to maintain BMMSC stem cell properties.
miRNAs accumulate in Tet DKO. To explore the underlying molecular mechanisms, we performed RNA sequencing (RNAseq) using RNA from control and Tet1/Tet2 siRNA-treated BMMSCs and found that around 80% of altered genes (p < 0.05 and fold change >2) were decreased in Tet1/Tet2 siRNA-treated BMMSCs compared to control group ( Supplementary Fig. 3a).  (Fig. 4a, b). After 14-day osteogenic induction, the expression levels of Runx2, ALP, and OCN were decreased in Tet DKO BMMSCs compared to control group (Fig. 3d). We compared the expression levels of Runx2, ALP, and OCN on different days after osteogenic induction to investigate if ALP and OCN were potential direct targets or downstream molecules of Runx2. The result showed that the expression level of Runx2, but not ALP and OCN, significantly decreased without osteogenic induction and under 3 days of induction ( Supplementary Fig. 3c), suggesting that ALP and OCN may be the downstream target of Runx2. There were also less Runx2-positive cells in femur of Tet DKO mice compared to control ones ( Supplementary Fig. 3d). Next, to evaluate if Runx2 can be a direct target for Tet1 and Tet2, we analyzed the Runx2 promoter using Methprimer software and mice were used as Tet DKO mice in these experiments, and their littermates whose genetic status was Prx1 cre were used as controls. *p < 0.05, **p < 0.01, ***p < 0.001 (mean ± SD). Scale bars, 400 μm (a, b), 1 mm (c), and 25 μm (d). Results are from three independent experiments. p values were calculated using one-way ANOVA (a-c) and two-tailed Student's t test (d) found Runx2 promoter lacks CpG island ( Supplementary Fig. 3e). We next used chromatin immunoprecipitation (ChiP)-qPCR to analyze if Tet1/Tet2 could directly binding to the site where CpG was comparably rich in the Runx2 promoter, and the result showed that no binding site was detected ( Supplementary Fig. 3f). Thus, we search potential molecules that may connect Tet with Runx2 27, 28 . Since osteogenic differentiation can be governed by microRNAs (miRNAs) via post-transcriptional regulation 29-31 , we investigated whether the variation of miRNAs in Tet DKO BMMSCs contributed to the altered Runx2 expression. We used MicroCosm Targets software to identify 19 miRNAs that may target Runx2 gene expression (Supplementary Table 1). Interestingly, qPCR analysis showed that the levels of miR-297a-5p, miR-297b-5p, and miR-297c-5p in Tet DKO BMMSCs were significantly higher than in control BMMSCs (Fig. 4c). To further evaluate the functional role of miR-297a-5p, miR-297b-5p, and miR-297c-5p, we used mimics of these miRNAs to treat BMMSCs and inhibitors of these miRNAs to treat Tet DKO BMMSCs ( Supplementary Fig. 3g, h). Tet DKO BMMSCs treated with these miRNA inhibitors showed elevated mineralized nodule formation (Fig. 4d, e) and increased expression levels of Runx2, ALP, and OCN under 14 days of osteogenic induction ( Fig. 4f and Supplementary Fig. 3i). Moreover, under osteo-inductive conditions, BMMSCs treated with miR-297a-5p, miR-297b-5p, or miR-297c-5p mimics showed reduced mineralized nodule formation (Fig. 4g) and decreased expression levels of osteogenic-related   Supplementary  Fig. 3j). The proliferation rate of Tet DKO BMMSCs was decreased by the miR-297b-5p or miR-297c-5p inhibitor treatment ( Supplementary Fig. 3k), while the proliferation rate was significantly upregulated in BMMSCs treated with miR-297b-5p or miR-297c-5p mimics ( Supplementary Fig. 3l). Collectively, these data indicate that the accumulation of miR-297a-5p, miR-297b-5p, and miR-297c-5p in Tet DKO BMMSCs blocks Runx2 signaling and results in impaired osteogenic differentiation and elevated proliferation.
Tet1 and Tet2 control exosome/miRNA secretion in BMMSCs. Next, we examined how miR-297a-5p, miR-297b-5p, and miR-297c-5p accumulated in Tet DKO BMMSCs. We could not detect CpG island on the promoter of these three miRNAs, indicating  The expression of Runx2 in control and Tet DKO BMMSCs was analyzed by western blotting and qPCR. c The levels of miR-297a-5p, miR-297b-5p, or miR-297c-5p in control and Tet DKO BMMSCs assessed by qPCR. d Mineralized nodule formation under osteogenic inductive conditions of Tet DKO BMMSC after miR-297a-5p, miR-297b-5p, or miR-297c-5p inhibitor treatment. e, f The expression levels of osteogenic markers Runx2, ALP, and OCN in Tet DKO BMMSCs after miR-297a-5p, miR-297b-5p, or miR-297c-5p inhibitor treatment, as assessed by qPCR (e) and western blotting (f). g-i Mineralized nodule formation and the expression of Runx2, ALP, and OCN in BMMSCs after miR-297a-5p, miR-297b-5p, and miR-297c-5p mimic treatment as assessed by alizarin red staining (g), qPCR (h) and western blotting (i). *p < 0.05, **p < 0.01, ***p < 0.001 (mean ± SD). Results are from three independent experiments. p values were calculated using two-tailed Student's t test (b, c) and one-way ANOVA (d, e, g, h) that these may not be directly targeted by Tet1 and Tet2. MSCs can secrete a large number of exosomes 32 , which can transfer proteins, mRNAs, and miRNAs to mediate physiological and pathological processes 19,33,34 . We therefore sought to test whether exosome secretion is involved in miRNA accumulation in Tet DKO BMMSCs. We found that the amount of exosome-like extracellular vesicles in Tet DKO BMMSCs was significantly decreased when compared to the control BMMSCs, as quantified using a Bradford assay (Fig. 5a). We verified that the extracellular vesicles identified in this assay were indeed exosomes, as they expressed exosome markers CD9 and CD81 (Fig. 5b). Exosomes derived from Tet DKO BMMSCs expressed reduced levels of CD9 and CD81 compared to control BMMSCs (Fig. 5b). Furthermore, we used an EXOCEP exosome quantitation kit to show that the exosome secretion was decreased in Tet DKO BMMSCs (Fig. 5c) 35 . More intracellular exosomes accumulated in Tet DKO BMMSCs compared to control BMMSCs, as indicated by more CD9-positive and CD81-positive intracellular exosomes of Tet DKO BMMSCs analyzed by immunofluorescence staining and more vesicle accumulation analyzed by transmission electron microscopy (Fig. 5d, e). Previous studies indicated that P2rX7 is capable of controlling exosome release [36][37][38] . Immunocytofluorescence staining showed the co-localization of P2rX7 with MSC surface molecule CD146, along with reduced expression level of P2rX7 in Tet DKO BMMSCs (Fig. 5f). We showed that the expression level of P2rX7 was decreased in Tet DKO BMMSCs compared to the control group (Fig. 5g, h). In addition, we discovered decreased P2rX7 expression levels in Tet1 and Tet2 siRNA-treated BMMSCs ( Supplementary Fig. 4a). Previous studies indicated that Tet1 and 5-hmC prefer to co-localize at transcriptional start sites of CpG-rich promoters 27,28 . In order to investigate whether Tet1 and Tet2 directly regulate P2rX7, we used ChIP-qPCR analysis to show that Tet1 and Tet2 bind to the CpG island of P2rX7 promoter ( Supplementary Fig. 4b, Fig. 5i, j). We next examined whether depletion of Tet1 and Tet2 affected the enrichment of 5-hmC level at the P2rX7 promoter. Hydroxymethylated DNA immunoprecipitation (hMeDIP)-qPCR analysis revealed that Tet DKO BMMSCs showed a remarkably decreased 5-hmC level compared to control BMMSCs (Fig. 5k).
Consistently, Tet1 and Tet2 siRNA-treated BMMSCs also displayed reduced 5-hmC enrichment at the P2rX7 promoter ( Supplementary Fig. 4c). MeDIP-qPCR analysis revealed that Tet DKO BMMSCs showed increased 5-mC level compared to control BMMSCs ( Supplementary Fig. 4d). OxBS sequencing analysis also showed that Tet DKO BMMSCs displayed elevated methylation in the promoter of P2rx7 locus compared to control BMMSCs ( Supplementary Fig. 4e). These data suggest that P2rx7 is a direct target of Tet1 and Tet2 for DNA demethylation. Moreover, we detected the overall level of 5-hmC and 5-mC in BMMSCs by dot blot assay and showed that the level of 5-hmC was lower in Tet DKO BMMSCs than the control group, while the level of 5-mC was elevated in Tet DKO BMMSCs (Supplementary Fig. 4f, g). Furthermore, we overexpressed wildtype Tet1 and Tet2 plasmid and catalytic domain inactive Tet1 and Tet2 plasmid on Tet-DKO BMMSCs to analyze whether overexpression could rescue the decreased expression of P2rX7 and osteogenic differentiation. The results showed that wildtype Tet1 and Tet2 plasmid, but not catalytic domain inactive ones, overexpression rescued decreased expression of P2rX7 in Tet DKO BMMSCs ( Supplementary Fig. 4h). Wildtype, but not catalytic domain inactive, Tet1 and Tet2 plasmid overexpression also elevated the mineralized nodule formation and expression levels of Runx2, ALP, and OCN under osteogenic induction (Supplementary Fig. 4h-j).
We next investigated the effect of overexpressing P2rX7 in 8-10-week-old Tet DKO mice by systemic injection of adenoassociated P2rX7 overexpression virus (P2rX7 AAV). P2rX7 AAV or control AAV were injected once a week for up to 4 weeks ( Supplementary Fig. 5c). To analyze the efficiency of P2rX7 AAV treatment, we used immunostaining to show the increased P2rX7 expression in P2rX7 AAV-treated Tet DKO mouse bone marrow  Fig. 5d-f). P2rX7 AAV-treated Tet DKO BMMSCs were collected and found to have reduced intracellular levels of miR-297a-5p, miR-297b-5p, and miR-297c-5p (Fig. 6f), improved capacity to secrete exosomes into the culture supernatant (Fig. 6g), and elevated expression of CD9 and CD81 ( Fig. 6h and Supplementary, Fig. 5g). Micro-CT and histological analysis showed that P2rX7 AAV-treated Tet DKO mice displayed significantly increased BMD, Ct.Ar, Ct.Th, and distal femoral trabecular bone volume compared to control AAVtreated Tet DKO mice (Fig. 6i-l). Under osteo-inductive conditions, the osteogenic differentiation of BMMSCs derived from P2rX7 AAV-treated Tet DKO mice showed marked improvement when compared to control AAV-treated Tet DKO BMMSCs, as indicated by elevated mineralized nodule formation (Fig. 6m) and increased expression of the osteogenic genes Runx2, ALP, and OCN (Fig. 6n). The cell proliferation rate was decreased after the P2rX7 AAV treatment relative to Tet DKO BMMSCs, as assessed by BrdU-labeling assay (Supplementary, Fig. 5h). To verify the role of P2rX7 on bone formation, we used P2rX7 AAV to treat osteopenia (OVX) mice. Micro-CT and histological analysis showed that P2rX7-treated OVX mice had significantly increased BMD, Ct.Ar, Ct.Th, and the distal femoral trabecular bone volume when compared to OVX mice (Supplementary Fig. 6a, c). Moreover, P2rX7 AAV treatment could rescue decreased osteogenic differentiation capacity, but not the elevated proliferation rate in OVX BMMSCs ( Supplementary  Fig. 6d, f). Collectively, these data indicate that P2rX7 is capable of regulating intracellular levels of miR-297a-5p, miR-297b-5p, and miR-297c-5p through the control of exosome release. Elevating P2rX7 levels in Tet DKO mice rescued BMMSC function and osteopenia phenotype (Supplementary Fig. 7).

Discussion
The Tet family contributes to DNA demethylation and acts as a key regulator of epigenetic modulation 1, 2, 5 . In this study, we generated Tet1 −/− ;Prx1 cre Tet2 fl/fl (Tet DKO) conditional knockout mice to reveal the functional role of Tet1 and Tet2 in maintaining BMMSC properties and bone homeostasis. Previous studies showed that loss of Tet2 leads to increased replicating capacity in hematopoietic stem cell and myeloid transformation 39,40 . Inhibition of Tet1/2 suppresses proliferation of umbilical cord (UC)derived MSCs 41 and Nanog expression in embryonic stem cells (ES) 1 . Here we showed that Tet1 and Tet2 depletion elevated the proliferation of BMMSCs. Previous study reported that MSCs derived from adipose tissue, BM, and UC origin showed differential pattern of IGF2BP1 expression 41 , implying that the different roles of Tet1/2 on stem cell proliferation may be regulated by IGF2BP1 or other molecules. Tet DKO BMMSCs exhibited a significant impairment in osteogenic differentiation, which contribute to the osteopenia phenotype observed on Tet DKO mice. It was reported that newborns of Tet1 and Tet2 conventional knockout mice did not show developmental delay although a fraction of these double knockout mice were smaller in size 24 , suggesting that the osteopenia phenotype observed in Tet DKO mice may be attributed to the impairment of BMMSC osteogenic differentiation. Bone homeostasis is a tightly regulated process, balancing new bone formation by BMMSC-derived osteoblasts with bone resorption by osteoclasts. Signals that determine the recruitment, replication, apoptosis, and differentiation of cells of both lineages may increase osteopenia risk 42,43 . For instance, alteration of histone H3K9 acetyltransferase PCAF (p300/CBPassociated factor) 44 , mammalian target of rapamycin (mTOR) signaling 45 and hydrogen sulfide levels 46 may lead to osteopenia by impairing osteogenic differentiation of BMMSCs. Our previous studies demonstrated that the depletion of Tet1 and Tet2 leads to impairment of regulatory T cell differentiation and immune system disorders 47 . Mutation of Tet2 in HSCs has been reported to impair their function and leads to diverse myeloid malignancies 48 .
Except the direct effect of Tet on BMMSCs, whether immune disorder or abnormal of HSCs caused by Tet mutation contributed to bone disorders and BMMSCs impairment need to be further investigated 49 .
In this study, we proposed a mechanism by which Tet1 and Tet2 may function as upstream epigenetic regulators of P2rX7controlled exosome release. Exosomes play an important role in intercellular communication because they are able to transfer proteins, mRNAs, and miRNAs 50 , contributing to a variety of physiologic processes, including cell homeostasis, differentiation, immune responses, and neuronal signaling 51,52 . Aberrant exosome secretion may be associated with various human diseases, such as Parkinson's disease, multiple sclerosis, and Alzheimer's disease [52][53][54] . It is known that multiple microenvironment factors including inflammatory cytokines, hypoxia, and infectious agents are capable of promoting MSC secretion through the activation of an exocytotic pathway 55 . However, whether MSC exosome secretion is modulated by epigenetic modification has remained unclear. In this study, we showed that Tet1 and Tet2 directly bind to the CpG island of P2rX7 promoter to directly regulate its demethylation activity [36][37][38] . P2rX7 is an ionotropic receptor that is controlled by adenosine triphosphate. Activation of P2rX7 triggers a remarkably diverse range of membrane trafficking responses in leukocytes and epithelial cells 56 . The significance of P2rX7 in regulating bone development and homeostasis has been demonstrated by previous studies. Bone deficiency was observed in P2rX7 knockout mice 57,58 . The combination of this evidence prompted us to further examine the functional role of P2rX7 in regulating MSC secretion and differentiation. In this study, we demonstrated that the activation of P2rX7 increased the amount of exosome secretion from Tet DKO BMMSCs. P2rX7 deficiency led to the impairment of osteogenic differentiation of BMMSCs. Collectively, our studies reveal that P2rX7 is a target of Tet demethylation, and that Tet1 and Tet2 can directly modulate P2rX7, thereby controlling BMMSC exosome release to maintain bone and BMMSC homeostasis. Tet1 and Tet2 deficiency also reduced the overall level of 5-hmC in BMMSCs, downregulating expression of other potential molecular targets. Further investigation is required to identify and characterise other downstream target molecules of Tet1 and Tet2 in BMMSCs. Fig. 6 P2rX7 activation rescues impaired Tet DKO BMMSCs and osteopenia phenotype in Tet DKO mice. a The levels of miR-297a-5p, miR-297b-5p, and miR-297c-5p in control, Tet DKO BMMSCs, and P2rX7 CRISPR activation plasmid-treated Tet DKO BMMSCs. b Bradford assay showed the exosome secretion in control, Tet DKO BMMSCs, and P2rX7 CRISPR activation plasmid-treated Tet DKO BMMSCs. c Western blotting showed the expression of CD9 and CD81 in control, Tet DKO BMMSCs, and P2rX7 CRISPR activation plasmid-treated Tet DKO BMMSCs. d, e Mineralized nodule formation and the expression of Runx2, ALP, and OCN in control, Tet DKO BMMSCs, and P2rX7 CRISPR activation plasmid-treated Tet DKO BMMSCs, as assessed by alizarin red staining (d) and western blotting (e). f The expression of miR-297a-5p, miR-297b-5p, and miR-297c-5p in control, Tet DKO BMMSCs, and adenoassociated P2rX7 overexpression virus (P2rX7 AAV)-treated Tet DKO BMMSCs, as assessed by qPCR. g Bradford assay showed the exosome secretion level in control, Tet DKO BMMSCs, and P2rX7 AAV-treated BMMSCs. h Western blotting showed the expression of CD9 and CD81 in exosomes derived from control, Tet DKO BMMSCs, and P2rX7 AAV-treated BMMSCs. i Bone mineral density (BMD) of trabecular bone (TB) area in distal femurs of control, Tet DKO, and P2rX7 AAV-treated-mice, as assessed by micro-CT. j The cortical bone area (Ct.Ar) and cortical thickness (Ct.Th) of the femurs of control, Tet DKO, and P2rX7 AAV-treated-mice analyzed by micro-CT. k, l H&E staining showed the TB volume (yellow-circled area) of control, Tet DKO, and P2rX7 AAV-treated-mice. m The capacity of BMMSCs isolated from control, Tet DKO, and P2rX7 AAV-treated-mice to form mineralized nodules as assessed by alizarin red staining. n Western blotting analysis showed the expression levels of the osteogenic markers Runx2, ALP, and OCN in control, Tet DKO, and P2rX7 AAV-treated Tet DKO BMMSCs. The 8-10-week-old Tet1 −/− Prx1 cre Tet2 fl/fl mice were used as Tet DKO mice in our experiments, and their littermates whose genetic status was Prx1 cre were used as controls. *p < 0.05, **p < 0.01, ***p < 0.001 (mean ± SD). Scale bars, 400 μm (i, j), 1 mm (k). Results are from three independent experiments. p values were calculated using one-way ANOVA The miRNAs are small non-coding RNAs with a length of 18-23 nucleotides that function as post-transcriptional regulators of gene expression 59 . Herein, we computationally identified 19 miRNAs that may target Runx2. Tet DKO BMMSCs showed a different miRNA expression profile than that of control BMMSCs. Previous studies showed that miR-29b, miR-26a, and miR-767 can target Tet1 to inhibit its expression, leading to a change in 5-hmC levels [60][61][62] . Conversely, the Tet family can also directly modulate miR-200 by demethylating the CpG islands of the miR-200 promoter 63 . In this study, as there is no CpG island on the promoter of miR-297a-5p, miR-297b-5p, and miR-297c-5p, we reveal a mechanism of interaction between the Tet family and miRNAs and show that Tet1 and Tet2 depletion leads to P2rX7 promoter hypermethylation and reduced exosome secretion. This results in intracellular accumulation of miR-297a-5p, miR-297b-5p, and miR-297c-5p in Tet DKO BMMSCs. In turn, the intracellular accumulation of miR-297a-5p, miR-297b-5p, and miR-297c-5p decreases the osteogenic differentiation of Tet DKO BMMSCs by inhibiting Runx2 signaling. Interestingly, we also found that the proliferation rate of BMMSCs is elevated after miR-297b-5p and miR-297c-5p mimic treatment, which is consistent with our findings on Tet DKO BMMSCs. Previous studies indicated that the biological role of Runx2 is not restricted to activating osteoblast lineage-specific genes; it also contributes to negative control of cell proliferation 64 . Our finding may help to explain why the Runx2-deficient in Tet DKO BMMSCs and the intracellular accumulation of Runx2-targeting miRNAs results in the impairment of osteogenic differentiation and elevated proliferation rate. Overexpression of P2rX7 rescues the exosome release and accumulation status of miR-297a-5p, miR-297b-5p, and miR-297c-5p in Tet DKO BMMSCs. In vivo overexpression of P2rX7 rescues the osteopenia phenotype and BMMSC function in Tet DKO mice, suggesting that P2rX7 may be a potential therapeutic target for treating Tet-associated skeletal disorders. Based on the previous studies, deletion of Tet proteins might lead to other pathogenic process other than osteopenia 65 . Thus, the effect of the exosome secretion impairment caused by Tet depletion might have a broad effect on miRNA release, which may not only specify to the miRNAs targeting Runx2. Global gene analysis showed Runx2 cluster genes were altered after Tet1 and Tet2 knockdown in BMMSCs, implying that the Tet/P2rX7/ Runx2 cascade may be one of critical regulating mechanism to maintain bone homeostasis and BMMSC function. As 5-hmC may be involved in establishing and maintaining chromatin structure for both actively transcribed genes and PcG-repressed regulators, it also affects both transcriptional activation and repression roles in a context-dependent manner. The detailed mechanisms of how 5-hmC regulates other genes altered by Tet in BMMSCs may need further investigation 66,67 In conclusion, depletion of Tet1 and Tet2 results in BMMSC impairment and an osteopenia phenotype. An increase in exosome secretion and reduction of intracellular accumulation of miR-297a-5p, miR-297b-5p, and miR-297c-5p, achieved by restoring the level of P2rX7 in Tet DKO BMMSCs, can rescue BMMSC function and bone homeostasis.
Isolation and culture of human BMMSCs. We purchased human bone marrow from AllCells LLC (Alameda, CA, USA), which were aspirated from healthy human adult volunteers (20-35 years of age). The dissociated cell suspension was filtered through a 70 μm cell strainer, then cultured with α-MEM containing 10% FBS, 10 mM L-ascorbic acid phosphate, 100 U ml −1 penicillin, 100 μg ml −1 streptomycin, and 2 mM L-glutamine, incubated at 37°C under 5% CO 2 condition. After 48 h, the non-adherent cells were removed. The adherent cells were passaged with 0.05% trypsin containing 1 mM EDTA after being cultured for 14 days.
Western blotting. Total protein was lysed in M-PER mammalian protein extraction reagent (Thermo, Rockford, IL, USA). After the proteins (20 μg) were loaded and separated to 4-12% NuPAGE gel (Invitrogen), they were transferred to 0.2 μm nitrocellulose membranes (Millipore, Billerica, MA, USA). Then, 0.1% Tween-20 and 5% non-fat dry milk were provided to block the membranes for 1 h, followed by overnight incubation with primary antibodies (1:1000) diluted in blocking solution. The membranes were washed and incubated for 1 h in horseradish peroxidase (HRP)-conjugated secondary antibody (Santa Cruz Biotechnology; 1:10,000). SuperSignal West Pico Chemiluminescent Substrate (Thermo) and BioMax film (Kodak) were used to detect the immunoreactive proteins. All full blots presented in the main figures are displayed in the Supplementary Information.
Real-time PCR. We used miRNeasy Mini Kit (Qiagen, Valencia, CA, USA) to isolate total RNA from the cultured cells according to the manufacturer's instructions. For qPCR of mRNA, we used SuperScript III Reverse Transcriptase (RT) kit (Invitrogen) to prepare the complementary DNA (cDNA). qPCR was performed using SYBR Green Supermix (Bio-Rad, Hercules, CA, USA) and genespecific primers. The level of mRNA expression for each gene was normalized to glyceraldehyde 3-phosphate dehydrogenase (GAPDH). For qPCR of miRNA, miScript II RT Kit (Qiagen) was used to synthesize the cDNA. miScript SYBR Green PCR Kit (Qiagen) was used to perform real-time PCR. We used RUN6 as an endogenous control for BMMSCs. A CFX96 Real-Time PCR System (Bio-Rad) was used for qPCR analysis.
Pre Kit to prepare a sequencing library from 1 μg of total RNA, and 2 × 100 pairedend sequencing in fat run mode was performed using the HiSeq 2500 and Illumina TruSeq SBS-Kit v2 (200 cycles).
Bioinformatics analysis. Illunina's bcl bcl2fastq v1.8.4 software (Illumina,San Diego, CA) was used to convert the resulting base calling (.bcl) to FASTQ files. Mapping RNA-seq reads on the mouse genome (Mus musculus GRCm38) after trimming the adaptors, transcript assembly, and abundance estimation were performed using DNASTAR Lasergene v15.0 (DNASTAR, Madison, WI) and reported using FPKM (fragments per kilobase of exon per pillion fragments mapped). The Gene Pattern and WebGestalt were used for functional analysis.
In vitro osteogenic differentiation. BMMSCs were cultured in osteogenic medium containing 100 μM L-ascorbic acid 2-phosphate (Wako), 2 mM βglycerophosphate (Sigma-Aldrich), and 10 nM dexamethasone (Sigma-Aldrich) in the growth medium. After 14 days of osteogenic induction, western blotting was performed to detect the expression levels of Runx2, ALP, and OCN. After 4 weeks of induction, mineralized nodule formation was detected by staining with 1% Alizarin Red S (Sigma-Aldrich) for the cultures. The stained positive areas were quantified using the National Institutes of Health (NIH) ImageJ software and shown as a percentage of the total area.
Cell proliferation assay. Mouse BMMSCs (10 × 10 3 /well) were cultured on 2-well chamber slides (Nunc, Rochester, NY, USA) for 2-3 days. After being incubated with BrdU solution (1:100) (Invitrogen) for 20 h, the cultures were stained with BrdU antibody (1:200, Invitrogen) at 4°C overnight, then treated with Alexafluoro 568 conjugated secondary antibody for 1 h at room temperature. Finally, Vectashield mounting medium containing DAPI (Vector Laboratories, Burlingame, CA, USA) was used to mounted slides. BrdU-positive cells were counted and indicated as a percentage of the total cell number, and 10 images were analyzed per subject. Three independent samples were repeated for each experimental group.
Isolation and characterization of exosomes. Exosomes were isolated from BMMSCs by ultracentrifuge. Exosome-depleted medium (complete medium depleted of FBS-derived exosomes by overnight centrifugation at 100,000 × g) was used to culture cells for 48 h. We used differential centrifugation to extract exosomes from culture supernatants of 12 × 10 6 BMMSCs 68 , at 300 × g for 10 min, 3000 × g for 10 min, 20,000 × g for 30 min, and 120,000 × g for 70 min. For analysis of exosome protein secretion, Bradford protein assay (Bio-Rad Laboratories) was used to measure the amount of total protein in purified exosomes from 12×10 6 BMMSCs. Total exosome protein was normalized to 1×10 6 cells to show the amount as micrograms per 1×10 6 cells. EXOCEP exosome quantitation kit (System Biosciences Inc.) was used to quantitate the number of vesicles, following the manufacturer's instruction 35 .
Methylated and hydroxymethylated DNA immunoprecipitation (hMeDIP). Immunoprecipitation of 5-mC and 5-hmC was performed using Active motif MeDIP kit and hMeDIP Kit with minor modifications. We used Branson sonicator to sonicate DNA into short fragments (100 to 1000 base pairs (bp)) for 20 min with 15 s on, 15 s off cycles at low power. We heat-denatured sonicated DNA at 95°C for 10 min. Sonicated DNA (1 μg) was immunoprecipitated with 4 μl of mouse anti-5-methylcytosine or anti-5-hydroxymethylcytosine monoclonal antibody (Active Motif, 1 μg μl −1 ). After overnight incubation at 4°C, we added magnetic beads to the DNA-antibody mixture and samples were incubated for 2 h at 4°C. Isolation of immunoprecipitated DNA was performed according to the kit instructions. SYBR® Green Supermix (Bio-Rad) was used to perform qPCR on a Bio-Rad CFX96 Real Time system, as indicated by the manufacturer's protocol.
The percentage of enrichment was calculated relative to the amount of DNA used in the IP reaction.
Chromatin immunoprecipitation-quantitative PCR (ChIP-qPCR). 5×10 6 cells were cross-linked and used for each immunoprecipitation, and chromatin was sheared with a Branson sonicator. Millipore ChIP kit was used to perform immunoprecipitation according to the manufacturer's protocol. To precipitate DNA-protein complexes, Tet1 and Tet2 antibodies were used and IgG was used as an isotype control. Percentage input was determined by removing an aliquot of sheared chromatin prior to immunoprecipitation and comparing amplification of this DNA to amplification of the precipitated chromatin. Putative Tet1 and Tet2 binding sites in the P2rX7 promoter were investigated using MethPrimer software to predict the CpG island 69 .
Transmission electron microscopy. For transmission electron microscopy (TEM), control and Tet DKO BMMSCs were collected by centrifugation and fixed with 2.5% glutaraldehyde overnight and then fixed with 1% osmium tetraoxide for 2 h, dehydrated in a graded series of ethanol concentrations, and embedded in SPIPON812 resin and polymerized. The block was sectioned by microtome (Leica EM UC6). The ultrathin sections approximately 70 nm, mounted on copper grids, uranyl acetate, and lead citrate were used to stain, and examined, and photographed with a FEI Tecnai spirit TEM (FEI Tecnai Spirit 120 kv).
Histology. To assess trabecular bone and bone marrow areas, femurs were fixed in 4% PFA, decalcified with 5% EDTA (pH 7.4), and embedded in paraffin. For histological analysis, sections were deparaffinized and stained with hematoxylin and eosin (H&E). The trabecular bone area percentage was then calculated using NIH ImageJ software.
Micro-CT analysis. Femurs were fixed in 4% PFA, and analyzed using a μCT scanner (micro-CT 35, Scanco Medical AG, Brüttisellen, Switzerland). Briefly, the distal end of the femur corresponding to a 0 to 4.1 mm region above the growth plate was scanned at 6 μm isotropic voxel size. For the trabecular bone analysis, the images of the secondary spongiosa regions 0.6 to 1.8 mm above the highest point of the growth plate were contoured. BMD and BV/TV were calculated using standard 3D microstructural analysis. After smoothing the image with a 3D gaussian lowpass filter (Sigma 0.7 voxels), the CT images were segmented into bone and marrow regions by applying a visually chosen for the cortical bone assessment. The outer and inner perimeter of the cortical midshaft was determined by a 3D triangulation of the bone surface of the 20 slices, and the formula Ct.Th = 1/2 × BS/ BV was used to calculate cortical thickness. Ct.Ar was calculated by the formula Ct. Ar = thickness of ring × length of middle line = thickness × (outer circumference + inner circumference)/2.
P2rX7 AAV treatment in Tet DKO mice. P2rX7 overexpression adeno-associated virus (P2rX7 AAV, Applied Biological Materials Inc., Richmond, BC, Canada) was used to overexpress P2rX7 in vivo. P2rX7 AAV (1×10 9 GC) or AAV control virus was injected into the Tet DKO mice once a week for up to 4 weeks. After the therapy, samples were harvested immediately for histological assessments and BMMSC isolation.
In vivo BMMSC-mediated bone formation. We subsequently implanted 4.0×10 6 BMMSCs mixed with 40 mg HA/TCP ceramic powder (Zimmer Inc., Warsaw, IN, USA) under the dorsal surface of 8-week-old immunocompromised mice. The transplants were harvested at 8 weeks post implantation, fixed in 4% PFA, and decalcified with 5% EDTA, then embedded in paraffin. The sections were stained with H&E. For quantification of in vivo new bone regeneration, we calculated the area of bone formation with 10 representative images from different regions of the BMMSC implants using ImageJ software (National Institutes of Health, Bethesda, MD). Each experimental group was repeated with three independent implants. Dot blot assay. DNA was diluted to 100 ng ml −1 and a dilution series was performed; 1 μl was spotted on a 0.45 μm pore size positively charged nylon membrane and wait for drying in the air, then the membrane was baked in 80°C vacuum oven for 2 h. The membrane was blocked using 5% milk/0.1%Tween-20/ Tris-buffered saline for 1 h, and being incubated with 5-hmC (1:1000) or 5-mC (1:1000) antibodies overnight at 4°C. Then, the membrane were washed and incubated for 1 h in HRP-conjugated secondary antibody (Santa Cruz Biotechnology; 1:10,000). Finally, SuperSignal West Pico Chemiluminescent Substrate (Thermo) and BioMax film (Kodak) were used to do the detection.
Methylation-specific PCR and OxBS sequencing. To assess P2rX7 methylation status, optimized oxidative bisulfite sequencing (OxBS sequencing) was used. A highly selective chemical oxidizes 5-hmC to 5-formylcytosine (5fC) and then bisulfite treatment deformylated and deaminated 5fC to uracil, and appears as thymine (T) in sequencing analysis. The 5-mC is not deaminated and appears as a cytosine (C) in sequencing analysis. Oxidative bisulfite conversion, which is able to produce an accurate readout distinguishing 5-mC from 5-hmC, was performed using a TrueMethyl™ kit (Cambridge Epigenetix, UK), according to the manufacturer's instructions 70 . For methylation-specific PCR, pretreated DNA was amplified with methylation-specific primers for P2rX7 as listed in the supplementary table 2 and were sequencing as we previously reported 47 .
Statistics and reproducibility. For animal study, we used at least five mice and experiments were performed three times, unless otherwise stated in the figure legends. No specific statistical test was used to predetermine the sample size. Comparisons between two groups were analyzed using independent unpaired twotailed Student's t-tests, and comparisons between more than two groups were analyzed using one-way analysis of variance with the Bonferroni correction if the data did not met the normality distribution assumption. The p values of less than 0.05 were considered statistically significant. In vitro and in vivo experiments were studied in a non-blinded fashion, no method of randomization was used, and typically no sample was excluded for the analysis.
Data availability. The data that support the findings in this study are available from the corresponding author upon reasonable request. RNA-sequencing data have been deposited in the Gene Expression Omnibus (GEO) database by GSE108872.