Specific functions of TET1 and TET2 in regulating mesenchymal cell lineage determination
The 5 hydroxymethylation (5hmC) mark and TET DNA dioxygenases play a pivotal role in embryonic stem cell differentiation and animal development. However, very little is known about TET enzymes in lineage determination of human bone marrow-derived mesenchymal stem/stromal cells (BMSC). We examined the function of all three TET DNA dioxygenases, responsible for DNA hydroxymethylation, in human BMSC cell osteogenic and adipogenic differentiation.
We used siRNA knockdown and retroviral mediated enforced expression of TET molecules and discovered TET1 to be a repressor of both osteogenesis and adipogenesis. TET1 was found to recruit the co-repressor proteins, SIN3A and the histone lysine methyltransferase, EZH2 to osteogenic genes. Conversely, TET2 was found to be a promoter of both osteogenesis and adipogenesis. The data showed that TET2 was directly responsible for 5hmC levels on osteogenic and adipogenic lineage-associated genes, whereas TET1 also played a role in this process. Interestingly, TET3 showed no functional effect in BMSC osteo-/adipogenic differentiation. Finally, in a mouse model of ovariectomy-induced osteoporosis, the numbers of clonogenic BMSC were dramatically diminished corresponding to lower trabecular bone volume and reduced levels of TET1, TET2 and 5hmC.
The present study has discovered an epigenetic mechanism mediated through changes in DNA hydroxymethylation status regulating the activation of key genes involved in the lineage determination of skeletal stem cells, which may have implications in BMSC function during normal bone regulation. Targeting TET molecules or their downstream targets may offer new therapeutic strategies to help prevent bone loss and repair following trauma or disease.
KeywordsMesenchymal stem cells Osteogenesis Epigenetics Hydroxymethylation TET Osteoporosis
bone marrow stromal/stem cells
ten-eleven translocation proteins
embryonic stem cells
hematopoietic stem cells
- H3K4, H3K27
histone 3 lysine 4 or 27
transcription start site
DNA methylation and hydroxymethylation have been shown to regulate stem cell maintenance and differentiation via the activation or suppression of pluripotency and lineage-specific gene expression . Although DNA methylation was one of the first epigenetic modifications studied, 5 hydroxymethylation has only recently been identified as an intermediate step in the process of DNA demethylation. Rao and colleagues  were the first to describe that ten-eleven translocation-1 (TET1) proteins can catalyse conversion of 5 methylcytosine (5mC) into 5 hydroxymethylcytosine (5hmC), as an intermediate step to the removal of the methylation mark on DNA, where 5hmC opposes the repressive effects of 5mC and can prevent methylation of DNA . The TET protein family are comprised of 2-oxoglutarate and iron-dependent dioxygenases, where TET1 is mainly expressed by ESC, while TET2 and TET3 are more ubiquitously expressed across different cell lineages [2, 3]. ESC studies have shown that a key event in differentiation is the repression of pluripotency factors such as Oct4 and Nanog partly by DNA methylation . TET1-deficient ESC display impaired self-renewal, diminished proliferation, with reduced Nanog expression coinciding with increased DNA methylation of its promoter . Hematopoietic stem cells (HSC) derived from conditional TET2 knockout mice display an enhanced ability to reconstitute haematopoiesis in vivo but prevent HSC from undergoing differentiation, providing evidence that TET2 restricts aberrant self-renewal and expansion of hematopoietic stem cells (HSC) . In support of this notion, one of the most common mutations in myeloid malignancies is present in TET2, where TET2-deficient HSC show a skewed differentiation towards myeloid lineages at the expense of lymphoid lineages [7, 8]. Recent studies of global TET1 and conditional TET2 knockout mice in the mesenchyme lineage showed impaired bone-forming capacity in BMSC . In other systems, conditional knockout of TET2 in smooth muscle demonstrated that TET2 is essential for smooth muscle cell differentiation and that loss of TET2 expression results in de-differentiation . Other studies reported that TET1 and TET2 mediate Foxp3 demethylation to drive regulatory T cell differentiation . A combined loss of TET1 and TET2 results in depleted 5hmC levels , with most mice exhibiting midgestation defects and perinatal lethality. Triple knockouts of TET1/TET2/TET3 display a complete loss of 5hmC and increase in 5mC . Differentiation of embryoid bodies is grossly impaired with a lack of mesoderm and endoderm markers. Global knockdown of all three TET molecules identified 1072 downregulated genes and 729 upregulated genes, illustrating that TET proteins can activate or repress transcription . Furthermore, reprogramming of fibroblasts into iPSC results in increased levels of TET1 and TET2 and a decrease in TET3 .
Bone marrow mesenchymal stem/stromal cells (BMSC) exhibit the capacity for multi-lineage differentiation and self-renewal [15, 16, 17, 18, 19]. BMSC maintenance and cell fate determination have previously been shown to be mediated, in part, by the activity of the histone 3 lysine 4 (H3K4) methyltransferase, MLL1/2 , the H3K27 methyltransferase, Ezh2 , and associated demethylases, KDM6A  and KDM6B [22, 23, 24], via the regulation of key lineage-associated transcription factors [25, 26, 27]. In an effort to further identify epigenetic enzymes involved in BMSC lineage determination and growth, we examined the function of TET DNA hydroxymethylases in human BMSC lineage determination. Previous studies have shown that TET1 can influence recruitment of Ezh2 to promoters , and plays a role in stem cell self renewal. In this study, we have identified a function role for both TET1 and TET2 in regulating human BMSC differentiation, by acting on genes involved in lineage determination. Moreover, we discovered that the expression of TET1 and TET2 is grossly deregulated in osteoporosis leading to deregulated 5hmC levels on promoters of genes controlling stem cell renewal and lineage determination in osteoporosis.
Materials and methods
Cell culture and antibodies
Human BMSC were derived from bone marrow aspirates from posterior iliac crest of normal adult volunteers after obtaining informed consent according to procedures approved by the Human Ethics Committee of the Royal Adelaide Hospital, South Australia (protocol# 940911a). Immunoselected STRO-1+ BMSC were cultured in regular growth medium as previously described .
In vitro differentiation assays
Human BMSC were cultured in either normal growth conditions, osteogenic inductive conditions (control growth media + 10−7M dexamethasone, 10 mM HEPES buffer and 2.6 mM potassium phosphate) or adipogenic inductive conditions (control growth media + 0.5 mM, methylisobutylmethylxanthine, 0.5 μM hydrocortisone and 60 μM indomethacin) for up to 28 days as previously described . Mineralised bone matrix formation was identified with Alizarin red (Sigma Aldrich Inc.) staining . Extracellular calcium was measured in triplicate samples and normalised to DNA content per well as previously described . Lipid formation was assessed and quantitation of lipid was performed by Nile red (Sigma Aldrich Inc, St Louis, MO) fluorescence staining, normalised to DAPI (Invitrogen/Life Technologies Australia, Mulgrave, VIC, AUS) stained nuclei per field of view in triplicate wells as previously described [29, 30].
Lentiviral transductions were performed by transfecting 5 μg of Lv105 (cat:Ex-Neg-Lv105; Geneocoepia, Rockville, MD), Lv105-TET2 , Lv231 (Ex-Neg-Lv231), Lv231-TET1 (Ex-E2856-Lv231) into HEK293 T cells together with 5 μg of packaging vector psPAX2 and VsVG using lipofectamine 2000 (Life Technologies, Carlsbad, CA). After 48 h, 5 × 104 BMSC were infected with the supernatant for the HEK293 T cells three times every 12 h in the presence of 4 mg/ml polybrene. Transduced BMSC were selected with 1 μg/ml puromycin for 7 days and then maintained in 200 ng/ml puromycin.
BMSC were transfected with 12 pmol siRNA targeting either TET1 (s37193; Ambion, Foster City, California), TET2 (cat: s29443), TET3 (s47238) or scramble control siRNA (AM4613), with RNAiMax lipofectamine (56532) in 100 μl media without foetal calf serum for 20 min. After 72 h, the media were replaced with either control growth media, osteogenic or adipogenic inductive media [29, 31, 32].
RNA extractions, cDNA synthesis and real-time PCR
Total RNA from approximately 1.5 × 105 human BMSC was isolated using Trizol (Invitrogen, Grand Island, NY) according to manufacturers’ instructions. cDNA and real-time PCR was performed as previously described in triplicate . Changes in gene expression were calculated relative to β-actin using the 2−dCT method.
Oligonucleotides and primers
β-ACTIN, F: 5′gatcattgctcctcctgagc3′, R: 5′gtcatagtccgcctagaagcat3′. PPARγ2, F: 5′ctcctattgacccagaaagc3′, R: 5′tcaaaggagtgggagtggtc3′; ADIPSIN, F: 5′gacaccatcgaccacgac3′, R: 5′ccacgtcgcagagagttc3′. C/EBPα, F: 5′gggcaaggccaagaagtc3′; R: 5′ttgtcactggtcagctccag3′. LEPTIN, F: 5′gaaccctgtgcggattcttgt3′; R: 5′tccatcttggataaggtcaggat3′. RUNX2, F: 5′gtggacgaggcaagagtttca-3′, R: 5′catcaagcttctgtctgtgcc3′. OSTEOPONTIN, F: 5′acatccagtaccctgatgctacag3′, R: 5′gtgggtttcagcactctggt3′. OSTEOCALCIN, F: 5′atgagagccctcacactcctcg3′, R: 5′gtcagccaactcgtcacagtcc3′. TET1, F: 5′gcagcgtacaggccaccact3′; R: 5′agccggtcggccattggaag3′. TET2, F: 5′ttcgcagaagcagcagtgaagag3′; R: 5′agccagagacagcgggattcctt3′. TET3, F: 5′gacgagaacatcggcggcgt3′; R: 5′gtggcagcggttgggcttct3′.
RUNX2 TSS F: 5′aggccttaccacaagccttt3′, R: 5′agaaagtttgcaccgcactt3′; RUNX2 Exon F: 5′gcaaaatgagcgacgtgag3′, R: 5′acaggaagttggggctgtc3′; RUNX2 Intron F: 5′cgattcaagagctgctcaca3′, R: 5′cctgttttgccgtcaatttt3′; RUNX2 UTR F: 5′tttgcactgggtcatgtgtt3′, R: 5′tggctgcattgaaaagactg3′; BMP2 TSS F: 5′atgcggagcacctactgc3′, R: 5′ccgcatcactctgccttact3′; BMP2 Exon F: 5′cttctagcgttgctgcttcc3′, R: 5′agtgcctgcgatacaggtct3′; BMP2 Intron F: 5′aaacaggccaaacacagtcc3′, R: 5′agccagggtctcagaacaga3′; BMP2 UTR F: 5′tgcaggaaagtgaatgatgg3′, R: 5′tgcataattttgctgcgtgt3′; PPARγ TSS F: 5′agcaaacgacaccaggtagc3′, R: 5′ggcacccgtactctgacct3′; PPARγ Exon F: 5′gctgtgcaggagatcacaga3′, R: 5′gggctccataaagtcaccaa3′; PPARγ Intron F: 5′tcctctccagcgtctgtttt3′, R: 5′cccatctgacaaagggctaa3′; PPARγ UTR F: 5′cacagatccaccgtttcctt3′, R: 5′acacggtgaaaccctgtctc3′; ADIPSIN TSS F: 5′cctccaccctcataaaagca3′, R: 5′gcgttcagagccttccatta3′; ADIPSIN Exon F: 5′ctggggcatagtcaaccac3′, R: 5′atcaagcgctcggtgatg3′; ADIPSIN Intron F: 5′ggaagagaaggggtcctgag3′, R: 5′tccaagccctttccagtatg3′; ADIPSIN UTR F: 5′tcaggagttcgagatcagca3′, R: 5′ctacaagcacccacctccat3′.
Chromatin immunoprecipitation (ChIP)
Cultured human BMSC were used for ChIP using the Magna ChIP kit (Millipore Corporation, Billerica, MA) according to manufacturers’ instructions using 2 μg of TET1 (MABE 1034; Millipore Corporation) or 1 μg of TET2 (MABE 462) antibody. Purified DNA was then analysed by qPCR as previously described .
Human BMSC were cultured in normal growth media or in osteogenic inductive media for 1 week. Genomic DNA was prepared using the Qiagen genomic DNA isolation kit and then sonicated to generate 200–500 bp fragments. DNA was denatured in 10 mM Tris–HCl, 1 mM EDTA, pH 7.5 (TE buffer), for 10 min at 98 °C and then chilled on ice. 1% of denatured DNA was retained as the input control. Denatured 5hmC DNA was immunoprecipitated for 2 h at 4 °C with 2 μg of monoclonal antibody against 5hmC (Ab106918; Abcam, Cambridge, UK) in a final volume of 500 μl IP buffer (10 mM sodium phosphate (pH 7.0), 140 mM NaCl, 0.05% Triton X-100). The immunoprecipitation was incubated with magnetic IgG beads for 2 h at 4 °C and washed with IP buffer and then in high salt buffer (10 mM sodium phosphate (pH 7.0), 400 mM NaCl, 0.05% Triton X-100). Immunoprecipitated bead–DNA complex was treated with proteinase K for 3 h at 50 °C in elution buffer and hydroxymethylated DNA purified by using the Qiagen PCR clean-up kit (Qiaquick). DNA was quantitated using qPCR.
Global 5hmc and 5mc analysis
The levels of 5mc and 5hmc were assayed by ELISA using either the Quest 5-hmc DNA ELISA Kit (Cat# D5425, Zymo Research) or 5-mc DNA ELISA Kit (cat# D5325, Zymo Research), respectively. Briefly, 96 well plates were coated with either anti-5-hydroxymethylcytosine polyclonal antibody or anti-5-methylcytosine polyclonal antibody (Zymo Research) (1 ng/μl) in coating buffer. BMSC were cultured for 2 weeks under normal, osteogenic or adipogenic inductive conditions. 100 ng of denatured DNA and standard DNA with known 5hmc or 5mc concentrations was added to the wells and incubated for 1 h. After washing, the cells were incubated with anti-DNA HRP conjugated antibody for 30 min. Wells were washed and HRP developer was added. The plates were read at 415 nm.
Ovariectomised mouse model
Animal experiments and analyses were approved by the SA Pathology (23/11) and University of Adelaide (M-2013-144) Animal Ethics Committees. Twelve-week-old C57BL/6 female mice were anesthetised, and then their ovaries were either sited (sham surgery) or ovariectomised (OVX) as previously described . The dorsal skin was then sutured and the mice were allowed to recover under normal housing conditions, with food and water provided ad libitum.
Three-dimensional micro-computed tomography (µCT) was performed as previously outlined  using Micro-CT (Skyscan 1076 X-ray Microtomography SkyScan, Bruker MicroCT, Kontich, Belgium, http://bruker-microct.com). Briefly, femora isolated from sham and OVX mice were scanned at 9 µm resolution, Al 0.5 mm filter, excitation 5890 ms, voltage 48 kV, current 110 mA, rotation step 0.6 and two-frame averaging. NRecon software was used to reconstruct the femora.
The incidence of murine clonogenic osteogenic progenitor cells was assessed using single-cell suspensions derived from bone marrow, and collagenase-treated crushed femoral bone chips were plated in colony-forming unit-fibroblast assays (CFU-F) which were performed as described previously .
Immunohistological staining was performed on sections (5 μm) of paraffin-embedded femora from OVX or sham surgery mice blocked using 5% horse serum (PIRL/SAHMRI; Gilles Plains, SA, Australia). Citrate antigen retrieval (DAKO; Glostrup, Denmark) was performed for 20 min at 95 °C, and endogenous peroxidase activity was quenched using 0.5% H2O2 in methanol for 40 min in the dark. IHC staining was performed using primary anti-5hmC antibody (Abcam; Cambridge, UK; 1/500 dilution), secondary biotinylated anti-rabbit IgG (Vector Labs; Burlingame, CA, USA; 1/250 dilution) and tertiary streptavidin conjugate (Life Technologies; Carlsbad, CA, USA; 1/100 dilution). Sections were stained using Liquid DAB+ Substrate Chromogen System (DAKO) and counterstained in Mayer’s haematoxylin.
Data analysis, graph generation and statistical analysis were carried out using Microsoft graphpad prism 6. Differentiation, ChIP assays and gene expression analysis were assessed by one-way ANOVA. Statistical differences (*) of p ≤ 0.05 between samples are shown.
BMSC differentiation is dependent on DNA methylation
The role of hydroxymethylation during BMSC differentiation
Since DNA methylation is inhibitory to BMSC development and the enzymatic activity of TET family proteins results in the hydroxylation of methylated DNA (5mC) to 5hmC, which opposes the repressive functions associated with 5mC, we investigated whether the DNA hydroxylase TET family of enzymes regulate differentiation of BMSC. Analysis of TET1, TET2 and TET3 gene expression during BMSC differentiation showed that TET1 transcript levels were significantly decreased at 2 and 3 weeks post-osteogenic induction, when compared to BMSC cultured under normal growth conditions (Fig. 1c). In contrast, TET2 transcript levels were significantly increased at 2 weeks post-osteogenic induction (Fig. 1d) and then transcript levels decreased. TET3 expression levels only showed a significant increase at 4 weeks of osteogenic induction compared to control-treated cells (Fig. 1e). Gene expression levels measured during adipogenic differentiation demonstrated that TET1 transcript levels did not significantly change compared to BMSC cultured under normal growth conditions (Fig. 1f). Levels of TET2 transcript did not show any significant changes during adipogenesis (Fig. 1g), whereas TET3 showed a significant decrease at 1-week post-adipogenic induction (Fig. 1h). These data suggest that TET molecules may play a role in BMSC cell fate determination with more dynamic changes in gene expression evident during osteogenesis.
Given that 5hmc is an intermediate that leads to DNA demethylation and that we saw effects on BMSC differentiation when using the DNA methyltransferase inhibitor, we next assessed global levels of 5mc and 5hmc following osteogenic and adipogenic induction. The data showed a significant increase in 5hmc levels under osteogenic conditions for 1 week, with an associated downregulation of 5mc levels (Fig. 1i). Similarly, global 5hmc levels appeared elevated correlating with decreased 5mc levels post-1 week under adipogenic inductive conditions (Fig. 1j). Western blot analysis revealed that TET1 protein levels were relatively unchanged during BMSC differentiation, while TET2 protein levels were elevated following 2 weeks of osteogenic or adipogenic inductive conditions (Fig. 1k).
Functional studies showed that siRNA TET1-treated BMSC exhibited an enhanced osteogenic potential, compared to scramble siRNA-treated cells (Fig. 2c), whereas siRNA knockdown of TET2 resulted in a significantly reduced osteogenic potential, as assessed by Alizarin red staining (Fig. 2c) and quantitation of extracellular calcium levels (Fig. 2d). However, siRNA knockdown of TET3 showed no significant difference in mineral formation compared to scramble siRNA control cells (Fig. 2c, d). Supportive studies found that siRNA TET1-treated BMSC expressed significantly higher transcript levels of the osteogenic genes RUNX2, BMP-2, but not OSTEOPONTIN and OSTEOCALCIN compared to scramble siRNA-treated cells, whereas siRNA TET2-treated cells displayed a significant reduction in RUNX2, BMP-2, OSTEOPONTIN and OSTEOCALCIN gene expression levels (Fig. 2e).
Assessment of the role of TET proteins during BMSC adipogenic differentiation showed that siRNA knockdown of TET1 resulted in an increase in Nile red-positive adipocytes, compared to the scramble siRNA control (Fig. 2f, g). Conversely, siRNA knockdown of TET2 resulted in a substantial decrease in Nile red-positive adipocytes compared to the scramble siRNA-treated cells (Fig. 2f, g). However, siRNA knockdown of TET3 showed no significant effect when compared to scramble siRNA-treated BMSC. Gene expression studies found that siRNA TET2-treated BMSC expressed significantly lower transcript levels of the adipogenic genes PPARγ2, C/EBPα, ADIPSIN and LEPTIN compared to scramble siRNA-treated cells (Fig. 2h) with no significant effect seen when TET1 or TET3 was knocked down.
We next sought to determine the contribution of both TET1 and TET2 to global 5hmC and 5mC demethylation. Human BMSC were transfected with either siRNA targeting TET1 or TET2 alone or both TET1/TET2 in combination and then cultured under control or osteogenic inductive media. The data showed significant effects on global 5hmc levels only when both TET1 and TET2 were knocked down in combination under normal and osteogenic conditions (Additional file 1: Figure S1a). Examination of 5mC revealed that dual knockdown of both TET1 and TET2 caused significantly higher levels of global 5mC under both normal and osteogenic conditions (Additional file 1: Figure S1a). Functional analyses found that osteogenesis was impeded in the presence of siRNA targeting TET2, negating the increase in BMSC mineralisation following treatment with siTET1 (Additional file 1: Figure S1c). Therefore, both TET1 and TET2 appear to contribute to global 5hmC and 5mC demethylation but work through different mechanisms to regulate BMSC osteogenic differentiation.
Under adipogenic inductive conditions, overexpression of TET1 inhibited the amount of Nile red-positive, lipid-containing adipocytes compared to BMSC infected with empty vector (Fig. 3g). Gene expression analyses demonstrated significantly lower levels of PPARγ2 and transcript in TET1-overexpressing BMSC cultured under adipogenic inductive conditions compared to vector control BMSC (Fig. 3h). Conversely, overexpression of TET2 promoted higher numbers of lipid-forming adipocytes compared to vector control BMSC (Fig. 3i). Supportive studies showed that TET2-overexpressing BMSC exhibited significantly higher levels of PPARγ2 and ADIPSIN under adipogenic inductive conditions, compared to vector control BMSC (Fig. 3j). Collectively, these findings indicate that TET1 represses differentiation of BMSC into osteoblasts or adipocytes and that TET2 promotes BMSC differentiation.
Deregulation of TET1 and TET2 expression and activity during osteoporosis
TET1 and TET2 are responsible for 5hmC on lineage genes
TET1 and TET2 directly bind lineage genes
In this study, we confirmed that inhibition of DNA methylation promotes human BMSC differentiation as treatment with 5-Aza2 showed increased osteogenesis and adipogenesis. This prompted us to examine the importance of 5hmc in BMSC differentiation, which has been shown to counteract the effects of DNA methylation by removing the methylation mark or inhibiting further methylation. Functional assessment of TET molecules found that TET1 inhibits both osteogenesis and adipogenesis by repressing transcription of RUNX2 and PPARγ2, two essential master regulators of BMSC osteogenic and adipogenic commitment, respectively. Even though TET2 is the main enzyme responsible for 5hmC on osteogenic and adipogenic genes, we found that TET1 is also partly responsible for 5hmC directly on osteogenic and indirectly on adipogenic genes as we did not detect recruitment of TET1 on adipogenic genes. However, TET2 was recruited on the adipogenic genes; hence, it is likely that TET1 possibly is in a complex with TET2 or it influences the recruitment of TET2 to the adipogenic genes.
Functionally, however, TET1 represses transcription of osteogenic and adipogenic genes, which conflicts with its ability to influence 5hmC. Further analysis revealed that TET1 influences the recruitment of SIN3A and EZH2 to the osteogenic promoters. This was not evident for TET2. We attribute the repressive function of TET1 to its ability to promote recruitment of the repressors, SIN3A and EZH2, to osteogenic genes. We showed that TET1 and not TET2 is responsible for their recruitment. Previous studies have reported that TET1 can exist in a complex with Sin3A and HDAC, recruiting them to promoters to influence the recruitment of Ezh2 [28, 38, 39, 40]. This is consistent with our previous findings showing that Ezh2 is recruited to osteogenic genes and inhibits osteogenesis . This finding therefore has provided some insight into how Ezh2 is recruited to the promoters of osteogenic genes.
We previously did not detect Ezh2 on adipogenic genes nor have we now found TET1 to be recruited. It is likely that TET1 is repressing an upstream activator of adipogenesis, and this will be the basis of future investigations. Secondly, despite the reduction in 5hmC when TET1 is knocked down, the increase in gene expression and osteogenic and adipogenic differentiation indicates that its repressive function is the most prominent feature in relation to differentiation. TET1 genomic location in ESC has been shown to overlap with PRC2, Ezh2, Sin3A and H3K27me3, where it can be recruited by the PRC2 complex to bivalent sites. TET1 has also been found to co-immunoprecipitate with PRC2 complex proteins, Ezh2 and SUZ12c. It was reported that with TET1-dependent demethylation and Ezh2-mediated repression through Ezh2, the TET1–PRC2 complex maintains developmental genes in a poised state of activation and is thought to contribute to plasticity . Genome-wide studies show that TET1 has a preference to bind gene-rich regions with CpG-rich promoters with highest levels of recruitment along the transcription start sites and less so throughout the gene body . A majority of TET1-bound promoters are also marked with H3K4me3, high CpG content which are hypomethylated in support of the role of TET1 in removing DNA methylation . A second set of TET1-bound genes was associated with H3K27me3 and corresponded to the binding of the PRC2 complex . A significant amount of these genes are involved in differentiation and development, with TET1 associated with H3K27me3 or bivalent domains. 5hmC is enriched at promoter regions and TSS and mainly on regions enriched in CpG islands, which are devoid of 5mC implying that TET1 via its ability to hydroxylate 5mC contributes to the hypomethylated signature of CpG islands in ESC .
In our study, TET2 was found to promote BMSC cell lineage commitment towards either an osteoblast or adipocyte fate, partly via transcriptional activation of Runx2 or PPARγ2, respectively, and it was the main protein of the TET family responsible for 5hmC on both osteogenic and adipogenic genes. Interestingly, the third TET family member, TET3, appeared to have no effect on human BMSC differentiation. Our finding that TET1 is a potential inhibitor of human BMSC differentiation is in accord with previous studies identifying a functional role for TET1 in ESC, iPSC and neuronal stem cell development [44, 45]. TET1-deficient ESC display impaired self-renewal capacity, partly because of increased methylation and decreased expression of Nanog . Another study, however, reported that TET1-deficient ESC displayed normal levels of pluripotency genes ; however, studies consistently show an increase in lineage-specific markers, such as trophoectoderm and neuroectoderm, associated with developmental skewing towards endoderm and mesoderm fates . Triple knockout of TET1-3 in mESC shows results in an overactive Wnt signalling which results in skewed differentiation to mesoderm at the expense of neuroectoderm . Wnt signalling is important in osteogenesis and adipogenesis, and it would be of interest to determine the effect of TET1–TET3 on Wnt signalling and differentiation in bone marrow-derived MSC. These findings correlate with our observations that reduction of TET1 expression resulted in an increase in osteogenesis and adipogenesis, while overexpression of TET1 inhibited BMSC differentiation.
In agreement, with our data a recent study found that TET2 is mainly responsible for inducing adipogenesis of a murine preadipocyte cell line 3T3-L1 and was responsible for 5hmC of the PPARγ locus . In another study, TET2 and TET1 were found to promote osteoblast formation of murine mesenchymal multipotent C3H cells using short hairpin RNA. Both TET1 and TET2 were found to promote 5hmC . TET2 has been found to be responsible for neurogenic differentiation of neuronal stem cells and is primarily responsible for the 5hmC dynamics . Another study showed that knockdown of TET2 but not TET1 inhibited myogenic differentiation of myoblasts and was responsible for 5hmC . In agreement, previous studies have found that TET2 is master regulator of smooth muscle cell differentiation .
In mESC, it has been shown that TET2 knockout results in 90% reduction in 5hmC, whereas knockout of TET1 results in 44% reduction. TET2−/− mESC showed delayed gene transcription kinetics during ESC differentiation to the neuronal lineage . Although the TET2 data are in agreement, the TET1 functional data contradict ours. This could be due to TET1 having a different function in the murine MSC cell line as opposed to primary human MSC. Another study has shown using a murine teratocarcinoma cell line, ATDC5, that TET1 promotes chondrogenesis and 5hmC by short hair pin RNA . Similarly, TET1 knockdown using shRNA lentiviral vectors shows that TET1 promotes odontogenic differentiation of dental pulp stem cells . A recent study by Yang et al. found that TET1 and TET2 deficiency in mouse bone cells reduced demethylation of the P2rX7 promoter and thus downregulated exosome release, leading to intracellular accumulation of miRNAs which inhibits Runx2 signalling to impair osteogenic differentiation . It therefore appears that TET molecules may act through direct and indirect mechanisms to regulate BMSC cell fate determination.
Our studies revealed the occupancy of TET2 on bone-associated genes, RUNX2 and BMP2, with TET2 levels increasing during osteogenic induction across exons and introns, whereas the appearance of 5hmC remained relatively the same, even though TET2 was shown to be a transcriptional activator of these genes based on knockdown experiments. TET2 knockdown studies showed that the presence of 5hmC on osteogenic and adipogenic promoters was mainly attributed to TET2. A recent study reported that stable expression of the osteoblast transcription factor, Osterix, during BMP2-induced osteoblast differentiation involved TET-dependent DNA demethylation and SWI/SNF-associated nucleosome remodelling at the Osterix promoter, in mouse embryonic fibroblasts . Therefore, while TET2-mediated conversion of 5mC into 5hmC is needed for transcriptional activation, this occurs in conjunction with other recruited epigenetic enzymes and chromatin remodelling enzymes . It is therefore likely that TET2 can interact with other enzymes to modify the nucleosomes and increase transcription. This is evident in smooth muscle cells, where removal of TET2 results in a decrease in chromatin accessibility and increased H3K27me3 on Myocd, Srf and Myh11 gene loci , leading to suppression of smooth muscle differentiation. Genome-wide studies of TET1 deletion in ESC correlated with diminished 5hmC levels and a change in the histone methylation status, supporting the notion that TET proteins and 5hmC can regulate other nucleosome modifications [43, 55]. When considering the gene binding profile of TET1, TET2 and 5hmC in BMSC, we believe that TET-mediated 5hmC is a marker for poised transcription and is present on genes that are active or can be transcriptionally activated in BMSC. This is supported by other studies showing that 5hmC is biased towards activated genes involved in differentiation . Genome-wide studies have discovered that genes with 5hmC at their TSS tend to also be CpG rich and contain bivalent domains, and are more likely to be upregulated during differentiation; thus, 5hmC is likely to contribute to the poised state of activation .
While the present study failed to observe any effect of TET3 on BMSC differentiation, there may be some possibility of redundancy between TET molecules. We showed that dual knockdown of both TET1 and TET2 in human BMSC was required to significantly decrease global levels of 5hmC and increase 5mC under either normal or osteogenic inductive conditions. However, we failed to observe any additive effects of knocking down both TET1 and TET2 on BMSC osteogenic differentiation, when compared to TET2 knockdown alone. Similar studies examining TET3 knockdown in combination with TET2 knockdown showed no significant additive effects on BMSC osteogenic differentiation compared to TET2 alone (data not shown), implying that TET2 has a dominant effect on osteogenesis.
In an attempt to initiate studies examining the function of TET enzymes during bone disease, we examined expression levels of TET molecules in ovariectomised osteoporotic mice. The data demonstrated a significant decrease in expression of TET1 and TET2 and overall 5hmC levels in the bones of ovariectomised mice. These results strengthen our conclusions that TET1 and TET2 have important functions in 5hmC and bone biology Their concomitant reduction in the murine osteoporotic model also suggests that both TET1 and TET2 could have similar functions in murine bone marrow stromal cells. A recent mouse study showed that depletion of both TET1 and TET2 resulted in impaired self-renewal and differentiation of BMSC, which was associated with a significant osteopenia phenotype and supports the observation that TET1 and TET2 expression is reduced in OXV mice leading to decreased 5hmC levels in BMSC and impaired bone formation . Given that both TET1 and TET2 are severely reduced in osteoporosis, pharmacological re-induction of TET1/TET2 may be of therapeutic benefit to individuals suffering from age-related bone loss or osteoporosis .
DC, SH and SG are involved in study design. DC, SH and KG conducted the study. DC, SH and KG collected data. DC, SH, KG, RL and SG analysed the data. DC, RL, AZ, SS and SG interpreted the data. DC and SG drafted the manuscript. DC, SH, KG, RL, AZ, SS and SG revised the manuscript content. DC, SH, KG, RL, AZ, SS and SG approved the final version of manuscript. All authors read and approved the final manuscript.
The authors declare that they have no competing interests.
Availability of data and materials
All data generated or analysed during this study are included in this published article.
Consent for publication
Ethics approval and consent to participate
Human BMSC were derived from bone marrow aspirates from posterior iliac crest of normal adult volunteers after obtaining informed consent according to procedures approved by the Human Ethics Committee of the Royal Adelaide Hospital, South Australia (protocol# 940911a). Animal ovariectomy experiments were approved by the SA Pathology (23/11) and University of Adelaide (M-2013-144) Animal Ethics Committees.
This work was supported in part by the National Health and Medical Research Project Grant APP1141361 and fellowship APP1042677.
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
- 32.Cakouros D, Isenmann S, Hemming SE, Menicanin D, Camp E, Zannettino AC, Gronthos S. Novel basic helix loop helix transcription factor Hes4 antagonizes the function of twist-1 to regulate lineage commitment of bone marrow stromal/stem cells. Stem Cells Dev. 2015;24:1297–308.PubMedCrossRefGoogle Scholar
- 44.Koh KP, Yabuuchi A, Rao S, Huang Y, Cunniff K, Nardone J, Laiho A, Tahiliani M, Sommer CA, Mostoslavsky G, et al. Tet1 and Tet2 regulate 5-hydroxymethylcytosine production and cell lineage specification in mouse embryonic stem cells. Cell Stem Cell. 2011;8(2):200–13.PubMedPubMedCentralCrossRefGoogle Scholar
- 49.Sepulveda H, Villagra A, Montecino M. Tet-mediated DNA demethylation is required for SWI/SNF-dependent chromatin remodeling and histone-modifying activities that trigger expression of the Sp7 osteoblast master gene during mesenchymal lineage commitment. Mol Cell Biol. 2017;37(20):e00177.PubMedPubMedCentralCrossRefGoogle Scholar
- 56.Serandour AA, Avner S, Oger F, Bizot M, Percevault F, Lucchetti-Miganeh C, Palierne G, Gheeraert C, Barloy-Hubler F, Peron CL, et al. Dynamic hydroxymethylation of deoxyribonucleic acid marks differentiation-associated enhancers. Nucleic Acids Res. 2012;40(17):8255–65.PubMedPubMedCentralCrossRefGoogle Scholar
Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.