TET3 Mediates Alterations in the Epigenetic Marker ...

JBMR

ORIGINAL ARTICLE

TET3 Mediates Alterations in the Epigenetic Marker 5hmC and Akt pathway in Steroid-Associated Osteonecrosis Jie Zhao,1,2 Xin-long Ma,1,2 Jian-xiong Ma,2 Lei Sun,2 Bin Lu,2 Ying Wang,2 Guo-sheng Xing,2 Yan Wang,2 Ben-chao Dong,2 Li-yan Xu,1,2 Ming-Jie Kuang,1,2 Lin Fu,1,2 Hao-hao Bai,2 Yue Ma,3,4 and Wei-lin Jin3,4 1

Tianjin Medical University General Hospital, Tianjin, China Institute of Orthopedics, Tianjin Hospital, Tianjin, China 3 Department of Instrument Science and Engineering, Key Lab. for Thin Film and Microfabrication Technology of Ministry of Education, School of Electronic Information and Electronic Engineering, Shanghai Jiao Tong University, Shanghai, China 4 National Centers for Translational Medicine, Shanghai Jiao Tong University, Shanghai, China 2

ABSTRACT Steroid-associated osteonecrosis (SAON) is one of the common complications of clinical glucocorticoid (GC) administration, with osteocyte apoptosis appearing as the primary histopathological lesion. However, the precise mechanism underlying SAON remains unknown. Epigenetic modification may be a major cause of SAON. Recently, cumulative research revealed that Ten-Eleven Translocation (TET) proteins can catalyze the conversion of 5-methylcytosine (5mC) to 5-hydroxymethylcytosine (5hmC) and then alter the epigenetic state of DNA. Here, we report that TET3-5hmC was upregulated in the femoral head tissues of SAON patients and MLO-Y4 cells with dexamethasone (Dex) treatment. Knockdown of TET3 in MLO-Y4 cells decreased 5hmC enrichment and rescued Dex-induced apoptosis. Meanwhile, the local intramedullary injection of TET3 siRNA in Sprague-Dawley rats abrogated GC-induced osteocyte apoptosis, histopathological changes, abnormal MRI signals, and bone microstructure declines in the femoral head in vivo. Moreover, a hydroxymethylated DNA immunoprecipitation (hMeDIP)-chip analysis of Dex-treated osteocytes revealed 456 different 5hmC-enriched genes. The Akt pathway was found to mediate the functional effect of Dex-induced dynamic 5hmC change; this was further verified in clinical samples. The loss of TET3 in MLO-Y4 cells abrogated Dex-induced Akt signaling pathway inhibition. Therefore, our data for the first time identify the effect of TET3-5hmC on the Akt pathway and the necessity of this signaling cascade in SAON, identifying a new potential therapeutic target. © 2016 American Society for Bone and Mineral Research. KEY WORDS: OSTEONECROSIS; EPIGENETICS; DNA DEMETHYLATION; TET3; APOPTOSIS

Introduction

S

teroid-associated osteonecrosis (SAON) is considered a multifactorial disease that leads to subchondral collapses and total joint replacement during later stages.(1,2) SAON ranks first among the known risk factors of nontraumatic femoral head necrosis, which is associated with the wide use of glucocorticoids (GCs) to treat many diseases in the clinic (eg, rheumatoid arthritis and systemic lupus erythematosus).(3) There are several alternative mechanisms underlying SAON, such as fat embolization, intramedullary pressure changes, modified artery constriction, circulatory impairment, coagulation disorders, and cell dysfunction.(4,5) However, osteocyte apoptosis has recently been shown to be the primary histopathological change in osteonecrosis.(1,6–8) GCs induce osteocyte apoptosis, which disrupts the mechanosensory function of the osteocyte lacunar–canalicular system and

ultimately leads to the collapse of the femoral head,(1) but the molecular mechanism of osteocyte apoptosis remains elusive. Epigenetics, an effective method for studying the interplay between environmental signals and the genome, has received a great deal of attention recently.(9) It is becoming apparent that methylation modification plays an important role in SAON. Aberrant methylation of the ABCB1 gene has been shown to be responsible for the pathogenesis of SAON,(4) and icariin may benefit the mesenchymal stem cells of patients with SAON through ABCB1 promoter demethylation.(10) However, in the past few years, it has become apparent that DNA methylation is not a static epigenetic mark but is highly dynamic and is governed by a precise molecular network of regulators.(11) 5-Methylcytosine (5mC) can be further oxidized to 5-hydroxymethylcytosine (5hmC), 5-formylcytosines (5fC) and 5-carboxylcytosines (5caC) by the Ten-Eleven Translocation (TET) protein family including TET1, TET2 and TET3, resulting in

Received in original form May 10, 2016; revised form September 1, 2016; accepted September 9, 2016. Accepted manuscript online September 13, 2016. Address correspondence to: Xin-long Ma, MD, Tianjin Medical University General Hospital, 154, Anshan Street, Heping District, Tianjin 300052, China. E-mail: [email protected] Additional Supporting Information may be found in the online version of this article. Journal of Bone and Mineral Research, Vol. xx, No. xx, Month 2016, pp 1–14 DOI: 10.1002/jbmr.2992 © 2016 American Society for Bone and Mineral Research

1

active demethylation.(12,13) The dysregulation of 5hmC levels may lead to neurological diseases, cancer, arthritis, and some other diseases.(14) In view of the recent advances in the understanding of DNA demethylation pathways and increasing evidence of the function of stable 5hmC and its derivatives in the epigenome, we sought to investigate the effect of TET and 5hmC in SAON. In this study, we therefore first measured the expression of TET1, TET2, TET3 and 5hmC in bone tissues from SAON and femoral neck fracture patients. Then, we examined the changes in 5hmC and the TET family in dexamethasone (Dex)-treated MLO-Y4 cells. A hydroxymethylated DNA immunoprecipitation (hMeDIP)-chip analysis was performed to explore the functional effects of dynamic 5hmC change, which were further studied in clinical samples. Knockdown of TET in MLO-Y4 cells or in vivo siRNA technology were used to verify the effect of TET-5hmC on Dex-induced osteocyte apoptosis and SAON. We aimed to determine the effects of GCs on the epigenome of osteocyte and define the role of TET-5hmC in a SAON model.

Materials and Methods SAON and femoral neck fracture bone tissue preparation Human SAON and femoral neck fracture samples were obtained from surgical procedures in the Department of Joint Surgery and Department of Traumatic Orthopedics at Tianjin Hospital, respectively. All procedures regarding obtaining patient samples were approved by the ethics committee of the Tianjin Hospital Institutional Review Board and complied with the World Medical Association Declaration of Helsinki. Six SAON and six femoral neck fracture patients were recruited. The severity of SAON and femoral neck fracture was evaluated using the Ficat and Garden staging system, respectively.(15,16) SAON operative indications included a collapsed necrotic femoral head with intractable pain. Patients with a demonstrable history of direct trauma or with the possibility of a combination of causes were excluded.(17) During surgery, the femoral head was extracted and then the necrotic bone around the center of femoral head was harvested for study. Bone specimens from six patients who underwent hip arthroplasty for femoral neck fractures were harvested from the same place and used as controls. No agematched controls could be obtained because of the absence of a surgical indication for hip arthroplasty in young patients with femoral neck fractures.

Cell cultures, Dex treatment, and transfection MLO-Y4 osteocyte-like cells (a gift from Dr. Lynda Bonewald, University of Missouri-Kansas City, Kansas City, MO, USA) were cultured as described.(18) For dose-dependent experiments, cells were treated with varying concentrations (1 10–8 M to 1 10–5 M) of Dex (Sigma-Aldrich, St. Louis, MO, USA). For timedependent experiments, cells were treated with 1 10–6 M Dex for various durations of time.(19,20) To activate the Akt signaling pathway, cells were exposed to the phosphatase and tensin homolog (PTEN) inhibitor bpV(phen) (1 10–6 M; Sigma-Aldrich) for 30 min. Afterward, the cells were incubated with 1 10–6 M Dex for up to 12 hours. For knockdown of TET, TET1-specific siRNA (TET1-1 GCAGATGGCCGTGACACAAAT, TET1-2 GCAGCTAGCTATAGAGTATAG), TET2-specific siRNA (TET2-1 CTCAGGGATGTCCTATTGCTAAA, TET2-2 GGATGTAAGTTTGCCAGAAGC), TET3-specific siRNA (TET3-1 GCTCCAACGAGAAGCTATTTG, TET32 AAGCGCAACCTATTCTTGGAA), and control scramble siRNA

2

ZHAO ET AL.

(ACGUGACACGUUCGGAGAATT) were synthesized by Biotend (Biotend Biotechnology Co., Ltd, Shanghai, China) based on published studies.(21) Transfections were performed using Lipofectamine 2000 Transfection Reagent (Invitrogen, Carlsbad, CA, USA) according to the manufacturer’s protocol. Twenty-four hours thereafter, cells were treated with 1 10–6 M Dex for an additional 12 hours.

hMeDIP-chip and data processing MLO-Y4 cells were treated with 1 10–6 M Dex for 0 hours, 4 hours, and 12 hours. Thereafter, genomic DNA (gDNA) was extracted and sonicated to random fragments of 200 to 1000 bp. One microgram of sonicated gDNA was immunoprecipitated with 1 mL of 5hmC antibody overnight at 4°C with rocking agitation. Then, antibodyDNA complexes were captured with protein A/G beads (ThermoFisher Scientific, Waltham, MA, USA). Then, 5hmC-containing DNA fragments were purified using Qiagen MinElute columns (Qiagen, Hilden, Germany). For DNA labeling, the NimbleGen Dual-Color DNA Labeling Kit was used according to the manufacturer’s guidelines (Nimblegen, Roche, Basel, Switzerland), and the labeled DNA was purified using isopropanol/ethanol precipitation. A mouse RefSeq Promoter Array (Arraystar lnc, Rockville, MD, USA) was hybridized with the labeled DNA at 42°C for 16 to 20 hours in a hybridization chamber (Nimblegen).(22) The hMeDIP chip data were analyzed with NimbleScan v2.5 (NimbleGen). A normalized peakMvalue (peakMvalue ¼ log2hMeDIP/Input) was created for each sample to quantify the relative value of 5hmC, which is proportional to 5hmC expression. The threshold set of peakMvalue changes 0 represented the different 5hmC-enriched genes (DEG). Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway and gene ontology (GO) analysis of DEG were performed using the database for annotation, visualization, and integrated discovery (DAVID) website.(23) The –log10(pValue) was used as a measure of the significance of the pathway associated with the 5hmC change. A heat map and Venn diagram showing the dynamic change in 5hmC were constructed with MultiExperiment Viewer v4.9 and Venny 2.1 website, respectively.

Establishment of the SAON model and knockdown of TET3 in vivo The animal procedures were conducted according to the Guide for the Care and Use of Laboratory Animals: 8th Edition, and the study protocol was approved by the Ethical Committee of the Tianjin Medical University General Hospital and Tianjin Hospital. Twenty-four 3-month-old adult male weight-matched Sprague Dawley (SD) rats (203 14 g) were randomly divided into four experimental groups of six rats each and kept under the same standard conditions; water and food were available ad libitum. The SAON model was established based on a protocol reported previously(24); methylprednisolone sodium (MPS; Pharmacia & Upjohn, Peapack, NJ, USA) was subcutaneously injected at a dose of 21 mg/kg per day for 4 weeks. The siRNA transfection complex was prepared according to the manufacturer’s instructions. Briefly, 5 nmol siRNA (Biotend), 40 mL Entranster-in vivo transfection reagent (Engreen, Beijing, China), and 40 mL of 10% glucose were mixed for 15 min at room temperature. TET3 or scramble siRNA-Entranster-in vivo complex was injected into medullary cavity of bilateral distal femur twice over the course of 4 weeks.(2,25,26) Euthanasia was

Journal of Bone and Mineral Research

performed at week 4. The left femoral heads were used for mCT scan (n ¼ 6/group), histological evaluation (n ¼ 6/group), and TUNEL analysis (n ¼ 6/group).(27,28) The right femoral heads were separated into two parts for Western blot (n ¼ 4/group) and dot blot analysis (n ¼ 4/group). Details regarding the analyses can be found in the Supplemental Methods.

mCT-based trabecular architecture assessment The femoral heads were scanned with an Inveon micro PET/CT manufactured by Siemens (Berlin, Germany) at a voltage of 80 kV and a current of 500 mA, with an entire scan length of 20 mm from the top of the femoral head to the femoral shaft in a spatial resolution of 10 mm. 3D structures were reconstructed using the Inveon analysis workstation. The ROI was determined as an irregular anatomic contour adjacent to the endocortical surface and epiphyseal line in proximal epiphysis. The cortical bone and spongy bone were separated manually by auto trace; later, the trabeculae and the bone marrow were separated using the threshold function.(29) The bone volume/total volume (BV/TV), bone surface area/bone volume (BS/BV), trabecular thickness, trabecular number, and trabecular separation were calculated.(30)

Statistical analysis Osteonecrosis was defined as previously described(27,28,31) and the incidence of SAON in rats was defined as the number of femoral heads with predominant empty lacunae divided by six femoral heads in each group. Number data were expressed as the mean SD with ANOVA (Fisher’s least significant difference [LSD]) method to assess significant differences between groups. Analysis of covariance (ANCOVA) was used to correct the age bias between SAON and fracture patients. All experiments were repeated at least three times. Statistical analysis was performed using SPSS 20.0 software (IBM Corp., Armonk, NY, USA); p < 0.05 was considered significant.

Results TET3-5hmC level was upregulated in SAON femoral heads The patient characteristics are listed in Supplemental Tables 1 and 2. In the SAON group, six SAON patients ranging from 59 to 77 years old (66.2 6.6 years) were enrolled. In the control group, six patients ranging from 68 to 85 years old (76.2 5.5 years) were included. The representative radiographs showed prominent femoral head erosion and collapse in the SAON group and evident femoral neck fracture in the control group (Fig. 1A). TUNEL analysis was performed to measure cell apoptosis in the bone tissues. In the SAON group, osteocytes embedded in the trabecula displayed strong TUNEL staining, which is consistent with previous studies (Fig. 1B, C).(32) It has been reported that TET3 mediates GC-induced harmful neurodevelopmental effects.(33) To explore the role of TET in the pathogenesis of SAON, TET1, TET2, and TET3 levels were analyzed. Real-time PCR showed that the levels of TET1 and TET2 were unchanged, whereas the expression of TET3 was increased in SAON tissues (Fig. 1D–F), which was confirmed by Western blot (Fig. 1G, H). TET3 mediates 5mC conversion to 5hmC, 5fC, and 5caC. As the steady-state level of 5hmC is considered much higher than the pool of 5fC and 5caC,(34) we evaluated only the 5hmC level. DNA dot blot indicated strong enrichment of 5hmC in SAON (Fig. 1I, J). Journal of Bone and Mineral Research

The TET3-5hmC increase was involved in Dex-induced osteocyte apoptosis To determine the effect of GC treatment on osteocyte viability, we first performed an in vitro analysis in MLO-Y4 cells. A TUNEL assay indicated that MLO-Y4 cells treated with Dex demonstrated a dose-dependent increase in apoptosis (Fig. 2A, B). Furthermore, MLO-Y4 cells treated with a time course of 1 10– 6 M Dex (a range from 2 hours, 4 hours, 8 hours, to 12 hours) showed increased apoptosis (Fig. 2C, D). To explore the mechanism of Dex-induced apoptosis, Bcl-2, Bax, and cleaved caspase-3 protein levels were quantified. The data showed that proapoptotic Bax and cleaved caspase-3 were upregulated, whereas the anti-apoptotic Bcl-2 protein level was lower at 8 hours and 12 hours (Fig. 2F). In the CCK-8 and trypan blue exclusion assay, the number of MLO-Y4 normal cells was found to be reduced by Dex treatment in both a dose-dependent and a time-dependent manner (Supplemental Figs. 1A, B, 2A, B). In addition, an ethynyl deoxyuridine (EdU) incorporation assay indicated that DNA replication and cell proliferation were suppressed by Dex (Supplemental Figs. 1C, D, 2C, D). To study whether TET and 5hmC were involved in Dexinduced apoptosis in vitro, genome-wide 5hmC levels and the expression of TET1, TET2, and TET3 were quantified. IF (Immunofluorescence) and dot blot showed that global 5hmC was upregulated by Dex in a time-dependent manner (Fig. 2E, G). Real-time PCR revealed that TET3 increased after Dex treatment while TET2 remained unchanged. Although TET1 increased slightly, it was not as significant as TET3 (Fig. 2H–J), which was similar to the results of the protein expression analysis (Fig. 2K). The upregulation was observed in both total and nuclear TET3, as revealed by subcellular extraction analysis (Fig. 2L). These data indicated that TET3 may localize in nucleus after Dex treatment to induce 5hmC enrichment and initiate osteocyte apoptosis.

The Akt pathway mediated the functional effect of Dexinduced dynamic changes in 5hmC To understand the functional effect of the 5hmC change, we performed a hMeDIP-chip assay to examine the 5hmC epigenome in MLO-Y4 cells over the course of Dex treatment. Following the hMeDIP-chip assay and purification of 5hmCenriched DNA fragments, the ArrayStar Mouse RefSeq Promoter Array was used to characterize the distribution of 5hmC within promoters. A heat map and Venn diagrams were used to depict the dynamic change in Dex-induced 5hmC expression. Four hundred and fifty-six genes predicted to have different enrichment levels of 5hmC were selected. Of these genes, 92 genes were 5hmC-upregulated and 98 genes were 5hmCdownregulated in all Dex-treated MLO-Y4 cells (4 hours and 12 hours). Moreover, 122 5hmC-upregulated genes and 144 5hmCdownregulated genes occurring in Dex treatment for 12 hours were also identified (Fig. 3A). Then, we used Venn diagrams to further show the dynamic 5hmC change in different types of genes. The genes were classified as high CpG genes (HCG), intermediate CpG genes (ICG), and low CpG genes (LCG) based on the CpG content in the promoter. The number of genes with differential 5hmC enrichment is listed in Fig. 3B. GO analysis revealed significant accumulation of DEG involved in several important cellular processes, such as development, localization, metabolism, and differentiation (Fig. 3C). KEGG pathway analysis showed that signaling pathways, including the PI3K-Akt pathway, Notch pathway, apoptosis

TET3-5HMC-AKT MEDIATES GC-INDUCED OSTEOCYTE APOPTOSIS AND SAON

3

Fig. 1. TET3-5hmC is upregulated in the femoral head tissues of patients with SAON. (A) Representative radiograph of SAON and femoral neck fracture patients. (B, C) TUNEL analysis of osteocyte apoptosis in SAON and femoral neck fracture. TUNEL positive osteocytes appear red. Nuclei (blue) are counterstained with DAPI. The merged image is shown in the bottom panel. Scale bar ¼ 30 mm. The TUNEL results are expressed as the percentage of apoptotic cells relative to total cells. Data are expressed as the mean SD from three independent biological replicates. (D–F) Quantification of relative mRNA levels of TET1, TET2, and TET3 in SAON and femoral neck fracture. Data are represented as the mean SD from three independent biological replicates. In all graphs, the fracture control group was set to 1, and SAON is expressed as the fold change relative to fracture. (G, H) Western blot analysis of TET3 in SAON and femoral neck fracture; a total of 20 mg protein was loaded (b-actin was used as a control for loading). The relative expression level of TET3 was calculated as TET3/b-actin. (I, J) Quantification of global 5hmC in SAON and femoral neck fractures with dot blot; a total of 500 ng DNA was loaded (DNA was stained with methylene blue as a control for loading). The relative level of 5hmC was calculated as 5hmC/DNA. p < 0.05, p < 0.01. DAPI ¼ 4,6-diamidino-2-phenylindole.

4

ZHAO ET AL.


Fig. 2. Dex induces TET3-5hmC upregulation in a time-dependent manner. Also see Supplemental Figs. 1 and 2. (A, B) TUNEL analysis of MLO-Y4 cell treated with varying concentrations (1 10–8 M to 1 10–5 M) of Dex for 12 hours. (C, D) TUNEL analysis of MLO-Y4 cell over the course of Dex (1 10–6 M) treatment. Data are represented as the mean SD from three independent biological replicates. Scale bar ¼ 30 mm. (E) Immunostaining of 5hmC in MLO-Y4 cells with Dex treatment (green). Nuclei (blue) are counterstained with DAPI; the merged image is shown in the bottom panel. Scale bar ¼ 30 mm. (F) Western blot analysis of Bcl-2, Bax, and cleaved caspase-3 in Dex-treated MLO-Y4 cells. In all graphs, the control is set to 1, and treatments are referred to as the fold change relative to control. (G) Dot blot analysis of global 5hmC in Dex-treated MLO-Y4 cells. Left panel: dot blot of 5hmC, right panel: methylene blue staining. (H–J) Relative gene expression levels of TET1, TET2, and TET3 in MLO-Y4 cells with Dex treatment. Data are represented as the mean SD from three independent biological replicates. In all graphs, the control is set to 1, and treatments are referred to as the fold change relative to control. (K) Western blot analysis of TET1, TET2, and TET3 in Dex-treated MLO-Y4 cells; a total of 50 mg protein was loaded. (L) Subcellular analysis of nuclear TET3 with Dex treatment (histone H3 served as a loading control). Twenty micrograms of nuclear protein were loaded. p < 0.05, p < 0.01. DAPI ¼ 4,6-diamidino-2-phenylindole.



5

Fig. 3. Dynamic changes in global 5hmC with Dex treatment. (A) A heat map of 456 genes with differential 5hmC enrichment in MLO-Y4 cells with Dex treatment at different times. Red indicates upregulation of 5hmC; green indicates a reduction. ¼ 5hmC upregulation at 4 hours and 12 hours, ¼ 5hmC upregulation at 12 hours, ¼ 5hmC downregulation at 4 hours and 12 hours, ¼ 5hmC downregulation at 12 hours. (B) Venn diagrams showing dynamic 5hmC change in HCG, ICG, LCG, and total genes. The number of genes with 5hmC upregulation is listed on the upper panel. The number of genes with 5hmC downregulation is listed on the lower panel. ¼ 5hmC upregulation at 12 hours but not 4 hours, ¼ 5hmC upregulation at 4 hours but not 12 hours, ¼ 5hmC upregulation at both 12 hours and 4 hours, ¼ 5hmC downregulation at 12 hours but not 4 hours, ¼ 5hmC downregulation at 4 hours but not 12 hours, ¼ 5hmC downregulation at both 12 hours and 4 hours. (C) Representation of GO term results from DEG for biological process. Upper panel: 5hmC upregulated genes; Lower panel: 5hmC downregulated genes.

pathway, and Wnt pathway, exhibited dynamic changes in 5hmC levels (Fig. 4A, B). The PI3K-Akt and Notch are signaling pathways showed the most significant 5hmC upregulation and downregulation, respectively. Eleven genes associated with the PI3K-Akt signaling pathway and the corresponding peakMvalues are listed in Table 1, which shows the 5hmC expression in these gene promoters. Then, we used real-time PCR to quantify the mRNA expression of these genes. The results showed that PTEN, Ppp2r5d, and JAK3 were upregulated with different

6

ZHAO ET AL.

patterns in Dex-treated cells relative to control cells, whereas PIK3r5, PIK3cd, TCL1, and PDK2 were downregulated (Fig. 4C–M). The mRNA levels of PIK3c2b, Itga1, Itga11, and Tlr4 remained almost unchanged, though the hMeDIP-chip assay revealed different enrichment with 5hmC (Fig. 4C–M). According to the Western blot assay, PTEN was upregulated and p-Akt was downregulated in all Dex-treated cells. Meanwhile, the total Akt level did not change significantly (Fig. 4N). To further explore whether the Akt pathway mediates the functional effect of


Fig. 4. The Akt signaling pathway is modified by Dex-induced TET3-5hmC change. (A, B) KEGG pathway analysis of DEG during Dex treatment. Left panel: 5hmC upregulated genes, Right panel: 5hmC downregulated genes. Corresponding –log10 (pValue) are shown. Data analysis was performed using DAVID (https://david.ncifcrf.gov/). (C–M) Relative gene expression levels of PIK3r5, PIK3cd, PIK3c2b, Ppp2r5d, PTEN, Itga1, Itga11, TCL1, Tlr4, PDK2, and JAK3 in Dex-treated MLO-Y4 cells. Data are represented as the mean SD from three independent biological replicates. In all graphs, the control is set to 1, and treatments are expressed as the fold change relative to control. (N) Western blot analysis of PTEN, p-Akt, and Akt in MLO-Y4 cells with Dex treatment. p < 0.05, p < 0.01.

Dex-induced 5hmC change, bpV(phen) was used to activate the Akt pathway. We found that the osteocyte apoptosis induced by Dex was reversed significantly by activating Akt signaling, and Western blot revealed the same pattern (Fig. 5A–C). Significantly, the levels of Akt and apoptosis were also changed in SAON tissues. The expression of PTEN was increased in SAON, while p-


Akt and p-PI3K decreased. Meanwhile, Bcl-2 decreased at the protein level in SAON, but Bax and active caspase-3 increased (Fig. 5D). These data verified that the PTEN-Akt pathway mediates the functional effect of Dex-induced TET-5hmC change in osteocyte and is critical for osteocyte apoptosis and SAON.


7

Table 1. The peakMvalue of 11 Genes Associated With Akt Pathway at Indicated Time Points Gene PIK3r5 PIK3cd PIK3c2b Ppp2r5d PTEN Itga1 Itga11 TCL1 Tlr4 PDK2 JAK3

Control

4 Hours

12 Hours

0.676239 1.043655 1.307617 1.612871 1.273141 2.082474 1.397681 1.071684 1.334823 0.947982 1.292034

1.128706 1.356383 1.223627 1.133585 1.283806 1.983941 1.586548 1.199659 1.270801 1.235555 1.407195

1.036246 0.977372 1.313212 1.649453 0.988902 1.605688 1.605688 1.231157 1.635225 1.283706 1.448056

The peakMvalue reveals the relative quantification of 5hmC.

Additionally, we also found changes in Notch4 (not Notch2) expression in SAON tissues (Supplemental Fig. 3). Thus, Notch signaling may also partially mediate the functional effect of 5hmC change and could be further explored in the future.

Knockdown of TET3 abrogated Dex-induced osteocyte apoptosis and Akt pathway inhibition To further confirm the effect of TET-5hmC on Dex-induced osteocyte apoptosis, we performed siRNA-mediated knockdown of TET1-3 expression in MLO-Y4 cells. We first tested the efficacy of several reported TET1-3 sequences in knocking down TET expression. After 24 hours of transfection, real-time PCR data showed that five sequences (siTET1-1, siTET2-1, siTET2-2, siTET3-1, and siTET3-2) decreased TET mRNA expression (Fig. 6A). The cells transfected with siTET1-1, siTET2-2, and siTET3-1 were used in the subsequent experiments. Significantly, only knockdown of TET3 reversed Dex-induced apoptosis, whereas knockdown of TET1 or TET2 had no effect (Fig. 6B). Twelve hours after Dex treatment, 30.00% 1.91% of transfected control cells, 30.05% 2.47% of siTET1-transfected cells, and 30.75% 3.46% of siTET2transfected cells were TUNEL positive. In contrast, the number of apoptotic cells decreased to 17.14% 1.06% in siTET3transfected cells (Fig. 6C). Western blot indicated that only knockdown of TET3 inhibited caspase-3 activation, although knockdown of TET1 and TET2 had a slight effect on Bcl-2 (Fig. 6D). In addition, transfection of siTET3 reversed Dex-induced TET3 upregulation at the protein level. Subcellular extraction analysis indicated that nuclear TET3 was also downregulated (Fig. 6E, Supplemental Fig. 4A). Meanwhile, dot blot and IF showed that the global enrichment of 5hmC was abrogated after knockdown of TET3 (Fig. 6F, G). These results further verified that TET3 is specific and indispensable for Dex-induced 5hmC enrichment and osteocyte apoptosis. We further explored the activity of the Akt signaling pathway. Compared with the Dex-treated group, knockdown of TET3 reversed the phosphorylation of Akt and downregulated PTEN levels; thus, the suppression of the Akt signaling pathway decreased (Fig. 6H). The data showed that the number of normal MLO-Y4 cells in the siTET3 group was larger than that in the Dex-treated group (Supplemental Fig. 4B, C). Further EdU incorporation assays showed that the number of EdU-positive cells increased significantly after TET3 knockdown compared with Dex treatment (Supplemental Fig. 4D, E).

8

ZHAO ET AL.

Knockdown of TET3 alleviated SAON in vivo TET3 siRNA transfection in vivo was performed as illustrated in Fig. 7A. Western blot showed that siTET3 treatment, which did not alter the levels of TET1 and TET2, abrogated MPSinduced TET3 expression in femoral head bone tissues (Fig. 7B) and further reversed MPS induction of global 5hmC accumulation (Fig. 7C). We next explored the therapeutic effect of TET3 knockdown on SAON by performing H&E staining, TUNEL assay, mCT and MRI scans in rats. H&E staining showed obvious empty bone lacunae and adipocytes in the bone marrow of the SAON group; these effects were rescued by siTET3 treatment. The incidence of osteonecrosis at week 4 was 83.3% (5/6) in the SAON group and 0 (0/6) in the siTET3 group by the standard of H&E-stained histological examination (Fig. 7D). TUNEL analysis showed that the osteocytes embedded in the trabeculae at the proximal epiphysis of the femoral head displayed intense staining in the SAON group, whereas few osteocytes displayed positive staining with siTET3 treatment (Fig. 7E, F). mCT showed that MPS treatment decreased the BV/TV, trabecular number, and trabecular thickness and increased trabecular separation, indicating apparent trabecular bone loss in SAON femoral heads; this change was significantly rescued by siTET3 treatment (Fig. 7G, H). Moreover, MRI scans showed heterogeneous T1WI (T1-weighted imaging) signals and high-intensity T2WI (T2-wighted imaging) signals in the femoral head in the SAON and scramble groups. In contrast, the siTET3 group showed a homogeneous T1WI signal and a low-intensity T2WI signal, similar to the results observed in a normal femoral head (Fig. 7I). Collectively, these data suggest that knockdown of TET3 alleviated osteocyte apoptosis and SAON in vivo.

Discussion Our results uncovered a direct impact of the TET3-5hmC-Akt pathway on the pathogenesis of SAON (Fig. 8). This deduction is based on the following. First, we observed that a global enrichment of 5hmC and TET3 increase in SAON bone specimens. Second, we found that Dex-induced TET3-5hmC upregulation in MLO-Y4 osteocytes. Third, a hMeDIP-chip analysis exhibited dynamic 5hmC change in the promoter of genes associated with the Akt pathway, and pharmacological activation of the Akt pathway rescued Dex-induced osteocyte apoptosis. Fourth, Akt and apoptosis signals were altered in SAON samples. Fifth, knockdown of TET3 with specific siRNA reversed the GC-induced inhibition of Akt signaling and osteocyte apoptosis in vitro and SAON in vivo. An earlier study reported that GCs elicit strong and persistent effects on DNA hydroxymethylation in neural stem cells, with TET3 being a key factor for harmful neurodevelopmental effects,(33) thereby implicating alterations in the TET3-5hmC pathway in SAON. The pattern of DNA methylation at cytosine bases in the genome is tightly linked to gene expression, and DNA methylation abnormalities are often observed in diseases. The TET proteins promote the reversal of DNA methylation.(34) TET1 is indispensable for maintaining embryonic stem cell pluripotency and exerts tumor suppressing functions in gastric cancer.(35,36) TET2 was positively correlated with the regulation of neuron survival.(37) TET3 is required for the normal survival, proliferation, and differentiation of neural progenitor cells.(33) Nevertheless, the expression pattern and exact function of these


Fig. 5. The Akt signaling pathway mediates the functional effect of Dex-induced TET3-5hmC change. (A, B) TUNEL analysis of MLO-Y4 cells pretreated with 1 10–6 M bpV(phen) for 30 min followed by 1 10–6 M Dex for 12 hours. Data are represented as the mean SD from three independent biological replicates. Scale bar ¼ 30 mm. (C) Western blot analysis of Bcl-2, Bax, and cleaved caspase-3 in MLO-Y4 cells. (D) Western blot analysis of p-Akt, Akt, PTEN, p-PI3K, PI3K, Bcl-2, Bax, and cleaved caspase-3 in femoral neck fracture and SAON bone tissues. p < 0.05, p < 0.01.

three important proteins have not been reported in bone tissues previously. In our study, we first found the expression of TET1, TET2, and TET3 in bone tissues. TET3 function appears to be distinct from that of TET1 and TET2 in SAON pathogenesis, as our data showed that only TET3 expression was significantly upregulated and accompanied by 5hmC accumulation in SAON. TET1 was also slightly increased by Dex treatment in vitro, but not as obviously as TET3. In addition, only knockdown of TET3 could reverse Dex-induced MLO-Y4 cell apoptosis; thus, TET3 is specifically involved in Dex-induced osteocyte apoptosis and SAON. Apart from a conserved core catalytic region in the C terminus of each TET protein, TET1 and TET3 have an N-terminal CXXC zinc finger domain that can bind DNA and facilitate recruitment to target genes in the genome.(38,39) Thus, GC-induced TET, especially TET3 expression, may aid in the demethylation of target genes. In addition to enzymatic roles, TET proteins serve as transcriptional co-activators/ co-repressors by interacting with some transcriptional regulators and scaffolding proteins.(40) Further studies are needed to clarify the effects of such nonenzymatic activity of TET proteins induced by GCs. GCs, including cortisol as the predominant GC in human, regulate the expression of a wide array of target genes through


binding to glucocorticoid receptor a (GRa) and GRb. Both are members of the nuclear receptor superfamily and present in osteocytes.(41,42) GCs lead to multiple changes of osteocytes; however, studies regarding the roles of GRs in mediating these effects are scarce.(42) It has been reported that GRs mediate GC promotion of osteocyte apoptosis by activating Pyk2 and JNK, followed by inside-out signaling that leads to anoikis.(43) Our demonstration that GCs induce osteocyte apoptosis is consistent with previous studies.(18,44–46) We found that GCs induced TET3 upregulation and localization to the nucleus, as well as 5hmC enrichment in a time-dependent manner, in osteocytes. Thus, it is suspected that GRs have the potential to mediate the expression of TET3 and initiate genomic hydroxymethylation and osteocyte apoptosis. However, how TET3 is modulated by GRs and recruited into the nucleus to modify methylation needs to be further studied. A hMeDIP-chip assay was performed to understand the functional effect of 5hmC enrichment. The dynamic changes in 5hmC distribution with Dex treatment are similar to the 5hmC distribution in progenitor differentiation during chondrogenesis and neurogenesis.(12,21) KEGG pathway analysis indicated that the PI3K-Akt signaling pathway could mediate the functional effect of Dex-induced 5hmC change, a finding that was verified


9

Fig. 6. Knockdown of TET3 abrogates Dex-induced osteocyte apoptosis and Akt pathway inhibition. Also see Supplemental Fig. 4. (A) Relative expression level of TET1, TET2, and TET3 in MLO-Y4 cells with specific siRNA treatment. Data are represented as the mean SD from three independent biological replicates. In all graphs, scramble is set to 1, and treatments are expressed as the fold change relative to scramble. (B, C) TUNEL analysis of MLOY4 cells with knockdown of TET1, TET2, and TET3. Data are represented as the mean SD from three independent biological replicates. Scale bar ¼ 30 mm. (D) Western blot analysis of Bcl-2, Bax, and cleaved caspase-3 in MLO-Y4 cells with knockdown of TET1, TET2, and TET3. (E) Western blot and subcellular analysis of TET3 in MLO-Y4 cells with knockdown of TET3. Fifty micrograms of total protein and 20 mg nuclear protein were loaded. (F) Quantification of 5hmC levels in MLO-Y4 cells with knockdown of TET3 by dot blot. Left panel: dot blot of 5hmC; Right panel: methylene blue staining. (G) Immunostaining of 5hmC in MLO-Y4 cells with knockdown of TET3 (green). Nuclei (blue) are counterstained with DAPI. The merged image is shown in the bottom panel. Scale bar ¼ 30 mm. (H) Western blot analysis of PTEN, p-Akt, and Akt. p < 0.05, p < 0.01. DAPI ¼ 4,6-diamidino-2-phenylindole.

10

ZHAO ET AL.


Fig. 7. Knockdown of TET3 alleviates SAON in rats. (A) SAON was induced in SD rats, and in vivo TET3 siRNA transfection was performed as indicated. (B) Western blot analysis of TET1, TET2, and TET3 in the femoral head tissues of rats. A total of 50 mg protein was loaded. (C) Dot blot analysis of 5hmC in the femoral head tissues of rats. A total of 500 ng DNA was loaded. Left panel: dot blot of 5hmC; Right panel: methylene blue staining. (D) Representative H&E staining of trabeculae at the proximal epiphysis of the femoral head. Left panel: Scale bar ¼ 500 mm; Right panel: magnification of boxed area in the left panel, Scale bar ¼ 50 mm. (E, F) TUNEL analysis of osteocyte apoptosis in trabeculae at the proximal epiphysis of the femoral head (red). Data are represented as the mean SD from three independent biological replicates. Scale bar ¼ 30 mm. (G) Representative mCT scan of the femoral head. (H) Quantification of trabecular bone structure within the ROI in G. Trabecular BV/TV, BS/BV, trabecular thickness, trabecular number, and trabecular separation were calculated. (I) Representative MRI T1WI and T2WI scans of the femoral head. The boxed areas indicate different signal in T1WI and T2WI scan. p < 0.05, p < 0.01.



11

Fig. 8. Schematic illustration of the effect of the TET3-5hmC-Akt pathway on SAON pathogenesis. These observations led us to hypothesize that GC-induced TET3-5hmC change could upregulate PTEN expression and downregulate PI3K expression in addition to regulating other target genes to inhibit the Akt pathway and promote osteocyte apoptosis in SAON.

when the pharmacological activation of the Akt pathway abrogated Dex-induced osteocyte apoptosis. Significantly, we further found alterations in Akt signaling in clinical SAON samples. The modulation of Akt by GCs is a cell autonomous mechanism of Wnt/b-catenin antagonism that contributes to the adverse effects of excess GCs. Akt attenuation promotes the binding of b-catenin to forkhead box Os (FoxOs) and inhibition of b-catenin/TCF (T cell factor) activity.(44) PTH counteracts the adverse effect of GCs by abrogating the suppressive effect of GCs on Akt phosphorylation and the Wnt pathway.(47) Moreover, inhibition of Akt-mTORC1 signaling by GCs induces autophagy, resulting in connexin 43 degradation and subsequently impairing osteocyte cell-cell communication.(20) Interestingly, knockdown of TET3 in vitro can block the Akt activity inhibition induced by Dex. Thus, our data for the first time indicate that Dex-induced Akt pathway suppression is mediated by TET3. In addition to Akt pathway-associated genes, GC-induced TET3-5hmC changes regulate the expression of a large number of genes, such as Notch, Dkk1, and Dkk4 (data not shown). Notch signaling plays a critical role in osteoblast cell fate and function, and endothelial Notch activity promotes angiogenesis and osteogenesis.(48) Dkk1 is involved in the inhibition of the Wnt signaling pathway, and its expression is associated with osteocyte apoptosis in SAON patients.(49) Bose and colleagues(33) reported that TET3 mediates stable GC-induced alterations in DNA methylation and Dkk1 expression in neural progenitors. This indicates that not only the Akt pathway but also other target genes may mediate the functional effect of GC-induced TET3-5hmC change and could be further explored.

12

ZHAO ET AL.

Then, we found that knockdown of TET3 with specific siRNA transfection, which has been used in many studies, could abrogate SAON in vivo.(2,25) In another SAON rabbit model, the blockade of Src with specific siRNA was used as a novel therapeutic strategy to prevent destructive repair.(2) In our study, the effective knockdown of TET3 in vitro confirmed the specific mouse TET3 siRNA sequence. Considering the fact that we used rats to establish an SAON model, a Basic Local Alignment Search Tool (BLAST) search was performed in GenBank (http://www.ncbi.nlm.nih.gov/ genbank/) to validate that the sequence was conserved and also specific for rats. Meanwhile, the efficiency of TET3 knockdown in vivo was confirmed with a Western blot assay of the entire femoral head bone tissue. Furthermore, with the aim of exploring whether knockdown of TET3 could prevent the development of SAON, we performed TET3 siRNA injection on the first and 15th day of SAON induction. Interestingly, knockdown of TET3 in vivo prevented GC-induced osteocyte apoptosis, histopathology changes, trabeculae decline, and high-intensity T2WI MRI signals in rat femoral heads. Unfortunately, our in vivo transfection method could not ensure the specific knockdown of TET3 in osteocytes. Liang and colleagues(50) developed CH6 aptamer-functionalized lipid nanoparticles (LNPs) to encapsulate siRNA and boosted in vivo osteoblastspecific gene silencing. To understand the exact function of TET3 in osteocytes, further osteocyte-specific deletion of TET3 with a corresponding aptamer delivery system is warranted. In conclusion, the current results indicate that TET3-5hmC-Akt signaling is critical for GC-induced osteocyte apoptosis and SAON and uncover a new mechanism of SAON. This finding indicates that inhibition of TET3 expression in osteocytes might be a new strategy to treat SAON early.

Disclosures All authors state that they have no conflicts of interest.

Acknowledgments This work was supported by grants from the National Natural Science Foundation of China (No. 81572154 and 81272801), the Scientific and Technological Project of Tianjin Public Health Bureau (No. 2014KY31), and the Health Care Key Project of Tianjin (No. 14KG123). We are grateful to Prof. Lynda Bonewald for the kind gift of MLO-Y4 cells. Authors’ roles: Study design: XLM, JXM, JZ, and WLJ. Study conduct: JZ, LS, BL, LYX, MJK, LF, and YM. Data collection: JZ, BL, LYX, MJK, YW, BCD, and LF. Data analysis: JZ, YW, GSX, HHB, YM, and WLJ. Drafting manuscript: JZ and JXM. Revising manuscript content: JZ, JXM, XLM, and WLJ. Approving final version of the manuscript: All. JZ, JXM, and XLM are responsible for the integrity of the data analysis.

References 1. Weinstein RS. Glucocorticoid-induced osteonecrosis. Endocrine. 2012;41(2):183–90. 2. Zheng LZ, Cao HJ, Chen SH, et al. Blockage of Src by specific siRNA as a novel therapeutic strategy to prevent destructive repair in steroidassociated osteonecrosis in rabbits. J Bone Miner Res. 2015;30(11): 2044–57.


3. Strehl C, Bijlsma JW, de Wit M, et al. Defining conditions where longterm glucocorticoid treatment has an acceptably low level of harm to facilitate implementation of existing recommendations: viewpoints from an EULAR task force. Ann Rheum Dis. 2016;75(6):952–7. 4. Sun Z, Yang S, Ye S, et al. Aberrant CpG islands’ hypermethylation of ABCB1 in mesenchymal stem cells of patients with steroidassociated osteonecrosis. J Rheumatol. 2013;40(11):1913–20. 5. Lee YK, Ha YC, Cho YJ, et al. Does zoledronate prevent femoral head collapse from osteonecrosis? A prospective, randomized, openlabel, multicenter study. J Bone Joint Surg Am. 2015;97(14):1142–8.

25. Li CJ, Cheng P, Liang MK, et al. MicroRNA-188 regulates age-related switch between osteoblast and adipocyte differentiation. J Clin Invest. 2015;125(4):1509–22. 26. Yan W, Fang Z, Yang Q, et al. SirT1 mediates hyperbaric oxygen preconditioning-induced ischemic tolerance in rat brain. J Cereb Blood Flow Metab. 2013;33(3):396–406. 27. Yang L, Boyd K, Kaste SC, Kamdem Kamdem L, Rahija RJ, Relling MV. A mouse model for glucocorticoid-induced osteonecrosis: effect of a steroid holiday. J Orthop Res. 2009;27(2):169–75.

10. Sun ZB, Wang JW, Xiao H, et al. Icariin may benefit the mesenchymal stem cells of patients with steroid-associated osteonecrosis by ABCB1-promoter demethylation: a preliminary study. Osteoporos Int. 2015;26(1):187–97.

28. Liu C, Janke LJ, Kawedia JD, et al. Asparaginase potentiates glucocorticoid-induced osteonecrosis in a mouse model. PLoS One. 2016;11(3):e0151433. 29. Dong Y, Li Y, Huang C, Gao K, Weng X. Systemic application of teriparatide for steroid induced osteonecrosis in a rat model. BMC Musculoskelet Disord. 2015;16:163. 30. Bouxsein ML, Boyd SK, Christiansen BA, Guldberg RE, Jepsen KJ, Muller R. Guidelines for assessment of bone microstructure in rodents using micro-computed tomography. J Bone Miner Res. 2010;25(7):1468–86. 31. Ma XL, Liu ZP, Ma JX, Han C, Zang JC. Dynamic expression of Runx2, Osterix and AJ18 in the femoral head of steroidinduced osteonecrosis in rats. Orthopaedic surgery. 2010;2(4): 278–84. 32. Jilka RL, Noble B, Weinstein RS. Osteocyte apoptosis. Bone. 2013; 54(2):264–71.

11. Tahiliani M, Koh KP, Shen Y, et al. Conversion of 5-methylcytosine to 5-hydroxymethylcytosine in mammalian DNA by MLL partner TET1. Science. 2009;324(5929):930–5.

33. Bose R, Spulber S, Kilian P, et al. Tet3 mediates stable glucocorticoidinduced alterations in DNA methylation and Dnmt3a/Dkk1 expression in neural progenitors. Cell Death Dis. 2015;6:e1793.

12. Taylor SE, Li YH, Smeriglio P, Rath M, Wong WH, Bhutani N. Stable 5-hydroxymethylcytosine (5hmC) acquisition marks gene activation during chondrogenic differentiation. J Bone Miner Res. 2016;31(3): 524–34.

34. Rasmussen KD, Helin K. Role of TET enzymes in DNA methylation, development, and cancer. Genes Dev. 2016;30(7):733–50. 35. Ito S, D’Alessio AC, Taranova OV, Hong K, Sowers LC, Zhang Y. Role of Tet proteins in 5mC to 5hmC conversion, ES-cell self-renewal and inner cell mass specification. Nature. 2010;466(7310):1129–33. 36. Fu HL, Ma Y, Lu LG, et al. TET1 exerts its tumor suppressor function by interacting with p53-EZH2 pathway in gastric cancer. J Biomed Nanotechnol. 2014;10(7):1217–30.

6. Weinstein RS. Glucocorticoid-induced osteoporosis and osteonecrosis. Endocrinol Metab Clin North Am. 2012;41(3):595–611. 7. Kerachian MA, Seguin C, Harvey EJ. Glucocorticoids in osteonecrosis of the femoral head: a new understanding of the mechanisms of action. J Steroid Biochem Mol Biol. 2009;114(3–5):121–8. 8. Weinstein RS. Clinical practice. Glucocorticoid-induced bone disease. N Engl J Med. 2011;365(1):62–70. 9. Minor EA, Court BL, Young JI, Wang G. Ascorbate induces ten-eleven translocation (Tet) methylcytosine dioxygenase-mediated generation of 5-hydroxymethylcytosine. J Biol Chem. 2013;288(19): 13669–74.

13. Song CX, Szulwach KE, Dai Q, et al. Genome-wide profiling of 5-formylcytosine reveals its roles in epigenetic priming. Cell. 2013; 153(3):678–91. 14. Sun W, Zang L, Shu Q, Li X. From development to diseases: the role of 5hmC in brain. Genomics. 2014;104(5):347–51. 15. Frandsen PA, Andersen E, Madsen F, Skjodt T. Garden’s classification of femoral neck fractures. An assessment of inter-observer variation. J Bone Joint Surg Br. 1988;70(4):588–90. 16. Ficat RP. Idiopathic bone necrosis of the femoral head. Early diagnosis and treatment. J Bone Joint Surg Br. 1985;67(1):3–9. 17. Kim T, Hong JM, Lee J, et al. Promoter polymorphisms of the vascular endothelial growth factor gene is associated with an osteonecrosis of the femoral head in the Korean population. Osteoarthritis Cartilage. 2008;16(3):287–91. 18. Kitase Y, Barragan L, Qing H, et al. Mechanical induction of PGE2 in osteocytes blocks glucocorticoid-induced apoptosis through both the beta-catenin and PKA pathways. J Bone Miner Res. 2010;25(12): 2657–68. 19. Sato AY, Tu X, McAndrews KA, Plotkin LI, Bellido T. Prevention of glucocorticoid induced-apoptosis of osteoblasts and osteocytes by protecting against endoplasmic reticulum (ER) stress in vitro and in vivo in female mice. Bone. 2014;73C:60–8. 20. Gao J, Cheng TS, Qin A, et al. Glucocorticoid impairs cell-cell communication by autophagy-mediated degradation of connexin 43 in osteocytes. Oncotarget. 2016;7(19):26966–78. 21. Zhao X, Dai J, Ma Y, et al. Dynamics of ten-eleven translocation hydroxylase family proteins and 5-hydroxymethylcytosine in oligodendrocyte differentiation. Glia. 2014;62(6):914–26. 22. Wang JH, Chen WL, Li JM, et al. Prognostic significance of 2-hydroxyglutarate levels in acute myeloid leukemia in China. Proc Natl Acad Sci U S A. 2013;110(42):17017–22. 23. Huang da W, Sherman BT, Lempicki RA. Systematic and integrative analysis of large gene lists using DAVID bioinformatics resources. Nat Protoc. 2009;4(1):44–57. 24. Han N, Yan Z, Guo CA, et al. Effects of p-glycoprotein on steroidinduced osteonecrosis of the femoral head. Calcif Tissue Int. 2010; 87(3):246–53.


37. Mi Y, Gao X, Dai J, Ma Y, Xu L, Jin W. A novel function of TET2 in CNS: sustaining neuronal survival. Int J Mol Sci. 2015;16(9): 21846–57. 38. Xu Y, Xu C, Kato A, et al. Tet3 CXXC domain and dioxygenase activity cooperatively regulate key genes for Xenopus eye and neural development. Cell. 2012;151(6):1200–13. 39. Hu L, Lu J, Cheng J, et al. Structural insight into substrate preference for TET-mediated oxidation. Nature. 2015;527(7576):118–22. 40. Gao J, Ma Y, Fu HL, et al. Non-catalytic roles for TET1 protein negatively regulating neuronal differentiation through srGAP3 in neuroblastoma cells. Protein Cell. 2016;7(5):351–61. 41. Surjit M, Ganti KP, Mukherji A, et al. Widespread negative response elements mediate direct repression by agonist-liganded glucocorticoid receptor. Cell. 2011;145(2):224–41. 42. Moutsatsou P, Kassi E, Papavassiliou AG. Glucocorticoid receptor signaling in bone cells. Trends Mol Med. 2012;18(6):348–59. 43. Plotkin LI, Manolagas SC, Bellido T. Glucocorticoids induce osteocyte apoptosis by blocking focal adhesion kinase-mediated survival. Evidence for inside-out signaling leading to anoikis. J Biol Chem. 2007;282(33):24120–30. 44. Almeida M, Han L, Ambrogini E, Weinstein RS, Manolagas SC. Glucocorticoids and tumor necrosis factor alpha increase oxidative stress and suppress Wnt protein signaling in osteoblasts. J Biol Chem. 2011;286(52):44326–35. 45. Plotkin LI, Mathov I, Aguirre JI, Parfitt AM, Manolagas SC, Bellido T. Mechanical stimulation prevents osteocyte apoptosis: requirement of integrins, Src kinases, and ERKs. Am J Physiol Cell Physiol. 2005;289(3):C633–43. 46. Wang Z, Guo J. Mechanical induction of BMP-7 in osteocyte blocks glucocorticoid-induced apoptosis through PI3K/AKT/GSK3beta pathway. Cell Biochem Biophys. 2013;67(2):567–74. 47. Weinstein RS, Jilka RL, Almeida M, Roberson PK, Manolagas SC. Intermittent parathyroid hormone administration counteracts the


13

adverse effects of glucocorticoids on osteoblast and osteocyte viability, bone formation, and strength in mice. Endocrinology. 2010;151(6):2641–9.

49. Ko JY, Wang FS, Wang CJ, Wong T, Chou WY, Tseng SL. Increased Dickkopf-1 expression accelerates bone cell apoptosis in femoral head osteonecrosis. Bone. 2010;46(3):584–91.

48. Ramasamy SK, Kusumbe AP, Wang L, Adams RH. Endothelial Notch activity promotes angiogenesis and osteogenesis in bone. Nature. 2014;507(7492):376–80.

50. Liang C, Guo B, Wu H, et al. Aptamer-functionalized lipid nanoparticles targeting osteoblasts as a novel RNA interferencebased bone anabolic strategy. Nat Med. 2015;21(3):288–94.

14

ZHAO ET AL.
