The T2DM-BMSCs obtained in our previous study (24) were cultured in MEM supplemented with 10% fetal bovine serum (Gibco, CA, USA) and 1% streptomycin/penicillin (Gibco). The 293T and L929 cells (mouse fibroblasts), purchased from the Cell Bank of the Chinese Academy of Sciences (Shanghai, China), were cultured in DMEM (Hyclone) supplemented with the same supplements. All cells were maintained at 37°C in a 5% humidified CO₂ atmosphere.

Total RNA of cells or bone tissue was extracted using a FastPure^® Cell/Tissue Total RNA Isolation Kit (Vazyme, Nanjing, China) and 1 μg RNA was used for cDNA synthesis by using a HiScript^® III RT SuperMix for qPCR (Vazyme). The ChamQ SYBR qPCR Master Mix (Vazyme) was used for qPCR. GAPDH was used as the internal control for mRNA and lncRNA expression. The MiPure Cell/Tissue miRNA Kit (Vazyme) was used to extract total miRNA. The miRNA 1st Strand cDNA Synthesis Kit (Vazyme) and MiRNA Universal SYBR qPCR Master Mix (Vazyme) were used for cDNA synthesis and qPCR, respectively. U6 was used as an internal control for miRNAs. All procedures were performed according to the manufacturer’s instructions. Relative expression level of gene was calculated using the 2^−ΔΔCt method. The primer sequences used for qPCR and miRNA reverse transcription are listed in Supplementary Table S1 .

Cells were lysed using RIPA lysis buffer (Beyotime, Shanghai, China), and protein concentration was quantified using a BCA kit (Beyotime). SDS-PAGE (10%) was used to isolate the denatured protein (20 μg), which was then transferred into the 0.22μm PVDF membrane. The membrane was then incubated with protein-free rapid blocking buffer (Servicebio, Wuhan, China) for 5 min at room temperature. After washing with TBST for 2 min, the membranes were incubated with corresponding antibody at 4°C overnight, followed by 1.5 h-incubation with the secondary antibody at room temperature. The bands were visualized using an ECL detection system (Beyotime). The antibodies: GAPDH, EHMT2/G9a, GLP, H3K9me1, H3K9me2, H3, RUNX2, BGLAP and COL1A1 were purchased from ABclonal Biotechnology Co., Ltd. (Wuhan, China), and diluted at a ratio of 1:1000 prior to utilisation.

Cells were seeded in 96-well plates at a density of 4000 cells/well. Different concentrations (0, 0.1, 0.5, 1, 5, 10, 25, 50, and 100 μM) of UNC0638 (Medchemexpress, Shanghai, China) were incubated with cells for 3 or 7 days respectively. Then, 5 mg/ml MTT (Sigma-Aldrich, Shanghai, China) was incubated with the cells for 3 h, crystal violet was dissolved in DMSO (Sigma-Aldrich), and the optical density was detected at 492 nm using a microplate reader.

After 7 days of osteogenic induction, as described in a previous study (24), the cells were fixed and stained with BCIP/NBT stain (Beyotime) for 1 h following the manufacturer’s protocols. An ALP detection kit (Beyotime) was used to detect ALP activity according to the manufacturer’s instructions.

The cDNA sequences of EHMT2, LINC00657, or IGFBP5 were amplified and subcloned into a pcDNA3.1 vector. An empty vector was used as the negative control (NC). Overexpression vectors for EHMT2, LINC00657, or IGFBP5 were transfected into the cells as previously described (24). miR-204-5p and NC mimics were synthesized by GenePharma (Shanghai, China), and transfected into cells using Lipofectamine 2000 (Invitrogen, Thermo Fisher Scientific) following the manufacturer’s instructions. After 48 h, cells were collected for qPCR analysis.

Cells overexpressing the LINC00657 overexpression vector or the NC vector were prepared, and total RNA was extracted using TRIzol (Invitrogen), then purified using an RNeasy Mini Kit (Qiagen, Germany). The samples were subjected to cDNA library construction and sequencing by Sinotech Genomics Co. Ltd. (Shanghai, China). The R package edgeR was used to analyze differential gene expression. |Fold change (FC)| value > 2 and p-value < 0.05 were used for screening differentially expressed genes (DEGs).

MiRNAs of LINC00657 were predicted using LncBase Predicted v.2 (36) and a score ≥ 0.8 was used as the filter. TargetScan (http://www.targetscan.org/vert_72/) was used to predict the miRNAs of candidate genes, and score percentile ≥ 80% was used as the threshold. The ceRNA network of LINC00657 was constructed using the Cytoscape software (version 3.5.1). GO analysis of biological processes, cellular components, and molecular functions was performed using the enriched R package.

The 3′-UTR of LINC00657-WT, LINC00657-MUT, IGFBP5-WT, and IGFBP5-MUT was amplified using PCR and inserted into the pmirGLO vector (Invitrogen, CA, USA). LINC00657-WT or LINC00657-MUT vectors, IGFBP5-WT or IGFBP5-MUT vectors, and miR-204-5p or NC mimics were co-transfected into 293T cells using Lipofectamine 3000 (Invitrogen). Forty-eight hours post-transfection, dual-luciferase reporter gene assays were performed to measure luciferase activity using a Dual-Luciferase Reporter Assay System kit (Promega, USA) according to the manufacturer’s instructions.

pAAV-U6-sh-LINC00657-EGFP, constructed by GenePharma (Shanghai, China), and packaging plasmids (pHelper and pAAV2, GenePharma) were co-transfected into 293T cells to produce recombinant AAV2 (AAV2-U6-sh-LINC00657-EGFP, AAV2/sh-LINC00657), as described in a previous study (37). A blank AAV (AAV2/sh-NC) was produced using the same method.

Twenty-five specific-pathogen-free male SD rats (220 ± 20 g), purchased from Shanghai Lab Animal Research Center (Shanghai, China), were randomly divided into a normal diet group (control, n=5) and high glucose and high fat (HGHF) diet group (n=20). Rats in the HGHF group received the HGHF diet for 12 weeks, followed by an intraperitoneal injection of STZ (Sigma-Aldrich) at a dose of 30 mg/kg (38). Animals with blood glucose levels >16.7 mmol/L for 3 days were considered T2DM models. Blood glucose levels were measured using a glucose meter (Omron, Shanghai, China). The control rats were fed a normal diet and intraperitoneally injected with normal saline. In addition, all rats were maintained under appropriate conditions (20–25°C with 65–80% humidity) with free access to water and food.

Next, the T2DM rats were randomly divided into four groups. Model group (n=5): intraperitoneal injection of normal saline; UNC0638 group (n=5): intraperitoneal injection of UNC0638 (5 mg/kg) twice a week for 12 weeks, according to previous studies (39, 40) with some modifications; UNC0638 + sh-NC group (n=5): intraperitoneal injection of UNC0638 (5 mg/kg) twice a week for 12 weeks, AAV2/sh-NC (3 × 10¹² vg/kg, 50 μl) (41); and UNC0638 + sh-LINC00657 group (n=5): intraperitoneal injection of UNC0638 (5 mg/kg) twice a week for 12 weeks, with intravenous injection of AAV2/sh-LINC00657 (3 × 10¹² vg/kg, 50 μl). All rats with T2DM were fed an HGHF diet for 12 weeks. At the end of the assay, all rats were anesthetized using intraperitoneal injection of chloral hydrate (400 mg/kg) and euthanized using cervical dislocation. Bilateral femurs were isolated for further analysis. This animal study was approved by the Ethics Committee of Zhoupu Hospital (ZPYYLL-018-02). All procedures were performed in accordance with the guidelines of the Care and Use of Laboratory Animals (Ministry of Science and Technology of China, 2006).

The blood of rats was collected, and the serum was obtained after centrifugation at 1,100×g for 10 min at 4°C. ALP and OCN levels were detected using a rat alkaline phosphatase ELISA kit (Elabscience, Wuhan, China) and osteocalcin ELISA kit (Elabscience), respectively, following the manufacturer’s instructions.

The femurs of rats were collected and fixed with 4% paraformaldehyde for 24 h, and the distal femur was scanned using an X-ray microtomography imaging system (SkyScan1276, Bruker, Germany). Trabecular bone volume (BV/TV; %), trabecular thickness (Tb.Th; mm), BMD (mg/cm²), structure model index (SMI), trabecular number (Tb.N; mm^–1), bone surface/volume ratio (BS/BV; mm^–1), and trabecular separation (Tb.Sp; mm) were evaluated.

After fixation, decalcification, embedding, sectioning, and deparaffinage, the slices of distal femurs were blocked with 10% goat serum, and incubated with G9a (1:200, ABclonal), OPN (1:200, ABclonal), and COL1A1 (1:200, ABclonal) at 4°C overnight. The sections were then incubated with a secondary antibody (ABclonal) for 2 h at room temperature, followed by nuclear staining with hematoxylin (Servicebio). The IHC results were visualized using a microscope (Olympus, Tokyo, Japan).

The prepared slices of the distal femurs were stained with H&E (Servicebio) or Masson’s (Servicebio)dye, as previously described (42), and then visualized using a microscope (Olympus).

The experimental data were expressed as mean ± standard deviation and analyzed using SPSS 17.0 software (SPSS, Inc., Armonk, USA). The unpaired Student’s t-test was used to analyze the differences between two groups. One-way analysis of variance (ANOVA) with Tukey’s post hoc test was used to compare multiple groups. P<0.05 was considered statistically significant.

As shown in Figure 1A , the EHMT2 mRNA and G9a protein expressions were significantly enhanced in T2DM-BMSCs compared with those in the controls ( Figure 1B , p < 0.05). The protein levels of GLP and H3K9me1 exhibited minimal alterations, while H3K9me2 level demonstrated a notable increase in T2DM-BMSCs ( Figure 1B , p < 0.05). The alterations in EHMT2 and H3K9me2 expression may be related to the weaker osteogenic differentiation potential of T2DM-BMSCs. Next, the cytotoxicity of UNC0638, a G9a inhibitor, in BMSCs was evaluated. The MTT result indicated that, 0.1–0.5 μM of UNC0638 was non-cytotoxic on T2DM-BMSCs after treatment for 3 or 7 days, and the cell viability reduced by 40–80% when the dose was ≥ 25 μM ( Figure 1C ). For the normal cells (L929), the safe dose of UNC0638 was 0.1-1 μM ( Figure 1D ), higher than that observed in the T2DM-BMSCs. Then the safe doses (0.1 and 0.5 μM) were used for further detection. Treatment with 0.1 or 0.5 µM of UNC0638 gradually reduced H3K9me2 levels, without affecting G9a, GLP, or H3K9me1 ( Figure 1E ). As demonstrated in Figures 1F–I , treatment with UNC0638 resulted in a significant enhancement of RUNX2, COL1A1, and BGLAP expression and protein levels, as well as ALP activity and calcification levels. In addition, 0.5 μM of UNC0638 showed a better promotion and significance than that at 0.1 μM.

Previous transcriptome sequencing detected thousands of lncRNAs in T2DM-BMSCs and 88 significantly altered lncRNAs (FC > 2, p < 0.05) were screened, including 35 known and 53 unknown lncRNAs ( Supplementary Table S2 ). Figure 2A shows the expression profiles of 35 known lncRNAs in each cell sample. The top 10 lncRNAs were selected for qPCR validation. As shown in Figures 2B, C , six lncRNAs (LINC00278, LINC02848, LINC00844, B3GALT5-AS1, LINC00657, and LINC00852) were significantly altered in T2DM-BMSCs (p < 0.05), and UNC0638 treatment caused significant upregulation of LINC00278 and LINC00657 [also known as NORAD (43)] (p < 0.05). Therefore, UNC0638 reversed the downregulation of these two lncRNAs in T2DM-BMSCs.

Next, LINC00657, which has a higher sensitivity to UNC0638 than that of LINC00278, was selected to investigate its role in G9a-mediated osteogenic differentiation. qPCR detection confirmed that EHMT2 mRNA levels were significantly enhanced after T2DM-BMSCs transfection with the EHMT2 overexpression vector (p < 0.05) and overexpression of EHMT2 caused a decrease in LINC00657 ( Figure 2D ), consistent with the results regulated by UNC0638. LINC00657 expression was enhanced after T2DM-BMSCs were co-transfected with the LINC00657 overexpression vector ( Figure 2D ). In addition, overexpression of EHMT2 inhibited the mRNA and protein levels of RUNX2, COL1A1, and BGLAP, as well as ALP activity and calcification levels, whereas upregulation of LINC00657 partially restored osteogenic marker expression, ALP activity and mineralization capacity ( Figures 2E–H ). These results indicated that the upregulation of EHMT2 in T2DM-BMSCs were responsible for the weaker osteogenic potency, which was exerted through the downregulation of LINC00657.

To screen the miRNAs sponged by LINC00657, the miRNA of LINC00657 was predicted and 296 miRNAs were obtained. Next, we focused on the mRNAs regulated by LINC00657. T2DM-BMSCs overexpressing LINC00657 were confirmed using qPCR ( Figure 3A ) and 433 DEGs (FC > 2, P < 0.05) were screened using transcriptome sequencing ( Figure 3B ). Next, osteogenically relevant GO terms were screened, including 20 biological processes ( Figure 3C ) and 40 DEGs enriched in these processes ( Figure 3D , Supplementary Table S3 ). Next, the miRNAs of the top 10 DEGs were predicted and a ceRNA network of LINC00657 was constructed. As shown in Figure 3E , three shared miRNAs (miR-204-5p, miR-211-5p, and miR-144-3p) and two potential target genes (IGFBP5 and NOG) were identified. qPCR validation indicated that miR-204-5p expression was significantly reduced (p < 0.05), whereas IGFBP5 mRNA levels were increased by approximately five folds in the LINC00657-overexpression group ( Figures 3F,G ). Downregulation of miR-204-5p and upregulation of IGFBP5 indicated that they might act downstream of LINC00657.

Next, T2DM-BMSCs were co-transfected with miR-204-5p mimics to investigate their roles in osteogenic differentiation. The qPCR results indicated that co-transfection with miR-204-5p reversed its downregulation compared with the LINC00657 overexpression group ( Figure 3H ). A dual-luciferase reporter assay showed that the miR-204-5p mimic decreased the luciferase activity of cells transfected with the LINC00657 wild-type vector but not the mutant vector ( Figure 3I ), indicating that LINC00657 could sponge miR-204-5p. Additionally, co-transfection with miR-204-5p mimics inhibited the expression of osteogenic markers, including RUNX2, COL1A1, BGLAP, ALP, and calcification, compared with that of the LINC00657 overexpression group ( Figures 3J–M ).

Next, T2DM-BMSCs were transfected with miR-204-5p mimics or co-transfected with an IGFBP5 overexpression vector. The qPCR results indicated that miR-204-5p levels were enhanced and IGFBP5 levels were decreased in the miR-204-5p overexpression group compared with those in the NC group ( Figure 4A ). Further, co-transfection with the IGFBP5 overexpression vector caused a significant increase in IGFBP5 expression ( Figure 4A , p < 0.05), indicating a potential sponge relationship between miR-204-5p and IGFBP5. The results of the dual luciferase reporter assay showed that luciferase activity was significantly decreased in the IGFBP5-WT group (p < 0.05), whereas it did not change in the mutant group ( Figure 4B ). These results confirm the sponging relationship between the two RNAs. Furthermore, we found that the overexpression of miR-204-5p inhibited the osteogenic potential of T2DM-BMSCs, and co-overexpression of IGFBP5 reversed this inhibitory effect ( Figures 4C–F ).

The flow chart of animal assay was presented in Figure 5A. We established a rat model of DOP to investigate the in vivo effects and mechanisms of UNC0638 treatment. As shown in Figure 5B , the blood glucose level of rats in the DOP group was >16.7 mM, which was significantly higher than that of rats in the control group (p < 0.05). Serum ALP and OCN levels in the DOP group were reduced by approximately 50% compared with those in the control group ( Figures 5C, D ). Furthermore, the bone trabecular structure became fewer and sparser, based on micro-CT analysis ( Figure 5H ), and H&E and Masson staining ( Figure 6A ). Compared with the control group, BMD, BV/TV, Tb.Th, and Tb.N were significantly decreased, and BS/BV, Tb.Sp and SMI were increased in the DOP group ( Figure 5H , p < 0.05). In addition, OPN and COL1A1 levels decreased in the distal femur of DOP rats ( Figure 6C ). These changes indicated that the DOP model was successfully established.

Next, G9a and H3K9me2 expression was detected by IHC and enhanced in the distal femur of DOP rats ( Figure 5F ), consistent with the in vitro results. The qPCR results indicated a significant decrease of LINC00657 and IGFBP5 expression, and an increase of miR-204-5p level, in the DOP group in comparison with the control group ( Figure 5E , p < 0.05). After 3 months of treatment with UNC0638, no significant alterations in body weight were observed among the three groups, compared with the DOP group ( Figure 5G ). Osteoporosis progression was significantly inhibited by UNC0638, including significantly increased levels of BMD, BV/TV, Tb.Th, and Tb.N, and decreased levels of BS/BV, Tb.Sp and SMI compared with those in the DOP group ( Figure 5H , p < 0.05). The bone trabecular structure, and OPN and COL1A1 levels in the distal femur increased after UNC0638 treatment ( Figures 5H , 6A, C ). Further, qPCR results showed that LINC00657 and IGFBP5 were notably enhanced, and miR-204-5p was reduced in the femoral head of DOP rats after treatment with UNC0638 ( Figure 6B , p < 0.05). This indicated that the in vivo anti-osteoporotic effect of UNC0638 was related to the upregulation of LINC00657 and its downstream molecules, miR-204-5p and IGFBP5.

DOP rats treated with UNC0638 were also intravenously injected with lentiviral sh-LINC00657 or sh-NC. The qPCR results confirmed that LINC00657 levels were lower in the UNC0638 + sh-LINC00657 group compared with that in the UNC0638+ sh-NC group ( Figure 6B ). Micro-CT, and H&E, and Masson staining results showed that knockdown of LINC00657 significantly inhibited the anti-osteoporosis effect of UNC0638, including a decrease in four bone indices (BMD, BV/TV, Tb.Th, and Tb.N) and an increase in three indices (BS/BV, Tb.Sp and SMI) ( Figure 5H , p < 0.05; and Figure 6A ). The IHC results indicated that OPN and COL1A1 were both decreased after LINC00657 knockdown ( Figure 6C ). Furthermore, downregulation of LINC00657 caused a significant increase in miR-204-5p and a decrease in IGFBP5 in the femoral head ( Figure 6B , p < 0.05). The above results indicate that the upregulation of LINC00657 was responsible for the anti-osteoporosis effect of UNC0638 in DOP rats, and UNC0638 treatment also regulated the expression of miR-204-5p and IGFBP5.

In summary, G9a inhibited the osteogenic potential of T2DM-BMSCs by regulating the LINC00657/miR-204-5p/IGFBP5 axis and UNC0638 reversed DOP by inhibiting G9a/LINC00657/miR-204-5p/IGFBP5 axis ( Figure 6D ).