Advanced glycation end products impair bone marrow mesenchymal stem cells osteogenesis in periodontitis with diabetes via FTO-mediated N⁶-methyladenosine modification of sclerostin

The animal experiments were complied with the ARRIVE (Animal Research: Reporting of In Vivo Experiments) guidelines and approved by the Ethics Committee of the College of Stomatology, Chongqing Medical University. 6-week-old male C57BL/6 mice were purchased from the SJA laboratory animal CO. LTD (Hunan, China) and were housed in a specific pathogen-free condition: room temperature 22 ± 2 °C, 12 h light/dark cycle and water ad libitum. 36 mice were randomly divided into two groups (n = 18/group) and fed with standard diet or high fat/high glucose diet (HFHG, D12492, Research Diets, NJ, USA) respectively. After 8 weeks, HFHG-fed mice weighed up to 30–35 g and were intraperitoneally injected with streptozotocin (STZ, 35 mg/kg, Sigma-Aldrich, MO, USA) for 3 consecutive days, while mice on the standard diet were given citrate buffer. 7 days after the last injection and then once a week after that, blood glucose levels were measured. T2DM was diagnosed in mice with fasting blood glucose levels higher than 16.7 mM and clinical symptoms (increased food and water consumption, increased urine output and weight loss).

Periodontitis was induced in Control and T2DM groups as described previously after blood glucose had been stable for 2 weeks [19]. The primary result was the distance of cemento-enamel junction (CEJ)-alveolar bone crest (ABC). Significant intergroup differences of 0.25 mm (standard deviation [SD] = 0.05 mm) in control and 0.45 mm (SD = 0.2 mm) in periodontitis were determined on the basis of published study [19]. With an 80% power, α of 0.05 and 2-tailed test, 6 mice per group were recommended. Totally 36 mice were randomly divided into four groups: control, DM, experimental periodontitis (EP), and diabetes-associated periodontitis (DP) (n = 9/group). The 5–0 silk ligature was bound around the maxillary right first molar and remained in place for 14 days. The left first molar was not ligated as self-control for bone loss measurement. 7 days after the ligature was removed, the mice were euthanized and the maxillae were obtained for micro-computed tomography (micro-CT) and histological analysis. The maxillae were trimmed and fixed in 4% polyformaldehyde buffer for 48 h and scanned by micro-CT. After being decalcified with 20% EDTA at 4℃ for 30 days until the bone could be readily pierced, the specimens were routinely dehydrated and paraffin embedded for histochemical analysis. For micro-CT and histological measurement, the interproximal area between the maxillary first and second molars was selected as the region of interest (ROI).

BMSCs were prepared according to previously published protocols [20]. The mandible was separated from male 6-week-old mice. Whole bone marrow cells were flushed out with α-MEM (C12571500BT, Gibco, NY, USA), supplemented with 1% penicillin/streptomycin and 10% FBS (BBP5, Moregate, New Zealand), and cultured in the incubator. The medium was changed every 2–3 days until the cells reach at 80% confluence for 7 days approximately. The cells were seeded into the plates at ratio of 1:3. The isolated cells were characterized by flow cytometric analysis of specific surface antigens, including CD29, CD44, CD105 and CD45. BMSCs were cultured in osteogenic or adipogenic induction medium for 21 days to detect the differentiation potential. The osteogenic induction medium was supplemented with 7% FBS, 1% penicillin/streptomycin, 10 mM glycerophosphate, 50 mg/l ascorbic acid and 100 nM dexamethasone. The adipogenic induction medium was supplemented with 10% FBS, 1 μM dexamethasone, 0.5 mM 3-isobutyl-1-methylxanthine, 0.1 M indomethacin and 10 μg/ml insulin.

The mouse pre-osteoblast line MC3T3 cells were purchased from Chinese Academy of Sciences (Shanghai, China). The cells were cultured in α-MEM, supplemented with 1% penicillin/streptomycin and 10% FBS. Cells were cultured with induction medium for osteogenic differentiation when they achieved 70–80% confluence.

BMSCs were exposed to AGEs (bs-1158P, Bioss, Beijing, China) at various amounts (0, 50, 100 and 150 μg/ml) for a period of 3 to 21 days. The induction medium was replaced every other day.

Three siRNAs targeting sclerostin (SOST, Gene ID: 74499) (siSOST) and negative control RNAs (siNC) were synthesized by Sangon (Shanghai, China). At 16–18 h of seeding, the BMSCs approximately at 80% confluence were transient transfected with siRNAs by Lipofectamine3000 (L3000015, Invitrogen, CA, USA) according to the manufacturer’s protocols.

The LV-Fto-RNAi (LV-shFTO) and LV-negative control (LV-shNC) were constructed by Genechem (Shanghai, China). BMSCs reached 30% confluent were infected with lentivirals by HitransG with MOL 10 according to the manufacturer’s instructions.

MC3T3 cells were chosen to establish stable FTO knockdown models, that were infected with LV-shFTO (Mol = 5) and were selected using puromycin (2 μg/ml). Real-Time quantitative Polymerase Chain Reaction (RT-qPCR) and Western blot (WB) were used to identify the efficiency of FTO knockdown. The primer sequences for PCR are provided in Additional file 1. The targeted sequences for siRNAs and shFTO are listed in Additional file 2.

Full details of the materials and methods are provided in Additional file 3.

The mice were fed with high fat/high glucose (HFHG) diet and their body weight reached 30–35 g at 8 weeks. They developed symptoms one week after receiving STZ intraperitoneally injection. The increased fasting blood glucose and abnormal glucose tolerance indicated that T2DM was established (Fig. 1A, B). Micro-computed tomography (CT) scan showed that the CEJ-ABC distance in both experimental periodontitis and diabetes-associated periodontitis groups were increased, and the alveolar bone loss in diabetes-associated periodontitis group was significant (Fig. 1C, E). Bone volume/tissue volume (BV/TV) and trabecular thickness (Tb.Th) of the DM group were lower than those in the control group, but the CEJ-ABC distance did not change significantly (Fig. 1D, F). The HE staining of the periodontal tissue showed that the gingival epithelium of control and diabetes groups were complete, no alveolar bone absorption was found, but the bone trabeculae in DM group were sparse. With regard to experimental periodontitis and diabetes-associated periodontitis groups, the integrity of the gingival epithelium was destroyed, the combined epithelium proliferated to the root, a large number of inflammatory cells infiltrated, and the alveolar bone was obviously absorbed. The destruction of periodontal tissue in diabetes-associated periodontitis group was more serious than that in experimental periodontitis group (Fig. 1G). The FTO expressions in alveolar bone were examined by immunofluorescence (IFC). Compared with that in control group, the diabetes-associated periodontitis group showed the largest increase in FTO expression, followed by that in the diabetes group (Fig. 1H).

Considering previous results suggesting that diabetes impaired the microstructure and reduced the repair and regeneration ability of alveolar bone, in order to dissect the m⁶A methylation mechanism of alveolar bone in diabetes, we conducted MeRIP-Seq to map the m⁶A modification and identify the critical m⁶A targets. The m⁶A-seq analysis identified 20611 m⁶A peaks representing 10469 gene transcripts in control and 12476 m⁶A peaks representing 7956 gene transcripts in DM, of which 9858 m⁶A peaks corresponding to 7224 gene transcripts were common between the two groups (Fig. 2A, B). The consensus sequence RGACU was significantly enriched in the m⁶A sites in both groups, which was consistent with previous research and indicated the sequence conservation of m⁶A motif (Fig. 2C). The accumulation of m⁶A peaks in different regions of mRNA was showed by density curve. The ordinate represented the proportion of peaks in the region at the corresponding position in relation to all peaks. The m⁶A peaks of two groups were particularly abundant near the end of coding sequences (CDS) and the beginning of 3ʹ untranslated region (3ʹUTR) (Fig. 2D). The m⁶A peaks distribution patterns on the functional region of mRNA were further analyzed based on the m⁶A-seq results. The pie chart showed the peaks distribution in 3'UTR region accounted for the largest proportion, and the histogram revealed the peaks were mostly enriched in the stop-codon. Meanwhile, similar patterns of m⁶A peaks distribution were found in control and DM groups (Fig. 2F). As Fig. 2E displayed, most genes have only one m⁶A modified peak. Especially in DM group, the proportion of genes containing unique peak modification increased than control.

Volcano plot showed that DM had 310 differentially hypermethylated peaks and 2153 differentially hypomethylated peaks compared with control group (Fold change > 2 and p < 0.05) (Fig. 3A). Bean plot showed that the overall RPM distribution of the differential m⁶A peaks in control (blue part) and DM (red part) group was almost consistent, and the median and average values of RPM in DM group were lower than control group (Fig. 3B). The differential m⁶A peaks based on RPM counts were exhibited in the form of heat map (Fig. 3C). To determine the potential biological significance of changes in m⁶A modification associated with DM, we conducted GO and KEGG analysis of differentially methylated genes. The results of GO analysis revealed the differential m⁶A modified transcripts were particularly enriched in “cellular macromolecule metabolic process”, “posttranscriptional regulation of gene expression” and “nucleic acid binding” (Fig. 3D). Further, with respected to the KEGG analysis of genes with differential m⁶A peaks, we found that the genes were enriched in “Signaling pathways regulating pluripotency of stem cells” and “AGE-RAGE signaling pathway in diabetic complications”. The enrichment of PI3K-Akt, Wnt signaling pathway indicated the correlation with osteogenesis and bone homeostasis (Fig. 3E).

Characterization of cultured BMSCs was identified by flow cytometry and differentiation induction (Additional file 4). The serum AGEs levels in mice were quantified and significantly increased in the DM mice compared to control (Fig. 4A). To explore the influence of diabetes microenvironment on periodontal tissue, different concentrations of AGEs were used to stimulate BMSCs. Cell counting kit-8 (CCK8) suggested that various concentrations of AGEs had no significant influence on cell growth within 72 h (Fig. 4B). TdT-mediated dUTP nick-end labeling (TUNEL) staining showed that AGEs (150 μg/ml) did not significantly affect the apoptosis of BMSCs (Fig. 4C). The results of Real-Time quantitative Polymerase Chain Reaction (RT-qPCR) showed that the mRNA expression of RUNX family transcription factor 2 (Runx2) and bone gamma-carboxyglutamate protein (Bglap) were inhibited at different degrees, while the expression of sclerostin (SOST) and dickkopf WNT signaling pathway inhibitor 1 (Dkk1) were up-regulated, suggesting that AGEs might affect the osteogenic differentiation of BMSCs, and 150 μg/ml was selected as the working concentration for subsequent experiments (Fig. 4D). Alkaline phosphatase (ALP) staining and ALP activity detection suggested that AGEs inhibited the synthesis and activity of ALP (Fig. 4E). Alizarin red S (ARS) staining showed that AGEs inhibited the calcium deposition of BMSCs (Fig. 4F). Dot blot assay presented the m⁶A methylation level of total RNA in BMSCs stimulated by AGEs decreased (Fig. 4G). The expression of the m⁶A-methylation related enzymes were verified by qPCR, METTL3, METTL14 and ALKBH5 were down-regulated, while the expression of FTO was up-regulated (Fig. 4I). WB verified the decrease of Runx2, Bglap and the increase of SOST and FTO in AGEs treatment group (Fig. 4H), which suggested that FTO might have a regulatory function in the process of AGEs prevented BMSCs from differentiating into osteoblasts. The uncropped blots of WB are provided in Additional file 5.

To observe the role of FTO-mediated m⁶A modification in AGEs impairing the osteogenic differentiation of BMSCs, lentivirus-FTO-shRNA-PURO (LV-shFTO) were constructed to knockdown the expression of FTO. Immunofluorescence, qPCR and WB confirmed the knockdown efficiency of LV-shFTO, by which effectively reduced the upregulation of FTO caused by AGEs stimulation (Fig. 5A–C). As expected, dot blot showed that FTO knockdown increased the total m⁶A level in BMSCs, confirming the biological function of demethylase FTO (Fig. 5D). ALP activity and ALP, ARS staining showed that FTO knockdown reduced the inhibition of bone formation potential of BMSCs stimulated by AGEs (Fig. 5E). The qPCR and WB also confirmed that FTO knockdown reversed the down-regulation of Runx2 and Bglap expression caused by AGEs, and inhibited the upregulation of SOST induced by AGEs (Fig. 5F, G). Double-label immunofluorescence proved that AGEs increased the expression of intracellular FTO and SOST, but FTO knockdown could reduce the level of SOST, suggesting that FTO knockdown might partially alleviated the inhibition of osteogenesis of BMSCs induced by AGEs (Fig. 5H).

To clarify the role of SOST in AGEs inhibiting osteogenic differentiation of BMSCs, siRNAs were used to silence the SOST expression. The siRNAs targeting SOST (siSOST) were verified according to qPCR and used for subsequent experiments (Fig. 6A). RT-qPCR and WB showed that SOST interference increased the expression of Runx2 and Bglap in BMSCs of control. In BMSCs treated with AGEs, siSOST rescued the expression of Runx2, a marker gene of early osteogenesis, but had limited effect on the recovery of Bglap, a marker gene of mineralization (Fig. 6B, E). ALP activity assay, ALP and ARS staining exhibited that SOST interference improved the alkaline phosphatase activity and mineralized deposition of BMSCs. In BMSCs treated with AGEs, SOST siRNAs partially restored the compromised osteogenic potential (Fig. 6C). Double-label immunofluorescence displayed that AGEs increased the intracellular SOST and decreased Runx2, siSOST reduced the SOST expression and alleviated the suppression of Runx2 caused by AGEs (Fig. 6D). Since SOST is an inhibitor of Wnt signaling and its expression was induced by AGEs, we next used TOPFlash luciferase assay to detect the SOST interference or FTO knockdown on the impact of β-catenin activation. Data suggested that AGEs suppressed TOPFlash/Renilla activity in negative control groups and this activity was increased in presence of siSOST, and shFTO attenuated the suppressive effect of AGEs exposure (Fig. 6F). WB was performed to verify the effect of FTO knockdown and SOST interference on Wnt pathway respectively. It was found that AGEs inhibited GSK phosphorylation and reduced β-catenin level, while FTO-knockdown and SOST interference partially relieved inhibition of Wnt pathway induced by AGEs (Fig. 6G, H).

To elucidate the underlying mechanism by which FTO regulates the osteogenic differentiation of BMSCs, the enrichment of m⁶A peaks in SOST transcripts from MeRIP-seq data was visualized by Integrative Genomics Viewer (IGV). Notably, the m⁶A modification in SOST was markedly decreased in alveolar bone of diabetes (Fig. 7A). Bioinformatics analysis was performed to investigate the interaction between m⁶A modification and demethylase FTO to the SOST transcript. The m⁶A motif and metagene analysis was detected using RMBase V2.0 (http://​rna.​sysu.​edu.​cn/​rmbase/​). It revealed that the sites of m⁶A modification on SOST mRNA were GGACU and mainly situated in the CDS-3ʹUTR junction region, which was consistent with our IGV results (Fig. 7B). Then, potential m⁶A binding sites on SOST mRNA were further investigated using the online m⁶A site prediction tool SRAMP (http://​www.​cuilab.​cn/​sramp/​). SRAMP analysis revealed that two potential m⁶A sites with high prediction scores were located in the CDS-3ʹUTR junction of SOST. Next, we selected the sequence of the CDS-3ʹUTR junction as a template to design primers for m⁶A-RIP-qPCR (Fig. 7C). Methylated RNA immunoprecipitation qPCR (MeRIP-qPCR) was performed in MC3T3 cells stable infected with LV-shFTO and indicated that m⁶A-specific antibodies significantly enriched on SOST transcripts compared with the IgG, which verified the predicted m⁶A sites existed on the selected region. FTO knockdown significantly increased the m⁶A enrichment in SOST transcripts compared with the cells infected with LV-shNC, suggesting the m⁶A modification of SOST transcripts was regulated by FTO (Fig. 7D). To determine whether SOST is regulated by FTO and YTHDF2, RPISeq (http://​pridb.​gdcb.​iastate.​edu/​RPISeq) and PRIdictor (http://​www.​rnainter.​org/​PRIdictor/​) were used to predict the interaction probability and binding sites. RPISeq analysis showed that the binding probability of FTO and YTHDF2 protein to SOST transcript were more than 0.5, which was considered to be highly possible for direct regulation (Fig. 7F). PRIdictor analysis showed FTO had higher binding residues than YTHDF2 (559 vs. 77), and the protein-binding sites location in 3ʹUTR was surprisingly consistent (Fig. 7E). We concentrated on the predicted binding sites and created primers to ensure the anticipated binding sites were included in the target sequence. Subsequently, the RIP-qPCR identified SOST transcripts as FTO substrates, particularly under AGEs exposure condition (Fig. 7G). Furthermore, YTHDF2-RIP-qPCR was performed and found that the SOST transcripts were not significantly enriched with YTHDF2-antibody or IgG in shNC group. However, FTO-knockdown increased SOST transcripts enrich to YTHDF2 compared to IgG, but not significantly affected the SOST transcripts enrichment to YTHDF2 compared to shNC group (Fig. 7H). In addition, RNA stability assay demonstrated that the half-lives of the SOST mRNA in control, AGEs and AGEs with shFTO group were 5.99, 7.79 and 5.53 h respectively, suggesting that AGEs exposure retard the decay of SOST mRNA, while FTO knockdown reduced the stability of the mRNA (Fig. 7I). Moreover, the subcellular distribution assay revealed that FTO knockdown had no discernible impact on the distribution of SOST transcripts in the cytoplasm and nucleus, indicating that FTO knockdown had no significant effect on the nuclear export of SOST transcripts (Fig. 7J).