m6A hypomethylation of DNMT3B regulated by ALKBH5 promotes intervertebral disc degeneration via E4F1 deficiency

NP tissues were obtained from 52 patients (29 females and 23 males; aged 55.5 ± 3.55 years; Grades IV (n = 35) and V (n = 17)) with degenerative disc diseases undergoing surgery. The control samples were taken from 43 patients (26 females and 17 males; aged 15.5 ± 1.68 years; Grade I (n = 23) and II (n = 20) undergoing surgery due to scoliosis or thoracolumbar fracture after informed consent was obtained. The specimen information was provided in Table S4. All specimens were obtained from lumbar discs. Randomly selected normal (n = 16) and degenerative (n = 16) samples were used to perform in vivo tissue analysis (5 normal and 5 degenerative samples for RNA Scop; 5 normal and 5 degenerative samples for IHC; 6 normal and 6 degenerative samples for western blot). The degenerative changes of the IVDs were evaluated according to the Pfirrmann grade of patients’ magnetic resonance images.8 This study protocol was approved by the Ethics Committee of Tongji Medical College, Huazhong University of Science and Technology (No. S341).

The NP tissues were cut into pieces after collection during surgery, and collagenase type II (Invitrogen, Carlsbad, CA, USA) was used to digest for 8 h at 37°C at the concentration of 0.2%. Then, cells from the digest were collected by centrifuging at 1200 rpm and then was cultured in DMEM (Gibco, Grand Island, NY, USA) supplied with 10% Fetal Bovine Serum （FBS) (Invitrogen), 1% penicillin–streptomycin (Sigma), 2 mM glutamine (Sigma), and 50 μg/ml l‐ascorbic acid (Sigma) at 37°C in 5% CO₂. When grown to confluence, 0.25% trypsin/1 mM EDTA was used to digest the cells for passing for expansion. Passage 2 cells were plated into experimental plates or bottles for following experiments. In some experiments, cells were stimulated by 20 ng/ml of TNF‐α (Peprotech, 300–01A, ROCKY HILL, New Jersey) in the culture medium for 48 h. Unstimulated cells were used as controls.

Samples were lysed by RIPA (Beyotime, P0013B, Shanghai, China) and the Micro BCA Protein Assay Kit (Beyotime, P0010S, Shanghai, China) was used to measure the protein content. Sodium dodecyl sulfate (SDS)‐polyacrylamide gel electrophoresis gels were used for electrophoresis and then transferred to polyvinylidene difluoride (PVDF) membranes. Then, blocking buffer was used to block for 1 h and following the specific primary antibodies (the primary antibodies used were listed in Table) were used to incubate with the membranes overnight at 4°C. Then, diluted horseradish peroxidase (HRP)‐conjugated Affinipure Goat Anti‐Rabbit IgG (SA00001‐2, Proteintech) and HRP‐conjugated Affinipure Goat Anti‐Mouse IgG (SA00001‐1, Proteintech) using antibody dilution were used to incubate the membranes. And GAPDH or H3 was used a loading protein control. At last, enhanced chemiluminescence reagents (Affinity, KF001, Nanjing, China) used to visualize the protein expression using the ChemiDoc MP Imaging System (Bio‐Rad, 12003154 Hercules, CA, USA). And ImageJ was used for semi‐quantification of the expression of proteins. S1

NPCs were isolated from specimens collected during surgery using collagenase. NPCs from four normal IVD samples were considered normal and NPCs from degenerated samples were considered senescent NPCs. Then TRIzol™ Reagent (Thermo Fisher, 15596026) was used to extract RNA. Then, library preparation and sequencing were conducted using Illumina HiSeq 2000 and analysis was carried out by Biocame. The NGS data of NPCs transcript in this study are available under the accession identifier GSE167931↗.

Cultured NPCs at passage 2 were treated with TNFα for 48h or not with three samples for each group, followed by RNA extraction using TRIzol™ Reagent (Thermo Fisher, 15596026). Then, Dynabeads® mRNA purification kit (Invitrogen) was used to isolated polyadenylated RNA from total RNA. On the basis of previously published protocols (https://doi.org/10.1038/nature11112↗.), we performed RNA fragmentation, Me‐RIP and library preparation. Sequencing was conducted using Illumina NovaSeq 6000 and analysis was carried out by Epibiotek. The raw data from the Me‐RIP‐Seq analysis of NPCs were deposited in the Gene Expression Omnibus database under the accession code GEO: GSE169484↗.

RNA scope was performed according to RNA scope Fluorescent Multiplex kit instructions (Advanced Cell Diagnostics, Hayward, CA). Single RNA molecules can be detected by way of its zz oligo pair design and DNA‐based amplification methods (https://doi.org/10.1016/j.jmoldx.2011.08.002↗). The DNMT3B probe was designed against nucleotides 1379‐2861. Slides were washed with 10% formamide in 2X SSC and staining with 4',6‐diamidino‐2‐phenylindole (DAPI). Then, we captured the images using microscope.

After corresponding treatment of TNFα or not, β‐galactosidase substrate C₁₂FDG (Fluorescein di‐B‐D‐galactopyranose) was used to incubate NPCs at the concentration of 33 mM in 2 ml medium for 2 h, pretreated with 100 nM bafilomycin A1 for 1 h at 37°C. After being washed with PBS, NP cells were fixed with 4% paraformaldehyde for 15 min at room temperature. Then, 0.1 g/ml DAPI (Beyotime, Shanghai, China) was used to co‐stain the nucleus. This was followed by visualization and capturing of the images under a microscope. The experiments were replicated three times.

EdU labelling was performed to examine the proliferation status of NPCs. NPCs were exposed to 25×10⁻⁶ M of 5‐ethynyl‐2′‐deoxyuridine (EdU, RiboBio, C10338↗, Guangzhou, China) for 2 h at 37°C and fixed in 4% paraformaldehyde. NPCs were then permeabilised using 0.5% Triton‐X‐100 and then reacted with Apollo488 for 30 min. Subsequently, Hoechst 33342 was used to stain the DNA contents of the cells for 30 min, and images were visualized and captured using a microscope (Olympus, BX53). EdU positive cells were analysed using Image J. The experiments were replicated three times.

The biotinylated DNA probes containing T7 and SP6 promoter complementary to DNMT3B (Table) were synthesized and dissolved in 500 μl of lysis buffer (0.5 M NaCl, 20 mM Tris‐HCl, pH 7.5, and 1 mM EDTA). Then, following steps of RNA‐pulldown experiments were performed according to the manufacturer's instructions of a MagCapture™ RNA Pull Down kit (Millipore Corporation, USA). The cell lysates were incubated with probe‐coated beads. The pull‐down mixture was then used for following western blot analysis. S2

RNA extraction was performed using PARIS™ kit (Invitrogen™: AM1921) and real‐time PCR were performed according to the protocol of instructions, GAPDH was used as endogenous control for the cytoplasmic RNA, while 18S RNA was selected as endogenous control for the nuclear RNA.

Data are presented as the mean ± SD of at least three independent experiments. GraphPad Prism 8 software (La Jolla, CA, USA) was used for statistical analysis. Two‐tailed unpaired Student's t‐test was used to measure the statistical significance of the difference. P < 0.05 was considered statistically significant while P > 0.05 was considered none significant (ns) (^#P > 0.05, *P < 0.05, **P < 0.01, ***P < 0.001 and ****P < 0.0001).

More Materials and Methods are available in Supporting information.

The relationship between disc degeneration and NP cellular senescence has been widely reported and to explore the aging status of NPCs in the degeneration process of IVDs, we verified the expression of aging‐associated markers p16 and p21 by immunohistochemistry (IHC) in NP tissues with various degrees of disc degeneration (Figure 1A). Furthermore, increased levels of represent senescence‐associated secretory phenotype (SASP) and senescence indicators were detected by western blot in NPCs from normal and degenerated discs, indicating NPC cellular senescence occurred during IVD degeneration (Figure 1B). To further explore the underlying pathological mechanism of NPC senescence, we isolated human NPCs from normal and degenerated discs to perform RNA sequencing, and gene set enrichment analysis (GSEA) results showed that degeneration of NP was accompanied with NPC senescence (Figure S1A). Subsequently, TNFα administration was performed to establish the degenerated in vitro model of NPCs. And the treatment of TNFα could induce senescence of NPCs according to increased SA‐β‐gal activity and decreased proportion of Ki67‐positive cells (Figure 1C). Meanwhile, western blot and immunofluorescence analysis of representative senescence markers verified the former results (Figure 1D, E and Figure S1B).

To study the role of m6A modification in IVD degeneration and NPC senescence, we examined the expression levels of methyltransferases and demethylases, the main regulators of m6A modification, in the NPCs by western blot and RT‐qPCR, and identified that ALKBH5, the core component of demethylase, was increased in senescent NPCs (Figure 1F, G and Figure S1C). Consistent with our findings, the expression of ALKBH5 was found significantly up‐regulated in the transcript sequencing data of NPCs from degenerated and normal discs, which was further confirmed by immunofluorescence staining of NP tissues (Figure S1D and Figure 1H). To investigate the functional roles of ALKBH5 and m6A modification in NPC senescence, we silenced ALKBH5 in NPCs using lentivirus‐based shRNA, and the knockdown efficiency was verified by western blot analysis and RT‐qPCR (Figure 1I). The higher level of m6A modification in NPCs was observed in ALKBH5 knockdown NPCs, though treated with TNFα, indicated by Elisa‐based m6A colorimetric assay (Figure 1J). We evaluated the aging‐related characteristics of NPCs after ALKBH5 silencing, and found expression of senescence markers and SA‐β‐gal activity were decreased compared with wild type NPCs treated by TNFα (Figure S1E and Figure 1K). Furthermore, overexpression of ALKBH5 in ALKBH5‐silenced NPCs attenuated senescent phenotype to some extent, while overexpression of mutant ALKBH5 disrupted with enzymatic activity by carrying the H204A mutation failed to rescue senescent phenotype34 (Figure 1L), indicating that the enzymatic activity of ALKBH5 is required for cellular senescence. Besides, we established a rat model of IVD degeneration using the needle puncture and injected into the IVD using a 29‐gauge fine needle with injection lentivirus containing shRNA against ALKBH5 or vector to testify the role of ALKBH5 in vivo. First, transcript expression analysis of ALKBH5 by RT‐qPCR in Rat NP tissue reflects the silencing efficiency (Figure S1F). The radiographic imaging analysis combined with histological staining indicated intervention of ALKBH5 could ameliorate the aging and degeneration of IVD (Figure 1M, N and Figure S1G, H). These results suggest ALKBH5‐mediated lower level of m6A modification contributes a lot to the senescence of NPCs and IVD degeneration.

Histone modification is another widely regulatory mechanism that could regulate gene expression by modulating the accessibility of chromatin, and epigenetic alteration of histone modification has been described as major contributors to cellular aging.35, 36 Increased histone modifications of H4K16ac, or H3K4me3, as well as decreased H3K9me3 or H3K27me3, constitute aging‐associated epigenetic marks.37 We reasoned ALKBH5 could be regulated in an epigenetic regulation manner by histone modification. To verify whether expression of ALKBH5 could be regulated by histone epigenetic modification, we used online tool (ENCODE) to predict above four kinds of histone modification of ALKBH5 promoter (Figure 2A), and ChIP‐qPCR was performed to identify their change levels. Results showed the H3K9me3 at the promoter changed most significantly with downregulation after treatment with TNFα (Figure 2B‐E). Data from transcriptome sequencing revealed, KMT1A and KDM4A, two transferases mediating H3K9me3 modification, significantly changed in NPCs (Figure 2F). And western blot of normal and degenerated tissues further confirmed above results, accompanied with a more significant change of KDM4A than KMT1A (Figure 2G, H). Additionally, slightly increased KMT1A functions as a methyltransferase, which would lead to the opposite direction of H3K9me3 modification. Therefore, KDM4A was chosen as the potential contributor to decrease H3K9me3 modification of ALKBH5, in accordance with the transcript sequencing data in senescent NPCs (Figure S2A). And then gain of function assays of KDM4A in NPCs further confirmed KDM4A was an effective contributor to H3K9me3 (Figure 2I, J), and DNA‐pulldown assay also revealed KDM4A bound more to the promoter region of ALKBH5 in senescent NPCs38 (Figure 2K, L). Further ChIP‐qPCR showed, compared with control NPCs, more KDM4A combined with ALKBH5 at the promoter region in senescent NPCs (Figure 2M). Loss of H3K9me3 correlates with the block of gene expression by inhibiting Pol II recruitment to the promoter sites. In normal NPCs, overexpression of demethylase KDM4A could down‐regulate H3K9me3 and promote Pol II recruitment and expression of ALKBH5, while inhibition of H3K9me3 demethylases using JIB‐04 in senescent NPCs leads to the opposite outcome (Figure 2N–P). Consistently, DNase I sensitivity assay also showed consistent change of chromatin accessibility of the ALKBH5 promoter in KDM4A‐overexpressed and TNFα‐induced senescent NPCs39 (Figure 2Q). Collectively, the above data demonstrated upregulation of ALKBH5 in the senescence process of NPCs was due to epigenetic decrease of H3K9me3 in the promoter.

RNA methylation could affect the turnover of mRNA transcripts via regulating the methylation level of target transcripts. To characterise the potential targets involved in m6A‐regulated senescence of NPCs, we performed Me‐RIP‐seq (m6A‐seq) to map the m6A methylomes of NPCs undergoing senescence with three independent biological replicates. And data manifested m6A sites were highly enriched in consensus GGAC motif in both senescent and control NPCs, especially abundant in the vicinity of start and stop codons (Figure 3A, B). To characterize the functional role of m6A modification, we identified the related pathways and enriched cellular processes of genes harbouring different m6A peaks, and “cell cycle” was demonstrated among the top significant biological processes (Figure 3C). By integrating analysis data of m6A‐seq and corresponding RNA‐seq data, we identified that DNMT3B was modified with m6A and one of the top genes with significant fold change of m6A modification during cellular senescence, which was confirmed by Me‐RIP‐qPCR analysis (Figure 3D, E). Moreover, consistent with RNA‐seq data, the mRNA expression of DNMT3B increased significantly during NPC senescence (Figure 3F, G). Furthermore, expression variance of DNMT3B transcripts in normal and degenerated NP tissues was confirmed by RNAScope ISH (Figure 3H and Figure S3A). Western blot was performed to find the protein level that was increased as well, in line with immunochemistry results of NP tissues from degenerated discs (Figure 3J, K and Figure S3B).

To further illuminate the role of m6A modification on DNMT3B expression, we examined the methylation level of DNMT3B transcripts by Me‐RIP‐qPCR in ALKBH5 knockdown NPCs and found that the m6A modification of DNMT3B was obviously increased under inhibition of ALKBH5, in concordance with the change of global m6A modification (Figure 3L and Figure S3C). Meanwhile, the expression of DNMT3B at both mRNA and protein level declined accompanied with ALKBH5 knockout‐induced higher methylation (Figure 3M and Figure S3E), indicating DNMT3B was regulated by m6A modification during NPC senescence. Therefore, the mechanisms regulating DNMT3B expression through m6A methylation demand further investigation.

First, we tested the promoter activity of DNMT3B using dual luciferase reporter assay in wild type and ALKBH5 knockdown NPCs and found that there was no obvious difference, suggesting that the transcription process of DNMT3B was not influenced by m6A (Figure 3N). Expression of pre‐mRNA (precursor) detected by RT‐qPCR showed no difference was observed in pre‐mRNA expression (Figure S3D). Then, we separated RNAs in nuclear and cytoplasm, and observed no different subcellular localization of DNMT3B mRNAs, which explains the fact that m6A could not affect the exportation of transcripts (Figure S3F, G). Furthermore, NPCs were treated with Cycloheximide (CHX) to block translation and western blot results verified that half‐life of DNMT3B proteins were similar, suggesting that m6A‐associated DNMT3B expression was not related to protein stability (Figure S3H). Moreover, ribosome–nascent chain complex qPCR (RNC–qPCR) was used to assess the translational efficiency of DNMT3B, whereas no difference of polysomes bound to EGFR mRNA in wild type and ALKBH5‐silencing NPCs showed a negative result, revealing translation efficiency of DNMT3B in ALKBH5‐silencing NPCs was consistent with that in wild type cells40 (Figure S3I). Above data indicated that there is no significant relationship between m6A modification and DNMT3B translation. Next, we treated NPCs with Act‐D to block the transcription, and the half‐life of pre‐RNA (precursor) and mat‐RNA (mature) was detected by RT‐qPCR at multi‐point time. The stability of mature mRNA of DNMT3B in ALKBH5‐knockdown NPCs was declined with shorter half‐life (Figure 3O). These results demonstrated that m6A modification may affect the degradation of mature mRNA but not the splicing of precursor mRNA of DNMT3B during NPC senescence.

To further clarify the regulatory mechanism of m6A on DNMT3B transcripts, we analysed the Me‐RIP‐seq data which indicated the m6A peak of DNMT3B was enriched in 3′UTR regions. Analysis of the sequencing data combined with bioinformatic analysis showed one GGACU motif in the 3′UTR region of DNMT3B mRNA (Figure 4A), consistent with the position of peaks identified by Me‐RIP‐seq. Furthermore, we generated luciferase reporters containing a firefly luciferase placed before the wild type and mutant DNMT3B‐3′UTR with substitution of A of the m6A sites with G. The dual luciferase reporter assay revealed that TNFα treatment could downregulate the luciferase activity, which could be partially abrogated by knockdown of ALKBH5 (Figure 4B), and Me‐RIP‐qPCR showed mutation could decrease the m6A modification of DNMT3B (Figure 4C, D), further confirming the methylation site of DNMT3B.

Previous study verified “readers” that could influence the stability of m6A‐modified transcripts include YTHDF2/3 and IGF2BP1/2/3.41 To identify the molecular mechanism by which m6A modification regulates the stability of DNMT3B mRNA, RNA‐pull down was performed followed by western blot analysis of the isolated proteins in senescent NPCs. Results manifested YTHDF2 interacted stronger with DNMT3B mRNA, and additional RIP‐qPCR analysis further revealed above recognition (Figure 4E, F), consistent with the prediction result of RNA interaction database RNA Intern (Figure 4G). In both normal and senescent NPCs, inhibition of YTHDF2 could lead to higher stability and expression of DNMT3B mRNA (Figure 4H). We then introduced the YTHDF2 truncation mutants (YTHDF2‐N and YTHDF2‐C) and m6A recognition site mutant (YTHDF2‐WA) in the NPCs (Figure 4I), and results at protein and transcript level implied that overexpression of full‐length but not mutants could promote the mRNA decay of DNMT3B (Figure 4J, K). These results demonstrated that the effect of m6A on DNMT3B was triggered by YTHDF2‐mediated decay of transcripts.

To further characterize the function of DNMT3B in the NPC senescence and disc degeneration, we knocked down the expression of DNMT3B with siRNA in NPCs, and knock down‐efficiency was verified by western blot and RT‐qPCR (Figure 5A). Compared with wild‐type NPCs treated with TNFα, the senescence of NPCs with deficiency of DNMT3B expression was alleviated. NPCs with DNMT3B silencing showed decreased SA‐β‐gal activity (Figure 5B, C), consistent with EdU incorporation analysis, indicating increased ratio of EdU positive cells and stronger ability of self‐renewal (Figure 5D, E). Furthermore, overexpression of DNMT3B could abrogate the anti‐senescence effect of ALKBH5 knockdown in NPCs, manifested by relative up‐regulation of senescence markers (Figure 5F, G). In the rat model of IVD degeneration, we revealed silencing of DNMT3B using lentivirus containing shRNA could alleviate the degeneration of IVD and the aging process of NPCs in vivo, reflected by MRI imaging and histological analysis (Figure 5H‐J and Figure S4A).

DNMT3B is a DNA methyltransferase that could suppress the expression of target genes via hypermethylation of the CpG islands in the promoter regions. Analysing the sequencing data combined with bioinformatic analysis of genes associated with cell cycle and senescence (GO:0006260: DNA replication; GO:0008283: cell proliferation; GO:0007049: cell cycle), we found that E4F1 had a significant difference (Figure 6A, B). Furthermore, the variant expression of E4F1 in senescent NPCs was confirmed by western blot and immunofluorescence, consistent with the results analysed in NP tissues (Figure 6C, D). To further investigate whether the E4F1 could be regulated by DNA methylation, we examined the methylation status of E4F1 promoter (−2000/+1000) in NPCs by Methylation‐specific PCR (MSP) analysis according to MethPrimer software recommendations (Figure 6E). The results showed, compared with normal NPCs, that senescent NPCs from degenerated individuals exhibited increased level of methylation at E4F1 promoter (Figure 6F, G). Furthermore, human NPCs treated with TNFα also manifested the same methylation tendency in E4F1 promoter, whereas additional administration of Nanaomycin A, a specific inhibitor of DNMT3B, and siRNAs against DNMT3B in NPCs could abolish the methylation level to some extent, and increase the expression of E4F1 meanwhile (Figure 6H–K). To gain additional molecular mechanism insight into the role of DNMT3B in the regulation of E4F1 gene, ChIP‐qPCR was performed to investigate the recruitment of DNMT3B to the promoter region of E4F1. Increased occupancy of DNMT3B in the CpG region of E4F1 was observed in senescent NPCs by ChIP assay, which could be abolished by DNMT3B silencing (Figure 6L). Furthermore, a notable decrease of chromatin accessibility of the E4F1 promoter in senescent NPCs was observed by DNase I sensitivity assay, which could be alleviated by inhibitor or small interference RNA of DNMT3B (Figure 6M). Collectively, these results demonstrate that DNMT3B elevation causes E4F1 promoter hypermethylation and suppresses E4F1 expression in NPC senescence.

E4F1 is a key regulator involved in proliferation and survival of cells, and is an essential gene in embryonic stem cells and during early embryogenesis.42, 43, 44 However, the roles of E4F1 in disc degeneration and NPC senescence remain unclear. Therefore, we investigated whether E4F1 repression devoted to the senescence of NPCs using lentivirus containing shRNA. As shown in Figure 6N–Q, downregulation of E4F1 led to autogenous senescence of NPCs, while overexpression of E4F1 in NPCs could reverse the senescence induced by TNFα. Furthermore, enforced expression of E4F1 in NPCs could abolish the pro‐senescence effect of ALKBH5 and DNMT3B to some extent (Figure 6R–U), demonstrating deficiency of E4F1 contributes to senescence of NPCs and revealing the critical role of m6A/DNMT3B/E4F1 axis in NPC senescence.