Clinical specimen collection

The experimental protocol was approved by the Ethics Committee of Peking University Third Hospital, and informed consent was obtained from each included patient. Control normal LF tissues were collected from patients undergoing percutaneous endoscopic lumbar discectomy for lumbar disc herniation. OLF tissues were collected from patients who underwent posterior laminectomy for thoracic spinal stenosis caused by OLF. The clinical specimens were either fixed in 4% paraformaldehyde or immediately frozen in liquid nitrogen after the surgery. The details of included patients in this study were listed in Supplementary Table 1.

LF cell isolation and culture

LF cells were isolated from freshly obtained LF tissues from OLF patients. First, the LF tissues were cut into small pieces of approximately 0.5 mm3 and washed twice with phosphate buffered saline (PBS). The clipped tissue pieces were then digested with 0.25% trypsin (Gibco, USA) followed by 250 U/mL type I collagenase (Solarbio, China). After centrifugation, the sediment was resuspended in Dulbecco’s modified Eagle’s medium (Procell, China) containing 10% foetal bovine serum (Gibco, USA) and 1% penicillin/streptomycin antibiotics (Procell, China), which was subsequently placed in 10 cm culturing dishes in a humidified atmosphere with 5% CO2 at 37 °C. The second and third passages of LF cells were used for subsequent experiments. For osteogenic differentiation, LF cells were cultured in commercial osteogenic medium provided by Procell (China).

RNase R treatment and RNA stability assay

Total RNA was extracted using a SteadyPure Universal RNA Extraction Kit (Accurate Biology, China). A total of 2 µg of RNA was incubated with or without RNase R at 37 °C for 10 min, which was used for the qRT‒PCR assay. RNA stability was measured using the actinomycin D assay. LF cells were treated with 1 µg/mL actinomycin D (MCE, USA) for 0, 8, and 16 h, and then RNA was extracted from LF cells to determine the abundance of circCDK14.

Subcellular fractionation

The Ambion PARIS Kit (Invitrogen, USA) was applied for subcellular fractionation according to the manufacturer’s instructions. GAPDH and U6 were employed as the cytoplasmic control and nuclear control, respectively.

RNA isolation and quantitative real time polymerase chain reaction (qRT-PCR) assay

Total RNA was extracted from LF cells using a SteadyPure Universal RNA Extraction Kit (Accurate Biology, China) following the manufacturer’s protocol. First-strand cDNA was synthesized from 1.2 µg of total RNA using the Evo M-MLV RT Kit with gDNA Clean for qPCR kit (Accurate Biology, China). Then, the qRT‒PCR assay was performed using the SYBR Green Premix Pro Taq HS qPCR Kit (Accurate Biology, China) on the ABI QuantStudio 5.0 version (Thermo Fisher Scientific, USA). The relative transcription levels were standardized to GAPDH using the 2−ΔΔCt method. The primer sequences are listed in Supplementary Table 2.

Western blot and immunohistochemistry

LF cells or tissues were lysed with RIPA buffer (Epizyme, China), and the concentration of proteins was measured using the BCA kit (Epizyme, China) according to the manufacturers’ protocol. After being denatured with loading buffer (Epizyme, China) at 99 °C for 15 min, protein samples were separated on 10% SDS–PAGE gels (Epizyme, China) and then transferred to polyvinylidene fluoride membranes (Millipore, USA) using Bio-Rad devices (Bio-Rad, USA). Then, the membranes were blocked with 5% skim milk powder (Epizyme, China) for 1.5 h and incubated with primary antibodies for 2 h at room temperature. The PVDF membranes were washed with TBST (Epizyme, China) and incubated with HRP-conjugated secondary antibodies (Epizyme, China) for 1 h at room temperature. An ECL kit (Epizyme, China) was used to visualize the immune complexes, and the relative protein levels were qualified using the iBright CL1000 device (Thermo Fisher Scientific, USA). The details of the preliminary antibodies used in this study were as follows: OPN (ImmunoWay, USA), RUNX2 (Abcam, USA), AFF4 (Proteintech, USA), AGO2 (Cell Signaling Technology, USA), and WTAP (Abcam, USA). For immunohistochemical staining, the formalin-fixed, paraffin-embedded LF tissues adjacent OLF tissues and normal LF tissues were analysed as previously reported [16], with the following preliminary antibodies: AFF4 (Proteintech, USA) and WTAP (Abcam, USA).

CircCDK14 pull-down assay and mass spectrometry analysis

The MS2 bacteriophage coat protein (MS2-CP) circRNA pull-down assay was conducted by using the MS2 tagging technique, and the MS2 tagging technique is developed for the natural binding between the stem-loop structure of MS2 and MS2-CP [28, 29]. In this study, the RBPs associated with circCDK14 were explored by using a circRNA pulldown assay with the MS2-CP technique. In brief, two specified plasmids were constructed: one plasmid with circCDK14-MS2 labelled with GFP, and another plasmid with MS2-CP-Flag labelled with mCherry. The HEK293T cells were transfected with these two plasmids and precipitated circCDK14 by the pulling down using anti-Flag antibodies. Besides, the lysates derived from the cells without the MS2 tagging system were used as the controls. The cell lysates were incubated with Protein A/G beads overnight at 4 °C. Then, the circCDK14-MS2-bound proteins were eluted. The bound RNA was reverse transcribed and analysed by qRT-PCR, and the bound proteins were determined by label-free mass spectrometry (Thermo Scientific, USA) and Western blot.

RNA immunoprecipitation (RIP) assay and methylated RNA immunoprecipitation (MeRIP) assay

The RIP experiment was conducted with the PureBinding® RNA Immunoprecipitation Kit (Geneseed, China) with anti-AGO2 (Cell Signaling Technology, USA) and WTAP (Abcam, USA) following the manufacturer’s instructions. IgG (Cell Signaling Technology, USA) was used as a negative control. The coprecipitated circCDK14 level was further evaluated by qRT–PCR assay.

To detect the level of m6A modification in the circCDK14, MeRIP was conducted using the Magna MeRIP™ m6A kit (17–10–499, Merck Millipore, USA) according to the manufacturer’s instructions. The m6A level of circCDK14 was evaluated using the qRT-PCR assay.

Dual-luciferase reporter assay assays

Human embryonic kidney (HEK) 293T cells were used for luciferase activity analysis and were plated on 96-well plates and cultured to 60–70% confluence. For circCDK14 and miR-93-5p, the circCDK14 fragments containing the complementary binding sites for miR-34a-5p were wild-type (WT) pMIR-REPORT-circCDK14 or mutant-type (MT) luciferase reporter pMIR-REPORT-circCDK14 (OBIO, China). Then, the luciferase reporter vectors of circCDK14-wt and circCDK14-mut, miR-93-5p mimics, and NC mimics were transfected using Lipofectamine 3000 (Invitrogen, USA). For AFF4 and miR-93-5p, the 3’UTR of AFF4 containing putative binding sites for miR-93-5p was cloned and inserted into the wild-type pMIR-REPORT-AFF4-3’UTR or mutant luciferase reporter pMIR-REPORT-AFF4-3’UTR (GemmaPharma, China). Then, the luciferase reporter vectors of AFF4-wt and AFF4-mut, miR-93-5p mimics, and NC mimics were transfected into HEK 293T cells. Luciferase activity was examined with a Dual Luciferase Reporter Assay kit (Promega, Beijing) following the manufacturer’s instructions 48 h after transfection.

Small interfering RNA and plasmid construction and cell transfection

All small interfering RNAs (siRNAs) were provided by OBiO Technology (Shanghai, China). The miR-93-5p mimics were provided by OBiO Technology. The details of these siRNAs or mimics are listed in Supplementary Table 2. For the overexpression of circCDK14 and WTAP, the cDNA of circCDK14 and WTAP were synthesized, cloned, and inserted into the overexpression vector pcDNA3.1 by OBIO Technology as follows: pcDNA3.1(+), pcDNA3.1(+)-S-hsa_circ_0001721, pcDNA3.1(+)-3xFLAG-P2A-EGFP, and pcDNA3.1(+)-WTAP-3xFLAG-P2A-EGFP. All siRNAs, mimics, and vectors were transfected into LF cells with lipo3000 reagent following the manufacturer’s protocols when LF cells reached 70% confluence. The LF cells were used for further experiments after the transfection.

Alkaline phosphatase (ALP) staining and activity assay

LF cells were seeded into 12-well plates at a density of 2.0 × 104 cells/well. ALP staining of LF cells was conducted using the BCIP/NBT Kit (Beyotime, China). According to the instructions, the cells were washed 3 times with cold PBS and then fixed in 4% paraformaldehyde for 15 min. Then, the fixed LF cells were washed 2–3 times with PBS again and added to ALP solution. After incubation in the dark for 30 min at room temperature, a stereomicroscope was used to capture the ALP staining images. The ALP activity of LF cells was assessed using the ALP/AKP kit (Nanjing Jiancheng, China) according to the instructions, and the OD was detected with a multiscan spectrum at an absorbance of 520 nm with an equal amount of protein.

Alizarin red s (ARS) staining and quantitative assay

Calcium deposition was assessed by ARS staining and quantitative assays during osteogenic differentiation. LF cells were inoculated into 12-well plates at a density of 2.0 × 104 / well. In brief, LF cells were fixed in 4% paraformaldehyde for 15 min 3 times with cold PBS. After rinsing with PBS 2–3 times, the cells were exposed to 0.20% ARS (Procell, China) solution and incubated shielded from light at room temperature for approximately 30 min. When the mineralization-positive cells were stained red, the cells were washed twice with PBS again, and photographs were captured under a stereomicroscope. After photographing, 10% cetylpyridine (Aladdin, China) was added to each well, followed by incubation under light protection at room temperature for 30 min. Then, the samples were transferred to 96-well plates at 100 µl per well, and the OD was detected with a microplate analyser (Thermo Fisher Scientific, USA) at an absorbance of 562 nm.

Statistical analysis

All analyses were conducted using GraphPad Prism 9.0 (GraphPad Software, USA), and all data are presented as the mean ± standard deviation (SD) of at least three independent experiments. Two-tailed unpaired Student’s t test and one-way ANOVA followed by the Tukey–Kramer test were used to evaluate the statistical significance of any differences. p < 0.05 was considered statistically significant.