Preosteoclast plays a pathogenic role in syndesmophyte formation of ankylosing spondylitis through the secreted PDGFB — GRB2/ERK/RUNX2 pathway

Five Chinese AS patients with a first diagnosis and five sex- and age-matched healthy subjects were recruited in the discovery stage. Forty Chinese AS patients and twenty gender- and age-matched healthy patients were included in the validation I stage. The validation II stage included 40 Chinese AS patients who had stopped drug use for at least 1 month, 60 Chinese late-stage AS patients, and 120 healthy Chinese subjects. The 40 AS patients in validation II underwent X-rays of the spine at baseline and at the 2-year follow-up. AS patients with a first diagnosis were those who had received an AS diagnosis but had never been treated according to AS treatment guidelines. AS patients with late-stage disease had a radiological score of the sacroiliac joints of bilateral grade 3–4 and a radiological score of the spine (modified Stoke Ankylosing Spondylitis Spinal Score (mSASSS)) greater than 20, which requires bridging of at least 3 vertebrae or syndesmophytes at the scored vertebral sites [26]. All of the imaging scores were evaluated by two experienced physicians [27]. The diagnosis of AS was performed according to the 1984 modified New York criteria of AS [28]. All participants signed written informed consent forms, and this study was approved by the institutional review boards of Guanghua Integrative Medicine Hospital (2021-K-31) and Shanghai Pudong Hospital (2021-DS-Q-26). The details of the subjects are shown in Table S1.

Protein expression in human plasma was detected by the Human Antibody Array 1000 produced by RayBio (RayBiotech; Cat# AAH-BLG-1000; USA) in the discovery stage. It was performed according to the standard protocol provided by the manufacturer (https://​www.​raybiotech.​com/​files/​manual/​Antibody-Array/​AAH-BLG-1000.​pdf).

Protein quantification of plasma in the validation I stage was performed using ProcartaPlex immunoassays (Invitrogen). We designed the panels (6 cytokines: IL-6, IL-23, IFN-γ, PDGFB, VEGF-D, and VEGF-R) and performed the assays according to the manual provided by the manufacturer. Further quantification was performed on a Luminex 200, and the data analysis was performed according to a standard protocol (https://​assets.​thermofisher.​com/​TFS-Assets/​LSG/​manuals/​ProcartaPlex_​Analyst_​1.​0_​SW_​Manual.​June2014.​pdf). PDGFB quantification of cultured cells and all patients in the validation II stage was performed using an ELISA kit (Invitrogen, BMS 2071).

Interlaminar ligamentum flavum, interspinous ligament, or supraspinal ligament tissue was obtained from AS patients and traumatic injury patients who had spinal fractures when undergoing spinal surgery as controls. The tissues were fixed in 4% paraformaldehyde for 4–6 h and then embedded in paraffin. The paraformaldehyde-fixed paraffin-embedded sections were cut into 5 μm using a Leica RM2235. Paraffin-embedded sections were first deparaffinized twice in a series of 100% xylene, rehydrated in a series of graded ethanol (100%, 100%, 95%, and 80%), and then washed briefly in distilled water and PBS, respectively. Antigen thermal repair was performed with 0.01 M citrate buffer (pH6.0) or EDTA buffer (pH9.0) at high pressure, washed in PBS, sections were treated with 3% hydrogen peroxide-methanol for 10 min and blocked with normal goat serum for 30 min. Then, the sections were incubated overnight with PDGFB (1:100, Abcam, catalog: ab23914); GRB2 (1:50, Abcam, catalog: ab32037); P-ERK (1:100, Abcam, catalog: ab278538); RUNX2 (1:50, Abcam, catalog: ab76956) antibody at 4 °C. Goat anti-rabbit IgG secondary antibody (JACKSON, catalog: 111–035-003) was incubated and DAB solution (Sigma, catalog: D8001) was used for color development. All images were obtained using an Open-field slice scanner (NanoZoomer S210).

The osteoclastogenesis assays were performed according to our previous study [29]. In brief, human peripheral blood mononuclear cells (PBMCs) were isolated by Ficoll reagent (GE Healthcare, Buckinghamshire, UK) and then monocytes were purified by anti-CD14-conjugated magnetic microbeads (Miltenyibiotech, Bergisch Gladbach, Germany). To induce osteoclast differentiation, monocytes were stimulated with M-CSF (25 ng/mL, R&D Systems, 216-MC-010) and RANKL (40 ng/mL, R&D Systems, 390-TN-010) supplemented with 10% FBS and 10% human serum. Tartrate-resistant acid phosphatase (TRAP) staining was performed with Acid Phosphatase Leucocyte Kit (ZuoChengBio, ZCIC216), and TRAP-positive multinucleated cells (MNC) containing three or more nuclei were counted as preosteoclasts.

Human adipose tissue was obtained through elective liposuction with informed consent. The isolation and expansion of adipose-derived MSCs (ADSCs) was conducted according to previously published techniques [30]. Briefly, lipoaspirate was transferred into 50 mL tubes and centrifuged at 400 × g for 5 min. After digestion with collagenase I and filtration through a 100-μm filter, the stromal vascular fraction (SVF) was obtained. The cells were cultured in 175 cm² flasks until the fifth passage and then used for cell therapy. The ADSCs were characterized by flow cytometry and were found to express CD73, CD90, and CD105 but not express CD31, CD34, CD45, or HLA-DR, and the detector gain optimization and data quality checks were performed according to Gao et al. [31]. The ADSCs were first isolated and cultured in α-modified Eagle’s medium (α-MEM) supplemented with 10% FBS and antibiotics (100 U/mL penicillin and streptomycin) at 37 °C in a 5% CO₂ humidified incubator for approximately 3 days (coverage > 80%) after which the medium was changed to osteogenic medium (α-MEM plus 50 μM vitamin C, 10 mM β-phosphoglycerol and 100 nM dexamethasone) supplemented with 6% FBS and incubated for 21 days. The hFOB1.19 cells were obtained from the Shanghai Institute of Cell Biology, Chinese Academy of Sciences (Shanghai, China). The hFOB1.19 cells were cultured in Dulbecco’s modified Eagle’s medium/nutrient mixture F-12 (DMEM/F12) supplemented with 10% FBS, 2 mmol/L L-glutamine, 100 U/mL penicillin and streptomycin, 0.3 mg/mL G418, and 1.5 g/L NaHCO3 at 33.5 °C in a 5% CO₂ humidified incubator and were stimulated with recombinant PDGFB (Abcam, ab259425) for 14 days. For transfection experiments, cells were transfected with 2.5 nmol/ml GRB2 or RUNX2 small interfering (si)RNA mixed with 2 nmol/ml Lipofectamine RNAiMAX transfection reagent (Thermo Fisher Scientific, Waltham, MA, USA). The ERK inhibitor was purchased from Selleck (SCH772984) and was worked in a concentration of 300 nM.

ALP staining was performed according to a published article [32]. ARS staining was performed with Alizarin red S kit (ZuoChengBio, ZCIC123) according to the manufacturer’s instructions. The proliferation, migration, and chemotaxis experiments were also performed with the xCELLigence system according to previously published articles [33, 34].

TRIzol reagent was used to extract total RNA from the cells according to the manufacturer’s instructions (Invitrogen). A High Capacity cDNA Reverse Transcription Kit (Applied Biosystems, Foster City, CA, USA) was used to perform reverse transcription according to the manufacturer’s protocol. Real-time PCR samples were mixed with SYBR Premix Ex Taq (TaKaRa Biotech, Tokyo, Japan) and analyzed with an ABI Prism 7900 detector system (Applied Biosystems). β-actin was used as an internal control, and all primers used in this study are shown in Table S2. The assay statistics were analyzed with SDS 2.3 software (Applied Biosystems).

The cellular lysates extracted from cells were used for protein assays. The protein concentration was detected by a spectrophotometer according to the BCA protein assay kit procedure. Equal amounts of protein were subjected to SDS‒PAGE on a 10% polyacrylamide gel and transferred to a polyvinylidene difluoride membrane (Millipore). The membrane with blotted protein was blocked for 1 h at room temperature with a blocking buffer containing 5% BSA and then incubated with antibodies overnight at 4 °C. A housekeeping gene (GAPDH) was used as an internal control. The membrane was then incubated for 1 h at room temperature with secondary horseradish peroxidase-conjugated goat anti-rabbit or anti-mouse IgG after three 10-min washes with TBST. The protein bands were visualized with an ECL solution. Antibodies against ERK1/2 (ab201015), p-ERK1/2 (ab278538), GRB2 (ab32037), RUNX2 (ab76956), ALP (ab83259), type I collagen (ab260043), and type III collagen (ab184993) were purchased from Abcam.

Total RNA was converted into a DNA library according to the Illumina protocol (TruSeq Stranded Total RNA Sample Preparation Guide). The quality and the intact DNA were examined with an Agilent 2100, and subsequent DNA sequencing was performed with a HiSeq X Ten (Illumina) according to the standard protocol. The RNA sequencing data analysis was performed using TopHat and Cufflink according to a previously published article [35]. Pathway enrichment, including GO and KEGG analysis, was performed with the Functional Interpretation of Differential Expression Analysis (FIDEA) database and KOBAS 3.0 [36, 37].

Phosphorylation characterization of the peptides was performed according to a previously published article [38]. Briefly, the samples were loaded onto a microcolumn packed with TiO₂. The phosphopeptides were then eluted from the column. The collected phosphopeptides were then eluted from a microcolumn packed with R3. Finally, the phosphopeptides were analyzed by LC–MS.

All of the experiments were performed in triplicate. The statistical analyses were performed with R software (v3.6.0). Unpaired sample t-test was used for unpaired samples and two-way ANOVA was used to evaluate the difference between the two treatments and their effect size. P values < 0.05 following FDR adjustment for multiple testing corrections were considered to be statistically significant. All of the plots were generated by ggplot2 in R software.

A protein array (1000 cytokines) was performed in 5 AS patients with a first diagnosis and 5 healthy samples (Fig. S1A, Table S3). Following the screening approach, several AS-associated pathways, such as Toll receptor homology, CC chemokine, and TNFR, were significantly clustered (Fig. 1A). Interestingly, the vascular-related pathway was also abnormally expressed in AS (Fig. 1A and S1B). We then detected the plasma expression of three vascular-related cytokines along with three known associated cytokines in 40 AS patients and 20 healthy subjects in the validation I stage. The results showed that three known associated cytokines were also significantly elevated in the AS patients. The fold change (FC)/p values of IL-6, IL-23, and IFN-γ were 1.75/8E − 3, 1.68/9E − 4, and 1.27/8E − 3, respectively. PDGFB was significantly higher in AS patients (Fig. 1B, FC/p value = 5.44/6E − 4). In the validation II stage, we detected the plasma expression of PDGFB in a Chinese cohort (100 AS patients and 120 healthy subjects). PDGFB was consistently more highly expressed in the AS patients (FC/p value = 2.67/2.2E − 14). In the Chinese cohort, AS patients with a first diagnosis had much higher PDGFB expression levels than AS patients with a long disease duration (> 5 years) (Fig. 1C).

PDGFB expression was positively correlated with the BASFI. The Pearson coefficient was 0.62 (p value = 6.7E − 8, Fig. 2A). PDGFB also showed marginal associations with C-reactive protein (CRP) and the pain score (Pearson’s coefficient/p value = 0.16/0.02 and 0.20/0.01, respectively), while it showed no association with the disease activity index (BASDAI), patient global assessment (PGA) or AS disease activity score (ASDAS) (Fig. 2B). Interestingly, we found that the plasma expression of PDGFB at baseline was associated with variance in the mSASSS (mSASSS _{2 years later − baseline}) in AS patients in validation II-40 (Pearson’s coefficient/p value = 0.76/8.75E − 10; Fig. 2C). To explore the origin of the excess PDGFB, monocytes were harvested from AS patients and healthy controls, and osteoclastogenesis assays were performed. Following stimulation for 3 days, the monocytes from AS patients secreted more PDGFB than those of healthy subjects (Fig. 2D, p value = 1.16E − 2). The result of the osteoclastogenesis assay was examined by TRAP staining and the monocytes were differentiated into pre-osteoclast and osteoclast 7 days after the stimulation (Fig. 2E).

PDGFB promoted the differentiation of ADSCs into osteoblasts. The expression levels of OSN, COL1, and RUNX2 were significantly increased after the ADSCs were stimulated with PDGFB (Fig. 3A). Cell staining also showed that ADSCs stimulated with PDGFB had more intense ALP staining and Alizarin red S staining than controls (Fig. 3B–D). In addition, to exclude the possibility of the unequal number of cells in the two groups, normalization was performed with GAPDH and then the expression values of ALP and RUNX2 were quantified. The results showed that ALP and RUNX2 protein expression was much higher in the PDGFB ( +) group than in the PDGFB ( −) group (FC/p value = 3.6/0.05 and 5.2/0.009, respectively, Fig. 3E and F). Cell proliferation and migration activities were examined by the xCELLigence system. As shown in Fig. 3G and H, PDGFB increased the proliferation of ADSCs. In addition, PDGFB recruited ADSCs and increased their migration (Fig. 3I and J).

The results of the whole-transcriptome analysis showed that PDGFB activated the MAPK pathway to strengthen the osteogenic differentiation of ADSCs (Fig. 4A, Table S4). In particular, the expression of GRB2 was significantly increased following stimulation (FC/p value = 2.1/0.008) (Fig. 4B). To determine the key phosphorylated molecule in the MAPK pathway, highly selective phosphorylated peptides were analyzed by titanium dioxide chromatography. ERK, MEK1, and MKNK were abnormally phosphorylated after stimulation with PDGFB (Fig. 4C). By considering both the RNA sequencing results and peptide phosphorylation data, we determined that PDGFB stimulates ERK phosphorylation and enhances RUNX2 expression through GRB2. The expression of GRB2 is then inhibited by siRNA in the ADSCs (Fig. 4D). The western blot results showed that in the si-GRB2 group, ERK phosphorylation and RUNX2 expression were decreased, and the enhancement of GRB2/p-ERK/RUNX2 triggered by PDGFB was also inhibited by silencing the expression of GRB2 (Fig. 4E and F, Table S5). The ARS staining also showed that si-GRB2 inhibited the mineralization of ADSCs, and the enhancement of mineralization triggered by PDGFB was also decreased by the treatment of si-GRB2 (Fig. 4G and H, Table S6). To further explore the effects of the GRB2/p-ERK/RUNX2 pathway triggered by PDGFB, we inhibited the expression of ERK by using the ERK inhibitor. The results showed that when the ADSCs were treated with both PDGFB and ERK inhibitor, the axis was significantly inhibited by ERK inhibition compared with that after stimulation with PDGFB alone (Fig. 4I and J, Table S7, mRNA levels were shown in Fig. S2A). In addition, when the ADSCs were stimulated with PDGFB, the GRB2/p-ERK/RUNX2 axis was activated in dose- (Fig. 4K and L, mRNA levels were shown in Fig. S2B) and time- (Fig. S2C and D, mRNA levels were shown in Fig. S2E) dependent manner, which strengthen the role of PDGFB to this axis. Furthermore, the enhancement of PDGFB and the activation of the GRB2/p-ERK/RUNX2 pathway was found in the spinal entheseal tissues of AS patients (Fig. 5A and B).

FOB1.19 cells had no response to PDGFB regarding further osteogenic differentiation, proliferation, chemotaxis, or migration (Fig. S3A and B). First, we detected the expression of the axis pathway (GRB2/p-ERK/RUNX2) and found that GRB2 protein expression was relatively low in FOB1.19 cells even after stimulation with 60 ng/mL PDGFB; thus, the MAPK pathway could not be activated after stimulation with PDGFB (Fig. 6A). To further explore the impact of PDGFB on osteoblasts, RNA sequencing was performed. The pathways of focal adhesion and ECM-receptor interaction were significantly enriched according to the KEGG analysis (Fig. 6B and S4B). The detailed genes are presented in Fig. S4C and D, and the fully enriched pathways are presented in Fig. S4A, B and Tables S8 and S9 for both GO and KEGG. The western blot results also showed that the expression of type I and III collagen was significantly enhanced after stimulation with PDGFB (Fig. 6C and D). In addition, to explore whether the enhancement of type I and III collagen was RUNX2 dependent, we silenced the expression of RUNX2 by siRNA. The results showed that the enhancement of COL1 and COL3 triggered by PDGFB were both decreased after silencing the expression of RUNX2 (Fig. 6E and F, Table S10, mRNA levels were shown in Fig. S5), suggesting that the overexpression of COL1 and COL3 by PDGFB stimulation was RUNX2-pathway dependent.

Several limitations of this study need to be addressed. Firstly, the specific mechanisms of how PDGFB triggered the GRB2-pERK-RUNX2 pathway were not completely identified. GRB2 was found to contain an SH2 domain which is flanked by two SH3 domains [52, 53]. The SH2 domain recognizes the phosphotyrosine residues of activated EGFR, and the two SH3 domains bind to proline-rich sequences. Therefore, GRB2 can link EGFR and downstream signaling molecules (i.e., pERK), and may also link to PDGF [54]. In addition, the phosphorylation of ERK was reported to regulate skeletal development and bone formation [55]. As pERK is essential for bone formation, the interaction between GRB2 and pERK and how they interacted with RUNX2 were not fully discovered. The activation of the GRB2-pERK-RUNX2 axis triggered by PDGFB warranted deep exploration. Second, the enhancement of the GRB2/p-ERK/RUNX2 axis triggered by PDGFB was examined by normal human ADSCs, and the enhancement or the enhanced osteogenesis in ADSCs from AS patients was not identified and compared. However, Sungsin Jo et al. [25] have demonstrated that after PDGFB stimulation, the mineralization ability of enthesis cells from AS patients was stronger than the healthy ones. In addition, although PDGFB affects the axis in ADSCs, it does not have such effects on BMSCs (data not shown), suggesting its role in bone formation is limited to specific cell types and could not fully explain the mechanisms of new bone formation in AS.