By analyzing data from the NCBI-GEO platform, it was found that, based on data from the GPL96, GPL570, GPL11154, GPL1708, GPL10558, and GPL91 platforms, and compared with the synovial samples from healthy people, those from RA patients had significantly higher expression of genes encoding CD47, TSP1, SIRP-α, and integrin subunits α4, αM, αv, αL, β1, β2, and β3 (RA=230, healthy=59) ( Figures 1A–J ). Moreover, based on the data from GPL6947, GPL570, GPL20171, and GPL13158 platforms, the expression of genes encoding CD47, TSP-1, SIRP-α, and integrin subunits α4 and αL was significantly higher in the peripheral blood samples of RA patients than in healthy people (RA=1238, healthy=120) ( Figures 1K–O ). Based on the NCBI/GEO/GSE55457/77298/55235 data, in the synovial tissue samples of RA patients, the expression of genes encoding several integrin subunits (α4, αM, β1, β2, and αL), ICAM-1, and SIRP-α was significantly and positively correlated with CD47 expression ( Supplementary Figure S1 ). These results indicated that CD47 may play a pivotal role in RA through related integrin molecules.

Collagen-induced arthritis model in wild-type (WT) rats was established, which included two animal groups: the normal group and the model group (WT-CIA) ( Supplementary Figure S2 ). Testing from day 13 of modeling found that the rats in the model group had markedly increased swelling of left and right hind paws, and higher arthritis and clinical scores compared with the normal group ( Supplementary Figures S2A–E ). On day 28, the animal blood samples were collected for testing, which revealed that the levels of CIA-associated autoantibodies (including rheumatoid factor, anti-cyclic citrullinated peptide (anti-CCP) antibody, and anti-collagen II (anti-CII) antibody), TNF-α levels, total leukocyte counts, neutrophil counts, and platelet counts were all substantially increased in the blood of the model group compared with the blood samples from rats in the normal group ( Supplementary Figures S2F–L ). These results indicated that rat CIA modeling was successfully set up. We sacrificed the experimental animals on day 28 of modeling, harvested the joint synovial tissues from the hind paws, and performed quantitative real-time polymerase chain reaction (qPCR) and western blot analyses. It was found that in the hind paw joint synovial tissue of the model group animals, the expression of CD47, its two ligands SIRP-α and TSP-1, and integrin subunits α4, αM, β1, and β2 was remarkably higher ( Supplementary Figures S2M–T ). These findings were consistent with the analysis results of the synovial specimens from RA patients, indicating that CD47 may also play an important role in CIA.

CRISPR/Cas9 technology was employed in this study to delete the 31 bp target site of the second exon of the Cd47 gene in Sprague Dawley (SD) rats. After breeding and screening, the homozygous Cd47 knockout rats were derived ( Supplementary Figure S3 ). The systematic protein expression analysis of the tail tissue of a wild-type and a Cd47 knockout rat revealed that CD47 was minimally expressed in the Cd47 knockout rat, while the expression of other related molecules did not differ between the wild-type and Cd47-deficient rats ( Supplementary Figure S3D ). We utilized wild-type rats and Cd47 knockout rats to establish the CIA models, which included four groups of animals: the wild-type rat normal group (WT), the Cd47 knockout rat normal group (CD47-KO), the wild-type rat model group (WT-CIA), and the Cd47 knockout rat model group (CD47-KO-CIA) ( Figure 2 ). Compared with the WT-CIA group, the CD47-KO-CIA group exhibited significant decreases in the swelling of the left and right hind paws, arthritis and clinical scores, as well as levels of rheumatoid factor (RF), TNF-α, total leukocytes, neutrophils, and platelets in the blood ( Figures 2A–J ). There was no significant difference in levels of anti-CCP and anti-collagen II antibodies between WT-CIA and CD47-KO-CIA groups.

On day 28 of modeling, the hind paw joints from animals were collected to perform histopathological analysis. Compared with the WT normal group rats, severe synovial tissue hyperplasia, pannus formation, inflammation, and bone erosion could be seen on the hind paw ankle joint sections of animals in the WT-CIA group, whereas on the hind paw ankle joint sections of the CD47-KO-CIA group, these symptoms were alleviated ( Figures 2K, L ). Immunohistochemical analysis of ankle joint sections for CD34, VCAM-1, and ICAM-1 expression revealed there was less vascular endothelial cell proliferation of the ankle joint synovial tissue in the CD47-KO-CIA group than in the WT-CIA group animals ( Figures 2M, N ). RNAseq analysis results of the synovial tissue in the experimental animals also demonstrated that compared with the WT-CIA group, the CD47-KO-CIA group had reduced expression of these vascular marker molecules in the ankle joint synovial tissue as well ( Figure 2O ).

Hind paw synovial tissue of animals was also collected on day 28 of modeling to perform western blot and qPCR analyses. Compared with the WT normal rats, the WT-CIA animals had significantly elevated expression of CD47, several integrin subunits, SIRP-α, and TSP-1 in the hind paw synovial tissue. In comparison with animals in the WT-CIA group, the hind paw synovial tissue expression of CD47, TSP-1, and integrin subunits α4, αL, and β1 decreased in the CD47-KO-CIA group animals, while there were no changes in the expression of SIRP-α and integrin subunits αM, αv, β2, and β3 ( Figures 3A–K ).

RNAseq analysis on the synovial tissue of the experimental animals from each group was used to find that compared with the animals of the WT-CIA group, those of the CD47-KO-CIA group had decreased expression of the genes encoding TSP-1 and integrin subunits α4, αL, and β1 in synovial tissue, while the expression of genes encoding SIRP-α and of the integrin subunits αM, αv, β2, and β3 were not altered ( Figure 3L ).

The above experiments analyzed the overall clinical appearance of the animals, tested for key factors and cells in blood, and analyzed experimental animal joint tissue slices, all of which suggested that Cd47 deficiency significantly reduced the disease severity of model rats, synovial inflammation levels, and vessel hyperplasia. Molecular analysis of the experimental animals’ articular synovial tissue indicated that Cd47 knockout significantly decreased the expression of TSP-1 and integrin subunits α4, αL, and β1 in the synovial tissue of model animals.

Immunofluorescence analyses were performed on sections of the experimental animals’ hind paw joints for a T cell marker (CD3), a neutrophil marker (Ly6G), a macrophage marker (CD68), and a B cell marker (CD45RA). It was found that compared with the WT-CIA group, the CD47-KO-CIA group had significantly reduced CD3 expression, whereas there were no obvious changes in Ly6G, CD68, or CD45RA expression ( Figures 4A–E ). RNAseq analysis on the joint synovial tissue of experimental animals confirmed the above conclusion at the mRNA level ( Figure 4F ).

Immunofluorescence analysis was further performed for a T cell marker (CD3), a CD4+ T cell marker (CD4), a CD8+ T cell marker (CD8), a Th17 cell marker (IL-23R), and a Treg cell marker (CD25) in the experimental animal joint tissue. It was found that compared with the wild-type rat normal group, the WT-CIA group exhibited significantly increased expression of these markers in the joint tissue sections. In contrast to the WT-CIA group, the CD47-KO-CIA group showed significantly reduced expression of these markers on the joint tissue sections ( Supplementary Figure S4 ). We performed RNAseq analysis on the synovial tissue of the experimental animals and found that the change in gene expression encoding marker molecules for various T cell subgroups (CD3D, CD4, CD8a, IL-17A, and FoxP3) agreed with the immunofluorescence analysis results ( Supplementary Figure S4F ). In addition, western blot analysis was further performed for the expression of relevant marker molecules for CD4+T cells, Th17 cells, and Treg cells in the experimental animal synovial tissue, which revealed that in comparison to the WT-CIA group, the CD47-KO-CIA group had substantially decreased expression of these marker molecules in synovial tissues ( Supplementary Figures S4G–I ).

Data shown in Figure 4 and Supplementary Figure S4 demonstrated that after modeling of Cd47-deficient rats, the number of T cells and several T cell subgroups, including Th17 cells, which infiltrated into joint synovial tissues markedly declined. According to reports in literature, in a CIA model, deficiency of IL-17A (17) or blocking its function by IL-17 antibody (18) considerably attenuated the development of disease symptoms. Thus, Cd47 knockout could mitigate the severity of arthritis by reducing the infiltration of T cells (especially Th17 cells) into articular synovial tissues.

In the last two parts of study, it was found that Cd47 deficiency significantly reduced T cell infiltration into articular synovial tissue, and that the expression of TSP-1, CD47, and integrin subunits α4, αL, and β1 noticeably declined in the synovial tissues of model animals. In this part of study, through the interactions between TSP-1, CD47, and integrin α4β1, the important role of CD47 in the process by which T cells pass across vessel walls and infiltrate into synovial tissues was investigated.

In order to clarify the details of these mechanisms, we first employed laser confocal observations and found that in human T cells (Jurkat cells), the signals from CD47 molecules labeled with red fluorescence and integrin subunit α4 molecules labeled with green fluorescence merged into yellow fluorescence, confirming the interaction between CD47 and integrin α4β1 ( Figure 5A ). Co-immunoprecipitation (co-IP) analysis using CD47, integrin subunit α4, or β1 antibodies also confirmed that CD47 interacted with integrin α4β1 ( Figures 5B–D ).

TSP-1 was continuously present in the synovial tissues of RA patients (19). The peptide 4N1K was derived from TSP-1 and directly participated in the interaction with CD47 (20). Cellular function assays revealed that 4N1K notably promoted Jurkat cell adhesion to immobilized VCAM-1 and migration to immobilized VCAM-1 induced by SDF-1, while the negative control peptide 4KGG did not have these effects ( Figures 5E, F ).

After activation, integrin α4β1 changed from its original folded state to an extended state (15), and the activated integrin α4β1 was detected by an antibody against the extended form of integrin α4β1 in combination with flow cytometry. Peptide 4N1K treatment significantly increased the percentage of the active form of integrin α4β1 on the surface of Jurkat cells ( Figures 5G–K ), while the total cell surface expression of integrin α4β1 remained normal ( Figures 5L, M ).

The above data suggest that by interacting with integrin α4β1, CD47 activated integrin and switched it from the original folded state to the extended state that had a high binding affinity to VCAM-1, enhancing the adhesive and migratory capacity of T cells mediated by VCAM-1.

Integrin αLβ2 is another integrin molecule expressed by T cells, which plays an important role during the special stage when leukocytes pass through the vessel wall into inflamed tissue (21). Through inspection by laser confocal microscopy and co-IP detection, the CD47 on human Jurkat cells was not found to interact with integrin αLβ2 ( Supplementary Figures S5A–D ). Functional assays at the cellular level showed that the 4N1K peptide did not enhance either the adhesion or SDF-1-induced migration of Jurkat cells to immobilized ICAM-1 ( Supplementary Figures S5E, F ). In addition, the proportion of integrin αLβ2 in the active form on Jurkat cells was also not increased after the action of peptide 4N1K on CD47 ( Supplementary Figures S5G–L ). These results indicated that the increased T cell adhesion and migration resulting from the interaction between CD47 and TSP-1 were achieved through the integrin α4β1-VCAM-1 interaction.

The 4N1K peptide significantly increased the wild-type rat CD3+ T cell adhesion to immobilized VCAM-1 ( Supplementary Figure S6A ) and migration induced by SDF-1 ( Supplementary Figure S6C ). The Cd47-deficient rat CD3+ T cell adhesion to immobilized VCAM-1 ( Supplementary Figure S6B ) and migration to SDF-1 chemoattractant ( Supplementary Figure S6D ) did not display any differences between the 4N1K treatment group, negative peptide 4KGG treatment group, or normal control cells. Examination by flow cytometry showed that peptide 4N1K significantly increased the amount of the active form of integrin α4β1 on the CD3+ T cell surface of wild-type rats ( Supplementary Figures S6E–K ). In contrast, 4N1K treatment of CD3+ T cells of Cd47 knockout rats did not noticeably change the quantity of integrin α4β1 in the active form ( Supplementary Figures S6L–O ). These results further verified that the TSP-1 activation of integrin α4β1 on the T cell surface and enhancement of the T cell adhesive and migratory capacity was CD47-dependent.

This set of results explained the critical function exerted by CD47 in T cell infiltration into synovial tissues by establishing that TSP-1-CD47-integrin α4β1 had mutual interactions. The CD47-TSP-1 interaction activated integrin α4β1 on the surface of T cells, and the activated integrin α4β1 had a higher affinity to VCAM-1 on the surface of the vascular endothelial cells, which consequently facilitated T cells to pass through the vascular wall and enter the synovial tissue.

Phosphorylation of the integrin α4 subunit cytoplasmic domain modulated its interaction with the cytoskeletal protein paxillin (22), and paxillin is an important cytoskeletal protein for cell adhesion and mobility induced by integrin α4β1 (23). Although the phosphorylation of the α4 subunit cytoplasmic domain at the basal level was undetectable by western blot, 4N1K or thrombospondin (TSP) treatment of Jurkat cells induced the cytoplasmic domain phosphorylation of the α4 subunit ( Figure 6A ).

Previous research indicated that PKA played a central role in the cytoplasmic domain phosphorylation of the integrin α4 subunit. In addition, the cytoplasmic domain phosphorylation of integrin α4 subunits was shown to be Gi protein-dependent, and the Src kinase in the downstream signaling pathway of Gi protein was necessary for this process (14). An examination in this study revealed that the cytoplasmic domain phosphorylation of integrin α4 subunit was completely blocked by the PKA inhibitor PKAI ( Figure 6B ). In contrast, the cytoplasmic domain phosphorylation of the α4 subunit was not affected by the Gi protein inhibitor pertussis toxin PTX ( Figure 6B ). The CD47-induced signaling pathways did not cause changes in the overall levels of cAMP and IP3, which are the second messengers for the two principal signal transduction pathways involving the G protein-linked receptors: the cAMP signal pathway and the phosphatidylinositol signal pathway (24) ( Figures 6C, D ). Furthermore, the Src kinase family inhibitor PP2 markedly repressed the cytoplasmic domain phosphorylation of the α4 subunit induced by 4N1K in Jurkat cells ( Figure 6B ).

4N1K or TSP treatment of wild-type rat CD3+ T cells noticeably increased the phosphorylation of the α4 subunit cytoplasmic domain ( Figure 6E ). Co-incubation with PKAI or PP2 substantially reduced the phosphorylation of the cytoplasmic domain of the α4 subunit induced by 4N1K, while there was no effect with PTX co-incubation ( Figure 6F ). Correspondingly, cAMP and IP levels in wild-type rat T cells did not vary under treatment with 4N1K, 4KGG, or without peptide treatment ( Figures 6H, I ). In Cd47-deficient rat CD3+ T cells, intracellular phosphorylation of the α4 subunit cytoplasmic domain was not detected under all the above conditions ( Figures 6J, K ). The α4 subunit expression levels within T cells of wild-type rats ( Figure 6G ) and Cd47 knockout rats ( Figure 6L ) did not change under 4N1K or 4N1K together with PKAI, PTX or PP2 treatments.

Peptide 4N1K treatment prominently increased the wild-type rat CD3+ T cell adhesion and migration towards the immobilized VCAM-1. 4N1K-stimulated cellular adhesion and migration were suppressed by co-incubation with PKAI or PP2, while co-incubation with PTX did not show an inhibitory effect ( Supplementary Figure S7 ).

These results indicate that the activation of integrin α4β1 by CD47 or the phosphorylation of the α4 subunit cytoplasmic domain in T cells were Gi protein-independent, and PKA and Src kinase were the key enzymes in these activation processes.

Next, we locally injected into CIA rats the CD47 antibody or inhibitors of key pathways in which CD47 activated integrin α4β1 (phosphorylating the α4 subunit cytoplastic domain) in T cells, and examined their effects on arthritis development. Experimental data showed that subcutaneous injection at the bottom of the left hind paw with the CD47 antibody, PKA inhibitor H89, or Src kinase inhibitor PP2 resulted in a significantly lower degree of swelling of the left hind paw compared with the WT-CIA group (subcutaneous injection of PBS after modeling), whereas subcutaneous injection of a G protein inhibitor (PTX) did not induce this effect ( Figure 7A ). The arthritis score ( Figure 7C ) and clinical score ( Figure 7D ) also showed similar outcomes. The animal groups did not differ in the swelling of the right hind paw as a control ( Figure 7B ). Western blot analysis of the left hind paw joint synovial tissues revealed that following treatment with the PKA inhibitor H89 or Src kinase inhibitor PP2, the expression of CD47, TSP-1, and integrin subunits α4, αL, and β1 noticeably decreased compared with the WT-CIA group, while the expression of SIRP-α and integrin subunits αM, αv, β2, and β3 showed minimal change ( Figure 7E ). Local injection of CD47 antibody also had similar results ( Figure 7F ). Furthermore, after treating with the PKA inhibitor H89, Src kinase inhibitor PP2, or CD47 antibody, the expression of the T cell marker molecule CD3 decreased in the left hind paw joint synovial tissue of the experimental animals, while the expression of the neutrophil marker (Ly6G), macrophage marker (CD68), and B cell marker (CD45RA) did not change significantly ( Figures 7G, H ). These results paralleled the previous findings showing that Cd47 deficiency remarkably reduced the T cell infiltration into joint synovial tissues, and that the Cd47 knockout significantly decreased the levels of TSP-1, CD47, and integrin subunits α4, αL, and β1 in the synovial tissue of model animals. In western blot assays, the levels of various molecules in the left hind paw joint synovial tissues of the animals treated with G protein inhibitor PTX did not significantly differ from those of the WT-CIA group. These experimental results suggested the TSP-1-CD47 interaction and that the key enzymes (PKA and Src) in the intracellular signaling pathways activating integrin α4β1 may become potential targets for RA therapeutics.

Expression profiles of CD47, TSP1, SIRP-α, and integrin subunits (α4, αM, αv, αL, β1, β2, and β3) in the synovial tissue or peripheral blood samples of RA patients and healthy people were analyzed and established based on published NCBI-GEO profiles (GPL96, GPL570, GPL11154, GPL1708, GPL10558, GPL91, GPL6947, GPL570, GPL20171, and GPL13158 platforms; for synovial tissue sample RA=230, Normal=59; for peripheral blood sample RA=1238, Normal=120).

Gene chip data from human synovial tissue and peripheral blood tissue in public databases were retrieved and analyzed. All the GEO project data related to “Homo sapiens” and “synovial or Blood” from the NCBI database were retrieved and the corresponding CEL files for each GEO dataset were downloaded. These CEL files were then subjected to preprocessing using the affy package in R language. Subsequently, the integrated data were subjected to batch correction using the sva package in R language. Differential expression analysis between the RA group and the normal group was conducted utilizing the limma package. Only dual-channel chip data were retained for further analysis.

Violin Plot, Heatmaps, and tabulated values (LogFC and adjusted P values) were calculated and presented. Threshold values for differential expression are logFC≥0.263 and adj.P.Val ≤ 0.01.

The linearity of CD47 and relevant integrin subunit α4, αM, αL, β1, β2, SIRP-α, and ICAM-1 was analyzed based on the published GEO profiles (NCBI/GEO/GSE 55457/77298/55235).

Human Jurkat cells (SCSP-513) were purchased from the Cell Bank of the Chinese Academy of Sciences (Shanghai, China) and cultured in RPMI-1640 medium (Gibco, Grand Island, NY) with 10% FBS (Life Technologies, Carlsbad, CA) and 1% penicillin/streptomycin (Invitrogen, USA). The non-tumorigenic rat CD3+ T cells were isolated from wild-type or CD47 knockout rats using anti-PE MicroBeads and PE-Anti-CD3 antibodies (cat log: 130-048-801 and 130-103-130, Miltenyi, China). Cells were cultured in RPMI-1640 culture medium with 10% FBS and 1% penicillin/streptomycin containing recombinant rat IL-2 (400-02, Peprotech).

Complete Freund’s adjuvant (M. tuberculosis H37RA, 10 mg·mL⁻¹) and bovine type II collagen were purchased from Chondrex (Redmond, WA, USA). The kits for detection of rat TNF-α (ml002859), rat IL-6 (ml102828), rat rheumatoid factor (ml059541), anti-CCP antibody (ml059150), anti-collagen antibody (ml058275), human cAMP (ml064278), human IP3 (ml060362), rat cAMP (ml002907) and rat IP3 (ml059408) were from Shanghai Meilian Biotechnology Co., Ltd. (Shanghai, China). TRIzol was purchased from Invitrogen Trading Co., Ltd (Waltham, MA, USA). A PrimeScriptTM RT reagent kit with gDNA Eraser and SYBR^® Premix Ex Taq™ II (Tli RNaseH Plus) was purchased from TaKaRa (Kyoto, Japan). HRP conjugated anti-GAPDH mouse mAb (70-ab011-040), HRP conjugated goat anti-rabbit IgG (70-GAR007), and HRP conjugated goat anti-mouse IgG (70-GAM007) were purchased from Multi sciences (Hangzhou, China). Monoclonal antibodies specific for CD47 (ab108415), SIRP-α (ab191419), and polyclonal antibodies specific for integrin subunit αM (ab133357) were purchased from Abcam (Cambridge, UK). A monoclonal antibody specific for integrin subunit β1 (34971S) was from Cell Signaling Technology (Danvers, MA, USA). Monoclonal antibodies specific for integrin subunit α4 (sc-376334), β2 (sc-8420), p-α4 (sc-23943), and IL-23 (sc-271279) were from Santa Cruz Biotechnology (Santa Cruz, CA, USA). Polyclonal antibodies specific for IL-17A (DF6127), IL-1β (AF5103), IL-12A (AF5133), and IL-6 (DF6087) were from Affinity. Polyclonal antibodies specific for FoxP3 (22228-1-AP) and TGF-β1 (21898-1-AP) were from ProteinTech (Wuhan China). PKAI (476485) and PTX (516560) were from Merck&Millipore (Burlington, MA, USA). PP2 (13198) was from Cayman (Ann Arbor, MI, USA). Human Thrombospondin-1 (SRP4805) was from Sigma-Aldrich (St. Louis, MO, USA).

Peptide 4N1K (KRFYVVMWKK) and 4KGG (KRFYGGMWKK) were synthesized by GL Biochem Co., Ltd. (Shanghai, China) with a purity of more than 95%.

Cd47 knockout rats were generated under SD rat gene background under SPF conditions by Nanjing Biomedical Research Institute of Nanjing University (NBRI). Project number: XM200765. Strain ID: T004140. Genetic background: SD. Cd47 gene includes 9 exons. The second exon of Cd47 was used as a sgRNA target and sgRNA primers were designed using the Cas9 target design software (http://crispr.mit.edu/). The sequence of sgRNA (5’→3’) was ACAATGGATACCCATG for S1 with PAM as AGG and TCTCAGACTTGCTCAA for S2 with PAM as AGG. After the design and clone of sg RNA vector, Cas9 mRNA and sgRNA were in vitro transcribed. 100 ng/μL Cas9 mRNA and 40 ng/μL sgRNA were micro-injected into fertilized eggs and the eggs were put back in the uterus of pseudo-pregnant rats. After DNA extraction from F0 rats and gene sequencing, positive F0 rats were selected and were crossed with wild-type rats of the same genetic background. Positive F1 rats were selected by DNA extraction and gene sequencing. Male and female positive F1 rats mated to generate F2 rats. Homozygous Cd47 knockout F2 rats were selected by DNA sequencing and protein identification. Male and female homozygous Cd47 knockout rats mated to generate more Cd47 knockout rats for the experiments in the current study in comparison with wild-type rats of the same age. Sequence (5’→3’) for primers for gene sequencing were GTGTCTGTATTCCAGGTTCAGCTC for primer T1 and TCCAGATGGCCAGTCCACCT for primer T2.

During the breeding process, Cd47 knockout rats did not exhibit any obvious differences from wild-type rats in terms of food intake, growth rate, body weight, and litter size.

CapitalBio Technology (Beijing, China) was commissioned to conduct this experiment according to the protocol. Briefly, rat tail tissues from a Cd47 knockout rat and a wild-type rat were ground in liquid nitrogen and lysed using protein extraction buffer (8 M urea, 0.1% SDS) containing 1 mM PMSF and a cocktail of protease inhibitors (Roche, USA). BCA assay was used for protein concentration measurement. Relative quantification of proteins was performed by Tandem mass tag TMT6/10 (Pierce, USA) labeling, which uses different reporter ions (126-131 Da) as isobaric tags. After lyophilization, protein samples were dissolved in solvent A (2% ACN, pH10) and RPLC was developed on Xbridge PST C18 Column (130 Å, 5 μm, 250×4.6 mm) (Waters, USA) with a gradient of 5 to 95% solvent B (90% ACN, pH10). Fractions were collected and vacuum dried, and LC-MS/MS analysis was performed using a Q Exactive mass spectrometer (Thermo Scientific, USA). The peptide mixture was loaded onto an Acclaim C18 PepMap100 nano-Trap column (75 μm×2 cm, 2 μm particle size) connected to an Acclaim PepMap RSLC C18 analytical column (75 μm×25 cm, 2 μm particle size) (Thermo Scientific, USA). The nano-LC was coupled online with the Q Exactive mass spectrometer. Mass spectrometry analysis was performed with full scans (350-1,600 m/z) acquired at a mass resolution of 70,000 at 400 m/z in Q Exactive. The twenty most intense precursor ions were selected for MS/MS and detected at a mass resolution of 35,000 at m/z of 400 in the Orbitrap analyzer. Database searching against the NCBI’s RefSeq human protein sequence database was performed using proteome discoverer software (version 1.4) (Thermo Scientific, USA). Trypsin was used as the digesting enzyme with a tolerance of 2 missed cleavages. Fixed modifications include cysteine carbamidomethylation and TMT modifications (N-terminus and lysine residues) and variable modifications include methionine oxidation. For protein quantitation, unique peptides were used to quantify proteins.

RNAseq analysis was performed by CapitalBio Technology (Beijing, China). Briefly, synovial membrane samples were prepared from the hind paws of rats in the four groups (n=3 for each group). For mRNA library construction and deep sequencing, RNA samples were prepared by using the TruSeq RNA Sample Preparation Kit according to the manufacturer’s protocol. Briefly, the poly-A containing mRNA molecules were purified from the 3 μg of total RNA by using poly-T oligo-attached magnetic beads. The cleaved RNA fragments were reversely transcribed into first-strand cDNA using random hexamers, followed by second-strand cDNA synthesis using DNA Polymerase I and RNase H. The cDNA fragments were purified, end-blunted, ‘A’ tailed, and adaptor-ligated. PCR was used to selectively enrich those DNA fragments that have adapter molecules on both ends and to amplify the amount of DNA in the library. The number of PCR cycles was minimized to avoid skewing the representation of the library. The library was qualified by Agilent 2100 bioanalyzer and quantified by Qubit and qPCR. The produced libraries were sequenced on the Novaseq 6000 instrument (Illumina).

The data was presented as the mean ± SEM or mean ± SD. All statistical tests were performed using SPSS19.0 (IBM) and GraphPad Prism 8.0 software. These results were performed by using a two-tailed Student’s t-test or one way ANOVA. P<0.05 was considered statistically significant.