Identification of MYH9 as a key regulator for synoviocyte migration and invasion through secretome profiling

Objectives ‘Invasive pannus’ is a pathological hallmark of rheumatoid arthritis (RA). This study aimed to investigate secretome profile of synovial fibroblasts of patients with RA (RA-FLSs), a major cell type comprising the invasive pannus. Methods Secreted proteins from RA-FLSs were first identified using liquid chromatography-tandem mass spectrometry analysis. Ultrasonography was performed for affected joints to define synovitis severity at the time of arthrocentesis. Expression levels of myosin heavy chain 9 (MYH9) in RA-FLSs and synovial tissues were determined by ELISA, western blot analysis and immunostaining. A humanised synovitis model was induced in immuno-deficient mice. Results We first identified 843 proteins secreted from RA-FLSs; 48.5% of the secretome was associated with pannus-driven pathologies. Parallel reaction monitoring analysis of the secretome facilitated discovery of 16 key proteins related to ‘invasive pannus’, including MYH9, in the synovial fluids, which represented synovial pathology based on ultrasonography and inflammatory activity in the joints. Particularly, MYH9, a key protein in actin-based cell motility, showed a strong correlation with fibroblastic activity in the transcriptome profile of RA synovia. Moreover, MYH9 expression was elevated in cultured RA-FLSs and RA synovium, and its secretion was induced by interleukin-1β, tumour necrosis factor α, toll-like receptor ligation and endoplasmic reticulum stimuli. Functional experiments demonstrated that MYH9 promoted migration and invasion of RA-FLSs in vitro and in a humanised synovitis model, which was substantially inhibited by blebbistatin, a specific MYH9 inhibitor. Conclusions This study provides a comprehensive resource of the RA-FLS-derived secretome and suggests that MYH9 represents a promising target for retarding abnormal migration and invasion of RA-FLSs.


ELISA
CCL2, IL6, IL8, TGFβ, IL1β, and TNFα concentrations in RA (n = 80) and OA (n = 40) SFs were measured using ELISA kits (R&D systems). MYH9 levels in RA and OA SFs or culture supernatants of human monocytes and RA-FLSs were also determined using ELISA kits (MyBiosource, San Diego, CA, USA) according to the manufacturer's instructions.

Protein digestion
Proteins were digested using a filter-aided sample preparation method [4] with slight modifications. Briefly, proteins were reduced at 37 °C using SDT buffer (4% SDS in 0.1M Tris-HCl [pH 7.6] and 0.1 M DTT) for 45 min and then boiled at 95 °C for 10 min. Subsequently, protein samples were sonicated in a bath sonicator (vibra cell; Sonics) for 10 min and centrifuged at 16,000 ×g for 5 min. Protein samples were transferred to a membrane filter device (YM-30; Millipore) and mixed with 200 μl of 8 M urea in 0.1M Tris-HCl (pH 8.5). The device was centrifuged at 14,000 ×g for 60 min at 20 °C to remove SDS, which was repeated three times. Subsequently, proteins were alkylated with 100 μl of 50 mM iodoacetamide in 8 M urea for 25 min at room temperature in the dark, followed by centrifugation at 14,000 ×g for 30 min. The filter was washed with 200 μl of 8 M urea four times and then washed with 100 μl of 50 mM NH 4 HCO 3 twice for buffer exchange. Trypsin (Promega) was added to proteins at an enzyme-to-protein ratio of 1: 50 (w/w). The filter device was placed in a thermomixer (Eppendorf) and incubated at 37 °C overnight. After the first digestion, the second digestion was performed using additional trypsin (1:100 enzyme-to-protein ratio) at 37 °C for 6 h. After digestion, tryptic peptides were eluted via centrifugation at 14,000 ×g for 30 min at 20 °C. After collecting the tryptic peptides, the filter was rinsed with 60 μl of 50 mM NH 4 HCO 3 and centrifuged at 14,000 ×g for 20 min at 20 °C. The eluent was combined with the first eluent. The combined eluent was dried via vacuum centrifugation. Peptide concentration was determined using a BCA assay. The peptide sample was aliquoted into Eppendorf tubes (10 µg per tube), dried using vacuum centrifugation, and stored at -80 °C.

LC-MS/MS analysis
Peptides (1 μg) from each of 24 fractions were dissolved in solvent A (2% acetonitrile and 0.1% formic acid). Nano-LC-MS/MS analyses were performed using a Q Exactive Mass Spectrometer (Thermo Fisher Scientific) equipped with an EASY-Spray Ion Source and coupled to an EASY-nLC 1000 chromatograph (Thermo Fisher Scientific). Peptides were loaded onto an Acclaim PepMap 100 pre-column (75 μm × 2 cm, C18, 3-μm particles, 100 Å pore size) and separated on an ES800 Easy-Spray column (50 cm × 75 μm inner diameter, PepMap C18, 3-μm particles, 100 Å pore size). A 120-minute gradient was used at a flow rate of 300 nl/min: from 2-40% solvent B (98% acetonitrile and 0.1% formic acid) over 90 min, from 40 to 80% solvent B over 10 min, 80% solvent B for 10 min, and 2% solvent B for 10 min. The temperature of the column was maintained at 35 °C. The electrospray voltage was set at 1.9 kV. MS precursor scans were acquired using the following settings: m/z range = 450-2000 Th, automated gain control (AGC) target value = 1.0 × 106, resolution = 70,000, and maximum ion injection time = 100 ms. For each MS scan, MS/MS data were acquired up to the ten-most abundant ions in a data-dependent mode using higher energy collisional dissociation with the following settings: normalized collision energy = 25, resolution = 17,500, AGC target value = 1.0 × 105, and maximum injection time = 50 ms.

LC-MS/MS data analysis
For each MS/MS dataset, post-experiment monoisotopic mass refinement was applied to accurately assign precursor masses to MS/MS spectra. [5] Resulting MS/MS spectra (i.e., mgf files) were subjected to a database search using the MS-GF+ search engine (v2017.01.13) [6] in BMJ Publishing Group Limited (BMJ) disclaims all liability and responsibility arising from any reliance Supplemental material placed on this supplemental material which has been supplied by the author(s) Search results from 24 MS/MS datasets were combined. Peptide spectrum matches were obtained using an FDR of 1%. The proteomics data were deposited to the ProteomeXchange with accession ID: PXD041077

Identification of subtypes of RA based on RNA-seq data
We first identified "expressed genes" as genes with TPM values of > 1 in more than 50% of the 152 patients with early and established RA using previously reported RNA-seq data (GSE89408). [7] To select genes with significant variations, we calculated median absolute deviations (MADs) and then selected the expressed genes with top 10% (MAD10), 20% (MAD20), and 30% (MAD30) of MADs. Next, we performed an orthogonal non-negative matrix factorization (ONMF) clustering [8] to the log2-fold-change matrix for the selected genes. ONMF was performed iteratively to produce consensus matrices which were used to calculate the cophenetic correlation coefficient for varying number of clusters (k = 2-6). Based on the cophenetic coefficient, we determined the final number of clusters (i.e., subtypes of RA) as previously described. [9] PRM analysis of RA-FLS secretome in SFs Depletion of abundant proteins SF samples from 117 patients with RA and 45 patients with OA were centrifuged at 1,500 ×g at room temperature for 15 min. Supernatants were filtered using 0.22 μm filters (Millipore, Burlington, MA, USA). Filtered samples were stored at -80 °C until further analysis. Samples were depleted to remove the 14 most abundant proteins (albumin, haptoglobin, transferrin, IgA, IgG, alpha 1-antitrypsin, alpha 2-antitrypsin, alpha 1-acid glycoprotein, apolipoprotein A1, apolipoprotein A2, complement C3, IgM, transthyretin, and fibrinogen) using the Human 14 multiple affinity removal spin cartridge (Agilent Technologies, Santa Clara, CA, USA). Depleted proteins were then washed and concentrated using 3 kDa MWCO filters (Amicon, Millipore). Depleted SF protein concentration was estimated using a bicinchoninic acid (BCA) assay and then subjected to filter-aided sample preparation digestion.

PRM analysis
PRM analyses were performed at the PRM mode using the same instruments used for global secretome profiling as described above. The inclusion list included m/z values, charge states, and retention time for 436 precursor ions of 151 target proteins, which were obtained from the global secretome analysis. The solvent compositions, gradient, and MS operation parameters for global secretome profiling were used here as described above. For quantification of target peptides, MS raw files were imported to Skyline (ver. 19.1.0). Precursor-product ion chromatograms were extracted, and peak areas were estimated using the targeted acquisition method. [10] Peptides with at least four fragment ions (dot-product score of ≥ 0.8) were used for peak area quantification. The abundance of a target peptide was then estimated as the sum of peak areas for individual fragment ions. Peak picking and integration boundaries were manually inspected. The PRM analysis data were deposited to the ProteomeXchange with accession ID: PXD041077

Identification of DEPs between RA and OA SFs
Peak areas of target peptides were converted to log 2 -areas and then normalized using the average chromatographic precursor intensity of each dataset. Using normalized peak areas while excluding zero values, differentially expressed peptides between different conditions were identified using a statistical method described previously. [11] For each peptide, two-tailed Student's t-test was applied to calculate the T-values for each comparison. To compute p-values for these T-values, we estimated an empirical distribution of T-values for the null hypothesis (a peptide is not differentially expressed) by performing 100,000 random permutations of the samples and then by applying the Gaussian kernel density estimation method to T-values obtained from random permutations. [12] For each peptide, FDRs were computed via the two-BMJ Publishing Group Limited (BMJ) disclaims all liability and responsibility arising from any reliance Supplemental material placed on this supplemental material which has been supplied by the author(s) sided test using the empirical null distribution using Storey's method [13]. Differentially expressed peptides were selected based on an FDR ≤0.05 and absolute log 2 -fold-changes of  0.33 (1.26fold change), which corresponds to the mean of 2.5 th and 97.5 th percentiles of the null distribution of log 2 -fold changes.

Subtype identification
We first defined "core samples" for each subtype as those with positive silhouette scores. To identify signature genes defining each subtype of the core samples, we compared log2-foldchanges in the core samples of the subtype with those of the other subtypes. The comparison was performed using a previously reported integrative statistical hypothesis testing method [14] that computes adjusted p values from two sample t-test and the median ratio test and then combines the adjusted p values into an overall p value. From the comparison, the final sets of gene signatures were selected as the genes with 1) overall p value of < 0.05, 2) median value of patients in the subtype larger than zero, and 3) median value of the patients in the subtype larger than that of the patients in the other subtypes. We next performed gene set enrichment analysis of the gene signatures for each subtype using ConsensusPathDB [15] and identified cellular pathways significantly (P < 0.01) represented by the gene signature.

Cell-type deconvolution analysis
A previously reported single cell RNA-seq dataset of synovial tissues from patients with RA was obtained from the Import database (accession ID: SDY998). [16] We used a unique molecular identifier (UMI) count matrix including molecules with ≥ 10 reads per UMI. The count matrix was filtered and analyzed using Seurat (version 4.1.0). [17] Quality control, normalization, and clustering were performed as described in the original study. [16] Briefly, we excluded low-quality cells (genes/cell <200, genes/cell > 1,810, cells/genes < 3, and cells with > 25% of molecules derived from mitochondrial genes). Gene expression levels were normalized across the filtered cells using default parameters of the Seurat-intrinsic LogNormalize function, and significantly variable genes were detected using the FindVariableFeatures function. The cells were then clustered using a default method in Seurat into five populations: fibroblasts, monocytes, T cells, B cells, and plasmablasts. Finally, for deconvolution of the five cell types in bulk mRNA data, CIBERSORTx was applied using the five cell populations and the count matrix with the absolute mode, S-mode batch correction, and 500 permutations.

Immunohistochemical analysis
Paraffin-embedded blocks of RA and OA synovial tissues were sectioned (5 μm thickness). These tissue sections were deparaffinized in xylene and rehydrated in a graded series of ethanol solutions. Antigen retrieval was performed by heating sections in citrate buffer using a microwave. Endogenous peroxidases were blocked by incubating sections in 3% H 2 O 2 solution for 30 min at room temperature. These tissue sections were then blocked with 10% normal donkey serum (Jackson ImmunoResearch, West Grove, PA, USA) for 1 h at room temperature followed by incubation with rabbit anti-MYH9 Ab (1:3000; catalog 11128-1-AP, Proteintech) as a primary Ab at 4 °C overnight. Slides were washed three times with PBS and then incubated with a biotinylated anti-rabbit IgG secondary Ab (VECTASTAIN Elite ABC HRP kit; Vector Laboratories). A chromogenic substrate 3,3′-diaminobenzidine tetrahydrochloride (Vector Laboratories) was used for visualization. Nuclei were stained with Mayer hematoxylin. Slides were dehydrated, cleared, and mounted. Images were obtained using a Panoramic MIDI slide scanner (3DHISTECH, Budapest, Hungary).

Cell-spreading assays
For cell-spreading assays, RA-FLSs (1 × 10 4 cells per well) were dissociated by Accutase (catalog AT-104; Innovative Cell Technologies, San Diego, CA, USA) and seeded onto 12 mm-pi cover glasses (catalog 0111520; Marienfeld, Lauda-Königshofen, Germany) in a 24-well cell culture plate as previously described. [19]. The cover glasses were coated with fibronectin (20 μg/mL; catalog 1105147001; Merck, Kenilworth, NJ, USA) for 1 h at room temperature before used. The cells were incubated for 20, 40, 60 and 120 min in DMEM supplemented with 10% FBS. The spreading of the cells was stopped by fixation using 4% paraformaldehyde in PBS (Wako Pure Chemicals) for 20 min at room temperature. The staining process to visualize F-actin and MYH9 was the same as described above in "Immunocytochemistry".

Cell migration and invasion assays
Wound migration of RA-FLSs was measured as described previously. [20] In brief, RA-FLSs were seeded onto 6-well plates at a density of 1.5 × 10 5 cells per well. After reaching approximately 90% confluence, the cells were transfected with MYH9 or control siRNA (50 nM) for 24 h. After transfection, RA-FLSs were wounded by scratching cells with 200 μL sterile pipette tips and incubated with DMEM supplemented with 1% FBS containing IL-1β (1 ng/mL; Thermo Fisher Scientific) and TGFβ (10 ng/mL, Peprotech) for 12 h. For blebbistatin treatment conditions, RA-FLSs seeded onto 6-well plates were wounded as described above and treated with DMEM supplemented with 5% FBS containing 0.5 μM or 1 μM blebbistatin (Selleckchem) or with DMSO for 12 h. Cells were fixed with 4% paraformaldehyde (Wako Pure Chemicals) and stained with crystal violet solution. Images were captured using an EVOS Cell Imaging System (Thermo Fisher Scientific). Cells that migrated over the reference lines were manually counted in three random microscopic fields. Live cell images were obtained using a Lionheart FX automated microscope (Agilent). The percentage of closure area was calculated using the following formula: A t=0 is the initial wound area and A t t is the wound area after n hours of the initial scratch. For Matrigel invasion assays, RA-FLSs were transfected with MYH9 or control siRNA (50 nM) for 24 h or were not transfected for non-muscle myosin II-inhibiting conditions. These cells were then detached and loaded (1.0 × 10 5 cells per well) into upper chambers of a 24-well Matrigel-coated Transwell plate (8 μm pore size; Corning, Corning, NY, USA) in a serum-free medium according to the manufacturer's instructions. Lower chambers of the Transwell plate were filled with DMEM supplemented with 1% FBS to attract cells in upper chambers for siRNA-treated conditions or were filled with DMEM supplemented with 5% FBS containing 0.5 μM or 1 μM blebbistatin (Selleckchem) or containing DMSO. Non-invading cells that remained at the top of the membrane surface were removed by scrubbing with a cotton swab after 12 h of incubation. Cells at the lower surface of the membrane were fixed with 4% paraformaldehyde in PBS (Wako Pure Chemicals), permeabilized with 100% methanol, and stained with 0.1% Crystal violet solution (Sigma-Aldrich). The number of migrating cells was manually counted in three random fields. Images were obtained using a Panoramic MIDI slide scanner (3DHISTECH).

Humanized synovitis in vivo model using SCID mice
Male SCID mice (age, 8 weeks) were obtained from Jackson laboratory and maintained in a specific pathogen-free facility. RA-FLSs were implanted together with normal human cartilages as described previously. [21] For developing a blebbistatin-injected model, RA-FLSs (2 × 10 6 cells per 100 μL) were resuspended in culture medium containing 1 M of blebbistatin (Selleckchem) or DMSO and then implanted with human cartilage into the left flanks of SCID mice. For right flanks, only cartilages were implanted. SCID mice were then intraperitoneally injected with blebbistatin at a dose of 10 mg/kg twice a week for 60 d after the implantation. Implanted tissues were fixed with 4% paraformaldehyde in PBS (Wako Pure Chemicals) for 24 h and embedded in paraffin according to standard procedures. The degree of cartilage destruction was evaluated via hematoxylin and eosin (H&E) staining. Evaluation criteria were the same as those described in a previous study. [22] Briefly, scores were defined as follows: 0 = no invasions, 0.5 = invasion of one to two cell layers, 1 = invasion of three to five cell layers, 1.5 = invasion of three to five cell layers at three independent sites of the cartilage, 2 = invasion of 6 to 10 cell layers, 2.5 = invasion of 6 to 10 cell layers at three independent sites, 3 = invasion of >10 cell layers, 3.5 = invasion of >10 cell layers at two independent sites, 4 = invasion of >10 cell layers at three or more sites of the cartilage.

Induction and evaluation of methylated bovine serum albumin (mBSA)/IL-1β-induced arthritis in mice
Female C57BL/6 mice aged 6 weeks were obtained from Orientbio (Seongnam-si, Gyeonggido, Korea) and maintained in a specific pathogen-free facility. Methylated bovine serum albumin (mBSA) (200 µg; Sigma-Aldrich) was intra-articularly injected into the knee joints using a Hamilton syringe. From the day after the mBSA injection, recombinant IL-1β (250 ng; R&D systems) was administered subcutaneously into the hind footpads for three consecutive days. Then blebbistatin (10 mg/kg; Selleckchem) was intraperitoneally administered once a day for 10 days. Two weeks after mBSA injection, the mice were sacrificed, and the knee joint tissues were obtained. For histological analysis, the tissues were fixed in 4% paraformaldehyde (Biosesang, Seongnam-si, Gyeonggi-do, Korea) overnight at 4°C. The tissues were decalcified with Decalcifying Solution-Lite (Sigma-Aldrich) for 36 h at room temperature, sectioned, stained with hematoxylin and eosin (H&E), and then subjected to histological analysis. The degree of inflammation and bone destruction was evaluated for each joint as described previously: 0 = normal, 1 = weak, 2 = moderate, and 3 =severe. [20] The extent of cartilage destruction also was determined in Safranin O-stained tissues using a scale of 0 to 6 as described previously. [20]

Induction and evaluation of collagen-induced arthritis in mice
Male heat-killed M. tuberculosis at a 1:1 ratio. The mice were intradermally injected with 50 μL of the emulsion (containing 100 μg of CII) at the base of the tail as a primary immunization. Two weeks after the primary immunization, the mice received booster injections with 100 μg of CII in incomplete Freund's adjuvant (catalog 7002, Chondrex) via footpad. From three weeks after primary immunization, the mice were intraperitoneally injected with blebbistaitn (Selleckchem) at a dose of 1 or 10 mg/kg twice a week for 3 weeks. The arthritis severity of each limb was assessed on a 0-to 4-point scale using visual inspection, as previously described. [18]

Statistical analyses
Data are presented as the mean ± SEM. To compare numerical data between groups, the Mann-Whitney U test and the Wilcoxon matched pairs tests were performed. Spearman correlation test was performed to analyze associations between cytokines and MYH9 levels in RA SFs. Two-way ANOVA followed by Sidak's post-test was used to evaluate significance between the curves of wound closure area at different time points. All statistical analyses were performed using GraphPad Prism software v8 (GraphPad Software, San Diego, CA, USA). The p-values < 0.05 were considered significant. CRP, C-reactive protein; GSUS, grey-scale ultrasonography; PDUS, power Doppler ultrasonography; SF, synovial fluid; OA, osteoarthritis; RA, rheumatoid arthritis; DAS, disease activity score; ESR, erythrocyte sedimentation rate; GPI, glucose-6-phosphate isomerase; S100A9, S100 calcium binding protein A9; and SOD2, superoxide dismutase. Consensus clustering heat maps generated from 100 ONMF clustering trials using genes with MAD30%. Notably, cluster memberships for the three subtypes determined using genes with MAD30% were used for the following analyses. "Red" indicates that the corresponding pairs of samples were clustered mostly to the same cluster in the 100 clustering trails, whereas "blue" indicates that the pairs were rarely clustered together.   S  Table S1. Selection of RA-FLS-derived secretory proteins related to "invasive pannus" among the whole RA-FLS secretome (n=843) using gene ontology biological process analysis RA-FLS, rheumatoid arthritis-fibroblast-like synoviocytes.