Synovitis is a key contributor to the inflammatory environment in osteoarthritis (OA) joints. Currently, the biological therapy of OA is not satisfactory in multiple single-target trials on anti-TNF agents, or IL-1 antagonists. Systems biological understanding of the phosphorylation state in OA synovium is warranted to direct further therapeutic strategies. Therefore, in this study, we compared the human synovial phosphoproteome of the OA with the acute joint fracture subjects. We found that OA synovium had significantly more phosphoproteins, and 82 phosphoproteins could only be specifically found in all the OA samples. Differentially expressed proteins of the OA synovium were focusing on endoplasmic reticulum–/Golgi-associated secretion and negative regulation of cell proliferation, which was verified through an IL-1β–treated human synoviocyte (HS) in vitro model. With data-independent acquisition–based mass spectrometry, we found that IL-1β could induce HS to secrete proteins that were significantly associated with the endosomal/vacuolar pathway, endoplasmic reticulum/Golgi secretion, complement activation, and collagen degradation. Especially, we found that while specifically suppressing HS endocytosis, IL-1β could activate the secretion of 25 TNF-associated proteins, and the change of SERPINE2 and COL3A1 secretion was verified by immunoblotting. In conclusion, our results suggest that OA synovium has a polarized phosphoproteome to inhibit proliferation and maintain active secretion of HS, whereas IL-1β alone can transform HS to produce a synovitis-associated secretome, containing numerous TNF-associated secretory proteins in a TNF-independent mode.

Osteoarthritis (OA) is one of the most common and debilitating joint diseases. Other than the typical manifestations, such as cartilage destruction and subchondral bone remodeling, the synovium inflammation or synovitis has now been accepted as a source of proinflammatory mediators that are difficult to contain in progressive OA (13). In this regard, synoviocytes have autoimmune cell properties as they respond to certain stimuli, such as IL-1β and TNF-α, in a destructive manner to become activated and attack bystander healthy cells. Meanwhile, synovium will be abnormally thickened, which in turn harms joint structures and functions (4).

It is known that synovitis in OA can be initiated by cartilage breakdown, with the subsequent phagocytosis of debris, and activation of synoviocytes, leading to the formation of an inflaming microenvironment and synovial fluid (5). Interestingly, Melin et al. (6) recently found with a quantitative secretome analysis that the cartilage itself represents a major inducer of complement proteins upon IL-1α stimulation. Indeed, monocytes/macrophages and dendritic cells are the primary producers of both IL-1 and TNF in the OA joint, but they have a classical self-containing mechanism with NO-induced apoptosis and/or anergy (7). Unfortunately, the excessive NO accumulation can selectively kill chondrocytes and induce cartilage degradation (8). In contrast, inflaming synovial cells tend not to be self-limited by the NO system, and the synovitis is rarely self-healing. All these findings implicate that a continuously homeostatic state of inflammation that lacks negative feedback signals exists in the OA synovium. However, the systems biological feature of such a state has not been fully understood. Specifically, although phosphoprotein changes in chondrocytes have been associated with the mechanism of OA inflammation (9), no phosphoproteome data in human OA synovitis has been published to date.

Particularly, we previously found that the synovium tissue of OA and fractures had differential IL-1β persistence, whereas TNF remained the same (2). With a functional proteomics strategy, we have found that IL-1β can activate the TNF-downstream signals in synoviocytes, which requires no existence of TNF itself (2). Indeed, clinical trials of both anti-TNF and anti–IL-1β therapies on OA synovitis have shown certain efficacy; however, the evidence is not strong enough to draw generalized conclusions on whether such biological therapies are beneficial (5). As IL-1β can bypass TNF to function just like TNF in synoviocytes, OA synovitis tends to be a specialized inflammation and far more complex than a disease with single etiology, which can be solved by a single-target treatment. Hence, investigation of the systems biology feature of OA synovitis is warranted.

As most signaling proteins are phosphoproteins, in this study we analyzed the phosphoproteome of the knee joint synovium tissues of subjects with OA and acute fractures. We recognized that the phosphoproteome of OA synovium had a unique pattern shift from the fracture synovial tissues, focusing on regulating the secretion and inhibiting cell proliferation. Accordingly, with an in vitro IL-1β–treated human synoviocyte (HS) model, we characterized the proliferation, secretome change, and endocytosis capacity of HS. We found that IL-1β can bypass TNF to induce the release of secretory proteins detrimental to cartilage, while reducing the endocytotic capacity of HS.

The scientific and ethics review committee of Jinan University approved this study, and written informed consent was obtained from all study participants. All methods were performed in accordance with the relevant guidelines and regulations. Synovium tissue samples from 13 subjects were obtained from The First Affiliated Hospital of Jinan University. Among them, seven subjects were diagnosed with knee joint OA and sampled upon synovectomy; the other six with acute knee fractures were included for synovial tissue acquisition during surgical operation. For the tissue phosphoproteome analysis, four OA and three fracture samples were used; the other three OA and three fracture samples were used for the immunohistochemistry (IHC) verification. Detailed clinical features and experimental information for each subject can be found in Table I. Specifically, all OA and fracture subjects had no prior history of corticosteroid injection or nonsteroidal anti-inflammatory drug use during the preceding 8 wk.

Prepuce dermis samples from three healthy donors were acquired upon circumcision, and human primary fibroblasts were isolated and cultured as described previously (10).

Synovial tissues were mixed with the 1% SDS lysis buffer, supplemented with 1 mM PMSF, PhosSTOP Phosphatase Inhibitor Mixture (Roche, Shanghai, China), and phosphatase inhibitor (Roche), followed by grounding in liquid nitrogen. Postincubation on ice for 30 min, tissue lysates were centrifuged at 17,000 × g, for 30 min. Supernatants were collected, and the protein concentration was determined by a BCA kit (Thermo Fisher Scientific, Shanghai, China).

We next performed in-solution protein digestion following our published procedure (11, 12). The phosphopeptide enrichment was conducted as we previously reported (12). Briefly, Titansphere TiO particles (GL Sciences, Tokyo, Japan) were used to trap phosphopeptides, which were eluted by 5% NH3·H2O. Phosphopeptides were then fractionated by gradient elutions with 0, 2, 5, 8, 10, and 40% acetonitrile in 5% NH3·H2O. Mono Tip C18 Pipette Tips (GL Sciences) were employed for desalting.

Phosphopeptides were analyzed by a TripleTOF 5600 mass spectrometer (AB Sciex, Framingham, CA). The exact mass spectrometry (MS) parameters can be found in our published work (12). The raw MS data are publicly available through the iProX database (accession number IPX0000873000).

The MS raw data were first searched with the Mascot server version 2.5.1 (Matrix Science, London, U.K.) against the Swiss-Prot HUMAN FASTA database (downloaded on Jan 4, 2016, 20,193 entries). Searching parameters were used as we previously described (12). Mascot search-result DAT files were then loaded into the Scaffold software version 4.5.0 (Proteome Software, Portland, OR) for controlling the protein level false discovery rate (FDR) to <0.01. The normalized spectral abundance factor method (13), as provided by the Scaffold software, was used for phosphoprotein quantification. We employed the Scaffold PTM software version 3.0 (Proteome Software) to determine the confidence of phosphosites using the Ascore algorithm, and only the phosphopeptides with a localization probability >99% were considered confident identifications (12).

We employed the power law global error mode (PLGEM) algorithm (14) to compare the abundance difference at the proteome level in both phosphoproteome and secretome analyses. We have found this method a useful tool for comparative proteomics per our published work, which addresses the detailed PLGEM method (15, 16). A protein was deemed a significantly differentially expressed protein (DEP) when p < 0.05 per PLGEM.

Upregulated DEPs were subjected to the gene ontology (GO) and pathway term network analysis with CluGO + CluePedia (17, 18) in Cytoscape software version 3.4.0, as we previously described with minor modifications (12, 15). Parameters were set as follows: Pathways, Reactome pathway (data: 09.11.2016); evidence code, All_Experimental_(EXP, IDA, IPI, IMP, IGI, IEP); GO term fusion, applied; pathways p value cutoff, p ≤ 0.01; p value correction algorithm, Benjamini–Hochberg; κ Score ≥0.7; leading group term was based on percentage of genes per term.

IHC was performed as we previously described with minor modifications (2). In brief, sections were sequentially treated with the primary Ab rabbit anti-pFLNC(S2233) pAb (1:100; Biorbyt, San Francisco, CA), HRP-conjugated goat anti-rabbit mAb (1:3000; Bioworld), the Vectastain ABC kit (Vector Laboratories, Burlingame, CA), and the ImmPACT Diaminobenzidine (DAB) Substrate (Vector Laboratories). ImageJ (19) and the IHC Image Analysis Toolbox (20) were used to quantify the grayscale of DAB staining in the synovium tissue. The thickness of the synovium was determined by the Measure tool provided by ImageJ.

Primary HS were purchased from ScienCell (Carlsbad, CA; lot number 2070), and maintained in Synoviocyte Medium (ScienCell) consisting of 500 ml of basal medium, 10 ml of FBS, 5 ml of synoviocyte growth supplement, and 5 ml of penicillin/streptomycin solution (all from ScienCell), as we previously described (2). Per the manufacturer’s instructions, these primary HS were isolated from a healthy nonarthritic subject; they were CD90 and fibronectin positive, although negative for HIV type 1, HBV, HCV, mycoplasma, bacteria, yeast, and fungi. Only HS with <5 population doublings were used. In addition, culture surfaces were treated with poly-l-lysine (ScienCell) at 2 μg per cm2. Either IL-1β or TNF-α (Sino Biological, Beijing, China) was used at 10 ng/ml per cell modeling needs.

The immunoblotting (IB) analysis was performed as we previously described (16). The primary Abs and their working dilution are listed as follows: mouse anti-human proliferating cell nuclear Ag (PCNA) mAb (1:5000; Thermo Fisher Scientific), rabbit anti-human p38 MAPK pAb (1:1000; Cell Signaling Technology, Shanghai, China), rabbit anti-human phospho-p38 MAPK (Thr180/Tyr182) mAb (1:1000; CST), rabbit anti-human JNK1/2 mAb (1:1000; CST), mouse anti-human phospho-SAPK/JNK (Thr183/Tyr185) mAb, rabbit anti-human ERK1/2 (phospho-T202/Y204) mAb (1:1000; CST), rabbit anti-human Collagen Type III (C-terminal) pAb (1:1000; Proteintech, Wuhan, China), rabbit anti-human SERPINE2 pAb (1:1000; Proteintech), and rabbit anti-human α-tubulin pAb (1:1000; CST). Secondary Abs were HRP-conjugated goat anti-rabbit mAb (1:2000; Bioworld) and HRP-conjugated goat anti-mouse mAb (1:2000; Bioworld).

The proliferation of HS was analyzed by a WST-1 Assay Kit (Beyotime, Shanghai, China), following the manufacturer’s instructions. In detail, cells were cultured in 96-well plates at 2000 cells per well in 100 μl culture medium. Cells were treated with 10 ng/ml IL-1β for 24 or 48 h, followed by addition of 20 μl WST-1 solution and 2 h incubation. The light absorbance was measured at 450 and 630 nm, respectively.

Cytometric bead array was used to measure the supernatant concentration of IL-8, IL-1β, IL-6, IL-10, TNF, and IL-12p70 by a Human Inflammatory Cytokines Kit (BD, Guangzhou, China), as we previously described (2, 15). Samples were analyzed by using a C6 flow cytometer (BD) and the FCAP Array software (version 3.0.1; BD).

HS cells were cultured to reach 60–70% confluency, and washed four times with the serum- and phenol red–free DMEM medium (Invitrogen), and incubated in the same medium at 37°C for 24 h. Supernatants were harvested, with addition of PMSF to reach a final concentration of 1 mM, followed by sequential centrifugations at 300 × g, for 10 min, and 15,000 × g, for 30 min, to remove debris. The supernatant was passed through a 0.22 μm filter, and concentrated to 500 μl using an Amicon Ultra 3K device (Merck Millipore, Guangzhou, China), followed by three washes with 50 mM triethylammonium bicarbonate for buffer change. Post–freeze-dry, secretory proteins were resuspended by the SDS lysis buffer (Beyotime), supplemented with 1 mM PMSF and protease inhibitor mixture (Roche). The protein samples were then digested with the same method as the phosphoproteome analysis described above.

HS secretome peptides were mixed with the retention time of standard peptides provided by the iRT-Kit (Biognosys, Schlieren, Switzerland) at a volume ratio of 9:1. To build the precursor ion library, a pooled sample was prepared by mixing the same amount of protein from each sample of the IL-1β untreated and treated groups. This pooled sample was analyzed by an Orbitrap Fusion Lumos (Thermo Fisher Scientific) in a data-dependent acquisition (DDA) mode for four repetitive runs. The parameters for the precursor ion analysis (MS) were set as follows: ion source type, NSI; spray voltage, static; positive ion, 2000; negative ion, 600; ion transfer tube temp, 320°C; detector type, orbitrap; orbitrap resolution, 60,000; mass range, normal; scan range, 400–1500 m/z; RF lens, 30%; AGC target, 4.0 × 105; maximum injection time, 50 ms; include charge state(s), 2–7; intensity threshold, 5.0 × 104; data-dependent mode, top speed. For the product ion analysis (MS/MS), the parameters were as follows: MS/MS isolation mode, quadrupole; activation type, HCD; HCD collision energy, 30%; detector type, orbitrap; orbitrap resolution, 15,000; AGC target, 5.0 × 104; maximum injection time, 30 ms.

Next, individual samples were analyzed in the data-independent acquisition (DIA) mode with the same instrument. The parameters were largely the same as the DDA analysis, except the following: MS scan range, 350–1200 m/z; MS/MS scan range, 200–2000 m/z; MS/MS maximum injection time, 50 ms.

To build the spectral library, the DDA raw files were searched using the Sequest HT (v2.5) engine in the Proteome Discoverer software version 2.1 (Thermo Fisher Scientific) against a customized reference database incorporating the Swiss-Prot HUMAN FASTA database (downloaded on Nov 24, 2016, 20,120 entries) and the iRT standard peptides sequence in the FASTA format. Search parameters were set as follows: MS tolerance, 10 ppm; fragment mass tolerance, 0.02 Da; enzyme, trypsin; static modifications, carbamidomethyl (C); dynamic modifications, oxidation of methionine, deamination of Q and N, and acetyl of the N terminus. Peptides were filtered for high confidence (FDR < 1%), and the minimum peptide length was 9 aa. Confident protein identification should meet the following criteria: 1) protein level FDR < 1%; 2) unique peptides ≥2. The pdResult files of the DDA searches were exported from the PD software, and imported into the Spectronaut software version 10 (Biognosys) to build the ion spectral library.

Next, we converted the DIA raw files into an htrm format by using the HTRMS Converter provided by the Spectronaut. Finally, the DIA htrm files, DDA raw files, DDA pdResult files, and the customized Swiss-Prot HUMAN database + iRT standard peptides FASTA file were collectively loaded into the Spectronaut, and processed with the BGS factory settings with minor modification by changing the enzyme from trypsin/P to trypsin. The proteins were inferred by the software, and the quantification information was acquired at the protein level by using the q-value <0.01 criteria, which was used for subsequent analyses.

The amino acid sequence and mass were downloaded from the Swiss-Prot human protein database. The MATLAB bioinformatics toolbox (MathWorks, Natick, MA) was employed to calculate the isoelectric point (pI) and the charge at the physiological conditions (pH 7.4) of each protein as we previously described (11). In addition, we used SignalP 4.1 server and TMHMM 2.0 to predict signal peptides and transmembrane domains as we previously described (21).

We performed Ingenuity Pathway Analysis (IPA; www.ingenuity.com, Qiagen, Shanghai, China) with the Core Analysis module to discern the bioprocess, disease, and upstream regulators, as we described previously (2, 12, 22). The Connect Tool provided by IPA was used to find any association between proteins per IPA knowledge base.

The network analysis was performed in the Cytoscape environment with the ReactomeFIPlugIn (version 5.1.0.β) (23, 24), as we previously described, with minor modifications (15). The 2015 version of functional interactions was used.

The fluorophore conjugated beads with sizes of 0.2 μm (fluorescent at 660/680; Thermo Fisher Scientific) and 0.5 μm (fluorescent at 505/515; Thermo Fisher Scientific) were used to treat the HS for 2 h prior to the flow cytometry analysis with a C6 flow cytometer (BD). Cells were gated based on forward scatter (FSC) and side scatter to exclude debris, and bead-untreated cells were used as the negative control for fluorescence determination.

The bootstrap comparison was performed with the MATLAB R2016b software (MathWorks). Intergroup difference was examined by either parameter or nonparameter tests based on the normality tests by using GraphPad Prism software version 6.02 (GraphPad Software, San Diego, CA) or MATLAB. Statistical difference was accepted when p < 0.05.

With liquid chromatography MS/MS analysis, a total of 951 phosphoproteins could be identified as a union set, i.e., a protein would be counted if it was identified from any of the seven samples (http://www.iprox.org/). Among them, 537 and 239 phosphoproteins were confidently identified and quantified across all OA samples, and all fracture samples, respectively (Fig. 1A; http://www.iprox.org/). In overlap, 219 phosphoproteins were identified across all seven samples (Fig. 1A, Table I). Interestingly, when analyzing tissue samples at the individual donor level, the OA samples had significantly more phosphoproteins than the fracture samples (Fig. 1B). The abundance of the 219 overlapped phosphoproteins could be well fitted by the PLGEM model (r2 = 0.984) (Fig. 1C), with no observed bias in terms of rank of mean (Fig. 1D). In addition, the residues were found to be normally distributed (Fig. 1E). With these features, we could only discern 12 significant DEPs (p < 0.05) per PLGEM (Table II). But we found that 82 phosphoproteins were consistently and exclusively expressed in the OA synovium tissues (Fig. 1F; http://www.iprox.org/), whereas there were 645 proteins that could be identified in one of the three fracture, but not the OA, samples (Fig. 1F). Hence, these 82 proteins were also treated as DEPs for the subsequent analyses.

FIGURE 1.

Phosphoproteome comparison of human synovium tissues from OA and fracture subjects. (A) Venn diagram comparison of phosphoprotein identifications. Numbers indicate the phosphoproteins identified across all samples within each group, respectively. (B) Comparison of the phosphoprotein numbers for MS analyses on individual subjects. As data did not follow normal distribution, the nonparameter Kolmogorov–Smirnov test, two-tailed, was used. All data are shown as mean ± SEM, OA donor n = 4, and fracture donor n = 3. (C) PLGEM fitting of the abundance of the 219 phosphoproteins quantified across samples from all subjects. (D) Residuals versus rank of protein mean abundances. (E) Quantile-quantile (Q-Q) plot. (F) Venn diagram comparison of phosphoproteins identified across all OA subjects, but not detected in any subject in the fracture group.

FIGURE 1.

Phosphoproteome comparison of human synovium tissues from OA and fracture subjects. (A) Venn diagram comparison of phosphoprotein identifications. Numbers indicate the phosphoproteins identified across all samples within each group, respectively. (B) Comparison of the phosphoprotein numbers for MS analyses on individual subjects. As data did not follow normal distribution, the nonparameter Kolmogorov–Smirnov test, two-tailed, was used. All data are shown as mean ± SEM, OA donor n = 4, and fracture donor n = 3. (C) PLGEM fitting of the abundance of the 219 phosphoproteins quantified across samples from all subjects. (D) Residuals versus rank of protein mean abundances. (E) Quantile-quantile (Q-Q) plot. (F) Venn diagram comparison of phosphoproteins identified across all OA subjects, but not detected in any subject in the fracture group.

Close modal
Table I.
Donor clinical features and sample applications
Donor No.AgeGenderDiagnosisApplication
35 Male Fracture MS 
58 Female Fracture MS 
85 Male Fracture MS 
89 Male OA MS 
76 Male OA MS 
80 Male OA MS 
67 Male OA MS 
28 Male Fracture IHC 
56 Male Fracture IHC 
10 64 Male OA IHC 
11 66 Female OA IHC 
12 53 Female OA IHC 
13 39 Male Fracture IHC 
Donor No.AgeGenderDiagnosisApplication
35 Male Fracture MS 
58 Female Fracture MS 
85 Male Fracture MS 
89 Male OA MS 
76 Male OA MS 
80 Male OA MS 
67 Male OA MS 
28 Male Fracture IHC 
56 Male Fracture IHC 
10 64 Male OA IHC 
11 66 Female OA IHC 
12 53 Female OA IHC 
13 39 Male Fracture IHC 
Table II.
Differentially expressed phosphoproteins of human synovium
Swiss-Prot IDProtein NameOA/Fracture RatioPLGEM p ValueMass (kDa)Isoelectric PointChargeHGNC Gene Name
Q9UPQ0 LIM and calponin homology domains-containing protein 1 4.08 1.374 × 10−2 121.9 6.41 −8.67 LIMCH1 
P35606 Coatomer subunit β′ 3.23 2.899 × 10−2 102.5 4.93 −33.89 COPB2 
Q63ZY3 KN motif and ankyrin repeat domain-containing protein 2 2.82 4.468 × 10−2 91.2 5.33 −18.83 KANK2 
Q14315 Filamin-C 2.33 4.530 × 10−2 291.0 5.82 −49.46 FLNC 
O75475 PC4 and SFRS1-interacting protein 0.42 4.939 × 10−2 60.1 9.86 14.57 PSIP1 
Q92954 Proteoglycan 4 0.46 4.917 × 10−2 151.1 10.16 66.60 PRG4 
Q8WVC0 RNA polymerase-associated protein LEO1 0.37 2.822 × 10−2 75.4 4.12 −99.87 LEO1 
P68871 Hemoglobin subunit β 0.27 1.917 × 10−2 16.0 7.32 0.37 HBB 
Q969G5 Protein kinase C δ-binding protein 0.3 1.851 × 10−2 27.7 6.39 −2.18 PRKCDBP 
Q08495 Dematin 0.21 1.007 × 10−2 45.5 9.36 6.38 DMTN 
P02794 Ferritin H chain 0.1 5.541 × 10−3 21.2 5.29 −9.51 FTH1 
P24821 Tenascin 0.11 2.257 × 10−3 240.9 4.52 −112.17 TNC 
Swiss-Prot IDProtein NameOA/Fracture RatioPLGEM p ValueMass (kDa)Isoelectric PointChargeHGNC Gene Name
Q9UPQ0 LIM and calponin homology domains-containing protein 1 4.08 1.374 × 10−2 121.9 6.41 −8.67 LIMCH1 
P35606 Coatomer subunit β′ 3.23 2.899 × 10−2 102.5 4.93 −33.89 COPB2 
Q63ZY3 KN motif and ankyrin repeat domain-containing protein 2 2.82 4.468 × 10−2 91.2 5.33 −18.83 KANK2 
Q14315 Filamin-C 2.33 4.530 × 10−2 291.0 5.82 −49.46 FLNC 
O75475 PC4 and SFRS1-interacting protein 0.42 4.939 × 10−2 60.1 9.86 14.57 PSIP1 
Q92954 Proteoglycan 4 0.46 4.917 × 10−2 151.1 10.16 66.60 PRG4 
Q8WVC0 RNA polymerase-associated protein LEO1 0.37 2.822 × 10−2 75.4 4.12 −99.87 LEO1 
P68871 Hemoglobin subunit β 0.27 1.917 × 10−2 16.0 7.32 0.37 HBB 
Q969G5 Protein kinase C δ-binding protein 0.3 1.851 × 10−2 27.7 6.39 −2.18 PRKCDBP 
Q08495 Dematin 0.21 1.007 × 10−2 45.5 9.36 6.38 DMTN 
P02794 Ferritin H chain 0.1 5.541 × 10−3 21.2 5.29 −9.51 FTH1 
P24821 Tenascin 0.11 2.257 × 10−3 240.9 4.52 −112.17 TNC 

With GO analysis of the DEPs, we found that the upregulated phosphoproteins of the OA groups were significantly enriched in the terms of Golgi-associated vesicle biogenesis and the downregulation of EGFR and MET (Fig. 2A). These bioprocesses correspond to secretion and compromised cellular proliferation. As a plasma membrane protein, filamin C (FLNc) is potentially associated with intracellular trafficking and secretory processes (25). Among the 12 DEPs shown in Table II, the only commercialized Ab available was against phosphorylated FLNc at Ser2233. We then obtained three additional pairs of clinical samples to perform IHC analysis (Fig. 2B). We found that the OA group had significantly more pFLNC (s2233) staining than the OA group (Fig. 2C), which was consistent with the MS quantitation. Furthermore, we found that the synovium was significantly thicker in the OA group (Fig. 2D).

FIGURE 2.

GO analysis and biological verification of the phosphoproteome. (A) ClueGO annotation of DEPs. Functionally grouped GO/pathway term networks were computed with ClueGO referenced by the Reactome database. The circular size depicts the statistical significance based on percentage of genes per term. Edge thickness represents the degree of connectivity between terms based on the κ score. (B) IHC verification of phosphorylated FLNC. Scale bar, 40 μm. (C) Statistical comparison of the pFLNC expression. The mean grayscale intensity of the DAB staining was analyzed by ImageJ software. The horizontal line indicates the median value of each group, donor n = 3, two-tailed Kolmogorov–Smirnov test. (D) Synovial tissue thickness assay. Data are shown as mean ± SEM, donor n = 3, unpaired Student t test.

FIGURE 2.

GO analysis and biological verification of the phosphoproteome. (A) ClueGO annotation of DEPs. Functionally grouped GO/pathway term networks were computed with ClueGO referenced by the Reactome database. The circular size depicts the statistical significance based on percentage of genes per term. Edge thickness represents the degree of connectivity between terms based on the κ score. (B) IHC verification of phosphorylated FLNC. Scale bar, 40 μm. (C) Statistical comparison of the pFLNC expression. The mean grayscale intensity of the DAB staining was analyzed by ImageJ software. The horizontal line indicates the median value of each group, donor n = 3, two-tailed Kolmogorov–Smirnov test. (D) Synovial tissue thickness assay. Data are shown as mean ± SEM, donor n = 3, unpaired Student t test.

Close modal

As we had shown that tissue phosphoproteome was focusing on endoplasmic reticulum (ER)–/Golgi-associated secretion and proliferation, we next used our well-established IL-1β–induced synovitis in vitro model for further systems analysis (2). We could well reproduce the upregulation of IL-6 and IL-8 in the IL-1β–treated HS (Fig. 3A), justifying the consistency of the cellular model (2). We found that the PCNA expression in HS was significantly suppressed by IL-1β (Fig. 3B), implicating that the proliferation was inhibited in this cellular model. Favorably, after the HS was treated by IL-1β for 48 h, the OD of WST-1 showed significant decline (Fig. 3C). These results showed that IL-1β inhibited HS proliferation, which was consistent with the GO analysis of tissue synovium phosphoproteins.

FIGURE 3.

In vitro modeling of synovitis with IL-1β–treated human synoviocytes. (A) Cytometric bead analysis. (B) Proliferation analysis. Synoviocytes were treated with IL-1β for 24 h, followed by the IB analysis of PCNA. (C) WST-1 assay. (D) Flow cytometry analysis of the cellular size as determined by the FSC. (EG) IB assay on MAPKs, including total and phosphoproteins of p38 (E), ERK1/2 (F), and JNK 1/2 (G). All experiments shown were acquired from three independent experiments, data are shown as mean ± SEM, unpaired Student t test. All IB blots are shown as cropped images, the whole-blot raw images can be found in Supplemental Fig. 1.

FIGURE 3.

In vitro modeling of synovitis with IL-1β–treated human synoviocytes. (A) Cytometric bead analysis. (B) Proliferation analysis. Synoviocytes were treated with IL-1β for 24 h, followed by the IB analysis of PCNA. (C) WST-1 assay. (D) Flow cytometry analysis of the cellular size as determined by the FSC. (EG) IB assay on MAPKs, including total and phosphoproteins of p38 (E), ERK1/2 (F), and JNK 1/2 (G). All experiments shown were acquired from three independent experiments, data are shown as mean ± SEM, unpaired Student t test. All IB blots are shown as cropped images, the whole-blot raw images can be found in Supplemental Fig. 1.

Close modal

Furthermore, we found that the FSC of the IL-1β–treated HS was significantly greater than the untreated HS (Fig. 3D), showing that the cells became bigger post–IL-1β treatment. Such a size change echoed the thickness increase of the OA synovium as shown in Fig. 2D. Additionally, we found that the phosphorylated p38 level was significantly higher in the IL-1β–treated HS (Fig. 3E), whereas p-ERK1/2 (Fig. 3F) and p-JNK2 (Fig. 3G) showed no significant difference. These results on the three MAPKs were comparable with the GO analysis shown in Fig. 2A. The above results indicate that the in vitro cellular model has certain in vivo relevance to OA synovitis.

We next analyzed the secretome of the IL-1β–treated HS following the workflow shown in Fig. 4A. The harvested secretory proteins were generally separated by the SDS PAGE gel in a band manner, suggesting that no obvious degradation had occurred (Supplemental Fig. 1A). With DIA-based liquid chromatography MS/MS analysis of three biological replicates, we observed high reproducibility in protein identifications (http://www.iprox.org/); among the ∼1400 identified proteins in the HS group, 1233 were identified across all MS analyses (Fig. 4B). Similarly, 1206 secretory proteins were consistently identified in all samples of the HS+IL-1β group (Fig. 4B). In total 86.6% proteins (1132) could be identified across all of the six samples of the two groups. These results justified that DIA MS in secretome analysis had excellent reproducibility. Furthermore, 11 secretory proteins were identified in all three biological replicates of the IL-1β–treated HS group, but not detected in any sample in the untreated HS group (Fig. 4C).

FIGURE 4.

IL-1β induces system changes of the human synoviocyte secretome. (A) Workflow of the functionally proteomic analysis of IL-1β–induced synovitis. (B) Venn diagram comparison of the secretory proteins identified from the three independent experiments. (C) Venn diagram showing the proteins constantly identified in the secretome of IL-1β–treated HS. Numbers indicate the secretory proteins identified across all samples within each group, respectively. (DF) Bootstrap analysis of the physical-chemical features of proteins. A total of 10,000 protein resamplings were performed from the secretomes of the HS group, the HS + IL-1β group, and the human proteome background (PE1 proteins in the neXtProt database), respectively. The isoelectric point (D), charge (E), and molecular weight (F) of these proteins were compared. Stars indicate the mean values, whereas the horizontal lines depict 95% confidence intervals. (G and H) Comparison of the proteins with signaling peptides (G) and transmembrane proteins (H). Statistical difference was examined by the χ2 test.

FIGURE 4.

IL-1β induces system changes of the human synoviocyte secretome. (A) Workflow of the functionally proteomic analysis of IL-1β–induced synovitis. (B) Venn diagram comparison of the secretory proteins identified from the three independent experiments. (C) Venn diagram showing the proteins constantly identified in the secretome of IL-1β–treated HS. Numbers indicate the secretory proteins identified across all samples within each group, respectively. (DF) Bootstrap analysis of the physical-chemical features of proteins. A total of 10,000 protein resamplings were performed from the secretomes of the HS group, the HS + IL-1β group, and the human proteome background (PE1 proteins in the neXtProt database), respectively. The isoelectric point (D), charge (E), and molecular weight (F) of these proteins were compared. Stars indicate the mean values, whereas the horizontal lines depict 95% confidence intervals. (G and H) Comparison of the proteins with signaling peptides (G) and transmembrane proteins (H). Statistical difference was examined by the χ2 test.

Close modal

With bootstrap analyses, we found that as compared with the background human proteome, the DIA MS tended to identify significantly more acidic proteins (Fig. 4D), with more charges (Fig. 4E); however, there was no significant difference in identifying proteins with different masses (Fig. 4F). We found that ∼38% of all secretory proteins had signaling peptides as predicted by the SignalP software, significantly more than the background human proteome (∼18%). Notably, ∼81% of the upregulated DEPs in the secretome were predicted to have signaling peptides (Fig. 4G). However, no significant difference was observed in the analysis of the transmembrane domain-containing proteins (Fig. 4H). These analyses favored the validity of DIA MS on secretome characterization in our cellular model.

We found that the secretory protein abundance determined by the DIA MS could also be well fitted by the PLGEM model, from which we discerned 68 up- and 13 downregulated DEPs (p < 0.01; http://www.iprox.org/). In addition, we counted the 11 secretory proteins shown in Fig. 4C as upregulated DEPs. With the IPA analysis, we found that the 79 upregulated DEPs significantly promoted the diseases and functions that were associated with known OA synoviocyte phenotypes, including cell movement, inflammatory response, organismal injury and abnormalities, and immune cell trafficking (z-score >1.96, Fig. 5A). However, these significantly upregulated DEPs significantly focused on the cell death and survival with z-score <−1.96 (Fig. 5A), suggesting that these DEPs have a negative impact on the function. These results reflected the inflammatory environment contributed by the secretome of inflaming synoviocytes, relevant to the HS invasion and detrimental immune responses.

FIGURE 5.

Pathway analysis of the human synoviocyte secretome. (A) IPA on the upregulated DEPs in the secretome of the IL-1β–treated HS. The significantly involved disease and functions are shown with the z-score heat map. (B) Reactome network analysis.

FIGURE 5.

Pathway analysis of the human synoviocyte secretome. (A) IPA on the upregulated DEPs in the secretome of the IL-1β–treated HS. The significantly involved disease and functions are shown with the z-score heat map. (B) Reactome network analysis.

Close modal

We found that these 79 DEPs could be enriched in four leading GO terms toward essential features of synovitis, including the endosomal/vacuolar pathway, ER/Golgi secretion, complement activation and collagen degradation, and chemokine receptors binding chemokines (Table III). With the Reactome analysis, these DEPs could be enriched and incorporated into a network with 23 nodes with a modularity of 0.7824 (Fig. 5B), suggesting that diversified protein modules were initiated by IL-1β, and the connection between these modules was loose. Most of the modules were also comparable with GO analysis, especially regarding the term-associated genes (Table III), but with tight intramodule protein-protein interactions (Fig. 5B).

Table III.
GO enrichment analysis of HS secretome upon IL-1β treatment
No.GO IDsaLeading GOTermaTerm p ValuebGroup p Valueb% Associated GenesAssociated Genes Found
GO:0000096 Endosomal/vacuolar pathway 3.10 × 10−4 4.60 × 10−4 9.30 [CTSS, MMP1, MMP3, TIMP1] 
GO:0000233 Transport of γ-carboxylated protein precursors from the ER to the Golgi apparatus 1.30 × 10−3 1.40 × 10−3 9.68 [IGFBP1, IGFBP4, MMP1] 
GO:0000399 Activation of C3 and C5 7.40 × 10−5 3.40 × 10−4 14.29 [C3, CD55, CFB, CFH] 
GO:0000166 Collagen degradation 3.90 × 10−7 4.20 × 10−8 5.09 [AGRN, COL16A1, COL3A1, CTSD, CTSS, HAPLN1, ICAM1, LAMC2, MMP1, MMP19, MMP3] 
GO:0000893 Chemokine receptors bind chemokines 4.00 × 10−3 4.20 × 10−3 6.25 [CXCL10, CXCL3, CXCL6] 
No.GO IDsaLeading GOTermaTerm p ValuebGroup p Valueb% Associated GenesAssociated Genes Found
GO:0000096 Endosomal/vacuolar pathway 3.10 × 10−4 4.60 × 10−4 9.30 [CTSS, MMP1, MMP3, TIMP1] 
GO:0000233 Transport of γ-carboxylated protein precursors from the ER to the Golgi apparatus 1.30 × 10−3 1.40 × 10−3 9.68 [IGFBP1, IGFBP4, MMP1] 
GO:0000399 Activation of C3 and C5 7.40 × 10−5 3.40 × 10−4 14.29 [C3, CD55, CFB, CFH] 
GO:0000166 Collagen degradation 3.90 × 10−7 4.20 × 10−8 5.09 [AGRN, COL16A1, COL3A1, CTSD, CTSS, HAPLN1, ICAM1, LAMC2, MMP1, MMP19, MMP3] 
GO:0000893 Chemokine receptors bind chemokines 4.00 × 10−3 4.20 × 10−3 6.25 [CXCL10, CXCL3, CXCL6] 
a

Ontology source was based on Reactome pathway (date: 09.11.2016).

b

Benjamini–Hochberg corrected p values are shown.

With the upstream analysis of IPA, IL-1β was discerned as a significant upstream regulator for 15 upregulated secretory DEPs in the IL-1β–treated HS (z-score = 3.099; Fig. 6A). This meant that from these DEPs, IPA suggested IL-1β should have been added into the cell culture, which was true. Thus, such a prediction could serve as a positive control to justify the validity of our IL-1β–induced synoviocyte inflammation model. In addition, other interesting upregulators included TNF, INFG, MAPKs (p38 and JNK), IKBKB, STAT3, and IL-6. Among them, IL-6 (Fig. 3A) and phosphorylated p38 MAPK (Fig. 3E) had been experimentally verified. We next used the ConsensusPathDB tool to analyze the hierarchy of different upstream regulators by using the shortest interaction path module (26). The results showed that IL-1β could also serve as an upstream regulator of IFNG genes via STAT3 or GATA3 based on the previous source literature (2729). No shortest interaction path between IL-1β and TNF could be built with the same ConsensusPathDB analysis. These results suggested that IFNG tended to be a downstream gene of IL-1β, whereas IL-1β and TNF were independent upstream regulators.

FIGURE 6.

Upstream regulator analysis and verification. (A) Upstream regulators were predicted by IPA based on the secretome DEPs. Activated (z-score >1.96) and inhibited (z-score <−1.96) regulators are highlighted in orange and blue, respectively. The IL-1β– (B) and TNF-regulated (C) HS secretory protein networks are shown. The HGNC names are provided for all node proteins. (D and E) IB verification on SERPINE2 and COL3A1 secretion from HS treated with TNF-α (D) and IL-1β (E), respectively. HS were treated with either TNF-α or IL-1β for 24 h prior to secretory protein harvests. Results were acquired from four independent experiments. All IB blots are shown as cropped images, the whole-blot raw images can be found in Supplemental Fig. 1.

FIGURE 6.

Upstream regulator analysis and verification. (A) Upstream regulators were predicted by IPA based on the secretome DEPs. Activated (z-score >1.96) and inhibited (z-score <−1.96) regulators are highlighted in orange and blue, respectively. The IL-1β– (B) and TNF-regulated (C) HS secretory protein networks are shown. The HGNC names are provided for all node proteins. (D and E) IB verification on SERPINE2 and COL3A1 secretion from HS treated with TNF-α (D) and IL-1β (E), respectively. HS were treated with either TNF-α or IL-1β for 24 h prior to secretory protein harvests. Results were acquired from four independent experiments. All IB blots are shown as cropped images, the whole-blot raw images can be found in Supplemental Fig. 1.

Close modal

Such a prediction of TNF is comparable with our previous cellular proteome analyses, showing that IL-1β can activate intracellular TNF signaling in a TNF-independent manner (2). In this study with the secretome analysis, other than the known IL-1β downstream proteins (Fig. 6B), we detected 25 TNF-associated secretory proteins that could also be induced by IL-1β (Fig. 6C), among which nine proteins were not associated with IL-1β (Fig. 6B, 6C). These nine proteins were CP, SERPINE2, COL16A1, SFRP1, HAPLN1, COL3A1, CTSC, MCAM, and SVIL (Fig. 6B, 6C). In this study, we would emphasize that IL-1β could not induce increased TNF secretion in this cellular model according to Fig. 3A and our previous report (2). With the IPA Connect tool, we found that no connection between IL-1β and the nine TNF-associated secretory proteins could be formed per IPA knowledge base. This meant that according to current knowledge, no findings proposed the association of IL-1β with any of these nine proteins. We also validated this with the STRING analysis (https://string-db.org/) and manual literature searches. To favor our proof of concept, we showed that TNF could promote the secretion of SERPINE2 from the HS through four independent experiments (Fig. 6D). Interestingly, we found that other than SERPINE2, IL-1β could also induce the upregulated secretion of COL3A1 from the HS (Fig. 6E).

Exocytosis and endocytosis are coupled and highly interactive bioprocesses in numerous cell types (30). We then used 0.2 and 0.5 μm fluorophore-conjugated microbeads to test the clathrin- and caveolae-mediated endocytosis, respectively (31, 32). Both types of endocytosis were found to be significantly decreased in the IL-1β–treated HS (Fig. 7A, 7B). However, IL-1β did not mediate changes in human normal fibroblasts in either type of endocytosis (Fig. 7C, 7D).

FIGURE 7.

Endocytosis analysis. (A and B) Endocytosis of 0.2 μm (A) and 0.5 μm fluorophore-beads (B) was evaluated by flow cytometry. Cells were gated in the FSC–side scatter plot to avoid analyses on debris. The untreated human synoviocytes were used as a negative control for gating. (C and D) Endocytosis assay on human fibroblasts. The experimental settings and gating strategies are the same as (A and B). All results were acquired from three independent experiments; p values were acquired by the paired t test, fibroblast donor n = 3, all data are shown as mean ± SEM. (E) Representative chart showing the gating strategy of flow cytometry.

FIGURE 7.

Endocytosis analysis. (A and B) Endocytosis of 0.2 μm (A) and 0.5 μm fluorophore-beads (B) was evaluated by flow cytometry. Cells were gated in the FSC–side scatter plot to avoid analyses on debris. The untreated human synoviocytes were used as a negative control for gating. (C and D) Endocytosis assay on human fibroblasts. The experimental settings and gating strategies are the same as (A and B). All results were acquired from three independent experiments; p values were acquired by the paired t test, fibroblast donor n = 3, all data are shown as mean ± SEM. (E) Representative chart showing the gating strategy of flow cytometry.

Close modal

In this study, we report for the first time, to our knowledge, that OA synovium tends to have a completely different phosphoproteome from the fracture synovium. The differential phosphoproteins focus on regulating the ER-/Golgi-associated secretion, and inhibiting HS proliferation. As a known inducer of OA synovitis, IL-1β can significantly promote the protein secretion from HS, with diversified functional units to regulate endocytosis, secretion, complement activation, and collagen degradation. Especially, we found that IL-1β could also bypass TNF to induce TNF-associated protein secretion, whereas suppressing the endocytosis of HS.

To have direct clinical relevance, we used surgery synovium samples to characterize the inflamed state maintained by the phosphoproteome. As compared with other human cells and tissues, the phosphoprotein number is quite small, ∼400 in the fracture and ∼800 in the OA groups. For example, according to our previous report, colorectal cancer cells can hold >1700 phosphoproteins, and the colorectal cancer cell–secreted exosome can pack >300 phosphoproteins (12). However, our results are comparable with a relevant report on cultured primary rheumatoid arthritis and OA synoviocytes (∼300 phosphoprotein spots in a difference gel electrophoresis analysis) (33). Even with such a small phosphoproteome size, it was interesting that we still found 82 phosphoproteins that were consistently identified in all the OA samples with shotgun MS analysis.

Although HS have very similar features to fibroblasts, they are also professional secretory cells that maintain the homeostasis of synovial fluid in the joint, releasing hyaluronan (34), and chondrocyte-protective proteins such as lubricin (35) and gelsolin (36). However, with the inflaming OA microenvironment, our phosphoproteome results implicated several critical changes of OA synovium. Among them, the enhanced secretory behavior and activation of p38 MAPK were verified in our in vitro model. Interestingly, the downregulation of proliferation predicted by the downregulation of MET and EGFR phosphoproteins was also validated in the IL-1β–treated HS model. This is different from numerous other studies showing that IL-1β promotes the proliferation of human primary rheumatoid arthritis synoviocytes (3739). Our findings suggest that the proliferation of synoviocytes in OA tends to be compromised, whereas the HS cell size can be enlarged by IL-1β.

Our in vitro model indicated that IL-1β alone could switch on the massive changes in HS secretory behavior toward harming the joints. This included the significantly increased secretion of C3 and C5 complement activation regulating proteins, and numerous collagen degradation-associated proteins. Among these secretory proteins, CD55, also known as complement decay-accelerating factor, is a specific marker of HS (40). These results are comparable with another proteomics study by Melin et al. (6) showing that IL-1α–stimulated bovine cartilage releases intensively more activated C3. In numerous proteomics studies on OA synovial fluid (4143), complement components have been deemed primary factors to differentiate OA from the health. Thus, the formation of an inflaming environment in joints involves many cell types, although the synoviocyte after exposure to IL-1β is an important contributor.

Interestingly, we found that IL-1β could partially function just like TNF to regulate HS secretome. This echoes our previous comparative proteomics report on cellular analysis; in addition, such a feature of IL-1β is independent from TNF, as we have proven that IL-1β cannot induce TNF secretion in HS (2). This finding helps to understand the results from current clinical trials on biological OA therapy. Most of these trials targeted either IL-1β or TNF alone. For example, a very recent clinical trial on the TNF blocker, etanercept, has concluded that it has no significant different efficacy from the placebo in erosive inflammatory hand OA (44). As to the anti–IL-1 therapy, although most clinical trials failed to see efficacy, a fully humanized mAb to IL-1R1 AMG 108 showed improvements in the pain of patients with OA in the knee (45). Our current results and previous findings suggest that inhibiting TNF alone cannot completely block its downstream signals and resulting secretory proteins, because IL-1β could still activate TNF-associated secretions. Such a functional degeneracy of the two cytokines suggests that single-target therapy tends not to be sufficient in containing synovitis in OA.

In this study, to our knowledge we not only reported the first secretome of IL-1β–induced HS, but also noted the technological advantage of DIA MS for the confident identification and quantification of secretory proteins. As the identification was referenced by the ion library, we realized 86.6% of reproducible identifications across all six samples in the three independent experiments. This led us to report a confident dataset containing >1100 secretory proteins from HS, which represent the deepest secretome coverage of synoviocytes to date, to our knowledge. The performance of DIA MS is comparable with our previous analysis of nasopharyngeal carcinoma tissues with another DIA platform, SWATH, which identified 1414 proteins across 11 tissue samples (16). In addition, DIA can accurately quantitate the protein abundance (46), and SWATH preferentially detects proteins with lower molecular weights (16). However, we did not observe such a molecular weight preference in this study with the DIA MS analysis by an LTQ Fusion Lumos mass spectrometer.

Finally, we found that IL-1β specifically downregulates the endocytosis capacity of HS, but does not affect fibroblasts. Human normal fibroblasts did not respond to IL-1β in a 24 h treatment duration, which served as a negative control. Endocytosis and exocytosis balance commonly exists in synaptic vesicle trafficking of human cells (47). Our results suggest that IL-1β can polarize the HS to behave more like a secretory cell type, and to function less as a phagocyte.

In conclusion, OA synovium tissue has a completely different pattern of phosphoproteome from the fracture synovium tissue, focusing on promoting secretion and inhibiting proliferation. IL-1β is sufficient to transform HS to a secretory cell type with compromised endocytosis capacity while releasing detrimental proteins. IL-1β can behave like a functional analog of TNF while stimulating the protein secretion of HS in a TNF-independent mode.

This work was supported by the National Basic Research Program 973 of China (Grant 2014CBA02000), the National Natural Science Foundation of China (Grant 81372135), the Key Project for Research and Development of Guangdong Province (Grant 2016B020238002), and the Key Special Project on the Integration of Industry, Education and Research of Guangzhou (Grant 201604020002) to T.W., the Fundamental Research Funds for the Central Universities of China (Grant 21615461), and the National Key Research and Development Program of China (Grants 2017YFA0505100 and 2017YFA0505000).

The mass spectrometry data have been submitted to the iProX database (http://www.iprox.org/) under accession number IPX0000873000.

The online version of this article contains supplemental material.

Abbreviations used in this article:

DAB

diaminobenzidine

DDA

data-dependent acquisition

DEP

differentially expressed protein

DIA

data-independent acquisition

ER

endoplasmic reticulum

FDR

false discovery rate

FLNc

filamin C

FSC

forward scatter

GO

gene ontology

HS

human synoviocyte

IB

immunoblotting

IHC

immunohistochemistry

IPA

Ingenuity Pathway Analysis

MS

mass spectrometry

OA

osteoarthritis

PCNA

proliferating cell nuclear Ag

PLGEM

power law global error mode.

1
Kapoor
,
M.
,
J.
Martel-Pelletier
,
D.
Lajeunesse
,
J. P.
Pelletier
,
H.
Fahmi
.
2011
.
Role of proinflammatory cytokines in the pathophysiology of osteoarthritis.
Nat. Rev. Rheumatol.
7
:
33
42
.
2
Shen
,
S.
,
J.
Guo
,
Y.
Luo
,
W.
Zhang
,
Y.
Cui
,
Q.
Wang
,
Z.
Zhang
,
T.
Wang
.
2014
.
Functional proteomics revealed IL-1β amplifies TNF downstream protein signals in human synoviocytes in a TNF-independent manner.
Biochem. Biophys. Res. Commun.
450
:
538
544
.
3
Mathiessen
,
A.
,
P. G.
Conaghan
.
2017
.
Synovitis in osteoarthritis: current understanding with therapeutic implications.
Arthritis Res. Ther.
19
:
18
.
4
Sellam
,
J.
,
F.
Berenbaum
.
2010
.
The role of synovitis in pathophysiology and clinical symptoms of osteoarthritis.
Nat. Rev. Rheumatol.
6
:
625
635
.
5
Wenham
,
C. Y.
,
M.
McDermott
,
P. G.
Conaghan
.
2015
.
Biological therapies in osteoarthritis.
Curr. Pharm. Des.
21
:
2206
2215
.
6
Melin Fürst
,
C.
,
E.
Åhrman
,
K.
Bratteby
,
S.
Waldemarson
,
J.
Malmström
,
A. M.
Blom
.
2016
.
Quantitative mass spectrometry to study inflammatory cartilage degradation and resulting interactions with the complement system.
J. Immunol.
197
:
3415
3424
.
7
Zeng
,
X.
,
T.
Wang
,
C.
Zhu
,
X.
Xing
,
Y.
Ye
,
X.
Lai
,
B.
Song
,
Y.
Zeng
.
2012
.
Topographical and biological evidence revealed FTY720-mediated anergy-polarization of mouse bone marrow-derived dendritic cells in vitro.
PLoS One
7
:
e34830
.
8
Abramson
,
S. B.
,
M.
Attur
,
A. R.
Amin
,
R.
Clancy
.
2001
.
Nitric oxide and inflammatory mediators in the perpetuation of osteoarthritis.
Curr. Rheumatol. Rep.
3
:
535
541
.
9
Melas
,
I. N.
,
A. D.
Chairakaki
,
E. I.
Chatzopoulou
,
D. E.
Messinis
,
T.
Katopodi
,
V.
Pliaka
,
S.
Samara
,
A.
Mitsos
,
Z.
Dailiana
,
P.
Kollia
,
L. G.
Alexopoulos
.
2014
.
Modeling of signaling pathways in chondrocytes based on phosphoproteomic and cytokine release data.
Osteoarthritis Cartilage
22
:
509
518
.
10
Cheng
,
B.
,
H. W.
Liu
,
X. B.
Fu
,
Z. Y.
Sheng
,
J. F.
Li
.
2008
.
Coexistence and upregulation of three types of opioid receptors, mu, delta and kappa, in human hypertrophic scars.
Br. J. Dermatol.
158
:
713
720
.
11
Yang
,
L.
,
X.
Lian
,
W.
Zhang
,
J.
Guo
,
Q.
Wang
,
Y.
Li
,
Y.
Chen
,
X.
Yin
,
P.
Yang
,
F.
Lan
, et al
.
2015
.
Finding missing proteins from the epigenetically manipulated human cell with stringent quality criteria.
J. Proteome Res.
14
:
3645
3657
.
12
Guo
,
J.
,
Y.
Cui
,
Z.
Yan
,
Y.
Luo
,
W.
Zhang
,
S.
Deng
,
S.
Tang
,
G.
Zhang
,
Q. Y.
He
,
T.
Wang
.
2016
.
Phosphoproteome characterization of human colorectal cancer SW620 cell-derived exosomes and new phosphosite discovery for C-HPP.
J. Proteome Res.
15
:
4060
4072
.
13
Zhang
,
Y.
,
Z.
Wen
,
M. P.
Washburn
,
L.
Florens
.
2010
.
Refinements to label free proteome quantitation: how to deal with peptides shared by multiple proteins.
Anal. Chem.
82
:
2272
2281
.
14
Pavelka
,
N.
,
M. L.
Fournier
,
S. K.
Swanson
,
M.
Pelizzola
,
P.
Ricciardi-Castagnoli
,
L.
Florens
,
M. P.
Washburn
.
2008
.
Statistical similarities between transcriptomics and quantitative shotgun proteomics data.
Mol. Cell. Proteomics
7
:
631
644
.
15
Chen
,
Z.
,
L.
Yang
,
Y.
Cui
,
Y.
Zhou
,
X.
Yin
,
J.
Guo
,
G.
Zhang
,
T.
Wang
,
Q. Y.
He
.
2016
.
Cytoskeleton-centric protein transportation by exosomes transforms tumor-favorable macrophages.
Oncotarget
7
:
67387
67402
.
16
Luo
,
Y.
,
T. S.
Mok
,
X.
Lin
,
W.
Zhang
,
Y.
Cui
,
J.
Guo
,
X.
Chen
,
T.
Zhang
,
T.
Wang
.
2017
.
SWATH-based proteomics identified carbonic anhydrase 2 as a potential diagnosis biomarker for nasopharyngeal carcinoma.
Sci. Rep.
7
:
41191
.
17
Bindea
,
G.
,
B.
Mlecnik
,
H.
Hackl
,
P.
Charoentong
,
M.
Tosolini
,
A.
Kirilovsky
,
W. H.
Fridman
,
F.
Pagès
,
Z.
Trajanoski
,
J.
Galon
.
2009
.
ClueGO: a Cytoscape plug-in to decipher functionally grouped gene ontology and pathway annotation networks.
Bioinformatics
25
:
1091
1093
.
18
Bindea
,
G.
,
J.
Galon
,
B.
Mlecnik
.
2013
.
CluePedia Cytoscape plugin: pathway insights using integrated experimental and in silico data.
Bioinformatics
29
:
661
663
.
19
Schneider
,
C. A.
,
W. S.
Rasband
,
K. W.
Eliceiri
.
2012
.
NIH Image to ImageJ: 25 years of image analysis.
Nat. Methods
9
:
671
675
.
20
Shu
,
J.
,
G. E.
Dolman
,
J.
Duan
,
G.
Qiu
,
M.
Ilyas
.
2016
.
Statistical colour models: an automated digital image analysis method for quantification of histological biomarkers.
Biomed. Eng. Online
15
:
46
.
21
Chen
,
Y.
,
Y.
Li
,
J.
Zhong
,
J.
Zhang
,
Z.
Chen
,
L.
Yang
,
X.
Cao
,
Q. Y.
He
,
G.
Zhang
,
T.
Wang
.
2015
.
Identification of missing proteins defined by chromosome-centric proteome project in the cytoplasmic detergent-insoluble proteins.
J. Proteome Res.
14
:
3693
3709
.
22
Wang
,
T.
,
Y.
Cui
,
J.
Jin
,
J.
Guo
,
G.
Wang
,
X.
Yin
,
Q. Y.
He
,
G.
Zhang
.
2013
.
Translating mRNAs strongly correlate to proteins in a multivariate manner and their translation ratios are phenotype specific.
Nucleic Acids Res.
41
:
4743
4754
.
23
Milacic
,
M.
,
R.
Haw
,
K.
Rothfels
,
G.
Wu
,
D.
Croft
,
H.
Hermjakob
,
P.
D’Eustachio
,
L.
Stein
.
2012
.
Annotating cancer variants and anti-cancer therapeutics in Reactome.
Cancers (Basel)
4
:
1180
1211
.
24
Fabregat
,
A.
,
K.
Sidiropoulos
,
P.
Garapati
,
M.
Gillespie
,
K.
Hausmann
,
R.
Haw
,
B.
Jassal
,
S.
Jupe
,
F.
Korninger
,
S.
McKay
, et al
.
2016
.
The reactome pathway knowledgebase.
Nucleic Acids Res.
44
(
D1
):
D481
D487
.
25
Allikian
,
M. J.
,
E. M.
McNally
.
2007
.
Processing and assembly of the dystrophin glycoprotein complex.
Traffic
8
:
177
183
.
26
Kamburov
,
A.
,
U.
Stelzl
,
H.
Lehrach
,
R.
Herwig
.
2013
.
The ConsensusPathDB interaction database: 2013 update.
Nucleic Acids Res.
41
(
D1
):
D793
D800
.
27
Kanhere
,
A.
,
A.
Hertweck
,
U.
Bhatia
,
M. R.
Gökmen
,
E.
Perucha
,
I.
Jackson
,
G. M.
Lord
,
R. G.
Jenner
.
2012
.
T-bet and GATA3 orchestrate Th1 and Th2 differentiation through lineage-specific targeting of distal regulatory elements.
Nat. Commun.
3
:
1268
.
28
Unlu
,
S.
,
A.
Kumar
,
W. R.
Waterman
,
J.
Tsukada
,
K. Z.
Wang
,
D. L.
Galson
,
P. E.
Auron
.
2007
.
Phosphorylation of IRF8 in a pre-associated complex with Spi-1/PU.1 and non-phosphorylated Stat1 is critical for LPS induction of the IL1B gene.
Mol. Immunol.
44
:
3364
3379
.
29
Kusaba
,
H.
,
P.
Ghosh
,
R.
Derin
,
M.
Buchholz
,
C.
Sasaki
,
K.
Madara
,
D. L.
Longo
.
2005
.
Interleukin-12-induced interferon-gamma production by human peripheral blood T cells is regulated by mammalian target of rapamycin (mTOR).
J. Biol. Chem.
280
:
1037
1043
.
30
Gundelfinger
,
E. D.
,
M. M.
Kessels
,
B.
Qualmann
.
2003
.
Temporal and spatial coordination of exocytosis and endocytosis.
Nat. Rev. Mol. Cell Biol.
4
:
127
139
.
31
Rejman
,
J.
,
V.
Oberle
,
I. S.
Zuhorn
,
D.
Hoekstra
.
2004
.
Size-dependent internalization of particles via the pathways of clathrin- and caveolae-mediated endocytosis.
Biochem. J.
377
:
159
169
.
32
Gratton
,
S. E.
,
P. A.
Ropp
,
P. D.
Pohlhaus
,
J. C.
Luft
,
V. J.
Madden
,
M. E.
Napier
,
J. M.
DeSimone
.
2008
.
The effect of particle design on cellular internalization pathways.
Proc. Natl. Acad. Sci. USA
105
:
11613
11618
.
33
Katano
,
M.
,
M. S.
Kurokawa
,
K.
Matsuo
,
K.
Masuko
,
N.
Suematsu
,
K.
Okamoto
,
T.
Kamada
,
H.
Nakamura
,
T.
Kato
.
2017
.
Phosphoproteome analysis of synoviocytes from patients with rheumatoid arthritis.
Int. J. Rheum. Dis.
20
:
708
721
.
34
Momberger
,
T. S.
,
J. R.
Levick
,
R. M.
Mason
.
2005
.
Hyaluronan secretion by synoviocytes is mechanosensitive.
Matrix Biol.
24
:
510
519
.
35
Rhee
,
D. K.
,
J.
Marcelino
,
M.
Baker
,
Y.
Gong
,
P.
Smits
,
V.
Lefebvre
,
G. D.
Jay
,
M.
Stewart
,
H.
Wang
,
M. L.
Warman
,
J. D.
Carpten
.
2005
.
The secreted glycoprotein lubricin protects cartilage surfaces and inhibits synovial cell overgrowth.
J. Clin. Invest.
115
:
622
631
.
36
Kaneva
,
M. K.
,
K. V.
Greco
,
S. E.
Headland
,
T.
Montero-Melendez
,
P.
Mori
,
K.
Greenslade
,
C.
Pitzalis
,
A.
Moore
,
M.
Perretti
.
2017
.
Identification of novel chondroprotective mediators in resolving inflammatory exudates.
J. Immunol.
198
:
2876
2885
.
37
Lee
,
Y. R.
,
S. H.
Kweon
,
K. B.
Kwon
,
J. W.
Park
,
T. R.
Yoon
,
B. H.
Park
.
2009
.
Inhibition of IL-1beta-mediated inflammatory responses by the IkappaB alpha super-repressor in human fibroblast-like synoviocytes.
Biochem. Biophys. Res. Commun.
378
:
90
94
.
38
Chang
,
Y. W.
,
Y. F.
Zhao
,
Y. L.
Cao
,
W.
Gu
,
J.
Pang
,
H. S.
Zhan
.
2014
.
Bufalin exerts inhibitory effects on IL-1β-mediated proliferation and induces apoptosis in human rheumatoid arthritis fibroblast-like synoviocytes.
Inflammation
37
:
1552
1559
.
39
Chen
,
S.
,
Y.
Yang
,
H.
Feng
,
H.
Wang
,
R.
Zhao
,
H.
Liu
.
2014
.
Baicalein inhibits interleukin-1β-induced proliferation of human rheumatoid arthritis fibroblast-like synoviocytes.
Inflammation
37
:
163
169
.
40
Bartok
,
B.
,
G. S.
Firestein
.
2010
.
Fibroblast-like synoviocytes: key effector cells in rheumatoid arthritis.
Immunol. Rev.
233
:
233
255
.
41
Gobezie
,
R.
,
A.
Kho
,
B.
Krastins
,
D. A.
Sarracino
,
T. S.
Thornhill
,
M.
Chase
,
P. J.
Millett
,
D. M.
Lee
.
2007
.
High abundance synovial fluid proteome: distinct profiles in health and osteoarthritis.
Arthritis Res. Ther.
9
:
R36
.
42
Mateos
,
J.
,
L.
Lourido
,
P.
Fernández-Puente
,
V.
Calamia
,
C.
Fernández-López
,
N.
Oreiro
,
C.
Ruiz-Romero
,
F. J.
Blanco
.
2012
.
Differential protein profiling of synovial fluid from rheumatoid arthritis and osteoarthritis patients using LC-MALDI TOF/TOF.
J. Proteomics
75
:
2869
2878
.
43
Balakrishnan
,
L.
,
R. S.
Nirujogi
,
S.
Ahmad
,
M.
Bhattacharjee
,
S. S.
Manda
,
S.
Renuse
,
D. S.
Kelkar
,
Y.
Subbannayya
,
R.
Raju
,
R.
Goel
, et al
.
2014
.
Proteomic analysis of human osteoarthritis synovial fluid.
Clin. Proteomics
11
:
6
.
44
Kloppenburg
,
M.
,
R.
Ramonda
,
W. Y.
Kwok
,
K.
Bobacz
,
D.
Elewaut
,
P.
Frallonardo
,
T. W.
Huizinga
,
F.
Kroon
,
J.
Smolen
,
B.
Vander Cruyssen
,
R.
Wolterbeek
,
L.
Punzi
,
G.
Verbruggen
,
R.
Wittoek
.
2016
.
OP0095 randomized, placebo controlled trial to evaluate clinical efficacy and structure modifying properties of subcutaneous etanercept (ETN) in patients with erosive inflammatory hand osteoarthritis (OA).
Ann. Rheum. Dis.
75
(
Suppl. 2
):
90.93
91
.
45
Cohen
,
S. B.
,
S.
Proudman
,
A. J.
Kivitz
,
F. X.
Burch
,
J. P.
Donohue
,
D.
Burstein
,
Y. N.
Sun
,
C.
Banfield
,
M. S.
Vincent
,
L.
Ni
,
D. J.
Zack
.
2011
.
A randomized, double-blind study of AMG 108 (a fully human monoclonal antibody to IL-1R1) in patients with osteoarthritis of the knee.
Arthritis Res. Ther.
13
:
R125
.
46
Huang
,
Q.
,
L.
Yang
,
J.
Luo
,
L.
Guo
,
Z.
Wang
,
X.
Yang
,
W.
Jin
,
Y.
Fang
,
J.
Ye
,
B.
Shan
,
Y.
Zhang
.
2015
.
SWATH enables precise label-free quantification on proteome scale.
Proteomics
15
:
1215
1223
.
47
Schweizer
,
F. E.
,
T. A.
Ryan
.
2006
.
The synaptic vesicle: cycle of exocytosis and endocytosis.
Curr. Opin. Neurobiol.
16
:
298
304
.

The authors have no financial conflicts of interest.

Supplementary data