Proteomic Analysis of Synovial Fluid From the Osteoarthritic Knee: Comparison With Transcriptome Analyses of Joint Tissues




The pathophysiology of the most common joint disease, osteoarthritis (OA), remains poorly understood. Since synovial fluid (SF) bathes joint cartilage and synovium, we reasoned that a comparative analysis of its protein constituents in health and OA could identify pathways involved in joint damage. We undertook this study to perform a proteomic analysis of knee SF from OA patients and control subjects and to compare the results to microarray expression data from cartilage and synovium.


Age-matched knee SF samples from 10 control subjects, 10 patients with early-stage OA, and 10 patients with late-stage OA were compared using 2-dimensional difference-in-gel electrophoresis and mass spectrometry (MS). MS with a multiplexed peptide selected reaction monitoring assay was used to confirm differential expression of a subset of proteins in an independent OA patient cohort. Proteomic results were analyzed by Ingenuity Pathways Analysis and compared to published synovial tissue and cartilage messenger RNA profiles.


Sixty-six proteins were differentially present in healthy and OA SF. Three major pathways were identified among these proteins: the acute-phase response signaling pathway, the complement pathway, and the coagulation pathway. Differential expression of 5 proteins was confirmed by selected reaction monitoring assay. A focused analysis of transcripts corresponding to the differentially present proteins indicated that both synovial and cartilage tissues may contribute to the OA SF proteome.


Proteins involved in the acute-phase response signaling pathway, the complement pathway, and the coagulation pathway are differentially regulated in SF from OA patients, suggesting that they contribute to joint damage. Validation of these pathways and their utility as biomarkers or therapeutic targets in OA is warranted.

The pathophysiology of osteoarthritis (OA), the most common joint disease and a significant cause of disability in the elderly, remains poorly understood. Development of this disease is multifactorial, with local mechanical factors, such as malalignment and internal derangements, and systemic factors, including genetics and obesity, each contributing to loss of joint integrity (1). Although OA is characterized by progressive loss of cartilage within the joint, substantial changes occur in the synovium and subchondral bone. However, the relative contribution of these tissues to disease pathogenesis is unresolved. Our poor understanding of the pathways that drive joint damage in OA is a major impediment to the development of disease-modifying agents.

Advances in the sensitivity and throughput of mass spectrometry (MS)–based proteomic techniques have made feasible their application as a discovery method in complex biologic fluids. Proteomic approaches have the advantage over nucleic acid expression profiling in that their interpretation is not limited by a possible disconnect between gene and protein expression levels. As such, this technology has emerged as a powerful method to identify proteins involved in disease etiology and pathogenesis, as well as potential biomarkers (2, 3).

To date, proteomic studies in OA using cartilage, chondrocytes, synovial fibroblasts, and bone marrow mesenchymal stem cells have provided novel insights into joint pathophysiology (4–7). While these studies have been informative, an inherent limitation of these tissue- and lineage-based analyses is their inability to fully represent the physiology of the intact joint. Since synovial fluid (SF) bathes all the intrinsic structures of diarthrodial joints, an analysis of its constituents offers a unique opportunity to study the entire diseased OA joint.

In a pilot study using a relatively insensitive liquid chromatography tandem MS (LC-MS/MS) method, we analyzed SF proteins in healthy individuals and OA patients and identified 18 differentially expressed proteins (8). In the current study, we used a more sensitive method based on gel electrophoresis and MS to quantitatively probe deeper into the SF proteome of OA. A subset of these dysregulated proteins was confirmed in OA SF using an MS assay that detects representative peptides from these proteins. Finally, a comparison of our proteomics results with messenger RNA (mRNA) expression profiling of joint tissues suggests that some of these proteins are derived from synovium or cartilage.


Human subjects and samples.

For the 2-dimensional (2-D) difference-in-gel electrophoresis (DIGE) experiments, patients ranged in age from 45 to 65 years. SF was obtained from 10 patients with late OA (4 males and 6 females) at the time of total knee arthroplasty. Ten patients with early OA (3 males and 7 females) were undergoing surgery for knee pain due to meniscal tear documented with magnetic resonance imaging; these patients had mild cartilage degeneration visualized at surgery. SF was obtained at the time of arthroscopy. Ten control subjects (6 males and 4 females) were asymptomatic individuals without radiographic OA who were paid to undergo arthrocentesis; these subjects were matched by age to the 10 patients with early OA. Exclusion criteria included inflammatory arthritis, steroid injection within 6 weeks, blood dyscrasias, and active malignancy. All patients and volunteers gave informed consent, and the University Hospital Medical Ethics Committee at Case Western Reserve University approved the study.

For the selected reaction monitoring assays, SF was obtained from 13 healthy individuals with no symptoms of knee OA; from 16 patients with <1 year of symptoms and early OA, as assessed by the presence of cartilage lesions visible on arthroscopy or by radiographic imaging; and from 119 patients with late OA who were undergoing total knee replacement. Samples for this cohort were obtained under protocols approved by the Partners Healthcare Institutional Review Board (IRB).

Synovial tissue biopsy samples for mRNA analysis were obtained from patients with early and late knee OA recruited at the Hospital for Special Surgery as described previously (9). Briefly, we recruited 9 patients with advanced (late) knee OA according to the American College of Rheumatology (ACR) classification criteria (10) who were undergoing total knee replacement surgery. Patients with late knee OA were required to have a preoperative Kellgren/Lawrence (K/L) grade ≥2 (11) and intraoperative evidence of full-thickness cartilage loss. We also recruited 10 patients with early knee OA who were undergoing arthroscopic procedures for degenerative meniscal tears, with documented cartilage fibrillation or fissuring by intraoperative inspection but no full-thickness cartilage loss and K/L grade ≤2. Biopsy samples for mRNA isolation were obtained from the suprapatellar pouch. Details of the transcriptional profiling were previously described (12). The IRB at the Hospital for Special Surgery approved the study, and informed consent was obtained from all patients.

The cartilage samples used for gene expression profiling were obtained with IRB approval and were previously described (13). Using a freezer/mill, RNA extracts were obtained directly from cartilage samples from 13 normal individuals and 12 patients with late-stage OA.

Two-dimensional DIGE.

SF preparation.

SF samples were flash-frozen in liquid nitrogen after the addition of protease inhibitors (Roche Diagnostics) and stored at −80°C. Prior to 2-D DIGE, SF was thawed on ice, centrifuged at 400g at 4°C to remove blood cells, treated with bovine hyaluronidase (Sigma), and depleted of abundant proteins with an Agilent High Capacity Multiple Affinity Removal Spin Cartridge Human 6. This cartridge depletes albumin, IgG, IgA, transferrin, haptoglobin, and antitrypsin and eliminates 85–90% of protein.

Gel electrophoresis.

Two experiments were conducted following the design described in Supplementary Table 1, available on the Arthritis & Rheumatism web site at The quenched Cy3- and Cy5-labeled samples were combined and mixed with an aliquot of Cy2-labeled standard specific to each set. The remaining protocol for gel electrophoresis and gel visualization was as previously described (14).

Image analysis.

A representative gel image used for analysis is shown in Supplementary Figure 1, available on the Arthritis & Rheumatism web site at To compare protein spots across the 6 gels and 9 gels (set 1 and set 2 in Supplementary Table 1, available on the Arthritis & Rheumatism web site at, DeCyder software, version 6.5 2 (GE Healthcare) was used for image analysis. Images consisting of 4 biologic replicates (4 early OA and 4 late OA SF samples) from set 1 and 6 biologic replicates (6 early OA and 6 late OA SF samples) from set 2 were loaded into the differential-in-gel analysis algorithm within the DeCyder software, and intragel spot detection and quantification were performed. The entire set of protein spots was matched with the Biological Variation Analysis module. The standardized volume ratio for each standard image from the different gels was set to the value 1.0 in order to compare ratios between matched protein spots in the different gels (sets). Student's t-test was used for statistical analyses. P values less than 0.05 were considered significant.

In addition, the results related to control and disease samples were compared and statistically evaluated by one-way analysis of variance (ANOVA) with the DeCyder Biological Variation Analysis module, applying the false discovery rate mode to minimize false-positive results. Protein spots with statistically significant variation (P < 0.05), showing a difference in volume of >1.0-fold above and <1.0-fold below, were selected as differentially present. For spots meeting these requirements, a pick list was generated and transferred to the automated Ettan spot picker, and gel plugs were excised and recovered into 96-well plates for in-gel digestion and MS analysis.

In-gel digestion and protein identification.

Recovered gel plugs were digested with trypsin (Promega) and analyzed by MS as described (14, 15). The tandem mass spectra were annotated, and peak list .dta format files were generated using the Mascot search engine. The peak lists were compared to the Swiss-Prot database, 2010 (human) (20,278 sequences) using the Mascot search engine.

Criteria for protein identification.

Scaffold (version Scaffold_3_00_08; Proteome Software) was used to validate MS/MS-based peptide and protein identifications. X! Tandem software and the Mascot search engine were set up to search a subset of the uniprot_sprot_5_2010. Mascot and X! Tandem were searched with a fragment ion mass tolerance of 0.9 Da and a parent ion tolerance of 10.0 parts per million. Iodoacetamide derivative of cysteine was specified in Mascot and X! Tandem as a fixed modification and oxidation of methionine was specified as a variable modification. Peptide identifications were accepted if they could be established at >95.0% probability as specified with the Peptide Prophet algorithm (16). Protein identifications were accepted if they could be established at >99.0% probability and contained at least 1 unique identified peptide. Protein probabilities were assigned with the Protein Prophet algorithm (17). Proteins that contained similar peptides and could not be differentiated based on MS/MS analysis alone were grouped to satisfy the principles of parsimony. Spots containing keratin species (as determined by molecular weight and peptide identification) were removed from further consideration. Eighty-four spots from set 1 and 44 from set 2 were removed based on this analysis. Expression values of each protein spot were represented as the fold change. The data were transferred into R-Bioconductor, and a heatmap was generated using a “gplots” package (18).

Network analysis.

The list of regulated proteins identified by 2-D DIGE and MS was analyzed by pathway analysis using the network-building tool, Ingenuity Pathways Analysis (IPA) (Ingenuity Systems, The analysis identified the biologic functions and/or diseases most significant within the data set.

Selected reaction monitoring assays.

Selected reaction monitoring assays were developed on a TSQ Vantage triple quadrupole mass spectrometer (Thermo Fisher Scientific), as previously described (19–21). Each sample was digested and analyzed 3 times in the selected reaction monitoring assay to total 3 technical replicates. One-way ANOVA was performed, with P values less than 0.05 considered significant. A Bonferroni post test was used to evaluate differences between groups.

Microarray data collection.

Published microarray expression data for OA patients and normal controls used in this study included OA synovial tissue from the Hospital for Special Surgery (GEO accession no. GSE32317) (12), normal synovial tissue from the public GEO (GEO accession no. GSE12021) (22), and a cartilage data set (13). See cited references for details regarding RNA extraction.

Messenger RNA expression level analysis and clustering.

Primary expression level data files (.cel files) from these studies were merged and normalized using the robust multiarray average method in MatLab (MathWorks). Thirty-one patient samples and 22 control samples from the 2 tissue types were included in this study. Genes of interest identified from the proteomic analysis were mapped to their probes, and heatmaps of an unsupervised hierarchical clustering analysis were generated using MatLab (23).


Two-dimensional DIGE analysis of SF proteins.

SF from 10 control subjects, 10 patients with early OA, and 10 patients with late OA were subjected to 2-D DIGE and MS analysis. In this technique, samples are labeled with spectrally different fluorescent dyes (Cy3 or Cy5) prior to electrophoresis in combination with an internal standard, which allows for intragel and intergel matching between samples (24, 25). Samples were batched into 2 experimental sets (see Supplementary Table 1, available on the Arthritis & Rheumatism web site at On average, >2,200 individual spots per SF gel image were detected and matched across subjects. Data filtering was performed using a volume ratio change of >1.0-fold (either increased or decreased) and a P value less than 0.05. From these initial spots, 199 spots in set 1 and 138 spots in set 2 satisfied these criteria and were picked for protein identification. If the same protein was identified in different spots across the 2-D gel, suggesting posttranslational modifications, these replicates were omitted and the protein was reported only once.

Using the fold change from the fluorescent images as a representation of relative protein abundance, protein spots were compared for changes between healthy subjects and patients with late OA. A total of 43 and 42 proteins were differentially present across sets 1 and 2, respectively (Tables 1 and 2) (see full data set in Supplementary Tables 2 and 3, available on the Arthritis & Rheumatism web site at To develop a larger protein list for further analysis, we combined these 2 data sets, noting that 19 proteins were observed in both sets (annotated in Table 2 and Supplementary Table 4, available on the Arthritis & Rheumatism web site at and the fold change was similar, except for annexin A2 and apolipoprotein A-I, which were up-regulated in set 1 and down-regulated in set 2. Combining the data from these 2 experiments produced a final list of 66 spots that were differentially present between healthy subjects and patients with late OA (see Supplementary Table 4, available on the Arthritis & Rheumatism web site at A subset of these proteins in the complement pathway was previously reported by Wang et al (12), although this previous publication did not report quantitative differences between SF from healthy subjects and SF from OA patients.

Table 1. Proteins differentially expressed between patients with late-stage osteoarthritis and healthy controls in set 1*
Accession no.SpotProtein descriptionGene nameFold change
P073551,055Annexin A2ANXA22.34
P010081,019Antithrombin IIISERPINC11.18
P026472,285Apolipoprotein A-IAPOA12.23
P027491,126β2-glycoprotein IAPOH1.82
P04003803C4b binding protein α-chainC4BPA1.46
P22792597Carboxypeptidase N subunit 2CPN22.14
Q9NQ79727Cartilage acidic protein 1CRTAC13.12
P027462,158C1q subcomponent subunit BC1QB1.34
P00736616C1r subcomponentC1R2.71
P09871654C1s subcomponentC1S2.31
P00751588Complement factor BCFB2.07
P08603276Complement factor HCFH1.62
P051561,214Complement factor ICFI2.05
P026751,158Fibrinogen β-chainFGB3.48
P02679721Fibrinogen γ-chainFGG4.83
P05546760Heparin cofactor IISERPIND13.55
P19827752Inter-α trypsin inhibitor heavy chain H1ITIH14.01
Q14624457Inter-α trypsin inhibitor heavy chain H4ITIH42.03
P01042966Kininogen 1KNG11.76
Q96PD5802N-acetylmuramoyl-L-alanine amidasePGLYRP21.26
P05155663Plasma protease C1 inhibitorSERPING12.04
Q151131,326Procollagen C-endopeptidase enhancer 1PCOLCE1.67
P525662,609Rho GDP dissociation inhibitor 2ARHGDIB−1.62
P02768937Serum albuminALB1.52
P271691,242Serum paraoxonase/arylesterase 1PON12.76
P027741,187Vitamin D binding proteinGC1.81
Table 2. Proteins differentially expressed between patients with late-stage osteoarthritis and healthy controls in set 2*
Accession no.SpotProtein descriptionGene nameFold change
P607091,563Actin, cytoplasmic 1ACTB−1.45
P073551,778Annexin A2ANXA2−1.39
P095251,868Annexin A4ANXA4−1.36
P010081,105Antithrombin IIISERPINC11.5
P026471,915Apolipoprotein A-IAPOA1−1.39
P067271,472Apolipoprotein A-IVAPOA41.58
P050901,938Apolipoprotein DAPOD−1.43
Q9NQ79765Cartilage acidic protein 1CRTAC12.04
P027472,090C1q subcomponent subunit CC1QC−1.49
P09871712C1s subcomponentC1S2.34
P08603305Complement factor HCFH1.62
Q035911,611Complement factor H–related protein 1CFHR1−1.46
P02679849Fibrinogen γ-chainFGG7.05
P223522,146Glutathione peroxidase 3GPX3−1.27
P01871864Ig μ-chain C regionIGHM3.45
P35858776Insulin-like growth factor binding protein complex acid-labile subunitIGFALS2.38
P19827856Inter-α trypsin inhibitor heavy chain H1ITIH14.98
Q14624394Inter-α trypsin inhibitor heavy chain H4ITIH41.85
P01042983Kininogen 1KNG12.12
P246662,208Low molecular weight phosphotyrosine protein phosphataseACP1−1.73
P300861,503Phosphatidylethanolamine-binding protein 1PEBP11.58
P80108385Phosphatidylinositol-glycan–specific phospholipase DGPLD11.87
Q8IV081,338Phospholipase D3PLD3−1.4
P369551,346Pigment epithelium–derived factorSERPINF11.23
P03952805Plasma kallikreinKLKB11.26
P027601,867Protein AMBPAMBP1.92
Q132281,297Selenium-binding protein 1SELENBP11.44

To determine whether the proteomic signatures of early OA and late OA were overlapping or unique, expression of the differentially regulated proteins from our primary analysis (Tables 1 and 2) was compared between these samples relative to samples from healthy controls. This has been graphically represented with heatmaps for proteins in patients with early OA and proteins in patients with late OA versus proteins in healthy controls, plotted side by side for each protein identified in sets 1 and 2 (Figures 1A and B, respectively). In set 1, there was 100% concordance between expression of proteins differentially regulated in patients with late OA and those with early OA compared to healthy controls. In set 2, the concordance was also strong, with only 3 proteins (C9, protein α1-microglobulin/bikunin precursor [AMBP], and apolipoprotein A-IV) showing discordant expression. These data suggest that the proteomic signature of OA is firmly established in patients with early OA.

Figure 1.

Heatmap representation of differentially expressed proteins across the patient subsets (late-stage osteoarthritis [LOA] and early-stage OA [EOA]) from the 2 experimental sets (see Supplementary Table 1, available on the Arthritis & Rheumatism web site at Proteins are listed by their respective gene names. Represented are fold changes in protein expression in subsets of patients with early-stage OA and late-stage OA compared to healthy control subjects in set 1 (A) and set 2 (B).

Pathway analysis of SF proteins.

An unbiased bioinformatics approach using the IPA tool was used to identify dysregulated functional pathways in the OA SF proteome. Proteins with fold changes of at least ±1.2 were taken as significant for pathway analysis. We focused the pathway analysis on proteins in patients with late OA versus those in healthy controls, as the highest number of differentially present proteins was contained within this group. The top 5 molecular and cellular functions were antigen presentation, cell-to-cell signaling and interaction, lipid metabolism, molecular transport, and small molecule biochemistry (Table 3), similar to functional classifications identified in smaller OA proteomic studies (5, 26). The canonical pathways associated with our proteins were acute-phase response signaling, the complement system, and the coagulation system. Both the intrinsic and extrinsic prothrombin activation pathways, which highly overlap with the coagulation system, were also represented. The individual canonical pathways and their statistical analyses are depicted in Table 3 and Supplementary Figure 2 (available on the Arthritis & Rheumatism web site at

Table 3. Details of Ingenuity Pathways Analysis (IPA) of top molecular and cellular functions and top canonical pathways*
IPAPNo. of proteins or ratio
Molecular and cellular functions  
 Antigen presentation2.21 × 10−9 to 1.68 × 10−216
 Cell-to-cell signaling and interaction1.08 × 10−8 to 2.51 × 10−217
 Lipid metabolism3.95 × 10−7 to 2.33 × 10−216
 Molecular transport3.95 × 10−7 to 2.09 × 10−219
 Small molecule biochemistry3.95 × 10−7 to 2.33 × 10−223
Top canonical pathways  
 Acute-phase response signaling4.38 × 10−3325/172
 Complement system1.17 × 10−2313/33
 Coagulation system1.52 × 10−149/35
 Intrinsic prothrombin activation pathway1.22 × 10−96/29
 Extrinsic prothrombin activation pathway5.05 × 10−74/16

Validation of protein targets by quantitative selected reaction monitoring assay.

To validate differences in protein expression between healthy and OA SF samples identified by 2-D DIGE, we developed specific MS assays using selected reaction monitoring. This technique measures peptides representative of proteins of interest generated by proteolysis of biologic samples. Levels of target peptides are quantified using otherwise chemically identical peptide standards synthesized with heavy isotopes (27). Reactions were developed for 10 peptides representative of 5 proteins with increased expression in the SF of OA patients by 2-D DIGE: afamin, clusterin, insulin-like growth factor binding protein complex acid-labile subunit (IGFALS), lumican, and pigment epithelium–derived factor. Concordant with the 2-D DIGE results, in an independent cohort of SF samples from 13 healthy subjects, 16 patients with early OA, and 119 patients with late OA, all 10 peptides showed a significant increase in early OA and late OA (Figure 2 and Supplementary Figure 3, available on the Arthritis & Rheumatism web site at

Figure 2.

Validation of differentially expressed proteins in OA synovial fluid (SF) by selected reaction monitoring. Shown are relative concentrations of the indicated peptides in SF from healthy controls, patients with early-stage OA, and patients with late-stage OA representative of afamin (A), clusterin (B), insulin-like growth factor binding protein complex acid-labile subunit (C), lumican (D), and pigment epithelium–derived factor (E). Symbols represent SF samples from individual subjects; horizontal lines show the mean. P < 0.001 for all peptides in SF from patients with early OA and SF from patients with late OA versus SF from healthy controls. See Figure 1 for other definitions.

Cartilage and synovial tissue gene expression profiling.

SF components may originate from multiple joint tissues, including cartilage and synovium. To gain insight into the relative contribution of the 2 tissues to the OA proteome, expression data sets from normal and OA synovial tissue and cartilage were probed for mRNAs corresponding to the SF proteins. Probes for the mRNAs encoding 64 of the 66 SF proteins were assessed in each microarray data set, and unsupervised hierarchical clustering analysis was performed to identify global expression patterns in health versus disease. Specifically, the synovial tissue contribution to the OA SF proteome was assessed using Affymetrix mRNA expression data sets for healthy synovium that were publicly available merged with expression data sets of early OA and late OA synovium as described in Patients and Methods (9, 12).

OA and control synovial tissue clustered separately, with early OA and late OA synovial tissue also clustering (Figure 3A). Comparing differential expression from the transcriptome analysis in synovium to the proteomic expression in SF, a few interesting trends were noticed. Gelsolin (GSN) and many complement proteins showed concordant expression in both synovium and SF, suggesting that the synovium may be the primary source of these proteins in the SF. Of the 14 complement pathway hits from the IPA of the SF proteome, 11 showed concordant expression in synovial tissue (C1R, C1QB, C1S, C3, C4A, C7, C9, CFB, CFH, CFHR1, and CFI). For the other Ingenuity pathways (coagulation and acute phase), less than one-third of the proteins were concordant in synovial tissue. Other proteins such as lumican (LUM), plasma protease C1 inhibitor (SERPING1), and apolipoprotein A-I (APOA1) were discordant, suggesting that they may not be derived from a synovial source.

Figure 3.

Examination of synovial or cartilage tissue expression of mRNA transcripts encoding proteins differentially present in synovial fluid (SF) from patients with late-stage osteoarthritis (OA) compared with SF from healthy control subjects. Shown is an unsupervised cluster analysis of mRNA from either synovium (A) or cartilage (B) encoding for proteins differentially expressed in SF from patients with late-stage OA. Disease status of individual samples is listed on the x-axis of each panel. In A, early-stage OA samples are denoted as “Early,” late-stage OA samples are denoted as “End,” and healthy control samples are denoted as “Normal.” In B, late-stage OA samples are denoted as “OA,” and healthy control samples are denoted as “normal.”

In parallel studies, the contribution of cartilage tissue to the proteins expressed in SF was investigated using a microarray data set from cartilage specimens from healthy subjects and OA patients (13). Similar to the synovial tissue analysis, unsupervised clustering analyses differentiated healthy versus OA cartilage tissue (Figure 3B). Comparing differential expression of transcripts in cartilage to proteomic expression in SF, we noted concordant up-regulation of the extracellular matrix proteins cartilage acidic protein 1 (CRTAC1) and lumican (LUM). We also noted that transaldolase (TALDO1) and glutathione peroxidase 3 (GPX3) were down-regulated in OA cartilage and OA SF. Comparisons of the Ingenuity pathways to the cartilage microarray data were less clear than in synovium. Only 5 of the 14 complement pathway targets showed concordant expression in cartilage tissue. For the coagulation system, only 2 of the 9 factors were concordant in SF and cartilage tissue.

Furthermore, we formally compared the differentially expressed transcripts in the microarrays to the 66 proteins identified in the SF proteome. Using a cutoff of fold change > ±2 and P < 0.05, cartilage had 17 of the 66 transcripts differentially expressed and synovium had 15 of the 66. We focused on the transcripts that had differential expression in cartilage and synovium concordant with our proteome, as these may be the most biologically relevant. In cartilage, 10 transcripts were concordant: ATRN, C1S, CFI, CRTAC1, FN1, GNS, GPX3, LUM, PCOLCE, and SERPINF1. In synovium, 11 transcripts were concordant: ATRN, C1R, C7, CFB, CFI, CP, CRTAC1, FN1, GSN, ITIH4, and PCOLCE. Several of these are members of the complement and acute-phase response pathways. While it is challenging to directly compare transcript and protein levels, due to regulation of mRNA translation and posttranslational protein modifications, our analysis suggests that synovium and cartilage contribute to the differential regulation of proteins in OA SF.


In this study, we performed 2-D DIGE and MS analysis to identify proteins differentially present in SF from OA patients as compared to age-matched controls. We hypothesized that an examination of protein expression in the fluid that bathes joint structures would provide insight into potential pathogenic mechanisms of this prevalent condition. Among these differentially present proteins, 3 dominant pathways were identified: the acute-phase response signaling pathway, the complement pathway, and the coagulation pathway.

The acute-phase response signature may be a reaction to tissue injury in the damaged joint. While it is appreciated that OA is associated with prior joint trauma, some have assumed that the ensuing joint damage results from aberrant mechanical loading (28). However, other studies have shown synovial inflammation to be present after trauma, as well as later in the course of OA (29, 30). In particular, acute-phase reactants such as C-reactive protein (CRP) and interleukin-6 are elevated in OA patients undergoing surgery (31), and proinflammatory cytokines and chemokines have been identified in synovial tissues in both early and late OA (32, 33). When tissues are damaged, the body attempts to heal the architectural disruption by initiating a remodeling and reconstruction program. Since this acute-phase response pattern was present to a similar degree in early OA and late OA, our data suggest that the chronic progressive articular damage of OA occurs through an ongoing inflammatory response associated with unresolved tissue injury. This observation is congruent with previous work identifying wound physiology and associated inflammation in OA (for review, see ref.34) and supports therapeutic approaches that interrupt tissue injury pathways.

Our proteomic analyses also identified up-regulation of complement proteins in OA SF, extending our initial description of the OA SF proteome, wherein several members of the complement pathway emerged (8). Supporting a functional role for complement in OA pathogenesis, Wang et al recently showed that mice lacking C5 and components of the membrane attack complex were protected from experimental OA (12). Thus, the presence of elevated levels of complement pathway proteins in OA may reflect an important pathophysiologic aspect of the disease. Furthermore, the concordant expression of complement components in SF and their corresponding mRNAs in synovium suggests that these factors are made, at least in part, by the joint lining.

The final pathway our analysis identified as dysregulated in OA SF was the coagulation cascade. While fibrin deposition in joint tissues is common in inflammatory arthritis (35–37), its involvement in OA is less appreciated. One proteomic study comparing SF to plasma showed fibrinogen degradation products in OA SF, but not plasma (38). Likewise, another analysis demonstrated that the balance between fibrin activation and fibrinolysis is perturbed in OA, with increased fibrin deposition and degradation products in SF (35, 39). A recent study showed that plasma proteins, including fibrinogen, were present in SF, and the authors hypothesized that this was either from plasma exudation or production by synovial tissues (40). Our data provide evidence that some components of the coagulation pathway may be produced locally. As there is overlap between the proteins in the coagulation and the acute-phase response pathways, this observation may reflect OA wound physiology discussed above. It remains to be determined how the coagulation pathway contributes to joint damage, inflammation, and pain in OA.

Our findings should be put in perspective with other proteomic analyses of OA SF. Yamagiwa et al used 2-D polyacrylamide gel electrophoresis to compare OA SF from 4 subjects and found 18 proteins with >5-fold change in spot intensity, 2 of which were likely haptoglobin α2-chains (41). Kamphorst et al identified 40 proteins by nanoscale LC-MS in 1 OA patient compared to 1 healthy subject and found that most of the proteins either were structural proteins or were related to the coagulation cascade and immune response (42). While these general categories are similar to those found in our study, there is <25% identity with our protein list. In our previous study, we identified 18 proteins differentially expressed in OA SF samples compared to control SF samples, some of which were replicated in the current study, such as fibrinogen, β2-glycoprotein I, C3, vitamin D binding protein, and protein AMBP (8). The use of an affinity depletion column to remove high-abundance serum proteins was one difference between the current study and our prior one and may have increased the sensitivity.

Finally, Mateos et al recently used LC–matrix-assisted laser desorption ionization–time-of-flight/time-of-flight MS to analyze the proteome of pooled OA and RA SF and found 136 different proteins, with 17 more abundant in OA SF (43). Eight of these 17 proteins were differentially expressed in our experiment (tetranectin, inter-α trypsin inhibitor heavy chain H1, gelsolin, plasma protease C1 inhibitor, cartilage acidic protein 1, fibronectin, pigment epithelium–derived factor, and α1B-glycoprotein), validating our approach and providing more evidence for these proteins in the pathogenesis of OA.

Differential expression of a subset of the OA SF proteome was validated in an independent cohort using a complementary MS technique, selected reaction monitoring, to provide precise quantification and additional specificity. While protein quantification in biologic fluids has been dominated by immunoassays, selected reaction monitoring assays, which quantify surrogate target peptides, are increasingly being applied in research and clinical settings (44). Selected reaction monitoring–based assays are typically robust and selective, even in complex matrices, and can be multiplexed to measure dozens of peptides simultaneously. Such assays could be used to quantify peptide biomarkers for disease diagnosis and prognosis (27, 45, 46).

One of our goals was to determine the likely tissue source(s) of the differentially expressed proteins. As discussed above, our analyses suggest that the synovium and cartilage are important sources of some of the differentially regulated proteins in OA. Some of the transcripts were not clearly differentially regulated in the tissue microarrays that we tested, so they may have been derived from bone, plasma, or cells infiltrating the joint (40). This may be the case for the coagulation pathway proteins, as there was low concordance between the microarray and proteomic data for this pathway. However, inherent technical limitations make it difficult to draw any conclusions from poor correlations between mRNA and protein levels.

First, the microarray data may not have been sufficiently sensitive to detect differential expression of some of the mRNAs. Second, the relationship between mRNA and protein levels is affected by complex regulatory mechanisms that cannot be captured by the techniques used in this study. Indeed, others have shown a weak correlation between mRNA and protein abundance (47). Another factor that limits our power to draw conclusions from our data is that some of the transcriptional and protein changes in the joint may reflect processes occurring more globally in the patient. For example, Pan et al found elevations in lumican in pancreatic cancer and chronic pancreatitis, suggesting that other diseases may affect lumican levels (48). Another study showed that pigment epithelium–derived factor correlates with high-sensitivity CRP levels in normal subjects (49). These studies suggest that tissues besides synovium and cartilage, as well as inflammation, could explain the protein variation in our study.

Limitations of the sensitivity of our study include the stringent definition of differential protein abundance and the elimination of any spot in the 2-D DIGE analysis with keratin contamination. Although we likely missed proteins using this approach, it probably enhanced the specificity. In addition, our sample of 30 subjects, although larger than that in any previous similar study, was still rather small but had the advantage of containing well-characterized patients and age-matched control subjects. Finally, we cannot exclude the possibility that some of the controls had occult OA, but none met the ACR classification criteria for knee OA (10).

In summary, to our knowledge, the current study has identified the largest number of differentially expressed proteins in OA SF compared to healthy SF. Analysis of these expression profiles uncovers signals from functional pathways not widely appreciated in OA pathophysiology. Validation of the differential expression of proteins identified in our study in larger cohorts is mandated, as are functional studies to dissect the roles of these pathways in OA pathogenesis. Finally, assessment of potential biomarkers and therapeutic targets for drug development within the OA proteome identified in the current study is warranted.


All authors were involved in drafting the article or revising it critically for important intellectual content, and all authors approved the final version to be published. Dr. Aliprantis had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Study conception and design. Ritter, Bebek, Crish, Krastins, Sarracino, Lopez, Crow, Aigner, M. Goldring, S. Goldring, Lee, Gobezie, Aliprantis.

Acquisition of data. Ritter, Crish, Scanzello, Krastins, Sarracino, Lopez, Crow, Aigner, M. Goldring, S. Goldring, Gobezie.

Analysis and interpretation of data. Ritter, Subbaiah, Bebek, Crish, Krastins, Sarracino, Lopez, Crow, Aigner, M. Goldring, S. Goldring, Lee, Gobezie, Aliprantis.


Author Lee is an employee of Novartis Institutes for Biomedical Research, Basel, Switzerland. Authors Krastins, Sarracino, and Lopez are employees of Thermo Fisher Scientific, BRIMS Center, Cambridge, MA. The BRIMS Center paid for half of the control peptides for the selected reaction monitoring assays and ran all of the selected reaction monitoring assays at no cost to the study.


The authors wish to acknowledge Elizabeth Yohannes (The Case Center for Proteomics and Bioinformatics, Case Western Reserve University, Cleveland, OH) and Vinay Vishwanath (Universität für Bodenkultur Wien, Vienna, Austria) for their advice on interpreting the 2-D DIGE data. We also thank Alejandra Garces (Thermo Fisher Scientific, Biomarkers Research Initiatives in Mass Spectrometry [BRIMS] Center, Cambridge, MA) for performing and analyzing the selected reaction monitoring assays of SF.