Abstract

Introduction

Osteoarthritis (OA) is a degenerative disease of the joints that is characterized by destruction of articular cartilage, inflammation of the synovial membrane (synovitis), and remodeling of periarticular bone. Which of these pathogenic processes occurs first is unknown. One proposed scenario is that cartilage breakdown (due to injury or mechanical stress) releases components of the damaged extracellular matrix (ECM) into synovial fluid, and that these ECM components elicit the local production of inflammatory molecules by binding to receptors on resident synovial cells or infiltrating inflammatory cells [1,2]. The inflammatory molecules produced may in turn stimulate production of cartilage-degrading enzymes and recruit inflammatory cells to the affected joint [3,4], thus establishing a vicious cycle of cartilage destruction and inflammation that perpetuates and promotes the OA pathology. Therefore, OA has been described as a chronic wound in which molecules in synovial fluid function as damage-associated molecular patterns (DAMPs; that is, endogenous molecules produced during injury that signal through inflammatory toll-like receptors (TLRs) to effect tissue remodeling) [2,5,6]. Although the identities of the endogenous molecules that mediate synovial inflammation have yet to be confirmed in OA patients or animal models, a continuous supply of DAMPs could perpetuate the early response to injury and thereby damage the joint.

Besides ECM components, many other molecules may act as DAMPs [2]. One such molecule is fibrinogen, which stimulates macrophage production of chemokines in a TLR4-dependent manner [7-9]. Fibrinogen is present at abnormally high levels in OA synovial fluid [10], and the amount of fibrin (the thrombin-cleaved form of fibrinogen [11]) deposited in the synovial membrane correlates with the severity of OA [12]. Although classically a plasma protein, fibrinogen exudes from the vasculature at sites of inflammation, such as the inflamed OA joint, owing to the retraction of inflamed endothelial cells [11]. Fibrinogen is not the only protein to extravasate at sites of inflammation, however, and several other plasma proteins have been detected in OA synovial fluid [10,13]. The extravascular function of most of these plasma proteins is unclear. It is possible that, like fibrinogen, some of these plasma proteins could have an immunoregulatory role at sites of inflammation or tissue damage.

Inflammation is present even in the early stages of OA [14,15], and clinical signs of synovitis correlate with radiographic progression of knee OA [16]. Insight into the cause of synovial inflammation is therefore important in understanding the pathogenesis of OA. Here we used proteomic techniques to survey the proteins present in OA synovial fluid and to evaluate levels of inflammatory cytokines in OA serum and synovial fluid. We then determined whether a subset of the identified proteins could promote inflammation by functioning as immunostimulatory DAMPs.

Material and methods

Results and Discussion

We first used mass spectrometry to survey the proteins present in the synovial fluid of patients with knee OA. Synovial fluid proteins from five OA patients were separated by 1D- or 2D-PAGE and then identified by LCMS. Analysis of all five samples identified a total of 111 unique proteins; three of these were keratin proteins, skin proteins most likely obtained as a result of the cutaneous puncture performed during aspiration of the synovial joints. Eliminating these keratins left 108 unique proteins (Tables ​(Tables22 and ​and3),3), most of which were detected in all synovial fluid samples analyzed. Of these, 44 were identified in a previous proteomic survey of highly abundant proteins in OA synovial fluid [10] (Table ​(Table2).2). Thus, we confirmed the presence of serine protease inhibitors (for example, antithrombin III, α₁-antitrypsin, α₁-antichymotrypsin, kininogen 1) and of proteins important in regulating proteases that degrade cartilage ECM. We also confirmed the presence of proteins involved in cartilage (for example, fibronectin) and/or collagen (for example, gelsolin and collagen α₁, α₂, and α₃ chains) metabolism, and of proteins involved in inflammation or immunity (for example, fibrinogen, AGP 1, complement factors, immunoglobulins, cytokines) (Table ​(Table2),2), findings consistent with the inflammation, ECM degradation, and immune-cell infiltration that characterize OA. Among the 64 proteins that we newly identified (Table ​(Table3)3) were histone-related proteins, macrophage-related proteins, proinflammatory receptors, and proteins related to the proinflammatory transcription factor nuclear factor kappa B (Table ​(Table4),4), presumably reflecting the turnover of resident synovial cells or infiltrating inflammatory cells.

Our mass-spectrometric findings revealed the presence of many molecules associated with inflammation. Although cytokines are also classically associated with inflammation, PAGE-based mass spectrometry is not well suited to the detection of small proteins such as cytokines. We therefore used a multiplex immunoassay to measure levels of inflammatory cytokines and chemokines in synovial fluid samples from 12 patients with knee OA and 14 patients with RA, as well as in serum samples from 24 patients with knee OA, 23 patients with RA, and 35 healthy individuals. Samples from patients with RA, a classic inflammatory arthritis, were used as a comparator. Figure ​Figure11 shows a heatmap of the relative levels of cytokines in the five groups of samples. Compared with cytokine levels in normal sera, cytokine levels in OA sera were generally slightly higher, and those in RA sera were much higher (Figure ​(Figure1).1). SAM analysis revealed that levels of several inflammatory cytokines (for example, IL-1β and IL-6), chemokines (for example, IP-10 (also known as CXCL10), MCP-1, IL-8, MIG, and MIP-1β), and growth factors (for example, VEGF and SCGF-β) were significantly higher in OA sera than in normal sera (FDR < 10%; Figure ​Figure2),2), consistent with previous reports of the association of OA with such inflammatory mediators [27]. Interestingly, we also found OA-associated elevations in levels of IL-9 and cutaneous T-cell attracting chemokine (CTACK), supporting the concept that T cells play a role in OA [28]. As expected, cytokine levels were significantly higher in RA sera than in OA sera (Figure ​(Figure3;3; FDR < 10%).

Open in a separate window

Figure 1

Inflammatory cytokines are associated with osteoarthritis. Relative cytokine levels in serum and synovial fluid (SF) samples from patients with osteoarthritis (OA) or rheumatoid arthritis (RA) and in serum samples from healthy individuals (normal sera). Cytokine levels were measured with a multiplex bead-based immunoassay. Samples from individual patients are listed above the heatmap, and the individual cytokines are listed to the right of the heatmap. IL, interleukin; IFN, interferon; MIG, monokine induced by IFN-γ; IP-10, interferon gamma-induced protein 10; IL-1ra, interleukin-1 receptor antagonist; VEGF, vascular endothelial growth factor; GM-CSF, granulocyte macrophage colony-stimulating factor; FGF, fibroblast growth factor; MCP, monocyte chemotactic protein; IL-2Rα, interleukin-2 receptor α chain; HGF, hepatocyte growth factor; GROα, growth-regulated oncogene α; MIP-1, macrophage inflammatory protein; β-NGF, β nerve growth factor; SCF, stem cell factor; M-CSF, macrophage colony-stimulating factor; SCGF-β, stem cell growth factor β; LIF, leukemia inhibitory factor; SDF-1α, stromal cell-derived factor 1α; G-CSF, granulocyte colony-stimulating factor; CTACK, cutaneous T-cell attracting chemokine.

Open in a separate window

Figure 2

Levels of inflammatory cytokines are higher in OA compared with healthy sera. Cytokines whose levels differ significantly between sera from individuals with osteoarthritis (OA) and sera from age-matched healthy individuals (FDR < 10%). Significance analysis of microarrays (SAM) was used to identify statistically significant differences, and the SAM-generated results were subjected to unsupervised hierarchic clustering. Cytokine levels were measured with a multiplex bead-based immunoassay. Samples from individual patients are listed above the heatmap, and the individual cytokines are listed to the right of the heatmap. IL, interleukin; MIG, monokine induced by IFN-γ; IP-10, interferon gamma-induced protein 10; IL-1ra, interleukin-1 receptor antagonist; VEGF, vascular endothelial growth factor; GM-CSF, granulocyte macrophage colony-stimulating factor; MCP, monocyte chemotactic protein; HGF, hepatocyte growth factor; GROα, growth-regulated oncogene α; MIP-1β, macrophage inflammatory protein 1β; M-CSF, macrophage colony-stimulating factor; SCGF-β, stem cell growth factor β; G-CSF, granulocyte colony-stimulating factor; CTACK, cutaneous T-cell attracting chemokine.

Open in a separate window

Figure 3

Levels of inflammatory cytokines are higher in RA compared with OA sera. Cytokines whose levels differ significantly between sera from individuals with osteoarthritis (OA) and sera from individuals with rheumatoid arthritis (RA) (FDR < 10%). Significance analysis of microarrays (SAMs) was used to identify statistically significant differences, and the SAM-generated results were subjected to unsupervised hierarchic clustering. Cytokine levels were measured with a multiplex bead-based immunoassay. Samples from individual patients are listed above the heatmap, and the individual cytokines are listed to the right of the heatmap. IL, interleukin; IL-1ra, interleukin-1 receptor antagonist; FGF, fibroblast growth factor; MCP, monocyte chemotactic protein; HGF, hepatocyte growth factor; MIP-1, macrophage inflammatory protein; M-CSF, macrophage colony-stimulating factor; SCGF-β, stem cell growth factor β; G-CSF, granulocyte colony-stimulating factor.

Unlike RA, OA is considered a disorder that is restricted to the joints. Indeed, levels of multiple cytokines were much higher in OA synovial fluids than in OA sera (Figure ​(Figure11 and Table ​Table5);5); levels of TNF were negligible in OA sera but substantial in OA synovial fluid (Figure ​(Figure11 and Table ​Table5).5). Our findings suggest that the abnormally high levels of cytokines in OA sera largely reflect overproduction of these cytokines in the joint, consistent with the finding that levels of high-sensitivity C-reactive protein in the serum of OA patients correlate with the degree of inflammatory infiltrate in the patients' joints [29]. Thus, OA is associated with low-grade inflammation that may originate in the joints.

Interestingly, 39 (36%) of the proteins we identified in OA synovial fluid are classically considered plasma proteins (Tables ​(Tables22 and ​and3).3). Indeed, plasma proteins form a large proportion of the proteins enriched in OA synovial fluid relative to healthy synovial fluid [10]. What might these plasma proteins be doing in the OA joint? Like certain products of ECM breakdown [2,5,6], the plasma protein fibrinogen can function as a DAMP and has been proposed to contribute to the pathogenesis of inflammatory arthritis [7-9]. We therefore examined whether other plasma proteins in OA synovial fluid can function as immunostimulatory DAMPs that could contribute to the low-grade inflammation associated with OA.

Key players in OA-associated inflammation are the macrophages [27,30,31]. The cell infiltrate in human OA joints consists mainly of macrophages, and mice depleted of macrophages are relatively resistant to collagenase-induced OA [30]. Macrophages from OA joints produce a number of growth factors, such as VEGF, and inflammatory cytokines, such as the major OA-associated cytokines IL-1β and TNF [30]. We detected VEGF, IL-1β, and TNF in OA synovial fluid in our cytokine screen (Figure ​(Figure11 and Table ​Table5)5) and found that levels of VEGF and IL-1β are significantly higher in OA sera than in normal sera (Figure ​(Figure2).2). VEGF may promote OA pathology by inducing angiogenesis (and thereby osteophyte formation) and by inducing matrix metalloprotease production (and thereby cartilage degradation) [32]. The cytokines produced by macrophages amplify the inflammation in the joints by inducing synovial cells to produce further cytokines and chemokines, as well as matrix metalloproteases [30]. Moreover, macrophages express many of the receptors that mediate DAMP signaling, and they can thus trigger an inflammatory cascade in response to DAMPs present in OA synovial fluid [7-9].

We therefore assessed whether a subset of the identified plasma proteins could induce macrophages to produce TNF, a key cytokine that is thought to drive the inflammatory cascade in OA [27]. We tested α₁-microglobulin (α₁m), α₁-acid glycoprotein 1 (AGP 1; also known as orosomucoid 1), α₂-macroglobulin (α₂m), Gc-globulin (also known as vitamin D-binding protein), albumin, and haptoglobin, all of them plasma proteins detected in our survey of synovial fluid proteins (Table ​(Table2)2) and shown to be enriched in OA synovial fluid [10]. With mouse macrophages, we found that α₁m, α₂m, and Gc-globulin, at concentrations similar to those measured in synovial fluid [33-35], each dose-dependently stimulated the production of TNF, whereas AGP 1, albumin, and haptoglobin did not (Figure ​(Figure4a).4a). The plasma proteins ceruloplasmin, complement component C3, complement component C4, β₂-glycoprotein (also known as apolipoprotein H) also did not stimulate TNF production (data not shown).

Open in a separate window

Figure 4

Plasma proteins detected in osteoarthritic synovial fluid are immunostimulatory. Mouse bone-marrow-derived macrophages (BMMs) and human monocyte-derived macrophages (MDMs) were stimulated for 24 hours with plasma proteins detected in osteoarthritic synovial fluid (Table 2), after which cytokine levels in the supernatants were measured with ELISA or Luminex immunoassay. (a) Levels of TNF produced by mouse BMMs stimulated with the indicated concentrations (in μg/ml) of Gc-globulin, α₁-microglobulin (α₁m), α₂-macroglobulin (α₂m), α₁-acid glycoprotein 1 (AGP 1), albumin, or haptoglobin. Lipopolysaccharide (LPS) was used as a positive control. (b) Levels of TNF produced by human MDMs stimulated with 50 μg/ml of α₁m, α₂m, or Gc-globulin in the presence or absence of 10 μg/ml of polymyxin B, an inhibitor of LPS. LPS (1 ng/ml) was used as a positive control for the efficacy of polymyxin B, and peptidoglycan (PGN; 5 μg/ml), as a negative control. (c) Levels of interleukin-1β (IL-1β), interleukin-6 (IL-6), and vascular endothelial growth factor (VEGF) produced by human MDMs stimulated with 50 μg/ml of Gc-globulin, α₁m, or α₂m. LPS (1 ng/ml) was used as a positive control. Results are representative of experiments performed at least twice. In (a), data are shown as the mean ± SEM of duplicates. In (b) and (c), data are shown as the mean ± SEM of triplicates. *P < 0.05; **P < 0.01, ***P < 0.001; NS, not significant.

We next examined the effect of α₁m, α₂m, and Gc-globulin on cytokine production in human macrophages. Because the endotoxin LPS is a common contaminant and is itself an agonist of TLR4, we tested the stimulatory properties of the plasma proteins in the presence of polymyxin B, a compound that neutralizes LPS. In the presence of polymyxin B, α₁m-, α₂m-, and Gc-globulin-induced TNF production was not significantly reduced, whereas LPS-induced TNF production was abrogated (Figure ​(Figure4b).4b). Additionally, pretreatment with proteinase K significantly abrogated TNF production induced by the plasma proteins but not TNF production induced by LPS (Figure ​(Figure5).5). Although we cannot exclude the possibility that a small component of the observed stimulation is due to endotoxin, this result confirms that the plasma proteins are themselves immunostimulatory. Gc-globulin, α₁m, and α₂m were also able to induce the production of several other inflammatory cytokines that were upregulated in OA serum and synovial fluid (Figures ​(Figures11 and ​and2):2): IL-1β, IL-6, and VEGF (Figure ​(Figure4c).4c). Thus, Gc-globulin, α₁m, and α₂m can each induce the production of TNF, IL-1β, IL-6, and VEGF, all molecules implicated in the pathogenesis of OA [27,30,31].

Open in a separate window

Figure 5

Induction of TNF production by plasma proteins is not due to endotoxin contamination. RAW264.7 macrophages were stimulated for 24 hours with 50 μg/ml of Gc-globulin, α₁-microglobulin (α₁m), or α₂-macroglobulin (α₂m) that had been incubated with proteinase K (20 μg/ml) at 55°C for 4 hours in the presence of β-mercaptoethanol and then heated to 100°C for 10 minutes. TNF levels in the supernatants were determined with ELISA. Lipopolysaccharide (LPS; 1 ng/ml) was used as a positive control. Data are shown as the mean ± SEM of duplicates from one of two representative experiments. *P < 0.05; **P < 0.01.

But how do these plasma proteins stimulate cytokine production? To determine whether these immunostimulatory plasma proteins signal through TLR4, we examined whether Gc-globulin, α₁m, and α₂m could also induce TNF production in TLR4-deficient macrophages. TLR4 deficiency inhibited Gc-globulin-, α₁m-, and α₂m-induced TNF production (Figure ​(Figure6).6). Confirming that the defect in inflammatory signaling in the Tlr4^lps-del macrophages was specific to the TLR4 pathway, theTLR2-specific agonist peptidoglycan was able to induce TNF production in these cells-in fact, to a greater degree than in wild-type cells (possibly because of compensatory mechanisms operating within the TLR family) (Figure ​(Figure6).6). Thus, Gc-globulin-, α₁m-, and α₂m-induced production of TNF is dependent on TLR4.

Open in a separate window

Figure 6

Plasma proteins detected in osteoarthritic synovial fluid stimulate macrophage TNF production via TLR4. Levels of TNF produced by wild-type (wt) or TLR4-deficient (Tlr4^lps-del) mouse bone-marrow-derived macrophages stimulated for 24 hours with 50 μg/ml of Gc-globulin, α₁-microglobulin (α₁m), or α₂-macroglobulin (α₂m), after which TNF levels in the supernatants were determined with ELISA. Lipopolysaccharide (LPS; 1 ng/ml) was used as a positive control for TLR4-dependent TNF production, and peptidoglycan (PGN; 5 μg/ml) as a positive control for TLR4-independent TNF production. Data are shown as the mean ± SEM of triplicates from one of three representative experiments. *P < 0.05; **P < 0.01.

Interest in the putative immunomodulatory effects of α₁m, α₂m, and Gc-globulin is increasing, with both proinflammatory and antiinflammatory properties suggested for each of them [36-38].

For example, α₁m has been shown to bind to the surface of various inflammatory cells and to either stimulate or inhibit the activation of human lymphocytes [38]. The immunoregulatory role of α₁m in health and disease is likely to be context dependent. Gc-globulin, however, appears to be primarily proinflammatory: it enhances the neutrophil- and monocyte-chemotactic activity of the anaphylatoxin C5a [36] and, in its sialic-acid-free form, activates macrophages [39]. Here, we uncover an additional mechanism by which these plasma proteins could promote inflammation. We speculate that exudation into extravascular spaces at sites of tissue damage and inflammation may render these plasma proteins inflammatory by bringing them into contact with TLR-expressing macrophages. Our finding that certain plasma proteins present in OA synovial fluid can induce macrophage production of inflammatory cytokines supports the model of local production of inflammatory mediators in the joints in OA.

Conclusions

We identified 108 proteins in OA synovial fluid and showed that OA is associated with low-grade inflammation. We found that plasma proteins form a large proportion of the proteins present in OA synovial fluid and that certain of these plasma proteins can signal through TLR4 to induce the production of an array of inflammatory cytokines, including those upregulated in OA. Our findings suggest that plasma proteins present in OA synovial fluid, whether through exudation from the plasma or production by synovial tissues, could contribute to low-grade inflammation in OA by functioning as DAMPs.

Abbreviations

α₁m: α₁-microglobulin; AGP 1: α₁-acid glycoprotein 1; α₂m: α₂-macroglobulin; BMM: bone-marrow-derived macrophage; DAMP: damage-associated molecular pattern; ECM: extracellular matrix; IL-1β: interleukin-1β; IL-6: interleukin-6; IP-10: interferon gamma-induced protein 10; LCMS: chromatography tandem mass spectrometry; LPS: lipopolysaccharide; MCP-1: macrophage chemotactic protein-1; MDM: monocyte-derived macrophage; MEM: minimal essential medium; MIG: monokine induced by interferon-γ; MIP-1: macrophage inflammatory protein-1; OA: osteoarthritis; PAGE: polyacrylamide gel electrophoresis; RA: rheumatoid arthritis; TLR: Toll-like receptor; TNF: tumor necrosis factor; VEGF: vascular endothelial growth factor.

Competing interests

The authors declare that they have no competing interests.

Authors' contributions

JS and WHR conceived the studies. OS performed the mass spectrometric analysis of synovial fluid. PEC, LJL, and JCE performed the multiplex cytokine analysis. JCE, KAB, and TPA collected and provided OA sera. DHS performed the in vitro macrophage stimulation assays. JS, WHR, DHS, OS, TML, and IH analyzed the resulting datasets. TML and JS wrote and edited the manuscript with the input of WHR, OS, and DHS. All authors read and approved the final manuscript.

Notes

See related editorial by Goldring and Scanzello, http://arthritis-research.com/content/14/2/111

References