To identify potential molecular mediators and biomarkers for osteoarthritis (OA), through comparative proteomic analysis of articular cartilage tissue obtained from normal donors without OA (n = 7) and patients with OA (n = 7).
To identify potential molecular mediators and biomarkers for osteoarthritis (OA), through comparative proteomic analysis of articular cartilage tissue obtained from normal donors without OA (n = 7) and patients with OA (n = 7).
The proteomic analyses comprised extraction of soluble proteins from cartilage, separation of the protein mixtures by sodium dodecyl sulfate–polyacrylamide gel electrophoresis followed by in-gel digestion, and subsequent nano-liquid chromatography–tandem mass spectrometry analysis in conjunction with a database search for protein identification and semiquantitation.
A total of 814 distinct proteins were identified with high confidence from 14 samples; 420 of these proteins were detected with ≥3 unique peptides in at least 4 samples from the same group. Using stringent criteria, 59 proteins were found to be differentially expressed in OA cartilage. Gene Ontology and Ingenuity pathway analysis tools were used to characterize these proteins into functional categories. One of the up-regulated proteins, HtrA1, a serine protease, was detected at high levels in cartilage.
Altered protein expression in the disease state is associated with many aspects of the pathogenesis of OA, such as increased proteolysis, lipid metabolism, immune response, and decreased signal transduction. To our knowledge, this is the first time that a large portion of these proteins and their expression patterns were identified in cartilage, thus providing new insights for finding novel pathologic mediators and biomarkers of OA.
Osteoarthritis (OA) is a prevalent joint disease characterized by irreversible erosion and destruction of articular cartilage and subchondral bone. OA is a progressive, degenerative joint disease that has a major impact on joint function and the patient's quality of life, due to chronic pain. Cartilage is a connective tissue consisting of a single cell type (the chondrocyte, comprising 2–5% of the tissue weight) embedded in a dense extracellular matrix (ECM; primarily type II collagen, anionic aggrecan molecules, and hyaluronic acid) (1–6). It is generally accepted that, at the molecular level, cartilage degeneration is characterized by a general failure of chondrocytes to maintain an appropriate balance between synthesis and degradation of ECM macromolecules. Anabolic factors, such as various growth factors, bone morphogenetic proteins, and enzyme inhibitors, promote ECM production, while catabolic factors, including proinflammatory mediators (i.e., cytokines), chondrocyte apoptosis, and degradative enzymes and their activators, lead to the detrimental turnover of the ECM (6).
Differential proteomics aims to identify global protein components in biologic systems and provide a snapshot of the protein quantity change in response to pathologic stress, disease progression, or drug treatment. Both 2-dimensional (2-D) gel and automated multidimensional high-performance liquid chromatography (HPLC) in combination with mass spectrometry (MS) have been used with success for global profiling of complex protein mixtures such as cell lysates, tissues, and body fluids (7–11). For relative quantitation, proteins or their corresponding peptides in samples used for comparison can be differentially labeled, either metabolically or chemically, with stable isotopes, and the intensity ratio of the corresponding isotope-labeled peptide pairs can be measured by MS (12–14).
Recently, the capability of this approach has been extended from labeled pairs to multiplex labeling, allowing simultaneous quantitation of protein abundance in >2 samples (15–17). Alternatively, several investigators have demonstrated the feasibility of label-free approaches based on the measurement and comparison of peptide ion intensities for global semiquantitation of proteins (8, 11, 18–23). Chelius et al (19) showed that peak areas of peptide precursor ions, as calculated using alternating MS full scans in LC–tandem MS (LC-MS/MS) experiments, were closely correlated with protein concentrations, even in complex proteomes such as human serum. More recently, our group and several other groups of investigators (8, 11, 21) demonstrated that the peptide total ion intensity of a protein, based on MS scanning in data-dependent LC-MS/MS experiments, can be used as a measure of its relative abundance across multiple samples, to identify differentially expressed proteins. More importantly, the results obtained using this label-free approach correlated well with those obtained by independent proteomic studies based on 2-D gel or enzyme-linked immunosorbent assay (8, 11, 24).
Several studies have been conducted to search for pathologic mediators and/or biomarkers of OA using proteomic characterization of synovial fluids, serum, cultured chondrocytes, and cartilage tissue (2, 5, 25–32). Hermansson et al (2) used 2-D gel and MS to analyze newly synthesized proteins secreted by human OA cartilage explants cultured in medium containing 35S-methionine/cysteine. Those studies revealed increased synthesis of type II collagen and activin A proteins in OA cartilage. Ruiz-Romero et al (5) described protein profiles of human articular chondrocytes isolated from the cartilage of normal individuals, from which 93 unique proteins were identified. However, due to inherent alterations to the microenvironment, proteomic profiling of cultured chondrocytes may not fully reflect the spectrum of cellular proteins made in the cells surrounded by cartilage matrix in situ.
More recently, Vincourt et al (31) reported a method for direct proteomic characterization of human articular cartilage using 2-D electrophoresis and matrix-assisted laser desorption ionization−time-of-flight MS, leading to the identification of 127 proteins with diverse functions. Garcia et al (32) also reported proteomic profiling of OA cartilage using 1-D sodium dodecyl sulfate–polyacrylamide gel electrophoresis (SDS-PAGE) and tandem MS, which resulted in the identification of ∼100 proteins. The majority of these proteins were located in the ECM, since their sample preparation protocol excluded chondrocytes. To date, differential proteomic profiling of articular cartilage tissue from healthy individuals and patients with OA has not been reported.
To better understand the molecular mechanism and identify pathologic mediators and biomarkers for OA, we carried out a differential proteomic study to identify and semiquantitate proteins in articular cartilage tissue obtained from 7 patients with OA and 7 normal individuals. Our sample preparation protocol was designed to extract both extracellular and intracellular proteins from cartilage depleted of highly abundant matrix proteins such as collagens and aggrecan, thus allowing detection of many functionally important proteins that are usually less abundant in articular cartilage. A total of 814 distinct proteins were identified with 2 or more unique peptides, among which 59 proteins, including several that are implicated in OA pathology (such as matrix metalloproteinases [MMPs]), were found to be present at significantly different levels in OA cartilage compared with normal cartilage. The proteomic analysis also revealed several proteins that had not been previously reported to be associated with the pathology of OA.
Specimens of normal cartilage obtained from tissue donors with no history of joint disease were received within 36 hours following autopsies (performed by National Disease Research Interchange, Philadelphia, PA) and examined macroscopically to ensure that the cartilage specimens had smooth, intact surfaces without OA-like lesions. Specimens of OA cartilage with visible lesions were obtained from patients (ages 50–82 years) undergoing knee replacement surgery at New England Baptist Hospital, Boston, MA. These specimens were stored on ice for ∼36 hours before being processed for protein extraction.
Within this time frame, the majority (>90%) of chondrocytes were still alive, because the specimens were kept in Dulbecco's modified Eagle's medium–Ham's F-12 medium on ice. Applicable regulations and guidelines regarding patient consent and donor confidentiality were followed. Upon receipt, cartilage slices were carefully dissected and separated from other bony, bloody, or fibrous pieces. Cartilage tissue (1.2 gm wet weight) from each study subject was minced into ∼2 × 2 × 2–mm pieces and extracted for 4 days at 4°C with 12 ml of extraction buffer (4M guanidine HCl, 50 mM sodium acetate [pH 5.8]) containing a proteinase inhibitor cocktail (Roche, Grenzach, Germany). The resulting mixture was centrifuged at 3,000 revolutions per minute for 5 minutes at 4°C to remove the insoluble matrix (mostly collagen).
Solid CsCl was added to cartilage extracts to achieve a final density of 1.45 gm/ml, and the mixture was ultracentrifuged at 100,000g for 72 hours at 4°C. Three fractions of equal volume were collected, from top to bottom, for each tube. Essentially all of the sulfated proteoglycan (mainly aggrecan) fractionated in the bottom third of the tube, as measured by dimethylmethylene blue assay. Nonaggrecan cartilage proteins were found exclusively in the top fraction, as confirmed using SDS-PAGE gels. The top fraction was collected and desalted by dialysis at 4°C against 20 mM (pH 8.2) Tris HCl buffer containing proteinase inhibitor cocktail, using Slide-A-Lyzer 7K MWCO dialysis cassettes (Pierce, Rockford, IL). The concentration of the extracted proteins was measured by bicinchoninic acid assay (Pierce), following the manufacturer's protocol.
Each sample (containing 110 μg protein) was concentrated in a vacuum concentrator (SpeedVac; Thermo, San Jose, CA) to a final volume of ∼40 μl. The resulting solution was mixed with 5× SDS loading buffer. Proteins were reduced with 20 mM dithiothreitol at 90°C for 5 minutes and alkylated with 50 mM iodoacetamide at ambient temperature in the dark for 20 minutes. Samples were then loaded onto a 5-well 10–20% tricine mini-gel (Invitrogen, Carlsbad, CA). After staining with Coomassie blue, each gel lane was horizontally divided into 29 slices. Each slice was then minced into 1 × 1–mm pieces and subsequently subjected to in-gel digestion with sequencing-grade modified trypsin (0.5 μg/gel slice; Promega, Madison, WI) in a digestion robot (DigestPro; AbiMed Analysen-Technik, Langenfeld, Germany) for 18 hours. The tryptic digests were concentrated in a SpeedVac to a final volume of 30 μl, prior to MS analysis.
Tryptic digests of proteins were analyzed with an automated nanoLC-MS/MS system, using a Famos autosampler (LC Packings, San Francisco, CA) and an 1100 HPLC binary pump (Agilent, Wilmington, DE) coupled to an LTQ ion trap mass spectrometer (Thermo Finnigan, San Jose, CA) equipped with a nanospray ionization source. Ten microliters of the digest solution was injected onto a reverse-phase PicoFrit column (New Objective, Woburn, MA) packed with Magic C18 media (5-μm particle, 200-angstrom pore size, 75 μm × 10 cm). Peptides were eluted at a flow rate of 0.2 μl/minute, using a 90-minute linear gradient from 2% to 55% B (mobile phase A: 0.1% formic acid aqueous solution; mobile phase B: 90% acetonitrile and 0.1% formic acid). The spray voltage was 1.8 kV, the heated capillary temperature was maintained at 180°C, and the collision energy for MS/MS was set at 35 units.
Automated data-dependent MS analysis was carried out using the dynamic exclusion feature built into the MS acquisition software (Xcalibur 1.3; Thermo Finnigan). Each MS full scan (mass/charge [m/z] 350–1,600) was followed by MS/MS acquisition of the 4 most intense precursor ions detected in the prior MS scan, to obtain as many collisionally induced dissociation spectra as possible.
SpectrumMill version 3.1 software (Agilent) was used for database searches and for data processing. The mass spectrometric raw data were searched against the human subset of the National Center for Biotechnology Information (NCBI) nonredundant protein database (135,279 protein entries, updated as of July 2005). Search parameters included a static modification on cysteine residues (carbamidomethylation), signal-to-noise ratio >25, sequence tag length >1, precursor ion mass range 600–3,500 daltons, retention time 0–135 minutes, 70% minimum matched peak intensity ± 2.5-dalton tolerance on precursor ions and ± 0.7-dalton tolerance on product ions, 1 missed tryptic cleavage, and electrospray ionization trap scoring parameters as defined by the searching algorithm. Scans for the precursor ion m/z ± 0.7 dalton within a ±45-second time window were merged.
All of the database search results were further validated by applying the designated protein and peptide scores as well as the following user-defined criteria: for the protein validation mode, protein score >20, peptide scored percent intensity (SPI) >70% for all charge states, peptide score >7 for peptide charge +1, peptide score >8 for peptide charge +2, and peptide score >9 for peptide charges ≥+3; for the peptide validation mode, peptide SPI >70%, and peptide score >13 for all charge states. The search criteria used here would result in a false-positive rate of <4%, as previously described (8). The proteins identified from individual samples were reassembled to generate a summary that displays the results for all 14 samples. To address the database redundancy issue, proteins that share common peptides were grouped together and displayed as a single protein group in “Protein Centric Columns” mode, as described in the SpectrumMill search engine. Within any given protein group, the protein with the highest score was selected as the most likely correct search result.
To semiquantitate the protein relative abundance in multiple samples, the peak area of the extracted ion chromatograms for each peptide precursor ion in the full scan was calculated in the region ±1.4 m/z and ±75 scans, using SpectrumMill. An individual protein's abundance was then calculated as the sum of the total ion current (TIC) values measured for all peptide precursor ions derived from that protein. Thus, the relative concentration of each protein was determined by comparing total MS intensities of all identified peptides from that protein in 1 sample versus those from other samples. The annotation of protein cellular localization and biologic function was performed using Ingenuity (www.ingenuity.com) and Gene Ontology (GO; http://amigo.geneontology.org) software.
The TIC for all peptides identified for a specific protein was used as a measure of relative abundance of that protein in the sample. Differential expression is based on the comparison of the mean TIC values for protein in OA and normal samples. Only proteins identified from ≥3 unique peptides were qualified for selection. A protein was considered to be differentially expressed if it was identified in ≥4 samples in at least 1 group (normal or OA) and the change in its mean TIC was >5-fold in either direction, from normal to OA tissue or from OA to normal tissue. The significance of such changes was measured by Student's t-tests of log10-transformed TIC values. The false discovery rate (FDR) adjustment to the resulting raw P values was calculated using the Benjamini-Hochberg procedure (33).
A reproducible procedure to remove highly abundant matrix macromolecules is critical to successful proteomic analysis of physiologically relevant proteins that may be present at very low levels. Following dissection and washing to remove extraneous proteins, cartilage specimens were extracted using 4M guanidine HCl in the presence of a proteinase inhibitor cocktail. Extraction using 4M guanidine HCl had been shown to be able to solubilize both the cellular and matrix proteins from cartilage (34). Collagens (primarily type II collagen), the most abundant cartilage matrix proteins, are insoluble under these conditions and were removed by centrifugation.
The cartilage extracts were further fractionated by CsCl gradient centrifugation, and the bulk of the aggrecan, the second most abundant protein component of cartilage, was subsequently separated from other cartilage proteins. SDS-PAGE analysis of the fractions from the CsCl gradients demonstrated that all of the cartilage proteins were accumulated in the top one-third fraction, while few or no proteins were detected in other fractions (data not shown). Thus, only the top one-third fraction (lowest buoyancy density) was processed for proteomic analysis. Cartilage sample information (donor age and sex) as well as the concentration of extracted cartilage protein from each study subject are listed in Supplementary Table 1 (available online at http://bioinfo.public.wyeth.com/WHB). To compare relative protein expression in different samples, an equivalent amount of proteins from each sample (110 μg) was analyzed. As shown in the typical SDS-PAGE gel in Figure 1, the expression pattern of the abundant proteins in OA and normal samples appeared similar.
The number of proteins identified by LC-MS/MS in conjunction with database searching against human proteins in the NCBI nonredundant database is shown in Supplementary Table 1. Supplementary Tables 2 and 4 (available online at http://bioinfo.public.wyeth.com/WHB) summarize the detailed information for each of the assigned proteins. There was no significant bias in the number of proteins identified in the OA or normal groups. A total of 814 distinct proteins representing a wide spectrum of functional classes were identified with high confidence from 14 cartilage samples (see Supplementary Table 2. Supplementary Tableq 4 lists an additional 400 proteins that were identified in this study, as assigned with a single unique peptide; however, these 400 proteins are not discussed in this report). Many of the identified proteins have not been previously reported to be components of articular cartilage. The cellular localization and molecular function of the identified proteins were categorized by GO and are shown in Figure 2. Both intracellular and extracellular proteins with various functions were identified. As expected, ECM proteins such as fibronectin, cartilage oligomeric matrix protein, thrombospondin 1, cartilage intermediate-layer protein, and decorin were among the most abundant, as measured by the sums of TIC values of their corresponding peptides. The protein list also contains several proteases and their inhibitors, as well as cytokines, kinases, and other signaling molecules that are usually present in very low amounts in cartilage.
To evaluate the relative changes in the levels of protein expression, a variety of criteria were applied, as follows: 1) the protein must be identified from ≥3 unique peptides; 2) the protein must be identified in ≥4 samples in at least 1 group; 3) the difference in the mean TIC values in 2 groups must be ≥5-fold (if the protein was not detected in a sample, a TIC value of 104, approximately equivalent to the noise level of the ion trap instrument, was arbitrarily assigned); and 4) FDR-adjusted (33) P values of ≤0.10 by t-test (corresponding to raw P values less than 0.03) were applied to further narrow down the list of differentially expressed proteins. Among the 814 unique proteins, 595 were identified from ≥3 peptides, and 420 of these 595 proteins were identified in ≥4 samples from at least 1 group (see Supplementary Table 3, available online at http://bioinfo.public.wyeth.com/WHB). One hundred eighty-three of these proteins exhibited a mean TIC change of ≥5-fold in either direction. However, only 59 met the criteria after applying the FDR-adjusted P value cutoff (Table 1). The changes in these 59 proteins between normal and OA samples are also shown in the clustergram in Supplementary Figure 1 (online at http://bioinfo.public.wyeth.com/WHB). For example, HtrA serine protease 11 (HtrA1; also called PRSS11), the most abundant protease in cartilage, was identified in all normal and OA samples from a total of 29 unique peptides (Table 1). Its expression in OA cartilage was ∼8-fold higher than that in normal cartilage (mean TIC1.8E + 09 versus 2.3E + 08).
|Protein name||Accession no.||No. of unique peptides||Log, normal mean||Log, OA mean||Fold change||FDR|
|Thrombospondin 1 precursor||40317626||90||9.468||8.553||−8.2||0.041|
|Fibrinogen α-chain isoform||11761629||44||6.354||9.086||539.7||0.030|
|Serine protease 11||4506141||29||8.224||9.133||8.1||0.030|
|Fibrinogen, γ-chain isoform||11761633||25||5.915||8.687||590.8||0.030|
|EGF-containing fibulin-like ECM protein 1||15072400||20||4.832||7.566||542.5||0.047|
|Annexin VI, isoform 1||16877589||16||7.220||5.433||−61.3||0.093|
|Complement component 4 binding protein||4502503||14||4.741||7.144||252.7||0.045|
|Matrix metalloproteinase 2 preproprotein||11342666||14||5.456||7.554||125.3||0.062|
|Complement factor H–related 5||13540563||14||5.321||7.651||213.6||0.031|
|Nucleolin (protein C23)||128841||13||6.535||4.514||−105.0||0.073|
|Latent TGFβ-binding protein||1082570||13||7.124||5.268||−71.9||0.065|
|Tissue inhibitor of metalloproteinases 2||1517893||12||6.020||7.531||32.4||0.087|
|Anti-HBsAg immunoglobulin Fab κ-chain||3721651||11||7.869||8.640||5.9||0.073|
|Inter-alpha-trypsin inhibitor heavy-chain H2||125000||11||4.000||7.458||2,869.7||0.000|
|Apolipoprotein H (β2-glycoprotein I)||32165624||10||5.493||7.840||222.7||0.064|
|Similar to 60S acidic ribosomal protein P2||55635097||10||5.661||7.157||31.3||0.094|
|F-box only protein 2||15812198||10||7.497||4.962||−343.0||0.030|
|Complement factor B||14124934||10||4.749||6.756||101.6||0.065|
|Ig λ light-chain VLJ region||21669507||9||7.356||8.073||5.2||0.073|
|Coagulation factor II||30802115||9||4.000||6.313||205.8||0.030|
|Ig α-1 heavy-chain constant region||184749||9||4.000||6.540||347.1||0.030|
|Transmembrane protein 4||37183321||7||5.280||6.881||39.9||0.093|
|Glutathione S-transferase M2||4504175||6||6.317||4.380||−86.5||0.039|
|Inositol(myo)-1 (or 4)-monophosphatase 1||5031789||6||6.023||4.000||−105.4||0.030|
|H1 histone family, member X||5174449||6||6.868||4.364||−319.2||0.003|
|Unknown (protein for IMAGE:4901992)||30353943||6||6.921||4.957||−92.2||0.094|
|Acrosomal serine protease inhibitor||13195769||6||4.000||6.151||141.7||0.030|
|Proteasome β1 subunit||12653473||5||6.303||4.257||−111.4||0.030|
|α1 acid glycoprotein 1||1197209||4||4.000||5.857||71.9||0.073|
|Acid phosphatase 1 isoform c||4757714||4||4.357||5.948||38.9||0.090|
|Similar to fatty acid–binding protein, adipocyte||55630858||4||4.000||5.738||54.7||0.081|
|Small inducible cytokine B10||15012099||3||5.571||4.267||−20.1||0.087|
|Branched-chain aminotransferase 1, cytosolic||38176287||3||5.820||4.351||−29.5||0.093|
|Transforming growth factor β1||10863873||3||5.210||4.000||−16.2||0.073|
The differentially expressed proteins shown in Table 1 represented a wide range of biologic categories, including major serum proteins such as apolipoproteins, hemopexin, fibrinogen, complement components; degradative proteases and protease inhibitors such as HtrA1, MMP/tissue inhibitor of metalloproteinases (TIMP), acrosomal serine protease inhibitor, and cystatin C; cartilage anabolic effectors such as different forms of transforming growth factor β (TGFβ) and TGFβ binding proteins; extracellular matrix proteins such as thrombospondin 1; as well as several interesting proteins whose properties and functions in OA pathophysiology have not been well characterized, such as fibulin 3, osteonectin, secreted modular calcium-binding protein 2, vitrin, and tenomodulin.
The proteins identified in cartilage tissue were analyzed by a variety of bioinformatics tools. Based on the TIC values of the 420 proteins (see Supplementary Table 3) identified in each experimental sample, the 14 cartilage samples could be separated into 2 distinct groups of OA and normal samples (Figure 3). This result demonstrated that OA cartilage samples have distinct protein expression profiles that can be distinguished from the protein profile of normal cartilage. Pathway analysis using tools provided by Ingenuity (www.ingenuity.com) showed that the 59 differentially expressed proteins in OA (those with an FDR-adjusted P value of less than or equal to 0.10) cover a broad range of protein functional classes (Figure 4). A significant portion of proteins were involved in tissue development (16 of 59), cell–cell signaling and interaction (17 of 59), lipid/carbohydrate metabolism (18 of 59), molecular transport (13 of 59), cellular movement (12 of 59), cell death (7 of 59), and immune response (8 of 59).
In this study, we performed differential proteomics for the analysis of proteins in human articular cartilage, to elucidate changes in protein expression in patients with OA. The unique features of our study are as follows. First, this is, to our knowledge, the first application of comparative proteomics to cartilage tissue (as opposed to cultured chondrocytes) from normal individuals as well as patients with OA. Second, a total of 814 distinct proteins were identified, and a large proportion of these proteins and their expression changes was identified for the first time from cartilage tissue, thus providing a comprehensive cartilage proteome database that is essential in understanding the molecular mechanisms underlying OA. Third, 59 proteins displayed significant differences in abundance in association with OA. The functions of the proteins with significantly different abundance in OA versus normal articular cartilage may be associated with the pathogenesis of OA, such as proteolysis, lipid metabolism, immune response, cell death, and signal transduction.
A major obstacle in applying proteomics to cartilage tissue is the heterogeneous nature of cartilage proteins. Articular cartilage contains mainly ECM proteins, of which type II collagen and aggrecan comprise >90% of the total dry mass, whereas chondrocytes, the only cell type in cartilage, account for <5% of the total cartilage volume. As a result, the separation of crosslinked collagen and highly sulfated proteoglycan aggrecan was critical to the overall success of the strategy. In fact, neither collagens nor aggrecan was present as predominant cartilage proteins in our proteomic analyses.
Proteomic analysis of human articular cartilage demonstrated variability in the abundance of many proteins between different individuals, as indicated by highly variable TIC values for an individual protein in different samples. The criteria for selecting differential proteins included not only the confidence of the protein identification and the relative TIC changes but also the occurrence frequencies of the proteins within a group. These stringent criteria ensured that a selected protein was truly present in a majority of samples within a group, and that the altered abundance was not random. To facilitate the data processing, undetected (“absent”) proteins were arbitrarily assigned a TIC value of 104, which was approximately equivalent to the instrument noise level and significantly lower than most of the TIC values observed in the data set. Protein expression changes from normal to OA groups were assessed by the logarithmic difference of their mean TIC values, and the significance of such changes was assessed. The P value for the mean TIC change for each protein was calculated, and only those with an FDR-adjusted P value less than or equal to 0.1 were selected, to ensure that the changes in expression for individual proteins between the normal and OA groups were truly significant. This resulted in 59 proteins that were significantly changed (Table 1).
The etiology and pathophysiology of OA disease are complex and only partially understood. However, it is generally accepted that an imbalance between anabolic and catabolic pathways in matrix turnover contributes significantly to the mechanism (1, 3, 4, 6). The increased synthesis of degradative enzymes such as MMP and ADAMTS may largely account for the unbalanced breakdown of the cartilage ECM. Several cartilage proteases and protease inhibitors were identified in the current study. Among these, the levels of HtrA1, TIMP-2, MMP-2, cystatin C, and acrosomal serine protease inhibitor were all significantly elevated in OA cartilage, while the levels of many others were also increased, but to a lesser extent (Table 1 and Supplementary Table 3). HtrA1 was the most abundant protease identified in cartilage tissue, as estimated by the TIC value generated by unique peptides. This result confirmed previous findings reported by Hu et al (35) and Grau et al (36), showing elevated expression of HtrA1 at the transcriptional level in OA articular cartilage, and elevated levels of HtrA1 protein (∼7-fold above normal) in synovial fluids obtained from patients with OA.
In vitro studies, including those by our group, have shown that HtrA1 is able to digest several cartilage matrix proteins, including aggrecan, decorin, fibromodulin, and soluble type II collagen (37), and to inhibit TGFβ-induced matrix synthesis by chondrocytes (Yang Z, et al: unpublished observations). Chondrocytes also synthesize a variety of MMPs, including collagenases (MMP-1, MMP-8, and MMP-13), gelatinases (MMP-2 and MMP-9), stromelysin 1 (MMP-3), and membrane-type MMPs. MMP-1, MMP-2, and MMP-3 were identified in this study, but only MMP-2 was altered in quantity in OA cartilage. Cysteine proteinases (such as cathepsins B, H, K, L, and S) represent another important enzyme family that might play a role in cartilage matrix destruction. Several cathepsins, particularly cathepsin K, have been shown to participate in OA cartilage degradation (38–40). Cystatin C, a potent extracellular inhibitor of cysteine proteinases, has also been linked to arthritis. Results from transcriptional profiling of cysteine proteinases and their inhibitors in a transgenic mouse model of OA showed that up-regulation of cathepsin K, but not cystatin expression, coincided with the onset of articular cartilage damage (41). However, our data strongly indicated that the level of cystatin C is increased in human OA cartilage.
The levels of several plasma proteins such as apolipoproteins (Apo A-I, II, IV, and Apo H) are significantly elevated in OA cartilage. The functional role of lipoproteins and their receptors in OA and other joint diseases has been described in the literature (42, 43). The observed up-regulation of these proteins in OA may also imply their involvement in disease mechanism that has not yet been fully investigated.
Finally, another of the interesting protein families identified in cartilage tissue comprised the fibulins. Fibulins are ECM proteins (44), and 6 different family members have been reported. All are characterized by a unique C-terminal fibulin-type module and several repeated epidermal growth factor–like domains. We identified 3 different fibulin proteins in cartilage: fibulin 1D, fibulin 2, and fibulin 3 (Figure 5). Fibulin 2 was identified in only 2 OA samples and 1 normal sample, at very low levels. In contrast, fibulin 1D and fibulin 3 were identified with strong abundance and were primarily observed in gel bands consistent with their intact molecular masses. The expression of fibulin 1D was slightly elevated in OA cartilage (mean TIC3.1E + 08 in OA cartilage versus 1.2E + 08 in normal cartilage), whereas expression of fibulin 3 was strongly increased in OA cartilage compared with normal cartilage (mean TIC3.6E + 08 versus 3.9E + 06).
Fibulin 1D binds to proteoglycans and was recently reported as being a cofactor of ADAMTS-1, an inhibitor of angiogenesis and a mediator of proteolytic cleavage of aggrecan. When colocalized in vivo with ADAMTS-1, fibulin 1D was shown to enhance the capacity of ADAMTS-1 to cleave aggrecan (45). A recent study demonstrated that fibulin 3 is a binding partner of TIMP-3, which is a known inhibitor of MMPs and aggrecanases (46). Collectively, such evidence indicates that the fibulin family might play an important role in cartilage metabolism.
To our knowledge, the current study is the first comprehensive proteomic analysis of soluble proteins extracted from normal and OA cartilage. It provides the most complete catalog of the proteins involved in cellular metabolism and organization, as well as proteins participating in the regulation of ECM synthesis and turnover of articular cartilage. Although the present findings are based on a limited number of study subjects, the results, from the protein level, clearly reflect the complex pathophysiology of OA. This study provides new insights for identifying novel OA mediators that can be potential therapeutic targets or biomarkers for OA.
Dr. Yang 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 design. Wu, Bemis, Qiu, Flannery, Yang.
Acquisition of data. Wu, Bemis, Wang, Yang.
Analysis and interpretation of data. Wu, Liu, Bemis, Morris, Flannery, Yang.
Manuscript preparation. Wu, Liu, Flannery, Yang.
Statistical analysis. Wu, Liu, Bemis, Yang.
We thank our colleagues Edward LaVallie, Lisa A. Collins-Racie, Maya Arai, Katy Georgiadis, Manas Majumdar, Sonya Glasson, Vishnu Daesety, Priya Chocklingam, and Aled Jones for constructive comments, and Ioannis Moutsatsos and Patrick Cody for software support. We also are grateful to the New England Baptist Hospital (Boston, MA) and the National Disease Research Interchange (Philadelphia, PA) for providing cartilage tissue.