All amino acid numbering of aggrecan is herein based on full-length human aggrecan, accession number [Swiss-Prot:P16112], starting with the N-terminal ¹MTTL-amino acid sequence.

Alcian blue 8GS (C.I. 742240) was from Chroma-Gesellschaft (Köningen, Germany). 4-(2-aminoethyl)-benzenesulfonyl fluoride (AEBSF), 6-aminohexonic acid (EACA), benzamidine-HCl, BSA, chondroitin sulfate type C from shark cartilage (no. C4384), ethylenediaminetetra acetic acid (EDTA), N-ethylmaleimide (NEM), 2-(N-morpholino) ethanesulfonic acid (MES), phenylmethylsulfonyl fluoride (PMSF), and phosphate buffered saline with TWEEN (PBST) buffer (0.01 M sodium phosphate, 0.138 M sodium chloride, 0.0027 M potassium chloride, 0.05% TWEEN 20; pH 7.4) were from Sigma (St. Louis, MO, USA). Cesium chloride and guanidinium hydrochloride were from Merck (Darmstadt, Germany). Molecular weight markers 10 to 250 kDa (no. 161-0373) were from BioRad (Hercules, CA, USA). Human recombinant ADAMTS-4 (a disintegrin and metalloproteinase with thrombospondin motifs, aggrecanase-1) was from GlaxoSmithKline (Collegeville, PA, USA) 23. ECL Plus detection was from Amersham Biosciences (Buckinghamshire, UK). Polyvinylidene difluoride (PVDF) membranes, Tris-acetate mini gels (3 to 8%), LDS sample buffer, Tris-acetate SDS running buffer and transfer buffer were from Invitrogen (Carlsbad, CA, USA). Non-fat dry milk by Semper (Sundbyberg, Sweden) was from the local supermarket.

Quick-Seal centrifuge tubes (2 ml no. 344625, 12.5 ml no. 342413), tube sealer (no. 342428), tube slicer (no. 303811) were from Beckman Coulter (Palo Alto, CA, USA). The monoclonal antibody (MAb) OA-1, with or without biotinylation, recognizing the neoepitope sequence ARGSVIL (representing the N-terminus of human aggrecan cleaved between ³⁹²Glu and ³⁹³Ala in the interglobular domain) was kindly provided by Michael Pratta (GlaxoSmithKline, Collegeville, PA, USA) 19. Tetramethylbenzidine (TMB)-hydrogen peroxidase (H₂O₂) solution (no. 50-76-00) and peroxidase labeled streptavidin (no. 14-30-00) were from KPL (Gaithersburg, MD, USA). Hyaluronidase from Streptomyces hyalurolyticus (EC 4.2.2.1), chondroitinase ABC protease free (EC 4.2.2.4), keratanase (EC 3.2.1.103) and keratanase II (from Bacilus species Ks 36) were from Seikagaku (Tokyo, Japan). Keratan sulfate (KS) capture 96-well plates (no. 42.146.08) were from Biosource International (Camiro, CA, USA).

Knee SF from 26 knee healthy volunteers and 269 patients were obtained from a cross-sectional convenience cohort, where each individual, after informed consent, provided a sample at one time point only. Diagnosis was made by arthroscopy, radiography, assessment of SF and clinical examination 1. Samples were centrifuged at 3000 g and aliquots of the supernatant were stored at -80°C. All patient-related procedures were approved by the ethics review committee of the Medical Faculty of Lund University.

Diagnostic groups were healthy knee references (REF), acute inflammatory knee arthritis (AA), knee OA, and injured knee (anterior cruciate ligament rupture and/or meniscus tear) grouped as acute knee injury (AI; 0 to 12 weeks after injury) or chronic knee injury (CI; > 12 weeks after injury; Table 1). Joint changes, assessed by arthroscopy and radiography, were scored ranging from 1 to 10, where 1 represents a normal joint by arthroscopy and radiography; 2 to 5 represents an increasing extent and severity of fibrillation and clefts in the joint cartilage by arthroscopy in joints appearing normal on radiographs; and 6 to 10 represents increasing degrees of radiographic joint space narrowing consistent with OA 24. Thirty-one samples lacked arthroscopic and/or radiographic data needed for assessment of OA score.

To study injury-dependent aggrecan fragment release at different times after injury, these samples were grouped as meniscal injury alone (MEN) or cruciate ligament rupture with or without an associated meniscus injury (ACL), stratified by time after injury (0 to 4, 4 to 12, 12 to 26, 26 to 52, or > 52 weeks).

Patient samples were selected from a biobank by one of the authors (LSL) on the basis of clinical diagnosis, without reference to any previously available assay data.

From the pool of human knee OA cartilage (10 patients) proteoglycans were extracted with guanidinium hydrochloride (4 M) in the presence of proteinase inhibitors (10 mM EDTA, 100 mM EACA, 10 mM NEM, 5 mM benzamidine-HCl and 5 mM PMSF) and aggrecan was then isolated by associative-dissociative cesium chloride density gradient centrifugation in the presence of the proteinase inhibitors 25. Fraction A1D1 was collected and dialyzed against Millipore-water prior to freeze drying 18. As described, this fraction contains only large aggrecan fragments, containing the IGD, without G1-IPEN and G1-TEGE fragments 18. Human aggrecan monomers were quantified based on dry weight assuming a molecular weight of 1.5 × 10⁶ g/mol.

Full-length human recombinant ADAMTS-4 was cloned, expressed, and purified at GlaxoSmithKline (Collegeville, PA, USA) 23. ADAMTS-4 (3.1 nM) was incubated with the A1D1 fraction of human aggrecan (346 nM) for 30 hours at 37°C in 50 mM Tris-HCl, 100 mM sodium chloride (NaCl), 10 mM calcium chloride (CaCl₂), pH 7.5, achieving complete conversion of the G1-containing starting material to G1-TEGE fragments and the corresponding ARGS fragments. The digest was quenched with 25 mM EDTA and monitored for complete digestion by G1, TEGE, and ARGS western blots (data not shown). The digest was used as an ARGS standard in the aggrecan ARGS ELISA.

Quantification in SF of aggrecan fragments with the N-terminal ³⁹³ARGS was by a sandwich ELISA using an anti-KS antibody as capture and the monoclonal neoepitope antibody OA-1 for detection of specific fragments 19. After modification for use in SF, the assay was conducted as follows:

Aggrecan content was analyzed by a slightly modified competition ELISA using the mAb 1-F21 recognizing a protein sequence within or close to the KS domain 2026. The 1-F21 ELISA differed from the original 20 as follows: concentration of chondroitinase-digested A1D1 was 1.25 μg/ml when coating; all washes were 3 × 200 μl; plates were blocked after coating (1% BSA, 200 μl/well, 30 minutes at room temperature); the primary antibody 1-F21 was diluted to 1:10,000; the secondary antibody (Dakopat nr. P447) was diluted to 1:2000.

Concentration of sGAG was measured by Alcian blue precipitation modified from Björnsson 22. Samples and chondroitin sulfate standards (25 μl) were precipitated for two hours at 4°C with 0.04% w/v Alcian blue, 0.72 M guanidinium hydrochloride, 0.25% w/v Triton X-100, and 0.1% v/v H₂SO₄ (0.45 ml). The precipitates were collected after centrifugation (16,000 g, 15 minutes, 4°C), then dissolved in 4 M guanidinium hydrochloride, 33% v/v 1-propanol (0.25 ml), and transferred to 96-well micro-titer plates prior to absorbance measurement at 600 nm.

These data were available from previous studies using these samples 262728.

For molar comparison of ARGS fragments and aggrecan, conversion from microgram sGAG/ml to pmol aggrecan/ml was made assuming an average aggrecan molecular weight of 1.5 × 10⁶ g/mol and assuming that 75% of this weight was sGAG 4.

Aggrecan fragments captured in the ARGS ELISA by the anti-KS antibodies were analyzed by western blot. Following a completed ARGS ELISA, plates were washed with PBST and incubated with 4 M guanidinium hydrochloride (150 μl/well) for 30 minutes at room temperature on a plate shaker. To obtain enough material for western blot analysis, the well contents of standard wells (74 wells) and wells of SF from 152 patients (152 wells) were pooled separately and dialyzed in 10,000 kDa cut-off dialysis cassettes (Slide-A-Lyzer, Pierce, Rockford, IL, USA) against Millipore water containing protease inhibitors 18. Samples were freeze dried, dissolved in deglycosylation buffer and digested by chondroitinase, keratanase and keratanase II 18. Samples were precipitated in ice-cold acetone, and pellets were dissolved in two times concentrated sample buffer.

ADAMTS-4 digested aggrecan (used in the ELISA as standard) and a D1 fraction of pooled SF from 40 OA patients were chondroitinase, keratanase and keratanase II digested 18.

All samples were run on a 3 to 8% Tris-acetate SDS-PAGE gel, transferred to a PVDF membrane and ARGS fragments were visualized using the MAb OA-1 18.

Quantification of ARGS fragment in SF by western blot was performed as described 4 using the same mAb for detection (OA-1) and the same standard as in the ARGS ELISA.

For some patients the available volume of SF was not large enough to perform all assays, which explains the variation in numbers between assays. Of the 295 subjects, 113 had ARGS fragment values below the level of detection (i.e. < 1 pmol ARGS/ml SF). Each was assigned a value of 0.5 pmol ARGS/ml, or half the lower limit of detection. To assess differences among the study groups, either a two-tailed Mann-Whitney U rank sum test with Bonferroni correction was used after Kruskal-Wallis testing, or a Chi-squared test, as appropriate. For correlation analysis Spearman's rank order correlation (r_S) was used. P values below 0.05 were considered significant unless otherwise noted. Statistical calculations were performed using Statistical Package for the Social Sciences (SPSS, Chicago, IL, USA) for Windows version 15.0.

SF samples needed to be diluted 1:50 or more for a linear recovery at different dilutions; at dilutions below 1:50 the signal was reduced due to unknown matrix effects (results not shown). With a linear measuring range for the standards of 0.02 to 1 pmol ARGS/ml, and a minimal dilution of SF of 1:50, the lower limit of detection was then recalculated to undiluted SF 1 pmol ARGS/ml SF. Intra assay coefficient of variation (CV) was 6% (n = 10), the inter assay CV for the two groups of KS capture plates used were 12% (n = 5) and 16% (n = 23), respectively, and the total inter assay CV for the control SF sample included on all plates was 20% (n = 28; Table 2). The mean spiking recovery at dilutions 1:50 to 1:1600 was 99% (range 75 to 121%; Table 2).

Anti-ARGS western blot analysis of aggrecan fragments captured by the ELISA plates, showed that the ARGS fragments present in the standard were also captured by the anti-KS plates (Figure 1). The SF ARGS fragments captured by the plates showed the same fragment pattern as those detected in an SF D1 control sample and in the two standard samples.

The concentrations of aggrecan measured as 1-F21 reactivity, sGAG, and aggrecan fragments bearing the ARGS neoepitope are summarized in Table 3. As shown 26, there was a strong correlation between aggrecan fragment concentration measured by the 1-F21 ELISA and the concentration of sGAG (r_S = 0.82, data not shown). The ARGS concentration showed a more moderate correlation with the concentrations of sGAG (r_S = 0.69; Figure 2a) and aggrecan (r_S = 0.66; Figure 2b).

To validate the identity of the fragments responsible for the signal below the detection limit of the ARGS ELISA, 32 samples, of which 10 were below ELISA detection, were analyzed by western blot quantification 4, using the same mAb for detection, and compared with aggrecan fragment content as measured by sGAG. The molar proportion of aggrecan fragments bearing the ARGS neoepitope (i.e. ARGS/aggrecan) as measured by western blot was calculated. In the 10 samples below the detection limit of the ELISA, the median proportion of ARGS-bearing fragments out of aggrecan was 1.8% (range 1.2 to 6.4%) and in the samples above the detection limit, the median proportion was 23.8% (2.6 to 59.2%). Conversion from μg sGAG/ml to pmol aggrecan/ml is described in Material and Methods.

Concentrations of aggrecan fragments carrying the neoepitope ARGS were elevated in all groups compared with the healthy knee reference group (Figure 3a). The median levels (in pmol ARGS/ml SF) were: REF 0.5 (range 0.5 to 3.3), AA 88.5 (0.5 to 961), AI 53.9 (0.5 to 946), CI 0.5 (0.5 to 266), and OA 4.6 (0.5 to 318). Similarly, all patient groups differed from the reference (P < 0.001) regarding the proportion of samples in each diagnostic group with ARGS concentration above the lower limit of detection (1 pmol ARGS/ml) as tested by Chi-squared tests. The percentages of detectable samples were 96% (AA), 87% (AI), 46% (CI), and 62% (OA) compared with 7.7% in REF. The sensitivity of ARGS fragment concentration as a marker for joint disease was 67% with a specificity of 92% (Table 4). We found no significant influence of age or sex on the SF levels of ARGS fragments (data not shown).

Median concentrations of sGAG were elevated only in the AA (P = 0.004) and AI groups (P < 0.001) compared with the REF group (Figure 3b). Similarly, concentrations of aggrecan measured by the 1-F21 ELISA were different from REF only in AA (P = 0.002) and AI (P = 0.026; Figure 3c). The sensitivities of sGAG and aggrecan fragment concentrations as markers for disease were 40% and 32% with specificities of 92% and 91%, respectively (Table 4). We found no significant influence of sex on the SF levels of sGAG or aggrecan (data not shown). However, age correlated negatively with sGAG concentration in the AA group (r_S = -0.292) and aggrecan concentration in the AI and CI groups (r_S = -0.335 and r_S = -0.230 respectively).

After knee injury involving either a MEN alone (Figures 4a,b), or an ACL injury with or without associated MEN (Figures 4c,d), the SF levels of both sGAG and ARGS fragments were elevated within the first four weeks compared with REF (P < 0.001). Notably, the median elevations for MEN and ACL patients were more than 200-fold compared with REF for ARGS, but only two- to three-fold for sGAG. For time spans more than four weeks after injury, the sGAG levels of the MEN and ACL groups were not different from the REF group, whereas the ARGS levels continued to differ from REF (except for 26 to 52 weeks after meniscal injury). At none of the time intervals were there any differences between MEN and ACL groups for ARGS or sGAG in SF.

The proportions of aggrecan fragments in SF detected as ARGS neoepitope fragments out of all SF proteoglycan (measured by Alcian blue precipitation) were increased in all study groups compared with REF (P < 0.001; Figure 5). The proportion in the AA group was the highest, 111-fold elevated compared with REF, and in the CI group the least increased, two-fold compared with the REF group. The median proportion of ARGS in REF was 1% and ranged in the diagnostic groups from 2% (CI) to 110% (AA). The proportion of ARGS of all aggrecan as a marker for joint disease had a sensitivity of 65% and a specificity of 96% (Table 4).

The range of ARGS concentrations within each study group was substantial, with the exception of the healthy knee reference group. In part, this can be explained by the cross-sectional study design, with the grouping together of individuals with varying severity of injury and disease activity. However, it is also known that the variability of SF markers is greater than for serum and urine markers 32. Despite the considerable range observed, we note that all study groups differed significantly from the reference group regarding ARGS concentrations. Based on previous studies it is most likely that the knee injury groups are not homogenous regarding progression of OA, but are comprised of progressors and non-progressors 3334. It is plausible that heterogeneities like these also influence the ARGS concentrations, and could partly explain the variations seen in these groups.

The lower limit of linearity of the ARGS ELISA in SF was 1 pmol/ml SF, and samples below this level were assigned half that value to allow statistical analysis. All study groups had significantly lower proportions of samples below the lower limit of detection compared with the knee healthy reference group.

As a validation of the ARGS ELISA, we analyzed a subset of SF samples, purified by dissociative cesium chloride density gradient centrifugation, with quantitative western blots using the same ARGS antibody and standard as in the ELISA. The results verified that samples below the detection level of the ARGS ELISA had very low levels of ARGS.

The similarity in the Western blot analysis of loaded and captured aggrecan ARGS fragments show that the ARGS fragments present in the standard and the cesium chloride D1 preparation of an SF sample are captured by the anti-KS plate. The weaker immuno-reaction seen for SF ARGS fragments captured by the ELISA plate, compared with fragments captured from the standard, is most likely a reflection of a lower total ARGS concentration in these SFs compared with the standards.

The strategy applied in the ARGS ELISA of capturing fragments with the anti-KS antibody limits detection of ARGS fragments to those also containing part of the KS domain. Although there are known cleavage sites for proteases such as MMPs, cathepsins, and calpains between ³⁹³ARGS and the KS domain stretching from amino acid 676 to 848, these cleavage sites were all confirmed to occur by in vitro experiments 5; Sandy and Verscharen showed in SF the presence of a 100 kDa ARGS band ('Species f') estimated to stretch to amino acids 800 to 900 17. We detected small amounts of a similar band in SF purified by chromatography or by associative A1 fractioning which, when deglycosylated, migrated to 50 to 70 kDa; the intensity of the band corresponded to about 3% of the total ARGS signal (data not shown). By use of a calculation model 35, we estimate these ³⁹³ARGS fragments to stretch to amino acids 690 to 750 (data not shown). With the KS domain starting at amino acid 676, these fragments contain part of the domain necessary for capture. We can not, however, completely rule out the presence of SF ARGS fragments not containing the KS necessary for detection.

The inter assay CV for the ARGS ELISA was to a large part caused by the use of two different lot numbers of the KS capture plates supplied by Biosource. The samples were, however, analysed blinded with diagnostic groups spread evenly, so the change of lot numbers had no effect on the observed group differences in ARGS concentrations.

As reported 1224, the group differences in aggrecan content determined by Alcian blue precipitation or by ELISA with the 1-F21 antibody were small with a maximum of a two-fold increase compared with the REF. Only AA and AI were shown to have significantly elevated levels of sGAG and aggrecan compared with REF.

In contrast to the Alcian blue precipitation method and the 1-F21 ELISA, the ARGS ELISA is highly specific regarding neoepitope and presence of KS on the fragments. Even so, the ARGS neoepitope concentrations correlated with both sGAG and 1-F21 aggrecan concentrations in SF, consistent with previous findings showing that a significant portion of the sGAG and aggrecan content in human SF consisted of neoepitope fragments such as the ARGS fragments measured here, or MMP-generated ³⁶¹FFGV fragments 1836. The differences in group median values of ARGS were, however, much greater than for either sGAG or 1-F21 aggrecan. All disease groups were significantly different from the REF, with as much as 177-fold increased levels of ARGS in the AA group. With specificities of 91 to 92%, the concentration of ARGS neoepitope fragments had a sensitivity of 67% in differentiating diseased from healthy patients, compared with sGAG or 1-F21 aggrecan, which had lower sensitivities of 40% and 32%, respectively. Quantification in SF of ARGS fragments generated by aggrecanases by a neoepitope-specific ELISA is clearly a more powerful tool to distinguish diseased and injured joints from healthy than quantification of aggrecan fragments either by 1-F21 ELISA or by measuring sGAG concentrations.

Acknowledging that there are uncertainties in our assumptions of molecular weight and degree of glycosylation of the average aggrecan fragment in SF, and of the molecular weight of the standard, uncertainties that make ARGS proportions of aggrecan greater than 100% possible, the diagnostic groups nevertheless showed large differences in the proportion of SF aggrecan fragments generated by aggrecanase IGD activity. In the two groups most strongly associated with high joint disease activity, acute inflammatory arthritis and acute knee injury, a majority of the aggrecan fragments were indeed shown to be the result of aggrecanase IGD activity, whereas the other groups had much lower proportions. These results corroborate those previously obtained by western blots 4.

Based solely on data available in this paper, the elevated SF levels of ARGS in disease, particularly in the acute inflammatory arthritis and acute injury samples, could be explained by enhanced aggrecanase activity against aggrecan resident in the joint cartilage matrix, or against newly synthesized and secreted aggrecan 37. ADAMTS-5 (aggrecanase-2) was shown to co-localize with hyaluronan surrounding chondrocytes in both normal and osteoarthritic cartilage 38. However, if enhanced synthesis of aggrecan in combination with aggrecanase activity were to explain the enhanced SF levels of ARGS, an equal increase in the SF levels of G3 was to be expected. This is not the case; we have in quantitative western blot analysis of 30 of these samples seen no significant difference in SF levels of G3 of any of the diagnostic groups compared with healthy knee references 4. We therefore suggest that an increased aggrecanase activity against the IGD domain of resident aggrecan best explains the enhanced SF levels of ARGS seen in these diagnostic groups.

SF is more proximate to the location of joint cartilage and aggrecan degradation than serum or urine, and may therefore better reflect local pathologic processes in the joint being studied. The observed group differences are thus likely to reflect differences in local knee joint pathology. The fragments observed in SF originate in a major part from the joint cartilage, while minor proportions may be released from menisci and ligaments 3940.