Aborted human embryos were obtained according to the regulations of the Ethics Committee of the Medical Faculty of the University of Göttingen, Germany. The embryos were classified as follows: three embryos of gestational week (gw) 8, three embryos of gw 10, and three embryos of gw 12. The ages were determined from histological data according to Carnegie stages 28. No malformations or anomalies were observed in these specimens.

Adult human articular cartilage from the knee joint was obtained from 12 patients (ages 55–75) with OA who were undergoing total knee replacement operations. The patients met the American College of Rheumatology classification criteria for OA of the knee 29. All patients gave their written informed consent according to the Ethics Regulations of the Medical Faculty of the Georg-August-University Göttingen. Four healthy control cartilage samples from accident victims (ages 31–50) were also investigated.

The abortion material and the cartilage specimens were transported to the laboratory in histidine-tryptophane-ketoglutarate solution at 4°C to ensure good preservation of the tissues 30. The samples were fixed in 4% paraformaldehyde in phosphate-buffered saline (PBS), pH 7.4, at 4°C overnight. Bone-containing samples were decalcified with buffered EDTA for 14 days. For light microscopy, specimens were dehydrated, embedded in paraffin wax, and cut with a Reichert's microtome. For the staging of the embryos, every fifth section was stained with hematoxylin and eosin. Longitudinal sections of the cartilage specimens stained with Alcian blue were classified as stage IV according to the OA grades (I – IV) proposed by Collins and McElligott 31 in the case of the 12 patients and classified as age-dependent healthy in the case of the control cartilage samples. None of the cartilage specimens showed any signs of rheumatoid involvement or exhibited any osteophytes. From the 12 patients, cartilage samples from the deep cartilage zones near the tidemark were obtained from two different regions of the OA knee joints. One sample, with a macroscopically normal appearance, was taken from the lateral aspect of a condyle. The other one was taken from the area adjacent to the main defect at a maximum of 0.5 cm away. All cartilage specimens were also processed for ultrastructural analysis. Samples (1 mm³) were embedded in LR-Gold^® (London Resin Company, Berkshire, England) according to standard procedures, and ultra-thin sections were cut with a Reichert's ultramicrotome and collected on nickel grids coated with Formvar^® (Serva, Heidelberg, Germany).

The anti-COMP antibody is a polyclonal rabbit-anti-bovine antibody that has been affinity-purified 1. Affinity-purified sheep-anti-digoxigenin (DIG) antibodies were purchased from Quartett (Berlin, Germany), an anti-DIG peroxidase labeled antibody from Dakopats (Hamburg, Germany), and the secondary antibodies from Medac (Hamburg, Germany).

Healthy cartilage and OA cartilage samples from the area adjacent to the main defect were pulverized. Proteins were extracted using 5 M guanidine hydrochloride and protease inhibitors NEM (N-ethylmaleimide), EDTA, benzamidine hydrochloride, and amino caproic acid, precipitated in ethanol, washed in PBS, precipitated again, and finally dissolved in PBS containing 0.4% SDS. All experiments were carried out under reducing and denaturing conditions. Protein separation was performed applying SDS-PAGE and using systems containing 6% acrylamide in stacking gels and 12% in the separation gel. Tris-glycine was applied as electrophoresis buffer, and separation was carried out at 100–120 V.

After the electrophoresis, the proteins were blotted onto nitrocellulose membranes. Transfer was controlled by Ponceau S staining. Thereafter, membranes were washed until no color was left and then blocked overnight in PBS + 10% (w/v) milk powder at room temperature. Immunoreactions were performed applying the anti-COMP antibody for 2 hours, diluted 1:100 in PBS. The secondary goat-anti-rabbit antibody coupled to alkaline phosphatase was diluted 1:500 and incubated for 1 hour at room temperature. Three 5-minute washes with PBS were carried out between all incubation steps. Visualization was achieved using NBT/BCIP (nitrobluetetrazoline chloride/5-bromo-4-chloro-3-indolyl toluidine) coloring agent (Roche, Heidelberg, Germany).

Immunoperoxidase staining was performed on paraffin-embedded tissue sections as follows: The tissues were deparaffinized, rehydrated, and rinsed for 10 minutes in PBS. Endogenous peroxidase was inhibited by a 45-minute treatment with a solution of methanol and 3% H₂0₂ in the dark. Each of the reactions was followed by rinsing for 10 minutes in PBS. The sections were pre-treated for 5 minutes with 10 μg/ml protease XXIV (Sigma, Deisenhofen, Germany). The anti-COMP antibody was applied at a dilution of 1:100 in PBS for 1 hour at room temperature. A standard peroxidase-anti-peroxidase procedure followed, applying a peroxidase-coupled goat-anti-rabbit antibody (Dako, Hamburg, Germany) at a dilution of 1:150 in PBS for 1 hour at room temperature. The color reaction was carried out with DAB (diaminobenzidine) substrate (Sigma).

As negative controls, each immunoreaction was accompanied by a reaction omitting the primary antibodies and applying rabbit serum diluted 1:100 in PBS instead. All controls proved to be negative.

As secondary antibody, an anti-rabbit immunoglobulin G (IgG) (Medac) was labeled with gold particles according to standard procedures. Ultrathin tissue sections were incubated with the anti-COMP antibodies diluted 1:100 in PBS for 16 hours at room temperature. The secondary gold-coupled antibodies, diluted 1:300 in PBS, were applied for 20 minutes at room temperature. Staining with uranyl acetate followed, and reactions were examined with the help of a Zeiss EM Leo 906E electron microscope (Carl Zeiss, Jena, Germany).

The grids were incubated with pure gold solution in order to exclude unspecific binding of free colloidal gold. Furthermore, the reactions were performed with gold-coupled goat-anti- rabbit IgG, omitting the primary antibody to exclude non-specific IgG binding.

RNA was isolated as described below and reverse-transcribed into COMP-specific cDNA. Polymerase chain reaction (PCR) was performed with primers specific for COMP (forward AGGGAGATCGTG CAGACAA and reverse AGCTGGAGCTGTCCTGGTAG) to generate a 154 bp product. They were designed with the help of the primer³shareware 32. Corresponding primers with the appropriate SP6/T7 promoter sequences were applied. In vitro transcription of non-radioactive sense and antisense RNAs with a DIG labeling kit (Boehringer DIG-RNA labeling kit, Boehringer, Mannheim, Germany) was performed applying SP6- and T7-polymerases (Gibco/BRL, Heidelberg, Germany). After extraction of the probes with phenol-chloroform, these were precipitated with absolute ethanol and the pellet was dissolved in DEPC-H₂O (diethyl-pyrocarbonate).

For light microscopic investigations, paraffin sections were deparaffinized, rehydrated, and pre-treated with proteinase K. The probe concentration was 100 ng of DIG-labeled antisense probes in 100 μl hybridization solution (50% formamide, 5 × SSC, 1 μg/μl yeast-RNA, 10 ng/μl probe) for each section. Hybridization was carried out for 18 hours at 45°C. Posthybridization treatment included a washing procedure with 2 × SSC (3 × 5 minutes, at 50°C), 1 × SSC (1 × 5 minutes, at 60°C), 0.1 × SSC (1 × 15 minutes, at 60°C) and 0.05 × SSC (1 × 15 minutes, at 60°C). Afterward, the incubation with the anti-DIG peroxidase-labeled antibody diluted 1:300 in PBS was started. Finally, color reactions were started with AEC (3-amino-9-ethylcarbazol) substrate. For electron microscopy, nickel grids were incubated for 19 hours at 50°C with the same hybridization solution as described above. The probe concentration was 100 ng of DIG-labeled antisense probes in 20 μl hybridization solution per grid. Rinsing steps were the same as described above. Afterward, sections were incubated with a gold-coupled anti-DIG antibody in PBS (diluted 1:60) for 1 hour at room temperature. The specimens were rinsed with PBS, contrasted, and analyzed with the Zeiss EM Leo 906E.

Each of the hybridizations was accompanied by one with an equivalently labeled amount of sense probe. Furthermore, hybridizations were performed without any RNA probes. Additionally, for the ultrastructural controls, tissue sections were treated with pure gold solution or the coupled anti-DIG antibody alone.

For in situ hybridization at the ultrastructural level, randomly chosen micrographs of cartilage tissue with a normal appearance which were taken from the lateral aspects of a condyle and tissue samples taken from the area adjacent to the main defect from OA cartilage (n = 10) were pooled and counted for gold particle contents. Mean values of the numbers of gold particles per cell were analyzed in the area of 5,000 nm² in 10 cells from each patient. Significant differences in the number of gold particles were noted for p values (p ≤ 0.01), using the Wilcoxon-Mann-Whitney test for unpaired samples.

Pieces (2 mm thick) of OA cartilage tissue taken from the area adjacent to the main defect and pieces of tissue with a macroscopically normal appearance of the lateral aspect of a condyle from each of the 12 patients were minced, and RNA was isolated according to a protocol combining Trizol^® and RNeasy kit (RNeasy^® Mini Kit, Qiagen, Hilden, Germany), following the manufacturer's instructions, and then treated with DNAfree^® (Ambion, Austin, TX, USA). The quality of the RNA was tested with an Agilent 2100 Bioanalyser RNA chip (Agilent, Böblingen, Germany). The RNA was reverse-transcribed with the help of the Advantage^® RT-for-PCR kit (BD Biosciences, San Diego, CA, USA) by applying Moloney Murine Leukemia Virus reverse transcriptase and oligo-(dT)₁₈-primer.

PCR conditions were optimized by applying the gradient function of the DNA engine Opticon™ 2 (Bio-rad, München, Germany) for HPRT-1 (NM_000194) as housekeeping gene and for COMP. The PCR was performed in a total volume of 50 μl with 150 ng cDNA, 5 μl 10× reaction buffer, dNTP 10 μmol each, 20 pmol of each primer, and 2.5 U HotStarTaq^® DNA polymerase (Qiagen) with the DNA engine Opticon™ 2. After an initial activation step of 15 minutes at 95°C, further steps were as follows: 35 cycles of denaturing 30 seconds at 94°C, annealing 30 seconds at 61°C, elongation for 30 seconds at 72°C, and (lastly) extension of 10 minutes at 72°C. Ten microlitres of each sample were loaded onto a 1.5% agarose gel and were visualized by ethidium bromide after electrophoresis.

To optimize the real-time reverse transcription (RT)-PCR conditions for quantification, the optimal MgCl₂ concentration was determined. Twelve point five microlitres of 2xQuantiTect™ SYBR^® Green PCR Master Mix (Qiagen), 20 pmol of each primer, and 250 ng of cDNA were added to a final volume of 25 μl. Cycling was performed with the protocol described above. Data acquisition was carried out after each extension step, and a melting curve was performed in 0.1°C steps from 50°–95°C. Real-time RT-PCR efficiencies were calculated from the given slopes in Opticon™ 2 Monitor software. Real-time RT-PCR efficiency rates were high (values of 2.00). Experiments were performed three times in triplicate, the inter-test variation was ≤ 2%, and the intra-test variation ≤ 1%.

During human embryonic development from gw 8 to gw 12, basement membrane zones of the developing skin stained positive for COMP whereas the mesenchyme remained unstained (Figure 1a). In limb buds, staining for COMP was found in the basement membrane zone of the apical ectodermal ridge (AER), and the condensed mesenchyme was not stained (Figure 1b). During further development of the long bones at gw 10, staining for COMP was seen throughout the extracellular matrix of the cartilage (Figure 1c). Later, at gw 12, staining for COMP became restricted to the margins of the developing epiphysis (Figure 1d), the developing joint surface (Figure 1e), and the diaphysis of long bones. COMP was seen mostly pericellularly around hypertrophic chondrocytes along the edges of the shaft of the diaphysis (Figure 1f).

The anti-COMP antibody 1 cross-reacted with human COMP from healthy (Figure 2, lane 3) and OA cartilage tissue extracts taken from the area adjacent to the main defect (Figure 2, lane 2). The 105 kDa band for a monomer was seen in both extracts, whereas a second band was found only in the OA cartilage sample and might represent a covalently bound binding partner of COMP (for example, one of the matrilins). This phenomenon has been observed with COMP in other instances. The smear in the blot of healthy cartilage tissue (Figure 2, lane 3) probably results from the high aggrecan content, which is missing in OA tissue. This is why this smear is not found in Figure 2, lane 2, where aggrecan is lost (M. Paulsson, personal communication). With the help of light microscopic immunohistochemistry, COMP was localized in healthy knee joint cartilage tissues in the pericellular, territorial, and interterritorial matrix compartments. This was seen in the superficial and middle zone. In contrast, in the deep zone near the tidemark, COMP was found only in the pericellular space (Figure 3a and inset). In OA cartilage, in the area adjacent to the main defect, pronounced staining for COMP was seen (Figure 3b), especially in the pericellular matrix of cell clusters (Figure 3b, inset). With the help of light microscopic in situ hybridization, the mRNA for COMP was detected in the cytoplasm of chondrocytes of the superficial and middle zones of healthy cartilage tissue (data not shown) and also in chondrocytes mainly found in clusters in the area adjacent to the main defect in OA cartilage (Figure 3c).

To elucidate which components in the differing matrix compartments stain for COMP, an ultrastructural analysis was performed. In healthy cartilage specimens, COMP was associated mainly with the fine fibrillar structures in the pericellular space (Figure 4a). In OA cartilage taken from the area adjacent to the main defect from patients in the late stages of OA, an increase in staining intensity was found in the pericellular space (Figure 4b). In healthy cartilage, COMP staining was also found in the territorial and interterritorial matrix (Figure 4c), whereas in OA cartilage specimens, staining for COMP was seen mainly on fibers but also next to them (Figure 4c, inset).

From earlier investigations on the pathogenesis of OA, we are aware of two different cell types found in the late stages of the disease 3334. Type 1 cells are the diseased chondrocytes found in regions with a macroscopically normal appearance of the OA cartilage, and type 2 cells are elongated, fibroblast-like cells found mainly in the area adjacent to the main defect. A small number of type 2 cells can also be found in the regions with a macroscopically normal appearance in OA cartilage and vice versa: a few type 1 cells are also present in the area adjacent to the main defect. To elucidate which cells, type 1 or type 2, produce COMP mRNA, we performed in situ hybridization at the electron microscopic level. In cartilage tissue with a normal appearance from the lateral aspects of a condyle of the OA patients, COMP mRNA was detected in type 2 cells (Figure 5a and inset) and less staining was seen in type 1 cells (Figure 5b,c). In contrast, in tissue samples from the area adjacent to the main defect of OA cartilage of late stages of the disease, strong staining for COMP mRNA was detected in the cytoplasm of type 2 cells (Figure 6a) and type 1 cells (Figure 6b,c).

The number of gold particles detected in the samples with a macroscopically normal appearance from OA tissue revealed staining intensities of approximately 42 (SEM = 3.4) in type 1 cells and 66 (SEM = 4.1) in type 2 cells. This represents a significant difference (p ≤ 0.01). In contrast, in both cell types found in the areas adjacent to the main defect of OA tissue, approximately 320 gold particles (SEM = 13.4) were detected (Figure 7). This represents a statistically significant (p ≤ 0.01), approximately 83% difference in staining intensity for the cells taken from the two areas.

To validate the semi-quantitative results from the ultrastructural in situ hybridization, we performed quantitative real-time RT-PCR. The mean threshold cycle value for COMP cDNA detected in tissue samples from patients with late stages of OA taken from the area adjacent to the main defect is 16.2, representing a relative ratio of 8.28, and the value detected in samples of cartilage tissue with a macroscopically normal appearance is approximately 27.5 (Figure 8a), representing a ratio of 0.16. The relative ratios were calculated according to the algorithm of Pfaffl. The relative ratio for COMP in normal cartilage tissue is approximately 98% lower when compared with OA tissue. The calibrator curve obtained by the correlation of the threshold cycle values with the dilution series of the housekeeping gene exhibited a low (≤ 1%) intra-test variation (Figure 8b,c). The validity of the PCR results was confirmed by sequencing and by the melting curves performed for each PCR (data not shown).