Cartilage oligomeric matrix protein induction of chronic arthritis in mice




To develop a new mouse model for arthritis using cartilage oligomeric matrix protein (COMP) and to study the role of major histocompatibility complex (MHC) and Ncf1 genes in COMP-induced arthritis (COMPIA).


Native (pentameric) and denatured (monomeric) COMP purified from a rat chondrosarcoma was injected into mice with Freund's adjuvant to induce arthritis. C3H.NB, C3H.Q, B10.P, B10.Q, (B10.Q × DBA/1)F1, (BALB/c × B10.Q)F1, Ncf1 mutated, H-2Aq, H-2Ap, and human DR4+-transgenic mice were used. Anti-COMP antibodies and COMP levels in the immune sera were analyzed, and passive transfer of arthritis with purified immune sera was tested.


Immunization with rat COMP induced a severe, chronic, relapsing arthritis, with a female preponderance, in the mice. The disease developed in C3H.NB mice, but not in B10.P mice, although they share the same MHC haplotype. Both H-2q and H-2p MHC haplotypes allowed the initiation of COMPIA. Using H-2Aq–transgenic and H-2Ap–transgenic mice, we demonstrated a role of both the Aq and Ep class II molecules in this model. Interestingly, the introduction of a mutation in the Ncf1 gene, which is responsible for the reduced oxidative burst phenotype, into the COMPIA-resistant B10.Q mouse strain rendered them highly susceptible to arthritis. In addition, the transfer of anti-COMP serum was found to induce arthritis in naive mice. Mice transgenic for the rheumatoid arthritis (RA)–associated DR4 molecule were found to be highly susceptible to COMPIA.


Using rat COMP, we have developed a new and unique mouse model of chronic arthritis that resembles RA. This model will be useful as an appropriate and alternative model for studying the pathogenesis of RA.

Development of rheumatoid arthritis (RA) involves an early erosive inflammatory attack on peripheral cartilaginous joints concomitant with an autoimmune response to normal and modified joint and cartilaginous proteins. Both T cell and B cell autoimmune responses to the major cartilage protein type II collagen (CII) occur in a subset of RA patients, predominantly associated with the expression of HLA class II molecules with a unique peptide-binding site, the so-called shared epitope (1, 2). There are obvious parallels with autoimmune responses in the mouse, since immunization with CII produces collagen-induced arthritis (CIA), a disease associated with the murine class II major histocompatibility complex (MHC) molecule Aq (3). The Aq molecule displays a peptide-binding site that is quite similar to the shared epitope. Consequently, transgenic expression of human class II molecules DR4 and DR1, which have the shared epitope, allows the induction of CIA.

Importantly, all shared epitope molecules, including murine Aq and human DR4*0401 and DR1*0101, present a glycosylated peptide of CII (amino acids 260–271). Although the immune response to CII is pathogenic in mice, this has not been determined in humans. In fact, immune responses to CII may also develop in arthritic mice without CII immunization (4, 5). It is likely that the immune response to CII, and possibly its pathologic consequences, is genetically controlled. However, arthritis can also be induced after immunization with several other proteins, some of which originate from cartilage. These responses may well be associated with other variants of RA. One such molecule is cartilage oligomeric matrix protein (COMP).

COMP is a structural cartilage protein composed of 5 identical subunits, with disulfide bonds near the N- terminal, and with a globular domain at the C-terminal (6–8). Mouse COMP complementary DNA has been cloned and sequenced, and its distribution pattern during early skeletal development and tissue expression levels have been studied (9, 10). COMP is synthesized by chondrocytes and is localized interterritorially. In young tissue and in the growth plate, it is found preferentially in the pericellular (territorial) matrix of the chondrocytes, while in adult cartilage, the protein is found primarily in the interterritorial matrix at some distance from the chondrocyte (11). It has a similar distribution as CII, being found in nasal, tracheal, and meniscal cartilage, and apart from the vitreous body of the eye (12), it is found most prominently in articular cartilage (8). COMP may have critical functions in cartilage, since a mutation in the zinc-binding domain of the human COMP gene is phenotypically manifested as pseudochondroplasia and multiple epiphyseal dysplasia (13). However, COMP-deficient mice showed normal skeletal development (14).

Recent studies have suggested an important role of COMP in chondrocyte proliferation mediated by granulin-epithelin precursor, an autocrine growth factor, in the cartilage (15). An interesting feature of COMP is its release into the blood from cartilage during the erosion of tissue in arthritis pathogenesis (16). COMP is also released in rats developing chronic arthritis after induction with pristane, a synthetic low molecular weight adjuvant, and the levels in serum correlate strongly with the occurrence of erosive arthritis (17). Thus, it is likely that COMP is extensively exposed to the immune system, leading to tolerance and preventing an autoimmune response, with the subsequent development of arthritis. This tolerance can be broken, since immunization with rat COMP leads to induction of both an autoimmune response and the development of arthritis in rats (18). In contrast to CIA, the arthritis developed with a late onset with a self-limiting disease course in rats.

To further investigate the mechanisms of COMP-induced pathogenic autoimmunity, we developed a mouse model of COMP-induced arthritis (COMPIA). As it turns out, this model is of interest for providing not only a better analytic tool with which to study the pathogenic mechanisms of RA, but also a better understanding of the factors involved in the development of a chronic relapsing disease course. Using class II MHC–transgenic mice, we also identified both mouse and human class II MHC molecules that are involved in this type of arthritis. Finally, we also identified the importance of the Ncf1 polymorphism, one of the major genetic factors that control arthritis severity and chronicity, in breaking tolerance in the B10.Q mouse.



Mice were maintained at the animal facility of the Division of Medical Inflammation Research at Lund University, where they were kept in a climate-controlled environment, with 12-hour cycles of light and dark. Mice were housed in polystyrene cages containing wood shavings and were fed standard rodent chow and water ad libitum. All experiments were performed on age-matched mice between the ages of 8 and 16 weeks.

Founders of the B10.Q/Rhd mouse strain were obtained from Prof. Jan Klein (Professor Emeritus, Tübingen University, Tübingen, Germany), and DR4/huCD4-transgenic mice with H-2 deficiency were from Lars Fugger (Department of Neuropathology, University of Oxford, Oxford, UK). Breeding pairs of BALB/c mice were obtained from The Jackson Laboratory (Bar Harbor, ME). B10.Q mice with a mutated Ncf1 gene (B10.Q.Ncf1*/*) have been described previously (5). B10P/Rhd, DBA/1/Rhd, and C3H.NB/Rhd mice were bred at the animal facility of the Division of Medical Inflammation Research at Lund University. The C3H.Q/Rhd strain was established in our laboratory by inserting the H-2q haplotype from an older C3H.Q strain into mice of the C3H.NB background (19). Rhd indicates that these classic inbred mouse strains have been maintained in our laboratory for more than 2 decades. Intercross mice (B10.Q × DBA/1)F1 or (BALB/c × B10.Q)F1 were also used in this study. To analyze the role of class II MHC genes, we used B10.QGBP mice with a transgenic H-2Aβp genomic construct, leading to the expression of Ap molecules on B10.Q (H-2q) mice (19). Similarly, we also used B10.PGBQ mice with a transgenic H-2Aβq genomic construct, leading to the expression of Aq molecules on B10.P (H-2p) mice (19). To increase arthritis susceptibility, the experiments were performed in F1 intercross mice with C3H.Q (H-2q) or C3H.NB (H-2p). HLA–DR4 mice, being triple-transgenic for HLA–DRA*0101, HLA–DRB*0401, and human CD4 (20), but lacking murine class II MHC expression (21), were backcrossed to the B10 background for 10 generations and then intercrossed and selected for the H-2–/– genotype. DR4 mice were also crossed with B10.Q.Ncf1*/* mice and subsequently intercrossed to generate HLA–DR4.Ncf1*/* mice.

The Lund-Malmö laboratory animal ethics committee approved the animal experiments described in this article.

COMP purification.

Purification of COMP from the Swarm rat chondrosarcoma line was performed as previously described (22), with some modifications. Rat chondrosarcoma (50 gm) was dissected and homogenized on ice at moderate speed in 5 volumes/weight of 150 mM NaCl, 50 mM Tris, pH 7.4 (TBS), containing 10 mMN-ethylmaleimide (NEM), 5 mM benzamidine, and 1 mM phenylmethylsulfonyl fluoride (PMSF) using a Heidolph homogenizer Diax 900 with the tool 30 DG (Sigma-Aldrich, Stockholm, Sweden). The homogenate was stirred for 1 hour at 4°C and then centrifuged for 20 minutes at 38,000g using a Sorvall SS-34 rotor (Kendro Laboratory Products, Upplands Väsby, Sweden). The pellet was washed 3 additional times but with 15 minutes of stirring. Next, the pellet was extracted with TBS containing 10 mM EDTA, 1 mM PMSF, and 5 mM NEM and stirred for 30 minutes at 4°C. The extraction was repeated once more, and both of the supernatants were pooled. The pooled fractions were then applied to a 65-ml DEAE-Sepharose fast-flow media column (GE Healthcare, Uppsala, Sweden) equilibrated with 150 mM NaCl, 50 mM Tris, 2 mM EDTA, and 5 mM benzamidine, pH 7.4.

After washing with equilibration buffer, COMP was eluted as the first peak, with a 6.7–column volumes linear 0–0.5M NaCl gradient in the buffer. COMP-containing fractions were evaluated by sodium dodecyl sulfate–polyacrylamide gel electrophoresis (SDS-PAGE) under reducing and nonreducing conditions (with and without β-mercaptoethanol) using 4–20% ready-made gradient minigels (Bio-Rad, Sundbyberg, Sweden). Pooled fractions were then ultrafiltrated in Amicon Ultra-15 Centrifugal Filter nominal molecular weight limit 30 kd (Millipore, Solna, Sweden) for buffer exchange to 150 mM NaCl, 10 mM Tris, and 2 mM EDTA, pH 7.4, and the same buffer was used to equilibrate a 5-ml HiTrap heparin–Sepharose high-performance column and a 5-ml HiTrap Q Sepharose fast-flow column (GE Healthcare), which were connected in a series. COMP was further negatively purified through the heparin column (trapping contaminants such as other thrombospondins and proteins with a heparin site[s]), followed by adsorption for concentration onto the Q column. After the heparin column was disconnected, COMP was pulse-eluted with 0.5M NaCl in the buffer. Further purification of proteins with extra purity >95% (polishing) was achieved using Superose 6 and Superdex 200 columns (GE Healthcare) in a series, with a bead volume of 75 ml each, and then equilibrated and eluted isocratically with 7 mM NaCl, 200 mM Tris, and 100 mM EDTA, pH 7.4.

Peaks were analyzed by SDS-PAGE (with and without SH) as above, and fractions containing denatured (monomeric) and native COMP were pooled separately and concentrated using Amicon Ultra-15 ultrafiltration filters (Millipore). Concentrations were colorimetrically determined with the Lowry-based RC DC Protein Assay (Bio-Rad). Purity was determined with silver staining of SDS-PAGE gels (standard protocol with silver nitrate), with a sensitivity of 5 ng of protein, and was estimated to be >95% pure for both of the COMP forms.

Induction and evaluation of arthritis.

Mice were immunized intradermally at the base of the tail with 100 μg of COMP emulsified in 100 μl of Freund's complete adjuvant (Difco, Detroit, MI). After 35 days, a booster injection of 50 μg of protein in 50 μl of Freund's incomplete adjuvant (Difco) was given intradermally at the base of the tail. The mice were monitored 3 times each week for the development of arthritis for up to 160 days. To induce arthritis with anti-COMP antibodies, IgG from the sera of mice that had been immunized with COMP were pooled, purified using a GammaBind Plus Sepharose column (GE Healthcare), sterile-filtered, and injected intravenously into male (BALB/c × B10.Q)F1 mice (2 mg per mouse). The control group received equal amounts of purified polyclonal IgG from strain- and age-matched normal mice. On day 5, lipopolysaccharide from Escherichia coli O55:B5 (25 μg per mouse) was injected intraperitoneally into all mice to enhance the incidence and severity of the passively transferred arthritis.

The arthritis score was determined through assessment of the number and extent of joint involvement, as described in detail elsewhere (22), using a scale of 0–15 points per paw. Briefly, each paw was divided into 3 parts: the ankle, midpaw with knuckles, and toe joints. The maximum score for each part was 5 points. Because of migration of edema, the midpaw could be scored as 5 even if the knuckles were unaffected; otherwise, each affected knuckle and toe joint was given 1 point until the maximum score was reached. A swollen ankle was always scored as 5 points, regardless of the degree of swelling.

Antibody measurements.

Levels of antibodies specific for COMP and for CII were measured essentially as described previously (23), using mouse sera obtained via retroorbital plexus bleeding. Briefly, Immulon 2HB microtiter plates (Thermo Electron Corporation, Milford, MA) were coated with purified proteins in phosphate buffered saline (PBS) at 10 μg/ml (50 μl/well) overnight at 4°C in a moisturized chamber. Then, 1% bovine serum albumin (BSA) in PBS was used to block unbound sites. After washing in 0.1M Tris, 150 mM NaCl, with 0.1% Tween buffer, the sera were prediluted in washing buffer 10 or 100 times in duplicate in the first row and then titrated down 4 rows with a 10-fold dilution per step. Each plate contained blank, positive (standard), and negative (pooled normal sera from relevant strains) controls. After 1 hour at room temperature in a moisturized chamber and following washing, goat anti-mouse IgG (heavy and light chains) peroxidase-conjugated antibodies (Jackson ImmunoResearch, West Grove, PA) were applied at a concentration of 1–2,000 in washing buffer.

After incubation and washing, plates were developed using ABTS (Roche Diagnostics, Bromma, Sweden) and read at 405 nm. Enzyme-linked immunosorbent assay (ELISA) values are expressed as arbitrary units (AU), where pooled sera from all individual mice in the strain-screening performed on day 35 (native COMP) were used as the standard. The absorbance value of pooled sera was set as value of 1 and was used as a standard for all other individual titers (1 AU/μl). The standard and pooled anti–denatured COMP sera were cross-tested and showed equivalent binding to both native and denatured COMP in absolute titer values (standardized time and preset absorbance level). The mean values reported are the geometric mean of the log-transformed data.

Similarly, COMP–IgG immune complexes were measured using a sandwich ELISA and by capturing COMP with polyclonal rabbit anti-COMP sera. Premixed COMP and pooled COMP-immunized mouse sera were used as the standards. Peroxidase-conjugated goat anti-mouse IgG (heavy and light chains) was used for detection, similar to the anti-COMP ELISA described above.

Serum COMP measurements.

COMP protein in sera was measured by using polyclonal rabbit anti-COMP sera. Immulon 2HB microtiter plates were coated with COMP as described above. In a separate low protein–binding plate (Sero-well; Bibby Sterilin, Staffordshire, UK), a mixture of prediluted mouse sera and a fixed amount of rabbit anti-COMP were added. Serial dilutions of purified native rat COMP were included as positive standards for calculations. After incubation for 1 hour at room temperature, the mixed sera and standard were applied to the COMP-coated plate. Plates were incubated and developed using alkaline phosphatase–conjugated swine anti-rabbit antibody (Dako, Glostrup, Denmark) and p-nitrophenyl phosphate disodium (Sigma-Aldrich) dissolved in diethanolamine buffer, thus measuring the added rabbit antibodies that were initially unbound to COMP that was present in the sera (16).

Flow cytometry.

Cells to be stained were collected on day 159 postimmunization from the peripheral blood of C3H.NB and C3H.Q mice, as well as from naive controls. Peripheral blood samples were hemolyzed, washed, and then stained with the following fluorescein isothiocyanate, phycoerythrin, or biotinylated conjugates for 20 minutes at 4°C: anti-CD19 (B cell marker; clone 1D3), anti-CD4 (T cell subset marker; clone H129.19), anti-CD8a (T cell subset marker; clone 53-6.7), anti–Ly-6G/C (peripheral neutrophil/monocyte marker; clone RB6-8C5) (all from BD PharMingen, Franklin Lakes, NJ), and anti–I-A (class II MHC marker; clone Y3P) (from our hybridoma collection). Streptavidin–allophycocyanin was used as a secondary reagent. Propidium iodide was added before acquisition to sort out the dead cells. Data from 50,000 cells were acquired with a FACScan instrument (BD Biosciences, Mountain View, CA), with gates set to include total viable leukocytes, which were further analyzed using CellQuest software (BD Biosciences). Values are expressed as the frequency (number in the cell population/total viable leukocytes among 50,000 cells).

Histologic analysis.

For histologic assessments, the paws were dissected, decalcified, dehydrated, and then paraffin-embedded, as described previously (24). Sections were then stained with hematoxylin and eosin.

Statistical analysis.

All ELISA data were first log-transformed and then used for calculation of the mean and for calculation of the independent-samples t-test, except for the SD. Mean ELISA values are expressed as the antilog of the transformed mean data (geometric mean). The Mann-Whitney U test was used to analyze the arthritis scores, and one-way analysis of variance was used to examine the arthritis onset data. Pearson's chi-square with Yates' correction was calculated for the incidence and fluorescence-activated cell sorter data, or if the number of animals in each group was <5, Fisher's exact test was used.


Induction of COMPIA in mouse strains.

To test the possibility that COMP immunization induces arthritis in mice, we selected the C3H.NB and C3H.Q mouse strains, since mice with the C3H background are highly susceptible to autoimmune disease (25). Although both mouse strains developed arthritis after immunization with native COMP as well as with denatured COMP (Figures 1A and B and Table 1), C3H.NB mice had more severe disease, as well as a relapsing disease course. The arthritis onset was sudden and was characterized by the appearance of erythema and swelling of both the front and rear paws. The arthritis typically began distally in a toe joint and simultaneously, or within a few days, flared in the ankle/wrist joints. In severe cases, it spread to the metatarsal/metacarpal region and to the knuckles and toes. Although both mouse strains developed arthritis, the C3H.NB strain had a more severe arthritis and a higher incidence than did the C3H.Q strain. After the first period of acute arthritis, at 5–6 weeks after immunization, the disease was transformed into a chronic relapsing pattern in the C3H.NB strain (Figure 1A).

Figure 1.

Cartilage oligomeric matrix protein (COMP)–induced arthritis in mice. The disease course in female C3H.NB and C3H.Q mice immunized with A, native COMP and B, denatured COMP shows a chronic relapsing arthritis in the C3H.NB strain, which was scored until termination of the study. Only 1 female mouse in the C3H.Q group developed arthritis (A). Each data point represents the mean and SD total arthritis score in arthritic mice at each assessment.

Table 1. Induction of arthritis with native COMP, denatured COMP, or anti-COMP antibodies*
Mouse strain, sexTreatmentIncidence of arthritisDay of onset, mean ± SDRange of disease daysMaximum arthritis score, mean ± SD
  • *

    Mice were treated by active immunization with either native (N) or denatured (D) cartilage oligomeric matrix protein (COMP), or they were treated by passive transfer with either purified polyclonal IgG from mice immunized with native COMP (A) or normal purified polyclonal IgG (G), at 2 mg intravenously per mouse. Maximum arthritis score represents the mean of the maximum scores for each mouse per group. HLA–DR4 mice were observed for arthritis development for 10 weeks postimmunization. Ncf1*/* and Ncf1+/* denote mice that were homozygous and heterozygous, respectively, for the Ncf1 gene.

  • Two mice in this group died early in the experiment and were excluded from the analysis.

  • Two mice in this group died: one on day 43 (after booster immunization; arthritis score 13), and the other on day 36 (no overt clinical arthritis). The latter mouse was excluded from the analysis.

  • §

    Mice with class II major histocompatibility complex (MHC) Aq and Ap haplotypes and no E molecule.

  • Mice with class II MHC Aq haplotype and no E molecule.

  • #

    Mice with class II MHC Aq and Ap haplotypes and the Ep molecule.

 FemaleD3/648 ± 247–502 ± 2
 FemaleN3/446 ± 1239–16024 ± 14
 FemaleD5/657 ± 1045–10923 ± 15
 MaleN3/549 ± 2539–16028 ± 19
(B10.Q × DBA/1)F1     
(BALB/c × B10.Q)F1     
 MaleN2/539 and 5239–9013 and 9
 MaleA3/1010 ± 19–154 ± 2
 FemaleN5/649 ± 843–16023 ± 14
(BQ × CQ)GBPF1§     
 Male and femaleN4/1139 ± 831–624 ± 2
(BQ × CQ)F1     
 Male and femaleN6/1329 ± 2215–908 ± 3
(BP × CP)GBQF1#     
 MaleN10/1154 ± 1438–9631 ± 15
 MaleN4/545 ± 828–7034 ± 21
 MaleN3/439 ± 1119–7045 ± 5

Histologic examination of inflamed paws obtained on day 35, before the booster injection, revealed moderate hypertrophic and hyperplastic synovium, accompanied by pannus and new vessel formation (Figure 2). During the declining phase of arthritis on day 90, a partial healing process was seen before the onset of relapse (results not shown). Moderate synovitis with pannus formation, extensive cartilage and bone erosion, and new cartilage formation were prevalent in the inflamed paws at this stage.

Figure 2.

Histologic features of normal and arthritic mouse joints. Representative photomicrographs of hind paw joint sections are shown. A and B, Metatarsal joints obtained on day 35 (before booster immunization) from normal (A) and native cartilage oligomeric matrix protein (COMP)–immunized (B) C3H.NB mice. In the section from the native COMP–immunized mouse, there is synovitis, with hypertrophic pannus and new vessel formation, extensive cartilage and bone erosion, and disruption of the joint architecture. C and D, Metatarsal interphalangeal joints obtained on day 15 from naive (BALB/c × B10.Q)F1 mice injected on day 0 with purified serum IgG derived from normal mice (C) or from native COMP–immunized mice (D) to generate passively transferred arthritis. In the section from the native COMP–immunized mouse, there is hypertrophic synovium, extensive erosion of bone and cartilage, and disruption of the joint architecture. (Hematoxylin and eosin stained; original magnification × 10.)

Flow cytometric analysis of blood cell populations obtained at the completion of the experiment (159 days after immunization) showed an increase in circulating B cells, neutrophils, and macrophages (Figure 3), most likely reflecting the chronic inflammatory process. Interestingly, there was a higher percentage of CD4+ T cells in arthritic C3H.Q mice than in control mice. In contrast, a significant decrease in the CD4+ T cell population was detected in C3H.NB mice as compared with the control group, but the CD4:CD8 T cell ratio in C3H.NB mice was significantly higher (P < 0.05) than that in C3H.Q mice. More experiments using mice with different MHC haplotypes are needed to clearly ascertain the role of T cells in COMPIA.

Figure 3.

Phenotypes determined by flow cytometry. Cells derived from the peripheral blood of native cartilage oligomeric matrix protein (COMP)–immunized and age-matched naive C3H.NB (A) and C3H.Q (B) mice were obtained on day 159 postimmunization. Significantly higher percentages of B cells (Bc) and neutrophils (Neu), as well as higher percentages of monocytes (Mo), were observed in arthritic C3H.NB mice as compared with age-matched naive mice. In arthritic C3H.Q mice, there were significantly higher percentages of CD4+ T cells and neutrophils as compared with age-matched naive mice. Values are the mean and SD percentage of leukocytes (among 50,000 cells assessed). = P < 0.05; ∗∗ = P < 0.01 by t-test for independent samples.

We found elevated and sustained serum levels of COMP protein, reflecting the severity of inflammation, in female C3H.NB and C3H.Q mice (Figure 4). In addition, high levels of COMP in the BP mice as compared with the BQ mice (especially on day 70) additionally confirmed the H-2Ap association with COMPIA and indicated major disturbances in cartilage homeostasis. Interestingly, an increased or steady-state level of COMP on day 70 seemed to be associated with arthritis severity. On the other hand, a decrease in COMP levels in all of the nonsusceptible strains except mice with the Ncf1 mutation could be interpreted as representing tolerance to arthritis induction that was broken by the homozygous Ncf1 mutation.

Figure 4.

Levels of cartilage oligomeric matrix protein (COMP) in sera obtained on day 35 (d.35) and day 70 (d.70) from mice immunized with native COMP. One female B10.P mouse had a value below the detection limit (<100 ng/ml), as in the normal mice; this value is not shown. Female C3H.NB mice had overall higher levels of COMP on both days 35 and 70 (P < 0.05) as compared with the other mouse strains. Male or female mice of the following strains were studied: B10.P (BP), B10.Q (BQ), C3H.NB (CP), C3H.Q (CQ), B10.Q (Ncf1*/* and Ncf1+/*), (BALB/c × B10.Q)F1 (QB), and (B10.Q × DBA/1)F1 (QD). Each symbol represents an individual mouse; horizontal bars show the geometric mean for the group. = P < 0.05; ∗∗ = P < 0.01; ∗∗∗ = P < 0.001 by t-test for independent samples.

However, no measurable amounts of IgG–COMP complexes could be detected in the sera (data not shown), suggesting the inability of the released COMP fragments to form complexes with the anti-COMP IgG produced by the host. Thus, epitopes present on native COMP used for immunization might not be available on the COMP fragments released from cartilage, which suggests that there is either destruction of these epitopes during the fragmentation of cartilage COMP or tolerization to epitopes unique to fragmented COMP.

Pathogenic role of anti-COMP antibodies.

The COMP-immunized mice developed a strong and specific IgG autoantibody response to COMP (Figure 5), with no notable cross-reactivity with CII (data not shown). Antibodies to COMP reflected the observed strain susceptibility to clinical arthritis. In general, female mice had a higher antibody response than did male mice. To directly address the pathogenic role of anti-COMP antibodies, we injected naive (BALB/c × B10.Q)F1 mice, which are highly susceptible to collagen antibody–induced arthritis (CAIA) (26), with purified IgG fraction from COMP-immunized mice. A group of 10 male mice were injected intravenously with 2 mg of purified IgG. These antibodies bound equally well to both native and monomeric COMP, with a titer that was 6-fold higher than the standard (1 AU/μl). The control IgG fraction did not bind to either native or monomeric COMP. Interestingly, 3 of 10 mice injected with anti-COMP IgG developed arthritis (Table 1), as confirmed by histologic analysis (Figure 2), with mild to severe synovitis, infiltration of inflammatory cells, erosion of bone and cartilage, and in the mouse with severe arthritis, disrupted joint architecture.

Figure 5.

Cartilage oligomeric matrix protein (COMP)–specific antibody responses in mouse sera. A, Titers of antibody against native COMP (anti-nCOMP) IgG in sera obtained on day 35 (d.35) and day 70 (d.70). B, Titers of antibody against denatured COMP (anti-dCOMP) IgG in sera obtained on day 35. Mice used for the anti-dCOMP analysis were the same as those used for the anti-nCOMP analyses. Male or female mice of the following strains were studied: B10.P (BP), B10.Q (BQ), C3H.NB (CP), C3H.Q (CQ), B10.Q (Ncf1*/* and Ncf1+/*), (BALB/c × B10.Q)F1 (QB), and (B10.Q × DBA/1)F1 (QD). Each symbol represents an individual mouse; horizontal bars show the geometric mean for the group. Female C3H.NB mice had significantly higher antibody titers as compared with female C3H.Q, female B10.Q.Ncf1*/*, and male C3H.NB mice on day 35 (P < 0.04) (A). No other significant differences within or between groups were noted for either of the antibodies.

Induction of COMPIA by denatured COMP.

The susceptibility to CIA is critically dependent on the nature of the CII used for immunization. In its native (triple-helical conformation) form, but not its denatured form, CII induces clinically severe arthritis. To test the corresponding situation for immunogenicity to part of the COMP molecule in COMPIA, we used denatured COMP (monomeric form) to immunize the mice. The results from this experiment showed that immunization with denatured COMP induced arthritis with similar severity as that induced by native COMP (Figure 1B and Table 1). A strong antibody response to both denatured and native COMP was induced (Figure 5), suggesting that the major epitope(s) is distant from the joining region at the N-terminus. Similar to native COMP–induced COMPIA, there was a clear difference in the severity of the arthritis between the C3H.NB and the C3H.Q strains (P < 0.03) (Figure 1B and Table 1), with a more severe arthritis developing in C3H.NB mice.

Role of non-MHC genes in COMPIA.

To further determine the genetic components involved in the development of arthritis, we next addressed the influence of a non-MHC gene background on COMPIA. We screened a series of inbred mouse strains for COMPIA susceptibility (Table 1). The C57BL/10 background provided resistance, since neither B10.P (H-2p) nor B10.Q (H-2q) mice developed arthritis. An F1 cross of the B10.Q and DBA/1 strains, (B10.Q × DBA/1)F1, which is highly susceptible to CIA, was also resistant to COMPIA. On the other hand, (BALB/c × B10.Q)F1 mice developed arthritis. Similar to arthritis susceptibility, the antibody response is controlled by both the MHC (higher in H-2p than in H-2q) and the non-MHC (higher in C3H than in B10) genetic backgrounds of the mouse strains tested.

We previously identified an association between an Ncf1 polymorphism responsible for reduced oxidative burst and arthritis severity in the rat, and we confirmed this observation using B10.Q mice with a mutation in the Ncf1 gene, which led to a greater susceptibility to CIA (5, 27). Hence, to test the possibility whether introduction of the Ncf1 mutation in B10.Q mice could break the tolerance for COMPIA, we performed arthritis experiments using a B10.Q strain harboring the Ncf1 mutation (B10.Q.Ncf1*/*). Interestingly, immunization with COMP induced both severe and chronic arthritis in the B10.Q.Ncf1*/* mice, but not in their heterozygous littermate controls or in B10.Q wild-type animals (Table 1). The COMPIA in the B10.Q.Ncf1*/* mice was also associated with a higher antibody response to COMP than that of their heterozygous littermate controls and the B10.Q wild-type mice (Figure 4A).

MHC genetic control of COMPIA.

In the previous experiments, the presence of the H-2p haplotypes appeared to be correlated with the development of severe disease, while mice with the H-2q haplotypes were less susceptible. Mice transgenic for genomic constructs of the class II MHC genes H-2Aβq and H-2Aβp (designated GBQ and GBP, respectively) were used to address the contribution of the H-2A and H-2E molecules in COMPIA. Both these transgenic lines were generated on the B10 background. Since neither of the B10 strains mentioned above developed arthritis, F1 crosses with C3H mice were used in these experiments. Animals from the first cross we tested, (B10.P × C3H.NB)F1, carry the H-2p haplotype. Mice with the H-2p haplotype express both H-2A and H-2E class II molecules. When expressing the GBQ transgene, this line (denoted as [BP × CP]GBQ) also expressed a functional H-2Aq molecule on antigen-presenting cells. Similarly, animals from the second cross we tested, (B10.Q × C3H.Q)F1, carry the H-2q haplotype and express a functional H-2Ap, but not Ep, molecule on antigen-presenting cells, and when transgenic with the GBP construct (denoted as [BQ × CQ]GBP), they also express the Ap haplotype.

As observed with the C3H.NB strain, mice with the H-2p haplotype also developed severe arthritis even when transgenic for GBQ, which is evidence against a protective role of Aq. Furthermore, H-2q mice showed a low incidence and a mild form of arthritis even when transgenic for GBP, which is evidence against a promotive role of Ap. Disease was more severe in the presence of an Ep molecule, as was seen both in C3H.NB versus C3H.Q mice and in the F1 crosses (BALB/c × B10.Q)F1 versus (B10.Q and DBA/1)F1 mice.

Finally, we investigated the role of the human DR gene associated with RA. Thus, HLA–DR4–transgenic B10 mice lacking endogenous class II MHC molecules were immunized with COMP and observed for arthritis development. HLA–DR4 mice were found to be highly susceptible to COMPIA and developed severe arthritis about a week after the booster immunization (Table 1). Introduction of the Ncf1 mutation into HLA–DR4 mice resulted in an even more severe arthritis, with a disease onset that preceded the booster immunization by 1–2 weeks in 3 of the 4 mice.


We present herein a new mouse model of RA with chronic relapsing disease that develops after immunization with COMP, a unique noncollagenous cartilage protein. This model has some advantages over previously described models in the mouse, such as CIA, aggrecan-induced arthritis, and glucose-6-phosphate isomerase (GPI)–induced arthritis (28–30). Importantly, the chronic relapsing disease pattern observed in this model is more useful for comparisons with the RA disease course. Similar to other arthritis models, the choice of the mouse strain seems to be essential for the development of chronic disease. Clearly, the C3H strain background allows a more chronic relapsing disease pattern in both experimental autoimmune encephalomyelitis and in arthritis induced with GPI or CII (19, 31). In CIA, the B10.Q strain is more prone to the development of chronic arthritis, although initially, it is more resistant, and it develops a mild acute arthritis as compared with that in the C3H.Q strain (25). Interestingly, CIA in B10.Q.Ncf1*/* mice is severe and chronic (5), similar to the findings in the present study, suggesting that the introduced Ncf1 mutation responsible for reduced production of oxidative radicals could also break the tolerance to COMP immunization in naive B10.Q mice.

The development of arthritis involves immune responses to many self antigens. So far, the cartilage proteins that have been reproducibly characterized to be arthritogenic in experimental animals involve types II and XI collagen, COMP, and aggrecan (18, 28, 29, 32, 33). It is likely that these are also recognized at different stages in the development of RA or in subtypes of RA (34–36) and that the autoimmune-specific component of the disease is genetically controlled. To date, however, the only genetic association that has been identified to be of importance for autoimmune specificity in humans or in experimental animal models is the MHC region. As expected, the MHC region is strongly associated with each of the self-antigen–specific arthritis models in mice and rats (3, 18, 33, 37). A more surprising observation is that in many cases, the identified class II MHC molecules mimic the shared epitope DR4 and DR1 molecules in humans (2, 38–42). The critical peptide of importance in CIA is the mouse Aq and DR4–restricted glycosylated CII peptide, both in Aq-expressing and in DR4-expressing mice (2). Clearly, COMP also harbors arthritogenic peptides, and as shown in the present study using HLA–DR4 transgenic mice, it is an intriguing possibility that such peptides are also relevant for the human class II molecule shared epitope structure.

Interestingly, T cell recognition of COMP is more complex, since other class II molecules (presumably, Ep, which is a DR homolog in the mouse) are also involved. Presentation of antigens through the E molecule has previously been suggested to be regulatory (43, 44), although in the case of Ep and COMP, this seems not to be the case. A unique property of COMP is that fragments of COMP are released into the bloodstream from cartilage, both physiologically and pathologically during the process of cartilage destruction in arthritis (45, 46). An explanation could be that the arthritogenic T and B cell epitopes are absent from COMP, but are present on circulating COMP fragments as neoepitopes. If that is not the case, then it is also important to raise the question of why the circulating COMP fragments could not induce a stronger regulatory tolerance. These questions can be answered in studies of COMP-deficient mouse models, which have shown a surprisingly normal development (14).

An apparent similarity with the CIA model is the arthritogenic potential of antibodies. This indicates a strong antibody component in the induction of acute arthritis. However, the role of antibodies in the development of chronic arthritis is still unresolved, although single CII epitope–specific monoclonal antibodies have been shown to induce relapses in a mouse model of chronic arthritis (47). Interestingly, a critical role of B cells in RA has recently been highlighted in studies showing that treatment with antibodies to the B cell–specific receptor CD20 is effective in ameliorating the disease (48). Recent discoveries of autoantibodies to citrullinated epitopes that have a similar or higher correlation with RA as rheumatoid factors also emphasize the importance of antibodies as prognostic and diagnostic markers for RA (49, 50).

The mouse arthritis models are efficient tools with which to analyze some of the precise mechanisms of disease development. The new COMPIA model described herein will provide an alternative model by which we can not only analyze the chronicity of the disease course in patients with RA, but also screen candidate drugs for their effective intervention in the disease process.


Drs. Holmdahl and Carlsen had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.

Study design. Carlsen, Nandakumar, Bäcklund, Holmberg, Holmdahl.

Acquisition of data. Carlsen, Nandakumar, Bäcklund, Holmberg.

Analysis and interpretation of data. Carlsen, Nandakumar, Bäcklund, Holmberg, Hultqvist, Vestberg, Holmdahl.

Manuscript preparation. Carlsen, Nandakumar, Bäcklund, Holmberg, Hultqvist, Vestberg, Holmdahl.

Statistical analysis. Carlsen.


We thank Carlos Palestro, Rebecca Lindqvist, and Isabel Bohlin for taking care of the animals, Emma Mondoc for performing the histopathologic analyses, Solveig Borgehammar for help with the COMP purification, and Dick Heinegård, Krisztina Halasz, and Ros-Mari Sandfalk for technical support with the COMP purification.