Type II collagen–specific antibodies induce cartilage damage in mice independent of inflammation




Murine collagen antibody–induced arthritis (CAIA), like collagen-induced arthritis, has clinical and immunopathologic features that parallel those in human rheumatoid arthritis (RA). This study was undertaken to examine the effects of autoantibodies to type II collagen (CII) on articular cartilage in the paws of mice, under conditions in which other factors that may influence joint pathology could be excluded.


Mice of 2 different strains, B10.QC5δ and the parental strain B10.Q, were injected intravenously with either saline or arthritogenic monoclonal antibodies (mAb) to CII. B10.QC5δ mice lack complement factor C5 and do not develop CAIA when injected with arthritogenic mAb, whereas B10.Q mice have C5 and develop CAIA when administered the mAb and a subsequent injection of lipopolysaccharide. Three days after injection the paws of the mice were examined by standard histologic methods to assess morphologic appearance and proteoglycan loss, and by synchrotron-enhanced Fourier transform infrared microspectroscopy to assess chemical evidence of structural change.


No macroscopic or microscopic signs of inflammation were evident in the mice. However, in contrast to the saline-injected controls, all mAb-injected mice exhibited cartilage damage in all joints, with loss of proteoglycans and collagen, chondrocyte hyperplasia and/or loss, and surface damage in the interphalangeal joints.


The implication of these findings is that an autoimmune response to CII can disrupt articular cartilage, particularly that of the small joints, and the subsequent integrity of the cartilage depends on a balance between breakdown and repair. This has relevance with regard to RA, in which such autoantibodies occur but the inflammatory response may dominate clinically and mask underlying features of the autoimmune response.

Collagen-induced arthritis (CIA), which results from immunization of susceptible animals with type II collagen (CII), is an inflammatory arthritis with clinical and immunologic features that parallel those in human rheumatoid arthritis (RA) (1). The preferred animal for use in the study of events in this model is the mouse, because of the number of strains that are available for examination of genetic susceptibility to the autoimmune response and ensuing arthritis. In addition, there are well-characterized murine monoclonal antibodies (mAb) to defined epitopes on CII (2, 3), and these can, on passive transfer, result in collagen antibody–induced arthritis (CAIA), which has all the features of CIA (4–6).

CAIA occurs independently of any direct activity of B and T cells; thus, with this model, effector processes can be studied independently of events that occur during disease induction (6). It is not major histocompatibility complex restricted. Most mouse strains are susceptible, but the severity varies, and lipopolysaccharide (LPS) injection may be required for enhancement of arthritis. Articular inflammation and cellular infiltration characteristic of the effector stage are attributable to deposited immune complexes and activation of complement and Fc receptors (FcR) (6, 7). Cartilage and bone destruction follows the activation of macrophages, lymphocytes, and synoviocytes and production of matrix metalloproteinases (MMPs) and cytokines, including tumor necrosis factor and interleukin-1β (7).

Studies of the in vitro effects of mAb to CII have shown that treatment with 2 mAb (M2139 and CIIC1) that are potently arthritogenic in vivo affected cartilage matrix integrity in cartilage explant cultures (8, 9), impaired synthesis of new matrix, and disrupted collagen fibril formation in chondrocyte cultures (10, 11). Notably, in the culture systems used in those in vitro studies, cartilage damage occurred independently of immune cells or their small molecular mediators. These findings suggest that antibodies to CII may participate directly in the cartilage damage that accompanies articular inflammation. The hypothesis for the current study was, therefore, that mice that do not develop the macroscopic inflammation and cellular infiltration characteristic of CAIA would nonetheless show the same microscopic evidence of cartilage damage as that seen in vitro, after receiving arthritogenic mAb in vivo.

To test this we transferred arthritogenic mAb, without LPS, to mice of 2 strains with reduced capacity to mount an inflammatory response. We examined changes in the cartilage on day 3; this time point was chosen to avoid changes secondary to any cellular inflammatory response and to reduce the confounding effects of regeneration and repair seen in healthy cartilage treated with mAb in vitro (9). We used standard procedures to assess histologic changes in articular cartilage and used Fourier transform infrared microspectroscopy (FTIRM) to evaluate changes in the chemistry of collagen. With this approach, we demonstrated that in the absence of macroscopic inflammation or the cellular infiltration characteristic of CAIA, mice nevertheless exhibited loss of cartilage structure including collagen and proteoglycans, and changes to chondrocytes.


Monoclonal antibodies.

CIIC1, UL-1, M2139, and CIIC2 are arthritogenic mAb derived from hybridomas developed from CII-immunized mice (2, 12, 13) that bind to separate well-defined conformational triple-helical epitopes of native CII. CIIC1 binds to the C1I epitope GARGLTGROGDA (amino acids [aa] 358–375) (14), UL-1 to the U1 epitope GLVGPRGERGF (aa 494–504) (13), M2139 to the J1 epitope MPGERGAAGIAGPK (aa 551–564), and CIIC2 to the D3 epitope RGAQGPPGATGF (aa 687–698) (15). The experiments were performed using a mixture of 2 mAb (M2139 plus CIIC1) or 4 mAb, (M2139 plus CIIC1 plus UL1 plus CIIC2) as in previous in vivo experiments (5, 15).


Two strains of mice were studied: C5-congenic B10.Q mice that lack complement factor C5 and hence cannot develop CAIA (B10.QC5δ) (16) and the parental strain (B10.Q) that are C5 sufficient and can develop CAIA (although only ∼10% do so without an injection of LPS) (5). The B10.QC5δ mice were generated using the speed congenic technique (13); the congenic fragment from NOD on the B10.Q genetic background was ∼54 Mb, ranging from D2Mit116 to D2Mit91. Permission for all animal experiments was granted by the local (Lund-Malmö region, Sweden) animal welfare authorities (permit no. M7-01).

Effects on cartilage after passive transfer of CII mAb.

Changes in cartilage after passive transfer of mAb to CII in the absence of inflammation were examined in 4-month-old mice that were injected intravenously with 2 mAb (CIIC1 plus M2139), 4 mAb (CIIC1 plus M2139 plus UL1 plus C2), or phosphate buffered saline (PBS). For histologic analysis, each of these treatments was administered to 4 B10Q mice and 4 B10.QC5δ mice (total 24 mice). The mixture of M2139 and CIIC1 contained 4.5 mg of each of the sterile-filtered antibody solutions in a final volume of 0.4 ml, and the mixture of CIIC1, M2139, UL1, and C2 contained 1 mg of each mAb in a final volume of 0.4 ml. Hind paws only from 8 additional mice (4 from each strain) injected with 2 mAb were used for a comparative study of changes in the small joints of the phalanges and the larger tarsal joints.

Mice were killed on day 3 after injection (5). Paws were fixed in 4% phosphate buffered paraformaldehyde solution for 24 hours, decalcified for 3–4 weeks in EDTA–polyvinylpyrrolidone–Tris solution (pH 6.9), dehydrated, and embedded in paraffin for processing for histologic and synchrotron FTIRM analysis. All 4 paws from each mouse were collected and processed, with the hind paws embedded in 2 blocks to allow examination of both phalangeal and tarsal joints (6 blocks per mouse). Each paw was sectioned longitudinally to a depth that allowed examination of both bone and cartilage surfaces of multiple joints within each section.

Histologic analysis.

For histologic examination, 5-μm serial sections of tissue were cut and stained with hematoxylin and eosin (H&E), and with toluidine blue to allow semiquantitative examination of proteoglycan loss from the cartilage (17). As a control for proteoglycan staining, the staining intensity of the cartilage at the surfaces of the carpal, metacarpal, and phalangeal joints was compared with that of cartilage within growth plates of the radial and ulnar epiphyses in the same section.


Histomorphometry to measure cartilage thickness and mean chondron size was performed on H&E-stained sections, using MCID software (M4 3.0 Rev 1.1; Imaging Research) on images captured at 200× magnification, as described previously (8). Individual chondrocytes were manually outlined using the MCID software, which then was used to calculate chondrocyte area.

Scoring of cartilage changes.

Scoring of cartilage changes was based on a modified Mankin score derived from the procedure that was used for grading cartilage changes in osteoarthritis (OA) (18, 19) (Table 1). The scale was reduced to exclude scoring of extreme structural change, such as pannus and fissures into the calcified cartilage layer, seen in longstanding OA. A combined score was derived for the joints by assessing structure (0–2 points), cellular abnormalities (0–3 points), and matrix staining (0–4 points); a total score of 0 reflected normal cartilage and 9 the most severe cartilage lesions. Scores for the structure of the cartilage and cellular abnormalities were derived from assessment from H&E-stained sections, and scores for matrix staining were derived from assessment of toluidine blue–stained sections. For each mouse, scores were derived from at least 3 different sections (111 joints in total), with selection based on the quality of the histologic sample.

Table 1. Grading scale for cartilage, based on modified Mankin scoring
 Point score
 Irregular surface, including bumps, fissures into radial layer1
 Slight disorganization2
Chondrocyte abnormalities 
 Hypercellularity, including small superficial clusters1
 Clusters or protruding cells2
Matrix staining 
 Slight reduction1
 Moderate reduction2
 Severe reduction3

Assessment of chemical composition of cartilage by synchrotron FTIRM.

Detailed statistical analysis of the differences in single point spectra from cartilage matrix and chondrocytes from the mice, obtained by FTIRM, has been published (20). In the present study chemical changes in areas of the cartilage, particularly changes to CII in the histologic samples, were examined. Sections of paraffin-embedded tissue (5 μm) on MirrIR low-e microscope slides (Kevley Technologies) were processed as described previously (8, 9, 20). To obtain the requisite spatial resolution, measurements were performed on the infrared beamline at the Australian Synchrotron (Melbourne, Victoria, Australia), with an aperture of 5 μm (9). Grids that encompassed regions of the cartilage surface and included the bone beyond the cartilage surface were defined. Spectra were analyzed over the region 940–1800 cm−1 using Cytospec (CytoSpec; http://www.cytospec.com), with unsupervised hierarchical cluster analysis (21) performed with second-derivative spectra where minima in the second-derivative spectra correlate with the maxima peaks in the nonderivatized spectra (22). This allowed the spectral resolution of bands that contribute to inflection points or shoulders in the original spectra, and eliminated problems with sloping baselines arising from loss of cartilage components and scattering effects.

Analysis of CII concentrated on the location of the amide I peak (1640–1670 cm−1) which, for native triple-helical collagen, is >1660 cm−1. This amide I peak is complex due to contributions from the secondary structure of the proteins; for collagen a band shift to lower wave number is characteristic of denaturation (20), and a shoulder at ∼1637 cm−1 has been attributed to the triple helix absorption (23), or possibly to new collagen synthesis (20). For proteoglycans, analysis was based on peaks within the region of 960–1175 cm−1 derived from carbohydrate moieties.

Statistical analysis.

Statistical analysis was performed using Statistica for Windows, version 4.5 (StatSoft). Differences in mean cartilage thickness and chondrocyte area obtained by histomorphometry were assessed by Student's t-test. With regard to scores for cartilage changes, the nonparametric Kruskal-Wallis analysis of variance (ANOVA) by ranks was performed to determine the significance of differences between groups, and the Mann-Whitney U test was used to compare individual differences. P values less than 0.05 were considered significant.


Histologic features of joints.

On day 3, none of the 32 mice studied had developed macroscopic arthritis. None of the joints from any mouse showed synovitis or histologic evidence of inflammation or cellular infiltration. However, there were clear histologic differences between antibody-injected and control mice, with abnormalities in the cartilage, varying in degree for different joints, even within individual animals.

Cartilage structure.

In the control mice injected with PBS, the cartilage surface was smooth and was highly cellular, with evenly spaced chondrocytes containing rounded nuclei with 2–3 nucleoli (Figure 1A). In the mAb-injected mice, areas of the cartilage appeared normal, but in many joints the cartilage surface was irregular, with surface fissures and loss of cells, or unevenness with protrusions of chondrocytes (Figures 1B and C). The appearance of the chondrocytes varied, with regions of hypercellularity and cell clusters suggestive of cell division, or hypocellularity with areas of empty chondrons, particularly deep in the cartilage (Figure 1D).

Figure 1.

Representative hematoxylin and eosin–stained sections of joints from control and monoclonal antibody (mAb)–injected B10.QC5δ mice, illustrating cartilage changes seen in mice injected with mAb. A, Joint from a mouse injected with phosphate buffered saline, in which the cartilage surface (CA) is smooth and chondrocytes are evenly spaced and have round nuclei. B, Joint from a mouse injected with 2 mAb, in which the cartilage surface is uneven. C and D, Joints from mice injected with 4 mAb, with protruding cells (arrows in C) or hypocellularity and many empty chondrons (arrows in D). Neither the number of mAb nor the mouse strain used affected the patterns of reactivity observed. All images shown are on the same scale. BO = bone.

Proteoglycan loss.

Staining of the articular cartilage from control mice was strong and of even intensity, similar to that of the growth plate (Figure 2A). In contrast, the intensity of staining of cartilage from the mAb-injected mice varied, with articular cartilage staining being substantially weaker than growth plate staining (Figure 2B). This was particularly evident in the distal interphalangeal joints of the forepaws, where there were regions in which there was an intensely stained ring of matrix around chondrocytes, often when there were protrusions of chondrocytes from the cartilage surface (Figure 2C). In other areas there was complete loss of staining over part of the cartilage, usually corresponding to areas containing empty chondrons (Figure 2D).

Figure 2.

Representative toluidine blue–stained sections of joints from control and monoclonal antibody (mAb)–injected B10.QC5δ mice. A, Joint from a mouse injected with phosphate buffered saline, in which the cartilage (CA) stained evenly and with the same intensity as the growth plate (GP). B–D, Joints from mice injected with 4 mAb, in which the cartilage stained weakly when compared with the growth plate (B) and protruding chondrocytes on the surface of phalangeal joints showed haloes of intense staining (arrows in C), whereas there was complete loss of staining in areas containing empty chondrons (arrows in D). Neither the number of mAb nor the mouse strain used affected the patterns of reactivity observed. BO = bone.

Comparison of histologic changes in groups of mice.

A modified Mankin score was used to examine variations in the degree of cartilage abnormality in multiple joints treated with different combinations of mAb, and in different mouse strains. For each mouse, cartilage structure, chondrocyte abnormalities, and degree of proteoglycan loss in the matrix of multiple joints were scored. There were highly significant differences in the overall scores between the 6 groups (P = 0.0009 by Kruskal-Wallis ANOVA by ranks). Although there was considerable variation both in the score derived from different joints in an individual mouse and between mice, for each of the features examined the percentage of joints with abnormalities (Figure 3A) and the median scores for the joints (Figure 3B) in mice of either strain that received mAb exceeded those in PBS-injected mice. For the 30 joints examined from the 4 PBS-injected mice of each strain, the median score for each mouse and for each characteristic was 0 (range 0–1) (Figure 3B). In contrast, the groups of mice that received either 2 or 4 mAb had significantly higher scores (median 5, range 4–7 [P = 0.00108 by Mann-Whitney U test] in mice that received 2 mAb; median 6, range 3–7 [P = 0.00015 by Mann-Whitney U test] in mice that received 4 mAb). There were no significant differences in the results between the 2 mouse strains, and although the median score was slightly higher among the mice injected with 4 mAb than among the mice injected with 2 mAb, the difference was not statistically significant.

Figure 3.

Histologic analysis of the groups of mice, with findings in structure, cells, or matrix scored using the grading scale shown in Table 1. A, Percent of joints with abnormalities (percent positive) in each group. B, Median score for each individual characteristic analyzed and median total score, from the joints for which data from both hematoxylin and eosin staining and toluidine blue staining were available (n values in A and B are the number of joints). Vertical lines show the range. C, Percent of tarsal joints and of distal interphalangeal (DIP) joints, from 8 mice injected with 2 monoclonal antibodies (mAb), with abnormalities in structure, cells, or matrix. D, Median score for each individual characteristic analyzed and median total score, from the joints analyzed in C. Vertical lines show the range. PBS = phosphate buffered saline.

Comparison of changes in different joints.

As in human RA, arthritic changes in both CAIA and CIA are particularly associated with the small peripheral joints. To compare the effects of mAb on cartilage in smaller and larger joints, serial sections were examined from the hind paws of 8 mice (4 from each strain) that had been injected with 2 mAb, for comparison of findings in the small distal interphalangeal joints and the larger tarsal joints.

The median and range of the overall Mankin scores for the 2 joints were similar (median 5, range 2–6 for the tarsal joints; median 5.5, range 4–6 for the interphalangeal joints), but differences in the abnormalities were observed (Figures 3C and D). In the larger tarsal joints of 7 of the 8 mice, the surface of the cartilage was normal, although samples from most of the mice exhibited areas containing empty chondrons and in such areas there was complete loss of proteoglycan staining. In the interphalangeal joints, in contrast, the cartilage surface was irregular in samples from 5 of the 8 mice, the cells protruded, and around these there was strong staining, possibly representing new proteoglycan synthesis, although there was loss of staining deeper in the matrix. Chondrocytes within the cartilage that showed an uneven surface were significantly larger than those within cartilage with a smooth surface (mean ± SD 145 ± 35 μm2 versus 74 ± 30 μm2; P < 0.0001), and in the small distal interphalangeal joints the thickness of the cartilage was significantly reduced in areas where the surface was uneven, compared with smooth cartilage (33.9 ± 7.4 versus 48.9 ± 6.4 μm; P = 0.0039) These changes are consistent with osmotic swelling of the chondrocytes as a result of loss of matrix (24, 25) seen previously in vitro (10).

Cartilage changes measured by synchrotron FTIRM.

Spectral grids that spanned the cartilage surface and bone were compared with light microscopy images of the same section. From previous analysis of single point spectra (20), the major changes were in the amide I region, with loss of the band above 1660 cm−1 and appearance of a band at 1635–1640 cm−1 associated with activated chondrocytes (9), possibly representing newly synthesized collagen (20). There was no apparent contribution from the proteoglycans at 960–1175 cm−1, so analysis was performed using 2 clusters defined by differences in the spectral region 1600–1750 cm−1.

For PBS-injected control mice, in 4 of 5 grids examined, the major amide I band in each cluster was located above 1660 cm−1, characteristic of native, triple-helical collagen, whether type II collagen in cartilage or type I in bone (Figure 4). Only 1 showed a cluster defined by a band at 1635–1640 cm−1, which appeared as a narrow border at the cartilage surface.

Figure 4.

Representative section of cartilage from a control mouse injected with phosphate buffered saline, examined by synchrotron Fourier transform infrared microspectroscopy. The infrared image of the area of the unstained tissue shown by light microscopy (A) was processed by unsupervised hierarchical cluster analysis over the range 1600–1750 cm−1 using second-derivative spectra with 2 clusters (coded red and blue) to identify changes in collagen (B). In control mice the second-derivative spectra for the 2 clusters (C) were almost identical; in both cases the wave numbers of the amide I bands at 1664 cm−1 and 1670 cm−1 in the blue and red clusters were characteristic of undenatured collagen and did not distinguish between cartilage (CA) and bone (BO). Lines in A and B show the cartilage border.

The 5 grids collected from the mAb-injected mice were more complex, and only 1 appeared normal. Overall there was more disruption at the cartilage surface, leading to loss of pixels in the region of the superficial cartilage where spectra failed the quality test in Cytospec. Where superficial cartilage remained, the most striking feature was a decrease in the amide I band above 1660 cm−1 and the development of a substantial band at 1635–1640 cm−1, which often became the most prominent band in the spectrum (Figure 5). There were no changes in the location of bands in the region 960–1175 cm−1 that would indicate alterations in proteoglycan composition, but decreases in heights of peaks in that region corresponding to proteoglycan loss were seen in the nonderivatized spectra (data not shown). Taken together, these results obtained by synchrotron-enhanced FTIRM confirm the cartilage damage and proteoglycan loss seen histologically, and provide evidence of marked cellular activation and changes to the collagenous matrix.

Figure 5.

Representative section of cartilage from a monoclonal antibody–injected mouse, examined by synchrotron Fourier transform infrared microspectroscopy. The infrared image of the area of the unstained tissue shown by light microscopy (A) was processed by unsupervised hierarchical cluster analysis over the range 1600–1750 cm−1 using second-derivative spectra with 2 clusters (coded red and blue) to identify changes in collagen (B). The mean spectrum for the bone (BO) and cartilage (CA) close to the bone identified by the red cluster was similar to that of the control tissue, with an amide I band at 1670 cm−1, but the amide I band for the areas near the surface of the cartilage, identified by the blue cluster, was shifted to 1640 cm−1 (C), previously associated with activated chondrocytes (9) and possibly new collagen synthesis (20). Lines in A and B show the cartilage border.


As constituents of immune complexes, autoantibodies to CII are essential for the development of inflammation in CAIA, either by activation of complement or by direct engagement and activation of FcR-bearing inflammatory cells (6, 7), but the detrimental effects of these antibodies on the integrity and function of articular cartilage has rarely been considered. Previously, we developed strategies to isolate deleterious effects attributed to antibody in vitro from those caused by the inflammatory response that occurs in vivo. In those experiments, mAb to CII that were arthritogenic in vivo, in the absence of inflammation in vitro, induced subtle morphologic changes in chondrocytes and abnormalities in collagen fibrils in the newly synthesized matrix (10, 11, 26); in mature cartilage, loss of proteoglycans, denaturation of CII, and overall disruption of the topography and chemistry of articular cartilage in situ were observed (8, 9, 26). The strength of the effects was dependent on the concentration of antibody and duration of exposure, and the balance between breakdown and repair was dependent on the presence of viable chondrocytes (9).

In the present study 2 strains of mice with limited capacity to develop CAIA were injected with arthritogenic mAb to determine whether autoantibodies to CII cause similar damage unrelated to inflammation in vivo. B10.Q mice may develop arthritis without additional LPS treatment but not as early as within 3 days (5), and B10.QC5δ mice lack C5 and do not develop CAIA (16). We examined the joints on day 3, before inflammatory sequelae were anticipated and before an injection of LPS is usually given. The mAb used bind cartilage both in vitro in explant cultures, where penetration of mAb increases as matrix is destroyed (8), and in vivo, where injected biotinylated mAb bind even more readily than on cartilage in vitro (27, 28). Despite the absence of any overt macroscopic or histologic evidence of inflammation, cartilage damage occurred in the mAb-injected mice, with proteoglycan loss and associated histologic changes including disruption of the cartilage surface, thinning of the cartilage in small joints, and evidence of chondrocyte hyperplasia and activation and/or empty chondrons. Changes demonstrated by FTIRM included denaturation and loss of collagen and abnormalities consistent with cellular activation and new collagen synthesis. The changes in vivo were similar to those seen previously in vitro in cartilage explants (8, 9, 26). There were no significant differences between the parent and congenic animals, indicating that the effects were not due to complement activation.

Overall, the most prominent feature observed in the joints of the mice injected with arthritogenic mAb was loss of proteoglycan from the cartilage. Proteoglycan loss is an early marker of cartilage disruption in both human OA and RA, and is usually attributed to the activation of MMPs and aggrecanases that are released after disruption of the molecular interactions between matrix constituents (29, 30). As in OA, the proteoglycan loss was accompanied by evidence of compensatory repair, including chondrocyte clusters and increased matrix synthesis, and areas of further damage marked by areas of empty chondrons (31, 32), changes that also occur in vitro (8, 9, 26). The empty chondrons may reflect the loss of CII, as the mAb are not directly cytotoxic (10, 11, 26), but increased chondrocyte death is associated with lack of CII in transgenic mice (33).

Although the mAb to CII do not react directly with proteoglycans, CII is the major protein of the extracellular matrix, forming the fibrillar skeleton within which are enmeshed chondrocytes, proteoglycans, and other matrix proteins (34). The importance of such interactions for the maintenance of cartilage integrity is highlighted by our previous observation that the arthritogenic mAb CIIC1 penetrates more readily and deeply into cartilage that lacks the important structural protein type IX collagen (35). The epitopes on CII are sites of interactions with other matrix components that are critical for the structural integrity of the cartilage matrix (3, 11, 36), and disruption of these interactions by mAb would enhance mechanical damage and proteoglycan loss. This loss of proteoglycans which provide “cushioning” of the cartilage in the joint would result in greater susceptibility to damage from compressive forces and greater penetration of degradative molecules, whether these be enzymes released by inflammatory cells or the arthritogenic mAb used in the present study.

The apparent thinning of the cartilage in the small joints at the tips of the paws correlates with the observation of in vitro collapse of the cartilage matrix that follows loss of proteoglycans and the denaturation of collagen demonstrated by FTIRM (8). In cartilage explant cultures, penetration of the mAb extended as the cartilage matrix was destroyed. In vivo, disruption of the matrix would result in direct mechanical loss of cartilage protein that could then be readily taken up into the surrounding synovial tissue. Such uptake has been shown for citrullinated CII (37) as well as for native triple-helical CII (38), with fragments of CII being detectable by immunohistochemistry in both synovial fluid and synovial tissue in human RA. Such release could provide a further antigenic stimulus, and would contribute to the development of CAIA by facilitating immune complex deposition. In this study the inflammatory changes that normally accompany immune complex deposition and complement activation were abrogated, but in most mouse strains, particularly following LPS injection, the proteoglycan loss induced by the arthritogenic mAb would enhance immune complex deposition within the joint and potentiate the degradative response.

The mechanical instability associated with loss of proteoglycan from cartilage contributes to development of noninflammatory arthritis, and OA can be induced in animals by intraarticular injection of papain, which causes proteoglycan degradation (39–41). Changes in the knees of mice 3 days after injection of papain, described by van der Kraan et al (41), closely resembled those in the joints of the mice on day 3 in the present study, and included pronounced proteoglycan loss, damage at the cartilage surface, and chondrocyte proliferation and/or cell death. Residual chondrocytes were often surrounded by a pericellular halo of strongly staining matrix. Notably, the cartilage damage after papain injection progressed without further treatment, and by day 42 was typical of early OA, including cartilage fragmentation and osteophyte formation. However, there was no inflammation. Possibly mAb to CII could induce changes akin to those of OA in the longer term, but in the animal models of CIA and CAIA, and in human RA, the inflammatory response dominates the pathology. The prospect of ongoing antibody-mediated damage leading to changes in the absence of inflammation is, however, important in light of the increasingly successful therapy of human RA with new biologic agents targeting inflammation. Extension of the current work to include longitudinal investigations in mice that did not develop CAIA after injection with mAb to CII might be informative, but was beyond the scope of the present study.

Antibodies to CII are found in the serum, cartilage, and synovial fluid of patients with RA (42–44). However, despite a prevailing view that they occur in only a minority of patients (36), several studies have shown a high frequency, particularly in early RA (45–47), with frequency decreasing as the disease progresses (46), and these antibodies have been eluted from cartilage (44). Their later disappearance from serum may reflect binding to the cartilage matrix and immune complex formation as more epitopes are exposed due to cartilage damage. Possible explanations for variable assay results include contamination of the purified CII by bound matrix molecules, which block the epitopes, or by pepsin, which may increase background levels in the assay. Use of purified antigens might increase sensitivity. For example, the frequency of antibodies detected by enzyme-linked immunosorbent assay increased from 24% when intact CII was used to 88% when the CB10 fragment of CII was used (48), and the use of synthetic triple-helical peptides of CII resulted in an even higher frequency (49). Since antibodies of the same epitope specificity as those in mice also occur in RA (49), the detrimental effects on the integrity and functions of articular cartilage observed in the present study may well contribute to the perpetuation and chronicity of the human disease.

Apart from anti-CII antibodies, a wide array of autoantibodies (50), some of which may function similarly to CII-specific antibodies in disrupting the architecture of articular cartilage, have been found in patients with RA. In particular, antibodies to citrullinated protein antigens are clearly important in disease pathogenesis, being particularly associated with erosive arthritis. Citrullinated CII occurs in the joints of RA patients, and antibodies reactive with it may mediate inflammation by formation of immune complexes (51). However, citrullination involves the modification of arginine, and the major B cell epitopes of CII contain arginine and are surface exposed on the collagen fibril (3). At least some antibodies to citrullinated CII react with the collagen fibril in the same regions as the antibodies to CII, and hence could contribute to cartilage breakdown in the same way as do CII antibodies. Indeed, mAb to the citrullinated C1 epitope on CII have been derived and shown to bind cartilage and to induce or enhance arthritis in mice (37), although their effects on cartilage have not been tested in vitro.

In essence, the importance of our findings lies in the implication that the autoimmune response in RA originates and persists in articular cartilage and there is a balance between disruption and repair regulated by chondrocytes. Many of the sequelae that develop in RA are a consequence of the inflammatory response associated with immune complex formation. It has been shown that CII-containing immune complexes obtained from RA patient sera induced the production of proinflammatory cytokines from peripheral blood monocytes via FcγRIIA (44), suggesting that this could be one of the possible detrimental mechanisms by which immune complexes induce inflammation. Therapeutic suppression of inflammation may restore quality of life for patients with RA but does not address the underlying problem of the autoimmune reaction and the balance between disruption and repair. Collectively, the present observations suggest the importance of understanding autoimmune responses and their consequences in order to design effective treatments to suppress the specific autoimmune response causing damage, and to promote repair and healing.


All authors were involved in drafting the article or revising it critically for important intellectual content, and all authors approved the final version to be published. Dr. Rowley 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 conception and design. Croxford, Whittingham, Nandakumar, Holmdahl, Rowley.

Acquisition of data. Croxford, Rowley.

Analysis and interpretation of data. Croxford, McNaughton, Rowley.


We thank Dr. Ian Mackay for helpful discussions and editorial comments. We are grateful for having been provided access to the Australian Synchrotron infrared microspectroscopy beamline for use in this study.