ARTHRITIS & RHEUMATISMVol. 63, No. 5, May 2011, pp 1312–1321DOI 10.1002/art.30261© 2011, American College of Rheumatology

Proteoglycan-Induced Arthritis and Recombinant HumanProteoglycan Aggrecan G1 Domain–Induced Arthritis in

BALB/c Mice Resembling Two Subtypes ofRheumatoid Arthritis

Tibor T. Glant, Marianna Radacs, Gyorgy Nagyeri, Katalin Olasz, Anna Laszlo,Ferenc Boldizsar, Akos Hegyi, Alison Finnegan, and Katalin Mikecz

Objective. To develop a simplified and relativelyinexpensive version of cartilage proteoglycan–inducedarthritis (PGIA), an autoimmunity model of rheuma-toid arthritis (RA), and to evaluate the extent to whichthis new model replicates the disease parameters ofPGIA and RA.

Methods. Recombinant human G1 domain ofhuman cartilage PG containing “arthritogenic” T cellepitopes was generated in a mammalian expressionsystem and used for immunization of BALB/c mice. Thedevelopment and progression of arthritis in recombi-nant human PG G1–immunized mice (designated re-combinant human PG G1–induced arthritis [GIA]) wasmonitored, and disease parameters were compared withthose in the parent PGIA model.

Results. GIA strongly resembled PGIA, althoughthe clinical symptoms and immune responses in micewith GIA were more uniform than in those with PGIA.Mice with GIA showed evidence of stronger Th1 andTh17 polarization than those with PGIA, and anti-mouse PG autoantibodies were produced in different

isotype ratios in the 2 models. Rheumatoid factor (RF)and anti–cyclic citrullinated peptide (anti-CCP) anti-bodies were detected in both models; however, serumlevels of IgG-RF and anti-CCP antibodies were differentin GIA and PGIA, and both parameters correlatedbetter with disease severity in GIA than in PGIA.

Conclusion. GIA is a novel model of seropositiveRA that exhibits all of the characteristics of PGIA.Although the clinical phenotypes are similar, GIA andPGIA are characterized by different autoantibody pro-files, and the 2 models may represent 2 subtypes ofseropositive RA, in which more than 1 type of autoan-tibody can be used to monitor disease severity andresponse to treatment.

Rheumatoid arthritis (RA) is an autoimmunedisease in which chronic inflammation of the synovialjoints leads to cartilage destruction and bone erosion.Although the etiology of RA is unknown, both environ-mental and genetic factors are thought to be involved inthe pathogenesis of the disease (1). Animal models ofarthritis, particularly those that allow for investigation ofjoint pathology in genetically susceptible rodents (2), areinvaluable tools for RA-related research. Among theanimal models of systemic RA, human cartilage proteogly-can (PG) aggrecan–induced arthritis (PGIA) in BALB/cmice is a T cell–dependent and (auto)antibody/B cell–driven disease (3–5). In addition to the major histocom-patibility complex, PGIA is controlled by multiple ge-netic loci (5), many of which are located in chromosomalregions that are analogous to human regions identifiedin genome-wide association studies of RA (6,7).

The BALB/c mouse strain is genetically predis-posed to the development of arthritis. In addition tocartilage PG aggrecan, immunization with human carti-

Supported in part by the NIH (grants AR-040310, AR-045652, AR-051163, and AR-047657) and the Grainger Foundation,Forest Park, Illinois. Dr. Glant holds the J. O. Galante Endowed Chairin Orthopedic Surgery at Rush University Medical Center.

Tibor T. Glant, MD, PhD, Marianna Radacs, PhD (currentaddress: University of Szeged, Szeged, Hungary), Gyorgy Nagyeri, MS,Katalin Olasz, BS (current address: University of Pecs, Pecs, Hun-gary), Anna Laszlo, BS (current address: Calculus Ltd., Budapest,Hungary), Ferenc Boldizsar, MD, PhD (current address: University ofPecs, Pecs, Hungary), Akos Hegyi, MS, Alison Finnegan, PhD, KatalinMikecz, MD, PhD: Rush University Medical Center, Chicago, Illinois.

Address correspondence to Tibor T. Glant, MD, PhD, Sec-tion of Molecular Medicine, Department of Orthopedic Surgery, 1735West Harrison Street, Cohn Research Building, Room 708, Chicago,IL 60612. E-mail: Tibor_Glant@rsh.net.

Submitted for publication September 2, 2010; accepted inrevised form January 13, 2011.

1312

lage link protein (8) or human cartilage gp-39 (9), butnot with type II collagen (10,11), induces arthritis inBALB/c mice. The BALB/c mouse strain is highly sus-ceptible to serum-transfer arthritis induced by injectionof serum from arthritic K/BxN mice (12,13), and thisstrain also develops arthritis in response to streptococcalcell wall injection (14). Moreover, interleukin-1 (IL-1)receptor antagonist protein–deficient mice (15) andSKG mice that carry a point mutation in the geneencoding ZAP-70 both develop spontaneous arthritis(16), although this only occurs in mice with a BALB/cbackground.

Several years ago, we “simplified” the PGIAmodel by replacing purified human fetal cartilage PG(3,4) with PG isolated from human osteoarthritic carti-lage (11,17) and by replacing Freund’s adjuvant with asynthetic adjuvant (18); this simplification allowed us touse the model for more extensive studies, includinggenome-wide screening of arthritis-associated chromo-somal regions (5). However, the source of antigen(human cartilage) and the cost of antigen preparationwere still limiting factors in the wide-range applicationof the PGIA model.

More recently, we mapped the T cell epitoperepertoire of the core protein of human cartilage PGaggrecan in BALB/c mice and found that 3arthritogenic/dominant and 4 subdominant T cellepitopes were located in the G1 domain of PG (19–21).Immunization of BALB/c mice with short syntheticpeptides corresponding to these epitopes (either aloneor in combination) failed to induce arthritis, promptingus to generate a recombinant human G1 domain thatcontained all of the dominant and subdominant T cellepitopes. Unfortunately, when expressed in prokaryoticcells, the nonglycosylated G1 domain was completelyinsoluble. The use of a baculovirus expression systemseemed to be a more promising approach (22,23), al-though difficulties with the virus titration and antigenpurification procedures precluded the generation oflarge quantities of G1 protein. Finally, using a mamma-lian expression system, we succeeded in a large-scaleproduction of recombinant human G1–recombinantmouse IgG-Fc fusion protein containing enzymaticcleavage site(s) for easy purification.

Herein we describe a new, improved model ofPGIA that is induced by immunization of BALB/c micewith the recombinant human G1 domain of cartilagePG; thus, we have designated this new model GIA (G1domain–induced arthritis). We used both the recombi-nant human G1–recombinant mouse IgG-Fc fusion pro-tein and purified recombinant human G1 (without the

recombinant mouse IgG-Fc partner) to induce GIA inBALB/c mice, and we then compared this model withthe “parent” PGIA model. The GIA model exhibits allof the characteristics described so far for systemic auto-immune arthritis models. Although we did not expectGIA to be more robust than PGIA, side-by-side com-parisons revealed that mice with GIA developed arthri-tis more uniformly and with higher overall inflammationscores than those with PGIA. We found that both GIAand PGIA were associated with abundant production ofautoantibodies to mouse (self) PG. However, the 2models could be distinguished on the basis of “RA-specific” serologic markers. Mice with GIA producedhigher quantities of rheumatoid factor (RF) than thosewith PGIA, whereas mice with PGIA had significantlyhigher serum levels of anti–cyclic citrullinated peptide(anti-CCP) antibodies than those with GIA.

MATERIALS AND METHODS

Generation of a recombinant G1 fusion protein. A753-bp complementary DNA (cDNA) fragment of the mouseIgG2a heavy chain was obtained by reverse transcription ofRNA purified from a monoclonal antibody (mAb)–producingB cell hybridoma. The cDNA fragment was amplified bypolymerase chain reaction (PCR) using primers with linkersfor restriction enzyme (Eco RI and Bcl I) cleavage sites. ThecDNA included the hinge region of the heavy chain and wasinserted into a Lonza pEE14.1 mammalian expression vector(Lonza Biologics). Total RNA was extracted from humanchondrocytes using TriReagent (Sigma-Aldrich), reverse-transcribed, and amplified by PCR. Subsequently, the cDNAfragments (with Eco RI sites) coding for the 351–amino acidG1 domain and 59 amino acids of the interglobular domain(IGD) of PG (23) were cloned into a pBlueScript S/K vector(Stratagene) (Figures 1A–C). The 5� end of the IGD regionwas mutated to carry the cleavage site for endopeptidase factorXa, and the construct was inserted into the Eco RI site inframe with the mFc2a-Lonza construct (Figure 1C). Appropri-ate orientation was determined by PCR, and the entire con-struct was sequenced. The IGD included the natural cleavagesites for stromelysin (matrix metalloproteinase 3) and 2 aggre-canases (ADAMTS-4 and ADAMTS-5) as well as the factorXa cleavage site that was inserted between the IGD and theheavy chain of mouse IgG2a (Figure 1D).

Semiconfluent CHO-K1 (Chinese hamster ovary) cellswere transfected with the hG1-Xa-mFc2a-Lonza constructusing a CaCl2 precipitation method followed by positive selec-tion according to the manufacturer’s protocol (Lonza Bio-Whittaker). First-step (i.e., cloning directly from petri dishes)and second-step (i.e., repeated cloning by limiting dilution)cloning procedures were performed in Ultra-CHO serum-freemedium (CHO-SFM; Lonza BioWhittaker). Irradiated mouseembryonic fibroblasts were used as feeder cells. G1-expressingclones were identified by incubating CHO-SFM supernatant–coated plates with biotinylated mAb G18, followed by detec-tion using peroxidase-labeled streptavidin and tetramethylben-

HUMAN PG G1 DOMAIN–INDUCED ARTHRITIS 1313

zidine substrate (BD Biosciences). The mAb G18 (IgG1) isspecific for an epitope in the A loop of the G1 domain (Figures1A and B) of human cartilage PG aggrecan and does notcross-react with mouse PG (11,23).

Alternative wells were tested with anti-mouse IgG2a,and the clones expressing the highest levels of G1 wereselected. The rhG1-Xa-mFc2a fusion protein was purified

from CHO-SFM using protein G–Sepharose 4B (Pierce). Theyield of the rhG1-Xa-mFc2a fusion protein was �8–10 mg perliter of CHO-SFM. The expression level, quality of purifica-tion, and amounts of the G1 domain expressed (relative to theamounts of native G1 domain isolated from human PG) weredetermined by sodium dodecyl sulfate–polyacrylamide gelelectrophoresis (SDS-PAGE) and Western blotting analysis

Figure 1. Schematics of the cartilage proteoglycan (PG) aggrecan, the G1 domain (which contains an immunoglobulin-like A loop andhyaluronan-binding B and B� loops), the mammalian expression vector containing the recombinant human G1 (rhG1) fusion construct(rhG1-Xa-mFc2a), and the analysis of expressed recombinant proteins. A, The PG molecule, consisting of a protein core to which hundreds ofglycosaminoglycan side chains (chondroitin sulfate or keratan sulfate) are attached together with O-linked and N-linked oligosaccharides. The B andB� loops of the G1 domain of the aggrecan PG core protein (PG monomer) interact with hyaluronan in cartilage; this interaction is stabilized bylink protein. The core protein structure is as follows: G1, G2, and G3 are the globular domains; IGD is the interglobular domain; “KS” indicatesthe keratan sulfate–rich domain; there is also a CS attachment region. Adapted, with permission, from ref. 19. B, Detailed structure of the G1domain, including the major cleavage sites for stromelysin (matrix metalloproteinase 3 [MMP-3]) and 2 aggrecanases (ADAMTS-4 andADAMTS-5). Three dominant/arthritogenic and 4 subdominant epitopes are located in the G1 domain (20,21). Thus, �0.02% of the molecularmass, or �15% of the core protein (i.e., the G1 domain), drives the arthritogenic response to PG in genetically susceptible BALB/c mice. C, TherhG1-Xa-mFc2a construct in a Lonza pEE14.1 mammalian expression vector. D, Schematics of the “double-chain” rhG1-Xa-mFc2a fusion protein.The C-terminal end of the heavy chain of mouse IgG2a (Fc tail) is linked, via the hinge region, to the G1 domain of PG. The heavy chains are ableto reform the disulfide bridges. A properly folded Fc tail binds to protein A or protein G, thus allowing for purification by affinity chromatography.E, Detection of the rhG1-Xa-mFc2a protein (rhG1-mFc2a), separated by 12% sodium dodecyl sulfate–polyacrylamide gel electrophoresis(SDS-PAGE) and stained with Coomassie blue G-250 (left) and separated by Western blotting and stained with monoclonal antibody (mAb) G18(right). Lane 1, Unpurified Ultra-CHO serum-free medium (CHO-SFM) harvested from rhG1-Xa-mFc2a-Lonza–transfected and cloned Chinesehamster ovary cells (�30 �g protein); lane 2, protein G–purified rhG1-Xa-mFc2a fusion protein from the same CHO-SFM serum-free medium (5�g protein); lane 3, purified rhG1-Xa-mFc2a fusion protein after cleavage with factor Xa; lane 4, recombinant human G1 protein repurified usingprotein G–Sepharose; lane 5, highly purified native human G1 domain (�42 kd) isolated from human cartilage PG as previously described (19).Mwt � molecular weight. F, Lower molecular mass protein yielded by in vitro deglycosylation of recombinant human G1. Lane 1, Purifiedrecombinant human G1 without Fc tail; lane 2, recombinant human G1 digested with N-glycosidase F, thus removing all N-linked oligosaccharides;lane 3, recombinant human G1 digested with keratanases 1 and 2; lane 4, recombinant human G1 digested with all enzymes (N-glycosidase F,keratanases, sialidase, and O-glycosidase), thus removing both N-linked and O-linked oligosaccharides.

1314 GLANT ET AL

using mAb G18. Purified rhG1-Xa-mFc2a fusion protein wascleaved with factor Xa (1,000 units per 10 mg fusion protein)(Novagen). The enzyme was removed by incubation withXarrest Sepharose (Novagen), and the IgG2a-Fc fragment wasabsorbed with protein G–Sepharose. The type and level ofglycosylation of purified recombinant human G1 were deter-mined using a deglycosylation kit that included N-glycosidase F(QA-Bio LLC) according to the manufacturer’s instructions.

Immunization of BALB/c mice with cartilage PG orrecombinant human G1. Cartilage from knee joints was ob-tained from consenting osteoarthritis patients who had under-gone joint replacement surgery. The use of human cartilage forPG isolation was approved by the Institutional Review Boardof Rush University Medical Center. The PG isolation anddeglycosylation procedures have been described in detail inprevious reports (11,17). Female BALB/c mice (ages 16–24weeks) were purchased from the National Cancer Institute. Allanimal experiments were approved by the Institutional AnimalCare and Use Committee of Rush University Medical Center.

For immunization with PG, 100 �g of PG core proteinwas emulsified with 2 mg dimethyldioctadecylammonium bro-mide (DDA) adjuvant in phosphate buffered saline (PBS; pH7.4) according to a standard procedure (11). After titration forthe optimal dose, 40 �g of recombinant human G1 (orrhG1-Xa-mFc2a fusion protein equivalent to 40 �g of purifiedrecombinant human G1) was emulsified using 2 mg of DDA inPBS (in a total volume of 300 �l) and injected intraperitoneallyon days 0, 21, and 42 for induction of GIA (18). Mice injectedwith 2 mg of DDA in 300 �l PBS or CHO-SFM were used asnegative controls.

Clinical and histologic assessment of arthritis. Arthri-tis severity was determined using a visual scoring system basedon the degree of swelling and redness of the front and hindpaws (3,4,11). Animals were examined at least 3 times a week;the degree of inflammation was scored from 0 to 4 for eachpaw, resulting in a cumulative arthritis score ranging from 0 to16 for each animal (3,11). All mice were scored by 2 differentinvestigators (TTG, KO, AL, or KM) in a blinded manner. Theincidence of arthritis was expressed as the percentage ofimmunized mice that were arthritic. After the mice were killed,limbs were removed, fixed in 10% formalin, acid-decalcified,embedded in paraffin, and processed according to standardhistologic procedures (3,5,11,24).

Measurement of antigen-specific T cell responses andserum levels of cytokines and anti-PG antibodies. Antigen-specific lymphocyte responses were determined in spleen cellcultures in the presence or absence of 50 �g/ml human PG or10 �g/ml recombinant human G1 domain. Cells were culturedin HL-1 serum-free medium (Lonza BioWhittaker) for 2 days,and antigen-specific IL-2 production was measured using theCTLL-2 bioassay (3,4,11). Cell proliferation was assessed onthe fifth day of culture by measuring 3H-thymidine incorpora-tion (4,11), and antigen-specific proliferation of spleen T cellswas expressed as a stimulation index (3,4,11). The serumcytokines IL-1�, IL-4, IL-6, IL-17, interferon-� (IFN�), andtumor necrosis factor � (TNF�) were measured using enzyme-linked immunosorbent assay (ELISA) kits purchased fromR&D Systems or BD Biosciences. In vitro production of thesecytokines was also assessed by measuring their concentrationsin supernatants of antigen (PG or recombinant human G1)–stimulated spleen cell cultures on the fifth day of culture by

ELISA. Cytokine concentrations were normalized to cell num-ber (pg or ng of cytokine per million spleen cells), as describedpreviously (5,25).

PG-specific serum antibodies were quantified byELISA using serially diluted serum samples. Purified human ormouse cartilage PG, or recombinant human G1 without the Fctail, was immobilized in Maxisorp 96-well plates (Nunc Inter-national) at a concentration of 0.1 �g/well each (11). ForPG-specific IgG isotype assays, peroxidase-labeled goat anti-mouse IgG1 antibody (Zymed) or IgG2a (BD Biosciences) wasused after incubation with serum. Serum PG-specific antibodylevels were calculated using serial dilutions of pooled sera ofmice with PGIA and known antibody titers (11).

Measurement of RF and anti-CCP antibodies in se-rum. Mouse IgG- and IgM-type RFs were measured in mouseIgG2a-Fc–coated plates; the IgG2a-Fc fragment was isolatedfrom the mTSG6-Xa-mFc2a fusion protein (another fusionprotein where the recombinant protein is different; thus, nocross-reactivity might be present using the mouse Fc2a tail)after cleavage with factor Xa and purification on a proteinG–Sepharose column. RFs (mouse Fc-binding autoantibodies)were measured in serially diluted (1:500–1:2,000 for IgM-RFand 1:2,000–1:8,000 for IgG-RF) serum samples collected atdifferent time points during the immunization period. Fc-bound IgG-type and IgM-type RFs were detected using poly-clonal antibodies to mouse IgG1 and IgM (�1-chain and�-chain specific, respectively). The results were validated(units/ml serum) by retesting representative samples usingcommercially available mouse IgG-RF and mouse IgM-RFELISA kits (Shibayagi).

Anti-CCP antibody levels were quantified usingQuanta Lite CCP-3 ELISA kits (Inova Diagnostics) with aminor modification to the manufacturer’s protocol to detectmouse anti-CCP antibodies. Briefly, serially diluted serum(previously pooled from mice with PGIA) was titrated tocorrespond to the highest arbitrary (“calibrator”) unit of thehuman reference sample in the kit by adjusting the dilution ofthe peroxidase-labeled secondary (anti-mouse IgG, IgA, andIgM) antibody. Our mouse serum with anti-CCP antibodieswas calibrated to contain 250 arbitrary units/�l serum and wasthen used as the mouse reference sample/standard in allsubsequent experiments.

Statistical analysis. Descriptive statistics were used todetermine group mean � SEM values. Differences between 2groups were tested for statistical significance using Student’st-test, and differences between �3 groups were tested forstatistical significance by one-way analysis of variance followedby the least significant difference post hoc test. Fisher’s exactchi-square test was used to evaluate statistical significancewhen disease parameters were compared. All statistical analy-ses were performed using the SPSS version 16.0 statisticalsoftware package. P values less than 0.05 were consideredsignificant.

RESULTS

Characterization of recombinant human G1 andthe clinical phenotype and histopathology of GIA. Wedeveloped a mammalian expression system for the large-scale production of the G1 domain of human cartilage

HUMAN PG G1 DOMAIN–INDUCED ARTHRITIS 1315

PG; our goal was to replace the antigen (PG isolatedfrom human cartilage) for the immunization of, andinduction of arthritis in, genetically susceptible BALB/cmice. Figure 1D shows representative schematics of therhG1-Xa-mFc2a fusion protein. The Fc tail of the IgG2aheavy chain was properly folded; therefore, the pairedheavy chains were able to bind to protein G. As demon-strated by SDS-PAGE and Western blotting analysisusing the human G1-specific mAb G18 (Figure 1E), therhG1-Xa-mFc2a fusion protein was effectively purifiedfrom the supernatants of a CHO-K1 transfectant clone.Cleavage of the fusion protein with factor Xa andremoval of the Fc tail on protein G–Sepharose yieldednearly 100% pure recombinant human G1 domain. Areduction in the molecular mass of the purified recom-binant human G1 fragment after digestion with variousglycosidases indicated that the protein was secreted bythe mammalian cells (CHO) in a glycosylated form thatcontained both N-linked and O-linked oligosaccharides(Figure 1F).

Purified rhG1-Xa-mFc2a fusion protein that had

not been cleaved with factor Xa was used first forimmunization of BALB/c mice, which were then com-pared with mice in which arthritis had been induced viathe “parent” PGIA model (Figures 2A and B). Approx-imately 10.0–12.5 �g of purified recombinant human G1(either fusion protein or single G1 domain) in DDAadjuvant was the lowest threshold dose for the inductionof arthritis, while �100 �g of recombinant human G1resulted in an “overdose” that actually delayed the onsetof arthritis (data not shown). The optimal dose wasfound to be between 25 �g and 50 �g of recombinantprotein. Therefore, we routinely used 40 �g of recom-binant human G1 (emulsified in DDA) per injection.

As in the PGIA model (11), the earliest onset ofarthritis (occurring in �10% of all immunized mice) wasnoted after the second injection and was mostly re-stricted to the interphalangeal and metacarpophalangealand/or metatarsophalangeal joints. In a side-by-sidecomparison, GIA reached 98–100% incidence levels9–10 days after the third immunization (Figure 2A), andthe cumulative arthritis score was consistently higher in

Figure 2. Comparison of disease development and immune responses in mice with recombinant human G1–induced arthritis (GIA) andPG-induced arthritis (PGIA). A and B, Incidence (A) and severity (B) of arthritis in BALB/c mice immunized with either rhG1-Xa-mFc2a fusionprotein (40 �g of recombinant human G1/injection) or cartilage PG (100 �g/injection) isolated from osteoarthritic cartilage (11,17,48). Arrowsindicate the third injection, administered on day 42. Each animal was scored for arthritis 3 times a week. C, T cell proliferation and interleukin-2(IL-2) production in response to stimulation with recombinant human G1 or PG. D, In vitro antigen (recombinant human G1 and humanPG)–induced cytokine production by spleen cells isolated from mice with GIA and PGIA, respectively. E and F, Serum levels of cytokines (E) andof antibodies to recombinant human G1, mouse PG, and human PG (F) in mice with GIA and PGIA. Values are the mean � SEM. � � P � 0.05;�� � P � 0.01. PBS � phosphate buffered saline; DDA � dimethyldioctadecylammonium bromide; TNF� � tumor necrosis factor �; IFN� �interferon-� (see Figure 1 for other definitions).

1316 GLANT ET AL

the GIA model than in the PGIA model (Figure 2B).Overall, the 2 models were very similar; however, in-flammation was initiated more robustly and involvementof the front paws was more extensive in the GIA model.Consistent with this finding, cartilage and bone destruc-tion in the front paws appeared to be more progressivein the GIA model than in the PGIA model, although theclinical symptoms (i.e., redness and swelling of affectedjoints) were comparable between the 2 groups. Thejoints became severely deformed due to the inflamma-tory destruction of cartilage and bone (Figure 3), similarto results observed in previous investigations of PGIA(3,5,11). At the late, chronic stage of arthritis (2–3months after onset), there were no clinical or histo-pathologic differences between the 2 models (results notshown).

T and B cell–mediated immune responses in GIAand PGIA. In vitro tests (i.e., T cell proliferation andcytokine production) showed evidence of robust T cellresponses to recombinant human G1 or PG stimulation.As expected, the magnitude of the responses to theseantigens was different in the 2 models. As shown inFigure 2C, mice immunized with recombinant humanG1 exhibited higher levels of T cell proliferation (ex-pressed as a stimulation index) and IL-2 productionwith recombinant human G1 than with PG, and PG-immunized mice exhibited higher levels of T cell prolif-eration and IL-2 production in response to stimulationwith PG than in response to stimulation with recombi-nant human G1. Antigen-specific production of IL-6,TNF�, and IL-4 was higher in spleen cell cultures frommice with PGIA than in spleen cell cultures from micewith GIA, although significantly more IFN� and IL-17were secreted by spleen cells of mice with GIA (Figure2D). Serum levels of IL-17 and IL-1� were �2-foldhigher in mice with GIA than in mice with PGIA,although essentially no TNF� was detected in the sera ofarthritic mice immunized with recombinant human G1(Figure 2E).

Serum levels of antibodies to recombinant humanG1 or human cartilage PG (especially the IgG1 isotype)were �3–4-fold higher in mice with PGIA than in thoseimmunized with recombinant human G1 (Figure 2); thiswas probably due to the presence of multiple epitopes(including immunogenic carbohydrate stubs) in full-length PG obtained from cartilage. However, in the seraof mice with GIA, unusually high concentrations ofanti-mouse PG IgG2a isotype antibodies were detected(Figure 2F). The IgG2a:IgG1 ratio of anti-mouse PGautoantibodies was 4.62 in mice with GIA and 0.75 inmice with PGIA, suggesting a strong Th1 polarization

Figure 3. Macroscopic images of hind limbs (insets) and histopathol-ogy of corresponding ankle joints of a normal (nonimmunized) mouse(A), a mouse with recombinant human G1–induced arthritis (GIA)(B), and a mouse with proteoglycan-induced arthritis (PGIA) (C).Sections of decalcified hind paws were stained with hematoxylin andeosin. There were no clinically or histologically detectable differenceswhen arthritic limbs were compared between animals immunized withrhG1-Xa-mFc2a or recombinant human G1 (GIA) and animals im-munized with human cartilage PG (PGIA). In contrast to the normaljoint, ankles in mice with GIA and PGIA showed histologic evidenceof inflammation, synovial pannus formation, and cartilage and bonedestruction.

HUMAN PG G1 DOMAIN–INDUCED ARTHRITIS 1317

(also indicated by the cytokine profile [Figures 2D andE]) in recombinant human G1–immunized mice.

Serum autoantibody levels as disease biomarkersof PGIA and GIA. The unusually high serum levels ofanti-mouse PG autoantibodies (especially of the IgG2aisotype) (Figure 2F) prompted us to investigate whetherother arthritis-related (“RA-specific”) antibodies, suchas RF or anti-CCP antibodies, were also produced inmice with GIA or PGIA. To this end, we collected bloodsamples every 7–8 days during the experimental periodfrom mice immunized with recombinant human G1,rhG1-Xa-mFc2a, mFc2a, or human PG. Mice injectedwith PBS in DDA adjuvant served as negative serumcontrol donors (only results from experiments withrecombinant human G1 fusion protein– and humanPG–immunized mice are shown).

As shown in Figure 4, serum levels of all autoan-tibodies, including anti-mouse PG (IgG1 and IgG2aisotypes) (Figures 4E and F), IgG-type and IgM-typeRFs (Figures 4C and D), and anti-CCP antibodies(Figure 4B), started to rise weeks before the firstappearance of clinical symptoms of arthritis (Figure 4A).While IgM-type RF and IgG2a anti-mouse PG antibodylevels were similar in the 2 models (Figures 4D and F),the levels of IgG1 anti-mouse PG autoantibodies andanti-CCP antibodies were �4–5-fold higher in the seraof mice with PGIA than in the sera of mice with GIA(Figures 4B and E). In contrast, IgG-type RF was themost prominent autoantibody in mice with GIA (Figure4C). The serum levels of these autoantibodies (exceptfor IgM-type RF in mice with GIA) showed a significantpositive correlation with increasing arthritis severityscores (Figure 4A) in both the GIA and PGIA models(data not shown).

We next investigated whether the serum levels ofany specific autoantibody could “predict” the onset timeor severity of arthritis in individual animals with eitherGIA or PGIA. Because all recombinant human G1– andhuman PG–immunized mice had developed arthritis bythe end of this experiment, we selected the mice with thelowest (late onset, low severity) and highest diseaseresponse from each group and compared their arthritisonset time and severity scores with their serum autoan-tibody levels over time (Figures 4G–J). The levels of allautoantibodies rose quickly in the serum of the high-responder mice from both the GIA and PGIA groups,and we did not observe any difference between the 2models (Figures 4H and J). However, low serum levelsof IgG-RF and anti-CCP antibodies were observed inthe low-responder mouse (delayed onset and low diseaseseverity) from the GIA group (Figure 4G) but not in thelow-responder mouse from the PGIA group (Figure 4I).

Figure 4. Kinetics of autoantibody production in the context of arthritisdevelopment in recombinant human G1–induced arthritis (GIA) andproteoglycan-induced arthritis (PGIA). A–F, Arthritis scores (A), serumlevels of anti–cyclic citrullinated peptide (anti-CCP) antibodies (Ab) (B),rheumatoid factors (IgG-RF and IgM-RF) (C and D), and autoantibodiesto mouse cartilage PG (E and F) were monitored in recombinant humanG1– and PG-immunized mice between days 7 and 58 after the firstimmunization. Arrows indicate second and third injections, administeredon days 21 and 42, respectively. Results from 17 mice with GIA and 14mice with PGIA are shown. Values are the mean � SEM. G–J, Shown areindividually analyzed arthritis scores and corresponding autoantibodytiters in low- and high-responder mice in the GIA and PGIA groups.Although the results for only 1 low-responder mouse and 1 high-responder mouse from each group are compared here, similar resultswere obtained in a replicate experiment. Note that the right y-axis scalesin G–J are different from the y-axis scales in B, C, E, and F.

1318 GLANT ET AL

DISCUSSION

In this report, we have described GIA, a simpli-fied and improved version of cartilage PGIA (3,5). GIAwas induced by systemic immunization of femaleBALB/c mice with a recombinant protein (G1 domain ofhuman PG) in synthetic DDA adjuvant, and the para-meters of arthritis in these animals were compared withthose in animals with PGIA. The incidence of arthritisreached 95–100% in both GIA and PGIA, althoughdisease severity was consistently higher in the GIAmodel than in the parental PGIA model, primarily dueto a more extensive involvement of the front paws in theinflammatory process. However, joints commonly af-fected in both models (e.g., the ankle) did not showsignificant macroscopic or microscopic differences at thepeak of arthritis. As expected, T cells from mice withGIA reacted better (in terms of proliferation and IL-2production) to in vitro stimulation with recombinanthuman G1 than with PG; likewise, T cells from mice withPGIA responded more robustly to stimulation with PGthan with recombinant human G1. However, both mod-els showed evidence of PG-specific autoimmunity, sinceautoantibodies to mouse PG were produced in compa-rable quantities. We found that there were differencesbetween PGIA and GIA in the production levels ofsome disease-associated cytokines and antibodies. Forexample, higher serum levels of IL-1�, IL-17, and IgG-type RF and a higher IgG2a:IgG1 ratio of anti-mousePG autoantibodies were detected in GIA, but higheramounts of anti-CCP antibodies were detected in PGIA.

Spleen cells from mice with GIA that were stim-ulated in vitro with recombinant human G1 producedlarger quantities of IFN� and IL-17 but less TNF� thanPG-stimulated splenocytes from mice with PGIA. PGIAhas been postulated to be a Th1-type disease (26–29),with significant IL-17 production (30), in which IFN�determines the requirement for IL-17 (29). Because theproduction of both IFN� and IL-17 was more robust inGIA than in PGIA, GIA could represent an intermedi-ate form of the disease between the Th1- and Th17-mediated forms. In addition, the propensity of T cells inmice immunized with recombinant human G1 to pro-duce higher amounts of IFN� and IL-17 than T cells inPG-immunized animals could contribute to the develop-ment of a slightly more severe form of arthritis in micewith GIA.

PGIA and GIA are closely related, although theycould be distinguished on the basis of select clinical andimmunologic parameters. Therefore, these models mayresemble 2 subtypes of seropositive RA with slightlydifferent disease phenotypes and cytokine/autoantibody

profiles. However, RA is a heterogeneous disease, andwhile the various animal models are tremendously help-ful for investigating certain aspects of the human dis-ease, none of these models embodies the full spectrumof diseases collectively called RA. Notably, thousands ofinvestigators and pharmaceutical companies use animalmodels of RA, perhaps without understanding the dif-ferences among the different subtypes and forms of thisdisease and the corresponding animal models (2,31–33).Moreover, while animal experiments represent acceler-ated forms of the disease, RA develops insidiously fordecades, and only a few serum markers are indicative ofimminent disease.

GIA represents a simplified, uniform, and afford-able version of the murine PGIA model. The recombi-nant human G1 domain contains arthritogenic T cellepitopes (20) and lacks the undesirable variability of theepitope repertoire that might be found in differentpreparations of human cartilage–derived PG. GIA ex-hibits most of the disease characteristics of the parentalPGIA model and is now available for laboratories thatdo not have access to human cartilage or that lackexperience in PG preparation. The homogeneity of theantigen and the presence of arthritogenic epitopes inrecombinant human G1 would also allow for sophisti-cated in vivo studies on the involvement of autoreactiveT cells (T helper cell subsets) and B cells in the diseaseprocess. Immunization with a protein that containsmultiple T cell epitopes of the PG molecule, such asrecombinant human G1, is necessary for the investiga-tion of disease-related alterations in T cell function.

An additional value of GIA is the presence oflarge quantities of IgG-type RF in the circulation (atlevels higher than those observed in PGIA), whichare absent in commonly used autoimmune models ofRA, such as collagen-induced arthritis (CIA) in theDBA/1 mouse strain or the K/BxN transgenic model ofspontaneous arthritis (2). To our knowledge, antiimmu-noglobulin (RF) and anti-CCP antibodies have beenobserved only in humanized HLA-transgenic mice im-munized with type II collagen (34–36), while autoanti-bodies to citrullinated filaggrin were observed in anotherstudy of CIA (37).

The prominence of RF and anti-CCP antibodiesin RA raises the possibility that these antibodies play arole in the pathogenesis of the disease. Indeed, anti-bodies with RF activity have been found depositedon the cartilage surface in patients with RA (38,39).Autoantibodies in the CIA model or in K/BxN trans-genic mice have been shown to be pathogenic, sincepassive transfer of serum from these mice inducestransient arthritis in naive recipients (40,41). However,

HUMAN PG G1 DOMAIN–INDUCED ARTHRITIS 1319

the characteristic autoantibodies in these models(against mouse type II collagen and glucose-6-phosphate isomerase, respectively) are detected only ina very small proportion of RA patients (2). Conversely,RF, a characteristic and abundant autoantibody in RA,is conspicuously absent in CIA and in K/BxN mice(2,41); however, anti-CCP antibody injection has beenshown to increase the severity of CIA (42). The produc-tion of high amounts of RF in GIA would make thismodel suitable for in vivo studies investigating theemergence of RF-secreting B cells and potential forma-tion of RF deposits in the joints in the context of arthritisinduction.

GIA and PGIA also appear to be the first animalmodels in which both RF and anti-CCP antibodies aredetected in the serum. Moreover, serum levels of RFand anti-CCP antibodies correlate significantly witharthritis scores and may predict disease severity inindividual mice with GIA. Together, RF and anti-CCPantibodies are considered important diagnostic/prognostic biomarkers of RA (43–47). Therefore, thepresence and prognostic potential of both of thesemarkers in mice with GIA should increase the relevanceof this model to RA and its usefulness in preclinicalstudies monitoring the efficacy of emerging drugs inestablished arthritis.

ACKNOWLEDGMENTS

The authors would like to thank Yanal Murad, ZsuzsaGyorfy, Oktavia Tarjanyi, Katalin Kis-Toth, Balint Farkas, andGabor Hutas for technical assistance.

AUTHOR CONTRIBUTIONS

All authors were involved in drafting the article or revising itcritically for important intellectual content, and all authors approvedthe final version to be published. Dr. Glant had full access to all of thedata in the study and takes responsibility for the integrity of the dataand the accuracy of the data analysis.Study conception and design. Glant, Finnegan, Mikecz.Acquisition of data. Glant, Radacs, Nagyeri, Olasz, Laszlo, Boldizsar,Hegyi, Finnegan, Mikecz.Analysis and interpretation of data. Glant, Laszlo, Boldizsar, Hegyi,Mikecz.

REFERENCES

1. Gregersen PK. Genetics of rheumatoid arthritis: confrontingcomplexity. Arthritis Res 1999;1:37–44.

2. Bevaart L, Vervoordeldonk MJ, Tak PP. Evaluation of therapeutictargets in animal models of arthritis: how does it relate torheumatoid arthritis? [review]. Arthritis Rheum 2010;62:2192–205.

3. Glant TT, Mikecz K, Arzoumanian A, Poole AR. Proteoglycan-

induced arthritis in BALB/c mice: clinical features and histopa-thology. Arthritis Rheum 1987;30:201–12.

4. Mikecz K, Glant TT, Poole AR. Immunity to cartilage proteogly-cans in BALB/c mice with progressive polyarthritis and ankylosingspondylitis induced by injection of human cartilage proteoglycan.Arthritis Rheum 1987;30:306–18.

5. Glant TT, Finnegan A, Mikecz K. Proteoglycan-induced arthritis:immune regulation, cellular mechanisms and genetics. Crit RevImmunol 2003;23:199–250.

6. Etzel CJ, Chen WV, Shepard N, Jawaheer D, Cornelis F, SeldinMF, et al. Genome-wide meta-analysis for rheumatoid arthritis.Hum Genet 2006;119:634–41.

7. Choi SJ, Rho YH, Ji JD, Song GG, Lee YH. Genome scanmeta-analysis of rheumatoid arthritis. Rheumatology (Oxford)2006;45:166–70.

8. Zhang Y, Guerassimov A, Leroux JY, Cartman A, Webber C,Lalic R, et al. Induction of arthritis in BALB/c mice by cartilagelink protein: involvement of distinct regions recognized by T and Blymphocytes. Am J Pathol 1998;153:1283–91.

9. Verheijden GF, Rijnders AW, Bos E, Coenen-de Roo CJ, vanStaveren CJ, Miltenburg AM, et al. Human cartilage glycopro-tein-39 as a candidate autoantigen in rheumatoid arthritis. Arthri-tis Rheum 1997;40:1115–25.

10. Adarichev VA, Valdez JC, Bardos T, Finnegan A, Mikecz K,Glant TT. Combined autoimmune models of arthritis revealshared and independent qualitative (binary) and quantitative traitloci. J Immunol 2003;170:2283–92.

11. Glant TT, Mikecz K. Proteoglycan aggrecan-induced arthritis: amurine autoimmune model of rheumatoid arthritis. Methods MolMed 2004;102:313–38.

12. Ohmura K, Johnsen A, Ortiz-Lopez A, Desany P, Roy M, BesseW, et al. Variation in IL-1� gene expression is a major determi-nant of genetic differences in arthritis aggressivity in mice. ProcNatl Acad Sci U S A 2005;102:12489–94.

13. Uysal H, Nandakumar KS, Kessel C, Haag S, Carlsen S, BurkhardtH, et al. Antibodies to citrullinated proteins: molecular interac-tions and arthritogenicity. Immunol Rev 2010;233:9–33.

14. Koga T, Kakimoto K, Hirofuji T, Kotani S, Ohkuni H, WatanabeK, et al. Acute joint inflammation in mice after systemic injectionof the cell wall, its peptidoglycan, and chemically defined pepti-doglycan subunits from various bacteria. Infect Immun 1985;50:27–34.

15. Horai R, Saijo S, Tanioka H, Nakae S, Sudo K, Okahara A, et al.Development of chronic inflammatory arthropathy resemblingrheumatoid arthritis in interleukin 1 receptor antagonist–deficientmice. J Exp Med 2000;191:313–20.

16. Sakaguchi N, Takahashi T, Hata H, Nomura T, Tagami T,Yamazaki S, et al. Altered thymic T-cell selection due to amutation of the ZAP-70 gene causes autoimmune arthritis in mice.Nature 2003;426:454–60.

17. Glant TT, Cs-Szabo G, Nagase H, Jacobs JJ, Mikecz K. Progres-sive polyarthritis induced in BALB/c mice by aggrecan fromnormal and osteoarthritic human cartilage. Arthritis Rheum 1998;41:1007–18.

18. Hanyecz A, Berlo SE, Szanto S, Broeren CP, Mikecz K, Glant TT.Achievement of a synergistic adjuvant effect on arthritis inductionby activation of innate immunity and forcing the immune responsetoward the Th1 phenotype. Arthritis Rheum 2004;50:1665–76.

19. Glant TT, Buzas EI, Finnegan A, Negroiu G, Cs-Szabo G, MikeczK. Critical role of glycosaminoglycan side chains of cartilageproteoglycan (aggrecan) in antigen recognition and presentation.J Immunol 1998;160:3812–9.

20. Buzas E, Vegvari A, Murad YM, Finnegan A, Mikecz K, GlantTT. T-cell recognition of differentially tolerated epitopes of carti-lage proteoglycan aggrecan in arthritis. Cell Immunol 2005;235:98–108.

21. Szanto S, Bardos T, Szabo Z, David CS, Buzas E, Mikecz K, et al.

1320 GLANT ET AL

Induction of arthritis in HLA–DR4–humanized andHLA–DQ8–humanized mice by human cartilage proteoglycanaggrecan but only in the presence of an appropriate (non-MHC)genetic background. Arthritis Rheum 2004;50:1984–95.

22. Zhang Y, Guerassimov A, Leroux JY, Cartman A, Webber C,Lalic R, et al. Arthritis induced by proteoglycan aggrecan G1domain in BALB/c mice: evidence for T cell involvement and theimmunosuppressive influence of keratan sulfate on recognition ofT and B cell epitopes. J Clin Invest 1998;101:1678–86.

23. Murad YM, Szabo Z, Ludanyi K, Glant TT. Molecular manipula-tion with the arthritogenic epitopes of the G1 domain of humancartilage proteoglycan aggrecan. Clin Exp Immunol 2005;142:303–11.

24. Glant TT, Bardos T, Vermes C, Chandrasekaran R, Valdez JC,Otto JM, et al. Variations in susceptibility to proteoglycan-inducedarthritis and spondylitis among C3H substrains of mice: evidenceof genetically acquired resistance to autoimmune disease. ArthritisRheum 2001;44:682–92.

25. Szabo Z, Szanto S, Vegvari A, Szekanecz Z, Mikecz K, Glant TT.Genetic control of experimental spondylarthropathy. ArthritisRheum 2005;52:2452–60.

26. Finnegan A, Mikecz K, Tao P, Glant TT. Proteoglycan (aggrecan)-induced arthritis in BALB/c mice is a Th1-type disease regulatedby Th2 cytokines. J Immunol 1999;163:5383–90.

27. Hollo K, Glant TT, Garzo M, Finnegan A, Mikecz K, Buzas EI.Complex pattern of Th1 and Th2 activation with a preferentialincrease of autoreactive Th1 cells in BALB/c mice with proteogly-can (aggrecan)-induced arthritis. Clin Exp Immunol 2000;120:167–73.

28. Doodes PD, Cao Y, Hamel KM, Wang Y, Farkas B, Iwakura Y, etal. Development of proteoglycan-induced arthritis is independentof IL-17. J Immunol 2008;181:329–37.

29. Doodes PD, Cao Y, Hamel KM, Wang Y, Rodeghero RL, MikeczK, et al. IFN-� regulates the requirement for IL-17 in proteogly-can-induced arthritis. J Immunol 2010;184:1552–9.

30. Boldizsar F, Tarjanyi O, Nemeth P, Mikecz K, Glant TT. Th1/Th17 polarization and acquisition of an arthritogenic phenotype inarthritis-susceptible BALB/c, but not in MHC-matched, arthritis-resistant DBA/2 mice. Int Immunol 2009;21:511–22.

31. Ahlqvist E, Hultqvist M, Holmdahl R. The value of animal modelsin predicting genetic susceptibility to complex diseases such asrheumatoid arthritis. Arthritis Res Ther 2009;11:226–35.

32. Cai C, La Cava A. Mimicking self-antigens with synthetic peptidesin systemic autoimmune rheumatic diseases. Curr Clin Pharmacol2009;4:142–7.

33. Van den Berg WB. Lessons from animal models of arthritis overthe past decade. Arthritis Res Ther 2009;11:250.

34. Taneja V, Taneja N, Paisansinsup T, Behrens M, Griffiths M,Luthra H, et al. CD4 and CD8 T cells in susceptibility/protectionto collagen-induced arthritis in HLA-DQ8-transgenic mice: impli-cations for rheumatoid arthritis. J Immunol 2002;168:5867–75.

35. Taneja V, Behrens M, Mangalam A, Griffiths MM, Luthra HS,David CS. New humanized HLA–DR4–transgenic mice that

mimic the sex bias of rheumatoid arthritis. Arthritis Rheum2007;56:69–78.

36. Taneja V, Behrens M, Basal E, Sparks J, Griffiths MM, Luthra H,et al. Delineating the role of the HLA-DR4 “shared epitope” insusceptibility versus resistance to develop arthritis. J Immunol2008;181:2869–77.

37. Kidd BA, Ho PP, Sharpe O, Zhao X, Tomooka BH, Kanter JL, etal. Epitope spreading to citrullinated antigens in mouse models ofautoimmune arthritis and demyelination. Arthritis Res Ther 2008;10:R119.

38. Jasin HE. Autoantibody specificities of immune complexes seques-tered in articular cartilage of patients with rheumatoid arthritisand osteoarthritis. Arthritis Rheum 1985;28:241–8.

39. Monach PA, Hueber W, Kessler B, Tomooka BH, Benbarak M,Simmons BP, et al. A broad screen for targets of immunecomplexes decorating arthritic joints highlights deposition ofnucleosomes in rheumatoid arthritis. Proc Natl Acad Sci U S A2009;106:15867–72.

40. Stuart JM, Dixon FJ. Serum transfer of collagen-induced arthritisin mice. J Exp Med 1983;158:378–92.

41. Korganow AS, Ji H, Mangialaio S, Duchatelle V, Pelanda R,Martin T, et al. From systemic T cell self-reactivity to organ-specific autoimmune disease via immunoglobulins. Immunity1999;10:451–61.

42. Kuhn KA, Kulik L, Tomooka B, Braschler KJ, Arend WP,Robinson WH, et al. Antibodies against citrullinated proteinsenhance tissue injury in experimental autoimmune arthritis. J ClinInvest 2006;116:961–73.

43. Nielen MM, van Schaardenburg D, Reesink HW, Twisk JW, vande Stadt RJ, van der Horst-Bruinsma IE, et al. Simultaneousdevelopment of acute phase response and autoantibodies inpreclinical rheumatoid arthritis. Ann Rheum Dis 2006;65:535–7.

44. Goldbach-Mansky R, Lee J, McCoy A, Hoxworth J, Yarboro C,Smolen JS, et al. Rheumatoid arthritis associated autoantibodiesin patients with synovitis of recent onset. Arthritis Res 2000;2:236–43.

45. Andersson AK, Li C, Brennan FM. Recent developments in theimmunobiology of rheumatoid arthritis. Arthritis Res Ther 2008;10:204–12.

46. Aletaha D, Neogi T, Silman AJ, Funovits J, Felson DT, BinghamCO III, et al. 2010 rheumatoid arthritis classification criteria: anAmerican College of Rheumatology/European League AgainstRheumatism collaborative initiative. Ann Rheum Dis 2010;69:1580–8.

47. Aletaha D, Neogi T, Silman AJ, Funovits J, Felson DT, BinghamCO III, et al. 2010 rheumatoid arthritis classification criteria: anAmerican College of Rheumatology/European League AgainstRheumatism collaborative initiative. Arthritis Rheum 2010;62:2569–81.

48. Cs-Szabo G, Roughley PJ, Plaas AH, Glant TT. Large and smallproteoglycans of osteoarthritic and rheumatoid articular cartilage.Arthritis Rheum 1995;38:660–8.

HUMAN PG G1 DOMAIN–INDUCED ARTHRITIS 1321