Humoral immune response to citrullinated collagen type II determinants in early rheumatoid arthritis



Collagen type II (CII) is a relevant joint-specific autoantigen in the pathogenesis of rheumatoid arthritis (RA). Whereas the reasons for the breakage of self tolerance to this major cartilage component are still enigmatic, T cell responses to glycosylated CII determinants in RA patients indicate that post-translational modifications play a role. Since the conversion of arginine into citrulline by peptidylarginine deiminases (PAD) in some non-joint-specific antigens such as filaggrin or fibrin has been shown to give rise to RA-specific humoral immune responses, we investigated whether PAD modification of cartilage-specific CII might affect its recognition by circulating autoantibodies in early RA. In vitro treatment with purified PAD led to arginine deimination of native CII or of synthetic CII peptides as evidenced by amino acid analysis. The citrullination resulted in modified recognition of the immunodominant CII epitope C1III (amino acid residues 359–369) by murine and human antibodies. In a cohort of early RA patients (n=286), IgG antibodies directed toward a synthetic citrullinated C1III peptide (citC1III-P) were detectable with a prevalence of 40.4%. The partial autoantibody cross-reactivity between citC1III-P and citrullinated peptides mimicking epitopes of the cytoskeletal autoantigen filaggrin suggests that autoimmunity to cartilage-specific modified self might be a critical intermediate bridging recognition of PAD-modified extra-articular autoantigens with the disruption of tolerance to native cartilage constituents.


Arbitrary unit


Collagen type II


Cyclic citrullinated peptide


Cyclic citrullinated peptides of the diagnostic kit Immunoscan RA Mark 2


Synthetic citrullinated C1III peptide


Peptidylarginine deiminase


Rheumatoid arthritis


In rheumatoid arthritis (RA) circulating autoantibodies of different antigen specificities reflect the activation of humoral immune responses in chronic joint inflammation. Although the pathogenic role of these autoantibodies still remains elusive, their analysis has proven useful in the development of disease markers. More recently, autoantibodies that bind to citrullinated proteins, including so-called anti-keratin, anti-fillagrin and anti-cyclized citrullinated peptide (anti-ccp) antibodies, have attracted much attention due to their considerably increased specificity as diagnostic markers 1, 2 as well as their detectability at very early stages of RA 3. Antigen recognition by these RA-specific autoantibodies is critically dependent on the presence of the non-standard amino acid citrulline, which is generated by an enzymatically catalyzed deimination (citrullination) of arginine residues. The enzyme involved in the citrullination of arginine residues is peptidylarginine deiminase (PAD), which exists in different isotypes (reviewed in 4). In the inflamed synovium, predominant expression of PAD type 2 and type 4 in CD14+ monocyte/macrophages has been described 5. The local occurrence of the citrullinated proteins vimentin 6 and fibrin 7 indicate an ongoing post-translational arginine conversion by these enzymes in the arthritic joint. However, in contrast to its well-established relevance for diagnostic procedures, relatively little is known about the significance of protein citrullination and the resulting autoimmune responses to altered self determinants in the pathogenesis of joint inflammation. Accordingly, attempts have so far failed to demonstrate arthritogenicity of citrulline-directed immunity in murine models of joint inflammation or to provide convincing evidence for the induction of autoimmunity to deiminated arginines in experimental arthritis models 8. On the other hand, much more is known about the potential of autoantibodies directed toward collagen type II (CII) epitopes, which are recognized in cartilage-specific autoimmune responses in RA patients, to induce an erosive arthritis in naive mice upon antibody transfer 9, 10. However, anti-CII autoantibodies are detectable in the circulation in only 5–15% of RA patients, and divergent data has been published on the issue of whether CII autoimmunity represents an early or late event in the disease course 11, 12. In the present investigation, we tested the hypothesis of whether CII might be a substrate for structural modification by PAD enzymes, which could result in a breakage of self tolerance to this cartilage-specific autoantigen. We provide first experimental evidence for a PAD2 (PAD isotype 2)-catalyzed conversion of arginine to citrulline in a CII region of immunodominance for autoantibody formation 9 and show the presence of circulating autoantibodies that specifically bind to the citrullinated triple helical collagen peptide CII359–369 in 40.4% of 286 early RA patients (duration of arthritis symptoms <12 months). Thus, the prevalence of anti-citrullinated CII IgG is considerably increased compared to the frequencies of autoantibodies directed to the non-modified CII (19.7%). Our results indicate that autoimmune recognition of citrullinated collagen is a frequent event in early RA. Moreover, the anti-citrullinated CII IgG correlates with the occurrence of so-called anti-ccp autoantibodies, which are widely used as early markers of RA 3 and can be determined by a commercially available ELISA kit that is based on the detection of IgG binding to synthetic citrullinated filaggrin peptide derivatives. Thus, the formation of anti- citrullinated CII IgG is part of an initial adaptive immune response that is directed toward modified self components in the joint and may later spread to the recognition of native structures.


Modification of CII by PAD treatment

First we investigated the question of whether PAD can modify arginine residues in CII, which is usually rather resistant to enzymatic modification due to the stable triple helical arrangement of its three α1 chains. For this purpose purified human CII was PAD treated and upon acidic hydrolysis subjected to amino acid analysis. Representative chromatograms of hydrolyzed CII that had previously been treated with either active or heat-inactivated PAD are shown in Fig. 1. The occurrence of an additional peak with a retention time on the reverse phase column corresponding to that of the citrulline standard is detectable in samples treated with active but not heat-denatured PAD (as a negative control), thus providing clear experimental evidence for the catalytic conversion of arginine residues into citrulline in triple helical CII.

Figure 1.

Amino acid analysis of CII upon treatment with either active or heat-inactivated PAD2. The chromatograms of the derivatized amino acid residues eluting from the reverse phase column at different retention times are presented. Precalibration of the column with the respective amino acid standards allows for identification of the proline and citrulline peaks by their characteristic retention times. As depicted in the enlarged sections of the chromatograms, citrulline was only detectable in the CII sample that was treated with the catalytically active PAD.

Altered immunorecognition of citrullinated CII by murine and human autoantibodies

The consequences of PAD treatment of CII on its immune recognition by a CII-specific murine mAb (CIIC1) and human RA sera were investigated in ELISA as well as Western blot experiments. As shown in Fig. 2A, PAD treatment of CII resulted in a complete loss of binding by CIIC1 antibodies. Since the fine specificity of this antibody is known (aa 359–369), including the critical importance of an arginine residue in position 361 of CII for binding 13, this result indicates the destruction of the CIIC1 epitope by the enzymatic conversion of arginine 361 into citrulline. In contrast, PAD treatment of CII can also facilitate its recognition as an autoantigen; as demonstrated in Fig. 2B, IgG antibodies in selected human RA sera did not recognize unmodified CII in ELISA but did react upon pretreatment of CII with PAD. The Western blot in Fig. 2C shows a representative experiment for RA sera that specifically recognize PAD-modified CII in ELISA. Whereas the serum IgG antibodies did not recognize unmodified CII or CII incubated with heat-denatured PAD, they bound to a band corresponding in electrophoretic mobility to CII upon pretreatment with enzymatically active PAD. Additional electroblotted protein bands in lanes b and c with electrophoretic mobilities between the 67 kD and 83 kD marker bands (indicated by the thin arrow in Fig. 2C) represent the purified PAD enzyme that is not recognized by the serum autoantibodies, thereby clearly excluding that PAD-specific antibodies might account for the positive ELISA results with PAD-treated CII in Fig. 2B.

Figure 2.

Modulation of CII recognition by IgG antibodies following treatment of CII with PAD. (A) Binding of the murine mAb CIIC1, which is known to recognize the CII epitope aa 359–369 with critical dependence on the arginine residue 361, was tested by ELISA on microtiter plates coated with BSA (negative control), CII pretreated with active PAD or CII that had been incubated with the heat-inactivated enzyme (positive control). (B) ELISA results for IgG binding to CII pretreated with either catalytically active or heat-inactivated PAD are shown for representative RA sera (1:240 dilution). RA sera S1–S3 did not contain significant titers of IgG specific for native CII, in contrast to PAD-modified CII. Examples of RA sera that did not bind to CII at all (S4) or recognized both native and PAD-modified forms (S5) are shown. (C) Western blot was developed with an RA serum that specifically recognized PAD-modified CII in ELISA. The following protein samples were electroblotted to the different lanes: CII (lane A), CII + catalytically active PAD after 16 h incubation (lane B) and CII + heat-inactivated PAD after 16 h incubation (lane C). The electrophoretic mobilities of the marker, PAD (thin arrow) and CII (bold arrow) are indicated. The stained band in lane B represents the PAD-modified CII according to its electrophoretic mobility.

Autoantibody recognition of citrullinated synthetic CII peptides

Based on earlier findings showing that the B cell epitope C1III (CII 359–369) is immunodominant in human CII autoimmune responses 13 and harbors an arginine-containing motif shared with other CII epitopes (Fig. 2A), we decided to analyze the response to this epitope in more detail. Thus, we synthesized the peptide in both its native (C1III-P) and its citrullinated (citC1III-P) triple helical form within the constant frame of repetitive glycine-proline-hydroxyproline repeats (see Materials and methods and Fig. 3A). Using CD-spectra analysis, we could demonstrate for the modified synthetic peptide citC1III-P that the replacement of the arginine residues in the positions corresponding to aa 361 and 368 in CII by citrulline did not disturb the collagen-specific triple helical conformation (data not shown). For the synthetic collagen peptide C1III-P, we could show by amino acid analysis that PAD treatment leads to the enzymatic conversion of arginine residues to citrulline (results not shown), thus confirming our results on the modification of native collagen (Fig. 1). In similar agreement with analogous experiments on native CII (Fig. 2A), the PAD-treated synthetic collagen fragment C1III-P was also shown to escape immune recognition by the murine mAb CIIC1 but remained detectable by selected human RA sera that, in contrast to the murine antibody, did not bind to the unmodified peptide (results not shown). Fig. 3B shows the titration curves of IgG binding to the synthetic CII peptides C1III-P and citC1III-P for three selected RA sera, thereby providing experimental evidence for the existence of circulating IgG autoantibodies in RA patients that specifically recognize arginine to citrulline conversions in the CII domain aa 359–369 up to rather high serum dilutions. The differences between the titration curves for the wild-type and the citrullinated collagenomimetic peptides in the two upper panels emphasize this clear distinction. However, the figure also provides an example of serum that recognizes both variants of the synthetic CII peptide equally well.

Figure 3.

The structures of the triple helical CII peptides C1III-P and citC1III-P are shown in the upper panel (A). The amino acid residues are depicted in the one letter code (P* = hydroxyproline). citC1III-P differs from C1III-P by the replacement of all arginine residues (R) with citrulline (cit). (B) Titration curves for the recognition of wild-type (C1III-P) and citrullinated (citC1III-P) synthetic triple helical CII peptides by representative RA sera are shown. The sera contain high titers of IgG that specifically recognize either of the CII peptides variants (sera 1 and 2). However, some sera exhibit similar titration curves of IgG binding for both peptides (serum 3).

Analysis of circulating IgG autoantibodies with specificity for citrullinated and “wild-type” CII 359–369 in early RA patients

Sera from a cohort of early RA patients (n=286) were analyzed by ELISA for IgG binding to the synthetic CII peptides C1III-P and citC1III-P. As controls for RA specificity, additional cohorts of healthy donors (n=70) and patients suffering from other rheumatic diseases such as osteoarthritis (OA; n=72), SLE (n=147) and Wegener′s granulomatosis (WG; n=44) were analyzed in parallel. In Fig. 4A and 4B, box plots of the results for the different cohorts of patients are shown. The shaded areas indicate the mean of healthy controls + two SD. As depicted in Fig. 4A, sera from the early RA patients exhibited increased IgG binding to the unmodified peptide C1III-P in ELISA. In this respect the RA results differed significantly from healthy donors and from all investigated control cohorts of OA, SLE and WG patients (Fig. 4A). However, at a quantitative level, the mean ELISA values [19.3 arbitrary units (AU)] determined for the early RA sera exceeded the mean of the healthy controls (6.0 AU) only slightly, resulting in a relatively low prevalence of positive sera (19.7%) as defined by the threshold of healthy control + two SD. Thus, these data indicate that autoimmune responses with high titers of autoantibody specific for unmodified CII remain restricted to a subgroup of patients. A likewise considerably increased frequency of sera (40.4%) fulfilling the criterion for positive IgG autoantibody testing was identified for the citrullinated collagen peptide citC1III-P in the cohort of early RA patients (Fig. 4B). The mean value for citC1III-P in the early RA cohort was 81.8 AU and thus clearly raised above the normal range (mean + two SD in healthy donors =10.5 AU). The determined quantitative differences from sera from healthy donors or respective samples from control cohorts of SLE, OA and WG patients were statistically highly significant (Mann Whitney, p<10–10), indicating the RA specificity of the detected anti-citC1III-P autoantibodies. In the investigated early RA cohort, the prevalence of seropositivity for anti-citC1III-P autoantibodies (40.4%) was quite comparable to that determined for the anti-ccp-marker (49.8%) using the very sensitive commercial CCP2 test and a threshold of 25 U for positive results. Due to these similarities in the serological testing, it was not surprising to find a correlation (correlation coefficient =0.752, p<0.01) between the anti-ccp marker and the detection of anti-citC1III-P autoantibodies in the early RA cohort (Fig. 4C). However, Fig. 4C also provides evidence for the existence of RA sera containing IgG autoantibodies with specificity for either ccp or citC1III-P, thereby indicating that the hypothetical concept of immunological cross-reactivity cannot provide an adequate and sufficient explanation for the serological data.

Figure 4.

Serum IgG autoantibody recognition of the wild-type CII peptide C1III-P (A) and the citrullinated C1III peptide citC1III-P (B) in different cohorts of patients: osteoarthritis (OA, n=72), early rheumatoid arthritis (RA, n=286), Wegener′s granulomatosis (WG, n=44) and systemic lupus erythematosus (SLE, n=147). The ELISA results in AU are depicted as box plots. Each box represents the 25th to 75th percentiles. Lines outside the boxes represent the 10th and 95th percentiles. The bold lines inside the boxes highlight the median values. Mild (▵) and extreme (H) outliers are indicated. The shaded areas in (A) and (B) indicate the average + two SD determined in a cohort of healthy donors (n=70). p values indicate significant differences by the Mann-Whitney U-test. C) The correlation between citC1III-P ELISA (AU) on the x-axis and CCP2 ELISA (U) on the y-axis in 286 RA patients is shown. The correlation coefficient of 0.752 (p<0.01) suggests a statistical connection between the serological parameters. The shaded areas highlight the respective normal ranges (citC1III-P, see B; CCP2, manufacturer's definition for the CCP2 assay with a cut off at 25 U) and contain a variety of sera that exhibit seropositivity for only one of the two parameters under investigation.

Preabsorbance on CCP2 ELISA plates reveals serological specificity as well as cross-reactivity of anti-citrulline CII IgG antibodies

In order to investigate whether autoantibodies cross-react between CCP2 and citrullinated collagen, we performed preabsorption experiments. Representative RA sera were preabsorbed on commercially obtained CCP2-coated plates or on self-coated microtiter plates containing citC1III-P, linear control CII peptide (aa 359–369) or BSA for 2 h prior to the determination by ELISA of IgG binding to citC1III-P. Sera had been preselected for investigation of cross-reactive antibodies by the criterion of a positive reactivity to citC1III-P as well as CCP2 by ELISA. For quantification of antibody cross-reactivities, the relative inhibitions of IgG binding to citC1III-P by different preabsorption procedures were determined and are depicted in Fig. 5A. The decrease in OD in the citC1III-P ELISA resulting from a given preabsorption procedure was normalized to the maximum of reduction caused by preabsorbance on the specific antigen citC1III-P itself in order to obtain a relative measure of IgG cross-reactivity. The results expressed as relative inhibitory capacities in Fig. 5A clearly reveal the existence of RA autoantibodies that exhibit cross-reactivity between CCP2 and citC1III-P. For these sera CCP2- and citC1III-P-coated microtiter wells were shown to be equipotent in their abilities to preabsorb ELISA-detectable anti-citC1III-P IgG. Nonspecific multi-reactivity cannot account for this cross-reactive binding, since neither BSA- nor control CII peptide-coated wells exhibited any measurable preabsorbing capacity of anti-citC1III-P IgG in parallel experiments. A lack of citC1III-P inhibition (0%) was noted for IgG preabsorbance on BSA-coated wells in all RA sera tested (data not shown). For some sera hardly any detectable suppression of anti-citC1III-P ELISA reactivity could be obtained by CCP2 prebsorption procedures, indicating the occurrence of a specific humoral immune response to citC1III-P without CCP2 cross-reactivity. However, the identification of some patient sera exhibiting intermediate degrees of preabsorbable anti-citC1III-P IgG antibodies are reconcilable with the additional occurrence of specific as well as CCP2 cross-reactive anti-citC1III-P autoantibody responses in the same patient.

Figure 5.

Specificity and cross-reactivities of anti-citC1III antibodies. (A) Representative RA sera containing IgG antibodies specific for CCP2 and citC1III-P were preabsorbed on commercial CCP2-coated plates or microtiter plates self-coated with citC1III-P, a linear CII control peptide or BSA and subsequently assessed by ELISA for citC1III-P reactivity. The relative suppression of IgG binding to citC1III-P by the preceding absorption on different antigen matrices is expressed as the percentage of maximum that was obtained in all cases with the specific antigen citC1III-P (100% per definition). Preabsorbance on BSA did not result in any detectable inhibition (0%, data not shown). (B) Representative RA sera containing IgG antibodies to citC1III-P were preabsorbed on microtiter plates coated with citC1III-P, CII or BSA and subsequently assessed by ELISA for citC1III-P reactivity as described in (A). BSA did not exhibit any inhibitory effect (0%, data not shown).

Moreover, the detection of anti-citC1III-P antibodies that can also bind CCP2 inspired us to further investigate the possible occurrence of cross-reactive humoral immune responses that recognize not only the citrullinated collagen structures but also the respective native CII with the unmodified arginine residues. Sera that had been selected according to their pretested content of anti-citC1III-P were preabsorbed on BSA-, citC1III-P- or CII-coated microtiter plates and subsequently analyzed for anti-citC1III-P IgG by ELISA as described above. The results of the preabsorption experiments presented in Fig. 5B show that in addition to selective citC1III-P antibody recognition, varying degrees of IgG cross-reactivity to the native collagen are detectable.


In this study we show that CII as a cartilage-specific extracellular matrix component can serve as a substrate for PAD-catalyzed conversion of arginine residues to citrulline. This post-translational modification was demonstrated for the purified native protein as well as for synthetic triple helical collagen peptides. We further investigated the hypothesis that the PAD modification is likely to affect CII recognition by the adaptive immune system. It was previously shown that arginine residues seem to be of crucial importance for humoral CII autoimmunity in rodent arthritis models as well as in human RA 9, 13, 14, since the identified immunodominant CII domains recognized by circulating autoantibodies harbor the amino acid consensus motif R-G-hydrophobic 13. The occurrence of autoantibody responses, in particular those involving IgG antibodies, are usually bound to triggering by T cells. In this respect the formation of autoantibodies specific for CII is likely to follow the rule, as T cells reactive to CII, and in particular to post-translationally modified CII, occur in RA 15. Thus, both autoreactive T and B cells specific for CII escape deletion in RA, and these are apparently not properly regulated. It is not clear how immunological tolerance to CII is broken and whether this breakage of self tolerance represents an early disease-driving event or an epiphenomenon of late-stage disease 11, 12, 13. In this respect our present data provide first evidence that already at very early stages of RA, CII is targeted by autoantibody responses. It is of note that the triple helical collagen peptide C1III-P 9 recognized by the sera from our RA cohort represents an evolutionary conserved immunodominant native CII structure that was originally identified as the major target of arthritogenic humoral autoimmunity to CII in the mouse 9, 13. However, our data also suggest that the initial immune recognition of collagen in RA is not likely to be preferentially directed towards this earlier described native CII determinant but rather to citrullinated derivatives; close to one-half of the patients with a disease duration of less than 6 months already had detectable IgG that specifically bound to the immunodominant citrullinated CII peptide, while only 19.7% of individuals in the same RA cohort had autoantibodies recognizing the same CII peptide in its unmodified form. These data are in good agreement with the assumption that CII modification by PAD released from stressed and damaged cells in nonspecific inflammatory responses in the joint space 5 could lead to the formation of citrullinated neoepitopes and thus provoke a joint-specific immune response. Of particular importance is that the B cell specificity for citrullinated epitopes on CII could reflect a linked response in which T cells recognize a glycosylated epitope on another part of the same collagen molecule, at positions 260–271 15. This opens up a new possible explanation for the initiation of a response specific for citrullinated epitopes, as B cells binding and processing citrullinated CII will predominantly present the galactosylated CII epitope 260–271. In fact, we have recently shown that the CII epitope 260–271 is predominantly glycosylated in human cartilage and that B cells recognize, process and present this CII-derived glycopeptide 16. Clearly, the same type of linked presentation can be achieved through immune complex formation of citrullinated proteins and protein containing neoepitopes for T cell recognition. Intriguingly, such a scenario resembles what seems to occur in celiac disease, in which the trans-glutaminase-specific B cell may present gliadin with post-translationally modified T cell epitopes 17, 18. This does in fact explain the MHC class II association in celiaci 18, 19, and the known association of the response to the CII glycopeptide with the MHC class II shared epitope 15, 20 could represent a parallel explanation. Depending on the genetic background of the susceptible individuals, including the MHC class II haplotype, the linked recognition of the initial response to the altered cartilage matrix component could further direct the epitope spreading cascades. B cells recognizing citrullinated epitopes on various joint antigens or other antigen-presenting cells processing citrulline-containing immune complexes may also break T cell tolerance to other joint antigens and thus further spread the pathogenic autoimmune response.

The recognition of the citrullinated peptide citC1III-P by circulating IgG antibodies was demonstrated to be a RA-specific serological phenomenon, since the detection of significant autoantibody titers in control cohorts of patients suffering from other rheumatic diseases such as SLE, osteoarthritis or Wegener′s granulomatosis remained a rare exception. With regard to RA specificity, anti-citC1III-P antibodies did not differ from anti-CCP autoantibodies, which could also be detected in very few patients suffering from rheumatic conditions other than RA, in good agreement with earlier studies 21, 22. In addition, the frequency of positive testing for anti-citC1III-P antibodies (40.4%) in our RA inception cohort was only slightly reduced compared to anti-CCP IgG results (49.8%) that confirmed published studies in early RA 23. Accordingly, a significant correlation was obtained between serum anti-CCP-IgG and anti-citC1III-P-IgG titers in the RA patients. However, simple immunological cross-reactivity between anti-CCP and anti-citC1III antibodies, although detectable in selected RA sera in our investigation, is not a sufficient explanation for the occurrence of circulating IgG antibodies that bind in high serum dilutions to citrullinated CII in patients with early RA. Patient sera containing IgG specific for either CCP or citC1III-P could be identified, thereby providing unequivocal evidence for the existence of independent immune responses to PAD-modified determinants in CII and other self proteins.

Our findings suggest a new possible explanation that links the occurrence of an antibody response to citrullinated CII and a T cell response to glycosylated CII. As B cell tolerance is more easily broken (may occur by subclinical inflammation), this could result in breakage of tolerance of the better controlled T cells. These T cells are difficult to detect but do occur in RA and are restricted by shared epitope MHC class II molecules (such as DR4*0401) and predominantly recognize galactosylated CII 15. Thus, in this respect autoimmunity to cartilage-specific modified self might be a critical intermediate step that helps to lower the threshold for the final breakage of tolerance to other joint proteins in chronic inflammatory joint disease.

Materials and methods

Study populations

Sera from 286 RA patients [78% female, mean age 58±12 years (SD)] with an average disease duration of 6 months since the onset of first symptoms were collected in 20 collaborating rheumatological centers within the German Network for Competence in Rheumatology (see Acknowledgements). All patients fulfilled at least 4 of the 1987 revised classification criteria for RA 24 at study inclusion. Sera of control cohorts were collected at the Department of Internal Medicine III, Friedrich-Alexander University Erlangen, Germany and consisted of 147 patients who met the requirements of the American College of Rheumatology classification criteria for SLE 25, 72 patients with osteoarthritis of the hip or knee, 44 patients with Wegener′s granulomatosis and 70 healthy donors. The study protocol was approved by the review board of the Friedrich-Alexander-University Erlangen-Nuremberg, and informed consent was obtained from all individuals before entering the study.

Modifications of CII and synthetic triple helical collagen peptides with PAD

Type 2 rabbit skeletal muscle PAD (EC was purchased from Sigma-Aldrich (Taufkirchen, Germany) and used for enzymatic treatment of human CII (Chondrex, Redmond, WA) following a protocol that had been established for other substrates 26. Briefly, CII and synthetic CII peptides were incubated at a concentration of 20 μg/ml with 2 U/ml PAD at 37°C for 18 h in a buffer of 20 mM Hepes (pH 8.8), 0.3 M NaCl, 1 mM EDTA, 1 mM DTT and 10 mM CaCl2. The reaction was terminated by the addition of EDTA (10 mM final concentration).

Amino acid analysis

Amino acid analysis of PAD-treated CII and synthetic CII peptides was performed according to well-established standard procedures. Briefly, upon gas-phase hydrolysis of the CII probe in 6 M HCl at 110°C for 24 h, derivatization of amino acids for subsequent analytic chromatographical separation and fluorescence detection was performed using ortho-phtalaldehyde (OPA reagent, Agilent Technologies, Palo Alto, CA) for primary amino groups and 9-fluoromethylchloroformiat (FMOC reagent, Agilent Technologies) for secondary amino groups. Reverse phase separation of the derivatized amino acids (1 μl volume) was performed on a Hypersil-AA-ODS,200* column (2.1×200 mm, 5 μ; precolumn: ODS-Hypersil (C18), Agilent Technologies) using a high pressure liquid chromatograph (HP 1090, Hewlett Packard, Palo Alto, CA) equipped with a programmable fluorescence detector (Hewlett Packard). The separation column was equilibrated with 20 mM sodium acetate/0.018% triethylamine (pH 7.2) and eluted by the application of an acetonitrile-methanol gradient. Calibration of the retention times for the amino acids L-citrulline, L-proline and L-arginine was performed using the respective commercial standards (L-citrulline and L-proline, Fluka, Buchs, Switzerland; L-arginine, Merck, Darmstadt, Germany).

Synthesis of triple helical collagen peptides

Triple helical collagen peptides were synthesized as described earlier 9, 27. The sequences of the homotrimeric collagen peptides containing either the unmodified C1III epitope (C1III-P, CII aa residues 359–369 13) or its citrullinated variant (citC1III-P) are shown in Fig. 3a. MALDI-Tof analysis was performed on tryptic peptide fragments. The expected mass-to-charge ratios (m/z) of the fragments were found in the mass spectra. Peptide synthesis was performed with an ABI 430 peptide synthesizer using Fmoc chemistry. For MALDI-Tof analysis, a Kratos Kompakt IV mass spectrometer was used. Amino acid derivatives were obtained from Alexis Corporation (Switzerland) or NovaBiochem (Switzerland). Fmoc-Gly-p-benzyloxybenzyl alcohol resin (Fmoc-Gly-Wang-resin) and Fmoc-Lys(Dde)-OH (Fmoc-Lys(-1(4,4-dimethyl-2,6-dioxocyclohexylidene)-1-ethyl) were purchased from NovaBiochem (Switzerland).

ELISA for detection of collagen peptide specificity

Microtiter plates (Nunc, Wiesbaden, Germany) were coated at 4°C overnight with entire collagen type II (CII) (10μg/well), the synthetic CII peptides C1III-P or citC1III-P (1 µg/well) and blocked with 2% BSA in phosphate buffered saline (PBS) for 1 h. For blank controls, wells were coupled with 4 µg BSA/well. For the determination of CII-specific IgG, the sera were diluted in RIA buffer (1% BSA, 350 mM NaCl, 10 mM Tris-HCl pH 7.6, 1% vol/vol Triton X-100, 0.5% wt/vol Na-deoxycholate, 0.1% SDS) supplemented with 10% normal rabbit serum and incubated for 2 h at room temperature. A standard dilution of 1:120 was used if not otherwise stated. After washing (three times with PBS/0.05% vol/vol Tween-20), 100 µl peroxidase-conjugated rabbit anti-human IgG (Jackson ImmunoResearch Laboratories, West Baltimore, USA) diluted 1:5000 in RIA buffer was added to each well. Following a 1 h incubation period and another washing procedure, bound antibodies were detected using OPD solution (o-phenylenediamine dihydrochloride, Sigma, Taufkirchen, Germany) as the chromogenic substrate. The reaction was allowed to develop for 15 min and stopped by adding 50 µl 0.1 M HCl per well. Plates were read at a wavelength of 490 nm. All sera were tested in duplicate, and the results were averaged. A control serum was included on each plate to monitor plate-to-plate variation. For the evaluation of the ELISA data, the OD measurements were transformed into AU and calculated in relation to the respective standard control of the positive serum:

AU = [average ODCII peptide – average ODBSA]test serum / [average ODCII peptide – average ODBSA]standard serum × 100

The positive sera for C1III-P or citC1III-P selected for assay standardization gave a mean difference of 0.5 in OD between the wells coupled with the respective CII peptides and blanks (BSA-coupled) with an interassay variation of 5%. Thus, the CII peptide-specific reactivity of the standard sera was defined as 100 AU for the relative unit calculations of the serum samples.

ELISA for detection of cyclic-citrullinated peptide specificity (CCP2)

Anti-CCP2 antibody assays were performed using the commercially available diagnostic kit Immunoscan RA Mark 2 (Euro-Diagnostica, Arnhem, The Netherlands) according to the manufacturer's instructions with the cutoff at 25 units. The assay is based on solid phase-coupled cyclized citrullinated peptides that mimic naturally occurring epitopes in PAD-modified profilaggrin.

Western blot analysis

Western blot analysis was performed according to standard protocols. For detection of CII and PAD-modified CII, the nitrocellulose membrane was incubated with selected human sera or the CII-specific mouse mAb CIIC1,which recognizes the C1III epitope 13. As peroxidase-conjugated secondary antibodies, either a goat anti-mouse IgG (Bio-Rad, Hercules, CA) or an anti-human IgG (Jackson ImmunoResearch Laboratories) at a dilution of 1:1000 was used. Protein bands were visualized using chemiluminescence-enhancing reagents (Amersham Pharmacia Biotech, Freiburg, Germany).


We are grateful to the patients with RA, osteoarthritis, Wegener's granulomatosis and SLE as well as to the control individuals and the collaborating clinicians for participation in this study. Collaborating members of the multicenter study on RA of the “German Competence Network Rheumatology” (funded by the German Ministry of Research and Education, Grant 01 GI 9948) are the following persons: Prof. Dr. G. R. Burmester, Medizinische Klinik III mit Poliklinik Charité, Humboldt-Universität zu Berlin; Prof. Dr. J. Braun, Rheumazentrum Ruhrgebiet, St. Josefs-Krankenhaus Herne; Dr. E. Edelmann, Rheumatologische Schwerpunktpraxis Bad Aibling; Dr. A. Ehlert, Städtische Kliniken Duisburg; Prof. Dr. E. Gromnica-Ihle, Rheumaklinik Berlin-Buch; Prof. Dr. H. Häntzschel, Medizinische Klinik und Poliklinik IV Universität Leipzig; Dr. U. v. Hinüber, Hildesheim; Dr. J. Listing, DRFZ Berlin; Prof. Dr. H. Nüsslein, Medizinische Klinik I Städtisches Klinikum Dresden Friedrichstadt; Prof. Dr. H. H. Peter, Abt. Rheumatologie und Klinische Immunologie Klinikum der Albert-Ludwigs-Universität Freiburg; Dr. D. Pick, Grafschaft-Holzweiler; Prof. Dr. R. Rau, Evangelisches Fachkrankenhaus Ratingen; Prof. Dr. J. Sieper, Universitätsklinikum Benjamin-Franklin Berlin; Dr. F. Schuch, Rheumatologische Schwerpunktpraxis Erlangen; Dr. S. Wassenberg, Evangelisches Fachkrankenhaus Ratingen, Prof. Dr. H. Zeidler, Abteilung Rheumatologie Medizinische Hochschule Hannover; Prof. Dr. A. Zink, DRFZ Berlin. This work was supported by grants from the Deutsche Forschungsgemeinschaft (DFG, BU 584-2/1), The German Ministry for Research and Education (BMBF 01GI9948), the Swedish Association against Rheumatism, Arexis AB, the Swedish Medical Research Council, the Swedish Society for Medical Research, the Strategic Research Foundation and the EU project EUROMA. The authors are grateful to Christine Zech and Eva Bauer for excellent technical assistance.


  1. 1


  2. 2


  3. 3


  4. 4


  5. 5