A vast majority of patients with rheumatoid arthritis (RA) experience pain, functional deterioration, rigidity and work disability due to atrophy and irreversible joint destruction if not treated efficiently and early. Several different autoantibodies such as rheumatoid factor (RF) [1] and antibodies against citrullinated proteins/peptides (ACPAs), like anti-cyclic citrullinated peptide antibodies (anti-CCP) [2, 3] and antibodies against modified citrullinated vimentin (anti-MCV) [4] that have been identified in the serum of patients with RA have a negative prognostic impact on future joint destruction. In earlier studies of a Swedish RA cohort investigated before the systematic introduction of biological agents, we have demonstrated that RF, anti-CCP and anti-MCV detected in serum from patients with RA were associated with late inflammation and late increased rate of radiographic damage [5, 6]. In a recently published study we discovered that high levels of anti-native human collagen type II (anti-CII) antibodies in the same group of patients with RA were, in contrast, associated with laboratory measures of inflammation at disease onset [7], which can be explained by pro-inflammatory cytokine induction driven by surface-bound immune complexes (IC) containing anti-CII [8]. We therefore hypothesized that anti-CII antibodies were also associated with early joint destruction in this group of patients with RA.

To address this question, we performed the present study in which we focused on joint destruction in a prospective early RA cohort (n = 256), utilizing radiological data from multiple occasions, with parallel investigations of RF, anti-CCP, anti-MCV and anti-CII antibody serum levels.

Results about the prognostic impact of anti-CCP [6], anti-MCV [5] and anti-CII on acute inflammation [7], based on a somewhat different patient selection, have been published previously. The 256 patients included in this present analysis represent individuals for whom complete data for RF, anti-CCP, anti-CII and consecutive radiographs were available. Anti-MCV levels were analyzed at a later time point than the other analyses, when 2 out of 256 baseline serum samples were no longer available.

For the anti-CII ELISA that was performed as previously described [7], Maxisorb ELISA plates (Nunc, Roskilde, Denmark) were coated with human native CII (ELISA grade, Chondrex, Redmond, Washington DC, USA, diluted to 2.5 μg/ml in ice-cold PBS immediately prior to coating. Blocking was done with PBS with 1% ELISA grade bovine serum albumin. Serum samples were diluted at 1:100, and antibodies were detected with a F(ab')2 fragmented antibody against human gamma chain that had been pre-adsorbed against bovine proteins (Jackson, Cambridgeshire, UK). Internal controls were investigated together with patient samples on each occasion. The intra-assay coefficient of variation for the internal control close to the cutoff value was 15%.

Radiographs were scored blinded to treatment, in pairs (hands and feet), and in chronological sequence, by an experienced investigator (MCW) using the Larsen method [10]. In each case, 32 joints were scored. The original Larsen score [10] was modified slightly by excluding grade 1 [11], so that the scale became 0 to 4, as described previously [6]. Thus, the maximum possible score was 160. Larsen scores were obtained at baseline, and after one and two years. The change in Larsen score (Δ Larsen score) was calculated by subtracting the baseline score values from the respective annual scores.

Statistical analysis

Non-parametric tests were employed due to the bimodal appearance of anti-CII antibodies. The Mann-Whitney U test was used for the comparison of Larsen score between different groups. P values less than 0.05 were considered significant. The cutoff value was placed between anti-CII levels supporting or not supporting IC-induced cytokine production in vitro (200 AU/ml) as described previously [7, 8]. As can be seen from Figure 1, this cutoff also delimited a patient group with high discrete outlier values.

As has been shown earlier, autoantibodies traditionally associated with RA, such as RF [5], anti-CCP [6] and anti-MCV [5], were associated with late increased rates of radiographic destruction in the early RA cohort reported here. In this paper, we have demonstrated that patients with increased levels of anti-CII autoantibodies at the time of diagnosis have more radiographic destructions measured by median Larsen scores at baseline. This is accompanied by a significantly higher HAQ score among anti-CII-positive patients at the time of diagnosis, but not at any later time point.

We can thus distinguish two different phenotypes of patient with RA, viz. a group of patients with antibodies traditionally characterized by late inflammation and late increased radiographic destruction, and another group of patients who are positive for anti-CII antibodies, which is characterized by early inflammation [7] and early radiographic destruction.

The Larsen and the Sharp-van der Heide scoring systems are highly correlated [12, 13], especially when omitting Larsen grade 1 (soft tissue swelling and periarticular osteoporosis), as was advocated by Arvi Larsen in the modified Larsen scoring system [11]. It is therefore plausible that raised anti-CII levels are associated with breakdown of CII. Such an association might occur either due to a direct effect of anti-CII on cartilage as has been claimed by others [14], or indirectly via induction of pro-inflammatory cytokines by macrophages in the pannus tissue [7, 8].

The anti-CII cutoff used in the present study (200 U/mL) was based on the almost dichotomous distribution of anti-CII antibodies among patients with RA shown in Figure 1B, where anti-CII-positive and -negative subjects were clearly separated. Another argument for the use of this cutoff is that the dichotomous distribution of anti-CII among patients with RA clearly distinguished patients whose anti-CII levels permitted IC-induced cytokine induction from those with autoantibody levels too low to induce IL-1β or tumor necrosis factor (TNF) [8]. All anti-CII-negative patients had levels below 60 AU/mL, whereas the anti-CII-positive patients had levels between 471 and 3,520 AU/mL, thus clearly within the range of cytokine induction by the corresponding surface-bound IC [8].

Another major difference between ACPA and anti-CII concerns the kinetics of appearance. Whereas ACPA and RF levels slowly increase years before diagnosis [15, 16] and thereafter remain rather stable for years [3, 6], anti-CII levels seem to be highest around the time of the beginning of clinical disease. Elevated anti-CII levels do not precede clinical onset of RA [17], and after peaking around the time of diagnosis, levels drop during the first year, according to our own [7] and other [18] investigations. This short window of appearance of anti-CII in early RA, together with our earlier studies linking anti-CII to cytokine-driven inflammation using both functional in vitro experiments [8] as well as clinical follow-up studies [7], imply that the long-term effect of initially raised anti-CII levels might be limited, and that the major impact is around the time of disease onset. Initial appearance of anti-CII might in fact be prognostically advantageous. As initially raised anti-CII levels are associated both with raised CRP and ESR [7] and with augmented joint destruction around the time of diagnosis, and as these differences are no longer evident after one year, patients with RA and an early appearance of anti-CII might erroneously give an impression of a more severe prognosis than is actually the case. The general belief among rheumatologists is that a certain subgroup of patients with acute disease onset eventually have a rather good prognosis and need only rather limited treatment. This has been formally proven, as a recent study showed that both acute onset of RA and short symptom duration before inclusion in the study are associated with drug-free remission [19]. We speculate that at least some of these patients belong to the initially anti-CII-positive RA subgroup. This hypothesis should be investigated in other follow-up cohorts.

In this study, early RA was defined as < 12 months of symtoms before enrolment. Given the fact that anti-CII levels often drop sharply within the first months [7], we cannot rule out that some initially anti-CII-positive patients might have dropped to levels below the cutoff before diagnosis of RA.

The more active onset of destructive joint disease in anti-CII-positive RA patients might also initiate an earlier health care contact by patients after the first apearance of joint symtoms. This was also the case in our earlier larger study, where anti-CII-positive patients had a significantly shorter duration of joint symtoms before diagnosis of RA [7]. Acute onset RA is common among older individuals [20], and the median age among our anti-CII-positive patients was 12 years higher than among anti-CII-negative patients (Table 1).

One earlier study has also implicated an association between anti-CII and simultaneously occurring inflammation in RA. In the study by Kim et al [21], patients with anti-CII had, like our patients [7], higher levels of CRP and ESR than antibody-negative patients with RA, and also had higher levels of TNF and IL-6. We presume that both these cytokines have been produced by macrophages stimulated with anti-CII-containing IC. In our studies we have shown that anti-CII containing IC induces the production of TNF, IL-1β and IL8 from monocytes/macrophages [7, 8]. In these experiments we chose not to study IL-6, due obscuring of the in vitro signal by high basal IL-6 production. In other systems we have, however, found patient-derived IC to induce IL-6 in our in vitro models [22–24]. In contrast to our present investigation, the study by Kim et al could not show increased radiologic progression among anti-CII-positive patients with RA, a finding that can be explained by the fact that the patients had long-standing disease with a mean duration of 75.8 months [21]. In our study the significant difference in Larsen score between anti-CII-positive and -negative patients was only evident at the time of diagnosis, and was no longer visible after one year. Even if our patients were all included within 12 months of first symtoms, the mean interval between the first symtoms and diagnosis of RA was 5.8 months [7]. It is therefore possible that we would have found bigger differences between the groups if patients with shorter duration of joint symtoms had been included.

In conclusion, anti-CII antibodies are associated with early radiological destruction at the time of diagnosis, while autoantibodies traditionally associated with RA, such as RF, anti-CCP and anti-MCV are later associated with increased rates of radiographic destruction. This finding is in agreement with our earlier discovery that the same groups of antibodies are inversely associated with early [7] and late [5, 6] signs of inflammation in the same patient cohort. It is intriguing to note that the phenotypically opposite ACPA and anti-CII phenotypes, at least in this patient cohort, are also statistically inversely related (P = 0.04) as previously described [7]. We speculate that CII antibodies might stimulate macrophages and fibroblasts within the synovium to produce the metalloproteinases that are responsible for being the first enzymes to cleave the interstitial collagens, resulting in bone and cartilage destruction. We are currently performing studies to address this issue.

The occurrence of high levels of anti-CII antibodies characterizes a small group of patients with RA who have an elevated degree of joint destruction and elevated HAQ score at the time of diagnosis. Anti-CII characterizes an early inflammatory/destructive RA phenotype, in contrast to the late inflammatory/destructive phenotype that is associated with ACPA.