Multiplex Analyses of Antibodies Against Citrullinated Peptides in Individuals Prior to Development of Rheumatoid Arthritis




The presence of antibodies against cyclic citrullinated peptides has been demonstrated to precede the onset of symptoms of rheumatoid arthritis (RA) by several years. The aim of this study was to analyze antibodies against 10 citrullinated autoantigen-derived peptides for reactivity before the onset of RA symptoms.


A case–control study was conducted within the Medical Biobank of Northern Sweden. The study was performed in 409 individuals, 386 of whom donated 717 blood samples before the onset of symptoms of RA (pre-patients). The median period of time predating the onset of RA was 7.4 years. A total of 1,305 population-based control subjects were also studied. Antibodies to 10 citrullinated peptides, fibrinogen α573 (Fibα573), Fibα591, Fibβ36–52, Fibβ72, Fibβ74, α-enolase (citrullinated α-enolase peptide 1 [CEP-1]), triple-helical type II collagen peptide C1 (citC1III), filaggrin, vimentin 2–17 (Vim2–17), and Vim60–75, were analyzed using a microarray system.


The fluorescence intensity of antibodies against Fibβ36–52, Fibβ74, CEP-1, citC1III, and filaggrin was significantly increased in pre-patients compared with controls (P < 0.001). The levels of the earliest-detectable antibodies (Fibα591 and Vim60–75) fluctuated over time, with only a slight increase after the onset of disease. The frequency of antibodies against Fibβ36–52, CEP-1, and filaggrin increased gradually, reaching the highest levels before symptom onset. The frequency of a cluster of antibodies, citC1III, Fibα573, and Fibβ74, increased only slightly before the onset of symptoms but increased prominently after disease onset. The odds ratio for the development of RA in individuals expressing both CEP-1 and Fibβ36–52 antibodies (using data from samples obtained <3.35 years predating symptom onset) was 40.4 (95% confidence interval 19.8–82.3) compared with having either antibody alone.


Development of an immune response toward citrullinated peptides is initially restricted but expands with time to induce a more specific response, with levels, particularly those of antibodies against CEP-1, Fibβ36–52, and filaggrin, increasing during the predating time period closer to the onset of symptoms.

Rheumatoid arthritis (RA) is a chronic autoimmune disease characterized by joint inflammation involving the synovial tissue and eventually leading to destruction of cartilage and bone. The etiopathogenic processes leading to the development of RA are not fully understood. We and other investigators have shown that the presence of anti–citrullinated protein antibodies (ACPAs), detectable as anti–cyclic citrullinated peptide (anti-CCP) antibodies of the IgG, IgA, and IgM isotypes, precedes the development of RA by several years (1–4). Multiple studies have demonstrated that levels of proinflammatory cytokines and chemokines are elevated prior to the onset of symptomatic RA (5, 6).

However, because of the lack of joint disease symptoms, the initiation and location of ACPA production before disease onset are uncertain (1). Systemic autoimmunity has been suggested to precede the development of synovial inflammation in ACPA-positive patients with arthralgia (7). It has been proposed that the development of ACPAs before disease onset may result from an inflammation-initiated process at other locations, such as the lungs (8). Furthermore, agents that cause inflammation in the lungs, including cigarette smoke and silica dust, are associated with the occurrence of ACPA-positive RA (9, 10). Inflammation in the lungs, possibly due to infection, has been reported to precede the development of ACPA-positive and/or rheumatoid factor–positive joint disease, even before the presence of any signs of synovitis (9, 10). Additionally, an increased frequency of periodontitis and the presence of Porphyromonas gingivalis has been reported in ACPA-positive patients with RA and is suggested to contribute to the emergence of an autoimmune response to citrullinated peptides (11).

An alternative series of citrullinated proteins/peptides from antigens that are potential targets of immune reactions important for the development of ACPA-positive RA has been identified in inflamed tissue, mostly in joints (12–16). Currently, the major proteins that have been identified are fibrinogen (12), α-enolase (13, 14), vimentin (15), and type II collagen (16).

Among patients with RA, antibodies reactive against several citrullinated peptides or proteins, e.g., against α-enolase (citrullinated α-enolase peptide 1 [CEP-1]) (13), vimentin, fibrinogen (17, 18), and triple-helical type II collagen peptide C1 (citC1III) (16, 17), have been detected. These antibodies display both some cross-reactivity and unique reactivities against single citrullinated peptide epitopes (17, 19). However, there are different associations between HLA shared epitope alleles and subsets of positive ACPAs in patients with RA that are defined by different antibody profiles (20–23). This suggests the involvement of diverse mechanisms in the development of ACPAs (20–23). The role of individual antibodies against different citrullinated antigens in the actual triggering and maintenance of arthritis is largely unknown. A better understanding of when the separate antibodies appear (before and/or at the onset of disease) would be an important step in defining the possible pathogenic role of the different ACPAs.

In the present study, using blood samples from individuals (designated pre-patients) stored in the Medical Biobank of Northern Sweden, the various patterns of antibodies directed against 10 different citrullinated peptides during the time period predating the onset of symptoms of joint disease were analyzed using a recently described multiplex assay system (24). In addition, we analyzed blood samples from 204 of these individuals who had provided blood samples at the time of RA diagnosis, and samples from matched population controls maintained in the same biobank.



A case–control study was performed using the Medical Biobank of Northern Sweden cohorts and the Maternity cohort. The cohorts within the Medical Biobank are population based, and all adult individuals residing in the county of Västerbotten are continuously invited to participate. Recruitment, blood sampling, and storage conditions (−80°C) have been described previously (1). The Maternity cohort is a collection of serum samples obtained from pregnant women in northern Sweden who have undergone screening for rubella; these samples are stored at −20°C (1). For coanalysis with the registers of the Medical Biobank cohorts, we used the register of the patients at the Department of Rheumatology, University Hospital, Umeå, comprising patients who fulfilled the American College of Rheumatology 1987 classification criteria for RA (25) with the date for the onset of symptoms of joint disease given.

In all, 409 individuals (73 men and 336 women; designated as pre-patients) were identified as having donated a total of 742 blood samples (481 from the Biobank cohorts and 261 from the Maternity cohort) before the onset of any symptoms of joint disease. Samples from 23 individuals (9 from the Biobank cohorts and 14 from the Maternity cohort) could not be used (25 samples from 18 individuals could not be located; in 2 cases, the storage tubes were empty; and 2 individuals, each of whom donated 2 samples, were excluded as being misdiagnosed), resulting in a study population of 386 pre-patients (total of 717 samples). Up to 7 samples were identified for any given individual, and the median period of time predating the onset of symptoms was 7.4 years (interquartile range [IQR] 3.3–12.6 years). At least 1 sample was identified for 386 individuals, 2 samples for 178 individuals (24.8%), 3 samples for 92 individuals (12.8%), 4 samples for 38 individuals (5.3%), 5 samples for 14 individuals (2.0%), 6 samples for 6 individuals (0.8%), and 7 samples for 3 individuals (0.4%) (additional information available from the corresponding author upon request).

Control subjects were randomly selected from the same Biobank cohorts as the pre-patients and matched for sex and age at the time of blood sampling. A total of 1,308 control subjects (367 men and 941 women) were selected; however, 1 man and 2 women were excluded due to a prior diagnosis of RA revealed by coanalysis of registers with a given date for the onset of symptoms.

All donors were classified as either a nonsmoker or an ever smoker. Of the individuals identified as pre-patients, 204 had also subsequently provided blood samples when attending the clinic at the time of being diagnosed as having RA. The median time between the onset of symptoms and diagnosis of RA was 7.2 months (IQR 4.7–10.6 months). The demographic data on the presymptomatic individuals, patients with RA, and control subjects are shown in Table 1.

Table 1. Characteristics of the presymptomatic patients (pre-patients), patients with rheumatoid arthritis (RA), and control subjects*
 Pre-patients (n = 386)RA patients (n = 204)Controls (n = 1,305)
  • *

    Pre-patients were defined as individuals in whom the onset of joint symptoms had not yet occurred. The HLA–DR shared epitope (SE) was defined as 0401/0404/0405/0408. IQR = interquartile range.

  • Median age, as calculated for all samples when collected (n = 717).

Female sex, %827772
Age, median (IQR) years49.9 (30.1–58.8)56.5 (48.7–63.7)51.0 (40.3–60.2)
Ever smoker, no./total (%)238/358 (66.5)136/200 (68)522/1,106 (47.2)
HLA–DR SE carrier, no./total (%)222/344 (64.5)129/204 (63.2)122/294 (41.5)
PTPN22 T variant carrier, no./total (%)130/364 (35.8)70/202 (34.7)237/998 (23.7)

HLA–DRB1 genotyping for 0401/0404/0405/0408 and the PTPN22 1858C/T polymorphism (rs2476601) was performed as previously described (26, 27).

The Regional Ethics Committee at the University Hospital, Umeå, Sweden approved this study, and all participants provided written informed consent when donating samples.

Multiplex assay.

Serum/plasma samples were analyzed for the presence of IgG-specific ACPA, using a custom-made microarray based on the ImmunoCAP ISAC system (PhaDia), containing 10 different citrullinated peptides and their native arginine-containing counterpart. A full description of this technology with extensive validation of the chip-based technique in comparison with enzyme-linked immunosorbent assay (ELISA)–based technology and diagnostic performance was recently published (24). The citrullinated antigens investigated were fibrinogen α573 (Fibα573), Fibα591, Fibβ72, Fibβ74 (28), Fibβ36–52 (20), CEP-1 (14, 29), citC1III (16, 30), filaggrin (31, 32), vimentin 2–17 (Vim2–17), and Vim60–75 (15, 20, 33) (for review, see ref.24). The CEP-1 and filaggrin peptides were circular, whereas the remaining peptides were used in a linear form (24). No comparison of the diagnostic performance of these linear peptides in circular forms has been undertaken. In all but 1 case, the difference in fluorescence intensity between the citrullinated peptide and the arginine-containing peptide was calculated for all samples studied, and the resulting delta value was used in the subsequent calculations. The exception was the citC1III peptide, for which the uncorrected values were used, because the ArgC1 peptide is an autoantigen in its own right, with conformational epitopes differing from those of the citC1III peptide (30). Overall, results were available for 691 samples from 374 pre-patients (before the onset of symptoms), 199 patients with RA, and 1,278 control subjects.

In the present study, the cutoff levels for each antibody were defined as the antibody reactivity expressed at the optimal level of sensitivity and specificity, using receiver operating characteristic (ROC) curves based on the concentrations observed in the patients with RA included in this study and the controls from the Medical Biobank.

Anti–CCP-2 antibody analysis.

Detection of anti-CCP antibodies was successfully performed in all but 4 samples, using ELISAs according to the manufacturer's instructions (Euro-Diagnostica). The cutoff for positivity was set at 25 arbitrary units/ml.

Statistical analysis.

Continuous data were compared (pre-patients versus controls or versus patients) by nonparametric analysis using the Mann-Whitney U test for 2 groups, the Kruskal-Wallis test for comparing several groups, and the Friedman's test for matched pairs (e.g., several samples from the same individual). The chi-square test for trend was used to compare, over time, categorical data. Relationships between categorical data (positive versus negative) were compared using chi-square analysis or Fisher's exact test as appropriate. Correlation analyses were performed using Spearman's rank correlation. Logistic regression analyses were performed to identify associations between combinations of ACPAs and the development of RA and are presented as odds ratios (ORs) and 95% confidence intervals (95% CIs). Backward stepwise regression analyses were performed to determine the strongest ACPA reactivities associated with the future development of RA. The time point of the first appearance between individual ACPAs was compared using a Mann-Whitney U test for 2 groups. Considering the study to be explorative, P values less than or equal to 0.05 were considered significant.

Statistical calculations were performed using SPSS for Windows, version 20. Sensitivity, specificity, ORs, and 95% CIs were calculated with the XLSTAT program (version 2011.4.01) in Microsoft Excel 2010 (Addinsoft).


Fluorescence intensities of the detected ACPAs.

Fluorescence activity, indicative of antibody levels in multiplex assays, was significantly increased for several of the citrullinated peptides, i.e., Fibβ36–52, Fibβ74, CEP-1, citC1III, and filaggrin, in pre-patients when all 717 samples were compared with controls (P < 0.001 for all 5 analytes). The fluorescence intensities were further increased after RA had developed in the pre-patients. Levels of all measured antibodies were significantly increased in samples after a diagnosis of RA compared with those in samples before the onset of symptoms (P < 0.05–0.001), when analyzed at a group level (Mann-Whitney U test) or compared as matched pairs (Wilcoxon's rank sum test), except for anti–Vim2–17. Antibodies against all of the citrullinated peptides investigated were significantly increased in the patients with RA compared with population-based control subjects (P < 0.05–0.001) (additional information available from the corresponding author upon request).

Frequencies of antibodies in predating samples.

The frequency of ACPAs in the pre-patients was highest for anti–Fibβ36–52, anti–CEP-1, and antifilaggrin antibodies (30.2%, 29.4%, and 26.2%, respectively), when analyzed once for each individual, e.g., among those ever being positive and at a specificity defined by the ROC curve for each antibody (Table 2). Analysis of samples obtained at the time points representing the quartile closest to the onset of symptoms (<3.35 years) showed that some of the antibodies were present at a high frequency during that quartile, and that the presence of these antibodies was associated with a high risk (OR) for developing RA. The most prominent of these antibodies were against CEP-1 (frequency 39.7%; OR 12.5 [95% CI 8.4–18.5]), followed by Fibβ36–52 (frequency 33.9%; OR 9.7 [95% CI 6.5–14.5]) and filaggrin (frequency 31.0%; OR 14.7 [95% CI 9.3–23.1]). During this time preceding the onset of RA, the frequency of anti–CCP-2 antibodies analyzed in parallel by ELISA was 48.0% (OR 45.5 [95% CI 28.1–73.8]) (Table 2).

Table 2. Frequencies of antibody positivity and odds ratios for development of rheumatoid arthritis (RA) in relation to antibodies against citrullinated peptides in presymptomatic patients (pre-patients; ever positive and stratified by quartiles of time predating onset of RA) and in individuals after the diagnosis of RA, and specificity of antibody positivity for RA according to the receiver operating characteristic curve for each antibody*
Ever positive (n = 386/374)≤−12.56 years (n = 180/166)−12.55 to −7.40 years (n = 178/175)−7.39 to −3.36 years (n = 180/176)−3.35 to −0.2 years (n = 179/174)RA patients (n = 200/199)
Frequency, % (95% CI)OR (95% CI)Frequency, % (95% CI)OR (95% CI)Frequency, % (95% CI)OR (95% CI)Frequency, % (95% CI)OR (95% CI)Frequency, % (95% CI)OR (95% CI)Frequency/sensitivity, % (95% CI)OR (95% CI)Specificity, % (95% CI)
  • *

    The antibodies are presented in consecutive order according to the odds ratio (OR) (95% confidence interval [95% CI]) for the frequency of ever being positive. Group values (n/n) are the no. of individuals analyzed for anti–cyclic citrullinated peptide 2 (anti–CCP-2) using enzyme-linked immunosorbent assay/individuals analyzed using multiplex assay. CEP-1 = citrullinated α-enolase peptide 1; citC1III = triple-helical type II collagen peptide C1.

CCP-233.59 (29.1–38.4)24.88 (16.06–38.54)9.44 (5.93–14.72)5.13 (2.75–9.59)19.66 (14.48–26.18)12.04 (7.07–20.49)29.44 (23.28–36.51)20.53 (12.45–33.84)48.04 (40.85–55.33)45.49 (28.06–73.75)74.5 (68–80.04)143.72 (87.32–236.54)98.01 (97.08–98.64)
Filaggrin26.2 (22.01–30.9)11.59 (7.81–17.19)6.63 (3.65–11.64)2.32 (1.17–4.57)22.86 (17.26–29.67)9.67 (6.01–15.55)15.91 (11.22–22.11)6.17 (3.69–10.32)31.03 (24.64–38.28)14.68 (9.33–23.1)46.23 (39.45–53.16)28.06 (18.35–42.89)97.03 (95.93–97.83)
Fibrinogen β7415.78 (12.43–19.85)9.39 (5.81–15.17)5.42 (2.77–10.17)2.87 (1.34–6.16)14.29 (9.84–20.33)8.35 (4.7–14.84)10.23 (6.52–15.7)5.71 (3.07–10.62)17.82 (12.83–24.25)10.87 (6.27–18.83)34.17 (27.94–41.02)26.02 (15.96–42.42)98.04 (97.11–98.68)
Fibrinogen β36–5230.21 (25.79–35.06)8.21 (5.88–11.46)10.24 (6.44–15.91)2.16 (1.24–3.77)21.71 (16.24–28.45)5.26 (3.4–8.14)17.05 (12.19–23.36)3.9 (2.45–6.2)33.91 (27.3–41.24)9.73 (6.52–14.53)64.82 (57.95–71.11)34.96 (23.83–51.28)95.0 (93.64–96.06)
CEP-129.41 (25.03–34.23)7.9 (5.66–11.05)7.23 (4.1–12.37)1.48 (0.79–2.77)17.71 (12.76–24.12)4.08 (2.58–6.47)21.59 (16.15–28.3)5.22 (3.38–8.08)39.66 (32.69–47.08)12.47 (8.41–18.47)67.34 (60.53–73.46)39.1 (26.55–57.6)95.0 (93.64–96.06)
CitC1III15.78 (12.43–19.85)4.51 (3.04–6.68)6.63 (3.65–11.64)1.71 (0.88–3.31)13.71 (9.36–19.69)3.82 (2.3–6.37)8.52 (5.17–13.72)2.24 (1.24-4.05)17.24 (12.34–23.62)5.01 (3.1–8.1)32.2 (26.1–39.0)11.4 (7.59–17.13)96.01 (94.78–96.96)
Fibrinogen β7212.03 (9.11–15.76)4.46 (2.86–6.97)4.82 (2.35–9.42)1.65 (0.77–3.54)6.86 (3.89–11.75)2.4 (1.24–4.64)6.82 (3.87–11.69)2.39 (1.24–4.61)10.92 (7.06–16.54)4 (2.26–7.07)14.57 (10.32–20.22)5.57 (3.36–9.23)97.03 (95.93–97.83)
Fibrinogen α57311.23 (8.41–14.87)3.04 (1.99–4.65)1.2 (0.07–4.63)0.29 (0.08–1.05)5.71 (3.04–10.37)1.46 (0.74–2.89)8.52 (5.17–13.72)2.24 (1.24–4.05)12.07 (7.99–17.85)3.3 (1.94–5.61)34.17 (27.94–41.02)12.49 (8.34–18.7)96.01 (94.78–96.96)
Vimentin 60–7513.37 (10.28–17.23)2.93 (1.99–4.32)6.02 (3.21–10.91)1.22 (0.62–2.38)7.43 (4.32–12.44)1.52 (0.83–2.8)10.23 (6.52–15.7)2.16 (1.26–3.72)10.92 (7.06–16.54)2.33 (1.36–3.96)29.15 (23.28–35.83)7.8 (5.26–11.58)95.0 (93.64–96.06)
Vimentin 2–178.02 (5.66–11.27)2.85 (1.74–4.64)2.41 (0.76–6.3)0.81 (0.3–2.17)5.14 (2.63–9.66)1.77 (0.85–3.66)2.84 (1.06–6.71)0.95 (0.38–2.36)8.62 (5.23–13.87)3.08 (1.67–5.68)11.06 (7.38–16.26)4.06 (2.36–6.98)97.03 (95.93–97.83)
Fibrinogen α5918.56 (6.11–11.88)1.77 (1.14–2.75)3.61 (1.52–7.89)0.71 (0.31–1.62)8.57 (5.2–13.79)1.78 (1–3.17)5.68 (3.02–10.31)1.14 (0.58–2.24)6.9 (3.91–11.82)1.41 (0.75–2.63)14.07 (9.9–19.66)3.11 (1.94–4.96)95.0 (93.64–96.06)

Notably, the frequencies of antifilaggrin, anti-Fibβ74, anti–Fibβ36–52, anti–CEP-1, anti-citC1III, anti-Fibα573, and anti–Vim2–17 antibodies increased significantly during the whole predating time (stratified by quartiles before the onset of RA) (P < 0.05–0.001 by chi-square test for trend). The frequencies of positivity for antibodies against Fibβ72, Vim60–75, and Fibα591 were, on a group level, unchanged during the predating time (P > 0.05, by chi-square test for trend).

The highest frequency of positivity for antibodies against citrullinated peptides among pre-patients during the time period <3.35 years before disease onset (the quartile closest to onset of symptoms) could be increased to 59.2% by combining the outcome of anti–CCP-2 frequency (48%) with positivity for Fibα72 and CEP-1. The cumulative frequency of each antibody found to be positive is presented in Figure 1. The graph clearly illustrates that antibodies against CEP-1, Fibβ36–52, filaggrin, and CCP-2 were, as they progressed, grouping together years before the onset of symptoms.

Figure 1.

A, Cumulative percentage of positivity for antibodies against citrullinated peptides in blood samples from individuals (n = 375) obtained during the time period predating the onset of symptoms of rheumatoid arthritis (RA) and at diagnosis of RA. B, Cumulative percentage of positivity for antibodies against citrullinated peptides in individuals having donated multiple samples (n = 696) during the time period predating the onset of symptoms of RA and at diagnosis of RA. CCP-2 = cyclic citrullinated peptide 2; CEP-1 = citrullinated α-enolase peptide 1; Fib = fibrillarin; VIM = vimentin; citC1 = triple-helical type II collagen peptide C1.

Frequencies of antibodies in patients with RA.

After disease onset, when RA was diagnosed, the frequency of antibodies against the various citrullinated peptides was highest for anti–CEP-1 (67.3%), followed by antibodies against Fibβ36–52 (64.8%) and filaggrin (46.2%). All ACPA frequencies, except those for Fibβ72 and Vim2–17, increased significantly after the onset of RA, when all samples were compared with those from pre-patients (P < 0.01–0.001) (additional information available from the corresponding author upon request). All antibodies assayed were significantly increased in patients with RA compared with controls (Table 2).

Antibody development/epitope spreading before onset of symptoms and at diagnosis of RA.

Overall, at least 1 of the 10 ACPAs analyzed with the multiplex assay was present in 50.1% (346 of 691) of the samples from the pre-patients, in 81.9% (163 of 199) of the samples from the patients with RA, and in 23.2% (297 of 1,278) of the samples from the controls. However, among the samples from the controls, only 7.7% (98 of 1,278) had ≥2 detectable ACPAs, while 27.4% (189 of 691) of the samples from the pre-patients and 71.9% (143 of 199) of the samples from the patients with RA were positive for ≥2 antibodies. The cumulative percentage of positivity for the different numbers of citrullinated antibodies over time is presented in Figure 2A.

Figure 2.

A, Cumulative percentage of positivity for different numbers of antibodies against citrullinated peptides present before the onset of symptoms and at diagnosis of rheumatoid arthritis (RA) (referred to as 0.6 years after onset of symptoms). B, Number of antibodies against citrullinated peptides present in relation to predating time (in years). 0 years refers to the time point at onset of symptoms.

The number of positive ACPAs increased significantly over time until the onset of symptoms, when data were analyzed on a group basis and after inclusion of data acquired at the time of RA diagnosis (P < 0.001 for both) (Figure 2B). There was also a significant increase in the number of positive ACPAs when data were calculated on an individual basis (P < 0.001) (data not shown).

Progress of antibody levels, as measured by fluorescence intensity, during the time predating symptoms and at the time of diagnosis of RA.

In the individual patients who donated multiple samples prior to the onset of symptoms (additional information available from the corresponding author upon request), the fluorescence intensity of the antibodies during the pre-patient phase increased significantly as the samples were collected at time points closer to the onset of symptoms (P < 0.05–0.001 by Friedman's test), except for antibodies against filaggrin, Fibβ72, Vim2–17, Vim60–75, and Fibα591. These latter antibodies did not increase significantly during the predating time (Figures 1A and B; additional information available from the corresponding author upon request) and their levels were more variable when compared with antibodies to the other peptides (Table 3).

Table 3. Time before the onset of symptoms in all antibody-positive samples by the first appearance of each antibody and in all antibody-positive samples overall, and frequency of stable positive samples in subsequent predating samples, analyzed in 374 individuals with 696 samples*
AntibodyFirst positive sampleAll positive samplesStable positive results in subsequent samples, no./no. assessed (%)
 No. of years predating symptom onset No. of years predating symptom onset
No. of samplesMedian (IQR)Mean ± SDNo. of samplesMedian (IQR)Mean ± SD
  • *

    The antibodies are presented in the order of first appearance and positivity in subsequent samples. IQR = interquartile range; citC1III = triple-helical type II collagen peptide C1; CEP-1 = citrullinated α-enolase peptide 1; CCP-2 = cyclic citrullinated peptide 2.

Fibrinogen α591326.0 (7.6)7.4 ± 5.3437.4 (7.2)7.3 ± 4.94/17 (23.5)
Vimentin 60–75514.8 (8.3)7.4 ± 6.1604.7 (8.3)7.1 ± 5.96/18 (33.3)
Fibrinogen β72455.9 (8.5)6.5 ± 5.1515.8 (8.3)6.4 ± 4.93/22 (13.6)
citC1III595.9 (9.3)6.4 ± 5.1805.6 (8.0)6.2 ± 4.813/27 (48.1)
Fibrinogen β74594.9 (7.9)6.7 ± 5.7834.9 (7.6)6.2 ± 5.217/25 (68.0)
Filaggrin985.2 (8.2)6.3 ± 4.71334.6 (7.8)5.8 ± 4.417/44 (38.6)
Fibrinogen β36–521134.8 (8.5)6.3 ± 5.31444.6 (7.9)6.0 ± 5.124/37 (64.9)
CEP-11103.5 (7.9)5.5 ± 5.01503.6 (6.8)5.3 ± 4.729/41 (70.7)
Fibrinogen α573423.9 (6.9)4.9 ± 4.2483.8 (6.3)4.7 ± 4.15/13 (38.5)
Vimentin 2–17303.5 (7.6)5.2 ± 4.4334.3 (7.5)5.6 ± 4.53/12 (25.0)
CCP-21323.8 (6.0)5.7 ± 4.61913.7 (6.4)5.1 ± 4.439/48 (81.3)

Anti-Fibα573 antibodies were also of low frequency but with stable positivity (23.1%), and the levels increased significantly during the predisease time period. When the samples collected following disease onset were included in the analyses, all levels of all antibodies increased significantly, except for anti-Fibα591 (P < 0.05–0.001 by Friedman's test). Analyses of the fluorescence intensity of samples at a group level from the whole period predating symptoms, whether or not data after the onset of RA were included, showed a significant increase in the levels of all 10 ACPAs analyzed (P < 0.05–0.001).

Time-related development of antibodies.

Although only one-half of the initial samples were ACPA positive, all samples were evaluated for subsequent development of antibodies. The distribution of the initially positive samples was proportionally equivalent among all of the antibodes at first appearance. The time point of first appearance of individual ACPAs either differed significantly or was close to being significantly different, i.e., antibodies against Fibα591 when compared with anti–CEP-1 (P < 0.05), anti–CCP-2 (P < 0.05), anti-Fibα573 (P < 0.05), and anti-Vim2–17 (P = 0.076). Anti–Vim60–75 appeared significantly earlier than anti–CEP-1 (P < 0.05), anti-Fibα573 (P < 0.05), and anti–CCP-2 (P < 0.05) and nearly significantly earlier than anti–Vim2–17 (P = 0.087). The 2 antibodies appearing first, Fibα591 and Vim60–75, had among the lowest stable positivity in consecutive samples, and the frequency of positivity hardly increased closer to symptom onset (Table 3 and Figure 1), although the frequency of antibodies against Vim60–75 increased significantly after RA was diagnosed.

In a separate analysis in which we included only those individuals who had an antibody-negative sample that preceded a sample showing a positive reaction, the predating time was evaluated. Thus, with the numbers of comparable positive samples reduced, the only significant difference in the time point of first appearance was between antibodies against Fibα591 and anti–CEP-1 antibodies (P < 0.05). The sample sizes were too small to assess the significance of results concerning the order of appearance of the other antibodies.

Relationships between antibodies against CEP-1, Fibβ36–52, filaggrin, and CCP-2.

The ACPAs most frequently coexisting within the same sample from individuals before the onset of symptoms were anti–CEP-1 and anti–Fibβ36–52, occurring in 20% of individuals, followed by the coexistence of antifilaggrin with either of these antibodies in 16% of individuals. However, the strongest coexistence within the same sample was between anti–CCP-2 and either CEP-1 or Fibβ36–52 antibodies, each occurring in 24% of individuals, or between anti–CCP-2 and antifilaggrin, occurring in 19%.

Calculations based on all samples from pre-patients, irrespective of time point, revealed a high covariance when considering anti–CEP-1, anti–Fibβ36–52, and antifilaggrin versus anti–CCP-2; i.e., 77% of individuals ever positive for anti–CCP-2 were also ever positive for CEP-1. The corresponding values for the other 2 antibodies were 76% and 72%, respectively. For anti–CEP-1 versus anti–Fibβ36–52 and anti–CEP-1 versus antifilaggrin, the covariance between the different antibodies was 50–59%. Stepwise regression analysis including all antibodies from the multiplex assay and stratified for the period of time predating symptoms (>3.35 years versus ≤3.35 years) was performed. Although the antibodies coexisted, significant predictive values were observed for CEP-1, Fibβ36–52, filaggrin, and Fibβ74.

The OR for the development of RA in individuals expressing the combination of anti–CEP-1 and anti–Fibβ36–52 was 40.4 (95% CI 19.8–82.3) when the data from the quartile <3.35 years were analyzed using simple logistic regression modeling. As a comparison, this OR for the development of RA was similar to that obtained when assessing anti–CCP-2 antibody (45.5 [95% CI 28.1–73.8]). Combinations of the 3 antibodies (i.e., against Fibβ36–52, CEP-1, and filaggrin), yielded a very high OR for the development of RA when compared with negativity for these antibodies or positivity for only 2 of them, as analyzed using simple regression analyses stratified for the predating time in quartiles (Table 4). The combination of the 3 antibodies was present in only 5 of the control samples (0.4%) compared with 50 of the pre-patient samples (9%) and 76 of the patient samples (44.7%). However, the OR was even higher when the presence of anti–CCP-2 was included, instead of antifilaggrin, in the triplet in combination with Fibβ36–52 and CEP-1 (Table 4).

Table 4. Results of simple logistic regression analyses performed within each predating time quartile for the predictive capacity of antibodies against fibrinogen β36–52 (Fibβ36–52), citrullinated α-enolase peptide 1 (CEP-1), and filaggrin and for the combination of antibodies against Fibβ36–52, CEP-1, and cyclic citrullinated peptide 2 (CCP-2)*
 No. of years predating symptom onset
  • *

    Values are the odds ratio (95% confidence interval).

Fibβ36–52, CEP-1, and filaggrin
 Negativity for all 3 antibodiesReferentReferentReferentReferent
 Fibβ36–52+/CEP-1+/filaggrin−3.34 (0.64–17.39)10.42 (2.97–36.56)17.94 (5.91–54.43)57.56 (20.91–158.42)
 Fibβ36–52+/CEP-1−/filaggrin+5.57 (0.92–33.62)13.9 (3.07–62.89)16.61 (3.92–70.4)40.39 (10.5–155.39)
 Fibβ36–52−/CEP-1+/filaggrin+2.78 (0.56–13.93)3.47 (0.69–17.42)3.32 (0.66–16.65)25.24 (8.93–71.33)
 Fibβ36–52+/CEP-1+/filaggrin+3.34 (0.64–17.39)33.35 (11.99–92.79)17.94 (5.91–54.43)69.68 (25.76–188.44)
Fibβ36–52, CEP-1, and CCP-2
 Negativity for all 3 antibodiesReferentReferentReferentReferent
 Fibβ36–52+/CEP-1+/CCP-2−0.0 (0–0)6.31 (1.04–38.16)0.0 (0–0)5.23 (0.54–50.93)
 Fibβ36–52+/CEP-1−/CCP-2+5.74 (0.95–34.69)22.10 (5.64–86.57)30.16 (8.05–112.98)52.33 (14.1–194.27)
 Fibβ36–52−/CEP-1+/CCP-2+12.93 (2.14–78.05)14.21 (2.35–85.86)55.29 (12.11–252.51)125.59 (28.34–556.63)
 Fibβ36–52+/CEP-1+/CCP-2+4.92 (1.42–17.04)25.71 (10.59–62.39)25.85 (10.57–63.2)91.94 (39.87–212.08)


This study addressed the questions of whether and when specific antibodies to citrullinated peptides representing potential autoantigens in patients with RA occur prior to the first symptoms of RA. Our choice of antigenic peptides encompassed citrullinated epitopes previously described to be autoantibody targets, or peptides corresponding to protein epitopes shown to be citrullinated in the joints of patients with RA in vivo. The novelty of our approach is that we had access to a much larger number of presymptomatic individuals and samples than were used in previously published studies, and that we were able to analyze an extended number of peptides using a multiplex assay system requiring minute volumes of sera. The major new result is the demonstration of previously unrecognized patterns of antibodies at different time points prior to the onset of symptoms. This may enhance our ability to predict the future onset of RA to a greater extent than previously. We also believe that this may contribute to our understanding of the immune responses against citrullinated peptides and, in particular, our ability to identify those autoantigens or autoantigenic epitopes contributing to the initiation of disease.

The methods of obtaining blood samples and the certainty of a diagnosis of RA allowed a comprehensive analysis of many samples obtained sequentially before disease onset, as well as samples from the same individuals after diagnosis. This type of comparative analysis has rarely been possible in other studies of individuals before the development of RA; however, there are also limitations. Although the present cohort is larger than those presented in previous publications, i.e., 386 individuals providing 717 samples, this still represents a relatively small number for development of predictive algorithms. Also, the samples were not donated at prespecified time points, but rather on an irregular basis, resulting in variable numbers of samples from the individuals, i.e., a given individual may have provided 1 sample years back, while another individual may be represented by up to 7 samples predating disease. Furthermore, samples were only available from a limited number of patients after disease onset. Our study complements and extends results from a recent study on a smaller collection of samples, in which a number of antibodies to citrullinated as well as other potential autoantigens relevant to RA were analyzed (34). A limited number of these antigens overlap with those analyzed in the present study. However, there is a need for further and larger collaborative studies using multiplex analysis of antibodies as well as other biomarkers in order to achieve sample sizes large enough to permit the development and validation of multiplex assay–based algorithms for predicting development of RA.

The multiplex antibody assay has the advantage of using very minute amounts of sera/plasma, i.e., <10 μl for analysis of reactivity against several different fine specificities. The sensitivity for ACPAs in samples from the patients with RA was consistent with the findings of other studies using ELISAs; for example, for anti–Fibβ36–52, our observed frequency of 65% was consistent with the frequency of 68% (23) or 73% (20, 22) reported by other investigators. For anti–CEP-1, the cyclic peptide, we observed a frequency of 67%, which was reasonably similar to the 61% reported for citrullinated α-enolase peptide, a linear peptide (35). Moreover, the previously reported frequency of antifilaggrin was 58% (35) compared to 46% in our study, and anti-citC1III was reported to be 40.4% (16) compared to 32% in our analysis. The largest disparity was for anti–Vim60–75, which was reported to be 49% (20, 22) or 59% in RA (35), whereas we observed a rate of positivity of 29%. In the present study, we chose cutoff values from the optimum sensitivity and specificity defined by ROC curves and the comparison of RA patients with controls, yielding negative values for 95–98% in controls. We considered this method to be robust and appropriate for this study.

The development of an anticitrullinated antibody immune response was observed to present in a restricted manner with 3 dominating patterns. One pattern was the clustering of antibodies against CEP-1, Fibβ36–52, and filaggrin represented by an expanding number of individuals converting to seropositivity over time, and with increasing levels of these antibodies present closer to the time of disease onset. At the time of RA diagnosis, the levels of these antibodies were further increased. Although our pre-patients were asymptomatic, this antibody clustering could suggest an involvement in the process by which clinical symptoms and subsequent disease will develop in an antibody-positive individual. The strongest association with subsequent development of RA was observed with the combination of anti–Fibβ36–52, anti–CEP-1, and antifilaggrin. The association with the development of RA was similar for the triplet combination of anti–Fibβ36–52, anti–CEP-1, and antifilaggrin and the combination of anti–Fibβ36–52 and anti–CEP-1, implying that antifilaggrin reactivity itself does not have any pathogenic impact on joint involvement. Citrullinated filaggrin, in contrast to the other autoantigens chosen, has not primarily been identified in rheumatoid joints but was first described as a dermal autoantigen. Citrullinated filaggrin constituted the target antigen in the very early variants of ACPA tests that preceded the development of currently used ELISAs, e.g., the anti-CCP assay (36).

A second antibody pattern was represented by the earliest detectable antibodies, i.e., those directed against Fibα591 and Vim60–75. These antibodies and Fibβ72 had among the highest degree of fluctuation when analyzed in consecutive samples collected before disease onset. The increase in fluorescence intensity was modest in samples collected close to the time of disease onset, although the levels of these antibodies, in particular anti–Vim60–75, increased further after the onset of clinical symptoms. We speculate that these results indicate a lesser pathogenic importance as compared with other antibodies that vary in intensity more dramatically during the pre-disease time period.

A third clustering of citrullinated antibodies could be identified in which the antibodies occurred at low frequencies before the onset of symptoms but were more consistently positive over time, namely anti-citC1III, Fibα573, and Fibβ74. The concentration of these antibodies increased significantly after disease onset. This pattern of antibody development suggested that these citrullinated antibodies may not be involved in the initial phase of autoimmunity to citrullinated proteins occurring systemically, but may be related to the clinical onset of joint disease.

The present study does not address the issue of cross-reactivity between the antibodies against the different citrullinated peptides, although the fact that the antibodies occurred in different patterns indicates a limited degree of cross-reactivity. A few studies have addressed the issue of cross-reactivity, demonstrated by cross-absorption experiments that provided evidence that there is some cross-reactivity, but also that a substantial proportion of the antibodies against α-enolase, vimentin, fibrinogen, and type II collagen do not cross-react when their respective citrullinated target antigens are used as a target in assays (17, 19, 23). In the present study there was an evident covariance with anti–CEP-1, anti–Fibβ36–52, and antifilaggrin, which we conclude may essentially be real covariance rather than cross-reactivity. Immunity to some citrullinated autoantigens, but not to others, may have different genetic as well as environmental determinants (17, 20–23).

The initiation of ACPA reactivity has been suggested to occur in the lungs and possibly in the gums (9, 10); consequently, it would be interesting to identify which of the currently described target antigens are present in these 2 sites concurrently with initiation of reactivity of the different ACPAs. Furthermore, it would be of interest to investigate which specific ACPAs, alone or in combination, are able to trigger different symptoms associated with RA. There is already evidence that certain ACPAs, alone or in combination, may contribute to arthritis, as seen in immunization and transfer experiments in rodents (14, 30, 37–39). Therefore, we conclude that it is essential to know when and why different ACPAs occur in order to better understand events governing the transition from a healthy state to an autoimmune state and subsequently to clinical disease.

Based on the results of the present study we propose the following hypothesis regarding antibodies against citrullinated peptides and the evolution of ACPA- positive RA. Some anticitrulline antibodies appear early, but with a low frequency and with fluctuating levels over time until the onset of disease, and increase discretely after disease onset. We consider it less likely that these antibodies are involved in triggering clinical symptoms, but they may be part of an initial breakage of tolerance against citrullinated proteins and facilitate further epitope spreading. Another group of antibodies, possibly occurring as a result of epitope spreading, are more permanently positive and increase significantly after disease onset. A third group includes those antibodies that develop later and increase markedly both before and after the onset of disease. We may speculate that antibodies from these last 2 clusters, and in particular those from the third group, may be involved in triggering disease symptoms. In experiments designed to test hypotheses on RA pathogenesis suggested by these and other data, this hypothesis on the role of antibodies against different citrullinated peptides may be combined with emerging evidence of major histocompatibility complex–restricted T cell reactivity toward the same citrullinated or otherwise modified potential target autoantigens in ACPA-positive patients with RA (40, 41).

In conclusion, the presence of elevated levels of antibodies against several citrullinated peptides in individuals before the onset of RA symptoms appeared in various patterns during the period of time before symptom onset. A better understanding of these patterns and their individual component is important for elucidating their involvement in the pathogenesis of RA. Such knowledge could provide a key to the understanding of which specific immune reactions, alone or in combination, may cause arthritis in different subsets of patients with RA.


All authors were involved in drafting the article or revising it critically for important intellectual content, and all authors approved the final version to be published. Dr. Rantapaa-Dahlqvist had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Study conception and design. Brink, Klareskog, Rantapaa-Dahlqvist.

Acquisition of data. Brink, Hansson, Mathsson, Hallmans, Rönnelid, Klareskog, Rantapaa-Dahlqvist.

Analysis and interpretation of data. Brink, Hansson, Mathsson, Jakobsson, Holmdahl, Stenlund, Klareskog, Rantapaa-Dahlqvist.


PhaDia/Thermo Fisher Scientific had no role in the study design or in the collection, analysis, or interpretation of the data, the writing of the manuscript, or the decision to submit the manuscript for publication. Publication of this article was not contingent upon approval by PhaDia/Thermo Scientific.


Dr. Mathsson is currently an employee of PhaDia/Thermo Fisher Scientific and is also affiliated with Uppsala University, where a substantial part of her contribution to the present work was made. Drs. Hansson, Jakobsson, Holmdahl, Rönnelid, and Klareskog, as members of the Rheumatology Unit at the Department of Medicine at Karolinska Institutet, are partners with PhaDia/Thermo Fisher Scientific within the Innovative Medicines Initiative /EU program on RA, where relationships for “precompetitive research” are regulated by a consortium agreement approved by the IMI organization (see Thermo Fisher Scientific contributes to this IMI consortium with in-kind contributions for the development of the multiplex ISAC assay used in the current study. Dr. Klareskog is administrative coordinator of the BTCure project. Drs. Holmdahl and Klareskog are cofounders of a company, Curara, which has an agreement with Thermo Fisher to contribute to the development of the multiplex ISAC assay. No commercial agreements were, however, involved with the use of the ISAC assay in the current study.


We thank Urban Kumlin and Monika Nimrodsson (Department of Virology, University Hospital, Umeå) for providing sera from the Maternity cohort, Per Matsson, Mats Nystrand, Christian Harwanegg, and Thomas Schlederer (Thermo Fisher) for contributions to establishment of the ISAC platform, and Lena Israelsson (Karolinska Institute) for technical assistance.