Autoimmunity to peptidyl arginine deiminase type 4 precedes clinical onset of rheumatoid arthritis


  • The views expressed herein are those of the authors and do not reflect the official policy of the Department of the Army, Department of Defense, or United States government.



To determine whether antibodies against peptidyl arginine deiminase type 4 (PAD-4) are present in the preclinical phase of rheumatoid arthritis (RA) and to compare the timing and extent of their appearance with those of other preclinical autoantibodies.


Prediagnosis serum samples from 83 patients with RA were evaluated for the presence of anti–PAD-4 antibody, anti–cyclic citrullinated peptide (anti-CCP) antibody, and rheumatoid factor. In addition, a control cohort (n = 83) matched by age, sex, race, number of serum samples, and duration of serum storage was tested for the presence of anti–PAD-4 antibody to determine its sensitivity and specificity for the subsequent development of RA.


Fifteen of 83 patients with RA (18.1%) had at least 1 prediagnosis sample positive for anti–PAD-4. One of 83 control subjects (1.2%) had at least 1 positive sample, resulting in a sensitivity and specificity of 18.1% and 98.8%, respectively, of anti–PAD-4 for the future development of RA. The mean duration of anti–PAD-4 positivity prior to clinical diagnosis was 4.67 years. Anti–PAD-4 positivity was associated with anti-CCP positivity (odds ratio 5.13 [95% confidence interval 1.07–24.5]). In subjects with prediagnosis samples that were positive for both antibodies, anti-CCP positivity predated anti–PAD-4 positivity in 9 of 13 cases (69%).


Autoantibodies to PAD-4 are present during the preclinical phase of RA in a subset of patients and are associated with anti-CCP positivity. Further exploration is needed regarding the timing of appearance and disease-related effects of PAD-4 autoimmunity.

Studies using stored prediagnosis specimens have demonstrated the presence of autoantibodies, cytokines/chemokines, and markers of inflammation years prior to the clinical onset and diagnosis of rheumatoid arthritis (RA) (1–5). These findings suggest that there is a preclinical period in RA during which immunologic and inflammatory changes occur that may subsequently lead to symptomatic disease. As such, biomarkers that are present in this preclinical period are of great interest and may aid in the understanding of disease pathogenesis.

Autoantibodies against peptidyl arginine deiminase type 4 (PAD-4) have recently been described as a specific biomarker in patients with clinically apparent RA (6). PADs are a family of enzymes responsible for the posttranslational modification of the amino acid arginine to citrulline. This process is likely to be of significance in patients with RA, given the established association of anti–citrullinated protein antibodies (ACPAs) with disease presence and severity. Several studies have identified an association between genetic polymorphisms of the PADI4 gene and RA (7–10), although it has not been confirmed across all racial and ethnic groups (11, 12). Subsequent studies showed that PAD-4 may also function as an antigen, generating antibody responses in patients with RA (13, 14). Recently, researchers demonstrated the presence of specific anti–PAD-4 antibodies in patients with RA, as well as an association with disease severity (6, 15, 16). However, the role of PAD-4 in the development of RA has not been fully elucidated. In the present study, we tested for the presence of anti–PAD-4 antibodies in prediagnosis serum samples from patients with RA in order to determine whether these autoantibodies play an early role in disease evolution. In addition, we sought to describe the timing of anti–PAD-4 antibody appearance in the preclinical period, its relationship to anti–cyclic citrullinated peptide (anti-CCP) autoimmunity, and potential associations with a more severe RA phenotype.


Study population.

We used stored serum samples previously obtained from 83 subjects in the military who had subsequently been diagnosed as having RA—a cohort previously identified through the Walter Reed Army Medical Center Rheumatology Clinic (4). RA patients included in this analysis met ≥4 of the 1987 revised classification criteria of the American College of Rheumatology (ACR; formerly, the American Rheumatism Association) (17) or were diagnosed as having RA by a board-certified rheumatologist. Information on age, sex, race, and symptom onset at the time of RA diagnosis was obtained by chart review. The presence or absence of radiographic erosions had been determined by a radiologist as part of clinical care. In addition, a control cohort of 83 subjects in the military without RA was matched to the RA patients by age, sex, race, number of serum samples, and duration of serum storage. The study protocol was approved by the Institutional Review Boards at Walter Reed Army Medical Center and the University of Colorado. Further details on the repository and RA cohort are described elsewhere (4).

Autoantibody testing.

Rheumatoid factor (RF) and anti-CCP antibody testing was performed at the University of Colorado Division of Rheumatology Clinical Research Laboratory. RF was measured by nephelometry according to the specifications of the manufacturer (Dade Behring). The ACR classification criteria for RA specify that an RF level is considered positive if it is present in <5% of control subjects (17). Accordingly, we determined a general cutoff level for RF positivity of >15.2 IU/ml, using a 95% cutoff point established from the 83 healthy control subjects in the military. Anti-CCP antibodies were tested by enzyme-linked immunosorbent assay (ELISA) using the anti–CCP-2 kit (Diastat; Axis-Shield). In accordance with the manufacturer's specifications, a positive test result was defined as >5 units/ml.

Anti–PAD-4 antibody testing was performed in the Johns Hopkins University Rheumatic Disease Research Core Center, using an immunoprecipitation method as previously described (6, 18). Briefly, 35S-methionine–labeled human PAD-4 was generated by coupled in vitro transcription/translation. Immunoprecipitation was performed in a controlled setting over 1 hour at 4°C and in the absence of calcium. Anti–PAD-4 antibody was detected by fluorography; densitometric scanning allowed for quantitative comparison between patient serum and established positive and negative controls. A positive reference serum was assigned a value of 1.0; patient sera with values >0.1 were categorized as positive. A semiquantitative scale (0, 1, 2, and 3+) based on densitometry of scanned immunoprecipitation autoradiograms was used to assign a value to each serum sample.

Statistical analysis.

The prevalence of preclinical autoimmunity as well as sensitivity and specificity values of the individual autoantibody tests were calculated. The mean duration of autoantibody positivity in the prediagnosis period (to time of diagnosis) was calculated based on the earliest serum sample that was positive for RA-related autoantibodies. In a subset of cases, the earliest available serum sample was autoantibody positive (representing left censorship of data); thus, the true duration of antibody positivity may have been underestimated. Survival analysis was not used to adjust for left censorship because of the relatively small number of antibody-positive cases in the cohort (4, 19). Differences in the mean duration of antibody positivity were determined by the t-test and adjusted for left censorship (20). The signed rank test was used to determine whether anti-CCP antibody appeared prior to anti–PAD-4 antibody.

Logistic regression was used to analyze the relationship of age, sex, and race with anti–PAD-4 positivity and to analyze the relationship between anti-CCP and anti–PAD-4 antibodies. Factors associated with radiographic erosions (at or after diagnosis of RA) were determined using logistic regression with the following predictor variables included in the model: anti–PAD-4 and anti-CCP antibodies, RF, age, sex, and the interaction of anti–PAD-4 with anti-CCP. All statistical analyses were performed using SAS software, version 9.2 (SAS Institute).


The demographic and clinical characteristics of the RA patients are presented in Table 1. Anti–PAD-4 was present in at least 1 prediagnosis serum sample from 15 RA subjects but in a serum sample from only 1 control, resulting in a sensitivity and specificity for subsequent RA of 18.1% and 98.8%, respectively (Table 2). The mean duration of anti–PAD-4 positivity prior to clinical diagnosis was 4.67 years (Table 2). Anti-CCP antibody was present in 51 RA subjects (61.4%), with a sensitivity and specificity for subsequent RA of 61.4% and 100%, respectively. The mean duration of anti-CCP antibody positivity prior to diagnosis was 3.49 years. RF was present in 47 RA subjects (56.6%), with a sensitivity and specificity for subsequent RA of 56.6% and 86.7%, respectively. The mean duration of RF positivity prior to diagnosis was 3.8 years. In postdiagnosis samples, the prevalence for RF and anti–PAD-4 and anti-CCP antibodies increased to 83%, 26%, and 68%, respectively (Table 2).

Table 1. Demographic and clinical characteristics of the patients with rheumatoid arthritis (n = 83)*
  • *

    Except where indicated otherwise, values are the number (%) of subjects. Adapted, with permission, from ref.4.

Age at diagnosis, mean ± SD years39.9 ± 10.0
Male49 (59)
 White57 (69)
 Black21 (25)
 Other5 (6)
 Present42 (51)
 Absent34 (41)
 Unclassified7 (8)
Preclinical serum samples (n = 243) 
 Samples per case, mean ± SD2.9 ± 1.2
 Years between first sample and diagnosis, mean ± SD6.6 ± 3.7
Table 2. Characteristics of preclinical autoantibodies (RF, anti–PAD-4, and anti-CCP)*
AutoantibodyPositive test result before diagnosis, no. (%)Sensitivity for subsequent RA, %Specificity for subsequent RA, %Duration of antibody positivity prior to diagnosis, mean (95% CI) yearsPositive test result after diagnosis, no. (%)
  • *

    RF = rheumatoid factor; RA = rheumatoid arthritis; 95% CI = 95% confidence interval.

  • There were 47 subjects with available postdiagnosis samples.

  • Patients positive for both anti–peptidyl arginine deiminase type 4 (anti–PAD-4) and anti–cyclic citrullinated peptide (anti-CCP) antibodies in the preclinical period (double-positive patients).

  • §

    Patients with anti-CCP antibodies but without anti–PAD-4 antibodies in the preclinical period.

RF >15.2 IU/ml by nephelometry47 (56.6)56.686.73.80 (2.83–4.76)39 (83)
Anti–PAD-415 (18.1) (2.54–6.82)12 (26)
Anti-CCP51 (61.4)61.4100.03.49 (2.59–4.39)32 (68)
 Anti–PAD-4 present13 (15.7)15.7100.06.16 (4.29–8.02)
 Anti–PAD-4 absent§38 (45.8)45.8100.02.58 (1.68–3.47)

Anti–PAD-4 positivity was significantly associated with anti-CCP positivity, with an odds ratio (OR) of 5.13 (95% confidence interval [95% CI] 1.07–24.5). Anti–PAD-4 and anti-CCP positivity (double positivity) was seen in 13 of the 83 subjects with RA but in no controls; double positivity was 15.7% sensitive and 100% specific for the future development of RA (Table 2). In 9 of 13 double-positive subjects, anti-CCP positivity predated anti–PAD-4 positivity, while only 1 subject developed antibodies to PAD-4 prior to antibodies to CCP, suggesting that anti-CCP antibody tends to appear prior to anti–PAD-4 antibody in these subjects (P = 0.027) (Table 3). Only 2 subjects with anti–PAD-4 positivity prior to diagnosis were anti-CCP negative.

Table 3. Timing of antibody appearance in the 13 patients with both anti–PAD-4 and anti-CCP in prediagnosis samples*
  • *

    Anti-CCP antibody tended to appear prior to anti–PAD-4 antibody in these subjects (P = 0.0273 by signed rank test). See Table 2 for definitions.

Anti-CCP preceded anti–PAD-49 of 13 subjects
Anti–PAD-4 preceded anti-CCP1 of 13 subjects
First appearance in the same sample3 of 13 subjects

The mean time between appearance of anti-CCP and clinical diagnosis in 13 double-positive subjects was 6.16 years (Table 2), while in subjects without preclinical anti–PAD-4 the mean time was 2.58 years (P < 0.0007). The mean anti-CCP antibody titer in all available samples in double-positive subjects was 206.3 units/ml, while in subjects without evidence of preclinical anti–PAD-4 the mean anti-CCP antibody titer was 82.7 units/ml (P = 0.03).

To evaluate the persistence of autoantibodies over time, autoantibody positivity was assessed in subjects who had additional samples collected after their initial positive test results. Samples in the pre- and postdiagnosis period were used for this analysis. Six of the 14 anti–PAD-4–positive subjects (43%) with multiple samples had reversion from anti–PAD-4–positive status to persistent seronegative status, while 8 (57%) remained positive. Six of the 44 subjects assessed (14%) had anti-CCP reversion to seronegative status, while 38 (86%) remained positive in followup samples. Three of the 42 subjects assessed (7%) had RF reversion to seronegative status, while 39 (93%) remained positive. The proportion of subjects with RF reversion was significantly less than the proportion with anti–PAD-4 reversion (7% versus 43%; P < 0.01), while the difference between anti-CCP and anti–PAD-4 reversion bordered on significance (14% versus 43%; P = 0.05). Several subjects had reversion in the postdiagnosis time period: all 3 subjects with RF reversion, 4 of 6 subjects with anti-CCP reversion, and 2 of 6 subjects with anti–PAD-4 reversion. Changes in anti–PAD-4 antibody level between sample collections, based on densitometry and using a semiquantitative scale (0–3+), are outlined in Figure 1. Changes in anti-CCP titer from the same serum samples are plotted alongside changes in anti–PAD-4 titer for comparison.

Figure 1.

Longitudinal pattern of anti–peptidyl arginine deiminase type 4 (anti–PAD-4) and anti–cyclic citrullinated peptide (anti-CCP) antibodies in 15 patients (ID 1–15) with rheumatoid arthritis whose serum was positive preclinically for anti–PAD-4. Testing for anti–PAD-4 and anti-CCP antibodies was performed on successive samples from each individual. The time between samples varied among subjects (not shown). Anti–CCP titers separated into the following categories: 0 = 0–5 units/ml, 1 = >5–50 units/ml, 2 = >50–100 units/ml, 3 = >100–150 units/ml, and 4 = >150 units/ml. The level of anti–PAD-4 antibodies was determined using a semiquantitative scale (0, 1, 2, and 3+) based on densitometry of scanned immunoprecipitation autoradiograms. Diamonds represent anti–CCP antibody titers (ranging from 0 to 4); x's represent anti–PAD-4 antibody levels (ranging from 0 to 3). The vertical line marked Dx represents the time of diagnosis of rheumatoid arthritis.

No association was seen between age, sex, or race and the presence of anti–PAD-4 antibody (Table 4). In multivariate analysis, after accounting for anti–PAD-4 autoantibody presence, an independent association was identified between anti-CCP antibody positivity in the prediagnosis period and subsequent erosions (OR 12.1 [95% CI 2.34–62.9]). There was a nonsignificant association between anti–PAD-4 and subsequent erosive disease (OR 1.71 [95% CI 0.45–6.54], P = 0.43). A subgroup analysis of subjects (n = 8) with persistent anti–PAD-4 positivity over time demonstrated similar results (OR 2.11 [95% CI 0.41–13.9], P = 0.48).

Table 4. Association of anti–PAD-4 antibody with patient demographics*
VariableOR (95% CI)
  • *

    The analysis was based on logistic regression. No association was seen between age, sex, or race and the presence of anti–PAD-4 antibody. OR = odds ratio (see Table 2 for other definitions).

  • Analyzed as a continuous variable.

  • Analyzed using non-Hispanic whites as the reference category; other race categories included Asian and Hispanic.

Age0.99 (0.93–1.06)
Female sex1.77 (0.30–10.4)
Black race1.06 (0.09–12.1)
Other race1.37 (0.33–5.76)


We have found that autoantibodies against the PAD-4 enzyme are present in the prediagnosis period and are specific for the future development of RA. Anti–PAD-4 antibodies were evident as early as 4 years prior to clinical diagnosis, similar to findings reported in other studies for preclinical anti-CCP and RF (1, 3, 4, 21, 22). In addition, the presence of anti–PAD-4 antibody was significantly associated with anti-CCP antibody. Patients with both anti–PAD-4 and anti-CCP present in the preclinical period appeared to have a longer time interval between the appearance of anti-CCP and clinical disease.

To detect antibody against the PAD-4 enzyme, we used a novel approach using immunoprecipitation with in vitro–transcribed/translated PAD-4. While others have used ELISAs to detect anti–PAD-4 antibody (15, 16, 23, 24), we have found a significant false-positive rate (∼10%) in RA using this method compared with immunoprecipitation. One may question whether or not the anti–PAD-4 antibodies detected in the above experiments were directed against citrullinated targets on the PAD-4 enzyme. A recent study by Andrade and colleagues demonstrated the presence of antibodies against both citrullinated and noncitrullinated epitopes on the PAD-4 enzyme in a subset of patients with RA (25). As such, the conditions of our study were chosen to minimize the risk of in vitro–transcribed/translated PAD-4 autocitrullination, which could lead to identification of antibodies against citrullinated residues. Calcium is required as a cofactor for proper functioning of the PAD-4 enzyme (26); therefore, the assay in our study was performed at 4°C in the absence of calcium. Under similar conditions, Andrade and colleagues demonstrated an inability of PAD-4 to autocitrullinate, providing further evidence that the antibodies identified in our study are directed against noncitrullinated residues.

The prevalence of anti–PAD-4 antibody detected in our study deserves additional comment. Data from prior studies have determined the prevalence of anti–PAD-4 antibody in established RA to be 35–45% (6, 15, 16, 24), in contrast to the prevalences in the pre- and postdiagnosis time periods noted in our study (18% and 26%, respectively). However, the duration of disease in these cohorts ranged from 7 years to 13.5 years, in contrast to the mean disease duration in our cohort of 3.3 years at the time of last postdiagnosis sample. In addition, a prior study by Harris et al discovered a decreased prevalence of anti–PAD-4 antibody in early RA compared with established disease (6). Taken together, these findings suggest that the prevalence of anti–PAD-4, while low in the prediagnosis period, may increase over time. The prevalence of anti–PAD-4, anti-CCP, and RF all increased in our cohort between the prediagnosis and early postdiagnosis time period, supporting a hypothesis of amplification of multiple autoimmune pathways. These latter data also support prior descriptions of increasing anti-CCP prevalence in the prediagnosis time period as patients draw closer to the time of RA symptoms (1).

Another area of particular interest is the description of autoantibody reversion over time. Reversion to a seronegative status occurred for each autoantibody tested, although the percentage of subjects with anti–PAD-4 reversion was higher than that of subjects with either anti-CCP or RF reversion. Most instances of RF and anti-CCP reversion occurred in the postdiagnosis period, as noted in the Results. Studies in established RA have shown declining RF and anti-CCP antibody titers in patients with good clinical response to ongoing therapy (27, 28). The loss of seropositivity in subjects during the postdiagnosis period may reflect improved disease control upon initiation of immunosuppressive therapy; however, supporting data are not available regarding disease activity, medication use, and other factors after the time of diagnosis.

Also of interest is the presence of a significant association between anti–PAD-4 and anti-CCP antibodies, a finding similar to that in cohorts with established RA (6, 16, 29). Epitope spreading may be a potential hypothesis to explain this association. According to this model, anti-CCP antibodies develop first due to novel antigen structures generated by PAD activity. The PAD-4 enzyme might then become an antigen itself through association with its substrates. A similar mechanism has been proposed in the development of celiac disease, in which a positive correlation has been found between autoantibodies directed against tissue transglutaminase and its substrate, gliadin (29). However, this hypothesis would not clearly explain the 2 subjects in our cohort who developed anti–PAD-4 antibodies in the absence of anti-CCP. These 2 cases suggest either that the PAD-4 enzyme can serve as an autoantigen independently of its interaction with citrullinated residues or that ACPAs that precede anti–PAD-4 antibody are present that are not detected by the commercial anti–CCP-2 assay. Antigen formation through PAD-4 automodification is another interesting possibility, especially given the recent observation that PAD-4 can be autocitrullinated and that this may modify the enzyme's tertiary structure (25).

We noted a prolonged latency between the appearance of anti-CCP antibodies and clinical disease in patients with anti–PAD-4 positivity. On average, this time period was >3 years longer in double-positive subjects compared with anti-CCP–positive subjects without anti–PAD-4. The mechanism responsible for this finding is unknown, but these data generate hypotheses regarding the influence of anti–PAD-4 antibody on subsequent PAD enzyme function. A gain or loss of function in response to antibody binding could explain several features relevant to PAD-4 autoimmunity in RA. For example, a loss of function following antibody binding could result in decreased levels of protein citrullination. This could result in a longer period of preclinical autoimmunity, during which additional genetic or environmental factors are needed for transition to clinical disease. In contrast, antibody binding could potentially lead to altered substrate specificities, generating additional citrullinated epitopes, such as vimentin, fibrinogen, and other peptide targets. This could lead to increased levels of autoimmune activation and could account for the association between anti–PAD-4 antibody and decreased functional status, increased disease activity, and advanced radiographic progression (15, 16, 24). A recent study by Auger et al has provided some insight into this area by demonstrating that autoantibodies to PAD-4 can inhibit PAD-4–mediated citrullination (23). These hypotheses remain an area where additional research is necessary.

This study has generated important initial data on the presence of anti–PAD-4 antibodies in early RA development; however, the analysis has several limitations. The RA cohort examined here consists of military personnel, resulting in an increased proportion of male subjects compared with that in the general RA population. However, the age at diagnosis and prevalence of RF, anti-CCP, and radiographic erosions are similar to those in cohorts previously described in the literature (30–32). Control subjects were selected from military personnel as well and may represent a younger and healthier cohort compared with the general non-RA subject population. This could affect our specificity analysis, since older subjects and those with coexistent disease may potentially have a higher prevalence of RA-related autoimmunity.

Additionally, analysis describing the timing of antibody appearance in relation to diagnosis is limited in subjects with antibody positivity in the first available serum sample. In these subjects, we were unable to determine the interval of conversion from seronegative status to seropositive status because of left censorship of the data. Statistical adjustment was performed for analyses that included these data (see Patients and Methods). Our study is also limited by the relatively small number of subjects in our RA cohort, resulting in only 15 subjects with anti–PAD-4 antibody in the preclinical period. As a result, we were limited in our ability to identify an association with clinical end points such as erosive disease, particularly in the subset of double-positive subjects. Furthermore, the prevalence of erosions in our study may be an underestimate, since subjects without available radiographs were considered not to have erosions, and some subjects may have radiographs from the time of diagnosis. This will be an important area for further study, since evidence suggests that the coexistence of both anti-CCP and anti–PAD-4 is associated with more severe disease (6, 15).

In conclusion, this study has identified anti–PAD-4 antibodies in the preclinical period of RA development, which appear to be specific for the subsequent development of clinically apparent disease. Further research is needed to investigate the relative timing of the appearance of multiple autoantibodies in preclinical RA, the physiologic implications of autoantibody binding to the PAD-4 enzyme, and the potential influence on disease development and severity.


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. Holers 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. Kolfenbach, Deane, Derber, Gilliland, Rosen, Norris, Holers.

Acquisition of data. Deane, Derber, Edison, Darrah, Holers.

Analysis and interpretation of data. Kolfenbach, Deane, Derber, O'Donnell, Rosen, Darrah, Holers.


We would like to thank David Hines of the Johns Hopkins University Rheumatic Disease Research Core Center for technical contributions in performing the PAD-4 immunoprecipitations, as well as Kristin Braschler for her work in sample processing for antibody tests at the University of Colorado Division of Rheumatology Clinical Research Laboratory.