Article Text

Serum IgG antibodies to peptidylarginine deiminase 4 predict radiographic progression in patients with rheumatoid arthritis treated with tumour necrosis factor-α blocking agents
  1. E H Halvorsen1,
  2. E A Haavardsholm2,
  3. S Pollmann1,
  4. A Boonen3,
  5. D van der Heijde2,4,
  6. T K Kvien2,
  7. Ø Molberg1,5
  1. 1
    Institute of Immunology, Faculty of Medicine, University of Oslo and Rikshospitalet University Hospital, Oslo, Norway
  2. 2
    Department of Rheumatology, Diakonhjemmet Hospital, Oslo, Norway
  3. 3
    Department of Rheumatology, University Hospital Maastricht, Maastricht, The Netherlands
  4. 4
    Department of Rheumatology, Leiden University Medical Center, Leiden, The Netherlands
  5. 5
    Department of Rheumatology, Rikshospitalet University Hospital, Oslo, Norway
  1. Eirik Hornes Halvorsen, Institute of Immunology, University of Oslo, Rikshospitalet University Hospital, Oslo, Norway; eirikhha{at}


Background: Peptidylarginine deiminase 4 (PAD4) may generate epitopes targeted by anticitrullinated protein antibodies in rheumatoid arthritis (RA). A subset of patients with RA has serum autoantibodies to human recombinant PAD4 (hPAD4). Here, we assessed whether anti-hPAD4 status in RA predicted disease outcome after antitumour necrosis factor (anti-TNF)-α therapy.

Methods: We analysed RA sera obtained at baseline (n = 40) and after 1 year on anti-TNF-α therapy (n = 33) for anti-hPAD4 IgG. Association analyses between baseline anti-hPAD status and disease progression were performed.

Results: We found that 17 of 40 patients (42.5%) were serum anti-hPAD4 positive at baseline, and the anti-hPAD4 IgG levels were stable over 1 year on anti-TNF-α therapy. At baseline, there were indications that anti-hPAD4 positive patients had more severe disease than the negative patients. After 1 year on anti-TNF-α therapy, the anti-hPAD4 positive patients displayed a persistently elevated disease activity score using 28 joint counts score and increased progression in the van der Heijde–modified Sharp erosion score. Accordingly, more anti-hPAD4 positive than negative patients presented an increase in van der Heijde–modified Sharp erosion scores >0 over 1 year.

Conclusions: Anti-hPAD4 IgG can be detected in a subset of RA sera and the levels are stable after initiation of anti-TNF-α therapy. Serum anti-hPAD4 may predict persistent disease activity and radiographic progression in patients with RA receiving anti-TNF-α therapy.

View Full Text

Statistics from

Rheumatoid arthritis (RA) is a complex disease causing destructive joint inflammation with synovial infiltration of activated immune cells secreting pro-inflammatory cytokines, including tumour necrosis factor (TNF)-α.1 The introduction of agents blocking the effects of TNF-α (TNF-inhibitors) has revolutionised the RA treatment, but unfortunately, the TNF inhibitors come with high costs, potential adverse effects and fail to induce clinical improvement in a substantial number of patients.1 2 Very few parameters associated with the clinical response to TNF inhibitors have been reproducibly identified,35 and the effects of TNF inhibitors on radiographic progression may not be correlated to the clinical response.6 It is critical to identify novel biomarkers associated with progressive joint destruction despite TNF-inhibitor therapy.

We and others have demonstrated that a subset of patients with RA have serum autoantibodies against human recombinant peptidylarginine deiminase 4 (hPAD4).79 The citrullinated epitopes targeted by the RA-specific anticitrullinated protein antibodies may be generated in vivo by PAD4.10 Here, our previously developed time-resolved immunoassay was used to assess serum anti-hPAD4 IgG in patients with RA at baseline and after 1 year on TNF inhibitors. We evaluated the stability of the autoantibody levels and analysed whether baseline anti-hPAD4 status was associated with clinical and radiographic outcome after 1 year on TNF-inhibitor therapy.


Patients and sera

Forty patients with RA (32 females and eight males, mean age 54.3 years and mean disease duration 11.5 years) receiving treatment according to clinical practice, were referred to the study when TNF-inhibitor therapy (adalimumab (n = 13), etanercept (n = 12) or infliximab (n = 15)) was initiated. The patients were assessed at baseline and after 3, 6 and 12 months, using a previously described protocol.11 Briefly, clinical examinations included swollen (28) and tender (28) joint counts and global assessments of pain and fatigue. Laboratory analyses included erythrocyte sedimentation rate, C-reactive protein, anti-cyclic citrullinated peptide (anti-CCP)2 and rheumatoid factor. The disease activity score using 28 joint counts (DAS28) was computed using the erythrocyte sedimentation rate and physical disability assessed by the Modified Health Assessment Questionnaire. Imaging procedures included digital radiographs of both hands, scored according to the van der Heijde–modified Sharp (vdHSharp) method.12 Radiographs at baseline, 3, 6 and 12 months were available from 39, 35, 31 and 30 patients, respectively, and scored by one trained observer blinded for clinical characteristics and time order.

Anti-hPAD4 antibody assays

The expression of hPAD4 and the detection of IgG autoantibodies to hPAD4 were done as described.7 For six patients, baseline serum was unavailable, and serum obtained after 3 months was used instead. For seven patients, serum obtained after 12 months was unavailable.

Statistical methods

Anti-hPAD4 levels at baseline and after 12 months were compared using Wilcoxon signed ranks test. Baseline variables were compared across anti-hPAD4 status by the Mann–Whitney U test or the Student t test.

For the 29 patients with radiographs available both at baseline and after 1 year, the change in the vdHSharp scores was assessed by Wilcoxon signed ranks tests, and compared across baseline anti-hPAD4 status. Possible predictors of radiographic progression were included in two linear multivariate regression analyses.

The frequencies of patients with erosive progression were compared between the anti-hPAD4 positive and negative patients using Fisher’s exact test and adjusted by bivariate exact logistic regression analyses.


Serum anti-hPAD4 IgG levels

For 33 patients, paired samples were available. Analyses of sera obtained at baseline and after 1 year on TNF-inhibitor therapy demonstrated stable levels of anti-hPAD4 (Z = −0.24, p = 0.809) (fig 1) and anti-CCP (Z = −1.14, p = 0.253) (data not shown). Seventeen of 40 (42.5%) and 15 of 33 (45.5%) patients were anti-hPAD4 positive at baseline and after 1 year respectively (fig 1).

Figure 1 Sera from a rheumatoid arthritis cohort starting on tumour necrosis factor inhibitors were analysed for anti-hPAD4 IgG by our time-resolved fluorometric immunoassay. Sera were obtained at baseline (n = 40) and after 12 months on tumour necrosis factor inhibitors (n = 33). All sera were tested in duplicates and the anti-hPAD4 IgG level is expressed as the mean emission signal after subtraction of the background signal. The dotted line indicates the assay’s cut-off level for positivity, defined as the 95th percentile of previously tested healthy controls sera. For the 33 patients with paired samples available, anti-hPAD4 IgG levels were stable over 1 year (p = 0.809 by Wilcoxon signed ranks test).

Serum anti-hPAD4 IgG status and disease variables

At baseline, the anti-hPAD4-positive patients had higher 28 “swollen joint count”, DAS28 and vdHSharp joint space narrowing scores than the negative patients. Otherwise, the two groups were comparable at baseline (table 1). After 1 year, the difference in DAS28 was enhanced (p = 0.016), due to persistently high scores in the baseline anti-hPAD4-positive patients and decreased scores in the negative patients (fig 2D).

Figure 2 Increased radiographic progression and persistently high disease activity in baseline anti-hPAD4 IgG-positive patients (▾ and shaded bars) compared with anti-hPAD4 IgG-negative patients (▾ and white bars) during 1 year on tumour necrosis factor inhibitors. (A) Longitudinal radiographic erosions expressed as the vdHSharp erosion scores at baseline and after 3, 6 and 12 months (median, IQR). (B) Change in vdHSharp erosion score after 12 months compared with baseline (mean, 95% CI). (C) Frequency of patients with change in vdHSharp erosion score >0 after 12 months compared with baseline. (D) DAS28 score at baseline and after 12 months (median, IQR). DAS28, disease activity score using 28 joint counts; vdHSharp, van der Heijde–modified Sharp method.
Table 1 Baseline description of patients with RA who were positive and negative for serum anti-hPAD4 IgG

Serum anti-hPAD4 IgG status and radiographic progression

The anti-hPAD4 -positive and -negative patients had similar vdHSharp erosion scores at baseline (table 1), but only those who were anti-hPAD4-positive displayed increased longitudinal erosion scores during 1 year (Z = −2.14, p = 0.032 vs Z = −0.71, p = 0.478) (fig 2A). The mean (95% CI) change in the erosion score from baseline to 1 year were 1.27 (0.11 to 2.44) vs −0.32 (−1.17 to 0.54) (p = 0.023) (fig 2B). Two multivariate linear regression analyses, each including the anti-hPAD4 status and two independent variables, showed borderline significant associations between anti-hPAD4 positivity and the change in erosion score (supplemental table 1).

When the patients were dichotomised based on the presence or absence of an increase in vdHSharp erosion score >0 from baseline, 11 of 29 (38%) patients displayed progressive erosion scores. The frequency of progressors was higher among the baseline anti-hPAD4-positive than among the negative patients (73% vs 17%, p = 0.005) (fig 2C). Significance was maintained in bivariate exact logistic regression analyses adjusting for possible independent predictors of erosive progression (supplemental table 2).


Enzymes with key roles in the pathogenesis of immune-mediated inflammatory diseases are often targeted by distinct serum autoantibodies, and this is also the case for PAD4 in RA.79 Here, we demonstrate stable levels of anti-hPAD4 in RA sera during TNF-inhibitor therapy. Moreover, we found different outcomes after 1 year on TNF inhibitors when comparing patients with RA who were anti-hPAD4 positive and negative at baseline.

All patients fulfilled the clinical criteria for TNF-inhibitor therapy. Consistently, they displayed high baseline values for 28 ‘swollen joint count’, 28 ‘tender joint count’, DAS28 and Modified Health Assessment Questionnaire, and 85% of them were anti-CCP positive. We believe that this very high frequency of patients who were anti-CCP-positive, combined with the relatively limited size of this study cohort, may explain why the current study, in contrast to previous observations,7 did not find a correlation between anti-hPAD and anti-CCP.

Almost all patients (95%) had radiographic joint damage at baseline according to the vdHSharp method. The anti-hPAD4-positive patients had higher baseline joint space narrowing scores than those that were negative, probably reflecting increased cartilage damage. This difference remained, but did not further increase significantly over 1 year. However, the erosion scores differed both in the rate of progression and the percentage of progressors. The fact that these differences were detectable in our limited cohort after 1 year only, should lead to further scrutiny.

From our observations, TNF-inhibitor therapy was inadequate to suppress disease activity and prevent radiographic progression in the majority of the patients with RA who were anti-hPAD4 -positive. The underlying mechanisms are unclear. It is possible that blocking of the TNF-α induced nuclear translocation of PAD413 by TNF inhibitors may skew the balance from nuclear to cytoplasmatic PAD4, possibly more susceptible to leak from dying cells and easier targeted by autoantibodies. The anti-PAD4 autoantibodies or anti-PAD4/PAD4 immune complexes could have direct pathogenic effects in RA. Moreover, in vivo interactions between anti-PAD4 IgG and PAD4 could increase the stability or activity of the enzyme, as shown in vitro for autoantibodies to tissue transglutaminase in coeliac disease.14 This would lead to increased substrate citrullination, fuelling the RA-specific immune responses contributing to the disease pathogenesis and progression. Noteworthy, continued inflammation and bone destruction in patients with RA on TNF inhibitors could be mediated by other proinflammatory cytokines, as interleukins 1 and 17.15

This observational study shows persistently elevated DAS28 and progressive vdHSharp erosion scores in most anti-hPAD4-positive patients on TNF inhibitors, suggesting that anti-hPAD4 IgG could potentially serve as a biomarker predicting severe and destructive disease despite TNF-inhibitor therapy. As our study is rather small, these results must be reproduced in larger RA cohorts, and further research on the potential roles of the anti-hPAD4 autoantibodies in the pathogenesis of RA is needed.


We thank Michiyuki Yamada for the hPAD4 construct, Inge C Olsen for statistical advice and Ludvig M Sollid for support and critical reading of the manuscript.


View Abstract

Supplementary materials


  • Competing interests: Hans Bijlsma was the Handling Editor for this article.

  • Funding: This work was supported by grants from Rikshospitalet University Hospital and the Research Council of Norway. EHH is financed by a PhD stipend from the University of Oslo.

  • ▸ Additional supplemental tables 1 and 2 are published online only at

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.