Article Text

Extended report
Environmental and genetic factors in the development of anticitrullinated protein antibodies (ACPAs) and ACPA-positive rheumatoid arthritis: an epidemiological investigation in twins
  1. Aase Haj Hensvold1,
  2. Patrik K E Magnusson2,
  3. Vijay Joshua1,
  4. Monika Hansson1,
  5. Lena Israelsson1,
  6. Ricardo Ferreira3,
  7. Per-Johan Jakobsson1,
  8. Rikard Holmdahl4,
  9. Lennart Hammarström3,
  10. Vivianne Malmström1,
  11. Johan Askling1,5,
  12. Lars Klareskog1,
  13. Anca Irinel Catrina1
  1. 1Rheumatology Unit, Department of Medicine, Karolinska University Hospital, Karolinska Institutet, Stockholm, Sweden
  2. 2Swedish Twin Registry, Department of Medical Epidemiology and Biostatistics, Karolinska Institutet, Stockholm, Sweden
  3. 3Division of Clinical Immunology, Department of Laboratory Medicine, Karolinska Institutet at Karolinska University Hospital Huddinge, Stockholm, Sweden
  4. 4Medical Inflammation Research, Department of Biochemistry and Biophysics, Karolinska Institutet, Stockholm, Sweden
  5. 5Clinical Epidemiology Unit, Department of Medicine, Karolinska Institutet, Stockholm, Sweden
  1. Correspondence to Dr Anca I Catrina, Rheumatology Unit, Department of Medicine, Karolinska University Hospital, Karolinska Institutet, Stockholm 171 76, Sweden; anca.catrina{at}


Objective To investigate the role of genetic and environmental factors in the development of anticitrullinated protein antibodies (ACPA) and ACPA-positive rheumatoid arthritis (RA) in a twin cohort.

Methods A total of 12 590 twins were analysed for the presence of ACPAs (CCP2 ELISA), HLA-DRB1 shared epitope (SE) gene alleles, and exposure to smoking. Twins with established RA were identified in national public care registers. Antibody reactivities against citrullinated and native forms of α-enolase, vimentin, fibrinogen and type II collagen peptides were tested by ELISA in anti-CCP2-positive subjects and their cotwins. Structural equation models and ORs for the development of ACPA and ACPA-positive RA were computed for smokers and SE carriers.

Results A total of 2.8% (350/12 590) of the twins were ACPA positive, and 1.0% (124/12 590) had ACPA-positive RA. Most of the variability in the ACPA status was accounted for by non-shared environmental or stochastic factors (78%, 95% CI 55% to 100%) rather than shared environmental and genetic factors. Analysis of specific risk factors revealed an association between smoking and SE and the presence of ACPAs. Twins with ACPA-positive RA were more frequently SE positive than twins with ACPAs without RA. Reactivities against multiple citrullinated peptides were present in most twins with ACPA-positive RA but in fewer twins with ACPAs without RA.

Conclusions Environment, lifestyle and stochastic factors may be more important than genetics in determining which individuals develop ACPAs. Genetic factors (particularly SE) may have a relatively larger role in determining which ACPA-positive individuals will ultimately develop arthritis.

  • Ant-CCP
  • Smoking
  • Rheumatoid Arthritis
  • Autoantibodies

Statistics from


Rheumatoid arthritis (RA) is a chronic inflammatory disease resulting from complex interactions between genes and the environment.1 The presence of specific immunity against citrullinated proteins defines a clinical RA phenotype called anticitrullinated protein antibody (ACPA)-positive RA. Smoking may induce post-translational modifications in the lungs, resulting in the local accumulation of citrullinated proteins,2 ,3 which may be presented to T cells in the context of specific HLA-DRB1-shared epitope (SE) alleles with the subsequent initiation of a specific anti-citrulline immune response.4 This immune response can be present long before disease onset.5–8

Most of the epidemiological studies contributing to our current understanding of ACPA-positive RA have been conducted in cohorts of individuals with established RA.9–11 By contrast, few studies have addressed ACPAs development in the absence of RA, hampering the identification of distinct risk factors responsible for the development of the antibody response as opposed to those responsible for the development of antibody-positive disease. Studies aimed at identifying determinants for the development of ACPAs per se have mainly been performed on healthy relatives of RA patients. These studies have demonstrated a familial aggregation of ACPAs, but have not been designed to determine to which extent this familial aggregation is due to shared genetic and/or environmental risk factors.12–16 The risk factors responsible for development of ACPAs in the healthy general population, and how these risk factors relate to the development of RA in ACPA-positive individuals remain unclear. A European League Against Rheumatism (EULAR) task force group recently has recognised the importance of studies that aim to understand the gradual development of RA.17

Studies of twins provide a unique opportunity to assess genetic and environmental contributions to the occurrence of a certain biological trait. Two previous large twin studies have estimated the heritability of RA irrespective of ACPA status.18–20 Further analysis of a selected number of twins derived from a British twin cohort was performed to estimate the heritability of ACPA-positive and ACPA-negative RA,21 while a larger investigation of Danish twins estimated the concordance of ACPA-positive RA.22 However, twin methodology has not been used to address genetic and environmental contributions to the development of ACPAs per se in the absence and presence, respectively, of clinical RA.

Therefore, we aimed to estimate the role of genetic and environmental factors in the development of ACPAs in the presence and absence of clinical RA. To this end, we used a large, population-based sample of Swedish twins derived from the national Swedish twin registry and the associated Twingene biobank.23

Subjects, material and methods


A total of 22 390 twins from the national Swedish twin registry, born in 1958 or earlier, were contacted between 2004 and 2008 and asked to participate in the Twingene study and donate blood to the Twingene biobank. For details of the subject recruitment, sampling, and sample storage, see online supplementary text file 1 and references.23–25 The response rate was 56%, resulting in the recruitment of 10 002 twins in complete pairs, and 2588 without a participating cotwin, with available blood samples. Among twins in complete pairs, 28% were monozygotic (MZ) pairs, and 72% were dizygotic pairs. The ethics review board at Karolinska Institutet approved this study, and all participants gave written informed consent.

Evaluation of smoking habits

According to answers given in a health questionnaire collected once between 1998 and 2002, individuals were defined as smokers if they smoked or had smoked cigarettes, pipe tobacco, or cigars, and non-smokers if they reported never smoking, one-time smoking, or party smoking. The cumulative dose of smoking was measured as number of pack years. Further details are given in online supplementary text file 1.

RA diagnosis

Linkage between the study database and the Swedish National Patient Register from the National Board of Health and Welfare was performed using national social security numbers. The Swedish National Patient Register covers all public care and contains information about diagnosis at hospital discharges since 1964 and visits at outpatient non-primary care specialist clinics since 2001. An individual was considered to have RA if an RA diagnosis was listed at least once in 2009 or a prior year in the patient register (diagnose codes: ICD7: 72 200; ICD8: 712,38, 712,39; ICD9: 714A, 714B, 714C, 714W; ICD10: M05, M060, M068, M069). Further details are provided in online supplementary table S1. Previous validation using RA classification criteria and information from underlying medical files suggest that an RA diagnosis at hospital discharge or outpatient specialist care visit is correct in 90% of cases.26 ,27

ACPAs assays

All serum samples were analysed in duplicate for the presence of ACPAs using a commercial CCP2 ELISA (Immunoscan CCPlus, kindly provided by Euro Diagnostica, Malmö, Sweden). Samples were defined as ACPA positive if anti-CCP2 levels were ≥25 arbitrary units (AU)/mL. All anti-CCP2-positive samples were rerun at a later time point, and any sample with discordant results in the first two runs was tested a third time and defined as positive if two of the three assays were positive. All the samples with discordant results had anti-CCP2 levels lower than 75 AU/mL.

All anti-CCP2-positive serum samples (n=350) and, when available, corresponding serum samples from twin siblings (n=273) were analysed by an inhouse ELISA for the presence of specific reacitvities against the citrullinated and native forms of α-enolase peptide 1 (CEP-1, aa 5–21), triple-helical peptides of the C1 epitope of type II collagen (CII C1, aa 359–369), vimentin peptide (aa 60–75), and/or fibrinogen peptide (aa 563–583),28–33 (see online supplementary text file 1). The cut-off was established using an AU value at the 98th percentile for the control sera from 152 healthy donors.

An individual was defined as having ACPA-positive RA when the anti-CCP2 ELISA was positive in the presence of an RA diagnosis; as ACPAs without RA when the anti-CCP2 ELISA was positive in the absence of an RA diagnosis; and as any ACPA-positive when either anti-CCP2 or any of the other ACPAs were positive.

GWAS and imputation of HLA-DRB1 SE

DNA extracted from whole blood was genotyped with the OmniExpress bead chip (SNP&SEQ Technology Platform, Uppsala). Genotypes from the bead chip data were used for imputing HLA-DRB1 gene alleles (see online supplementary text file 1).34 Previous validation using information from high-resolution and low-resolution HLA-DRB1 genotyping compared with imputed data revealed a detection accuracy of 84–98%.35 HLA-DRB1 0101, 0102, 0401, 0404, 0405, or 1001 alleles were classified as SE alleles.35

Statistical analysis

We performed multiple univariate analyses to examine the relationships between ACPA positivity and gender, different categories of age, smoking status, different categories of smoking, and the presence and number of SE alleles using a binomial logistic regression model to adjust for the clustering of data and sex when appropriate. p Values were estimated by Wald and χ2 statistics.

The casewise concordances for ACPAs, ACPA-positive RA and ACPAs without RA were calculated for MZ and same-sexed dizygotics (SSDZ). No ACPA-concordant opposite-sexed dizygotic pair (OSDZ) was identified, and therefore no estimation of concordance was performed for OSDZ twins. The correlation of the liability within twins was estimated by calculation of tetrachoric correlation coefficients. Within-pair associations were estimated by tetrachoric correlation coefficients (r) and 95% CIs. Heritability was estimated by quantitative genetic analyses using structural equation models accounting for the contribution of additive genetic, shared environmental, and non-shared environmental factors (the so-called ACE model) or additive genetic, dominant genetic and non-shared environmental factors (the so-called ADE model). More details of the statistical analysis are provided in online supplementary text file 1.

Statistical analyses were performed using SAS 9.1 at a significance level of 0.05. Tetrachoric correlations were calculated with the PROC CORR procedure, and the binominal logistic regression for adjustment due to data clustering was calculated via the PROC GENMODE GEE procedure. Heritability estimations were conducted in the statistical package Mx ( V.1.7.03 May 17, 2009).


Prevalence of ACPA and ACPA-positive RA in the Twingene cohort

The demographic characteristics of the cohort are summarised in table 1. A total of 2.8% (350/12 590) of the twin individuals were ACPA positive, including 1.0% (124/12 590) who had ACPAs and RA, and 1.8% (226/12 590) who had ACPAs without RA. ACPAs were more common in females than males (OR=1.5, 95% CI 1.2 to 1.9). Half the ACPA-positive individuals (180/350, 51%) had at least one detectable reactivity against tested citrullinated peptides. The occurrence of multiple reactivities (at least 2) was more common in ACPA-positive individuals with RA compared with those without RA (p<0.05), and among individuals with high ACPA concentrations compared with those with low (<75 AU/mL) ACPA concentrations (p<0.05). Presence of multiple reactivities associated with smoking (OR=1.8, 95% CI 1.20 to 2.64, p<0.05) and SE (OR=11.5, 95% CI 5.52 to 24.93, p<0.05). Among the tested reactivities, antibodies against citrullinated-enolase were the most commonly detected (148/350, 42%).

Table 1

Demographic characteristics of the twin cohort

ACPA prevalence in relation to smoking status and the presence of SE

A total of 50% (6254/12 521) of all twins were smokers. Smoking was significantly associated with the presence of ACPAs (OR=1.30, 95% CI 1.05 to 1.61). Heavy smoking (more than 10 pack-years) was associated with ACPA-positive RA (OR=1.52, 95% CI 1.01 to 2.27) and ACPAs without RA (OR=1.42, 95% CI 1.05 to 1.93) (table 2). Twins with ACPA-positive RA, and twins with ACPAs without RA, were equally exposed to smoking and/or heavy smoking (see online supplementary table S2).

Table 2

Association between ACPAs and smoking

A total of 52% (5612/10 802) of all twins were SE positive. The presence of SE was associated with ACPAs (OR=2.21, 95% CI 1.72 to 2.84) and was associated more strongly with ACPA-positive RA (OR=7.24, 95% CI 3.96 to 13.22) and more weakly with ACPAs without RA (OR=1.42, 95% CI 1.06 to 1.90), with no changes following the imputation of missing data (n=1788, 14%). The presence of two SE alleles increased the strength of the association in all categories (table 3). Even with the imputation of missing SE data, twins with ACPA-positive RA were more frequent carriers of the SE alleles than twins with ACPAs without RA (p<0.05, see online supplementary table S3).

Table 3

Association between ACPAs and HLA-SE

Twin concordance rates for ACPA

Among 4998 complete twin pairs, six pairs were concordant and 279 pairs were discordant for ACPAs. Two ACPA concordant twin pairs were discordant for RA. The casewise concordance rates, and tetrachoric correlation coefficients for ACPA and ACPA-positive RA were low, but consistently higher for MZ compared with SSDZ twins (table 4). By contrast, the casewise concordance rates for ACPAs without RA were largely similar for MZ (3.7%) and SSDZ (2.8%) twins (table 4).

Table 4

Number and % of twin pairs concordant (+/+) or discordant (+/−), and tetrachoric correlations for ACPA positive, ACPA-positive RA and ACPA positive without RA

Heritability estimations

The low but consistently higher tetrachoric correlation coefficients for MZ compared with SSDZ twin pairs prompted us to estimate heritability. Reflecting the low correlation within twin pairs, a large proportion of the variability in the presence of ACPA (with or without RA) was accounted for by non-shared environmental factors and/or stochastic factors than shared environmental and genetic factors. The estimated variance assigned to genetic factors (heritability) in the full ACE model was 23% (95% CI 0 to 45%) for ACPA, 41% (95% CI 0 to 74%) for ACPA-positive RA, 10% (95% CI 0 to 43%) for ACPA without RA, 26% (95% CI 0 to 52%) for any-ACPA (table 5), and 39% (95% CI 0 to 66%) for RA (see online supplementary Table S4). The low correlation within MZ twin pairs for ACPA-positive RA and the Akaike criterion value favour a model that includes shared environmental factors (ACE model) rather than dominant genetic factors (ADE model). Details of the model-fitting results for complete and reduced ACE models and ADE models are summarised in online supplementary table S4.

Table 5

Heritability, shared environmental factors and non-shared unique environmental factors and their contribution to the variance of susceptibility of developing ACPA


The complex interactions between genetic (particularly SE) and environmental (of which smoking is the most prominent) factors are essential for the development of ACPA-positive RA, while less is known about the importance of these interactions for developing ACPAs. We took advantage of a unique population-based twin cohort, and demonstrated that despite the importance of genetic factors, environmental and stochastic factors are the main contributors to ACPA development.

We used a large population-based twin cohort at a high-risk age for RA development.36 The observed RA prevalence is 1.5%, which is in the same range as previously published Swedish prevalence estimates in this age group,37 suggesting that there was no selection toward/away from twins with RA during the stepwise recruitment of participants. The occurrence of SE is in accordance with a previously reported prevalence for the Swedish population,38 and no differences were observed when missing SE data (14%) were imputed. Some misclassification of individuals as non-smokers because smoking status was recorded, on average, 6 years prior to blood sample collection cannot be excluded, but the proportion of ever smokers in our cohort is in the range of those previously reported for the general Swedish population of comparable age.39 ,40 Additionally, the association of smoking with ACPAs and ACPA-positive RA is strongly related to duration of smoking, minimising the risk of bias due to this misclassification.

An important strength of our study is the large number of tested individuals, resulting in the largest population-based cohort in which ACPA prevalence has been estimated.41 ,42 The ACPA prevalence was 2.8% in the entire group of twins, and 1.8% when excluding individuals with established RA. The latter is in the higher range of previously reported prevalences in smaller population-based studies13 ,41 of healthy individuals, but lower than the previously reported prevalence for non-diseased family members of RA patients.12 ,14 ,16 In ACPA-positive individuals without RA, specific reactivities against citrullinated antigens were sparse, whereas the frequency of these reactivities in twins with RA was similar to previously reported frequencies in other cohorts of established RA.9 ,33 ,43

A central finding of our study comes from statistical modelling, which revealed that non-shared environmental and/or stochastic factors substantially contribute to the susceptibility of developing ACPAs. Previous twin studies have focused on estimating the twin concordance rates for and heritability of RA,18–20 ,22 ,44–46 and, to a lesser extent ACPA-positive RA, while none have studied ACPAs per se. Based on two of these previous studies, the heritability of RA was estimated to be 40% and 77%, respectively.18 Here, we report a non-significant estimated variance of 39% due to assigned genetic factors, which is somewhat lower but in line with older reports. A more recent Danish population-based twin study reported a lower heritability of RA of 12%, while suggesting (without calculating estimates) a higher heritability of ACPA-positive RA.22 However, this particular cohort also had a lower prevalence of RA (0.37%), most likely due to the younger age of the cohort (a mean age of 52 years, and inclusion of twins born between 1921 and 1982) compared to that used in our study (twins born before 1958) and previously published twin studies.19 One previous study attempted to calculate the heritability of ACPA-positive RA and reported an estimated heritability of 68% using a reduced AE model.21 This value is higher than the 44% heritability (95% CI 2 to 74) we find when using the same model. However, by contrast with our population-based cohort, this previous study was conducted in a subgroup of the original RA twin cohort, which might have resulted in a positive selection for diseased individuals.21

Another important finding in our study is the identification of smoking and SEs as risk factors for ACPA development. This finding is in some contrast to previous studies of healthy relatives and RA probands, which failed to detect a significant association between ACPA and SE or smoking.12–14 ,16 However, the numbers of individuals included in these previous studies were much smaller than the numbers of individuals included in this study, affecting the statistical power. Additionally, by contrast with our population-based study, these others studies were based on the active recruitment of individuals from selected families.

Interestingly, SE was mainly associated with high rather than low ACPA concentration, and with multiple rather than single ACPA reactivities (data not shown), suggesting a potential dichotomy in ACPA production based on concentration and epitope spreading. The higher risk of SE positivity among individuals with ACPA-positive RA compared with individuals with ACPAs without RA suggests that this genetic factor may be more important in determining which ACPA-positive healthy individuals will eventually develop RA than in determining which individuals will become ACPA-positive at the outset. The fact that only a minority of the twins with ACPAs without RA had high ACPA concentrations, or displayed reactivity to multiple autoantigen-derived citrullinated peptides further suggests that epitope spreading may be an important step in the longitudinal development of RA.7 ,43 ,47 Based on these results, we propose a stepwise model in which individuals will initially acquire reactivity toward single citrullinated epitopes as a direct consequence of environmental challenge with consecutive HLA class II-dependent epitope spreading and disease initiation and chronicity (figure 1).

Figure 1

Schematic representation of a model for stepwise development of ACPA and ACPA-positive RA.

In conclusion, we report a relatively low heritability of ACPA. Despite the importance of genetic factors, non-shared environmental and/or stochastic factors are essential for developing ACPAs and ACPA-positive RA. Our data support and add to the previously proposed stepwise model for developing ACPA-positive RA. According to this model, it appears that environment, lifestyle and stochastic factors are central in determining which individuals develop ACPAs. Genetic factors (particularly SE) may have a relatively larger impact in determining which ACPA-positive individuals will ultimately develop arthritis.


Supplementary materials


  • Handling editor Tore K Kvien

  • Contributors All authors had have substantial contributions to (1) conception and design, or analysis and interpretation of data, (2) drafting the article or revising it critically for important intellectual content and (3) final approval of the version to be published.

  • Funding This study was supported, in part, by research funding from the Swedish Foundation for Strategic Research, the European 7th framework program (FP7/2007–2013) Euro-TEAM (305549), Innovative Medicine Initiative Be the CURE, the Swedish Research Council, the European Research Council (ERC), Karolinska Institutet, and through a regional agreement on medical training and clinical research (ALF) between Stockholm County Council and Karolinska Institutet.

  • Competing interests AIC and LK report personal fees from Euro Diagnostica, outside the submitted work.

  • Ethics approval Ethic Review Board at Karolinska Institute.

  • Provenance and peer review Not commissioned; externally peer reviewed.

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.