Article Text

Extended report
Smoking, the HLA-DRB1 shared epitope and ACPA fine-specificity in Koreans with rheumatoid arthritis: evidence for more than one pathogenic pathway linking smoking to disease
  1. Benjamin A Fisher1,
  2. So-Young Bang2,
  3. Muslima Chowdhury3,
  4. Hye-Soon Lee2,
  5. Jae-Hoon Kim2,
  6. Peter Charles3,
  7. Patrick Venables3,
  8. Sang-Cheol Bae2
  1. 1Rheumatology Research Group, University of Birmingham, Birmingham, UK
  2. 2Department of Rheumatology, Hanyang University Hospital for Rheumatic Diseases, Seoul, Republic of Korea
  3. 3Nuffield Department of Orthopaedics, Rheumatology and Musculoskeletal Sciences, Kennedy Institute of Rheumatology, University of Oxford, London, UK
  1. Correspondence to Dr Benjamin A Fisher, Rheumatology Research Group, Centre for Translational Inflammation Research, School of Immunity and Infection, College of Medical and Dental Sciences, University of Birmingham Research Laboratories, New Queen Elizabeth Hospital, Birmingham, B15 2WB, UK; b.fisher{at} Prof Sang-Cheol Bae, Department of Rheumatology, Hanyang University Hospital for Rheumatic Diseases, Seoul 133-792, Korea


Objectives Data from North European rheumatoid arthritis (RA) populations has suggested a particularly strong association of gene-environment interaction between smoking and HLA-DRB1 shared epitope (SE) with antibodies to citrullinated α-enolase (CEP-1) and vimentin (cVim) peptides. We investigated this further by examining anticitrullinated peptide/protein antibody (ACPA) fine specificity in a Korean cohort, where there are notable differences in the RA-associated HLA-DRB1 alleles.

Methods Antibodies to fibrinogen (cFib), α-enolase (CEP-1) and vimentin (cVim) peptides and cyclic citrullinated peptide (CCP) were measured in 513 cases. The Mann-Whitney U test was used to compare antibody levels. Logistic regression generated ORs for RA in a case-control analysis with 1101 controls. Association of ACPA status and erosion in patients with RA was examined by logistic regression.

Results Anti-CCP, CEP-1, cVim and fibrinogen peptides were found in 86.7%, 63.9%, 45.5% and 74.7%, respectively. The number of ACPA and their levels were associated with SE, with evidence of a gene-dosage effect. There was a particular association of smoking with levels of anti-CEP-1. However, a gene-environment interaction was associated with all the ACPA positive subgroups, albeit the highest OR was seen with the anti-CCP+/cVim+ subset. In the absence of SE, smoking only conferred risk for anti-CCP negative subsets. The presence of erosions was not associated with the number of positive ACPA or specificity.

Conclusions The SE governed the magnitude and diversity of the ACPA response, but its interaction with smoking did not exclusively segregate with any of the ACPA specificities studied here. Smoking was associated with RA by SE-dependent and independent effects.

  • Ant-CCP
  • Rheumatoid Arthritis
  • Epidemiology

Statistics from

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.


Rheumatoid arthritis (RA) is an autoimmune disease of unknown aetiology. Genetic and environmental risk factors are important with heritability estimated at 60%.1 The most prominent genetic risk factor is the HLA-DRB1 shared epitope (SE) which encompasses alleles with a similar amino acid sequence at the P4 pocket of the peptide binding region.2 ,3 Smoking is the most well-established environmental factor.4 Antibodies to citrullinated peptides/proteins (ACPA) have been demonstrated to have a high diagnostic specificity for RA (95–98%), and initial studies suggested that the SE and smoking were only associated with ACPA-positive RA.5 ,6 Furthermore, these risk factors showed ‘biological interaction’,6 ,7 meaning that the risk for ACPA-positive RA associated with having both is more than the sum of the parts. These studies used the widely available anticyclic citrullinated peptide (anti-CCP) assay which is a generic test based on non-physiological peptides and capable of detecting ACPA of different specificities.8 Recent attention has focussed on the ‘fine specificity’ of the ACPA response using peptides derived from in vivo citrullinated autoantigens such as α-enolase, vimentin and fibrinogen.9–14 This pattern of reactivity varies between patients with many showing multiple ACPA specificities. We and others previously reported that the gene-environment interaction between smoking and the SE was strongly associated with the ACPA-positive subset characterised by antibodies to a citrullinated epitope from α-enolase (citrullinated enolase peptide 1; CEP-1).10 Similar findings were then reported for a partly overlapping subset characterised by antibodies to a peptide from citrullinated vimentin (cVim).11 The association of this gene-environment interaction with antibodies to specific citrullinated autoantigens promised insight into the aetiopathogenesis of this disorder. However, a recent detailed investigation by Willemze et al15 concludes that while smoking, the SE, and the gene-environment interaction between them, confer risk for ACPA-positive RA, only the SE governs the diversity of the ACPA immune response.

It is notable that much of the above data has been derived from North European or North American cohorts. In Asian populations there are important differences in the prevalence of the various HLA-DRB1 SE alleles,16 ,17 and the RA-risk allele *0901, which lacks the SE motif, is found more commonly.16 ,18 ,19 Xue et al20 found that the SE was associated with RA but not with ACPA production in a small Chinese RA cohort, whereas such associations have been reported in Malaysian, Chinese and Japanese patients.21 ,22 In a mixed Malaysian population, smoking, the SE and the interaction between them, only conferred risk for ACPA-positive RA.23 However, in a Korean population, Bang et al16 reported that the SE was associated with ACPA-negative and ACPA-positive RA. Given these differences we have investigated the association of the SE, HLA-DRB1*0901 and smoking with ACPA fine specificity in a Korean population with RA.



Serum was available for study from 513 patients with RA enrolled in the BAE RA cohort of Hanyang University Hospital for Rheumatic Diseases, which has previously been described.16 In brief, patients fulfilled the 1987 American College of Rheumatology (ACR) classification criteria for RA. Cigarette smoking before the onset of RA was recorded as never or ever. HLA genotype and smoking data was also available from 1101 controls. All subjects were native Koreans and gave informed consent. The study had a favourable ethical opinion from the Institutional Review Board of Hanyang University Hospital in Korea.


Numerous citrullinated antigens have been reported to be reactive with RA sera but only a few have been mapped to immunodominant peptides and their specificity and sensitivity examined in several laboratories. These include citrullinated α-enolase (amino acids 4-21: KIHACitEIFDS-Cit-GNPTVE; CEP-1),10 ,14 ,24–26 vimentin (amino acids 60-75: VYAT-Cit-SSAV-Cit-L-Cit-SSVP)9 ,11–14 ,27 and fibrinogen (cFib) β chain (amino acids 36-52: NEEGFFSA-Cit-GHRPLDKK).9 ,12–14 Serum samples were analysed for antibodies to the above peptides by ELISA. Cysteine residues were added at the amino-termini and carboxy-termini of each peptide to facilitate cyclisation, and antibodies to the corresponding arginine-containing control peptides were measured in parallel (see online supplementary text).

Anti-CCP antibodies were measured with the ImmuLisa CCP assay (IMMCO Diagnostics, Buffalo, New York, USA), according to the manufacturer's instructions.

HLA-DRB1 genotyping

The HLA-DRB1 genotypes of cases and controls were defined by direct DNA sequencing as previously described,16 using dye terminator chemistry and an ABI3100 Genetic Analyser, following PCR-based amplification according to the method of Kotsch et al.28 The HLA-DRB1 alleles considered to be SE were *0101, *0102, *0401, *0404, *0405, *0408, *0410, *1001, *1402 and *1406.

Radiological assessment

All patients with RA were classified into Steinbrocker stages as a marker of RA severity at the time of enrolment.29 Stage I is the absence of destructive changes, stage II, the presence of osteoporosis with or without slight cartilage or subchondral bone destruction, stage III requires evidence of cartilage and bone destruction and stage IV, the addition of bony ankylosis. For the purposes of this study, stage I was considered non-erosive and stages II–IV erosive, as previously described.30

Statistical analysis

The Mann Whitney U test was used to compare median antibody levels. Logistic regression, adjusted for age and sex, generated ORs for RA in the case-control analyses. The relationship of ACPA status to erosions in patients with RA was examined by logistic regression, adjusted for age, sex and disease duration. The presence of a smoking-SE gene-environment interaction was tested using the attributable proportion,10 ,31–33 with 95% CI,34 and also with the relative excess risk due to interaction and the synergy index.15 All statistical analyses were performed with PASW Statistics V.18.0 for Windows (IBM SPSS, USA).


Baseline characteristics

Baseline characteristics of the patients and controls are detailed in table 1. Anti-CCP, CEP-1, cVim and cFib were found in 86.7%, 63.9%, 45.5% and 74.7%, respectively. Among anti-CCP negative patients (n=68, 13.3% of patients), 39.7% had antibodies to either CEP-1, cVim or cFib (25.0%, 8.8% and 13.2%, respectively) so that only 8% of the whole RA population was ACPA-negative. Only one of four ACPA (anti-CCP, CEP-1, cVim, cFib) was found in 9% of patients with RA compared with 20% with two, 32% with three and 31% with four. Seven patients with anti-cFib antibodies, one with anti-cVim and nine with anti-CEP-1, were excluded from analyses as reactivity with the control peptide meant we were unable to be certain of citrulline specificity. Out of the 445 anti-CCP antibody positive subjects, anti-CCP antibody levels were available in 366.

Table 1

Basic demographics of RA cases and healthy controls

Relationship of specific ACPA to the HLA-DRB1 SE and smoking

The number of positive assays and the levels of specific ACPA were associated with the presence of the SE, with evidence of a gene dosage effect (table 2). It has previously been reported in a North European cohort, that the DR4 but not DR1 SE subtype was associated with the presence of anti-CEP-1 antibodies,10 and in our study only the DR4 SE subtype was associated with higher CEP-1 and cVim antibody levels. Higher levels to cFib were observed with DR1, DR4 and DR10 SE subtypes (table 3). The non-SE risk allele HLA-DRB1*0901 had no influence on specific ACPA levels when stratified for the presence of SE and non-risk alleles (data not shown).

Table 2

Antibody levels to cyclic citrullinated peptide (CCP) and citrullinated α-enolase (CEP-1), vimentin (cVim) and fibrinogen (cFib) peptides, among antibody positive patients with rheumatoid arthritis and in relation to the presence of the shared epitope

Table 3

Antibodies to citrullinated peptides/proteins (ACPA) levels according to HLA-DRB1 susceptible subtypes and smoking

Smoking was independently related to the development of RA using a multiple logistic regression analysis, which was performed with adjustments for age, sex and HLA-DRB1 SE alleles (OR 2.40; 95% CI 1.50 to 3.82) in the case-control analyses. This compared with an OR of 2.11 (95% CI 1.28 to 3.49) for anti-CCP positive RA, and for anti-CEP-1, anti-cVim and anti-cFib positive RA of 1.87 (95% CI 1.06 to 3.31), 2.16 (95% CI 1.13 to 4.16) and 2.04 (95% CI 1.19 to 3.48), respectively. Among antibody-positive subjects, only antibody levels to CEP-1 were significantly associated with a history of smoking (table 3).

Using a case-control approach, it has previously been demonstrated in this cohort that the SE and smoking are also risk factors for anti-CCP antibody negative RA.16 We therefore questioned whether these were risk factors for RA alone, or carried additional risk for ACPA-positive RA. We restricted analysis to subjects with RA and found that the SE, but not smoking, was associated with an increased risk for ACPA-positive RA, compared with ACPA-negative (table 4).

Table 4

Multiple logistic regression analysis, limited to patients with rheumatoid arthritis, of variables affecting ACPA formation

We further investigated whether this association with anti-CCP negative RA might be due to the presence of specific ACPA, which we found in 40% of this group. However even among the patients negative for all the ACPA tested (n=41), smoking and the SE still conferred risk with ORs of 2.54 (95% CI 1.24 to 5.18) and 2.35 (95% CI 1.25 to 4.40), respectively in an univariate analysis, and 4.89 (95% CI 1.74 to 13.75) and 2.38 (95% CI 0.40 to 3.39) in a multivariate analysis. There was no association of HLA-DRB1 *0901 with ACPA-negative RA in this analysis, but the number of cases who were *0901-positive but SE-negative was too small to draw conclusions (see online supplementary table S1).

Influence of the HLA-DRB1 SE on individual ACPA specificities

Willemze et al investigated whether the strong SE associations seen with some specific ACPA might simply be explained by the known association between the SE and ACPA status.15 By limiting the analysis to ACPA-positive patients in a Dutch cohort they found that the SE did indeed shape the specificity of the ACPA response with the strongest association being found with the vimentin peptide studied here. In our case control study, the OR for cVim-positive RA was 6.02 (95% CI 4.31 to 8.40), which did not differ much from that seen with anti-CEP-1 (5.66; 95% CI 4.25 to 7.56) and anti-cFib (4.54; 95% CI 3.49 to 5.91) positive RA. However when limiting the analysis to anti-CCP positive patients, we found a significant association between the SE and anti-CEP-1 (OR 2.45; 95% CI 1.53 to 3.79) and anti-cVim (OR 2.14; 95% CI 1.41 to 3.24), but not anti-cFib (OR 1.34; 95% CI 0.78 to 2.30) antibodies (see online supplementary table S2). In order to further analyse the relation between the SE and individual ACPA fine specificities, given that these overlap in many patients, we divided RA subjects into single, double and triple positive subsets (see online supplementary table S3). This revealed that the strongest associations were with the cVim peptide (OR 8.3; CI 1.7 to 39.3), and with double and triple positive subsets, however the small numbers in some of these groups mean that these findings should be interpreted with caution.

The relationship of the HLA-DRB1 SE and smoking gene-environment interaction and ACPA subsets

We next examined the interaction between smoking and the SE in relation to RA subsets defined according to the presence of anti-CCP and specific ACPA (figure 1 and online supplementary table S4). In contrast to previous findings in European populations,10 ,11 the SE-smoking gene-environment interaction did not segregate with individual ACPA-positive subsets albeit the highest OR was seen with the anti-CCP+/cVim+ subset (OR 15.8; 95% CI 7.3 to 34.0) highlighted by Willemze et al and Lundberg et al.15 ,35 However all the anti-CCP positive subsets showed evidence of a statistically significant gene-environment interaction, with attributable proportions ranging between 0.55 and 0.62, and similar relative excess risk due to interaction and synergy index values. In accordance with these findings, no evidence of a gene-environment interaction was observed with any subset when restricting the analysis to anti-CCP antibody cases only (see online supplementary table S5). Interestingly, in the absence of the SE, smoking only conferred risk for all ACPA double-negative subsets (anti-CCP-/CEP-1-, anti-CCP-/cVim-, anti-CCP-/cFib-) (see online supplementary table S4).

Figure 1

The relationship between smoking, the presence of the HLA-DRB1 SE and RA subsets defined according to the presence of anti-CCP and antibodies to CEP-1, cVim and cFib. *p<0.05, ***p<0.001. CCP, cyclic citrullinated peptide; CEP-1, citrullinated α-enolase peptide; cFib, fibrinogen peptide; cVim, vimentin peptide; RA, rheumatoid arthritis; SE, shared epitope.

Influence of ACPA fine specificities on erosive disease

Finally we analysed whether specific ACPA were associated with erosive RA (table 5). As might be expected, we found a shorter disease duration in patients labelled non-erosive and the subsequent analysis was corrected for disease duration, age and gender. We found no difference in the prevalence of any specific ACPA, or number of positive ACPA, between non-erosive (n=133) and erosive (n=380) RA (table 5).

Table 5

Relationship of specific ACPA to radiographic changes in patients with rheumatoid arthritis (RA)


We have studied, in a Korean population, antibodies to immunodominant epitopes from three of the most well characterised citrullinated autoantigens in RA. In contrast to our previous findings in a North European study, we did not find any ACPA-defined subset to have a pre-eminent association with a SE/smoking gene-environment interaction, albeit the highest OR was seen with the anti-CCP+/cVim+ subset. This is in accordance with the findings of Willemze et al who found this gene-environment interaction to be a risk factor for ACPA-positive RA as a whole, but no subset in particular.15 However, the SE alleles in general did govern the particular specificity of the ACPA response, as well as increasing its diversity and magnitude. Furthermore, the SE subtype had a differential effect on specific ACPA levels, with only DR4 alleles being associated with higher anti-CEP-1 and cVim antibody levels, whereas those of anti-cFib antibodies were positively associated with DR1, 4 and 10.

Our findings differ from the recent report of Lundberg et al, who found particularly strong gene-environment interactions associated with antibodies to cVim and CEP-1, and to combinations of ACPA when anti-CEP-1 was included.35 Their study differed from ours in being larger and an early arthritis inception cohort. They also found a much lower frequency of antibodies to the cFib peptide, which may reflect the use of a linear as opposed to a cyclic peptide.9 ,15 ,36 However it is important to note the population differences in SE alleles. The principal RA-associated DR4 alleles in Caucasian populations are *0401 and *0404. However these are rare among Koreans, each occurring in less than 2% of RA,17 whereas the predominant DR4 allele is *0405. It seems likely that the divergent make-up of the SE between populations will influence the measurable effect of the SE on individual ACPA specificities. Notably there was no influence of the non-SE RA risk allele *0901 on ACPA fine specificity.

An interesting finding in this cohort, as previously published, is the association of the SE and smoking with anti-CCP negative as well as anti-CCP positive RA.16 One possible explanation could be the presence of antibodies to other citrullinated epitopes among the anti-CCP negative population. Yet the association with the SE and smoking persisted even among those negative for all ACPA tested and raises the possibility of there being unmeasured antibodies to other citrullinated or even carbamylated autoantigens in this subset.37 It is interesting that an association of the SE with ACPA-negative RA has also recently been reported in a Caucasian population.19

The association of RA with smoking in this study is even more intriguing. First, the only subsets where smoking contributed risk in the absence of the SE, were the anti-CCP/specific ACPA double negative groups. Second, when restricting the analysis to patients with RA and controlling for the presence of the SE, smoking conferred no additional risk for ACPA-positive RA compared with ACPA-negative RA, unlike the SE. Third, there was evidence of an interaction between smoking and the SE which was only associated with the ACPA-positive subgroups. This data implies that there is more than one pathway by which smoking is associated with RA in this population.

It has been suggested that smoking confers risk for RA by inducing protein citrullination in the lung. In the context of inflammatory signals and the presence of the SE, this may lead to loss of immunological tolerance towards citrullinated proteins.6 ,7 While there is much data to support the presence of this gene-environment interaction, including our data here, there is also some evidence that smoking may operate as a risk-factor outside of this. Lee et al found evidence of smoking as a risk factor for anti-CCP antibodies, but no strong evidence for an associated gene-environment interaction.38 In a population-based study of older American women with RA, Criswell et al found smoking to only confer risk in SE-negative RA.39 Some of these discrepancies may be related to the quality of smoking data, for instance Karlson et al found no association with smoking until analysed according to dose.40 Furthermore, there are likely to be confounders such as air pollution41 or the presence of deletion polymorphisms in glutathione S transferases (GST) that reduce the capacity to metabolise tobacco toxins.42 An interaction between smoking and GSTT1-null polymorphisms has recently been associated with RA,43 and interestingly, in the Korean population, GSTT1-null homozygosity is present in more than 50%,44 whereas the frequency in Caucasian populations is only 15%.45 ,46 Nevertheless our data would suggest that smoking carries an additional risk for RA that is independent of the SE. This would seem likely since smoking is associated with other autoimmune diseases,47 and so its role in pathogenesis may not be limited to the citrullination of autoantigens.

An important consideration in this regard is the hypothesis that anti-CCP negative RA may consist of distinct diseases with a similar phenotype, although evidence to confirm this has not yet been provided.48 If anti-CCP negative RA is heterogeneous, it is possible that the make-up of this group differs between study populations, explaining the discrepant associations with the SE and smoking. The high anti-CCP antibody positivity we observed, although not different from that seen in the Korean Observational Study Network for Arthritis,49 may reflect it being a hospital-based cohort and so biased towards more severe disease.

In a case-only analysis, smoking influenced antibody levels to the CEP-1 (p=1.28×10−3). Interestingly, in an analysis limited to ACPA-positive patients, Willemze et al reported that the only association of tobacco exposure was with anti-CEP-1 antibodies.15 This, together with the data of Lundberg et al cited above,24 ,35 does suggest a particular influence of smoking on antibodies to this CEP-1.

The lack of association with erosions must be interpreted with caution because of the cross-sectional nature of the study, the lack of correction for disease-modifying antirheumatic drugs (DMARD) use and dose, and the limitations of the radiological scoring system employed. Nevertheless the results are consistent with North European longitudinal studies suggesting that ACPA fine specificity is not predictive of erosive change.50 ,51

In conclusion, we have provided evidence that the SE governs the specificity, magnitude and diversity of the ACPA response in a Korean RA population, but that its interaction with smoking does not have a dominant association with any of the ACPA specificities studied here. These findings differ from those in certain North European populations, and may reflect differences in the frequencies of alleles that comprise the SE. Our data on smoking implies first, that there are population differences in ACPA-negative RA, and second, that there are SE-dependent and independent pathways by which smoking confers risk in this RA cohort.


Supplementary materials

  • Supplementary Data

    This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.

    Files in this Data Supplement:


  • Handling editor Tore K Kvien

  • BAF and S-YB are joint first authors. BAF and S-CB are joint corresponding authors.

  • Correction notice This article has been updated since it was published Online First. The Objectives section of the Abstract has been corrected.

  • Contributors BAF, S-YB, S-CB and PV designed the study. H-SL, J-HK, MC and PC undertook laboratory work. BAF and S-YB interpreted results and drafted the paper, which was reviewed by all authors.

  • Funding This study was supported by the National Institute of Health Research (NIHR), Arthritis Research UK, and a grant from the Korea Healthcare technology R&D Project, Ministry for Health and Welfare, Republic of Korea (A102065).

  • Competing interests None.

  • Ethics approval Institutional Review Board of Hanyang University Hospital, Republic of Korea.

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