Production of anti–citrullinated protein antibodies (ACPAs) is an important biomarker for rheumatoid arthritis (RA). We undertook this study to determine whether genetic factors (HLA—DRB1 alleles) are associated with extreme ACPA levels in individuals with ACPA-positive RA, and to ascertain whether there are any phenotypic characteristics associated with these subgroups of RA.
HLA–DRB1 allelic groups were genotyped in 1,073 ACPA-positive RA patients from the Swedish Epidemiological Investigation of Rheumatoid Arthritis study. We found that 283 patients (26.4%) had high ACPA levels (defined as >1,500 units/ml using the Euro-Diagnostica anti-CCP2 test), while the rest of the patients had moderate ACPA levels and served as the comparison group. A replication group consisted of 235 RA patients.
No significant differences in baseline disease activity were observed between patients with high and those with moderate ACPA levels. However, the HLA–DRB1*15 allele was associated with high ACPA levels (P = 0.0002). A similar trend was detected in HLA–DRB1*15–positive patients in the replication cohort, with meta-analysis of the discovery and replication cohorts demonstrating an overall effect of HLA–DRB1*15 on development of high ACPA levels in both the discovery and replication cohorts (P < 0.0001 by Mantel-Haenszel test with a fixed-effects model).
Our data indicate that HLA–DRB1*15 may promote the production of high ACPA levels. Due to the high value of ACPA level scores in the 2010 American College of Rheumatology/European League Against Rheumatism classification criteria for RA, the presence of HLA–DRB1*15 may, at least in part, contribute to fulfilling the criteria for RA. This illustrates the complex nature of the genetic regulation of ACPA levels. Additional mechanistic studies of the regulation of ACPAs and ACPA-positive RA are pending.
Production of anti–citrullinated protein antibodies (ACPAs) is an important biomarker for a major subgroup of patients with rheumatoid arthritis (RA) (1). It has been demonstrated that the development of these antibodies in general is associated with HLA–DRB1 shared epitope (SE) alleles (2–5). ACPA positivity is known to be associated with more progressive, more destructive RA and extraarticular manifestations as well as cardiovascular disease (6–8). Furthermore, ACPA positivity has been reported to be associated with worse response to anti–tumor necrosis factor therapy (9).
Interestingly, the level of ACPAs in ACPA-positive RA does not correlate with disease activity (10). Nevertheless, there is a distinct group of ACPA-positive RA patients characterized by very high ACPA levels close to and even exceeding the upper detection limit of routinely available diagnostic enzyme-linked immunosorbent assays (ELISAs). Since ACPA is a very specific biomarker for RA, and since the presence or absence of ACPAs in RA influences the course of disease and depends on SE alleles, it is interesting to study the phenomenon of very high production of ACPAs from both a clinical and a genetic perspective. So far, no data are available to confirm whether there is a distinct genetic predisposition underlying very high ACPA levels. HLA–DRB1*03 has already been implicated in association with lower ACPA levels in previous studies (11, 12). However, due to the small size of these studies, the study design, and the strong influence of SE alleles on development of ACPA-positive RA overall, the question remains as to whether this is a true association (13, 14). Our aim was to determine the influence of HLA–DRB1 alleles on the production of very high levels of ACPAs in comparison with moderate levels in ACPA-positive RA and to ascertain whether there is any phenotypic characteristic associated with these subgroups of RA.
PATIENTS AND METHODS
This study was approved by ethics review boards at the Karolinska Institutet, and informed consent was obtained from all participating patients. The cohort of 1,073 ACPA-positive RA patients (764 women and 309 men) from the Swedish Epidemiological Investigation of Rheumatoid Arthritis study was included in the analysis (Table 1). RA patients with ACPA levels below the diagnostic threshold (25 units/ml) at baseline were not included in the study. All cases were newly diagnosed, and the diagnosis of RA was made according to the 1987 revised criteria of the American College of Rheumatology (ACR) (15). Data on the Disease Activity Score in 28 joints (DAS28) (16) from the time of diagnosis were available for 515 patients. Hand and foot radiographs were evaluated at baseline (i.e., time of diagnosis) and 1 and 2 years after diagnosis. Larsen scores (17) were currently available for 173 patients. Information about smoking exposure and detailed cigarette smoking history among patients was available for 979 of the 1,073 patients, obtained by questionnaire as described previously (5). The quantification of smoking was based on the smoking history before first symptoms of arthritis as described previously (18). All smokers were cigarette smokers; a minor number of subjects who had smoked pipes or cigars were excluded from the study (18).
Table 1. Baseline characteristics in the EIRA and replication cohorts of ACPA-positive patients with rheumatoid arthritis*
EIRA cohort patients
Replication (non-EIRA) cohort patients
High ACPA levels (n = 283)
Moderate ACPA levels (n = 790)
High ACPA levels (n = 48)
Moderate ACPA levels (n = 187)
There were no significant differences between the groups in either cohort. EIRA = Epidemiological Investigation of Rheumatoid Arthritis; ACPA = anti–citrullinated protein antibody; NA = not available; DAS28 = Disease Activity Score in 28 joints.
The replication cohort consisted of 235 ACPA-positive RA patients from Karolinska University Hospital who fulfilled the 1987 ACR criteria and were tested for ACPA and HLA–DRB1 alleles (see below and Table 1). There was no overlap between the cohorts. For 3 of the 235 patients in the replication cohort, genetic data were not available.
HLA typing and ACPA level.
HLA–DRB1 alleles were detected by sequence-specific primer–polymerase chain reaction genotyping with the DR low-resolution kit (Olerup SSP). ACPA concentrations were measured at the time of diagnosis by Immunoscan RA (Mark 2) anti-CCP2 ELISA (Euro-Diagnostica). A level of >25 units/ml was regarded as being positive according to instructions in the kit and as confirmed by the Clinical Immunology Laboratory at Uppsala University Hospital. The cutoff between moderately and highly elevated ACPA levels was defined as 1,500 units/ml. Using a provisional cutoff of 1,500 units/ml, we found that 283 patients (26.4%) had a high ACPA level; the remaining patients were considered to have a moderate ACPA level. In the replication cohort, a similar proportion of patients (48 of 235 individuals [20.4%]) had an ACPA level of >1,500 units/ml (P = 0.06 versus the discovery cohort, by chi-square test with Yates' correction).
The chi-square test was used to analyze the association of allele frequencies. Odds ratios (ORs) with 95% confidence intervals (95% CIs) were calculated to assess effect size. An unpaired t-test and Mann-Whitney U test were used to compare age at disease onset and DAS28 values. The Mantel-Haenszel test with a fixed-effects model was used for meta-analysis. The GraphPad Prism 4.0 software package and ManRev 5 were used for the analysis. P values less than 0.05 were considered significant.
Association of HLA–DRB1 alleles with high ACPA level in RA patients.
To address the possible association of HLA–DRB1 alleles with the level of ACPA production, we first tested it in a dominant model for RA patients with moderate ACPA levels in comparison with RA patients with high ACPA levels. Among all groups of DRB1 alleles, initially we observed HLA–DRB1*03 to be negatively associated with high ACPA levels (P < 0.0001, OR 0.38 [95% CI 0.23–0.60]) and HLA–DRB1*15 to be positively associated with high ACPA levels (P = 0.0002, OR 1.82 [95% CI 1.33–2.48]) (Table 2). However, no association between high ACPA levels and SE alleles was detected in comparison with the group with moderate ACPA levels.
Table 2. Association between HLA–DRB1 allelic groups and very high ACPA levels in patients with ACPA-positive RA*
RA patients with moderate ACPA levels, no. carriers/no. noncarriers of the allele
RA patients with high ACPA levels, no. carriers/no. noncarriers of the allele
Since we knew that SE alleles are strongly associated with development of ACPA-positive RA, we performed stratification for the carriage of SE alleles in DRB1*03-positive and DRB1*15-positive individuals for association with high ACPA levels. This analysis demonstrated that the protection against development of high ACPA levels by HLA–DRB1*03 was not independent from SE alleles (Table 3). In the replication cohort of 232 ACPA-positive RA patients, a similar trend was observed (data not shown).
Table 3. Test for independence from SE alleles of the association of HLA–DRB1*03 and of HLA–DRB1*15 with high levels of ACPAs in RA patients*
No. of patients with high ACPA levels/ no. of patients with moderate ACPA levels
NA = not applicable (see Table 2 for other definitions).
By chi-square test.
No SE/no HLA–DRB1*03
No SE/no HLA–DRB1*15
On the other hand, the HLA–DRB1*15 allele was more frequent in individuals with high ACPA levels, independent of SE alleles (P = 0.0136, OR 2.51 [95% CI 1.19–5.29]) or in combination with SE alleles (P = 0.0011, OR 2.94 [95% CI 1.51–5.70]) (Table 3). Again, SE alleles without the HLA–DRB1*15 allele did not show any association with high ACPA levels (Table 3).
Since an individual who carries an SE allele and an HLA–DRB1*03 allele at the same time cannot carry an HLA–DRB1*15 allele, we wanted to clarify the possible role of these alleles and to adjust effects from DRB1*03 and DRB1*15 in one test for SE alleles' presence. The data showed that only HLA–DRB1*15, alone (P = 0.0112, OR 3.21) or in combination with the SE (P = 0.0051, OR 3.13), remained associated with high ACPA levels (Table 4). HLA–DRB1*03 or SE alleles alone or in combination with each other did not show any association with ACPA level in ACPA-positive RA patients.
Table 4. Test for independence from SE alleles of the association of HLA–DRB1*15 with high levels of ACPAs in RA patients*
No. of patients with high ACPA levels/ no. of patients with moderate ACPA levels
Since the effect was truly significant only for association with HLA–DRB1*15, we decided to limit replication analysis to this specific finding. A similar tendency was observed in the replication cohort. There were 43 patients carrying the HLA–DRB1*15 allele (30 with moderate ACPA levels and 13 with high ACPA levels), while there were 189 noncarriers of the HLA–DRB1*15 allele (155 with moderate ACPA levels and 34 with high ACPA levels) (P = 0.0714 by chi-square test, OR 1.98 [95% CI 0.93–4.18]). There were 186 patients carrying the SE allele (146 with moderate ACPA levels and 40 with high ACPA levels), while there were 46 noncarriers of the SE allele (39 with moderate ACPA levels and 7 with high ACPA levels) (P = 0.3421 by chi-square test, OR 1.53 [95% CI 0.63–3.67]). (P values represent carriers versus noncarriers for both the HLA–DRB1*15 and the SE alleles.) A meta-analysis of the discovery and replication cohorts demonstrated a strong effect of HLA–DRB1*15 on the development of ACPAs (Figure 1), with no heterogeneity between the studies.
To address smoking status as a possible risk factor for the development of high levels of ACPAs in RA patients, we compared the proportion of ever smokers between those with high and those with moderate ACPA levels. Our data did not show a significant association of smoking with ACPA level (Table 1), although smoking was slightly more common in the group of RA patients with a high ACPA level (OR 1.35 [95% CI 0.97–1.87]).
However, when patients were stratified for smoking status, the carriage of the HLA–DRB1*15 allele was found to be more strongly associated with a high ACPA level (>1,500 units/ml) only in ever smokers. Among ever smokers who carried the HLA–DRB1*15 allele, we observed 64 patients with high ACPA levels and 97 patients with moderate ACPA levels. This contrasts with the finding of 50 patients with high ACPA levels and 176 patients with moderate ACPA levels in the reference group (never smokers who did not carry the HLA–DRB1*15 allele) (P = 0.0002, OR 2.32 [95% CI 1.49–3.63]). No other combinations differed from the reference group. Any conclusions should be made with caution, however, since no data about smoking status were available in the replication cohort and we were unable to repeat the analysis. In a similar analysis of SE alleles and smoking, no association of either single or combined factors with high ACPA levels was detected in comparison with the group with moderate ACPA levels (data not shown).
Age at onset of the disease.
We could not detect any significant difference in the age at onset of the disease between patients with high ACPA levels and those with moderate ACPA levels (Table 1). However, according to our observations in the discovery cohort, ever smokers who carried the SE allele, the HLA–DRB1*15 allele, or both alleles developed RA significantly later, at the mean ± SEM ages of 52.49 ± 0.48 years, 53.71 ± 1.48 years, and 52.74 ± 1.00 years, respectively (P = 0.0008, P = 0.0008, and P = 0.0013, respectively), than never smokers who carried neither the SE allele nor the HLA–DRB1*15 allele (mean ± SEM age 44.46 ± 2.36 years). However, in the replication cohort, the SE and HLA–DRB1*15 alleles alone or in combination were not associated with delayed onset of disease (data not shown).
Testing of disease activity markers.
When addressing possible differences in baseline disease activity and in bone erosion in relation to ACPA level, we could not find any significant difference in baseline DAS28 values between patients with high and those with moderate ACPA levels. When we compared RA patients with moderate and those with high ACPA levels, we found that smoking, the carriage of the SE allele, the carriage of the HLA–DRB1*15 allele, or any combination of these factors did not affect baseline DAS28 or Larsen scores significantly in our study (data not shown).
The novel observation of our study was that a non-SE HLA–DRB1 allele, namely, HLA–DRB1*15, is associated with production of high levels of ACPAs in RA patients. It is important to differentiate this issue from association with production of ACPAs in the general population. All previously reported studies have shown association of HLA alleles with production of ACPAs (12), while we report association with high production of ACPAs.
HLA–DRB1*15 has been previously reported to be associated with anti-SSA antibody positivity in primary Sjögren's syndrome (19, 20) and with a high level of natural antibodies in healthy individuals (21). HLA–DRB1*15 was also found to be associated with renal involvement (22) and with secondary Sjögren's syndrome (23) in RA. These observations together with ours support the hypothesis that HLA–DRB1*15 may play a specific role in the production of high levels of autoantibodies. In the latter study, HLA–DRB1*15 was also found to be associated with RA in RA patients lacking the SE (23), although we did not replicate this finding in our own previous work (13).
Cigarette smoking is the strongest proven environmental risk factor for the development of RA (24–26). Although interaction of smoking and HLA–DRB1 SE alleles has been shown to confer high risk for the development of rheumatoid factor–positive or ACPA-positive RA (5, 27, 28), we did not find this gene–environment combination to have any effect on ACPA levels. We may speculate that HLA–DRB1*15 serves as a second trigger that may aggravate autoimmune response in individuals with predisposition to RA caused by the SE, other genetic factors, or early life exposure to an environmental risk factor, when the first trigger (smoking, the SE, or another) breaks the tolerance to citrullinated protein antigens. However, we did not have a large enough cohort to confirm these findings in an independent study, and inferences about the influence of smoking on the level of ACPAs in ACPA-positive RA patients should be made with caution.
Irigoyen et al found HLA–DRB1*03 to be associated with a lower level of anti–cyclic citrullinated peptide (anti-CCP) antibodies, even when adjusted for the presence of the SE (11). Cui et al reported that a single-nucleotide polymorphism in linkage disequilibrium with HLA–DRB1*03 was associated with anti-CCP levels below the diagnostic threshold (12). They also considered HLA–DRB1*03 a factor that could potentially influence anti-CCP levels independent of HLA SE alleles (12). In fact, all of these studies compared anti-CCP–negative with anti-CCP–positive disease, but they did not address the question of ACPA levels in seropositive RA. Our data show that neither HLA–DRB1*03 nor the SE alone affected ACPA production in ACPA-positive RA. It is difficult to determine association with ACPA levels outside the range (upper and lower thresholds) of the ELISA kits due to their inaccuracy outside this range.
An interesting association was observed in our study. The SE allele and/or HLA–DRB1*15 allele together with smoking were significantly associated with late RA onset in the discovery cohort, but this finding was not replicated.
One may speculate that the relatively lower ACPA level in individuals who have earlier disease onset may be explained by more active “consumption” of autoantibodies in joints and more active disease and severe joint damage. However, our analyses failed to detect any associations of ACPA level, the presence of the SE allele and/or the HLA–DRB1*15 allele, and smoking with disease severity or radiographic progression. In a relatively small study, it was found that serial anti-CCP measurements during the first 3 years of RA may be better predictors of radiographic progression than baseline anti-CCP level (29). In accordance with our findings, however, Mattey et al failed to demonstrate any association between HLA–DRB1*15 and erosive disease or nodule formation (23).
A high ACPA level contributes 3 points toward the 2010 ACR/European League Against Rheumatism (EULAR) classification criteria for RA (30). In this clinical context, “highly positive” refers to detected values >3 times the upper limit of normal laboratory test results in healthy individuals (i.e., >75 units/ml). In our cohorts the definition of high ACPA levels meant remarkably higher ACPA levels (i.e., >1,500 units/ml). Nevertheless, the group of patients with high ACPA levels associated with the HLA–DRB1*15 allele is included in the group of patients who score the highest in the serology part of the new ACR/EULAR classification criteria for RA. Therefore, the presence of HLA–DRB1*15 may, at least in part, contribute to fulfilling the criteria for RA, which emphasizes the clinical significance of our observation. Our study shows that this contribution is independent of SE alleles, while there is a possible protective effect against development of high ACPA levels conferred by HLA–DRB1*03 linked to SE alleles.
In summary, we have found and confirmed that HLA–DRB1*15 alleles are increased in RA patients with high ACPA levels in comparison with patients with moderate ACPA levels. Our finding of a relationship between high ACPA levels and HLA–DRB1*15 further emphasizes the importance of genetic background and adaptive immunity in development of autoantibody-positive 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. Laki 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. Laki, Lundström, Snir, Bengtsson, Alfredsson, Klareskog, Padyukov.