Anti-citrullinated peptide antibody (ACPA) is a highly specific serological marker for rheumatoid arthritis (RA).1,–,3 Different HLA-DRB1 alleles have been shown to be associated with the susceptibility to ACPA-positive RA.4 5 Former studies demonstrated that HLA-DRB alleles carrying a shared epitope (SE),6 consisting of a conserved amino acid motif at positions 70–74 of the HLA-DRβ chain, were strongly associated with ACPA-positive RA and with higher ACPA levels in European and Japanese populations.7,–,9 On the other hand, HLA-DRB1*09:01 was recently found to be negatively associated with ACPA levels in the Japanese.9 These observations imply that combinations of HLA-DRB1 alleles differentially influence ACPA levels in ACPA-positive RA.

To address this question, we conducted a genetic association study employing 2457 ACPA-positive Japanese RA patients. ACPA was quantified by MESACUP CCP ELISA kit (MBL Co Ltd, Nagoya, Japan) with a cut-off level of 4.5 U/ml. The patients were then divided into three groups based on their ACPA titres: low (∼4.5–13.5 U/ml), intermediate (∼13.5–100 U/ml) and high (≥100 U/ml) groups. These groups were defined according to the 2010 ACR/EULAR classification criteria for RA and a measurement limit of the kit. HLA-DRB1 genotyping was carried out using either the Wakflow system (Wakunaga Pharmaceutical Co Ltd, Osaka, Japan) or the sequencing-based AlleleSEQR HLA-DRB1 typing kit (Abbott Japan, Nagoya, Japan). Frequencies of HLA-DRB1 alleles were compared among the three groups using the Cochran-Armitage Trend test. The relative predispositional effect (RPE) method was applied to identify the associations of more than one HLA-DBR1 allele sequentially according to their strength.10 Briefly, associations of HLA-DRB1 alleles with ACPA categories were estimated for each allele using the Cochran-Armitage Trend test. When we detected the strongest association with a significant p value, the allele was excluded from the whole data and the same steps were repeated until no further significant alleles were found.

As expected from the previous studies,9 HLA-DRB1*09:01 showed the strongest association with ACPA levels in a decreasing manner (p=1.0×10⁻²¹) and the SE alleles were significantly associated with an increasing effect (p=3.2×10⁻⁷) (table 1). In addition, HLA-DRB1*04:07 showed negative association with ACPA levels (p=0.0013), and HLA-DRB1*15:01 and HLA-DRB1*15:02 were positively associated with ACPA levels (p=2.3×10⁻⁵ and 0.0011, respectively) (table 1). Of note, the association between the SE and ACPA levels lost significance after stratification of HLA-DRB1*09:01 using RPE (p=0.16) whereas HLA-DRB1*04:07 and HLA-DRB1*15:01 remained significant after RPE (p=0.00034 and p=0.0011, respectively) (table 1). To confirm the dominant effect of HLA-DRB1*09:01 on ACPA levels over SE, we compared ACPA levels in two sets: first between HLA-DRB1*09:01/*09:01 and HLA-DRB1*09:01/SE or HLA-DRB1*09:01/X, and second between SE/HLA-DRB1*09:01 and SE/SE or SE/X. We found that HLA-DRB1*09:01 showed a significant association with low ACPA category compared with the other two groups in both sets of analyses (p<0.005, figure 1). On the other hand, we could not observe any difference between SE and the other alleles.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

Comparisons of blood anti-citrullinated peptide antibody (ACPA) levels among HLA-DRB1*09:01, shared epitope (SE) and other alleles in combination. Frequencies of three rheumatoid arthritis subgroups based on ACPA levels were compared among different HLA-DRB1 combinations containing HLA-DRB1*09:01 and/or SE. X indicates HLA-DRB1 alleles other than HLA-DRB1*09:01 and SE. ‘Low’, ‘intermediate’ and ‘high’ categories correspond to patients with ACPA titres of ∼4.5–13.5, ∼13.5–100 and ≥100 U/ml, respectively. **p<0.005 and ***p<0.00005. NS, not significant.

In this study, we aimed to identify HLA-DRB1 alleles showing quantitative effects on ACPA levels using a large collection of Japanese ACPA-positive RA patients. RPE was applied to avoid misleading frequency deviation by the allele with the strongest association to other associated alleles. We demonstrated that HLA-DRB1*09:01 was the strongest genetic determinant for lower ACPA levels, and the quantitative effects of HLA-DRB1 alleles carrying the SE were not a primary effect but merely an expected consequence of the decreased frequency of HLA-DRB1*09:01. We also identified two novel HLA-DRB1 alleles, HLA-DRB1*04:07 and HLA-DRB1*15:01, being associated with ACPA levels. It is interesting and feasible to perform similar studies in other populations and investigate whether or not the same set of HLA-DRB1 alleles are related to the quantitative effects beyond ethnicities and to examine if such alleles share conserved amino acid motifs.