A common functional variant of endoplasmic reticulum aminopeptidase 2 (ERAP2) that reduces major histocompatibility complex class I expression is not associated with ankylosing spondylitis

Sir, The strong genetic association between AS and HLA-B27 has defied explanation for nearly 40 years. However, the additional discovery of a strong association between AS and ERAP1 (endoplasmic reticulum aminopeptidase 1), a gene that almost certainly operates in the trimming of peptides for optimal binding to MHC class I molecules, has rekindled hopes of rapid advances in this field [1]. It has been suggested that another aminopeptidase, endoplasmic reticulum aminopeptidase 2 (ERAP2), may act in concert with ERAP1, trimming residues inefficiently removed by ERAP1 [2]. The association described recently between ERAP2 and Crohn’s disease, which shares many clinical and genetic overlaps with AS [3], suggests that ERAP2 is worthy of further study in AS. An experiment of nature allows us to do this relatively simply.

ERAP2 has evolved under balancing selection, similar to the MHC, and includes a high-frequency variant (∼50%) that influences antigen presentation [4]. A single nucleotide polymorphism (SNP), rs2248374 (A to G), located within the 5′ canonical splice site of exon 10, results in an alternatively spliced ERAP2 mRNA that is degraded by nonsense-mediated decay (NMD). Homozygosity for the minor G allele (carried by ∼25% of the population) results in failure to express ERAP2 protein; in turn this genotype is also associated with reduced surface MHC Class I expression on human B cells [4]. Such a dramatic phenotypic effect provides a great opportunity for studying the potential role of ERAP2 in AS, analogous to using a knockout model.

We therefore specifically tested for differences in frequency of rs2248374 in 470 sporadic AS cases and 420 healthy, ethnically matched blood donors to determine whether this functionally important variant is implicated in AS. All patients were Caucasians, of UK origin and fulfilled the modified New York criteria for AS. Eighty-four per cent of the AS patients were HLA-B27 positive. All patients gave informed consent and ethical approval was obtained (Multicentre Research Ethics Committee 98/5/23). Genotyping involved restriction fragment length polymorphism (RFLP) analysis of PCR products performed under standard conditions. Primers (forward 5′-GCATCCATGGCTAATGTGCR and reverse 5′-GTTGTGGGAAAGCCGAACTA) amplified a 370-bp product that, in the presence of the G allele, was digested to 214 and 156 bp products by HphI. Genotype and allele frequencies in AS cases and controls were compared using the Cochrane–Armitage test of trend and the chi-squared test, respectively. This study had 80% power under a log-additive model to detect an odds ratio (OR) of 1.3 with a population risk of AS of 0.04% and a risk allele frequency of 46% (frequency in our controls) at a significance level of 0.05. There were no significant differences between allele or genotype frequencies in AS cases and controls at rs2248374 (Table 1). Additionally, due to the potential functional significance of homozygosity at the minor G allele (i.e. no functional ERAP2 protein), we also compared the frequency of GG homozygotes in AS cases and controls, but these proved almost identical again (Table 1).

Table 1.

Genotypes, minor allele frequency (MAF), OR, 95% CI and P-value for the ERAP2 SNP rs2248374

Patients and controls	Genotype counts			MAF	OR (95% CI)	P-value
	AA	AG	GG
AS cases (n = 470)	125	245	100	0.47	1.07 (0.77, 1.14)	0.46
Controls (n = 420)	128	201	91	0.46

The association between ERAP1 and AS is the second only to HLA-B27, but formal testing for additional association with the neighbouring gene, ERAP2, had not been previously undertaken. A study of familial AS that typed one marker within ERAP2 reported that an ERAP1/ERAP2 haplotype was over-represented in affected family members [5]. However, such studies are frequently confounded by linkage disequilibrium effects. If genetic variation at ERAP2 were to play a role in susceptibility to AS, then a dramatic loss-of-function variant, such as rs2248374, would appear to be a prime candidate for the ‘causal variant’ at this locus. However, we have now excluded any significant effect arising from this important functional ERAP2 variant. Nonetheless, the possibility of other associations between ERAP2 and AS has not been excluded by this study. This could be done only by systematic mapping of the gene with tagging SNPs in a very large sample after controlling for the significant linkage disequilibrium between ERAP1 and ERAP2 would this be achievable. Our study was insufficiently powered to do this.