Balancing selection maintains a form of ERAP2 that undergoes nonsense-mediated decay and affects antigen presentation

Abstract

A remarkable characteristic of the human major histocompatibility complex (MHC) is its extreme genetic diversity, which is maintained by balancing selection. In fact, the MHC complex remains one of the best-known examples of natural selection in humans, with well-established genetic signatures and biological mechanisms for the action of selection. Here, we present genetic and functional evidence that another gene with a fundamental role in MHC class I presentation, endoplasmic reticulum aminopeptidase 2 (ERAP2), has also evolved under balancing selection and contains a variant that affects antigen presentation. Specifically, genetic analyses of six human populations revealed strong and consistent signatures of balancing selection affecting ERAP2. This selection maintains two highly differentiated haplotypes (Haplotype A and Haplotype B), with frequencies 0.44 and 0.56, respectively. We found that ERAP2 expressed from Haplotype B undergoes differential splicing and encodes a truncated protein, leading to nonsense-mediated decay of the mRNA. To investigate the consequences of ERAP2 deficiency on MHC presentation, we correlated surface MHC class I expression with ERAP2 genotypes in primary lymphocytes. Haplotype B homozygotes had lower levels of MHC class I expressed on the surface of B cells, suggesting that naturally occurring ERAP2 deficiency affects MHC presentation and immune response. Interestingly, an ERAP2 paralog, endoplasmic reticulum aminopeptidase 1 (ERAP1), also shows genetic signatures of balancing selection. Together, our findings link the genetic signatures of selection with an effect on splicing and a cellular phenotype. Although the precise selective pressure that maintains polymorphism is unknown, the demonstrated differences between the ERAP2 splice forms provide important insights into the potential mechanism for the action of selection.

Figure 1. Allele site-frequency spectrum (SFS) of ERAP2, control regions, and ERAP1 in each population.

The X-axis reflects the absolute frequency of the derived allele, while the Y-axis reflects the frequency of that allele frequency bin in the generated data set. To account for missing data, the frequencies were projected to a sample size of 15 chromosomes. See the SFS of only coding SNPs in Figure S2.

Figure 2. Haplotype network of ERAP2 and ERAP1.

Circles represent haplotypes, with the areas proportional to the frequency of the haplotype (color-coded by population). The lines connecting the haplotypes have a length proportional to the number of mutations that differentiate the two haplotypes. Reticulations reflect recombinations or recurrent mutations. The ancestral state was inferred using the chimpanzee sequence data. For ERAP2, the four coding diagnostic SNPs are shown as white boxes; one nearly diagnostic SNP, which appears four times in the network due to the reticulations, is marked as thinner horizontal boxes. The ERAP2 haplotype network that includes all SNPs (coding and non-coding) is shown in Figure S5, and the ERAP2 haplotype network that includes the chimpanzee sequence is shown in Figure S6.

Figure 3. Haplotype-specific splicing of ERAP2.

A, The genomic organization of the human chromosome 5q15 region containing ERAP1 and ERAP2 is included at the top. The two haplotype-specific ERAP2 spliced forms are shown for Haplotype A (in blue) and Haplotype B (in purple). The different alleles of rs2248374 are shown as a blue or purple base position, respectively. The red boxes represent the premature stop codons in the Haplotype B mRNA. B, PCR amplification of cDNA across the exon 10 splice junction (see Materials and Methods) from the indicated 16 LCLs, with the haplotype status of each cell indicated as homozygote (AA or BB) or heterozygote (AB). A negative control PCR, with no DNA template, was also performed (water).

Figure 4. Immunoblot analyses of ERAP2 using LCL protein extract.

Two LCLs of each ERAP2 genotype (AA, AB, and BB) were tested for protein using primary antibodies specific to: A, ERAP2 (goat polyclonal); B, ERAP2 (mouse polyclonal); and C, ß-actin (see Materials and Methods).

Figure 5. Quantification of allele-specific ERAP2 mRNA levels in LCLs.

A, Locations of the four coding diagnostic SNPs across ERAP2 are shown, of which three (in red) were used to test for allele-specific expression. B, The allelic ratio of Haplotype B to Haplotype A ERAP2 cDNA levels, which was measured using these three coding diagnostic SNPs in the indicated heterozygote LCLs treated/untreated with emetine (NMD blocked), are depicted with colored bars. The control represents the allelic ratio measured with genomic DNA (gDNA), expected to be 1.0. The average allelic ratio across all cell lines tested (for a given SNP) is indicated above each set of bars. The error bars represent the standard error of the mean.

Figure 6. Standardized HLA-ABC mean fluorescence intensity of B-cells with various ERAP2 genotypes.

The distribution of observed levels of surface-expressed HLA-ABC for B cells of AA, AB, and BB individuals are graphically represented as boxplots (the blue box containing the 25th–75th percentile of the distribution, the black horizontal line indicating the median, the red dot reflecting the mean, and black circles representing outliers). Data are shown for two independent experiments (left and right). For each experiment, the significance level of the comparison between AA and BB homozygotes (T-test) is shown within the plot; the significance level of the effect of genotype in the global comparison between AA and BB homozygotes (two-way ANOVA) is shown above. A representative HLA-ABC fluorescence intensity plot is shown in Figure S7, and the mean fluorescence intensity boxplots of HLA-ABC and CD19 are presented in Figure S8.