Divergent Allele Advantage at Human MHC Genes: Signatures of Past and Ongoing Selection

Abstract

The highly polymorphic genes of the major histocompatibility complex (MHC) play a key role in adaptive immunity. Divergent allele advantage, a mechanism of balancing selection, is proposed to contribute to their exceptional polymorphism. It assumes that MHC genotypes with more divergent alleles allow for broader antigen-presentation to immune effector cells, by that increasing immunocompetence. However, the direct correlation between pairwise sequence divergence and the corresponding repertoire of bound peptides has not been studied systematically across different MHC genes. Here, we investigated this relationship for five key classical human MHC genes (human leukocyte antigen; HLA-A, -B, -C, -DRB1, and -DQB1), using allele-specific computational binding prediction to 118,097 peptides derived from a broad range of human pathogens. For all five human MHC genes, the genetic distance between two alleles of a heterozygous genotype was positively correlated with the total number of peptides bound by these two alleles. In accordance with the major antigen-presentation pathway of MHC class I molecules, HLA-B and HLA-C alleles showed particularly strong correlations for peptides derived from intracellular pathogens. Intriguingly, this bias coincides with distinct protein compositions between intra- and extracellular pathogens, possibly suggesting adaptation of MHC I molecules to present specifically intracellular peptides. Eventually, we observed significant positive correlations between an allele's average divergence and its population frequency. Overall, our results support the divergent allele advantage as a meaningful quantitative mechanism through which pathogen-mediated selection leads to the evolution of MHC diversity.

Keywords: HLA; balancing selection; heterozygote advantage; human evolution; pathogen-mediated selection.

Fig. 1.

In silico evidence for divergent allele advantage across five classical human MHC genes. Correlation between pairwise genetic distances reported as Grantham distance (x axes) and number of bound peptides (y axes) counted for all possible pairs of common HLA alleles. Each dot represents an allele pair. Binding prediction analyses performed on the complete data set of pathogen proteins (n = 232). Linear model (red line) and smoothed lowess curve (dashed blue line), describing the association between the combined number of bound peptides and pairwise Grantham sequence divergence. Note the different axis scales.

Fig. 2.

Different origins of pathogenic peptides. Correlation between pairwise genetic distances reported as Grantham distance (x axes) and number of bound peptides (y axes) calculated for all possible pairs of common HLA alleles. Each dot represents an allele pair. Binding prediction analyses performed considering proteins within three groups of pathogens: extracellular (orange, n = 57 proteins), intracellular (green, n = 100), and intra-extracellular (purple, n = 75).

Fig. 3.

Multidimensional scaling plot of amino acid composition in intracellular and extracellular pathogen proteins. Multidimensional scaling (MDS) based on amino acid frequencies indicates similarity in amino acid composition among individual proteins (dots). Intracellular proteins (n = 100) are reported in green while extracellular proteins (n = 57) in orange. MDS enables a standardized unit-less representation of variation among data points in 2D space (along perpendicular axes MDS1 and MDS2): location of proteins within the plot is indicative of potential bias toward specific amino acids (blue characters in one letter code), proteins with more similar amino acid composition are displayed closer to each other. The dashed circles indicate 95% confidence intervals for each group. Stress for 2D representation: 0.21.

Fig. 4.

Population frequency of divergent HLA alleles. Correlation between the average Grantham pairwise divergence to the most common alleles and the allele frequency in the USA NMDP European Caucasian (N = 1,242,890), German (N= 39,689), and Polish (N= 20,653) populations, for four classical HLA loci with available allele frequency data in AlleleFrequencies.net. Significant associations that persisted after Bonferroni correction across populations are reported with a solid line, while dashed lines indicate associations that are only nominally significant (P < 0.05 before Bonferroni correction; for exact values see supplementary table S9, Supplementary Material online).