Functional characterization of antigen repertoires in HLA-associated complex diseases to investigate antagonistic selection on HLA genes

The interaction between hosts and pathogens represents a major driver of their evolution, and a core interest in the evolutionary biology. In vertebrate hosts, the classical genes of Major Histocompatibility Complex (MHC) make a central component of their adaptive immune system. The cell-surface molecules encoded by the MHC genes present peptide fragments (derived from both self-antigens and pathogens) to T-cells, which upon recognizing them as foreign, initiate a specific immune response. To explore the general MHC evolution, I used humans as a study system where MHC is designated as Human Leukocyte Antigen (HLA). Although a vast allelic diversity of HLA genes exists at the population level, potentially maintained by pathogen-mediated balancing selection, only a fraction of that is seen at the individual level in the form of a limited number of HLA genes and their alleles. This setting has been proposed as an optimum between HLAconferred resistance to pathogens and the risk of autoimmunity, which represent major antagonistic selection forces on HLA genes. The fine-mapping of various HLA’s association with different infectious and autoimmune diseases has suggested a major role of the peptide-repertoire of HLA alleles in determining their specific effect on diseases. However, the exact mechanisms that underlie HLA’s association with the diseases, and by that modulate the antagonistic selection on HLA genes remain elusive. In order to elucidate them, I started by investigating the functional basis of the previously known protective effect of HLA heterozygosity at the HLA class-I genes on HIV-1 progression. I used a dataset of 6,311 HIV-1 infected individuals and predicted the HLAbound peptides derived from HIV-1 proteome for each HLA allele represented in the dataset. The individual-specific repertoire of HLA-bound peptides suggested that HLA heterozygote advantage against HIV-1 could be mediated by both a broader array of HLAbound peptides and a higher likelihood of carrying specific protective alleles in heterozygotes compared to homozygotes. The comparison of the peptide-repertoire of risk and protective alleles suggested that individual alleles could confer disease control by binding either a large number of peptides or specific immunodominant peptides. The separate analysis of the individual HLA genes indicated that different mechanism for the heterozygote advantage might work at different genes, such as either T-cell or NK-cellmediated immune attack on the virus, possibly resulting in different evolutionary constraints on different HLA genes. Overall, the findings suggested that the pathogenmediated selection might favor both HLA heterozygosity and individual alleles. hypothesizing that not all HLA-bound peptides would be relevant for disease control, we developed a new approach, named Peptidome-wide association study (PepWAS), that can predict HLA-bound disease-associated epitopes from a given peptidome. The PepWAS-predicted HIV-1-associated epitopes accounted for as much variation (12%) in HIV-1 viral load as by the genetic variants in HLA class-I genes, providing a functional basis for the association between HLA and HIV-1 control. I then focused on the association between HLA class-II genes and Type 1 Diabetes (T1D). Using a case-control dataset of 16,029 individuals, I first showed that heterozygosity at the HLA class-II genes conferred T1D risk. To investigate functional basis of this HLA heterozygote disadvantage, I predicted individual-specific repertoires of HLA-bound peptides from 17 T1D-relevant human proteins. The comparison of individual-specific peptide-repertoires between HLA heterozygous and homozygous individuals suggested that both a broader array of HLA-bound self-peptides and higher odds of carrying risk alleles might contribute to HLA heterozygote disadvantage. The characterization of the allele-specific peptide-repertoire suggested that an allele might confer T1D risk due to its low peptide-binding affinity possibly contributing to inefficient removal of autoreactive T-cells in thymus and (or) by binding specific disease-causing peptides, e.g. posttranslationally deamidated peptides. The PepWAS-predicted T1D-associated epitopes accounted for even more deviance in T1D status than HLA class-II haplotypes (33.1 vs. 29.6%). Moreover, the sequence homology between predicted T1D-associated epitopes and pathogenic peptides suggested pathogens as the potential trigger of autoimmunity. Overall, the insights from this thesis shed light on different mechanisms that possibly underlie the differential association of HLA genes with infectious and autoimmune diseases, which, in turn, potentially shape the antagonistic selection on the classical HLA genes