81 The strongest genetic signal belonged to the HLA-class II region in chromosome 6, where we found 130 peptides associated with 134 different leading SNPs. Most of the associated peptides belonged to Streptococcus and Staphylococcus species, but we also found several peptides belonging to human viruses (adenoviruses or herpesviruses) and phages and to allergens (ovomucoid, barley, casein and wheat, amongst others) and gut microbiota. To dig into this genomic region, we conducted a specific imputation of HLA SNPs, indels, amino acids and gene isoforms and performed an association analysis with all peptides (see Methods) (Supplementary Table 2.3). This analysis substantially increased the number of associated peptides and the strength of associations. We discovered that a large number of peptides (530/2,813) had at least one significant (p < 1x10-6, after correction for number of independent tests, see Methods) association with HLA variants (amino acids, insertions, SNPs or genes). At HLA gene level, we identified 1,267 significant peptide–gene associations to 276 different peptides. Most of those associations (and the strongest) belonged to allelic variants of HLA-II (1,139 associations to 271 different peptides) in comparison to variants of HLA-I (128 associations to 41 different peptides). Within the HLA-II variations, most associations were observed for various alleles in DQ and DR beta chain genes. To determine whether these associations are due to the capacity of a specific HLA complex to present the peptide, we performed computational modeling of the HLA– peptide complex using some of our top associations. Here we identified that the predicted residues that are recognized from the peptide by a specific HLA complex51 can form stable structures with their associated HLA complexes. For instance, streptococcal C5a peptidase (TPSDAGETVADDANDLAPQAPAKTADTPATSKATIRDLNDPSQVKTLQEKAGKGAGTVVAVID A) is highly associated with DRB1*0301 (always bound to the alpha chain DRA*01, DR3 haplotype) (Odds ratio (OR) = 3.78, p = 1.65x10-31) and with DQB1*0201 (OR = 3.75, p = 5.16x10-31) and the alpha chain DQA1*0501 (OR = 1.91, p = 4.80x10-13), which together form the haplotype DQ2.5 that is highly linked to DR3. The predicted core recognized by the HLA complex was nearly identical for both DR3 and DQ2.5 (VADDANDL) and is very similar to the amino acid composition identified from HLA ligand elution experiments.71 This reveals a close relationship in the composition of the streptococcal C5a peptidase peptide sequence in comparison to chemically synthesized peptides tested for DR3 and DQ2.5. Additionally, we employed the predicted binding metric (percentage of elution score %Rank_EL, Methods) to assess the binding of the core peptide to the selected alleles. %Rank_EL is calculated as the percentile of the predicted binding affinity compared to the distribution of affinities calculated on a set of random natural peptides (%Rank_ EL; strong binding: ≤ 2.0, weak binding: 2.0-10.0, no binding: > 10). This analysis found a favorable binding prediction of the core to DR3 and DQ2.5 complexes, with a higher binding for DQ2.5 (2.39 and 0.65, respectively). We further compared the binding prediction for this epitope to a non-associated negative control (DR14) that was predicted to be non-binding (%Rank_EL 14.77). Structural modeling and analysis of the bindingmode of the peptide revealed a favorable binding energy with both DQ2.5 and DR3 (-9.1 and - 10.1 Kcal/mol, respectively) compared to the nonassociated structure DR14 (-7.7 Kcal/mol) (Figure 3B). Additionally, the computed dissociation constant (Kd) showed an order of magnitude less affinity for the non-associated allele (2.3x10-6 M) compared to DR3 (3.7x10-8 M) and DQ2.5 (1.7x10-7 M). As a result, the peptide core exhibited similar behavior and key stabilizing polar interactions when binding into the binding sites of DR3 Determinants of the human antibody epitope repertoire
RkJQdWJsaXNoZXIy MjY0ODMw