Aim: We aim to validate new tools for predicting HLA allele contributions to protective immunity.
Methods: USA300 S. aureus proteome was obtained from UniProt and a dataset of >450,000 overlapping 15 amino acid peptides was generated. Peptide binding predictions for five HLA alleles representing different ancestral haplotypes were performed using machine learning frameworks (NetMHCIIPan-4.0 and MixMHC2Pred-2.0). A binding prediction score of ≤ 2.0 was used as a biological threshold and data was analyzed using R (4.2.2). CD4 cells were isolated from five HLA-DRB1*04:01 / HLA-DRB1*15:01 heterozygous healthy controls and cocultured with S. aureus pulsed EBV transformed B cell antigen presenting cells homozygous for HLA-DRB1*04:01 (IHW 9090) or HLA-DRB1*15:01 (IHW 9008). Antigen presenting cells were pulsed with inactivated S. aureus across a gradient of multiplicity of infection (MOI): 0, 0.5, 5, 50. CD4 T cell activation (CD69+ CD25+) was determined via flow cytometry (BD Fortessa, FlowJo) using monoclonal antibodies specific for CD4 PerCP-Cy5.5, CD69 PE, and CD25-APC. Mean and standard deviation were determined using Prism software.
Results: We observed different S. aureus peptide binding score distributions across HLA-DRB1 alleles reflecting unique peptide binding pockets. Analysis revealed higher proportions of peptides predicted to bind HLA-DRB1*15:01 and 07:01 compared to DRB1*04:01, DRB1*03:01, and DRB1*01:01 (binding scores ≤ 2.0). Three-day cocultures revealed an HLA allele influence on CD4 T cell activation (CD69+ CD25+) across a gradient of S. aureus concentrations. DRB1*15:01 antigen presenting cells elicited a higher percentage of CD4 T cell activation (MOI 0=24% ± 6; MOI 0.5=42% ± 5; MOI 5=49% ± 5; MOI 50= 53% ± 5) when compared to S. aureus pulsed DRB1*04:01 antigen presenting cells (MOI 0=17% ± 4; MOI 0.5=30% ± 3; MOI 5=37% ± 3; MOI 50= 40% ± 4).
Conclusion: We provide evidence for inferior S. aureus peptide binding to explain the strong association between HLA-DRB1*04:01 haplotypes and S. aureus infection. Use of in silico tools to predict HLA-peptide binding may identify patients at risk for S. aureus complications.
