In Silico Derivation of HLA-Specific Alloreactivity Potential from Whole Exome Sequencing of Stem Cell Transplant Donor-Recipient Pairs

Graft vs. host (GVH) effect mediated by donor T cells are responsible for a significant measure of therapeutic benefit as well as toxicity observed following stem cell transplantation (SCT) between HLA-identical donors and recipients. Donor T cell-mediated GVH effects may be influenced by the aggregate alloreactivity to minor histocompatibility antigens (mHA) presented by the HLA in each donor-recipient pair (DRP). The cumulative mHA variation in each DRP may thus be regarded as the alloreactivity potential (AP) of that pair. To estimate AP in DRP, whole exome sequencing (WES) of 4 matched-related (MRD) and 5 unrelated (URD) pairs was performed and revealed extensive coding variation between them. To quantify the contribution of exome sequence variation to AP, data from each DRP was filtered to isolate non-synonymous single nucleotide polymorphisms (SNP) in the GVH direction (polymorphisms present in recipient and absent in donor). Logically, the SNP involved in encoding peptides presented on the HLA in each DRP will contribute to the AP in that pair. To identify these peptides, the nucleotide sequence flanking all of the several thousand SNP in each DRP was obtained with the ANNOVAR software package. All possible resulting nonameric-peptides were interrogated in-silico for their likelihood to be presented by each of the 6 HLA class I molecules in individual DRP, using the Immune-Epitope Database (IEDB) SMM algorithm. The IEDB-SMM algorithm predicted between 1,043,514 and 366,426 peptides/DRP (~ 18 peptides/SNP). Peptide-HLA binding affinity estimate was reported as an IC50 value, which when <500 nM, predicts that peptides may be presented by the HLA. Using this cutoff, from 18.9-1.2% of all the above peptides were categorized as presented by the HLA class I molecules in the DRP examined, with a smaller fraction of these peptides being strong binders (IC50 <50nM). (Fig 1) In our small cohort, MRD presented a median 3% (range 1-10) of all the predicted peptides, compared with 13.5% (4-20) for the URD. Further, plotting the IC50 values of the presented-peptides in each DRP, demonstrated a continuum of binding affinities (Fig 2) as opposed to discrete sets of values taken on by different peptides, implying a certain degree of plasticity in this system. Computing the area under the curve for these data to account for both number of and binding affinity of peptide-HLA complexes, demonstrated a marked difference between MRD (1x10^6 nM.Peptide-HLA) and URD (8x10^6). Our data demonstrates that a marked difference is evident between URD and MRD when WES is used to predict potential protein-coding differences - estimating an HLA-specific AP in unique DRP. This represents a paradigm shift away from looking for unique mHA to predict likelihood of developing GVH phenomenon in a population-based manner, and may eventually better define GVHD risk in HLA matched DRP.