Exome Sequencing Derived Alloreactivity Potential May Predict Gvhd in Allogeneic Stem Cell Transplantation

Allogeneic hematopoietic stem cell transplants may result in graft versus host disease, mediated by donor T cells alloreactivity towards minor histocompatibility antigens (miHA) in the recipient. Up to 60% of allogeneic stem cell transplants experience acute graft vs host disease (GVHD). Even in perfectly HLA-matched related transplants, GVHD is still unpredictable; conversely, haploidentical mismatched transplants with reduced intensity conditioning can have a GVHD-free course. As of 2015, only 54 miHA have been described, such as HY, shown to affect GVHD and outcomes in specific diseases and donor-recipient pairs.

To quantify potential miHA, whole exome sequencing was performed on 34 allogeneic transplants. For patient characteristics, 27 patients were molecularly matched at 8/8 loci and 7 patients were 7/8 HLA matched; 5 female to male donor transplants were included; as well as 10 African American and 24 Caucasian matches. Ten patients did not develop GVHD, while 24 patients experienced GVHD, either acutely or chronically. A bioinformatics pipeline was developed to discover all non-synonymous single nucleotide polymorphisms (nsSNPs) within unique donor-recipient pairs (DRP) in the graft vs host (GVH) direction. These nsSNPs were translated into 9-mer peptides, with the protein-coding change modeled at each site (position 1-9); this generated nine 9-mers per nsSNP. The peptides were then queried for binding affinity (represented as an IC50) to each patient's HLA with the NetMHC algorithm, with an IC50 < 500 called - presented and IC50 <50 called as - strong binder. On average, 7,525 nsSNP were discovered on exome sequencing each DRP in the GVH direction. This yielded an average of 169,807 peptides (DRP^-1) which were interrogated for binding to patient-specific HLA-A, B, and C alleles. Of those, an average of 2260 peptides/DRP were presented, and 432 peptides were strong binders.

Using logistic regression analysis, significant differences were found between matched-related and matched-unrelated donors' presented peptides: 1646 vs. 2249, respectively (p-value .036). There were no significant differences in the average number of predicted peptides presented by HLA-A, B, or C.

Importantly, the ratio of strong binders/presented peptides was a significant predictor of GVHD (p-value 0.046) with a model controlling for donor type, HLA-mismatch, patient race, and donor gender. There were 17,457 presented peptides shared between GVHD and GVHD-free patients. In 10 patients that did not develop GVHD 8,331 peptides were unique. In the 24 patients that did develop GVHD, 20,300 peptides were unique to that group (Figure 1).

Potential miHA had varying tissue distribution based on average tissue values from the GTEx RNA expression database. The data from this 34 patient cohort quantify the large magnitude of potential miHA present in every BMT and their HLA binding affinity distribution, which can be thought of as an alloreactivity potential. Additional DRP whose analysis is not yet complete may aid in refining the analysis presented here.

Current HLA matching strategies ignore genomic variation and strategies to account for this variation may aid both in better donor selection for optimal GVH/GVL effects and also in tailoring immunosuppression in an individualized, patient-specific manner.