Genome-Wide Association Study for Determinants of Acute Graft Vs Host Disease (aGVHD) After Allogeneic Hematopoietic Stem Cell Transplantation: Development of 7 SNP Model Predicting aGVHD

Numerous studies have been reported non-HLA genetic associations with acute graft vs host disease (aGVHD) developing after allogeneic hematopoietic stem cell transplantation (allo-HSCT). So far, however, relatively few genes have been examined in each cohort. Meanwhile, aGVHD represents one of the most attractive targets to study the effect of normally silent genetic polymorphisms. We intended to elucidate genetic determinants predicting aGVHD in patients undergoing allo-HSCT by genome-wide screening of genetic polymorphisms, and develop a SNP model to predict development of aGVHD.

Genetic polymorphisms predicting development of aGVHD were determined through successive 5 stages in a total of 429 patients who underwent allo-HSCT. Firstly, we screened genomic determinants predicting aGVHD using 109K SNP chip (Sentrix Human-1 GT Bead Chip, Illumina) in 15 patients (7 patients with aGVHD and 8 without aGVHD). At stage 2 and 3, low-density Immumina chips comprising 768 candidate SNPs and 96 SNPs, respectively, were manufactured and tested in independent cohorts of 91 patients (34 patients with aGVHD and 57 without aGVHD) and 84 patients (35 patients with aGVHD and 49 without aGVHD), respectively. At stage 4, 76 SNP chips comprising 33 SNPs discovered from the previous studies and also 43 SNPs reported to be associated with aGVHD in the literature were manufactured and evaluated in 36 patients (13 patients with aGVHD and 23 without aGVHD) undergoing allo-HSCT. Finally, 7 SNPs were chosen and validated in an independent cohort of 203 patients (50 patients with aGVHD and 153 without aGVHD) using TaqMan assay.

At stage 1, 955 SNPs were identified to be significantly associated with development of aGVHD by whole genome assay. At stage 2, candidate genetic determinants were narrowed down to 74 SNPs. Through stage 3 and 4, 7 SNPs including the two selected from the literature review (transforming growth factor beta receptor II (TGFBR2) and interleukin-18) were finally determined as predictive markers for aGVHD. The 7 SNP model was found to be effective in predicting the development of aGVHD after allo-HSCT. When decision tree model was applied, the sensitivity, specificity, accuracy, and area under ROC (AUR) of the 7 SNP model to predict aGVHD were 80.5%, 80.4%, 80.4% and 91.6%, respectively.

A genome-wide association study approach was successful in elucidation of genetic determinants associated with development of aGVHD in patients undergoing allo-HSCT. The 7 SNP model was effective in predicting the development of aGVHD after allo-HSCT.

No relevant conflicts of interest to declare.