Frequent alterations of HLA-A in rrDLBCL. (A) For each biopsy type (diagnostic DLBCL, diagnostic rrDLBCL and relapse rrDLBCL), MHC class I allele inference for HLA-A was performed using POLYSOLVER. Depicted along the vertical axis are all inferred alleles from the WES with matched normal cohort (n = 37; b = 45). Each box is colored to represent absence, presence, loss, or mutation of each allele in each biopsy. Above the grid, B2M protein-coding variants in each biopsy are shown as boxes colored by mutation type. (B) Mutation burden in HLA-A per sample by biopsy type. Comparing initial diagnostic biopsies that do (green) or do not (purple) go on to relapse, the relapsing cases show a significantly higher burden of HLA-A alterations (P < .02, unpaired Student t test with Welch correction). (C) Increased mutational burden in HLA-A in cases that go on to progress or relapse from R-CHOP–like therapies. Each circle represents a case where analysis was available, each X represents a mutation or LOH event. (D) Immunohistochemical analysis of HLA-A protein expression in the combined cohort of DLBCL samples from the WES and targeted sequencing cohorts. Classification into HLA-A negative and positive staining was performed for each biopsy type, n = 55, b = 58. (E) Representative images of IHC staining of HLA-A; HLA-A⁺ and HLA-A⁻ biopsies obtained on a Nikon Eclipse Ci at 60× objective.