Systematic Identification of Autosomal and Y-Encoded Minor Histocompatibility Antigens Reveals Predictors of Chronic Gvhd and Candidate GVL Targets

At the core of the therapeutic effect of HLA-matched allogeneic hematopoietic cell transplantation (allo-HCT) is T cell alloreactivity against minor histocompatibility antigens (mHAgs), polymorphic peptides resulting from donor-recipient disparity at sites of single nucleotide gene polymorphisms (SNPs). Despite their crucial role in graft-versus-leukemia (GvL) and graft-versus-host disease (GvHD), only few mHAgs have been characterized to date. To systematically identify autosomal and Y-chromosome encoded HLA-I restricted mHAgs, we devised a computational pipeline based on: (i) comparison of whole exomes (WES) from paired donor-recipient DNA to define recipient-restricted exonic non-synonymous polymorphisms; and (ii) generation of expression filters by incorporating published single cell expression profiles from normal and malignant hematopoietic cells and GvHD target organs. For each GvHD tissue (skin, liver, GI, lung, oral mucosa and lacrimal gland), a large single cell database combining ≥1 independent datasets was created to cluster and annotate tissue-resident cell types. Cell-type specific genes were defined based on their total number of reads (CPM>5 in any cluster), thus creating an expression atlas capturing less frequent albeit biologically relevant cell types underestimated in bulk expression profiles. For GvL predictions, we built an AML-specific filter by defining AML-expressed genes from single cell and bulk data, and by subtracting genes with expression in non-hemopoietic tissues, per transcriptome and proteome GTEx databases. Candidate GvL/GvHD mHAgs prediction was performed by applying the HLAthena binding prediction tool on recipient-restricted SNPs expressed in GvL or GvHD tissues. To validate the pipeline predictive power, we used a panel of 95 HLA class I monoallelic lines generated from B721.221 cells, which we previously profiled by WES, RNASeq and immunopeptidomics. By assigning the B721.221 genome as surrogate 'HCT recipient’ and the reference hg19 genome as surrogate 'donor', we identified recipient-restricted SNPs, whose class I-restricted presentation we confirmed by mass spectrometry for a median of 5 (range 1-16) predicted epitopes/ allele. For 2 HLA-A0201 epitopes encompassing SNPs with an allelic frequency <0.001, we validated immunogenicity through stimulation of PBMCs from HLA-A0201 healthy donors (HDs) with corresponding synthetic peptides. For Y-encoded mHAgs, we predicted a median of 62 (range 24-107) epitopes/HLA-I allele from 9 genes expressed (>1 TPM) in ≥1 non germline tissue. Analysis of 12 male immunopeptidomes confirmed presentation of 3 known Y mHAgs and 26 novel predicted epitopes across 18 alleles. To estimate the fraction of immunogenic predicted binders, we tested all Y epitopes predicted across 3 common alleles (A0201, B1801, C0501). Of 215 Y peptides tested, 53 (~25%) elicited an antigen-specific T cell response in PBMCs from female HDs, as per dextramer-based detection of ≥0.2% CD8 cells. To test the prognostic potential of the mHAg load on cGvHD risk, we performed WES on 220 donor-recipient pairs treated with matched-related-donor allo-HCT for myeloid disease at DFCI between 2013-2020. Our logistic regression analysis used organ-specific NIH moderate/severe cGvHD as outcome; as covariates, we included tissue-specific mHAg load and standard HCT variables (disease type and prognostic risk, conditioning intensity, graft type, GvHD prophylaxis, female into male [F→M] transplant, relapse). The burden of lung-expressed mHAgs was correlated with lung GvHD occurrence (P=0.023). Liver GvHD was associated with F→M transplants (P=0.023). Subgroup analysis further showed liver GvHD to be correlated with the number of epitopes predicted from the RPS14Y gene (P=0.037). For identification of AML-specific mHAgs, a median of 31 (range 13-50) mHAgs/patient were predicted in our cohort from 269 genes preferentially expressed in AML, with a trend for a protective effect of their load against relapse (P=0.058). Of note, 39 SNPs generated putative mHAgs shared by >10 patients and covering 8 common HLA-I alleles in Caucasians, thereby representing potential targets for post-HCT immunotherapy. Overall, we report the generation of a robust analytic pipeline which enables the broadening of the repertoire of mHAgs available for donor selection and for predicting, monitoring or manipulating GvL and GvHD after allo-HCT. Figure 1View largeDownload PPTFigure 1View largeDownload PPT Close modal