Quantifying HLA transcripts by genotype in chimeric mixtures at single-cell resolution [scrHLAtyping_MAS-seq]

Gene products from the highly variable major histocompatibility locus, including HLA, are essential for self-recognition and immune surveillance of malignancy. Following allogeneic hematopoietic cell transplantation (alloHCT), genetic and epigenetic alterations in HLA can drive disease recurrence, making precise HLA assessment critical for determination of future therapy. However, current methods lack the sensitivity to quantify HLA transcripts at the single cell level, limiting their clinical utility. We introduce scrHLA-typing, a novel method that accurately identifies and quantifies HLA transcripts in single cells using hybridization capture and long-read sequencing. When applied to samples from patients with post-transplant relapse, scrHLA-typing successfully detected allele-specific expression of MHC gene products at clinically actionable levels. By characterizing allele expression in residual leukemia cells, our assay identified differences in expression patterns among patients. This capability highlights scrHLA-typing’s potential to improve risk stratification and guide the selection of appropriate salvage therapies, enhancing personalized treatment strategies after post-transplant relapse. Mononuclear cells from bone marrow aspirates from patients with relapsing AML post-alloHCT were density-gradient collected and magnetically enriched for cells expressing on their surface the CD34 progenitor marker (in one patient, the CD117 marker was used instead). From both unsorted and enriched cells, full transcriptome single-cell libraries (10x Genomics) were generated using 5-prime chemistry capture (by default, unless otherwise indicated, i.e., prefix '3p' tagged to the sample name) and short-read sequenced to quantify global RNA expression. For most patients, cells were also stained prior to droplet capture with oligonucleotide-barcoded antibodies specific to cell surface/principal lineage antigens, and antibody-derived tags (ADT) were short-read sequenced and quantified. From the leftover transcriptomes, single-cell barcoded HLA full transcripts were magnetically enriched with biotinylated hybridization probes HLA-specific and streptavidin magnetic beads, afterwhich they were PCR-amplified and long-read sequenced. The HLA probe design incorporated all known HLA alleles available from the IMGT/HLA reference. In a few cases, long-read sequencing of HLA transcripts was augmented using ‘multiplexed arrays isoform sequencing’ (MAS-seq). The molecular method was supplemented with a computational pipeline we designed to iteratively align reads, predict genotypes, filter errors, and summarize HLA counts. Short-read and long-read sequencing were conducted on Illumina and PacBio platforms, respectively.