Unbiased linkage of SNV to expression in single cells for regulatory variant discovery influencing cancer drug resistance

I. Salas-Gonzalez, T. Morozova, D. Arvapalli, S. Velivela, J. West, G. Harton, J. Zawistowski, V. Weigman; BioSkryb Genomics, Durham, NC

Single-cell RNAseq methodologies have revolutionized the ability to define cell type identity and transcriptional phenotypic diversity within a tumor sample, yet in most studies DNA-level variation contributing to driving this diversity of gene expression remains uncharacterized. We employed ResolveOME™ chemistry to bridge this gap within an individual cell, by taking inventory of genome-wide single nucleotide variation (SNV) and cross referencing it with full-transcript RNAseq data in an acute myeloid leukemia (AML) cell line model of drug resistance. The catalog of genome-wide SNV in conjunction with transcriptomic data in this model powered an association screening to identify regulatory SNVs with biased prevalence in single cells resistant to the FLT3 inhibitor quizartinib vs treatment-naïve single cells, which then were correlated with differential expression of genes proximal to the variant in the same cell. Initially, we focused on nucleotide variation within core promoter regions and within gene bodies. Fitting a zero-inflated linear model to a matrix of expression and genotype across all single cells revealed an intronic heterozygous variant in MYC in parental single cells, while, in contrast, the alternate allele was absent in quizartinib-resistant cells. Upregulation of MYC transcript was correlated with the absence of this variant in resistant single cells, suggesting that the single nucleotide change may have intronic enhancer activity as opposed to intronic regulation of splicing. Additionally, the screen uncovered a heterozygous single nucleotide change within 5kb 5 of the transcriptional start site of the mRNA binding factor PABPC4 as a candidate promoter variant not present in parental single cells yet harbored by 50% of the quizartinib resistant single cells displaying PABPC4 expression relative to parental cells. Differential expression analysis uncovered the enhancer factor CEBPA as upregulated in the resistant single cells, and the genomic component of ResolveOME identified two SNVs 20 kb upstream of the locus representing putative CEBPA enhancer variants influencing the differential gene expression. The co-identification of CEPBA and PABPC4 through this multi-omic approach suggest that quizartinib resistance in this model is mediated in part by global gene regulation through enhancer modulation and through mRNA stability/translational regulation, respectively. We are furthering enhancer variant detection in intergenic space by overlaying ChIP-seq, chromatin accessibility, and transcription factor binding site data to prioritize candidates contributing to resistance in this AML drug resistance model.