Integrated Single-Cell Detection of Genotype and Phenotype in SF3B1-Mutated Chronic Lymphocytic Leukemia Cells

Cancer cell phenotype is controlled by both genetic composition and gene expression. Recent large-scale cancer sequencing studies have revealed extensive intratumoral genetic heterogeneity and have demonstrated its potential impact on clonal evolution and clinical outcome. The most direct approach to uncovering the impact of genetic heterogeneity on cellular phenotype requires integration of genetic and transcriptomic profiles of single cells. Currently, however, RNA and DNA cannot be reliably isolated from the same cell. Here, we demonstrate the feasibility for linking single-cell somatic mutation data with cellular transcriptional heterogeneity through a targeted RNA-based approach. By leveraging a microfluidic platform (Fluidigm BioMark^TMHD system) to perform multiplexed targeted amplification of RNA derived from hundreds of single cells, we have generated a versatile approach for the integrated detection of somatic mutations in relation to specific gene transcripts.

We focused on a series of chronic lymphocytic leukemia (CLL) B cells that were previously characterized by bulk whole-exome (WES) and RNA-sequencing (RNA-Seq). We developed 2 classes of assays. First, we generated multiplexed nested quantitative RT-PCR assays of 96 genes with known involvement in CLL biology. Second, to simultaneously detect patient-specific somatic mutations in the same cell, we devised multiplexed pre-amplification primers targeting transcribed regions containing somatic point mutations. These regions were then amplified using paired nested primers, for detection of the wild-type or mutant alleles. We focused on those somatic mutations with detectable expression in bulk CLL RNA (> 5 FPKM by RNA-seq). When applied to either artificial oligonucleotide templates or bulk patient cDNA, these paired wild-type and mutant allele detection assays reliably demonstrated consistent differences in DC_T values of >6 cycles.

In total, we designed expression assays for 96 genes and 46 mutation detection applied to 5 CLL samples (median of 9 assays/sample, range 6-13). We examined up to 384 single cells from each of 5 samples and from normal CD19+ B cells. Based on expression of housekeeping genes ACTB and B2M, we observed viable expression in 1951 of 2112 cells (92.4%). We could clearly discern that expression of the 96 genes was heterogeneous across 354 single CLL-B cells and could discriminate CLL from 174 normal B cells by principal component analysis. 32 out of 46 (70%) mutation detection assays successfully distinguished between wild-type and mutant alleles and the mutant allele was consistently observed in the originating CLL cells, but not in unrelated CLL or non-leukemic B cells. Our RNA-based estimates of allele frequency agreed with single-cell targeted DNA-based detection of somatic mutations conducted for 3 of 5 CLL samples as well as with frequencies estimated from bulk WES-based cancer cell fraction (CCF) measurements.

We applied our integrated assay design to 2 CLL samples known to harbor mutations in the putative CLL driver SF3B1: Patient 1 with bulk CCF of 17% (G742D) and Patient 2 with 87% (K700E). Mutation of this critical spliceosome component broadly changes RNA splicing profiles although the functional impact of these alternative splice variants on CLL biology remains unknown. We generated multiplex assays for SF3B1 mutation detection and for expression of mutation-associated alternative splice variants. Consistent with the bulk-sequencing results, we detected 50 of 373 (13.4%) single CLL cells from Patient 1 with SF3B1 mutation. Moreover, the subset of cells with SF3B1 mutation demonstrated high expression of splice variants relative to wild-type cells (GCC2 and MAP3K7, p< 0.000001). This SF3B1 mutated subclone also displayed reduced expression of RNA splicing factors (BTAF1, DDX17, SNW1, SRSF3, U2SURP; all p<0.05), cell cycle regulators (CDC27, PDS5A; p<0.015) and an inflammatory pathway gene (MALT1p=0.039), suggesting involvement of SF3B1 mutation in these biological processes. Analysis of Patient 2 is ongoing.

Taken together, our study demonstrates the feasibility of linking genotype with gene expression at the RNA level. Furthermore, these analyses reveal the potential for single cell RNA-based analysis to directly uncover the effects of driver mutations on the leukemia cell phenotype.