Correlation of gene expression and genome mutation in single B-cells

Abstract

High-throughput measurement of gene-expression and immune receptor repertoires have recently become powerful tools in the study of adaptive immune response. However, despite their now-widespread use, both tend to discard cell identity by treating cell populations in bulk, and therefore lose the correlation between genetic variability and gene-expression at the single cell level. In order to recover this information, we developed a method to simultaneously measure gene expression profiles and genome mutations in single cells. We applied this method by quantifying the relationships between gene expression and antibody mutation in ensembles of individual B-cells from immunized mice. The results reveal correlations reflecting the manner in which information propagates between a B-cell's antigen receptors, its gene expression, and its mutagenic machinery, and demonstrate the power of this approach to illuminate both heterogeneity and physiology in cell populations.

Figure 1. Experimental workflow.

Single PE+/B220+ and PE−/B220+ B-cells were FACS-sorted into 96 well plates, pre-amplified using a pool of gene-expression and sequencing-amplicon outer-primers for heavy- and light-chains (see Tables S4 and S5 in File S1). After exonuclease digestion of leftover single-stranded primer, PCR products were split between sequencing-amplicon amplification using inner-primers and quantitative PCR using EvaGreen dye on Fluidigm 48×48 Dynamic Array chips.

Figure 2. Interrelations between antibody mutations and immunological gene-expression in single B-cells.

Normalized gene expression values (red denotes up-regulation and blue denotes down-regulation) were hierarchically-clustered across 193 single PE+ and PE- B-cells belonging to BALB/c (WT) and TCRd−/− (KO) mice, and plotted alongside mutational content of antibody heavy- and light-chains expressed by each (A). For the latter, cells for which light-chains either could not be sequenced or were rejected by the quality-filter are color-coded on the bar-plot in grey. Spearman-correlations were calculated between each gene-expression value and mutation-count for wild-type (B) and knock-out (Figure S3A in File S1) mice. Data were further analyzed by calculating differences between Spearman correlations performed on non-synonymous and synonymous-mutations separately (C), with absolute values of these differences depicted on the vertical axis.