Visualizing SNVs to quantify allele-specific expression in single cells

Abstract

We present a FISH-based method for detecting single-nucleotide variants (SNVs) in exons and introns on individual RNA transcripts with high efficiency. We used this method to quantify allelic expression in cell populations and in single cells, and also to distinguish maternal from paternal chromosomes in single cells.

Figure 1

Toehold probes enable SNV detection on individual RNA molecules in situ. a. Schematic of the principle behind in situ SNV detection, using the T1799A mutation of BRAF as an example. b. Visualization of the guide probe detecting BRAF mRNA (ATTO488, left panel) and the wild-type and mutant detection probes (Cy5, Cy3, middle and right panels, respectively). c. Classification of RNA as being either wild-type or mutant using the detection probes. d. Quantification and classification of RNA as wild-type or mutant in a group of single cells. Each sample shown is one of a set of at least two biological replicates. Left: cells with only wild-type BRAF; middle: cells that are heterozygous for BRAF; right: cells that are mutant for BRAF. Scale bars are 5μm long.

Figure 2

Allele-specific expression at the population and single cell level in GM12878 cells. a. We quantified allelic imbalance in the population of the indicated genes by measuring the probability that a transcript comes from either the maternal or paternal allele. Error bars reflect 95% confidence intervals on counting statistics plus an 8 percentage-point differential between maternal and paternal detection efficiency; see methods for details. b. Diagram of single cell allelic balance and imbalance. c. Allelic imbalance in single cells. The solid black midline represents the average imbalance across cells (from b). The dashed black lines shows the 95% confidence interval on the imbalance for each cell with the null hypothesis that the probability of an RNA being maternal or paternal is independent of which cell it is in. The inset shows the likelihood of the observed population imbalance (red) compared to that of the null model (blue); see methods for details. Note that for EBF1, ~90% of cells expressed zero transcripts, so we excluded those cells from the figure. Each sample shown is one of a set of at least two biological replicates. ** represents cells with a p-value below 0.05, and * represents a p-value below 0.10 (p-value defined in methods and Supplementary Note).