Sydney Shaffer

Rare cell variability and drug-induced reprogramming as a mode of cancer drug resistance

Therapies that target signalling molecules that are mutated in cancers can often have substantial short-term effects, but the emergence of resistant cancer cells is a major barrier to full cures. Resistance can result from secondary mutations, but in other cases there is no clear genetic cause, raising the possibility of non-genetic rare cell variability. Here we show that human melanoma cells can display profound transcriptional variability at the single-cell level that predicts which cells will ultimately resist drug treatment. This variability involves infrequent, semi-coordinated transcription of a number of resistance markers at high levels in a very small percentage of cells. The addition of drug then induces epigenetic reprogramming in these cells, converting the transient transcriptional state to a stably resistant state. This reprogramming begins with a loss of SOX10-mediated differentiation followed by activation of new signalling pathways, partially mediated by the activity of the transcription factors JUN and/or AP-1 and TEAD. Our work reveals the multistage nature of the acquisition of drug resistance and provides a framework for understanding resistance dynamics in single cells. We find that other cell types also exhibit sporadic expression of many of these same marker genes, suggesting the existence of a general program in which expression is displayed in rare subpopulations of cells.

Turbo FISH: a method for rapid single molecule RNA FISH.

Advances in RNA fluorescence in situ hybridization (RNA FISH) have allowed practitioners to detect individual RNA molecules in single cells via fluorescence microscopy, enabling highly accurate and sensitive quantification of gene expression. However, current methods typically employ hybridization times on the order of 2–16 hours, limiting its potential in applications like rapid diagnostics. We present here a set of conditions for RNA FISH (dubbed Turbo RNA FISH) that allow us to make accurate measurements with no more than 5 minutes of hybridization time and 3 minutes of washing, and show that hybridization times can go as low as 30 seconds while still producing quantifiable images. We further show that rapid hybridization is compatible with our recently developed iceFISH and SNP FISH variants of RNA FISH that enable chromosome and single base discrimination, respectively. Our method is simple and cost effective, and has the potential to dramatically increase the throughput and realm of applicability of RNA FISH.

SAVER: gene expression recovery for single-cell RNA sequencing

In single-cell RNA sequencing (scRNA-seq) studies, only a small fraction of the transcripts present in each cell are sequenced. This leads to unreliable quantification of genes with low or moderate expression, which hinders downstream analysis. To address this challenge, we developed SAVER (single-cell analysis via expression recovery), an expression recovery method for unique molecule index (UMI)-based scRNA-seq data that borrows information across genes and cells to provide accurate expression estimates for all genes.SAVER accurately recovers expression values in single-cell RNA-sequencing data to improve downstream analysis.

Rare Cell Detection by Single-Cell RNA Sequencing as Guided by Single-Molecule RNA FISH

Although single-cell RNA sequencing can reliably detect large-scale transcriptional programs, it is unclear whether it accurately captures the behavior of individual genes, especially those that express only in rare cells. Here, we use single-molecule RNA fluorescence in situ hybridization as a gold standard to assess trade-offs in single-cell RNA-sequencing data for detecting rare cell expression variability. We quantified the gene expression distribution for 26 genes that range from ubiquitous to rarely expressed and found that the correspondence between estimates across platforms improved with both transcriptome coverage and increased number of cells analyzed. Further, by characterizing the trade-off between transcriptome coverage and number of cells analyzed, we show that when the number of genes required to answer a given biological question is small, then greater transcriptome coverage is more important than analyzing large numbers of cells. More generally, our report provides guidelines for selecting quality thresholds for single-cell RNA-sequencing experiments aimed at rare cell analyses.

Transcriptional bursting explains the noise–versus–mean relationship in mRNA and protein levels

Recent analysis demonstrates that the HIV-1 Long Terminal Repeat (HIV LTR) promoter exhibits a range of possible transcriptional burst sizes and frequencies for any mean-expression level. However, these results have also been interpreted as demonstrating that cell-to-cell expression variability (noise) and mean are uncorrelated, a significant deviation from previous results. Here, we re-examine the available mRNA and protein abundance data for the HIV LTR and find that noise in mRNA and protein expression scales inversely with the mean along analytically predicted transcriptional burst-size manifolds. We then experimentally perturb transcriptional activity to test a prediction of the multiple burst-size model: that increasing burst frequency will cause mRNA noise to decrease along given burst-size lines as mRNA levels increase. The data show that mRNA and protein noise decrease as mean expression increases, supporting the canonical inverse correlation between noise and mean.

SAVER: Gene expression recovery for UMI-based single cell RNA sequencing

Rapid advances in massively parallel single cell RNA sequencing (scRNA-seq) is paving the way for high-resolution single cell profiling of biological samples. In most scRNA-seq studies, only a small fraction of the transcripts present in each cell are sequenced. The efficiency, that is, the proportion of transcripts in the cell that are sequenced, can be especially low in highly parallelized experiments where the number of reads allocated for each cell is small. This leads to unreliable quantification of lowly and moderately expressed genes, resulting in extremely sparse data and hindering downstream analysis. To address this challenge, we introduce SAVER (Single-cell Analysis Via Expression Recovery), an expression recovery method for scRNA-seq that borrows information across genes and cells to impute the zeros as well as to improve the expression estimates for all genes. We show, by comparison to RNA fluorescence in situ hybridization (FISH) and by data down-sampling experiments, that SAVER reliably recovers cell-specific gene expression concentrations, cross-cell gene expression distributions, and gene-to-gene and cell-to-cell correlations. This improves the power and accuracy of any downstream analysis involving genes with low to moderate expression.

A Comparison Between Single Cell RNA Sequencing And Single Molecule RNA FISH For Rare Cell Analysis

The development of single cell RNA sequencing technologies has emerged as a powerful means of profiling the transcriptional behavior of single cells, leveraging the breadth of sequencing measurements to make inferences about cell type. However, there is still little understanding of how well these methods perform at measuring single cell variability for small sets of genes and what “transcriptome coverage” (e.g. genes detected per cell) is needed for accurate measurements. Here, we use single molecule RNA FISH measurements of 26 genes in thousands of melanoma cells to provide an independent reference dataset to assess the performance of the DropSeq and Fluidigm single cell RNA sequencing platforms. We quantified the Gini coefficient, a measure of rare-cell expression variability, and find that the correspondence between RNA FISH and single cell RNA sequencing for Gini, unlike for mean, increases markedly with per-cell library complexity up to a threshold of ∼2000 genes detected. A similar complexity threshold also allows for robust assignment of multi-genic cell states such as cell cycle phase. Our results provide guidelines for selecting sequencing depth and complexity thresholds for single cell RNA sequencing. More generally, our results suggest that if the number of genes whose expression levels are required to answer any given biological question is small, then greater transcriptome complexity per cell is likely more important than obtaining very large numbers of cells.

Memory sequencing reveals heritable single cell gene expression programs associated with distinct cellular behaviors

Non-genetic factors can cause individual cells to fluctuate substantially in gene expression levels over time. Yet it remains unclear whether these fluctuations can persist for much longer than the time for a single cell division. Current methods for measuring gene expression in single cells mostly rely on single time point measurements, making the time of a fluctuation of gene expression or cellular memory difficult to measure. Here, we report a method combining Luria and Delbrück’s fluctuation analysis with population-based RNA sequencing (MemorySeq) for identifying genes transcriptome-wide whose fluctuations persist for several cell divisions. MemorySeq revealed multiple gene modules that express together in rare cells within otherwise homogeneous clonal populations. Further, we found that these rare cell subpopulations are associated with biologically distinct behaviors in multiple different cancer cell lines, for example, the ability to proliferate in the face of anti-cancer therapeutics. The identification of non-genetic, multigenerational fluctuations has the potential to reveal new forms of biological memory at the level of single cells and suggests that non-genetic heritability of cellular state may be a quantitative property.

Gene expression distribution deconvolution in single-cell RNA sequencing

Significance We developed deconvolution of single-cell expression distribution (DESCEND), a method to recover cross-cell distribution of the true gene expression level from observed counts in single-cell RNA sequencing, allowing adjustment of known confounding cell-level factors. With the recovered distribution, DESCEND provides reliable estimates of distribution-based measurements, such as the dispersion of true gene expression and the probability that true gene expression is positive. This is important, as with better estimates of these measurements, DESCEND clarifies and improves many downstream analyses including finding differentially expressed genes, identifying cell types, and selecting differentiation markers. Another contribution is that we verified using nine public datasets a simple “Poisson-alpha” noise model for the technical noise of unique molecular identifier-based single-cell RNA-sequencing data, clarifying the current intense debate on this issue. Single-cell RNA sequencing (scRNA-seq) enables the quantification of each gene’s expression distribution across cells, thus allowing the assessment of the dispersion, nonzero fraction, and other aspects of its distribution beyond the mean. These statistical characterizations of the gene expression distribution are critical for understanding expression variation and for selecting marker genes for population heterogeneity. However, scRNA-seq data are noisy, with each cell typically sequenced at low coverage, thus making it difficult to infer properties of the gene expression distribution from raw counts. Based on a reexamination of nine public datasets, we propose a simple technical noise model for scRNA-seq data with unique molecular identifiers (UMI). We develop deconvolution of single-cell expression distribution (DESCEND), a method that deconvolves the true cross-cell gene expression distribution from observed scRNA-seq counts, leading to improved estimates of properties of the distribution such as dispersion and nonzero fraction. DESCEND can adjust for cell-level covariates such as cell size, cell cycle, and batch effects. DESCEND’s noise model and estimation accuracy are further evaluated through comparisons to RNA FISH data, through data splitting and simulations and through its effectiveness in removing known batch effects. We demonstrate how DESCEND can clarify and improve downstream analyses such as finding differentially expressed genes, identifying cell types, and selecting differentiation markers.

Corrigendum: Rare cell variability and drug-induced reprogramming as a mode of cancer drug resistance

This corrects the article DOI: 10.1038/nature22794