Single-cell transcriptomic methods classify new and existing cell types very effectively, but alternative approaches are needed to quantify the individual regulatory states of cells in their native tissue context. We combined the tissue preservation and single-cell resolution of laser capture with an improved preamplification procedure enabling RNA sequencing of 10 microdissected cells. This in situ 10-cell RNA sequencing (10cRNA-seq) can exploit fluorescent reporters of cell type in genetically engineered mice and is compatible with freshly cryoembedded clinical biopsies from patients.Through recombinant RNA spike-ins, we estimate dropout-free technical reliability as low as ~250 copies and a 50% detection sensitivity of ~45 copies per 10-cell reaction. By using small pools of microdissected cells, 10cRNA-seq improves per-cell reliability and sensitivity beyond existing approaches for single-cell RNA sequencing (scRNA-seq). Accordingly, in multiple tissue and tumor settings, we observe 1.5-2-fold increases in genes detected and overall alignment rates compared to scRNA-seq. Combined with existing approaches to deconvolve small pools of cells, 10cRNA-seq offers a reliable, unbiased, and sensitive way to measure cell-state heterogeneity in tissues and tumors.

Direct reprogramming of fibroblasts into cardiomyocytes is a promising approach for cardiac regeneration but still faces challenges in efficiently generating mature cardiomyocytes. Systematic optimization of reprogramming protocols requires scalable, objective methods to assess cellular phenotype beyond what is captured by transcriptional signatures alone. To address this question, we automatically segmented reprogrammed cardiomyocytes from immunofluorescence images and analyzed cell morphology. We also introduce a method to quantify sarcomere structure using Haralick texture features, called SarcOmere Texture Analysis (SOTA). We show that induced cardiac-like myocytes (iCLMs) are highly variable in expression of cardiomyocyte markers, producing subtypes that are not typically seen in vivo. Compared to neonatal mouse cardiomyocytes, iCLMs have more variable cell size and shape, have less organized sarcomere structure, and demonstrate reduced sarcomere length. Taken together, these results indicate that traditional methods of assessing cardiomyocyte reprogramming by quantifying induction of cardiomyocyte marker proteins may not be sufficient to predict functionality. The automated image analysis methods described in this study may enable more systematic approaches for improving reprogramming techniques above and beyond existing algorithms that rely heavily on transcriptome profiling.

The authors declare no potential conflicts of interest. Running title: Stochastic profiling of luminal breast cancer by 10cRNA-seqThe heterogeneous composition of solid tumors is known to impact disease progression and response to therapy. Malignant cells coexist in different regulatory states that can be accessed transcriptomically by single-cell RNA sequencing, but these methods have many caveats related to sensitivity, noise, and sample handling. We revised a statistical fluctuation analysis called stochastic profiling to combine with 10-cell RNA sequencing, which was designed for laser-capture microdissection (LCM) and extended here for immuno-LCM. When applied to a cohort of late-onset, early-stage luminal breast cancers, the integrated approach identified thousands of candidate regulatory heterogeneities. Intersecting the candidates from different tumors yielded a relatively stable set of 710 recurrent heterogeneously expressed genes (RHEGs) that were significantly variable in >50% of patients. RHEGs were not confounded by dissociation artifacts, cell cycle oscillations, or driving mutations for breast cancer. Rather, we detected RHEG enrichments for epithelial-to-mesenchymal transition genes and, unexpectedly, the latest pan-cancer assembly of driver genes across cancer types other than breast.Heterogeneous transcriptional regulation conceivably provides a faster, reversible mechanism for malignant cells to sample the effects of potential oncogenes or tumor suppressors on cancer hallmarks. Statement of significanceProfiling intratumor heterogeneity of luminal breast carcinoma cells identifies a recurrent set of genes suggesting sporadic activation of pathways known to drive other types of cancer.

Single-cell transcriptomic methods classify new and existing cell types very effectively, but alternative approaches are needed to quantify the individual regulatory states of cells in their native tissue context. We combined the tissue preservation and single-cell resolution of laser capture with an improved preamplification procedure enabling RNA sequencing of 10 microdissected cells. This in situ 10-cell RNA sequencing (10cRNA-seq) can exploit fluorescent reporters of cell type in genetically engineered mice and is compatible with freshly cryoembedded clinical biopsies from patients. Through recombinant RNA spike-ins, we estimate dropout-free technical reliability as low as ~250 copies and a 50% detection sensitivity of ~45 copies per 10-cell reaction. By using small pools of microdissected cells, 10cRNA-seq improves technical per-cell reliability and sensitivity beyond existing approaches for single-cell RNA sequencing (scRNA-seq). Detection of low-abundance transcripts by 10cRNA-seq is comparable to random 10-cell groups of scRNA-seq data, suggesting no loss of gene recovery when cells are isolated in situ. Combined with existing approaches to deconvolve small pools of cells, 10cRNA-seq offers a reliable, unbiased, and sensitive way to measure cell-state heterogeneity in tissues and tumors.

Cancer evolves from premalignant clones that adopt unusual cell states to achieve transformation. We previously pinpointed the oligodendrocyte precursor cell (OPC) as a cell of origin for glioma, but the early changes of mutant OPCs during premalignancy remained unknown. Using mice engineered for inducible Nf1-Trp53 loss in OPCs, we acutely isolated labeled mutant OPCs by laser-capture microdissection, determined global gene-expression changes by bulk RNA sequencing, and compared with cell-state fluctuations at the single-cell level by stochastic profiling, which uses RNA-sequencing measurements from random pools of 10 mutant cells. At 12 days after Nf1-Trp53 deletion, bulk differences were mostly limited to mitotic hallmarks and genes for ribosome biosynthesis, and stochastic profiling revealed a spectrum of stem-progenitor (Axl, Aldh1a1), proneural, and mesenchymal states as potential starting points for gliomagenesis. At 90 days, bulk sequencing detected few differentially expressed transcripts, whereas stochastic profiling revealed cell states for neurons and mural cells that do not give rise to glial tumors, suggesting cellular dead-ends for gliomagenesis. Importantly, mutant OPCs that strongly expressed key effectors of nonsense-mediated decay (Upf3b) and homology-dependent DNA repair (Rad51c, Slx1b, Ercc4) were identified along with DNA-damage markers, suggesting transcription-associated replication stress. Analysis of 10-cell transcriptomes at 90 days identified a locus of elevated gene expression containing an additional repair endonuclease (Mus81) and Rin1, a Ras-Raf antagonist and possible counterbalance to Nf1 loss, which was microdeleted or downregulated in gliomas at 150 days. These hidden cell-state variations uncover replication stress as a potential bottleneck that must be resolved for glioma initiation. Significance: Profiling premalignant cell states in a mouse model of glioma uncovers regulatory heterogeneity in glioma cells-of-origin and defines a state of replication stress that precedes tumor initiation. See related articles by Singh and colleagues, p. 1840 and Schaff and colleagues, p. 1853