The process of development in multicellular organisms may be specified by a lineage tree, whose structure represents the developmental trajectory of each cell, and by the dynamic gene expression changes that occur in individual cells progressing along these trajectories. Transcriptome measurements from single developing cells may provide sufficient information to infer an unknown lineage and estimate lineage-specific expression dynamics. Single-cell RNA sequencing provides a method to make such measurements; however, the adequacy of scRNA-seq measurements for developmental lineage reconstruction is unknown. In particular, technical noise may generate spurious expression similarity between cells and limited sensitivity may prevent detection of critical regulators that are expressed at low levels.
Here, we evaluate the adequacy of Drop-seq [1] for developmental lineage reconstruction. This recently developed scRNA-seq method provides high-throughput measurements at low cost and so could be used to measure the transcriptomes of individual cells densely sampled over developmental time, even for large lineage trees. We generated expression measurements for thousands of melanoma cells using both Drop-seq and single-molecule florescence in situ hybridization (smFISH) for hundreds of genes. We compare the population distribution of expression levels estimated using these methods and, treating smFISH as a gold standard, provide a gene-level assessment of Drop-seq measurements in terms of sensitivity, precision and accuracy. We consider the relative strengths of in situ imaging and sequencing methods, as well as the possibility of integration of these different data types. As proof of concept, we are generating Drop-seq measurements for individual C. elegans cells from dissociated mixed-stage embryos. We will compare these data to dynamic, single cell expression measurements of the developing worm generated by automated lineage-tracing microscopy. Through this comparison, we will further assess potential sampling bias in the Drop-seq method and the ability to detect key developmental genes.
1. Macosko, Evan Z., Anindita Basu, Rahul Satija, James Nemesh, Karthik Shekhar, Melissa Goldman, Itay Tirosh, et al. Cell 161, no. 5 (May 21, 2015): 1202–14.