Transcriptome and single-cell gene expression

Enabled by next-generation DNA sequencers, RNA-seq has supplanted earlier microarrays to allow routine quantitation of transcripts in biological samples (31, 32). Latest high-throughput sequencers from Illumina and other manufacturers can generate billions of sequences from a single experiment thanks to the massively parallel nature of sequencing reactions. Deducing the sequence of messenger RNAs with these sequencers allows rapid quantitative assessment of the expression of tens of thousands of transcripts within a cell or tissue sample. Measurements of bulk RNA expression on a large scale are now commonly deployed to query gene expression in iPSC-derived cells. From RNA-seq data, clustering and unsupervised classification analyses are commonly used to determine specific groups of genes or pathways that may be changed when comparing diseased over normal cells, and thus implicating their potential importance in disease origin or in explaining observed cellular pathologies. RNA-seq also provides context to functions of gene variants in association studies and can be used for fine-mapping and identification of causal variants, as well as the potential mechanisms by which they affect traits. Many GWAS variants function as expression quantitative trait loci (eQTLs) by affecting transcript level whereas other, exonic, variants affect splicing ratios that can likewise be discerned using RNA sequencing (54).

Single-cell RNA sequencing (scRNA-seq) is a recent development for resolving transcriptional heterogeneity within cell populations by allowing transcript expression from single cells to be characterized. Its emergence was driven by technical advances in constructing and amplifying sequencing libraries from the miniscule amounts of RNA, as well as microfluidic contraptions that allow separation of individual cells. To date, three major scRNA-seq approaches are in popular use. The first involves plate-based protocols that place individual cells into wells. The second involves automated microfluidic platforms that capture individual cells on microfluidic chips. The third involves droplet-based massively parallel technique (Table 1).

Comparison of common single-cell analysis techniques

Plate-based techniques such as SMART-Seq offer a fast and efficient method to analyze 50 to 500 single cells in one experiment with flexible experimental set-up (55). Current plate-based techniques boast increased accuracy and short processing time, and are compatible with automation by liquid-handling robotics. They also allow cells of any morphology and size to be analyzed and can read up to 5,000–10,000 genes per single cell. Commercially available, automated microfluidic platform (Fluidigm C1) allows 96 individual cells to be captured at a time on a microfluidic chip. It offers the option to evaluate the captured cells under the microscope before reverse transcription and is effective for comparison of homogeneous cell populations. However, the cost of reagents remains high and >10,000 cells are required as input, rendering analysis of rare or small cell populations possible only when multiple samples are pooled. Ineffective automated sorting of cells into singlets also has been reported, in which multiplets are falsely analyzed as single cells. To overcome such limitations, cell expression by linear amplification and sequencing (CEL-seq) has been developed by combining the two technologies (56). CEL-seq applies molecular barcode to cells at early stages, lowering the reagent cost and increasing the cell number per sample to 500 to 2,000 cells. Subsequently, massively parallel scRNA-seq (MARS-seq) has also been developed by combining single-cell barcoding with 384-well-plate fluorescence-activated cell sorting (FACS) to increase the scale and lower the costs (57). These pooled techniques allow the isolation of various cell types and enhanced throughput. Finally, droplet-based scRNA-seq can tackle tens of thousands of single cells per sample, using barcoded complementary DNAs to label single cells encapsulated in individual droplets (58).

Insights into single-cell transcriptomes have revealed hidden heterogeneity in cell types and cell states (59, 60), decoded dynamic processes and developmental timelines (61, 62), and uncovered disease markers that are masked when averaged in bulk sequencing (59, 63). In iPSC models, scRNA-seq has been used to understand spatial and temporal heterogeneity of reprogramming (64) and differentiation (65), and to identify novel surface markers for enrichment of induced cardiomyocytes (65). Machine-learning algorithms have also been used to predict functional states of iPSC-derived neurons based on single-cell transcriptomes, in which a combination of scRNA-seq and patch-clamping analyses allowed prediction of neuronal physiology and identified biomarkers for electrophysiologically active neurons (66).