Experimental design for single cell analysis

It is challenging to create generalizable designs for single-cell transcriptomic experiments because each one requires the user to make informed decisions in order to obtain interpretable results. These include the selection of sample types, cell numbers and preparation methods; the choice of scRNA-seq techniques and sequencing parameters; and the design of computational analysis strategies to generate insights from single-cell datasets.

Preparation of high-quality single-cell suspensions is key to successful single-cell studies. Although most methods use fresh viable single cells, alternatives include preserved samples and nuclear RNA from frozen tissue. It is important to minimize cellular aggregates, dead cells, noncellular nucleic acids and reverse-transcription inhibitors in single-cell preparations. To minimize these contaminants while maximizing the purity and unbiased recovery of different cell types, one may need to apply optimization (e.g., adjust the number of wash steps, the composition of the wash solution, centrifugation conditions and/or strainer type).

For isolation of single cells from suspensions, samples are density centrifuged, after which they can be used directly for single-cell capture. Solid tissues must first be dissociated via mechanical and enzymatic treatment. It is important to note that sample processing might introduce variation in the gene expression profile. Also, some more sensitive cell types might be damaged during sample preparation, so processing time should be kept to the minimum required. In contrast, too short digestion times could result in incomplete cell separation and the exclusion of tightly interconnected cells from subsequent single-cell analysis.

For transcriptome profiling in single cells, most methods require the physical isolation of cells in individual reaction volumes. Cells can be isolated by microdissection or pipetting, although high-throughput experiments use fluorescence-activated cell sorting (FACS) or microfluidics to guide cells into micro- or nanoliter reaction volumes, respectively. Microfluidic systems capture cells in integrated fluidics circuits, droplets or nanowells, thus allowing thousands of cells to be processed simultaneously while minimizing reaction volumes and reagent use. FACS sorts cells into microtiter plates ready for library preparation by manual or automated processing, and facilitates the exclusion of dead or damaged cells, as well as the enrichment of target cell populations.

To obtain an unbiased view of the cellular composition of a sample, one must capture all cells during the isolation process. Here attention must be paid to very small or large cells that may be excluded during FACS isolation or captured in microfluidic systems, respectively. However, for many experiments, it may be necessary to enrich for or exclude some cell types to increase the total number of cells of interest in the final scRNA-seq libraries. To define adequate cell numbers per experiment, one must consider sample heterogeneity and subpopulation frequency. In particular, larger cell numbers are required to resolve the structure of heterogeneous samples with many expected subpopulations. Also, the total number of cells required increases when rare cell types need to be identified. Because most experiments target poorly described systems, heterogeneity can only be estimated, so pilot experiments are recommended before large-scale data production.

Author: Marco Silvano (ESR7)