scRNAseq Basics: Advice for Common Concerns

Single-cell RNA sequencing (scRNAseq), also known as single-cell transcriptomics, is increasingly used to measure gene expression levels in individual cells. The method is instrumental in identifying distinct cell types in the complex tumor microenvironment, examining cell heterogeneity, or evaluating the effects of checkpoint immunotherapies on specific immune cell populations. The unique information derived from scRNAseq cannot be obtained from bulk transcriptomic methods on groups of cells, as such methods average out potentially important transcriptional differences across multiple cell types. This article offers expert advice about common concerns when performing scRNAseq.

Generally, scRNAseq protocols begin with isolating single cells and tagging their mRNAs (including a unique cell identifier or barcode), followed by reverse transcription, library preparation, and next-generation sequencing (NGS). After sequencing, data analysis determines the number of RNA molecules from each cell. Currently, three types of scRNAseq platforms are common, either based on wells, micro- or nanowells, or droplets. Platform choice depends on throughput levels and a researcher’s overall objective. For example, well-based platforms enable the study of full-length transcripts, but are lower in throughput. Conversely, micro/nanowell- and droplet-based platforms are higher throughput but lack the option to study full-length transcripts.

Isolating cells

A critical factor in the success of single-cell RNAseq is beginning with a high-quality single-cell suspension of high-viability (> 70%) cells—great sample is likely to yield quality data. “The majority of cells or nuclei from various tissue or cell types are compatible for single-cell research as long as they meet the viability and purity requirements,” says Jens Durruthy Durruthy, Associate Director in Product Management, Single Cell at 10x Genomics. “Cell lines or cells already in suspensions (e.g., blood) require typically fewer steps to process as they are already dissociated and/or in suspension.”

Cultured cells, peripheral blood mononuclear cells, and mouse splenocytes are amenable and widely used cells for scRNAseq, according to Frank Harrison, Product Manager in Proteogenomics at Biolegend, while “the highly sensitive nature of neutrophils makes them one of the least amenable cell types for scRNAseq.” Obtaining a single-cell suspension of intact cells when starting from a tissue sample is challenging, says Mike Garbati, Product Manager for Epigenetic Services at Active Motif, who recommends starting with flash freezing the tissue, and avoiding tissue fixatives and preservatives like RNAlater. “Determine the best dissociation method for your tissue type by trying several reagents and techniques, and measuring the viability upon obtaining a single-cell suspension,” he says. “Optimize that technique to obtain the highest percent viability reasonably achievable.”

Automated NGS library preparation can also include isolating cells within a droplet or using microfluidics. “These technologies allow researchers to load a bulk sample of cells, and the automated system does the isolation,” says Thomas Colunga, Commercial Product Manager for Genomics Solutions at Beckman Coulter Life Sciences. “Most of these systems are very easy to use and help to reduce some of the front-end work required to sort cells.”

Taking care to prepare the best starting sample is a worthwhile investment. “Researchers need to thoroughly plan and optimize their isolation protocols to preserve cellular and RNA integrity, and ensure they are obtaining high viability single-cell suspensions with minimal debris,” says Harrison. The presence of dead cells and other debris can contribute to background noise, which can be partially remedied using filtration, washing, or tools like BioLegend’s MojoSort™ Human or Mouse Dead Cell Removal Kits, which they recommend if cell death is greater than 30%.

Sometimes preparation of the starting sample may include an enrichment step such as flow cytometry, magnetic beads, or column-based enrichment methods “which can help to isolate specific subsets of cell types,” says Durruthy. Fluorescence-activated cell sorting (FACS) is particularly suited to sorting and selecting for specific cell types using antibodies. “When deciding to enrich a population, you must consider what cell surface markers to target on your population(s) of interest and design your panel such that you have both positive and negative markers,” says Colunga. “It’s also important to titrate your antibodies and reserve the brightest fluorophores for the markers that are in lowest abundance to ensure they are properly detected.”

How many cells and how deep?

The target number of sequenced cells, and sequencing depth, will vary according to different lab situations, objectives, protocols, and types of experiments. Several factors bear consideration, such as the sample’s cell heterogeneity or diversity, availability of cells, sample type, and the presence of any subpopulations of interest. “If the sample diversity is not known, a high number of cells at low sequencing depth may be the most flexible option to obtain a representative proportion of the cell population and meaningful biological information,” says Durruthy, noting that cell classification can improve with sequencing more cells, albeit at lesser depth. “For highly heterogeneous samples, thousands of cells may be required to resolve each subpopulation fully,” he says.

Colunga notes that because RNA yield varies by cell type, sequencing depth might vary too. “A general rule of thumb is to sequence at least 10k cells for your experiment with no less than 50k reads per cell,” he recommends. “If performing whole genome sequencing, then you want to aim for no less than 30x coverage.” Active Motif recommends that researchers start with 250,000 to 2 million cells. “From this starting material we get single-cell data on 4000 to 7000 cells,” says Garbati. “We perform paired-end sequencing to a read depth of 250 million, which gives us about 50,000 paired-end reads per cell, assuming about 5000 cells.”

To determine your sequencing needs, reviewing the literature and talking to experienced researchers can help you find the appropriate balance between cell numbers and sequencing depth. For example, sequencing fewer cells with less depth may be perfectly sufficient for targeted experiments using a homogeneous cell population. In contrast, “for experimental designs with complex heterogeneous cell populations, rare cell types, or cells with low RNA content, larger cell numbers and deeper sequencing will be needed,” says Harrison. “If the researcher is simply looking at differential expression between treatments of highly expressed genes, then larger cell numbers may not be needed and depth can be decreased.” He recommends running small pilot experiments, using a range of cell numbers and sequencing depths, to help you discover which combination of parameters is right for your situation.

Batch effect variability: sources and solutions

Sequencing cells in multiple runs can incur “batch effect” variability, which stems from a multitude of sources including different cell harvesting times or protocols, sample preparations, reagent lots, sequencing times or protocols, or being conducted in different labs, by different people, or on different sequencers.

Simple precautions can reduce some of this variability. Collect samples at the same time whenever possible; if not—as with patient samples, or different collection locations—then cryopreserve them until you can process all samples simultaneously. Use the same reagent lots, sequencing platform, cell numbers and depth. If possible, minimize the number of researchers conducting the experiments.

Researchers also use BioLegend’s TotalSeq™ Hashtag antibody oligo conjugates to keep track of “batch effects caused by users loading cells on different single-cell chips or different lanes on a chip,” says Harrison. “Each sample can be labeled with a different hashtag, pooled for single-cell processing, and then distinguished by using the hashtag barcodes post sequencing.” Biolegend developed the TotalSeq™ antibodies specifically for CITE-seq (cellular indexing of transcriptomes and epitopes by sequencing), a multiomics method that analyzes cell surface proteins and RNA by single-cell sequencing.

Bioinformatics can also identify and remove at least some batch effect variability from the data. “A number of computational tools, including Seurat, scran, and SCONE, can correct batch effects,” says Durruthy. “Cell Ranger can also perform batch correction for libraries generated with multiple versions of the same chemistry.”

As in many aspects of life, a little advance planning goes a long way with scRNAseq. “A majority of the issues we see stem from researchers jumping into experiments without proper planning,” says Harrison. “Run small pilot experiments to familiarize yourself with the protocol and identify issues early on to avoid wasting reagents, time, and potentially valuable samples.” Study the current literature, discuss with other researchers, and reach out to representatives from manufacturers (and especially their very knowledgeable technicians). You—and your research —will be better off for the efforts.