thebench.bio #1 Ecosystem for the Bio-Tech Industry
All Single-Cell Analysis questions

Question 9 of 20 Fundamentals Concept

When would you choose single-nucleus RNA-seq over single-cell RNA-seq?

Short answer

Choose nuclei when cells cannot be dissociated intact, when only frozen archived tissue exists, or when dissociation would bias which cells survive or trigger stress genes. The cost is fewer genes per cell, because cytoplasmic transcripts are lost, and the count matrix must include introns.

Why

Whole-cell methods need a suspension of living, intact cells. Some tissues resist dissociation, such as adult brain with large neurons, and banked frozen tissue cannot yield intact cells at all. Nuclei survive freezing and mechanical stress much better.

Bakken and colleagues, comparing matched mouse cortex data, report that nuclear profiling can be applied to archived frozen specimens, gives less biased coverage of cell types and avoids some isolation artifacts, such as immediate early gene induction. They also show the trade-off. Cells detected about 11,000 genes and nuclei about 7,000, and over half of nuclear reads were intronic, because the nucleus holds unspliced pre-mRNA.

Closely related neuron types were still separated well once intronic reads were counted. That is why the Bioconductor book says the main processing change is to include introns in the gene annotation. QC also changes.

Stripped nuclei should contain no mitochondrial transcripts, so mitochondrial counts measure how well the cytoplasm was removed, not cell stress. Nuclei are a poor choice when the question depends on cytoplasmic or mitochondrial transcripts.

A strong answer covers

  • Nuclei work for frozen and hard-to-dissociate tissue
  • Dissociation can bias survival and induce stress genes
  • Fewer genes detected because cytoplasmic RNA is lost
  • Count intronic reads, since nuclear RNA is largely unspliced
  • Mitochondrial reads become a measure of nuclear purity

Common mistakes

  • Counting only exonic reads for nuclei and losing much of the signal
  • Applying the same mitochondrial threshold logic used for whole cells
  • Assuming nuclear and whole-cell profiles can be merged without accounting for the technology