The challenges of single cell transcriptomics on difficult human tissue: the placenta
TL;DR
A comparison of five single-cell transcriptomics methods on human placenta reveals that sample processing drastically alters detected cell-type proportions, while in situ combinatorial indexing shows a strong, unexpected bias toward nuclear rather than cytoplasmic RNA.
Problem / question
Applying single-cell transcriptomics to difficult tissues like the placenta is challenging due to high RNase activity, hypoxia, and tissue trauma, making it unclear how different storage methods (fresh, frozen, FFPE) and sequencing chemistries (droplet versus plate-based, polyA versus probe-based) bias the resulting cell-type proportions and gene expression profiles.
Methods
The researchers tested five single-cell and single-nucleus modalities on two human placenta samples (delivered at 36 and 37 weeks): fresh droplet poly-A (10X Genomics), fixed in situ poly-A (Parse Biosciences), fixed droplet probe (10X Genomics), nuclei droplet poly-A (10X Genomics), and FFPE droplet probe (10X Genomics). They compared cell capture, sensitivity, and differential gene expression, using a 477-gene 10X Xenium in situ spatial transcriptomics sample as a baseline for true cell-type proportions.
Key findings
Sample processing drove the largest variance in cell-type proportions: freshly dissociated tissues heavily favored immune cells (like NK/T cells) and missed myofibroblasts, whereas stored tissues (FFPE and snap-frozen nuclei) favored non-immune cells and closely matched the spatial transcriptomics baseline. The fixed droplet probe method yielded the highest cell capture efficiency (57%) and sensitivity. Surprisingly, the fixed in situ poly-A method, designed for whole cells, correlated most strongly with the single-nucleus method (Pearson coefficient 0.95 for syncytiotrophoblast and NK/T cells) and captured high levels of nuclear MIR and LINC genes, indicating a significant loss of cytoplasmic mRNA.
Why it matters
These findings provide concrete guidance for experimental design, demonstrating that researchers must choose sample processing methods carefully to avoid artificially skewing cell populations or missing cytoplasmic disease markers in challenging clinical tissues.
Limitations
The study relies on a small sample size of only two biological replicates for the single-cell comparisons and one for spatial transcriptomics, and the comparative gene expression analyses were restricted to the most abundant cell types (syncytiotrophoblasts and NK/T cells).
Takeaway
Sample storage and processing methods skew single-cell population data far more than the sequencing chemistry itself, and some whole-cell plate-based methods primarily capture nuclear RNA.