NEW YORK – Researchers have described a new approach to measure both protein expression and RNA transcript levels within single cells, which they used to examine the heterogeneity of immune cells.
Recently developed tools like CITE-seq and REAP-seq have enabled scientists to measure protein and transcript levels in parallel, but these approaches can be resource intensive.
The Fred Hutchinson Cancer Institute's Florian Mair and his colleagues instead found that a targeted transcriptomics approach combined with simultaneous protein measurements did not need as high a read depth as other approaches and still maintained high sensitivity of low-abundance transcripts.
As the researchers reported in Cell Reports Tuesday, they applied this approach to study immune cell heterogeneity, and adapted a mass cytometry data-analysis tool called One-SENSE to visualize the protein-transcript datasets.
"We're working on making these techniques, which tell us a whole lot about how the immune system works, more affordable and easier to interpret," Mair, the study's co-first author and a postdoctoral research fellow at Fred Hutch, said in a statement.
For their approach, the researchers used oligonucleotide-barcoded antibodies for staining and BD Biosciences' Rhapsody platform to capture single cells in nanowells for targeted transcriptomics. In this way, they could interrogate 492 immune-related genes and 41 surface proteins simultaneously.
To test their approach, the researchers split peripheral blood mononuclear cell samples from three healthy controls into two batches, one that underwent the multi-omic workflow and one that underwent flow cytometry-based phenotyping. Both approaches, they found, could largely differentiate between various immune cell populations.
Additionally, when they compared their targeted transcriptomic approach to a common whole-transcriptome analysis approach, the researchers found both methods could capture the major peripheral blood mononuclear cell lineages.
Using PBMCs from a separate donor with about 27,000 reads per cell, the researchers examined the read depth needed to resolve these different signals using their multi-omic approach. By subsampling reads from this dataset, they found little difference in the protein signals when they used all 100 percent of the reads or only 20 percent of the reads. At the 10 percent level, though, differences became apparent.
This suggests that between 2,000 and 4,000 reads per cell — or about a tenth of the read depth necessary for whole-transcriptome approaches — are needed for this targeted approach. Additionally, they said 200 to 400 reads per antibody per cell are needed for the antibody portion of the library to deliver sufficient resolution. This, they estimated, would make their approach about five times cheaper.
By adapting One-SENSE, the researchers could visualize correlations between transcript and protein expression in their multiomic dataset. They mapped cells by plotting protein expression profiles on one axis and the differentially expressed gene expression profiles on the other. This, they said, enables easy identification of cellular clusters with similar transcripts but not proteins, and the converse. When they applied this visualization approach to their peripheral myeloid cell data, the researchers could identify different subsets of the peripheral CD14+ myeloid population.
In a statement, Mait and co-first author Jami Erickson, also a postdoc at Fred Hutch, said they hope their findings will enable more scientists to adopt multiomics approaches and, in the future, fuel personalized medicine efforts.