Two related but unique challenges – analyzing heterogeneous populations, and a scarcity of biological samples – challenge biomedical research. The inter-relatedness of these obstacles can be seen when identifying a hereditary mutation like BRCA1/2 that increases cancer risk. Such a mutation represents 1 base pair within the 3,000,000,000 contained in a single cell. One way to increase the amount of material analyzed is to isolate more cells; however, only 100-1000s of cells can be taken from a patient blood sample. PCR further increases detection by allowing specific amplification of specific DNA regions. These approaches overcome the sample limit and allow reliable detection of DNA sequence, but increase the challenges of working with large populations of molecules/cells. In the case of a tumor suppressor like BRCA1/2, if a mutation is detected in 50% of the DNA, the implications are unclear. Do 50% of the cells isolated contain this mutation on both copies of the gene? Or do 100% of the cells contain a single mutant copy? These questions have pushed scientists to develop techniques that measure molecules within a single cell.
As with most new scientific technologies, this is not the only approach to single cell RNA-seq (scRNA-seq) so researchers were less interested the novelty, but rather in the methods’ sensitivity and reliability. In particular, scientists need to detect rare cell populations in a reproducible manner. To do this researchers performed the scRNA-seq and then performed principal component (PC) analysis, a common technique for visualizing relatedness of high-dimensional data. Mixing two very different cell types (human kidney and mouse embryonic fibroblasts) revealed that 1.6% of the GEMs were doublets or contained two cells; however, with two distinct RNA profiles such errors were easily identified and could be excluded from further analysis. This approach was further validated using two different human cell line combinations (human kidney and T-cell lymphoblast), In this experiment, cells were mixed at varying ratios to test the limits of detection and PC visualization. Even when mixed at a 1:99 ratio, two clear populations emerged and were not obscured by doublets, which occur at the same frequency as before. Overall, these experiments show single-cell RNA profiles are distinguished reliably and with high sensitivity.
In all of these experiments the input data for PC analysis was the quantity of mRNA. This is a common approach for characterizing unique cell types, but cannot distinguish very similar cell types. In order to determine mRNA abundance, researchers actually sequence ~250 nucleotides of the amplified cDNA allowing even more information than mRNA abundance to be extracted. In this case Fred Hutch scientists searched the sequence data for single nucleotide variants (SNVs). SNVs are spots within the genome where single nucleotides vary among the population at an appreciable rate (>1%) thus they can be used to distinguish geographical or ethnic populations. The two human cell lines tested each had ~350 SNVs so researchers performed PC analysis using SNV data, rather than mRNA abundance. This approach also clearly distinguished the two groups even when mixed at a 1:99 ratio, though with less sensitivity because only SNVs were used as input data rather than the entire mRNA profile.
Artificially mixing samples allowed researchers to test the lower limits of detection but did not demonstrate real experimental application. For this researchers used peripheral blood mononuclear cells (PBMC). Blood is one of the most diverse biological samples as it contains red blood cells, and components of both innate and acquired immune response. PBMCs consist broadly of monocytes, T cells, B cells, and Natural Killer cells – however, these groups contain multiple subtypes with unique expression profiles making this a naturally heterogeneous population. Researchers identified 10 unique populations after single cell RNA-seq was performed on 68,000 PBMCs. One of the most impressive feats was identifying a megakaryocyte population that consisted of 0.5% of the PBMCs.
These experiments demonstrated the value of this technique in well-established systems, but to finish these researchers identified a new diagnostic/clinical use for this sensitive approach. Bone marrow transplant remains an effective cure for some forms of leukemia and lymphoma. The success of complication-free transplant is aided by monitoring the ratio of donor and host immune cells. This is difficult as they are the same cell type, so researchers used SNV data to correctly predict which fraction of the PBMC came from host and donor.
Dr. Bielas also emphasized a major strength to this approach over other single cell technologies, “its unbiased discovery - thousands of digital gene expression measurements are made per cell, then used to classify them. Whereas Flow cytometry or CyTOF you use pre-determined sets of antibodies and try to multiplex them.” The collaboration between 10x Genomics demonstrates the value of academic-industry partnerships. The next steps for this work will be to “integrate gene expression profiles from individual cells with their spatial location in a particular tissue/tumor” said Dr. Bielas.
Zheng GX, Terry JM, Belgrader P, Ryvkin P, Bent ZW, Wilson R, Ziraldo SB, Wheeler TD, McDermott GP, Zhu J, Gregory MT, Shuga J, Montesclaros L, Underwood JG, Masquelier DA, Nishimura SY, Schnall-Levin M, Wyatt PW, Hindson CM, Bharadwaj R, Wong A, Ness KD, Beppu LW, Deeg HJ, McFarland C, Loeb KR, Valente WJ, Ericson NG, Stevens EA, Radich JP, Mikkelsen TS, Hindson BJ, Bielas JH. (2017). Massively parallel digital transcriptional profiling of single cells. Nat Commun, 8, 14049.
Funding for this research was provided by the Canary Foundation, the Ellison Medical Foundation, National Institutes of Health, the US Department of Defense, and the Pacific Ovarian Cancer Research Consortium.