NEW YORK (GenomeWeb) – Two research teams have described new single-cell RNA-sequencing methods aimed at addressing the challenges of low sensitivity and scalability of existing approaches.
In separate papers, Baylor College of Medicine researchers describe their multiple annealing and dC-tailing-based quantitative single-cell RNA-seq (MATQ-seq) that they reported could find slight differences in gene expression, while researchers from 10X Genomics and the Fred Hutchinson Cancer Research Center reported on their Chromium Single Cell 3' Solution approach that they said enabled the profiling of thousands of immune cells.
As Baylor's Chenghang Zong and his colleagues reported in Nature Methods, MATQ-seq aims to increase the sensitivity of RNA-seq by boosting the reverse transcription efficiency and the subsequent production of PCR amplicons.
"Sensitivity has been a key limitation of single-cell RNA-seq methods to date, and the low sensitivity of these methods also limits accuracy," Zong and his colleagues wrote. "MATQ-seq provides high sensitivity and accuracy for detecting subtle differences in gene expression between single cells of the same type, and it gives researchers the ability to capture non-polyadenylated RNAs."
The method relies on primers composed of mostly G, A, and T bases inspired by those used in multiple annealing and looping-based amplification cycle (MALBAC) amplification. At the same time, to reduce the bias that is introduced during PCR amplification after second-strand synthesis, MATQ-seq uses random hexamer unique molecular identifiers to label the molecules before amplification and subsequent sequencing.
Using ERCC RNA spike-in controls, the researchers calculated that MATQ-seq has a capture efficiency of 89.2 percent and noted that the levels were correlated between samples. In comparison to the analysis of other single-cell RNA-seq approaches like SMART-seq2 or SUPeR-seq on HEK293T cells, they reported that MATQ-seq had neither 3'- nor 5'-end bias and had an increased detection efficiency of low-abundance genes.
In particular, Zong and his colleagues noted that MATQ-seq could detect differences in gene expression between single cells of the same type. For instance, they were able to detect bursts of transcription within a few cells. When they split 38 single cells into two random sets of 19, they noted that the average Fano factors — a ratio of variance to mean gene expression — of the pooled sets correlated. But when they compared the Fano factors of the single cells themselves to the single-cell averages, there was no correlation, indicating that their approach could capture variants in single-cell gene expression due to transcriptional bursting.
The 10X Genomics and the Fred Hutch team, meanwhile, turned to a droplet-based approach to enable single-cell transcriptomics to be ramped up. As they reported in Nature Communications, their approach uses a microfluidics platform that builds off the GemCode technology in which gel beads with barcodes, index molecules, primers, and more are combined with single cells in oil. Inside each droplet, after emulsion is broken, reverse transcription takes place to yield barcoded cDNA for sequencing.
According to the researchers, such cell encapsulation has about a 50 percent capture efficiency and can take place in about six minutes. It could also, they reported, be used to profile some 250,000 single cells from 29 different samples and distinguish individual cells based on their SNVs.
The team led by Fred Hutch's Jason Bielas used the approach to profile about 17,000 single bone marrow mononuclear cell samples obtained from two patients before and after they underwent allogeneic hematopoietic stem cell transplants for AML. Before treatment, they noted that these samples had three times to five times the unique molecular identifier counts per cell of healthy cells, which they noted likely reflected the abnormal transcriptional programs of the leukemic cells.
For one patient, they noted the presence of two genotypes in the post-transplant samples, one at 13.8 percent and the other at 86.2 percent, reflecting the host and donor genotypes. For the other patient, however, they only detected the genotype of the host. Both of these findings were validated with a clinical chimerism assay.
Based on an SNV analysis, the researchers also compared cell subpopulations across the samples. For the first patient, they uncovered a high level of erythroid cells before transplant — reflecting the patient's erythroleukemia diagnosis — and blast and immature erythroids after transplant — which they said was consistent with the relapse the patient experienced. They added that this would have been tricky to detect with FACS, as there are a limited number of markers for early erythroid cells.
"We were able to discover cells identities without looking at them under a microscope or knowing anything about the patterns of proteins on their surfaces," Bielas added in a statement. "That can be particularly helpful in finding leukemic cells from a particular lineage for which there are few surface markers."