In this issue of Genome Biology, Bose and colleagues 1] describe an important technical advance that marries the simplicity of microwells
and the early technique of barcoding of droplets with the parallelizability of microfluidics
to enable many single cells from multiple different samples or perturbations to be
profiled in parallel. The authors’ system consists of five parallel microfluidic channels,
each of which contains over 2000 microwells for cell capture and processing. To operate
the system, the authors load single cells into microwells by simply flushing in a
cell suspension. They then profile single-cell gene expression using either a RNA-printing-based
approach or a bead-based capture modality.
The RNA-printing approach is conceptually akin to microengraving 9]: lysis buffer is added and the microwells are quickly pressed against a slide that
contains covalently grafted oligo-dT primers. Mature cellular mRNA hybridizes to these
oligos and then, after washing, is reverse-transcribed on-chip by flowing appropriate
reagents through the device. Expression can then be quantified by hybridizing gene-specific
probes to these slide-delimited cDNAs. Importantly, this mode of operation maintains
spatial correspondence between cell and well, potentially enabling additional information
collected before lysis (e.g. cytokine secretion) 9] to be used in downstream analyses. In the bead-based capture approach, uniquely barcoded
oligo-dT beads are co-loaded into the microwells before performing lysis. After mRNA
capture, a modified variant of CEL-Seq 10] is performed in which reverse transcription and second-strand synthesis take place
on-chip, and other steps are performed off-chip after bead harvest (although, in principle,
other amplification strategies are possible).
To demonstrate the utility of their scalable platform, the authors use it in a bead-based
capture format to profile approximately 600 cells. In one of their five lanes, they
profile U87 human glioma cells; in another, MCF10a human breast cancer cells; and,
in the remaining three, a mixture of both. A second, similar experiment compares U87
and the human fibroblast cell line WI-38. Through sequencing the cells prepared during
these two runs, the authors detect an average of 635 and 530 genes, respectively,
enabling them to clearly distinguish each type from one another. Clearly, there is
room for improvement in the number of genes detected and in the fraction of high-quality
libraries made using the chip (currently 50 to 70Â % of cells); with future optimizations,
this strategy could potentially match other high-throughput, but serial, approaches
6], 7]. Most importantly, this method is inexpensive ($0.10 to $0.20 per cell) and can easily
be scaled to accommodate additional assay channels and further reduce costs.