When Alex Rosenberg, PhD, and Charlie Roco, PhD, were graduate students in Georg Seelig’s lab at the University of Washington, they drew out their idea for how to increase the scalablility of single-cell RNA sequencing (scRNA-seq) on a whiteboard. At that time, roughly five years ago, “large scale was about 100 cells,” said Rosenberg. They developed their idea into the technique known as Split Pool Ligation-based Transcriptome sequencing (SPLiT-seq).
According to Rosenberg, the emails started rolling in as soon as the proof-of-concept paper was published in Science in 2018. A “huge number” of groups reached out, he said, asking how to set up SPLiT-seq in their own labs. As soon as the two students earned their PhDs, they jumped on the opportunity they saw by starting a company named Split Biosciences (recently renamed Parse Biosciences) to commercialize the technology.
Last week, at the annual American Society of Human Genetics (ASHG) meeting, the team at Parse Bio released a new kit complete with data analysis software. The product allows any researcher with a single cell suspension, and access to DNA sequencing, to do scRNA-seq on up to one million cells.
A better mousetrap
Just three years before Rosenberg and Roco’s SPLiT-seq paper, two research groups from Boston—one from the Broad Institute, led by Evan Macosko, PhD, and the other from Harvard Medical School led by Marc Kirschner, PhD—published back-to-back papers in Cell describing microfluidic-based methods to isolate single cells in droplets. The next year, 10X Genomics launched their Chromium System based on these technologies. Since then, 10X Genomics has become the established leader in the field and helped launch a scRNA-seq revolution.
But Rosenberg argues that there is room for improvement, and that Parse Bio’s kits are better than the technologies that exist today. Because of SPLiT-seq’s simplicity, he said, it lacks some of the limitations found in droplet-based approaches. One example is that the size of the cell doesn’t matter in SPLiT-seq. This, and other advantages, makes scRNA-seq both more accessible and easier to scale up.
The cell is the compartment
Single cell means just that, with microfluidic-based systems facilitating the physical separation of cells. With each cell in its own, individual compartment, they can be individually labeled with a barcode that is specific to their section.
But SPLiT-seq works differently—the cells are not physically isolated from one another. The cell is the compartment. To do this, the cells are fixed and then permeabilized. The biochemical reagents enter the cells, the chemistry occurs in the cells, and the barcodes are added. Using combinatorial barcoding, the method can label huge numbers of cells.
Without the need for an additional benchtop machine, any molecular biology lab could theoretically do a scRNA-seq experiment using SPLiT-seq. After working through a kit, a library is prepared that can be sequenced on any sequencer. The data is then analyzed using the software included in the kit.
Parse Bio has had a kit available since February of this year (the Evercode Whole Transcriptome Kit) that could perform scRNA-seq on 100,000 cells. Last week, at ASHG, the company introduced the Evercode Mega that can do one million cells, and an Evercode Mini for smaller experiments.
It’s all about probability
How does SPLiT-seq parse out one million cells? Rosenberg explained it like this: If you start with one million cells, and split them into 96 wells, each well has roughly 10,000 cells. Every cell in one well gets one barcode. That’s 10,000 cells with one barcode—not exactly a unique label. Then, the cells are pooled together and randomly distributed again into a second plate with 96 new wells. A second barcode is added to every well as each cell starts to take a unique path through these wells. This process is done over again through a third and fourth round. The probability that any two cells get the same combination of barcodes becomes smaller and smaller as they progress through the protocol.
There is a chance, however, that two cells will travel together through all four rounds creating a doublet. In other, droplet-based techniques, there is also a probability that two cells will end up in the same droplet. When trying to scale up using droplets, as more cells are put in, the probability for doublets increases.
SPLiT-seq yields lower numbers of doublets, given the exponential nature of combinatorial barcodes. Doublets can be measured using a standard method involving the mixing of mouse cells and human cells. Using this technique, at a million cells, the observed doublet rate of Parse Bio’s kit is 3.2%. The real doublet rate is twice that.
To the victor go the spoils
Like any technology that breaks out into an already mature market, SPLiT-seq has an uphill road to overcome the market hurdle. In the past five years, scRNA-seq has exploded, largely due to 10X Genomics Chromium—which is up and running in core genomics facilities all over the country. One researcher at an academic institution on the West Coast told GEN that their lab has started the generation of a big atlas using the Chromium and does not intend to change now.
The game of developing new genomics technologies is not one played by those weary of competition, as any DNA sequencing company can attest. But when Rosenberg told GEN that Parse’s products, “are better than the competition on literally every single axis,” he made it clear that he is ready to get off the bench.