NEW YORK – Researchers from the Austrian Academy of Sciences' Research Center for Molecular Medicine (CeMM) have devised a new single-cell sample preparation method that raises throughput as much as fifteenfold on droplet-based platforms, such as 10x Genomics' Chromium.
Single-cell combinatorial fluidic indexing RNA sequencing (scifi-RNA-seq) employs a single round of barcoding prior to droplet encapsulation, barcoding the transcriptome molecules inside each cell or nucleus prior to "isolation," either on Chromium or another platform. That indexing allows massive overloading beyond 10x's recommendations, filling droplets with as many as 15 nuclei, the transcriptomes of which can be picked apart in downstream analyses.
"With 10x's standard protocol, most of the emulsion droplets remain empty, and the reagents are used very inefficiently," said Christoph Bock, a researcher at CeMM whose lab came up with the method. He's also senior author of a paper published this week in Nature Methods describing and benchmarking the method.
Normally, getting more than one cell in a 10x droplet — a doublet — leads to unusable results, so the firm recommends loading up to about 16,000 cells or nuclei per microfluidic channel to capture up to 10,000 cells or nuclei. With eight channels on its cartridges, that yields about 80,000 transcriptomes per run. But Bock's team found that not only can indexing on 384-well plates provide the mathematical feasibility to run more cells, but also that Chromium can physically handle up to 1.5 million nuclei per channel without clogging.
"Nothing is a bad idea in life sciences if you get closer to mastering biology," a 10x Genomics spokesperson said in an email. "Conceptually, the approach is interesting, but we have yet to see how it will have broader adoption, as it is still in the academic realm. It is an academic method that still requires more investigation and work to commercialize it."
The method offers another way to increase throughput in single-cell sequencing, which could help lower sample preparation costs, a major consideration with these studies. 10x is working on its own ways to increase throughput, including the forthcoming Chromium X instrument, which it says will be able to process 1 million cells at a time, and its recently released CellPlex kits, which uses oligo-conjugated lipids to label cells upstream from droplet encapsulation.
But CellPlex is a cell-hashing strategy that flags and discards droplets containing more than one cell, Bock said, providing about a twofold increase. Another cell-hashing method, cellular indexing of transcriptomes and epitopes by sequencing (CITE-seq), has enabled multiplex readouts from single-cell experiments.
Other groups have also come up with different combinatorial indexing approaches to improve throughput, though not necessarily on droplet-based single-cell platforms. Researchers led by Jay Shendure and Cole Trapnell of the University of Washington developed sci-Plex, a plate-based method for high-throughput screening that also uses nuclear hashing, publishing it in Science in 2019.
Bock said his team was spurred to create scifi-RNA-seq after some frustration with multi-round plate-based combinatorial indexing methods. "It turned out to be a hassle," he said, and tedious, since it required three rounds of barcoding to provide a benefit over Chromium's base throughput. "You lose a lot of material and the results we were getting from those processes were not that great," he said.
His team tried to create a hybrid approach and came up with the combination of one round of indexing on a 384-well plate paired with overloading. They settled on running experiments at around 750,000 nuclei per channel, about 50 times what's recommended. "That's mainly because we didn't need more," Bock said. That led to about five to 10 nuclei per droplet, with as many as 15. "If we were loading more, we could push that more," he said.
The method fixes cells to permeabilize the nuclei and reverse transcribes to cDNA prior to encapsulation. Thus, it uses the 10x ATAC-seq (assay for transposase-accessible chromatin by sequencing) kit for library prep, rather than the 3' gene expression kit, which also adds a specific primer binding site.
It also uses a lower level of sequencing due to the scale of experiments. Sequencing to saturation is simply not feasible, Bock said, and his lab recommends aiming for at least 10,000 reads per cell. He noted that the sequencing protocol is slightly different on Illumina's NovaSeq platform than it is on the NextSeq platform, so researchers attempting it should pay close attention to the steps.
Scifi-RNA-seq did not perform as well as 10x's gene expression assay on fresh cells, Bock said. "That's expected," he said, "but not necessarily true for the future." He plans to optimize the method to do combinatorial indexing without fixation. But for fixed cells, it performs as well as the standard 10x protocol, he said. "At this point in time, the goal with scifi-RNA-seq is not to replace 10x or microfluidics, but to reach applications where massive scale is needed," he said.
Those applications include high-throughput drug screens and CRISPR perturbation screens (Bock also is a founder of Aelian Biotechnology, a company that offers CRISPR screening services to the pharmaceutical industry).
"We're not interested in whether a particular transcription factor is expressed in one cell," he said. "This is mainly about cell counting applications or finding which cells are there or whether there are relatively broad changes in cell identity."
He suggested that the method could even be used as a replacement for flow cytometry. "Rather than count surface markers, you could do it with transcriptomes," he said. And ultimately, he wants to take scifi-RNA-seq into the clinic, perhaps in genetics or cancer diagnostics.
Bock and first author Paul Datlinger, a graduate student, have applied for a patent on the barcoding scheme and overall method. Bock noted that they've made the protocol available to academic researchers already. He's open to the idea of kitting the technology or integrating it into a platform and said he's had some preliminary discussions with companies but did not elaborate.