This article has been updated to include information about Dolomite Bio's plans to commercialize a chip that is compatible with DroNc-seq.
SAN FRANCISCO (GenomeWeb) – Recent advances in single-cell transcriptome sequencing are enabling scientists to analyze RNA from large numbers of cells or nuclei more cost effectively. The technology advances are also facilitating studies that previously weren't possible, such as measuring transcriptional differences between neurons in a healthy state versus in neurodegenerative disease, or to generate cell catalogs of entire model organisms.
This week, researchers from the Broad Institute described a single-nucleus RNA sequencing method called DroNc-seq in Nature Methods that builds on another method called Drop-seq. Also, last week, researchers from the University of Washington, in collaboration with Illumina, described in Science a single-cell RNA sequencing method that makes use of combinatorial indexing.
The Broad Institute researchers developed DroNc-seq to sequence RNA from nuclei as opposed to entire cells because theywanted to use single-cell techniques to study the brain, but found that the typical enzymatic dissociation techniques that are used to separate cells from tissue often destroyed the neurons and RNA they wanted to study.
Looking at nuclei as opposed to cells did not require "harsh enzymatic dissociation," said Naomi Habib, a co-lead author of the Nature Methods study. "And by looking at the nuclei, we can really start to understand the different cell types and dynamic processes, like neurogenesis," she said.
Previous single-nucleus sequencing techniques relied on sorting nuclei by FACS, a process that enabled a few hundred nuclei to be analyzed. But because the Broad team's goal was to study the human brain, which is very diverse, the team wanted a method that would enable them to analyze tens of thousands of nuclei in a way that wasn't too labor or time intensive, Habib said.
The researchers turned to the previously developed Drop-seq method, a droplet microfluidics-based approach developed in Steven McCarroll's laboratory at Harvard, and tweaked it so that it would work for nuclei.
Anindita Basu, co-lead author of the Nature Methods study, said that one change to the method was in the size of the droplets. Typically, nuclei contain less RNA than cells, so the researchers had to decrease the droplet size in order to maintain a high concentration of RNA within each drop. As a result, she said, the microfluidic device had to be modified slightly to accommodate the smaller droplet size.
Habib added that aside from the droplet size and microfluidics, protocols for isolating nuclei had to be scaled up. "It's important to have good, healthy nuclei, but you also want the process to be fast," she said.
Basu added that the group will continue to optimize the protocol, in particular the microfluidics portion, though it already works well.
She said she anticipates that the method would enable single-nucleus sequencing more broadly. The method does not require fresh samples, which makes it more amenable for research labs. For instance, she said, fresh tissues are often delivered at the end of the day and experiments needed to be started right away. But with DroNc-seq, samples can be preserved and experiments can be run when it's convenient.
In addition, Basu said, any lab that currently uses the Drop-seq method would just need to change that protocol slightly to enable DroNc-seq.
In the study, the researchers analyzed nuclei from a mouse cell line, frozen mouse brain tissue, and archived frozen adult human brain tissue collected post mortem. They used Drop-seq to sequence single cells from the same mouse cell line to compare the two methods. The researchers found that both methods generated high quality libraries and had similar throughput. In addition, the average expression profile for single nuclei was highly correlated to that of single cells. As expected, DroNc-seq yielded greater expression levels for genes that are known to be enriched in the nucleus. However, the DroNc-seq method did result in a higher proportion of reads mapping to introns — at almost 42 percent — compared to just 9 percent for Drop-seq. The authors wrote that this result reflected the "enrichment of nascent transcripts in the nucleus."
Importantly, the researchers demonstrated that DroNc-seq could work on archived human tissue. They analyzed seven frozen postmortem samples from the hippocampus and five samples from the prefrontal cortex. Analysis revealed clusters that corresponded to known cell types. In addition, Habib said, the team was able to "detect small differences in subpopulations of cells and find more rare cells, which usually we wouldn't be able to see." Going forward, Habib said, this ability "is very exciting when we're thinking in the context of disease."
Habib said that the method could have a number of different applications, including for studying neurodegenerative diseases like Alzheimer's or tumor heterogeneity in various cancers, and even normal aging processes in the brain. In addition, senior author Aviv Regev is spearheading an initiative to build a human cell atlas, and the method will likely be used in that project, Habib said.
The protocol is available for any lab to use. The researchers also have a patent application on the method and Habib noted that a number of companies have expressed interest in licensing the technology. In fact, UK firm Dolomite Bio, which focuses on single-cell applications, said it plans to launch a chip this fall that would be compatible with the DroNc-seq protocol.
The Nature Methods study is just one demonstration of increased interest in developing higher-throughput methods to profile single cells. "There's a big interest in sequencing nuclear RNA, mainly because we as a community are starting to build atlases of cell types and molecular states," said Cole Trapnell, an assistant professor in the department of Genome Sciences at the University of Washington and a co-lead author of the study published in Science. Echoing the Broad researchers, henoted that one advantage of single-nucleus sequencing is that it does not require dissociating tissues.
However, he said, single-cell RNA sequencing will continue to be important in research, and in the Science study, his group demonstrated that a combinatorial indexing strategy could be used to sequence RNA from single cells in high throughput at a low cost. In the study, the researchers sequenced nearly 50,000 cells from the roundworm Caenorhabditis elegans, covering the transcriptome for each cell at about 50x. It was the first time the transcriptome of every cell in an entire organism has been sequenced, Trapnell said.
The UW team has been developing combinatorial indexing strategies as part of a collaboration with Illumina. Last month, the groups developed a method known as contiguity-preserving transposition sequencing on beads (CPTv2-seq) in order to haplotype genomes, and earlier this year, they applied combinatorial indexing to single-cell Hi-C sequencing.
Trapnell said that as a result of the agreement with Illumina, he could not comment on the commercialization possibilities.
Similar to the DroNc-seq protocol, Trapnell said, the combinatorial indexing approach seeks to make single-cell sequencing scalable for tens of thousands of cells. One problem with typical single-cell sequencing methods is that they require sequencing libraries to be made for each individual cell, but the cost of making those libraries increases with more cells. "Sequencing twice as many cells costs twice as much and takes twice the effort," Trapnell said. With combinatorial indexing, the cost and effort increase less than linearly, and the more cells are analyzed, the more is gained in cost and time savings.
In the study, the team estimated that constructing libraries for tens of thousands of single cells could be completed by one person in two days at a cost of $.03 to $.20 per cell.
The way the method works is that instead of sorting individual cells, cells are distributed across either a 94-well or a 384-well plate, where more than one cell will land in an individual well. Then reverse transcription is performed using a primer that also contains a barcode specific to the well. Another unique aspect of the method is that the cells are not completely split open, Trapnell said. Instead, they are perforated and the libraries are made inside the individual cells. After the initial step, the cells are then pooled, mixed, and redistributed onto a new plate. Library construction is finished and the final PCR reaction includes a separate well-specific barcode. Again, more than one cell is in a well, but at a low enough concentration that there's a high probability that each cell will end up with a unique pair of barcodes. After the libraries are made, the cells can be pooled and sequenced.
In the Science study, the team applied the method to the model organism C. elegans, generating transcriptomes from more than 42,000 single cells. The researchers were able to identify 29 distinct groups of cells and 26 different cell types. In a separate experiment, they sequenced an additional 7,000 cells since in the first experiment intestinal cells were not represented. Because intestinal cells are polyploidy in C. elegans, the researchers hypothesized that they were excluded from the original analysis.
Trapnell said that the group is interested in continuing to apply the technique to C. elegans, adding that the group has been collaborating with Robert Waterston's UW lab, which focuses on C. elegans genomics. "Working with him made it possible to better understand the data," he said. Going forward, the team is interested in collaborating with Watterson's lab to apply the single-cell combinatorial indexing approach to the different developmental stages of C. elegans. "It would be amazing to have the entire developmental history of the worm," he said. In addition, the group plans to use the method on other model organisms, like mouse, and eventually in humans to "generate very large-scale catalogs of all the different cell types."