By modifying steps in an existing library preparation protocol, a Swedish team has found ways to sequence more of the RNA present in individual cells, while generating longer RNA transcripts.
As they reported online late last month in Nature Methods, researchers from the Ludwig Institute for Cancer Research and the Karolinska Institute tested a series of slight revisions to a template-switching-based single-cell RNA sequencing approach called Smart-seq, looking for ways to improve the throughput, coverage, sensitivity, and cost associated with the method.
The resulting protocol, dubbed Smart-seq2, involves tweaks to multiple steps of the original Smart-seq protocol, including changes to buffer composition, the template switching oligonucleotide, and reverse transcription steps.
With libraries generated from individual mouse or human cells using their kit-free Smart-seq2 approach, authors of the study saw a jump in the amount of RNA that could be captured from each cell, along with increases in transcript coverage and a dip in library prep cost.
"A lot of small things have been improved," the study's senior author Rickard Sandberg, a researcher affiliated with the Ludwig Institute for Cancer Research and the Karolinska Institute, told In Sequence. Even so, he said, the method "still has the same general steps" as the original Smart-seq approach.
Sandberg was the senior author on a Nature Biotechnology study out last summer that described the original Smart-seq method and highlighted its feasibility for assessing messenger RNA patterns in individual melanoma cells nabbed from circulation (CSN 7/25/2012).
Prior to that, Sandberg collaborated with Smart-seq developers from Illumina's research and development arm to evaluate and benchmark the Smart-seq protocol, which was adapted into a commercial kit from Clontech.
The approach is loosely based on Clontech's SMART ("switching mechanism at 5' end of RNA template") method for using template-switching technology to make complementary DNA.
Generally speaking, single-cell sequencing methods for gauging RNA profiles in individual cells had to make tradeoffs between throughput and transcript coverage, Sandberg noted.
Some approaches have been developed for systematically taking a snapshot of expression levels of known transcripts, for example, while others have focused on findings ways to get a more fine grain view of as much of each transcript as possible.
The template-switching scheme employed by Smart-seq lends itself to the latter type of analysis, since it produces cDNAs that span broad swaths of each transcript, allowing a look at alternative isoforms, mutations, and more.
"The Smart-seq that came out last year is the first one that really aims to cover the whole part of the transcript — towards full-length coverage," Sandberg said.
While his team has been generally pleased with that method, he noted that there were "just a few things that were a bit troubling for us."
In particular, he pointed to the cost associated with using a kit-based method, which can add up when preparing libraries for high-throughput experiments. The original Smart-seq protocol also includes a manual pipetting step that can slow down such studies.
"We need more automation, frankly," Sandberg said. "We wanted to remove parts of the protocol that made it hard to put it on a liquid handling robot to make it more automated."
As they began looking for ways to boost Smart-seq's throughput, he and his colleagues realized that a few relatively modest changes could conceivably bump up the coverage and sensitivity of the method, too.
Using hundreds of libraries made from small amounts of RNA, the team tuned various parameters of the Smart-seq protocol one after the other, considering minor modifications that might bolster sensitivity, transcript coverage, or other performance measures.
"We started to experiment with a lot of variations to the protocol," Sandberg said.
There have been a variety of optimization studies done to find library prep strategies that yield information on full-length RNA molecules, he noted, though most have focused on bulk samples rather than individual cells.
"We tried to borrow ideas that have been used in other contexts to see if some of them could actually be beneficial for single-cell level analysis," Sandberg said, explaining that sequencing RNA from individual cells presents a particular challenge because of the very small amounts of material available.
Among the Smart-seq protocol improvements that they settled on were subtle modifications to the template-switching oligo used for the method — a move that appears to have a pronounced effect on the amount of RNA that can be converted to cDNA.
For example, in experiments done using a set amount of purified input RNA, they found that they could double the amount of cDNA produced by swapping the conventional template-switching oligo for a version of the oligo that contains a locked nucleic acid guanylate at its 3' end.
In coming up with the complete Smart-seq2 protocol, the study's authors also fiddled with buffer solutions, modified the pre-amplification method, and found ways to bypass an early bead extraction step.
The group went on to validate that protocol through a series of experiments in individual mouse and human cells.
When they prepared single-cell RNA sequencing libraries from dozens of individual cells from the HEK293T human kidney cell line and compared them with libraries made using Clontech's SMARTer Ultra Low RNA kit for Illumina sequencing, meanwhile, the researchers saw an approximately three- to four-fold increase in RNA sensitivity.
"We think before that we were capturing around 10 percent of the RNA in each cell and now we've boosted it up to 40 [percent]," Sandberg said.
Smart-seq2 libraries sequenced for the study yielded an average of almost 2,400 more genes per cell than those prepared from the same cell line using Clontech's SMARTer protocol, he and his colleagues reported, while stretching out the sequences covered across RNA transcripts.
Such full-length coverage is important when trying to simultaneously track polymorphism or mutation patterns and gene expression with RNA from individual cells, Sandberg noted, which may be beneficial for some cancer studies, for instance.
It is also expected to be useful for characterizing different transcript isoforms and/or monitoring other forms of heterogeneity between cells.
On the cost side, meanwhile, Smart-seq2 offers "kind of dramatic" improvements, according to Sandberg, who pointed to the price advantage of preparing single-cell libraries with readily available reagents.
The new method costs 10 euros ($14) per library in Sweden, according to Sandberg, who said Smart-seq2 would be even more affordable in locales with lower starting reagent costs.
"The savings comes mostly from not having to buy any refined commercial kits," he noted. "We're just buying pure reagents — buying [reverse transcription] enzymes separately, buying the PCR enzymes separately, and getting all the buffers and all the nucleotides ourselves."
For the proof-of-principle single-cell RNA sequencing experiments included in the current analysis, the researchers relied on Illumina's HiSeq 2000 instrument.
But because "the important part of Smart-seq2 is what you do from cell lysis basically until you've amplified the cDNA," Sandberg said the library prep protocol is expected to be compatible with any of the available high-throughput sequencing platforms.
For their part, Sandberg and his colleagues are currently applying Smart-seq2 to early development studies, including experiments that use SNPs to determine whether each sequenced stretch of RNA originates from a maternal or a paternal chromosome.
Along with that type of allele-specific expression analysis, the group is starting to apply the approach to an increasingly broad set of cancer projects, in the hopes of detecting mutations and gene expression shifts associated with individual cells from primary tumors as well as single circulating tumor cells.
The researchers are interested in looking at additional tricks for further improving library prep protocols for single-cell RNA sequencing — particularly with respect to capturing even more of the RNA present in each cell.
Sandberg said his team has no plans to commercialize the Smart-seq2 method, though he noted that the Ludwig Institute has patented some of the core improvements and may explore such opportunities in the future.