NEW YORK (GenomeWeb) – A University of British Columbia team has pioneered a new approach to single-cell sequencing that allows analysis of copy number at the individual cell level with greater coverage uniformity and more reliable detection than other approaches.
The method, called direct library preparation, or DLP, uses nanoliter-volume transposition reactions to do single-cell whole-genome library preparation without preamplification or sorting.
In essence, the approach circumvents the two main methods currently used to study the clonal architecture or individual cellular makeup of tumors: bulk sequencing with computational methods to predict clusters of co-occuring mutations, and single-cell methods using whole-genome preamplification.
Describing the approach in Nature Methods today, investigators led by UBC's Carl Hansen tested the method using 782 cells — 268 from cell lines and 514 from breast tumor xenografts. They showed that the method could achieve greater coverage uniformity and more reliable detection of copy-number alterations (CNAs) than existing single-cell methods, allowing analyses of cell population diversity that would be impossible using bulk sequencing methods, but with the same high depth and uniformity of coverage.
According to Hansen and his coauthors, most studies so far that have examined the clonal architecture of tumors have used either bulk sequencing with informatics approaches to tease out the patterns of mutations associated with different clonal branches, or single-cell methods that necessitate some sort of amplification step before library creation.
Both approaches have important downsides. Although bulk methods can identify major tumor subclones, their capacity to resolve minor populations is limited by sequencing error rates, the team reported. They are also of limited utility when tumor cellularity is low, and they have difficulty adequately addressing subclonal CNAs at low prevalence.
On the other hand, single-cell approaches that use whole-genome amplification before library construction suffer from amplification biases, which decrease coverage uniformity and obscure the detection of CNAs, or polymerase errors that lead to false single-nucleotide variants.
Improved methods have been developed, according to Hansen and his colleagues, such as degenerate oligonucleotide-primed PCR, or DOP-PCR, which achieves higher coverage uniformity than do other approaches like multiple displacement amplification (MDA) and multiple annealing- and looping-based amplification cycles (MALBAC), making it more amenable to single-cell CNA inference.
However, DOP-PCR libraries are less suitable for SNV analysis because their coverage breadth saturates with deeper sequencing, the authors argued.
Researchers essentially must choose whether they want high-fidelity measures of copy number using WGA-based single cell methods or reliable detection of SNVs using bulk sequencing, Hansen and his colleague wrote. One method can't adequately give them both.
Hoping to create a middle ground, the team created a fluidic system that would allow them to construct indexed libraries directly from single-cell template DNA using direct tagmentation of single-cell DNA in nanoliter volumes, followed by PCR cycles to add sequencing adaptors and index barcodes.
The indexed libraries can then be pooled for multiplex sequencing at low depth, which enables single-cell copy-number detection and identification of clonal subpopulations.
Finally, sequencing reads from all cells are mergeable, producing the equivalent of a high-depth bulk genome for SNV, loss of heterozygosity (LOH), and breakpoint inference.
"The chemistry this uses, Tn5 transposase for library production, isn't novel, but the trick is getting that to work on a single cell's worth of DNA," Hansen said in an interview.
"The reason this works is that if you can process the cell in a very small volume you get into the range where the concentration of DNA templates in that cell is in the range for that Tn5 reaction to work efficiently."
"We shrunk it down about 10,000 times from a conventional reaction, so that the concentration is the same as [you would have] with the 10,000 cells in a normal reaction," Hansen added. "That was the fundamental technical breakthrough that allows you to take this approach to sequencing a large number of cells at low depth so that you can avoid amplification that obscures copy-number variation."
In their study, Hansen and his coauthors showed that by upping coverage uniformity, and increasing the reliability of copy-number analysis, they could outperform DOP-PCR, and identify important aspects of cell population diversity — like minor xenograft cell subpopulations that were undetectable by bulk sequencing.
When they then created their bulk genome equivalents for high-depth analysis of SNVs, they saw that the merged libraries had coverage uniformity comparable to that of an actual bulk genome at the same sequencing depth.
Interesting, the DLP method potentially also allows researchers to use the initial CNV results to identify and separate out clonal cell subpopulations, and then create individual bulk genomes for each of these groups for SNV and other higher-depth analyses of isolated lineages.
Hansen said that the IP related to DLP has been licensed to the immune repertoire sequencing company, AbCellera, of which he is president and CEO. Heand his colleagues initially formed a separate company to specifically develop and commercialize DLP, but decided more recently to wrap the technology into a single company.
Though AbCellera's activities have less direct relevance to DLP, Hansen said that there could be applications for the technology in immunotherapy, for example in detecting aspects of cancer clonality or differentiation that predispose patients to respond to a particular immunotherapeutic drug. However, the team is not looking at that actively at this time.
They are, however, optimizing the approach, including shifting the format to allow a significant scaling up of throughput.
"I won't say how we are doing that exactly, but we are on a good path to get up to thousands of cells per run," Hansen said.
In terms of ultimate application, Hansen said that the most immediate promising areas are obviously analysis of cancer cells, either in bulk tumor specimens, or in circulation.
According to the authors, another exciting advantage of DLP is that when single-cell genomes are merged, information about which cell each read originated from is preserved.
"Future computational methods may exploit this property to infer SNVs, LOH, and breakpoints in merged DLP genomes with improved power, and to comprehensively characterize differences in somatic genome variation between copy number subpopulations at lower sequencing depth," Hansen said. "We envision that direct single-cell library preparation may become a new standard approach to the sequencing of heterogeneous populations."
He further noted that at the same sequencing depth and cost of looking at a bulk tumor, "you can get both of these sets of information, the bulk representation, and the copy number information about the lineage relatedness of cells." You also get a massive boost in sensitivity in detecting CNVs."
For instance, the group showed that the method enabled detection of a single CNV at a prevalence of about 0.2 percent.. "With bulk you'd be lucky to pull something that is 20 percent prevalent," Hansen said. So for me, if we can get this standardized there is no reason to ever do a bulk genome anymore. This is more information at the same cost."