NEW YORK – Researchers at Oregon Health & Science University continue to improve the capabilities for their combinatorial indexing-based single-cell assays for chromatin accessibility and methylation.
Earlier this month, researchers led by OHSU's Andrew Adey published a paper in Genome Research describing an updated method for single-cell ATAC-seq (assay for transposase-accessible chromatin by sequencing) that boosts the number of open chromatin profiles obtainable from a single experiment by a factor of about 50, into the hundreds of thousands, while also reducing per-cell sample prep costs.
And in December, Adey's lab published an update to its single-cell DNA methylation profiling assay, dubbed sciMETv2, which leads to a 14-fold improvement in genome coverage and improves the workflow to make it more manageable.
"The protocol was miserable," said Adey. "It was 16 hours long with no stopping points and while it produced good data, it wasn't very robust." Moreover, it required custom sequencing primer and custom sequencing recipes.
Now, the longest step is only about 1.5 hours and there are multiple stopping points. Coverage is improved, while also reducing per-cell costs. The lab has developed two flavors of sciMETv2 with different ligation methods, including a "splint ligation" workflow that is cheaper and faster than the linear amplification-based workflow, making it especially appealing for studies with large numbers of cells.
This splint amplification-based methylation assay is being optimized in collaboration with Seattle-based startup Scale Biosciences, Adey said, as is a version of the ATAC-seq assay.
The update to sciMET "provides a good opportunity to finally commercialize single-cell methylome assays," Chongyuan Luo, a researcher at the University of California, Los Angeles, said in an email. Luo, who was not involved in the papers, led development of a similar assay, single nucleus MethylC sequencing (snMC-seq).
Existing methods require fluorescence-activated cell sorting to dispense cells onto plates, which has been a "major obstacle" for getting it used by more labs, Luo said. "If sciMETv2 can be successfully commercialized, I anticipate very broad application and rapid adoption, especially in the cancer genomics field," he added. "Tumors are known to associate with highly heterogeneous and unusual DNA methylation patterns across individual malignant cells."
Together, the new methods are indicative of two trends in single-cell sequencing: toward ever larger numbers of cells and a focus on multiomics. Commercial solutions from companies such as 10x Genomics and Parse Biosciences are now able to process as many as 1 million cells at a time and increasingly offer the ability to do more than just gene expression profiling.
Adey's research has focused on methods that make use of single-cell combinatorial indexing to increase the scale of experiments, including applications in single-cell whole-genome sequencing, 3D genome structure (Hi-C), RNA sequencing, and ATAC-seq, as well as spatial methods. This general strategy uses separate rounds of plating cells and tagmentation of DNA molecules of interest to help assign reads to a particular cell during analysis. Two rounds of Tn5 transposase-based barcoding is usually sufficient to ensure that molecules from any particular cell receive a unique tandem.
Previously, this has been done in standard 96- or 384-well plates. However, the new sciATAC-seq protocols use Takara Bio's ICell8 instrument and 5,184-nanowell plates, which enable a 52-fold improvement in scale.
Optimizing the protocol for the Takara platform was not trivial, Adey said, but unlocked the ability to prepare 250,000 cells in a single run, with a total cost of reagents of about $2,000. The nanowell plates use the same amount of PCR reagents as the 96-well plate and is technically reusable, he said. Per cell, it costs between a quarter and a half a cent, he said, not including the cost of transposase.
The paper also referenced a third version of the scATAC-seq assay that is run using reagents from Scale Biosciences. The per-cell cost for this method is about $.05, providing lower numbers of cells, but "really high coverage," Adey said. In the paper, his lab reported calling about 70,000 unique accessible chromatin regions, on average. "That's sampling an extensive fraction of the accessible chromatin landscape," he said.
Improving their bisulfite conversion-based, single-cell methylation assay was more challenging, Adey said, as the multiple wash and buffer steps mean that techniques used to scale up other omics technologies haven't translated easily.
The main advance was developing two new amplification strategies. One uses linear amplification (sciMETv2.LA) while the other uses a modified "splint" ligation technology originally developed by researchers at the University of California, Santa Cruz to study ancient DNA. UCSC spinout Claret Bioscience is commercializing this splint ligation technology as SRSLY (single-reaction single-stranded DNA library prep, pronounced "seriously").
The methods also make use of fully methylated adapters, which unlock standard sequencing recipes and optimizations, including improved nucleosome disruption methods, to boost tagmentation efficiency.
The splint-based amplification was simpler, faster, and cheaper, making it "more appealing for large numbers of cells," the authors wrote in Nature Communications, although it came with less coverage per cell.
"In terms of scale, we can go as high as we would ever dare to sequence, because sequencing cost is the biggest burden," Adey said, noting that he presented a 20,000-cell dataset at the Advances in Genome Biology and Technology annual meeting earlier this month. The method calls for at least 1 million to 2 million raw reads per cell. "That ends up being pretty huge," he said. "We have no interest in doing a 100,000-cell prep."
The difference in coverage of CG dinucleotides in a methylome between the two flavors is "considerable," Adey said. However, the linear amplification method takes 10 hours and costs about $2,300 per plate, versus just over two hours and $230 for splint ligation.
In the paper, the authors said that they're working to further optimize the splint amplification method to close the gap in coverage. But even in its current state, the method hits the "sweet spot" of between 200,000 and 500,000 CG dinucleotides per cell.
"That's plenty to do a cell type assignment," Adey said.