SAN FRANCISCO (GenomeWeb) – Researchers from Oregon Health & Science University have modified a method for sequencing the genomes of thousands of single cells to profile the methylomes of single cells.
The method could have implications for understanding the epigenetics of cell development, particularly as it relates to neurodevelopment and cancer, according to senior author Andrew Adey.
Methylation is "a really important mark in development, particularly neurodevelopment," Adey said, so having methods that are both high-throughput and cost-effective would be important for future research into these areas.
Adey, an assistant professor at OHSU, along with researchers at the University of Washington and Illumina, described the method this week in Nature Biotechnology. The OHSU and UW researchers have been collaborating with Illumina over the last several years on single-cell combinatorial indexing methods. Previously, they developed methods for single-cell whole-genome sequencing, Hi-C sequencing, RNA sequencing, and ATAC sequencing that all use combinatorial indexing. The researchers also hold several patents on the methods.
Adey said that the technique has commercial potential, but added that he is not involved in that process. Illumina declined to comment on whether it plans to commercialize the method.
The goal of the collaboration has been to develop methods to sequence thousands of cells cost effectively without requiring ancillary equipment. The single-cell methylation technique described in the latest study, costs just $.50 per cell, Adey said. The library prep does not include the cost of sequencing, he said, "which is still an expensive portion, but library prep costs are dramatically reduced."
The method itself is "a natural extension of the whole-genome approach," Adey said. "But there are substantial changes," mainly related to the adaptor strategy.
For instance, in the standard whole-genome approach, combinatorial indexing is achieved through a two-step process of transposase-based adaptor incorporation. Nuclei are distributed in a 96-well plate and each well contains a unique index. Then they are pooled, diluted, and redistributed in plate where each well has a different set of adapters, such that there is a very low probability that any two nuclei have the same set of barcodes.
For whole-genome bisulfite sequencing, the first set of barcodes is added prior to bisulfite conversion, while the second set of adaptors is added after bisulfite conversion using a linear amplification approach. Transposase is still used to add the first set of adaptors, but the adaptors themselves had to be redesigned, Adey said. The researchers designed adaptors that did not contain cytosine so that bisulfite conversion would not convert them to uracil. The second adaptor is incorporated after the nuclei are pooled, redistributed, and bisulfite converted using a random primer extension strategy like in single-cell whole-genome bisulfite sequencing protocols.
In the study, Adey's team first demonstrated the technique on a B-lymphoblast cell line, generating libraries for just over 700 single cells. Sequencing resulted in more than 55,000 unique reads per cell, which Adey said was actually a higher number of reads per cell than what they had achieved with the standard whole-genome approach.
"The changes we made to work with methylation sequencing also increased our read counts," he said. Aside from making small optimizations to the protocol, Adey said that the new strategy — adding a single adaptor for the combinatorial indexing step and then doing a linear amplification — "is more amenable to capturing more molecules."
As in the whole-genome combinatorial indexing strategy, the team tested two different approaches to remove nucleosomes — lithium-assisted nucleosome depletion (LAND) and crosslinking with SDS (xSDS). They found that the xSDS approach yielded fewer nuclei that had the same transposase barcode within the same PCR well, so they moved forward with that approach.
The researchers first validated that the method enabled them to pick out mouse and human cells from artificial mixtures of cell types. Next, they wanted to see whether it could discriminate between subpopulations of the same cell type. To test this, they sequenced the methylomes of 641 cells from mouse brain tissue. They found that the cells clustered depending on whether they were excitatory or inhibitory neuronal cells, and they also identified a group of cells that appeared to be non-neuronal cells.
Going forward, Adey said the team planned to continue to make improvements to the method and start applying it to study neurodevelopment. While the immediate applications of the method would be in basic science and clinical research, he said that ultimately the knowledge gleaned from that work could lead to insight on therapeutic options.
Aside from neurology, Adey said that the method could have applications in cancer, for studying clonal populations of cells, for instance. "Clonal populations of cancer cells might be different based on methylation state and it would be important to understand what's driving that state difference," he said.
Adey also noted that there are "aspects to the combinatorial indexing strategy that will enable sampling of different properties simultaneously in the same single cell," he said.