NEW YORK (GenomeWeb) – Researchers at the Hubrecht Institute in the Netherlands have developed a new method for sequencing the genome and transcriptome of the same single cell.
The method, called DR-Seq for gDNA-mRNA sequencing and published in Nature Biotechnology today, employs quasilinear amplification and does not separate genomic DNA and mRNA. It could help scientists better understand the effects of genomic variants on gene expression in various cell types.
Over the last several years, different groups have developed methods to sequence the genome or transcriptome of single cells, which has proven important for studying heterogeneous cell populations, such as those in tumors.
However, so far, these single-cell approaches have not allowed researchers to directly study how changes in the genome — be it single-nucleotide variants or copy number variants — are correlated with changes in gene expression.
To overcome this limitation, researchers at the Hubrecht Institute in Utrecht, the Netherlands, developed DR-Seq, allowing them to sequence the genome and transcriptome of the same cell in parallel.
"We wanted to develop a single-pot strategy," where genomic DNA and mRNA are initially not separated, in order to prevent loss of material and contamination, explained Siddarth Dey, the lead author of the paper and a postdoctoral fellow in Alexander van Oudenaarden's group at the Hubrecht Institute.
The method starts by lysing a manually picked cell and reverse transcribing its messenger RNA into single-stranded cDNA, using a primer that contains a cell-specific barcode, a 5' Illumina adaptor, and a T7 promoter overhang.
Genomic DNA and single-stranded cDNA then undergo quasilinear whole-genome amplification for seven cycles, using a primer that has a defined 27-base sequence followed by eight random nucleotides at the 5' end. According to Dey, this amplification step builds on the MALBAC method that was published two years ago by Sunney Xie's group at Harvard.
After that, the scientists split the sample into two halves, using one to amplify the genomic DNA in 21 cycles of PCR, the other to convert single-stranded cDNA into double-stranded cDNA, which is then amplified into RNA using in vitro transcription.
The amplified genomic DNA is then converted into a cell-specific indexed Illumina sequencing library, whereas the RNA undergoes NGS RNA library preparation.
One problem with single-cell transcriptome sequencing methods is amplification bias, which distorts the original ratio of mRNA molecules present in the cell. Other methods have overcome this issue by using random barcodes in the reverse transcription primer, which uniquely tags all molecules that derive from the same original mRNA molecule.
However, that approach does not work in DR-Seq, Dey said, so they had to come up with an alternative way, using the annealing position of the primer during the quasilinear amplification step to identify unique mRNA molecules.
This priming location, which they called "length-based identifier," has "the potential to reduce amplification biases and technical noise to enable quantification of the original number of cDNA molecules," the authors wrote.
To assess the performance of DR-Seq, they sequenced the mRNA of 13 single cells from a mouse embryonic stem cell line and the genomic DNA of three of these cells and compared the results to existing single-cell genome and transcriptome sequencing methods.
For mRNA sequencing, they compared the results to those from cell expression by linear amplification and sequencing (CEL-Seq), performed on 33 single cells, and found that both methods detected similar numbers of genes and had about 9,700 genes in common. Across several metrics, such as dynamic range and sensitivity, the two methods performed similarly, Dey said.
They also analyzed genomic DNA sequencing, comparing their results to those from the MALBAC method, which they used on three single cells. Both methods showed greater coverage biases than bulk sequencing, they found, but performed similarly to each other.
Next, they applied DR-Seq to a breast cancer cell line, sequencing the mRNA of 21 cells and the genomic DNA of seven of these cells. They found that the average expression of genes was strongly correlated to the copy number of that genomic region.
In addition, they discovered that cell-to-cell variability in transcript number generally increased with reduced copy numbers and decreased with higher copy numbers. A possible explanation for this could be the fact that gene expression is believed to happen not continuously but in bursts, Dey explained. If a gene has several copies, this "bursting behavior" would average out, resulting in less variation of gene expression, he said, adding that they have not proven this hypothesis.
Dey and his colleagues are currently in the process of automating DR-Seq, using a liquid dispensing robot, in order to make it more high throughput and to reduce technical variability further. They are also working with an undisclosed company to increase automation. "Now, we can look at maybe 800 cells in a single day," he said, whereas "a year or two back, when I did this manually, I would maybe do 20 cells."
Being able to analyze hundreds of cells would help with practical applications, for example, to identify subpopulations within a tumor and to study drug resistance. Besides tumor analysis, the method could be applied to neurons, Dey said, which are known to carry somatic copy number variants, though it is unclear how these affect gene expression.
Another area of interest are polyploid cells, such as hepatocytes, he said, where it is unknown what effect the amplification of chromosomes has on gene expression.
Dey is also working on other integrated approaches for studying single cells, such as methods to analyze DNA methylation and transcription, or nucleosome position and transcription in the same cell. "This would allow you, potentially, to correlate epigenetic changes with gene expression," he said.