NEW YORK (GenomeWeb) – Harvard Medical School researchers have unveiled a new approach that enables a genome-wide glimpse of DNA demethylation, as they reported in Nature Biotechnology this week.
DNA demethylation can occur through either a passive or active approach. The active pathway relies on the TET family of 5-methylcytosine(5mC)-modifying enzymes.
Harvard's Hao Wu and his colleagues added in a step to the methylase-assisted bisulfite sequencing, or MAB-seq, method to allow genome-wide mapping of 5-formylcytosine (5fC) and 5-carboxylcytosine (5ac) sites, products of successive oxidation of 5mC.
By reducing 5fC to 5-hydroxymethylcytosine (5hmC), Wu and his colleagues could distinguish between the two after bisulfite sequencing. Without this added step, unmodified cytosines cannot be told apart from either 5fC or 5aC, as they all appear as thymines after bisulfite conversion.
"Prior to our method, there's actually no report on whole-genome activity of demethylation in the mammalian genome," Wu told In Sequence.
Demethylation has increasingly been implicated in mammalian development and cancer, he and his colleagues noted.
MAB-seq draws on the bacterial DNA CpG methyltransferase M.SssI, which adds methyl groups to cytosines located within CpG dinucleotides, where the vast majority of 5hmCs are found. After M.SssI treatment, an unmodified cytosine within a CpG becomes a 5mC.
Then bisulfite conversion — which converts cytosines to uracils that are picked up as thymines during sequencing — of sequences treated with this methyltransferase leads to deaminated 5fC and 5aC. After sequencing, those deaminated 5fC and 5aC appear as thymines, while unmodified cytosines, which are converted to 5mCs by the methyltransferase, appear as cytosines after sequencing.
The Harvard team optimized this reaction to more than 99 percent efficiency, noting that complete conversion is needed to successfully detect 5fCs and 5aCs.
"If you really want to understand the complete picture of demethylation activity, you actually have to simultaneously measure the generation and the excision of the formyl-C and carboxyl-C in a single experiment," Wu added.
But 5fC is present in wild-type cells at very low levels. So, to boost its signal, the researchers combined MAB-seq with thymine glycosylase, or TDG, depletion.
TDG is involved in DNA demethylation, along with TET, and it mediates 5fC and 5aC excision. The researchers reasoned that its depletion should allow 5fC and 5aC to accumulate.
In mouse cells engineered to lack TDG, the researchers used MAB-seq to map at which CpG sites 5fC and 5aC were amassing to generate a picture of where DNA demethylation takes place.
From this, they reported that 5fC/5aC and 5hmC are linked with distinct CpGs within 5hmC/5fC/5aC-enriched regions.
This, they added, suggests that TET proteins typically stall at CpGs during the 5hmC step. But when there is better chromatin access at the CpG site, TET can oxidize it further to 5fC/5aC.
Still, this approach lumps 5fC and 5aC together. By adding in another step — sodium borohydride incubation — Wu and his colleagues could then tell them apart. With this added step, only 5aC is read as thymine after bisulfite sequencing.
"Through a clever reduction of 5fC to 5hmC, the authors could separate sequencing of 5caC from 5fC," the University of Chicago's Chuan He, who was not involved in the study, told IS via email. He has developed a modified Tet-assisted bisulfite sequencing approach, or TAB-seq, to identify 5hmC at single-base resolution.
Wu and his colleagues dubbed their modified approach caMAB-seq.
Adding MAB-seq to a 5aC mapping method like caMAB-seq also revealed, the researchers added, that 5fC and 5aC don't usually overlap at individual CpGs, further indicating TET processivity that's based on chromatin accessibility or other regulatory process.
Other integrated analyses — combining MAB-seq and BS-seq with Illumina deep sequencing — enabled the researchers to identify unmodified cytosines, 5mC/5hmC, and 5fC/5aC at more than 70 CpG sites within four 5fC/5aC enriched regions in mouse embryonic stem cells.
With this, they noted, they are able to get a peek into all the cytosine derivatives at the various stages of the TET/TDG-mediated active demethylation pathway.
Interestingly, an integrative analysis of BS-seq and MAB-seq results found that unmodified cytosine was significantly increased in response to TDG depletion — the opposite of what BS-seq alone found and of what others had previously reported.
"Based on this new method and improvement, we start to reveal some unexpected results," Wu said.
He added that he and his colleagues are planning to further explore this finding using mutant mice to generate in vivo active methylome maps.
An advantage of this approach over others for mapping 5fC and 5aC is that it doesn't rely on subtraction, Wu said. This way, he added, they can be measured in a single experiment.
In the future, he said that they want to use this approach to analyze demethylation in neurons and cancer cells.
"But the trick is you have to have genetically engineered cells," he said. "So therefore there is still some limitations with MAB-seq because on its own you cannot measure the formyl-C because they are dynamically being turning over, so you need to have a kinetic way to block this process."
University of Chicago's He added that combining an enrichment-based approach with a base-resolution approach like MAB-seq could offer the best strategy for genome-wide investigations of 5fC and 5aC. Though, he noted, that "this work really provided an easy and alternative method to map 5fC and 5caC in genomic DNA."