When someone slides into your DMs, it could mean that they are interested in you. Now, there’s a whole new meaning of “DM” for epigeneticists. Direct Methylation Sequencing (DM-seq) nondestructively maps 5mCs directly at single-base resolution with enzymes, and only needs nanograms of sample.
There are already lots of ways to get at 5mC, including:
- Old-school bisulfite sequencing (BS-Seq) that damages DNA, and can’t distinguish 5mC from 5-hydroxymethylcytosine (5hmC)
- OxBS-Seq that resolves 5mC and 5hmC, but requires a lot of sample
- TAPS that maps modified cytosines, but as we’ll see in the DM-Seq study by Rahul Kohli’s lab (U Penn), it has issues
Kohli’s lab has been in on the conversation for a while, developing the enzymatic strategy ACE-Seq a few years ago. That method leaves 5hmCs alone while deaminating unmodified Cs and 5mCs with the nondestructive DNA deaminase A3A. But it stays mum when it comes to exactly which cytosines are methylated. Taking inspiration from ACE-Seq, the Kohli team now spills the tea on their brand-new approach: DM-Seq. They reasoned that if they could protect unmodified Cs from deamination, they could use A3A again and this time, get at 5mC directly.
But as some might say, it was “complicated,” and the team’s plan of using a new CpG-specific carboxymethyltransferase called M.MpelN374K with an E. coli metabolite called carboxy-S-adenosyl-L-methionine (CxSAM) to protect unmodified Cs ran into trouble. After much sleuthing, the upshot was that these components work best on hemimethylated strands. So, here’s the 411 on the final DM-Seq workflow:
- Shear genomic DNA and add custom, deaminase-resistant adapters with 5-propynylcytosine (5pyC) instead of the traditional 5mC version
- Copying DNA with these adapters results in a methylated copy strand, and hence, hemimethylation
- Treat with M.MpelN374K and CxSAM to modify Cs to 5-carboxymethylcytosines and β-glucosyltransferase to glucosylate 5hmCs
- Deaminate with A3A, which turns 5mCs to Ts, then amplify with PCR and sequence
In a head-to-head comparison using unmethylated pUC19, T4-hmC, and methylated lambda DNA, DM-Seq and BS-Seq gave similar results. The researchers also compared those two methods with TAPS and TAPS-β, which, ICYMI, change 5mCs to dihydrouracil (DHU). They saw with their own data and by re-evaluating the TAP-developers’ data that TAPS significantly underestimates 5mC, and DHU is the source of this bias.
Finally, the team tested DM-Seq and BS-Seq on a human glioblastoma sample spiked with pUC19, T4, and lambda DNA. DM-Seq had a 2.8 times higher yield than BS-Seq, showing how much less destructive it can be. DM-Seq also outperformed BS-Seq by not confounding 5mC and 5hmC at clinically important sites.
For all the deets, direct yourself over to Nature Chemical Biology, June 2023.