Protocol Online logo
Top : New Forum Archives (2009-): : DNA Methylation and Epigenetics

Designing a simple assay for global DNA methylation - (May/20/2015 )

Hi all,


I would like to design a very simple, idiot-proof protocol for measuring the relative amount of methylation in the genome as a whole (working with non-model organisms). That is, I would like to be able to show that the genome of organism A has about 3% cytosine methylation and that of organism B has 3.5% (ideally with high accuracy, as my organisms are insects and thus have less methylation than mice/humans). I would also settle for having methylation in arbitrary units relative to some sort of reference. I don't care where the methylation is in the genome (yet) - I only want to know the genome-wide rate of methylation.


Here is my current plan - is there anything wrong with it? I mostly don't do molecular work, so it might be way off:


1. Extract the DNA and sonicate it into smallish pieces (150bp?).

2. Denature the DNA into single-stranded DNA at 95oC - this is needed because the antibody only binds ssDNA.

3. Add primary antibody that binds to 5-meC, and incubate for 2 hours at 4oC.
4. Add secondary antibody, and incubate - this will stick to the primary-labelled DNA, and presumably also to the unbound antibody in the supernatent.
5. Add AMPure beads - all the DNA sticks to them, both antibody-labelled DNA and the non-labelled DNA pieces.
6. Discard supernatent and wash the beads a few times. This removes the excess primary and secondary antibody.
7. Remove the DNA from the beads with elution buffer, and measure the amount of (methylated) DNA using a plate reader. Then, measure the total amount of DNA using Qubit ssDNA kit.

8. Take the ratio of the two readings in step 7 to get a number expressing the relative amount of genome-wide DNA methylation.

9. Repeat steps 1-8 with some reference samples which contain 100% methylated DNA and 100% non-methylated DNA, mixed in various ratios like 100:1, 50:1, 10:1 etc. These will allow the values in step 8 to be converted into estimates of the % methylation rate.


Any tips are much appreciated!




The "classic" method of determining total cytosine methylation is by HPLC. I haven't needed it myself, but it would be the method I'd go for.


Thanks - I have heard of that method but HPLC doesn't seem idiot proof! Requires specialised equipment and training that's not very common, as far as I understand. Any thoughts on my method?


It might work, yes. AMPure XP binds ssDNA, some variability in fragment size shouldn't matter too much, and the reference samples should account for differences in antibody binding (although you might lose some precision here). I don't have any specifics on how to properly label DNA with anti-methylcytosine antibodies (denaturation/renaturation, binding time), but I'm sure you checked it out; so, I don't see any obvious flaws.


I too would opt for degradation of the DNA to single nucleotides, followed by HPLC. But your scheme may well work. I'd worry about keeping your DNA single stranded, when you cool it. Another option would be PacBio sequencing of some random fragments. This wouldn't give you complete coverage, but would sample the space pretty efficiently.


the variation you get from an ELISA-like experiment, which there are kits for DNA methylation are quite wide ranging. I am not too sure you would see enough precision to determine 3% from 3.5%


HPLC can suffer the same thing, remember 0.5% difference in methylation is actually a hell of a lot of CpG's methylated in your system.


It is not as idiot proof but may improve on your precision, which is to use methyl-light or something along those lines to measure global repeat DNA methylation (eg: LINEs and SINEs) although they also have their own caveats.




Thanks for the input all. I am starting to think that non-sequencing methods won't give enough accuracy for my purposes (the above poster is right - 0.5% difference in methylation is a lot, and I expect the effect to be smaller than that), so I am looking into some BiS-seq (which does limit me to species with half-decent genomes available, I think).

All the best!