DNA methylation analysis
The common first half of almost every methylation assay: denature the DNA, deaminate unmethylated cytosine to uracil with sodium bisulphite, desulphonate and clean up, then amplify the region of interest with primers designed against the converted sequence, after which any sequence-reading method can report the methylation as a C/T ratio.
Our ruleSimulations prepare you for the bench but don't replace hands-on practice.
Bisulfite conversion + PCR
The common first half of almost every methylation assay: denature the DNA, deaminate unmethylated cytosine to uracil with sodium bisulphite, desulphonate and clean up, then amplify the region of interest with primers designed against the converted sequence, after which any sequence-reading method can report the methylation as a C/T ratio.
How it differs from the standard method
- Denature before adding the reagent. Bisulphite only reaches cytosine in single-stranded DNA, so a conversion started on native duplex DNA is patchy and leaves unconverted cytosines scattered through the molecule.
- The alkaline desulphonation step is not optional cleanup: until the sulphonate adduct is removed the polymerase cannot read through, so skipping it gives converted DNA that will not amplify.
- Expect to lose most of the input. Plan on 10–20% recovery from kit conditions and scale the input to the molecule count the downstream measurement needs, rather than to a mass.
- Design primers against the converted sequence of one chosen strand, off every CpG, and over several non-CpG cytosines so that unconverted genomic DNA is disfavoured.
- Converted DNA is single-stranded, partly abasic and already fragmented, so it keeps badly. Amplify the same day where possible and store the rest frozen.
- Methylation-specific PCR is the simplest readout of this step: two primer pairs, one matching the methylated and one the unmethylated converted sequence, read as presence or absence of each band. Stop at 30–35 cycles: the mismatched pair is slow, not blocked.
- Turning a chemical mark into a sequence difference
- Conversion has to be complete, and a control has to prove it
- The converted genome is a harder genome to work in
- Four readouts, four different questions
- Affinity methods trade resolution for simplicity
- What bisulphite cannot tell you
- Key terms
- When to use it
- Sub-techniques
1Turning a chemical mark into a sequence difference
5-methylcytosine pairs with guanine exactly as cytosine does, so no polymerase, probe or sequencer can see it directly. Every mainstream methylation assay therefore starts by converting the mark into something a sequence reader can detect.
Converted to U, read as T
Bisulphite deaminates it. The base is gone from the sequence, and the PCR product carries a T where the genome had a C.
Resists, still read as C
The 5-methyl group blocks bisulphite addition, so the base survives the reaction and is copied as C.
C / (C + T) at each CpG
A proportion of molecules, not a property of one molecule. Reported as % methylation or as a beta value.
2Conversion has to be complete, and a control has to prove it
An unconverted cytosine is indistinguishable from a methylated one: both read as C. Incomplete conversion therefore does not add noise, it adds methylation, and always in the same direction.
- Incomplete conversion
- Unconverted cytosines read as methylation. The error is one-directional and locus-independent: it raises every value, and it raises low-methylation samples proportionally most. Caught by a conversion control, not by replicates.
- Over-conversion
- Not a chemical over-shoot but DNA damage. Depurination and strand breakage destroy template, so yields fall, PCRs fail, and the percentages that do come out rest on too few molecules to be stable.
- Molecule count
- The number of converted molecules that reach the amplification. It, not the read depth or the number of technical replicates, sets the floor on how precisely a percentage can be measured.
- Conversion control
- A position that must read as T if conversion worked: a non-CpG cytosine inside the read, or a fully unmethylated DNA standard put through the same conversion.
3The converted genome is a harder genome to work in
Conversion does two things to the sequence that make the molecular biology awkward. First, it strips most of the cytosines out of one strand, so the template is effectively three-letter and its complexity collapses: stretches that were unique in the genome can become near-identical after conversion, and primers have far fewer distinguishing bases to work with.
4Four readouts, four different questions
- Methylation-specific PCR
- Two primer pairs on the converted sequence, one matching the methylated version and one the unmethylated version. Cheap, extremely sensitive, it can see roughly one methylated molecule in a thousand, and strictly qualitative. Band intensity is not a quantity, and the mismatched pair is slow rather than blocked, so extra cycles manufacture methylation.
- Pyrosequencing
- Sequence a short stretch of the converted amplicon by synthesis and read the C-to-T ratio at each CpG from relative peak heights. Gives a quantitative percentage at a handful of CpGs, to roughly one percentage point, with a conversion control inside the read. Limited to about 100 bases per assay.
- Methylation arrays
- Hundreds of thousands of designed CpGs typed at once by hybridising converted, amplified DNA to bead-bound probes, with a methylated and an unmethylated signal per site. Reports a beta value per CpG per sample. Genome-wide in coverage but only at designed positions, and the two probe chemistries on the same chip need normalising before samples can be compared.
- MeDIP and affinity capture
- No conversion at all: shear the genomic DNA and immunoprecipitate the fragments carrying 5-methylcytosine with an antibody, or capture them with a methyl-CpG-binding domain protein. Gives regional enrichment at the resolution of the fragment length, and is biased by CpG density, so it under-reports methylation wherever CpGs are sparse.
5Affinity methods trade resolution for simplicity
MeDIP sidesteps the whole conversion problem, and pays for it in resolution and in linearity. An antibody against 5-methylcytosine binds a fragment, not a base, so the answer is 'this 300 bp window was methylated', never 'this cytosine was'.
6What bisulphite cannot tell you
- Bisulphite does not distinguish 5-methylcytosine from 5-hydroxymethylcytosine: both resist conversion and both read as C. In tissues where 5hmC is abundant, brain especially, a bisulphite percentage is the sum of the two. Oxidative and chemical variants of the chemistry, and enzymatic conversion, were developed to separate them.
- A percentage says nothing about which molecules carried which pattern. Two samples can share an average and differ completely in whether methylation is uniform or confined to a subpopulation of cells; only reads that cover several CpGs at once show that.
- Nothing in the assay tells you which strand or which allele a molecule came from unless the assay was designed to: allele-specific methylation needs a linked variant in the amplicon, and strand-specific methylation needs separate assays for the two strands.
- Methylation is a property of the cells you put in. A tumour biopsy is a mixture, and an apparent 40% can be 80% methylation in 50% tumour cells. Purity and cell composition belong in the interpretation, not only in the methods.
Key terms
Practise these as flashcards →Cytosine methylated at the 5 position, in vertebrates almost entirely at CpG dinucleotides. It base-pairs normally, which is why it has to be converted chemically or captured by affinity before it can be read.
Treatment with sodium bisulphite that deaminates unmethylated cytosine to uracil while leaving 5-methylcytosine intact, turning methylation into a C/T sequence difference.
The fraction of unmethylated cytosines that were actually converted, measured at positions known to be unmethylated. Below about 99% the residual cytosine is read as methylation on every sample in the run.
A few hundred base pairs to a couple of kilobases of unusually CpG-dense, GC-rich sequence, often at a promoter. Island promoters are usually unmethylated in normal tissue, which is why their methylation is informative in cancer.
The array measure of methylation at a CpG: methylated intensity divided by the sum of methylated and unmethylated intensities, plus a small offset. It runs from 0 to 1 and is compressed at both ends, which is why differential testing is often done on the logit-transformed M value instead.
Two PCRs per sample on converted DNA, with primers that match either the methylated or the unmethylated converted sequence at several CpGs. Reports presence or absence, not quantity.
A cytosine outside a CpG dinucleotide. Essentially unmethylated in vertebrate somatic tissue, so it serves as the built-in conversion control inside a read.
An oxidation product of 5mC and an intermediate in demethylation. It also resists bisulphite, so standard bisulphite methods report 5mC and 5hmC together.
Enzymatically methylated and whole-genome-amplified DNA, usually also mixed to defined intermediate ratios, taken through the conversion alongside the samples to show that conversion and amplification both behaved.
When to use it
- Testing whether a specific promoter CpG island is methylated in a set of tumours and matched normal tissue, where a yes/no answer at one locus is enough: methylation-specific PCR on converted DNA is the cheapest route and the most sensitive to a small methylated fraction.
- Quantifying methylation at a few known CpGs across many samples (validating a biomarker, measuring a dose response, confirming an array hit) where pyrosequencing gives a per-CpG percentage with a conversion control in every read.
- Screening genome-wide for differentially methylated positions across a cohort, where an array types hundreds of thousands of designed CpGs per sample at a cost that makes a few hundred samples feasible.
- Mapping methylated regions in a genome with no array available, or where only regional information is needed, where an affinity method such as MeDIP gives enrichment maps from a few micrograms of DNA without any conversion chemistry.
- Checking whether an imprinted locus or a repeat family has kept its methylation after a long culture, an editing experiment or a reprogramming step: a question that a single well-chosen quantitative assay answers better than a genome-wide screen.
Compared with alternatives
Prefer it when you need per-base methylation genome-wide, phased patterns along single reads, or coverage of CpGs no array interrogates. Whole-genome bisulphite sequencing is the reference method; reduced-representation and targeted capture versions trade coverage for cost. The chemistry and all its controls are the same: it is the readout that changes.
Prefer it when the question is which protein is on the DNA rather than which cytosine is methylated. ChIP-based mapping shares MeDIP's affinity logic, its input and mock controls, and its fragment-length resolution limit, so the two are run and analysed in much the same way.
Prefer direct Sanger sequencing of a cloned bisulphite amplicon when you need the methylation pattern along individual molecules at a single locus: each clone is one original DNA strand, so you see whether methylation is patchy or all-or-none, which an averaged percentage hides.
Prefer it for detecting a rare methylated allele in a large excess of unmethylated DNA, such as a circulating tumour marker in plasma. Partitioning converts the measurement into counting positive partitions, which is far more robust at very low methylated fractions than reading a ratio off a bulk amplicon.
Prefer a methylation-specific qPCR when you want a semi-quantitative answer at one locus across many samples with no gel step; it is the natural upgrade from endpoint MSP, and the MeDIP readout is itself a qPCR against an input fraction.
Sub-techniques
Choosing one here switches the whole page, including the protocol and the lab, to it.
The common first half of almost every methylation assay: denature the DNA, deaminate unmethylated cytosine to uracil with sodium bisulphite, desulphonate and clean up, then amplify the region of interest with primers designed against the converted sequence, after which any sequence-reading method can report the methylation as a C/T ratio.
- Denature before adding the reagent. Bisulphite only reaches cytosine in single-stranded DNA, so a conversion started on native duplex DNA is patchy and leaves unconverted cytosines scattered through the molecule.
- The alkaline desulphonation step is not optional cleanup: until the sulphonate adduct is removed the polymerase cannot read through, so skipping it gives converted DNA that will not amplify.
- Expect to lose most of the input. Plan on 10–20% recovery from kit conditions and scale the input to the molecule count the downstream measurement needs, rather than to a mass.
- Design primers against the converted sequence of one chosen strand, off every CpG, and over several non-CpG cytosines so that unconverted genomic DNA is disfavoured.
- Converted DNA is single-stranded, partly abasic and already fragmented, so it keeps badly. Amplify the same day where possible and store the rest frozen.
- Methylation-specific PCR is the simplest readout of this step: two primer pairs, one matching the methylated and one the unmethylated converted sequence, read as presence or absence of each band. Stop at 30–35 cycles: the mismatched pair is slow, not blocked.