Bisulphite sequencing question - (Jan/09/2007 )

Hi guys

Ya'll have been really helpful so far and I had a couple of other questions. I'm a little bit confused about the difference between sequencing with TA cloning, and direct bisulfite sequencing from a product amplified with GC rich tags (Spivack et al). My understanding is that the reasons for cloning are two fold:

1) To increase the GC content of the amplified product, which is typically AT rich after bisulfite conversion, by compensating through the use of an external plasmid, and then using standard T7/SP6 sequencing primers.
2) To reduce the level of background resulting from incomplete bisulfite conversion, which has high efficiency but isn't 100%, in a heterogenous cell population by cloning individual DNA molecules into vectors and examining them separately.

I can understand how using GC rich tags to amplify your sequencing product could help in the first of the two scenarios. But wouldn't the background typically found in direct sequencing reactions still be a problem? If thats the case wouldn't cloning ALWAYS be necessary in a sequencing reaction. Even in pyrosequencing?

cancergeek,

cloning will give you the methylation pattern of one allele/DNA molecule/cell. The cloned sequence would be derived from the one DNA template from the one cell so you are able to determine the methylation status across the region of interest in one cell. You sequence many to get a representative picture of the population of cells you are looking at. Of course this assay is prone to PCR and cloning bias so you may not get the representative sample, hence the more you sequence the better (the costs rise in proportion to this).

With direct sequencing you are removing the possibility of PCR/cloning bias by directly sequencing the entire population of templates/cells/DNA molecules so you would get a consensus methylation pattern. You can't determine the methylation pattern of one cell from the results.

As an example take a normally inprinted region. You have two alleles of which one is hypermethylated and the other hypomethylated by cloning and sequencing this will become very apparent when you have half the clones completely unmethylated by sequencing and the other half of the clones are completely methylated. From this you can say yes it is imprinted. You can't say the same thing from a direct sequence, even though at each CpG you can see there is 50:50 methylation and non-methylation, because you are looking at all possible DNA templates in your sample, you can't assume that you have imprinting and it could be that the CpG's are variably methylated across the region to give rise to a 50:50 ratio.

There are pro's and con's to both methods and it's weighing them up for your application and choosing the most appropriate one for yourself.

nick

hello,

I´m new, young and inexperienced. I have two more question about this topic: If I want to establish the relation between methylated CpG (of a promotor region of a gene) and the quantitative gene-expression, should I first clone the PCR-products or may I directly sequence the PCR-products? (PS: I´want also to know which CpG´s are Methylated in case of low gene-expression; the promotor-region contain 23 CpG´s of interest ).

Second question: all primers I use contain 2-3 CpG, is it necessary to amplify with both (in the same or in different reactions?): primer for the methylated version and primer for the unmethylated version?

Hi Tharom,

if you want to establish a direct link between amount of methylation and gene-expression, I would suggest trying to perform direct cycle sequencing of PCR amplicons after bisulfite treatment. If you decide to do so, then your primers, which apparently have been designed for MSP, are not the best choices. Consult Nick's pinned threat regarding PCR conditions and primer design for BSP. BSP primers should not contain CpGs and you need only one set of primers that amplify methylated and unmethylated DNA - as you want to have both in the sequencing reaction.

Good luck!

Krümel

Oh Oh - that sounds like a serious flaw in the paper ... the author seems to have used MSP and sequenzed the amplicons, which doesn't make sense (see below). For BSP your prumers should not differ between methylated and unmethylated CpG's, if the primer sequnece has to contain a CpG than you MUST use degenareted primers (Y for C/T and R for G/A). Otherwise, your pimers pick only those templates that are either methylated or unmethylated. By using primers that are not biasing one over the other, you get sequence information, that contains the amount of T and C for every CpG in the amplicon - I paste a picture which should make it clear. If you use U and M primers and sequence those amplicons, you have one sequence likely to show a lot of unmethylated C's and one with a lot of methylated CpGs and it will be hard to average those ... leading to highly unreliable results.
USing BSP + direct sequencing you can calculate the amount of methylation for every CpG by the formula:
%methylated C= peak height C / (peik height C + peak height T) * 100. There are some reasonable thougts, whether or not to use these raw values, but some papers show, that those can be accurate. You should, however validate them by using positive and negative controls.

Hope that made it clearer - if not, please post the sequence + primers you want to use and maybe tell us the paper they are from?

Much look,

Krümel

Well there are some papers testing which method is more accurate. Cloning and seqeuncing has the advantage that you do not need to measure and compare peak heights, on the other hand are you selecting your clones by random... To avoid bias, you need to sequence at least 10, better 20 and some say 50 clones! Direct cycle sequencing is not prone to this selection bias but has the disadvantage of the not-so-exact-measurement and like you can see in the example, you easily get a G or C background, due to the algorithms of the basecalling software... You may transform your direct cycle measurement by some form of scoring to account for these problems.
The papers I know on that topic show, that both methods are comparingly accurate and reliable.
Crucial thing is primer design, for both, and of course an accurate bisulfite modification. It makes sense to use mixtures of unmethylated and methylated DNA to test your measurements.
I prefer direct cycle sequencing, as it saves a lot of time
But figure out for yourself what you like more! If you have the possibility to perfom pyrosequencing (or your boss is willing to pay for it) I would give that a try - it's convenient and the results are probably the most reliable.

K.

Ok, thanks, good look in your work!