Heavy BSP bias towards methylated DNA - Why??? (Aug/27/2006 )
I have been doing BSP on a number of imprinted genes and I'm getting a heavy bias (almost exclusive) towards methylated DNA following sequencing of clones.
For one DMR I'm investigating, the last two controls I sequenced showed 20/20 methylated strands and 22/24 methylated strands respectively (amplicon was 471bp and contained 30 CpGs). Two other DMRs I'm investigating gave similar results.
My primers contain no CpGs, my bisulphite conversion appears complete (98-100% conversion of non-CpG C to T), I'm ligating into pGEM T Easy and transforming with DH5A. The DMRs I'm looking at have been well investigated by others and should be showing both methylated and unmethylated strands.
Why am I having so much trouble cloning unmethylated strands? I keep reading that bias usually favours unmethylated over methylated DNA. Has anyone else had this experience??? Any thoughts much appreciated. 
Hi again
I'm moving this query up the list again in hope that someone (anybody!) has a suggestion or comment. Any advice or questions will be welcomed! 
I agree that BSP favors for unmethylated template. You said your controls showed almost 100% methylation, what are those control? What % of methylation you expect to see? Are you sure your DMRs are not 100% methylated? if Yes, how do you know?
Try direct sequencing your PCR product to see if you can get complete methylation. I did some comparisons between cloned and direct sequencing and found they are pretty much consistent. I use the TOPO TA cloning system from Invitrogen and have not noticed such bias.
Many thanks for your comments pcrman.
My controls consist of DNA from normal individuals without imprinting defects. I'm looking at well documented CpG islands within imprinted genes that are known to be differentially methylated (where one parental chromosome is methylated and the other is unmethylated).
I've done a TaqI (TCGA) restriction digest on my BSP products and I'm getting distinct bands suggesting there is both methylated and unmethylated DNA present (TaqI won't cut unmethylated TCGA sites as they are converted to TTGA). I've also used a patient with UPD (totally methylated DNA) and I don't get an unmethylated (uncut) band following TaqI digestion suggesting the enzyme is working as it should.
Although I don't have parental blood samples I'm keeping an eye out for informative SNPs and I have evidence that the methylated DNA from several samples contains the same SNPs suggesting I'm only seeing product from one parental c'some. I think the problem is with the ligation or the cloning. Somehow there is selection against the unmethylated molecules. I have experienced this with 3 different DMRs but I don't know why??? 
For each methylated clone, how many out of the 30 CpGs on the insert are methylated?
During ligation and transformation, the only difference between methylated and unmethylated alleles is GC content which is unlikely to affect ligation reaction. Whether long strings of "TA" in unmethylated molecules somehow makes the vector behave differently? For example, ATAAA or poly As can serve as transcription terminator.
Another member reported having problem sequencing cloned bisulfite product in pGEM vector here. I never have problem using Topo TA vector. Maybe you should switch to that.
Thanks again pcrman
I have attached a couple of jpegs - almost all the CpGs are methylated. In control2 there were two unmethylated sequences (miracle!). Both unmethylated sequences contain a documented G>T SNP which is not present in the methylated strands. I gather these two strands are from one parent and all the methylated strands are from the other.
I think I might take your advice and try another vector as I've been bagging my head against a wall for months.
.