False positives? MSP and BGS - (Jan/29/2007 )

Dear All

Could anyone tell me how they verify false positives from MSP? And are false positives due to false priming from doing too many cycles, or can you get false positives by some other means, in the setting of full bisulfite conversion?

I've been verifying my MSP bands by performing BGS on the MSP bands (ie cloning first then sequencing, as I can't seem to get direct sequencing to work in my hands). I'm trying to prove a theory to my bosses that when I do high number of cycles the methylated bands I see are false positives. However my results from BGS are confusing me:

Sample 1: Unmethylated band - all the CpGs are unmethylated. Methylated band - all the CpGs are methylated
Sample 2: Unmethylated band - all the CpGs are unmethylated. Methylated band - all the CpGs are covered by the MSP primers are methylated, but the remaining CpGs are unmethylated.
Sample 3 (methylated control): Unmethylated band - all the CpGs covered by the MSP primers are unmethylated, but the remaining CpGs are methylated. Methylated band - all the CpGs are methylated.

Does this make any sense to anyone? N.B All the non-CpG C's have been converted to T's (ie full conversion).
If I assume that my methylated control contains only methylated molecules then if the band with unmethylated primers is a false positive, would I not expect on BGS all the CpGs to be methylated here, rather than just the CpGs between those covered by my primers? I buy in the methylated control from Chemicon.
And vice versa for sample 2?....

I am very confused. Any help greatly appreciated!!

Thank you all in advance

---

Hi methstudent,

sounds confusing. I even did not understand what you were doing... sorry. Just to get it clear: Did you clone and sequence your MSP amplicons or did you use different primer pairs designed for bisulfite sequencing?
Where there any differences between the three samples or are they treated all exactly the same way?
How many cycles are a high number of cycles?

Sorry, no help from me at the moment

Krümel

---

Thanks Krumel
What I have been doing is testing different cycles for nested MSP: (Stage I uses unbiased primers and stage 2 uses U and M primers)
e.g.
1) 40 cycles stage 1, 30 cycles stage 2
2) 30 cycles stage 1, 30 cycles satge 2
3) 30 cycles stage 1, 25 cycles stage 2
4) 25 cycles stage 1, 30 cycles stage 2....etc

I then run a gel after my stage 2 PCR, and gel band extract my MSP bands. I then clone each of these cleaned up products (using pGem), select 10 clones and sequence.

With the samples I refer to in the first posting, these are all following 40 cycles stage 1 and 30 cycles stage 2. I personally think too many cycles (but this is what people say they have used in their papers?!) hence I wanted to confirm the presence of false positives.

hope this is clearer.....I am still confused!

---

Ok methstudent, that made it clearer.
Some thoughts:
- Your samples are all fine - MSP can only test those CpGs that are in the primer itself. The others are not tested and hence have no influence on the MSP results. Your "strange" results are a little funny, as you should also see something in between no and full methylation outside the primers...but don't have to.
- However, you only prove what you already knew before sequencing - your M primers amplify methylated and your U primers amplify unmethylated CpG in the sequence your primer binds to.
- I think you want to know, if these findings are true - if samples showing methylation after 70 cycles really are methylated???
- to test, if your MSP results are right or wrong, you have to perform a "real" bisulfite sequencing! This gives you the unbiased truth about all the CpG's and especially those your primers bind to! You can do this by cloning and sequencing, also.

Or, to make a long story short: You can't validate your MSP findings by sequencing your MSP products.

Or maybe, I have not properly understood your problem?

BSP works with direct cycle sequencing and will work in your hands - just stick to Nick's pinned notices - everything you need is in there!

Hope that helps

Krümel

---

Ah, and btw. for BSP you will have to design primers that amplify the region including your MSP primer binding sites...
K.

---

Thanks again!....that puts it all into perspective again

Could I ask a dumb question also.....I am trying to detect the presence of methylation in patient samples (eg exfoliated cells), so the number of abnormal cells here is actually very few amongst many "normal" cells. Hence I am using nested MSP. Now does that mean if I do high number of cycles (at a high enough annealing temperature to prevent mispriming) that any methylation I detect is true?....I'm struggling to get my head around this concept because of all the talk of false positives (eg amplifying non-specific product etc.)...and I have been through the whole of the forum in desperation!!!! i

e will increasing the annealing temp be enough to give me confidence in my MSP result?

I will be very grateful for any help with this.

---

I second krümelmonster but to add, other means of false positives is having a mixed population of cells in your initial sample.

For instance you are looking at a tumour biopsy sample there could be normal cell contamination there and this is where the false positives are arising.

Indeed, too many cycles in MSP can lead to your results.

As for the sequencing as Krumel said, it's not the best way to confirm your MSP results by sequencing the amplicons because your primers will be biased toward one template over the other. To confirm you need to pick BSP primers that don't bias towards methylation or non-methylation.

Good luck!!

Nick

---

Thanks Methylnick!
Could I just check that I understand your point....

...Are you trying to say that if I have a patient sample with a mix of tumour cells and normal cells, if there is NO methylation in the tumour cells but some methylation in the normal cells, then this will give me a false positive, because the tumour cells are actually not methylated? sorry to be a pain but the issue of false positives is driving me mad at the moment!

Thanks again in anticipation!

---

No, the other way round...
Given that you have a lot of unmethylated tumor cells and only a hand full methylated "normal" cells. With large cycle numbers you amplify also the methylation signal from the "normal cells" - leading to a false positive M-Band (!) So you'll see some methylation in your sample, even though the cells you are interested in are not methylated.
False positive and false negative can be confusing - you have to know exactly what a true positive should be - i.e. no methylation signal or methylation signal - that depends on your question!
In your case it is more difficult, as you have a lot of "normal" and only some "abnormal" cells. Large cycle numbers or increased annealing temperatures won't help in that case, as you can never be sure that you look only at the abnormal cells. I would look for a way to seperate normal and abnormal cells first and compare the methylation afterwards. But I don't know if this can be done?
Another way could be comparing patients vs. controls. Even if you only have a small percentage of abnormal cells, they could change methylation that you observe when compared to controls. But I think that it would make more sense to use direct cycle sequencing of bisulfite treated samples (BSP) - you may find a small but nevertheless significant difference in the promoter methylation of your gene. However, you cannot state by this way, that patients have more abnormal cells with an increase/decrease in DNA methylation. You can only say that there is a difference and postulate, that the abnormal cells are contributing to it...
oops...this all sounds even more confusing ...

Krümel

---

....thanks so much!!! I understand it now...although like you say there is no simple way to answer the question. Unless I LCM samples I can never be sure the methylation is definitely from the abnormal cells as opposed to the normal cells in a mixed population!....

---