Do you have the sequences of the mRNas - meaning stop codon to poly-A tail?

Yes. It seems that three variants with three alternative poly-A-sites exist. They are about 150, 380, and 1200 bp long (from the stop codon to the poly-A-tail).

Then I don't understand the problem. You want to quantify the ratio between these three variants?

Yes. The problem is, that the short variant has the same sequence as the beginning of the long variant. There is no problem for example to construct primers which bind exclusively at the end of the long version to amplify the long version alone. But to bind primers within the short version means, that all three products will be amplified together (because of the sequence homology). Therefore, the question is how could I analyse the three fragments seperately, if possible (ideally by TaqMan technology).

In an RNAseq sense you can use the average coverage of the bases in the first 150 bases compared to coverage from 150-380, compared to coverage from 380-1200. However, there is likely bias if the cDNA was primed with polyT, since there will tend to be more copies of cDNA of the 3'end, or bias against the extreme 3'end if random priming is used. For TaqMan assay, you could use probes in a similar way, subtracting the signal of the 380-1200 and 150-380 probes from the 0-150 probe to estimate the shortest 3'UTR, then subtract the 380-1200 probe signal from the 150-380 to get the expression of the midsize splice form. Keep in mind these also will be potentially biased by the cDNA synthesis primer.

Can you make three sets of primers, which amplify DNA pieces of approximately the same length (80-120 bp should be enough for qPCR) and amplify only one of the three 3' UTR? If this is possible, you could use these primers in qPCR to measure the copy numbers (it should also be possible with the relative expression) and calculate the ratios between them from this data. Any thoughts on that?

Along the lines of what Christian said - for qPCR, including using Taqman probes, there is no reason you need to amplify the entire mRNA in the qPCR reaction. You can just have primer sets that are specific to each variant. This is easy if the variants are actually splice variants within the 3'UTR (design primers or Taqman probes that span the splice sites). You could also consider the assay in terms of ratios if you can't design variant specific primers or probes - you can simply do qPCR on the entire 3'UTR with 3 Taqman probes that recognized: (a) just the longest, (b) the longest and the middle length, (c) all three forms. If you report the result as ratios you'll still see changes. Lastly, if you really need an unbiased quantitative method you can do a northern blot with a probe that binds to all three. This will be both unbiased because there are no primer/probe/amplification steps and highly informative because you can measure the amount of each form directly.

Dear Oliver,

the best method in your case would be an old-fashioned Northernblot. Hybridize with a Probe that equally anneals to all your mRNA splice variants. What you see is what you get.

I'm no expert, but surely you just design a short polyA tract to your primers specific to your three different 3' ends. That's where the difference is between the three transcripts - where the polyA region starts in the UTR.

@Christian: Not a bad idea. Then you can have a transcript specific forward primer and a general reverse. If the amplified DNA piece is rather short (which is advisable for qPCR) then it shouln't even matter, if the shortest 3'UTR ist present in all three versions of the 3'UTR. It will not get amplified.

Thanks for your answers!

Yesterday evening I found a nice paper I want to show you.

In the end, the authors descibe what you have already mentioned above.

However, some of you may find the paper interesting (like me).

The paper deals with the identification and quantification of various 3'-UTR lengh variants of the human cellular transporter ABCG2.

It is:

To et al. (2008): "Regulation of ABCG2 expression at the 3' untranslated region of its mRNA through modulation of transcript stability and protein translation by a putative microRNA in the S1 colon cancer cell line" Molecular and Cellular Biology p. 5147-61

PMID:18573883

Please look at page 5150, chapter: "Identification of the 3' end of the human ABCG2 mRNA by 3' RACE assay."

I think I will try it with different primer sets (TaqMan probes) and calculate the rations between the amplicons, as mentioned by you.

So I can see which variant is "mostly" expressed in a given tissue or cell or if an expression change happens if the physiological status of the cell changes.

@Markus: Northernblot maybe a solution. However, the hardware in our laboratory is focussed on qPCR. So I think I will prefer this technique.

@Christian: Special thanks to you! I think you will like the paper cited above. If you like, plaese let me hear what you think.

Interesting paper. And a nice solution for the problem.

I think the most quantitative option would be an RNAse protection assay.

Oliver,

just a cautionary note, you may want to consider using a primer pair for amplifying all transcripts and then see how well this fits to the sum of the differentially spliced variants. Keep in mind that PCR values or principally on a log-scale but what you aim for is a quantitative difference on a linear scale (i.e 40:60 or 20:80 ratios or something like that). Depending on what question you are trying to answer the qPCR may not have a high enough resolution.

@Markus

Thank you very much for your suggestions!

The primer pair for amplifying all lenght polymorphisms must -in this case- include a reverse primer complementary to a sequence tail of the poly-T RT-Primer. The forward primer, of course, should be located at the beginning of the 3'-UTR. This is, what I have already done, but short amplicons are highly preferred with that method. Maybe it is better to use standard curves generated with DNA fragments or what ever.

To incorporate a log scale into a linear scale is mainly a mathematical problem.

My question is finally, if there are differences of the amount of variants between various cells or tissues, and if there is a shift in the case of environmental changes.

The RNase protection assay, suggested by Margarida, may serve as control-experiment.

@Margarida

This sounds very interesting. I read the paper of Ma et al. "RNase protection assay" (1996) PMID: 8954837 describing the method. It should work (denpending on the labeled probes) with length polymorphisms. However, you will get also only a ratio. But it may serve as a nice control-experiment in addition to qPCR.

(It is not so easy as it seems)

Oliver, sure its easy to transform the log-scale to a linear scale, thats not what i have meant. I am refering to the errors associated with your numerical values (e.g. Ct values), how they propagate and how your final result is affected by that. Often this is ignored but has hige impact on the actual results. Then you see differences but effectively they are all arteficial! Just remember this before you jump to any conclusions.

Cheers

Dear Markus,

I agree completely.

I think this is the "endless" discussion how "ideal" an "ideal PCR reaction" really is, if you have really a "perfect amplification rate" etc.

I have learned on the 2012 AACR congress, that about 70% of the scientific results in the field of oncology are NOT reproducible.

This is unbelievable, but seems to be the truth.

Do we overrate the techniques we use?

But this is a cmpletely new topic.