This is a very interesting method, and your normalization problem is a difficult one. It is too bad the company did not think to add some foreign DNA to the samples during processing to at least help with some of the internal variability that arises as they put together the plates. I wonder if including some kind of spike-in control could be useful at this point, because at least you would have a standard of reference to normalize the well-to-well PCR variability. (For human samples I have used CAB and RCA plant genes as exogenous spike-in controls.) 

The major problem is that with siRNA treatment, especially affecting a transcription factor, you might expect there to be an increase in the relative amounts of unaffected mRNAs that results from a global decrease in the amount of mRNA being produced by that cell. Perhaps your best bet is to use a well-characterized gene, like GAPDH. As long as you are not looking at Sp1 or another transcription factor known to affect GAPDH, then it may be stable enough to be a suitable normalizer. Use the literature to select a few genes, such as GAPDH, that have a good chance of being stable in your system, and then do an experiment to select the best gene.

The way to select the best normalizing gene is based on stability, and there are a number of programs available to assist with identifying a stable gene. Please see research by Jo Vandesompele http://genomebiology.com/2009/10/6/R64 and CL Andersen http://cancerres.aacrjournals.org/content/64/15/5245.abstract?etoc. Please look for other references, as I am sure there are a number of more recent papers by these and other authors who have contributed so much to our ability to understand our gene expression results. I am sure you will find these approaches to be useful for helping you solve your problem. 

I hope that this helps you, and good luck with your project!

Thank you very much for the answer.  

The plates are expensive and I got few of them so there are not a lot of options for testing. I think I will spare one plate for assessing a gene like GAPDH. If the results obtained with GAPDH normalisation will differ a lot in comparison to vehicle treated control normalisation than I will look for the viarable in GAPDH Ct to exclude the possibility that GAPDH expression is regulated by a particular transcription factor. 

Could you please explain a little more the problem of global decrease in mRNA produced by the cell after siRNA treatment and how it would affect results ? You  stated that" there will be an increase in relative amounts of unaffected mRNA that results from a global decrease of mRNA"  It is not clear to me. 

I also thought that in the "treated wells" there might be less cDNA but the company assured that there are equal cDNA levels on the plate so this should not be a problem.

Regards,

Tomasz

What i am thinking of when I say  "there will be an increase in relative amounts of unaffected mRNA that results from a global decrease of mRNA" is essentially the opposite effect that is documented in lactation by this group:  Physiol Genomics. 2007 May 11;29(3):312-9. In lactation there is a large increase in highly expressed genes involved in milk protein synthesis. This causes an apparent decrease in the amount of various stable transcripts when the same amount of RNA is used for cDNA synthesis.

I am suggesting that in cells with transcription factor knockdown by siRNA treatment, there will be an impact on a large number of different mRNAs and that this could lead to a similar, but opposite effect where stable genes have an apparent increase. 

If a large enough percentage of the total mRNA pool is eliminated by knocking out the transcription factor, then the remaining mRNAs might be a larger portion of the cDNA pool, but these genes may not actually be increased or decreased on a per-cell basis. 

I hope this helps explain what I am thinking. It basically has to do with "per-cell copy number" versus 'fraction of 500 ng of RNA." You could think about all of these different siRNA conditions as different cell types producing a different amount of RNA per cell. This is why normalization to the no-siRNA control should not work, because all of these cell lines have different amounts of total RNA per cell. You need a stable mRNA that you expect to have the same copies per cell in order to normalize this effect.

I realize this is a very difficult issue that is not only confounded by the above problems, but is also made even more difficult because transcription factor genes themselves rely on other transcription factors for their expression. So how can you possibly find a "clean" control gene that is not affected by the knocked down  transcription factors?  

I have 2 more suggestions to help with this problem.

1) Use multiplex real-time PCR to test multiple normalizing genes in one reaction so that one plate is good enough for your control.

2) Improve your chances! Do several preliminary experiments. Find out what cell line is used to make the plates and determine if there are any genes that have already been shown to be relatively stable. (if not then get the original cell line and find stable genes yourself using geNORM or NormFinder types of programs.) You need more than one relatively stable gene.

This next part is hypothetical, maybe something like below can be done, and you should consider whether the suggestion below is actually appropriate. I am just throwing out an idea:

Maybe you can then use your knowledge that in untreated cells from this same line these genes are stable to SELECT the siRNA conditions for which THAT reference gene will work? What I am saying is that if you know a gene is stable in untreated cells, then you can use that gene to normalize only the siRNA treatment conditions for which its expression is still stable. You basically do the selection by eliminating the knockdown treatments that are outliers for a gene you expect to be stable, then perhaps you can use a couple of different genes to analyze different sets of siRNA knockdown experiments??

Hope this helps you! Good Luck!