Apologies to Uma Gautam, but I would not particularly recommend GAPdH, beta actin and B2M reference genes in isolation as in response to different treatments, expression of the the first two especially tends to be 'up and down like a hoors knickers', as my resident Scottish QRT-PCR expert is wont to say.   Instead, it is probably better to make a selection from the possible reference genes included in the references recommended above, alongside GAPdH, beta actin and B2M and 18S RNA, which to all intents and purposes should be treated as another reference gene. Bear in mind also that one will probably need to test up to 12 reference genes known to be expressed in the cell type or tissue of interest to get two or three that are relatively invariant and match all the sensitivity criteria in response to treatment and also that different reference genes may be needed for different tissues or treatments of interest.  The reason that GAPdH and beta actin turn up so often is that they are expressed fairly ubiquitously, but this in itself does not make then good reference genes and one has to analyse things on a case by case basis.

I don't want to make it sound too complicated but at the end of the day one needs to try to conform to the MIQE Guidelines_Minimum Information for Publication of Quantitative Real-Time PCR Experiments:

Bustin SA et al., The MIQE guidelines: minimum information for publication of quantitative real-time PCR experiments. Clin Chem. 2009 Apr;55(4):611-22. doi: 10.1373/clinchem.2008.112797.

Bustin SA et al., The need for transparency and good practices in the qPCR literature. Nat Methods. 2013 Nov;10(11):1063-7. doi: 10.1038/nmeth.2697.

Not straightforward!  Option 1 is obviously preferable and likely to become the gold standard for publication, unless there is an exceptional situation such as yours.  Do you not have the option of checking reference genes in your control samples or another situation where your gene of interest is expressed before analysing your treatment samples?

In the absence of Option 1, you probably need to pursue all of Options 2, 3 & 4, to determine the best alternative, using normalisation to 18S RNA for Option 3.  There are some fairly reliable commercial 18S RNA primer pairs that give a linear response over a wide range of RNA concentrations.  I have always found normalisation to 18S RNA to be highly reliable because it represents and internal standard, so pipetting accuracy  etc., becomes less of an issue.  However, it suffers the qualification of any internal standard in that any treatment or disease that affects the proportion of ribosomal RNA to mRNA is going to affect the standardisation, though rRNA seems to remain pretty rock-solid in most situations.

I hope that helps but I will also watch for other, possibly more informed, responses with interest.

Best iwhes

Steve Morley

Whoops, I meant to add the following references in which I have found quite useful information for optimising the selection of reference genes:

Rae M, Grace C, Hogg K, Wilson LM, McHaffie SL, et al. (2013) The Pancreas Is Altered by In Utero Androgen Exposure: Implications for Clinical Conditions

Such as Polycystic Ovary Syndrome (PCOS). PLoS ONE 8(2): e56263. doi:10.1371/journal.pone.0056263

Lin P, Lan X, Chen F, Yang Y, Jin Y, et al. (2013) Reference Gene Selection for Real-Time Quantitative PCR Analysis of the Mouse Uterus in the Peri-

Implantation Period. PLoS ONE 8(4): e62462. doi:10.1371/journal.pone.0062462

Finally, We used 18S primer sets from ABI.

Best wishes

Steve Morley

Hi,

I wouldn't use starting input RNA for normalization. "Normalising a sample against total RNA has the drawback of not controlling for variation inherent in the reverse transcription or PCR reactions".  (http://www.nature.com/gene/journal/v6/n4/full/6364190a.html). The idea of absolute quantification suffers from the same problem: there would be no way to account for sample-specific issues (for example with RT), which is, for instance, what reference genes are best for. 

One idea is to spike in a known specific amount of a foreign RNA into each of your samples (a synthetic spike-in control oligonucleotide). You´d then do reverse transcription and qPCR on all the samples and you would use your foreign gene as a reference (for instance, see http://journals.plos.org/plosone/article?id=10.1371/journal.pone.0089237).

-A

referecne genes- gapdh, beta actin, B2M

Alejandro Montenegro-Montero makes a perfectly valid point about using synthetic spike-in oligonucleotide to control for the inherent variation of reverse transcription efficiency.  It is important also to carry out a minimum of three determinations per RT reaction and a minimum of three independent RT reactions. Of course, variation in RT efficiency is to some extent controlled for by using an 18S internal control (NOT total RNA).  18S rRNA is probably the best internal control in relative RT-PCR because it shows less variance in expression across a variety of treatment conditions than many internal reference genes. However, there is always the worry that because 18S rRNA is so abundant, rare messages are both reverse-transcribed and amplified at reduced efficiencies due to exhaustion of reagents in RT and particularly QRT-PCR reactions. Consequently, this approach is most suited to the relative delta CT method.  It is also essential to use a starting RNA concentration in which the target has been shown to amplify linearly, which means qualifying the assay with e.g. 4x or 10x serial dilutions over perhaps 5-6 orders of magnitude of an RNA known to express the target of interest and only taking experimental samples as valid when they correspond to the liner part of the dilution curve.

Using a synthetic spike-in control oligonucleotide is a neat solution and a bit like normalizing to another region of your GOI, the advantage  in both cases being that the level of control is much more comparable to the target.

The bottom line is that on one hand no method is perfectly reliable for every situation, so that one must decide on absolute vs relative quantitation, use of specific or universal primers, determination of primer pair efficiency, choice of multiple internal or external controls, demonstration that controls are unaffected by treatment conditions, etc, etc., and time must be invested in rigorous and robust assay design, commensurate with the availability of experimental materials. Finally, one must be aware of the limitations of the assay method chosen to avoid the temptation of over-interpreting the data.

Hope that helps!

Best iwhes

Steve Morley

Apologies to Uma Gautam, but I would not particularly recommend GAPdH, beta actin and B2M reference genes in isolation as in response to different treatments, expression of the the first two especially tends to be 'up and down like a hoors knickers', as my resident Scottish QRT-PCR expert is wont to say.   Instead, it is probably better to make a selection from the possible reference genes included in the references recommended above, alongside GAPdH, beta actin and B2M and 18S RNA, which to all intents and purposes should be treated as another reference gene. Bear in mind also that one will probably need to test up to 12 reference genes known to be expressed in the cell type or tissue of interest to get two or three that are relatively invariant and match all the sensitivity criteria in response to treatment and also that different reference genes may be needed for different tissues or treatments of interest.  The reason that GAPdH and beta actin turn up so often is that they are expressed fairly ubiquitously, but this in itself does not make then good reference genes and one has to analyse things on a case by case basis.

I don't want to make it sound too complicated but at the end of the day one needs to try to conform to the MIQE Guidelines_Minimum Information for Publication of Quantitative Real-Time PCR Experiments:

Bustin SA et al., The MIQE guidelines: minimum information for publication of quantitative real-time PCR experiments. Clin Chem. 2009 Apr;55(4):611-22. doi: 10.1373/clinchem.2008.112797.

Bustin SA et al., The need for transparency and good practices in the qPCR literature. Nat Methods. 2013 Nov;10(11):1063-7. doi: 10.1038/nmeth.2697.

No reference gene is universal. One has to test under the specific conditions of interest, whether a particular set of genes are suitable for use in RT-qPCR as reference genes (I usually recommend geNorm to test for that).

I agree with Steven D Morley regarding following MIQE guidelines. 

Regarding the problem of choose the right gene, I can suggest for normalization of gene expression experiment the approach based inter-run calibration. Basically it is a deltadelta-Ct method that take into account for better normalization multiple (potential) reference genes and their specific amplification efficiencies .

Here the link of the publication:

http://www.genomebiology.com/content/pdf/gb-2007-8-2-r19.pdf

Dear colleagues,

Sorry for the delay in my response, but I have been really busy and travelling these last weeks.

I really do appreciate your suggestions and comments. I think it's important to open a discussion on this issue, because the answer in particular experimental settings is not straightforward, and little standarization has been achieved to date.

As some of you comment, I have read MIQE guidelines several times before asking this question (among with many many other papers and book chapters). But I cannot find the answer to my question there.

Just in case I was not clear enough:

- I would never use reference genes reported in the literature for tissues, species and/or experimental conditions different than my own conditions.

- My choice would be to test 10-12 Ref Genes in a large set of samples and then run bestkeeper, genorm and others to select the most stable ones. This is NOT POSSIBLE, I have access to a small set of samples (3 at this point, although I could certainly obtain a some more, but not much). This is because I am studying a rare disease and there is a little number of patients per country/year. Nevertheless, I work with immortalized cell cultures, thus, the RNA amount for each sample is not a limitation.

- My objective is to calculate the % of exon skipped mRNA with respect to the non-skipped after antisense oligonucleotide treatment. Some people in the field (I'm don't think this is technically correct) don't even consider including a normalization strategy, because you are already testing both transcripts (skipped vs non-skipped). What do you think about this idea? And about the idea of using a different exon of the transcript of interest as normalizer?

- I agree on that including a synthetic RNA would be a good solution. This will probably be my choice after careful consideration.

- Regarding to the publication posted by Alessandro, I think I still have the problem of not been able to test many different samples. Nevertheless, I will have a closer look to it.

I will keep on researching on this issue, and let you know about my final choice.

Thank you all for your useful comments and suggestions!

Iker

Let us know what you decide and how it goes!

First of all, sorry for the delay in my update, but this last year has been crazy at work and I had to deliver plenty more of things prior to my Ref Genes approach.

I finally decided to purchase a prevalidated reference gene panel from tataa biocenter, test it with all my samples, and then evaluate the stability of each candidate gene across all the samples with and without treatment. This was performed through geNorm and Normfinder algorithms. Both identified 3 genes (TBP and HPRT1 for geNorm, and PPIA for Normfinder) as the best candidates. Although both algorithms showed different best candidates, the best ones identified by one of the algorithms were also rated as good by the other.

This work was presented in a conference in the begining of october, and I will add the full poster in researchgate as quick as I am allowed to.

More importantly, B2M and GAPDH, 2 genes that are commonly used for gene expression normalization, were identified as the least stable through both algorithms.

It must be noted that these conclusions only apply a priori to our particular experimental settig: myoblast cell cultures (2 controls and 3 patients) that were obtained from a muscle biopsy and then immortalized, and which are treated or not with an antisense oligonucleotide for specific exon skipping.

I also plan to test the stability of an assay to detect an upstream region in my transcript of interest (DMD), which is commonly used in splicing studies that more or less resemble our experimental setting: we want to evaluate the ratio between skipped and non skipped transcript after treatment with antisense oligonucleotides. I did that with 2 pairs of primers, but they were discarded due to low efficiency and a new primer design must be performed an tested prior to the stability testing.

I will come back and add updates on this regard.

Thank you very much for your contributions! I hope my experience can serve to other peers with the same issues.