First, it is better to use the term "Endogenous Control" than "Housekeeping".

Second, as stated several times before, there is no "ideal" or "best" EC for ALL experimental settings. All genes vary to some extent, so a good normalizing gene for me might not be the best for you. So validation is a must.

I strongly recommend reading the MIQE guidelines:

http://www.ncbi.nlm.nih.gov/pubmed/19246619

www.clinchem.org/content/55/4/611.full.pdf

There is also a good overview Nature Protocol by Nolan et al.:

http://www.nature.com/nprot/journal/v1/n3/abs/nprot.2006.236.html

The Gene Quantification website is full of very high quality information regarding quantification PCR, including protocols and also theoretical considerations on the models for PCR data analysis.

http://www.gene-quantification.info/

http://www.gene-quantification.de/hkg.html#refgenes

Our workflow included a list of 8 genes reputed "stable" from the litterature in our system. We systematically included the 8, and retained the 3 most stable for normalization in our analyses. The advantage of this approach was, as we actually found, that in particular subanalyses we could normalize expression with a group of EC better suited for the comparative analysis of the subgroups. Obviously, comparisons were done between groups normalized with the same ECs.

Yes, the process might seem cumbersome. But working on conclusions from erroneously analyzed data can be more wasteful (thus, costly) in the long run.

http://www.clinchem.org/content/55/4/611.full.pdf

Dear Oskan,

There are species-specific and cell-specific house keeping genes. Therefore, there is no unique gene. However, in case of human T cells including Treg we tested several genes including b-actin (it's not good) and found that EF1-alpha (elongation factor alpha, see Kubach et al. Blood 2007) is the best one for rtRCR with human PBMC.

I agree with Helmut, the right housekeeping gene depends on the cell type and your treatment paradigm. Some frequently used housekeeping genes such as b-actin or Gapdh might be regulated by your treatment. When starting with a new cell type you should try some housekeeping genes to find the most suitable for your purposes and/or screen the literature.

Why is everybody predicting some genes? Normally you test all your different genes in all treatments/cell lines/... and then you use 2-3 genes which have the lowest difference in your experiment and use them as the "control genes".

Nadine is right. Actually Selection of Housekeeping gene greatly varies with cell type, treatment etc. It is essential to use few of them and select one with less variance.I ideally use 5 house keeping genes in my experiment and selection best three. use their avergae value.

Have you tried 18S ????

You might want to have a look at these websites and tools:

http://www.gene-quantification.info/

http://www.gene-quantification.de/hkg.html#refgenes

www.gene-quantification.de/Vandesompele-Kubista-Pfaffl-real-time-PCR-chapter-4.pdf

I find literature for 18S exactly used in macrophages, so it is a posible one!

Thanks a lot everybody!

If you have some other ideas, I'll be happy to hear them!

18 S will work fine. eF1alpha also, I disagree with Nadine on a simple point. Imagine the situation where the same OD is obtained by different amounts of RNA.... this will invalidate your case. I must say that in guanidinium-thyocianate preparations (RNAzol, Trizol, etc this is what you use) the OD readings sometimes give up to 1 log difference with the real amount of RNA. We solved this by resuspending the RNA on proteinaseK in the presence of 1%SDS and some Tris, incubating it during 30 min and then extracting again, but most people do not do this. This will not happen if you use NP40 lysis (Favaloro et al.,) but again most people do not do this.....

If you have enough target genes (at least five) and also enough replicates for one treatment (also five I think) you can use the software NORMA-Gene described in the article of Heckmann and colleagues (2011). It does not use housekeeping genes but rather calculates the average deviation from a sample to the mean for the specific genes, respectively. Finally, it gives you a normalization factor with wich you multiply the fluorescence data of your samples.

Dear Oskan, in our experience rplp0 (arbp, 36B4, prp0) works fine; however, you must look for the best choice for your specific cell type. We work with 3T3-F442A. Best wishes. Tonatiuh

First, it is better to use the term "Endogenous Control" than "Housekeeping".

Second, as stated several times before, there is no "ideal" or "best" EC for ALL experimental settings. All genes vary to some extent, so a good normalizing gene for me might not be the best for you. So validation is a must.

I strongly recommend reading the MIQE guidelines:

http://www.ncbi.nlm.nih.gov/pubmed/19246619

www.clinchem.org/content/55/4/611.full.pdf

There is also a good overview Nature Protocol by Nolan et al.:

http://www.nature.com/nprot/journal/v1/n3/abs/nprot.2006.236.html

The Gene Quantification website is full of very high quality information regarding quantification PCR, including protocols and also theoretical considerations on the models for PCR data analysis.

http://www.gene-quantification.info/

http://www.gene-quantification.de/hkg.html#refgenes

Our workflow included a list of 8 genes reputed "stable" from the litterature in our system. We systematically included the 8, and retained the 3 most stable for normalization in our analyses. The advantage of this approach was, as we actually found, that in particular subanalyses we could normalize expression with a group of EC better suited for the comparative analysis of the subgroups. Obviously, comparisons were done between groups normalized with the same ECs.

Yes, the process might seem cumbersome. But working on conclusions from erroneously analyzed data can be more wasteful (thus, costly) in the long run.

http://www.clinchem.org/content/55/4/611.full.pdf

I agree that HK genes should be established for each condition (cell type, treatment etc.) but I think the actual right way to do it would be to run a pre-experiment to see which genes are most stable. There are special kits that come with 12 different HK gene assays (at least we used it for mouse tissues, but I don't know if they have those for cells, chack with PrimerDesign or Applied Biosciences). You run these on your control and treated cells (or whatever conditions you plan to test gene expression in) and use the GeNorm software to estimate which genes are most stable. Then only you use those in the proper experiment. Of course, it is a bit expensive to do this for every treatment, if you only plan to do the experiment once in your lab, so I would agree that the second best way would be to run at least 4 HK genes you "predict" to be stable with your experiment and use the ones that don't change. But the problem I see with this is that what if say, 2 genes are stable and 2 genes show upregulation in your treated sample. How do you know it wasn't you extraction method that introduced bias, and you actually have more RNA in your treated sample, and the two HK genes that are upregulated are showing you the real amounts of RNA whereas the two that are stable would actually be dysregulated if you had the same RNA amounts (no extraction bias)? I agree that the chances of that happening are small, but it is possible.

Also, I think 18S should be avoided if possible, unless you have no other genes to use or your genes of interest are expressed at very high levels. The reason is that 18S is expressed at very high levels and if you are measuring a gene that is expressed at much lower levels, small differences in your RNA amounts between samples will not come up in the 18S, but might come up in you gene of interest, in which case these would look like your gene of interest is affected by treatment, where it really is just extraction/pipetting/cDNA synthesis variation. I would always advocate using a few HK genes with different expected Ct values (probably between 20 and 30 Ct) so that you don't mistake variabilities in the method for biological variability.

You can (and people do) use any gene they want for normalization, but that doesn't mean it will give you the true answer. I believe several individuals above mentioned the work by Hellemans et al. and Vondesompele et al. that describe a way to screen a panel of potential reference genes and pick those with the most expression stability across your samples. These papers also highly suggest picking at least three of the most stable reference genes and normalizing to the geometric mean of their relative expression values. The newer BioRad instrument softwares are already set up to do this if you have access, or you can use geNorm software (which I think is free to academia).

It certainly costs more money to normalize properly since you have to run quite a few more reactions in order to assess multiple reference genes. The alternative, however, is spending less money and producing data that's of questionable value, since it's nearly impossible to validate the stability of a single reference gene. My previous labs all had a "favorite" reference gene that, when I compared them for normalization purposes, gave up to a 10-fold difference relative expression. So appropriate normalization REALLY can make quite a large difference in the repeatability of the data.

A word of caution on using 18S rRNA as an Endogenous Control:

Ribosomal RNA is NOT a messenger RNA: it is a highly structured, highly abundant RNA that is not transcribed by the same polymerase as your target genes (presuming you are studying mRNA expression levels), is not spliced, and lacks a both the Cap and the polyA tail.

rRNAs do not behave as your usual mRNA. They may remain abundant and degrade with a different kinetics that your mRNAs in the mix, and their abundance makes normalization a problem when comparing to mRNAs with lower expression levels.

So, as Anna Bobrowska suggests, it would be best to look for ECs within the range of expression of the target genes in the experiment... and avoid rRNAs when possible.

p.s. If you're working in humans or mouse, one suggested panel of reference genes is Hprt1, Gadph, Tbp, Sdha, Ubc, Yhwaz, B2m, Rpl13a, and Hmbs. Unfortunately, I can't remember what paper these come from but they're used in the Hellemans and Vandesompele papers.

If you're short on RNA, I would suggest using one step RT-qPCR. Primer design is much trickier, but I've had great success with low well-to-well variance using one step RT-qPCR. Plus it costs less and you can start with lower RNA amounts than when you have to go through a separate cDNA production step. I work at Promega, so I'm of course a little partial, but in my hands, our Access One-Step RT-qPCR kit has been fantastic.

A word on MIQE:

The MIQE guidelines help establish a quality control method to produce the best qPCR results, and to be able to clearly document the process. To assist those who may be interested in setting up a MIQE workflow for their qPCR experiments, I attach a handy checklist (not of my production, so kudos to the folks at illumina http://www.illumina.com/).

And if you're tech savvy, there is an App for that.

On iPhone, iPod, iPad:

http://appshopper.com/reference/miqe-qpcr

On Android:

https://play.google.com/store/apps/details?id=com.biorad.miqeqPCR

Rafael thank you for guidelines!

At the moment I'm write an artical about modulation of gene expression by plant extract. I'm sure, that it will be very helpfull!

But what about Macrophages, Limphocytes (isolated from human PBMC). if I'm analyse the expression of antioxidant and inflammation related genes, wich endogenous controls are most abundand?

Thanks again?

Dear Oskan, as already described in the literature, there are no universal housekeeping genes. The best control depends on your experimental conditions. I would suggest you to look for the best control genes with the RefGenes tool: http://www.refgenes.org. There you can give your gene of interest and select in the Samples selections experiements or conditions which are as closely related as your own experimental conditions. Good luck with your experiements. Best Gaelle

What about normalizing to the average between biological repeats? Is that a valid approach for large data sets or is there a strict requirement for a reference gene?

I have used Actin, TBP and GAPDH in the past, depending on the cell lines under investigation. Actin levels seem to vary between cell lines , TBP on the other hand, seems to work very well with most cell lines.

I recommend Actin, Ubiquitin, PP2A & GAPDH. They work for me very well.

you just have to try the mentioned genes and see which is the most stable in your system.

We published a few years ago a simple analysis method for seeing which genes were the most stable. It can be found in:

A.B. Schote, J.D. Turner, J. Schiltz, C.P. Muller. Nuclear receptors in immune cells: expression and correlations. Mol. Immunol 44 (6) 1436-1445 (2007)

You can try a panel of housekeeping genes like this http://www.invitrogen.com/1/3/housekeeping-gene

The Eugene software automatically obtains house keeping genes for any genome: http://bioinformatics.ua.pt/eugene

The obtained/calculated housekeeping genes are then given in file format.

This application also allows you to harmonize codon usage.

The generalization of good normalization genes is still an empirical process and as noted there are no "universal" control genes. We have typically looked at genes with the following characteristics: very low replicate CVs and hence usually reasonable expression levels and independence from case:control status. Any specific tissue type or especially blood samples will yield different such genes, as with mixed cell populations this becomes especially challenging. We will often look at 5-10 candidates and pick the best ones based on the above criteria as well as sometimes also look for uncorrelated genes so we can balance out any slight effects.

GAPDH has work well in difficult human tissues as well for me. We found that even if you use acceptable endogenous controls, the selection may still effect your results. We have used RSP9 in addition to GAPDH, but in the end GAPDH has been better.

A comparison of the accuracy of Illumina HumanHT-12 v3 Expression BeadChip and TaqMan qRT-PCR gene expression results in patient samples from the Tampere Vascular Study.Raitoharju E, Seppälä I, Lyytikäinen LP, Levula M, Oksala N, Klopp N, Illig T, Laaksonen R, Kähönen M, Lehtimäki T.

Atherosclerosis. 2012 Nov 12. doi:pii: S0021-9150(12)00795-2. 10.1016/j.atherosclerosis.2012.10.078. [Epub ahead of print]

PMID: 23177970

I recommend beta-Actin or GAPDH. They work very efficient.

if you can afford it run a GeNorm analysuis. If not, make sure you're using at least three HKG and look for changes in expression from control to treatment. Pick the most stable ones and use the geomean to normalise.

hi,

18s and GAPDH work out fine for me

Which genes are suitable for data normalization can change from experiment to experiment, depending on your treatment and cell types. This is why reference genes have to be evaluated by including severeal potential candidates to the assay and evaluate afterwards which were the most stable expressed and hence moste suitable once to calculate the reference gene index (RGI), as mentioned before by Rafael Ramirez Moralez. Sometimes it happend to me that one of my target genes turned out belong to the most stable expressed genes. This highlights one fact which everyone should keep in mind: In nature, there are no things like reference genes. Every transcript of a coding gene can be influenced by a certain treatment. For example, beta actin which is involved in the the cytosceleton formation of non-muscular cells often turns out to be regulated when working with a treatment that leads to major shifts in protein metabolism e.g. in hepatic cells. This is not suprising, when we keep in mind, that up- and down-regulation of protein synthesis leads to major shifts in the cyto-skeleton because the intracellular motility has to change in order to adapt protein transport mechanisms within the cell. I just want to say, that a gene isn´t just suitable for data normalization just because the literature states it. Furthermore, as there are still many work groups are publishing gene expression data without proper reference gene evaluation e.g. with software packages like GeneEx, such references could not be trusted enough to avoid the "cumbersome" evaluation procedure, as Rafael Ramirez Moralez called it before.

more than a few housekeeping genes are no longer considered as stable as they were in the past. you can read all the authoritative answers and related references cited. moreover, you can easily find on pubmed lots of research works indicating variation of the endogenous control genes in lots of different treatments in cell lines or animal tissues.

in my experience, the ribosomal RNAs (18-28 S) are often very good endogenous controls, mainly based on the fact that their abundance is largely greater than most of the mRNA species; thus, their fluctuations are almost always slower than those of mRNA... almost always, not always. so you have to study and decide case by case which housekeeping is better for your conditions.

All the answers above, and from Rafael and Steven in particular, are very good. An important point not to miss though, is multiple endogenous control genes can be the best way to go.. If you have 2-3 that are kind-of stable, then run all three and use the geometric mean of them as your reference, This is mentioned in the MIQE (i think), but I thought it important to really note this fact, Its like biomarkers- the hunt for one has often failed us, but multiple indicators are more robust.

I recommend you GAPDH.

All reference genes should be identified emprically (with or without available software) depending on the system you're using. I agree with all of the above on that point.

But, I believe the MIQE guidleines suggests the term "Reference Genes" instead of HKG or EC. And one would use the arithmetic mean of the Cq values of 2 or 3 reference genes' technical replicates and the geometric mean of those values for the 2 or 3 different ref genes used.

See the attached file: these things can be figured out (relatively) easily in Excel; given enough time.

(Also see the post by MIQE co-author Jo Vandesompele on this site at):

https://www.researchgate.net/post/What_method_do_you_use_for_choosing_the_correct_refrence_gene?ev=tp_feed_post_xview

Another post here addresses other important facets of this topic:

https://www.researchgate.net/publication/237669540_New_Guidelines_Seek_to_Promote_Accurate_Interpretation_of_Data_and_Reliable_Results_Standardization_of_qPCR_and_RT_qPCR?accountKey=Jo_Vandesompele&ch=reg&cp=re215_x_p7&pli=1&loginT=fiFYQvhfVWkFym8ra-OXFk5zemV39_MO

Article New Guidelines Seek to Promote Accurate Interpretation of Da...

better to go for 2 or 3 EC, actin and GAPDH

@Utkarsh: Better to have a look when the answer was posted AND check other comments first. Repeating the same things again and again and again isn't helpful!

Cheers, Nadine

HPRT1 is good for EC, worked well for me.

see:Careful Selection of Reference Genes Is Required for

Reliable Performance of RT-qPCR in Human Normal and

Cancer Cell Lines

Francis Jacob1,2*, Rea Guertler2, Stephanie

www.plosone.org

http://www.biotechniques.com/news/MIQE-qPCR-Guidelines-Three-Years-Later/biotechniques-342718.html#.UacRr4WCNxt

@Shailendra: The question was posted in November last year! And GAPDH is not always the answer for every qPCR! It depends on samples etc. Cheers, Nadine