What is a housekeeping gene? Theoretically, these are genes that are constantly expressed in cells, whose expression is not altered by treatment with any chemical. By choosing an appropriate housekeeping gene, one hopes to find a baseline upon which the expression of other genes can be tested. In other words, a housekeeping gene is an internal loading control for cDNA added to the PCR reaction.
Of course, even the best housekeeping genes are affected by some treatments. Therefore, it is important to use multiple housekeeping genes to examine your changes in gene expression. Choosing a combination of ribosomal RNA and genomic transcripts can provide the best results. For example, I often use actin, GAPDH, and 18S-rRNA.
What to do if you suspect that your housekeeping gene is being affected by your treatment:
Best practice is to run at least two housekeeping genes per run. This way, if one of them is suspect, you have a second one to compare to. In the example below, I compare actin and GAPDH with HO-1. In this example, actin and GAPDH are both housekeeping genes, and therefore shouldn't be changed with treatment, while HO-1 is a gene that is changed. I'll show you some comparisons of the data so that you can tell if your housekeeping gene is not behaving well.
Once you have your data in terms of relative quantity (not Ct values, but calculated values of DNA), you can plot two different genes in a linear regression plot. For example, plot GAPDH versus actin.
The correlation coefficient is around 95%, which is pretty good. In other words, actin is a good predictor for how much GAPDH we have, regardless of how the samples were treated. However, when we compare HO-1 to actin, the correlation is terrible, suggesting that one of the two genes is changing with treatment.
The same is true when we compare HO-1 to GAPDH.
Below is an example of results taken from a comparison of actin, HO-1, and GAPDH from three different tissues: brain, liver, and lung.
You can see right away that there is a good correlation between the expression of actin and GAPDH in the brain and lung. These genes may not be changing in expression with whatever treatments were done, or they possibly are changing together in response to the treatments. However, because actin and GAPDH have such diverse functions, it is unlikely that they would change together in expression.
Comparing actin and GAPDH versus HO-1 in all three tissues we see that they are poorly correlated. I know from previously experiments that HO-1 does change in response to the treatment I've given, so this makes sense.
We run in to trouble when we look at the correlation of actin and GAPDH in the liver. We can see that they poorly correlate with one another. This suggests that the treatment is having an effect on one, but not the other. However, we cannot tell from this experiment which is changing! Therefore, I would need to run another experiment looking at a new control gene such as 18S rRNA, and compare these two against them.