Hi there,
When perfoming ChIP-qPCRs, we - as most of others - express the results in "input DNA %" or "fold enrichment":
- in the first case, normalizing the Ct values from experiments with antibody against our protein to those from negative controls (e.g.: "beads only" or IP with IgG), and express that in the % of the input DNA;
- while in the second case, we calculate the fold change comparing Cts with our antibody IPs to those with the negative controls. However, neither of these cases data is used from the (otherwise performed) positive control experiments, such as iPs with H3 histones.
The reason I am bringing up this issue is because of our recent experiments: we could prove the binding of a protein to certain enhacer DNA segments in placenta but not in embryonic stem cells - so the data looks convincing. However, when I look at the data (ChIP-qPCR) with the "positive control" H3 histone, although we see a uniform binding to all examined DNA segments, there is a high peak in one of the enhancers. Moreover, this coincides with ONE of the enhancers where our protein of interest binds to... My question is: why is that?? Could be that H3 histone does not bind unifomly to all DNA sequences? Or - I think more likely - the efficiency of IP on that region is significantly better than that on the other enhancers? If so, this should somehow be indicated in the calculations - but should we ALSO normalize the data (let's say the "input %") to the result with H3? Should we correct all IPs to the background first, than calculate fold change our input % of our protein comparing/normalizing to H3? Has anyone run into similar problems? Any comments or suggestion would be much appreciated!
Some additional info: our H3 antibody is not specific for any of the modified histone forms (e.g. H3K9meth).
Thanks in advance,
Tamas
Hi
I think you're right about the interpretation of the H3 control - something in the enriched region is suspicious. but then it should be also more enriched with the IgG no? anyway, I'm not sure what is the right solution.
What Ab are you using for the IP?
Anyway, I'm not sure about your analysis to begin with.
The way I see it, % Input is basically the input/bound results (2^-Ct), and in order to show enrichment, you have to check that a negative control gene in your target IP, AND the IgG/no Ab are much less enriched. Positive controls like the H3 and a positive gene (that should and known to be bound by the target IP protein, and validate that your Ab is working OK) are enriched.
In the normalized method, I basically normalize first to the input and then to the negative control gene, and then I get fold enrichment over something I know isn't bound at all by my Ab (like APRT for H3K9M3).
I always like to check the IgG control but I don't like to normalize to it. the only reason to do it is if you have huge background nonspecific enrichment and then i wonder if the chip is OK. The IgG 2^-Ct should be at least 100 fold lower than the bounds'.
This is how I do ChIP analysis. I'm not sure it's the only right way so I'll be happy to read other people's opinions on that.
Thank you for the responses.
Yes, the negative control is the "beads only", or the better is IgG IP, there's no doubt about that. I also agree with Sharon's thoughts; we eventually did the normalization in input % but for the manuscript figure, we also showed the input % for the neagtive control as well.
Regarding the positive control histone H3, and its enrichment in parallel experiments, I eventually came up with the idea that although the antibody we are using recognizes all modified protein variants (methylated, acetylated, etc.), it might have different affinities for the variants, causing fluctuations in the signal. We haven't specifically proved that but the other experimental data also supported this idea (we have a strong enhancer supported promoter with a high transcription rate, and in those cases, we see the enrichment for the positive control H3). So then, the conclusion is that we will (and we should) use H3 as a positive control - to show that the IP reaction is working - but due to this affinity difference, it is not correct to normalize the other IP values to this (as opposed to a normalization method widely used with positive endogenous controls in western blots, or analogously in real-time PCRs).
I am still happy to receive comments on this topic, and I will also update my responses if we something new or supportive comes out from our experiments.
- Badges
- Science method