% IP variations - (Dec/12/2008 )

Hi all,

I am facing a massive problem at the moment and have run out of ideas now, so I was hoping someone might know some suggestions?

I have been doing ChIP with various antibodies (predominantly against histone modifications) for several months now, and doing every antibody in 3 independent experiments on samples crosslinked on different days usually resulted in fairly good reproducibility in the % of Input value, suggesting the protocol and antibodies work(ed) fine. Now out of a sudden, using exactly the same protocol on exactly the same cell line, the final % of Input values are about 5-10 fold lower than before, while the overall 'profile' for each antibody across the different genomic loci seems to remain similar to my previous data. This is a big problem, as obviously now the standard deviations will be enormous if I compare current and past results.

I tried everything, from using starting from a new bottle of formaldehyde, including making all my buffers and reagents from scratch - nothing seems to help, which only leaves me with the assumption that the beads (different batch?) are different - I am using the sheep anti mouse Ig Dynabeads...

Has anyone maybe experienced a similar problem at all or any suggestions?

Thank you very much!

---

Is this a change for all of the antibodies you use or just one?

If it's all of the antibodies, is it possible something could have happened to your beads (did you switch to a new lot for instance)? Are you careful in using the same amount of input chromatin in each ChIP? Did the change correspond with the use of any new buffers? Are the pHs on all your buffers correct?

If it's just a change in one antibody, did you switch to a new lot for that antibody?

---

Well, I have a suggestion on data interpretation, although you might have done that already

what if you do 2-way ANOVA to take into account of experimental variation?

In that case, even if there's big inter-assay variation, as long as the trend is still there, your p-value might still be ok, although the error bar will not change

---

Hi KPDE,
Thanks for the feedback!
Yes, I have checked about 3 different antibodies (all newly thawed from -80C stocks), and all seem to show a reduction in %IP values. I don't think the amount of chromatin is the problem either, as I'm using a set number of cell to crosslink and the Input cycle thresholds are pretty much constant and similar across current and previous experiments.
The beads are in fact a different lot number, though at least in the initial experiments, I could not see the difference I am seeing now... Also, I was just running ChIPs with two different vials of beads to see if I killed mine off due to wrong storage, but the result is pretty much indistinguishable.
The only thing left I could think of now would be the sonication step - maybe the current samples are 'oversonicated' (would that cause such a difference?) - the reason for this assumption would be that the debris pellet when spinning down after sonication is smaller than what it used to be. (The sheared chromatin on a gel looks similar though, but then again it is pretty much low MW already... Maybe reducing from 10 to 9 sonication cycles might improve the signal if oversonication was an issue?)

---

The only thing I could imagine "oversonication" doing to your samples is overheating them. That could be responsible for the difference. I've never done an experiment to determine if the amount of sonication makes a difference over the point where your fragments won't get any smaller. Might be worth looking into.

---

I couldn't recall where exactly I read that now but I remember somewhere from Upstate, they mentioned that you cannot tell from a gel whether a sample is oversonicated

while under-sonication you can tell from smear in a gel, over-sonication is harder to judge,

say for example, if you sonicate with 10 strokes and 12 strokes, both resulted in smear of around 500bp (indistingishable in gel), the 12 strokes sample is in fact "over-sonicated" already, and Upstate also mentioned that over-sonication will lead to reduced signal in later steps

The "optimal" sonication condition is always preferred, which is defined as the least number of strokes that result in the desirable DNA fragment size.

(because there is a limit of how small the DNA fragment can get upon subsequent sonication, you will not see the DNA getting smaller and smaller and smaller when you continue to increase the no. of strokes)

---