OK, so here is my brief protocol. I use the SimpleChIP enzymatic magnetic kit from CST (#9003).
- Harvest 5 x 10(7) cells per experiment. For me that is about 4 x 10cm dishes of cells.
- Follow protocol exactly (except for 30 min protein lysis on ice), up until the addition of MNase
- I add just 0.3 ul of MNase/5 x 10(7) cells in 1 ml of buffer B and leave for 20 min at 37 degrees
- The protocol states that 5 ul of MNase is what CST used for HeLa cells. When I tried this amount, I had nothing but single nucleosomes.
- I then resuspend the nuclear pellet in 1 ml ChIP buffer as per the protocol and split into two 500 ul aliquots
- I then sonicate very mildy just to break open the nuclei. For this I use a Branson Sonifier on 20% and do 2 x 10 second pulses for each aliquot.
- 100 ul of the clarified chromatin is then used in each IP reaction (final volume 500 ul). Therefore, I use about 5 x 10(6) cells per IP reaction. This is pretty much per the protocol
- IPs are left O/N and then I follow the protocol exactly from there on in.
- For ChIP-Seq, I will usually do 4 - 5 individual ChIPs for my TF and then combine at the DNA purification step to get enough material for quantification and library prep.
Once you get down to single nucleosomes (<150 bp), you really start to see dramatic drops in % input values. I think that this is mostly because you are working blind in the primer design stage (at least I was) and so you may not be right on top of the binding site and you really have no flanking region to "buffer" the PCR. So your chances of capturing the binding event dimish, unless of course your primers are right on top of the site. If you go for slightly larger range of fragments (i.e. one to six nucleosomes) you see a huge jump in % input and I think it is related to the primer design point above. Note that there is really no parallel increase in background with less fragmentation.
I am convinced that this over-digestion problem is why MNase digestion has gotten a bad rap with TF ChIP. Most of the objections to it are also theoretical in nature and there is really little data to suggest it is inferior to sonication. In fact, all of my data suggests the complete opposite, as does a whole bunch of data from some vendors who are pushing it (i.e. Cell Signaling Technology). Once optimized, it is also very reproducible, assuming experimental conditions are kept consistent.
In terms of optimization, it is quite simple. You just have to titrate very carefully the MNase and be very careful to keep cell number consistent. In my hands MNase was very potent in chopping up my DNA and I ended up using a very small quantity in my reactions. One big simple optimization experiment should take care of it. I would be more than happy to give you some more details should you decide to switch.
I am submitting my ChIP-Seq data for publication this week and so we will see how it copes in peer review. I am not anticipating any major issues with the ChIP methodology.