A week or so ago, I posted an enquiry concerning practical
considerations in using EMSA. I received many very useful answers to my
enquiry, and I would like to thank all that participated in the
discussion. I would like to offer ABRF subscribers the following
compilation of responses, as it may be helpful to users of this
technique.
Mark
The questions
1) How long does it take to run an experiment?
2) How quantifyable are the results?
3) Can it be automated?
4) What are its limitations?
5) How reproducible are the results?
The answers
1) It takes about 1-2 hours to pour the non-denaturing gel out and
prepare the samples, plus 2-4 hours to run the bandshift (=
electro-mobility shift assay, EMSA), the time will vary depending on the
size of the free probe, the concentration of the PAGE gel and the
maximum current you would choose to use. This will be followed by
1/2-12 hours to expose the film, if the probe is P-32 labeled. However,
the typical experiment will take longer if you plan to use different
probes - the labeled DNA that you would like to use is probably
double-stranded, and has to be less than 150 bp, better several dozen
base pairs long. The latter means you have to choose one of the
following: a) synthesize (partially) complementary oligos, anneal them,
fill them with Klenow and P-32 dNTP, run a gel and isolate them from a
gel; b) use "hot" PCR to amplify the probe either on a plasmid template
or on a longer synthetic oligo that would serve as a template; c)
prepare a restriction fragment from a cloned DNA, label it with Klenow
and isolate from a gel. Of course, if you a planning to use the same
type of DNA, you will have to prepare your probe only once or twice a
month (for P-32).
2) Gel shift assay is a qualitative, at best semi-quantitative
method. It is quantifiable only if the affinity is very high. Otherwise,
you need
to have the drug in all the buffers, and run gels at different drug
concentrations. In principal, you are better off with a competition
experiment of your drug against a fluorescent drug, especially if both
are intercalators.
3) The automation is up to you to develop. My guess is that for the
start it is better to pay somebody $30K to do it manually for a year (or
contract the job out to a university professor who can put three 10K
students on the project) than buy a robot for 1000K.
4) I would say the biggest limitations are caused by the
"non-physiological" type of the experiment and the "false positive"
results. Almost anything can bind (and shift) your probe in vitro in a
"pure probe - pure ligand" setting, but when you try a more stringent
method, such as a footprint, the results are often negative.
Competitive binding and the use of extracts/mixes instead of pure
ligands can improve the situation, but peer review process will nail you
down for using only bandshifts, they will probably ask you to add the
following additional experiments: a) supershift (antibody against the
ligand should supershift the band "ligand-DNA"); b) footprint (with
DNaseI); c) affinity column chromatography (your DNA probe attached to
the resin) followed by a sell gradient elution; d) if possible,
immunoprecipitation.
All of the above is for protein-DNA interaction. If your drug is a
small 300 Da molecule rather than a 50-200 KDa polypeptide, the
bandshifts may appear less convincing, given the fact that each
nucleotide in you probe is as big as your whole ligand. In addition,
you may have no antibodies against the ligand.
5) The results are reproducible.