Best Method for Target Protein Quantification - Western Blot Accuracy Improvement Needed (Sep/11/2008 )
Greetings,
First off I am primarily trained as an Ecologist. I am grappling with identification of the proper techniques / strategies for assaying protein expression. Here is the problem:
I am attempting to compare population-level differences in expression of a particular HSP across a range of temperatures. I extracted total proteins from 4 plants (Arabidopsis) per population per treatment for a total of 192 samples. I then ran 16 SDS-PAGE gels with a Control sample for cross-gel comparison. I used Western Blotting, chemiluminescence, and the Kodak ImageStation camera and quantification software to obtain Intensity measurements for all bands. I used the Control sample that was on each gel to attempt to correct for differences across gels. An internal loading control was not possible as most proteins have altered expression levels under differing environments and I am interested in differences among populations (essentially genotypic differences).
The ultimate result is that there is tremendous variation in the readings within-populations and only somewhat clear differences between temperature treatments.
I ran 2 gels with multiple samples of my Control. There was a large amount of variation both across and WITHIN gels. The within-gel variation is the biggest problem. How can I trust my control as a method of standardizing across gels?
There are multiple other sources of variation. I fear this technique is not quantitative enough. I do not need to know how many ng of HSP are in the sample, but just the relative levels.
So I am seeking any advice I can get on: Improving the accuracy and repeatability of Western Blotting, Comparing across gels, other techniques such as ELIZA (which I know nothing about) that may be more accurate and comparable.
Part of the problem is that all of the molecular biologists in my department work with isogenic lines or genotypes that differ greatly in expression. Does anyone compare natural variation in protein expression?
Sorry for the long post!
The short answer is - you can't.
Western blot quantitation is notoriously difficult and (as you have seen) error prone. The majority of problems stem from that fact that transfer and running conditions are very variable between gels, even if the gels are run in the same tank with the same batch of buffer. Slight variations in the conditions lead to bands being more or less diffuse which affects the quantitiation. Even such simple things as exposure times and the size of the box you draw around the bands will alter the end result.
Your best bet is probably to normalise the data to a dilution series of standards (3 points should be enough) run on each gel and compare across gels from there. The results will be very variable but you should be able to get something out of them statistically. People who do a lot of quantitative real time PCR do something similar with the data out of the machine, if there is someone who does a lot of real time in your department, they may be able to help.
If you want to do this properly, you would probably need to purify your target protein from each sample and mass spec it (very expensive and time consuming), at a guess.
How about doing a BCA assay and then load equal amounts of protein from each group. You might have actually done this. Also as already described, WB is not a real quantitative method only an approximation.
This is just something which may help for quantification. You could run your samples with BSA of known concentrations. Go for 5-6 different concentrations and also have your samples in 2 different dilutions. After the gel and try coomassie staining & try to quantify the bands. Ofcourse if there are too many contaminating bands running along your HSP, then its problematic as well.
This is just something which may help for quantification. You could run your samples with BSA of known concentrations. Go for 5-6 different concentrations and also have your samples in 2 different dilutions. After the gel and try coomassie staining & try to quantify the bands. Ofcourse if there are too many contaminating bands running along your HSP, then its problematic as well.
So I quantified total proteins using the Bradford (Commassie) Assay. I loaded 5.0 ug of protein for each well for half the samples and for the higher temperature treatments where I expected the highest expression levels I loaded only 2.5 ug of protein. I ran a gel using my control sample at 2.5, 5.0 and 6.0 ug to develop a standard curve which I could use to normalize across loading quantities. Is this an acceptable technique?
What would Coomassie staining of the gel tell me? Just that I loaded evenly? If so, then it is certainly worthwhile as confirmation.
What is a BCA assay? Is that Bradford Coommassie?
I am not looking to determine the exact quantity of HSP101 in each sample, but rather accurately assess the relative amount in each sample.
Finally, if I were to do a set (or more than 1 set?) of technical replicates (essentially running all 16 gels again), would this be likely to improve my resolution/confidence in my measures? Or am I likely to see that Western Blot technique is extremely variable and anything not on the same gel is not comparable?
Thanks for all the advice so far.
P.S. Anybody know of a company that does protein analysis using something more accurate like an Odyssey? Or else another technique for quantifying proteins in this manner? Is ELISA a relevant technique? If so what is the cost per sample for any of these techniques?
Sorry for the shit-tons of questions.
What would Coomassie staining of the gel tell me? Just that I loaded evenly? If so, then it is certainly worthwhile as confirmation.
Coomassie staining is a non-reversible stain of the gel that binds in stoichiometric proportion to the number of arginine, aromatic AAs and histidine. Do not do this stain if you want to transfer and detect your protein by western blot. If you want to transfer, you can stain on the membrane with Ponceau S which is reversible (washes out in water, but probably not completely, so a little bit of blocking solution will take out all of the stain).
BCA is Bicinchoninic acid- a copper based quantitative protein detection method going from green to purple. Similar to the bradford assay, but more sensitive and probably more quantitative.
I am not looking to determine the exact quantity of HSP101 in each sample, but rather accurately assess the relative amount in each sample.
Technically yes, it should improve your resolution, but in practise it may not. I have found that you are unlikely to get between gel comparability unless you can treat all of the gels (you would probably need precast gels for this) and membranes exactly the same. By exactly, I mean exactly - not plus or minus 10 sec on a wash etc or using different batches of solution for anything.
Sorry for the shit-tons of questions.
ELISA is probably more sensitive. Radioimmunoassay (RIA) is the most sensitive and specific of these types of assay, but is fairly difficult and requires access to things like scintillation counters and radio-isotopes of iodine or something similar. Both quite expensive, ELISA will be cheaper and easier. Check out "Current Protocols in Molecular Biology" for details.
Good luck
Thanks for the advice to all who have contributed.
Regarding even loading of proteins:
I have been using a Ponceau S stain to verify satisfactory transfer. How would I use this to verify even loading quantitatively? Simply photographing the image and using densitometry? Do I quantify the entire lane?
Regarding accurate quantitation of target proteins:
I have done some reading on ELISA. But what I can't find is advice on porting one's anti-body from Western to ELISA. What I mean is that I have monoclonal anti-HSP101 produced in Rabbit. I have been doing Western blot using a Donkey anti-Rabbit secondary anti-body labeled with HRP. Obviously I will no longer use the same secondary for ELISA. What are the requirements for the final part of the "sandwich"? Does the labeled anti-body need to be designed to recognize my anti-gen (plant protein samples) or just the anti-HSP101 anti-body?
Peace,
M
Choose a consistent band and use that as reference. Not really any better than doing a loading control probe, but it does eliminate one source of error.
Sorry, can't help with the last bit really, having never done it... but I think from a quick read that you should be able to immobilise your samples on the plate and then probe with your standard primary and secondary. I am guessing it will require quite a bit of optimisation.
you can spike your samples with an equal amount of some innocuous protein to use as a normalizing control.
you can use quantification of the ponceau s staining to determine the normalizing factor for each sample.
btw, bca is similar to the lowry protein assay.