Wednesday, January 6, 2016

Western blots versus parallel reaction monitoring (PRM)!!

This week I visited a lab that has been doing some great validation work with parallel reaction monitoring (PRM). While preparing their work for publication, one of their collaborators began insisting that they "validate" their findings with western blots. I don't know about them, but I felt like I'd been playing Jumanji....

It is 2016 (despite what you've been writing at the top of every page!)!!!  Holy cow. I know I'm typing to my imaginary choir here, but I really want to get this out of my system. Surprisingly, though, no one has really done a head-to-head comparison of PRM versus Western Blots that I can find and I'd like this rant to pop up when I type the two terms into Google. There is, however, a ton of material to pull from to support my rant.

I hereby present: Western blots versus Parallel Reaction monitoring!

In the red corner we have Western Blots. There is a great wikipedia article on this methodology here.

Basically, though, you do this:
1) You run an SDS-PAGE gel
2) You transfer the proteins via more electricity to a membrane (sucks 'em right out and they stick to the membrane
3) You soak the entire membrane in a solution containing a commercial antibody raised against a peptide within your protein of interest. (They do this by injecting a peptide from your protein of interest into a rodent, or camel, or horse. Typically a rodent, though. For this example, lets say its a bunny rabbit.
4) You wash away bunny rabbit antibodies that don't stick to your membrane in a super tight way
5) Then you add a solution to your membrane that has an antibody with a detection region and a region that binds to any bunny rabbit antibodies
6) Then you activate your detection. That detection region might light up (fluorescence) or it typically causes a chemical reaction that makes a dark spot where stuff matches.

Amazing technique when it was developed in the 70s. It is, of course, still super powerful today, but it has weaknesses that have been addressed many times. First of all, it relies on the efficacy and specificity of two commercial antibodies. I know this is ancient history, but in 2008 Lisa Berglund et al., did a high-throughput analysis of commercial antibodies and found that a large number of them did not work at all. In fact, the average success rate of the 1,410 antibodies they tested was an awe-inspring 49%.  I'm sure those numbers have went way way up. However, according to this 2013 article in Nature Methods, the field of antibody production contains over 350 separate producers. Despite this level of competition, the paper appears to recommend returning antibodies as a step in normal lab practices. Hey, no one is perfect, but I'm just throwing these articles out there.

Let us assume that the antibodies you've ordered have been used by tons of groups and that they work just fine. Chances are that you can find a protocol that will give you a good method for step 4 above. If you don't wash away non-specific binding you will just get a blot full of signal. If you use a wash that is too stringent, you get nothing at all. Hopefully someone has done this work for you!  If not, you're on your own. And then you have to it the antibody? or is it me? Probably a good idea to run it a few more times. At 4-5 hours a pop with the newest technology and not one single glitch along the way, it might take a few days to optimize a new assay.

I found this nice picture on Google Images, I'd like to share (original source unknown):

The challenger in the blue corner -- Parallel Reaction monitoring!

The figures for PRM are taken from this great recent (open access!) review by Navin Rauniyar.

Here are the steps
1) You start with a pretty good estimate of the mass of some peptides from your protein of interest and you use that for the quadrupole (or ion trap on LTQ-Orbitrap instruments. Yup! You can totally do this on hybrids, but it works better on quadrupole-Orbitraps) isolation. You can easily refine your data acquisition by retention time or by focusing the isolation to reduce background and increase specificity

2) You fragment the ions you select

3) You match the high resolution accurate mass fragment ions to the theoretical (or your previously experimentally observed fragments) within 1 or 2 ppm (really, no reason to ever go above 3 ppm.)

4) Post processing allows you to drop fragment ions that might not be as specific as you'd like and you can use the intensity of your fragment ions cumulatively to score your signal.  Once you have your favorite fragment ions your rerun your PRM method for the samples you'd like to compare.

HEAD to HEAD time!

Category 1: Reproducibility
According to references in this paper from Dan Liebler and Lisa Zimmerman, the CV of a western blotting measurement ranges from 20-40%. In this paper from S. Gallien et al., all PRM measurements in unfractionated human body fluids were found with CVs less than 20%. 5% is common.
Winner?  PRM!

Category 2A: Time (single target)
There are a bunch of new technologies for western blots including fast transfers and fast blotting and direct signal measurements. If we assume you're using something like that and it takes you 2 hours to normalize, load and run a gel (you are faster than me). You can get this down to 4-5 hours. Now, you can use multiple lanes. But this also involves man hours, where you have to be moving things, transferring things, blah blah blah.
 Winner? In pure time with newest technology? Western blots, maybe. In time (measurements per work day? Definitely PRM.

Category 2B: Time (multiple protein targets)
Okay, imagine that you weren't just interested in the quantification of one protein or one phosphorylation site. What if you were interested in the quantification of EVER protein in a given pathway. Say, for example, MAPK or RAS (overlapping, I know).
 One Western blot can be maybe 20 patients for ONE protein (or phosphorylation).
 One PRM run can look at many proteins. How many? With a Q Exactive classic, up to 100 targets is pretty darned easy to set up. In the paper above from S. Gallien, they looked at something like 700 targets. So, if you're conservative and say that you had two targets per protein, then 350 Western blots on that patient. In one run. In under 70 minutes. Compare that to 350 Western blots.
 Winner? With multiple protein targets, it is PRM and it isn't even close. 

Category 3: Sensitivity
A chart on this page says you can get sensitivity down to the femto/pico range (in grams.) Since proteins have masses in the kDa range and we measure LC-MS sensitivity on new instruments in the femto and atto-mol range, unless I'm not awake yet, this seems pretty clear.
Winner? PRM by a bunch of zeroes!

Category 4: Specificity
Antibodies are awesomely specific. But they are often raised against one peptide. PRMs of multiple peptides can easily be set up. And you can choose targets AND fragment ions with the level of specificity that you need. Commercial antibody providers will obviously take this stuff into account, but this control is often out of your hands. And its one peptide.  You can see in the blot pictures above that you may often get multiple targets. You can narrow it down by SDS-PAGE determined average mass. OR you can choose multiple peptides that are unique in evolution and you can use the retention time of those peptides and fragments that are unique in evolution, within a few electrons in mass?? This one seems pretty darned clear to me..
Winner? PRM!!!!

Category 5: Cost
Starting from scratch? A full setup to do Western blots is gonna look a lot cheaper than an Orbitrap. But if you are doing a lot of them, man hours and western reagents are going to add up soon enough. (Side note: If you are reading this and you don't have a mass spec you just might have too much time on your hands.) If you already have a HRAM LC-MS setup, you don't need anything additional for relative quantification via PRM. For absolute quan you'd want some heavy peptide standards. If you have both the capabilities to do western blots and PRMs in your lab, the antibodies, gels and membranes are additional costs/experiment.
Winner? (If you already have a mass spec that can do PRMs for single targets?) PRM!
   If you want to look at an entire pathway, like RAS or MAPK? It doesn't take long before it is WAY cheaper to buy a Q Exactive than it is to do tens of thousands of Western blots.

TL/DR? Friends don't let friends western blot. (A friend summed this post up that way, so I stole it!)


  1. Hi Ben, is it an American idiosyncrasy to write femto as fempto ? ;-)

  2. Nope, I just end up with so many things flagged as misspelled on blogger that I can't even pay attention to them. Thanks for catching!

  3. Err..perhaps explain how PRM can demonstrate that a protein has been fragmented? Also, PRM is not that sensitive, compared to the very best antibodies.

    The beauty of orthogonal methods is that they give different information. The problem with 'traditional' western blots is the stone-age technology. I'm hard core 'quant by MS' guy, but we have be running comparison of SRM (='PRM-gold' :-) ) with a commercial western blot system (no connection) and we note assay performances that are very similar, and that the quantitative western blot can detect proteins we cannot measure my MS.

    Lastly, the loss of connectivity elicited by proteolysis is a huge problem. We forget that we digest to harmonise chemical properties so that a gradient RP-HPLC run will work for most - that has never been feasible at the protein level. But, we really need to understand proteins, not peptides, and there is no method that resolves proteins better than 2-d PAGE - again, compromised by poor implementation rather than the intrinsic failing of the method. We should not conflate the two.

    Rob Beynon