Calling all biochemists!
February 21, 2011 7:36 PM   Subscribe

Biochemistry filter: What do you consider the best evidence that would show a direct protein-protein interaction?

I have been contemplating several assays with which I hope to determine whether a protein complex interacts with another specific protein. I am hoping that there are a few biochemists out there who will tell me what you consider to be the best proof, and also if there are any assays that are commonly, yet mistakenly, used to show such a thing.
Thanks for your thoughts.
posted by nasayre to Science & Nature (17 answers total) 2 users marked this as a favorite
What kind of interaction is it? If it chemically modifies the protein in some way (changing the mass), then mass spectrometry, electrophoresis, or chromatography may be appropriate. If it either induces or removes some sort of activity (ability to interact with other species), then an enzyme kinetics assay would work. If it creates a conformational change, some more sensitive kinds of chromatography might be suitable, or crystallography. However, without some insight into what kind of interaction you're looking for, it's hard to determine what kind of assay would detect it.
posted by kagredon at 7:51 PM on February 21, 2011

Purification of a FLAG-tagged complex in vivo followed by protein sequencing. IP and then the reverse IP with antibodies to a member of the complex and then to the specific protein - if you have to make antibodies, make sure to get at least two to each protein because they don't always work well. GST pulldowns are fairly standard but I'm not a huge believer in them even with good controls. Far Western if you're feeling really silly.

Are we talking association or transient association? Modification?
posted by sciencegeek at 8:03 PM on February 21, 2011

Also: are there antibodies to any of the complex members or the putative interacting protein?
Do you need confirmation of an in vivo or in vitro interaction?
Can you express the complex and or the interacting protein?

Sorry to pepper you with questions but these are all relevant.

If you want vague in vivo information, you can always show colocalization if you have either antibodies that work for immunohistochemistry or a GFP or other FP fusion.

If you can purify the complex in vivo, you can change the stringency of the washes to show interaction. Of course if you're already down to no detergent, a limited volume and number of washes and less than 100mM salt, I'm not going to be glowingly convinced.
posted by sciencegeek at 8:10 PM on February 21, 2011

Well, that is part of the conundrum. If it does interact, it is most likely as a binding partner. I do not believe that any modification occurs (phosphorylation etc) as a result of this interaction. Really I would like to hear from someone with experience what would be taken as proof of direct binding. I have a protein (a), that when stimulated causes protein b to assemble into a complex. There are a number of possible mechanisms, but a leading hypothesis is that protein a acts as a scaffold.
If your tools include a (crappy) antibody to A, fusion protein expression systems, and a null cell line for A but not for B, what would you like to see (and what controls) that prove a direct binding?
posted by nasayre at 8:11 PM on February 21, 2011

To my thinking, I guess an IP isn't quite good enough, because though it shows binding in a complex, it would not necessarily show direct binding. Thoughts?
posted by nasayre at 8:14 PM on February 21, 2011

Isn't this done with chemical crosslinking? Also, the wikipedia page on protein-protein interactions may give you further ideas.
posted by Durin's Bane at 8:28 PM on February 21, 2011

Also, I am aware of several possible routes of investigation, including AUC, chromatography, electrophoresis, far westerns (far out!), FRET. but what is the most convincing of what you have seen?
posted by nasayre at 8:29 PM on February 21, 2011

This Pierce Protein Interaction Handbook should be useful.
posted by euphorb at 8:58 PM on February 21, 2011

IP certainly needs to be done, but so should colocalization by immunofluorescence to show that they could plausibly interact in live cells (easier and less perilous than FRET, for sure), as well as direct pulldown of purified proteins in a cell-free system (rules out the "in a complex" worry.)
posted by juliapangolin at 9:07 PM on February 21, 2011

Chemical crosslinking using crosslinkers of various lengths (available from Pierce) can tell you something about how close the reactive groups are. Are you concerned that there may be a protein C that mediates the binding of A to B, or something like that?
posted by Quietgal at 10:06 PM on February 21, 2011

Protein chemist here.

There really is no 'gold standard' experiment for protein-protein interactions, because a negative result does not necessarily indicate that a protein will not bind to a partner. An interaction might be apparent via a co-localisation experiment, but might not be apparent via, say, size exclusion chromatography. So the answer might very well - whatever works!

That said, some of the most respected protein-protein experiments are those where one can derive quantifiable data (Km, on-off rates, free energy changes). I have done a lot of work on a Biacore machine, which uses surface plasmon resonance to measure interactions and obtain these data: Assuming you have recombinant protein, as you will need quite a bit. It might need a little bit of optimisation, and someone who knows how to operate the machine. But if you are able to demonstrate an interaction using SPR - then I think it would be fair to say the vast majority of reviewers will be convinced.

But again, no one single experiment is 'better' than another. A well-run co-localisation experiment (with robust controls) can be just as convincing as an elaborate SPR assay.

(but i will go on record in saying that I am not a big fan of chemical cross-linking - too many false positives)
posted by TheOtherGuy at 1:20 AM on February 22, 2011

Former protein chemist here.

Seconding surface plasmon resonance. Like TheOtherGuy, I have also worked quite a bit with a Biacore instrument. If you do not have immediate access to one, ask around your department or university.

One way you could set up your experiment is to clone one of your interacting proteins into a vector that expresses a biotin tag. You could then use a streptavidin surface on your Biacore chip to immobilize that protein, then flow the other protein over the immobilized protein and record the binding as an increase in mass at the surface of the chip.

David Myzka at the University of Utah is perhaps the best practitioner of SPR, as well as other protein-protein interaction methods.

Good luck!
posted by Seppaku at 3:10 AM on February 22, 2011

So you don't have an antibody for B?
I'm not a huge fan of xlinking as proof of interaction either. I also find colocalization experiments to be somewhat disingenuous. Unless you have really tight localization, it gives you a pretty picture but doesn't convince me.

When you say that protein A is stimulated, what kind of stimulation are you talking about?

I spent a few years purifying large protein complexes from mammalian cells. For me, purification of a protein with a complex was verified by the reverse IP. The initial protein of interest was tagged with an epitope tag, made into a stable cell line and then subjected to a single step pull down with specific, stringent wash conditions. The resulting purified complex was TCA precip'd and then sequenced via mass spec. At the same time, a portion of the purified complex was run on a sizing column to show what was truly associated. A silver and a western were done and candidate associating proteins were westerned. The controls were showing that the protein didn't associate with the proteins that always came down with the complex purification protocol - we had a nice short list of the usual suspects. Then the confirmation of doing an IP with an antibody to the protein used for complex purification (with the epitope tag being used as, hopefully, a positive control) - controls being an unrelated antibody or two - as well as an IP with an antibody to the putative interacting protein.

Having a null cell line, or the ability to knock down expression, gives you a further way to do a negative control.

Biacore was something that occasionally happened after these steps.

Interaction of recombinant proteins was something that seemed to give a lot of false positives.
posted by sciencegeek at 4:22 AM on February 22, 2011

if you don't have an antibody to both, you can run the two (purified) proteins on blue native PAGE + western blot, with the proper controls of each protein alone, with/without stimulation of Protein A,etc. Then blot with the antibody you do have. Whether the apparent size ofProtein A changes under different conditions (like, especially when Protein A is stimulated),will tell you if they complex...under these rather artificial conditions. It's a start, especially if you have a control condition, like when Protein A *isn't* stimulated, that*doesn't* result in binding.

Asregards yourother question, I agree that IP is no demonstration of direct interaction. And colocalization is no indication of interaction.

And FWIW, both me andsomeone else in my lab tried doing blue native PAGE. He tried it the 'homemade'way, and turned everything blue and never reallygot anywhere with it. I bought Invitrogen's (overpriced) system, but I was up andrunning with the first gel and just published a paper with the data. Recommend paying a little more and saving lots of frustration.
posted by Tandem Affinity at 5:02 AM on February 22, 2011

Thanks to all for the great answers. Seppaku, your mention of Myzka is timely, as my reading of one of his retrospective-type papers in which he specifically points out instances of "you are doing it wrong" type SPR experiment got me thinking about this. Not being formally trained as a biochemist, I realize that I could do a lot of different techniques and never realize that I was missing some crucial control (or waste a lot of time doing an experiment that isn't very convincing to a biochemist, which is where the project stood until I arrived). Thus the general nature of the question. I kind of wish there were more of the Myzka-type papers for other sorts of techniques that people use incorrectly all of the time.

Sciencegeek, the "stimulation" of protein A occurs via binding of a small molecule ligand into a hydrophobic binding pocket.

This discussion helps me realize that there is no substitute for a mentor with experience! My boss is fantastic for a number of things, but a protein chemist he is not. It didn't occur to me that many biochemists might view cross-linking data warily. Good to know.
posted by nasayre at 7:17 AM on February 22, 2011

Chemical cross-linking isn't bad par se - but should always be complimented with other assays. It is usually used when trying to find unknown candidate binding partners (either in vivo or from an in vitro library), as opposed to characterizing an established interaction.
posted by TheOtherGuy at 8:34 AM on February 22, 2011

Sounds like you're doing preliminary experiments, so this isn't likely to be applicable. But somewhere down the line, NMR can do this if your system is amenable and the proof is quite convincing.

It'd require lots of purified protein, at least one 15N labeled, and sufficient concentrations (up to 1 mM)--also the mass of the complex can't be too huge for NMR. That eliminates a lot of systems upfront.

But if it works it can tell you a lot about interaction, including a sense of binding constants (fast exchange or slow exchange) and if the spectrum can be assigned, which part of the protein is at the interface.
posted by Sublimity at 9:35 AM on February 22, 2011

« Older Your Best Electropop Reccomendations   |   words, words, words, words Newer »
This thread is closed to new comments.