Southern Illinois University
Who: Luke Tolley
Position: Assistant professor, Southern Illinois University, 2003 to present
Background: PhD in analytical chemistry, University of North Carolina, Chapel Hill, 1996 to 2001; postdoc in analytical chemistry, Oak Ridge National Laboratories, 2001 to 2003; visiting scientist, GlaxoWellcome, 1998
Working with his colleague Matthew McCarroll at Southern Illinois University, Luke Tolley is developing an instrument that can identify protein targets for new drug-discovery applications.
Identifying such targets would allow greater insight into how drugs work, whether they work at all, and what adverse effects they could have. Yet, according to Tolley, no such genomic or proteomic approach for such identification currently exists.
Tolley and McCarroll, who is an associate professor of chemistry at SIU, have developed a prototype of their instrument, dubbed DIABLA for dynamic isoelectric binding ligand assay. They recently received a grant from the National Science Foundation to further develop DIABLA.
ProteoMonitor spoke with Tolley this week about the platform. Below is an edited version of the conversation.
DIABLA meets a need in the drug discovery and development process. [It’s goal] is to find the unknown protein targets of drugs. There are a good number of drugs or other chemicals of whatever sort … it’s clear that they have protein targets, and yet it’s not clear what those protein targets are.
So that makes it very difficult for us to optimize a drug, for instance, or to even maybe figure out why it has certain side effects, or other aspects of its mode of action. DIABLA is designed to find these unknown targets. As far as we know there aren’t other good ways of doing this.
How does it work?
DIABLA is based on a variation of isoelectric focusing and fluorescent anisotropy. What we do is we use drugs that either have an intrinsic fluorescence or that we’ve tagged with a fluorescent molecule. Then we separate the proteins using isoelectric focusing. Then we can scan to … check the rotational rate of the drug, and if it’s binding to the protein, then it will spin much more slowly than if it’s not.
Therefore we can find all of the protein bands that are interacting with that drug or compound, and then we can isolate these bands and test them to make sure that it’s a true positive. It allows us to screen very, very complex samples for the binding targets of drugs.
The detection is fluorescent polarization anisotropy, so as the molecules spin, if you excite them with polarized light, they lose their polarizations. If they’re spinning more slowly, they lose their polarization more slowly, and since drugs are very small in general, and proteins are huge, if a drug binds to a protein, it will greatly increase the time it takes to rotate, which means it will keep its polarization much longer than a drug that does not bind.
It can spend 1 percent of its time binding, and we can still see it because of the sensitivity of the fluorescence.
You said there’s no real good way of detecting these protein-drug interactions.
Well, if you have a pure protein, then it’s very easy to see if a drug binds to it. And that’s been done for a long time and there’s no difficulty there.
However … one of the test cases we did was with a cancer cell lysate … and we found the five proteins in there that bound to the drug that we were checking. In that sort of a situation where you have an incredibly complex sample, I haven’t been able to find any other good way to discover the targets like this.
I have found papers … where other people are also trying to do sort of the same thing, and they have a much worse way than what we’ve got.
So is this kind of protein target identification just not done, or is it done by guesswork?
Mostly by guesswork. What you’ll do is you’ll get a very skilled biochemist or somebody who has a lot of experience, and they’ll look at the structure and say, ‘Well it sort of looks like this and this other drug, and we know what this other drug binds to, so let’s see if it binds to the same thing.’
Even if you do that, you don’t know if it binds to something else in addition. Even some very common drugs – the example I use is Tylenol, or acetaminophen – they don’t know what it binds to, and it’s been around for 50 years. They’ve got some guesses, but they’re still not sure exactly how it works.
DIABLA would allow a researcher to see exactly what is binding to what?
It would definitely help you find the targets, and once you know the targets, you can figure out what the targets are doing. Before you know the target of a drug, it’s harder to figure out exactly how it’s doing whatever it’s doing.
And I say drugs but one of the [projects] we’re going to look at is environmental endocrine disruptors, pollution and things like this where they kill all sorts of amphibians and cause these bad problems, and yet they’re not sure exactly what it is that they’re doing.
At what stage is DIABLA?
We have a prototype that we use to gather data, and so we have tested it, as I said, to look at breast cancer lysate, to look for basically steroid hormone receptors, just as a proof of concept.
[We have a National Science Foundation] grant to develop it into two different platforms. One of them will be more suited for high-throughput screening. So you could take a drug and say, ‘Well, we’ll compare samples from different tissues and see if it binds to the same receptor in every tissue.’ If it binds to something different in one of them, then that’s of interest.
It could do that very quickly or it could also do it from different individuals. You could say, ‘Let’s see if this drug is going to bind to the right thing in you, because if it doesn’t, then it’s going to mess you up, or it may more likely do nothing.’
That’s one of the platforms, and the other one will be … it won’t be high-throughput, but it will be for very, very detailed or more sensitive and methodical isolation and identification of all the targets, rather than a screen.
Is the idea that someone who’s doing research in this area to use both platforms or one or the other?
Many people would use both, but it depends on what you’re doing. If you have a drug that you want to screen for side effects, then you could use just our screening technology or if, let’s say you have 30 percent of patients responding to drugs, and the others don’t: you would want to screen through that to see if it was a binding difference.
However, once you find something that’s different, they you would want to go to the other [platform] to get a very, very accurate and precise isolation of the protein to make sure it is what you think it is.
You gave an example of how the environment affects certain species. But what other applications would this platform have aside from drug discovery?
With safety. Right now, when [companies] come out with new chemicals, they don’t have to do nearly as many tests as people assume they do before they use them, like in plasticizers. With this, instead of having to do long-term animal studies to see the effect of this thing, you could screen and see if it’s going to bind to the same sort of things that these other pollutants [do]. If it binds to the same things, then we know it’s going to be bad, too.
Anytime you’re trying to find the protein target of a small molecule, DIABLA could greatly accelerate that process.
Has there been any commercial interest in DIABLA?
Sort of. We haven’t gone out to try to talk to companies, really. When I was at Pittcon last year, I mentioned it to some people, and they were very interested, but I didn’t really pursue it because it wasn’t ready at the time.
The university has filed a patent … so there has been some interest because the companies do see the enormous potential that it has. But we haven’t really pursued it because I wanted more solid data and better proof that it really works like we say it works.
We are always looking for new test cases. I don’t see papers where they say, ‘Hey here’s a new drug and we have no idea how it works. I wish someone would help us.’ You just don’t see that sort of thing published.
So, I’m talking to collaborators and trying to find good examples of drugs where we could really prove this thing solved the problem that was very difficult to solve any other way.
Can you share any of the data from your testing of the prototype?
We’re submitting it for publication, so I don’t think we’re allowed to … Basically, what we’ve done for some of this test data, for the cancer cells, for instance, is label the proteins with one fluorophor and had our drug labeled with a different fluorphor. If you look at it under the one excitation, then you can see all these proteins and they’re focused and separated.
And then you look at it under the excitation for the drug, and only one of them shows up. So that’s the target that we’ve identified.
How are you validating your data if you’re saying there’s no approach currently to identify these protein-molecule bindings?
That’s a concern. The first thing that that we’ll do is, after this potential [target] is isolated, then we’ll test it again in solution to make sure the preliminary reading was correct. Then we’ll basically digest it, run it through LC-MS, to get a positive identification of the target.
And then, you go to the biochemistry people and talk to them. ‘Look, does this make sense for a target? What would this do if this protein were activated or deactivated?’
Some of them turn out to be … even though this is binding, it’s a non-specific interaction, and so it’s not a target. We have in our protocol a way to identify that.
If it turns out to be … not a non-specific interaction, then that’s where you use the knowledge of biochemistry and say, ‘OK, we believe this is a target, we can then measure the binding coefficient and everything.’ And then we could check to see if this makes sense with the known physiological responses.
Have you come across any candidate protein that’s not identified in any database?
It hasn’t happened yet. That would be very exciting, but not yet. I think, more than likely, with most drugs [there] would be targets that were not known [to be] targets but had been identified previously for other things.
Would post-translation modifications affect the performance of DIABLA?
It doesn’t affect the efficacy [of DIABLA], [but] it does affect the binding. One of the great things is once we isolate the target, we have a very, very high resolution isolation of the target. Once we isolate it, with the LC-MS, we can also search not only for fragments [but also] post-translational modifications because it may turn out that it only binds to one that’s been modified in a certain way.
Even though LC-MS doesn’t find all the post-translational modifications, it can find the majority of them.
Whether there’s a PTM or not doesn’t affect whether it’s a candidate protein?
No, we can still find it even if it’s modified.
Does that apply to all modifications and splice variants of a protein?
I mean if it were so modified that we could never get it into solution or anything, then we’re not going to see [it].
[For example], membrane proteins are difficult for everybody to analyze, but in general, glycosylation or whatever you can do to that protein, phosphorylation, these things are just going to be fine.