Thomas J. Magliery was born in Oak Park, Illinois, in 1974 and grew up outside of Chicago and Indianapolis. He conducted medical genetics research with M. Ed Hodes at the Indiana University Medical Center in Indianapolis, leading to a semifinalist-winning project in the Westinghouse Science Search in 1992. Magliery majored in chemistry at Kenyon College in Gambier, Ohio and received his Ph.D. (2001) in chemistry from the University of California, Berkeley, under the direction of Peter G. Schultz. As an NSF Pre-Doctoral Fellow, he worked on several key aspects of engineering living bacteria for the site-specific insertion of unnatural amino acids. Joining Lynne Reganin Molecular Biophysics & Biochemistry at Yale University in 2001, Magliery introduced a cell-based screen for the four-helix bundle protein Rop and demonstrated its use in sorting libraries of protein variants with randomized hydrophobic cores. Magliery joined the faculty of the Ohio State University in the fall of 2005 as an Assistant Professor in the Department of Chemistry and the Department of Biochemistry. He is a member of the Ohio State Biochemistry Program, the Biophysics Graduate Program and the Chemistry-Biology Interface Training Program.
And so I literally drew on a napkin an idea for how you could tie Rop activity to GFP expression from a plasmid that’s regulated by Rop. I got there, and a few months later we had a picture of an LB agar plate that was exactly like that napkin picture, and it was one of those moments where you realize you can once in a while actually draw a picture on a napkin and have it come out being a really useful experiment. I mean that experiment alone is the basis of a big portion of the combinatorial work we do now in protein stability.
Interviewed by Barry Bunin, PhD, CEO, Collaborative Drug Discovery, Inc.
I thought an interesting first question, given most of our researchers work on drug discovery, would be to talk about your drug discovery screening work versus some of the protein engineering work you’ve done as well, in terms of what you do for each area, what’s different, what’s similar, how they might complement each other?
I think in terms of the drug discovery screening work that we’re doing, we’ve mainly been focusing on this protein called paraoxonase 1, which is a human serum protein that is associated with HDL cholesterol and is a hydrolase, but of an unknown target. But conveniently, at least at a low rate, it’s able to hydrolyze organophosphorus compounds. And so our efforts have been to modify the protein, to make it a realistic therapeutic against organophosphate, pesticides, and nerve agents, and that involves a number of different steps. One step is that the activity is too low and that we don’t really understand clearly how to change the specificity, particularly since we don’t even really know exactly what the mechanism or active side of the protein looks like. So we’ve been using structure activity relationships to try to understand that by basically making a large number of mutants around what we think is the active site, and studying it against a large number of analogs of both esters and organophosphates.
But that work has dovetailed with some of the other work in the lab, because the other aspect of this program has been to try to improve drugability, to make it a more drug-like molecule. It’s really kind of a bear to work with because it’s normally bound to these greasy HDL particles, and so we’ve been doing a number of things to try to make the protein more soluble and stable and easy to work with, for example by removing cysteines and reengineering around them. We are also doing things to improve the solubility while at the same trying to make the actual drug as much like the human protein on the surface as possible, to avoid immunological effects.
So the reason that dovetails with the other things we’re doing, is because actually most of my lab’s focus is on a number of different approaches to understand both protein stability and protein interactions. We’ve invented a number of ways to screen for, and then in high throughput fashion quantitatively measure, the stability of libraries of proteins. And we’ve also developed methods to be able to predict stabilizing mutations from sequence statistics, in addition to some things with understanding protein interactions using similar kinds of approaches. And so we’ve been able to take some of these fundamental things we’ve learned and apply them to making better therapeutics that are more stable or more soluble or things like that. So there are some themes that go throughout like using high throughput approaches and using structure activity relationships, but the two different things have really been able to inform each other quite a bit.
And if the research is modestly successful or extremely successful, what do you anticipate being the outcomes in the short- and long-term?
Well, it depends on which part of the project, because some of it is very fundamental and some of it is very applied. So I think that we’ve already had a number of preliminary successes. And I should say, by the way, this the paraoxonase work that we’re doing is part of a large center grant that’s led by David Lenz at the U.S. Army Medical Research Institute for Chemical Defense (ICD) and that includes a group at the Weizmann Institute, notably Danny Tawfik and Joel Sussman, who have done a lot of the activity engineering, and then people here as well – Chris Hadad, who’s a computational chemist, and George Wang who works on modification of proteins, so it’s a very big group of people.
But I think we’ve already made some substantial progress. I mean we’ve been able to, between the Weizmann and us and ICD, we’ve been able to engineer, produce and test in animals, variants of paraoxonase that are able to protect against lethal doses of authentic nerve agents. Not all of them yet, and they may not be exactly the ideal molecule yet, but I think just that is a success. So I think that if it was very successful, we would have either one molecule or a cocktail of molecules that’s able to hydrolyze a broad spectrum of organophosphates against high lethal doses, so that they’re really completely protective. But like I said, along the way there’s fundamental things to learn that dovetail more with some of the other work that we do.
In terms of engineering properties, our long-term goal is to be able, at least empirically, if not fundamentally, to predict better which mutations to proteins are going to change their physical properties, especially in terms of stability and solubility – things that are currently very difficult to compute and also have been difficult to meta-analyze from a bunch of different unrelated mutations of proteins. Our strategy is to hopefully generate large sets of highly related mutations and study their physical properties, so that we can get a sort of statistically significant answer to questions like: How do changes to the sequence of proteins and particular parts of the proteins, say the core or in loops or on the surface, how do those changes really correlate to changes in properties like stability and solubility that are very difficult to predict?
Great. So how have you used CDD? Where has CDD been useful for your research in science?
I think we’re using it a lot in one area and I think we have started to make some use of it in another area. To be honest, I think we need to spend more time on that, but I think it has great potential in that area. So the first area is in the structure activity relationships with paraoxonase. This enzyme and other enzymes like it, like phosphotriesterase and a couple other esterases, are excellent at turning over OPs [organophosphorus compounds] at least to some degree. There have been a lot of studies of these over the years, and so there’s a large set of plausible substrates including mimics of nerve agents, actual nerve agents, pesticides, mimics of pesticides, esters, lactones – there’s a huge knowledge base. I mean aside from the many mutants we’ve made in our lab, there’s a huge knowledge base that’s already out there of different mutations.
Just being able to collate all of that data, and then turn it into actual knowledge is a very significant challenge, and that’s an important way that we’ve been using the CDD resources: to be able to put those together in a way that’s searchable and in a way that we can organize, based on properties of the different substrates, and hopefully be able to draw conclusions, and I actually think that has been fruitful. I think now we understand a lot better which mutations are important for things like large versus small substrates for example. I think the ability to calculate molecular properties and sort by those molecular properties easily in CDD just makes that an invaluable resource for us.
The other thing that we’ve been trying to do and we haven’t explored as much, is seeing if we can do something similar with the large libraries of protein variants that we generate. Obviously it’s possible to put them in, including the sequences and even structures if we wanted to. In principle we can build in lots more simple calculations, because for a lot of our libraries that’s how we start: like for the places where we’ve made mutations, what’s the net charge? what’s the hydrophobicity? and things like that. And we’ve been doing that with a combination of CDD and Excel, but I’d like to see us be able to use your resource [CDD] a little more for that. Mainly because I think that it’s just easier to organize all the results that way. And this is my fifth year as a professor now, having a bunch of students make a bunch of libraries of even just a couple model proteins – we’re looking at thousands of variants of these proteins, and so it just gets to be an organizational nightmare.
Great. Moving back to science, because I know you have a broad background in chemistry as well as biology. Just talk about an “ah-ha” moment, which resonated with you for your career or just for curiosity, to explore things and catapult you in your development as a scientist.
Absolutely. You’re exactly right that, also as a scientist yourself, you know that there are lots of days where you are just… Well let me put it this way: from grad school alone, I have seven or eight notebooks with probably 30 publishable pages in them. There are lots and lots of days when you’re just trying to figure out how things are going, but there are really watershed events in your development, when you do have those moments, and I think at almost every level I’ve had things like that happen.
A really fundamental thing that shaped me when I got started in science was that I did a Westinghouse project when I was in high school and I had no idea what I was doing. I looked in a phone book and called Ed Hodes at Indiana University Medical Center, not knowing that he was a giant in medical genetics and in genetic testing, and he just very generously had me come to his lab, and I was working with a post doc on a project on single strained conformation polymorphism. I think when we visualized that first gel, and I really realized in practice the fact that it was going to work as a genetic testing method, that was extremely exciting to me and I think that experience was really formative. But I think I had experiences like that at almost every level.
Another one was when I started as a post doc with Lynne Regan. Lynne for a long time had been doing directed designs of this protein Rop, which is a bundle protein that we still work on quite a bit. One of the things that they didn’t have was a high throughput way to evaluate the function of Rop. And so I literally drew on a napkin an idea for how you could tie Rop activity to GFP expression from a plasmid that’s regulated by Rop. I got there, and a few months later we had a picture of an LB agar plate that was exactly like that napkin picture, and it was one of those moments where you realize you can once in awhile actually draw a picture on a napkin and have it come out being a really useful experiment. I mean that experiment alone is the basis of a big portion of the combinatorial work we do now in protein stability, so I think those things are really formative and driving moments.
This may be my last question: Just before you used CDD, how did you manage your data and why did you initially choose CDD?
I would say we did it pretty badly, is probably the answer. Mainly Excel, and we still use Excel for a lot of things because it’s very convenient for a lot of calculations, but the result is that our data would end up in a bunch of different worksheets and people’s group file folders. I think analysis between experiments and between projects was much for difficult. Not to mention the fact that you end up with nomenclature problems, because people aren’t going to necessarily enter the structure of the molecule in your Excel database, so you just name it some random way and then when you go back to analyze, you have to be able to decode all of those things. Actually I have to say I’m not exactly sure I remember the moment when I first heard about CDD. I think maybe you and I talked at an ACS meeting and then I went back and looked at it online and I thought: This is exactly what we need to be able to catalog molecules and things like that in a much more reasonable way, things that are really beyond the simple resources of Excel. In a way it’s actually a little bit weird, because what we do has a very significant bioinformatics component, and I have students using Perl and Java, but we hadn’t brought any of those resources to bear on the chemoinformatics end of it. That’s what CDD has really done for us, and it’s an amazing interface for that.
This blog is authored by members of the CDD Vault community. CDD Vault is a hosted drug discovery informatics platform that securely manages both private and external biological and chemical data. It provides core functionality including chemical registration, structure activity relationship, chemical inventory, and electronic lab notebook capabilities!
CDD Vault: Drug Discovery Informatics your whole project team will embrace!