At A Glance
Name: Katheryn Resing
Position: Director, chemistry department mass spectrometry facility, University of Colorado, Boulder, since 2003.
Background: Associate and assistant professor, department of chemistry and biochemistry, University of Colorado, Boulder, since 1989.
Research associate, post-doc, department of periodontics, University of Washington, Seattle, 1985 to 1989.
MS in botany, University of Kansas, Lawrence, 1969.
PhD in biochemistry, University of Washington, Seattle, 1985.
How did you get involved in mass spectrometry and proteomics?
I'm a protein chemist. I trained with Ken Walsh and Hans Neurath in Seattle, and they were one of the first people to get the SCIEX API III plus that basically introduced most biochemists to mass spec. It was the first really commercially viable electrospray instrument. And I was fortunate enough to work with a really good mass spectrometrist, Rich Johnson. We did some of the early phosphorylation studies and we also analyzed proteolytic processing of a protein I was working on at the time called profillagin, which is a really important skin protein and has a really complicated protein chemistry to it. We'd been trying to sort it out with standard techniques for a couple of years and made no headway, and then within a couple of weeks we had it all straight with mass spec. The standard techniques we used before mass spec included digesting peptides, resolving them on HPLC, running them by Edmund degredation and later by gas phase sequencers. And then I was using radio-labeled P32, trying to identify the phosphate sites, and I hadn't identified any. And then within a week [with mass spec] I had 21 sites. That must've been in 1990 or 1991.
What did you do after that? Were you convinced you wanted to stay with mass spectrometry?
Soon after that, I moved to Boulder. At that time it was pretty hard to get instrumentation. We worked out this really nice deal with one of the local biotech companies where I helped them use the machine, and then on the machine I got to use it for my own work. That worked out really nicely and then after about a year we managed to get our own instrument. I worked in phosphorylation and we did some of the early top-down proteomics, where you analyze the proteins instead of digesting them into peptides. We did some early top-down work on histones and ribosomes.
Why did you decide to use the top-down approach?
If you have a protein that has a lot of modifications and you want to try to understand combinatorial regulation between the sites, if you digest them then that combinatorial information goes away. So top-down is the way to get at that kind of combinatorial information. And we were particularly interested in the correlation between phosphorylation and acetylation. So we did some of that work and then we got into deuterium exchange mass spec which is a way of looking at dynamic motions in proteins and conformation changes in proteins. Basically what you do is you put the protein in D20, so you get deuterium exchanged into multiple sites in the protein, and when you drop the protein then into acid and cold, that stops the exchange and stabilizes the exchange of the backbone amide, the peptide bond. And then as it happens very conveniently, pepsin digests under those conditions. Even though you're on ice and at pH2.5, you can digest the protein into small pieces and then you put them onto a reverse phase column and analyze them and you can see the changes in conformation and structure of a protein. If you have a peptide that's buried inside the protein structure, it's not going to exchange. The ones on the surface are exchanging. And things like dynamic motion in the protein are important too, because if you have a site moving back and forth, it's partially accessible but not completely accessible.
So you used the deuterium exchange to study phosphorylation?
In collaboration with Natalie Ahn, we've done a lot of work on that, looking at allostery and dynamic motion changes when kinases are activated by phosphorylation. Then Matthias Mann came out with the paper showing in-gel digestions using 2D gels. We had already been doing in-gel digestions for doing de novo sequencing, so we knew it worked really well. And people who wanted to clone genes, we'd do in-gel digestions with the bands. I would identify the peptide sequences for the cloners. So we already knew it worked really well. Then we had a student named Tim Lewis, and Tim worked out how to do 2D gels. He was very good with his hands and he made beautiful 2D gels. And he did a really nice study looking at phorbol-ester induction of map kinase pathways and differentiation on erythroleukemia cell lines to a megakaryocyte phenotype. Megakaryocytes make platelets. And as far as I know, it's still one of the few studies where a large number of protein spots were identified by 2D gel analysis, and every protein that was targeted was identified. And there were like 104 of them. We were interested in those proteins because they changed when you induced this differentiation process.
How did you get into shotgun proteomics?
Three or four years ago, I went to a small meeting that Waters organized. I had been hearing about shotgun proteomics on yeast, and I thought this looks promising, but it's still not as good as 2D gels. I went to this little meeting, and they showed results on the yeast protein. They had like 2,800 or 3,000 proteins. And I did a little back-of-the-envelope calculation. We're mammalian people and I said this might just be doable for a mammalian system. And we had just purchased an ABI Pulsar which was supposed to be good for this kind of work, so I came home and I said ‘OK, let's do shotgun proteomics.’ To my distress, the first sample that we ran, we got 13 proteins, and I said ‘Uh-oh, we've got to fix this, make this work better.’ So then I spent a lot of time — it took us six months — well, part of the reason we got so few was we weren't really resolving the sample. So first thing we did was really what John Yates pioneered using the strong cation exchange to separate the peptides and longer reverse phase gradients, and then we got numbers that were comparable to what they were getting. But that first run was a bit of a shock.
But what surprised me when I looked at the data was how much of the data you couldn't account for with the search program. And at that time we were using Sequest. We were seeing on a good day in a good dataset only 25 percent of the MS/MS spectra were being accounted for. So I began to try to understand what was going on. Everyone was saying that's modified peptides, sequencing junk. So I developed assays for everything I could think of and I went through and I took a dataset and I identified everything, or almost everything — I think there was five or six percent I couldn't identify. It took me quite a long time, obviously. We noticed in the process of doing that that if we ignored the scores, and we just looked at the fact that Mascot and Sequest agreed on the sequence, often the peptide was correctly identified. But the scores were so low below the statistical threshold that you would never have caught them with the search program. So then we developed a whole new approach to validating peptide sequences that used consensus between search programs instead of scores. But that gives you a lot of false positives. So then we developed methods to filter the list using peptide properties, like how they eluted on a chromatography, and whether or not they were probably miscleavages. What we found was with those methods, we could get another 30, sometimes 40 percent more identifications. So in a dataset where we could identify 500, you could now identify 800 peptides. We called it the MS Plus and we just recently published that, and we're just about to ready to produce a production version that people can use.
So you kind of switched from phosphorylation studies to developing new bioinformatics approaches?
Yes. The other thing we discovered in this study was that the search program didn't handle protein isoforms or variants very well, and sometimes seemed to identify them just randomly. So if you had a peptide that came from two different parts of the protein, it might assign one isoform variant to one peptide, and another isoform variant to the other peptide. So we developed a method where we threw out the protein identification information too. From the search, all we saved was our validated peptide assignments. Instead of looking at the database from a protein viewpoint, we just look at it from the peptide viewpoint. And one peptide might come from one protein in the database, or two, three, however many. We use that database to take our peptide and pull in the protein information. We call the program Isoform Resolver. That program is available, and we've written it so it can be used with any data. It can compare between samples — it's a really good protein profiler. And when we applied isoform resolver to a set of Sequest-identified proteins, we eliminated 24 percent of the identifications by the application of the isoform resolving. We now have a good work flow that handles shotgun proteomics data really well that we think identifies more peptides than any other system out there. We think we have a better method of handling isoforms than other programs out there do.
What are you going to work on next?
We're getting back into phosphoproteomics. We just bought a new API 4000 Q-TRAP, which is supposed to be ideal for phosphoproteomics and we're getting ready to start making use of that.