Also, we want to avoid a lot of the less rigorous science. We want people to be rigorously evaluating things in the field. I think interacting in forums like this [conference] forces us to be.
Yale’s Snyder Talks About Microarrays and US HUPO
Name: Michael Snyder
Position: Chair of the department of molecular, cellular, and developmental
biology, Yale University, 1988-present; Director, Yale Center for Genomics
Proteomics, 2002-present; President, US HUPO, 2006-2008
Background: Postdoc, Ronald Davis' laboratory, Department of Biochemistry,
Stanford University School of Medicine, 1982-1986; PhD, California Institute
of Technology, 1983; BA in chemistry and biology, University of Rochester,
While for most proteomic researchers mass spectrometry is their technology of choice, a small handful has focused on microarray technology. In that space, Mike Snyder is considered one of the leaders.
For the past two years, Snyder has been the president of US HUPO. When its annual conference ended this week in Bethesda, Md., so did his presidency (William Hancock at Northeastern University became the organization’s third-ever president.).
ProteoMonitor spoke with Snyder this week at the conference to speak about his work and the role of US HUPO. Below is an edited version of the conversation.
The last time we spoke, you had just published a paper on microarray technology you had developed for detecting ovarian cancer biomarkers [See PM 11/01/07]. What kind of update can you give me on that research?
We’re working very hard. The first paper was obviously to see if we could use protein microarrays to find candidate markers for ovarian cancer. The hope, quite frankly, was that we could find an auto-antibody signature that would be diagnostic of either early detection or … prognosis or projected outcomes from treatments.
The first step was just to see if we could get a signature. We didn’t find the signature. We did find proteins that seemed to have higher reactivity to auto-antibodies than controls, although we couldn’t come up with a nice diagnostic signature like we had hoped.
It essentially became a discovery tool. We’ve now gone through about 16 of our top list [of antigens], and the vast majority are over-expressed or elevated in some fashion in the cancer versus normal. The first paper, I think had five [in fact, four] proteins. We’re now up to 16 [and] it’s very, very clear these things are elevated. Some of them, we think, are elevated just because of [the] kinds of cells that are there, epithelial cells are enhanced.
But others seem to be maybe more specific for ovarian epithelial versus others.
We’ve been pushing on this in two ways. One is we’ve taken this initial candidate list of 94 proteins, made a lot more of those, and spotted them down onto mini-arrays, if you will, to make a mini-array of 85 or so proteins. The goal there is once again to try the auto-antibodies, but this time with this more focused array, [with a] higher amount of protein [and] see if we could find this signature that eluded us the first time.
We’re just doing that now. We’ve just done all the first probing, so I can’t tell you the answer, but we’ll probably know next week, the first sign of this. We’re going to be doing a lot of probing, a lot of patients.
The second thing is to see if we can set up a sera-based assay because obviously, we can find these things over-expressed in tissue, [but] tissue is not going to be any way to do diagnostics. No one’s going to go around taking chunks of people’s ovaries to see if they have cancer.
A postdoc has been working on this, trying to set up a mass spec-based as well as an immunologic-based assay, and we’ve just got the assays working to see if we can see these things in plasma, and the next step is to look beyond that to see if we can see something different in cancer versus normal.
Do you have any idea how difficult it’s going to be to go from the tissue-based [assay] to a sera-based one?
I have no idea, quite frankly. I’ll be surprised if everything in tissue show up in sera. But on the other hand, if you think about how we got these antigens in the first place, they showed up because they are auto-reactive, because they’re active with antibodies, which suggests that at some level an antigen or some mimic [was] released into the sera, so I’m hopeful that that will actually be diagnostic.
Are you going to eventually do the assay for all 94 proteins or are you going to limit it to some smaller subset?
The biggest shortcoming for proteomics, and this is true for everybody in this area, is that we don’t have good antibodies for probably two-thirds of these 94 proteins, so we’re still missing many, many antibodies. We’re still going through to find out for those that exist, which ones are any good for staining or good for immunoblots, or what have you.
We are making antibodies for some of the things that we’re more excited about, as well.
So that’s with ovarian cancer. We’re going to apply this to other diseases as well, autoimmune diseases and probably some other cancers. We’re going to try to generalize the approach a little more.
Is this the main work you’re doing in proteomics?
We’re doing all kinds of different things. There’s a big push to work out the phospho network of yeast, and that involves a lot of things. One project right now is to map the phosphorylation or recognition motifs, the target motifs of all protein kinases. We’ve done about 60, so we’re halfway there for yeast kinases.
And then with that, we can actually map out where the phosphorylation sites are occurring on the substrates [that] we already know about, but we don’t know where the site is. And the other thing is … we can make antibodies of those motifs and, I hope, set up a mass spec-based assay to be able to map things in kinase mutants versus normal and try different, new screens to find all these things.
The ultimate goal is to try and work out the phospho network for yeast, which is going to be thousands, if not tens of thousands, of phosphorylations.
We’re also studying protein-protein interactions, which proteins interact with other proteins, particularly in the area of protein kinases, not just substrates of protein kinases, but interacting partners, presumably a lot of which are regulators and such. We’re also working on targets of small molecules for using protein chips, and we’re doing metabolomics-type work to figure out which small molecules bind which proteins. We just started that project.
Talk a little bit about why you decided to work with microarrays. Proteomics is dominated by mass spectrometry as a technology.
It is true for protein profiling, it used to be 2D gels; these days it’s mass specs, because you can see several thousand proteins by these technologies.
Why did you decide to work on microarrays?
They’re very different, they operate in a very different space in terms of the application than mass spectrometry. We set up protein microarrays to do a high-throughput characterization of proteins, although we recognize it has a lot of value in a lot of different applications.
We had been doing some DNA microarray work, and we knew how powerful those were. What we reasoned were that protein microarrays should be even more powerful because they can be used for lots of things. Over the years, we’ve set them up for looking for new interactions. That includes protein-protein interactions, interactions with phospholipids. Nobody had these kinds of studies: what are all the phospholipids binding proteins and which proteins like to bind which phospholipids.
We did a screen looking for new DNA-binding proteins. We found some new gene regulators using this method as well. And then the idea of using it as a target for post-translational modifications, or phosphorylation work that we did a few years [ago] and that we’re still continuing now, taking all the kinases to see what they’ll phosphorylate. It’s just something you can’t do easily by any other approach.
You can do complementary things, but you can’t interrogate what all of the possible substrates of a protein kinase are. It’s really quite valuable for that.
And one of the other things we did was use it to find targets of small molecules.
I always envision this will be just terrific for pharma companies, to be able to take lead compounds [forward]. It’s very easy to get leads, but it’s hard to figure out which ones to take to trial. If you knew something about what their reactivity profile was, you might know which compounds to go forward with and which ones you might shelve for a while.
We just think they have enormous utility, at least the ones that we’ve built. They’re really not for protein profiling, which is what mass spectrometry does.
Now antibody arrays, which is a different type of protein microarrays that we build, those can be used for protein profiling. I think there’s a lot of value to those arrays as well, but that’s a different space than the protein arrays we build. They’re for characterizing sets of proteins.
Why aren’t more people working in this field?
I don’t know. There should be. This meeting clearly has a huge biomarker focus. Biomarkers mean protein profiling. The stuff that we’re building is not for protein profiling. It does this have this angle, the auto-antibody profiling that has a disease angle, and building networks and pathways is pretty important.
And if you look in science in general, look at the RNA field. A lot more people look at RNA expression and study the biochemical activities of new RNAs. It’s just a very different space. So expression arrays, like mass spectrometry, it’s just a way of characterizing a system and quite frankly, the goal is to have the technology robust enough that anyone can do this.
On the protein chips, certainly as we build them, there are some technical issues. It is more technically challenging to build a chip with 20,000 proteins that to build a chip with 20,000 probes that follow gene expression. So the technical challenges are much higher. So there has to be a higher upfront resource investment to build these. Even if you do these nucleic acid programmable arrays, you’re going to need some sort of clone database to work with.
On the vendor side, there also hasn’t been a lot of work in this area, especially compared to mass specs. How much has that affected development of microarrays?
I’m sure that’s been a big factor. Companies like to go into something they think will profit.
It’s kind of a Catch-22.
It is. To me, it’s a no-brainer for any pharmaceutical [company] to take these arrays and put their small molecules on them to see their reactivity profile. But it was very, very difficult initially. Now, it’s not the case, there are some companies doing this, although still not all.
But initially pharma companies were not doing this, and their first reaction to me was, ‘Well, who else is doing this?’ Not so much, ‘Wow, this is a great idea, we should do this.’ It was, ‘Nobody else is doing this, why should we?’
Now, I can understand why they wouldn’t want to put their blockbuster drug on there now. They probably don’t want to find something out that, quite frankly, they don’t want to know. I don’t believe [that’s the right decision] but I can see why they might not do it.
But for their lead compounds coming through the pipeline, it’s a no-brainer to use these things.
What is going on at US HUPO? Are there any projects that you’re excited about?
Well, as far as the organization as a whole, there is this statistical proteomics initiative. I think that’s very, very valuable to the field because there isn’t the rigor you would like to see for analyzing data, just knowing what standard programs are the best ones to be using, and how they’re evaluating their performance, things like that.
There’s just a lot of basics that need to be done, and I think it’s great that it’s being taken on by US HUPO, and … I think they’re making some really nice contributions and strides in that area.
There are things I would like to see us do. I’m part of the international HUPO, and there’s a big antibody initiative, and I’m part of that. We really do need antibodies to all proteins. To me this is a no-brainer.
It’s funny. The need was recognized very, very early on. In 2000, we would come to these NIH meetings and we would tell them, ‘You need an antibody for all proteins because that’s the rate-limiting step.’
In ovarian cancer, you would get these hits. Then we want to get antibodies to characterize them further [but] they don’t exist.
Why is US HUPO necessary?
The reality is that proteomics should be a basic tool that everybody uses, and it’s not just yet. I think it is a new emerging area. These days, genomics, on some level, has become fairly routine, so anybody can do a gene expression microarray, any research lab.
Proteomics should be that way; it’s not there yet.
Also proteomics is a wide-open area, there’s no endpoint in proteomics. There are just lots of different initiatives that should be done. The whole area of post-translational modifications is so poorly investigated, and the classic example I like to use is that when we started doing our kinase initiative to see which kinases phosphorylate with proteins … when we entered the area, we actually counted up how many kinase substrate interactions were known for yeast, and we came up with 160, most of which were tentative, by the way.
You know that’s ridiculous because there are thousands and thousands of phosphorylations and we don’t know which kinases do any of these things. We don’t even know what they all are. We’re so early in this field.
So, I think stimulating research and bringing us to the next level is very valuable for science. I think if we creep along the way we were going, one protein at a time, it’s just not the best way.
Also, inventing new technologies, getting them so robust that any person can step up and do this [is important].