AT A GLANCE
Professor and chair of molecular, cellular and developmental biology at Yale.
Postdoc at Stanford from 1982 to 86 in Ron Davis’ lab.
PhD from Caltech in 1982. Worked in Norman Davidson’s lab.
Developed ChIP-Chip strategy to study transcription factor binding sites, microwell array for biochemical assays, and yeast proteome chip
In your work with microarrays, you have developed a number of different chips for various applications, One innovative use of arrays was the ChIp-Chip strategy you devised to study transcription factor binding sites. How did you come up with this chip?
We were studying this [yeast] transcription factor, Swi4, and we really wanted to know its targets. At the time we started this, for most transcription factors you knew one, maybe ten targets. This ChIP-Chip idea — [ChIP stands for chromatin immunoprecipitation] — was basically a great way to find out what all the targets are. It was a wonderful collaboration with Pat Brown’s lab because they are the microarray gurus.
Of course when you set this up, everybody comes up and says, that’s great for yeast, but we want it for human, so we got it working for human. We got a center grant to do this on a large scale for humans now, so we will be doing this in a big way. We built an array of chromosome 22, but we haven’t published anything yet.
Now you are also well-known for the yeast proteome chip your lab has built as well as your microwell arrays. Why go with protein chips to study protein binding and functions?
The nice thing about protein chips is that you really control the conditions, you are in charge. You can change concentrations, which lets you look at a whole spectrum of affinities. Obviously these are all in vitro assays, and you ultimately have to verify everything with in vivo assays, but it’s a great way to find [binding] candidates.
If you are interrogating individual proteins, quite frankly, all you do is pull out a chip, and your probe, and you are done. It’s just so much more rapid than two-hybrid for example. You can explore biochemical functions just as fast, you are only limited by your assays.
What are the advantages of your microwell arrays?
There are certain kinds of assays that are just better done in a well format because they involve several components. Not that you can’t do them on a surface, but you can control the environment a lot better. Those wells we used initially, they are 300 nanoliter volumes, so you really don’t need very much material; they also reduce evaporation. You can do multi-component reactions in those and keep them segregated from neighboring wells. It’s really nice for [kinase] inhibitor studies. We think it’ll be nice for certain small-molecule studies as well, where you could incubate small molecules with the wells in these compartments, wash them off and then, using mass spec, elute off the bound small molecules and find out what they are.
The yeast proteome chip developed in your lab received notice with the publication of the September 14 Science article describing it. What is special about this chip?
When I go to these proteomics meetings, a lot of these companies say ‘we spend all this time on making nice surface technology and arrays.’ But that’s not the rate-limiting step in the whole protein chip business. We figured that out in a week, by just testing lots of different conditions. The hard part was making a high-quality expression library, and that took a lot of time because getting the clones sequence-verified [and] making sure they are fused properly by sequencing across the junction takes time, and it’s not that cheap either. There were three things you needed to be able to do to accomplish this feat: The first was the high-quality expression library, the second was procedures for making lots of proteins at once, so we had to set up high-throughput protein production procedures. And the third is the array technology itself. The biggest thing about our protein chips is that the assays are exquisitely sensitive. We are putting down 10 million molecules in a typical spot, but we only need 1,000 to see a positive signal. If you have 99.9 percent of your material dead, you still see it.
What is version 2.0 of the proteome chip going to look like?
Even though we haven’t got 6,200 clones, we are almost there. Obviously in our case, you may wonder why we made aminoterminal tags, and the reason is, the primers existed for that. When we started this, we had no money and there was one postdoc doing this. Eric [Phizicky at the University of Rochester] and I are going to join forces to make C-terminal fusion proteins, which I think is going to be much more valuable for the membrane proteins.
At the end of the day we’ll have an N-terminal fusion set and a C-terminal fusion set. The N-terminal set actually fits on half of a slide, and so the C-terminal one could presumably fit on the other half. I don’t know whether version 2.0 will or will not be optimized for membrane proteins. There are probably always a few that will be very difficult to get, and it won’t be cost-effective to get them down as full-length.
The other thing you have to remember is, the annotation of yeast is not perfect. We will always be discovering new genes that were missed by the annotation, so there will always be new revised editions. But having said that, I’d be surprised if we don’t hit 95 percent of the information, I hope within a year. [With] the membrane proteins, we just have to see how challenging they are. But in the worst case, even the big ones we can break down into subdomains if we have to, and we get the information, just not in a fully functional fashion.
How difficult will it be to do the same with human proteins?
One big factor in our success for yeast was that we express yeast proteins in yeast. One thing that surprised me certainly was how many of our proteins were active. The fact that most of our protein kinases were active was pretty remarkable. It still remains to be seen what level of function we’ll have for the human proteins as we go about expressing these in various systems. That will be a little more challenging than yeast.
When I started this, especially with the [yeast] protein kinases, so many people came up to me and told me so many reasons why that wasn’t going to work. Everybody knows proteins are unstable, you can’t store them very well, they don’t work on solid supports. Had we listened to all these people back when we started, we would be nowhere. We just jumped in with both feet and did it. And I think the same will be true for human [proteins]. I think one could get a first human proteome chip without too much effort. Will it be 100 percent active in all assays? No. Will we get certain kinds of assays working, simple binding assays? The answer almost certainly is going to be yes.
How will you overcome the challenges of membrane proteins?
That remains to be seen. I think there are a number of obvious ways to try. Will you be able to set up lipid environments on a solid surface to do binding assays? The answer almost certainly is yes. I think it’s doable, but I think it remains to be shown. Some of the [yeast] membrane proteins, by the way, are active. Two of the protein kinases are actually membrane proteins. The polytopic membrane proteins will be the most challenging — the multiple spanners, the G-protein coupled receptors. But even in a membrane protein you have soluble domains, and even if they are aggregating, some of the soluble domains would probably still be showing up on the chip.
What kinds of assays are you going to do next with your proteome chip?
As an example, we probed with some lectins, like wheat germ agglutinin, which recognizes N-acetyl glucosamine. We can find all the modified proteins that way. And in principle, you can do this for phosphoproteins as well [using phospho-specific antibodies]. We have also probed with GTP and ATP.
There are many small molecules and many drugs that we would love to know the targets of. I am involved with a new company called Protometrix that’s involved in this whole area.
Where do you see the protein chip field going?
Ultimately there will be protein chips for every single organism. I think the first step will be sets of classes of proteins, then there will be uni-proteomes, where you have one representative of every single gene. The criticism that everybody used to raise, and they still do, [is that] there are probably millions of proteins in humans when you consider alternative splicing. I would say just having representatives is not going to have all the information but will be very valuable for lots of things. Another thing that would be very valuable is the uni-domain set. And the combination of the two would be a great first start for biology.
Ultimately you’ll buy these things from companies. I think the protein chip field will move faster [than the DNA chip field] because I think the market place is more aggressive, and we obviously would like to see Protometrix be a big player in this area.