Chair in Bioanalytical Chemistry
Barnett Institute of Chemical and Biological Analysis, Northeastern University
At A Glance
Name: William Hancock
Position: Bradstreet Chair in Bioanalytical Chemistry, Barnett Institute of Chemical and Biological Analysis, Northeastern University, since 2002; Editor-in-Chief of the Journal of Proteomic Research.
Background: Vice President of Proteomics, Thermo Electron, 2001
Staff scientist, Medicinal and Analytical Chemistry unit, Genentech, 1985-1994
PhD in chemistry, University of Adelaide, South Australia, 1970.
At the Cambridge Healthtech Institute's Protein Biomarkers 2005 conference held this week in Philadelphia, William Hancock gave a talk about using multi-lectin affinity chromatography as a platform for serum proteomics. ProteoMonitor caught up with Hancock before his talk to find out more about his background and his technology.
How did you get into proteomics, and why did you decide to focus on studying glycoproteins?
I've been at Northeastern now three winters. Before that I was vice president of proteomics at Thermo Electron, concentrating on the mass spec platform. That was the time of the development of the linear ion trap and the new FT-MS. So I could see the new mass specs were getting much more powerful, but the problem is that really for the discovery phase, it's much better to do it in academia, because of the long pipelines and that sort of thing.
I came to Boston, and I decided to come to Northeastern because they have a very good bioanalytical research group — the Barnett Institute. I had collaborated with Barry Karger on and off for a couple of decades, starting from when I was at Genentech. So being in Boston, I think it's pretty obvious your focus should be in the clinical area, with all the wonderful hospitals. So Barry and I actually started off with tissue proteomics with Dennis Sgroi at Massachusetts General Hospital. He's done a lot of very nice work in terms of expression profiling — this is breast cancer with laser capture microdissection.
Barry's really continued in that area, but I decided that since blood is the diagnostic fluid, I'd really concentrate my program around blood. Most diseases are not accessible with biopsy and in terms of early diagnosis, sometimes you don't have a biospy until things are really bad. So I decided to concentrate on blood. The problem, of course, is that plasma or serum is unbelievably complicated. So we've all been struggling with the problem of what is a good depletion strategy, and if you deplete what do you lose?
I'd used lectins way back when I was at Genentech. The problem with lectins is they have broad specificity, and they don't capture all of the glycoforms. So I said, 'OK, why not use a combination?' The simplest thing is just a physical mixture — just take lectin A, B, C, and physically mix them.
The other bad rap about lectins is they don't bind as tightly as antibodies — they don't have as high of an affinity constant. This is actually good news if you're going to do chromatography, because you don't want it to bind so tightly that you kill the column or kill the protein when you're trying to elute it.
Then you go on to blood, and it's very interesting because the pharmaceutical companies are really concentrating on the cell, and things like phosphorylation and dephosphorylation, which are very important in the signaling pathways. That's not the blood. Blood has a lot of phosphotases, so phosphorylation is not so important in blood. What's important is glycosylation, because secretion is often associated with glycosylation.
All the focus of the field on phosphorylation is fine if you're in small molecule research and you're looking for drug targets inside the cell. But if you say, 'No no, I want to be in the clinical diagnostic arena and I want to study blood,' then glycosylation becomes very important.
We produced a very shocking result — using the multiple lectins, we found that at least half the proteins in blood are glycosylated. I'm not talking on a mass basis, because most of blood is albumin, but if you just talk on the percentage of proteins that are glycosylated, it's about half of them.
I did check that out with Leigh Anderson. Leigh took a different approach — he just looked at the proteins that had been discovered in blood, added them up, and about half of those had been glycosylated. What we did is we removed albumin. Then after that, mass-wise, about half of the non-albumin fraction bound to the multiple lectin column and was glycosylated.
Is that when you started to study glycoproteins in disease?
Well, we're not carbohydrate chemists. We try not to get into the details of glycosylation too much. I think the Japanese are the real gurus in this field, and it can become unbelievably complex. We try to stay at a more global level, and not to descend into great detail with just one protein.
Firstly, it turns out that this is an excellent depletion method. If you remove the immunoglobulins, which are glycosylated, with a Protein A column, the bound fractions are now substantially purified because you've gotten rid of the immunoglobulins, you've gotten rid of the albumin and all the other things that are not glycosylated. The glycosylated fraction is being enriched very nicely. The non-bound fraction has a problem because it's still got albumin. So now you do an albumin depletion, which is relatively simple. You're now in very good shape to do some quite deep plasma proteomics. So we've been applying this to diseases such as breast cancer and cardiovascular disease in various collaborations, and we actually get good depth in terms of the proteomics.
Then the remaining key point is if you want to have an impact on clinical chemistry, I don't think you can do what I call 'heroic proteomics' — that's a six-month study for one sample. There's nothing wrong with that — you're going deep, you're doing discovery — but the problem with clinicians is they don't have a lot to begin with. Usually you can get a quarter of a milliliter, and they've got a lot of samples. The challenge is it needs to be robust, high throughput, and it needs to get the same results over and over. So that's the great thing about this approach — it's very simple, it's high throughput.
You see, you just can't afford to do a lot of pre-fractionation. Say you've got one plasma sample, and you do a separation into 20 fractions. So one sample has now become 20, and you should analyze each fraction in triplicate, so now that's 60 analyses for one sample. You can see you quickly lose your way in terms of throughput.
The final problem with all of this is that the amount of charge heterogeneity is unbelievable. I had experience with tissue plasminogen activator — that's the heart attack drug — it's a glycoprotein. Highly purified TPA has 21 charge forms. PSA has 50 charge forms. So if you do an ion exchange pre-fractionation step on plasma proteins, things just smear out. David Speicher has done a very good job of showing this as well.
That's the problem when you go to human samples. That's why a lot of the yeast work or the simpler systems are not relevant to this type of a study, because yeast does not put on charged carbohydrates. What worked fine for yeast is going to be compromised in humans by these proteins smearing out.
I think there was a general consensus at the Human Proteome Organization — we're a reference lab for the HUPO Plasma Proteome Project — that to prefractionate plasma, you want group-specific methods, like lectins, so you don't let this charge heterogeneity kill you. You're better off to select from a group property, like glycosylation, or if they were phosphorylation, you could do phosphorylation residues.
Of the biomarkers that you've found, have they been more glycosylated, or non-glycosylated?
That's interesting. I can't give you a statistically meaningful answer, but to date we've found more interesting biomarkers in the glycosylated fraction than the non-glycosylated fraction. What happens is, most people work with a protein, and they don't even know it's glycosylated. One example is the molecule of the year, published on the cover of one of the issues of Journal of Proteome Research, epidermal growth factor — EGFR. It's the drug target. That's a surface protein that spans the membrane. On the interior, there are some 13 sites of phosphorylation, and everyone's concentrating on drug targets for the phosphoryltion motifs. But it turns out that the external domain of EGFR has 11 sites of N-linked glycosylation.
Another example is HER2, one of the markers of breast cancer. Most people don't know HER2 is glycosylated.
Are a lot of people now using this multi-lectin affinity method?
It's interesting. We've been just starting to publish and get the method out. I think with most of the talks I hear now, it's with people using single lectins. The problem with single lectins is that they don't capture everything.
You know SILAC, the labeling technique? Well, there's another acronym, SLAC, and that's serial lectin affinity column. The Japanese use that a lot. You load your samples onto Lectin A, then what doesn't bind, you load onto Lectin B, and what doesn't bind to that you load onto Lectin C. The problem with that is the workflow is a disaster — you've got all these columns to run and fractions up the yazoo — so again, it doesn't fit the clinical arena. And you don't get the cooperativity.
How did your background in Genentech and Thermo Electron help you to develop this technique?
Thermo was really helpful in terms of giving me and my group the ability to perform good mass spectrometry measurements, which is so important with these samples. I recruited one of the best scientists from my proteomics group — Billy Wu — to my lab. And the ion trap is just a great platform for doing proteomics, particularly the new linear ion trap.
You've been studying mainly breast cancer and cardiovascular disease. Are you looking to get into other diseases in the future?
We're looking at getting into other diseases. Psoriasis is one, diabetes is another one, and we're just starting a prostate cancer sample. The other thing is, we would really like to be global. It's a challenging world, but the blood does report basically on all the organs of the body, so that allows you to have a global look. Currently the problem is that people are encouraged to specialize in breast cancer, or prostate cancer, or cardiovascular disease. Let's say you find a marker that really goes up or down in a disease — the problem is you have no idea if it has disease specificity, because you're just looking at that one disease.
So what we'd like to try to do is to look at a broader range of diseases to see if there are particular markers that are specific for one disease or another. For example, it could be that the protein is a marker of inflammation. Then we might see the same protein go up in cardiovascular disease and in cancer. Then we'd say ok, well, this is a more general marker.
The final thing which is a nice bonus is that we will have an opportunity to measure what we call 'silent changes' in biomarkers. That's where there's not a change in the amount of protein, but a change in the post translational modification. Of course you don't get that from expression profiling. In the glyco field, it could be that you get an increase in sialic acid. It's well known in cancer, when the cells are deranged with cancer, glycosylation does change.
So you want to measure 'silent change' as a biomarker in itself?
Well the problem is how do you do it in the global sense, because you can't afford to isolate every glycoprotein and then look. So one of the neat tricks with the multiple lectin columns is that there are two ways to run them. I was talking before about the more high-throughput method where you bind, then put in a cocktail of the three displacers for the lectins, and all the glycoproteins come off together.
The other approach is to first elute with displacer one for the first lectin, then displacer two, then displacer three. Now you start to get a separation into three displacer fractions. We have a lectin that's specific for sialic acid, so if the sialic acid in a group of proteins changes in the course of a disease, you get some shifts. The bulk of the protein may shift from the displacer group for sialic acid lectin to another lectin.
We published a proof-of-concept of this method in the Journal of Chromatography a while back. This approach doesn't give you everything. But it does give you the first clue that there's something going on in what I call 'silent changes.' We're trying to do something, but it's not going to be the full research study on glycosylation. In diagnostics you don't always have to know the full biology — you just have to know that this marker changes reproducibly and has predictive value.
Are you working on commercializing your technology?
We are going through the process of trying to figure out how to commercialize the technology. We believe that mass spectrometry has a future in clinical chemistry. Eventually, clinical chemistry will move away from ELISAs because you may want to have a panel of 50 proteins you're trying to follow. That becomes very difficult in an ELISA format, whereas mass spec is a measurement device for looking at a large numbers of proteins. That's where we figure the field of clinical chemistry is eventually going.