At A Glance
Name: Timothy Wells
Position: Head of research, Serono, Geneva, Switzerland, since 1997
Head of Biochemistry and Immunology, Glaxo Wellcome, Geneva, 1990-97
Molecular Enzymology Lab, SmithKline Beecham, 1987-1990
PhD, Imperial College, London, 1984-1987. Studied enzyme catalysis with Alan Fersht.
How did you get to be head of research of Serono?
I have a PhD in protein biochemistry from the University of London. After that I worked at SmithKline for three years on suicide inhibition of enzymes. Then I came here [to Geneva] in 1990 to work for Glaxo Wellcome as the head of protein chemistry. What we did there is a lot of classical purification of proteins plus purifications of recombinant proteins for structural studies and other things. At that stage, the biggest problem was that crystallographers couldn’t get enough protein of high quality. Over the next few years we solved three or four different structures, which was the first time Glaxo had been doing crystal structures. As a result of that, we ended up doing the structures of some proteins involved in the immune system, and after that I got viewed more as being an immunologist, though I am actually a protein chemist at heart.
Currently, I am head of research at Serono, the largest European biotech company, and the third largest in the world after Amgen and Genentech. We have a big project here on what we call ‘functional genomics,’ which you could argue is actually proteomics. It is expressing all secreted proteins in the human genome and trying to find out what they do, and also using protein structures and structure simulation techniques to try and find new proteins. We are trying to find new cytokines, growth factors, or hormones. Beyond that, we also do a little bit of intracellular target identification. We have a very good medicinal chemistry group, so we look for ways to make small molecules that upset cytokine signaling cascades.
How do you find all the secreted proteins in the human genome?
By looking at protein sequences that have classical and not-so-classical signal sequences. We have done a lot of bioinformatics work, rewriting neural networks for detecting signal sequences. We also made the assumption that what matters often in protein chemistry is the overall fold of the protein. If you look at a lot of growth factors, they actually fold into the same three-dimensional shape, even though their protein sequence identity can be as low as 15 percent. We have done a lot of work internally on that but we also had a very fruitful collaboration with a company called Inpharmatica in London.
We have also done a little bit of in-house work using classical HPLC or 2D gel separations of secreted proteins to try and pull out new proteins. That will tend to give us much more of the processed forms. Also, we just signed a deal with GeneProt. They have some lovely high-throughput mass spec technology, which they have been using to identify largely secreted proteins in biological fluids.
What is the next step after identifying the proteins?
We are trying to assemble a collection of full-length cDNAs for all secreted proteins. We have got 1,200-1,300 at the moment, and we will have a couple of thousand soon. Most people who have larger collections have actually not fully characterized them, so there is a lot of redundancy.
We took the decision to express those proteins in a mammalian cell system, and then purify them in a highly parallel manner, 20 at a time, not using standard equipment.
After that, we test them in various cell biology experiments — up to 100 different cell-based assays which are relevant to our disease areas. We are also testing GeneProt’s proteins and peptides in our assays. What we are looking for is, for example, ‘what are the proteins in multiple sclerosis which control new growth, remyelination, demyelination?’ It’s important that everybody remembers that what we are looking for is the one or two really exciting pieces of data that come out. Because the danger with all of these high-throughput experiments is that people get carried away by numbers. In the end, we are looking for 10 or 20 proteins which have an interesting activity in vivo, and a novel activity.
How do you get from cell biology assays to an application in a certain disease area?
Within the disease areas, from our understanding of biology, we know for example which cell populations we are trying to turn off and which ones we are trying to turn on. We have cell biological assays that look at inflammation, neurodegeneration, or, on the reproductive side, what controls female reproductivity.
But the other question is, how are we going to find the novel activities for well-known proteins? If you imagine that we sell a lot of beta interferon as a drug for multiple sclerosis, beta interferon was originally described as an antiviral agent. A lot of what we are trying to do is develop some quick ways of understanding what proteins will do in vivo. One way to do that is test how a mouse responds to having the protein, standard physiology. The other one is looking at responses from transgenic animals. We recently signed a deal with Regeneron to look at knockout mice, for example.
How successful have you been in making recombinant proteins?
Of the 1,200 cDNAs, we have expressed and purified about 700 or 800. With HEK [human embryonic kidney cells], it’s been relatively easy to express secreted proteins. That hides the real story, which is to say, one-third of them express really well, one-third of them express moderately well, and one-third of them express with difficulty. It seems that for secreted proteins, E.coli is not a very good way forward. If you talk to people who are doing similar experiments, then high-throughput expression in mammalian cells, using transient expression, seems to be working really well.
Why did you decide to partner with GeneProt instead of building up your mass spec expertise in-house?
We do have our own in-house proteomics expertise. However, one reason to parter with GeneProt is, they have a world class proteomics facility two miles down the road. Second, we have had a number of interactions with the people who work there throughout the year. The strength that we bring to the collaboration is the set of biology that we have. We don’t cover all biology, but it’s likely that some of their proteins and peptides will be interesting in some of our assays. Also, there is a difference in scale. GeneProt in Geneva has currently about 80 or 100 people; our total proteomics group — if you take the sort of classical definition of proteomics — is only six people. The expression and cell testing [groups are] way, way bigger than that, probably 70 or 80 people. Basically, we are interested in talking to anybody who is identifying interesting proteins.
Where do you see a need for new technologies?
I like the idea of protein chips. I think one thing protein chemistry failed to sell to the outside world is the fact that proteins are far more interesting than DNA or RNA, because you can actually ask them questions about what they do. If you immobilized protein collections on a chip, then you would be able to ask the proteins, ‘which of you binds ligand such, which of you catalyzes this reaction?’ I think protein chips will turn out to be much more interesting than DNA chips once we get over some of the technical problems. They will probably produce a lot less data, but it will be of higher [value].
Furthermore, if you look at what happened in the last 10 years, the real breakthroughs have been the improvements in sensitivity in mass spectrometry. I, as a non-mass spec person, don’t see any reason why those improvements in sensitivity should stop happening. The question is miniaturizing the detection systems so that we can detect proteins in much, much smaller samples.
Are you actually using protein chips in house?
We are using them, but I am not sure we are using them to full capacity. I think that certainly over the next year, we will be moving a lot more out of nucleotide chips and into protein chips. It always depends on how well the first six months or so go with protein chips.
Is what you do at Serono really proteomics?
Most of the meetings about proteomics are organized by either 2D gel people or mass spec people. I used to get into trouble a lot at these meetings. People would say, ‘what is Serono’s proteomics effort?’ and by that they would mean ‘how many people do you have working on 2D gels and LC-MS/MS systems?’ And the answer to that is, six people. But actually, we are one of the world’s leading proteomics companies — half to two thirds of our job is finding new proteins and new activities for these proteins, and taking those to the clinic, and seeing that they work.
For a while, I almost seemed to have to accept that when I talk about proteomics, I was talking about something completely different from everybody else, which is why we didn’t go to the proteomics meetings. Now we are starting to go, because they are becoming more relevant to us. More and more now, people are accepting that proteomics is much more than just identifying the sequence of the protein, [and that] the real question is to understand their function.
Where do you see proteomics going?
What we really need to move towards is more technology for understanding function. I think protein chips are going in that direction. It would be interesting to see what happens when protein chips, for example, can be mixed with cell biology. There are all sorts of possibilities.