This effort has made proteomics the newest buzzword in molecular biology, with entire companies and academic institutions springing up to take on the challenge. At the leading edge of this endeavor is Swiss biologist Ruedi Aebersold of the Institute for Systems Biology in Seattle, Washington. Last year, in one bimonthly update of the Hot Papers database, Aebersold fielded six hot reports published over the preceding two years, and in the last decade he has published 25 papers that have each been cited more than 100 times. He is a double threat, with as many high-impact articles characterizing the nature of proteins themselves as articles describing the technologies that make the elucidation of protein, sequence, structure, and function possible. His Nature paper, "Molecular characterization of mitochondrial apoptosis-inducing factor," has received more than 900 citations in barely five years, while his seminal paper "Quantitative analysis of complex protein mixtures using isotope-coded affinity tags," published in Nature Biotechnology, has racked up more than 600 (see table).

Aebersold, 50, received his undergraduate degree in biology from the University of Basel in Switzerland in 1979 and his Ph.D. in biology from the same institution four years later. Through 1988, he worked in Leroy Hood’s laboratory at the California Institute of Technology, and then spent five years at the University of British Columbia in Vancouver as an assistant professor and senior investigator at the Biomedical Research Center. In 1993, he joined the University of Washington, where he was a full professor and associate director of the Science and Technology Center for Molecular Biotechnology until 2000, when he became a co-founder of the Institute for Systems Biology. Later in 2004 he will move to the Swiss Federal Institute of Technology in Zurich.

Aebersold spoke to Science Watch from his office at the ISB.

  You first started sequencing proteins 20 years ago at Caltech. What did you see as your goal at the time?

The goal was always to sequence one protein that was isolated to virtual homogeneity and to do that at ever-increasing sensitivity and precision. This was the mindset that always applied. You would isolate a protein, purify it, and then sequence it. And we learned to do it better and better. Eventually, around 1990, mass spectrometry became really useful for the analysis of proteins. That was due to a number of innovations—in particular, the development of ionization sources, which was actually awarded the Nobel Prize in chemistry in 2002. So all of a sudden we could generically ionize relatively large molecules such as peptides and proteins. There was also a convergence at the time of various advances having to do with our ability to analyze data: a convergence of computer science, engineering, physical chemistry, and instrumentation.

  For the uninitiated, could you explain how mass spectrometry works and about the important about ionization?

The simplest possible way to think of mass spectrometry is as a balance that measures the mass of a molecule in an ionized form. So you have to attach a proton or an electron to your protein or peptide, then bring it into a gas phase in a high vacuum. This is what this ionization method allowed us to do. You can’t read out the sequence of peptide from the intact ion, but advanced instruments developed during that time in different laboratories gave us the ability to fragment a peptide ion. Basically, you accelerate the protein or peptide ion and run it into a wall of glass. This shatters the protein or peptide into fragments, then you use the mass spectrometer to measure all the masses of these fragments. From these masses we can then figure out the amino acid sequence of the peptide.

  How do you go from masses to protein sequences?

It’s via a somewhat lucky coincidence that these peptide ions don’t fragment at random. They always fragment with a certain order, and that order is reasonably well understood. It’s not like when you throw a piece of glass against the wall and it fragments randomly. If you were to fragment the same peptide 1,000 times, you would get essentially the same fragments each time. Once you understand the rules of this fragmentation process, you can start to piece together the amino acid sequence of the peptide from the masses of the fragment ions.

  Can you give us an idea of how protein sequencing alone has advanced since the mid-1980s when you started at this game?

If you were to sequence a protein using the chemical methods developed in the 1980s, it would take about a day to generate a sequence. With the mass spectrometer and today’s technology, we can do about one per second. Once we had the mass spectrometry methods better established, we realized we didn’t have to sequence proteins one a time but could work with large numbers of proteins simultaneously.

  When did the idea of sequencing every protein in the human body start percolating into molecular biology as a potential reality?

High-Impact Papers by Rusdi Aebersold, Published Since 1999 (Ranked by total citations) Rank Paper Citations 1  S.A. Susin, et al., "Molecular characterization of mitochondrial apoptosis-inducing factor," Nature, 397(6718): 441-6, 1999. 950 2  S.P. Gygi, et al., "Quantitative analysis of complex protein mixtures using isotope-coded affinity tags," Nature Biotechnology, 17(10): 994-9, 1999. 615 3  S.P. Gygi, et al., "Correlation between protein and mRNA abundance in yeast," Mol. Cell. Biol., 19(3): 1720-30, 1999. 531 4  T. Ideker, et al., "Integrated genomic and proteomic analyses of a systematically perturbed metabolic network," Science, 292(5518): 929-34, 2001. 295 5 S.P. Gygi, et al., "Evaluation of two-dimensional gel electrophoresis-based proteome analysis technology," Proc. Natl. Acad. Sci. USA, 97(17): 9390-5, 2000. 272 SOURCE: Thomson ISI Web of Science®

That actually goes way back to the 1970s. It was an idea a little ahead of its time, pushed by Leigh Anderson and Norman Anderson, father and son. They were advocating a concept they called the "human protein index," and they wanted to essentially establish a database that would describe every human protein. It didn’t happen, simply because it wasn’t feasible at the time. It still hasn’t happened, but now it’s getting feasible.

  Does it help to have the human DNA sequence available?

Absolutely. Now, rather than having to do a de novo sequence for every protein every time, you can take the information obtained from the mass spectrometer and correlate it with the human-genome database, using database search algorithms. Without these sequence databases, the field of proteomics would still be in the dark ages.

  There seems to be a wide spectrum of definitions for the term "proteomics." How do you define it?

All of the definitions floating around center on one idea, which is to identify and analyze all the proteins expressed by a cell or a tissue. For me, the key to proteomics is the systematic analysis of proteins present in a sample. Rather than picking out one protein or a few, you want to systematically measure and analyze all proteins. In our group we want to do these measurements quantitatively. We can then use the quantitative results to detect the differences in various proteomes—for instance, between the proteome of a healthy cell and a diseased cell, or an activated cell and a resting cell. And to do that we need very accurate quantification of all the proteins in the cell.

  Where do isotope-coded affinity tags come into this?

Well, with the mass spectrometer we are very good at identifying the proteins present and we can do it in a very short time, maybe a few seconds per protein. What we’re not as good at is providing quantitative information on this protein, unless we apply a certain trick. This trick is the isotope-coded affinity tag, and it’s basically what my colleagues and I contributed to the field. The idea is to generate two molecules that are chemically the same but which can be differentiated based on their mass. This is easily done by incorporating heavy stable isotopes in one molecule and light isotopes in the other. So we generate two molecules that are effectively indistinguishable by their polarity or anything else, except their mass. Then, since the molecules are in every other respect identical, we can assume that the signal detected from the heavy or light form of the particular molecule in the mass spectrometer is a true representation of its abundance. This trick is called stable isotope dilution, and it allows us to turn the mass spectrometer into a quantitative device. And that gives us a generic method to study the change in protein profile in proteomes of different cells or tissues.

  What are the major bottlenecks hindering the advance of proteomics?

Right now every technique has its bottlenecks, or its limitations. One obvious one at the moment is in our capacity to analyze data. We can now generate huge amounts of data, and currently there is an enormous challenge to figure out how to actually analyze this data and generate real biological insights. That’s not to say that we don’t need better and faster ways to actually acquire the data, but in practical terms the major bottlenecks today are related to data analysis. The whole idea behind the Institute for Systems Biology is to create an environment where computer scientists and biologists and the people who collect data can work closely together, so they can develop the necessary analytical tools that will help interpret the data and help put them in biological context. I think that’s absolutely necessary and, actually, it’s been very successful so far.

  What do you think are realistic accomplishments that we can expect from this proteomics/systems-biology effort in the short term, say, the next five years?

Well, in five years I hope that we will be able to very easily and very quickly and with very good precision analyze proteomes of any complexity. I think that within five years, for instance, we will be able to do clinical studies using proteomic approaches. Say you go out and collect blood samples or sera samples or spinal fluid samples or tissue samples from biopsies of a large number of patients and controls, and then you go through the samples and extract new insights into which proteins might be of diagnostic value for a particular disease. I think that’s entirely feasible and will probably happen within the next few years. That’s a near-term goal and it is achievable.

Another goal that I think is not quite so near-term but is certainly achievable or, at least, worth the attempt, is to use proteomics to eventually get enough information that we can understand how different biological processes are controlled within the cell and how they interconnect with each other. Eventually I think we will have sufficient data that we can construct computer models of how cells operate.

  Are you talking about one pathway within a cell, or all pathways?

All pathways. That is really an explicit goal of proteomics. That’s what we mean when we talk about systems biology. The idea is to go beyond one pathway to study the interconnectedness of different pathways concurrently active in a cell. There is lot of evidence indicating that cells don’t operate in isolated pathways and processes, but that all processes are interlinked and co-regulated. For example, it makes absolutely no sense for a cell to start dividing if it doesn’t have sufficient energy to do that, and doesn’t have sufficient molecules, say, to build new DNA strands, which are a requirement for cell division. So cell division is coupled to availability of energy, to availability of precursors to build DNA, proteins, and other biopolymers, etc. So the cell evolves checkpoints to make sure a particular status has been reached, and only then will it proceed. This is a clear indication that processes that we traditionally studied as isolated, independent pathways are actually all connected and interdependent in a living cell. That’s what systems biology and proteomics and also genomics are trying to address: Can we get a global overview that shows us all this interconnectivity and dependency?

The problem, of course, is the data-analysis challenge that I mentioned, which is enormous. Then there are the specific challenges solely related to proteins. So far I’ve just discussed the analysis of proteins as polypeptides. But proteins are actually more than that. The essence of a protein is a polypeptide chain. But these are modified and processed in complicated ways as they are synthesized. We are just starting to see techniques that allow us to also systematically study the processing and modification events to which proteins are subjected—glycosylation, phosphorylation, lipid attachments to the proteins. More than 200 different types of modifications have been described in the literature. It will be extremely important to study these modifications on something like a proteomic scale and in a quantitative way. How do these modifications change over the lifetime of the proteins and how do they affect their function? Methods to ask such questions are just beginning to appear.

Then proteins interact with other molecules, with DNA, with other proteins and lipids. There is a huge amount of important information contained in these interaction maps and protein networks, and, again, techniques to systematically analyze these interactions are just beginning to be useful or developed. And so there are enormous technical challenges that will have to be solved in this field.

  It sounds like the more you achieve, the greater the data-analysis problems you create. Is that accurate?

Yes, data analysis will get infinitely more complicated. Somehow we have to figure out how to take the data that result from all these different types of measurements and integrate them together, and that’s largely an unsolved problem. On the other hand, the most fun thing about this work is that we are constantly doing things that have never been done before or that we never thought were possible.