Pioneers of Proteomics

Dr. Ruedi Aebersold
Co-founder of the Institutes for Systems Biology in Seattle

In part six of our Pioneers in Proteomics series, Dr. Ruedi Aebersold discusses the application of mass spectrometry to the field of proteomics and the need for sample quality control, data sharing and improvements in instrumentation and reagents.

Dr. Aebersold co-founded the Institutes for Systems Biology in Seattle, Washington in 2000 with Drs. Leroy Hood and Alan Aderem. In November 2004, he assumed an appointment as Professor of Systems Biology, Institute of Biotechnology, ETH-Zurich and Faculty of Natural Sciences, University of Zurich. Dr. Aebersold's research focuses on developing quantitative proteomics methods and technologies and the application of these technologies to further the understanding of the structure, function, and control of complex biological systems.

On proteomics and systems biology

.most frequently systems biology is associated with the integrated analysis and comprehensive analysis of biological systems rather than the analysis of biological systems piece by piece, let's say protein by protein It is a non-reductionist, holistic approach we try to study the biological system.

.the question is how proteomics fits into that systems biology field. One can certainly state that virtually every biological process is essentially built or reliant on the presence of proteins and their function and their structure. So it is inconceivable that one can obtain a comprehensive model of a biological process without knowing which proteins are involved in the process, how do they function, how do they interact with each other, and how do they dynamically change. So this is a very strong impetus or motivation to study proteins in the context of systems biology. And so proteomics in that sense is a very central and integral part of systems biology research. It is by no means the only part. Of course you need computational biology. You need genomics. You need the biological system that is being studied. But proteomics is clearly a central component of systems biology toolbox.

On the challenge of studying the proteome

We would like to analyze every protein that is produced by a cell or by a tissue. Now, in contrast to genomics, proteomics is a much more challenging project because proteins are the molecules that actually carry out essentially all catalytic functions in cells that carry out virtually every biological function that we know of. And these functions of course are strongly regulated. So that means that proteome is the protein composition of a cell is very different from cell to cell. It is different from whether the cell is active, whether a cell is inactive, whether a cell is healthy or whether a cell is from a person with a particular disease. So the proteome is very dynamic and changes over time and changes in response to external influences. So, that means that we need to, we cannot just in proteomics just do the equivalent of the genome project and list all the proteins in that. It is useful but it is only a very small part of the interest in the protein field. What we also need to be able to do is measure quantitative differences between tissues and cells and to relate these profiles back to the physiological or pathological state of the tissue and the cells.

On the challenge of studying low abundance proteins

Biomarker research is an extremely challenging, analytical challenging, challenging analytical task. The samples that we have to analyze are among the most complex samples you can imagine. If you, especially if you try to define biomarkers from human serum, which is of course very easily accessible and has many other very interesting features. All the methods that are being used right now, mass spectrometry or otherwise, are only basically scratching the surface of the proteome. So the proteome in plasma is expected to have a dynamic range of ten or more orders of magnitude. Most current analytical methods that are being used to analyze or find biomarkers will see probably see measurements in the top two to three orders of magnitude. And the question is of course, is there any interesting biomarkers to be found in this top layer of the proteome of serum or do you expect that the useful biomarkers are further down, basically in the less abundant, among the less abundant species. And there are many kind of theoretical considerations that you could make that would indicate that you are more likely to find interesting biomarkers in the lower abundance proteins. And that seems also to be born out by the fact that in this top three, two, three, four order of magnitudes for proteins, very few useful biomarkers have been discovered to date

On the need for throughput

The obstacles are that we are currently in the proteomics field using as the main method mass spectrometry analysis and mass spectrometry is dealing with relatively short fragments of a protein. So, a typical experiment is that we isolate a proteome out of the cell or a tissue. We then digest these proteins with proteases into smaller fragments, referred to as peptides and then these peptides are being analyzed in the mass spectrometer. So, of course, if we start out with something like ten thousand proteins, and if each protein generates in the order of fifty to a hundred such peptide fragments, then the number of peptides we have to deal with is rather large. And so, one of the technical difficulties we are facing is that we don't have sufficient throughput in the analytical machinery to be able to analyze each one of these potentially hundreds of thousands or even millions of peptides if we deal with the case of a human tissue.

So we need to come up with a different strategy.

So what we are trying to work out now is a strategy which is very much following the strategy of the genomic scientists where you first would map out the space in the genomic sciences that was the genomic sequence and then you learn from that and you devise assays which take advantage of this map. And, this information can be used to devise very fast targeted assays to probe the space that you initially mapped out.

So we want to get rid of all this redundancy of the shotgun approach. So these strategies are fairly far ahead. It requires changes in way the samples are prepared. It requires changes in the way the data are analyzed. It requires build up of databases which essentially contain the proteomic map and it requires changes in the way the mass spectrometers are being operated and driven but this, all these things are really falling into place exactly now at this time. So, I think there is very, very, another leap in performance can be expected within the next year for sure.

On the role of informatics

In the field of proteomics, informatics is currently a tremendous challenge and it is a tremendous challenge at various levels. The situation is that every day gigabytes or terabytes of data are being generated, distributed in many laboratories and that these data are somewhere stored in some format that is usually proprietary because it is of course given by the type of instrument that is used to generate the data. The data generated in a format that really is not publicly known and then the data is stored somewhere on a hard drive in the respective laboratory. So collectively as a proteomics research community, a lot of data are being generated and the challenge is to figure out how to take the data from the individual experiments and hard drives into a domain where they are useful for the community as a whole and useful in a sense that they can be, that they are organized, have concrete quality parameters associated with the data and can be communicated and visible and accessible.

I think a lot of work remains to be done there but it is an important step to not just do proteomic data collection analysis in individual laboratories but also go the extra step to take the data out of the mutual labs and put them into generally accessible, well curated and well maintained databases.

On the need for high quality clinical samples

So the quality of the specimens is of course critical and it is even more critical in cases where we want to compare, to compare the analysis but let's say between patients.

.some samples are collected in clinical center A, some samples are collected in clinical center B by a completely different person. Samples may be surgically removed samples or the surgeon may have other things to do than worry a lot about the sample. The surgeon may be cutting out tissue and then the tissue may be standing around for a minute, maybe for a half an hour, and so the samples are very inconsistent. The methods to collect them and the methods to store them are inconsistent. And so that of course creates, downstream, an issue if you want to compare molecular pattern data from these samples because you need to make sure that you are not just measuring effects of the way the sample is treated but that you are measuring effects that tell you something about the state of the sample as related to disease or health. So it is very critical issue.

On the transition to clinical applications

.it is certainly not inconceivable that in the timeframe of ten years, one would be able to develop biomarker kits that define the well being or the state of a human being from tissue to tissue specific, tissue specific and disease specific matter. That one could take blood tests and then look for the presence of specific indicators of disease on a regular basis, let's say every half year with relatively modestly priced or very cheap high-throughput assays. One could monitor the health state of an individual over time and detect changes early so that one could with current or with therapeutic activities or regiments available at the time, hopefully revert the disease progression at a very early stage of its occurrence. I think it is a realistic goal if one goes about it systematically.

On the use of biomarkers in the clinical setting

We assume that in a normal living cell or tissue certain biological processes or physiological processes are active at the molecular level which are required to carry out normal physiology of that cell on a tissue. We furthermore assume that if the, if the person comes down with a disease, let's say cancer, that some of these networks of interacting processes and molecules are somehow perturbed.

.if we understand what the difference is between the healthy and diseased tissue, there is a realistic expectation that pharmacologic intervention targets could be identified to revert the physiology of the diseased tissue maybe back to the physiology of the healthy tissue, which basically means the disease would be pharmacologically interfered with, ideally cured.

Another area where proteomics is expected to make contributions is in the area of diagnostics. So, clearly it is useful for a physician to know very precisely what the disease that a patient will have that the physician is faced with - diagnosis today is not really carried out on the molecular level. It is dependent on the skill of the physician to integrate various signs or cues into a diagnosis. There is no doubt that the diagnosis would be much more precise if the physician had an array of molecular signatures and molecular cues at his disposable.

These biomarkers are also extremely useful for following a treatment. It is not just to detect whether a disease is present and what stage it is present but they are also expected to be extremely useful to determine whether a treatment that is applied by physicians is actually working, whether the dosage of a drug is correct, whether another regimen would have to be applied.