Designing a More Accurate Protein Census

Graphic: Leza BerardoneApplied Biosystems will soon begin testing new mass spectrometry machines designed to identify proteins in as many as 1,000 samples per hour. For the machines to work as planned, however, each sample must be prefractionated down to just a handful of proteins, according to Stephen A. Martin, director of the company's proteomics research center in Framingham, Mass. This example suggests how the slower protocols leading to mass spec are as important to the progress of proteomics as are mass spec's higher-throughput capabilities. These protocols have a serious limitation. Just as the Census Bureau's methods lead to the undercounting of certain population groups, the protocols lead researchers to under-identify whole classes of proteins. But clever strategies to attack this problem are starting to appear. Two papers published this month unveil chemistry-based methods that help recognize phosphorylated proteins.1,2 And a paper published last month presented a process to uncover many traditionally overlooked proteins.3 Phosphorylation is a central event in many biological processes, so why are researchers only now finding ways to study it proteomically? "One now has fast, effective mass spectrometric ways for identifying proteins that really only matured in the last few years," says Brian T. Chait, director of the mass spectrometry and gaseous ion chemistry lab at Rockefeller University and senior author of one of this month's papers.1 Ruedi Aebersold, senior author of the other paper2 and a professor at the Institute for Systems Biology in Seattle, explains that a "biochemical mindset" has characterized phosphorylation studies. Scientists identify phosphoproteins one at a time by slowly purifying each one and tracking it with radioactive phosphorus-32 (32P). In contrast, Aebersold recalls, "We tried to apply a proteomics mindset to the problem. We said we would like to establish profiles of protein phosphorylation in samples consisting of many different proteins without having to first purify each protein." Another goal was to avoid 32P, which can't be used in many human and animal studies. The application of mass spec, however, presented two problems. Before being subjected to mass spec, complex protein samples are digested into dozens to thousands of fragments, only a few of which might be phosphorylated. Mass spec often fails to detect scarce molecules. And attachment of a phosphate group to a peptide generally lowers the intensity of the peptide's spectrometric peaks. As a result, the spectra of unknown phosphorylated fragments are difficult to recognize. One solution to these problems is to enrich a sample for phosphorylated proteins or peptides, thereby cutting down on the background "noise" level. That's what Chait's and Aebersold's labs did. Chait's postdoc, Yoshiya Oda, replaced phosphate groups on serines and threonines with a biotin-containing group. Avidin was then used to isolate proteins and peptides that had been phosphorylated on these two residues. Aebersold's postdoc, Huilin Zhou, took a different tack to enrich for peptides phosphorylated on serines, threonines, or tyrosines. (Chait notes that phosphotyrosine-containing peptides can also be isolated by good antibody-based protocols.) Zhou attached a small molecular chain to the phosphate groups. This chain ended in a free sulfhydryl group, which stuck to modified glass beads as the peptides filtered through them. An acid wash freed the peptides from the beads. "This method, as powerful as it is, will find only limited penetrance unless it can be supported by an instrument that performs these steps," says Aebersold. Consequently, his lab is now trying to automate the method, using off-the-shelf parts. Chait is less interested in automation, explaining that his procedure yields "a huge amount of information from a single experiment" and thus doesn't have to be performed repeatedly by a machine. Instead, his lab is refining the technique by, for instance, testing tags other than biotin. The lab is also trying to quantify phosphorylation by applying reagents that contain deuterium. The benefits of refining a proteomics method were apparent in last month's paper, whose senior author was John R. Yates III, an associate professor of cell biology at Scripps Research Institute.3 Two years ago, Yates, then at the University of Washington, described a process for analyzing protein complexes. Proteins in the yeast ribosome were separated by two-dimensional liquid chromatography (LC) according to charge and hydrophobicity and then subjected to mass spec.4 Researchers loaded samples of digested proteins onto an LC column, and an automated system performed the rest of the analysis. After optimizing the solvents and improving sample preparation, Yates and his two postdocs, Michael P. Washburn and Dirk Wolters, applied the process, memorably titled MudPIT (for multidimensional protein identification technology), to a yeast whole-cell lysate. In a 24-hour run, the process identified 1,484 proteins in the approximately 350 micrograms of protein loaded onto the system.3 Among those proteins were many in categories often overlooked when two-dimensional gel electrophoresis precedes mass spec: proteins that are very large, highly basic, membrane-bound, or present at very low levels. Yates says that several companies are now trying to produce systems similar to his, which consists of commercially available parts. His recent paper states that the most yeast proteins found in a single previous study was 279. "One piece of information that everybody is missing is how many proteins really should be in a yeast cell," he observes, adding that estimates based on gene-expression studies range from 2,500 to 5,000. By combining his process with quantitative proteomic methods, Yates claims to have scoped out about 2,000 proteins. The process "can detect proteins down to about 100 copies per cell," he notes. "Realistically, we probably need to be down to the level of about 10 copies per cell on the assumption that if you have 10 copies in a cell, they're actually doing something of note." Douglas Steinberg (dougste@attglobal.net) is a freelance writer in New York. References 1. Y. Oda et al., "Enrichment analysis of phosphorylated proteins as a tool for probing the phosphoproteome," Nature Biotechnology, 19:379-82, April 2001. 2. H. Zhou et al., "A systematic approach to the analysis of protein phosphorylation," Nature Biotechnology, 19:375-8, April 2001. 3. M.P. Washburn et al., "Large-scale analysis of the yeast proteome by means of multidimensional protein identification technology," Nature Biotechnology, 19:242-7, March 2001. 4. A.J. Link et al., "Direct analysis of protein complexes using mass spectrometry," Nature Biotechnology, 17:676-82, 1999.

Designing a More Accurate Protein Census

Interested in reading more?

Become a Member of