Proteomic Players Pick Plasma

Courtesy of Stephen Naylor, Beyond Genomics Inc. Like any good board game, proteomics requires a blend of strategy and serendipity. But while the former is about winning, the latter is about achievement, and its rules are still being made. Without such rules, it's hard to measure success, which is why nobody at a recent proteomics conference in San Diego could gauge exactly how far the field has advanced since the human genome was sequenced almost two years ago. Nevertheless, several speakers affirmed that proteomicshas established its practical worth, particularly in disease detection and analysis. It's here to stay, the scientists firmly told one another, dismissing earlier concerns that the daunting task of identifying all proteins, their interactions, and functions in a given organism might prove to be insurmountable. One million is a frequently heard estimate of how many different molecules the human proteome holds, and an immediate challenge is to delve deeper, past the most plentiful proteins, in order to grow the known list. A principal effort to do that is being made by the Human Proteome Organization (HUPO), a nonprofit, international effort founded about two years ago in the United States to coordinate human proteomics research in the public and private sectors. HUPO, which boasts affiliates in 15 European and Asian-Oceanian nations, has established a three-pronged goal of consolidating all the national organizations, encouraging scientific and educational activities, and coordinating public proteomics initiatives. Its highest priorities include development of resources and technology, cell models, antibodies to all proteins, and bioinformatics. An initiative has been launched to standardize data worldwide, and a consolidated public database of all human proteins is a long-range objective. MULTIPLE PROTEOMES HUPO's other high priority is to develop research into the proteome "version" in human blood plasma. Unlike the singularity of one genome per person, the complexities of human proteomics begin with a different collection of proteins, or proteome version, for each tissue or fluid. A major reason that HUPO chose plasma is because of its outstanding potential for diagnostics. But Leigh Anderson, founder and CEO of the privately held, nonprofit Plasma Proteome Institute in Washington, DC, says that only 10 assays with new protein analytes were approved by the US Food and Drug Administration between 1992 and 2002. A graph of approvals over time intersects this year at zero, which he labels "clearly counterintuitive." Regulatory groups still must be convinced that multivariate protein markers offer the best diagnostic potential, he says. Yet, to conduct the multiparameter tests that can assess the utility of a protein panel requires statistical skills that extend beyond those of many life scientists, especially in showing the changes to a given protein over a range of diseases compared with controls. Also, even though these marker patterns can reveal much about the effects of different drugs on different diseases, current development costs are high. If technological shortcomings can be overcome, Anderson thinks that the ideal way to measure disease will be to use an individual as his or her own control. Serial measurements with protein panels of a patient's condition could be an excellent diagnostic tool, particularly for chronic disease. Plasma contains the most proteins of any human proteome version. When splicing and other variations are factored in, secreted proteins that function in plasma probably take 40,000 forms. Add to that another half-million forms of tissue proteins from cell leakage, especially following injury. Then there are perhaps 10 million clonal forms of immunoglobulins, plus aberrant secretions from disease states, temporary plasma passengers such as lysosomes, and foreign products from viruses or other sources. However, a literature search up to 2002 showed that only 289 proteins had been observed in plasma, according to Anderson. New detection techniques are responsible for a leap in discoveries over recent months to about 500 plasma proteins, a figure that should double within a year, he predicts. ONE BY ONE Currently, the best way to document new proteins is one at a time. No equivalent of PCR for genes has been devised for proteins, to multiply them for ease of study and manipulation. High-throughput analyses are plagued by artefacts, and variations in collection and handling of plasma samples for high-throughput are troublesome. "It's very important for all of us to get together and agree on a protocol [for sample collection and handling] and on a type of specimen we'll use," says Daniel W. Chan, director of the Biomarker Discovery Center at Johns Hopkins Medical Institutions. "Otherwise, it will be very difficult to compare the technologies." The fundamental challenge for technology is that, unlike the essentially unvarying nature of genes, proteins differ and change--they undergo modifications, form complexes, and degrade. Such deterrents to their analysis are compounded by constant fluxes in their concentrations. When a person stands up or lies down, various proteins dissolve. A viral infection can change normal protein levels. Genetics also plays a role in regulating protein abundance, which can vary within or between individuals, or not much at all. Courtesy of Stephen Naylor, Beyond Genomics Inc. A correlation network graphically illustrating interconnections among proteins (squares), metabolites (circles) and gene transcripts (hexagons). The color of each component reflects changes measured in the experiment. Strongly correlated components cluster together, reflecting common regulatory or functional pathways. Click to view larger version (51K) A main impediment to discovery of the less abundant proteins and new biomarkers is the huge amount of commonly occurring molecules that create a cacophony of data noise within plasma. The differing abundance of components in the proteome spans more than 10 orders of magnitude. "Albumin accounts for 50% of the mass in protein plasma, and the top 10 proteins account for 90% of the mass," notes the University of Michigan's Gilbert S. Omenn, who heads HUPO's plasma proteome project. But even if these most numerous proteins were removed, the "dynamic range" would still be too wide to cost-effectively detect relatively rare proteins. DEEP MINING New N-dimensional software can perform deep mining for uncommon proteins but, because inexpensive experimental platforms for those tools are not yet available, most scientists still rely on platforms such as antibody arrays and mass spectrometry (MS) to measure known proteins for the development of marker panels. MS is more sensitive in protein analysis than immunoassays, but it's not particularly good for quantitation. For separations, two-dimensional gel electrophoresis yields higher resolution than liquid chromatography (LC), but most of the proteins thus identified are abundant and have structural or housekeeping functions, of relatively low interest to development of disease biomarkers. Of course, many markers for various molecular diseases already can be found in the literature, and others are being generated with traditional methods, including immunohistochemistry, ELISA, and immunoblots, as well as more recent genomics techniques such as PCR and cDNA microarrays. In the absence of one method for establishing all biomarkers, these different approaches can collectively increase sensitivity and selectivity. But another problem arises. "How do you integrate all this?" asks Bill Hancock, vice president of proteomics at Thermo Finnigan, a laboratory products and services provider in San Jose, Calif. "You're measuring temporal effects and posttranslational effects, so bringing all this together will be quite a challenge." A cancer researcher, Hancock suggests a shortcoming of the plasma proteome. At some point, tumors become independent of hormones and growth factors, and resistant to drug treatment. This should be associated with molecular markers, but plasma markers generally detect only late-stage disease. Earlier-stage disease markers must be sought in tissue samples. Hancock is constructing a model system for LC/MS-based "direct cell analysis" from tissue rather than plasma. ENTER BIG PHARMA Major pharmaceutical companies already have incorporated proteomics into their technologies. "We are performing something like 6,000 to 10,000 protein identifications from 2-D gels per day," says Hanno Langen, proteomics initiative head at F. Hoffman-La Roche. But these are primarily the abundant proteins. Accordingly, his group is turning its highly automated process toward a now popular method of finding less common proteins: They're reducing the complexity and moving deeper into the proteome by fractionating samples before the application of gels. Others in Big Pharma agree with Langen that proteomics is already beginning to produce better drug targets and safer, more effective compounds. Stanley A. Hefta, director of proteomics for Bristol-Myers Squibb, says his group collects as many as 200,000 MS/MS protein spectra daily. Tape backups of this research carry more data than that produced by all the rest of the company. "About 80% of the work now is in biomarker discovery," he estimates. James R. Ludwig, proteomics research head at Eli Lilly and Co., is bullish but warns against overemphasis: "Proteomics is a tool, not the tool." Hancock provides a guideline for everyone who strives to shape the field's amorphous challenges: "The problem becomes to find the few important things among all the others that aren't important--or not this year, anyway." Steve Bunk (sbunk@the-scientist.com) is a contributing editor.

Proteomic Players Pick Plasma

Interested in reading more?

Become a Member of