For several years, as knowledge of the human genetic complement expanded, promises were made with regard to our understanding of how cells function normally and how disease progresses when they do not.

To a large extent, this promise was based on genomic analysis using DNA or gene chips, tools that distinguish the slightest changes in gene sequence or RNA expression patterns. But as researchers learned more about the biochemistry of metabolism and disease, they began to realize that this promise would go largely unrealized if they relied on gene chips alone.

Every cell in a population of individuals contains largely the same genes, and although the expression of these genes may change with time or conditions, the main participant in cellular metabolism is not nucleic acid but protein. It only makes sense, therefore, to look to proteins to convey messages about human health and disease. To that end, a new industry has followed on the heels of DNA microarrays, and we are entering the era of protein chips and antibody arrays.

Chip chemistry
The first step in developing a protein chip is the attachment of the proteins or antibodies to a solid support, and this can be accomplished by using a variety of surface chemistries (see also “A binding proposition”). Some chips use ionic charge or hydrophobicity to bind the proteins of interest. Such is the case with some of the protein chips available from Biacore International (Uppsala, Sweden, www.biacore.com) and Ciphergen Biosystems (Fremont, CA, www.ciphergen.com).

Unfortunately, although these chips are useful for global protein studies, they are not very specific. For that reason, companies such as Prolinx (Bothell, WA, www.prolinxinc.com) have developed affinity systems such that a particular moiety on the chip binds directly to a partner attached to the protein of interest. In the Prolinx chip, the chemical pair is phenylboronic acid and salicylhydroxamic acid; other pairs include biotin and streptavidin, and nickel and hexahistidine.

For even more specific binding, however, nothing can compete with the use of antibodies or antibody mimics. Within the range of antibodies, monoclonals offer the tightest binding; polyclonals offer diversity of binding, which may become critical if you still want your bound protein to function; and single-chain or phage display molecules offer greater stability. Following in the single-chain vein, antibody mimics are small proteins that have been engineered to interact with their partner protein. These include the trinectins of Phylos (Lexington, MA, www.phylos.com) and the affibodies of Affibody (Stockholm, www.affibody.com).

Detection

Figure 1. The SELDI process

Regardless of which attachment method is used, detecting interactions between proteins and ligands or other proteins is the name of the game. To that end, various technologies have been established to identify points of interaction, some of which are derivative of DNA chip technologies. Companies such as Ciphergen and LumiCyte (Fremont, CA, www.lumicyte.com) rely on surface-enhanced laser desorption ionization-mass spectrometry (SELDI-MS), whereby laser light desorbs and ionizes proteins bound to a chip (Figure 1). The ionized molecules then enter a mass spectrometer, where their composition and mass are detected (www.seldi.org). With SELDI-MS, the bound proteins do not have to be labeled.

Another technique that does not require protein labeling is surface plasmon resonance (SPR), as exemplified by the system developed by Biacore. In chip-based SPR, target proteins are bound to the chip surface, and a baseline refractive index is established. A mixture of potential ligands or proteins is then passed over the chip, and interactions between the solution-phase molecules and the chip-bound proteins are measured as a change in the refractive index.

Alternatively, many laboratories rely on the use of fluorescent or radioisotope tags. The tag can be attached directly to the proteins in the solution being tested or can be a component of a secondary antibody such as those used in Western analysis.

To a large extent, though, how the proteins are detected depends on what instruments are available to a researcher and, even more so, what application the researcher has in mind.

Diagnostics
One of the prime areas in which researchers are looking to apply protein chips is the development of diagnostic tools. Unlike the use of MS for diagnosis and biomarker discovery (see “Diagnosing newborns”), the application of protein chips does not necessarily require the financial investment inherent in the purchase of a mass spectrometer, making it an attractive alternative. Furthermore, improvements in detection limit (approaching femtomolar levels) and their ability to pull biomarkers directly from complex mixtures without resorting to preliminary purification have established protein chips as a hot area of research.

In February, Emanuel Petricoin, a researcher at the U.S. FDA (Bethesda, MD), and his colleagues presented their experiences in using serum protein patterns to identify individuals with ovarian cancer (1). Petricoin’s group incubated Ciphergen’s C16 hydrophobic interaction protein chips with serum from women with and without ovarian cancer and elucidated patterns of proteins that were specific to the cancer. These patterns then served as a template against which blind serum samples were tested. The pattern matching correctly classified 63 of 66 noncancerous or benign samples and all 50 cancerous samples.

“After proper validation, serum proteomic pattern analysis might ultimately be applied in medical screening clinics as a supplement to diagnostic workup and assessment,” Petricoin wrote. “A negative value, if the sensitivity remains at 100% on further trials, could be used for reassurance, whereas a positive value might be sufficient to warrant further investigation.”

While Petricoin’s group only looked at protein patterns, other groups have been vocal about the need to identify the actual proteins associated with the disease. For some of these people, protein chips present an ideal method.

“Take prostate cancer, where there is only the prostate-specific antigen (PSA), which is secreted in sera,” says Moncef Jendoubi, founder of Milagen, Inc. (Richmond, CA, www.milagen.com). “The PSA is secreted as a result of the enlargement of the prostate as we age and is not indicative of cancer. We applied almost 15,000 antibodies, each recognizing a separate protein, and identified, in individuals suffering from prostate cancer, 164 proteins that are not PSA.”

While Milagen is making strides with its antibody program, other companies are entering the diagnostic protein chip market. In early 2001, Large Scale Biology Corp. (LSBC, Germantown, MD, www.lsbc.com) announced a collaborative agreement with Biosite Diagnostics (San Diego, CA, www.biosite.com). By combining LSBC’s Human Protein Index with Biosite’s high-throughput Omniclonal phage display technology, the companies hope to generate antibody arrays against various human diseases.

Similarly, this past January, NextGen Sciences (Alconbury, U.K., www.nextgensciences.com) initiated a collaborative project with Cytomyx (Cambridge, U.K., www.cytomyx.com) and Carlos Caldas of the University of Cambridge (U.K.) to generate a range of protein chips targeted at breast cancer.

“My research group has been studying breast cancer at the gene and mRNA levels using human biopsy cell lines as well as primary tumors with linked, appropriately anonymized clinical information,” says Caldas. “This initiative allows us to access new technology in protein expression analysis and to combine this with the genomics and transcriptomics information we are already gathering.”

Prognosis and pharmacoproteomics
But diagnosis is a retroactive event, and many researchers believe that the population would be better served with a tool that allows disease prognosis.

“You can imagine having an antibody chip for predictive medicine,” says Jendoubi. “Once you have identified a number of proteins secreted in sera or urine, you can segregate the proteins by which are linked to early disease, the onset of metastasis, who does and does not tolerate treatment, toxic effects, and who is prone to resistance or relapse.”

Fundamentally, you establish a pharmacoproteomic profile of an individual. Like pharmacogenomics, which allows researchers and clinicians to predict the response of an individual to drug treatment on the basis of his or her genetic profile, the evolving field of pharmacoproteomics allows drug developers and clinicians to further subdivide the treated population.

“Like others, we believe that the availability of protein biochips designed to evaluate toxicity and efficacy may dramatically change the research process,” says Gunars Valkirs, vice president and chief technical officer of Biosite Diagnostics.

Proteomics
Although much attention is being focused on the clinical application of protein chips, future developments will largely hinge on their use in proteomics. As more proteins are identified and become available for use in various applications, pressure will mount to determine their roles in the cell and how they relate to other partners, whether small molecules, nucleic acids, or other proteins.

In the construction of protein interaction maps, protein chips offer advantages over the more traditional methods. In yeast two-hybrid screening, one protein is fused to the DNA-binding domain of a transcription factor (the bait), and the proteins in the mixture to be analyzed are fused to the transcriptional activation domain of another protein (the prey). When the bait and prey interact in vivo, they activate the transcription of a reporter gene.

One of the main advantages of a protein chip is that whereas the researcher has little control over the conditions under which the fusion proteins interact in the yeast cell, there is ample opportunity to modify the reaction conditions on a chip. This effectively eliminates many of the false positive interactions that are inherent in the yeast two-hybrid system. Furthermore, although the expression of a natural transcription factor can trigger a reaction in the yeast system, this same protein is just another spot on the protein chip, again limiting the risk of false positives.

Figure 2. Bound to deliver. Researchers tested chips carrying histidine- and GST-tagged yeast proteins (top) to identify polypeptides that bind calmodulin and phospholipids (bottom). (Adapted with permission from Reference 3.)

Recently, Michael Snyder and his colleagues at Yale University and North Carolina State University (Raleigh) developed a yeast protein chip by tagging the 5800 genes of the microbe with sequences encoding six histidine residues and attached the resulting proteins to a Ni⁺-coated slide (2). The researchers then probed the array with calmodulin, a Ca²⁺-binding protein involved in calcium-regulated cellular processes. Beyond identifying six of the known calmodulin-binding proteins in yeast, the array identified another 33 potential partners (Figure 2).

Protein arrays also facilitate the identification or characterization of the enzymatic activity of proteins. This is potentially useful to the pharmaceutical industry, in which companies are searching for small molecules that bind to specific proteins in the proteome or interrupt the protein’s activity. In 2000, Gavin MacBeath and Stuart Schreiber of Harvard University (Cambridge, MA) described their experiences in developing an array of peptides and determining which one was a substrate for a particular kinase (3). They incubated the slides in the presence of a kinase and radiolabeled ATP. As anticipated, each of the substrates was radiolabeled by its respective kinase. Similarly, Snyder’s group probed the yeast array described in the preceding paragraph with various phospholipids and identified 150 lipid-binding proteins (Figure 2).

Drug development
Although the flow of protein chips to market has been slow, the market is beginning to pick up. In January, Sense Proteomic (Cambridge, U.K., www.senseproteomic.com) announced the release of its ExpressArray p53, a collection of normal and mutant human p53 proteins, the first commercial result of its collaboration with the robotics company Genetix Group plc (Congleton, U.K., www.genetix.com). The protein is a transcription factor that acts as a natural tumor suppressor in human cells. In more than half of all human cancers, p53 function is lost and the regulation of cell death (apoptosis) is disrupted. If function could be returned to the protein, then perhaps the cancers could be reversed. Thus, it is important to find compounds that will bind to the protein. To that end, protein chips such as this serve as a scaffold on which to test small-molecule ligands for potential therapeutic activity.

Prospects
So what is next for protein chips and antibody arrays? Although there are still many technological barriers to overcome—including protein production and stability, as well as sensitivity—the market is expanding, as is the number of commercial chips and chip systems. According to Stanford University professor Pat Brown, the success of the protein chip is virtually a sure thing (4). “But what will be the best technology, and how soon,” he adds, “remains to be seen.”

References

1. Petricoin III, E. F.; et al. Lancet 2002, 359, 572–577.
2. Zhu, H.; et al. Science 2001, 293, 2101–2105.
3. MacBeath, G.; Schreiber, S. L. Science 2000, 289, 1760–1763.
4. Service, R. F. Science 2001, 294, 2080–2082.

Randall C. Willis is a senior associate editor of Modern Drug Discovery. Send your comments or questions regarding this article to mdd@acs.org or the Editorial Office by fax at 202-776-8166 or by post at 1155 16th Street, NW; Washington, DC 20036.