Phylos, Inc.

Protein chips and antibody arrays are beginning to make serious headway in diagnostics, biomarker discovery, and proteomics. By studying the changes in protein expression patterns during the course of a disease, researchers are elucidating the metabolic mechanisms behind the disease. This will allow them to develop therapies targeted to the specific molecular cause of the disorder. Similarly, chips are being used to identify previously unknown interactions between proteins, divulging information about cellular signaling pathways that was heretofore unknown. (For more information on protein chip applications, see “The matching game”.)

In part, these achievements are possible because protein chips offer a degree of versatility that is not found in their DNA-based siblings. Whereas cellular DNA and RNA are relatively static, varying only in amount from cell to cell, proteins can take on a wide variety of chemical masks—posttranslational modifications—that can severely alter their activity. By adjusting your array system, you can identify not only whether the protein of interest is in the sample, but also in what form.

But for all of its potential, protein array technology is still in its infancy, and there are hurdles to be overcome.

The hardware
An advantage for laboratories moving into the arena of protein arrays is that much of the equipment used in the manufacture of DNA chips can be tweaked for use in protein chips. Fundamentally, the same liquid-handling systems and arrayers can be used for nucleic acid and protein work; it becomes a matter of adjusting some of the surface chemistry and buffer conditions to place proteins where oligonucleotides once stood. This was the case at Genomic Solutions (Ann Arbor, MI; www.genomicsolutions.com), a company well known for its work in DNA array development, which recently adjusted its GeneTAC DNA array system to handle proteins.

“The ability to produce both protein and DNA microarrays extends the utility of our system and allows researchers to carry our proteomic and genomic research on a common platform,” says Michael Kane, vice president for R&D at Genomic Solutions.

For other companies, however, the shift into protein arrays can be an all-or-none process. In a recent interview with Electronics Journal, Rich Fisler, marketing director of Packard Biochip Technologies (Billerica, MA; www.packardbioscience.com) explained that the company’s move from DNA to protein arrays was prompted by the saturation of the gene chip market (1). “We started to make modifications to the technology that we licensed,” Fisler said, “did testing and validity checks with people in the field to make this a great substrate for protein arrays, and launched our HydroGel product.”

Direct binding
But regardless of what equipment you have in your laboratory, how you stick the protein to the chip surface might be just as critical to your experiments as what proteins you choose. The surface chemistries available to researchers looking to use protein arrays are almost as diverse as the proteins themselves.

One way of attaching proteins to a chip is to simply let them interact directly with a surface. This method is exemplified by some of the systems available from Ciphergen Biosystems (Fremont, CA; www.ciphergen.com) and Biacore International (Uppsala, Sweden; www.biacore.com). These companies offer chips that isolate proteins on surfaces that are hydrophobic or ionic, or that carry metal ions (immobilized metal affinity). Last year, Michael Snyder and his colleagues at Yale University and North Carolina State University (Raleigh) used “homemade” Ni⁺-coated slides and hexahistidine-tagged proteins to generate a yeast proteome array, which they used to identify a series of novel calmodulin- and phospholipid-binding proteins (2).

The main drawback of these systems, however, is the lack of specificity and affinity, and what proteins you find bound to the chip depends largely on the stringency of the washes. As a result, some companies have altered their chips to carry ligands to which specific proteins or protein families bind.

Using ligands
SomaLogic (Boulder, CO; www.somalogic.com) is developing an array system that relies on small oligonucleotides or aptamers that contain brominated deoxyuridine (BrdU) to bind proteins.

“We first take a tiny amount of protein that has been identified and purified, either through classical biochemical research or the Genome Project,” said company founder Larry Gold in an interview with Biophotonics International (3). “We use that tiny amount of protein to prepare an aptamer, and it is that aptamer that will go on the array to seek protein signatures.”

After a protein solution is flowed over the aptamer chip and the chip is rinsed, the chip is exposed to UV light. This activates the BrdU in the aptamer and cross-links it with proteins within an angstrom or so of the modified nucleoside. The photoactivation allows the array to be washed stringently such that proteins interacting nonspecifically can be removed, increasing the signal-to-noise ratio. The chip is then stained with a label that interacts with free amines, which exist on the protein but not on the aptamers.

Using antibodies
Other researchers, however, believe that rather than find a molecule that binds like an antibody, why not just use an antibody? Thus, several scientists have followed the lead of enzyme-linked immunosorbent as says, replacing the multiwell plates with glass slides. In the most direct example of this, Pierce Boston Technology (Woburn, MA; www.piercenet.com) miniaturized the assay to form their SearchLight arrays. Zyomyx (Hayward, CA; www.zyomyx.com) combined its expertise in surface chemistry and microfluidics with Cambridge Antibody Technology’s (Melbourn, U.K.; www.cambridgeantibody.com) expertise in phage-display library development to generate a series of arrays based on single-chain antibody fragments. But others worry that the use of phage-display products limits the conditions under which an assay can be run.

“If people work with a phage-display antibody, that antibody has one single affinity to one form of its target,” says Moncef Jendoubi, founder of Milagen, Inc. (Richmond, CA; www.milagen.com). “If that antibody, because of its inherent amino acid content, doesn’t like to be in a specific [assay] condition, then it is very hard to reproduce the data.”

For this reason, Milagen decided to go with arrays based on polyclonal antibodies, such that a given array spot contains antibodies directed against several aspects of its target.

Antibody mimics
Some groups have decided that antibodies are too large and unwieldy to suit their purposes and have therefore chosen to develop systems based on antibody mimics, small proteins that bear no sequence similarity with antibodies but bind proteins with similar affinity and specificity.

“One of the hurdles that we see to conventional antibody arrays or affinity arrays is the lack of scalability,” says Robert Kumelis, associate director of microarray and assay technology at Phylos (Lexington, MA; www.phylos.com), “and the problem of what to do once you have found and used the 50 or so antibodies that work well in a solid-phase format.”

To address this problem, researchers at Phylos developed a system of antibody mimics that they call trinectins.

“In developing the trinectin system, we searched structural databases and tried to identify small single-domain structures to find something that would resemble an antibody in terms of the [complementarity-determining] regions [CDRs]. Human fibronectin was identified,” says Kumelis. “The strategy was to randomize the loop regions that are analogous to the CDRs of an antibody and, from that, we generated the 10¹³ pools of antibody mimics.”

Another company that has attacked the protein array issue in a similar way is Affibody AB (Stockholm; www.affibody.com). Using combinatorial protein engineering techniques, the Swedish company randomized the surface residues of an -helical receptor domain from staphylococcal protein A. The mutant sequences were inserted into phagemids (plasmids used for protein expression by bacterial viruses or phages). The phages are then used to “pan” for proteins of interest.

Through this method, the company developed a series of short (58-residue) polypeptides or “affibodies” that are highly specific (affinities in the micromolar range) and stable under various conditions (e.g., high temperature or pH extremes), and that can be produced in large quantities by bacteria (1–6 mg/L). When attached to a slide, the affibodies form a protein array.

Technical challenges
But regardless of how the proteins are attached to the slides, inherent problems arise from the biochemical nature of proteins. The DNA molecules used on gene chips are relatively simple (if large) molecules composed of deoxyribose, a phosphate backbone, and four bases that are chemically similar. These components make the DNA on the chip hardy—temperatures approaching 70–80 °C are needed to denature the two strands.

Proteins, however, are complex molecules composed of 20 amino acids that can be mixed and matched in a seemingly endless variety of ways. This diversity makes it difficult to study a population of proteins under the same conditions, because its activity and/or stability can be altered by the slightest of sequence differences. Rather than denature at 70–80 °C, some proteins lose their native structure at temperatures as low as 10 °C or when the buffer solution in which they sit moves outside a specific range of pH or moisture.

Packard BioScience has tried to address this to some extent with the HydroGel slides mentioned earlier. With HydroGel, a polyacrylamide matrix supports the protein on the slide, allowing it to interact with whatever substrate is being tested. For others, the stability problem was the reason behind their choice of what proteins they attached to the array.

“The particular scaffold that we use is a very robust, thermally stable molecule that is particularly well suited for solid-phase applications where you may be required to print an array, package it, and ship it around the world,” says Phylos’s Kumelis. “When we use these approaches side-by-side with antibodies, we find that trinectin is physically more stable.”

When the stability issue is finally eliminated, however, look for protein and antibody arrays to explode onto the market, leaving their DNA forebears far behind.

References

1. Sterlicchi, J.; Wales, E. Electronics Journal; www.electronicsjournal.com/content/trends/trends.html.
2. Zhu, H.; et al. Science 2001, 293, 2101–2105.
3. Lewis, R. A. Biophotonics International 2001, 8, 40–41.

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.