Protein Microarrays Mature

MANUFACTURING HURDLES

Victor Morozov of the National Center for Biodefense at George Mason University, Manassas, Va., says that in order for protein arrays to be competitively priced, developers will have to turn to parallel, industrial-scale manufacturing technologies similar to those employed in the production of microelectronic chips. Or, they will need to develop in situ synthesis methods such as that pioneered by DNA array manufacturer Affymetrix of Santa Clara, Calif.

With current technologies, though, on-chip synthesis of whole proteins is impractical, as the chemistry involved is more complex and costly than oligonucleotide synthesis. Further, proteins synthesized on-chip may lose tertiary structure during the multistep fabrication process. "In situ [synthesis] doesn't really work," says Schena. "You really need to presynthesize the proteins and deliver them to the surface postsynthesis."

Parallel fabrication of multiple protein spots can be achieved by microcontact printing and electrospray deposition, but these methods currently work best for arrays comprising a limited number of proteins. Because protein synthesis currently cannot be amplified, parallelization of heterogeneous array manufacturing remains problematic owing to the time and expense involved in individual protein synthesis. Protometrix of Bradford, Conn., addresses the problem with a proprietary high-throughput process for cloning, validating, and expressing proteins. But most arrays, whether commercially available or user-made, are manufactured serially, spot-by-spot, and will likely continue to be done this way until a "protein PCR" is invented.

THE TROUBLE WITH ANTIBODIES

In the infant stages of protein array development, format was the main concern: two-dimensional slides vs. 3-D hydrogel-based substrates, and bead-based vs. microfluidics vs. standard chips. "Since then this has totally changed; the focus is now much more on the content rather than on the platform," says Steffen Nock, cofounder of QualityProteomics, Mountain View, Calif., a provider of research agents for the proteomics market. The reason: Several formats have proven successful. "People see differences in terms of background and in some cases reaction kinetics, but those seem to be pretty much secondary to the fact that all of the surfaces work," says Schena.

Though the content of protein arrays ranges from antibodies and antigens to recombinant proteins and peptides, the majority of arrays use monoclonal antibodies for target capture. But monoclonal antibodies with the affinity and specificity necessary for multiplexing can take up to nine months and tens of thousands of dollars to develop. "The biggest challenge out there right now is how do we develop these high-affinity antibodies for general proteins in a cost-efficient way," Nock says. Some companies are currently trying to develop alternatives.

Boulder, Colo.-based SomaLogic, for example, employs photoaptamers, single-stranded nucleic acids selected to bind specifically to target proteins. These aptamers, selected via a process called PhotoSELEX, form covalent bonds with their targets when irradiated with ultraviolet light. "That allows them to do much more thorough washing of the array," than would otherwise be possible, says Nock.

SomaLogic's current array format, which the company uses internally in its diagnostics-development programs, combines photoaptamer capture and antibody-based detection. In the long term, though, the company is working to develop a universal protein stain that does away with antibodies altogether. Instead, the stain specifically binds to the proteins, not the photoaptamers, providing a signal directly proportional to the amount of bound protein, says Todd Gander, SomaLogic's senior director of corporate development and strategic planning. The advantage: better signal-to-noise. "You don't get the cross-talk, you don't have the problem of getting large numbers of secondary antibodies into solution. Plus you don't have to identify all the secondary antibodies in the first place, and then find compatible pairs," Gander explains.

THE SENSITIVITY CONUNDRUM

Courtesy of Biacore

Surface plasmon resonance (SPR) arises when light is reflected from a conducting film at the interface between two media of different refractive index. In Biacore systems the media are the sample and the glass of the sensor chip, and the conducting film is a thin layer of gold on the chip surface. When molecules in the sample bind to the sensor surface, the concentration and therefore the refractive index at the surface changes and an SPR response is detected – specifically, a reduction in the intensity of reflected light at a specific angle of reflection. Plotting the response against time during the course of an interaction provides a quantitative measure of the progress of the interaction.

Improving signal-to-noise ratios in protein arrays remains challenging across the board. Signals from high-abundance proteins can drown out those from low-abundance analytes. Or, such proteins can cross-react with nonspecific antibodies on the array. Technologies such as Rolling Circle Amplification from New Haven, Conn.-based Molecular Staging offer enzyme-based amplification of a target molecule or probe, but these techniques go only so far, says Nock. "You have to keep in mind if you amplify the signal, you amplify the background at probably the same kind of rate, so at the end you don't really gain a lot in overall sensitivi ty," Nock says, adding that reducing background and boosting reagent specificity ultimately will improve sensitivity.

Companies such as SomaLogic reduce background noise by changing the capture agent. Others employ novel detection sys tems. Witterswil, Switzerland-based Zeptosens, for example, has developed a method based on planar waveguides (PWGs), modi fying the standard glass-slide substrate with a thin film of tantalum pentoxide (Ta₂O₅). This high-refractive-index material guides laser light on the surface of the chip only, permitting selective detection of captured labels. "We can discriminate between sur face-bound and free labels in solution, and therefore we can detect binding reactions in situ in solution without having an additional wash," says Michael Pawlak, director of protein microarrays. The company uses PWG technology in its Zep toMARK CeLyA cell lysate protein-profiling system and arrays.

Another approach is to eliminate labels altogether. The ProteinChip system from Ciphergen of Fremont, Calif., uses a MALDI mass spectrometry variant called SELDI (surface-enhanced laser-desorption ionization) to identify bound proteins Alternatively, Uppsala, Sweden-based Biacore uses surface plasmon resonance (SPR) technology to monitor protein-binding events. Biacore's detection instrument plots time versus changes in the refractive index at the surface of a chip coated with a thin layer of gold, allowing real-time measurement of the progress of interactions and providing information about the specificity, kinetics, and strength of binding. One potential drawback: The technology lacks the target-multiplexing capabilities usually associated with the concept of arrays, which can be problematic for high-throughput drug discovery.

Sensitivity can also be improved within the bounds of con ventional array technologies. One problem, for instance, is that some 3-D substrates can diffuse when aliquotted, reducing spot circularity and protein density. Anaheim, Calif.-based Mira gene's Zeta-Grip polymer reduces this diffusion problem, how ever, boosting assay sensitivity into the femtogram range, according to Stewart Lebrun, director of R&D.

ENSURING PROPER FOLDING

For accurate measurement of binding events, surface-bound proteins must be correctly folded and fully functional. Some companies follow rigorous quality-control procedures to ensure this. Protometrix (a subsidiary of Carlsbad, Calif.-based Invitrogen), which sells an 11,000-spot yeast whole-proteome array, looks for reciprocal interactions between proteins on the array and in solution to indicate proper folding. Ninety percent of the time, proteins that bind in solution retain their ability to interact on the array, says Paul Predki, vice president for R&D at Protometrix.

UK-based Procognia bypasses this type of quality check by incorporating biotinylated tags into each protein during synthesis; these tags serve both to tether the protein to streptavidin-coated slides and to ensure that the arrayed proteins are folded in their active states. Because biotinylation depends on the enzymatic recognition of a 3-D epitope, the proteins must be folded correctly for the tags to be accessible for modification, says Johanna Griffin, the company's chief compliance officer. The strong binding affinity between the labeled proteins and strepta-vidin-coated surface enables proteins to be spotted directly from cell lysates without prior purification. The combined surface and capture chemistries enable Procognia to spot hundreds of proteins per array.

Another workaround to the protein-folding problem is to simply array only the functional domains of interest themselves, rather than the whole protein. Amy Keating of Massachusetts Institute of Technology employs domain-based arrays in her research on coiled-coil protein interactions. Observing an interaction or activity on a domain-level protein array allows researchers to localize that activity to a specific part of the protein more rapidly, Keating explains. Further, domains often have better autonomous folding properties than their full-length counterparts, which may need chaperones to fold. And domains can be better behaved. "Proteins are prone to oxidation and degradation via proteolysis, issues of nonspecific aggregation. All these can be made a lot worse if you bring parts of the protein that aren't important for the properties you're assaying along for the ride," she says.

ALL-IN-ONE SYSTEMS VS. DO-IT-YOURSELF

Like many researchers, Keating spots her own arrays on commercially available derivatized slides. For specialized applications, do-it-yourself arrays are the logical choice. But prefabricated arrays are growing in popularity, especially for routine applications in drug discovery and development. In addition, a few companies sell complete protein biochip platforms that take the customer through an entire experiment but require specialized instrumentation.

Supporters of the closed-platform approach emphasize their ease of use and increased walk-away time. Others are less impressed. Todd Martinsky, vice president of ArrayIt, says open platforms are key to commercial success. "I don't think that a closed platform is going to be successful. Sticking with the 1-by-3-inch substrate that's commercially compatible with all the different [equipment], that's critical," Martinsky says.

But closed-platform proponents argue their products attract a more lucrative customer base: Big Pharma. "I think the do-it-yourself mode from a market side is not that attractive. Serious screeners that have to compare thousands of samples, they won't go that road," says Peter Wagner, chief technology officer of Hayward, Calif.-based Zyomyx, which markets the closed Protein Profiling Biochip System.

Researchers who decide to go with a closed platform have two options. One, typified by PerkinElmer Life and Analytical Sciences of Boston, offers researchers the tools to make, scan, and analyze the arrays but leaves array design to the user. The alternative, represented by Biacore and Zyomyx, is an integrated system based on prefabricated arrays. If the protein array market parallels that of DNA arrays, the prefabricated chip may ultimately prevail.

"People first started making their own [DNA arrays] because they needed to identify what they wanted on the chips, and then eventually I think scientists started to turn more towards either core facilities that would print them for them, or commercially available chips. And I think the protein array market, on a smaller scale, would be very similar," says Sandra Rasmussen, business manager for functional genomics and proteomics, PerkinElmer Life and Analytical Sciences.

One current challenge that remains to be addressed is that of membrane proteins, considered by many to be the most clinically relevant class of proteins – and also the hardest to array. "Most of modern pharmacology is based on about 700 targets in the human body, and most of them are those membrane proteins," notes Morozov. "[But] they are very hard to introduce into microarray format, because it's not easy to organize an environment on microarray that would keep them mobile and alive." Membrane proteins ultimately will require new technologies to make them amenable to arraying, Morozov says. Zyomyx, for one, plans to introduce a new biochip for the study of ion-channel proteins that detects electronic signals instead of fluorescent labels.

Whole-proteome arrays are also in the works. Protometrix/Invitrogen recently released a whole yeast proteome array and plans in the near future to release human proteome subset arrays. And novel applications continue to be addressed, including Procognia's glycoanalysis array and a "dissociable protein array" that Hypromatrix of Worcester, Mass., is developing as a high-throughput substitute for standard immunostaining.

All of which is good news for the burgeoning protein array market. Says Schena: "As more researchers realize they can do things in a protein microarray format, I think we're going to see an expansion of the field."

Aileen Constans aconstans@the-scientist.com.