Bring on the Protein Biochips

If you had a protein microarray, what would you put on it?

If only the answer were simple. Unfortunately for researchers looking to scan the proteome the way DNA microarrays scan the genome — and for the companies hoping to supply the required tools — assembling the content necessary to make protein microarrays useful is currently the most formidable obstacle to putting the technology in the hands of scientists.

Take Zyomyx, for example. The Hayward, Calif.-based startup has spent four years developing a robust platform for tethering antibodies to the etched surface of silicon wafers, devising fluorescence-based sandwich assays for detecting the presence and quantity of specific proteins in a sample, and manufacturing the protein arrays and instrument systems en masse. But currently Zyomyx offers only two versions of such an array: one containing antibodies against 30 human cytokines, and one against 30 mouse cytokines. Needless to say, a slow start for a field that many expected to blow up the same way DNA microarrays burst upon the scene almost 10 years ago.

The problems are nowhere near straightforward. Proteins are far less well-behaved than strands of DNA, of course, and finding a specific capture agent for a protein is a lot more complicated than designing a nucleic acid with a sequence complementary to the mRNA of interest. In defense of those trying to develop the technology, comparing protein arrays to DNA arrays may seem logical, says Matt Yoshikawa, Zyomyx’s product development manager — but a more accurate comparison is with the decades-old technology for performing assays such as the ubiquitous ELISA, or enzyme-linked immunosorbent assay. Multiplexing 30 ELISAs, he says by way of comparison, “is a huge analytical and informatics challenge.”

In any case, researchers working with protein arrays agree that the content issue is currently the most significant. The problem also varies according to the type of protein array under consideration. Creating arrays of antibodies or other capture agents presents the biggest challenge, but cloning and expressing proteins for use with arrays of proteins — so-called protein interaction arrays — also presents a vexing issue. Lastly, chips designed for custom antibody arrays involve another set of problems. Here Genome Technology takes you through the trials and tribulations of manufacturing array content, what’s available on the market currently, and what you can expect to see in the future.

Capture Arrays

Of the various types of protein chips, arrays of capture agents such as antibodies seem to hold the most promise for aiding researchers studying protein expression. With applications in basic biological research, discovering new disease pathways, and evaluating drug toxicity, to name just a few, protein capture arrays could vastly improve the speed and biological insight of protein expression experiments.

But what exactly do researchers want to study? Producing a universal panel of capture agents is the simplest approach from a commercial perspective, and would also provide an easy option for researchers looking to purchase a tool off-the-shelf. Given the wide variation in protein modification, however, researchers at the moment don’t even know how many types of proteins actually exist, much less how to produce antibodies against them. So many companies are starting simple. Cytokines, a set of proteins associated with inflammation, are relevant to researchers and clinicians studying conditions ranging from asthma to arthritis, and type 1 diabetes to general mechanisms of immune response. It doesn’t hurt that cytokines are a well-characterized class of proteins, and that many antibodies specific to various cytokines are already available commercially, according to researchers involved in the field.

No big surprise then that Zyomyx and other biochip manufacturers such as Novagen and Schleicher & Schuell have launched initial products that focus on capturing a collection of therapeutically relevant cytokines. Zyomyx, which launched its complete system for performing experiments with its antibody arrays in February of 2003, says it settled on antibodies against cytokines as the content for its first arrays after determining that the diseases of most interest to pharma, according to rates of spending on R&D and number of drug development collaborations, are cancer, central nervous system disorders, and inflammation.

Likewise, Schleicher & Schuell is using its FastQuant platform, which relies on conventional scanners for readout, to bind antibodies to a chip designed with the same footprint as a microtiter platform. The current version of the chip contains antibodies to nine human cytokines, according to Michael Harvey, vice president for R&D at Schleicher & Schuell in the US.

The content, he says, “is not tremendously original,” but the chip should appeal to the large number of scientists interested in cytokines because of their role in inflammation and immune disorders.

Despite the obvious utility of these array products, however, cytokine arrays represent just the low-hanging fruit of protein capture agent arrays. Expanding beyond well-characterized, commercially available antibodies presents a serious challenge — first, in determining which proteins you want to capture, and second, in producing the capture agent and validating its specificity and reproducibility.

“The reason we have cytokine arrays is because all the cytokine antibodies existed already, and nobody had to spend any money to make them for this application,” says Leigh Anderson, who led a now-defunct effort at Large Scale Biology to commercialize antibody arrays. “Once you want to step beyond the cytokines, you have to start thinking about what you really want to measure, and that’s hard.”

Anderson, who now leads the Plasma Proteome Institute, a non-profit initiative to study all the proteins found in plasma, not surprisingly advocates creating arrays of antibodies that detect all of the proteins in plasma that are useful as drug targets or biomarkers for disease. At Zyomyx, on the other hand, Yoshikawa, the product manager, and CEO Robert Monaghan say in the near term that they plan to create antibody arrays targeting relatively small sets of human growth factor proteins and markers associated with human tumors.

Novagen, a unit of EMD Biosciences, which currently sells an antibody array against a set of 16 cytokines, also plans to venture into markers for apoptosis and Alzheimer’s disease, according to Karin Hughes, a product manager at Novagen. At NextGen Sciences in Cambridgeshire, UK, CEO Kevin Auton says his company is developing an array of antibodies targeting proteins associated with breast cancer. The strategy, these and other companies say, is to create sets of antibodies targeted at a particular subset of basic or clinical research, as opposed to creating one huge array containing a broad spectrum of antibodies.

In contrast, BD Biosciences Clontech is one of the few groups tackling the more discovery-oriented realm of the market. Sailaja Kuchibhatla, Clontech’s VP for business development, says from her perspective the market dictates that “the more content, the better.” Researchers want to first discover new proteins involved in any system under investigation, she says, before narrowing their focus to quantitatively measure the concentrations of small sets of proteins in later-stage studies.

Clontech currently offers an array containing up to 500 antibodies immobilized on glass slides, although the content consists of well-characterized antibodies available commercially through its Pharmingen subsidiary. First launched in January of 2002, these arrays contain antibodies for measuring the relative concentration of a wide variety of proteins, including many cell cycle and signal transduction proteins, cytokines, kinases, and proteins involved in mechanisms of apoptosis.

Molecular Staging, in New Haven, Conn., is also making strides in developing antibody chips containing a broad array of proteins. In a collaboration with Eli Lilly, the company is employing a chip containing 130 cytokines, interleukins, and apoptosis-related proteins, among others, to screen samples from a phase III clinical trial of a Lilly sepsis drug, according to Steven Bodovitz, a consultant with San Francisco-based BioPerspectives.

The problem is that manufacturing capture agents like antibodies is time-consuming and expensive. For lack of a better alternative, monoclonal antibodies remain the de facto standard, but require several weeks to produce and isolate them from mouse serum. Clontech and Novagen say they have the advantage of being able to access antibody development expertise within their own companies, but other players in protein capture arrays say that expertise is just as easily outsourced. Several dozen companies specialize in producing monoclonal antibodies — each with their own proprietary twist — and in addition, there are several academic initiatives to systematically create new antibodies for proteins and provide public access to their data, if not the antibodies themselves.

However, just because monoclonal antibodies are accepted as the best capture agents today doesn’t mean they always will be. Larry Gold, CSO of Boulder, Colo.-based SomaLogic, says his company has spent several years optimizing aptamers — single-chain DNA molecules that crosslink to form the shape of a protein’s active binding site — and remains optimistic that researchers will come around to accepting that aptamers perform equally if not better as capture agents than monoclonal antibodies. Countless other academic groups and companies, such as Archemix and Phylos, are also trying to convince the scientific community that alternatives to antibodies don’t necessarily suffer from poor performance.

And ultimately, validating the antibodies for specificity and stability in an array experiment still presents a signficant hurdle. “You have to put in an enormous effort in quality control to ensure that customers get reliable data,” says Zyomyx’s Yoshikawa. “The performance requirements are much higher for protein chips than DNA chips,” because in a DNA array the variability of any one probe is averaged out, he adds.

Interaction Arrays

At the other end of the spectrum lie arrays of proteins themselves, which have the potential to help researchers study networks of protein-protein interactions and identify new drug-protein interactions, among other applications. While antibody arrays are conventionally thought to represent a larger market because they could find use in clinical diganostics, market projections for protein interaction arrays and protein capture arrays “come out to be roughly in the same ballpark,” says Bodovitz.

The reason for the optimism stems primarily from the success of several academics, most notably Mike Snyder at Yale and Dolores Cahill at the Royal College of Surgeons in Ireland, in arraying thousands of yeast proteins on a single array. Snyder, who published a paper in Science two and a half years ago describing his method for arraying 5,800 proteins on one array, has since co-founded Branford, Conn.-based Protometrix in an attempt to commercialize the technology. The yeast chips have turned out to be incredibly useful, Snyder says. “You wouldn’t believe the number of requests we’ve received [to access the technology].” (Take his word for it: Protometrix has garnered enough attention that it was just acquired by Invitrogen.) Adds Bodovitz: “In theory it’s a hugely informative experiment.”

But again, content rears its ugly head. Snyder says Protometrix is working to release chips containing human proteins, but has had to experiment with various platforms to find one that reliably expresses functional human proteins. And at least initially, Snyder says the company plans to release arrays containing focused sets of human proteins, such as protein families of interest to pharmaceutical researchers. “If these work, then it would represent a huge quantum leap on the interaction side [of protein microarrays],” says Bodovitz. In the long run, Snyder hopes to develop a “uniproteome” chip containing proteins representative of each gene.

Protometrix envisions small molecule screening as a potentially powerful application of its interaction arrays, as well as experiments that investigate the ability of a particular enzyme to phosphorylate or otherwise modify proteins on the array, says Paul Predki, Protometrix’s vice president of R&D. “A lot of our content is driven by these considerations,” he adds.

A Simple Solution?

Ciphergen CEO Bill Rich believes his company has the ideal solution for researchers who want to study sets of proteins that may only be of interest to them. Effectively a form of separation medium, Ciphergen’s ProteinChip pulls down proteins with specific functionalities and holds them in place on the chips while the company’s SELDI mass spectrometry platform attempts to identify them. If the proteins are significant enough to merit further study, a researcher could acquire the expressed protein or its antibody (if it’s commercially available) and buy a different type of chip from Ciphergen designed to function as a do-it-yourself antibody array. In this way, scientists can create a customized antibody or capture agent array, Rich says. Perkin Elmer also currently sells a do-it-yourself protein array system.

Of course, if you knew what you were looking for, it would be a lot easier just to buy a protein capture array off the shelf. And someday, perhaps, it might be possible to scan the entire proteome in one experiment using an array containing antibodies for every human protein. But skeptics abound. In the meantime, be ready for the next small panel of antibodies to emerge — with any luck, it’ll be something you can use.