by Jeff Perkel
Suppose you want to know the substrate for a particular kinase. Or probe how a certain drug treatment affects protein levels directly, instead of using mRNA levels as a proxy. Or perhaps—taking a page from the “guilt by association” school of functional genomics—you wish to know which proteins interact with your interesting, but functionally mysterious, protein.
Each of these questions can be solved using protein microarrays, proteomics' answer to the DNA chip. Yet despite their apparent similarity to DNA arrays, protein arrays represent a much smaller market, precisely because they are as finicky as the protein elements spotted on their surfaces.
"It is very difficult to make protein arrays," says Grigoriy Tchaga, R&D director at Clontech Laboratories. "The dynamic range of proteins is enormous, 1014 in serum, so the complexity of the sample you use on the array is orders of magnitude more complex than when you analyze gene expression. In addition, you have to preserve the biological activity of the capture elements, whether antibodies or proteins. You have to be able to print them while preserving their activity, and for a commercial product, the arrays have to be stable for years."
As a result, there are relatively few protein arrays on the market. Those that are available come in two basic flavors: general protein arrays, and antibody arrays.
Discovery through protein arrays
Michael Snyder, professor of molecular, cellular, and developmental biology at Yale University and director of the Yale Center for Genomics and Proteomics, has published extensively using the former class. In 2004, for instance, Snyder used a proteome array containing 5,800 yeast proteins to search for DNA-binding transcriptional regulators. What he found was a DNA-binding, metabolic enzyme. A year later, he used protein arrays to map protein kinases to their respective targets.
Though some of these studies involved homemade arrays—Snyder's group has considerable experience with these, having developed (in 2001) an array of 5,800 yeast proteins for proteomics analyses—the team also uses commercial ProtoArray microarrays from Invitrogen.
"These [Invitrogen arrays] work great for us," Snyder says. "Interaction assays and kinase assays work really, really well."
Invitrogen's ProtoArray product line originates from the company's 2004 acquisition of ProtoMetrix, a Connecticut-based biotech built around Snyder's original protein microarray technology. The flagship product in the line is a high-density array containing more than 8,200 full-length human proteins from a broad range of classes, including secreted factors, transcription factors, and kinases, on a single microscope slide.
According to Sunitha Baskaran, general manager for protein arrays at Invitrogen, ProtoArray microarrays are finding utility in various applications, including discovery of autoantibody biomarkers, identification of novel kinase and ubiquitin ligase substrates, and elucidation of protein interactions.
The former application enables the profiling of biological fluids such as sera, plasma, urine, and saliva, to reveal differential autoantibody profiles in cancer and autoimmune disease. Jennifer Cannon, business area manager, says the sensitivity and broad dynamic range of the platform allows one to use small amounts of crude sample without preparation or labeling.
For identification of kinase and ubiquitin ligase substrates and interaction partners, however, individual protein probes (that is, proteins used to query the array) must first be purified and tagged, either chemically with biotin, or genetically, with a V5 or FLAG peptide. Binding events can then be detected using an Alexa-fluor-conjugated streptavidin (for biotin tags) or secondary antibody to peptide tags.
Antibody Arrays
Antibody arrays serve a different purpose. Available from such suppliers as Clontech, Thermo Scientific, and Sigma-Aldrich, these products enable protein profiling, examining how levels of several hundred proteins change in response to some treatment.
According to Richard Pembrey, global market segment manager for functional proteomics for Sigma-Aldrich, Sigma’s 725-element Panorama Antibody Microarray-XPRESS Profiler725 Kit includes the largest collection of arrayed antibodies available commercially. (Thermo Scientific's LabVision antibody microarray includes 720 antibodies, and Clontech's Ab Microarray, 512.)
Protein preparation and labeling
Typically, protein samples are labeled with Cy3 and Cy5 dyes, using a kit available from GE Healthcare (formerly Amersham Biosciences). This allows the arrays to be visualized using most standard DNA array scanners. "A range of biological samples can be used," Pembrey says. "The only thing when using serum [as opposed to tissue or cell culture extracts] is you may need to deplete the high abundance proteins, like the immunoglobulins and the albumin, in order to see the lower abundance signals."
According to Pembrey, "optimization is key" when working with antibody arrays.
During protein labeling, for instance, try to get the dye-to-protein ratio as close to 2-to-1 as possible, in order to maximize signal-to-noise. "Fluorescence labeling technologies are notorious for inconsistent conjugation rates when used with proteins," Pembrey says. Plus, he adds, nitrocellulose membranes are inherently autofluorescent, particularly in the Cy5 emission range. (Sigma 's antibody arrays, and some other commercial products as well, are arrayed on nitrocellulose filters attached to glass slides.) So, he says, it is important to start with lower gain settings (around 40%) than would be used for a DNA array.
Clontech's Tchaga also advises watching the dye-to-protein ratio during labeling. Cy dyes are charged and hydrophobic, he says, and incorporation of too many dye molecules can block the antigen epitope and/or change its solubility.
Tchaga further advises researchers to consider the efficiency of protein extraction steps during probe preparation: "It is very important to show that you actually extract representative amounts of the proteins from all the different compartments of the cell." Clontech's extraction, labeling, and incubation buffer kit, which the company offers separately from the Ab Microarray itself, extracts protein at about 95% the efficiency of boiling in SDS, according to Tchaga.
When it comes time to actually applying the sample to the array, Tchaga says, the company recommends using from 10 to 50 micrograms of labeled protein per dye. For most protein preparations, the company advises using the lower limit. But for serum, where 80% of the protein is immunoglobulins and albumin that can mask the lower abundance proteins, he suggests using 50 micrograms of labeled protein instead.
In the end, protein preparation steps—even the array experiment itself—are just the prelude to the real challenge: validating the data and understanding what they say about the underlying biology. Says Snyder, "You do an experiment, and then it takes a few years to figure out what the data means."
