Julio Celis: keeping track of proteins. Credit: T. BERTELSEN

Mapping the protein universe is a daunting task. Charting the interactions of proteins in the human proteome, studying their function and how their expression changes in health and disease, may take decades. Compared with about 40,000 genes in the genome, the human proteome is predicted to top 1 million. “The existence of more than a million protein isoforms in a single human being, in different cells, during his or her lifetime is not unreasonable,” speculates Rolf Apweiler, Swiss-Prot co-ordinator at the European Bioinformatics Institute in Hinxton, UK. “These are all guesses,” he cautions, but, so far, over 300 known post-translational modifications have been discovered, and these can occur in different combinations, on different proteolytically processed forms and splice variants. A growing array of tools is allowing scientists to advance the study of the proteome, boost drug discovery and pave the way to personalized medicine.

Traditional proteomics uses techniques such as two-dimensional gel electrophoresis or chromatography, combined with mass spectrometry, to separate and identify proteins on a large scale (see Nature 424, 581–587; 200310.1038/424581b). Tracking protein interactions and activity and determining function at this scale is an even greater task, but is beginning to be tackled by protein arrays and multiplexed assays.

To cancer researcher Julio Celis, coordinator of the Danish Centre for Translational Breast Cancer Research in Copenhagen, protein arrays are a breakthrough because they allow many different proteins to be tracked simultaneously. “Discoveries have to be made using very small amounts of clinically relevant sample, and you will want to probe, say, thousands of tissues, both normal and pathological. Multiplexing allows that; it makes life a lot easier,” he says.

The commonest type of protein array, or protein chip, has a large number of spots of either proteins or their ligands arranged in a predefined pattern, arrayed by robots onto coated glass slides, microplates or membranes. The array may consist of antibodies to bind proteins of interest (see Nature 426, 725–731; 200310.1038/426725a), enzymes that will interact with substrates, or substrates or ligands that will interact with applied proteins.

Arrays made easy

One technical hurdle to producing a reliable array is shelf-life. Most protein arrays use antibodies as the immobilized ‘probe’. “The process of depositing an antibody onto a solid surface can denature the antibody and affect its recognition properties,” says Michael Hadjisavas, strategic marketing manager for Sigma-Aldrich in St Louis, Missouri. “We've developed the know-how to select the most desirable antibodies and maintain their biological activity on an array for up to two years.” Sigma's off-the-shelf chips provide a hassle-free introduction to protein arrays. The Panorama Ab microarray kit for cell signalling has 224 different antibodies, arrayed on nitrocellulose-coated glass slides. Proteins extracted from cells or tissues are labelled with the fluorescent dyes Cy3 and Cy5, and the protocol is similar to that of a DNA microarray, with the two labelled protein pools being mixed and applied to the array. The assay takes only 4–5 hours, and can be measured using any standard fluorescence scanner.

Fluorescence is not the only detection system on offer in kit arrays. The range of chemiluminescence-based TransSignal protein-domain arrays from Panomics in Redwood City, California, is expanding, with the addition of SH2 and WW domains. The FASTQuant kit from Schleicher and Schuell in Keene, New Hampshire, combines traditional enzyme-linked immunosorbent assay (ELISA) methodology with the power of multiplexing. Antibodies to nine different human cytokines are incorporated into an array on the company's FASTSlides — nitrocellulose-coated glass slides — and 64 arrays are arranged in a microtitre plate footprint. The company plans to release six FASTQuant kits for human and mouse cytokines, angiogenesis factors and human chemokines this year.

Adding content

One bottleneck that chip manufacturers face is content. “Getting an antibody against every protein in the human proteome is a gigantic task, more so than sequencing the genome. This is not something that will happen in the next decade,” says Jan van Oostrum, head of proteome sciences in functional genomics at Novartis in Basel, Switzerland.

Increasing the content is a matter of competitive advantage for Molecular Staging in New Haven, Connecticut, which uses protein chips in their work with biotech and pharma companies to identify biomarkers for diseases and help reanalyse promising drugs that, although highly effective in some patients, failed clinical trials because of unacceptable side-effects in others. “As soon as we see that a pathway is upregulated or downregulated we gain a mechanistic understanding of the disease, its progression, and patients' response to drugs,” says Peter Fuller, senior vice-president of business development. “We work constantly to increase content — every four months we validate a new chip.” The company's antibody-based chips use a signal-amplification technology developed at Yale University, New Haven. DNA circles attached to the detector proteins are replicated to produce a long DNA molecule that can be detected by attachment of large numbers of small detector molecules. Molecular Staging is collaborating with Eli Lilly to identify biomarkers for sepsis, a potentially deadly reaction to a blood infection that develops so rapidly that physicians find it difficult to diagnose quickly enough.

Multiplexing in solution

The challenges of constructing solid-surface arrays holding thousands of proteins with different properties are fuelling interest in protein-interaction assays in solution. Suspension-bead assays are particularly flexible, and can be adapted to both proteins and nucleic acids. The Bio-Plex system from Bio-Rad Laboratories in Hercules, California, uses Luminex's bead-based xMAP technology (see Nature 422, 917–923; 200310.1038/422917b), as does the LiquiChip system from Qiagen Instruments in Hombrechtikon, Switzerland. For protein analysis, Bio-Plex uses antibodies from Cell Signaling Technologies in Beverly, Massachusetts, to offer ready-to-go cytokine assays, and assays for phosphorylated signal transduction molecules, as well as customized arrays. Suspension-bead arrays are flexible enough to tackle any sort of protein–ligand interaction by simply coupling the required proteins or ligands to different bead populations. Luminex beads, for example, enable simultaneous quantitation of up to 100 different biomolecules in a single microplate well.

Another versatile platform is that of ACLARA Biosciences in Mountain View, California, which can support gene-expression and protein assays run side-by-side on the same sample. Detection is by proprietary eTAG reporter molecules, which are coupled to the ligands (for protein analysis these are usually antibodies), with a different eTAG for each ligand. Labelled ligands are applied to the sample in microtitre plates, or tissue sections on slides, and if the ligand binds to its target, the tag is released. Released tags are resolved, identified and quantified by capillary electrophoresis. An additional tacophoresis step concentrates the tags , taking detection sensitivity into the attomolar range. “It is possible to detect just 10 molecules,” says Sharat Singh, chief technology officer at ACLARA. “In testing patient samples, we can predict whether a patient will respond or not by analysing a few analytes related to the mechanism of action of the drug and the biology of the specific target in the tumour.”

The 3D HydroArray platform from Biocept in Carlsbad, California, uses a proprietary hydrogel polymer that is 95% water, enabling arrayed biomolecules (DNA or proteins) to keep their native conformation. Probe biomolecules are mixed with the prepolymer, and nanolitre droplets are arrayed on a glass slide. During curing, covalent reactions simultaneously crosslink the prepolymer, tether the biomolecules to the gel matrix and bind the droplets to the slide. The result is many levels of probes, giving high sensitivity and a broad dynamic range. Biocept has developed genomic arrays and protein arrays are coming soon.

Mapping interactions

Zeptosens: microarrays must be made under rigorous clean-room conditions.

“If you are talking proteomics, what you are looking for must not be predefined, it has to be an open approach,” says van Oostrum. Reverse arrays, where the cell extract itself is spotted on a chip and probed with a large number of antibodies, is such an approach. They allow researchers to track the activation or perturbation of cellular pathways, to monitor expression of disease-related proteins, and to investigate the cellular effects of drugs, all using very small amounts of sample. Zeptosens in Witterswil, Switzerland, for example, has developed the ZeptoMark CeLyA protein-profiling system, a sensitive reverse array that uses the company's planar waveguide technology. Each chip can contain the protein content of up to 384 different cell lysates for high-throughput profiling.

“We thought that after we had identified how many genes, how many proteins, the problem would go away,” says Celis. “But there are protein–protein interactions that give you different functions. Unless you know those differences, it is difficult to come up with a drug therapy,” he points out.

Giulio Superti-Furga, vice-president of biology at drug-development company Cellzome in Heidelberg, Germany, says: “We look at proteins in their own juice, with their own partners, with their own post-translational modifications.” Cellzome uses two technologies based on liquid chromatography/tandem mass spectrometry to study the protein context of drugs and drug-like molecules. Potential drug compounds are immobilized and used as affinity reagents to identify binding proteins. Candidate drug targets are then ‘positioned’ onto cellular pathways using tandem affinity purification, originally developed by researchers at the European Molecular Biology Laboratory in Heidelberg. “You can almost derive an organizational chart, as you'd draw for a company, but for a proteome, with its drug-binding components earmarked,” says Superti-Furga. Key to the approach is the compilation and visualization of data by informatics. Such maps should help identify points of intersection between different pathways, such as apoptosis and stroke for the TNF-α pathway and lipid metabolism for the APP pathway.

MDS Proteomics in Toronto, Canada, takes a chemiproteomics approach to identify primary and secondary protein targets of bioactive compounds, including those whose mechanism of action is unknown. Protein-drug complexes are pulled out of cells using a tag present on the drug. The proteins are then processed on a ‘proteomics reactor’, developed at the company, that simulates the physiological protein concentrations found in cells, making it more likely that the protein complexed with the drug can be chemically or biochemically processed and analysed by mass spectrometry.

A tough challenge for proteomics is the detection of integral membrane proteins which include important drug targets. By focusing on quantitative proteomic profiling of plasma-membrane proteins, drug-discovery company Caprion in Montreal, Canada, hopes to home in on the 1% of proteins in the cell that are likely to be the most relevant drug targets. Its proprietary Cell-Carta platform is an integrated and automated suite of technologies for fractionation of human tissue samples and plasma, quantitative analysis and directed mass spectrometry-based identification of protein targets.

Inside the intact cell

“The big issue is, what is physiological? What is going on in vivo? How do proteins come together and interact in intact cells?” says Sam Hanash of the University of Michigan Medical School, Ann Arbor, and president of the Human Proteome Organization. This international effort in proteomics focuses on organ systems and biological fluids relevant to diseases. “Some exciting technologies are becoming available that may not be necessarily high-throughput but are more physiological.” He cites methods that allow visualization of protein modifications in living cells, such as genetically encoded fluorescent indicators that detect protein-phosphorylation in signal transduction, for example.

Sam Hanash: keeping it real.

A well-established technique for detecting protein–protein interactions is yeast two-hybrid analysis. “It's very powerful and you generate many interactions. But it is the first level of proteome mining; you then require more detailed follow-up and validation,” says David Litman, senior vice-president of R&D and chief technology officer at BD Biosciences in San Jose, California, which makes the BD Matchmaker system for yeast two-hybrid analysis.

“A lot of proteomics is done in bulk, people taking cells and cracking them open to see what proteins are in there,” says Litman. “But after scanning the whole proteome, one would want to analyse smaller sets of proteins.” Antibodies to detect phosphorylated proteins are being commercialized by BD Biosiences as multiplexed immunoassays to interrogate intracellular pathways using flow cytometry. The new BD FACSArray bioanalyser accepts 96-well plates for high-throughput cellular analysis or multiplexed protein analysis using cytometric beads. A microtitre plate can be run in less than 35 minutes, detecting up to 1,000 events per well and reporting up to four fluorescent and two optical scattering parameters. With this benchtop system, researchers can study protein interactions within a given pathway without having to separate cell populations before running the assay. Reagents to detect kinases, phosphorylated proteins and other activated protein states are available, and BD Biosciences is developing preconfigured kits.

Protein-capture microarrays are limited as they stand, partly because of nonspecific binding by the capture agents, says James Wang, chief technology officer at Hypromatrix in Worcester, Massachusetts. “Additional protein microarray formats are needed,” he says. To increase specificity and expand array applications, Hypromatrix has developed the Staining AntibodyArray to simultaneously detect and localize large numbers of cellular antigens in intact cells. Detector antibodies are arrayed on one support, while the cells are attached to a second. When the two surfaces are apposed, antibodies dissociate from the array and bind to cell-surface antigens. After the supports are separated, secondary antibodies are used to profile the antigens revealed and their cellular location.

The proteomics boom continues. And although cutting-edge proteome analysis is expensive, the growing array of proteomics tools promises to supply researchers with a method suited to every experimental need.