The Case for Activity-Based Proteomics

By Jason Socrates Bardi

No doubt owing to the success and hype of the Human Genome Project, several other "'omes"—the glycome, the transcriptome, the proteome—have been popping up in the English lexicon, like so many decorative statues on a green lawn.

These other 'omes are no mere ornaments, though. As scientific endeavors go, unraveling the secrets of the carbohydrate in the human glycome, the proteins in the human proteome, and the mRNA in the human transcriptome follow naturally from the human genome. In fact, understanding them may even be necessary for the realization of the promise of the human genome—that mapping will reveal much about human biology, including possible targets and tools for addressing human disease.

Perhaps the most immediately important of these post-genome 'omes is the proteome, which by some estimates is at least an order of magnitude more vast than the human genome. With multiple, alternative RNA splices, post-translational modifications, and protein folding controlled by localization, protein interactions, cofactors, temperature, pH, and a dozen other factors, the 40,000 or so genes in the human genome multiply into millions of possible protein forms and states in the cell.

What is really interesting is to look at a portion of the proteome—those gene products that are specifically involved in some discrete pathology. What could offer more insight than looking at the protein "profile" of a cancerous cell compared to a normal cell, of a cancerous cell soaked in a drug compared to one that is not?

These are applications within the scope of proteomics, which attempts to elucidate the actual expression and action of the genome in different cell types and states. One important aspect of this is where and when the proteins are working in the human body, and how they are turned on or shut off in certain cell types, times, or conditions like infection or cancer.

A new proteomics technology, which is being developed and applied by two investigators at The Scripps Research Institute (TSRI) seeks to answer an even more profound question—which proteins are active in a given cell, tissue, or disease state?

"It's a chemical approach to understanding protein function," says Cell Biology Associate Professor Benjamin Cravatt, who began developing the approach about three years ago. Soon afterwards, he began collaborating with his colleague and fellow investigators in the Skaggs Institute for Chemical Biology, Chemistry Associate Professor Erik Sorensen.

The Problem With Proteomics

Generally, proteomics is still an application in search of its technology. While there are powerful methods for the global analysis of mRNA levels in cells based on gene chip technology. There are also tools for looking at proteins—using two-dimensional electrophoresis coupled with mass spectrometry, for instance. These techniques have lacked luster when it comes to profiling protein activities, though. Methods that are based on profiling expression levels may not be sensitive to changes in the activity of a protein. And methods that do directly look at protein activity do not necessarily do so in the native state of the cell.

Current proteomics methods may also miss proteins that are only expressed in low abundance, which may be exactly the proteins that are of interest. Most of the proteins expressed in a cell, after all, are structural and housekeeping proteins. And they may miss proteins that are associated with membranes, proteins with extreme isoelectric points—either very basic or very acidic proteins—and proteins that have very high or very low molecular weights.

Cravatt believes that activity-based profiling of proteins expressed in living cells is the next step

"This is a way to accelerate the process [of finding] active enzymes that correlate with disease [and] circumvent major challenges that face proteomics technologies," says Cravatt.

Based on Fundamental Enzymology

The basis of Cravatt's technique is thoroughly grounded in over a half-century of classical enzymology, the biochemical field concerned with the activity and rates of catalytic biological molecules.

Over the decades, biochemists have advanced their understanding of various enzymes by developing chemical probes called "affinity labels." These are simple, small chemicals with a reactive moiety that has the ability to attach to the active sites of protein enzymes. These chemical moieties are used to try to understand the mechanism through which enzymes work.

Active-site proteomics takes this classical approach a step further. Rather than employing chemical probes to profile one enzyme, the technology uses probes to assay the activity of entire enzyme families in complex proteomes. These probes combine a reactive group that binds and covalently modifies the active sites of many members of a given enzyme class, with a chemical (e.g., fluorescent) tag for the detection and isolation of reactive enzymes.

The idea is simple: throw these probes in together with different types of cells, let them label their enzyme targets, and then do a separation. Usually this separation involves first splitting cells into the extracellular fraction (which contains the secreted proteins), the lipidic fraction (containing the membrane-bound proteins) and the cytosolic fraction (with the soluble, intracellular proteins). Then the tagged enzymes in each portion can be further separated and identified by time-tested methods like running a gel.

By taking these labels and applying them in cases where there is some phenotype of interest—a metastatic cell, for instance—activity-based proteomics can give a readout of which proteins in the cell are active and which are not. Simple.

But nothing is ever that simple.

The real trick is to find the right probes. The best experimental design is to cover the maximum number of enzymes in a cell with the minimum number of what they call "promiscuous" probes. The point is not necessarily to look at the absolute activity of proteins in the mix but to interrogate broadly and see whether proteins are on or off.

Luckily, when Cravatt started developing the technology, he already had one such probe—an affinity agent for a family of enzymes called the serine hydrolases. He had been using the probe in the course of his studies on the fatty acid amide hydrolase (FAAH) and other members of the serine hydrolase family. It was known that the FAAH probes were promiscuous and reacted broadly with other serine hydrolases, and it was a relatively simple matter to doctor these known affinity labels with readout tags.

"The beauty of it is that it reacts broadly within that family and it tends not to react with enzymes of other classes," says Sorensen.

So far, they have begun using the technology with this probe to study melanoma, breast carcinoma, and ovarian carcinoma cells. They are even able to detect proteins at the femtomole level—as few as a hundred or so copies of per cell.

Activity-Based Profiling of Cancer Lines

By far the most interesting application of the technology is comparative profiling, which aims to detect disease-related protein activities and to establish the identity of proteins that are involved in the pathogenesis of diseases by comparing, for instance, the active proteins in tumor cells to normal cells.

In what Cravatt calls the first practical application of the new technology, a paper in this week's Proceedings of the National Academy of the Sciences describes the application of the approach to cancerous cells.

Using the serine hydrolase probes, Cravatt and his student Nadim Jessani looked for proteins that could uniquely identify different cell lines, like a fingerprint or an eye scan. They chose well-characterized cell lines to start with, so that they would have a baseline from which to judge the effectiveness of the technology.

What Jessani and Cravatt found were clusters of proteases, lipases, and esterases that they could use to distinguish human breast cancer cells from other types of cancer cells. They also found that they could use these markers to distinguish invasive tumor cells—those that can migrate to new tissues after metastasis—from non-invasive ones.

More importantly, they also found that invasive cells from different cell lines looked more like each other than they do like the non-invasive cells that they derived from.

"It's not just the upregulation of new enzymes," says Cravatt. "The cells shut off the original markers that would lead you to believe that they are, for example, a melanoma."

Furthermore, they identified an enzyme, the membrane-associated hydrolase KIAA1363, that had previously not been associated with cancer. Based on the fact that KIAA1363 is up-regulated in invasive cancer cell lines, Cravatt and Jessani suggest that this enzyme may represent a new marker of tumor progression.

They were even able to predict invasiveness in other cells based on the readout of their activity-based assays and the upregulation of KIAA1363.

Significantly, many of the proteins that they were able to profile using the technology are secreted, found in the membrane, or are expressed in low abundance—difficult to study using traditional proteomics methods.

"[The paper] shows that we can go into cell lines that have been characterized for a decade and see new stuff, such as enzymes that correlate with invasiveness that have never been seen before," says Cravatt.

As insightful as these studies are, they are nevertheless limited by the breadth of the probe, which has the ability to label all the active serine hydrolases but none of the active proteins in other families. And while the serine hydrolases as an incredibly large family, comprising fully one percent of all predicted gene products, that still leaves the glass 99 percent empty.

There are many other important families of enzymes that would be interesting to study using the method, but for which suitable probes are lacking. The kinases, the metalloproteinases, the phosphatases, for instance, are all large families with many enzymes relevant to numerous disease pathologies waiting to be interrogated.

"Can we generalize the concept to address other important enzyme families?" asks Sorensen. "That is the question."

A More Generalized Approach

It's a question that's simple to pronounce but hard to solve.

Indeed, any chemical that can attach itself to a residue in the active site of a protein could be a potential probe. The problem is finding ones that react broadly.

In order to address this, Cravatt, Sorensen, and their student Greg Adam took what they call a non-directed strategy to come up with more candidate probes.

They made combinatorial libraries of probes based on a common "chemotype"—sulfonyl esters, which are commercially available in hundreds of variations, linked to a variable alkyl/aryl binding group. By attaching different binding groups to various sulfonyl esters, the nature of the protein activity profile can be changed. By varying these binding groups, the researchers constructed a library of compounds and then screened these against complex proteomes, looking for activity-dependent protein reactivity.

In a Nature Biotechnology article appearing this month, Sorensen and Cravatt describe how one probe in their library can be used to detect six or seven mechanistically distinct, known enzyme families.

"This is the first evidence that this technology can be expanded to cover a significant portion of the proteome," says Cravatt. "The challenge [now] is to cover the maximum amount."

Sorensen is pursuing projects in total synthesis to produce additional probes, looking towards bioactive natural products for hints on how to design them.

"There are a lot of natural products that nature designed to interact with biological molecules," he says.

One of these, a compound called fumagillin, was discovered in a fungus that inhibits a virus that infects the bacterium Staphylococcus aureus. Fumagillin, as it turns out, reacts with certain metalloproteinases and is a powerful anti-angiogenic and anti-tumor agent.

Sorensen reasoned that the reactivity of the compound might be a good scaffold upon which to build a more broadly reactive probe. By building these molecules from commercially available precursors, Sorensen and his group can design precursors and suitable structural alternatives.

By varying the chemical nature of one of its chains, Sorensen has been able to redirect fumagillin-like compounds to other proteins in the metalloproteinase family. Now they are seeing if they can address metalloproteinases in a global fashion.

Working with Students

Cravatt says projects like these are perfect for students at TSRI because they offer a twofold education. Students learn both how to make molecules and test them in living cells to see how they work—a truly interdisciplinary approach that is rewarding for the students and the instructors alike.

"That's what really drew me—the ability to do both chemistry and biology," says Adam, who is beginning his fifth year in the Chemistry Program.

Jessani, who starts his fifth year in the Macromolecular and Cellular Structure and Chemistry Program this fall, was immediately drawn to the biological applications of the work. "The serine hydrolases are such a huge family," he says. "They had developed the probes, and it was obvious they would be good at looking at biological systems."

"He has really done tremendous interface work," Cravatt adds. And, he says, TSRI is perfect for investigators like Cravatt and Sorensen who design such projects because there are students who legitimately want to work in both fields.

"It's a lot of fun to work with these students," notes Sorensen.



Go back to News & Views Index

Associate Professor Benjamin F. Cravatt says of the proteomics technology he is developing with TSRI colleague Erik Sorensen, "It's a chemical approach to understanding protein function." Photo by Jeff Tippett.


Associate Professor Erik Sorensen asks, "Can we generalize the concept [we developed] to address other important enzyme families? That's the question.


Graduate Student Nadim Jessani is first author on the recent Proceedings of the National Academy of Sciences article.
Photo by Jason S. Bardi.

Graduate Student Gregory C. Adam is first author on the recent Nature Biotechnology article. Photo by Jason S. Bardi.