In signal transduction research, protein kinase inhibitors help scientists tease out the vagaries of complex signaling pathways, but anecdotal evidence suggests that they are flawed tools at best, lacking the specificity necessary to draw conclusions from their use. Though experts agree that inhibitors have a legitimate place in science as screening tools, can researchers conduct good science on the carbon backbones of such agents?
Philip Cohen decided to find out. Cohen, Royal Society research professor and director of the Medical Research Council's protein phosphorylation unit, University of Dundee, Scotland, and colleagues published a study that finally quantified the specificity of a battery of commercially available, allegedly selective inhibitors.1 For the most part, Cohen found that the rumor of their precision was greatly exaggerated. "It became clear that many of them were absolutely hopeless," says Cohen. This study, which became a Hot Paper, lays out guidelines for validating kinase inhibitor data.
"This is extremely invaluable information for people who are looking through catalogs, trying to decide whether one of the hundreds of compounds that are purported to be specific inhibitors will solve their problems," Cohen says. The guidelines that he proposes provide "a way to decide whether the compound is any good or not."
Photo: Courtesy of Philip Cohen
JOLLY POWERFUL REAGENTS Cohen has studied protein phosphorylation since 1969, but it took a 1994 meeting with SmithKline Beecham to jump-start his interest in kinase inhibitors. The company had identified a compound that blocks production of tumor necrosis factor by bacterial endotoxin; this compound could help cure inflammatory diseases like rheumatoid arthritis. More importantly, SmithKline also identified the drug's target. "To my amazement," Cohen recalls, "it turned out to be a protein kinase that we'd just discovered ourselves a couple of months earlier."2
What the company lacked was a way to test the enzyme and determine how the drug worked; Cohen provided one. It turned out to be a very potent inhibitor, and a very specific one. "To my great astonishment, at the end of the day, none of the other 20 protein kinases that we were working on at the time were inhibited."3
This experience, among others, convinced Cohen to focus on protein kinase inhibitors. "From my standpoint, these compounds were not only potentially important for treating disease, but they were jolly powerful reagents for use by the cell signaling community." Cohen negotiated with several pharmaceutical companies to set up the Division of Signal Transduction Therapy at the University of Dundee, a laboratory that he hoped would prove mutually beneficial. The facility helps companies speed up the development of potential protein kinase inhibitors by providing sufficient amounts of numerous kinases for high-throughput screening, and by testing the specificity of the leads his team identified. In return, Cohen gains access to some interesting compounds. The laboratory is supported by, and for, pharmaceutical giants AstraZeneca, NovoNordisk, Pfizer, GlaxoSmithKline, and Boehringer-Ingelheim.
As part of the collaboration, Cohen established a kinase profiling service in which his group ran company compounds against a panel of protein kinases. Despite the work's confidentiality, Cohen still wanted to publish at least some of the group's work. Using open time slots in the testing schedule, Cohen's team studied commercially-available protein kinase inhibitors about which they could publish. Currently, the group runs about 240 company compounds per month against a panel of 30 protein kinases.
ROOM TO DOUBT Of the 28 inhibitors tested, only a few were as selective as previously discovered; the rest inhibited at least two different kinases. The work threw a large body of scientific knowledge into doubt. "Literally hundreds of papers had been published with the use of these compounds, with the conclusions that the effect that they observed was specific," says Cohen. "It really meant actually that a huge number of papers really have to be completely written off." Cohen is not aware, however, that any publications have actually been retracted.
It was from this work that Cohen's team devised its proposals. The team recommends that researchers first test a drug against a drug-resistant enyzme mutant. If such a mutant is unavailable, Cohen suggests trying to show that the inhibitor functions at the same concentration necessary to block phosphorylation of an authentic physiological substrate. Another avenue is to show that the effect is observed using at least two structurally unrelated inhibitors, whose specificities have been tested against a wide range of kinases. The remaining guidelines--observations, really--note that different inhibitor concentrations will be necessary in vivo as compared to in vitro, and that even nonspecific protein kinase inhibitors have value as screening tools.
Natalie Ahn, associate professor, chemistry and biochemistry, University of Colorado at Boulder, who reviews manuscripts on protein kinases for Science, adds another suggestion: Show that a dominant-negative mutant of the kinase has the same effect, and that an active mutant has the opposite effect as the drug. Like many of the signal transduction researchers whose work she evaluates, Ahn uses kinase inhibitors carefully. "I used to not trust them," she says. "But sometimes, you've got to use them." For example, she points out that as scientists test inhibitors against more enzymes, the inhibitor specificity diminishes: "I've never seen a drug that didn't become less specific with time." She also follows some of Cohen's suggestions. Her skepticism about these compounds--tempered by pragmatism--carries over into her duties as a manuscript reviewer.
TWO-PRONGED ATTACK The question for scientists following in Cohen's footsteps is: How does one go about publishing work using protein kinase inhibitors? Bernd Pulverer, senior editor at Nature in charge of manuscripts on signal transduction, DNA repair, and the cell cycle, says that Nature has no specific policy regarding publication of work based on kinase inhibitors. "But as a general rule," he says, "we would definitely be extremely unlikely to accept something that's based on a single inhibitor, even if the inhibitor is thought to be rather specific." L. Bryan Ray, a senior editor of Science and its signal transduction knowledge environment editor, echoes that sentiment.
Both editors say that, to publish in their publications, researchers must complement pharmacological data with other lines of research. Though Cohen advises scientists to use several unrelated inhibitors, Pulverer notes that a researcher is unlikely to have a manuscript accepted, even using multiple unrelated inhibitors, without corroboration. "We would still like to see some other genetic means ... to support these conclusions."
Both Ray and Pulverer suggest using RNA interference (RNAi) as one possible approach. RNAi is a genetic technique in which short RNAs, which are complementary to an mRNA, block that transcript's expression. Cohen concedes that RNAi is likely to become a useful tool for inhibitor research, but he cautions that currently it cannot eliminate a protein's expression. If 20% of the protein activity remains, he says, "You actually may see no effect whatsoever."
Cohen suggests using a complementary approach consisting of inhibitors and drug-resistant mutants on one hand, and kinase knockout mice on the other. Ideally, he says, the drug-resistant mutant studies should be conducted using a knock-in strategy, rather than with overexpression constructs. "I think a whole battery of knock-in mice expressing different drug-resistant mutants is going to be very important to the pharmaceutical companies as well as basic scientists," says Cohen; he doesn't know if such a panel is in the pipeline.
GAINING ACCEPTANCE? Ahn agrees that drug-resistant mutants provide definitive answers, but says such mutants are not always an option. That's probably because they're not easy to make, says Cohen. Scientists produce drug-resistant mutants by cocrystallizing the kinase with the drug, then making "informed guesses" about which mutations will eliminate drug binding without simultaneously killing the enzyme. Consequently, Ahn doesn't always expect to see such data in the papers that she reviews.
Ahn does suggest using multiple inhibitors of the kinase, if they exist, but adds that, very often the two drugs have differing specificities, so they are not equally valuable. Ahn also looks for evidence that the drug works at physiological concentrations, as Cohen suggests, and that active and dominant-negative mutants produce the expected phenotypes. But, she says, "There are many experiments where mutant expression is difficult." It may not be feasible to obtain 100% transfection efficiency, she explains, meaning that the mutant's activity may be masked by low signal-to-noise. The best approach is therefore a coordinated one, in which pharmacological and molecular studies complement one another, she concludes.
Admitting his uncertainty, Cohen believes his recommendations are taking hold. "I don't know whether it's my imagination, but I think people are trying to use more than one compound with a different structure ... a lot more." His group, he says, receives many requests for drug-resistant mutant-expressing cell lines. Several messages noted: "It's a pity you didn't publish this paper several years earlier; you would have saved us an enormous amount of time."
Jeffrey M. Perkel can be contacted at firstname.lastname@example.org.
1. S.P. Davies et al., "Specificity and mechanism of action of some commonly used protein kinase inhibitors," Biochemical Journal, 351:95-105, Oct. 1, 2000. (Cited in 191 papers)
2. J. Rouse et al., "A novel kinase cascade triggered by stress and heat shock that stimulates MAPKAP kinase-2 and phosphorylation of the small heat shock proteins," Cell, 78:1027-37, 1994.
3. A. Cuenda et al., "SB 203590 is a specific inhibitor of a MAP kinase homologue which is stimulated by cellular stress and interleukin-1,"FEBS Letters, 364:229-33, 1995.