In the 1980s, scientists discovered a fascinating family of powerful molecules and gave them a long name: extracellular signal-regulated kinases, or ERKs. Kinases are enzymes, and while many kinases are cellular workhorses that catalyze energy transfers and keep the cell’s internal machinery humming, ERKs in particular — as their name suggests — respond to signaling molecules from outside the cell that, in effect, relay biochemical messages telling cells to grow, die or change function.
When all goes well, ERKs operate like a network hub and routing system of astonishing resourcefulness, receiving signals from a host of molecules and transmitting information to at least 70 different cellular proteins. Because this process works, cells grow and differentiate into healthy tissue, networks of nerves communicate, and the body’s immune system springs into action when bacteria or other pathogens invade.
But ERKs have a dark side. They can spark the wild fire of cellular chaos. For reasons not fully understood — but often due to genetic mutations in the proteins that signal ERKs, some ERKs become over-energized and, in this state, activate biochemical pathways that trigger uncontrolled cell proliferation — disease states we experience as various forms of cancer. As scientists have learned more about these processes, ERKs have become immensely interesting to researchers; they are one of the most promising targets available for anti-cancer drugs.
The trick, notes University of Maryland, Baltimore (UMB) biophysicist Alex MacKerell, is to design “smart drugs” – drugs that can block the over-energized ERKs while allowing normal ERKs to go about their essential cellular business. MacKerell, who directs the Computer-Aided Drug Design Center of the University of Maryland School of Pharmacy, and his UMB biochemist colleague Paul Shapiro have attacked this challenge with a formidable one-two punch of computation and laboratory experiments.
The ERK protein (blue) represented as a surface with a potential binding-pocket (yellow spheres) on the docking domain and the predicted orientation of a compound (ball and stick representation, balls colored by atom type) that would inhibit the binding of ERK with its partner proteins.
Over the last four years, with multi-year support from the National Institutes of Health and tens of thousands of hours worth of computing on several PSC systems, starting with LeMieux and evolving to BigBen, MacKerell has applied powerful molecular-search techniques. From millions of possible candidates, he has identified about 100 drugs that have potential to stop only the bad ERKs.
In laboratory studies, Shapiro has taken the next step. Working from MacKerell’s list of candidates, he found several compounds with the ability to stop malignant growth in a variety of human cancer cells, including breast and lung cancer. “We’re very optimistic,” says MacKerell. “We have patents pending on some of these compounds, and people in Paul’s lab are testing more of them. Companies want to talk to us about licensing.”
In recent years, researchers around the world have confirmed ERKs’ role in the generation of cancer. In 2001, for example, scientists working in Chicago, New York and Israel found that one of the most used cancer-fighting drugs, Taxol, has its effect through interaction with ERKs. Such intriguing finds help to suggest the prospect of drugs that selectively target particular sites within the complexly coiled structure of ERKs.
You can think of a protein as a big machine with many functions, says MacKerell, and some of them may not be good. “You can break the machine in many places, but that might mean you’d lose some of the good things the machine does. So our job is to be extremely specific about how we break the machine, and to do that you have to understand exactly how the machine is built, its precise shape.”
ERKs, notes MacKerell, can be thought of as a crossroad protein. They are activated by a range of different signaling molecules and in turn can communicate with (via a reaction called “phosphorylation”) up to 70 different proteins. “We want to block just a subset of those 70 proteins — so that we block one pathway and not others. This work falls into the realm of ‘chemical biology,’ an important area right now in biomedical research, and it helps us to understand in detail how ERKs function.”
Across the atomic landscape of ERKs are specific places — “docking domains” — where other molecules can link up. For ERKs, a partner protein first docks and then a chemical change occurs at the ERK’s “active site.” The method to MacKerell’s and Shapiro’s madness is that experiments show different parts of the docking domain link with different proteins.
“It’s like if you jump into a moving truck,” says MacKerell. “You’ve got to latch onto it, and then you can pull yourself in. Basically, 70 different proteins can latch onto ERKs, and one part of an ERK is involved in the latching on, and another part handles the phosphorylation. And that hand-hold where proteins latch on involves different parts of the ERK docking domain for different proteins.”
View from the interior of the ERK protein with a bound inhibitor (red), where the ribbons represent the protein backbone and the mesh represents the protein surface.
Each of the hand-holds — better known as binding pockets — represents the start of a distinctive biochemical pathway that controls a specific task in cellular life. Finding these pockets is a first step. The next is to find a compound that can fill the right pocket and block a cancer-causing pathway. Ultimately, such compounds have the potential to become new drugs that can treat certain types of cancer.
The trick is to find molecules that are the right size, shape and electronic charge at the right places to latch on and fill the pockets. This quest puts MacKerell squarely into the prediction business, and it would be impossible without computational tools. With “molecular dynamics” — simulations that track the atom-by-atom position of a molecule as it changes its shape over time, MacKerell identifies likely binding pockets.
The next step is to use this computer-drawn blueprint to search through a database of small organic compounds to find candidates that are likely fits to a binding-pocket target. To match a single molecule with a single binding-pocket within an ERK’s docking domain requires screening nearly a million compounds one-by-one. When a promising compound is found, it takes up to 20 “docking runs” — computational screens that test to see how snug is the fit, how well the compound matches structurally and biochemically with the pocket. Each of these runs can require up to 100-billion calculations.
Without massively parallel systems — such as PSC’s BigBen — that make it possible to employ hundreds of processors simultaneously, this work would not be feasible. “What PSC does,” says MacKerell, “by allowing us access to so many processors, is make it possible for us to screen through our database multiple times in a very short period of time.” Searches that otherwise would take months can be done in as little as a couple of hours.
When the data storm of the first round of screening settled, MacKerell had found about 100 compounds worth focusing on — his best picks for Shapiro to take into his laboratory for in vivo biological testing with cancer cells. “Alex gave me a list of 100 or so compounds,” says Shapiro, “and I bought a small amount of each of them. In my lab, we have several cancer (cell) lines that we use for our preliminary studies and then the most biologically active compounds are tested in animal models. Here’s where we evaluated whether the compounds Alex predicted could actually be useful by biological standards.”
From the original round of 100 compounds, Shapiro found 10 that showed significant promise as ERK inhibitors, molecules capable of turning off taps of cancer at their source. “It’s exciting,” says Shapiro “that some of these compounds are showing they have an effect on stopping cancer cell proliferation.”
Shapiro counts the computer-based methodology as highly “cost and person-power effective.” His findings to date led two years ago to renewal of NIH support for his work with MacKerell’s lab. “Computational modeling has identified a number of compounds with biological activity,” says Shapiro. “Some of them are now under patent protection.”