The biological processes underlying diseases such as rheumatoid arthritis and cancer are fundamentally linked, and should be linked in how they are treated with drugs, a series of MIT studies indicates.
Key to the work: The researchers applied an engineering approach to cell biology, using mathematical and numerical tools normally associated with the former discipline.
In a series of three papers, the latest of which appeared in the March 24 issue of Cell, Professors Douglas A. Lauffenburger, Peter K. Sorger and Michael B. Yaffe, all members of MIT's Center for Cancer Research, led a team of scientists and engineers in looking at how cells make life-or-death decisions. Understanding what tips a cell toward survival or death is key to treating diseases and fighting cancer through radiation, drug therapy and chemotherapy.
The researchers looked at tumor necrosis factor (TNF), a substance produced by the immune system that promotes cell death, and two prosurvival hormones, epidermal growth factor (EGF) and insulin. TNF and EGF induce conflicting prosurvival and prodeath signals, and the "crosstalk" between these signals is not well understood. The MIT studies provide the first big picture of how these two key factors interact in time and space.
The studies uncovered a surprising link between inflammatory diseases and cancer that may change how these diseases are treated in the future.
More effective drugs
Researchers have been exploring ways to use drugs in combination to increase their therapeutic value in fighting tumors. The results of the three MIT studies have implications for how two classes of drugs involving TNF and EGF affect common biological processes in the body.
Drugs that inhibit TNF are used to treat debilitating chronic inflammatory diseases such as rheumatoid arthritis. Yet TNF, which causes inflammation, also leads to generation of the EGF signals that play a role in many cancers. (The breast cancer drug Herceptin, for example, works by blocking EGF-induced signals.) "TNF is supposed to kill cells. It's counterintuitive that it simultaneously promotes cell survival by sending an 'autocrine' EGF signal to itself," said Sorger, a professor of biology and head of MIT's Center for Cell Decision Processes. Autocrine EGF messages are analogous to mailing yourself a letter. In the case of TNF, cells also mail back the response via another hormone, IL-1.
"With what we now understand about the interactions between these two factors, we should aim for increasing the therapeutic value from these drugs," Sorger said. "There is a fundamental molecular connection between diseases such as rheumatoid arthritis and cancer. Their protein cascades are connected; one stimulates the process of the other."
"In drug development, we want to identify the really important hubs in the network we should be targeting and when we should be targeting them," said Yaffe, an associate professor of biology and biological engineering. "It's key to figure out the most critical point in the cell cycle for the drug to intervene. This work will help accomplish that goal."
Protein wars
Among the cells lining the intestines of a person with inflammatory bowel disease, two different camps are at war. TNF launches an attack, killing many of the epithelial surface cells, while EGF struggles to keep the cells alive and dividing to repair the damage.
In every cell, genes create proteins, the building blocks of life. Besides carrying out the functions of keeping the cell alive, some proteins such as TNF and EGF work as signals, turning on or off other genes. In a cascade effect, the proteins from these genes may affect still more genes. What's more, a single protein behaves differently at different points in time: A protein may do one thing early after stimulation and something else later on.
Researchers want to be able to predict how cells will respond to tiny molecular changes that spur them to develop, multiply or die. If researchers knew exactly how much of a certain protein was needed to kill a cancer cell and exactly when in the cell's life cycle it would be most lethal, drugs could be custom-designed to destroy malignant cells while leaving normal cells intact, Yaffe said.
But for many cell-decision networks, there is simply not enough information about the signaling proteins and reactions to construct a believable model that would allow accurate predictions to be made. Can you design an effective model without measuring every one of the tens of thousands of proteins in a cell?
Hidden connections
"There are a lot of variables and a limited set of observations," Lauffenburger said of cell biology. "How can you abstract what's going on underneath the surface? We're never going to have complete knowledge, but the question was, could we construct a model that admits that we don't know everything, but we know enough to do something useful?" Lauffenburger is head of MIT's Biological Engineering Division and is the Whitaker Professor of Bioengineering.
To answer that question, the team used an engineering approach typically applied to manufacturing or software. "At some point, we need to bring new tools to bear on complexity, and those new tools are engineering-based mathematical and numerical tools," Sorger said. "Just as we can engineer extremely complicated systems like jets that we can't understand in their totality just by looking at them, we can do the same thing in biology." Modeling signaling pathways with computers is one of the tactics of MIT's Center for Cell Decision Processes.
"Models store our aggregate biological knowledge in a tractable way and are used to identify which proteins and pathways are most critical for mediating cell responses," Yaffe said. The research team plugged measurements of thousands of signaling proteins gathered in painstaking laboratory experiments into the models, providing a "firm theoretical grounding" to intuit how protein network interactions affect cell behavior, Yaffe said.
Yaffe divided cell signals into two major dimensions that can be plotted on a graph with a stress/death axis and a survival/growth axis. Where the conflicting factors fall on the graph determines whether the cell upon which they are acting lives or dies. "Our study gives us a broader functional sampling of a lot of things at the same time," he said.
Using this new approach involves teams of researchers, a concept unusual in traditional cell biology. Working as a team, computational scientists remain in close touch with their laboratory-based collaborators. Interdisciplinary scientists working at the interface of biology and computation is the way of the future, according to Kevin A. Janes, graduate student in biological engineering and one of the study's co-authors.
The payoff is high. Combining broad protein-based measurements and computation revealed the big picture, uncovering connections between spheres of biology previously believed to be distinct. "We are finding that things that once appeared to be biologically independent are closely connected," Sorger said. "We are not just collections of independent parts."
This work is supported by the National Institutes of Health and the Whitaker Foundation.
A version of this article appeared in MIT Tech Talk on April 5, 2006 (download PDF).
Key to the work: The researchers applied an engineering approach to cell biology, using mathematical and numerical tools normally associated with the former discipline.
In a series of three papers, the latest of which appeared in the March 24 issue of Cell, Professors Douglas A. Lauffenburger, Peter K. Sorger and Michael B. Yaffe, all members of MIT's Center for Cancer Research, led a team of scientists and engineers in looking at how cells make life-or-death decisions. Understanding what tips a cell toward survival or death is key to treating diseases and fighting cancer through radiation, drug therapy and chemotherapy.
The researchers looked at tumor necrosis factor (TNF), a substance produced by the immune system that promotes cell death, and two prosurvival hormones, epidermal growth factor (EGF) and insulin. TNF and EGF induce conflicting prosurvival and prodeath signals, and the "crosstalk" between these signals is not well understood. The MIT studies provide the first big picture of how these two key factors interact in time and space.
The studies uncovered a surprising link between inflammatory diseases and cancer that may change how these diseases are treated in the future.
More effective drugs
Researchers have been exploring ways to use drugs in combination to increase their therapeutic value in fighting tumors. The results of the three MIT studies have implications for how two classes of drugs involving TNF and EGF affect common biological processes in the body.
Drugs that inhibit TNF are used to treat debilitating chronic inflammatory diseases such as rheumatoid arthritis. Yet TNF, which causes inflammation, also leads to generation of the EGF signals that play a role in many cancers. (The breast cancer drug Herceptin, for example, works by blocking EGF-induced signals.) "TNF is supposed to kill cells. It's counterintuitive that it simultaneously promotes cell survival by sending an 'autocrine' EGF signal to itself," said Sorger, a professor of biology and head of MIT's Center for Cell Decision Processes. Autocrine EGF messages are analogous to mailing yourself a letter. In the case of TNF, cells also mail back the response via another hormone, IL-1.
"With what we now understand about the interactions between these two factors, we should aim for increasing the therapeutic value from these drugs," Sorger said. "There is a fundamental molecular connection between diseases such as rheumatoid arthritis and cancer. Their protein cascades are connected; one stimulates the process of the other."
"In drug development, we want to identify the really important hubs in the network we should be targeting and when we should be targeting them," said Yaffe, an associate professor of biology and biological engineering. "It's key to figure out the most critical point in the cell cycle for the drug to intervene. This work will help accomplish that goal."
Protein wars
Among the cells lining the intestines of a person with inflammatory bowel disease, two different camps are at war. TNF launches an attack, killing many of the epithelial surface cells, while EGF struggles to keep the cells alive and dividing to repair the damage.
In every cell, genes create proteins, the building blocks of life. Besides carrying out the functions of keeping the cell alive, some proteins such as TNF and EGF work as signals, turning on or off other genes. In a cascade effect, the proteins from these genes may affect still more genes. What's more, a single protein behaves differently at different points in time: A protein may do one thing early after stimulation and something else later on.
Researchers want to be able to predict how cells will respond to tiny molecular changes that spur them to develop, multiply or die. If researchers knew exactly how much of a certain protein was needed to kill a cancer cell and exactly when in the cell's life cycle it would be most lethal, drugs could be custom-designed to destroy malignant cells while leaving normal cells intact, Yaffe said.
But for many cell-decision networks, there is simply not enough information about the signaling proteins and reactions to construct a believable model that would allow accurate predictions to be made. Can you design an effective model without measuring every one of the tens of thousands of proteins in a cell?
Hidden connections
"There are a lot of variables and a limited set of observations," Lauffenburger said of cell biology. "How can you abstract what's going on underneath the surface? We're never going to have complete knowledge, but the question was, could we construct a model that admits that we don't know everything, but we know enough to do something useful?" Lauffenburger is head of MIT's Biological Engineering Division and is the Whitaker Professor of Bioengineering.
To answer that question, the team used an engineering approach typically applied to manufacturing or software. "At some point, we need to bring new tools to bear on complexity, and those new tools are engineering-based mathematical and numerical tools," Sorger said. "Just as we can engineer extremely complicated systems like jets that we can't understand in their totality just by looking at them, we can do the same thing in biology." Modeling signaling pathways with computers is one of the tactics of MIT's Center for Cell Decision Processes.
"Models store our aggregate biological knowledge in a tractable way and are used to identify which proteins and pathways are most critical for mediating cell responses," Yaffe said. The research team plugged measurements of thousands of signaling proteins gathered in painstaking laboratory experiments into the models, providing a "firm theoretical grounding" to intuit how protein network interactions affect cell behavior, Yaffe said.
Yaffe divided cell signals into two major dimensions that can be plotted on a graph with a stress/death axis and a survival/growth axis. Where the conflicting factors fall on the graph determines whether the cell upon which they are acting lives or dies. "Our study gives us a broader functional sampling of a lot of things at the same time," he said.
Using this new approach involves teams of researchers, a concept unusual in traditional cell biology. Working as a team, computational scientists remain in close touch with their laboratory-based collaborators. Interdisciplinary scientists working at the interface of biology and computation is the way of the future, according to Kevin A. Janes, graduate student in biological engineering and one of the study's co-authors.
The payoff is high. Combining broad protein-based measurements and computation revealed the big picture, uncovering connections between spheres of biology previously believed to be distinct. "We are finding that things that once appeared to be biologically independent are closely connected," Sorger said. "We are not just collections of independent parts."
This work is supported by the National Institutes of Health and the Whitaker Foundation.
A version of this article appeared in MIT Tech Talk on April 5, 2006 (download PDF).
