Volume 4, Issue 28 May 2007
Proteins as Shape-Shifters
Jhih-Wei Chu joined the UC Berkeley Chemical Engineering faculty in 2006. Photo courtesy of Jhih-Wei Chu.
The devil is in the details, or so the old adage goes. Jhih-Wei Chu understands that better than anyone. A UC Berkeley professor of chemical engineering and QB3 faculty affiliate, Chu models the behavior of proteins atom by painstaking atom. By understanding how proteins interact with other molecules, as well as larger structures in the cell, he is developing a new way to target medicines, design novel materials, and ultimately improve our understanding of cell behavior.
"More than ninety percent of the work in biology is done by proteins," Chu says. The majority function like nanoscale Rube Goldberg machines. When the appropriate target molecule, or ligand, floats by, the active site of a protein will envelop it like a hand closing over a ball. The act of binding then triggers another part of the protein to change shape. Such conformational changes, scientists believe, is how proteins communicate with other proteins and cell elements.
Chu studies how these conformational changes occur. His findings could help develop a more targeted generation of medicines. "A lot of ligands are common from one protein family to another. If you design a drug to target a protein where the ligand binds, you might bind many other proteins as well. This leads to side effects," Chu says. "We'd like to design molecules that don't target active sites but modulate the protein's conformational change instead."
Chu's computer models can track between 30,000 and 1,000,000 atoms, including the protein itself, water molecules, and target substrates. Following that many particles produces an incredibly detailed picture of protein behavior. The tradeoff, however, is that atomistic simulations require massive amounts of processing power. For example, one recent simulation of 100 nanoseconds (ten millionths of a second) took Chu a month and a half to complete.
If designed carefully, 100 nanoseconds is sufficient to watch a protein binding and closing around one ligand. Chu can even hurry the process along by pushing and pulling on different portions of the protein with precise amounts of force. But Chu wants to understand how those conformational changes affect the cell and its external environment. And these processes can require microseconds to milliseconds to occur. Running an atomistic model of such behaviors could occupy a year an a half of computer time.
A constituent of muscle fiber, the protein actin self-assembles into long filaments (above). With the help of computer models, Chu showed in detail how converting the molecule ATP to ADP elongates each actin molecule from 8 micrometers (left column, and a) to 15 micrometers (middle column, and b). Image courtesy of Jhih-Wei Chu
To sidestep this problem, Chu uses an approach called multiscale modeling. He uses information obtained from finer-grained models to set the parameters of more coarse-grained simulations. Though not as detailed as atomistic models, these larger-scale simulations can follow longer reactions and involve structures such as cell membrane molecules – without taking months to complete.
"The approach we're taking will enable us to transfer the most important pieces of information to models at other scales. Only then can we correctly characterize the behavior of the system," Chu says. "Can there be a systematic way of doing such a transfer of information? If we have a successful methodology for one application, maybe there's a way to generalize that and apply it to many cases."
Chu's simulations have already yielded many useful insights into protein behavior. For example, proteins are often capable of twisting into several different poses, or conformations, while completing a reaction. Chu's models can not only identify which reaction pathway is most likely, but help him analyze how best to modulate that conformational change. Such drugs could prevent one portion of the protein from closing around the ligand, accelerate its gripping behavior, or block the process altogether.
While scientists have a fairly good understanding of how single proteins function, they remain more mystified at how protein assemblages work in concert. "Once they're linked together, proteins can walk, pull things from one place to another. How do different proteins get together to form a machine? How do these molecules communicate with each other at the molecular level? We think conformational change is a part of this," Chu says.
Chu has already found this to be true for actin molecules, which help muscles contract and cells move. In order to link into a long filament, individual actin molecules must "burn" a unit of chemical energy known as ATP. Using molecular simulations, Chu found that this reaction causes a coiled portion of each actin protein to loosen, transforming the structure of each molecule from rigid to floppy, and short to long. This relaxation is what allows actin to move.
"I'm a chemical engineer. We're very good at breaking big problems into small ones, understanding each step very carefully, and at the end putting things together to form a factory. So if we understand how the different units of a cell work, and how they connect to each other, we can try to understand the cell itself–the most efficient factory," Chu says.