There’s long been a standard tradeoff in biochemistry: You can study overall shape (of, say, a macromolecule or an organelle) or high-resolution detail within, say, at the 10-40-nanometer scale. Seeing both at the same time would be one of today’s scientific Holy Grails.
But new work by a research team at the National Biomedical Computation Resource (NBCR) at UC San Diego has taken a big step in this Grail’s direction. In basic terms, they are studying the biochemistry of how proteins move, how that movement relates to their conformational shapes (which determine their functions), and how an allosteric ligand can influence those movements.
If a researcher wants to understand a protein’s shape, current tools only allow freezing it in place, studying it for a moment in time. But a protein experiences various kinds of movements in various places, and these ensembles of movements are not captured in such static structures. And there might be many ways for a protein to move from one shape to another, not just a single linear path.
By way of analogy, think of the movements of an overeager puppy. It’s the sum of these movements over time that determines the puppy’s charm. In the same way, the sum of movements over time determine a protein’s shape and function. So, clearly a static snapshot is limited in the information it yields. Worse, it may even be misleading.
Using another analogy, a researcher on the team, Robert Malmstrom, puts it this way: “You can liken a protein to a pack of cars in rush hour. Each car—that is, protein movement—is separate from all the others but its movements respond to the movements of the others. Cars are interacting with each other, speeding up, slowing down, and changing lanes. Until this recent work, we’ve only been able to take the equivalent of a snapshot of the movement of one hour of traffic. But with the computational technologies we’re now employing, we can see traffic movement for an entire day or even a week.”
Proteins are important because they function as enzymes that catalyze nearly all biological reactions and transmit biological information within an organism. And regulating their activity is key to governing cell behavior. Susan Taylor, a member of the National Academy of Sciences and NBCR collaborator, says, “Protein kinases are key molecular switches that regulate much of biology, so they are a good target for exploring this dynamic behavior.”
In biochemistry, a particular type of protein regulation is called allostery. In this process, an effector molecule binds to a protein at a site away from where protein function occurs. This allosteric binding process often results in a change to the conformational shape of the protein, which alters the protein function, typically by promoting or inhibiting activity. That’s why this process has been a topic of ongoing study dating back to the first canonical results published in 1961.
Computational researchers have determined that the character of the mechanism of allostery results from the underlying free-energy landscape. This term describes the relationship of conformation to energy. Each conformation of a molecule is associated with a particular energy. The lower the energy, the more likely the molecule is to assume that conformation. Plotting the various energies for the various possible conformations produces the protein’s free-energy landscape.
More specifically, scientists know that protein allostery induced by ligand binding plays a central role in regulating cellular signaling pathways. In protein-ligand binding, a small molecule ligand binds to a site on a target protein distinct from the active site and triggers a biologically significant signal.
The NBCR team applied cutting-edge computational approaches to explore the conformational ensemble of a protein with and without a ligand bound to it to gain insight into how the protein ensemble gives rise to function. Intercellular activation of the protein PKA by the ligand cyclic adenosine monophosphate (cAMP) is a prototypical example of ligand-protein allostery. The inactive PKA holoenzyme contains a regulatory subunit dimer and two catalytic subunits. Allostery takes place when cAMP cooperatively binds to two cAMP binding domains, referred to as A and B, in the PKA regulatory subunits. The result is that PKA’s catalytic activity is unleashed, which activates its signal.
The research team believed that achieving an all-atom description of these conformational ensembles and resolving their corresponding free-energy landscapes would clarify the mechanisms by which cAMP modulates the function of the binding domain. To explore these conformational ensembles, they focused initially on one specific part of PKA and asked what space this domain could explore.
This work combined molecular dynamics (MD) simulations with Markov state models to explain the conformational variations of binding domain A for two states: when the binding domain was free of cAMP and when it was bound to cAMP. MD is a technique to simulate the physical movements of atoms and molecules, in this case in a protein or protein ensemble (if the cAMP is bound). By assigning individual frames extracted from the MD trajectories to discrete conformational states, it is possible to integrate sampling from many trajectories into one coherent framework that captures the kinetics and thermodynamics of the conformational ensemble at atomic resolution. Markov state models depict the interaction dynamics of these discrete conformational states. MD and Markov state models, working together, produce conformation plus movement in a single analysis.
While similar approaches have been used to study topics such as protein folding, this work is unique: It assesses the atomic conformational landscape using initial unbiased, long-timescale MD simulations augmented with adaptive sampling of the same protein in the two functional states (bound and unbound), as opposed to work that combines many short-timescale simulations or sampling along a predetermined reaction coordinate.
The research team’s results showed that binding of cAMP modifies the principal motions of the binding domain, which correspond to transitions between active and inactive states (the only high-resolution conformational states observed so far, as determined by crystallographic analysis). The results also showed that both the unbound and bound states demonstrate shallow free-energy landscapes that link different functional states through various transition pathways. Further, the researchers were able to follow propagation of the allosteric signal through key structures in the binding domain and explore how kinetics relate to the binding domain’s function. Their models support existing structural and dynamical experimental data obtained with nuclear magnetic resonance imaging and fluorescence.
What’s especially exciting, though, is that this new method can be used to integrate disparate experimental data sets into a single framework that can be applied, more generally, to describe structure, dynamics, and function in complex biological systems. Taylor says, “In the past, we have thought that structure defines function, but here we see a huge range of dynamic information that we need to interrogate before we fully understand function. We can no longer think in terms of one crystal structure, as it is only a single snapshot—hardly the whole story.”
“We can use this new approach,” says Rommie Amaro, Director of NBCR, “to quantify the dynamics and kinetics that take place at longer timescales, by ‘stitching together’ different kinds of data sets taken at much shorter timescales. This allows us to create new linkages across biological time and length scales that are otherwise enormously challenging for researchers to bridge.”
Malmstrom points out that such linkages across multiple scales is an important part of NBCR’s charter. He says, “We can evaluate how an atomic mutation changes the conformational ensemble, then predict the function of the protein. This knowledge can be integrated into larger models of how the cell or an organ, like the heart, works. Together, our models generate new hypotheses and ways of testing them, gradually extending the scale and detail of our understanding. In the case of the PKA, we can also map known disease mutations that lead to ‘gain of function’ or ‘loss of function’ phenotypes.”
What Malmstrom loves about this work is that it’s theoretical basic science yet it has obvious applications to improving health care. It’s a first important step in meeting his personal goal: Helping advance the efficacy of drug design by moving the target focus from static to dynamic protein structures. This work has produced a paper (with Malmstrom as lead author) that has just been accepted for publication in Nature Communications.
References:
Researchers: UCSD: Robert D. Malmstrom, Alexandr P. Kornev, Susan S. Taylor, Rommie E. Amaro