Physical Mechanism
Proteins are often described as molecular machines, carrying out complex chemical reactions through cycles of conformational change between different functional states. Like macroscopic machines, the essence of understanding how proteins work lies in defining their mechanics – the motions they can perform and the forces constraining these motions. Unlike man-made machines, however, proteins are the product of an evolutionary process that produces marginally stable materials that display a vast number of degrees of freedom operating over many time scales. Somewhere embedded within this sea of high-dimensional and likely irrelevant fluctuations, we expect a low-dimensional pattern of collective functional motions that are under selection and that represent the biological reaction coordinate. Isolating and visualizing these relevant motions is the essence of understanding protein function – the concept of a protein as a machine.
How can we “see” the biologically relevant motions of proteins? How can we use these data to infer the pattern of forces constraining the motions? How will that pattern relate to the pattern of statistical interactions between amino acids that we learn from coevolution based methods such as SCA or DCA? To address these fundamental questions, our strategy over the past several years has been to collect a ton of data using multiple technical approaches, all focused on a single model system (the PDZ domain). The idea is that such a systematic dataset in one protein family will provide a basis for a rigorous evaluation of computational and statistical models for protein mechanism:
(1) Global NMR chemical shift perturbation analysis. In one approach, we examined the effect of mutating each position in the PDZ domain, individually, on chemical shifts of nuclei distributed throughout the rest of the protein – a sensitive, global structural response profile for perturbation of each site. A correlation analysis of the response profiles shows (a) that most positions display idiosyncratic, ligand-independent, local coupling within the tertiary structure, but (b) that a few positions display highly correlated, ligand-dependent, and distributed response to mutation. The latter links the ligand binding pocket of the PDZ domain to two known allosteric sites – the α1-β4 surface, and the β2-β3 loop.
(2) High-order structural cycles. In this approach, we studied the context dependence of mutational effects within the PDZ domain using high-resolution (<2Å) crystal structures of single and double mutants. Remarkably, the difference in the structural effect of mutation at an allosteric site (G330T, in the β2-β3 loop) in the context of a mutation in the ligand (T-2F) reveals a contiguous allosteric pathway of amino acids that connects the β2-β3 loop to the α1-β4 surface through the ligand-binding pocket.
(3) Homolog averaged room-temperature X-ray structures. In yet another approach, we solved very high-resolution (<1.5Å) structures of 11 sequence-diverged homologs of the PDZ family at “room temperature” (277K). By carrying out ensemble refinement against the data for each homolog, we compute the correlated motion of every pair of amino acids. Averaging over all 11 homologs and examining the top significant eigenmodes, we deduce the existence of a small number of evolutionarily conserved collective mechanical modes within the ground state of the PDZ domain. These modes include the ligand-binding pocket and known allosteric sites.
(4) Electric field-stimulated protein mechanics (EFX). The previous approaches represent the state-of-the-art in probing the internal functional mechanics of proteins, but ultimately, are insufficient for making physical models for proteins. Most importantly, there is no direct visualization of dynamics and thus no way to classify time-scales or causality of motions, and there is no obvious way to go from observation of collective modes to the underlying forces that drive those motions. Starting in 2011, we began developing a new approach, called electric-field stimulated X-ray crystallography or EFX. The idea is that the random distribution of charged species (fixed and dipolar) throughout protein molecules can be seen as actuators for driving motions with applied external electric field pulses. If this can be done in conjunction with time-resolved X-ray diffraction in crystals, we can subtly excite and record the natural internal motions of proteins with high spatial and temporal resolution. In essence, we can make movies of protein motions driven by applied external forces of known magnitude and orientation. For the PDZ domain, EFX reveals time-resolved motions in the hundreds of nanosecond time-scale that include both binding-site and allosteric surfaces in the β2-β3 and the α1-β4 loops. EFX is reported in Hekstra DR, White KI, et al. Electric-field-stimulated protein mechanics. Nature. 2016 540:400. 
The bottom line is that all four approaches tell basically the same story about the internal mechanics of the PDZ domain: A small number of evolutionarily conserved collective mechanical modes exist in the PDZ domain, with a functional mode that connects the ligand-binding pocket to at least two biological allosteric sites, the β2-β3 loop and the α1-β4 surface. These very same regions are linked by the protein sector, the statistically coevolving unit of amino acids in the PDZ family, provide strong support that protein sectors are indeed collective mechanical modes within proteins.The future: For us, the direction of physical mechanism of proteins lies in direct observation of intramolecular dynamics with atomic resolution and over as broad a range of time-scales as possible. We will focus on two approaches: (1) EFX, and its derivatives, and (2) time-resolved serial microcrystallography (TR-SMX). EFX has considerable promise in exposing the picosecond to microsecond time-scale dynamics of proteins and in principle, for any protein for which well-diffracting crystals can be grown. This opens the possibility of making movies of protein motions while varying the amplitude and orientation of the applied electric field. In principle, this means we can carry out force-extension analyses with atomic resolution in proteins, and try to map the shape of free-energy landscapes governing specific motions. TR-SMX is an approach for examining nanosecond to much slower motions in proteins in which reactions are initiated within proteins by either temperature jumps or diffusion of small molecules, and structures solved as a function to time delay after reaction initiation. Together, these approaches have the potential to directly expose functional motions in proteins, a starting point for making proper physical models.
We note that both EFX and TR-SMX require either specialized synchrotron facilities for time-resolved X-ray diffraction or X-ray free-electron laser (XFEL) sources. We are closely partnering with BioCARS (sector 14, Advanced Photon Source, Argonne National Labs) and LCLS (Stanford Linear Accelerator Center) to make these experiments possible and broadly available to the scientific community.
Center for Physics of Evolution
Biochemistry & Molecular Biology
The Institute for Molecular Engineering
The University of Chicago
929 E. 57th Street Chicago, IL 60637
