To gain insights into the function of enzymes, much effort has gone into the determination of their three-dimensional structures. This venture has been highly successful: more than 22,000 proteins structures are now deposited in the Protein Data Bank, of which nearly 4,500 are characterized as enzymes. These static structures provides hints into how proteins function; however, the side chains surrounding the active site are not static spectators but are active participants in the choreographed motions that mediate chemical transformation. To fully understand how an enzyme functions at the molecular level, it is crucial to know the structural changes that ensue as it executes its designed function. With this knowledge, researchers will be better poised to rationally engineer proteins and peptides with therapeutic value. We have been working to develop and refine the method of picosecond time-resolved X-ray crystallography, a technique that allows us to literally ?watch? a protein as it functions. This technique is based on the pump-probe method where a laser pulse (pump) triggers a reaction in a protein crystal and a delayed X-ray pulse (probe) takes a ?snapshot? of the protein?s structure. The ID09B time-resolved beamline at the European Synchrotron and Radiation Facility (ESRF) in Grenoble, France is still the only beamline in the world capable of recording time-resolved macromolecular structures with 150 ps time resolution and < 2 ? spatial resolution. One of the challenges we have recently dealt with is how to assess structural changes in a visually intuitive fashion. Crystallographers often compare structures by generating difference electron density maps from two different structures. The difference density is portrayed as a contour map with the threshold set at an appropriate density. This approach leaves much to be desired. The electron density differences are not localized on the atoms that move but on their edges, which can make it difficult to assign the difference density ellipsoid to the atom or group of atoms that move. Only differences that exceed some specified threshold are plotted, so this approach is insensitive to small conformational changes. As the magnitude of the structural change diminishes, the difference density dips below the threshold and the feature abruptly disappears, creating a nonlinear response to the change. To surmount these limitations, we developed a novel approach for visualizing structural changes in proteins. Our approach is based on a color-coded rendering of the electron density distribution for the unphotolyzed (magenta) and photolyzed (green) states. Where these two colors overlap, they blend to white. Consequently, the magenta-to-green color gradient depicts the direction of atomic motion, and the brightness of the color gradient correlates with the amplitude of the motion. By mapping the electron density to brightness in a nonlinear fashion, weak features are not obscured by intense features. This nonlinear mapping enhances the dynamic range of structural changes that can be observed in a single image. By stitching a sequence of time-resolved images into a movie, subtle time-dependent changes in the conformation are much more apparent. This approach allows us to intuitively visualize conformational changes at an unprecedented level of detail [Schotte et al., JSB, (2004)]. While we have been engaged in studies of several different protein systems including photoactive yellow protein, here we highlight our efforts to probe ligand migration and correlated structural changes in proteins. Myoglobin, a heme protein found in muscle that reversibly binds small ligands such as O2, CO, and NO has proven to be a useful model system for pursuing these studies. Owing to the photosensitivity of the ligand bond and the reversibility of ligand binding, ligand dissociation can be triggered repeatedly by a laser pulse without damage to thhhe protein. Moreover, mutant forms of this protein can be over expressed in E coli and purified protein can be coaxed to form highly ordered crystals that diffract to atomic resolution. We have studied the L29F mutant of myoglobin (Mb), where the leucine (L) in the 29 position is replaced by phenylalanine (F). According to femtosecond time-resolved IR measurements of photolyzed L29F MbCO, the rate of ligand escape from its primary docking site is accelerated approximately 1000-fold compared to wild-type MbCO. We have acquired time-resolved structures of this mutant at time delays spanning 100 picoseconds to 3 microseconds [Schotte et al., Science (2004)]. The structural rearrangements triggered by ligand dissociation are striking, and involve correlated motion of the heme and numerous side chains. A comparison of the structural changes occurring in wild-type and L29F MbCO provide a structural explanation for the dramatic differences in the rates at which these two proteins excrete toxic CO [Schotte et al., JSB, (2004)]. Our successes in time-resolved X-ray crystallography couldn?t have been realized were it not for our ability to study the photophysics of protein crystals at the NIH with a state-of-the-art microfocusing femtosecond spectrometer. By analyzing femtosecond time-resolved spectra of photoexcited proteins in crystals, we have been able to develop protocols for attaining more efficient photoexcitation without damaging the chromophore. Currently, our analysis of time-resolved Laue diffraction images employs routines from a hodge-podge of disparate software packages and requires weeks to months of effort to fully process a few days worth of data. To eliminate the data processing inefficiencies inherent in this approach, we have been developing a stand-alone software suite capable of processing time-resolved Laue diffraction images in real time and translating those results into movies of protein motion at atomic resolution. Dr. Eric Henry (LCP) has been assisting us by developing code to automate and/or speed up several critical steps in our data processing algorithm. Preliminary tests suggest that we will be able to analyze diffraction images as fast as they are generated on the beam line. This ?real-time? feedback will help us make much more efficient use of the limited and precious beam time allocated to our research. We have also initiated a collaboration with Dr. Gerhard Hummer (LCP) to perform molecular dynamics simulations of proteins. By comparing experiment and theory at comparable resolution in space and time, we are able to gain a single-molecule perspective into mechanisms of protein function. Our combination of spectroscopic, crystallographic, and computational tools are paving the way to explore functionally-important structure transitions at an atomistic level, from which a far more meaningful mechanistic description of protein function will be achieved.
Showing the most recent 10 out of 12 publications
