Watching a Protein as it Functions with 150-ps Time-Resolved X-ray Crystallography

See allHide authors and affiliations

Science  20 Jun 2003:
Vol. 300, Issue 5627, pp. 1944-1947
DOI: 10.1126/science.1078797


We report picosecond time-resolved x-ray diffraction from the myoglobin (Mb) mutant in which Leu29 is replaced by Phe (L29Fmutant). The frame-by-frame structural evolution, resolved to 1.8 angstroms, allows one to literally “watch” the protein as it executes its function. Time-resolved mid-infrared spectroscopy of flash-photolyzed L29F MbCO revealed a short-lived CO intermediate whose 140-ps lifetime is shorter than that found in wild-type protein by a factor of 1000. The electron density maps of the protein unveil transient conformational changes far more dramatic than the structural differences between the carboxy and deoxy states and depict the correlated side-chain motion responsible for rapidly sweeping CO away from its primary docking site.

To gain insight into the function of enzymes, much effort has gone into the determination of their three-dimensional structures (1). However, enzymes are dynamic molecules whose mechanistic description also requires information about intermediates along the reaction pathway. We have developed the method of picosecond x-ray crystallography and used this technique to characterize the structure of the L29F mutant of myoglobin as it evolves from the carboxy to the deoxy state. We chose this model system because transient infrared (IR) spectra of photolyzed L29F MbCO revealed a short-lived intermediate along the ligand escape pathway. Its 140-ps lifetime is comparable to the current time resolution of macromolecular x-ray crystallography and thereby provides a stringent test of its capabilities. The time-resolved “snapshots” of this mutant's crystal structure capture a “transition state” whose highly strained side chains apparently sweep CO away from its primary docking site in less than a nanosecond.

Myoglobin is an oxygen-binding heme protein (Fig. 1A) found in muscle that has long served as a model system for investigating the dynamics of ligand binding (2) conformational relaxation in proteins (3, 4). With the heme in its chemically reduced (Fe II) ferrous state, Mb reversibly binds O2, as well as other physiological ligands such as CO and NO. Of the three adducts, MbCO is most stable and easily photolyzed (5), which makes it an ideal model for time-resolved studies. Time-resolved IR investigations of photolyzed wild-type MbCO showed that CO becomes temporarily trapped in a nearby ligand-docking site before escaping into the surrounding solvent (6). This site presumably mediates the transport of ligands to and from the active binding site. Molecular dynamics simulations suggested a location for this docking site (7), which was found to be consistent with low-temperature crystal structures of photolyzed MbCO (8, 9) where “docked” CO was displaced about 2 Å from the active binding site. The docking site is fashioned by the heme and amino acid side chains valine (Val68), isoleucine (Ile107), and leucine (Leu29), all of which are highly conserved in mammalian Mb. Therefore, site-directed mutagenesis at these sites would likely influence the function of the Indeed, in the L29F mutant (where Leu29 is replaced by phenylalanine), the O2 affinity is elevated by an order of magnitude, and the autooxidation rate is lowered by a comparable amount (10).

Fig. 1.

Structure and time-resolved mid-IR spectra of wild-type and L29F MbCO. (A) The 29 position side chains for both wild-type (leucine, blue) and L29F Mb (phenylalanine, green) are shown. The heme and key nearby residues are modeled as sticks, the iron and CO as space filling spheres, and the protein as ribbon with its space-filling volume outlined. (B and C) The mid-IR spectral evolution, offset for clarity, is well described by consecutive equilibria (ABCS), where A corresponds to the CO-bound form (MbCO), B corresponds to CO trapped in a primary docking site (Mb·CO), C corresponds to CO elsewhere inside the protein (Mb··CO), and S corresponds to CO in the surrounding solvent (Mb + COaq). All rate coefficients were treated as time-independent, except kAB in wild type, which changes because of protein conformational relaxation on the same time scale as ligand escape. The least-squares determination of the time constant for B-state escape was found to be 190 ns and 140 ps for wild-type and L29F Mb, respectively (measured at 10°C). For clarity, sub-nanosecond spectra of wild-type Mb are not shown. [Fig. 1A was generated with BobScript 2.5, author R. Esnouf.]

We determined the dynamics of ligand translocation in wild-type and L29F MbCO using femtosecond time-resolved mid-IR spectroscopy (11). Briefly, the samples were photolyzed by an orange laser flash, after which a time-delayed IR pulse probed the CO stretch band around 4.8 μm. The transient IR absorbance spectra of photodetached CO (Fig. 1, B and C) reveal two promptly appearing bands that are interpreted as CO at opposite orientations within the primary docking site (6). The two bands in wild-type Mb persist for a few hundred nanoseconds ,whereas those in the L29F mutant evolve into a single, broad band within a few hundred picoseconds. The disappearance of the two bands corresponds to escape CO of from nearby primary docking site to other locations within the protein, with the translocation being faster in the L29F mutant that in wild-type Mb by a factor of 1000.

To determine three-dimensional time-resolved electron density maps of photolyzed L29F MbCO, we used picosecond Laue crystallography at the European Synchrotron and Radiation Facility (ESRF) (11). Briefly, the protein crystal was photolyzed with an orange laser flash, after which a time-delayed x-ray pulse was scattered by the sample, and its diffraction pattern was recorded (Fig. 2). Crucial to these experiments is the ability to isolate a single x-ray pulse from the synchrotron pulse train. When operated in the socalled “single-bunch” maps of photolyzed mode (implemented at the ESRF for 12 days in 2002), the time spacing between x-ray pulses is sufficient to isolate a single, intense x-ray pulse with a high-speed chopper.

Fig. 2.

Sample geometry used to acquire picosecond time-resolved xray diffraction data. The crystal was photolyzed by a laser pulse and then probed by a time-delayed x-ray pulse, whose diffraction pattern was recorded on a charge-coupled device (CCD) area detector. A custom fixture supported the MbCO crystal (0.2 by 0.2 by 0.2 mm) in a sealed capillary at a 45° angle relative to the goniometer rotation axis. The laser beam was directed along the rotation axis with the intersection of the laser and x-ray beams centered 50 μm inside the pump-illuminated face of the crystal. The laser pulse was focused to a 0.2by 0.12 mm spot (full width at half maximum, FWHM), which modestly overfilled the volume probed by the 0.1 by 0.1 mm x-ray beam. A 10°C cooling stream from an Oxford Cryostream extracted excess heat from the crystal, permitting laser photolysis at 3.3 Hz with no ill thermal effects.

The diffraction image presented in Fig. 3 contains about 3000 diffraction spots, each of which corresponds to a Fourier component (structure factor) of the electron density within the unit cell of the crystal. By recording diffraction images over a range of crystal orientations, a nearly complete set of structure factors is experimentally determined. The inverse Fourier transform of the structure factors reconstructs the electron density within the unit cell of the crystal (11). Because the level of photolysis is incomplete (∼23%), the measured structure factors have contributions from both photolysed and unphotolysed states. To generate electron density maps of the photolyzed state, we extrapolate the partially photolyzed structure factors to complete photolysis. High-resolution electron density maps of L29F MbCO before and after photolysis are shown in Fig. 4; color contrast is used to highlight the structural differences (12). This color-coded presentation of the experimentally determined electron density requires an estimate of the level of photolysis; however, this estimate need not be precise, as minor variations in the degree of photolysis alter only the absolute magnitude of the motion, not the relative magnitude nor its direction. Because each structure is determined from data collected before and after photolysis in an interleaved fashion on the same crystal, the photolysis-induced structural differences are revealed with high precision, despite the relatively low level of photolysis.

Fig. 3.

One of 155 time-resolved Laue diffraction patterns recorded from a single, low-mosaicity P6 crystal of L29F MbCO. This image contains about 3000 diffraction spots, each of which corresponds to a Bragg reflection from a set of crystallographic planes. Many spots are simultaneously excited by the undulator radiation, whose useful intensity range spans about 0.72 to 1.24 Å. Only a small fraction of these spots would appear when probed with monochromatic x-ray radiation. The enlarged view identifies Miller indices associated with the crystallographic planes that generated the corresponding Bragg reflections. This 10-s exposure, acquired midway through the series, represents the cumulative diffracted intensity from 32 x-ray pulses with the pump-probe delay set to 100 ps.

Fig. 4.

(A) Experimentally determined electron densities within a 6.5-Å-thick slice through the myoglobin molecule before (magenta) and 100 ps after (green) photolysis. Where these densities overlap, they blend to white. The white-stick model corresponds to the unphotolyzed structure and is included to guide the eye. The direction of molecular motion follows the magenta-to-green color gradient. Three large-scale displacements near the CO-binding site (large arrows) are accompanied by more subtle correlated rearrangements throughout the entire protein (small arrows; not drawn to scale). (B to G) Enlarged views of the boxed region in panel (A) at time delays specified in the lower left of each panel. (B) Upon photolysis, the bound CO (magenta; site 0) dissociates and becomes trapped about 2 Å away in a docking site (green; site 1), close to the Phe29. To accommodate the docked CO, the Phe29 is displaced and rotates, pushing His64 outward, which in turn dislodges a bound surface-water molecule (magenta). The electron density surrounding Phe29 appears to arise from a mixture of two conformations, one rotated and displaced toward His64 and the other already relaxed to a conformation near the deoxy state. (C to E) From 316 ps to 3.16 ns, the structural changes are largely confined to the vicinity of the binding site. As the CO migrates from site 1 to site 3 (Xe4), the Phe29 and His64 relax toward their deoxy conformations, which are similar to their unphotolyzed states. (F and G) Structures acquired after nanosecond photolysis of a second protein crystal. By 31.6 ns, CO has migrated to sites 4 and 5, where it remains trapped for microseconds. The magenta and green maps were contoured at the same absolute level (1.5 σ of the unphotolyzed density) using O 7.0, author A. Jones. [See Movie S1 to view the time-dependent electron density changes.]

Correlated displacements of the heme the protein backbone, and other side chains are evident throughout the protein at 100 ps (see Fig. 4A), which is consistent with spectroscopic studies that report significant conformational relaxation by that time (13, 14). The heme tilts toward the distal residues, with a hinge point located near the propionate side chains. The driving force for this motion resides in a photolysis-induced displacement of the theme iron: Atomic resolution structures of Mb show the iron in the plane of the heme in MbCO but displaced 0.30 Å (15) or 0.36 Å (16) in the proximal direction in Mb. This displacement generates downward pressure on the His93 and upward pressure on the heme. The strain is relieved largely by the tilt of the heme and, to a lesser extent, by displacement of the proximal histidine. As the heme tilts, it drags with it the side chains Leu104 and Ile107 and the G helix to which they are attached. The largest amplitude motion occurs near the binding site and involves displacement of CO, Phe29, and His64. The displacement of these side chains is much more dramatic than the static structural differences between deoxy Mb (Protein Data Bank 1MOA] and MbCO (Protein Data Bank 2SPL). To accommodate the ∼2 Å(15) or displacement of the docked CO (labeled 1 in Fig. 4), the Phe29 twists and moves upward, pushing His64 away from the ligand-binding site, a direction of motion nearly opposite that observed on longer nanosecond time scales or in the equilibrium state with no bound ligand (15, 16). Concomitant with the ejection of CO from the primary docking site, the Phe29 and His64 residues relax toward their deoxy Mb configuration. Apparently, the strain energy retained by these side chains accelerates the departure of docked CO 1000 times compared with the rate of CO escape from wild-type Mb; this strain energy corresponds to an Arrhenius energetic bias of 16 kJ mol1 at 283 K (about half the wild-type activation energy barrier for ligand departure from the primary docking site).

As CO escaped from site 1, we observed electron density appearing in sites 2 and 3. Site 3 corresponds to the so-called Xe4 site, one of several internal hydrophobic cavities found to harbor Xe under pressure (17). The CO residence time in the Xe4 site is rather short-lived: It departs by 32 ns and migrates around the heme to sites 4 and 5 (Xe1), where it persists for microseconds. Ligand migration from Xe4 to Xe1 presumably requires a protein fluctuation to permit passage around the heme, an event that occurs within a few tens of nanoseconds. The Xe4 intermediate observed here has not been seen in wild-type Mb, where the rate of CO escape from the primary docking site is slow compared with the rate of CO migration from Xe4 to Xe1. Therefore, L29F-induced destabilization of CO in the primary docking site made it possible to characterize the protein dynamics associated with CO passage around the heme. Because CO remains sequestered in hydrophobic cavities at 3.16 μs (see Fig. 4G), time-resolved structures at longer times will be needed to identify pathways for ligand escape into the surrounding solvent.

In earlier work, the Xe1 site was shown to harbor CO indirectly through mutagenesis and Xe studies (18), as well as directly by pioneering nanosecond time-resolved crystallographic studies (19). In addition, cryocrystallography studies of photolyzed L29W (20) and L29Q MbCO (21) found CO in the Xe4 site, whereas studies of wild-type horse MbCO found CO in the Xe1 site (22). When photolysis of L29W MbCO was followed by thermal cycling above the Mb glass transition temperature, CO was found to migrate from Xe4 to Xe1 (20), as observed here in real time. The highly strained intermediate observed in our 100-ps snapshot has not yet been trapped under cryogenic conditions, and it is unlikely that it could be. Moreover, cryocrystallography cannot assess protein dynamics under physiologically relevant conditions where thermally driven conformational fluctuations lead to rapid interconversion among conformational substates (23). Nevertheless, the method of cryocrystallography, with its higher spatial resolution, is a useful complement to time-resolved crystallographic methods.

The static structures of thousands of proteins are known, yet there is no single protein whose function is fully understood at an atomistic and deterministic level. In this time-resolved study, the interplay between a protein and a docked ligand is clearly revealed, with picosecond correlated side-chain motion providing a structural explanation for the ultrafast departure of docked CO. By extending the time-resolution of crystallography into a time domain readily accessible to molecular dynamics simulations, we pave the way for exploring functionally important structure transitions both theoretically and experimentally. Whereas the time resolution of macromolecular crystallography at the ESRF is, for all practical purposes, currently limited by the x-ray pulse duration to ∼100 ps, the chemical time scale will be accessible to x-ray free electron lasers, which promise to deliver intense x-ray pulses shorter than 100 fs (24).

Supporting Online Material

Materials and Methods

Table S1

Movie S1


References and Notes

View Abstract

Navigate This Article