Time-resolved crystallography reveals allosteric communication aligned with molecular breathing

See allHide authors and affiliations

Science  13 Sep 2019:
Vol. 365, Issue 6458, pp. 1167-1170
DOI: 10.1126/science.aaw9904

Active sites that move together

Enzymes often form dimers or higher-order oligomers, even when each active site is isolated and the reactions are simple. But the effect of a neighbor can be profound. Mehrabi et al. used a photolabile compound to initiate a reaction in the enzyme fluoroacetate dehalogenase, which they could follow by time-resolved serial synchrotron crystallography. Snapshots of the reaction revealed coupled allosteric motions between the two active sites of the dimeric enzyme. Each active site traded the ability to bind substrate and catalyze the reaction, such that only one was engaged at a time. This behavior is common in enzymes but is rarely visualized and still poorly understood.

Science, this issue p. 1167


A comprehensive understanding of protein function demands correlating structure and dynamic changes. Using time-resolved serial synchrotron crystallography, we visualized half-of-the-sites reactivity and correlated molecular-breathing motions in the enzyme fluoroacetate dehalogenase. Eighteen time points from 30 milliseconds to 30 seconds cover four turnover cycles of the irreversible reaction. They reveal sequential substrate binding, covalent-intermediate formation, setup of a hydrolytic water molecule, and product release. Small structural changes of the protein mold and variations in the number and placement of water molecules accompany the various chemical steps of catalysis. Triggered by enzyme-ligand interactions, these repetitive changes in the protein framework’s dynamics and entropy constitute crucial components of the catalytic machinery.

Understanding the structural dynamics of biomolecular catalysis requires atomically resolved structures acquired continuously along an enzymatic reaction pathway—a “molecular movie.” One of the few methods that approaches providing this information is time-resolved crystallography (1, 2). Whether performed at synchrotrons or, more recently, x-ray free-electron laser sources, the most challenging experimental task is homogeneous reaction initiation throughout the entire crystal (3). This is commonly achieved by exploiting naturally occurring light-sensitive proteins, in which reactions can be triggered by ultrashort laser pulses to then be observed at time scales typically ranging from femtoseconds up to milliseconds (16). However, in irreversible systems that are not naturally light sensitive, i.e., most enzymes, reaction initiation is more complicated. Most enzymes are rather slow to turn over, with a median of 10 s−1 (7). For these systems, a photolabile caged substrate is an elegant way to probe conformational changes associated with substrate binding and catalysis over milliseconds to seconds (4, 8).

A well-studied model system of the structural basis of enzyme dynamics is Rhodopseudomonas palustris fluoroacetate dehalogenase (FAcD), a homodimeric α/β-hydrolase that displays half-of-the-sites reactivity and cooperativity, features commonly observed in enzymatic catalysis (911). FAcD hydrolyzes the C–F bond of fluoroacetate (FAc), with the help of a conserved catalytic triad (Asp110, Asp134, and His280). With a turnover rate in the tens of seconds, it is one of the slower enzymes (9). Studies into the role of this enzyme’s oligomeric state, its allosteric behavior, and the defining elements of its rate-limiting step have provided first, but still indirect, answers (10, 11). Analysis of enzyme mutants can create structural and functional artifacts, and freeze-trapping of enzyme states likewise will inevitably miss dynamic motions present at higher temperatures and can perturb the protein energy landscape (12). We recently demonstrated the potential for time-resolved serial synchrotron crystallography (TR-SSX) to efficiently collect snapshots along reaction coordinates of wild-type enzymes at room temperature (13) and now apply this methodology to understand the allostery and dynamics of fluoroacetate dehalogenase over the course of four full catalytic turnovers.

We collected data on FAcD crystals at time points ranging from 30 ms to ~30 s after initiation of the reaction by light, all at resolutions between 1.7 and 1.9 Å. An unreactive, caged substrate soaked into the solvent channels of the FAcD crystals was uncaged by an ultrashort ultraviolet-laser pulse in a reaction that is orders of magnitude faster than the enzymatic reaction (11). The liberated substrate is free to diffuse into the active site as the catalytic steps are observed by TR-SSX (Fig. 1, A and B). Eighteen consecutive structures from this data-collection strategy reveal initial binding of the substrate, its reorientation in the active site, a sequential covalent ester intermediate, positioning of a water molecule for hydrolysis of the intermediate, and product release and enzyme regeneration (figs. S1 to S3). Although coupled, these processes occur sequentially and in mutually exclusive reaction centers in a visual representation of half-of-the sites reactivity.

Fig. 1 TR-SSX reveals key biochemical events of FAcD catalysis.

(A) Reaction diagram of FAcD uncaging and FAc hydrolysis. (B) Cartoon overview of the key biochemical states resolved by TR-SSX. (C) Representative structural snapshots of FAcD’s active sites for the key intermediates shown in (B).

After photolytic release of the substrate, FAc, the first weak coordination to the active site, can only be observed at the 100-ms time point. By contrast, calculations of typical diffusion times from solvent channels to substrate-binding sites estimate that this process should occur on the order of nanoseconds (4, 14). The delay in this case may indicate that FAc binding into the first active site is not a diffusion-limited process.

After 188 ms, FAc is clearly bound in the active site of subunit A, closely recapitulating the Michaelis–Menten (MM) or substrate-bound complex previously observed in a catalytically inactive mutant of the enzyme (9); at the same time, the active site of subunit B remains empty. The MM complex stays largely unchanged for ~1 s except for a subtle (0.5 Å) relocation of the ligand deeper into the active-site pocket. The longevity of this complex suggests that it displays a predominant ground state conformation before further substrate reorientation (Fig. 1C and figs. S2 and S5). Molecular dynamics simulations indicate that the ligand can sample additional conformations in the active site, the majority of which are nonproductive, as indicated by their orientation or distance to the catalytic residues (fig. S4 and movie S1). This is consistent with the crystal structure of the 2052-ms time point, which clearly shows an additional, reoriented ligand conformer where the fluorine has rotated ~160° around the axis of FAc’s C–C bond (Fig. 1 and figs. S2 and S5). These structures demonstrate the conformational flexibility of the substrate molecule in the active site, likely mirroring the enzyme’s sampling of conformational space before formation of the covalent intermediate. These multiple orientations might reflect a still not evolved “fit” between FAcD and FAc and explain the slow turnover time. Indeed, FAcD’s active site is large given the size of its substrate, indicating that FAc might not be its natural substrate (see supplementary text).

Approximately 2 s after substrate binding (2256 ms), the electron density of a covalent intermediate is visible alongside two other FAc conformations (Fig. 2 and fig. S1). We postulate that the latter conformations display the MM complex and the FAc conformation poised for second-order nucleophilic displacement (SN2) attack, respectively (fig. S6 and supplementary text). Regeneration of the free enzyme through ester hydrolysis involves a water nucleophile that adopts the correct geometry for nucleophilic substitution (at the 2256-ms time point). This requirement is met best by a water molecule near the base of the catalytic triad (His280) and the carbonyl-carbon of the covalent ester intermediate (Fig. 2 and fig. S6). At the same time, active-site B is now populated with electron density that largely corresponds to the conformation observed previously during the binding period in active-site A (Fig. 1 and figs. S3 and S8).

Fig. 2 Strong electron density unambiguously defines covalent intermediates.

(A) Polder-OMIT maps (2σ) for covalent intermediates in both subunits at 2256 and 6788 ms, respectively. (B) Active-site residues of FAcD (yellow) with interatomic distances and angles to the covalent intermediate (green) and an active-site water (red sphere), prepared to hydrolyze the Asp110-FAc ester bond.

About 2 s after formation of the covalent intermediate (4512 ms), product glycolate is bound and it is released ~2 s later (6156 ms) (Fig. 1, fig. S2, and supplementary text). Seven seconds (6788 ms) after initiation, active-site B contains the covalent intermediate and a properly positioned water (Fig. 2 and figs. S6 to S8). Active-site A is now again occupied by an FAc substrate (Fig. 1 and figs. S2 and S5), repeating the allosteric process already observed during formation of the first covalent intermediate.

Yet another ~2 s later (9024 ms), active-site A still contains the MM complex, whereas the substrate in active-site B has now been hydrolyzed. The structure of FAcD active-site A at the following ~12 s time point (12312 ms) matches exactly the conformation observed at 2052 ms after reaction initiation (fig. S2). The observation of the full catalytic cycle in both active sites illustrates the reproducibility of conformational sampling during the sequential catalytic events. The turnover kinetics in the crystal are in rough agreement with solution measurements (9) despite the very different environment. We therefore interpret ligand density of the last two time points (13,536 and 27,072 ms) as glycolate, corresponding to four full catalytic turnovers of FAcD (Figs. 1 and 3 and figs. S2 and S3).

Fig. 3 Molecular breathing correlates catalytic snapshots with structural dynamics.

Bar graphs for each time point show the correlative behavior of molecular breathing, single point energies, intermolecular contacts, normalized water content, and interfacial waters. Colors of cartoons are the same as in Fig. 1B.

The FAcD structure displays global-breathing motions correlated with biochemical events. The lateral distance between the Cα atoms of the two symmetrically related Pro32 residues in FAcD’s two subunits changes by ~1 Å (~1.3%). After contracting during the binding of FAc (0 to 100 ms), the distance reaches a lower limit of 81.7 to 81.8 Å (188 to 1182 ms) before increasing markedly to 82.4 to 82.9 Å (2052, 6156, and 12,312 ms) immediately before covalent intermediate formation (2256 and 6788 ms), after which it decreases again to 81.8 Å (Fig. 3 and fig. S9). The normalized water content of the FAcD dimer fluctuates with a very similar periodicity (Fig. 3 and fig. S9). The water count substantially increases upon ligand binding and dimer contraction (~50% increase, 0 to 282 ms time points) but decreases markedly during FAcD expansion immediately before covalent intermediate formation (~60% decrease, 2052 ms) (Fig. 3 and fig. S9). Then, upon covalent intermediate formation and FAcD contraction, it rises again (~20% increase, 2252 ms). This anticorrelation of molecular breathing and water content continues, although it is less pronounced for the later time points. Such “washing out” of detailed observations is expected, as they become more spurious at longer time points owing to the increasing inherent heterogeneity that arises with catalysis in ensembles (15).

Focusing on the interface between the two subunits of FAcD, where water networks could bridge functional areas and facilitate information transfer (16, 17), reveals that before MM complex formation (0 to 100 ms), there is only a single tightly bound water molecule between Ser157 and its symmetry mate from the other subunit (Fig. 4). Upon substrate binding (188 to 752 ms), this number rises to four, accompanied by an increase in intermolecular contacts and resulting in rigidification of the structure (Figs. 3 and 4 and fig. S9). These water molecules are located on the same plane as the substrate molecules. The loops accommodating them connect directly to the active sites (Fig. 4 and fig. S9). Upon covalent intermediate formation (1128 to 2256 ms), fewer water molecules are found, and the network of intermolecular contacts is broken.

Fig. 4 Interfacial waters and hydrogen bond network connecting FAcD’s active sites correlate with catalytic turnover.

(A) Model of the position of the water molecules at the dimer interface generated by superposing the MM complex (752 ms) with the product complex (4512 ms). (B) Magnification of the dimer interface at different catalytic states: relative position of water molecules, a cartoon representation of the structural coordinates (waters at different time points are displayed in different colors), and intermolecular contacts (hydrogen bonds are shown in green).

Fluctuations in the hydrogen bond network and the water content should also be reflected in the systems’ total energy. We calculated the single-point energy for each time point as these correlate with the enzyme’s catalytic state. The system energy is reduced upon substrate binding and increases before covalent intermediate formation. A sharp reduction of the system energy can be observed for the covalent intermediate structure, again followed by an increase during product formation (Fig. 3 and fig. S9), consistent with previous quantum mechanical and molecular mechanical calculations on a homolog of FAcD (18).

We conclude that the catalytic cycle begins with productive binding of substrate, dimer contraction, and an increase in structural waters in the interfacial region accompanied by rising structural rigidity, demonstrated by increasing intermolecular contacts (Figs. 3 and 4 and fig. S9). During an extended lag phase, the dimer remains in a contracted state and molecular breathing communicated allosterically between the two active sites precedes covalent intermediate formation. FAcD shows its largest lateral dimension before productive binding and before covalent intermediate formation. The number of structural water molecules at the dimer interface is lowest at a time when the system is most dynamic. A plausible trajectory of information transfer runs across two symmetry-related α-helices, which are bridged by four structural water molecules located in an interfacial cavity. Residues from these helices directly link to active-site residues involved in substrate binding and catalysis (Figs. 3 and 4). The water content in the interface and the normalized water content decrease as the dimer expands, potentially modulating substrate dynamics, thereby sampling states potentially more suited to an SN2 attack (Fig. 3 and fig. S9). Overall, there is no extensive movement in the protein frame throughout the whole catalytic cycle, consistent with the idea of “allostery without conformational change” (19) (supplementary text). In dimeric FAcD, allosteric regulation does not involve large structural motion, but rather dynamic long-range effects facilitated by water molecules at the subunit interface.

Supplementary Materials

Materials and Methods

Supplementary Text

Figs. S1 to S9

Table S1

References (2057)

Movie S1

References and Notes

Acknowledgments: We acknowledge DESY (Hamburg, Germany), a member of the Helmholtz Association HGF, for the provision of experimental facilities. TR-SSX diffraction data were collected at the PETRA III storage ring at beamline P11 and at beamline P14, operated by EMBL Hamburg. We thank F. Diederich and R. De Gasparo, ETH Zurich, for the synthesis of caged fluoroacetate. We also thank T. Schneider, G. Bourenkov, O. Lorbeer (EMBL, Hamburg), and A. Burkhardt (DESY, Hamburg) for exceptional assistance during the beamtime. We gratefully acknowledge support in data processing by W. Kabsch (MPIMF, Heidelberg) and K. Diederich (University of Constance). We thank colleagues S. Hayes, S. Horrell, M. Agthe, F. Westermeier, K. Siddiqui, H.-G. Duan, A. Jha, A. Sarracini, J. Besaw, K. Krawczyk, and W. Stuart for assistance during data collection. Funding: Support was provided by the Max Planck Society and by the Cluster of Excellence “The Hamburg Centre for Ultrafast Imaging” of the Deutsche Forschungsgemeinschaft (DFG) - EXC 1074 - project ID 194651731 (R.J.D.M.), the People Programme (Marie Curie Actions) of the European Union’s Seventh Framework Programme (FP7/2007–2013) under REA grant no. 623994 (H.M.M.-W.), the Natural Sciences and Engineering Research Council of Canada (RGPIN-2015-04877), the Canada Research Chairs program, and the Burroughs Wellcome Fund via a Collaborative Research Travel Grant (E.F.P.). P.M. was the recipient of an Alexander von Humboldt-Stiftung postdoctoral research award. Author contributions: P.M., E.C.S., H.M.M.-W., R.J.D.M., and E.F.P. designed the experiment. E.C.S., P.M., and H.M.M.-W. performed the experiments with support from F.T. and E.F.P. P.M. prepared the protein crystals. F.T. designed the experimental end station and developed the interlacing algorithms. R.D. designed and performed molecular dynamics simulations. E.C.S. and P.M. processed and analyzed the diffraction data. P.M., E.F.P., and E.C.S. wrote the manuscript. All authors discussed and corrected the manuscript. Competing interests: The authors declare no competing interests. Data and materials availability: Coordinates and structure factors have been deposited in the Protein Data Bank under accession codes 6QHY, 6QHV, 6QHU, 6QHT, 6QHS, 6QHQ, 6QHP, 6QHW, 6QHX, 6QHZ, 6QI0, 6QI1, 6QI2, and 6QI3 (see supplementary materials). All other data are available in the main text or the supplementary materials.

Stay Connected to Science

Navigate This Article