Research Article

Proton uptake mechanism in bacteriorhodopsin captured by serial synchrotron crystallography

See allHide authors and affiliations

Science  05 Jul 2019:
Vol. 365, Issue 6448, pp. 61-65
DOI: 10.1126/science.aaw8634

Refilling the proton pump

Proteins are dynamic. Rearrangements of side chains, secondary structure, and entire domains gate functional transitions on time scales ranging from picoseconds to milliseconds. Weinert et al. used time-resolved serial crystallography to study large conformational changes in the proton pump bacteriorhodopsin that allow for redistribution of protons during the pumping cycle. They adapted methods used for x-ray free electron lasers to synchrotron x-ray sources. Large loop movements and a chain of water molecules were central to regenerating the starting state of bacteriorhodopsin.

Science, this issue p. 61


Conformational dynamics are essential for proteins to function. We adapted time-resolved serial crystallography developed at x-ray lasers to visualize protein motions using synchrotrons. We recorded the structural changes in the light-driven proton-pump bacteriorhodopsin over 200 milliseconds in time. The snapshot from the first 5 milliseconds after photoactivation shows structural changes associated with proton release at a quality comparable to that of previous x-ray laser experiments. From 10 to 15 milliseconds onwards, we observe large additional structural rearrangements up to 9 angstroms on the cytoplasmic side. Rotation of leucine-93 and phenylalanine-219 opens a hydrophobic barrier, leading to the formation of a water chain connecting the intracellular aspartic acid–96 with the retinal Schiff base. The formation of this proton wire recharges the membrane pump with a proton for the next cycle.

Proteins are intrinsically dynamic molecules that follow defined conformational rearrangements to drive the fundamental biochemistry of life. Structural studies of these dynamic processes are challenging, because structural intermediates are often short-lived. Pump-probe crystallography offers the opportunity to capture successive temporal snapshots of light-activated processes to study protein dynamics within the crystallographic lattice at the level of individual atoms, at physiological temperatures and on wild-type proteins (1).

X-ray free electron lasers (XFELs) and the emerging method of time-resolved serial femtosecond crystallography (TR-SFX) have brought exciting opportunities for structural biologists (2, 3). Recent studies on photoactive yellow protein (4, 5) show that TR-SFX is a powerful new tool for studying the structural dynamics of proteins within a crystalline environment. The ultrafast but bright x-ray pulses from these powerful linear accelerators outrun most radiation damage effects (6) and allow the use of small crystals, resulting in high protein activation levels (4). The serial injection of crystals further favors the study of nonreversible reactions and the collection of redundant datasets of high quality. The great potential of the TR-SFX method for studying membrane transport can be realized when comparing results (7) from freeze trapping studies of the light-driven proton pump bacteriorhodopsin (bR) to a whole series of structural states ranging from the femtosecond (8) to the early millisecond regime (9).

Primary energy conversion processes in biology are variations around one central theme: proton transfer across membranes. Time-resolved crystallographic studies provided much insight into how proton pumping is achieved in bR. In particular, they focused on the early events in the pumping cycle, including the ultrafast process of energy capture through photoisomerization of retinal (8) and the proton release step from the retinal Schiff base (SB) toward the extracellular side of the membrane (9). The fundamental piece missing in our dynamic view of proton transport is how the protein rearranges to reprotonate the SB from the cytoplasmic side to reload substrate for the next pumping cycle.

We answer this question by adapting injector-based serial crystallography for time-resolved measurements in the millisecond range at widely available synchrotron sources. Requirements for time-resolved crystallography using the Laue method were never met for bR (7). Therefore, pioneering works to determine the structures of spectroscopic intermediates were carried out using freeze-trapping and mutagenesis (10). Although mutagenesis and freeze trapping provide valuable insights and allow trapping of individual structural intermediates, major controversies remained because mutations modify critical activation switches and structures of mutants may not adequately reflect rearrangements in the native protein. Such structures may further suffer from radiation damage, are obtained at nonphysiological temperatures, and lack the temporal dimension critical to understanding protein dynamics. Serial millisecond crystallography (SMX) approaches work well with high-viscosity injectors (11), as the injectors extrude crystals slowly and allow sufficient x-ray exposure to collect high-resolution data with the photon flux available at synchrotrons (12). Our recent work has shown that modern high–frame-rate and low-noise detectors make SMX a viable method for routine room-temperature structure determination (13). By extending the setup with a simple class 3R laser diode, we enable time-resolved studies with millisecond time resolution (TR-SMX) and bring dynamic serial crystallography from the few operational XFELs to the large community of synchrotron users worldwide. We demonstrate the technology by recording the temporal evolution of two structurally distinct intermediate states of bR and their decay back to the dark state. The closed state obtained by TR-SMX at 0 to 5 ms shows the structural rearrangements necessary for proton release that are associated with the spectroscopic M-state of bR with comparable quality and in agreement with previous XFEL experiments. The open state that we obtained from data collected 10 to 15 ms after activation is characterized by large conformational rearrangements up to 9 Å on the cytoplasmic side, which have been associated with the spectroscopic N-state of bR and have not yet been resolved by dynamic crystallography. These open the protein for the formation of a Grotthuss proton wire, a transient water chain that allows the reprotonation of the SB from Asp96, the primary proton donor at the cytoplasmic side of the membrane. The observations align with the classical alternate access model of membrane transport and further complete the dynamic view of vectorial proton transport in bR by resolving the critical proton uptake step at the end of the pumping cycle.

Time-resolved serial millisecond crystallography

The most straightforward experiment possible with our TR-SMX setup (Fig. 1) leaves the laser diode on while data from continuously light-activated protein molecules within the crystal are being collected. By comparing these data to serial data collected without illumination (fig. S1), it is possible to reveal dominant conformational changes in a photocycle analogous to steady-state spectroscopic techniques. From bR microcrystals with a size of about 35 μm by 35 μm by 3 μm, we collected 114,052 light and 119,374 dark diffraction patterns in the steady-state mode with data extending to a resolution of 1.8 Å (table S1). The quality of difference electron density maps [Fo(light) − Fo(dark)] improved with the number of included images with characteristic difference density peaks starting to plateau at about 10,000 images, indicating data convergence (fig. S2). The estimated activation level was 38% (see materials and methods) and resulted in well-resolved difference density map features, with the strongest positive and negative peaks at 7.3 and 8.5σ, respectively. Structure factor extrapolation allows selective refinement of the activated species in a crystal by removing the contribution of nonactivated states from structure factor amplitudes (14) (fig. S3 and materials and methods). Refinement against extrapolated steady-state data revealed two activated species of bR (table S2). Although one subspecies had been observed in previous TR-SFX structures obtained at 1.7 and 8.3 ms after photoactivation, which we termed the “closed” state, the second subspecies featured large rearrangements on the cytoplasmatic side, indicating that the dynamic view of proton pumping in bR had not yet been completed in the previous studies; hence, we termed this intermediate the “open” state (table S2).

Fig. 1 Experimental setup and evolution of structural intermediates over time.

(A) Experimental pump-probe setup and TR-SMX data collection. The laser (green) originates from a class 3R laser diode and is controlled by way of two steering mirrors before being focused to intersect with the extrusion path of the high-viscosity injector (gray) and the x-rays (red). The laser and detector are triggered by a delay generator. Running the detector at 200 Hz, the data collection cycle consists of 40 frames each 5-ms long. The laser is activated for 5 ms during collection of the first data frame and is then turned off for the remaining frames, during which the illuminated bR completes its photocycle. (B) Large-scale conformational changes on the cytoplasmic ends of helices E, F, and G at different timepoints [dark state in purple, closed state (0 to 5 ms) in blue, and open state (10 to 15 ms) in red] are evident in difference electron density maps [left: Fo(0 to 5 ms) − Fo(dark); right: Fo(10 to 15 ms) − Fo(dark) at 3σ with positive difference density in blue and negative difference density in yellow]. (C) Displacement occurs on the cytoplasmic side at the end of helix F. (D) Combining the dark-, closed- and open-state models included as alternative conformations into a single model and refining their occupancies against collected time-resolved data provide a measure of the structural evolution of the two intermediates over time. Biexponential fits against the normalized data are shown with solid lines. The fit resulted in decay times of 23.8 ± 2.7 ms for the open conformation and 4.7 ± 0.3 ms for the closed conformation.

To follow the rise and decay of these intermediates and to obtain separate structures, we developed a data collection scheme, which triggers the laser diode using a transistor–transistor logic signal issued from a delay generator to achieve laser exposure synchronized with data collection at the detector (Fig. 1). In this setup, bR crystals were exposed for 5 ms with laser light at the x-ray interaction region, while the photoreaction was tracked by collecting 5-ms-long timepoints over 40 consecutive frames. This data collection sequence was continuously repeated while new crystals were resupplied by the high-viscosity injector (11). After about 5 hours of data collection, each of the timepoints contained ~13,000 indexable diffraction patterns with the resolution extending to 2.3 Å (table S1).

Analysis of the data provides a “continuous” molecular movie of structural changes over 200 ms in time during which bR molecules complete their photocycle within the crystal. Within this time frame, we observe a long-term decay of difference map signal and complete restoration of the dark state within about 100 ms (movie S1). The rate of this long-term decay (fig. S4) and the evolution of structural states as determined by occupancy refinements (Fig. 1) occur within the temporal range established by time-resolved spectroscopy on bR crystals (9, 15, 16). Successive structural intermediates typically overlap, which presents a general problem for time-resolved experiments. Following the rise and decay of intermediates in small steps, as demonstrated here, allows one to choose time delays at which occurring intermediates can best be separated. These results demonstrate that time-resolved serial crystallography is not limited to XFEL sources but can also be done efficiently at synchrotrons. Results are of a quality comparable to that of our previous time-resolved studies at XFELs on the same crystals (fig. S5), even though the resolution was limited by the available photon flux. However, the resolution is compensated by high levels of activation from the millisecond-long light exposure, more accurate data from the use of a photon counting detector, and a more stable monochromatic beam, as previously observed during de novo phasing (13). Owing to the slow extrusion rate, less than 20 μl of prepared microcrystals were necessary to obtain these data, instead of milliliters needed for the XFEL experiments. The reduced sample consumption substantially lowers the entry barrier for time-resolved crystallographic studies. Although the simple TR-SMX setup implemented at the Swiss Light Source delivers excellent results, there is great potential to improve the method further. Pink beams (17) and diffraction-limited sources (18) optimize beam characteristics for faster data acquisition. The Hadamard transform method (19) and continued software development (20) are means to improve data analysis. Together with stronger nanosecond lasers, as well as new fast and accurate detectors (21), these improvements will allow the time resolution to be pushed from milliseconds into the lower microsecond regime to cover the majority of biologically relevant protein dynamics.

Completing the bR pumping cycle

The molecular movies of structural changes in bR are an impressive example of how XFEL sources can be used to understand biology (7). Our TR-SMX measurements extend the dynamic view of bR into the second half of the photocycle (Fig. 2), where the proton substrate is reloaded.

Fig. 2 Proton pumping in bacteriorhodopsin proceeds through three principal steps.

(A) The proton pumping cycle in bR with spectroscopic intermediates (I, J, K, L, M, N, O) and the relevant time domains for the individual steps. (B) Overview of the three principal steps in the proton pumping cycle of bR. (C) At 10 to 15 ms, we observe the opening of the hydrophobic barrier between the primary proton donor Asp96 to the SB [dark state in purple, open state (10 to 15 ms) model in red]. The changes are accompanied by the ordering of three water molecules [positive Fo(10 to 15 ms) − Fo(dark) density in blue at 3σ]. (D) The 0- to 5-ms timepoint shows the rearrangements necessary for proton release, in agreement with previous XFEL studies (8, 9) [dark state in purple, closed state (0 to 5 ms) model in blue]. Arrows indicate the direction of proton transfer reactions.

Refining the closed-state structure obtained by steady-state SMX against extrapolated data from the 0- to 5-ms timepoint (table S2), we obtained a good fit to M-state structures obtained by cryotrapping (22) or TR-SFX (8, 9) [root mean square deviation (RMSD) of 0.46 Å for the cryo-trapped structure 1IW9, 0.40 Å for the 1.7-ms structure 5B6Z, and 0.27 Å for the 8.3-ms structure 6G7L]. Difference maps obtained from 10 to 15 ms onward, however, showed clear additional features at the cytoplasmic side of the protein (Fig. 1). When refining the open-state structure obtained by steady-state SMX against extrapolated 10- to 15-ms data, we obtained a better fit in the dark-state structures of the constitutively open Asp96→Gly/Phe171→Cys/Phe219→Leu triple mutant [RMSD of 0.59 Å for the x-ray structure 4FPD (23) and 0.85 Å for the electron diffraction structure 1FBK (24)], which are considered analogous to the wild-type N-state obtained by photoactivation (fig. S6). The structural rearrangements in the open state (10 to 15 ms) with respect to the dark and closed states (0 to 5 ms) occur in the cytoplasmic ends of helices E, F, and G. The largest change in helix F corresponds to a movement of 9 Å measured at the Cα position of Glu166 (Fig. 1 and movie S2), which resembles the movement of helix 6 upon activation of structurally related G protein–coupled receptors (fig. S7). The magnitude of the motion is consistent with transient movements of helices F and G in the millisecond range observed by time-resolved spin labeling studies in purple membranes (25), suggesting that bR can complete the motion inside the crystal lattice (fig. S8). The final O-state does not accumulate well in bR crystals (16), possibly because crystal contacts, which change upon formation of the open state (fig. S8), affect the kinetics of the N-state to O-state transition inside the crystal. However, the O-state is in a fast equilibrium with the N-state (26), suggesting that no large structural rearrangements occur in the transition. The small changes upon relaxation of retinal from the cis-isomer back to the trans-isomer are spread out in time, with associated difference map features disappearing around the 35- to 40-ms timepoint (movie S1). Our TR-SMX experiment thus completes the crystallographic view of the bR photocycle, by revealing the structural rearrangements over the timeframe in which the M-state to N-state transition occurs inside crystals until the dark state is recovered.

Structural changes in helices E and F involve large side-chain motions. The isomerizing retinal pushes against Trp182 in helix F starting at earlier timepoints (9). This motion continues in the transition from the closed state to the open state, driving the displacement of helix F and allowing the aromatic ring of Phe156 to move 9 Å and replace the ring of Phe171. This alters the hydrophobic interaction network holding helix F in its original position. Moving helix F occurs concurrently with the disordering of the C terminus starting from Arg227 at the end of helix G (Fig. 3). By contrast, the structural changes in helix C are dominated by smaller side-chain rearrangements. The largest conformational change in helix C is a rotamer change of Leu93 matching a displacement of the nearby Phe219 side chain in helix G (movie S2). These changes create space for a chain of water molecules (Wat404, Wat453, Wat454) observed as strong positive difference map peaks [Fo(10 to 15 ms) − Fo(dark) above 3σ] connecting Asp96 on the cytoplasmic side with the retinal SB (Fig. 2). Additional experimental evidence for such waters originates from a combination of Fourier transform infrared spectroscopy, site-directed mutagenesis, and molecular dynamic simulations (27, 28). Our experimental results are consistent with molecular dynamic simulations showing that three waters are sufficient to close the gap and enable the proton transfer step to the SB (28). The positions of the three water molecules differ from those of the four waters observed in a freeze-trapping structure of the Val49Ala mutant (29), and the state trapped by mutagenesis lacks the large conformational changes on the cytoplasmic side. The constitutively open triple mutant Asp96→Gly/Phe171→Cys/Phe219→Leu, by contrast, allowed trapping of N-like features without light activation (23). However, altering the side chains of these residues prevents formation of the water chain in the x-ray structure of the triple mutant (23) (fig. S5). Such controversies have long hindered the emergence of a consensus concerning large conformational changes in the late bR photocycle (10). The useful mutations by their very nature had to target functionally critical regions, which complicated interpretation. Our study resolves these discrepancies by showing the relevant arrangements in light-activated native bR measured at physiological temperatures and with real-time resolution.

Fig. 3 Opening of the cytoplasmic side facilitates reprotonation.

Rearrangements of helices E, F, and G between the closed state (0 to 5 ms) (A) and the open state (10 to 15 ms) (B) lead to disordering of the C terminus (green dotted spheres). Charged residues on the cytoplasmic surface are shown as sticks. The presumed proton position is indicated on either Asp96 or the SB, and the chain of ordered water molecules forming at 10 to 15 ms is shown in spheres.

Molecular mechanism of proton uptake

The structural transition fits well with the molecular mechanism of vectorial proton pumping (movie S2) (30). Mechanistically, bR can be divided into an extracellular half and a cytoplasmic half with retinal positioned at the center (Fig. 2). Light-induced isomerization of the retinal chromophore is the first principal step in the reaction and provides the energy for proton pumping. The energy is used in the second principal step to drive protein conformational changes on the extracellular side to transfer a proton from the retinal SB toward the extracellular release group via a hydrophilic interaction network of residues and water molecules. Transferring the proton to the release group removes the substrate from the center of the membrane.

In the third principal step, the protein has to adopt a different conformation to allow the SB to accept a proton from the intracellular side of the membrane. Preventing proton back leakage to the cytoplasmic side is critical for vectorial transport against a concentration gradient. In bR, this is achieved through a hydrophobic barrier between the primary acceptor of protons in the center (the SB) and the proton donor on the cytoplasmic side of the membrane (Asp96). This 12-Å barrier has to be opened at the right time during each cycle to allow the SB to accept a new proton and recharge the system. Such a reprotonation switch has been suspected to involve larger conformational changes of the protein chain (31) in the late photocycle (i.e., M to N transition). Electron crystallography on bR mutants suggested opening on the cytoplasmic side and formation of a water channel to the retinal (24, 32). The formation of a water chain, oriented by the protein environment, could in principal act as a Grotthuss proton wire (33), bridging the hydrophobic barrier to deliver a new proton to the SB. Yet, despite numerous molecular dynamics simulations and structural studies (10), it remained unclear whether proton transfer occurs via rapidly exchanging waters in a channel (23) or via distinct water molecules rearranged by the protein environment (28). In addition to differences in methodology, discrepancies originate from the use of different starting models (fig. S5). The transition from the closed to the open bR conformation dominating the 10- to 15-ms timepoint provides direct evidence for the transient formation of such a chain of water molecules (Fig. 3). The rotation of Leu93, together with Phe219, creates space for three water molecules that connect the SB with the intracellular proton donor Asp96, opening the hydrophobic barrier by altering the hydrophobic interaction network. Transferring the proton to the SB replenishes the substrate at the center of the membrane.

The structural rearrangements that we observe leave Asp96 covered by a single hydrophobic lid formed by Phe42, making it accessible for reprotonation. Furthermore, structural changes to a cluster of residues acting as a proton funnel at the cytoplasmic side of the protein aid reprotonation of Asp96 (Fig. 3) (34), which may transiently proceed via an intracellular uptake cluster (35).

Time-resolved serial crystallography has revealed the structural reorganizations during proton uptake and release in bR with astounding detail and over many orders of magnitude in time. The alternating rearrangements on the extracellular and intracellular sides during the pumping cycle of bR agree with an alternate access model and may provide a template for understanding the principal transport steps in other membrane transport proteins. The large-scale motion of helix F further resembles the outward movement of transmembrane helix 6 in G protein–coupled receptors. Bringing time-resolved serial crystallography developed at XFELs to the broader scientific community using synchrotrons should provide tools to answer many other pressing questions in structural biology.

Supplementary Materials

Materials and Methods

Supplementary Text

Figs. S1 to S8

Tables S1 and S2

References (3649)

Movies S1 and S2

References and Notes

Acknowledgments: We thank D.Oesterhelt and G.Schertler for providing purple membranes. Funding: For financial support, we acknowledge the Paul Scherrer Institute and the Swiss National Science Foundation for grants 31003A_141235, 31003A_159558 (to J.S.), and PZ00P3_174169 (to P.N.). This project received funding from the European Union’s Horizon 2020 research and innovation program under the Marie-Sklodowska-Curie grant agreement no. 701646. Author contributions: T.W., P.S., M.W., and J.S. conceived the research with suggestions on laser setups from D.J. and F.D. Protein and microcrystals were prepared by A.F. and P.N. Slow crystal extrusion for the use at synchrotrons was optimized by P.S. and D.J. Synchronized diode and detector triggering was implemented by E.P. The pump-probe setup was built by D.J. and F.D. with suggestions from M.W. Data collection was done by T.W., P.S., D.J., A.F., D.K., S.B., F.D., D.O., P.N., M.W., and J.S. Data processing and structural refinements were done by T.W. The manuscript was written by T.W. and J.S. with contributions from all authors. Competing interests: The authors declare no competing interests. Data and materials availability: Coordinates, light amplitudes, dark amplitudes, and extrapolated structure factors for light activated data have been deposited in the Protein Data Bank (PDB) database under accession codes 6RNJ (TR-SMX closed state), 6RPH (TR-SMX open state), 6RQO (steady-state SMX activated state). Coordinates and dark state structure factors (steady-state SMX dark) have been deposited in the PDB database under accession code 6RQP.
View Abstract

Stay Connected to Science

Navigate This Article