|Home | About | Journals | Submit | Contact Us | Français|
X-ray free electron laser (X-feL)-based serial femtosecond crystallography is an emerging method with potential to rapidly advance the challenging field of membrane protein structural biology. here we recorded interpretable diffraction data from micrometer-sized lipidic sponge phase crystals of the Blastochloris viridis photosynthetic reaction center delivered into an X-feL beam using a sponge phase micro-jet.
Studying membrane proteins remains a major challenge for structural biologists. These proteins contain hydrophobic and hydrophilic surfaces, are typically scarce and highly flexible, frequently become unstable when removed from their natural membrane environment and are usually difficult to grow into large, well-ordered crystals suitable for conventional crystallography. Serial femtosecond crystallography1, which allows X-ray diffraction data to be recorded from thousands of sub-micrometer- to micrometer-sized crystals, holds great promise for the structural analysis of membrane proteins. This emerging method has been first demonstrated1 using sub-micrometer crystals of cyanobacterial photosystem I (PSI) delivered to the X-ray beam using a thin liquid jet2. PSI, which is a large membrane protein complex, is exceptional in that it crystallizes overnight at low ionic strength3 (8 mM MgSO4). More representative membrane protein crystallization conditions involve higher concentrations of salt (typically 50-300 mM) and polyethylene glycol (PEG; typically 10-35%), which pose challenges for micro-jet injection owing to higher viscosity and the risk that salt crystals or aggregates may clog the micro jet nozzle.
Lipidic cubic phase (LCP) crystallization of membrane proteins4-6 is a generic crystallization method developed to mimic the natural lipid bilayer of membrane proteins and thereby enhance their stability during crystallization. In the original formulation4, solubilized membrane proteins were first mixed with the lipid monoolein in the ratio 60:40 to form the semisolid LCP, to which crystallization agents were then added. LCP crystallization rapidly led to X-ray structures of archaeal rhodopsins6 and more recently structures of G protein-coupled receptors7 (Supplementary Note and Supplementary Table 1).
Because of its semisolid nature, the LCP does not readily form a micrometer-sized jet required to deliver microcrystals in serial femtosecond crystallography1. In contrast, the closely related lipidic sponge phase (LSP), made by mixing monoolein and water with a third agent such as jeffamine or PEG that swells the cubic phase into a liquid phase6,8, can be adapted to serial femtosecond crystallography. LSP crystallization was originally developed from the observation that the LCP crystallization of the Rhodobacter sphaeroides reaction center proceeded via an LSP8. Four independent LSP membrane protein structures have been reported to date (Supplementary Table 1); it appears likely that crystals used for determining the structures of several G protein-coupled receptors and one bacterial oxidase grew via a cubic-to-sponge-phase transition (Supplementary Note), and an LSP crystallization screen gave leads from 8 of 11 membrane proteins tested9. In this work we adapted the LSP crystallization to yield showers of membrane protein microcrystals that we injected across an X-ray free electron beam using microjet technology2. From the diffraction data we recovered a new crystal form of the Bl. viridis photosynthetic reaction center (RCvir) solved using serial femtosecond crystallography.
For batch LSP crystallization experiments we modified earlier hanging-drop conditions10 (Online Methods): we dispensed 250 μl aliquots in septum-sealed glass vials (Fig. 1a). Because the sponge phase is less dense than water, it spontaneously separates8, with the LSP floating on the top (Fig. 1a). We used optical microscopy to identify crystallization conditions under which showers of microcrystals grew in the LSP over 2 weeks (Fig. 1b) and used cross-polarization analysis to confirm their ordered nature through birefringence (Supplementary Fig. 1). We shipped these sealed glass vials at room temperature to the Linac Coherent Light Source (LCLS) at SLAC National Laboratory, USA. We collected the colored upper phase (Fig. 1a) and removed large crystals that could potentially block the micro-jet by passing the LSP through a 10-μm-cutoff metal filter (Online Methods and Supplementary Fig. 2). We injected these suspensions as a rapidly flowing liquid micro-jet2 across a focused X-FEL beam11. The sponge phase (Fig. 1a) readily flowed at a rate similar to that of water (10 μl min−1) and produced a stable jet ~4 μm in diameter (Fig. 2a). Because the diameter of this jet was of the same order of magnitude as that of the RCvir microcrystals, the LSP provided a low-background environment for serial femtosecond crystallography.
We collected diffraction data at the Atomic Molecular and Optical (AMO) beamline12 of the LCLS on the Center for Free-Electron Laser Science Advanced Study Group multipurpose chamber (CAMP) pn-junction charge-coupled devices13 (pnCCDs) with two vertically offset panels that we read out at 60 Hz, which was the repetition rate of the FEL X-ray pulses. Each X-FEL pulse was nominally 70 fs in duration and contained up to 1013 2-keV X-rays (6.17 Å wavelength) focused into a 10 μm2 spot. The maximum resolution accessible with this energy and detector geometry was 7.4 Å, and reaction center microcrystals repeatedly diffracted to this resolution limit (Fig. 2b). Of the 365,035 recorded frames (~100 min of data collection), we observed 1,542 diffraction patterns of which we processed 265 in the space group P212121 (Fig. 2c) with unit cell axes a = 57.6 Å, b = 84.6 Å, c = 375.8 Å and unit cell angles α = β = γ = 90° (Supplementary Fig. 3), representing a new crystal form of RCvir.
As we processed 265 diffraction patterns from 365,035 images, the hit rate was below 0.1%, approximately an order of magnitude lower than that obtained for PSI1. A probable explanation for this difference is that the RCvir microcrystals tended to have a high mosaic spread (Supplementary Fig. 4), which, in combination with one cell axis being longer than 300 Å, made data-processing challenging. Because we grew PSI microcrystals overnight on-site at the LCLS1, it is reasonable to expect that the quality of RCvir microcrystals would improve if long-distance shipping at room temperature could be avoided. Further improvements include more robust diffraction-spot finding and indexing routines and changes in crystallization conditions to maximize the number of diffracting microcrystals.
We processed these diffraction data using Monte Carlo methods14,15 because each diffraction pattern recorded only partial reflections from randomly oriented microcrystals. We summarize the crystallographic data recovered from this analysis in Supplementary Table 2 and provide crystallographic statistics for each resolution shell in Supplementary Table 3. The best molecular replacement solution (Online Methods) showed the crystals to pack as stacked layers of 2D crystals typically found in lipidic phase crystallization6 (Supplementary Fig. 5). During structural refinement we cut the diffraction data to 8.2 Å resolution for which the multiplicity was greater than 4 and completeness was above 95% (Supplementary Table 3). Structural refinement yielded crystallographic R factor Rfactor and Rfree values of 35% and 38%, respectively (Supplementary Table 2; in Supplementary Table 4 we provide a breakdown by resolution shell) and both the 2Fobs - Fcalc (where Fobs and Fcalc are the observed and calculated structure factor amplitudes) electron density map (Fig. 3a) and composite omit map (Supplementary Fig. 6) clearly indicated transmembrane α helices. Moreover, when we removed all four heme groups of the RCvir cytochrome subunit from the structural model, we recovered positive Fobs - Fcalc electron density associated with each of these cofactors in the resulting omit map (Fig. 3b and Supplementary Fig. 7).
We also recovered well-performing micro-jets using LSPs derived from monoolein, water and PEG 400, PEG 1500 or PEG 4000 (Supplementary Fig. 8). As jeffamine and PEG conditions form the basis of a validated LSP crystallization screen9, and PEG 400 has been a crystallization agent in all recent LCP crystal structures of G-protein-coupled receptors (Supplementary Table 1), the adaption of LSP crystallization to serial femtosecond crystallography appears promising for solving membrane protein targets of unknown structure.
Shorter X-ray wavelength beamlines (λ of ~1.5 Å) and higher repetition rates (120 Hz) have recently become available at the LCLS. As such, high-resolution membrane protein crystal structures should soon be achievable using X-FEL radiation. We estimate that an order of 10,000 processed diffraction images will be needed for high-resolution electron density maps to be recovered using serial femtosecond crystallography, which would require three 12-h shifts at the hit rate reported here. Nevertheless, because we solved the structure of RCvir to 8.2 Å resolution by molecular replacement with only 265 processed images, this bodes well for future applications of lipidic phase serial femtosecond crystallography to membrane protein structural biology.
We cultivated the photosynthetic reaction center from Bl. viridis as described10. We modified the purification by using 250 ml POROS ‘50 micron’ HQ media (Applied Biosystems Europe BV) packed in an XK 50/20 column (GE Healthcare) and a HiPrep 26/60 Sephacryl S-300 column (GE Healthcare). This protocol yielded ~3 mg of pure RCvir per liter of cell culture.
We prepared LSPs as previously described10. We set up batch crystallizations in septum-sealed glass vials (Sigma-Aldrich) containing 100 μl protein (20-30 mg ml−1), 100 μl LSP (12% monoolein, 17.5% jeffamine M-600, 1.0 M Hepes (pH 8.0), 0.7 M (NH4)2SO4 and 2.5% 1,2,3-heptanetriol) and 50 μl of 1.0-1.2 M trisodium citrate. These setups were left to equilibrate for 2-4 weeks at 20 °C. We diluted crystals 4:1 in a solution containing 0.1 M Hepes (pH 8.0), 0.1% lauryldimethylamine-oxide (LDAO) before filtering in a 10-μm-cutoff titanium filter (VICI AG International). We developed the conditions for large-batch crystallization setups by initially screening for batch crystallization conditions using smaller batch crystallization volumes (60 μl to 100 μl setups).
We delivered samples to the injector nozzle via a sample loop, and we injected these into the X-FEL beam at a flow rate of 10 μl min−1. The liquid capillary of the nozzle had an inner diameter of 50 μm and the liquid was focused by coaxially flowing helium gas to a continuous jet-stream of ~4 μm diameter2. We aligned the X-ray beam to hit the liquid in the continuous jet region before the breakup into droplets occurred (Fig. 2a). For PEG-based LSPs, we required an inner diameter of 100 μm for the jet to flow (Supplementary Fig. 8). These LSPs consisted of 40% PEG 400, PEG 1500 or PEG 4000 mixed with 30% Monoolein and 30% buffer (0.1 M Hepes, pH 8.0, 0.1 M NaCl and 0.1 M MgCl2).
We collected diffraction data at the atomic molecular and optical (AMO) beamline at the Linac Coherent Light Source12 using the CAMP instrument13. We recorded diffraction data on two 76.8 mm by 38.4 mm pnCCDs located 64.7 mm and 67.7 mm from the sample position. The pnCCDs we offset asymmetrically such that X-ray diffraction was sampled from 3.5° to 49.0°, which maximized the sampled resolution (7.4 Å at the outer corners; Fig. 2b). The X-ray wavelength at AMO was 6.17 Å, and we focused the beam to a 10 μm2 spot.
We processed diffraction data using in-house code that called DirAx16 and MOSFLM17 for automated indexing1. We indexed data from 265 diffraction images in P212121 with unit cell axes a = 57.6 Å, b = 84.6 Å, c = 375.8 Å and α = β = γ = 90°. We did not observe subpopulations of other crystal forms (Supplementary Fig. 3). Small variations in the length of the c axis, however, arose perhaps owing to pulse-to-pulse fluctuations in the X-FEL wavelength. Because every observation is a partial reflection, we scaled and merged this integrated data using Monte Carlo methods14,15. We estimated I/σ values for each resolution bin (Supplementary Table 3) from reflections with both positive intensity and a multiplicity higher than 1. NZ-test and L-test plots showed the expected distributions (Supplementary Fig. 9).
We obtained phases by molecular replacement using Phaser18 2.3.0 with PDB entry 2WJN10 as the search model. The best solution had a translation function Z score (TFZ) of 8.5, the rotation function Z score (RFZ) of 5.8 and a log likelihood gain (LLG) of 81. This was well discriminated from second best solution with scores of TFZ, 4.6; RFZ, 5.8; and LLG, 39. We evaluated the crystal packing and confirmed that only the best molecular replacement solution was physically meaningful. We performed 20 cycles of rigid body and restrained refinement using REFMAC19, converging to Rfactor = 35% and Rfree = 38%, and an overall figure of merit of 0.62. During refinement we used simple Wilson scaling and a constant density (using default values) we assigned to the region of the unit cell not occupied by protein atoms. We calculated the solvent mask using default parameters: increase van der Waals radius of non-ion atoms by 1.2 Å, increase ionic radius of potential ions by 0.8 Å, shrink the area of the mask by 0.8 Å after calculation. Crystallographic data statistics are summarized in Supplementary Table 2, and resolution shell breakdowns of the crystallographic data and refinement statistics are given in Supplementary Tables 3 and 4, respectively.
We performed test calculations using two control datasets: one with the serial femtosecond crystallography data randomly shuffled and another where all observations were set equal. Molecular replacement failed with both control datasets. When we combined the phases generated using the molecular replacement solution recovered against the experimental data with these control datasets, the resulting electron density maps did not show α-helical structure (Supplementary Fig. 10).
Experiments were carried out at the LCLS, a national user facility operated by Stanford University on behalf of the US Department of Energy, Office of Basic Energy Sciences. We acknowledge financial support from the Swedish Research Council (Vetenskapsrådet), the Swedish Foundation for International Cooperation in Research and Higher Education, Stiftelsen Olle Engkvist Byggmästare, the Max Planck Society for funding the development and operation of the CAMP instrument, the US National Science Foundation grant MCB 0919195, the US Department of Energy Office of Basic Energy Sciences through the Photon Ultrafast Laser Science and Engineering Center Institute at the SLAC National Accelerator Laboratory and the Energy Frontier Research Center for Bio-Inspired Solar Fuel Production (award DE-SC0001016), the Hamburg Ministry of Science and Research and Joachim Herz Stiftung as part of the Hamburg Initiative for Excellence in Research and the Hamburg School for Structure and Dynamics in Infection, US National Science Foundation (awards 0417142 and MCB-1021557), US National Institutes of Health (awards 1R01GM095583-01 and 1U54GM094625-01), the Deutsche Forschungsgemeinschaft Cluster of Excellence at the Munich Center for Advanced Photonics, Center for Biophotonics Science and Technology at the University of California (cooperative agreement PHY 0120999).
Accession codes. Protein Data Bank: 4ac5.
Note: Supplementary information is available on the Nature Methods website.
L.C.J., R.N., H.N.C., J.C.H.S. and P.F. conceived the experiment, which was designed with A.B., R.A.K., J.C.H.S., D.P.D., U.W., R.B.D., M.J.B., D.S., I.S., S.M. and J.H. S.W.E., R.H., D.R., A.R., L.F., N.K., P.H., B.R., B.E., A.H., C.R., G.W., L.S., G.H., H.G., J.U., I.S., H.S., M.B., H.H., L.G. and H.G. designed and set up the CAMP instrument and/or developed and operated the pnCCD detectors. C.B. and J.D.B. set up and aligned the beamline; L.C.J. and D.A. grew sponge phase crystals; U.W., R.B.D., J.C.H.S., D.P.D., R.L.S. and L.L. designed and operated the injector; H.N.C., A.B., A.A., J.S., D.P.D., U.W., R.B.D., S.B., M.J.B., L.G., J.H., M.M.S., N.T., J.A., S.S. and J.C.H.S. developed diffraction instrumentation; L.C.J., D.A., T.A.W., D.P.D., U.W., R.B.D., R.L.S., L.L., E.M., J.D., K.N., M.L., A.A., M.B., A.B., M.J.B., C.B., J.D.B., C.C., R.C., N.C., T.E., H.F., P.F., C.Y.H., M.S.H., S.K., R.A.K., F.R.N.C.M., A.V.M., I.S., M.M.S., R.G.S., F.S., N.T., X.W., C.W., H.N.C., J.C.H.S. and R.N. collected data; L.C.J., D.A., T.A.W., G.K., W.Y.W. and K.N. analyzed the diffraction data using software developed by J.C.H.S., T.A.W., R.A.K., A.B. and H.N.C.; L.C.J. and R.N. wrote the paper with discussion and improvements from all authors.
COMPETING FINANCIAL INTERESTS
The authors declare no competing financial interests.