Two states of a light-sensitive membrane protein captured at room temperature using thin-film sample mounts

High-resolution X-ray diffraction data were collected at room temperature for light- and dark-adapted states of the archaerhodopsin-3 photoreceptor using transparent and opaque polymer-film sample mounts.


Introduction
Room-temperature diffraction methods are highly desirable for dynamic studies of biological macromolecules, since they allow high-resolution structural data to be collected as proteins undergo conformational changes (Shoemaker & Ando, 2018;Ren et al., 2020). Fixed-target approaches, in which crystals are loaded onto chips that can be scanned across the X-ray beam, have proven to be successful for soluble proteins (Martiel et al., 2019;Mehrabi et al., 2020). For membrane-protein crystals grown in lipidic cubic phase (LCP), an extruder is commonly used to pass a stream of microcrystals through the X-ray beam; however, this method requires a considerable amount of protein sample, the production of which may not be feasible for many membrane proteins (Weierstall et al., 2014;Ren et al., 2020). New approaches, including those using thin polymer films (Huang et al., 2015;Axford et al., 2016;Doak et al., 2018;Lieske et al., 2019), are therefore required to allow complete data sets to be obtained from small sample volumes, especially if the quantity of crystals is limited, without recourse to specialized sampledelivery mechanisms and data-collection environments.
The photoreceptor archaerhodopsin-3 (AR3) from the archaebacterium Halorubrum sodomense is a light-driven proton transporter. Like other microbial rhodopsins, it has seven transmembrane helices and a retinylidene chromophore created by the covalent conjugation of retinal to Lys226 via a Schiff base . Absorption of a photon of appropriate wavelength initiates a highly ordered sequence of conformational changes (the photocycle) which results in the movement of one H + ion from the cytoplasm to the extracellular medium (Chow et al., 2010;Saint Clair et al., 2012). Interest in the transporter has grown over the past decade, primarily because of its widespread applications in optogenetics (Guru et al., 2015;McIsaac et al., 2015); mutants of AR3 are now routinely used as silencers of mammalian neurons (El-Gaby et al., 2016;Rost et al., 2017) and as fluorescent sensors of transmembrane potential McIsaac et al., 2014).
AR3, along with many other microbial rhodopsins, has a desensitized, dark-adapted (DA) state which is formed from the resting, light-adapted (LA) state in the absence of light. The DA state is characterized by a thermal equilibrium between at least two conformations: one possessing all-trans retinal, in common with the LA state, and one with 13-cis 15-syn retinal (Saint Clair et al., 2012). Our understanding of the transition between the DA and LA states has recently been advanced with new atomic resolution cryo crystal structures .
In this study, microcrystals of AR3 were prepared for data collection in light and dark environments by sandwiching them between 13 mm thick sheets of polymer film with high X-ray transmittance (Axford et al., 2016). Transparent cyclic olefin polymer (COP) and opaque black Kapton provided light-permissive and light-tight containment, respectively (Supplementary Fig. S1), and enabled diffraction data to be collected at room temperature, significantly reducing the risk of crystal dehydration. The opaque sample mount enabled AR3 desensitization to the DA state, which was maintained and captured at 1.9 Å resolution in the diffraction data. This research not only demonstrates that AR3 can convert from LA to DA in a crystalline environment, but also that conversion to the DA state can take place at room temperature, with no apparent input of energy. Here, we show that the thin-film mounts are suitable for collecting high-resolution room-temperature diffraction data for membrane proteins, for which the supply of crystals is limited. This approach also provides a simple and accessible method for the X-ray scattering analysis of light-sensitive samples.

Protein expression and purification
H. sodomense (ATCC 33755) cells were purchased from LGC Standards Ltd (Teddington, UK) and were grown without any genetic modification. Initially, cells were grown in 4 Â 25 ml liquid culture medium at pH 7.4 with high salt concentration [125 g NaCl, 160 g MgCl 2 , 0.13 g CaCl 2 , 5 g K 2 SO 4 , 1 g bacteriological peptone (Oxoid, UK), 1 g yeast extract (Melford, UK) and 2 g soluble starch (Sigma) per litre] at 45 C with shaking at 170 rev min À1 for 5-7 days or until the OD 600 reached 1.2. Fresh culture medium (of identical composition) was inoculated from the 25 ml cultures and was incubated at 45 C with shaking at 170 rev min À1 for approximately three weeks until an OD 600 of 1.9 was reached.
Cells were harvested by centrifugation (8000g, 30 min, 4 C) and the pellets were resuspended in 4 M NaCl. DNAse I (Sigma, UK) was added and the solution was stirred for 2 h before being manually homogenized using a Potter-Elvehjem homogenizer. The preparation was dialyzed overnight in 0.1 M NaCl and centrifuged (70 000g, 50 min, 4 C). Sucrose-density gradient ultracentrifugation (110 000g for 15 h at 15 C) was used to isolate the AR3-rich membrane, using a step gradient consisting of layers of 4 ml sucrose solution at densities of 30, 40, 50 and 60%(w/v). The lower band with a pink/ purple colour was collected, and the sucrose remaining in the sample was removed through overnight dialysis against distilled water. The sample was then further centrifuged (70 000g, 50 min, 4 C) and the pellet was resuspended in distilled water to a final concentration of 20 mg ml À1 and stored at 4 C prior to crystallization. A yield of 7 mg protein was produced per litre of cell culture and, using SDS-polyacrylamide gel electrophoresis, the AR3  content was estimated to be 78 AE 2%(w/w) of the total protein.

Crystallization procedure
The crystallization process for AR3 has been reported previously . Briefly, LCP was prepared by mixing the non-delipidated protein sample with monoolein (Nu-Check Prep Inc., Elysian, Minnesota, USA) in a 2:3(v:v) ratio using two gas-tight Hamilton syringes connected by an SPT Labtech syringe coupler (Caffrey & Cherezov, 2009). All crystallization procedures were performed under dim light and the growing crystals were incubated at 20 C in the dark. Crystals grew from a precipitant solution consisting of 33%(v/v) polyethylene glycol 600 (Fluka Analytical), 100 mM MES buffer pH 5.5, 150 mM NaCl, 150 mM Ca 2+ to form rods with typical sizes of 15-20 mm in the shortest dimensions and 100-150 mm in length.

Data collection
For the LA structure, a small amount of crystal-containing lipidic sponge phase was sandwiched between two 13 mm thick transparent cyclic olefin polymer (COP) films (Zeon, Dusseldorf, Germany) separated by a 50 mm thick adhesive spacer containing a 5 mm diameter aperture (4titude, Wotton, UK; Fig. 1a; Supplementary Figs. S1 and S2). This 'sandwich' of crystals was cut to a $7 Â 7 mm square using a scalpel and was mounted by hand onto the beamline goniometer using a DiffraX pin (Molecular Dimensions, Newmarket, UK) as described previously (Axford et al., 2016). For the DA structure, the COP film was replaced by light-impermeable 12 mm black Kapton (Dupont; Fig. 1a), which has a maximum transmittance of <0.6% for all wavelengths in the range 200-640 nm and a maximum transmittance of <1.5% for all wavelengths in the range 640-840 nm (Dupont, personal communication). These samples were stored at room temperature in the dark for at least 30 min before data collection to allow the protein in the opaque Kapton film sandwich to change conformation to the DA state.
The final LA data set was formed by merging data from three crystals. The first was oriented parallel to the spindle axis and a helical collection of 50 rotation was performed while translating along the length of the crystal, with the X-ray beam focused to 8 Â 8 mm; the second and third crystals were orientated orthogonally to the spindle axis and a defocused beam, $40 Â 8 mm, was used to approximate the sample crosssection during rotation. The nature of the thin-film sandwich forces the rod-shaped crystals of AR3 to align parallel with the plane of the film. In order to collect complete data, the second and third crystals were rotated up to 60 and 50 away from face-on and with total angular ranges of 30 and 50 , respectively. The latter crystal received the largest X-ray dose, which was calculated to be 80 kGy using RADDOSE3D (Bury et al., 2018).
To collect the data for the DA structure, crystals were mounted between two layers of black Kapton. A raster scan was performed with an attenuated beam (one third of the per-image dose used for data collection) to locate the crystals beneath the opaque film (Fig. 1b). Small wedges of data were then collected from each crystal 'hit' and the starting angle was varied to maximize sampling of reciprocal space. 18 partial data sweeps (10-15 ) were collected in total using a defocused 30 Â 30 mm beam, with the maximum dose per wedge calculated to be 15 kGy.

Data processing and structure solution
All diffraction data were integrated with DIALS (Winter et al., 2018). For the three LA partial data sets AIMLESS was used to merge and scale the measurements . For the DA data, xia2.multiplex was used to merge and scale the 18 partial data sets together. A summary of the dataprocessing statistics is presented in Table 1  98.21 98.21 † See Section 3 for an explanation of the relatively low completeness. ‡ R p.i.m. is the multiplicity-weighted, precision-indicating merging R factor for comparing symmetryrelated reflections (Weiss & Hilgenfeld, 1997). § R meas is the redundancy-independent multiplicity-weighted R factor for comparing symmetry-related reflections (Diederichs & Karplus, 1997). (Liebschner et al., 2019) and structural models were built using Coot (Emsley et al., 2010).

Results and discussion
Two AR3 structures were solved, the first to 1.85 Å resolution (PDB entry 6s63) in the opaque-film sandwich and the second to 1.9 Å resolution (PDB entry 6guz) in the transparent-film sandwich. Data-processing statistics are reported in Table 1. Although the tolerance of the crystals to X-ray dose is lower at room temperature than under cryo conditions, the structures are sufficiently well resolved to allow all of the key structural elements of AR3 to be identified. Data completeness is somewhat compromised at around $90% due to the enforced systematic orientation of the rod-shaped crystals when squeezed between the two film sheets, creating what is effectively a two-dimensional sample mount (Supplementary Fig.  S2). Although it might be possible to limit the amount of flattening of the sample to maintain a greater range of crystal orientations, this would have led to a much longer beam path through the LCP, resulting in higher background scatter, thereby compromising the diffraction resolution.
All of the major structural features visible in the high-resolution cryo crystal structures were also resolved at room temperature, including the retinylidene chromophore, which is formed from the covalent conjugation of retinal to Lys226, and the conversion of the N-terminal Gln7 to a pyroglutamyl group . The most striking difference between the two room-temperature structures is the contrasting retinal conformation, which is resolved as the 13-cis isomer in PDB entry 6s63 but as the all-trans isomer in PDB entry 6guz, consistent with these structures representing the DA and LA states, respectively ( Fig. 2; Supplementary Fig. S3). Two different retinal conformers are modelled for each structure and the greatest apparent movement is around the -ionone ring. Although the roomtemperature structures are at lower resolution, the conformations of retinal observed here are broadly consistent with those in the previously reported cryo structures (Bada Juarez et al., 2021). The 1.3 Å resolution DA structure (PDB entry 6gux) has both 13-cis and all-trans retinal modelled with relative occupancies of 70:30, and the lower resolution of the corresponding room-temperature structure (PDB entry 6s63) is likely to explain why only the 13-cis isomer could be modelled. Two forms of the all-trans isomer were modelled into the 1.07 Å resolution LA cryo structure (PDB entry 6s6c), consistent with the room-temperature LA structure reported here (PDB entry 6guz; Bada Juarez et al., 2021).
Analysis of the protein structure surrounding the Schiff base identifies Retinal-binding pocket. Conformations of retinal (coloured in light and dark pink) in (a) the DA state (PDB entry 6s63) and (b) the LA state (PDB entry 6guz) of AR3. The blue mesh in (a) and (b) around retinal and Lys226 represents the 2F obs À F calc electron-density map contoured at 1.5. Omit maps are shown in Supplementary Fig. S3. (c) Overlay of composite omit maps (2F obs À F calc ) of the DA (PDB entry 6s63; grey) and LA (PDB entry 6guz; white) states for retinal and selected amino acids from the surrounding binding pocket. The electron-density maps (LA, blue mesh; DA, green mesh) for the two structures were created using Phenix and are contoured at 1.0. The differences in the distributions of electron density surrounding the -ionone ring and the C 13 C 14 bond are indicated by yellow arrows. Water molecules are represented by grey and white spheres. All images were created using PyMOL.
further differences between the two room-temperature structures and their cryo counterparts. All of the side chains in this region are resolved in only one conformation in the two AR3 structures reported here (Fig. 3); however, in the LA cryo structure (PDB entry 6s6c) the Asp95 side chain is resolved as two rotamers, consistent with partial breakage of the hydrogen bond to Thr99. Arg92, which is resolved as four rotamers in the two cryo structures, is only resolved in one position in the two room-temperature structures .
There are several bound water molecules within the half channels that link the Schiff base to the intracellular and extracellular sides of the membrane. In general, the positions of these molecules do not vary significantly between the DA and LA states reported here. One exception is Wat401, which is hydrogen-bonded to Asp95 and to Wat406 (Fig. 3)   Structures of the pentagonal hydrogen-bond networks in AR3 at room and cryogenic temperatures. The pentagonal hydrogen-bond networks in the DA (PDB entry 6s63) and LA (PDB entry 6guz) states of AR3 at room temperature are shown in (a) and (b), respectively, while in (c) and (d) the pentagonal hydrogen-bond networks in the DA (PDB entry 6gux) and LA (PDB entry 6s6c) states of AR3, respectively, at cryogenic temperatures are shown. Water molecules are shown as red spheres and predicted hydrogen bonds are shown as dashed yellow lines. The blue mesh represents the 2F obs À F calc electrondensity map contoured at 1.5. Omit maps are shown in Supplementary Fig. S4. which is modelled in two positions in the DA structure (PDB entry 6s63) but in only one position in the LA structure (PDB entry 6guz). The single Wat401 position in PDB entry 6guz is at the approximate midpoint between the two positions in PDB entry 6s63, and the relatively high B factor (23.64 Å 2 ) suggests that this molecule is highly disordered.
Supplementary Table S1 shows the B factors for Wat401 and the neighbouring atoms in the LA and DA structures. There is no significant difference in the apparent order of the adjacent Wat406; however, the B factor of the OD2 atom of Asp95, to which Wat401 is hydrogen bonded, is higher in the DA state than in the LA state. In contrast, there is little difference in the B factors of the Asp95 OD1 atom, which may interact directly with the Thr99 side chain and indirectly with the Schiff base N atom via Wat402. The disorder observed in Wat401 is consistent with previous Fourier-transform infrared spectroscopy experiments, which have suggested greater movement of Wat401 in AR3 than in bacteriorhodopsin (Saint Clair et al., 2012). Wat401 is implicated in determining the pK a of the equivalent Asp side chain in bacteriorhodopsin (Chang et al., 1988), and it is possible that a weaker hydrogen bond between Wat401 and Asp95 in the LA form reflects a higher pK a for the side chain than in the DA form (Bada Juarez et al., 2021).

Conclusion
Together these two structures show that high-resolution diffraction data may be obtained at room temperature using polymer films, and that the two different light-and darkadapted states may be differentiated, despite the increased kT relative to cryogenic conditions. This simple method provides an alternative to freeze-quenching approaches that rely on trapping the conformation of the protein by shifting it far from its working temperature. The use of a light-tight sample support meant that the beamline could be operated in a completely normal configuration, rather than under 'darkroom' conditions that would require the blocking of all ambient light. Furthermore, the practicality of using pin-based sample mounts allows the method to be used on regular goniometers, including those used on laboratory X-ray sources.