Cryo-EM structure of the photosynthetic RC-LH1-PufX supercomplex at 2.8-Å resolution

Atomic structure of the first RC-LH1-PufX supercomplex provides new insights into the assembly and function of photosynthetic systems.


INTRODUCTION
Photosynthesis, performed by plants, algae, and many bacteria, is one of the most important biological processes on Earth as it harnesses solar energy to provide energy, oxygen, and food for life (1,2). In purple bacteria, photosynthesis requires close connectivity between several membrane protein complexes: the peripheral lightharvesting complex 2 (LH2), the central LH1, the photochemical reaction center (RC), the proton translocating cytochrome (Cyt) bc 1 complex, and adenosine 5´-triphosphate (ATP) synthase (3,4). Typically, the RC and LH1 form a photosynthetic core supercomplex that is central to bacterial photosynthesis (5). The photoinduced charge separation results in the reduction of quinone to quinol in the RC, which is then translocated across LH1 and diffuses through the membrane to Cyt bc 1 , where a proton-motive force is generated to trigger ATP synthesis. Cyt bc 1 reoxidizes the quinol, and an electron is shuttled via the soluble cytochrome c 2 back to the RC.
In the past few years, several RC-LH1 complexes from different purple photosynthetic bacteria have been resolved, exhibiting the distinctive molecular architectures of the photosynthetic supercomplexes. The RC-LH1 complexes from Thermochromatium (Tch.) tepidum and Thiorhodovibrio (Trv.) strain 970 contain a closed LH1 ring of 16 -heterodimers enclosing the RC with an extra periplasmic tetrahaem cytochrome subunit (4Hcyt) (6)(7)(8). In the bacteriochlorophyll (BChl) b-producing bacterium Blastochloris (Blc.) viridis, the 4Hcyt-containing RC is encircled by an LH1 ring that consists of 16 -heterotrimers and one -heterodimer, and the missing -polypeptide creates a small gap in the LH1 ring (9,10). In Rhodopseudomonas (Rps.) palustris, the RC is encircled by an open LH1 ring consisting of 14 -heterodimers and protein W (11,12), whereas the LH1 ring of Roseiflexus (Rfl.) castenholzii is interrupted by the transmembrane helix of Cyt c and subunit X (13). While it has been proposed that quinone/quinol can diffuse across the LH1 ring through specific channels between LH1 heterodimers (6)(7)(8), the gap in the LH1 ring potentially offers unique routes to facilitate quinone/quinol exchange between the RC and Cyt bc 1 .
The RC-LH1 core complex of most Rhodobacter species contains an additional transmembrane polypeptide PufX of ~80 amino acids, which is involved in determining the quinone diffusion and dimerization of the RC-LH1 complex (14)(15)(16)(17). An exception to this arrangement is Rhodobacter (Rba.) veldkampii (18), in which the RC-LH1 core complexes form only monomers in the presence of PufX, as characterized by atomic force microscopy (AFM) (5) and low-resolution cryo-electron microscopy (cryo-EM) (19,20). However, because of the limited resolution, the molecular basis underlying the assembly and arrangement of proteins and cofactors within this specific RC-LH1-PufX supercomplex to ensure efficient electron transfer remains unclear. Here, we report a cryo-EM structure of this RC-LH1-PufX photosynthetic complex from Rba. veldkampii at 2.8-Å resolution. Analysis of the supercomplex structure reveals how PufX functions as a molecular cross brace to interact with the RC, LH1, and cofactors, and thereby plays roles in determining the overall architecture and quinone transport pathways of the RC-LH1-PufX complex.

Overall structure
The intact RC-LH1-PufX core complexes were isolated from photoheterotrophically grown Rba. veldkampii cells (21). The Rba. veldkampii RC-LH1-PufX complexes are present exclusively as monomers with a single conformation, as displayed by reference-free two-dimensional (2D) and subsequent 3D classifications ( fig. S1). From a total of 1,168,396 particles, 184,921 good particles were used for final 3D reconstruction. Cryo-EM single-particle analysis determined the structure of the RC-LH1-PufX core complex at 2.84-Å resolution ( fig.  S2). The cryo-EM map clearly defined most of the amino acid side chains (Fig. 1, A to C, and fig. S3) and enabled the generation and refinement of the atomic model of the RC-LH1-PufX complex ( Fig. 1, D to F, and table S1).
The overall structure of the Rba. veldkampii RC-LH1-PufX complex is composed of H, L, and M subunits for the RC,  15  15 subunits for LH1, the transmembrane peptide PufX, and 62 cofactors ( Fig. 1 and table S2). The total molecular mass of RC-LH1-PufX is ∼285 kDa. The height of the core complex from the C terminus of PufX on the periplasmic side to the bottom of the H subunit on the cytoplasmic side is 81.1 Å (Fig. 1D). The RC contains three protein subunits (H, L, and M), four BChls a, three bacteriopheophytins (BPhes), one carotenoid, one Fe 3+ ion, and six ubiquinone molecules tentatively assigned as ubiquinone-10 (UQ-10) due to its high content in previously studied Rhodobacter species (22). The RC is surrounded by the LH1 ring composed of 15 pairs of transmembrane helices of and -apoproteins (Fig. 1, E and F), in agreement with the structures determined by AFM and low-resolution cryo-EM (19,20). The overall LH1 ring forms a slightly elliptical architecture, with the lengths of 113.7 and 107.6 Å for the major and minor axes, respectively ( fig. S4).

Interactions between PufX and RC-LH1
A unique feature of the Rba. veldkampii RC-LH1-PufX complex is a marked gap (30 Å in distance) with the LH1 ring interrupted by PufX ( Fig. 2 and fig. S4). The PufX peptide is composed of a central bent transmembrane helix (from Thr 11 to Gln 49 ), a short N-terminal tail and a C-terminal loop. The transmembrane helix of PufX has a tilt angle of 43.0° to the membrane plane and an angle of 68.5° to the orientation of the 15th LH1 -heterodimer (Fig. 2, A and B), instead of being parallel to the LH1 -peptides (20). The superposition with the nuclear magnetic resonance structure of isolated PufX from Rba. sphaeroides [Protein Data Bank (PDB) ID: 2DW3] (23) revealed that the transmembrane helices of the two structures, which are conserved among the PufX-containing species, are well fitted, whereas the extended N-and C-terminal regions show flexible conformations ( fig. S5).
The N-terminal tail of PufX is exposed on the cytoplasmic surface of the photosynthetic membrane, close to the N terminus of LH1-1 (Fig. 2, C and D, box 3). The C-terminal domain of PufX is exposed on the periplasmic side of the RC-LH1-PufX complex L-Tyr 74 and forms a salt bridge with the side chain of L-Glu 83 . PufX Phe 81 is exposed to a hydrophobic patch formed by Pro 71 , and Tyr 74 of the L subunit. Other interactions in this interface are listed in table S3. In addition, the M subunit of the RC partially contributes to the PufX-RC interface as PufX Phe 65 hydrophobically interacts with M-Ala 306 and M-Leu 307 (Fig. 2E).
The PufX-binding sites in the L subunit are largely conserved among purple bacteria ( fig. S6), suggesting high compatibility of PufX binding to the RC. This is indicated by the superimposed structures of Rba. veldkampii PufX-RC-L and Rba. sphaeroides Cyt c 2 -RC (PDB ID: 1L9J) (24). Moreover, the PufX-RC association does not affect the binding of the water-soluble electron carrier Cyt c 2 to the RC, which occurs at the center of the periplasmic surface of the RC ( fig. S7). Therefore, the specific PufX-RC binding does not affect reduction of the oxidized special pair in the Rhodobacter genus of purple photosynthetic bacteria. However, the superpositions of the 4Hcyt subunit cause a steric hindrance to the binding of PufX C-terminal loop to the RC ( fig. S7), explaining why many purple bacteria (especially those containing PufX) do not have a 4Hcyt subunit and rely on Cyt c 2 or other electron carriers (i.e., Cyt c y ) for electron donation (10).
The specific orientation of the transmembrane helix of PufX is mediated by interactions with adjacent LH1 -heterodimers located next to the gap in the LH1 ring. The middle region of PufX forms close contacts with the periplasmic side of the LH1-15 apoprotein (Fig. 2D, box 2). An inter-subunit hydrogen bond between PufX and LH1-15 is formed between -Ser 37 and PufX-Phe 50 at the inner side of the LH1 ring ( Fig. 2F and table S4). No direct interactions are formed between PufX and the LH1-15 apoprotein. At the interface between PufX and LH1-1 on the other side of the gap, PufX-His 5 forms a salt bridge and hydrogen bond with -Asp 12 and -Arg 14 , respectively. The PufX-LH1-1 interface forms hydrogenbonding networks between Lys 4 , Tyr 6 , Thr 11 , and Lys 12 residues of PufX and Thr 9 , Gln 15 , Glu 18 , and Ser 21 of LH1-1, respectively ( Fig. 2G and table S4). The specific molecular architecture suggests the assembly pathway of the RC-LH1-PufX supercomplex. Presumably, the assembly of the photosynthetic core complex initiates with the association of PufX and the RC in the membrane, and then PufX serves as an anchoring site for LH1 /-subunits to trigger LH1 assembly and encirclement.
PufX in Rba. sphaeroides is crucial for mediating the dimerization of the RC-LH1 core complex (15,16). The cryo-EM structure of the Rba. veldkampii RC is similar to the crystal structure of the Rba. sphaeroides RC (PDB ID: 2J8C) ( fig. S8), except for their lipid compositions (25,26). The superimposed structures of two Rba. veldkampii RC-LH1-PufX monomers and a Rba. sphaeroides RC-LH1-PufX dimer (7.78 Å; PDB ID: 4V9G) (27) depicted that two PufX proteins might be located at the center of the RC-LH1 dimer ( fig. S8). This implies that the outward-facing N-terminal regions of two PufX peptides may form an interface at the cytoplasmic side, suggesting the importance of the PufX N terminus in the dimerization of Rba. sphaeroides RC-LH1-PufX complexes as previously assumed (28). A higher-resolution structure of the dimeric RC-LH1-PufX complex is required to provide mechanistic details.

RC-LH1 interactions
LH1 generates six binding interfaces with the RC at both the periplasmic and cytoplasmic sides (Fig. 3, A and B). At the periplasmic side, L-Trp 52 , L-Trp 60 , L-Ala 79 , M-Asp 80 , M-Ala 108 , M-Met 111 , and H-Phe 7 interact with the Ser 37 residues of neighboring LH1 -subunits (Fig. 3A, fig. S9, and table S6). At the cytoplasmic side, two arginine residues (Arg 14 and Arg 15 ) of the -subunit helices have a major role in interacting with the RC subunits (table S6). On the outer side of the LH1 ring, LH1-15 is the only -subunit that directly interacts with the RC via hydrogen bonds of Ser 7 and Phe 8 residues with His 201 and Lys 199 of the RC-H subunit, respectively (Fig. 3B, box 5, fig. S9, and table S6). Moreover, the integration of the molecular cross brace PufX leads to a strong association between the RC and LH1. All of these interactions provide the foundation for forming a robust RC-LH1-PufX supercomplex.

Intra-and inter-subunit interactions within LH1
The LH1 ring is formed by the transmembrane helices of the inner and outer -apoproteins, both with the N and C termini located at the cytoplasmic and periplasmic sides, respectively (Fig. 1D). Both N and C termini of the LH1 -apoprotein form additional helices; the cytoplasmic helix of the LH1 -apoprotein is perpendicular to the -apoprotein (Fig. 3, C and D). Each LH1 -heterodimer contains two B880 BChls a at the periplasmic side and one carotenoid molecule, assigned as spheroidene based on the spectral properties (absorption maxima at 446, 471, and 505 nm) (Fig. 3C, fig. S10, and table S2) (18,21,29). The cryo-EM structure does not identify the spheroidene molecule within the 15th LH1 -heterodimer that is next to the LH1 opening. The B880 pigments bound in the same and adjacent LH1 -heterodimers overlap with each other in parallel to form the circular array of pigments (Figs. 3D and 4A). All 15 -heterodimers can be superimposed over the transmembrane region with good agreement (fig. S11). However, the N-terminal domain of the -apoprotein shows relatively large deviations, especially those located next to the gap (LH1-1 and LH1-15), presumably due to the changes in the local membrane environment resulting from the specific interactions between PufX and LH1 subunits.
Within the individual LH1 -heterodimer, -Tyr 5 , -Trp 8 , and -Leu 9 form hydrogen bonds with -Asp 13 , -His 20 , -Phe 8 , and -Thr 9 ( Fig. 3C and table S7). The -subunit Lys 6 interacts with -Asp 13 via salt bridge at the cytoplasmic side. Hydrogen bonds were also found between -Gln 20 and -Tyr 23 in the transmembrane region and between -Gly 40 and -Arg 45 at the periplasmic side. In addition, -Met 34 and -Val 22 are in close proximity to the ends of the spheroidene molecule (Fig. 3C). Both -Trp 43 and -Trp 47 are hydrogen bonded to the keto-oxygens of BChl a, and the Mg atoms of BChls a are coordinated by their neighboring -His 32 and -His 38 residues. Collectively, these interactions ensure tight associations of the apoproteins and pigments within the LH1 -heterodimer assembly.
The cryo-EM structure also reveals intermolecular interactions between LH1 -heterodimers in the Rba. veldkampii RC-LH1-PufX complex. Extensive inter-subunit hydrogen bonds between LH1 -heterodimers are formed in the C-terminal regions at the periplasmic side ( Fig. 3D and table S7).  (n) -Leu 48 is hydrogen bonded to its neighboring  (n)+1 residues Lys 50 and Tyr 51 ; the  (n) -Arg 45 residue, which is highly conserved in purple bacteria, forms hydrogen bonds with  (n) + 1 -Arg 53 and  (n)+1 -Val 54 . At the cytoplasmic side,  (n)−1 -Ser 2 is hydrogen bonded to  (n) -Gln 25 , and  (n)−1 -Lys 3 forms a interaction with the neighboring  (n) -Glu 18 residue via salt bridge. In addition, interactions between adjacent -subunits are formed between  (n) -Asn 38 and  (n)+1 -Asp 45 at the periplasmic side. Hydrogen bonds are also formed between adjacent -subunits through the interaction between  (n) -Leu 6 and  (n)+1 -Gln 15 in the N-terminal region at the cytoplasmic side. As a consequence, the LH1 and -subunits form extensive intra-and inter-subunit interactions with their neighboring and -subunits (fig. S12). These interactions sufficiently stabilize the LH1 protein-protein association even without Ca 2+ that plays roles in the LH1 assembly of RC-LH1 complexes from Tch. tepidum (6,7) and Trv. strain 970 (8). This provides the structural basis for the formation of a rigid LH1 ring architecture, integration of cofactors within the LH1, and the association of LH1 and the RC.
Cofactors in the RC include four B800 BChls a, three BPhes, one 15-cis-carotenoid, one Fe 3+ ion, and six UQ-10 molecules (Fig. 4D). The BChls of the RC are aligned on the same level as the LH1 B880 ring in the transmembrane region and have relatively equal distance with the closest LH1 BChls (Fig. 4B and fig. S13), providing the foundation for efficient energy transfer from LH1 to the RC. Unlike the RCs in other purple bacteria that have two BPhes (6-9), the cryo-EM structure reveals that the Rba. veldkampii RC L and M subunits accommodate three BPhes, which was confirmed by pigment extraction and high-performance liquid chromatography (HPLC) quantification analysis using the Rhodospirillum (Rsp.) rubrum RC-LH1 complexes as a reference ( fig. S15; see Materials and Methods). The extra BPhe was assigned to a well-resolved planar density toward the outside of the M subunit with its center seemingly coordinated by a nearby peptide bond oxygen ( fig. S16).
The cryo-EM map also shows three potential densities of lipid molecules between the RC and LH1 subunits. The most abundant lipids in Rba. veldkampii membranes are phosphatidylethanolamines (PEs) and phosphatidylglycerols (PGs), whereas glycolipids, phosphatidylcholines (PCs), and diphosphatidylglycerols, including cardiolipins, are absent (25). Therefore, we tentatively assigned these lipid densities to PE with C18:0 side chains (Fig. 4, A and C) (32). These putative lipid molecules have similar locations as those observed in the Tch. tepidum RC-LH1 complex (7) and form close contacts with the neighboring residues in the RC and LH1 ( fig. S17 and table S6). Phe 56 of the RC-H subunit may form interactions with both LH1 (-Arg 15 ) and lipids.

Multiple pathways of quinone/quinol exchange
Six molecules of UQ-10 were identified in the density map (Figs. 4, C and D, and 5A). Two UQ-10 molecules function as the primary (Q A ) and secondary (Q B ) quinone acceptors, with similar organizations to those of Tch. tepidum ( fig. S18) (7). The head of Q A is hydrogen bonded to His 220 and Ala 261 residues of the M subunit, and the head of Q B is hydrogen bonded to the L-subunit residues His 191 and Val 225 . Three additional putative UQ-10 molecules (Q 3 , Q 4 , and Q 6 ) are located in the gap between LH1 and the RC and are mainly surrounded by nonpolar residues of the L and M subunits (Fig. 5B  and fig. S18). The head of Q 3 forms hydrogen bond with the L subunit Ser 179 . Q 3 and Q B have the same orientation and Q 3 is in close proximity to the isoprenoid tail of Q B (Figs. 4D and 5A and fig. S18), suggesting that Q 3 is in a position appropriate for the exchange of Q B after double reduction and protonation. A quinone molecule has also been identified at a similar position in the RC-LH1 complex of Tch. tepidum ( fig. S22) (7). Compared to a closed RC-LH1 monomer, an open RC-LH1 monomer exhibited increased diffusion rates of quinol and photosynthetic efficiency (33,34). The gap between the 1st and 15th LH1 -heterodimers in the LH1 ring from Rba. veldkampii created by PufX (30 Å in distance) is the largest among those reported for purple bacterial RC-LH1 complexes (Fig. 5B), including the small gaps in the LH1 rings of Rps. palustris interrupted by protein W (11), of Blc. viridis by the missing -subunit (9), and of Rfl. castenholzii by the transmembrane helices of Cyt c and subunit X (13). These gaps have relatively similar locations within the LH1 ring ( fig. S19 and movie S1), suggesting a specific channel for quinol/quinone exchange. The large space between PufX and the 15th LH1 -subunits close to the cytoplasmic surface represents a channel for quinol diffusion, facilitating the transfer of reduced ubiquinone from the RC into the quinone pool outside the LH1 complex (Fig. 5C, circle 1). The space between PufX and the first LH1 -subunits close to the periplasmic surface could be another route for quinone diffusion (Fig. 5C, circle  2, and fig. S20). Consistently, spectroscopic analysis has shown that PufX could facilitate the quinone-mediated redox interaction between the RC and Cyt bc 1 in Rba. veldkampii and Rba. sphaeroides (21).
Molecular dynamics (MD) simulations on a 1-s time scale reveal that Q 3 moved toward the Q B site and had a similar orientation as the Q B quinone. Meanwhile, the Q B quinone left the L subunit of the RC and moved toward the positions of Q 6 and Q 4 ( fig. S21 and movie S2). In addition, Q 4 displayed a remarkable movement in the lipid region between the RC and LH1 toward the PufX-formed large gap in the LH1 ring ( fig. S21 and movie S3). The results suggest a potential quinone diffusion route: A quinone molecule enters the RC-LH1 complex through the large gap formed by PufX and accesses the position of Q 3 ; the Q 3 quinone replaces Q B , and the reduced Q B moves to the Q 6 and Q 4 sites and subsequently diffuses out of the complex through the large opening in the LH1 ring (Fig. 5B).
Moreover, the isoprenoid tail of Q 5 was found to be inserted into a space between the first and second LH1 -heterodimers ( Fig. 5D  and fig. S20), representing a quinone in transit through a putative shuttling channel between the LH1 subunits. This channel is surrounded by hydrophobic residues of LH1 subunits and was also found in the closed LH1 rings of the RC-LH1 complexes from Tch. tepidum (6,7) and likely in the LH1 ring of Rps. palustris (12). Computational simulations on a modeled LH1 structure also suggested the possibility that ubiquinone can diffuse through the closed LH1 ring (35). In addition, the cryo-EM structure identified a conformational alternation at the phytol tail of the BChl close to Q 5 (inward facing) in contrast to the tails of other LH1 BChls (outward facing) ( Fig. 5E and fig. S22), presumably critical for generating a specific environment within the LH1 ring to facilitate the diffusion of Q 5 . Similar conformational variation of the phytol tails of BChls occurred also in the Tch. tepidum RC-LH1 structure (PDB ID: 5Y5S) (7), in which three BChls have inward-facing tails while those of the remaining LH1 BChls face outward ( fig. S22).
Overall, the cryo-EM structure of the Rba. veldkampii RC-LH1-PufX supercomplex reveals the unique organization of PufX and the LH1 ring architecture with a large opening. This provides the structural basis for forming a stable photosynthetic core complex and ensures efficient proton transfer and quinone exchange across the LH1 ring through multiple pathways (Fig. 5B) necessary for anoxygenic photosynthesis and photosynthetic competence in the changing environment. The new structural model also highlights the natural variations of the photosynthetic RC-LH1 architectures.

Protein purification
Wild-type Rba. veldkampii DSM-11550 [Deutsche Sammlung von Mikroorganismen und Zellkulturen (DSMZ), Germany] cells were grown under 1120 lumen light at 30°C for 7 to 14 days in flat glass bottles filled with anoxic sodium succinate 27 (N medium, DSMZ, Germany) medium to the top and tightly closed with magnetic stirring of 150 rpm (21). Cells were harvested by centrifugation at 5000g for 10 min. After washing twice with tris-HCl buffer (pH 8.0), cells were resuspended in the working buffer (20 mM HEPES-Na, pH 8.0) and disrupted by passage through a French press three times at 16,000 psi. Unbroken materials were removed by centrifugation at 20,000g for 30 min. Membranes were collected by centrifugation at 125,000g for 90 min and solubilized by 3% (w/v) n-dodecyl -d-maltoside (DDM) for 15 min in the dark at 4°C with gentle stirring. After the unsolubilized materials were removed by centrifugation at 21,000g for 30 min, the clarified supernatant was applied onto the 10 to 25% (w/v) continuous sucrose gradient made with the working buffer containing 0.01% (w/v) DDM. Gradients were centrifuged at 230,000g for 19 hours. The core complex, represented by the heaviest red/brown pigmented band, was collected and further purified by a Sephacryl S200 gel filtration column (Cytiva). The pooled fractions used for cryo-EM data collection show an absorbance ratio at 871/805 nm, which represents the absorbance ratio of RC-LH1/LH2 of 2.84.

Cryo-EM data collection
Three microliters of the purified RC-LH-PufX complex was applied to the glow-discharged holey copper grids (Quantifoil Cu R1.2/1.3, 300 mesh) with a thin carbon-supported film. The grid was plungefrozen in liquid ethane using a Vitrobot Mark IV (Thermo Fisher Scientific). Parameters for plunge-freezing were set as follows: blotting time, 3 s; waiting time, 30 s; blotting force, 0; humidity, 100%; and chamber temperature, 4°C. Data were collected at the University of Tokyo on a 300-kV Titan Krios electron microscope (Thermo Fisher Scientific) with a K3 direct electron detector (Gatan) in counting mode. A total of 5022 movies were recorded at nominal magnification of ×105,000 and a pixel size of 0.83 Å/pixel, with a defocus range between 0.8 and 1.8 m and a dose rate of 1.0 electrons per Å 2 per frame. A typical motion-corrected cryo-EM image is shown in fig. S1A.

Data processing
The movie stacks were motion corrected using MotionCor2 (36). Contrast transfer function (CTF) and defocus values were estimated by CTFFIND-4.1 (37). A total of 1,168,396 particles were automatically picked using crYOLO (38) with the input box size of 250 × 250 pixels. Particles were then extracted in a box size of 64 pixels in a 3.24-Å/pixel size using RELION 3.1 (39,40) to speed up initial dataset cleanup consisting of two subsequent reference-free 2D classifications followed by three rounds of 3D classifications, for which a 3D initial model was calculated in RELION 3.1. During each classification step, the particles that were categorized into poorly 9 of 11 defined classes were rejected. The remaining 644,294 particles were reextracted in a 0.83-Å/pixel size. After another initial 3D model calculation and the subsequent four rounds of 3D classification, the resulting 184,921 good particles were refined at 3.7-Å resolution on the basis of Fourier shell correction (FSC) 0.143 criterion in RELION 3.1. Per-particle CTF refinement and beam tilt estimation were performed. Bayesian polishing and an additional round of 3D autorefinement generated the final maps at 2.84-Å resolution on the basis of the FSC 0.143 criterion. Local resolution was estimated in RELION 3.1.

Model building and refinement
For the RC, the crystal structure of Rba. sphaeroides RC (PDB ID: 1M3X) (41) was fitted to the cryo-EM map as a rigid body using "fit in map" in Chimera (42). For the LH1 ring, a single pair of LH1 from the crystal structure of Tch. tepidum RC-LH complex (7) was fitted to the cryo-EM map. The RC and LH1 structures were then manually adjusted using Coot (43). Then, 30 BChls (PDB ligand ID: BCL) of LH1, 15 spheroidene pigments (14 belonging to LH1 and 1 to RC, PDB ligand ID: SPO), and 4 UQ-10 molecules (PDB ligand ID: U10) outside of the RC were fitted to the cryo-EM map. PufX was built manually using Coot.
Rba. veldkampii and Rba. sphaeroides differ in their polar lipids. The most abundant lipids in Rba. veldkampii membranes were previously found to be PE, PG, and an unidentified lipid, whereas glycolipids, PCs, and diphosphatidylglycerols, including cardiolipins, were absent in Rba. veldkampii (25). Therefore, we assigned the potential lipid densities to the major and mostly fitted lipid PE in the Rba. veldkampii RC-LH1-PufX structural model.
The final model was refined by Phenix (44), and the stereochemistry was assessed by MolProbity (45). Statistics for cryo-EM data collection and model refinement are summarized in table S1. Amino acid sequences of protein polypeptides in the RC-LH1-PufX complex from Rba. veldkampii are shown in fig. S23. Images were generated and analyzed by UCSF Chimera. The mapping of electrostatic potential was achieved using PyMOL with the Adaptive Poisson-Boltzmann Solver (APBS) Electrostatics plugin (https://pymolwiki.org/index.php/ APBS_Electrostatics_Plugin). The potential tunnels for quinone diffusion were calculated by Hollow 1.3 (46) and the channel volumes were calculated by 3V server (http://3vee.molmovdb.org/volumeCalc.php).

Pigment extraction from RC-LH1 complexes
The RC-LH1 complexes from Rba. veldkampii and Rsp. rubrum were isolated from sucrose density gradients as described above. The samples were concentrated in an Amicon Ultra 15-ml centrifugal filter with a 100-kDa membrane at 3000g. Pigments were extracted from concentrated complexes by the addition of 5 volumes of 7:2 (v/v) acetone:methanol and incubated on ice in the dark for 10 min. Precipitated proteins were separated from the extracted pigments by centrifugation at 14,000g for 10 min, and the solvent was further clarified by passage through a 0.22-m hydrophobic polytetrafluoroethylene (PTFE) syringe filter. Pigments were then immediately analyzed by reversed-phase HPLC.

Analysis and quantification of pigments
BChls and BPhes were separated at 1 ml·min −1 at 40°C on a Supelco Discovery HS C18 (5-m particle size, 120-Å pore size, 250 × 4.6 mm) on an Agilent 1100 HPLC system using a program modified from a previously published method (47). Solvents A and B were 64:16:20 (v/v/v) methanol:acetone:H 2 O and 80:20 (v/v) methanol:acetone, respectively. Pigments were eluted at 50% solvent B held for 2 min, followed by a linear gradient to 100% solvent B over 10 min, and held at 100% solvent B for 25 min. Elution of BChl and BPhe was monitored by checking the absorbance at 747 nm. Peak areas for BChl and BPhe in each complex were calculated using the integration function in Agilent Chemstation software. BChl peak areas from each chromatogram were normalized, and relative BPhe peak areas were used to calculate the BPhe content of the Rba. veldkampii RC-LH1 complex based on the known ratio of BChl:BPhe in the RC-LH1 complexes of Rsp. rubrum (36:2) (48) as a reference. Three independent biological replicates were measured and analyzed.

MD simulation
The initial structure for MD simulation was prepared as follows: The structures of hydrocarbon chains of UQ-10 and BPhe (BPH), which are missing in the cryo-EM structure, were modeled manually. The topologies and the force field parameters of BPH, spheroidene (SPO), and UQ-10 were generated using the "Ligand Reader & Modeler" function (49) of the CHARMM-GUI server (50) and the CHARMM general force field (CGenFF) (51). The topology of BChl a (BCL) was generated by replacing the two protons of the bacteriochlorin moiety of BPH with a Mg 2+ ion. The charge of the Mg 2+ ion was the sum of the charges of the two protons. The rotational and translational position of the protein complex in a lipid bilayer was determined using the Positioning of Proteins in Membrane (PPM) server (52). The protein complex was embedded in a solvated lipid bilayer consisting of 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphoethanolamine (POPE) molecules using the "Membrane Builder" function (53) of the CHARMM-GUI server. The system was composed of 34 protein chains, 34 BCL, 15 SPO, 3 1,2-distearoyl-sn-glycero-3-phosphoethanolamine, 3 BPH, 6 U10, and 551 POPE; an Fe 3+ , 179 K + , and 183 Cl − ions; and 65,692 water molecules. The total number of the atoms was 312,747, and the size of the initial system was 16.0 nm × 16.0 nm × 12.9 nm. Transferable Intermolecular Potential 3-Point (TIP3P) model (54) was used for the water molecules. The CHARMM36m force field (55) was used for the protein chains, and the CHARMM36 force field (56,57) was used for the other molecules. Distance restraints with the force constant of 1.0 × 10 5 kJ•nm −2 were imposed between the metal ions Mg 2+ and Fe 3+ and their coordinating atoms. After the system was energy minimized and equilibrated, 1-s MD simulations were performed twice with different initial velocities. During the MD simulations, the temperature was kept at 303.15 K using the Nosé-Hoover method (58), and the pressure was kept at 1.0 × 10 5 Pa using the Parrinello-Rahman method (59,60). Bond lengths involving hydrogen atoms were constrained using the Linear Constraint Solver (LINCS) algorithm (61,62) to allow the use of a large time step (2 fs). Electrostatic interactions were calculated with the particle mesh Ewald method (63,64). All MD simulations were performed with GROMACS 2020 (65), with coordinates recorded every 10 ps.

SUPPLEMENTARY MATERIALS
Supplementary material for this article is available at http://advances.sciencemag.org/cgi/ content/full/7/25/eabf8864/DC1 View/request a protocol for this paper from Bio-protocol.