Structure of the far-red light utilizing photosystem I of Acaryochloris

Acaryochloris marina is a cyanobacterium that can, uniquely, use far-red light for oxygenic photosynthesis. Here, we report the structure of the photosystem I reaction center of A. marina determined by cryo-electron microscopy at 2.5 Å resolution. The structure reveals a unique arrangement of electron carriers and light harvesting pigments. The primary electron donor P740 is a dimer of chlorophyll d/d′ and the primary electron acceptor pheophytin a, a metal-less chlorin different from the chlorophyll a common to all other oxygenic type I reaction centers. The architecture of the 11 subunits and identity of key components help explain how the low energy yield from far-red light is eciently utilized for driving oxygenic photosynthesis.

The loop regions of PsaB (residues 290−320 and 465−510) were not identi ed due to disorder because of the absence of PsaX, suggesting that PsaX stabilizes these regions in other cyanobacteria. Most of subunits PsaE, PsaF, and PsaK were modeled using polyalanine because of poor regional map quality (Extended Data Figs. 12, 13, and Extended Data Table 1), possibly suggesting that some part of those subunits dissociated during sample preparation. The cofactors assigned (Extended Data Fig. 14) are 71 Chl d (Extended Data Table 6), one Chl d′ (an epimer of Chl d), 11 α-carotenes (α-Car) but no β-Car (Extended Data Table 7) 6,10 , two pheophytins (Pheo) (a derivative of Chl; however, unlike Chl, no Mg 2+ ion is coordinated by the tetrapyrrole ring), two phylloquinones (PhyQs), three iron-sulfur clusters, two phosphatidylglycerols, one monogalactosyl diacylglycerol, and 84 water molecules. The Chls and α-Car pigments are mainly involved in the electron transfer reactions and light-harvesting.

Electron transfer components
The important cofactors involved in electron transfer (Fig. 1B) are four Chls (three Chl d and one Chl d′), two Pheos, two PhyQs (A 1 ), and three iron-sulfur clusters (F X , F A , and F B ) (Fig. 3A). The Chls and PhyQs are arranged in two branches, the A-branch and the B-branch, which are related by a pseudo-C 2 axis as in other type I reaction centers, and are stabilized by amino acid residues of subunits PsaA and PsaB. The reaction center, P740, is composed of Chl d′ (P A ) and Chl d (P B ) (Fig. 4A), which are coordinated by residues His678/A and His657/B, respectively (capitalized letters following the slash (/) indicate the subunit names such as PsaA or PsaB) ( Fig. 3B and C). The distance between the ring planes (the π-π interaction distance) of Chl d′ and Chl d is 3.5 Å (Fig. 3D) (3.6 Å in T. elongatus). Two Tyr residues, Tyr601/A and Tyr733/A, are positioned within hydrogen bonding distance of P A . These con gurations are similar to those in the PSI from T. elongatus (Fig.1), although cofactor compositions are different.
In A. marina PSI, we uniquely identi ed three water molecules (W1−W3) around P A , forming hydrogen bonds with surrounding amino acid residues (Tyr601/A, Ser605/A, Asn608/A, Ser741/A, and Try745/A) and one Chl d (Chl d 32) ( Fig. 3B and C). For the T. elongatus P A , only one water molecule forms hydrogen bonds with surrounding amino acid residues (Tyr603/A, Ser607/A, Ile610/A, Thr743/A, and Phe747/A) and Chl a′ (P A ). There are no water molecules around P B in A. marina, but some water molecules surround P B in T. elongatus without forming hydrogen bonds. This difference around P B may come from the difference in two amino acid residues: Val594/B in A. marina vs. Thr597/B in T. elongatus, and Asn598/B in A. marina vs. His601/B in T. elongatus. The hydrogen bonding pattern around P A probably contributes to the charge distribution ratio (P A˙+ /P B˙+ ) 1,25 , and therefore, is likely different in the two organisms. The midpoint potential values, E m (P A ) and E m (P B ), are in uenced by the protein environment, in particular, by the presence of charged residues, and the E m difference, E m (P A ) − E m (P B ) (= ΔE m ), is an important factor in determining the P A˙+ /P B˙+ ratio 26 .
Previous studies have suggested that the primary electron acceptor A 0 in A. marina PSI (Fig. 1A) is Chl a 9,27 , in line with other species 1,28 . However, we found that there was no mass corresponding to Mg 2+ at the center of the tetrapyrrole rings of A 0A and A 0B , but rather a hole in the cryo-EM density map (Fig. 4A).
This indicates that A 0 is actually Pheo, a derivative of Chl, which is consistent with pigment analysis identifying the presence of Pheo a (Extended Data Table 7 and Extended Data Fig. 15). Furthermore, the position of Leu665/B (Extended Data Fig. 16B) 29 , pointing to A 0B and in the same place as conserved Met688/B that ligates A 0B in T. elongatus PSI 1,29 , supports Pheo a as A 0B in A. marina, because Leu cannot serve as a ligand to Mg 2+ in Chl. Thus, surprisingly, A 0A and A 0B are assigned as Pheo a in A.
The amino acid residue nearest to A 0A in A. marina is Met (Met686/A) (Extended Data Fig. 16A) 29 and the identity of close amino acid residues may modify the reduction potentials of the two A 0 and their absorption peak wavelength(s). Two mechanisms have been proposed for electron transfer in PSI reaction centers, one using both A-and B-branches 30,31, and the other using the A-branch preferentially 32,33 . Only the A-branch may be active in A. marina PSI 29 . Details of the electron transfer mechanism in relation to the P A˙+ /P B˙+ ratio can now be lled in using the precise structure of A. marina PSI.
Why is Pheo a the A 0 in A. marina PSI? Pheo a has not been identi ed in any PSI before, although it is the primary electron acceptor in all type II reaction centers 3 . While the reaction center P740, which is composed of Chl d/d', uses low-energy far-red light of 740 nm (1.68 eV), it generates reducing power almost equivalent to that of the P700 (Chl a/a', 1.77 eV) in plants and most cyanobacteria 8 . Then, the excited state of P740 (P740*) produces 0.09 eV-weaker reducing power than P700* (i.e., 77 -1.68 eV).
This could result in a slower electron transfer rate to A 0 and an increase in reverse reaction, due to the smaller driving force. However, the rates of electron transfer from P740* to A 0 , and to PhyQ are actually comparable to those from P700* to A 0 and to PhyQ 27 , therefore, Pheo a as A 0 must achieve the same electron transfer e ciency as the Chl a-type PSI.
The midpoint potential value (E m ) vs. the normal hydrogen electrode (NHE) of P740 is 439 mV 8,9,14 , whereas that of P700 is 470 mV 34 . However, most of the amino acid residues around A 0 in A. marina PSI are similar to those in T. elongatus, so it is unlikely that the protein structure around A 0 in uences the E m value. We obtained E m (vs. NHE) values of puri ed Chl a, Chl d, and Pheo a in acetonitrile of −1100, −910, and −750 mV, respectively (Extended Data Fig. 17) 34,35 . Therefore, according to the E m value, it seems reasonable for the Chl d-driven PSI to use Pheo a as A 0 , as the energy gap is su cient for the primary electron transfer step. However, a possible effect on the rate of the following A 0 to PhyQ step should be investigated as the rate is also known to be comparable to that in Chl a-type PSI.
A. marina contains a limited but distinct amount of Chl a 8,10,29 , and we found a small amount in the PSI (1-2 Chl a per PSI monomer) (Extended Data Table 7). One possible locality is as the accessory Chls, Acc A and Acc B (Figs. 1 and 4). However, unfortunately, Acc could not be conclusively identi ed as Chl d or Chl a, because the functional group containing C3 1 of Acc Chl could not be precisely de ned from the cryo-EM density map. While quantum mechanical/molecular mechanical calculations show that the formyl group of Chl d adopts two orientations (Extended Data Fig. 1C, D) 36 , the oxygen atom of the formyl group on the Chl d molecule in a vacuum is more stable when oriented towards the C5 H atom, suggesting the conformer in Extended Data Fig. 1C. In contrast, the vinyl group of Chl a can adopt either orientation. The assignment of Chl d to the Accs in this study is actually due to the difference spectrum of P740, which suggests that the large electrochromic shift of Acc at 710 nm arises from Chl d 7 .
The Mg 2+ in Acc A and Acc B are coordinated by water molecules forming hydrogen bonds with Asn588/B and Asn602/A, respectively ( Fig. 4B and C). In T. elongatus PSI, the methyl ester groups of the Chl a in Acc A and Acc B affect the charge and spin distributions on P700 26,[37][38][39] . The distance between the carbonyl O atom in Acc A and the Mg 2+ in P A , an epimer of Chl a, is 5.4 Å, whereas in Acc B it is 7.1 Å.
In contrast, in A. marina, these distances were 6.2 Å (Acc A −P A ) and 6.6 Å (Acc B −P B ), respectively (

Pigment environments: antennas and red chlorophylls
In the PSI, a large number of Chl d and α-Car harvest low energy far-red light and nally transfer the energy to P740. The arrangements of Chl d and α-Car in A. marina are similar to those of Chl a and β-Car, respectively, in T. elongatus, as shown in Extended Data Fig. 19, in which, the Chl ds are numbered according to the nomenclature of Chl a in T. elongatus 40 . However, there are differences in some amino acid residues surrounding the cofactors, slight gaps in the arrangement of the cofactors, and lack of some Chls, which, we assume, re ects optimization for e cient use of far-red light.
According to the theoretical study of Adolphs et al. 40 , the changes in site-energy of the three Chl a molecules whose band positions are most signi cantly shifted in T. elongatus PSI are produced by the following factors: (1) interaction with the backbone of α-helices, in particular with that of residues Gly316/B−Met320/B, causing Chl a 65 to be the third-lowest site-energy site of all the pigments; (2) the amide dipole in the side-chain of Gln115/A causing a decrease of the site-energy of Chl a 12; and (3) the positive charge of Arg24/M that induces a signi cant increase in the site-energy of Chl a 94.
The rst factor is related to what are known as "red Chls" 1,41 . Red Chls are a dimer or trimer of Chls, whereby the peak wavelengths of their Qy bands are shifted to wavelengths longer than the 700-nm peak of P700. In T. elongatus PSI, 9−11 Chl a per PSI monomer are considered to be red Chls. In particular, the trimer of Chl a 77-79 positioned in the region of Gly316/B−Met320/B is proposed to be a red Chl Regarding the third factor above, Arg24/M in T. elongatus PSI corresponds to Asn (Asn24/M) in A. marina, which has a polar uncharged side-chain and no Chl d that corresponds to Chl a 94 of T. elongatus PSI could be identi ed in the vicinity of Asn24/M (Extended Data Fig. 19C). The A. marina PSI appropriately lacks the Chl with most increased site-energy that captures high-energy, short-wavelength light in T. elongatus.
Differences in some of the amino acid residues in the A. marina PSI compared with those of T. elongatus (Extended Data Table 5), result in different arrangements of some cofactors. For example, Phe49/J in A. marina, the counterpart of His39/J that is an axial ligand of Chl a 88 in T. elongatus PSI, explains the absence of the corresponding Chl d (Extended Data Fig. 19D). Owing to the absence of PsaX in A. marina, the Chl d corresponding to Chl a 95 in T. elongatus whose axial ligand is Asn23/X is also missing. These differences help explain the unique features of light harvesting in A. marina.
The arrangements of pigments adjacent to Psa27 (PsaI) and PsaL, whose sequence identity with other cyanobacterial PsaLs is low, are speci c to A. marina (Extended Data Fig. 19A, black dashed line region, and Extended Data Fig. 20, and 21). Chl d 38, 52, 53, and the ring (α or ε) of Car4007 near Chl d 53 are set within the surrounding structure differently from those in T. elongatus (Extended Data Fig. 20A). This impacts on the distances between Car4010 and Chl 92 (8.1 Å in A. marina vs. 4.9 Å in T. elongatus), and Car4010 and Chl 47 (7.9 Å in A. marina vs. 6.9 Å in T. elongatus) (Extended Data Fig 20B and C). In the A. marina PSI, Chls 37 and 91−93 are in close proximity to PsaL, and Gln92/L (Ala94/L of T. elongatus) and Trp27/L (Phe30/L of T. elongatus) are positioned within hydrogen-bonding distance of Chl d 91 and 92, respectively (Extended Data Fig. 20D and E). Moreover, the formyl group of Chl d 37 and 93 and Ser62/L may form hydrogen bonds via one water molecule (Extended Data Fig. 21). Again, these divergences must affect energy transfer e ciencies in the two organisms.
A. marina PSI is a trimer as is the case for other cyanobacterial PSIs, but the amino acid sequence identity of each protein subunit is low, and, not unexpectedly, the divergence is most pronounced around the unique pigments, for example Chl d, α-Car, and Pheo a-A 0 (Extended Data Table 5). The structure exhibits unexpectedly high exibility of photosynthetic machinery. Again, the alterations must re ect optimization to drive e cient photochemistry. The structure would also bridge the evolutionary gap between the Chl a-type oxygenic reaction centers (adapted to visible light) and the BChl-type reaction centers (adapted to 800-900-nm light).
In conclusion, the structure and identi cation of electron carriers and key light-harvesting pigments provide the architecture for understanding how low energy far-red light is utilized by A. marina PSI. However, the full picture must wait for the structure of PSII, which is responsible for oxygen evolution through water-splitting, to be elucidated 3 . 1.0 M betaine, and 5% (w/v) glycerol, and stored frozen until use. The cells were broken at 1.5 mg Chl d/mL with the same volume of glass beads (0.2 mm diameter) at 0°C by 24 cycles of 10-s breaking, and 2-min cooling in the presence of DNase I (0.5 μg/mL; Sigma, St. Louis, MO) and a protease inhibitor mixture (280 μL/100 mL cell suspension; Sigma). Thylakoid membranes were collected by centrifuging the broken cells at 31,000 × g for 10 min. The resulting pellets of thylakoid membrane were resuspended in buffer A.
Thylakoid membranes were solubilized with 1.0% (w/v) UDM at 1 mg Chl d/mL on ice for 10 min. The extracts were separated from insoluble membranes by centrifugation at 75,600 × g for 10 min, and layered onto a stepwise sucrose gradient [1.4, 1.2, 1.1, 1.05, 1.0, 0.9, 0.8, and 0.7 M sucrose in buffer A supplemented with 0.2% (w/v) DM] for ultracentrifugation at 100,000 × g for 20 h. After centrifugation, fractionation was performed from bottom to top, and the protein content of each fraction was investigated by blue native-polyacrylamide gel electrophoresis (BN-PAGE) using linear gradient gels of 3%-12% polyacrylamide (Extended Data Fig. 2). Fractions exhibiting a band of PSI trimer of ~720 kDa, corresponding to sucrose concentrations of 1.05-1.0 M, were collected.
For electron microscopic analysis, the PSI trimer in the above fractions was washed using Amicon centrifugal lters (pore-size 100,000 Da) with buffer A supplemented with 0.002% LMNG, and applied to a second stepwise sucrose gradient ultracentrifugation consisting of 1.4, 1.2, 1.1, 1.05, 1.0, 0.9, 0.8, and 0.7 M sucrose in buffer A supplemented with 0.002% LMNG. LMNG was used for electron microscopic analysis because of its low critical micellar concentration.
For biochemical analyses, the fractions obtained after the rst sucrose density gradient ultracentrifugation were applied to an anion exchange column (HiLoad 16/10 Q Sepharose HP) and eluted with a linear gradient of 0-600 mM KCl.

Polypeptide analysis
Denatured polypeptides were separated on a gel containing sodium dodecyl sulfate (SDS), 6 M urea, and 18% or 16%-22% (w/v) acrylamide as described previously 43 with or without dithiothreitol, and then stained with Coomassie Blue (Extended Data Fig. 5). Separated polypeptides were identi ed by mass spectrometry (MS) analysis after in-gel digestion by trypsin (Extended Data Table 2).

Spectroscopic analysis
Chl a and Chl d concentrations were determined as described by Porra et al. 44 and Li et al. 45, respectively, using methanol as the solvent and a UV-2700 spectrophotometer (Shimadzu; Kyoto, Japan) with a slit-width of 1 nm at 25°C. Absorption spectra of the PSI trimers were measured at 25°C using an UV-2700 spectrophotometer (Extended Data Fig. 6). For cell suspensions, an integrating sphere, model ISR-2600, was used with the UV-2700 spectrophotometer (Extended Data Figs. 2 and 6). Absorption spectra at 77 K were measured using an MPS 2000 spectrophotometer equipped with a low-temperature measurement unit, LTS-2000 (Shimadzu) (Extended Data Fig. 7).

Pigment analysis
Photosynthetic pigments and quinones were extracted in methanol and quanti ed using reversed-phase high-performance liquid chromatography (HPLC) carried out using an LC-20AT with a CBM-20A system controller (Shimadzu), equipped with a Kinetex C18 column (5 µm, 250 × 4.60 mm, 100 Å) (Phenomenex, Torrance, CA), as described previously 46 . The elution pattern of the pigments was detected using a Shimadzu photodiode array detector, SPD-M20A, with Shimadzu LabSolutions (ver. 5.82) analysis software at 430 nm for photosynthetic pigments (Extended Data Fig. 15) and 270 nm for PhyQ. Standard curves for Chl a and Chl d were obtained after quanti cation of Chl a from a cyanobacterium, Synechocystis sp. PCC 6803, and Chl d from A. marina as described above. The standard curve for Chl d was also used for Chl d'. A Standard curve for Pheo a was made using Pheo a obtained by acidi cation of quanti ed Chl a. PhyQ (Vitamin K 1 ) was obtained from Wako Pure Chemicals (Osaka, Japan).
Zeaxanthin and α-and β-carotene were purchased from the VKI Water Quality Institute (Hørsholm, Denmark). PhyQ bound to PSI trimer was quanti ed after the removal of Chl a contribution at almost the same retention time by acidi cation 47 .

DNA sequencing
Genes psaA and psaB were partially sequenced as follows (Extended Data Fig. 16). Pairs of primers for psaA (GTACAACTGCATCTCAATTG and CTATCCTAATGCGAGAATTC) and psaB (CCTTGCCTTCTTCTGGATGC and TTAGCCGAGAGGAGCTGTTG) were used to amplify part of the genes by PCR using genomic DNA as the template. Then, the DNA sequence of the ampli ed, puri ed DNA fragment was analyzed using BigDye Terminator ver. 3.1 (Applied Biosystems; Foster City, CA) using the same primers.

Cryo-EM sample preparation and data collection
Three microliters of puri ed PSI [0.4 mgChl d/mL in a buffer containing 50 mM MES-NaOH (pH 6.5), 5 mM CaCl 2 , 10 mM MgCl 2 , and 0.002% LMNG] were applied to a holey carbon grid (Quantifoil R1.2/1.3 200 mesh grids, Microtools GmbH, Berlin, Germany) that had been pretreated by gold-sputtering 48,49 and glow-discharging (JEC-3000FC, JEOL, Japan). The grid was blotted with lter paper for 4 s, then immediately plunge-frozen in condensed ethane using an FEI Vitrobot Mark IV (ThermoFisher Scienti c, Waltham, MA, USA) under 100% humidity at 4°C. The frozen grids were then introduced into a CRYO ARM 300 electron microscope (JEOL, Tokyo, Japan) equipped with a cold-eld emission gun, after which inelastic scattered electrons were removed using an in-column type energy lter with an energy slit of 20 eV. Dose fractionated images were recorded on a K2 summit camera in counting mode with a nominal magni cation of 60,000×, which corresponded to a physical pixel size of 0.823 Å. All image data were collected with a JEOL Automatic Data Acquisition System (JADAS) 50 with a dose rate of 11e − /Å 2 /s, 10 s exposures, 0.2 s/frame, and a defocus range of -0.8 to -1.8 µm. In total, we collected 3,225 (data 1) and 4,346 movie stacks (data 2) from two independent sample preparations.

Cryo-EM image processing
Cryo-EM movie stacks were grouped into 10 separate optics groups based on time of data collection.
Drift corrected and dose-weighted summation of movie frames was performed using MotionCor2 51 , and contrast transfer function (CTF) parameters were estimated from the correct images with CTFFIND4 52 . Based on Thon-ring patterns, 4,237 images from data 1 and 2 were selected for particle picking. Good PSI particles were manually selected and subjected to reference-free two-dimensional (2D) classi cation with RELION-3.1beta. Next, homogeneous 2D class averages were used as templates for reference-based autopicking, and a total of 774,416 particles were extracted with a pixel size of 1.646 Å and a box size of 220 pixels for further 2D classi cation. Good 2D class averages contained 242,550 particles, which were passed to three-dimensional (3D) classi cation with a de novo initial model constructed using cryoSPARCv2, and a well-aligned and symmetrical 3D class was reconstructed from 86,509 particles. The particles in the best 3D class were rescaled and re-extracted with a pixel size of 1.097 Å and a box size of 330 pixels, and 3D re nement with three-fold symmetry enforcement and post-processing yielded a 3.3-Å resolution map based on the gold standard Fourier shell correlation criterion. Particles in the cryo-EM density map were then subjected to Bayesian polishing for correction of particle-based beam-induced motion and CTF re nement of particle-based defoci and optics-group-based high-order astigmatisms. Finally, 86,419 particles were selected by excluding those showing unrealistic defocus values and then subjected to 3D re nement and post-processing again, which yielded a 2.59-Å resolution map with C3 symmetrization. After beam tilt estimation and correction for every micrograph, 3D re nement and postprocessing reached 2.48-Å and 2.97-Å resolution structures with and without C3 symmetrization, respectively. B-factor values for map sharpening were estimated as -83.67 Å 2 and -77.95 Å 2 , respectively.

Model building and re nement
An initial model of the PSI monomer from A. marina was built using homology modeling in MODELLER (version 9.23) 53 by referring to the structure of the PSI monomer from T. elongatus (PDB code: 1JB0). The identity of each amino acid in subunit proteins in A. marina PSI was determined using amino acid sequences obtained from UniProt (https://www.uniprot.org/). The resulting homology model of the PSI monomer was tted to the whole cryo-EM density map (C3 map) using a " t in map" program in UCSF Chimera (version 1.13) 54 , and an initial model of the trimeric form of PSI from A. marina was created.
Next, the created PSI trimer model was tted to the cryo-EM density map using a molecular dynamics simulation program (CryoFit) and simulated annealing in Phenix 55,56 . Each model of the three monomers in the PSI trimer was re ned separately, and the re ned model was modi ed using COOT 57 manually to t the cryo-EM density map. Subsequently, the obtained structure model was nally re ned using refmac5 58 (Extended Data Fig. 11). Re nement statistics of the re ned model was obtained using a comprehensive validation program in Phenix.
marina PSI, a cryo-EM density map of Mg 2+ of Chl d could not be identi ed unambiguously, and thus the two A 0 were identi ed as Pheo. Validation of the re ned structure model was assessed using Phenix, ModelZ 59 , and MapQ 60 . The re nement statistics of the models are summarized in Extended Data Table   1. Structure gures were generated and rendered with PyMOL and Chimera 54,61 .

Identi cation of cofactors and their geometry restraint information
The restraints needed for the pigments in PSI from A. marina were generated by Electronic Ligand Bond Builder and Optimization Workbench (eLBOW) 62 because the information was not registered in the restraints library in Phenix. The restraint information for Chl d was created from the model of Chl d identi ed in a high-resolution structure (PDB code: 2X1Z) 63 , and information for Chl d', an epimer of Chl d, was created with reference to the structures of Chl d and Chl a', an epimer of Chl a (PDB codes: 1JBO and 2X1Z). The orientation of the C3-formyl group of Chl d' was determined based on that of Chl d 63 . The PSI trimer from A. marina contained α-Car instead of β-carotene (β-Car). The α-Car molecule possesses two rings-known as α-and ε-rings-at each end, while the β-Car molecule possesses two βrings. Although the restraint information for α-Car was created using eLBOW, as for the other cofactors, α-Car was registered as an unknown ligand (three-letter code: UNL) in this study because it is hard to distinguish between α-and ε-rings from the present cryo-EM density map. Restraint information for Pheo, iron-sulfur clusters, MGDG (three-letter code: LMG), and PG (three-letter code: LHG) was created with reference to high-resolution structures (PDB codes: 1JB0 and 3WU2).

Identi cation of water molecules
Water molecules were identi ed manually from the cryo-EM density map at >2.0 root-mean-square level using COOT. Eighty-four water molecules were found in the re ned structure model. Most of the water molecules were identi ed as ligands to Chl d.   Geometry of PA/PB and Acc (AccA and AccB). Chl d, green; Pheo, orange; water molecule, red.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download.