Structure of plant photosystem I-plastocyanin complex reveals strong hydrophobic interactions

Biological energy conversion relies on a sequence of redox reactions within the membrane-embedded electron transport chains of chloroplasts and mitochondria. In plants, the photosynthetic electron transport between the two photosystems has to cover distances of a few hundred nanometers. This is due to the unique architectural features of the photosynthetic thylakoid membranes, which fold into stacked grana and unstacked stroma thylakoids [1]. However, the thylakoid architecture and the configuration of the protein complexes are subject to constant modulation in response to environmental and physiological stresses [2–4]. Plastoquinone and plastocyanin (Pc) shuttles electrons from PSII to PSI, one within the hydrophobic milieu of the membrane and the second in the hydrophilic environment of the lumen [5]. Pc exploits specific binding sites on its electron donor — cytochrome f — and its electron acceptor — PSI [6]. The former interaction is a crude protein–protein interaction [7], and the latter is quite involved due to the onset of P700 oxidation at the ns time scale and the electron transfer from Pc that take place at the µs time scale [8–11].

Binding and electron transfer between Pc and PSI is driven by hydrophobic and electrostatic interactions [12]. In plants and green algae, the electron transfer reaction between Pc and PSI exhibits two distinct kinetic phases. A first fast microsecond intra-molecular electron transfer can be explained by a stable complex between PSI and Pc already formed prior to a flash. In the second slower phase, the remaining PSI complexes are re-reduced by the soluble donor in a bimolecular reaction with second-order kinetics. The occurrence of a preformed complex is dependent on the presence of the eukaryotic positively charged domain of subunit PsaF. The dissociation constant of oxidized Pc is about six times larger than that observed for reduced Pc [13], resulting in a 50–60 mV higher midpoint redox potential of Pc bound to PSI as compared with soluble Pc. This translates into a decrease in the driving force within the intra-molecular electron transfer complex, which is consistent with the fact that the P700 population is not fully reduced in vivo [14]. Moreover, single turnover flashes of isolated PSI fail to oxidize the whole P700 population [15,16].

In this work, we used the single particle cryo-EM technique to obtain a high-resolution structure of reduced Pc in complex with plant PSI and analyzed the interaction between the two complexes. For kinetic studies, a highly purified pea PSI derived from PSI crystals and purified pea Pc were employed.

In all the kinetics measurements described in this work about 50 large pea PSI crystals were collected using 0.7–1.0 mm Cryoloop (Hampton research) and released into a 0.5 ml Eppendorf tube containing 0.4 ml solution of 50 mM K₂PO₄, 50 mM Tris (pH 8) and 20% PEG 400. Following centrifugation at 14 000 rpm for 1 min, the supernatant was carefully removed, and the crystals were solubilized in 60 µl of a solution containing 20 mM Tricine-Tris (pH 8) and 0.05% n-dodecyl-α-maltoside (αDM) at a chlorophyll concentration of 1–3 mg/ml.

For cryo-EM structural determination, about 50 large pea PSI crystals were collected using 0.7–1.0 mm Cryoloop (Hampton research) and released into a 0.5 ml Eppendorf tube containing 0.4 ml solution of 50 mM K₂PO₄, 50 mM Tris (pH 8) and 20% PEG 400. Following centrifugation at 14 000 rpm for 1 min, the supernatant was carefully removed, and the crystals were solubilized in 60 µl of a solution containing 20 mM MES-Tris (pH 7), 20 mM sodium ascorbate and 0.05% αDM at a chlorophyll concentration of 6 mg/ml. A concentrated pea plastocyanin was added into the PSI solution to give final concentration of PSI 2.8 mg Chl/ml and 5 mg/ml plastocyanin. The resulting solution of PSI–Pc (3 µl) was applied on glow-discharged holey carbon grids (Cu Quantifoil R1.2/1.3), before using a Vitrobot FEI (3 s blot at 4°C and 100% humidity). The frozen grids were sent to Dr. Alexey Amunts at the SciLifeLab. Images were collected in the Karolinska Institute (KI) facility using a K3-300 kV FEI Titan Krios electron microscope. The images were collected using a 300 kV FEI Titan Krios electron microscope, with an energy filter of 10 eV. A Gatan K3-Summit detector was used in fast acquisition mode at a magnification of ×130 000 (yielding a pixel size of 0.654 Å), with a total dose of 40.8 e Å⁻². EPU was used to collect a total of 5310 images, which were dose-fractionated into 51 video frames, with defocus values ranging from 0.3 μm to 1.5 μm. The collected micrographs were motion-corrected and dose-weighted using MotionCor2 [17]. The contrast transfer function parameters were estimated using CtfFind 4.1 [18]. A total of 276 815 particles were picked using reference-free picking using the KLT picker [19]. The picked particles were processed for reference-free 2D averaging, initial model building, 3D classification and refinement resulting in 45 376 particles for intermediate model building; All steps were performed using RELION v.3 [20]. The intermediate model was subsequently used for template-based autopicking in RELION, resulting in 359 977 picked particles. This set was used applied to 2D classification resulting in 141 959 particles, then followed by 3D classification resulting in six distinct classes in RELION. From these, four classes were selected, which contained a total of 104 127 particles. These particles were pooled together and processed for 3D homogeneous refinement and postprocessing using RELION. The reported resolution was based on a gold-standard refinement, applying the 0.143 criterion on the FSC between the reconstructed half-maps. Following data processing (see Methods section) the Pc–PSI supercomplex was modeled at 2.74 Å resolution with local resolution ranging from 2.5 to 4.5 (Supplementary Figure S1).

To generate the Pc–PSI, the cryo-EM structure of the plant Pc–PSI–Fd model PDB 6YEZ was selected. This model was fitted onto the cryo-EM density map using phenix.dock_in_map in the PHENIX suite [21], and manually rebuilt using Coot [22]. Stereochemical refinement was performed using phenix.real_space_refine in the PHENIX suite [21]. The final model was validated using MolProbity [23]. The refinement statistics are provided in Supplementary Table S1. Local resolution was determined using ResMap [24], and the figures were generated using PyMOL [25] and UCSF Chimera [26]. Representative cryo-EM densities are shown in Supplementary Figure S9.

Wild type and mutant C. reinhardtii strains were grown as described [27]. Tris-acetate-phosphate (TAP) medium was used to grow cells for PSI isolation. The growth tests were performed on solidified (1.5% agar) TAP medium for photoheterotrophic and high salt medium (HSM) for photoautotrophic growth.

Whole cells were used for Thylakoid membrane isolation according to established procedures [28]. Afterwards, the isolated membranes were diluted to a chlorophyll concentration of 0.8 mg/ml, solubilized and incubated on ice for 20 min in the presence of 0.9% (w/v) dodecyl β-maltoside. The photosynthetic complexes were separated using a linear sucrose density gradient [29]. The photosystem I was collected and concentrated by using ultra-filtration colums with a size exclusion of 100 000 MW.

Plastocyanin was heterologously expressed as described using the E. coli BL21 strain. The concentration of the isolated Pc was determined spectroscopically as described [30].

The cell growth at different light conditions was determined by spotting 20 µl of cells from log-phase cultures, which were diluted to the same cell-number (1 × 10⁶cells/ml) onto agar plates. These plates were kept under 400 µE m⁻²s⁻¹ termed as high light, 40 µE m⁻²s⁻¹ as normal light and 5 µE m⁻²s⁻¹ as low light, respectively. The Genebag Anaerobiosis system provided by BioMérieux was used to establish anaerobic growth.

The point mutations Asp647Arg, Arg648Asp and Asp647Arg/Arg648Asp in the photosystem I subunit PsaA were introduced by site-directed mutagenesis [31]. The transformation success was confirmed by sequencing as well as the expression and the expression level of the altered photosystem I by Western blot analysis.

SDS–PAGE (13% T, 3.45% C) was carried out as published [32]. Proteins were transferred onto a nitro-cellulose membrane by semidry electro-blotting. The immuno detection was performed using anti-PsaD and anti-PsaF antibodies for PSI detection in a 1 : 1000 dilution as well as anti-D1 and anti-LhcbM6 antibodies of PSII detection in a 1 : 2000 and 1 : 1000 dilution, respectively.

The kinetics of flash-induced absorbance changes at 817 nm were measured as described before [13] using a cuvette containing 200 µl of the sample with an optical path length of 10 mm. The chlorophyll content was determined according to the protocol published by Porra [33].

The light driven, PSI dependent reduction in NADP⁺ was performed as described by Finazzi et al. [34]. The sample was illuminated with saturating light supplied by a halogen lamp with a light intensity of ∼10⁴µE m⁻²s⁻¹.

One-hundred and fifty annotated and reviewed sequences from various photosynthetic phyla where used to generate a multiple sequence alignment using MUSCLE [35] (sequences taken from UniProt). The WebLogo [36] was generated taking subsequence representing the PsaA α-helix l’ from the previously generated alignment.

The availability of large amounts of Pisum sativum PSI crystals have offered a unique opportunity to solve the structure of photosynthetic supercomplexes by the cryo-EM technique [37]. We have utilized these crystals to form a plant Pc–PSI complex simply by adding purified Pc to pure PSI. To ensure the complex formation the concentration of Pc was kept way above the dissociation constant. Plant PSI was solubilized from stored crystals in a solution containing 20 mM MES-Tris (pH 7), 20 mM sodium ascorbate and 0.05% n-dodecyl-α-maltoside (αDM) at a chlorophyll concentration of 6 mg/ml. Concentrated pea Pc was added into the PSI solution to give final concentration of 2.8 mg Chl/ml PSI and 5 mg/ml Pc. The resulting solution of PSI–Pc (3 µl) was applied on glow-discharged holey carbon grids that were vitrified for cryo-EM structural determination (see Methods). Following data processing (see Methods section) the PSI–Pc supercomplex was solved at 2.74 Å resolution with local resolution ranging from 2.5 to 4.5 (Supplementary Figure S1, Supplementary Table S1, PDB 6ZOO). The structure of Pc–PSI complex is shown in Figure 1. The position on PSI of the reduced Pc is virtually identical with its position in the previously reported Pc–PSI–Fd complex ([37]; PDB 6YEZ). However, improved map densities of Pc (Supplementary Figure S2) allowed a closer look not only at the electrostatic but also potential hydrophobic interactions (Figure 1; Supplementary Movie S1; PDB 6ZOO). Figure 1 shows some of the identified surface interactions where previous molecular biology studies revealed deleterious mutations, with PsaA Trp651 and PsaB Trp627 (Trp658 and Trp625 in pea, respectively) at the center of the hydrophobic binding region [38,39]. Mutagenizing PsaB Glu613 and Asp612 (pea Glu611 and Asn612), which are located in the periphery of Pc binding domain, increased the affinity of Pc binding to PSI alongside with slower Pc release [10,38]. PsaF lysines were observed in close proximity to Pc negatively charged amino acids as reported in [37], with the addition of PsaA Arg117 identified 5.7 Å away from Pc Asp61. Substitution of these lysines disrupted Pc binding, release, and electron transfer to P700 [40]. On the Pc hydrophobic surface, mutating Pro36 — which faces the two tryptophans [41] — as well as Gly10, Leu12, Tyr83 and Ala90 [8,42,43] showed a decreased efficiency in Pc binding and activity. The hydrophobic Pc region involved in PSI–Pc association was suggested to serve a similar role in Pc binding to cytochrome f [44,45].

While structural information readily reveals electrostatic interactions, hydrophobic interactions are much more difficult to decipher [37]. One of the hallmarks of such interaction are the close contacts of the hydrophobic surfaces leaving no space for water molecules. Analysis of the solvent-accessible surface area (SASA) of the Pc binding domain formed by PsaA–PsaB and Pc surface using GetArea with default parameters [46,47] demonstrates that more than 50% of the binding regions become inaccessible to water molecules once the PSI–Pc complex forms, fitting a tightly packed hydrophobic binding (Figure 1B–E, Table 1). These regions are composed of four domains in Pc and five domains in PsaA and PsaB are presented in Figure 1, and several of these amino acids were previously identified as necessary for the transient binding of Pc [8,9,34,37,38,39]. These binding domains display a combination of electrostatic, polar, and hydrophobic interactions with varying magnitudes. Pc domains span from Asp8–Leu12 (domain I), Ala33–Pro36 (domain II), Glu60–Leu62, Asn64–Glu68 (domain III) and Pro86–Gly91 (domain IV). PsaA and PsaB association domains span more intermittently but form five distinct surfaces encircling PsaA Trp658 and PsaB Trp625. The first PsaA surface is formed by Thr635, Ile637–Gly639 and Asn641, the second by Asp655, Trp658, Ala659, Ser662, Gln663, Gln666, Gly669, Ser670 and Leu672 and the third by Arg753, Ala756 and Val757. PsaB first domain consists of Gln603, Asn605 and Gln608, and the second by Arg621, Asp622, Trp625–Asn627, Ser629, Gln630, Asn633, Asn636, Phe638, Asn641, Leu643 and Lys732.

As part of our efforts to describe Pc association and dissociation mechanism, we used in-silico modeling of oxidized Pc from Populus nigra ([48]; PDB 4DP9) as our reference. Sequence comparison of Pc from P. nigra and pea using BLAST [49] shows a 79% identity (Supplementary Figure S3). The sequence was changed according to pea Pc and then superpositioned on the reduced Pc in 6ZOO. Conformation changes which appeared significant were identified in both Pc acidic patches Asp42–Glu45 and Glu59–Glu60 (Figure 2). The short helix formed by Asp42–Glu45 in oxidized Pc brings Glu43 and Asp44 closer to PsaF Lys93, Lys96 and Lys100, and Glu60 drifts closer to PsaF Lys101 (Figure 2, Supplementary Movie S2). Next, we superimposed the two conformations based on the unchanged PSI interaction region (Asp61–Asn64). In this superposition, the negative patches mentioned before were closer to PsaF Lysines, and tilted relative to the hydrophobic binding region. With this, Pro86–Gly91 moved by 1–4 Å compared with the 6ZOO conformation, leaving ample room for water molecules to accommodate the space between Pc and PSI and drive Pc dissociation from PsaA–PsaB towards PsaF. Inspection of the in-silico model SASA shows that the oxidized Pc surface solvent accessibility increased by 5% compared with the reduced Pc, mostly in the hydrophobic binding patches (Supplementary Table S2).

To further assess the role of amino acids close to the hydrophobic core represented by PsaA-Trp658, residues PsaA-Arg647 and Asp648 (Arg654 and Asp655 in pea, respectively) were altered via a site-directed mutagenesis approach using Chlamydomonas reinhardtii as a model. Notably, Asp648 (pea Asp655) is found at the beginning of the second PsaA surface. Both amino acids PsaA-Arg647 and Asp648 are found very close to Trp651 (pea Trp658) and are also present right at the PsaA–PsaB border. PsaA-Asp648 faces Pc Phe35 directly, and the salt bridge formed by Arg647 and Asp648 (Supplementary Figure S4) prevents them from drifting towards PsaA Trp658 or PsaB Asn633 (pea annotations), the latter found at the interface of the first Pc domain formed by Asp8–Leu12. The coding sequence for both residues in the psaA gene was altered to generate a scenario where both encoded residues are positively charged (PsaA-Asp648Arg), both are negatively charged, (PsaA-Arg647Asp) and where the charge is switched (PsaA-Arg647Asp/Asp648Arg). SDS–PAGE fractionation of isolated thylakoid membranes followed by an immuno-blot analysis using anti-PsaD and anti-PsaF antibodies revealed diminished PSI contents of ∼25–75% as compared with wild type, while the PSII content was stable (Supplementary Figure S5).

The second-order electron transfer rates for the electron transfer between Pc and purified PSI complexes were determined via in vitro single turnover flashes for excitation at increasing ionic strength (Figure 3A). All genetically modified photosystems revealed a diminished second-order transfer rate as compared with wild type, independent of the salt concentration, with the strongest impact for PsaA-Asp648Arg (Supplementary Table S3).

These findings were supported by light-induced PSI dependent NADP⁺ reduction measurements in an independent approach. At low Pc concentrations (1.5 µM) the photoreduction rate was about 2-3 times (PsaA-Arg647Asp, PsaA-Asp648Arg) up to 5 times smaller (PsaA-Arg647Asp/Asp648Arg) for altered PSI was lower as compared with wild type (Figure 3B). Similar effects were observed at high Pc concentrations (80 µM), yet the difference was only about 1.5 times smaller for PsaA-Arg647Asp and PsaA-Arg647Asp/Asp648Arg and 2.5 times smaller for PsaA-Asp648Arg. Both independent approaches indicated an impaired electron transfer from Pc to the altered PSI, independent from the residue combination.

Absorbance transients at 817 nm, induced by fast single turnover flashes, were also recorded at increasing Pc concentrations (Figure 3C). The absorbance transients at high Pc concentrations can be separated into three kinetic components. The fast component A(1) with a constant half-life of 3–4 µs representing the first-order electron transfer reaction within a preformed complex, a second components A(2) with a half-life decreasing at higher concentrations indicative of a second-order binding reaction, and a very slow third component A(3) representing PSI complexes that lost their PsaF subunit during the isolation process making up to 25–40% of the total PSI signal [50]. The changes in amplitude of the two faster kinetic components in regard to the donor concentration reflect the binding equilibrium and allows the calculation of the dissociation constant K_D according to Drepper et al. [13]. The K_D value for WT PSI was measured at 88 mM, similar to previous work [40] Determination of the K_D values for all altered photosystems was not possible as the amplitude A(1) did not significantly increase at increasing Pc concentrations, visible in the absorbance transients measured in the presence of 50 and 250 µM Pc (Figure 3C). This reflects a drastically decreased affinity of Pc for the altered photosystems, indicating that changing charges at PsaA residues Arg647 and Asp648 omitted efficient pre-binding of reduced Pc.

Electron transfer measurements were performed using a uniform pea PSI population derived from solubilized diffracting crystals to substantiate that these PSI complexes are very competent in electron transfer. The light-induced P700 oxidation and re-reduction by various electron donors were followed by recording absorption changes at 705 nm using a Joliot-type spectrophotometer (JTS). The signals converted to assume a mM extinction coefficient of 64 at 705 nm [51]. The light-induced absorption change at 705 nm measured by JTS that was calibrated by using plant PSI crystals containing 156 chlorophyll molecules per P700. The standard variation of Chl/P700 ratio from five independent experiments is ±2%, caused mainly due to variation in chlorophyll determination and not the extent of P700 signal.

The data revealed that the reduction in light-induced oxidized P700⁺ is biphasic with a fast kinetics of ∼100 ms and a slow one of several minutes. The addition of ferredoxin eliminated the fast kinetics. Ascorbate failed to reduce all the population of P700⁺ and the idle P700 disappeared in the presence of PMS and thus the signal represents the entire P700 population. The presence of PMS eliminated the fast 10 ms back kinetics that was attributed to FB. Only partial elimination of the back reaction was observed in the presence of Pc.

No significant change in OD was observed during illumination from 0.1 to 1 s (Supplementary Figure S6). Similar to ferredoxin and PMS, the presence of methylviologen also eliminated the fast kinetics (Figure 4A). The addition of NaCl significantly decreased the rate of P700⁺ reduction by Pc (Figures 3A and 4B). Pre-incubation with sufficient Pc amounts should saturate the plastocyanin binding site even under conditions where the electrostatic interactions are eliminated by the presence of high salt concentration. Measurements at a shorter time scale revealed that the A(1) phase, the initial P700⁺ reduction by pre-bound Pc, is insensitive to the presence of high salt concentration in the reaction mixture (Figure 4C). Under these conditions hydrophobic interactions dominate the Pc binding to PSI, underpinning the importance of such interactions between PSI and Pc.

In PSI of oxygenic organisms, the binding site of the reduced electron carriers is mediated by both cytochrome c₆ and Pc. Vascular plants developed specificity for Pc, while cyanobacteria and algae use both as viable electron donors. Three subunits of PSI PsaA, PsaB and PsaF are directly involved in Pc binding [37]. It was reported that the PSI binding site involve both electrostatic and hydrophobic interactions [8,9]. The structure of Pc–PSI supercomplex reveals not only the electrostatic interactions reported before [37] but also hydrophobic interactions (Figure 1B–E, Table 1). It was suggested that the observed kinetics can be explained by a rate-limiting conformational change that occurs in the Pc–PSI complex before the electron transfer takes places [52,53]. We found no significant structural alterations of PSI in response to Pc binding. This is in line with an alternative model, that explains the kinetic properties of the electron transfer between Pc and PSI, where the driving force of the intercomplex electron transfer is decreased to ΔG = −55 meV and tuned for an optimized turnover of PSI [13]. The formation of the intermolecular PSI–Pc complex leads to an exclusion of water molecules as revealed via the structural data (Figure 1). Notably, this is in accordance with a low reorganization energy λ of ∼418 meV, that was determined for the intermolecular electron transfer between Pc and PSI [54], indicating a rather hydrophobic environment at the contact side of both proteins. Thus, electron transfer between Pc and PSI is only coupled to a small reorganization of solvent molecules supporting fast interprotein electron transfer despite relatively low driving force. According to Marcus and Sutin [55], such an exclusion of solvent molecules would result in a decrease in standard entropy of reduction ΔS° of bound compared with soluble Pc mainly from changes in freedom of the rotation and liberation of the solvent molecules which surround the reaction partners. Taking advantage of the thermodynamic parameters such as ΔG (free energy change) and λ (the reorganization energy) and an empirical approximation for intraprotein electron transfer [56], an edge-to-edge distance R between Pc and PSI redox centers of 14.5 Å were estimated [54]. This value is very close to the recent structural determination of this distance of 14.4 Å from Cu⁺ to P700 CBC_A atom. Therefore, in conclusion, the thermodynamic and structural data are consistent, both supporting the exclusion of solvent molecules from the intermolecular electron transfer complex between PSI and Pc. Interestingly, the exclusion of solvent molecules from the intermolecular complex also has a consequence of the pH-dependent redox midpoint potential of Pc [54].

The importance of the hydrophobic, polar and electrostatic properties formed by the two α-helices l’ and j’ for binding of Pc to PSI was further investigated by site-directed mutagenesis of a Arg/Asp residue pair PsaA-Asp648/Arg647. PsaA-Asp648/Arg647 are close to the Trp pair (PsaA-Trp651 and PsaB-Trp627), where indole groups are arranged in a sandwich-like structure above the P700 chlorophyll pair. An alignment of 114 eukaryotic and 36 cyanobacterial PsaA amino-acid sequences revealed a very high sequence similarity, especially for the part of the sequence representing the α-helix l’ (Supplementary Figure S7). The residues PsaA-Arg647 and PsaA-Asp648 are conserved through all sequences, thus only in Dinoflagellate the PsaA-Asp648 is substituted by Asparagine. The single flash absorption measurements revealed diminished second-order rate constants as compared with wild type. The lowest rate was measured for PsaA-Asp648Arg (6.539 × 10⁷ k₂/M⁻¹s⁻¹). On the other hand, K_D values could not be determined for the electron transfer between Pc and altered PSIs as no or only a minor increase in the A(1) amplitude with increasing Pc concentration was observed. Thus, indicating that the K_D value is substantially increased due to destabilized complex formation between Pc and the altered PSIs. A similar effect was observed for the PsaB-Trp627Phe mutant PSI, where the secondary electron transfer rate was also only slightly diminished, whereas the affinity was strongly impaired. Thus, stable complex formation between PS and Pc is not required for rather productive electron transfer, as revealed by less severe impact on second-order electron transfer constants. However, an exact positioning of electrostatic and polar properties of PsaA-Arg647/Asp648 are required for stable binding of Pc supporting the importance for a precise polar, electrostatic and hydrophobic landscape required for stable complex formation between PSI and Pc. This is also supported by the fact, that high salt concentrations did not impact the formation of the first-order electron transfer complex, while the second-order electron transfer constant was diminished (Supplementary Table S4). This clearly underpins the significance of hydrophobic interactions for stable complex formation between PSI and Pc. The importance of a stable intermolecular electron transfer complex between PSI and Pc is signified by a severe growth phenotype under high light of mutant strains PsaA-D648R, PsaA-R647D and PsaA-R647D/D648R (Supplementary Figure S8). While the polar, electrostatic and hydrophobic landscape is crucial for the tight and stable association between PSI and reduced Pc, it appears to be the weakening of the hydrophobic interaction caused by oxidation of Pc (Supplementary Movie S2), that allows water molecules to accommodate the space between Pc and PSI and that drives Pc dissociation from PSI. This in turn likely explains the six times larger dissociation constant of oxidized Pc as compared with reduced Pc [13], responsible for the 50–60 mV higher midpoint redox potential of Pc bound to PSI.

In summary, the electron transfer between Pc and PSI can be described by the following steps, (i) binding of Pc to PSI facilitated via electrostatic interactions between Pc and PsaF, (ii) stabilization of the reduced Pc–PSI complex due to the hydrophobic as well as polar and electrostatic interface provided by the PsaA and PsaB association domains, (iii) electron transfer from Pc to P700⁺ within the intermolecular complex, (iv) dissociation of bound oxidized Pc driven via its conformational change, that destabilizes its hydrophobic interaction with PsaA and PsaB interfaces additionally supported through electrostatic pulling towards PsaF.

In conclusion, our data provide mechanistic insights at molecular resolution revealing how the light-driven electron transfer between Pc and PSI was optimized for fast turnover.

The atomic co-ordinates have been deposited in the Protein Data Bank, with accession code 6ZOO. The cryo-EM maps have been deposited in the Electron Microscopy Data Bank, with accession codes EMD-11326. All other data are presented in the main text or supplementary materials.

The authors declare that there are no competing interests.

This work was supported by The Israel Science Foundation (grant no. 569/17), and by German-Israeli Foundation for Scientific Research and Development (GIF) to N.N and M.H., grant no. G-1483-207/2018. M.H. acknowledges support from German Science Foundation (DFG, Hi 739/13-2). N.B-T acknowledges the support of NSF-BSF grant 2019658 and Abraham E. Kazan Chair in Structural Biology, Tel Aviv University.

Open access for this article was enabled by the participation of Tel Aviv University in an all-inclusive Read & Publish pilot with Portland Press and the Biochemical Society under a transformative agreement with MALMAD.

I.C., M.F., S.K., A.B.-S., D.K., and N.N. performed the research. I.C., F.D., M.H. and N.N. analyzed the data. I.C., M.F., M.H. and N.N. wrote the manuscript. All the authors discussed, commented on and approved the final manuscript.

Nathan Nelson: Conceptualization, supervision, funding acquisition, investigation, methodology, writing — original draft, project administration, writing — review and editing. Ido Caspy: Data curation, software, formal analysis, validation, investigation, visualization, methodology, writing — original draft, writing — review and editing. Mariia Fadeeva: Validation, investigation, methodology. Sebastian Kuhlgert: Investigation, visualization, methodology. Anna Borovikova-Sheinker: investigation. Daniel Klaiman: Investigation. Gal Masrati: Investigation. Friedel Drepper: Investigation. Nir Ben-Tal: Investigation. Michael Hippler: Conceptualization, funding acquisition, validation, investigation, methodology, writing — original draft, project administration, writing — review and editing.

Dr Yael Levi-Kalisman is gratefully acknowledged and thanked for vitrifying the samples. We thank Dr. Alexey Amunts for data collection in the SciLifeLab Cryo-EM facility, Solna, Sweden. We would like to thank Kevin Redding from the Department of Chemistry and Biochemistry of the Arizona state University for providing the ΔpsaA strain as well as the plasmid used for site-directed mutagenesis and transformation. Molecular graphics and analyses were performed with UCSF Chimera, developed by the Resource for Biocomputing, Visualization, and Informatics at the University of California, San Francisco, with support from NIH P41-GM103311.

JTS

Joliot-type spectrophotometer

SASA

solvent-accessible surface area

TAP

tris-acetat-phosphate