Algal photosystem I dimer and high resolution model of PSI:plastocyanin 1 complex 2

21 Photosystem I (PSI) enables photo-electron transfer and regulates photosynthesis in the 22 bioenergetic membranes of cyanobacteria and chloroplasts. Being a multi-subunit complex, its 23 macromolecular organization affects the dynamics of photosynthetic membranes. Here, we reveal 24 a chloroplast PSI from the green alga Chlamydomonas reinhardtii that is organized as a 25 homodimer, comprising 40 protein subunits with 118 transmembrane helices that provide scaffold 26 for 568 pigments. Our cryo-EM structure identifies that the absence of PsaH and Lhca2 gives rise 27 to a head-to-head relative orientation of the PSI-LHCI monomers in a way that is essentially 28 different from the oligomer formation in cyanobacteria. The light-harvesting protein Lhca9 is the 29 key element for mediating this dimerization. The interface between the monomers is lacking PsaH, Together, the data explain how the PSI is modulated to perform its functional and structural roles in a chloroplast.

Photosystem I (PSI) enables photo-electron transfer and regulates photosynthesis in the 22 bioenergetic membranes of cyanobacteria and chloroplasts. Being a multi-subunit complex, its 23 macromolecular organization affects the dynamics of photosynthetic membranes. Here, we reveal 24 a chloroplast PSI from the green alga Chlamydomonas reinhardtii that is organized as a 25 homodimer, comprising 40 protein subunits with 118 transmembrane helices that provide scaffold 26 for 568 pigments. Our cryo-EM structure identifies that the absence of PsaH and Lhca2 gives rise 27 to a head-to-head relative orientation of the PSI-LHCI monomers in a way that is essentially 28 different from the oligomer formation in cyanobacteria. The light-harvesting protein Lhca9 is the 29 key element for mediating this dimerization. The interface between the monomers is lacking PsaH, 30 and thus partially overlaps with the surface area that would bind one of the LHCII complexes in 31 state transitions. We also define the most accurate available PSI-LHCI model at 2.3 Å resolution, 32 including a flexibly bound electron donor plastocyanin, and assign correct identities and 33 orientations of all the pigments, as well as 621 water molecules that affect energy transfer 34 pathways. 35 36

Main 37
A chloroplast PSI of green algae consists of the core complex and three antenna modules: inner 38 belt, outer belt, and Lhca2:Lhca9 heterodimer, which together comprise 24 subunits 1-4 . As a short-39 term light acclimation mechanism in response to fluctuating illumination and anoxia, the algal PSI 40 additionally associates with two LHCII trimers 5,6 . Structural studies have shown that the 41 oligomeric state of a chloroplast PSI is a monomer, due to the presence of the subunit PsaH, 42 whereas in cyanobacteria structures of dimers 7-10 and trimers 11 were also reported. Cyanobacterial 43 PSI oligomerizes via direct contacts between subunits PsaI and PsaL, however such an association 44 has been ruled out for a chloroplast PSI due to structural constraints of PsaH presence that imposes 45 an apparent rigidity 12,13 . Yet, recent structural studies of PSI from a chloroplast of a salt-tolerant 46 alga suggested that its functional core may vary more than previously believed 14 . Particularly, 47 functional PsaH-free particles were found, thus showing a potential architectural plasticity of PSI 48 in response to the ecological environment. On the macromolecular level, an atomic force 49 microscopy analysis of a plant thylakoid membrane showed that when its architecture is altered 50 upon transition from darkness to light, larger inter-membrane contacts are formed, leading to a 51 reduced diffusion distance for the mobile electron carriers 15 . The membrane architecture in dark-52 and light-adapted membranes contains ordered rows of closely packed PSI dimers, which are more 53 abundant in the dark state 15 . Similarly, closely associated PSI-LHCI complexes were detected in 54 plants by negative stain electron microscopy 16 , and dimers were found in a subpopulation of PSI 55 from a temperature-sensitive PSII mutant alga 17 . This suggests that reversible PSI dimer formation 56 may have a physiological role in thylakoid membrane structure maintenance in chloroplasts. 57 However, very little is known about PSI-LHCI dimers and information on their structures is 58 lacking. In the absence of high resolution data, no evidence is available on composition, elements 59 regulating and mediating dimerization, and how the arrangement would differ from the 60 cyanobacterial counterparts. 61 62

Structure determination 63
We grew C. reinhardtii cells containing a His-tag at the N-terminus of PsaB in low light and under 64 anoxic conditions (see Methods). The thylakoid membranes were solubilised with n-dodecyl-α-D-65 maltoside (α-DDM), followed by affinity purification, crosslinking via the chemically activated 66 electron donor plastocyanin (Pc) and sucrose density gradient centrifugation (Extended Data Fig.  67 1). Two PSI fractions were detected on the sucrose gradient, and 2D polyacrylamide gel 68 electrophoresis (native/reducing 2D-PAGE) of isolated thylakoids indicated the presence of PSI 69 dimers (Extended Data Fig. 2 and Supplementary Table 1). The heavier green band on the gradient 70 was subjected to single-particle cryo-EM analysis (Supplementary Table 2). We used 2D 71 classification to separate PSI dimers from monomers in a reference free manner, followed by 3D 72 classification leading to a subset of 14,173 particles, which were refined to an overall resolution of 73 2.97 Å by applying C2 symmetry (Extended Data Fig. 3). PSI dimers were also found in 2D class 74 averages in a dataset recorded from a sample without the use of crosslinker. Upon symmetry 75 expansion, the resolution was further improved to 2.74 Å (Extended Data Fig. 3 To derive a structure of the chloroplast PSI dimer, we first built an accurate model of one monomer 82 using the 2.74 Å resolution map, and then fitted it into the cryo-EM density of the C2 refined dimer. 83 Compared to the monomer, all but two core subunits (PsaH, PsaO) and one light-harvesting protein 84 (Lhca2) are found in the dimer (Fig. 1, Supplementary movie 1). The structure contains 40 protein 85 subunits, 398 chlorophylls a, 60 chlorophylls b, 56 beta-carotenes, 54 luteins, 2 violaxanthins, 2 86 neoxanthins, 4 phylloquinones, 6 iron-sulphur clusters, and 32 lipids (Fig. 1). In addition, two 87 unaccounted densities corresponding to two loops in the stromal side of Lhca9 and PsaG could be 88 interpreted in the dimer, due to the stabilization by the adjacent monomer (Extended Data Fig. 4). 89 The first better-defined density is the Lhca9 loop region 132-153 which is stabilised due to a direct 90 interaction with PsaL of the second monomer within the dimer. As a result, the area is closely 91 packed with PsaG, and therefore also the PsaG loop region 63-77 is better resolved in the dimer 92 (Extended Data Fig. 4)

Structural basis for PSI dimerization 104
The structural basis for the algal chloroplast PSI dimerization is fundamentally different from 105 cyanobacteria (Fig. 2a). In cyanobacteria, PSI dimerises via the stromal region of PsaL 7-10 and 106 trimerises via the lumenal C-terminus of PsaL, assisted by PsaI 11 . In our structure of the chloroplast 107 PSI-LHCI dimer, neither PsaL nor PsaI interacts with each other between the neighbouring units. 108 Instead, PsaH that normally preserves a monomer is not present, and Lhca9 with its associated 109 cofactors acts as a symmetrical linker between the monomers, highlighting the importance of the 110 light-harvesting antenna proteins for regulation of the macro-organisation. Lhca9 is distinct among 111 the light-harvesting proteins in our structure due to a truncated loop between helices A and C, and 112 lack of the associated chlorophyll 6 . As a result, it contains the fewest chlorophylls among Lhcas 113 (Supplementary Table 3). Based on this difference, we rationalised how Lhca9 allows for 114 dimerization, as a longer AC-loop would clash with the neighbouring PsaB (Extended Data Fig.  115 5). 116 The two Lhca9 copies tether the PSI monomers in a head-to-head fashion, resulting in a 340-Å 117 long structure (Fig. 1, Fig. 2b, Supplementary movie 1). They form interactions of four types 118 covering the entire membrane span: 1) a hydrogen bond of the backbone carbonyl of G148 with 119 S137 of PsaL in the stroma; 2) a hydrogen bond between the two Q109 of the Lhca9 copies; 3) 120 hydrophobic contacts via coordinated cofactors in the membrane that include a newly modelled, 121 beta-carotene 62 3 (N2 in nomenclature according to ref. 6 ), and five chlorophylls (604, 610, 611, 122 612, 810); 4) lipid-mediated hydrophobic interactions via monogalactosyl diglyceride LMG852 123 and LMU624. One acyl chain of lipid 852 associates with chlorophyll 810 from monomer-2, while 124 the other acyl chain associates with beta-carotene BCR623 from monomer-1 (Fig. 2b). Thus, lipids 125 contribute to the oligomerization of PSI, meaning that the membrane itself plays a role in the 126 association. The finding that specific carotenes and lipids enable inter-molecular contacts that 127 bridge the PSI monomers is of a particular interest, as it can only be detected by high-resolution 128 structural studies. Similarly, a recent structure of the RC-LH1 dimer from Rhodobacter 129 sphaeroides revealed a bound sulfoquinovosyldiacylglycerol that brings together each monomer 130 forming an S-shaped array 21 . The involvement of lipids in the oligomerization is consistent with 131 the formation of supercomplexes in other bioenergetic membranes 22-25 . 132 To solidify the structural observations, we engineered a lhca9 insertional mutant having the His-133 tag at the N-terminus of PsaB and repeated the purification procedure in the same way as for the 134 wild type. This time, no PSI dimer band could be found in the sucrose density gradient (Extended 135 Data Fig. 6). Notably, Lhca9 is present in Δlhca2, while Lhca2 is absent from Δlhca9 26 . Moreover, 136 Lhca9 stably associates with PSI-LHCI after sucrose density gradient centrifugation of solubilized 137 thylakoids isolated from lhca2 insertional mutant (Extended Data Fig. 7). PsaI/L. The PSI core is grey, LHCI light-blue. b, The dimer interface is formed by: hydrogen bonds 142 between PsaL and Lhca9, and between two Lhca9 copies (left); potential energy transfer paths 143 between the two monomers (centre); pigments and lipids (right). 144 145

Implications of PSI dimerization 146
The specific interactions between the monomers are enabled due to unoccupied positions of PsaH 147 and Lhca2. Since PsaH is also required for the lateral binding of LHCII to the PSI core in state 148 transitions 5,6 , we next compared the structure of PSI-LHCI dimer to the state transition complex 149 (Fig. 3). The superposition shows that Lhca9 from the neighbouring monomer is positioned in the 150 membrane, where Lhca2 resides in PSI-LHCI-LHCII, and their three transmembrane helices would 151 overlap with each other (Fig. 3b). The presence of the PsaH transmembrane helix is not compatible 152 with the Lhca9 2 -associated cofactors CLA9, LMG852, BCR9 that extend from the neighbouring 153 monomer in the dimer. In addition, the superposition shows that there would be a clash between 154 PsaG and Lhca1 of the inner belt with one of the LHCII trimers, but not the other (Fig. 3b). Since 155 Lhca2 and PsaH are absent, the structure of the algal PSI dimer would not facilitate LHCII binding 156 at this position. However, our 2D-PAGE indicated a comigration of LHCII polypeptides with the 157 dimer fraction, and therefore a structural adaptation cannot be excluded (Extended Data Fig. 2). 158 The antagonistic relationship of Lhca9 2 and Lhca2, and the assembly state of PsaH might further 159 reflect a regulation of PSI dimerization (Fig. 3a). 160 PsaH is a 11-kDa transmembrane protein that is imported into chloroplasts and peripherally 161 associates  To further extrapolate potential conformational changes during the dimerization of PSI, we applied 205 multi-body refinement analysis of the PSI dimer using the two monomers as bodies (Extended Data 206 Fig. 9). The analysis indicated no distinct conformational states, but instead revealed continuous 207 motions in the three eigenvectors describing a relative movement of the monomers in relation to 208 each other (Extended Data Fig 9a). The intrinsic flexibility is dominated by combinations of all 209 three rotations of one monomer with respect to the other up to 13° (Extended Data Fig. 9b-d). 210 Therefore, excitation energy transfer between the PSI monomers in the dimeric scaffold would also 211 depend on degrees of rotation around the identified pivot points. Specifically, three chlorophylls 212 are found within a potential cross-monomer excitation-sharing: CLA807 (PsaB), CLA604 213 (Lhca9 2 ), CHL606 (Lhca9 2 ), and the distance between them is ~20 Å in the consensus map ( Fig.  214 2). While such a positioning might suggest direct coupling, the multi-body analysis indicates 215 considerable variability (Extended Data Fig. 9e,f). Therefore, similar to the cyanobacterial PSI 216 dimer, an excitation coupling between the two monomers is less favourable in vitro, and this is 217 consistent with measurements of 77 K fluorescence spectra that showed only a minor shift between 218 monomer and dimer (Extended Data Fig. 1). However, in vivo the observed PSI-LHCI dimer 219 conformation, and therefore the distance between the chlorophylls at the interface, could also be 220 affected by a local membrane curvature. 221 222

High resolution features and solvation of PSI 223
In our PSI monomer reconstruction, the resolution in the core is ~2.1 Å, and in the LHCI inner belt 224 2.1-2.5 Å, revealing unprecedented structural details of chlorophylls, carotenoids, and 621 water 225 molecules (Fig. 4, Supplementary Table 2). The map can improve the level of detail not only when 226 compared to the previous cryo-EM studies of algal PSI 1-3,17 , but also the plant PSI maps obtained 227 by X-ray crystallography 19,20 (Extended Data Fig. 10). The quality of the data aided in improving 228 the previous models in functionally important regions. This includes the identification of nine Chl 229 b molecules, two newly modelled luteins, a beta-carotene and more accurate estimation of the 230 coordination of 53 chlorophylls (Fig. 4b, Extended Data Fig. 11). Particularly, Chl b molecules are 231 identified at positions 601 and 606 in Lhca4, Lhca5, Lhca6, Lhca7 and Lhca8. The two newly 232 modeled luteins 720 and 626 are in the N-terminus of Lhca3 next to Chl a 614, and in Lhca5 next 233 to Chl a 617, respectively (Fig. 4b). The newly modeled beta-carotene 622 is in Lhca9 and could 234 be identified due to structural stabilization of the interface region in the dimer (Fig. 4b). Since chl 235 b limits free diffusion of excitation energy 42 , some of the new assignments affect the energy 236 pathways between the antenna proteins. Together with the new structural data, this allowed us to 237 produce a more accurate map of the energy channelling in PSI based on the new model (Fig. 4a). 238 Another striking feature of the high resolution cryo-EM map is resolvable density for multiple 239 newly detected water molecules, which particularly aided in modeling the coordination of 240 chlorophylls (Fig. 4c, Supplementary movie 1). Thus, we report the most complete available 241 experimental picture of a chemical environment for chlorophyll binding (Supplementary Table 3). 242 Particularly, it allows distinguishing between mono-and di-hydrated forms, which largely escaped 243 detection by X-ray crystallography (Extended Data Fig. 12). This is mechanistically important 244 because the di-hydrated derivative is chemically more stable, as illustrated by quantum chemical 245 calculations 43 . We observe that, other than the previously reported CLA824 20 , only two waters 246 can be involved in penta-coordinated Mg for all the chlorophylls. Remarkably, waters play a 247 coordinative role for most of Chl b, for which the relative ratio of water coordination is four times 248 higher than for Chl a (Supplementary Table 3). The difference between Chl a and Chl b is a methyl 249 versus a formyl group, thus water serves as a hydrogen bond donor to the latter, while it also 250 interacts with charged/polar protein residues or lipids. Therefore, the immediate surrounding of 251 Chl b molecules is more enriched with non-protein material than previously thought, which plays 252 a role in tuning the photophysics and the transport properties of excitation energy in PSI. Together, 253 the presented model now allows for comparison of PSI phylogenetic conservation also on the level 254 of chlorophyll coordination and solvent positioning. phylloquinones purple, iron-sulphur clusters yellow-red. 263

Biochemical analysis of PSI 354
For SDS-PAGE (Extended Data Fig. 1c- Extended Data Fig. 3 shows the processing scheme applied. The pre-processing steps were 436 performed using cryoSPARC 3.1.0 61 . Movie stacks were motion corrected and dose weighted 437 using MotionCor2 62 . Contrast Transfer Function (CTF) of the motion corrected micrographs was 438 estimated using CTFFIND4 63 . Blob picker and then template picker were used to pick 440,494 439 particles and 2D classification in cryoSPARC was performed. Dimeric particles were separated 440 from monomeric by inspection of the 2D-class averages and for each sub-population an Ab initio 441 model was generated using cryoSPARC applying C2 and C1 symmetry, respectively. For each 442 model homogenous refinement was performed leading to a nominal resolution of 3.7 Å for the 443 dimer and 3.0 Å for the monomer. Particles (dimer: 69,144 monomer: 123,746) were converted 444 into a Star file format 64 and were imported into RELION 3.1.beta 65,66 . Particles were re-extracted 445 (un-binned) and processed in RELION using a box size of 700 pixel and 500 pixel for the dimer 446 and monomer, respectively. 3D Refinement followed by 3D classification was performed imposing 447 C2 symmetry for the dimer and C1 for the monomer. A subset of high quality particles was selected 448 for the dimer and monomer and subjected to 3D refinement which resulted in a resolution of 3.3 Å 449 for the dimer and 2.9 Å for the monomer. CTF refinement 67,68 followed by 3D refinement and 450 Bayesian polishing followed by another round of CTF refinement was performed for the dimer as 451 well as for the monomer. A final 3D refinement resulted in an overall resolution of 2.97 Å for the 452 dimer and 2.31 Å for the monomer. The resolution of the dimer could be further improved to 2.74 453 Å by using signal subtraction of one monomer followed by symmetry expansion and 3D refinement 454 applying C1 symmetry. 455 In order to increase the number of particles for classification on the Pc region, Dataset 2 was 456 collected from the same dimer band, but with a pixel size of 0.51 Å. The dataset was processed 457 with cryoSPARC 3.1.0 61 . After template picking 864,399 particles were extracted. With a small 458 subset of the extracted particles an Ab initio reconstruction was generated followed by 459 Heterogeneous Refinement using 5 classes, one of which contained Ab initio reconstruction as 460 reference. The class containing the PSI monomer was then subjected to Homogeneous Refinement 461 in cryoSPARC 3.1.0 resulting in a reconstruction at 3.88 Å resolution. The particles were then 462 exported to RELION 65 and 3D classification was performed. The class that contained 88,219 good 463 particles was used for further refinement which improved the overall resolution to 3.5 Å. Applying 464 CTF-refinement and Bayesian Polishing resulted in further improvement, and the final nominal 465 resolution is 2.68 Å. The data was then merged with the monomer (Dataset 1) and signal subtraction 466 followed by focused classification using a mask around the Pc region was performed. A class with 467 66,080 particles showed the best density for Pc which was used for model building. Initially, the available model of the PSI structure (PDB ID: 6JO5) of C. reinhardtii was rigid body 480 fitted into the 2.74 Å map of the symmetry expanded dimer using Chimera v 1.14 70 . Model 481 building and real-space refinement was then carried out using Coot v9. For plastocyanin, a model was generated using SWISS model 71 . The model was then rigid body 504 fitted using Chimera. Rotamers were corrected for the residues that were allowed due to the better 505 local densities. Self-restraints in Coot were activated followed by flexible fitting into the density. 506 All models were refined using Real-Space-Refine from the PHENIX suite 72 using the Grade server 507 restraint files for the ligands and a distance .edit file which was generated by Ready-set in PHENIX. 508 Further, hydrogen atoms were added for refinement to the model using Ready-set. The refinement 509 protocol was optimized using different weight parameters. The refinement statistics are shown in 510 Supplementary  Quantitative mass spectrometry analysis of the PSI-LHC monomer fraction. iBAQ values are 806 normalized to the PSI core subunit PsaB and values below 21 are excluded as they represent already 807 low intensity values, which might not be reliable. PSI-LHCI subunits PsaE and Lhca3 were not 808 detected at quantifiable intensities in this experiment. Detection of these small PSI and Lhca 809 subunits via mass spectrometry depends on the presence of only a few proteotypic tryptic peptides, 810 which could be missed, as measurements were done in a data dependent fashion (12 MS 2 per MS 1 ). 811 subunits PsaE, PsaK and Lhca3 were not detected at quantifiable intensities in this experiment. 823 Detection of these small PSI and Lhca subunits via mass spectrometry depends on the presence of 824 only a few proteotypic tryptic peptides, which could be missed, as measurements were done in a 825 data dependent fashion (12 MS 2 per MS 1 ). 826 827 836 Extended Data Fig. 9: Multi-body refinement analysis. a, Eigenvectors that explain the 837 variability of the data. The three major eigenvectors account for ~78% of the motion in the PSI 838 dimer. b-d, Stromal and side view of the model, showing the maximal motion along the three 839 vectors. e, f, The distance between the chlorophylls is plotted based on the relative motion from 840 the multi-body analysis. The y-axis shows the counts of particles that exhibit a certain distance. 841 with the corresponding density. Local resolution and map levels are indicated. 867