Crystal Structure of Monomeric Photosystem II from Thermosynechococcus elongatus at 3.6-Å Resolution*

The membrane-embedded photosystem II core complex (PSIIcc) uses light energy to oxidize water in photosynthesis. Information about the spatial structure of PSIIcc obtained from x-ray crystallography was so far derived from homodimeric PSIIcc of thermophilic cyanobacteria. Here, we report the first crystallization and structural analysis of the monomeric form of PSIIcc with high oxygen evolution capacity, isolated from Thermosynechococcus elongatus. The crystals belong to the space group C2221, contain one monomer per asymmetric unit, and diffract to a resolution of 3.6 Å. The x-ray diffraction pattern of the PSIIcc-monomer crystals exhibit less anisotropy (dependence of resolution on crystal orientation) compared with crystals of dimeric PSIIcc, and the packing of the molecules within the unit cell is different. In the monomer, 19 protein subunits, 35 chlorophylls, two pheophytins, the non-heme iron, the primary plastoquinone QA, two heme groups, 11 β-carotenes, 22 lipids, seven detergent molecules, and the Mn4Ca cluster of the water oxidizing complex could be assigned analogous to the dimer. Based on the new structural information, the roles of lipids and protein subunits in dimer formation of PSIIcc are discussed. Due to the lack of non-crystallographic symmetry and the orientation of the membrane normal of PSIIcc perpendicular (∼87°) to the crystallographic b-axis, further information about the structure of the Mn4Ca cluster is expected to become available from orientation-dependent spectroscopy on this new crystal form.

The primary processes of oxygenic photosynthesis takes place in the thylakoid membranes of plants, green algae, and cyanobacteria and involve the cooperation of several proteincofactor complexes. Among these complexes, only photosystems I (PSI) 6 and II (PSII) are able to perform light-induced charge separation necessary to convert photon energy into a biochemically amenable form.
The photosystem II core complex (PSIIcc) is a light-driven water-plastoquinone-oxidoreductase, which is characterized by the unique property of abstracting electrons from water molecules, the primary electron source in oxygenic photosynthesis (1)(2)(3). Several x-ray crystal structures of homodimeric PSIIcc from thermophilic cyanobacteria have been published with resolutions ranging from 3.8 to 2.9 Å (4 -9). The monomers in the dimer are related by a non-crystallographic C2 rotation axis. Each monomer contains 17 membrane-intrinsic and three membrane-extrinsic protein subunits, the latter being located at the lumenal side of the membrane. The redox-active cofactors are harbored by the heterodimeric protein matrix formed by subunits D1 (PsbA) and D2 (PsbD) and are arranged in two pseudo-C2 symmetric branches.
This entity, referred to as reaction center (RC), consists of four chlorophyll a (Chla) molecules, two pheophytins a (Pheo D1 , Pheo D2 ), two plastoquinones (PQ) Q A and Q B with one non-heme iron located in between, two redox active tyrosines (Y Z and Y D ), and the water oxidizing complex (WOC). Whereas the tightly bound Q A acts as electron transmitter, the mobile Q B is the substrate of the quinone reductase part of PSIIcc. An additional plastoquinone molecule (Q C ) of unresolved function has been located next to the Q B binding site in the recent crystal structure of dimeric PSIIcc at 2.9-Å resolution. In the RC, light-induced charge separation takes place leading to the oxidation of P D1 to the cation radical P D1 .
ϩ . The electron is transferred via Pheo D1 and Q A to Q B . P D1 . ϩ has an exceptionally strong oxidizing power (1.25 eV (10)) and is able to abstract electrons via Y Z from the WOC, a heteronuclear Mn 4 Ca cluster located at the lumenal side of PSII. After the accumulation of four redox equivalents in the so-called S-states (S 0 , S 1 , …, S 4 ) of the WOC (11), water is oxidized to molecular oxygen.
The crystals of dimeric PSIIcc belong to the orthorhombic space group P2 1 2 1 2 1 and diffract to a maximal resolution of ϳ2.9 Å (8). Their x-ray diffraction pattern is highly anisotropic (the attained resolution depends on the orientation of the crystal in the x-ray beam), which limits the useful resolution range (12,13). This feature seems to originate from the packing of the dimeric complexes in the unit cell, but no other crystal form has been found for cyanobacterial PSIIcc so far.
In addition to dimeric PSIIcc, a monomeric form with high oxygen evolution capacity has been prepared (14 -16). It is therefore attractive to crystallize the PSIIcc monomer in attempting to achieve a more suitable crystal packing for x-ray structure analysis. In earlier work, the monomer was found to aggregate in solution (15). Therefore, it was considered to be inappropriate for crystallization. We succeeded in developing a novel method for purifying an intact monomeric PSIIcc that crystallizes in a different form. The new crystals enabled us to obtain the first x-ray structural model of the PSIIcc monomer at 3.6-Å resolution. Based on these data, we address several open questions concerning the role of protein subunits and lipids for the oligomeric state of PSIIcc and the assembly/disassembly of PSIIcc during the repair cycle of photodamaged subunit D1 (17)(18)(19). We also discuss the possibility of improving the structural analysis of the WOC by orientation-dependent spectroscopy.

EXPERIMENTAL PROCEDURES
Protein Purification-The initial purification steps of PSIIcc followed the preparation protocol published in (14) with slight modifications. After two consecutive chromatography steps, the fraction containing the PSIIcc monomer was concentrated to ϳ5 mM Chla and loaded onto a third column (diameter of 15 mm, length of 410 mm, ToyoPearl DEAE 650S, Tosoh Bioscience) pre-equilibrated with 20 mM MES-NaOH (pH 5.0), 20 mM CaCl 2 , 0.5 M betaine monohydrate, 0.02% (w/v) n-dodecyl-␤-D-maltoside (DDM). After washing at a constant salt concentration for 6 cv with a flow rate of 3 ml/min, monomeric PSIIcc was eluted in a linear salt gradient (MgSO 4 , 0 -50 mM, 8 cv). The fraction was concentrated in Amicon stirring cells using a Millipore Biomax 100 membrane (Millipore, MA). For further concentration to volumes Ͻ1 ml and to change the buffer conditions after ion exchange chromatography, the sample was washed three times with 10 mM MES-NaOH (pH 6.0), 5 mM CaCl 2 , 5 mM MgCl 2 , 0.02% (w/v) DDM using Sartorius Ultra Free 100 concentrators in a centrifuge at 3000 ϫ g at 4°C. Samples were concentrated to 3 mM Chla and either directly used for crystallization or stored in liquid nitrogen.
Crystallization-A broad crystallization screen was set up using a robot and the sitting drop vapor diffusion method and yielded conditions that were further optimized. Crystals were grown using the microbatch method by mixing the protein solution (3 mM Chla corresponding to ϳ25 mg/ml of protein) with the same volume of precipitant solution. Between 4 and 10 l of the finally obtained solution (containing 1.5 mM Chla, ϳ19% (w/v) PEG 400, 0.1 M PIPES, pH 7.0, 0.2 M CaCl 2 , 0.01% (w/v) DDM) was placed either in a 96-well plate (IMP@CT, Greiner-Bio-one, Germany) or in the middle of a Teflon tube (inner diameter 1 mm; YCM Europe) and closed with sealing tape or sealing plaster. The crystals grew in 3 to 6 days at 18°C in the dark and were directly flash-cooled in a nitrogen gas stream at 100 K after soaking with 28% (w/v) PEG 400 to provide cryoprotection.
Crystallographic Data Collection and Analysis-The data set was collected at the European Synchrotron Radiation Facility (ESRF, beam line ID 29), integrated and scaled with XDS (20). The structure of the PSIIcc monomer was resolved by the molecular replacement method with the PHASER program (21) using one monomeric part of the 2.9-Å resolution structure of homodimeric PSIIcc (8) (PDB entry 3BZ1) as search model. Model rebuilding and refinement were done using COOT (22) and the CNS 1.2 package (23), respectively. The structure was refined with the rigid-body procedure for the protein part of the complex and the annealing procedure for the cofactors. The final model of PSIIcc monomer shows R/R free factors of 0.297/0.308, with root mean square deviations from ideal geometry of 0.013 Å for bond lengths and 2.1°for bond angles.
Analytical Ion Exchange and Gel Permeation Chromatography-For analytical anion exchange chromatography, a small column (5 mm diameter, 200 mm length, Toyopearl DEAE 650 S) was used connected to an Ä kta FPLC system (Ä KTA purifier, Amersham Biosciences) with simultaneous detection at 205, 280, and 680 nm. Gel permeation chromatography experiments were conducted as described in Ref. 15, with simultaneous detection at 222, 280, and 680 nm.
Dynamic Light Scattering-Dynamic light scattering was performed using a DynaPro Titan instrument with a tunable laser diode at 833 nm wavelength (Wyatt Technology Corporation, Santa Barbara, CA). The protein was dissolved in 100 mM PIPES (pH 7.0), 10 mM CaCl 2 , 0.5 M betaine monohydrate, 0.03% (w/v) DDM and filtered through a Millex sterile filter (0.22 m pore size) into a 100-l cuvette. To obtain the diffusion coefficient D Z , autocorrelation functions of 50 measurements were averaged and analyzed using the instrumental software (Dynamics 6.9.2.9, Wyatt Technology Corporation, Santa Barbara, CA).
Spectroscopic Quantitation of Carotenes-Pigments of redissolved crystals of monomeric and dimeric PSIIcc were extracted in 80% (v/v) aqueous acetone and spectra were recorded in the wavelength region from 800 to 400 nm, normalized to 664 nm, and the absorbance difference was calculated. By using the molar extinction coefficient for Chla (76,800 M Ϫ1 cm Ϫ1 at 664 nm (24)) and ␤-carotene (144,000 M Ϫ1 cm Ϫ1 at 454 nm (25)) a difference of 1.3 Ϯ 0.2 ␤-carotene/36 Chla between monomeric and dimeric PSIIcc was calculated (assuming that the absorption of pheophytins at 664 nm is reduced by ϳ50% compared with Chla).
Mass Spectrometry and SDS-PAGE-MALDI-TOF MS analysis was conducted using an Ultraflex II Spectrometer (Bruker Daltonics, Germany) in the linear mode using sinapinic acid as matrix. SDS-PAGE was performed with a Phast System (Amersham Biosciences) using precast HD-SDS gels. Gels were run and silver stained following the protocol of the manufacturer.
Oxygen Evolution Activity Assay-Oxygen evolution of PSII samples were measured at room temperature using a home built Clark-type electrode (26). The excitation was performed either with saturating continuous white light from a tungsten lamp passed through a heat filter or with repetitive 1-Hz flashes from a xenon flash lamp. The sample was diluted to 20 -50 M Chla in a buffer containing 20 mM MES-NaOH (pH 5.0-6.0), 20 mM CaCl 2 , and 10 mM MgCl 2 . Artificial electron acceptors added were either 2 mM 2,6-dichloro-p-benzoquinone for continuous excitation or 2 mM K 3 [Fe(CN) 6 ] and 0.4 mM phenyl-pbenzoquinone for single flash excitations. The electrode was calibrated using air-saturated and nitrogen-saturated water at atmospheric pressure.

Preparation and Characterization of PSIIcc Monomer-Fol-
lowing the published preparation protocol (14), which comprises two ion exchange chromatography steps, we obtained essentially equal amounts of PSIIcc monomer and dimer (the monomer: dimer ratio of 40 preparations was 1.05 Ϯ 0.45). The fraction of monomeric PSIIcc shows high oxygen evolution activity, but contains significant amounts of other proteins (as revealed by SDS-PAGE, supplemental Fig. S1, lane 1) not found in the fraction of dimeric PSIIcc. This contamination is due to phycobilisome proteins, a small amount of monomeric PSI, and ATP synthase. The latter occurs in nearly the same amount as the PSIIcc monomer (for characterization of ATP synthase from Thermosynechococcus elongatus, see supplemental data). The detection of ATP synthase by absorption at 280 nm is difficult due to the absence of bound pigments and its low tryptophan content (ϳ0.2% assuming a complex formed by ␣ 3 ␤ 3 ␥␦⑀-abbЈc 10 (27)). However, ATP synthase could be monitored at a wavelength of 205 nm (Fig. 1), where mainly the protein backbone absorbs (28).
To improve the purity of the PSIIcc monomer preparation, we applied an additional chromatography step using a weak anion exchange matrix. As under the formerly used chromatography conditions at pH 6.0 no further separation between the PSIIcc monomer and ATP synthase could be achieved (Fig. 1), the pH of the applied buffer was varied. Whereas at pH 6.8 the separation of the compounds remains incomplete, a clear separation was obtained by decreasing the pH value to 5.0. This condition was consequently used to obtain pure PSIIcc monomer that enabled us to grow single crystals suitable for x-ray diffraction experiments.
We found no indication that exposition of monomeric or dimeric PSIIcc to pH 5.0 has a deleterious effect on the oxygen evolution capacity, in agreement with the reversible inhibition of PSIIcc at acidic pH described earlier (29). Although the oxygen evolution activity of dimeric and monomeric PSIIcc was decreased when measured at pH 5.0, it was fully restored when the buffer was changed to pH 6.0 after long term incubation (4 or 24 h) at pH 5.0. The final monomeric and dimeric PSIIcc samples, as utilized for crystallization, show similar oxygen evolution activity rates ranging from 2400 to 3800 mol O 2 (mg Chla h) Ϫ1 . Furthermore, measurements of the oxygen evolving activity per single flash using the PSIIcc monomers prepared as described above revealed the same oxygen evolution capacity of ( 1 ⁄ 4 mol O 2 /(39 -68 mol Chla ϫ flash)) as known from dimeric PSIIcc used in crystallographic studies so far (( 1 ⁄ 4 mol O 2 /(37-73 mol Chla ϫ flash)) taken from Ref. 14). Neglecting double hits and misses (30), the theoretical limit for fully active centers is 1 ⁄ 4 mol O 2 /(36 mol Chla ϫ flash) taking into account the smaller absorption of 2 pheophytin a compared with 35 Chla per center.
A combination of MALDI-TOF-MS (supplemental Fig. S2) and SDS-PAGE (supplemental Fig. S1, lane 5) revealed that the PSIIcc monomer contains the same protein subunits as the PSIIcc dimer and is of sufficient purity for protein crystallization. Mass spectrometry also confirmed that the small membrane intrinsic protein subunits of the PSIIcc monomer are N-terminally processed (supplemental Table S1) in the same way as described for the dimeric form (8). Gel permeation chromatography showed a single peak corresponding to a molecular mass of 457 Ϯ 20 kDa for monomeric PSIIcc similar to the value reported earlier (15).
To further analyze the aggregation behavior of monomeric PSIIcc, samples were investigated by using dynamic light scattering at different protein concentrations that are in a suitable range for protein crystallization. The use of a laser at 833 nm wavelength avoids disturbance of the measurements due to light absorption by protein-bound chlorophylls and therefore allows for measurements at high protein concentrations. We obtained a monodisperse particle distribution (9.1 Ϯ 1.3% polydispersity) with a hydrodynamic radius of 5.9 Ϯ 0.1 nm that remained invariant at protein concentrations between 0.8 and 6.0 mg/ml (supplemental Fig. S3). In contrast, the crude fraction of monomeric PSIIcc isolated without the third chromatography step (as in Ref. 14) forms aggregates with higher molecular mass similar to those reported in Ref. 15.
Crystallographic Analysis of PSIIcc Monomer-The green colored, diamond-shaped and plate-like crystals of the PSIIcc monomer grew to maximum dimensions of ϳ1.0 ϫ 0.6 ϫ 0.2 mm (Fig. 2). The diffraction pattern showed a diffuse background scattering as observed for crystals of PSIIcc dimer (12), but was nearly isotropic (supplemental Fig. S4A). We were able to collect and process a dataset to 3.6-Å resolution ( Table 1). The crystals belong to the orthorhombic space group C222 1 with unit cell constants a ϭ 119.89 Å, b ϭ 224.69 Å, c ϭ 337.28 Å and one PSIIcc monomer in the crystal asymmetric unit (solvent content, 61.6%). The structure was determined by molecular replacement using one monomer of the 2.9-Å resolution structure of dimeric PSIIcc (8) as search model.
Crystal Packing of PSIIcc Monomer-Although the unit cell contains eight monomeric PSIIcc complexes, its volume is only slightly (ϳ3%) larger than the unit cell of crystallized dimeric PSIIcc with four dimers. This indicates an almost equally dense molecular packing in both crystal forms. Fig. 3A shows the packing of PSIIcc monomers in the unit cell viewed along the crystallographic b-axis. Three types of crystal contacts can be distinguished (Fig. 3A, circles A, B, and ellipse C). The major-ity of the crystal contacts are provided by the membrane-extrinsic subunit PsbV and the lumenal parts of PsbB, PsbE, and PsbH. This solvent-exposed region is characterized by polar side chains forming at least 24 hydrogen bonds between monomers that are rotated by 180°against each other along the crystallographic C2 axes (Fig. 3A, ellipse C). Two further contacts are provided by the membrane extrinsic subunit PsbO (Fig. 3A,  circles A and B). The extended lumenal loop of PsbO provides two hydrogen bonds with the neighboring monomer (Fig. 3A, circle A), thereby interacting with the N-terminal loop of PsbF (PsbO-Lys 112 -PsbF-Pro 10 ) and the N-terminal short ␣-helix of PsbE (PsbO-Ser 115 -PsbE-Asp 12 ). Another hydrogen bond is found between PsbO-Thr 51 and the N terminus of PsbA (PsbA-Asn 12 , Fig. 3A, circle B). The contacts along the b-axis are not resolved so far (Fig. 3B), but there are two regions, where protein subunits of different monomers approach each other by 5-6 Å. These comprise the C termini of PsbM (PsbM-Gln 33 , Fig. 3C) and the N terminus of PsbH (PsbH-Arg 4 ) facing the loop of PsbC that connects transmembrane ␣-helices c and d.
Because the monomers are oriented with their membrane normal (corresponding to the non-crystallographic C2 axis of the PSIIcc dimer) perpendicular (ϳ87°) to the crystallographic b-axis, their membrane planes are nearly parallel to this axis (Figs. 2 and 3B). Therefore, the detergent belt may prohibit close interactions between neighboring complexes in this direction. There is enough space between the PSIIcc monomers to accommodate a detergent belt in the form of a monolayer ring of 25-Å thickness in agreement with simple geometric models (31,32). The only exception is the region close to the C terminus of PsbM, where the detergent belts of the two adjacent monomers are probably partly fused or squeezed. Interestingly, in the crystals of monomeric PSIIcc a crystallographic C2 axis is located between the PsbM subunits of two neighboring monomers leading to an arrangement, in which their dimerization surfaces are facing each other (Fig. 3C). In the following, the term dimerization surface exclusively refers to monomeric PSIIcc and corresponds to the monomer-monomer interface in dimeric PSIIcc.
Protein Subunits and Cofactors-The quality of the data allows the unambiguous assignment of the main chain folding of 19 polypeptide subunits, and no major changes compared with the dimeric PSIIcc structure could be detected. No electron density was found for the peripheral subunit PsbY, although its presence would not lead to sterical conflicts within the crystal. Nevertheless, PsbY could be detected in the majority (80%) of redissolved crystals by MALDI-TOF-MS analysis (supplemental Fig. S2 and Table S1).
Iron is electron-rich and therefore its position can be easily mapped in the electron density. The non-heme iron is situated at the pseudo-C2 symmetry axis of the RC close to the cytoplasmic side as expected. The positions of the heme groups of cytochrome b 559 and cytochrome c 550 are in accordance with the above assignment of the corresponding protein subunits PsbE/F and PsbV, respectively.
Despite the limited resolution of 3.6 Å, the good starting phases provided by the model of dimeric PSIIcc allowed us to reliably assign cofactors in the PSIIcc monomer. Large cofactors such as Chla and pheophytin molecules are well defined by  the electron density (supplemental Fig. S5A). This fact enables us to confirm the arrangement of all chlorins in the RC and of all 29 Chla molecules bound to the core antenna proteins CP43 (13 Chla) and CP47 (16 Chla). In cases where flexible phytyl chains are not stabilized by contacts with protein and/or other cofactors, their assignment is rather difficult. Therefore, the similarity with the structure of the PSIIcc dimer was used as a guide. The reliability of the assignment of each cofactor was probed with the calculation of OMIT maps, which serve to reduce possible model bias problems introduced after molecular replacement. For most of the carotenoids, the electron density is not continuous. Their localization is mainly based on the electron density of the ionone rings and the analogy to the ␤-carotene positions in the PSIIcc dimer. This resulted in the assignment of 11 ␤-carotenes. Absorption spectroscopy of redissolved crystals confirmed this number (see "Experimental Procedures"). The ␤-carotene missing in the structure of the PSIIcc monomer is located at the monomer-monomer interface in dimeric PSIIcc and therefore may detach due to its exposed localization in the monomer. The best defined ␤-carotene in the PSIIcc monomer is Car D2 close to cytochrome b 559 .
In the PSIIcc monomer, electron density is found at the Q A , Q B , and Q C sites, but can be assigned unambiguously to a PQ only for Q A (supplemental Fig. S6A) revealing a position virtually identical to Q A in the dimer (8). The electron density at the Q B site cannot be assigned to a PQ molecule, but could be modeled as a buffer molecule MES (supplemental Fig. S6B). A patch of electron density is found in the vicinity of Car D2 and the heme group of cytochrome b 559 , which corresponds to the position of the head group of Q C identified in the dimer (supplemental Fig. S6C). However, the assignment of this electron density to a specific molecule is difficult.
Lipids and Detergent Molecules-The structure of the PSIIcc dimer (8) revealed the presence of 25 integral lipid molecules per monomer (11 monogalactosyldiacylglycerol (MGDG), 7 digalactosyldiacylglycerlol (DGDG), 5 sulfoquinovosyldiacylglycerol (SQDG), and 2 phosphatidyldiacylglycerol). Using this information about lipid positions, it was possible to assign 22 lipid headgroups in the electron density map of monomeric PSIIcc (supplemental Fig. S5B). This finding is in agreement with the similar lipid composition found for monomeric and dimeric PSIIcc from T. elongatus (33).
The positions of the lipid headgroups surrounding the RC are essentially the same as those of their counterparts in the dimer (8), forming three smaller clusters (2-3 lipids) around the RC and one larger lipid cluster consisting of seven lipids (missing one MGDG (LMG218E in PDB code 3BZ1) found in dimeric PSIIcc) at the plastoquinone/plastoquinol exchange cavity. The three lipids found at the periphery of the PSIIcc dimer were also found in the monomer with slightly shifted positions of the lumenal headgroups. The angle between the membrane normal and the b-axis is 87°(see Fig. 2). C, two monomers with their dimerization surfaces oriented toward each other. The monomers are related by the crystallographic C2-symmetry axis (a-axis) located between their PsbM subunits (colored white in space-filling representation). For clarity, only protein subunits, lipids, and detergent molecules are shown; protein subunits in schematic representation are colored as follows: D1 (yellow), D2 (orange), CP43 (magenta), CP47 (red), cytochrome b 559 (PsbF (cyan) and PsbE (green)), PsbU (pink), PsbV (blue), PsbO (green), and the remaining small subunits (gray). Lipids (cyan) and detergent molecules (brown) are drawn in space filling representation. The red circle marks a lipid/ detergent cluster located at the dimerization surface.
In dimeric PSIIcc, seven pairs of lipids are located at the monomer-monomer interface due to the non-crystallographic C2 symmetry (Fig. 4B). At the dimerization surface of the PSIIcc monomer, five lipids were found: two lipids pointing to the cytoplasmic side (SQDG 1 and MGDG 2, Fig. 4A, for nomenclature, see supplemental Table S2) and three to the lumenal side (MGDG 3 and DGDG 4 and 5). Whereas the headgroups of the lipids oriented toward the cytoplasm are at the same positions as in the dimer, the headgroups of the remaining lipids are found to be slightly shifted. Furthermore, the electron density suggests that one MGDG found in the dimer (MGDG 5) is replaced by DGDG (DGDG 5) or DDM in the monomer (Fig.  4A). By analyzing the lipid positions in the PSIIcc monomer, we found 10 lipids, including DGDG 4 and DGDG 5 located at the dimerization surface, which follow the pseudo-C2 symmetry of the RC.
Seven detergent molecules (DDM) per monomer could be assigned in the PSIIcc dimer. Three of them are located at the periphery and four at the monomer-monomer interface. We resolved seven DDM molecules in the PSIIcc monomer: three at the periphery and four at the dimerization surface. Two of the three peripheral DDM molecules are at similar positions as in Ref. 8, but the headgroup of the third is shifted by about 7 Å and contributes to crystal contacts. Two of the DDM assigned at the dimerization surface (DDM 6 and 7) are nearly at the same position as described for dimeric PSIIcc. Two further DDM (DDM 8 and 9) were located at new positions in close vicinity to DGDG 4. The headgroup of the latter is rotated and forms polar contacts (Fig. 3C, red  Mn 4 Ca Cluster-In the PSIIcc monomer, electron density arising from the metal ions of the Mn 4 Ca cluster was found in the same position as reported for the PSIIcc dimer (8). In agreement with EPR measurements on the PSIIcc monomer from T. elongatus (16), we assume that the structure of the cluster is essentially the same in monomeric and dimeric PSIIcc. Due to the limited resolution, the assignment of a chloride ion in the vicinity of the Mn 4 Ca cluster was not possible.

DISCUSSION
In this study, we present the first structure of a monomeric PSIIcc with high oxygen evolution capacity. The results show that in the PSIIcc monomer 19 of the 20 subunits are arranged identically to the corresponding subunits in the dimer. Although PsbY was found to be present in most of the redissolved crystals examined by MALDI-TOF-MS, we obtained no electron density for this subunit in the actual dataset. This finding may be attributed to variant, substoichiometric occupancy  Table S2. Lipids and DDM only found in homodimer are numbered according to the PDB file 3BZ1 as in supplemental Table S2. and larger disorder of PsbY within the PSIIcc monomer. In the PSIIcc dimer the electron density for this subunit was visible at 3.8-Å resolution (4), but appears to be absent in a dataset at 3.5 Å (6). In agreement with the poorly defined electron density for this subunit even at 2.9-Å resolution in the PSIIcc dimer (8), we conclude a loose association of PsbY with PSIIcc.
The almost identical structure of monomeric and dimeric PSIIcc is also reflected by the location of the tetrapyrrole cofactors and the non-heme iron. In agreement with this finding, the presence of 11 carotenes at positions expected from dimeric PSIIcc could be confirmed. As five of the 12 carotenes in the PSIIcc dimer are located at the monomer-monomer interface it is remarkable that only one of these five molecules is missing in the monomer. This carotene is found in a bridging position across the monomer-monomer interface.
Concerning plastoquinone cofactors, the assignment of Q A is reliable due to the presence of electron density for both the headgroup and the isoprenoid tail. In contrast, it was not possible to assign a PQ in the electron density at the Q B -site. Because Q B is the substrate of the quinone reductase part of PSIIcc, it may have left the Q B -site as plastoquinol. Our model suggests that a buffer molecule can occupy the Q B -site in the absence of competing PQ. The penetration of a buffer molecule would be facilitated, if the quinone exchange cavity is more accessible from the aqueous phase. This may be the case, because one of the lipids shielding the Q B -site from the cytoplasm appears to be absent in the PSIIcc monomer. The electron density found at the Q C -site is not yet assigned to any molecule in our model, but the presence of a PQ cannot be excluded (supplemental Fig. S6C).
The present data do neither allow the detection of fatty acid tails of lipids nor a distinction between the DGDG and the maltose headgroup of detergent molecules (DDM). Therefore, the assignment in the PSIIcc monomer is tentative and based on the 2.9-Å model of dimeric PSIIcc, which is used as criterion to discriminate these molecules. Consequently, the electron density of sugar headgroups found at positions that significantly differ from those in the model of the PSIIcc dimer was assigned to DDM molecules.
Despite this limitation, the majority of the lipids and detergent molecules described in the structure of dimeric PSIIcc are found to be also present in PSIIcc monomer. Due to the noncrystallographic C2 symmetry in dimeric PSIIcc, seven pairs of lipids and four pairs of DDM are located at the monomermonomer interface. In the PSIIcc monomer, lacking this local symmetry, five lipids and four DDM are found at the dimerization surface. These molecules are located next to the D1 protein (surrounding the RC), to PsbT and PsbM, but not at the surface provided by CP47 (Fig. 4). Apparently, these lipids form a stable complex only with parts of the dimerization surface and probably exert a functional or structural role in the RC or in the repair cycle of D1 rather than being only involved in the dimerization of PSIIcc.
In the dimer, the main protein-protein contact between the monomers is provided by subunit PsbM and its counterpart PsbMЈ. Our new data revealed that PsbM is at the same position in both, PSIIcc monomer and dimer. Furthermore, after deletion of PsbM in the mesophilic cyanobacterium Synechocystis PCC 6803, PSIIcc dimers are still formed (34). This supports our earlier suggestion that direct protein-protein contacts alone are not responsible for dimer formation (8).
The present structural model gives only two hints concerning the question of why the PSIIcc monomer does not dimerize. (i) One of the two expected binding niches for SQDG at the monomer-monomer interface in the dimer (8) is unoccupied in the monomer (denoted as SQDG 532B in Fig. 4B). This would support a direct role for SQDG rather than phosphatidyldiacylglycerol (35) in dimer formation of PSIIcc. (ii) Additional detergent molecules were found at positions where they would interfere with dimer formation (labeled 8 and 9 in Fig. 4A). In aqueous solution, the dimerization may be inhibited by the detergent belt.
The D1 protein is characterized by the highest turnover rate among all PSIIcc polypeptides (19). The repair cycle most probably involves a monomerization of PSIIcc and N-terminal proteolytic degradation of D1 by FtsH proteases (17). In the structure of the PSIIcc monomer, the N-terminal part of D1 (including transmembrane ␣-helix a) is highly exposed to the membrane phase and may be more accessible for FtsH proteases than in the dimer. However, we found no structural indications of a destabilization of D1 in the PSIIcc monomer and essentially the same oxygen evolution capacity as for the PSIIcc dimer. Therefore, it is unlikely that the PSIIcc monomer used for crystallization represents a photodamaged product of dimeric PSIIcc.
The crystals of the PSIIcc monomer represent an important step towards the structural elucidation of the Mn 4 Ca cluster as they provide a new, highly ordered arrangement of PSIIcc for orientation-dependent spectroscopy. The application of nondestructive spectroscopic methods is necessary, because the exposure to x-rays at doses used in crystallography rapidly reduces the four manganese ions of the cluster and leads to structural changes (36).
Polarized extended x-ray absorption fine structure experiments on oriented PSII membranes suggest a manganese-calcium vector (manganese-calcium distance 3.4 Å) to be oriented along the membrane normal and a manganese-manganese vector (mono--oxo-bridged manganese-manganese distance 3.2 Å) along the membrane plane in the dark-adapted S 1 state (37). Further constraints for modeling the structure of the cluster were derived from extended x-ray absorption fine structure studies on single crystals of the PSIIcc dimer (38). Three possible models were proposed based on these data, but additional structures derived from DFT calculations are also in agreement with the observed dichroism (39). The remaining ambiguity is caused inter alia by the arrangement of PSII in the oriented samples.
Although the crystals of PSIIcc dimer and monomer feature the same number of monomers in the unit cell, the new crystal form offers two main advantages: (i) the lack of the non-crystallographic C2 symmetry and (ii) the orientation of the membrane normal perpendicular to the crystallographic b-axis (Fig.  2). The absence of non-crystallographic symmetry reduces the ambiguity of the data, because the possible orientations of the Mn 4 Ca cluster within the unit cell are half as much as in the dimer. Due to the orientation of the membrane normal in the new crystal form, a better discrimination is possible between absorber-backscatter vectors oriented parallel and perpendicular to the membrane plane. Therefore, it is expected that a full set of polarized extended x-ray absorption fine structure spectra along the crystal axes will significantly expand the available structural information about the Mn 4 Ca cluster, especially with respect to the manganese-calcium interaction. With these additional constraints, a selection between the currently discussed models may become feasible. These experiments are in progress. The absence of non-crystallographic symmetry in crystals of monomeric PSIIcc will also reduce the complexity of the spectra derived from electron paramagnetic resonance spectroscopy on the paramagnetic S 2 -state of the Mn 4 Ca cluster in single crystals (40,41) and thus facilitate the assignments of spectral features.
Besides the still limited information concerning the Mn 4 Ca cluster, a higher resolved overall structure of PSIIcc is indispensable to clarify many questions still open in the present structural models. Therefore, further optimization of the diffraction quality of PSIIcc crystals is crucial. To achieve this goal, the new crystal form features the important advantage of significantly lower anisotropy in the diffraction pattern compared with crystals of dimeric PSIIcc. Moreover, an improvement of the crystal structure of monomeric PSIIcc is also a prerequisite to elucidate the origin of the different oligomerization states of PSIIcc, which may help to clarify the mechanism of D1 exchange.