Structural insights into photosystem II assembly

Abstract

Biogenesis of photosystem II (PSII), nature’s water-splitting catalyst, is assisted by auxiliary proteins that form transient complexes with PSII components to facilitate stepwise assembly events. Using cryo-electron microscopy, we solved the structure of such a PSII assembly intermediate from Thermosynechococcus elongatus at 2.94 Å resolution. It contains three assembly factors (Psb27, Psb28 and Psb34) and provides detailed insights into their molecular function. Binding of Psb28 induces large conformational changes at the PSII acceptor side, which distort the binding pocket of the mobile quinone (Q_B) and replace the bicarbonate ligand of non-haem iron with glutamate, a structural motif found in reaction centres of non-oxygenic photosynthetic bacteria. These results reveal mechanisms that protect PSII from damage during biogenesis until water splitting is activated. Our structure further demonstrates how the PSII active site is prepared for the incorporation of the Mn₄CaO₅ cluster, which performs the unique water-splitting reaction.

Photosystem II (PSII) is the only enzyme that catalyses the light-driven oxidation of water, a thermodynamically demanding reaction that drives photosynthesis, sustaining life on our planet^1–3. This multisubunit membrane protein complex is located in the thylakoid membranes of cyanobacteria, algae and plants. PSII strips electrons from water and injects them into the photosynthetic electron transport chain. It forms a homodimer with a molecular mass of ~500 kDa (ref. ⁴), with each monomer composed of at least 20 protein subunits and numerous cofactors, including chlorophylls, quinones, carotenoids, lipids, bicarbonate and the unique Mn₄CaO₅ cluster^5–7. The two core proteins, D1 and D2, form a central, membrane-intrinsic heterodimer, which binds all important redox cofactors involved in internal electron transfer⁸. Light excitation leads to a charge-separated state in which an electron is transferred from the reaction centre chlorophylls P₆₈₀ (ref. ⁹) to the nearby pheophytin¹⁰. Subsequently, the electron is passed to the bound plastoquinone (Q_A) and then to the mobile plastoquinone molecule (Q_B), which leaves the complex after accepting two electrons and two protons¹¹. The electron hole at P₆₈₀ is filled by oxidation of an adjacent tyrosine residue (Tyr_Z)¹² and finally by the oxygen-evolving complex (OEC), which contains the Mn₄CaO₅ cluster. In cyanobacteria, the cluster is shielded on the luminal side by the three extrinsic proteins, PsbO, PsbU and PsbV, which regulate access to the OEC by forming an intricate network of channels for different substrates and products¹³. Light energy is collected and funnelled towards P₆₈₀ by the two membrane-intrinsic antenna proteins CP43 and CP47. These proteins bind most of the chlorophyll molecules and are located on opposite sides of the D1/D2 heterodimer¹⁴. Moreover, at least 12 small transmembrane subunits with one or two transmembrane helices have been identified in PSII¹⁵, including cytochrome b₅₅₉ (ref. ¹⁶).

Structural and spectroscopic investigations have revealed these aforementioned comprehensive insights into PSII function^17–21 but we are still far from understanding PSII biogenesis in molecular detail. How nature facilitates the assembly of a multisubunit, multicofactor membrane protein complex is a fundamental, unsolved question. The biogenesis of PSII is even more challenging, as the mature complex performs sophisticated and extreme redox chemistry to catalyse the light-driven oxidation of water. This can easily lead to the formation of reactive oxygen species (for example, singlet oxygen is produced by triplet chlorophyll in the PSII reaction centre) and subsequent loss of function due to damaged proteins and cofactors^22,23. Biogenesis intermediates with only partially functional fragments of the redox chain are particularly prone to damage, thus demanding specialized protection mechanisms for the assembly process²⁴. Therefore, PSII biogenesis is not a spontaneous process but rather must be tightly regulated by the action of assembly factors. Thus far, >20 auxiliary proteins have been identified that guide the stepwise assembly of PSII subunits and cofactors via intermediate modules, which are assembled independently and then joined together to produce mature PSII^25–27. In cyanobacteria, PSII biogenesis begins with the formation of the D1/D2 heterodimer reaction centre (RC) complex from the D1 precursor protein (pD1) and the D2 protein. This is assisted by the PSII assembly factor Ycf48 after partial processing of the D1 C-terminal extension by the D1-specific peptidase CtpA^28,29. In the next step, the assembly factor Psb28 helps CP47 join the RC complex to form the RC47 complex, in which the intermediate iD1 is further processed into its mature form by CtpA^30,31. Almost all ligands of the Mn₄CaO₅ cluster are already present at this stage, except for those provided by CP43, which comes preconstructed with assembly factor Psb27 and several small subunits (together called the CP43 module)³². Psb28 is released as CP43 binds and the resulting Psb27-PSII monomer is activated by maturation of the OEC and the binding of the extrinsic proteins PsbO, PsbU and PsbV^33–35. OEC assembly is a multistep process that requires a functional upstream redox chain for the oxidation of Mn²⁺ to build up the cluster’s µ-oxo bridges between the manganese atoms^36,37. The mechanistic and structural details of this photoactivation process are not yet understood. In the consensus ‘two quantum model’^38–40, a single Mn²⁺ ion bound to the high-affinity site (HAS)⁴¹ is oxidized to Mn³⁺. This initiating, light-dependent step is followed by a slow, light-independent phase and further fast, light-dependent steps in which the remaining Mn²⁺ ions are oxidized and incorporated. Understanding the light-independent slow phase is key to unravelling the mechanism of photoactivation. Finally, PSII biogenesis completes with dimerization of two fully assembled monomers and attachment of the soluble phycobilisome antenna complexes. Interestingly, deletion of psbJ, which encodes a small single transmembrane helix protein at the entrance of the PSII plastoquinone channel, decreases PSII-mediated oxygen evolution and increases the lifetime of the reduced primary acceptor (Q_A⁻) in cyanobacterial and tobacco chloroplasts⁴². These effects are caused by a massive accumulation of an intermediate monomeric PSII complex, which contains both assembly factors Psb27 and Psb28 (ref. ⁴³). Physiological studies of Psb27- and Psb28-deletion strains point towards multifaceted functions. Cyanobacterial mutants lacking Psb28 exhibited slower autotrophic growth, particularly under stress conditions^31,44 and limited synthesis of chlorophyll-binding proteins but without decreased PSII functionality³¹. The Psb28 mutant also exhibited an overall increase in PSII repair and faster recovery from photodamage³¹. Chemical cross-linking combined with mass spectrometry revealed that Psb28 binds to the cytosolic side of CP47, close to cytochrome b₅₅₉ and the Q_B binding site. On the basis of this, researchers postulated a protective role for Psb28, where it blocks electron transport to the acceptor side of PSII to shield the RC47 complex from excess photodamage during the assembly process⁴⁵. This hypothesis is strengthened by the observation that Psb28 is also found in PSII repair complexes⁴⁶. The luminal PSII assembly factor Psb27 has been similarly well investigated. This lipoprotein is predominantly associated with inactive PSII fractions involved in assembly or repair^{32,34,46–50}, stabilizing the CP43 luminal domain and presumably facilitating the assembly of the OEC.

Our current knowledge of PSII biogenesis mainly describes the order of events and protein composition of each intermediate, as well as the general roles of PSII assembly factors. However, the precise molecular functions of these intermediate complexes and the involved assembly factors are still elusive due to their low abundance and intrinsic instability. High-resolution structural information is of vital importance to gain a deeper understanding of the molecular action of PSII assembly factors, as they are proposed to alter the structures of their associated PSII proteins to provide protection or facilitate specific biogenesis transitions.

Here, we use cryo-electron microscopy (cryo-EM) single-particle analysis to describe the molecular structure of a PSII assembly intermediate. This structure represents one of the key transitions in PSII biogenesis: the attachment of the CP43 module to the pre-assembled RC47 complex, which precedes incorporation and activation of the Mn₄CaO₅ cluster. We complement this structural data with spectroscopic analysis, revealing detailed insights into the molecular mechanisms of PSII assembly. Our study provides mechanistic answers to three long-standing questions: (1) How do assembly factors modulate the structures of PSII subunits to assist biogenesis? (2) How is PSII protected from photodamage during assembly? (3) How is the PSII active site prepared for incorporation of the Mn₄CaO₅ cluster?

Results

Stable PSII intermediates were purified from the Thermosynechococcus elongatus ΔpsbJ mutant⁴³ by affinity chromatography using a Twin-Strep-tag fused to the C terminus of the CP43 subunit and subsequent ion exchange chromatography (IEC) (Supplementary Fig. 1a). The main peak of the IEC profile corresponds primarily to monomeric PSII, which lacks the extrinsic subunits PsbO, PsbU and PsbV that are indicative for water-splitting activity (Supplementary Fig. 1b,c). Single-particle cryo-EM analysis of this PSII fraction resulted in three different high-resolution maps that allowed model building with high confidence and excellent statistics (Supplementary Fig. 2 and Supplementary Table 1). In addition to the protein subunits, we also assigned all essential non-protein cofactors, including chlorophylls, quinones, carotenoids and lipids, which are also present in the mature PSII complex (Supplementary Fig. 3). Consistent with previous biochemical studies^34,43,48, the EM density corresponding to the fully assembled, active Mn₄CaO₅ cluster is missing in the purified biogenesis intermediates. The first cryo-EM map (2.94 Å), which we call PSII-I (for PSII intermediate), provides a snapshot of the attachment of the CP43 module to the pre-assembled RC47 complex (Fig. 1). This PSII-I contains three assembly factors (Psb27, Psb28 and Psb34), as well as almost all the membrane-intrinsic subunits and cofactors found in mature PSII, although CP43 is bound in an immature conformation. Psb27 and Psb28 are well-known assembly factors^{31,32,35,43,51}, whereas the additional single transmembrane helix protein (tsl0063), which we named Psb34, has not been described before. The small subunit PsbY, which is known to be loosely bound⁵², is not resolved in our structure. In addition, PsbJ is not present, as the corresponding gene was inactivated to stall PSII assembly at this specific transition⁴³.

Fig. 1 |. Cryo-EM map of a PSII assembly intermediate (PSII-I) from T. elongatus, segmented into subunits.

a, Fifteen PSII subunits and three assembly factors are coloured and labelled (PSII subunits—D1, D2, CP43, CP47, PsbE, PsbF, PsbH, PsbI, PsbK, PsbL, PsbM, PsbT, PsbX, PsbZ and Psb30; assembly factors—Psb27, Psb28 and tsl0063, which we named Psb34) (front view). b, Parts of PSII that originate from the CP43 module (comprised of CP43, Psb27, PsbZ, Psb30 and PsbK) and the RC47 complex are indicated by dashed lines (back view). c, Schematic model of the PSII assembly process starting with the formation of PSII-I from the CP43 module and RC47. Small PSII subunits were omitted for simplicity.

The two additional maps serve as internal controls. PSII-I′ (2.76 Å) lacks Psb27 but is otherwise comparable to PSII-I; the root mean square deviation of the C_α atomic positions between similar subunits of the two complexes is 0.4 Å. Probably, Psb27 was partly lost during sample preparation. The third cryo-EM map (2.82 Å), which we call PSII-M (for PSII monomer), represents a monomeric PSII complex without bound assembly factors. Comparison of our PSII-M structure with a crystal structure of monomeric PSII⁵² (PDB-ID 3KZI, 3.6 Å) reveals only minimal differences between both structures, with a C_α root mean square deviation of 1.3 Å, which verifies that the structural changes observed in PSII-I are not caused by the deletion of PsbJ.

Psb34 binds to CP47 and is involved in the RC47 transition.

Our PSII-I structure provides identification of the single transmembrane helix protein Psb34 bound to a PSII assembly intermediate (Fig. 2a), which we also confirmed by mass spectrometry (Fig. 2b). Psb34 was probably overlooked previously due to its hydrophobicity and small size. It has a single transmembrane helix that binds to the CP47 antenna protein in close proximity to PsbH (Fig. 2a). Its conserved long N-terminal arm is located at the side and top of the D2 subunit (Fig. 2a). In addition, we independently confirmed the interaction of Psb34 with PSII assembly intermediates by isolation of Strep-tagged Psb34 complexes from otherwise unmodified cells, indicating a specific function of Psb34 in the attachment of CP43 to RC47 (Fig. 2c). In this case, two distinct PSII intermediates were isolated via pulldown of Strep-tagged Psb34: the RC47 complex with bound Psb28 and the subsequent PSII-I after attachment of CP43 and Psb27 (Fig. 2c). This observation implies that Psb28 is usually released from the PSII-I after attachment of CP43, probably after incorporation of PsbJ, as this trigger is missing in the analysed ΔpsbJ mutant. PSII-I, which only accumulates in the ΔpsbJ mutant, seems to represent a hybrid state of both PSII intermediates that might be stabilized by the incomplete binding of CP43. Psb34 shows sequence similarity to high-light inducible proteins (HLIPs), which play a role in transient chlorophyll storage and chlorophyll biosynthesis⁵³. However, the chlorophyll-binding motive is missing in Psb34 (Supplementary Table 2), suggesting a distinct function for this protein in PSII biogenesis.

Fig. 2 |. Psb34 binds to RC47 during attachment of the CP43 module.

a, Binding site of Psb34 at CP47, close to PsbH (top view), with extended binding of the Psb34 N terminus along the cytoplasmic PSII surface (dashed box). b, Matrix-assisted laser desorption/ionization time-of-flight mass spectrometry (MALDI-TOF MS) analysis of PSII assembly intermediates. Mass spectrum of Psb34 (tsl0063) from the PSII complex (inset) and the fragment spectrum obtained for m/z 5,936.356 with annotated b- and y-ion series matching the Psb34 sequence. Observed fragmentation sites are indicated by dashes in the sequence. Mascot score: 171. c, Subunit composition of Psb34-PSII assembly intermediates analysed by two-dimensional polyacrylamide gel electrophoresis (PAGE). The results are representative of three independent Psb34-TS preparations. a.u., arbitrary units.

Psb28 induces conformational changes in the D1 D-E loop.

Psb28 binds on the cytosolic faces of the D1 and D2 subunits, directly above the Q_B binding site (Fig. 3a), which differs from the position that was previously predicted by mass spectrometry⁴⁵. Its binding induces the formation of an extended β-hairpin structure that incorporates the central antiparallel β-sheet of Psb28, the C terminus of CP47 and the D1 D-E loop⁵⁴ (Fig. 3a). Binding of Psb28 to the C terminus of CP47 also imparts a directionality to the assembly process. In the Psb28-free complex (PSII-M), the CP47 C terminus blocks the Psb28 binding site by interacting with the D1 D-E loop, thus preventing the reverse process and perturbation of active PSII by Psb28. Using nuclear magnetic resonance (NMR) spectroscopy, we performed chemical shift perturbation (CSP) experiments with recombinant Psb28 and a synthetic peptide of the conserved CP47 C terminus to characterize this interaction in detail and determine the dissociation constant (K_D) (Fig. 3 and Supplementary Fig. 4). The CSP measurements indicated notable shifts with a chemical shift difference (∆δ) of >1 s.d. located at strands β3 and β4 as well at the C-terminal region of Psb28 (Fig. 3c,d). On peptide binding, resonances for several residues gradually appeared with increasing peptide concentration, which were line-broadened beyond detection for the free form of Psb28. This observation indicates a less dynamic and more rigid complex structure. This is further supported by the heteronuclear Overhauser effect (NOE) data, which show that the C terminus of Psb28 becomes rigid from L108 to K112 on CP47 peptide binding due to creation of an intermolecular β-sheet (Fig. 3e). Two-dimensional lineshape analysis was performed, yielding a K_D of 53.92 ± 0.41 µM and a dissociation rate k_off of 10.14 ± 0.16 s⁻¹, which is consistent with the observed slow exchange in the NMR spectra (Fig. 3b). The affinity of Psb28 for full-length CP47 and PSII might indeed be even higher due to additional contacts between Psb28 and the D-E loop of D1 (Fig. 3a).

Fig. 3 |. The role of the CP47 C terminus in binding of Psb28.

a, Binding of Psb28 at the cytoplasmic/stromal PSII surface (side view, colours correspond to Fig. 1) and continuation of the central Psb28 β-sheet by the CP47 C terminus and the D-E loop of D1 (dashed box). For comparison, mature monomeric PSII (PDB ID 3KZI) is shown in grey. b, Superimposed two-dimensional ¹H-¹⁵N-HSQC spectra of free Psb28 (blue) and Psb28 bound to the C-terminal peptide of CP47 (magenta). Upper left inset: representation of slow exchange behaviour for the proton amide resonance of T24, ranging from 126.9 to 128.6 ppm in the ¹⁵N dimension. c, CSPs of >1 s.d. projected onto the model representation of Psb28. d, Weighted ¹H/¹⁵N chemical shift perturbations observed for Psb28 on binding to the CP47 peptide. Red line indicates 1 s.d.; residues that yield resonances only in the complex form are indicated in orange. e, Backbone ¹⁵N {¹H}-heteronuclear NOE of free Psb28 (blue) and Psb28 bound to the C-terminal region of the CP47 peptide (magenta). I/I₀ data are presented as mean values ± s.d. derived from n = 3 independent experiments. Smaller I/I₀ ratios correspond to regions that exhibit dynamics on the pico- to nanosecond timescale.

Psb28 prevents full association of CP43 and distorts the QB-site.

Binding of Psb28—with support of Psb34—causes major structural perturbations at the PSII acceptor side (Supplementary Videos 1 and 2), which mainly involve the D-E loops of the central D1 and D2 subunits. Comparison of the CP43 structure in PSII-I with that in our PsbJ-free control PSII-M (Fig. 4a–d) or with that in mature monomeric PSII (PDB-ID 3KZI) (Fig. 4c,d) reveals several differences. The CP43 C terminus is not resolved in PSII-I, probably due to an immature position of the last transmembrane helix of CP43 and an altered conformation of the D1 D-E loop, which may prevent binding of the CP43 C terminus to the cytoplasmic PSII surface (Fig. 4b). This region is close to the loop between helices D and E of the D2 subunit, which is also altered by binding of Psb28, as clearly shown by movement of D2 Arg233 (Fig. 4b and Supplementary Fig. 5a,b). After dissociation of Psb28, the CP43 module undergoes a rigid body rotation where it clicks into place (Fig. 4b–d and Supplementary Video 1), whereas binding of PsbJ and the extrinsic proteins PsbO, PsbV and PsbU during further maturation has very little influence on the CP43-binding position (Fig. 4c,d). The part of PSII that originates from RC47 shows almost no difference between PSII-I and mature PSII (Fig. 4d), except for PsbE, which binds adjacent to PsbJ (Fig. 4c).

Fig. 4 |. Structural changes of the D1 and D2 D-E loops induced by binding of Psb28 and Psb34.

a, Side view of the CP43 antenna protein in PSII-I (teal) and the PSII-M control (light blue). b, Structural changes between PSII-I and the PSII-M control in the cytoplasmic D2 D-E loop (yellow, PSII-I; blue, PSII-M) and attachment of CP43 (teal, PSII-I; light blue, PSII-M control) (top view). Details of the structural changes in the D2 loop are shown in Supplementary Fig. 5a,b. c,d, Side view (c) and top view (d) of the PSII-I structure (orange) compared to the PSII-M control (light blue) and mature monomeric PSII (light red, PDB ID 3KZI). e,f, Side view (e) and top view (f) of the Psb28-induced structural changes in the D1 D-E loop (orange) and perturbation of the Q_B binding site compared to PSII-M (light blue), which lacks the assembly factors. Q_A is shown in yellow (PSII-I) or light blue (PSII-M), respectively. See Supplementary Fig. 5c–h for enlarged views of the Q_A and Q_B binding sites and the adjacent non-haem iron.

Most importantly, the structural changes in the D1 D-E loop may have a direct functional impact on PSII electron transfer (Fig. 4e,f), as this region coordinates several important PSII cofactors. In functional PSII, after charge separation at the RC P₆₈₀, electrons are transferred via pheophytin to the bound plastoquinone (Q_A) and further to mobile plastoquinone (Q_B). In our PSII-I structure, the Q_A site is fully assembled and a well-resolved Q_A molecule is bound (Fig. 4e,f and Supplementary Fig. 5c,d). The nearby non-haem iron is also already in place in PSII-I (Fig. 4e,f and Supplementary Fig. 5e,f). The Q_B binding site of the PSII-M control is comparable to mature PSII, although it is not occupied by Q_B in our preparation (Supplementary Fig. 5g). In contrast, the Q_B binding site of PSII-I is immature due to the Psb28- and Psb34-induced structural changes in the D1 D-E loop (Fig. 4e,f and Supplementary Fig. 5h). Notably, D1 Phe265, which coordinates the head group of Q_B in mature PSII, is clearly at a different position²¹ (Supplementary Video 2).

Binding of Psb28 protects PSII during biogenesis.

A more detailed analysis of the structural environment close to the Q_A/Q_B binding sites revealed differences in the coordination and the hydrogen-bond network of the adjacent non-haem iron, which also indicate functional consequences for PSII electron transfer and charge recombination processes. In mature PSII, the non-haem iron is coordinated by four histidine residues and bicarbonate as the fifth ligand (Fig. 5a,c), whereas in PSII-I, the bicarbonate molecule is replaced by the E241 side chain of D2 (Fig. 5b,d, Supplementary Fig. 5e,f and Supplementary Video 3). Other residues, including D1 E244 and Y246, which bind to the bicarbonate molecule in mature PSII (Fig. 5a), are also displaced in PSII-I due to the conformational change of the D1 D-E loop (Fig. 5b, Supplementary Fig. 5e,f and Supplementary Video 3). Binding of bicarbonate is important for PSII efficiency⁵⁵, as it lowers the redox potential of (Q_A/Q_A⁻) to favour forward electron transport^56,57. If charge recombination occurs, the lower redox potential favours indirect charge recombination via P₆₈₀^•+/Pheo^•−. This back reaction yields triplet chlorophyll and subsequently singlet oxygen⁵⁷, a highly reactive oxygen species. Changes in the redox potential of (Q_A/Q_A⁻) have been proposed to tune the efficiency of PSII depending on the availability of CO₂ as the final electron acceptor and thereby protect PSII under low CO₂ conditions⁵⁷. Therefore, we used flash-induced variable fluorescence to monitor the stability of Q_A⁻ in active PSII, in the PSII-I assembly intermediate and in inactivated, Mn-depleted PSII, both of which lack a functional OEC (Fig. 5e and Supplementary Fig. 6a,b). The results indicate that the unique fluorescence traces observed with PSII-I are a consequence of Q_A/Q_A⁻ redox tuning by an absent OEC, an immature Q_B binding pocket and probably the replacement of bicarbonate by glutamate as a ligand of the non-haem iron (Supplementary Discussion). To determine the functional consequences of this redox tuning in PSII-I, we measured the formation of ¹O₂ by electron paramagnetic resonance (EPR) spectroscopy using the spin probe TEMPD. The data clearly show that ¹O₂ formation is reduced by ~30% in PSII-I compared to inactivated PSII (Fig. 5f).

Fig. 5 |. Binding of Psb28 displaces bicarbonate as a ligand of the non-haem iron and protects PSII from damage.

a, The electron transfer from Q_A to Q_B is coordinated by the non-haem iron (Fe²⁺), with the binding of bicarbonate (Bic) serving as a regulatory mechanism⁵⁷ in mature PSII (PDB ID 3WU2). b, Binding of Psb28 to the PSII-I assembly intermediate induces a conformational change in the cytoplasmic D2 D-E loop, where the side chain of Glu241 replaces bicarbonate as a ligand of the non-haem iron. The respective fits of the non-haem iron binding sites are shown in (Supplementary Fig. 5e,f). A similar coordination is also found in non-oxygenic bacterial reaction centres⁷⁰ (Supplementary Fig. 6c). c, Electron transfer (purple arrows) in mature PSII. Light-induced charge separation at the RC chlorophylls (P_D1, P_D2, Chl_D1 and Chl_D2) leads to electron transfer via pheophytin (Pheo_D1) and plastoquinone A (Q_A) towards Q_B. The electron gap at the RC is filled by the OEC. d, Reoxidation of Q_A⁻ by direct and safe charge recombination is favoured in the PSII assembly intermediate, as indicated by the purple arrows. e, Flash-induced fluorescence decay of PSII. Blue lines represent active PSII and red lines correspond to PSII-I. Black and grey lines represent PSII control samples without a functional OEC (Apo-PSII, hydroxylamine-treated PSII; PSII (-OUV), extrinsic proteins removed). Full lines represent mean values calculated from three independent experiments (n = 3), with dotted corridors depicting s.d. f, The protective role of Psb28 binding was further confirmed by electron paramagnetic resonance (EPR) spectroscopy using the spin probe TEMPD, which is specific for ¹O₂, the major reactive oxygen species in PSII generated by triplet chlorophyll (³P).

Psb27 binds to loop E of CP43 at the luminal PSII surface.

Psb27 binds to the luminal side of the PSII complex, adjacent to loop E of the CP43 subunit (Fig. 6a,b). We could successfully model the Psb27 N terminus until Cys22, which is the first residue after N-terminal processing of Psb27 (ref. ³⁴). Psb27-Cys22 is located close to the membrane plane but the lipid moiety is not well enough resolved to build a model. In contrast to previously proposed models^58,59, the binding site of Psb27 has little overlap with the binding sites of the extrinsic subunits (PsbO, PsbV and PsbU) and has at least no direct contact to the Mn₄CaO₅ cluster binding site, particularly not to the D1 C terminus (Fig. 6a,b), which has been suggested previously^58,60. Instead, Psb27 is bound at a remote position that might be occupied by CyanoQ in the mature complex⁶¹. This position of Psb27 may still prevent proper binding of PsbO³⁴ by lowering its binding affinity due to the partial overlap of the binding sites, although both proteins can bind together, as PsbO-containing Psb27-PSII complexes have been isolated via His-tagged Psb27 (ref. ⁴⁹). Additionally, Psb27 might stabilize loop E of CP43 in the unassembled state and facilitate its binding to the D1 subunit for the assembly of the Mn₄CaO₅ cluster⁶². This is of particular importance, as loop E of CP43 provides Arg345 and Glu342 (Arg357 and Glu354 in previous publications, see caption of Fig. 6 for details) to position the Mn₄CaO₅ cluster in mature PSII (Fig. 6b, dashed box). Glu342 serves as a ligand, whereas Arg345 stabilizes the Mn₄CaO₅ cluster by coordinating Asp170 of D1 through a hydrogen bond. Moreover, in the Psb27-bound state (PSII-I), the D1 C terminus, which is directly involved in coordination of the Mn₄CaO₅ cluster²¹, is bound away from the cluster (Fig. 6c and Supplementary Fig. 7), which might be an additional consequence of Psb27-induced stabilization of the CP43 E loop. Thus, our PSII-I structure reveals not only how the Psb27 protein binds to CP43 and thus stabilizes it but also indicates a role for the maturation of the oxygen-evolving complex that is consistent with functional data from previous studies^32,35,48,62.

Fig. 6 |. The role of Psb27 in Mn₄CaO₅ cluster assembly.

a, Bottom view of the luminal PSII surface for PSII-I (orange), the PSII-M control (light blue) and mature monomeric PSII (PDB ID 3KZI) (light red). b, Side view of CP43 (teal) and Psb27 (brown) in PSII-I, as well as of CP43 (blue) and PsbO (purple) in mature monomeric PSII (PDB ID 3KZI). Dashed box: CP43 E loop with residues Arg345 and Glu342 (shown as sticks), which form the second coordination sphere of the Mn₄CaO₅ cluster. We changed the numbering of CP43 residues due to a corrected N-terminal sequence (www.UniProt.org). The residues correspond to Arg357 and Glu354 in previous publications. The high-resolution structure of the Mn₄CaO₅ cluster is taken from Umena et al.²¹ (PDB ID 3WU2). c, Position of the D1 C terminus in PSII-I (orange) and mature monomeric PSII (PDB ID 3KZI) (light red).

PSII-I contains a single cation at the immature OEC site.

The unique Mn₄CaO₅ cluster is a key feature of PSII that splits water into oxygen and protons. However, our PSII-I complex does not show any oxygen-evolving activity, suggesting that the OEC is not fully assembled. In mature PSII, the Mn₄CaO₅ cluster is submerged in the complex and additionally capped by the extrinsic subunits PsbO, PsbU and PsbV (Fig. 6a,b). In our PSII-I structure, these subunits are absent, which leaves two parts of the CP43 E loop (residues 320–327 and 397–404) in a flexible conformation, exposing the binding site of the Mn₄CaO₅ cluster to the lumen. There is no strong density feature at this position that would correspond to the fully assembled metal-redox cofactor. Thus, our PSII-I structure provides a model for an immature OEC. Comparing our structure with the high-resolution crystal structure of mature PSII²¹ (PDB-ID 3WU2) provides insights into the first steps of OEC biogenesis (Fig. 7).

Fig. 7 |. Conformational changes within the active site of the Mn₄CaO₅ cluster. The Mn₄CaO₅ cluster performs the unique water-splitting reaction of PSII.

a, The active site of the Mn₄CaO₅ cluster is resolved within our PSII-I structural model but is not yet oxygen evolving. b, Crystal structure of the oxygen evolving, mature PSII (PDB ID 3WU2, resolution 1.9 Å). c, Overlay of both structures, illustrating notable differences in the backbone conformation of the D1 and D2 C-terminal tails. d, Accompanying side-chain rearrangements of the D1 C-terminus. e–g, The Cl⁻ (e), ionⁿ⁺ (f) and Mn₄CaO₅ (g) cluster coordination partners are compared in detail. The validation of the fit to the density of structural details shown here is provided in Supplementary Fig. 7.

The D1 C terminus is an essential component for formation of the of OEC, as it provides several charged residues that are responsible for coordination of Cl⁻ and the Mn₄CaO₅ cluster (Fig. 7a,b,d). The density for these C-terminal residues is weak in our PSII-I map but traceable (Supplementary Fig. 7a), indicating a flexibility that confirms the absence of the OEC. Compared to the mature complex, the last 12 residues of the C-terminal tail of D1 would need to undergo large conformational changes to bring the side chains of Glu333, His337, Asp342 and the Ala344 C terminus into the correct position to coordinate the Mn₄CaO₅ cluster (Supplementary Video 4).

Moreover, we identify a clearly visible density at the position of Cl⁻, which is coordinated by Lys317 (D2) and the hydrogen atom of the backbone nitrogen of Glu333 (D1) in mature PSII (Fig. 7b,e). Despite the similar position, the Cl⁻ is coordinated by the nitrogen atom of the ring of adjacent His332 (D1) in PSII-I (Fig. 7a,e and Supplementary Fig. 7d). Surprisingly, we identified another density in the area where the Mn₄CaO₅ cluster is located in mature PSII (Fig. 7a–c,f and Supplementary Fig. 7c). However, this density is not large enough to reflect the whole cluster. Based on its size and interaction partners (Fig. 7f), it corresponds to one positively charged ion. In the structural context, this ion is probably Mn²⁺ but it could also be Ca²⁺ or any other positively charged ion.

This ion is coordinated by the side chains of D1 Asp170, Glu189 and His332, which are already in similar positions compared to mature PSII. Glu342 and Arg345 (Glu354 and Arg357 in previous publications) of CP43, which are both involved in the second coordination sphere of the Mn₄CaO₅ cluster, are also already prepositioned through the interaction between Arg345 with D1 Asp170 (Fig. 7g). However, there are still substantial conformational changes necessary for the transition from PSII-I to mature PSII, as highlighted in Fig. 7d,g, as well as in Supplementary Video 4. The D1 C-terminal tail must bring the side chains of Glu333, His337 and Asp342, as well as the C terminus of Ala344, into correct alignment to coordinate the Mn₄CaO₅ cluster. In addition, the C-terminal tail of D2 needs to flip towards the D1 C terminus (Fig. 7c, Supplementary Fig. 7b and Supplementary Video 4). In summary, PSII-I is characterized by only one positively charged ion bound instead of the complete Mn₄CaO₅ cluster, resulting in different conformations of the D1 and D2 C-termini compared to the structural model containing a mature Mn₄CaO₅ cluster. However, the PSII-I structure seems to be prepared to accept the Mn₄CaO₅ cluster, as indicated by the similarities in side-chain positioning described above.

Discussion

PSII biogenesis is a complex process that requires the action of specific assembly factors. These auxiliary proteins are not present in the mature complex and interact only transiently with specific subunits or pre-assembled PSII intermediates. Although >20 factors have been identified and assigned to specific transitions, their precise molecular functions in PSII assembly remain elusive in almost all cases. Our study provides detailed molecular insights into the function of PSII assembly factors Psb27, Psb28 and Psb34, which are involved in an important transition before activation of the OEC. The determined binding positions of Psb27 and Psb28, which are two of the most-studied PSII assembly factors, define the binding with atomic resolution, in contrast to previous models^{58,59,63–67}. This provides precise information about the molecular function of Psb27 and Psb28 in the assembly process.

Binding of Psb28 and Psb34 to the cytoplasmic side of PSII induces large conformational changes in the D1 D-E loop (Fig. 4), which has been identified previously as an important location for PSII photoinhibition and D1 degradation^68,69. Structural changes observed in the PSII-I Q_B binding pocket and coordination of the non-haem iron suggest a functional impact on PSII electron transfer to protect the immature complex until water splitting is activated. In particular, D2 Glu241 replacing bicarbonate as a ligand of non-haem iron suggests a regulatory role, as binding of bicarbonate was proposed to tune PSII efficiency by changing the redox potential of (Q_A/Q_A−)^56,57. As a functional consequence, PSII-I generates less singlet oxygen compared to inactivated PSII (Fig. 5f).

Interestingly, the coordination of the non-haem iron in PSII-I resembles that in non-oxygenic bacterial reaction centres (BRCs)⁷⁰ (Supplementary Fig. 6c). In BRCs, the fifth ligand of the non-haem iron is provided by E234 of the M subunit⁷¹ and mutagenesis of this residue induces changes in the free energy gap between the P•+/Q_A•− radical pair⁷². These findings indicate that the environment of the non-haem iron is important for the Q_A redox potential, which influences charge recombination and the formation of reactive oxygen species⁵⁶.

The Psb27-bound and -unbound structures do not differ substantially (Fig. 6b), suggesting a rather subtle action in PSII biogenesis. Previous work demonstrated that Psb27 is already bound to free CP43 (ref. ³²), where it might protect free CP43 from degradation or stabilize the E loop in a specific conformation to chaperone the subsequent association with the RC47 complex. This step is crucial for preparing the binding site of the Mn₄CaO₅ cluster, as the CP43 E loop is involved in its second coordination sphere⁷³. Moreover, a recent high-speed atomic force microscopy study with spinach PSII has shown fluctuations of the C43 E loop at room temperature after removal of the extrinsic subunits and destruction of the Mn₄CaO₅ cluster⁷⁴, which might suggest a role of Psb27 in restraining the scope of conformational fluctuations for further assembly of the OEC. Consistent with this model, recent kinetic analysis of PSII from Synechocystis sp. PCC 6803 revealed that Psb27 facilitates photoactivation, probably by stabilization of intermediates with increased accessibly for ions, especially Ca²⁺, to the site of cluster assembly⁶². However, our data indicate a self-organization of the CP43 E loop, at least in T. elongatus. The corresponding structural models for this region are almost identical, independent of binding of Psb27 or the extrinsic proteins, whereas the positions of the D1 and D2 C-termini are clearly different between intermediate and mature PSII. Binding of PsbO seems to be the main driver for these structural differences and its presence interferes with efficient photoactivation, probably by closing the cluster binding site and thus preventing the exchange of ions⁶². The overlap of the Psb27 and PsbO binding sites, although small, might decrease the PsbO affinity and prevent its premature binding, as initially suggested^34,37 or Psb27 may promote binding of PsbO in a different conformation (both proteins can bind simultaneously in a stable complex⁴⁹) that might support photoactivation.

Previous structural studies aimed to obtain mechanistic insights into the dark rearrangement during photoactivation by removing the Mn₄CaO₅ cluster from fully assembled PSII, either by depleting it directly from PSII crystals by chemical treatment⁷⁵ or by cryo-EM single-particle analysis in a manganese- and calcium-free buffer⁷⁶. The X-ray structure was indeed missing the Mn₄CaO₅ cluster but the D1 C terminus followed mostly the same trajectory as found in the mature PSII-dimer structure. The authors suggested that the D1 Cterminus might not rearrange during Mn₄CaO₅ cluster assembly. However, the crystal structure was dimeric and still had the extrinsic subunits PsbO, PsbU and PsbV bound. It is known that these subunits are typically not associated with Mn₄CaO₅ cluster-depleted PSII. Thus, the structure might be artificially stabilized by crystal packing forces. The cryo-EM structure, on the other hand, revealed a monomeric PSII that lacks extrinsic subunits and the Mn₄CaO₅ cluster⁷⁶. This structure is like our PSII biogenesis intermediate PSII-I, as PsbY, PsbZ and PsbJ are also missing. The PsbJ subunit is surprising; it is an integral subunit of PSII and should not be easily detached, yet it is missing from this structure, as we deleted it to stabilize our PSII-I complex. These observations might indicate a more specific and regulatory role of PsbJ in PSII biogenesis. Additionally, the D1 C terminus is disordered in this previous cryo-EM structure and the authors suggest that the dark rearrangement involves a transition from a disordered to an ordered state.

Our structure now reveals the fate of the D1 C terminus with the assembly factor Psb27 bound. The D1 C terminus follows a different trajectory compared to the mature PSII. Thus, we provide structural evidence that the slow dark rearrangement involves a conformational change of the D1 C terminus rather than the previously proposed disorder-to-order transition after initial photoactivation⁷⁶. Compared to mature PSII, 12 residues of the D1 C-terminal tail must undergo large conformational changes to bridge the side chains of Glu333, His337 and Asp342, as well as to bring the C terminus of Ala344 to the correct position to coordinate the Mn₄CaO₅ cluster (Figs. 6c and 7d and Supplementary Video 4), which is consistent with previous models^39,77,78. We also identified a single positively charged ion in our PSII-I structure, coordinated by Asp170, Glu189 and His332 of D1 (Fig. 7f), at the position of the Mn₄CaO₅ cluster of mature PSII. This binding site probably corresponds to the long-sought single HAS, where the first Mn²⁺ binds before the first photoactivation step in OEC biogenesis⁴¹. The presence of manganese in immature PSII, due to a PratA-facilitated preloading mechanism, has been previously suggested⁷⁹. However, we cannot exclude binding of Ca²⁺, which was shown to bind with a much lower affinity^77,80, or any other positively charged ion at this position. Nevertheless, Asp170 has been identified as the most critical residue for the HAS^81,82, which supports our hypothesis. Further photoactivation steps occur presumably after cooperative binding of calcium and manganese. The binding of the extrinsic subunit PsbO, potentially after release of Psb27 and maturation of the water-oxidizing complex, is the next step of the PSII assembly line in vivo, which leads to the next unsolved question in PSII biogenesis: what triggers the release of an assembly factor? For Psb27, its detachment might be promoted by the binding of PsbO, as their binding sites partially overlap.

Membrane protein complexes play a fundamental role in bioenergetics to sustain and proliferate life on Earth. They drive the light-to-chemical energy conversion in photosynthetic organisms and are essential for energy supply in heterotrophs. These intricate molecular machines are assembled from numerous single proteins in a spatiotemporally synchronized process that is facilitated by a network of assembly factors. These auxiliary proteins are the key players of nature’s assembly lines. Our PSII-I cryo-EM structure reveals a molecular snapshot of PSII biogenesis and, accompanied by our spectroscopic and biochemical analyses, provides clear mechanistic insights into how three assembly factors (Psb27, Psb28 and Psb34) coordinate the stepwise construction of this powerful catalyst of life.

Methods

Cultivation of T. elongatus BP-1.

Cell growth and thylakoid membrane preparation were performed as described previously⁸³. In brief, T. elongatus mutant strains (∆psbJ psbC-TS and psb34-TS) were grown in BG-11 liquid medium inside a 25-l foil fermenter (Bioengineering) at 45 °C, 5% (v/v) CO₂-enriched air bubbling and 50–200 µmol photons m⁻² s⁻¹ white light illumination (depending on the cell density). Cells were harvested at an optical density OD₆₈₀ of ~2 after 5–6 d of cultivation and concentrated to ~0.5 l, using an Amicon DC10 LA hollow fibre system, pelleted (3,500 relative centrifugal force (r.c.f.), 45 min and 25 °C) and resuspended in 150 ml of Buffer D (100 mM Tris-HCl, pH 7.5, 10 mM MgCl₂, 10 mM CaCl₂, 500 mM mannitol and 20% (w/v) glycerol). The harvested cells were flash-frozen in liquid nitrogen and stored at −80 °C until further use.

Preparation of T. elongatus mutant strains.

T. elongatus ∆psbJ psbC-TS was generated on the basis of the previously described strain T. elongatus ∆psbJ⁴³ that was transformed with the plasmid pCP43-TS. The plasmid is based on pCP34–10His³⁴. The His-tag sequence was exchanged with a Twin-Strep-tag by polymerase chain reaction (PCR) using the primers CP43TS_rev (5′CCCGATATCTTACTTCTCAAATTGCGGATGAGACCACGCAGAACCACCAGAACCACCGCCGCTGCCGCCGCCTTTTTCGAACTGCGGGTGGCTCC3′) and NTCP43 (5′TGCTCTAGAATGAAAACTTTGTCTTCCCAGA3′). The resulting PCR product was ligated back into an empty pCP34–10His backbone using XbaI and EcoRV restriction endonucleases. T. elongatus BP-1 cells were transformed as described previously⁸⁴. Mutant colonies were selected by frequent replating onto agar plates with increasing antibiotic concentrations, stopping at 8 µg ml⁻¹ of chloramphenicol and 80 µg ml⁻¹ of kanamycin. Complete segregation of the mutant was confirmed by PCR with the primers CTCP43DS (5′CCGCTCGAGACCATCCAAGCTTGGCAGCA3′) and NTCP43 (5′TGCTCTAGAATGAAAACTTTGTCTTCCCAGA3′).

T. elongatus psb34-TS was generated by transformation with the plasmid pPsb34-TS. The plasmid DNA was obtained from TwistBioscience. It encodes Psb34 (tsl0063) with a C-terminal Twin-Strep-tag and a kanamycin resistance cassette, flanked by tsl0063 upstream and downstream regions (900 basepairs each). T. elongatus BP-1 cells were transformed⁸⁴ and mutant selection took place³⁴. Complete segregation of the mutant was verified by PCR. The primers used were tsl0063-up-for (5′CATATGGTCTCGCAATTATTTGCCCATGC3′) and tsl0063-down-rev (5′GGTACCCCG ACACAGTTGATCACCGC3′).

Purification of PSII assembly intermediates.

Thawed cells were diluted in 100 ml of Buffer A (100 mM Tris-HCl, pH 7.5, 10 mM MgCl₂ and 10 mM CaCl₂) and pelleted again (21,000 r.c.f., 20 min and 4 °C). The pellet was resuspended in 100 ml of Buffer A with 0.2% (w/v) lysozyme and dark incubated for 75–90 min at 37 °C. This was followed by cell disruption by Parr bomb (Parr Instruments) and pelleting (21,000 r.c.f., 20 min and 4 °C). All following steps were performed under green illumination. The pellet was resuspended in 150 ml of Buffer A and pelleted again (21,000 r.c.f., 20 min and 4 °C). This step was repeated three times, with the last resuspension in 80 ml of Buffer B (100 mM Tris-HCl, pH 7.5, 10 mM MgCl₂, 10 mM CaCl₂ and 500 mM mannitol). The isolated thylakoid membranes were flash-frozen in liquid nitrogen and stored at −80 °C.

Strep-Tactin affinity purification of PsbC-TS and Psb34-TS assembly intermediates were performed under green illumination. Membrane protein extraction was performed as described previously⁸³, with certain adaptations. Isolated thylakoid membranes were supplemented with 0.05% (w/v) n-dodecyl β-maltoside (DDM) (Glycon) and pelleted (21,000 r.c.f., 20 min and 4 °C). The sample was resuspended in extraction buffer (100 mM Tris-HCl, pH 7.5, 10 mM MgCl₂, 10 mM CaCl₂, 1.2% (w/v) DDM, 0.5% (w/v) sodium cholate and 0.01% (w/v) DNase (AppliChem)) to a final chlorophyll concentration of 1 mg ml⁻¹ and incubated for 30 min at 20 °C. The solubilized membrane proteins were ultracentrifuged (140,000 r.c.f., 60 min and 4 °C) and NaCl was added to the supernatant to a final concentration of 300 mM.

The supernatant was filtered through a 0.45-μm syringe filter (Sarsted AG) and applied to a 5-ml Strep-Tactin Superflow HC column (IBA Lifesciences), equilibrated in Buffer W (100 mM Tris-HCl, pH 7.5, 10 mM MgCl₂, 10 mM CaCl₂, 500 mM mannitol, 300 mM NaCl and 0.03% (w/v) DDM) at a flow rate of 3 ml min⁻¹. The column was washed with Buffer W until a stable baseline (A₂₈₀) was reached. Strep-tagged protein complexes were eluted by an isocratic elution with Buffer E (100 mM Tris-HCl, pH 7.5, 10 mM MgCl₂, 10 mM CaCl₂, 500 mM mannitol, 300 mM NaCl, 2.5 mM desthiobiotin and 0.03% (w/v) DDM). The captured fractions were equilibrated in Buffer F (20 mM MES, pH 6.5, 10 mM MgCl₂, 10 mM CaCl₂, 500 mM mannitol and 0.03% (w/v) DDM) with a spin concentrator (Amicon, Ultra-15, nominal molecular weight limit (NMWL) of 100,000 Da), flash-frozen in liquid nitrogen and stored at −80 °C until analysis.

PsbC-TS-containing assembly intermediates were further separated by IEC. Captured elution fractions from the Strep-Tactin affinity purification were loaded onto an anion exchange column (UNO Q6, Biorad) with a flow rate of 4 ml min⁻¹, pre-equilibrated in Buffer F. Protein complexes were eluted by a linear gradient of MgSO₄ (0–150 mM) using Buffer G (20 mM MES, pH 6.5, 10 mM MgCl₂, 10 mM CaCl₂, 500 mM mannitol, 150 mM MgSO₄ and 0.03% (w/v) DDM). Fractions containing PSII assembly intermediates were collected, concentrated to 100–10 μM RCs, using a spin concentrator (Amicon, Ultra-15, 100,000 NMWL), aliquoted, flash-frozen in liquid nitrogen and stored at −80 °C until further analysis.

Cryo-EM.

For cryo-EM sample preparation, 4.5 µl of purified protein complexes were applied to glow discharged Quantifoil 2/1 grids, blotted for 3.5 s with force 4 in a Vitrobot Mark III (Thermo Fisher) at 100% humidity and 4 °C, then plunge frozen in liquid ethane, cooled by liquid nitrogen. Cryo-EM data were acquired with a FEI Titan Krios transmission electron microscope using the SerialEM software (v.3–8-0beta8_64; ref. ⁸⁵). Video frames were recorded at a nominal magnification of ×22,500 using a K3 direct electron detector (Gatan). The total electron dose of ~55 electrons per Ų was distributed over 30 frames at a calibrated pixel size of 1.09 Å. Micrographs were recorded in a defocus range of −0.5 to −3.0 μm.

Image processing, classification and refinement.

Cryo-EM micrographs were processed on the fly using the Focus (v.1.1.0) software package⁸⁶ if they passed the selection criteria (iciness <1.05, drift 0.4 Å < x < 70 Å, defocus 0.5 μm < x < 5.5 μm, estimated CTF resolution <6 Å). Micrograph frames were aligned using MotionCor2 (v.1.1.0; ref. ⁸⁷) and the contrast transfer function (CTF) for aligned frames was determined using Gctf (v.1.06; ref. ⁸⁸). Using Gautomatch (v.056, http://www.mrc-lmb.cam.ac.uk/kzhang/), 693,297 particles were picked template-free on 824 acquired micrographs. Particles were extracted with a pixel box size of 260 using RELION 3.1 (ref. ⁸⁹) and imported into Cryosparc 2.3 (ref. ⁹⁰). After reference-free two-dimensional classification, 675,123 particles were used for ab initio construction of initial models and subjected to multiple rounds of three-dimensional classification to obtain models with and without Psb28 density. Non-uniform refinement in Cryosparc resulted in models with an estimated resolution of ~3.2 Å. Particles belonging to three-dimensional classes with and without Psb28 (150,090 and 166,411 particles, respectively) were re-extracted in RELION with a pixel box size of 256 and subjected to several rounds of CTF refinement (estimation of anisotropic magnification, fit of per-micrograph defocus and astigmatism and beam tilt estimation) and Bayesian polishing⁹¹. Both classes were refined using the previously generated starting models. Three-dimensional classification without further alignment using a mask around the Psb27 region separated particles in the Psb28-containing class into distinct classes with and without Psb27 (57,862 and 91,473 particles, respectively). Final refinement of each of the three classes (with Psb27 and Psb28 (PSII-I), with Psb28 but without Psb27 (PSII-I′) and without Psb27 and Psb28 (PSII-M)) resulted in models with global resolutions of 2.94, 2.76 and 2.82 Å, respectively (Gold standard FSC analysis of two independent half-sets at the 0.143 cutoff). Local-resolution and 3DFSC plots (Supplementary Fig. 2) were calculated using RELION and the Remote 3DFSC Processing Server web interface (v.3.0; ref. ⁹²), respectively.

Atomic model construction.

The 3.6-Å resolution X-ray structure of monomeric PSII from T. elongatus with PDB-ID 3KZI⁵² was used as initial structural model that was docked as rigid body using UCSF Chimera 1.10.2 (ref. ⁹³) into the obtained cryo-EM densities for PSII-M and PSII-I. The cofactors that had no corresponding density were removed. The subunit PsbJ was also removed, as it was deleted in the experimental design. By highlighting the still unoccupied parts of the PSII-I density map, we identified densities that led to the structures of Psb27, Psb28 and Psb34.

The 2.4-Å resolution X-ray structures of isolated Psb28 from T. elongatus with PDB-ID 3ZPN⁹⁴ and the 1.6-Å resolution X-ray structure of isolated Psb27 from T. elongatus with PDB-ID 2Y6X⁶⁶ were docked as rigid bodies into the unoccupied densities. The 1.6-Å resolution X-ray structure of CyanoQ from T. elongatus with PDB-ID 3ZSU⁹⁵ does not fit into the density and was therefore not modelled.

As there was no experimentally resolved structural model of Psb34 available, we first used the sequence with UniProt-ID Q8DMP8 to predict structures using the webserver SWISS Model⁹⁶ and LOMETS⁹⁷. We also predicted the secondary structure through the meta server Bioinformatics Toolkit⁹⁸ and CCTOP⁹⁹. The results of the secondary structure prediction are summarized in Supplementary Table 4. Combining these predictions together with the unassigned cryo-EM density, we used COOT¹⁰⁰ included in CCP4 7.0 to build an initial model of Psb34 that has one α-helix from amino acid numbers 28 to 55.

Model refinement.

The initial model of the complex described above was refined in real space against the cryo-EM density of PSII-I and structural clashes were removed using molecular dynamics flexible fitting (MDFF)¹⁰¹. MDFF simulations were prepared in VMD 1.9.4a35 (ref. ¹⁰²) using QwikMD¹⁰³ and the MDFF plugin. The simulations were carried out with NAMD 2.13 (ref. ¹⁰⁴) using the CHARMM36 force field. Secondary structure, cis peptide and chirality restraints where used during 800 steps of minimization followed by a 40-ps MDFF simulation at 300 K. Due to the restraints used, only conformational changes of side chains and subunit movements compared to the initial structure were identified during the initial MDFF run. We checked the fit to density of the structure by calculating cross-correlation values of the backbone atoms. For PSII-I, we identified residues 217 to 269 from PsbA and residues 467 to 499 from PsbB and PsbZ as main regions where the structural model was not yet in accordance with the density after the initial MDFF run. For these three regions, we used an iterative combination of MDFF with Rosetta 3.8 (refs. ^105,106). Here, we used the optimized strategy as described for model construction of the 26S proteasome^107,108.

To obtain an atomic model that fits the PSII-M density, we used the initial model based on 3KZI described above but without PsbJ, Psb27, Psb28 and Psb34. After the initial MDFF run, the cross-correlation check did not reveal any regions with substantial deviation between model and density. Therefore, no further refinement was necessary. This fast convergence reflects that there are no crucial differences between the PSII-M model and the X-ray structure 3KZI.

To obtain the atomic model that fit the PSII-I′ density, we used the final PSII-I model without Psb27 for MDFF. After the initial MDFF run, the cross-correlation check did not reveal any regions with substantial deviation between model and density. This fast convergence reflects that there are no crucial differences between the PSII-I and PSII-I′ models, except for the presence of the Psb27 subunit.

Lastly, the PSII-M, PSII-I and PSII-I′ models were used to initiate one final round of real-space refinement in Phenix 1.17.1–3660 (ref. ¹⁰⁹) including structure validation checks using MOLPROBITY and EMRinger score.

Visual inspection of structural models and density maps was done with VMD 1.9.4a35 (ref. ¹⁰²), UCSF Chimera 1.10.2 (ref. ⁹³) and PyMol 2.2.0.

Reporting Summary.

Further information on research design is available in the Nature Research Reporting Summary linked to this article.

Data availability

The cryo-EM density maps are deposited in the Electron Microscopy Data Bank under accession numbers EMD-12335 (PSII-M), EMD-12336 (PSII-I) and EMD-12337 (PSII-I′). The atomic models of the cryo-EM structures are deposited in the Worldwide Protein Data Bank under accession numbers 7NHO (PSII-M), 7NHP (PSII-I) and 7NHQ (PSII-I′). The NMR backbone assignments for Psb28 bound to the C-terminal peptide of CP47 are deposited in the Biological Magnetic Resonance Bank under accession code 50747. Protein sequences with the following accession codes were downloaded from the UniProt database (P0A444, Q8DIQ1, Q8DIF8, Q8CM25, Q8DIP0, Q8DIN9, Q8DJ43, Q8DJZ6, Q9F1K9, Q8DIN8, Q8DHA7, Q8DIQ0, Q9F1R6, Q8DJI1, Q8DHJ2, Q8DG60, Q8DLJ8, Q8DMP8). Source data are provided with this paper. Additional data supporting the findings of this manuscript are available from the corresponding authors on reasonable request.