Crystal structure of cyanobacterial photosystem II at 3.2 Å resolution: a closer look at the Mn-cluster

In the crystal structure of photosystem II (PSII) from the cyanobacterium Thermosynechococcus elongatus at 3.2 Å resolution, several loop regions of the principal protein subunits are now defined that were not interpretable previously at 3.8 Å resolution.

The head groups and side chains of the organic cofactors of the electron transfer chain and of antenna chlorophyll a (Chl a) have been modeled, coordinating and hydrogen bonding amino acids identified and the nature of the binding pockets derived.

The orientations of these cofactors resemble those of the reaction center from anoxygenic purple bacteria, but differences in hydrogen bonding and protein environment modulate their properties and provide the unique high redox potential (1.17 V) of the primary donor.

Coordinating amino acids of manganese cluster, redox-active TyrZ and non-haem Fe2+ have been determined, and an all-trans β-carotene connects cytochrome b-559, ChlZ and primary electron donor (coordinates are available under PDB-code 1W5C).


Photosystem II (PSII) is located in the thylakoid membrane of higher plants, algae and cyanobacteria and captures sun light with two antenna proteins CP47 and CP43, which transfer the excitation energy to the photochemical reaction center with primary electron donor P680 formed by chlorophyll a (Chla) molecule(s).

The excited primary donor is oxidized to P680+˙, and the released electron travels along the electron transfer chain (ETC) featuring two pairs of Chla, one pair of pheophytin a (Pheoa) and two plastoquinones (QA and QB).

P680+˙ is re-reduced via redox-active tyrosine TyrZ by an electron from a Mn-cluster that catalyzes the oxidation of water to atmospheric oxygen.

After two cycles, doubly reduced and protonated QB is released as plastoquinol QBH2 into the plastoquinone pool in the thylakoid membrane and oxidized at the membrane-intrinsic cytochrome b6f complex to QB.

The electrons are transferred at the lumenal side of the membrane to soluble cytochrome c6 or plastocyanin and conveyed to photosystem I. Here the electrons are donated at the cytoplasmic (stromal) side to soluble ferredoxin that brings them to ferredoxin-NADP+ reductase where the reducing equivalents finally convert NADP+ to NADPH.

This whole process is associated with formation of a proton gradient across the membrane that drives synthesis of ATP.

Together with NADPH, it serves to reduce CO2 to carbohydrates.1

The structures of PSII subcore and PSII core complexes have been elucidated by electron diffraction2 and electron microscopy,3 respectively.

The crystal structure of oxygen-evolving PSII core from the thermophilic cyanobacterium Thermosynechococcus elongatus has been determined at 3.8 Å and 3.5 Å, respectively4,5 and that of Thermosynechococcus vulcanus at 3.7 Å resolution.6

The structures show that PSII occurs as homodimer with the monomers related by a non-crystallographic twofold axis (pseudo-C2).

Experimental procedures

PSII from T. elongatus was purified and crystallized as described.4

X-ray diffraction data were collected at ESRF beamline ID14-2 (λ = 0.933 Å) at 100 K and anomalous diffraction data at the Fe-edge (λ = 1.7330 Å) and beyond the Mn-edge (λ = 1.910 Å) were collected at the MPG/GBF beamline BW6/DESY at 100 K. Fine-tuning of crystallization conditions and an improved cooling protocol yielded low mosaic crystals diffracting to 2.9 Å maximum resolution.

However, due to strong anisotropic diffraction the final dataset had to be truncated beyond 3.2 Å resolution.

Data were processed with DENZO/SCALEPACK,7 the orthorhombic space group is P212121 and cell parameters are a = 127.5, b = 224.6, c = 305.6 Å.

Electron density maps calculated from experimental phases4 combined with phases derived from partial models were improved by solvent flattening, histogram matching and non-crystallographic symmetry averaging (CCP4 suite),8 and interpreted with O.9

The atomic model of PSII contains ca. 38 000 atoms corresponding to nearly 70% of the expected total number of PSII atoms.

Refinement of atomic coordinates against maximum-likelihood crystallographic target with CNS10 converged at R/Rfree of 0.36/0.42.

Despite relatively good quality of electron density maps, reliable sequence assignment was only possible in regions with specific sequence landmarks (bulky amino acid side chains and coordination of cofactors).

Therefore, amino acid side chains of fragments of the structure where sequence assignment was ambiguous were modeled as poly-Ala or Cα-trace, resulting in a final structure containing ca. 36 000 atoms.

All figures were drawn using DINO.11

See Table 1 for diffraction data.

Results and discussion

Here we describe the structure of Thermosynechococcus elongatus PSII at 3.2 Å resolution.

Of the at least 14 membrane-intrinsic protein subunits in the monomer,12 the six largest were identified: the reaction center proteins D1 and D2, core antennae CP43 and CP47, and α- and β-chains of cytochrome (cyt) b-559.

The remaining 12 transmembrane α-helices (TMH) belong to smaller subunits, some of which were tentatively assigned in previous structures based on biochemical data.4–6

In contrast to (5,6) no assignment of them was done in our present model, as the resolution does not allow a reliable tracing of side chains in these regions.

Also, the comparison of other structures 5,6 with our electron density calculated at 3.2 Å resolution does not support their assignments.

The membrane-extrinsic subunits PsbO, PsbU and PsbV (cyt c-550) were localized on the lumenal side of PSII (Fig. 1).

The core antenna subunits harbour 16 Chla in CP47 and 13 in CP43 that are arranged in two layers near the cytoplasmic and lumenal sides of the membrane.

All of them are in positions similar to the ones described by Fereirra et al.5

One additional Chla in CP43 shown in (5) is not present in our electron density maps.

However, one can not exclude that PSII contains several more Chla molecules, which are not visible due to limitations of current experimental data.

The β-carotene located in PSII connects cyt b-559 and ChlzD2 through ChlD2 with the ETC, so that secondary electron donation to the primary donor is possible.

This Car is found in a position similar to Car 48 reported in (5) but differs from the two Car given in (6).

The presence of the other six Car molecules shown in (5), although possible, is not supported by our X-ray diffraction data.

Chla and Pheoa in the ETC

In the Chla pair P680, Chla PD1 is coordinated by D1-His198Nε and PD2 by D2-His197Nε; the distance between the two Mg2+ ions is 8.3 Å (instead of the previously reported 10 Å)4,6 similar to the distance in (5,13).

The interplanar separation of the chlorin heterocycles is 3.6 Å (Fig. 2).

The environment of PD1 and PD2 is mainly hydrophobic with no evidence for direct H-bonding to protein.

D2-Trp191, shown to be important for the high redox potential of P680+˙,14 stacks with D2-His197 and is in van der Waals contact with PD2.

ChlD1 and ChlD2 are found in positions equivalent to the two accessory bacteriochlorophyll b in purple bacteria reaction center (PbRC).

In contrast to PbRC, they are not coordinated by histidines,15 but probably by water molecules H-bonded to D1-Thr179Oγ and D2-Val175O, respectively.

The binding pockets of ChlD1 and ChlD2 are both hydrophobic with two notable differences: first, in the ChlD1 binding pocket there are three Met residues but none in the ChlD2 pocket and second, there is no direct H-bond in the ChlD1 pocket but one in the ChlD2 pocket formed between D1-Gln199NεH and the carboxymethyl carbonyl oxygen of ChlD2.

PheoD1 is located between TMH c of D1 and d of D2, while PheoD2 between TMH c of D2 and d of D1.

Both Pheoa are in van der Waals contact with non-polar side chains.

PheoD1 forms three H-bonds: between D1-Gln130NεH and the keto oxygen; D1-Tyr126OηH and carboxymethyl carbonyl oxygen; D1-Tyr147OηH and carboxyphytyl carbonyl oxygen.

Conversely, PheoD2 is not involved in any direct H-bond to the protein matrix.

Although the arrangement and orientation of the cofactors constituting the two branches of the ETC are similar in PSII and PbRC, there are some notable differences.

The Mg2+⋯Mg2+ distance of PD1/PD2 in PSII, 8.3 Å, is only 0.7 Å longer than that of the two bacteriochlorophyll b of P960 (7.6 Å),15 in good agreement with the structures published at lower resolution,5,13 suggesting that P680 could likewise act as a special Chla pair.

However, according to the multimer model, the excited singlet state 1P680* is rather delocalized over all four Chla (PD1, PD2, ChlD1, ChlD2), and possibly could include the two Pheoa and the arrangement found in the structure does not contradict such model for the excited state.

This would allow these cofactors to act as a collective state, thereby promoting fast and efficient charge separation.16,17

The formed cation radical P680+˙ is localized on a monomeric Chla18 that was identified as PD1 by spectroscopic studies on D1-His198 mutants.19

An increased distance between PD1 and PD2 also reduces electronic coupling, which in turn probably contributes to the higher midpoint redox potential for the primary donor in PSII compared to PbRC (1.17 V in P680+˙/P680 versus 0.52 V in P960+˙/P960),20 finally allowing electron abstraction from TyrZ and the Mn-cluster.

The environment of PD1/PD2 is more hydrophobic than that of P960, where three H-bonds have been identified.15

In PSII one H-bond is found to ChlD2, but none to the accessory bacteriochlorophyll b in PbRC.

These differences are likely to influence the redox potential of P680.

Environment of the Mn-cluster

The cluster of four Mn cations located at the lumenal side of D1 is surrounded mostly by residues from this subunit and by some (not modeled) side chains from CP43 (Fig. 3A).

Its electron density contoured at the 3.5 σ level is Y-shaped with dimensions 5 × 7 × 3 Å (Fig. 3B).

This changes at the 5.5 σ level to two distinct peaks, one larger and the other smaller harboring probably 3 and 1 Mn cations, respectively.

Our X-ray diffraction data measured at and beyond the Mn-edge confirm the presence of Ca2+ near the larger density peak of Mn, making interpretation in terms of a Mn–Ca-heteronuclear cluster21 possible.

From this data the most probable of the four ions, which might be Ca2+, is Mn57 (Fig. 3).

Assuming that all four cations are Mn would be in agreement with the 3 + 1 model structures derived from EPR and EXAFS data.22

In contrast to Ferreira et al.,5 our electron density of the cluster accommodates only four metal ions.

Such a model of the oxygen evolving centre (OEC) might be incomplete and explained by radiation damage occurring upon long exposure of crystals to X-rays during data collection.

The radiation, especially at low energies around the Mn-edge, is likely to promote intrinsic disorder or even loss of some metal ions.

The two models of OEC agree in terms of having two moieties: one smaller containing one cation and one larger containing three or more cations.

The exact positions and identity of the cations are certainly affected by insufficient quality of data and different approaches in the interpretation of electron density.

Obviously, higher resolution (<2.9 Å) data are necessary to complete a reliable model of the OEC, as the Mn-Mn distance is about 2.9 Å as derived from EXAFS.23

The side chains of Asp170, Glu189, His332 and Glu333 of D1 are close enough for direct coordination of Mn cations (Fig. 3C and D), in agreement with mutation studies.24,25

Additional coordination could be provided by residues located in the loop aab and in the C-terminal region of D1, which are not clearly defined in the electron density.

According to biochemical data the likely ligands would be His337, Asp342, Ala.34424,25

Most of the protein ligands are in agreement to the recent 3.5 Å structure.5

Glu354 of CP43, proposed to coordinate Mn in (5) is close to the cluster but its side chain is not visible in our electron density.

Mn-cluster and phenolic TyrZOηH are bridged by the carboxylate of D1–Glu189 (Fig. 3C).

TyrZOηH is separated by 13 Å from the central Mg2+ of PD1 (closest edge-to-edge distance 6.9 Å).

Extended H-bonding of TyrZOηH (Fig. 3C and D) supports the proposal that an H-bond network facilitates deprotonation of the TyrZOηH group.26,27

The symmetry related TyrD (D2-Tyr160) is found in a hydrophobic environment with two H-bonds to D2–His189 and D2–Gln164, as predicted by spectroscopy28,29.

A word of caution

All thus far presented crystal structures of PSII from T. elongatus4,5 and T. vulcanus6 must be considered with caution.

The reason is that at resolutions in the range 3.2 to 3.8 Å the interpretation of electron density maps is only safe for polypeptide main chains whereas side chains can not be assigned unambiguously except if they are coordinated to ligands (Chla or Fe2+) that could serve as landmarks, and this holds even for main chains in loop segments.

The best defined sections in the electron density are membrane-intrinsic parts of the main subunits D1, D2, CP43, CP47 and cyt b-559.

The electron densities of the small subunits located at the periphery of PSII are less well defined and consequently difficult to interpret, especially since these subunits do not bind marker groups, and the size of their side chains is more uniform compared to the large subunits.

With this caveat we find that many sections of the 3.5 and 3.7 Å structures5,6 may have been over-interpreted especially with respect to sequence assignment in loop segments and small subunits.