Letter pubs.acs.org/JPCL
Electrostatic Asymmetry in the Reaction Center of Photosystem II Frank Müh,* Melanie Plöckinger, and Thomas Renger Institute of Theoretical Physics, Department of Theoretical Biophysics, Johannes Kepler University Linz, Altenberger Strasse 69, AT-4040 Linz, Austria S Supporting Information *
ABSTRACT: The exciton Hamiltonian of the chlorophyll (Chl) and pheophytin (Pheo) pigments in the reaction center (RC) of photosystem II is computed based on recent crystal structures by using the Poisson−Boltzmann/quantum-chemical method. Computed site energies largely confirm a previous model inferred from fits of optical spectra, in which ChlD1 has the lowest site energy, while that of PheoD1 is higher than that of PheoD2. The latter assignment has been challenged recently under reference to mutagenesis experiments. We argue that these data are not in contradiction to our results. We conclude that ChlD1 is the primary electron donor in both isolated RCs and intact core complexes at least at cryogenic temperatures. The main source of asymmetry in site energies is the charge distribution in the protein. Because many small contributions from various structural elements have to be taken into account, it can be assumed that this asymmetry was established in evolution by global optimization of the RC protein.
P
hotosystem II (PSII) is a key pigment−protein−complex (PPC) in oxygenic photosynthesis as it is the location of biological water oxidation.1−7 (For a list of abbreviations, see the Supporting Information S11.) The photosystem II core complex (PSIIcc) consists of at least 20 protein subunits (i.e., 19 to 20 subunits are visible in the more recent crystal structures of cyanobacterial PSIIcc8−12) and nearly 100 cofactors.1 From an enzymological point of view, PSIIcc is a light-dependent H2O:plastoquinone oxidoreductase that takes electrons from water2,4,5 and transfers them in a photontriggered process2,13 to plastoquinone7 under net proton transfer from the cytoplasmic/stromal side of the thylakoid membrane to the lumenal side. At the heart of PSIIcc is the reaction center (RC) containing two branches of cofactors arranged in an approximate C2 symmetry (Figure 1) and harbored by the two large protein subunits PsbA (D1-protein) and PsbD (D2-protein). The symmetry is apparent from the pairs of chlorophyll (Chl) a pigments PD1/PD2 (also referred to as “special pair”14) and ChlD1/ChlD2 (sometimes called “accessory” Chls) as well as the two pheophytin (Pheo) a pigments PheoD1/PheoD2 and the two plastoquinone cofactors QA/QB. In addition, there are two peripheral Chls termed ChlzD1/ChlzD2, which do not belong to the RC but are bonded to the D1−D2−complex. In terms of function, ChlzD1 and ChlzD2 are part of the core light-harvesting antennae CP43 and CP47, respectively.15 The symmetry of the RC is broken at the functional level; that is, electrons are transferred from the donor side, where the water-oxidizing complex (WOC) is situated, to the plastoquinones at the acceptor side only via the D1-branch of pigments (the “active” branch). Solar energy is absorbed by antenna pigments3 and funneled to the RC, where the energy is used to drive a charge-separation −• process,6,13 ultimately leading to the state P+• D1PheoD1. This symmetry break is related to the fact that the two © 2017 American Chemical Society
Figure 1. Arrangement of Chl a (PD1, PD2, green; ChlD1, ChlD2, orange; ChlzD1, ChlzD2, yellow), Pheo a (PheoD1, PheoD2, blue), and plastoquinone (QA, QB, cyan) cofactors as well as the water oxidizing complex (WOC, Mn4CaO5 cluster) and the nonheme iron (Fe) with the bicarbonate ligand (BCT) in the RC of PSII based on PDB 3WU2.10 The numbers in parentheses refer to the site energies (in nm) assigned to the RC pigments in the present work. Figure made with VMD.20
plastoquinones serve different purposes: Whereas QA is a one-electron transmitter taking an electron from Pheo−• D1 and transferring it to QB via the nonheme iron (Fe),16 QB is the substrate of the reductase part of PSIIcc and is doubly reduced and protonated before leaving the RC and being replaced by fresh plastoquinone.1,7,9 Another symmetry break is caused by the WOC, a Mn4CaO5 cluster,5 which is located close to PD1 to transfer the electrons extracted from water via a redox-active Received: December 2, 2016 Accepted: February 2, 2017 Published: February 2, 2017 850
DOI: 10.1021/acs.jpclett.6b02823 J. Phys. Chem. Lett. 2017, 8, 850−858
Letter
The Journal of Physical Chemistry Letters tyrosine4 (not shown in Figure 1) to PD1+•. The location of the WOC is unlikely to be the main cause of asymmetry, however. Reaction centers of purple bacteria (bRC) have the same architecture, albeit lacking the WOC, and the same type of acceptor side. Yet electron transfer goes only down the active branch, as is required for proper quinone reduction, and it is extremely difficult to force electron transfer down the inactive branch by introducing site-specific mutations in the protein matrix.17 Ever since the first crystal structures of bRC have revealed the structural symmetry decades ago,18,19 it has been an open question of how this structure can promote an asymmetric ET. In the present work, we show by structurebased simulations for the RC of PSII that an asymmetry already exists at the level of the excited states (exciton states) of the RC pigments, which will ultimately lead to charge separation only in the D1-branch, at least at cryogenic temperatures. To understand the exciton states of the RC, it is necessary to know the interactions of the RC pigments (Chls and Pheos) with each other and with the protein environment. The pigment−protein interaction is responsible for shifting the S0 → S1 (QY) transition energies of individual chlorin pigments in their binding sites in the PPC (site energies, Em) with respect to the transition energy in an organic solvent. As we shall show, these site energy shifts are different in the D1 and D2 branches. Besides, the interaction with the protein also couples the local QY transitions to low-frequency vibrational modes that determine optical band shapes and rates of excitation energy transfer within the RC.21,22 It is not possible to unravel individual optical bands in the RC spectra because of homogeneous and inhomogeneous broadening. However, even if this was possible, these bands could not, in general, be assigned to individual pigments. This important fact is related to pigment−pigment interactions.21,23 Because of these excitonic couplings (Vmn), which originate from the Coulomb interaction of electrons in different pigments, the excitation of one pigment causes the excitation of nearby pigments with a certain probability, so that the exciton states are, in general, delocalized. More precisely, the eigenstates |M⟩ of the exciton Hamiltonian Hex =
A recent attempt to compute site energies from crystal structure data employing a combination of extensive molecular dynamics (MD) simulations with semiempirical electronic structure methods35 confirmed that ChlD1 has the lowest site energy and PheoD2 a lower site energy than PheoD1 (see Table S9), but the site energies are overall too low by ∼2000 cm−1 (∼100 nm), which might be due to the quantum-chemical method used. In addition, it was found that site energies critically depend on the relaxation of the crystal structure by MD simulations. In general, it is, however, not clear, how well a pigment geometry obtained with a classical force field meets the requirement of an accurate quantum-chemical calculation of transition energies.36 To obtain reasonable structure-based exciton Hamiltonians in a cost-effective way avoiding the above geometry-mismatch problem, we have developed methods that combine a quantum-chemical computation of pigment charge and transition densities in vacuo with an electrostatic modeling of pigment−protein interactions in atomic detail.21,23 In particular, the method to compute site energy shifts37 is based on a numerical solution of the linearized Poisson− Boltzmann equation (LPBE)38 and therefore was dubbed Poisson−Boltzmann/quantum-chemical (PBQC) method.39 This approach worked well for the FMO protein of green sulfur bacteria37 and was also applied to the antenna system of PSII.40−43 Here we apply it for the first time to the RC of PSII. In the PBQC method, pigments and protein are modeled as sets of atomic partial charges (APCs) situated in a dielectric medium at atom positions inferred from crystal structure data. (For details, see the Supporting Information and a recent review.21) Each pigment is represented by two APC sets, one describing the pigment’s charge distribution in the electronic ground state and another in the first excited state. The site energy Em of a pigment in site m is computed from the interaction of these APCs with the environment in the PPC taking into account the Coulomb interaction with APCs on atoms in the environment as well as with the reaction potential that the pigment’s APCs induce in the polarizable medium. The computation is performed by solving the LPBE numerically employing well-established finite difference methods.38,44,45 A prerequisite for an application of this method is knowledge of APCs for each atom in the PPC. We determined the APCs for the ground and excited states of Chl a and Pheo a (Table S3) by fitting of quantum chemically computed electrostatic potentials, while the remaining APCs are taken from the CHARMM force field46 supplemented by literature data for other cofactors (see Supporting Information S1). A problem is that there are titratable groups in the protein, that is, side chains that can release a proton and hence change their charge state. Therefore, application of the PBQC method requires also knowing the protonation states of titratable groups, that is, their apparent pKa values in the protein. The latter differ from the pKa values of the respective molecular group in an aqueous solution due to interactions with the protein matrix. Similar to site energy shifts, pKa shifts can be computed by solving the LPBE and assigning different sets of APCs to protonated and deprotonated forms of the titratable groups.21,38,47 To obtain the protonation probability for each group in thermal equilibrium at a given temperature T and pH, the average over a canonical ensemble of protonation patterns (i.e., sets of protonation states assigned to individual groups) has to be determined taking into account electrostatic interactions between titratable groups. This average is computed by using a Monte Carlo (MC) method47,48 with importance sampling
∑ Em|m⟩⟨m| + ∑ Vmn|m⟩⟨n| m
m,n
(1)
are linear combinations |M⟩ = Σmc(M) m |m⟩ of the local excited states |m⟩ with coefficients c(M) m . In state |m⟩, only pigment m is in the S1 state, and all other pigments are in the S0 state. It is important to note that the exciton states |M⟩ and not the local excited states |m⟩ give rise to the bands observed in optical spectra. A difficulty is that the site energies and excitonic couplings cannot be determined experimentally. In the early multimer models,24−27 all site energies were assumed to be equal with slight variations25,26 (see Table S9), and the excitonic couplings were estimated based on the homology to bRC. The first high-resolution crystal structure8 of PSIIcc gave rise to a more precise determination of excitonic couplings. On the basis of these refined couplings, site energies were fitted by comparing the resulting spectra with experimental data.15,28−34 Although there seems to be a general consensus now that ChlD1 has the lowest site energy and that the special pair Chls PD1 and PD2 have the highest site energies, the values for the two pheophytins are still controversially discussed. Whereas in some analyses PheoD1 is red-shifted with respect to PheoD2,28,32 both pheophytins are assigned the same site energy in other works,30 or PheoD1 is slightly blueshifted33 or clearly blueshifted.15,29,31 851
DOI: 10.1021/acs.jpclett.6b02823 J. Phys. Chem. Lett. 2017, 8, 850−858
Letter
The Journal of Physical Chemistry Letters based on the Metropolis criterion.49,50 The site energy of pigment m is finally obtained as Em = ⟨Em′ (σ )⟩σ + E0
As shown below, the structure-based site energies indeed allow for a description of experimental data with only minor refinements. Deleting low-molecular-mass subunits from the structural model has only a marginal effect on site energies. Whereas PsbJ and PsbY have essentially no influence, PsbI and PsbX slightly perturb ChlzD1 and ChlzD2, respectively, as expected from their location close to these pigments (Table S5). The data summarized in Table S5 refer to pH 6.0, but we found no significant pH dependence of site energies, which is due to either fixed protonation states or a weak electrostatic interaction of the pigments with titratable groups. Deletion of the metal ions and μ-oxo bridges of the WOC has no influence on site energies, except for a slight red shift of the site energy of ChlD1. We note that the charge state of the WOC is still a matter of debate, and a more detailed theoretical account requires an advanced quantum-chemical approach that is beyond the scope of the present work. We used a very simple model, in which an APC of +3.5 is assigned to each Mn-ion, +2 to the Ca-ion, and −2 to each bridging oxygen, resulting in an overall neutral WOC together with the negative charges of the deprotonated amino acid side chains ligating the metal ions. As a consequence of removing the WOC, some of these ligands become protonated, so that the loss of positive charge in the WOC region is partly compensated. To estimate effects of net charge changes in the WOC region due to advances in the catalytic cycle, we placed two additional positive elementary charges on each one of the Mn ions but found only marginal site energy changes (
