Closing Kok's cycle of nature's water oxidation catalysis

Photosystem II (PSII) catalyzes biological water oxidation at the Mn4CaO5 oxygen-evolving complex (OEC) via Kok’s cycle, progressing through S0–S3 and a transient S4. Recent time-resolved XFEL crystallography provided structural snapshots of the S3 and early S4 intermediates and supports O–O bond formation via coupling of an extra oxygenic ligand (Ox) on Mn1 with the central O5. Kinetic studies indicate that S3→S4 (single-electron/multi-proton transfer to a Mn(IV)-O• species) is the bottleneck of the overall S3→S0 transition, while O2 formation and release appear faster. However, the atomic-scale steps by which the cluster reconstitutes the S0 state after O2 release, including which waters move where and in what order, remain unresolved. The present study aims to close this knowledge gap by modeling the immediate post-O2 events, identifying intermediates and mechanisms that restore the S0 state, and clarifying the origin of structural flexibility in the cycle.

Prior theory and experiment converge on the oxo–oxyl coupling mechanism in S4 for O–O bond formation with O5 and Ox as substrates. Water exchange experiments and earlier QM/MM and MD studies generally identify the Ca-bound W3 as the water that refills the O2-formed cavity, with stepwise O2 release and subsequent water insertion favored over a concerted process. Protonation states of key ligands (e.g., W2 as OH− vs H2O; O4/O5 as O2− vs OH−; Ox as OH/O−/O2−) remain debated, and different spectroscopies have supported different assignments in specific S-states. Previous computational work (e.g., Capone et al.) proposed that W3 moves directly to the Mn3/Mn4 bridge to yield an open-cubane S0, whereas XFEL structures have so far resolved open-cubane geometries for S0–S3. EPR and EXAFS have provided constraints consistent with the accepted open-cubane S0, but hints from EPR (e.g., methanol dependence of the S0 multiline signal and observations in thermophiles) suggest equilibria between closely spaced conformers. Structural isomerism is established in S1 and S2 (Jahn–Teller and valence isomerism). Whether closed-cubane forms transiently occur in S0 had not been explicitly established before this work.

The authors combined Born–Oppenheimer ab initio molecular dynamics (BO-AIMD) with minimum energy path (MEP) DFT calculations. BO-AIMD: A large full-QM cluster model (369 atoms) was built from room-temperature XFEL S3 data (PDB 6WIV), with O5/Ox removed to model the O2-released Im0-O2 state. The model included the Mn4CaO4 core, 20 protein residues near the OEC, 24 crystallographic waters, and a chloride ion, embedded in a dielectric (COSMO, ε=6.0). Protonation states followed widely accepted schemes. AIMD simulations were run in the NVT ensemble at 298.15 K for at least 30 ps with 1 fs time step using TeraChem (v1.94), employing B3LYP-D3, LanL2DZ for Mn/Ca and 6-31G*/3-21G for main group atoms, with GPU acceleration and backbone α-carbon constraints. Two spin manifolds (octet/αααβ and doublet/αβαβ) were simulated without steering forces to observe unbiased water insertion dynamics and cluster evolution. Snapshots were further geometry-optimized to assess relative free energies. MEP calculations: Truncated QM models derived from BO-AIMD snapshots were used to characterize specific reactions beyond the picosecond window: (i) deprotonation of W1 (the new W2) to Asp61 (defining a pre-S0 state), (ii) dissociation of W2(H2O) from Mn4(III) in the closed-cubane pre-S0, and (iii) µ3-W3(OH) ligand transfer from Mn1 to Mn4 to yield the open-cubane S0. Gaussian 16 with B3LYP* (15% exact exchange) and D3(BJ) dispersion was used for optimizations and frequencies (LanL2DZ for Mn/Ca, 6-31G* for H/C/N/O/Cl), with TS verification via IRC. Single-point energies employed SDD for Mn/Ca and cc-pVTZ for main group atoms with SMD (ε=6.0). Sensitivity analyses across multiple dispersion-parameterized functionals and alternate structural starting points (e.g., 6DHP) were performed. Mulliken spin populations and LOBA were used to confirm Mn oxidation states. Constraints and environmental effects (HB networks and protein) beyond the immediate cluster were approximated via the continuum model and constrained residues.

• Water insertion pathway after O2 release proceeds in three stages: Im0-O2 (O2-released cavity), Im1 (W3 moves from Ca toward Mn1, forming strong H-bond with W2 and binding between Ca and Mn1; concurrent motions of W5 and W6), and Im2 (closed-cubane with W3 deprotonated to W3(OH) acting as a µ3-bridge Ca/Mn1/Mn3; W7 binds Mn4; W8 follows; Mn4 becomes octahedral). These events occur spontaneously in BO-AIMD within ~10 ps, implying barrierless or near-barrierless dynamics.
• Electrostatics and geometry rationalize W3 binding first to Mn1 (closer to Ca3+ and a better Lewis acid than Mn4) and subsequent W3 deprotonation that promotes Mn3–W3 bonding and a pivot/carousel-like reorganization of Mn4 ligation, enabling W7 entry from the O4 channel.
• Across simulations in two spin states, the Mn oxidation assignments remain Mn1(III) Mn2(IV) Mn3(III) Mn4(III) and results are qualitatively identical, indicating weak dependence on ferromagnetic/antiferromagnetic coupling patterns.
• Optimized snapshots indicate an energetically downhill progression by about −30 kcal/mol from Im0 to Im2 (intrinsic/local energetics around the cluster).
• Deprotonation of W1 (trans to O4 in Im2) to the lumen via D1-Asp61 is facile, defining a proton-released pre-S0 intermediate.
• W2(H2O) dissociation from Mn4(III) in the pre-S0 closed-cubane is slightly endothermic by ~2–3 kcal/mol with a small barrier of ~4–5 kcal/mol. Jahn–Teller elongation at Mn(III) and coincident W2 protonation/W7 binding weaken Mn4–W2 (trans effect), enabling dissociation analogous to known substrate-water exchange behavior.
• After W2 release (forming S0 closed-cubane), µ3-W3(OH) shifts from Mn1 to Mn4 over a low barrier of ~4–6 kcal/mol to produce the open-cubane S0 structure. The two S0 isomers (closed, open) are quasi-isoenergetic and interconvertible.
• Predicted intermetallic distances in the resulting open-cubane S0 (2.77, 2.76, 2.84, 3.40 Å) agree with EXAFS constraints.
• Kinetics: water insertion and closed-cubane formation occur on picosecond timescales; subsequent S0 closed↔open isomerization occurs on nanosecond timescales (Eyring estimates), well within the experimental millisecond window attributed to the preceding S3→S4 step.
• Sensitivity analysis over functionals and an alternate structural model yields consistent qualitative conclusions with quantitative variations within ~1–2 kcal/mol.
• The work identifies a previously unrecognized structural isomerism in S0 (closed vs open cubane) that may underlie the structural flexibility throughout Kok’s cycle.

The simulations propose a mechanistically and kinetically plausible sequence of ultrafast events after O2 release that reconstitute S0, thereby closing Kok’s cycle. The closed-cubane intermediates explain how Ca-bound W3 enters and reorganizes the cluster: initial binding to Mn1, fast internal proton transfer to W2, ligand reorganizations on Mn4, and W7 ingress. The small barriers for W2 dissociation and µ3-W3(OH) transfer account for rapid attainment of the accepted open-cubane S0. The predicted picosecond-to-nanosecond times fit between XFEL-derived timepoints (onset of O2 evolution at ~1200 µs and evidence of cavity refilling at ~2000 µs) and are consistent with the experimental finding that S3→S4 (not S4→S0) is the rate-limiting step of S3→S0. Although XFEL has resolved only open-cubane S0, the near-isoenergetic closed-cubane S0 isomer could be transient and underpopulated, consistent with EPR observations (e.g., methanol dependence of S0 multiline signal) indicating equilibria between close-lying states. The proposed S0 isomerism complements established isomerisms in S1 and S2 and may be functionally important for catalytic efficiency, water exchange, and resetting of the HB networks and water channels. The pathway differs from earlier computational proposals that had W3 move directly to the Mn3/Mn4 bridge; here, a Mn1-first route reveals additional intermediates and rationalizes the roles of Ca, Mn4, and channel waters in orchestrating the reset.

This study provides a theoretical closure of Kok’s cycle by detailing the post-O2-release steps that reconstruct the S0 state. Ultrafast insertion of Ca-bound W3 leads to closed-cubane intermediates, accompanied by spontaneous W3(H2O)→W2(OH) proton transfer and ligand/water rearrangements. Subsequent low-barrier W2(H2O) dissociation and µ3-W3(OH) transfer to Mn4 rapidly yield the accepted open-cubane S0. The discovery of reversible closed↔open S0 isomerism, coupled to Jahn–Teller axis reorientation of Mn4(III), provides a structural basis for flexibility throughout the cycle and has implications for designing bio-inspired water oxidation catalysts. Future work should probe experimental signatures of the transient closed-cubane S0 isomer and further resolve protonation states and HB network dynamics that modulate these steps.

• The simulations directly model only the S4→S0 phase after O2 release, not the full millisecond S3→S0 transition; processes outside the local cluster (protein dynamics, extended water/proton networks, O2 diffusion) are approximated.
• Energetics reported are intrinsic/local to the cluster and not fully comparable to whole-PSII experimental thermodynamics and kinetics due to model truncation and continuum solvation.
• Protonation states of key ligands (e.g., W2, O4/O5, Ox) are debated; alternative assignments could modulate details though not the qualitative pathway.
• DFT method limitations introduce uncertainties of a few kcal/mol; sensitivity analyses show qualitative robustness but preclude absolute energetics.
• Finite QM model and fixed-backbone constraints limit explicit sampling of long-range HB network rearrangements and protein motions; microsecond experimental intermediates may not be captured in picosecond AIMD windows.
• MEP steps use truncated models for computational tractability, which may omit minor environmental couplings.