A single residue controls electron transfer gating in photosynthetic reaction centers

Interquinone QA− → QB electron-transfer (ET) in isolated photosystem II reaction centers (PSII-RC) is protein-gated. The temperature-dependent gating frequency “k” is described by the Eyring equation till levelling off at T ≥ 240 °K. Although central to photosynthesis, the gating mechanism has not been resolved and due to experimental limitations, could not be explored in vivo. Here we mimic the temperature dependency of “k” by enlarging VD1-208, the volume of a single residue at the crossing point of the D1 and D2 PSII-RC subunits in Synechocystis 6803 whole cells. By controlling the interactions of the D1/D2 subunits, VD1-208 (or 1/T) determines the frequency of attaining an ET-active conformation. Decelerated ET, impaired photosynthesis, D1 repair rate and overall cell physiology upon increasing VD1-208 to above 130 Å3, rationalize the >99% conservation of small residues at D1-208 and its homologous motif in non-oxygenic bacteria. The experimental means and resolved mechanism are relevant for numerous transmembrane protein-gated reactions.

on the quinones and their immediate vicinity. In particular, temperature increases were shown to result in elevated frequency and amplitude of individual atom displacements at and around the non-heme iron and quinone sites 4,16,26 . Decreased temperatures have been correlated with increased binding interactions in this domain, leading to less frequent displacements and consequentially reduced probability to attain the conformation favorable for ET 11 . Other experimental and theoretical studies have aimed at resolving the relative orientations, electron distribution, redox potentials, formation or elimination of hydrogen bonding to Q B − and changes in the amino acids that ligate the non-heme iron 9,11,13,22,[27][28][29][30][31][32][33][34] . However, to the best of our knowledge, none have successfully generated a model that quantitatively relates well-defined protein conformations with the ET frequency. Furthermore, none of the proposed models were evaluable in whole cells, where monitoring of photosynthetic ET and subsequent processes below the freezing point is impossible. Thus the RC gating phenomenon and its impact on whole-cell photosynthetic activity and physiology remained unresolved in oxygenic and non-oxygenic phototrophs alike.
In this study, we first set out to find a means of mimicking the temperature effect on protein conformations and the resulting Q A − → Q B ET rate in vivo, without cooling the cells. Once establishing such a model, the impact of "temperature-like" perturbations on light-induced ET, photosynthetic machinery and whole cell physiology could be quantitatively monitored.
We hypothesized that reversible thermal motion of the Q A and Q B binding helices (d2 and d1) relative to each other "opens" and "closes" a "gate" for ET tunneling from Q A to Q B . We further hypothesized that as in other systems undergoing "gating" [35][36][37] , a hinge controls the frequency of this movement. Following careful examination of Type-II RC structures, we speculated that the hinge is located within a GxxxG-like TM helix-helix interface motif, where G denotes small amino acid residues V res < 130 Å 3 (i.e., Gly, Ser, Cys, Thr and Ala) [38][39][40] . As a candidate we selected D1-208 within the TM G 208 xxxG 212 -like motif in the d1 helix of PSII RC. This motif faces a G 207 xxxG 211 -like motif in the d2 helix (Fig. 1b). Following transition state theory, residues that comprise The four-helix bundle includes helices d1, e1, d2, e2 of the D1 (green) and D2 (purple) subunits. It holds a central cluster of chlorophylls (P680), including a central pair (green) flanked by additional Chls (yellow), Phe (purple), quinones (red) and the non-heme iron (red sphere), that is further ligated by a bicarbonate (red). See Supplemental Fig. 1 for a depiction of all the 10 TM helices of subunits D1 and D2. Within the four-helix bundle, intersubunit interactions are found at the center of d1 (D1-208 and D1-209) and d2 helices, around the non-heme iron, where histidines (black) from the two subunits ligate an iron ion, and at the central P680 Chls, where histidines from the two subunits ligate the Mg ions. A 90 o turn (bottom) shows that the d helices form an 'X' shape with D1-208 at their center. (b) The intersubunit Cα -Cα distances between of the studied GxxxG-like motifs: D1 residues Gly208, Ala209, and Cys212 and their closest D2 residues Ala208, Gly207 and Cys211, respectively. The histidines ligating the non-heme iron (red sphere) and central P680 Chls (green) are depicted. (c) The intersubunit Cα -Cα distances (y-axis) between d1 (x-axis) and the closest Cα on d2 (see Supplementary Fig. 1, for detailed table). (d) The list of residues enabling photoautotrophic growth for the three studied amino acids includes 4 residues for D1-208, 9 for D1-209 and 14 for D1-212. a hinge should control the binding interactions among the d1 and d2 and consequently determine the ratio of ET-active versus inactive protein conformations. A similar GxxxG-like motif consisting of L-185 to L-189 in RC of non-oxygenic bacteria interfaces M-212 to M-216 at the crossing of d1 and d2 helices ( Supplementary Fig. 1A). The presented in silico, in vivo and in vitro studies focused on the PSII RC motif, with particular emphasis on D1-208.

Results
Evolutionary conservation of a small residue motif at the d1/d2 TM interface. D1-208 and D1-212 delimit a G 208 xxxG 212 -like motif. The motif is located at the membrane's hydrophobic core, with the C α centers of the terminal residues positioned at a membrane depth of 2.46 and − 6.73 Å (PDB: 3wu2, ref. 41) from the membrane center, respectively (Supplementary Table 1). These residues face a d2 helix G 207 xxxG 211 -like motif where D1-208Gly is in close contact with D2-208Ala (4.0 Å Cα to Cα , Fig. 1c and Supplementary Fig. 1B). Such an interhelical distance is regarded as an extremely and atypically close distance 39 that should be accompanied by strong interhelical interactions.
D1-208Gly is flanked by small residues D1-207Gly on the donor side and by D1-209Ala on the acceptor side of the PSII, forming a unique 'consecutive-small' residue motif. The distance between the respective d1 and d2 helices then increases as approaching the donor and acceptor sides of PSII (Fig. 1c). The Cα to Cα distance between residues D1-209 and D2-207Gly (acceptor side) is 4.4 Å and between D2-211Cys and the accepting D1-212 is 4.8 Å. The next interface pairs, namely D1-205Val and D2-208Ala and D1-202Val and D2-205Leu are positioned at Cα to Cα distances of 5.3 and 10.1 Å, respectively ( Fig. 1c and Supplementary Fig. 1B).
Screening of all available D1 PSII RC sequences demonstrated conservation of three consecutive small residues between D1-207 and D1-209 in 4665 sequences. The vast majority (89.4%) were GGS, while the remaining (10.6%) were GGA sequences. Overall, the few cases in which the motif was not fully conserved small residues still comprised the sequence, which were often non-viable, e.g. in viruses (Supplementary Table 2).
Small amino acid volume at the d1/d2 TM interface is essential for photoautotrophic growth of Synechocystis sp. PCC 6803. Using PCR-based saturation mutagenesis in Synechocystis sp. PCC 6803, we aimed to experimentally assess the role of most closely situated d1 TM sites that may be involved in intersubunit interactions with the d2 helix ( Fig. 1). All viable mutants contained functional RCs that maintained photoautotrophic growth and presented Q A -→ Q B ET. As shown in Fig. 1d, saturation mutagenesis at the D1-212 site yielded the wild-type (wt) (D1-Ser212Ser) as well as thirteen photoautrophic isogenic mutant strains bearing Gly, Ala, Cys, Thr, Asn, Asp, Pro, Val, Gln, Glu, Ile, Leu, or Met at the 212 position. Mutagenesis at the D1-209 site yielded the wt (D1-Ser209Ser) and eight (Gly, Ala, Cys, Thr, Asn, Asp, Pro, Val) mutants (Fig. 1d). The same mutagenesis procedure at the D1-208 site yielded the wt (D1-Gly208Gly) and only three photoautotrophically competent mutants, bearing Ala, Ser or Thr at the 208 position (Fig. 1d), hereafter denoted D1-Gly208Ala, D1-Gly208Ser and D1-Gly208Thr. The remaining 16 possible amino acid substitutions at D1-208 failed to support photoautotrophic growth. Interestingly, bulky aromatic amino acid residues (Tyr, Trp, Phe, and His) and positively charged amino acid residues (Arg and Lys) 42 failed to support photoautotrophic growth when substituted into these three target sites despite the fact that their codons were available. In summary, the allowed maximal residue volume decreases with decreasing distances from the crossing point of the d1/d2 helices (Fig. 1).  Supplementary Fig. 2), was followed by monitoring in vivo chlorophyll fluorescence decay in whole cells, at temperatures ranging from 0-50 °C. The procedures selected for the whole cell preparation assured the measurement of the first Q − A → Q B . These include adaptation to dark; fluorescence measurements modality and deconvolution of the collected data (see online methods). The in vivo ET rates of wt and all viable D1-212, 209 and 208 mutants were derived as described 43 . Mutants D1-Ser212Met and D1-Ser209Cys, could not be stabilized during growth with significant reversal to the wt strain hence excluded from the analysis. The experimentally derived standard deviations of the measured k values are presented for the three mutated D1 residues in Supplementary Fig. 2b-d. The Ln(k/T) displays a linear correlation with 1/T for wt Synechocystis 6803, till leveling off at, or slightly above T 0 , the physiological temperature of the organism habitat 43 as illustrated in Fig. 2a

Volume increase of residue
Next, for all D1-212, D1-209, and D1-208 mutants, Ln(k/T 298°K ), i.e. the ET rate at T = 298 °K, was plotted ( Fig. 2d-f) as a function of the residue volume V res till reaching T 0 . as it shown before 43 . The ET rate constants for the D1-212 mutants clustered into the previously 43 defined group I (V I res ≤ 130 Å 3 , green data-points in Fig. 2) and group II (Pro and residues with V II res > 130 Å 3 , red data-points in Fig. 2. The Ln(k/T 298°K ) values for group I were in the range of 2.1-2.4 (Fig. 2d), while those for group II ranged between 1.5 and 2.0. For D1-212, small changes in the packing values of group I residues were previously shown to facilitate the acclimation of mesophiles and thermophiles to their ambient temperatures 43 . For D1-209, a linear regression (Ln(k/T 298°K ) = − 5.9·V D1-209 + 2.9, R 2 = 0.87) was obtained for all viable mutants. The allowed maximal residue size was significantly smaller as compared to D1-212 (150 Å 3 and 170 Å 3 , respectively). Occupation of the D1-208 site was only feasible with small, group I residues (V res < 130 Å 3 ); the plot of Ln(k/T 298°K ) vs V D1-208 (Fig. 2f) presented a steep linear regression (Ln(k/T) = − 14.5·V D1-208 + 3.25, R 2 = 0.88). Taken together, the effect of V res on the rate of ET markedly increased as the site of mutation approached the closest d1/d2 contact point (crossing point). At D1-208, this effect appeared equivalent to the previously reported effect of reduced PSII RC temperatures in membrane fragments or thylakoids 4,16 . Both interventions strongly "cool down" the rate of the protein-gated Q − A → Q B ET (Fig. 3). To formulate this intriguing equivalence, we extrapolated the Ln(k/T) vs. (1/T) plot beyond the measured temperature range (dashed boxes in Supplementary Fig. 3 represent the extrapolated region) for the wt (gray diamonds in Supplementary Fig. 3). Then, the rate constants measured at 25 °C for the different viable mutants (k mut 298°K , Fig. 2) at D1-208, D1-209 and D1-212 sites, were positioned on the extrapolated wild type plot (colored squared, Supplementary Fig. 3). Next, for each V res , we found a corresponding 1/T that provided the same ET rate (arrows, Supplementary Fig. 3). The derived 1/T values were used to construct a plot of 1/T vs. V res for each site ( Fig. 2g-i). These plots provide the "effective" cooling effect introduced by a certain residue volume at this site.  (d-f) The measured Q A − → Q B ET rate at 25 °C is presented as a function of the residue volume (V res ) for the three sites. Utilizing the linear relationship found between Ln(k/T) and V res , for each of the three sites, each mutant was given a temperature that provides for the same ET rate as that of the wt at room temperature. (g-i) Correlation between the temperature of the "effective" cooling effect and the residue volume.
Scientific RepoRts | 7:44580 | DOI: 10.1038/srep44580 Previous works have shown that exposing Type-II RCs to low temperatures, freezes both Q − A → Q B ET and the average displacements of atoms throughout the PSII RC protein scaffold 4,16 . The impact of T reduction, measured for isolated PSII RCs 4,16 , and V res elevation, measured here in whole cells, on Q A → Q B ET were highly similar ( Fig. 3 and Supplementary Table 3), strongly suggests that the D1-208 residue controls the frequency of protein shifting from an ET-inactive to an ET-active conformation.
D1-209 mutants also displayed a linear regression relationship between 1/T and V D1-209 , albeit, more moderate when compared to D1-208 (Fig. 2h). For example, a volume increase of 24 Å 3 was equivalent to a temperature decrease of ~8 °K ( Supplementary Fig. 3B). Finally, as shown in Fig. 2g, no systematic "cooling effect" was demonstrated for increased V res at D1-212. Nevertheless, as we previously showed, changing the residue packing value modified the entropy of activation and thereby the ET rate 43 . The apparent "cooling" of the Q − A → Q B ET with increasing V res values, can be translated to increased activation free energy (∆G ‡ ), as shown for D1-208, as well as for D1-209, although at a lower rate and with a somewhat lower correlation ( Supplementary Fig. 4).
Detailed analysis of the enthalpic (∆H ‡ ) and entropic (∆S ‡ ) components of the activation free energy for each of the D1-208 mutants (Supplementary Table 4) indicated that the residue size mainly affected the entropy of activation (by up to ~3.7-4.8 kJ·mol −1 ), which was in agreement with the concept of local cooling and a previous study of the Q − A → Q B ET 17 . The ability to locally and systematically "cool down" the ET processes in vivo while leaving the rest of the cell at a physiological temperature, provides a new means of studying the physiological significance of Q − A → Q B ET rates in the context of whole cells.
D1-208 volume affects whole cell growth, RC pigment composition and RC integrity. The photoautotrophic growth rate was found to be the highest in the wt (Fig. 4a), which exhibited a doubling time of 0.8 days, while that of D1-Gly208Thr was 20% longer. Growth rate of the D1-Gly208Ala and D1-Gly208Ser mutants was 0.9 days. Under light conditions of 80 μ mol (photons)·m −2 ·s −1 , the D1-Gly208Ala mutant demonstrated a 50% decrease of D1 content, while D1-Gly208Ser and D1-Gly208Thr showed 60% and 70% reduction, respectively compared with the wt (Fig. 4b). Furthermore, the mutants exhibited larger amounts of the ~60 kDa cross-linked D1/D2 heterodimer products, relative to the wt. The steady state levels of the PsaC protein were equally high in the wt, D1-Gly208Ala and D1-Gly208Ser mutants, but were approximately 50% lower in the D1-Gly208Thr mutant (Fig. 4b). Xanthophyll carotenoids myxoxanthophyll 44 (Pearson's correlation R = 0.854) and echinenone (Pearson's correlation R = 0.643) levels increased with the size of the D1-208 residue (Fig. 4c), which paralleled a significant decrease in the content of zeaxanthin (Pearson's correlation R = − 0.895) and β -carotene (Pearson's correlation R = − 0.754). Accumulation of D1 degradation products was demonstrated in time-resolved D1 protein immunoblots of wt and mutated cells exposed to high light (500 μ mol (photons)·m −2 ·s −1 ) in the presence of the protein synthesis inhibitor, lincomycin (Fig. 4d,e). Specifically, the 16 kDa fragment of the C-terminus was detected in the wt but not in the mutants (Fig. 4b). D1 degradation (Fig. 4e) appeared almost five times faster in the D1-Gly208Thr mutant compared to the wt and three-fold faster in D1-Gly208Ala and D1-Gly208Ser as compared to wt.
Increased D1-208 volume affects d1/d2 binding interactions. To resolve the mechanism by which D1-208 V res affects the thermodynamics and kinetics of Q − A → Q B ET, we explored the freedom of motion and binding interactions between the d1 and d2 helices mutated at D1-208 using molecular dynamics (MD) simulations. The probability of specific intersubunit interactions is displayed in Fig. 5a   mol −1 , respectively (d H = 1.9 Å in both). The least probable interhelical H-bond formation in all D1-208 mutants (4% for wt) was between d2-211Cys(Sγ ) and d1-212Ser or d1-212Ser(Oγ ), which had projected binding energies of 11.5 and 15.7 kJ.mol −1 , respectively. Increased D1-208 volume stabilizes the d1/d2 binding. The MD simulations unveiled the impact of V res values on the probability of making intersubunit H-bonding between d1 and d2. In agreement with the local "cooling effect", increased V res may also strengthen the van der Waals interactions between the two helices, e.g. dipolar interaction. To test this hypothesis, AFM was applied to measure the forces required to unbind synthetic d1 and d2 TM analogues comprising wt and mutated amino acid sequences (Fig. 5b, Supplementary Fig. 5). When the D1-208 site was occupied with Gly (wt), Ser (d1-208Ser) or Val (d1-208Val), representing group I (Gly, Ser) and group II (Val), residues, the mean unbinding forces for the d1wt/d2wt complex decayed markedly faster with lowered loading rates, as compared to the d1-208Val/d2wt pair (Fig. 5b). Specifically, the rate of dissociation of the d1wt/d2wt complex at equilibrium was twice as fast as that of the d1-208Val/d2wt complex. The dissociation rate of the d1-208Ser/d2wt complex was intermediate. When normalizing and analyzing the data according to Dudko 45 , the extracted lifetimes for the d1wt/d2wt, d1-208Ser/d2wt and d1-208Val/d2wt, were 9, 23 and 52 ms, respectively.

Discussion
The long-unresolved protein gating mechanism of Q A − → Q B ET in Type II reaction centers has been suggested to reflect specific changes in binding interaction at the quinones' vicinity 46 . While recent serial time resolved crystallography of the PSII-RC complex appear to identify such changes, more recent study has challenged these findings 47 .
Unlike this and earlier studies, we did not attempt to solve the molecular details of the ET active and inactive conformations, e.g., the "light" and "dark" conformations 30 . Rather, we searched for specific residues in protein-protein interface region that control the frequency of shifting from one conformation to another, expecting that such information can be utilized in subsequent studies to further characterize these conformations. With this approach, we searched for RC structural elements that determine the protein flexibility and adaptability to structural changes as reflected in the frequency (rate) of the apparent Q A − → Q B ET. Overall, the similar dependency of RC flexibility and Q A − → Q B ET rates on temperature has already been recognized two decades  16 . Both were shown to freeze at a similar temperature: ~250 and ~200 °K for Type-II RC for oxygenic and non-oxygenic bacteria, respectively 4 . However, assignment of specific residue as the conformational gating regulators remained highly speculative, since the ambient temperature affects movement of multiple residues in protein-level experiments in vitro, let alone in organism-level experiments in vivo 9 . Nevertheless, the concept of residue and cofactor displacement at temperatures exceeding the protein freezing point, provided preliminary evidence of protein-gated ET in Type-II RC. Our results suggest that the temperature dependence of the ET rate in PSII RC within whole cells can be mimicked under 298 K, which is the physiological temperature of the organism, by changing the V res at D1-208, a single site at the d1/d2 crossing point located within the center of the helix-helix interface. In agreement with our working hypotheses, increasing V D1-208 linearly increased the energy barrier for d1/d2 unbinding and logarithmically decreased the ET rate constant. Indeed, the previously reported in vitro cooling effect on ET efficiency 4,16 coincides with our in vivo findings regarding ET rate attenuation due to V res changes at the D1-208 site (Fig. 3). Thus, D1-208 functions as a hinge controlling the frequency of thermal d1 and d2 fluctuations relative to each other.
Notably, the D1-208 site is not proximal to the Q A /Q B redox pair, yet, when examining the Type-II RC structures, the effect of such a remote site is not surprising. The d helices in both oxygenic and non-oxygenic organisms are non-covalently bound at the (Fig. 1): (a) non-heme iron, ligated by histidines of the two d helices, the same residues that also bind Q A and Q B , (b) d1/d2 central TM crossing interface where inter-helical H-bonds and hydrophobic interactions tie the helices to each other, and (c) central (B)Chls cluster 3,11,26 . Following this emerging insight of the d1 to d2 binding regions, interactions at the d1/d2 crossing-interface are expected to markedly impact the freedom of movement (flexibility) of these helices at their quinone binding residues. Similar remote effects on protein flexibility and functionality were previously reported for numerous protein systems 24,25 . The known sensitivity of the non-heme iron coordination and spin state to the collective motions of the RC protein core are in line with such a "remote" effect on protein flexibility 5,26,48,49 .
The transition state that enables ET is expected to be temporarily stable. This may be facilitated by an H-bond network re-arrangement along the GxxxG-like motif, such as exchange of the H-bond D1208Gly … D2-211Cys for D1-212Ser …. D1-271Met, which decreases the probability (lower entropy) of escape from the transition state 17,50 . Indeed, the entropy calculations show a ~3.7-4.8 kJmol −1 energy barrier between wt (Gly) and Ser and Thr mutants at the D1-208 site (Supplementary Table 4).
The importance of d1/d2 thermal fluctuations at the D1-208 site was further corroborated by MD analysis. According to our results (Fig. 5a), the wild type Gly at the D1-208 site does not participate in interhelical H-bonding. In contrast, Thr and even Val (as example of Class II, high-V res , residues) exhibited a nearly persistent H-bond (30% materialization). Moreover, the AFM analysis of d1/d2 TM peptides showed that the increased lifetimes of the paired peptides corresponded with increased stability of helix-helix binding, which positively correlated with D1-208 size (Fig. 5b). It also provided a metric of protein flexibility, estimated by the frequency of shifting from one d1/d2 conformation to another, manifested by unbinding and temporal rebinding at and around the D1-208 site. In contrast to most GxxxG-like motifs at TM helix interfaces, typically flanked by bulky β -branched residues that rigidify the protein contact domain 39,40 , D1-208 is flanked by conserved small residues, which extend the flexibility of this region (Supplementary Fig. 1A). Figure 6 and Supplementary Fig. 6 illustrate the proposed mechanism for protein-gated ET in Type II RC, which follows the transition state theory Eyring equation (equation (2)).
Binding interactions between the d1 and d2 helices at the crossing point (e.g., D1-208 in PSII RCs) preserve the Type II RC in an ET-inactive conformation. The collective motions of the protein scaffold in the dark below or at the physiological temperature, T, are insufficient to unbind the two helices. The additional energy required to release the binding, i.e., to "open" the gate, and achieve the ET active d1/d2 geometry, is provided by the exothermic ET from the (bacterio)chlorophyll cluster to Q A (dissipation of ~6 kCal/mol 51 . An increased 1/T reduces the value of the first exponent in equation (2) and thereby the value of "k". Increased V D1-208 increases the negative value of ΔS ‡ (Supplementary Table 1), resulting in a decreased value of the second exponent. Hence, increased V D1-208 is equivalent to increased 1/T, accounting for the equality presented in equation (1).
This mechanism provides a straightforward rationale for the conservation of small residues at the d1/d2 crossing of Type-II RC (Fig. 6). Such residues provide a hinge that maximizes the frequency of helices displacement relative to each other at physiological temperatures. Larger residues would markedly reduce (Fig. 6) or "cool down" the unbinding displacement frequency of d1 versus d2, and thereby, the frequency of Q − A to Q B ET. Moreover, slow gating may allow for charge recombination, generation of toxic reactive oxygen species and rapid degradation of the photosynthetic machinery, particularly, at high light intensities.
The ability to mimic the temperature effect at cryogenic temperatures by a single site modification, allowed us to isolate the in vivo impact of the conformation-gated ET frequency from unrelated enzymatic reactions, such as autotrophic growth rates, D1 and psaC protein content, carotenoid profiling, and PSII RC degradation (Fig. 4). For example, we showed that attenuation of the ET rate in vivo, following enhanced d1/d2 binding (introduced by D1-Gly208Thr), resulted in significant modification of the carotenoid profile. It was also found to facilitate cross-linking of the D1 and D2 protein subunits, thereby attenuating D1 repair (Fig. 4b). Similar crosslinking was previously observed under high-intensity light stress condition, where Q A reduction did not match Q A − oxidation 52 . The amino-acid sequences at the TM interface of the d1 and d2 helices from the L and M subunits in non-oxygenic bacteria (Supplementary Fig. 1) suggest a similar mechanism of conformation-gated ET. This includes similar intersubunit hydrogen-bonding patterns 17 and rare consecutive small residues within a Figure 6. Increased V res at D1-208 increases the interaction between the d1 and d2 helices. Using space filling models for different amino-acids at the D1-208 locus, the increased contact interaction between the two helices is depicted. See Supplementary Table 3 for the precise volume of D1-208 in the different mutants.
Scientific RepoRts | 7:44580 | DOI: 10.1038/srep44580 GxxxG-like motif (Supplementary Fig. 1). Importantly, the distance between the d1 and d2 helices at the crossing point of the L and M protein subunits is somewhat larger than in PSII RC (Supplementary Fig. 1B). Thus, group I residues larger than Gly (e.g. Ala) may provide a better hinge for the conformational change of the non-oxygenic RC, in agreement with the observed sequence ( Supplementary Fig. 1). 53 of the D1 PSII subunit was conducted allowing for 10,000 output sequences. All PSII RC resulting sequences were analyzed. These displayed an expectation value (e-value) of better than 9 × 10 −148 as well as 82% identity over 243 residues. This e-value was chosen as below it bacterial reaction centers were present in the BLAST results.

Sequence conservation. A BLAST sequence search
Photosystem II structural analysis. The PSII structure from T. vulcanus at a resolution of 1.9 Å (PDB code 3wu2) was studied as the highest-resolution available structure 41 . To validate the findings, other structures were analyzed including from T. vulcanus (PDB code 4il6 54 with 2.1 Å-resolution) and from T. elongatus (PDB codes 4pj0 55 , 3bz1 56 , and 2axt 57 displaying 2.44, 2.9 and 3.0 Å-resolution, respectively).
Alignment of the protein complexes relative to the membrane such that the z-coordinate of each atoms correlates with membrane depth, was conducted using the OPM 58 or Ez 59 servers. Hydrogens were added to the protein complex using REDUCE 60 . Rotamers were scanned using SCWRL 61 . Correlation of Gibbs activation energy to the volume of the corresponding residues was conducted using the residue volume scale of Wodak 42 .
Preparation of Synechocystis 6803 mutants. The D1-208, D1-209 and D1-212 mutant libraries were constructed using wt Synechocystis 6803 that contained intact psbAII gene with kanamycin resistance cassette introduced downstream of the psbAII at the StuI site. Two other gene copies psbAI and psbAIII were inactivated by spectinomycin and chloramphenicol resistance cassettes respectively 62 .
PCR-based saturation mutagenesis of Synechocystis 6803 psbAII gene that codes for the D1 protein was carried out as described ( Supplementary Fig. 7 and Supplementary Table 6). The 3′ fragment of the psbAII gene carrying the mutation at positions 622-624 corresponding to the D1-208 63 was made with P3 primer and the degenerated primer P208 coding for all amino acids in total. Similarly, the psbAII gene mutation at positions 625-627, corresponding to the D1-209 was introduced with the P3 and a degenerated primer P209. Additionally, the silent mutation was introduced in the third codon downstream the 208 or 209 mutation to create the NsiI restriction site used for screening of the transformant colonies. Primers P5and A208 for D1-208 library and P5 and A209 for D1-209 library were used to create the 5′ fragment of the gene required for homologous double recombination. Both the 5′ and the 3′ fragments were purified from agarose gel and fused in the final PCR to construct linear DNA product that was used to transform the recipient strain. The presence of the mutation was verified by restriction of the PCR product with the NsiI and by sequencing using P260 and P4 primers in both strands of the gene. Specific X208 and X209 oligonucleotides were designed to introduce codons in the D1-208 or D1-209 site, respectively corresponding to amino acids that were not obtained with the help of degenerated P208 or P209 primers (Supplementary Table 5). Construction of the mutant library at the D1-212 site was as described 43 . ET in vivo measurements. Cells re-suspended in fresh BG-11 medium were pre-incubated in darkness on ice for 30 minutes to completely reoxidize all PSII-RCs. Chlorophyll fluorescence decay following a single-turnover saturating flash (15 μ s, ~0.1 mol photon m −2 s −1 ) was measured by a series of short, weak measuring flashes (pulse duration 4 μ s), in the range of 0-50 ± 0.1 °C using FL-100 double-modulation fluorometer equipped with TR 2000 thermoregulator (PSI Ltd., Czech Republic) as described 43,50 . The normalized fluorescence was fitted to a three exponential decay function, using nonlinear least-squares fitting algorithm in MATLAB. The highest rate constant was assigned to the Q A − → Q B ET 64,65 . The ET rates in all mutants follow an Eyring exponential dependence on the temperature inverse. An example of the ET rate measured by chlorophyll fluorescence decay is provided in Supplementary Fig. 2. Growth conditions. Strains of Synechocystis 6803 were grown in 250 ml Erlenmeyer flasks containing BG-11 medium 66 under continuous light illumination of 80 μ mol (photons)·m −2 ·s −1 provided by Fluora TM fluorescent tubes (Osram GmbH, Germany). Photoinhibitory conditions were elicited by halogen lamp (500 W) delivering illumination of 500 μ mol (photons)·m −2 ·s −1 . The cultures were kept at 30 °C in an incubator, aeration was provided by magnetic stirring. Growth rates were estimated from changes in cell density following light attenuation at 730 nm (OD 730 ) and chlorophyll a concentration every 24 hours for 7 days. New cultures were inoculated from a starter culture and brought to an equal cell density (OD 730 of 0.2). For the physiological measurements, cells were harvested at a mid-exponential growth phase (OD 730 of 0.8-1) they were pelleted and re-suspended in a fresh BG-11 medium to reach 4 μ M·concentration of chlorophyll a. Pigment analysis and absorption spectroscopy. Chlorophyll a, β -carotene and xanthophylls (myxoxanthophyll, zeaxanthin, echinenone) extracted from the cells of the wt and the mutant strains grown at 80 μ mol (photons)·m −2 ·s −1 light intensity were separated and identified by HPLC analysis (Fig. 4c). The Philips PU 4100 HPLC included a Spherisorb ODS-1 reversed phase column (Waters, USA) according to the procedure described 67 . Chlorophyll a concentration was also determined spectroscopically using Jasco V-570 spectrophotometer (Jasco Inc., USA) from cells' methanol extract using extinction coefficient of Lichtentaller 68 .

SDS-PAGE and immunoblotting.
Before extraction of thylakoid membranes cells of all strains were brought to OD 730 = 1. Thylakoids were isolated as described 69 , chlorophyll was measured in membrane fraction, and protein extracts of all the strains were brought to total chlorophyll content 2.5 μ g before loading. Thylakoid proteins were solubilized in sodium dodecyl sulfate (SDS) buffer (0.5 M Tris-HCl pH 6.8, 1% SDS, 24% glycerol and 4% 2-mercaptoethanol) at room temperature for 1 hour, and separated by 12.5% SDS-PAGE. Separated proteins were transferred onto a polyvinylidene difluoride (PVDF) membrane (Hybond-P, Amersham). The immuno detection was performed with primary antibodies against global D1, D1 C-terminal and PsaC proteins (Agrisera, Sweden). Secondary HRP-conjugated antisera (Promega and Sigma) were visualized using enhanced chemoluminescence system (Pierce). The quantification of the amount of D1 and PsaC proteins was done by integrating variable pixel intensities with ImageJ and comparing them to a dilution series of samples.

Molecular dynamics simulations.
The 3D structures of the D helices in the D1 and D2 PSII subunits were obtained by in silico mutagenesis of PSII-RC structures (PDB: 3BZ1). The models of the D helices (D1-P196:221, D2-P195:N219) were placed in periodic boundary simulation boxes that were 1 nm larger than the peptides using YASARA (YASARA Biosciences GmbH, Vienna, Austria). To mimic the native structure of the Synechocystis 6803 in silico mutations were done as described 50 at D1-209Ala and D1-212Cys sites into Ser, and D2-204Val into Ile. In addition, two lysines were added to both termini to imitate the sequence of the peptides synthesized for the dynamic force spectroscopy d1wt: KKPFHMLGVAGVFGGSLFSAMHGSLVTSKK d2wt: KKPFHMMGVAGILGGALLCAIHGATVENKK. The wt structure of the d1 helix, d1wt was subsequently mutated at the D1-208Gly site into Ala, Ser, Thr and Val yielding four mutated structures: Hydrogens were added according to basic chemistry rules in pH = 8.0, the boxes were filled with TIP3P water model, and sodium atoms were iteratively placed at the coordinates with the lowest electrostatic potential until the cell was neutral. Molecular dynamics simulations were run with YASARA Structure (version 11.1.19), using a time step of 1.25 fs for intra-molecular and 2.5 fs for intermolecular forces. To remove energy hot-spots and correct the covalent geometry, the structures were first energy-minimized with the Yamber3 force field 70 , using an 8.0 Å force cut off and the Particle Mesh Ewald algorithm 71 . After removal of conformational stress by a short steepest descent minimization, the procedure was continued by simulated annealing (time step 2 fs, atom velocities scaled down by 0.9 every 10th step) until convergence was reached, i.e., the energy improved by less than 50 J·mol −1 during 200 steps. The simulations were then run at 300 K for 20 ns at a constant pressure (NPT) to account for volume changes due to fluctuations of peptides in the solution. Graphics were created with YASARA and POVRay (www.povray.org). The probability of interhelical H-bond formation was calculated as the ratio of the number of simulated snapshots featuring the bonded state over the total count of all simulated snapshots. AFM and Dynamic Force Spectroscopy. Peptides corresponding to d1 and d2 were synthesized chemically using standard Fmoc-protected amino acids as described 72 . Silicon nitride probes MSCT of 40 pN. nm −1 nominal spring constant (Bruker Inc., Camarillo, USA) and freshly cleaved sheets of muscovite mica were modified with ethanolamine in order to generate free -NH groups for coupling to the ~6 nm long flexible acetal-PEG-NHS linker as described 73 . Acetal function was cleaved with 1% citric acid in water and probes (mica sheets) with the cross-linker were then immersed in the aqueous solution of the peptide. The receptor peptides d2 (D-helices of the D2 protein) were immobilized to the mica surface via flexible PEG linker and the d1 peptides were attached to the AFM tip. Unreacted functional groups were passivated with aqueous solution of 1 M ethanolamine. Functionalized probes and mica sheets were then washed in the 0.1% SDS and stored at 4 °C for immediate use.
The interactions between the d1wt, d1-208Ser, d1-208Val peptides and the d2wt peptide were investigated by measurements of force distance cycles using the atomic force microscope PicoSPM (Agilent, Tempe, AZ). Measurements were carried out in a liquid cell enclosing 0.5 mL of 0.1% SDS at room temperature. The 0.1% SDS solution was optimized for retaining at least 90% helicity of the peptides as reflected by circular dichroism measurements (in the range of 180-320 nm) using Jasco J-715 spectropolarimeter (Supplementary Fig. 8).
The sweep amplitude of the force-distance cycles was 100 nm, with the sweep frequencies ranging from 0.3 to 10 Hz. Representative retrace data of the force distance cycles display unique unbinding events as shown in Supplementary Fig. 9. Two thousand force-distance cycles were performed with each of three individually functionalized tips. To measure the unbinding forces, spring constants of cantilevers were calibrated using thermal noise method 74 . Force spectroscopy curves were analyzed according to the formalism of Baumgarten et al. 75 utilizing a custom made MatLab ™ script 76 (MathWorks Inc., Natick, MA).
The probability of unbinding was ≤ 10% in all examined pairs of peptides. Extrapolation of the mean unbinding forces to a zero applied force was carried out using the phenomenological theory of Bell 77 . Application of Dudko's theory allowed to extract detailed information about the thermodynamics and microscopic properties of the dissociating complex from the unbinding force data 45 assuming that the complex dissociation proceeds through an escape from a one-dimensional harmonic potential well to rupture through a transition state.
Control measurements were performed with the d2 peptide (immobilized on the mica surface) blocked by an excess of the respective d1 peptide partner ( Supplementary Fig. 5). Force-distance cycles measured in decane ( Supplementary Fig. 5) performed at hydrophobic conditions akin to the interior of membrane yielded unbinding forces between d1 and d2 peptides indistinguishable from those measured in aqueous 0.1% SDS solution.