Double Mutation in Photosystem II Reaction Centers and Elevated CO2 Grant Thermotolerance to Mesophilic Cyanobacterium

Photosynthetic biomass production rapidly declines in mesophilic cyanobacteria grown above their physiological temperatures largely due to the imbalance between degradation and repair of the D1 protein subunit of the heat susceptible Photosystem II reaction centers (PSIIRC). Here we show that simultaneous replacement of two conserved residues in the D1 protein of the mesophilic Synechocystis sp. PCC 6803, by the analogue residues present in the thermophilic Thermosynechococcus elongatus, enables photosynthetic growth, extensive biomass production and markedly enhanced stability and repair rate of PSIIRC for seven days even at 43°C but only at elevated CO2 (1%). Under the same conditions, the Synechocystis control strain initially presented very slow growth followed by a decline after 3 days. Change in the thylakoid membrane lipids, namely the saturation of the fatty acids is observed upon incubation for the different strains, but only the double mutant shows a concomitant major change of the enthalpy and entropy for the light activated QA −→QB electron transfer, rendering them similar to those of the thermophilic strain. Following these findings, computational chemistry and protein dynamics simulations we propose that the D1 double mutation increases the folding stability of the PSIIRC at elevated temperatures. This, together with the decreased impairment of D1 protein repair under increased CO2 concentrations result in the observed photothermal tolerance of the photosynthetic machinery in the double mutant


Introduction
Photosystem II reaction center (PSIIRC) is a water/quinone oxido-reductase that catalyzes light-activated electron mobilization from the lumenal (water oxidation site) to the stromal (quinone reduction site) side of the photosynthetic membrane. The electrons are transferred through the PSIIRC in a multi-step process initiated by photoexcitation of the primary electron donor (chlorophylls). This process is concluded by the double, stepwise reduction of a mobile quinone, termed Q B , by another quinone termed Q A . Impairment of the Q A 2 RQ B electron transfer steps results in the recombination of electrons and holes, an increased probability of reactive oxygen species production, and attenuation of the PSIIRC repair [1,2,3]. The malfunctional PSIIRC undergoes repair that involves its partial disassembly, removal, and proteolysis of the D1 protein subunit, generation of a new D1 protein, and refolding of the repaired PSIIRC to a functional complex [4,5,6]. Under physiological light and temperatures, the rates of impairment and repair are balanced, and the steady-state concentration of the PSIIRC maintains continuous photosynthetic activity and growth. Failure to balance the two processes eventually results in cell death [2].
The activity of PSIIRC is highly sensitive to the ambient temperature [7,8,9,10,11]. Short-term temperature elevation was found to enhance the rate of light-induced oxygen evolution [12]. This phenomenon is reflected by a higher flux of electrons that traverse the PSIIRC complex. The increased flux is probably enabled by the enhanced rate of the Q B /Q B H2 turnover because of the increased membrane fluidity under short exposure to elevated temperatures and before membrane lipids saturation takes place [13,14,15,16]. The increased flux has a dual effect: [1] it enhances the probability of localized reactive oxygen species (ROS) generation by PSIIRC and the subsequent impairment and degradation of the D1 subunit; [2] the increased supply of electrons to PSI and thereby to the carbon fixation domain results in an increased probability of ROS generation at that site that can lead to the inhibition of protein synthesis and the consequent decrease of the D1 repair activity [17]. The oxidative stress imposed by the increased flow of electrons to the site of CO 2 fixation is further enhanced by the impairment and reduced activity of Rubisco (ribulose-1,5-bisphosphate carboxylase/oxygenase) at elevated temperatures [18,19,20,21,22]. Furthermore, at elevated temperatures the affinity of Rubisco for oxygen is increased relative to its affinity for CO 2 [19,23,24] resulting increased photorespiration and overwhelming production of deleterious ROS that impair the D1 repair activity [3,25,26].
Short-term (minutes to a few hours) exposure to elevated temperatures results in reversible effects on the photosynthetic activity [27]. However, prolonged exposure (hours to days) to temperatures above the physiological range causes a strong imbalance in the rates of PSIIRC impairment and repair, resulting in the collapse of the photosynthetic machinery and death of the photoautotrophic organism [17,28]. Therefore, photosynthetic organisms have had to develop strategies to sustain their growth in extremely hot (thermophiles), intermediate (mesophiles), and extremely cold (psychrophiles) habitats. Despite this overall plasticity, the individual strains maintain activity over a narrow range of temperatures, typically 65-10uC around their physiological optimum. Hence, prolonged global warming is expected to strongly diminish the PSII activity in mesophilic organisms, resulting in reduced biomass production, unstable ecosystems worldwide, as was already observed in oceanic coral populations, and in disruption of renewable energy and food resources [28,29,30,31,32]. Thus, maintaining a high rate of photosynthesis and biomass formation at elevated, non-physiological temperatures either by increasing the PSIIRC stability or enhancing the rate of D1 repair or both, represent major challenges in acclimatizing photosynthetic mesophiles to global warming [7,9,33,34].
Although many studies have aimed at resolving the role of thylakoids' fatty acid saturation in inducing thermotolerance to photosynthetic organisms [35,36], we have focused on proteins comprising the PSIIRC. Three major observations provided us with new clues for better understanding the strategy of PSIIRC adaptation to elevated temperatures. First, as we previously showed, the temperature dependence of the first Q A 2 RQ B electron transfer rate in mesophiles and thermophiles follows Arrhenius kinetics until it levels off at T o , which turned-out to be within the physiological temperature range of the examined mesophiles and thermophiles [37]. Second, screening the amino acid sequences in the D1 and D2 subunits of many photosynthetic thermophiles and mesophiles revealed consistent variations in two conserved sites: D1-212 and D1-209, within a GxxxG motif at the protein center [37]. More specifically, D1-Ser212 and D1-Ser209 in mesophiles are replaced by Cys and Ala in thermophiles. Third, single mutations at each of the aforementioned sites could increase the value of T o by up to 10uC, in line with the observed values in thermophilic strains [37].
In view of the aforementioned observations and considerations we hypothesized that concomitant D1-Ser209Ala and D1-Ser212Cys (AC) mutations may improve the functional stability of PSIIRC in mesophilic cyanobacteria grown at elevated temperatures. We further hypothesized that enhanced CO 2 concentration should compensate for the increased Rubisco affinity to oxygen at elevated temperatures and thereby reduce the impairment of the D1 repair mechanism by ROS as proposed by Murata et al [17] and Takahashi et al [26,28]. All together, we postulated that combination of double mutation and elevated CO 2 concentration would enable photoautotrophic growth and biomass production at temperatures that cannot be tolerated by the wild type.
The mutations were performed on the DKS strain of Synechocystis sp. PCC 6803 which was selected as control because it retains only the intact wild-type psbAII gene followed by a kanamycin resistance gene (Kmr) [38]. The absence of the other two gene copies, psbAI and III genes, which are replaced by spectinomycin (Smr) and chloramphenicol (Cmr) resistance cartridges, simplifies the interpretation of genetic modifications in the D1 protein subunit.

Results
The D1-S209A/D1-S212C double mutant (AC) grew photoautotrophically and produced biomass in 1% CO 2 atmosphere under continuous illumination at 43uC The growth of AC in liquid cultures under continuous illumination (40 mmol photons m 22 s 21 ) was monitored and compared with that of DKS at 30, 38, 40 and 43uC. There was no significant difference in the growth rates and pigmentation between the two strains when grown at 30uC (Fig. 1) under normal air bubbling or under 1% CO 2 . When incubated at 38 and 40uC the growth of the AC mutant was slightly slower (by 5 and 10% respectively), in comparison to the control strain that showed a much slower growth at both temperatures (by,25 and 29%, respectively) ( Fig. 1). However, when grown at 43uC, DKS biomass slightly increased in the first 3 days of incubation followed by complete bleaching after 4 days regardless of the CO 2 content, whereas the AC mutant, when grown at 43uC and under 1% CO 2 , exhibited a growth rate that was only 25% lower than the one measured at 30uC, resulting in a 15-fold increase in OD 730 (Fig. 1) and an almost 10-fold increase in the dry biomass after 7 days of incubation ( Fig. 2A).
Importantly, when grown at the same temperature but under stirring (no CO 2 supplement) the AC biomass increased at relatively slow pace and started to level off at the fourth day. Nevertheless, when transferred back to 30uC the growth was regained (Fig. 2B). In contrast, under the same conditions, the DKS cultures leveled off after three days of slow growth and could not recover when transferred back to 30uC (Fig. 2B). Thus, although the AC mutant can survive a prolonged incubation at 43uC, normal growth at such temperature requires CO 2 supplement. Notably, the growth of wild type Synechocystis sp. PCC 6803 (having all three psbA genes) at 43uC and 1% CO 2 showed similar kinetics to that of DKS with somewhat higher growth during the first three days of incubation followed by a slower decay from day 4 (Fig. 2C). The chlorophyll (Chl) content in DKS and AC increased by 10-fold after 7 days of incubation at 30uC and 1% CO 2 . However, when grown at 43uC, the Chl content sharply declined in DKS after 3 days, whereas that in the double mutant increased throughout the entire period of incubation to almost 3 times its initial value (Fig. 2D).
The D1 and Rubisco proteins exhibited higher steady-state levels in the AC mutant compared with DKS, under continuous illumination, 1% CO 2

and 43uC
At 30uC the D1 and Rubisco proteins in both strains maintained a constant steady-state concentration throughout 7 days of incubation under continuous illumination (data not shown). However, upon incubation at 43uC, the D1 and Rubisco proteins content declined ,2 and 2.2 times faster, respectively, in DKS compared with the AC (Fig. 3). Thus, on the fifth day of incubation the D1 content dropped to ,4% of its initial value (Fig. 3) and that of Rubisco reached non-significant levels for DKS. In contrast, at the 6 th day of incubation, the AC maintained ,15 and ,8% of its initial levels of D1 and Rubisco, respectively. At the 7 th day of incubation, no traces of D1 were detected in DKS, while ,8% of its initial level was found in AC (Fig. 3).
Temperature dependence of the photosynthetic oxygen evolution in cells acclimated at 30 and 43uC under 1% CO 2 Figure 4A illustrates the temperature dependence of oxygen evolution by the strains grown for three days at 30 or 43uC. Briefly, samples were taken from each culture and then incubated for 10 minutes at the measuring temperature of 30 or 43uC (white and gray bars, respectively). Oxygen evolution by DKS and AC grown at 30uC reached similar values when measured at 30uC (659 and 615 mmol O 2 mgChl 21 h 21 , respectively) with a twofold increase when measured at 43uC (1230 and 1290 mmol O 2 mgChl 21 h 21 , respectively). A remarkably different behavior was observed for DKS and AC grown at 43uC. After 3 days, the oxygen evolution declined in DKS to 305 and 485 mmol O 2 mgChl 21 h 21 when measured at 30 and 43uC, respectively. The oxygen evolution by the AC mutant was markedly higher after three days of growth at 43uC: 440 and 920 mmol O 2 mgChl 21 h 21 when measured at 30 and 43uC, respectively. All together, the rate of oxygen evolution by AC grown at 43uC and measured at 43uC (920 mmol O 2 mg Chl 21 h 21 ) was significantly higher than the activity of the DKS grown at 43uC and measured at 43uC (485 mmol O 2 mg Chl 21 h 21 ), suggesting that the PSII activity in the AC mutant underwent acclimation and optimization at 43uC that allowed for retaining activity similar to the one measured under the short term exposure of cells to the elevated temperature, while the control strain did not show such capacity. After 7 days of incubation at 43uC the oxygen evolution activity in the AC mutant approached 20% of its initial value. However, that of DKS dropped down to zero already at the fifth day of incubation (data not shown).
Temperature dependence of the Q A 2 RQ B electron transfer rate The Q A 2 RQ B electron transfer rate in DKS grown at 30uC leveled off already at 26uC at a value of ,3400 s 21 (Fig. 4B, closed circles) as previously reported [37]. The AC leveled off at a higher temperature of 35uC, though it reached only 2500 s 21 (Fig. 4B, closed squares). When grown at 43uC for three days, the small fraction of DKS with active PSIIRC (Fig. S1) exhibited a low rate (,1300 s 21 ) of Q A 2 RQ B ET at 30uC that increased to ,2100 s 21 at 43uC (Fig. 4B, open circles). In contrast, after .24 h of acclimation at 43uC, the Q A 2 RQ B electron transfer rate in AC increased to ,3800 s 21 when measured at 43uC and continued to rise exponentially reaching a rate of 5200 s 21 at 48uC (Fig. 4B, open squares and Fig. S3).
The dissimilarity in the temperature response of the Q A 2 RQ B rate constants between the AC mutant grown at 43uC, AC grown at 30uC, and DKS grown at 30uC, reflects upon their different enthalpies and entropies of activation for the electron transfer reaction ( Fig. 4C and Table 1). More specifically, the activation parameters of the AC mutant acclimated at 43uC (DH { = 7885 cal mol 21 , DS { = 217.2 cal mol 21 K 21 ) become close to those of the thermophilic T. elongatus grown at 56uC (DH { = 7142 cal mol 21 , DS { = 220.7 cal mol 21 K 21 ) and are markedly different from those measured for the AC and DKS grown at 30uC (Table 1).

D1 Degradation and Repair
The maintenance of electron transfer and oxygen evolution activity in the AC can be attributed to the higher stability and/or to the enhanced repair rate of their PSIIRCs. To decipher the predominant contribution, we monitored the D1 content as a function of time in cells exposed to high light irradiances, either in the presence or in the absence of the protein synthesis inhibitor lincomycin ( Fig. 5A and B).
The D1 protein content in both the AC and DKS decayed during incubation at 43uC and under high light conditions. However, the decay of the D1 protein content in the mutant was markedly slower, reaching 20% of its initial value after 6 h of illumination (Fig. 5A). Lincomycin markedly accelerated the decay of D1 content in both DKS and AC cultures. Nevertheless, after 6 hours the D1 protein content in the AC reached ,5-10% of the initial value whereas in DKS it dropped to zero (Fig. 5A). Moreover, the difference between the levels of D1 in the absence and presence of lincomycin are ,2 times larger for the AC compared with DKS (Fig. 5A, insert), suggesting that enhanced PSIIRC repair in the AC plays a key role in maintaining high photosynthetic activity at elevated temperatures under elevated CO 2 concentration.
Under the same conditions, the PSII activity was assessed by measuring the light-saturated steady-state rate of oxygen evolution in the presence of artificial electron acceptors (Fig. 5B). The oxygen evolving activity for both strains decreased during the treatment but the decrease was significantly more pronounced in DKS. Thus, after 6 h of exposure, the DKS and AC maintained ,20 and 34% of their initial activity, respectively. However, upon adding lincomycin and thereby preventing PSIIRC repair, DKS showed no oxygen evolution already after 4 h of incubation under high light while the AC mutant maintained 35% of its initial oxygen evolution activity at that time. After 6 h of illumination in the presence of lincomycin, the oxygen evolving activity of the AC mutant also dropped to a non-detectable value.

Conformational changes and related energies that involve inter helical H-bonding of the mutated residues in PSIIRC
In silico introduction of the double mutation D1-A209S/D1-C212S to the resolved structure of T. elongatus provides insight into the structural and energetic differences between PSIIRC of wild-type Synechocystis sp. PCC 6803 and the AC mutant. Hereafter, we will refer to the structure of PSIIRC from T. elongatus [39,40,41] as representing the putative structure of the AC, and the one obtained by in silico D1-A209S/D1-C212S double mutation, as representing the structure of DKS Synechocystis sp. PCC 6803.
According to the energy minimized structure, D1-Cys212Sc enters into an H-bond with the backbone carbonyl of D2-Met271. The DFT computations show that this conformation (Fig. 6B, GS = conf19) is at an energy minimum. The in silico mutation of D1-Cys212 to D1-Ser212 enables a similar ground state conformation, although the D1-Ser212Oc…D2-Met271 H-bond is longer than the D1-Cys212Oc…D2-Met271 H-bond (2.63 and 2.17 Å , respectively). The energy for this conformation is ,1 kcal smaller than the energy for conf1. The second low energy for this in silico mutant is one in which D1-Ser212Oc is H-bonded to the backbone carbonyl of D2-Gly207 at 2.2 Å (Fig. 6A, conf2). The DFT computations show that although the energy of conf2 is at a local minimum, it is ,5.75 kcal mol 21 higher than the energy acquired by conf1. The published structure of T. elongatus does not allow for conf2 because the distance between D1-Cys212Sc and the backbone carbonyl of D2-Gly207 is too short (1.7 Å ). Binding A. DKS and AC cells (circles and squares, respectively) that were grown in liquid medium for 7 days at 30u or 43uC (open and filled symbols, respectively) under 1% CO 2 . B. DKS and AC cultured in liquid medium under stirring and normal air. The strains were incubated at 43uC for 5 days (filled symbols) and then transferred to 30uC (open symbols) for 5 days to test their viability. C. Wild-type (triangles), DKS (circles) and AC cells (squares) were grown at 43uC under 1% CO 2 . D. Chlorophyll content in wild type, DKS and AC cells (triangles, circles and squares, respectively) that have been transferred to 43uC and 1% CO 2 after 3-4 days incubation at 30uC and 1% CO 2 . The values represent the mean 6 SD of three independent experiments. doi:10.1371/journal.pone.0028389.g002 at a slightly different geometry (Fig. 6B, conf29) is possible at an energy that is 6.75 kcal mol 21 higher than the conf1 energy for D1-212Cys.
Molecular dynamics simulations reveal substantial changes in the relative geometries and energies of D helices in PSIIRC during conf1(conf19)Rconf2(conf29) transitions The molecular dynamics simulations have provided insight into the interaction potential of the D helices of the D1 and D2 proteins (Fig. 7). Two interhelical hydrogen bonds were formed between the D helices of the D1 and D2 proteins of T. elongatus; the reciprocal bonding involved the D1-Cys212Sc donating hydrogen to D2-Gly207O along with D1-Gly208O accepting hydrogen from D2-Cys211Sc (hydrogen bonding energies were 3.661.2 and 3.561.3 kcal mol 21 , respectively, and hydrogen bond lengths were 2.160.2 and 2.260.2 Å , respectively). In Synechocystis sp. PCC6803, the two analogous interhelical bonds D1-Ser212Oc to D2-Gly207O and D1-Gly208O from D2-Cys211Sc had hydrogen-bonding energies 5.460.9 and 3.961.3 kcal mol 21 , respectively, and the hydrogen bond lengths were 1.860.2 and 2.160.2 Å , respectively). An additional hydrogen bond was observed between D1-Ser209Oc and D2-Ile204O (bond length 2.060.2 Å , bond energy 4.261.4 kcal mol 21 ). As a result of this change in the inter-helical hydrogen bond network, the D helices of the modeled mesophilic D1 and D2 proteins take on a conformation that is different from the thermophilic one. Namely, the average distance of the Ca atoms at the helix-helix interface for (Fig. 7). Thus, to form the new H-bond there is a need for a more relaxed environment. Also, the helix-helix contact area in Synechocystis hardly increased during the 20 ns simulation, whereas that of T. elongatus increased by The effect of temperature on lipids and fatty acid composition in the thylakoid membrane of DKS and AC It is well known that the photosynthetic membrane lipids present increased saturation with increasing temperature [42,43,44] in different strains of cyanobacteria and higher organisms. Such a change may affect the energy required for the D1 and D2 subunits to undergo the conformational changes required to enable the gating of the Q A 2 RQ B electron transfer which was observed for the AC strain following .24 h of incubation at 43uC. Hence, we followed the timeline for the increased saturation of the thylakoid membranes in DKS and AC and compared it with the observed changes in the thermodynamic parameters for the electron transfer (Table S1 and Table 2). In both DKS and AC the double bond index (DBI) value decreased   (Table S1). Importantly, the fatty acid composition in DKS that grew at 30uC was found to be similar to the one previously reported for Synechocystis sp. PCC 6803 [36,45], with a relatively high percentage of polyenic acids that provided a fairly high DBI value (96.8). The DBI value in the AC mutant that was grown at 30uC was higher (104.6), mainly because of an increase of linoleic (18:2) and a decrease of palmitic (16:0) fatty acid contents, reflecting a more fluid membrane environment at standard growing conditions.
Incubation of DKS and AC mutant at 43uC for .24 h was sufficient to decrease their DBI values to 78.9 and 86.2, respectively. A similar but somewhat lower value was reported by others for Synechocystis sp. PCC 6803 grown at 38uC [36]. The lower DBI is mainly the result of elevated levels of saturated palmitic (16:0) and a decrease of both unsaturated palmitoleic (16:1) and polyunsaturated gamma-linolenic (18:3 n6) fatty acid contents. Even though both strains showed a lower DBI after growth at 43uC, the AC exhibited a higher value than DKS, which reflects a higher fluidity in the thylakoid membrane.

Discussion
The goal of our research is to genetically engineer novel, mesophilic cyanobacteria that retain prolonged photosynthetic activity and biomass production under continuous illumination at elevated temperatures to which the ''wild type'' cannot adjust. We hypothesized that PSIIRC should be a major target for such engineering and that genetic differences between the PSIIRC in thermophiles and mesophiles provide clues for new strategies. Following sequence alignment analysis, we found two sites within a GxxxG-like motif in the D1 protein subunit that are consistently occupied by different residues in thermophiles and mesophiles. In a previous study, we focused on the effect of single mutation on the electron transfer dynamic with respect to the ambient temperature [37]. Here, we aimed at deciphering the mutations effect on the bacterium viability at above its physiological temperature. Therefore, a double mutation was performed in the GxxxG-like motif, which made it identical to the one found in the thermophilic cyanobacterium T. elongatus. Indeed, only a slight decrease in the growth rate was shown by the AC mutant at 38 and 40uC, temperatures at which the DKS grew at a much slower rate. More important, the double mutant presented prolonged photosynthetic activity and biomass growth during 7 days incubation at 43uC far above the growth temperature of wild-type Synechocystis sp. PCC 6803 or the DKS strain that was used as control, but only under elevated CO 2 conditions (1%). The DKS completely perished already after 6 days of incubation under identical growth conditions.
Notably, when DKS and AC mutant cells were grown at 43uC but at lower CO 2 supply both strains showed slower growth.  The temperature dependence of Q A 2 RQ B ET in DKS, AC and T. elongatus (Te) cells grown at the indicated temperature was measured as described in Materials and Methods. The activation parameters (DS { DH { ) were calculated from the Eyring plots (LN(k/T) vs. 1/T) in Figure 4C that represents the mean of 10-12 independent experiments. doi:10.1371/journal.pone.0028389.t001 However, in all tested CO 2 supply conditions (stirring, air bubbling and 1% CO 2 ) the AC mutant showed higher growth than the DKS (Fig. S2). Moreover, when the CO 2 concentration was increased to 3% the AC mutant showed growth even at 45uC, while the DKS grew only for 1 day (Fig. S2). The beneficial effect of the increased CO 2 concentrations could be attributed to two mechanisms. First, the increased affinity of Rubisco to oxygen at elevated temperatures decreases CO 2 fixation, which can subsequently increase the generation of ROS [19,23,24]. Furthermore, the reduction of molecular oxygen may lead into the formation of H 2 O 2 that inhibits the synthesis of PSII proteins and primarily of the D1 protein [3,25,26]. Second, CO 2 solubility drops by estimated 30% upon increasing temperature from 30 to 45uC, further increasing the probability of Rubisco to react with oxygen. Thus, the increased CO 2 concentration should help decreasing the probability of oxygen binding to the Rubisco and consequently the formation of radicals that could impair the D1 repair activity.
The longevity of the double mutant at the high temperature and elevated CO 2 conditions, appears to be correlated with the markedly slower decline of the PSIIRC activity monitored by the levels of electron transfer and oxygen evolution activities as well as the level of the D1 protein subunit and Rubisco ( Fig. 3 and 4). The relatively larger amplitude of the fastest component in the fluorescence decay curves (Fig. S1) provides an additional support for the enhanced functional stability of the PSIIRC in the AC mutant. Importantly, only when the level of D1 and Rubisco reach ,20% and 5-10%, respectively, of the content found at room temperature, the rate of biomass growth slowed down ( Fig. 1 and  3). This finding is in agreement with previous studies reporting that photosynthetic organisms with less than 50% of their steady state D1 level can still maintain the same rate of biomass formation because of the excess capacity of light-induced electron transfer in PSII [46,47]. Nevertheless, maintaining a normal growth rate, even at 10-20% Rubisco protein level, is an interesting finding that requires further investigation.
To elucidate the contributions of enhanced PSIIRC photo thermal stability and the rate of repair to the subunit steady-state concentration, we followed the decay of the D1 protein content and the concomitant PSII activity during exposure to high irradiance at 43uC in the presence and absence of lincomycin. The difference between the respective pair of measurements (represented by full and empty symbols, respectively, in Fig. 5) represents the contribution of D1 and PSIIRC repair to the measured quantity, as demonstrated by the insert in Fig. 5A. In both strains, the D1 protein content and oxygen evolution activity exhibited a decrease during the exposure and this decrease was enhanced in the presence of lincomycin ( Fig. 5A and B). Nevertheless, the AC mutant showed higher content of D1 protein and higher activity than the DKS during the course of the treatment both in the absence and presence of lincomycin. These results suggest that both the stability (Fig. 5A and B) and repair (Fig. 5A, insert) of the D1 protein are enhanced in the double mutant compared with DKS. The increased rate of repair in the AC mutant may reflect upon increasing rate of PSIIRC refolding with the mutated D1 protein or, upon higher photo/thermal stability of the mutated psbA mRNA. This question is currently being explored in our lab.
We previously showed that the rate of Q A 2 RQ B electron transfer levels off at T o , which was defined as the optimal temperature for electron transfer, and it was found to be within the physiological range of the examined strain, 26uC for DKS and ,59uC for T. elongatus [37]. Importantly, the rate of Q A 2 RQ B electron transfer at T o is similar for mesophiles and thermophiles, reaching a value of 3000-4000 s 21 . Apparently, at this rate the balance between PSIIRC degradation and repair, as well as other enzymatic processes that comprise photosynthetic charge separation and carbon fixation, is optimal. This observation is in line with the corresponding hypothesis, which suggests that psychrophilic (cold-adapted), mesophilic, thermophilic and hyperthermophilic homologous enzymes have comparable catalytic efficiencies (indicated by kcat/K M ) at their respective optimal temperatures because optimal activity requires a certain degree of conformational flexibility in the active site [48,49]. Hence, one can reasonably assume that for inducing thermotolerance and biomass generation, as has been sought in this study, Q A 2 RQ B electron transfer needs to reach and maintain a value of ,3500 s 21 when growing at the elevated target temperature. Following the transition state theory, the rate of Q A 2 RQ B electron transfer is given by (k et ) = (k b T/h)exp(2DE { /RT), where k b and h represent the Boltzmann and Max-Plank constants, respectively, and DE { = DH { 2TDS { is the activation energy [37]. Thus, maintenance of similar electron transfer rates at the physiological optima of mesophiles and thermotolerant/thermophiles can be achieved  Table 1). However, after $24 h of acclimation at 43uC, DH { and DS { of the AC are changed by +3 kcal mol 21 and 29.8 cal mol 21 K 21 , respectively (Table 1). With these new values, k et is equal to 3500 s 21 only when reaching T = 42uC, which becomes the new optimal temperature (T o ) for PSIIRC activity. Likewise, provided that the optimal rate for electron transfer in T. elongatus is also ,3500 s 21 , (Fig. 4C) acclimation at 43uC results in DH { and DS { values that provide the optimal k et = 3500 s 21 at T o = 52uC, whereas growth at 56uC acclimatizes the strain for k et = 3500 s 21 at T o = 57uC. Cumulatively, the change in thermodynamic data after .24 h of incubation combined with the DFT calculations and dynamic simulations suggest that the PSIIRC ground state is more stable by at least 2 kcal mol 21 in the AC (Fig. 6). The change in the Q A 2 RQ B electron transfer rate and the related thermodynamic parameters, appears to occur after .24 h of incubation at 43uC under 1% CO 2 (Fig. S3), this appears to correlate with the timeline for increased saturation of the membrane lipids in Synechocystis, already reported by others [36,44]. Loll et al (2007) identified six MGDG, four DGDG, three SQDG, one PG as well as three b-DM molecules per PSIIRC monomer. The most recent structure [50] reports 1 more DGDG (5 in total), 1 more SQDG (4 in total) and 4 more PG (5 in total). According to Sakurai et al (2006), only MGDG and DGDG contain the 18:3 fatty acids that undergo saturation and consequently rigidification, similarly to the trend observed here upon incubating DKS and AC at 43uC (Table 2). Moreover, the percentage of these lipids in thylakoid membranes and isolated PSIIRC is practically the same [45,51]. Hence, the MGDG and DGDG, which interact with PSIIRC, may experience the aforementioned 18:3R18:1 and 16:1R16:0 transition under prolonged temperature elevation. The resulting rigidification of the lipids interacting with D1/D2 proteins should modify any process involving a protein conformational change that requires displacement of these lipids. The possibility of such conformational changes in the studied strains is discussed in the following text.
The activation energy for the Q A 2 RQ B electron transfer process is indicative of dissociation of 1-2 H-bonds [37] upon transferring from a ground to a transition state during the Q A 2 RQ B electron transfer. The molecular dynamic simulations (Fig. 7) indicate that alternating H-bonding association or dissociation of the D1 and D2 protein subunits involve an average 0.7-1 Å expansion of the D1/D2 complex in the AC but not in DKS. At a high level of desaturation (e.g. following incubation at 30uC), reflected by a high DBI value (Table 2), the flexible lipid bed provides similar low resistance to conformational changes of the D1/D2 in DKS and the AC. The higher DBI value for AC compared to DKS, possibly reflect upon the larger membrane flexibility that is needed to allow the larger PSIIRC expansion during electron transfer. However, following acclimation at elevated temperatures, the increased saturation makes the lipid environment of the PSIIRC more rigid than at room temperature and the conformational change in the AC needs to overcome an additional energy barrier accounting for part of the markedly increased DH { . The enhanced rigidity of the lipids should decrease the entropy of the Q A 2 Q B state and therefore the value of DS { for the Q A 2 RQ B transition is reduced. The experimental values of the thermodynamic parameters (Table 1) fit the putative conformational changes and the involved energies that are presented in Figure 6, following the DFT computations. Thus the thermodynamic parameters show both stabilization of the ground state and enhanced conformational rigidity of the AC after acclimation at the elevated temperature. Such enhancement is often proposed to account for the increased thermal stability of thermophilic enzymes compared with mesophiles [52].
The significance of fatty acid saturation in regulating enzymatic reactions that depend on protein conformational changes, was explored using lipid specificity for the reconstitution of wellcoupled ATPase proteoliposomes [53]. Different approaches attempted to decipher the role of saturation/desaturation in adapting the photosynthetic machinery to temperature changes (for a recent review see: Allakhverdiev et al, 2008). However, a recent study claims that membrane protein stability does not depend on the lipid composition of the membrane [54] and arguments against lipid saturation per se as a regulator of thermotolerance were raised [36]. The present study supports the possibility that both lipid saturation/rigidification and point mutations that modify the protein structure at the transition state may be required for controlling the activation energy for the ratedetermining electron transfer and for inducing thermotolerance to the PSIIRC. Namely, as the growth temperature is increased to 43uC, DH { and DS { for the AC mutant should be changed for maintaining k et at 3000-3500 s 21 . To that end the DBI value decreases to counteract the increased fluidity of the membrane at elevated temperatures.
At the same time our study suggests that in addition to the enhanced functional stabilization of the PSIIRC complex, there is a need to attenuate the impairment of the D1 repair machinery at elevated temperature, possibly by providing more CO 2 as a sink for the accumulation of redox equivalents.
Furthermore, the sequence, structure, and thermodynamic similarities between the PSIIRC in the AC and T. elongatus suggest that the AC double mutation could account for the thermophilicity of existing strains and could provide a first step for adapting mesophilic photosynthetic organisms.  were measured spectroscopically by sampling aliquots from the liquid cultures every 24 hours as previously described [55]. In some experiments, the protein synthesis inhibitor lincomycin was added to the cell suspension (final concentration 200 mg ml 21 ) at the start of the treatment [6]. Thermosynechococcus elongatus BP-1 [56] cells were grown at 30, 43 or 56uC under a light intensity of 40 mmol photons m 22 s 21 white light in liquid BG-11 medium.

Isolation of proteins and Western blot analysis
Thylakoid membranes were prepared as previously described [58]. Whole-cell extract samples were obtained from the same preparation and used for Rubisco Western blot analysis. Proteins were solubilized in sample buffer (0.5 M Tris-HCl pH 6.8, 1% SDS, 24% glycerol, 4% b-mercaptoethanol, 0.001% (w/v) Bromophenol blue), incubated at room temperature for 1 hour and then separated on 12.5% SDS-PAGE. The equivalent of 1 mg of chlorophyll was loaded in each well. Proteins were electroblotted to PVDF (Hybond-P, Amersham, UK) using a BioRad Mini Transblot Cell (Bio-Rad, USA). The immuno detection was carried out using a chemiluminescence kit (SuperSignal West Pico, Pierce, USA). Antibodies against D1 and RuBisCO large subunit proteins were purchased from Agrisera (Umeå, Sweden). For quantification the bands from the scanned blots were quantified by integrating variable pixel intensities using the ImageJ software [59] and comparing them to a dilution series of samples (Fig. S4).

Oxygen evolution rate
Light-saturated (1500 mmol photons m 22 s 21 ) steady-state rate of oxygen evolution was measured using a Clark-type platinum silver electrode in a thermostated glass cuvette (Hansatech, Inc., England). Cells were harvested at the indicated times by centrifugation and re-suspended in fresh media to a final concentration of 10 mg chlorophyll ml 21 . The cells were kept in the dark for 10 min at the measuring temperature before being measured. A total of 3 ml of cells was added to the electrode chamber and the artificial electron acceptors 2,5-dimethyl-pbenzoquinone (DMBQ) (0.5 mM final concentration) and ferricyanide (1 mM final concentration) were added just before measurement. The temperature was maintained at 30 or 43uC with a circulating water bath. Calibrations at both measuring temperatures were made to adjust the sensitivity of the electrode.

Extraction, preparation, and analysis of thylakoid membrane lipids and their fatty acids
Thylakoid membranes were isolated as described [60]. Thylakoid lipid extraction was carried out according to [61]. Lipid classes were separated by thin-layer chromatography on silica gel (Merck 5721) with chloroform/methanol/acetic acid/water (90:9:12:2 v/v) as the developing solvent. TLC plates were sprayed with 0.05% solution of primuline. Lipid spots were visualized with a hand-held UV lamp VL-6.M (Vilber Loumart, France) in order to assess the quality of separation or to mark lipids for scraping and extraction. The plates were then scanned and the spots were quantified by integrating variable pixel intensities on ImageJ software [59] and comparing them to standard curves. Lipid spots on the TLC plates were scrapped-off and later subjected to transmethylation [62]. The esterified fatty acids were analyzed with a gas-liquid chromatograph (HRGC 5300, Carlo Erba instruments) equipped with a hydrogen flame-ionization detector. The double bond index (DBI) was calculated by dividing the sum of the percentages of the unsaturated fatty acids, each multiplied by the number of its double bonds, by 100.

Flash Fluorescence Measurements
The rate constant for the first Q A to Q B electron transfer was assessed as recently described [37]. The percentage of active PSIIRC was deduced from the relative contribution of the fast component to the overall decay of the intact cell fluorescence as previously described [27,37,63].

DFT calculation and protein dynamics
PSII structural coordinates at 2.9 Å resolution (PDB ID: 3bz1, [40]) were downloaded from the PDB and hydrogen atoms were added using REDUCE [64]. Relevant parts of helices D and E from D1 (aa 203-217, aa 268-281) and D2 (aa 202-216, aa 264-277) were extracted and capped with hydrogen atoms at the Nand C-termini; only this four-helix bundle was considered in the following calculations since D1-C212 does not interact with other parts of the protein. The possible H-bonds formed by this cysteine's side-chain were determined by optimizing its Sc, Hc, 1Hb, and 2Hb atoms (all other atoms were frozen) at 12 initial side-chain conformations, defined by combinations of x1 (180u, +60u, 260u) and x2 (180u, +90u, 0u, 290u) dihedral angles. These 12 optimizations resulted in a global minimum and several local minima. A similar procedure was applied for D1-C212S by adjusting the bond lengths and angles of its Oc. For an accurate calculation of the H-bond geometries and energies, while considering the four surrounding helices, a quantum mechanics/ molecular mechanics (QM/MM) hybrid method was utilized. Specifically, a two-layer ONIOM approach [65], as implemented in GAUSSIAN 03 [66], was applied. The QM layer for the D1-C212 (and D1-S212) optimizations consisted of the following atoms: Ca, Ha, Cb, (2X)Hb, Sc (or Oc), and Hc of D1-C212; Ca, (2X)Ha, C, and O of D2-G207; N and H of D2-A208 and D2-C211; Ca, Ha, C, and O of D2-M271; N and H of D2-L272; N, Cd, (2X)Hd, Cc, and (2X)Hc of D2-P275. The DFT-B3LYP/6-31+G** level of theory was used for the QM layer. The hybrid B3LYP [67,68] functional was used as an intermediate-level means of including electron correlation, since it has been shown to produce accurate geometries compared with protein structures [69] and realistic H-bond energies in ligand protein systems [70].
The double-f 6-31+G** basis set contains polarization functions on all atoms and diffuse functions on heavy atoms. The MM layer, consisting of all other atoms, was modeled by the Amber force field [71], and its partial charges were able to polarize the QM wavefunction (electronic embedding).
We did not find significant differences in the results when using the most recent PSII crystal structure [50].
Molecular dynamics simulations. The simulations were performed with YASARA [72]. The 3D structures of the D helices of the D1 and D2 proteins in thermophilic Thermosynechococcus elongatus were obtained from the crystal structure of cyanobacterial PSII (PDB ID: 3BZ1) [40] D1(196-221): PFHQLG VAGVF-GGALFCAMHGSLVTS; D2(195-219): PFHMMGVAGVLG-GALLCAIHGATVE. The in silico mutagenesis of two amino acid residues in the D1 sequence -A209S and C212S (underlined) mimicked the native structure of the mesophilic Synechocystis sp. PCC6803. The D2 sequence of the D helix differs only at the site 204 -the thermophile valin was replaced for isoleucine in the mesophile. Both models of the mesophilic and thermophilic D helices of the D1 and D2 proteins were placed in the periodic boundary simulation boxes that were 1 nm larger than the peptides along all three axes. After hydrogen atoms were added to the helices according to basic chemistry rules and the currently selected pH = 8.0, the boxes were filled with TIP3P water, and sodium atoms were iteratively placed at the coordinates with the lowest electrostatic potential until the cell was neutral. Molecular dynamics simulations were run using a multiple time step of 1.25 fs for intra-molecular and 2.5 fs for intermolecular forces. To remove bumps and to correct the covalent geometry, the structures were energy-minimized with the Yamber3 force field [73] using a 8.0 Å force cutoff and the Particle Mesh Ewald algorithm [74] to treat long-range electrostatic interactions. 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 0.012 kcal mol 21 during 200 steps. The simulations were then run at 300 K at a constant pressure (NPT ensemble) to account for volume changes due to fluctuations of peptides in the solution. The simulations were run for a total time of 20 ns. Molecular graphics were created with YASARA [72] and Persistence of Vision (TM) Raytracer (http://www.povray.org/). Table S1 Thylakoid membrane lipids composition in Synechocystis sp. PCC6803 and Thermosynechococcus elongatus grown at the indicated temperatures. Lipids were extracted from thylakoid membranes. Lipid classes were separated by thin-layer chromatography, sprayed with a primuline solution and quantified by integrating variable pixel intensities of the scanned plates. The values represent the mean of three independent experiments. The deviation of values was within 6 2%. Monogalactosyldiacylglycerol, MGDG; digalactosyldiacylglycerol, DGDG; phosphatidylglycerol, PG; sulfoquinosyldiacylglycerol, SQDG. (DOC)