Natural Variants of Photosystem II Subunit D1 Tune Photochemical Fitness to Solar Intensity*

Background: Cyanobacteria use multiple PSII-D1 isoforms to adapt to environmental conditions. Results: D1:2 achieves higher quantum efficiency of water oxidation and biomass accumulation rate at high light versus D1:1; the latter is more efficient at low light due to less charge recombination. Conclusion: A functional advantage for D1:1 is revealed for the first time. Significance: Improved photochemical efficiency at low light suggests an evolutionary advantage to retain D1:1. Photosystem II (PSII) is composed of six core polypeptides that make up the minimal unit capable of performing the primary photochemistry of light-driven charge separation and water oxidation in all oxygenic phototrophs. The D1 subunit of this complex contains most of the ligating amino acid residues for the Mn4CaO5 core of the water-oxidizing complex (WOC). Most cyanobacteria have 3–5 copies of the psbA gene coding for at least two isoforms of D1, whereas algae and plants have only one isoform. Synechococcus elongatus PCC 7942 contains two D1 isoforms; D1:1 is expressed under low light conditions, and D1:2 is up-regulated in high light or stress conditions. Using a heterologous psbA expression system in the green alga Chlamydomonas reinhardtii, we have measured growth rate, WOC cycle efficiency, and O2 yield as a function of D1:1, D1:2, or the native algal D1 isoform. D1:1-PSII cells outcompete D1:2-PSII cells and accumulate more biomass in light-limiting conditions. However, D1:2-PSII cells easily outcompete D1:1-PSII cells at high light intensities. The native C. reinhardtii-PSII WOC cycles less efficiently at all light intensities and produces less O2 than either cyanobacterial D1 isoform. D1:2-PSII makes more O2 per saturating flash than D1:1-PSII, but it exhibits lower WOC cycling efficiency at low light intensities due to a 40% faster charge recombination rate in the S3 state. These functional advantages of D1:1-PSII and D1:2-PSII at low and high light regimes, respectively, can be explained by differences in predicted redox potentials of PSII electron acceptors that control kinetic performance.

Photosystem II (PSII) is composed of six core polypeptides that make up the minimal unit capable of performing the primary photochemistry of light-driven charge separation and water oxidation in all oxygenic phototrophs. The D1 subunit of this complex contains most of the ligating amino acid residues for the Mn 4 CaO 5 core of the water-oxidizing complex (WOC). Most cyanobacteria have 3-5 copies of the psbA gene coding for at least two isoforms of D1, whereas algae and plants have only one isoform. Synechococcus elongatus PCC 7942 contains two D1 isoforms; D1:1 is expressed under low light conditions, and D1:2 is up-regulated in high light or stress conditions. Using a heterologous psbA expression system in the green alga Chlamydomonas reinhardtii, we have measured growth rate, WOC cycle efficiency, and O 2 yield as a function of D1:1, D1:2, or the native algal D1 isoform. D1:1-PSII cells outcompete D1:2-PSII cells and accumulate more biomass in light-limiting conditions. However, D1:2-PSII cells easily outcompete D1:1-PSII cells at high light intensities. The native C. reinhardtii-PSII WOC cycles less efficiently at all light intensities and produces less O 2 than either cyanobacterial D1 isoform. D1:2-PSII makes more O 2 per saturating flash than D1:1-PSII, but it exhibits lower WOC cycling efficiency at low light intensities due to a 40% faster charge recombination rate in the S 3 state. These functional advantages of D1:1-PSII and D1:2-PSII at low and high light regimes, respectively, can be explained by differences in predicted redox potentials of PSII electron acceptors that control kinetic performance.
Photosystem II (PSII) 3 is nature's sole enzymatic solution to the challenging chemistry of water oxidation (1,2). In PSII, solar energy is converted to chemical energy by splitting two molecules of water in the Mn 4 CaO 5 (3) water-oxidizing complex (WOC) and reducing two plastoquinone (PQ) molecules to plastoquinols (H 2 PQ). The oxygen (O 2 ) by-product from this reaction accounts for nearly all the O 2 in the atmosphere (2). Regulation of solar energy conversion to these products in PSII is essential for cell survival and is controlled at multiple levels.
The D1 protein of PSII is one of six core polypeptides that make up the minimal unit performing the primary photochemistry of light-driven charge separation and water oxidation in all oxygenic phototrophs. It provides most of the ligating amino acid residues for the WOC manganese core as well as binding pockets for the P 680 chlorophyll-a (Chl-a) special pair, pheophytin (Pheo), and the secondary PQ acceptor. Because of its proximity to the WOC, it is frequently damaged by reactive oxygen species and is turned over faster than any other PSII protein subunit (4).
In Synechococcus 7942 and Thermosynechococcus BP-1, PSII centers containing the D1:2 isoform have been shown to have higher O 2 evolution rates (25), faster photoautotrophic growth rates (26), more rapid tyrosine-Z (Y Z ) donation to P 680 ϩ (24), and less sensitivity to photoinhibition (14, 20, 26 -29). However, "low light" D1:1 remains the dominant isoform in many cyanobacteria. Why have cyanobacteria maintained this seemingly inferior D1 isoform over billions of years of evolution? We hypothesized that under very low light intensities, D1:1 may have a functional advantage over D1:2. By expressing cyanobacterial D1 isoforms in a model green alga, we avoided background fluorescence interferences commonly encountered in experiments with cyanobacteria (30) and were thus able to quantitatively compare the D1:1, D1:2, and algal isoforms both in vivo and in vitro. Here, we show that when compared with the D1:2 and algal isoforms, D1:1 has higher WOC cycling efficiency at low light intensities, which is supported by the observation of a more stable S 3 WOC intermediate. We attribute this improved efficiency to fewer competing PSII-cyclic electron transfers from native cofactors (including Q A Ϫ and cyt b 559 ).

EXPERIMENTAL PROCEDURES
Mutant Construction-Heterologous psbA genes from Synechococcus 7942 were expressed in the chloroplast genome of the model green alga Chlamydomonas reinhardtii (hereafter Chlamydomonas). Chlamydomonas contains two identical psbA copies in the inverted repeat region of its chloroplast genome. A strain generated from wild type 137c (CC-125 mt ϩ ) in which both psbA copies had been inactivated and the native gene reintroduced at a distal single copy site in the chloroplast genome was available from a previous study (31). This strain, psbA-m-saa3ϩpsbA-psbA, accumulates native D1 protein at wild type levels and does not express the mammalian protein introduced in the psbA-m-saa strain (31). Here, we refer to this strain as C. reinhardtii-PSII. Gene sequences for the Synechococcus 7942 D1:1 and D1:2 (9) were codon-optimized for Chlamydomonas chloroplast by modifying the endogenous psbA sequence only in the codons for which amino acid substitutions were required. Additionally, we replaced the cleaved C-terminal peptide with that of Chlamydomonas to avoid potential processing incompatibilities (supplemental Fig. S1). Gene syn-thesis was carried out by GeneArt (Germany). The synthetic genes were introduced into psbA-m-saa by particle bombardment into the same site as in C. reinhardtii-PSII (32). The coding sequences were under the control of the Chlamydomonas psbA promoter, 5Ј-and 3Ј-untranslated regions. Transformants were selected for resistance to kanamycin and confirmed by PCR through sequencing of the PCR products. Both strains were rendered homoplasmic for the transgene insertion by propagation in kanamycin and confirmed by PCR. The resulting strains are referred to in this work as D1:1-PSII and D1:2-PSII.
Culturing Conditions, Growth Measurements, and Chlorophyll Determination-Chlamydomonas strains were grown in HS medium (33)  For growth rate measurements, 40-ml cultures were grown in HS medium and continuously bubbled with 2% CO 2 in air. Growth was monitored as OD at 730 nm. Full growth curves were recorded and then fit to a four-component Gompertz function (34) to calculate doubling times.
For biomass accumulation experiments, 9-liter cultures were grown in 12-liter glass carboys in HS medium and continuously bubbled with 2% CO 2 in air. After the stationary phase was reached (monitored by OD 730 nm ), cells were harvested by centrifugation and dried at 90°C overnight. Total dry weight was determined gravimetrically.
Chl was extracted in methanol, and relative concentrations of Chl-a and Chl-b were determined spectrophotometrically using extinction coefficients reported by Porra et al. (35).
qPCR and Western Blots-All starting material corresponded to mid-log phase cultures (ϳ10 6 cells/ml) grown at 100 E m Ϫ2 s Ϫ1 and 25°C. The Concert Plant RNA Reagent (Invitrogen) was used for all RNA extractions, following the small scale protocol with 5 ml of culture. 320 ng of total RNA were reversetranscribed using the Verso cDNA synthesis kit (Thermo Scientific, Waltham, MA). 20-l reactions were diluted 10-fold, and 2 l were used for qPCR with the Absolute Fast kit for probes and without ROX (Thermo Scientific) according to the manufacturer's instructions. qPCR assays were done in a CFX96 thermal cycler (Bio-Rad). Samples from each strain were run with five replicates. The rbcL gene was used as a reference (36). Expression levels were calculated with the CFX manager software (Bio-Rad) using the ⌬⌬Ct method. Amplification efficiencies ranged from 90.7 to 93.2% and were also considered for the expression analysis. All qPCR primers and probes were purchased pre-mixed from Integrated DNA Technologies (Coralville, IA) and are shown in supplemental Table S1.
Westerns blots were performed as described previously (37) with slight modifications. Total protein extracts were used instead of insoluble fractions. Total protein was quantified using the DC Protein Assay (Bio-Rad). The loading buffer also included 2 M urea, and denaturation was performed on ice for 30 min. The primary antibody corresponded to anti-D1 rabbit polyclonal raised against a conserved N-terminal peptide of the protein (Agrisera, Sweden). Dylight 488-conjugated anti-rabbit was used as the secondary antibody (Thermo Scientific). Chl fluorescence from light-harvesting complex II monomers was used as a loading control (38). Fluorescence imaging was carried out with a Typhoon 8600 scanner (GE Healthcare). Image analysis and densitometry was performed with ImageJ (rsb.info.nih.gov).
Competition Assays-An equal number of D1:1-PSII and D1:2-PSII cells (determined by hemocytometer) were inoculated in either TAP medium (dark control) or HS medium supplemented with 5 mM NaHCO 3 in vented large surface area untreated tissue culture flasks. Mixed cultures in TAP medium were grown in complete darkness without shaking at 25°C. Mixed cultures in HS medium were grown at either 3 or 290 E m Ϫ2 s Ϫ1 with shaking (80 rpm) at 25°C. For each time point, the total cell count of each flask was determined by a hemocytometer. Each resulting PCR used ϳ1000 cells from the mixed culture for DNA template. Primers were designed that did not discriminate between the DNA sequences of D1:1-PSII or D1:2-PSII (supplemental Fig. S1) as follows: forward primer, TTCTAACGCAATCGGTCT, and reverse primer, GAATA-ATGAACCACCGAAT, which results in a 379-bp gene fragment.
By chance, the D1:2-PSII gene sequence contains a PvuII restriction site (CAG2CTG) that is not present in the D1:1-PSII gene sequence (supplemental Fig. S1) and was used for identification. The 379-bp PCR fragment from D1:2-PSII DNA can be cleaved into 167-and 212-bp fragments by PvuII nuclease, although the 379-bp PCR fragment from D1:1-PSII DNA is unaffected (supplemental Fig. S2). Following 32 PCR cycles, two 10-l aliquots were taken from each reaction mixture. 5 units of PvuII (New England Biolabs) were added to only 1 aliquot, and both were incubated at 37°C for 40 min. Then both aliquots were run on a 1.2% agarose gel at 80 V for ϳ45 min and visualized using ethidium bromide. The fluorescence intensity of the 379-bp fragment in the control aliquot (no PvuII) was proportional to the sum total of D1:1-PSII and D1:2-PSII genes in the original mixed culture. The fluorescence intensity of the 379-bp fragment in the PvuII-treated fragment was proportional to the concentration of D1:1-PSII genes in the original mixed culture. These intensities were quantified using ImageJ and used to determine the fraction D1:1-PSII in the mixed culture at each specific time point. Knowledge of the total cell count and the fraction of D1:1-PSII enabled the determination of the number of D1:1-PSII cells and the number of D1:2-PSII cells.
Isolation of Thylakoid Membrane Fragments-Thylakoid membrane fragments were isolated from Chlamydomonas strains using a modified procedure by Shim et al. and Gokhale and Sayre (39,40), which is based in turn on the method for spinach by Berthold et al. (41). Briefly, ϳ1-liter cultures were grown in HS medium supplemented with 5 mM NaHCO 3 on a 12-h/12-h light/dark cycle at 100 E m Ϫ2 s Ϫ1 and 25°C. To synchronize growth phase, cells were harvested by centrifugation (3,500 ϫ g, 10 min) at 4 h into the third light cycle (42). Cells were resuspended in Buffer 1 (20 mM HEPES, pH 7.5, 350 mM sucrose, 2.0 mM MgCl 2 ) at 1-2 mg of Chl/ml and then disrupted using a BeadBeater (BioSpec Products, Bartlesville, OK) with 0.5-mm zirconia beads. Thylakoid membranes were isolated by centrifugation (40,000 ϫ g, 20 min) to pellet thylakoids and whole cells, resuspended in Buffer 1, centrifuged (1200 ϫ g, 30 s) to remove cell debris and whole cells, and then centrifuged (40,000 ϫ g, 20 min) to pellet thylakoids. Purified thylakoids were resuspended in Buffer B (20 mM MES, pH 6.0, 15 mM NaCl, 5.0 mM MgCl 2 , 5.0 mM EDTA) at Ն2.86 mg Chl/ ml. 25% Triton X-100 (20 mM MES, pH 6.0, 15 mM NaCl, 5.0 mM MgCl 2 ) was slowly added while stirring to a final Triton X-100/ Chl ratio of 20:1 and Chl concentration of 2.0 mg/ml. Following slow stirring for 25 min, membrane fragments were washed three times with Buffer B (40,000 ϫ g, 20 min) and flash-frozen in liquid N 2 at ϳ2 mg of Chl/ml in 25% glycerol, 50 mM MES, pH 6.0, 300 mM sucrose, 35 mM NaCl. All steps following cell harvesting were carried out at 4°C under dim green light. All buffers contained 0.1 mM phenylmethanesulfonyl fluoride and 1.0 mM benzamidine as protease inhibitors.
EPR Spectroscopy-EPR measurements on thylakoid membrane fragments were made on a Bruker Elexsys E580 at 9.45 GHz frequency. Specific measurement conditions are provided in supplemental Fig. S3. Spin quantification was performed using Fremy's salt (K 2 NO(SO 3 ) 2 , Sigma) standards, as described by Babcock et al. (43).
Lithium Dodecyl Sulfate-PAGE-25 g of total protein of thylakoid membrane fragments were denatured in 2% lithium dodecyl sulfate and 0.5% ␤-mercaptoethanol on ice for 10 min. Samples were loaded into wells of a 4 -16% polyacrylamide gel (Bio-Rad) and run at 4°C at 1 watt constant power for ϳ26 h. Bands were visualized using silver staining.
Flash O 2 -Flash O 2 yields were measured amperometrically using a home-built Clark-type electrode (membrane-covered Pt-Ir electrode) as described previously (44). A red LED (6,200 E m Ϫ2 s Ϫ1 , max ϭ 627 nm) was used to provide single turnover flashes (STFs). Optimal STF duration for 0.5 M PSII was determined to be 30 s. Amperometric response to each STF was detected over a 1-s window and then integrated with baseline correction to determine the O 2 yield per flash. Thylakoid membrane fragments were diluted to 0.5 M PSII in 40 mM MES, pH 6.0, 200 mM sucrose, 10 mM CaCl 2 , 10 mM MgCl 2 , and 10 mM NaCl. Freshly prepared 2 mM K 3 Fe(CN) 6 and 0.1 mM DMBQ were added immediately before each experiment began. Samples were dark-adapted for 5 min, pre-flashed with one 30-s STF, dark-adapted for 5 min, and then subjected to 30 30-s STFs applied at 0.1-1 Hz.
Fast Repetition Rate (FRR) Fluorometry-FRR fluorometry measurements were performed on a home-built instrument as described previously (45). Samples of whole Chlamydomonas cells at 50 g of Chl/ml were dark-adapted for 120 s and then subjected to 50 STFs applied at 0.2-100 Hz. To prevent physiological adaptations to dark conditions during the ϳ3-h experimental period, a 1-s flash from a blue LED (200 E m Ϫ2 s Ϫ1 ) was applied prior to each 120-s dark adaptation period.
Fourier Transform Analysis-The temporal frequencies comprising the oscillating traces from flash O 2 and FRR fluorometry experiments were determined by Fourier transform as described previously (45,46) with minor alterations. Prior to Fourier transformation, constant steady-state yields were subtracted from full traces to provide a base line of zero. Discrete Fourier transforms were calculated and fit to a cubic spline. The cycle Ϫ1 value corresponding to the maximal Fourier amplitude in the physical range of 0 -0.5 cycle Ϫ1 was determined. The reciprocal of this value was defined as the period. The shortest theoretical period was 4.
Kok Cycle Modeling-The rates of damping of flash O 2 yield and flash Chl fluorescence ratio (F v /F m ) were estimated by fitting to an extended version of the Kok model (47,48), utilizing matrix analysis of the Markov process (49), extended to include S-state selective transitions. The transition probability matrix is shown in Scheme 1, where ␣, ␤, ␥, ␦, and ⑀ represent misses, double hits, hits, backward transitions, and inactivations, respectively. Each matrix element a ij represents the transition probability from S jϪ1 to S iϪ1 , where 1 Յ i,j Յ 5. The fifth S-state in this model is an inactive state S ⑀ accessed via inactivations of PSII centers during the experiment (50). Conservation of matter implies that the sum of each column must equal 1, and therefore the ␥ ?ϭn (hit) parameters are written as ␥ 0 ϭ ␥ 1 ϭ 1-␣-␤, ␥ 2 ϭ 1-␣-␤-␦, and ␥ 3 ϭ 1-␣-␤-␦-⑀. Thus, the average hit probability ␥ avg ϭ 1-␣-␤-␦/2-⑀/4. The yield of O 2 , Y O2 , produced by a given flash is proportional to the population passing through the S 3 3 S 0 transition (47). For this model, Y O2 (n) ϭ (1-␣-␦-⑀)S 3 (n Ϫ 1) ϩ ␤S 2 (n Ϫ 1). We note that this model is similar to the extended Kok model proposed by Shinkarev (51) with the following alterations. First, the thermodynamically unreasonable backward transitions of S 0 3 S 3 and S 1 3 S 0 have been removed. We hypothesize that inactivations of the PSII reaction centers are most probable during the release of O 2 , which is known to generate reactive oxygen species (1). Thus, the inactive state, S ⑀ , is accessible solely through the precursor to the O 2 -evolving step, S 3 . Direct inactivation of S 3 occurs with probability ⑀. Double hit transitions passing through S 3 access S ⑀ with probability ␤⑀. Limiting access to S ⑀ via S 3 instead of allowing all S-states to directly feed into S ⑀ does not affect the calculated magnitude of ⑀. 4 Full oscillating traces from flash O 2 or FRR fluorometry were normalized to achieve a steady-state value of 1. Fittings were numerically determined by minimizing an objective function proportional to the mean square deviation from the experimental curve using the BOBYQA algorithm (52) in the NLopt nonlinear optimization package (53). For flash O 2 experiments, Y O2 was followed as a function of flash number. For FRR fluorometry experiments, the population of S 1 was followed.
S 3 Lifetime Measurements-Changes in the flash O 2 oscillation pattern following two pre-flashes was used to monitor S 3 lifetimes (54, 55). PSII samples as described above were darkadapted for 5 min, pre-flashed with one 30-s STF, darkadapted for 5 min, pre-flashed twice with 30-s STFs at 1 Hz, and then subjected to 30 30-s STFs applied at 1 Hz. The resulting oscillation traces were fit to the extended Kok model as described above. The relative initial population of S 3 (S 3 /(S 0 ϩ S 1 ϩ S 2 ϩ S 3 )) was plotted as a function of the delay time between the two pre-flashes and the 30 STF train. Data were fit to the two-component exponential decay function y ϭ y 0 ϩ

RESULTS
Characterization of D1 Mutants-Chlamydomonas strains containing heterologous psbA genes from Synechococcus 7942 were photoautotrophic and grew at identical rates as C. reinhardtii-PSII at moderate light intensity (15 E m Ϫ2 s Ϫ1 , Fig. 1). At stationary phase, the OD 730 nm of D1:1-PSII cultures was significantly higher than C. reinhardtii-PSII or D1:2-PSII, indicative of more cells or more absorbance per cell. Table 1 shows the ratio of Chl-a to Chl-b during exponential growth in photoautotrophic conditions. In whole cells, this ratio reflects the relative amount of reaction centers (Chl-a) to antenna (Chl-aϩb). No significant difference in Chl-a/b was observed between D1:1-PSII and D1:2-PSII strains at any light intensity tested. However, both heterologous strains have a significantly higher Chl-a/b ratio than the C. reinhardtii-PSII strain.
To better understand this phenotype, large scale cultures (9 liters) were inoculated in HS medium and bubbled with 2% CO 2 in air. Upon reaching stationary phase (monitored by OD 730 nm ), the biomass yield was measured gravimetrically. As shown in Table 2, D1:1-PSII accumulates 11% more biomass than C. reinhardtii-PSII (p ϭ 0.0269) and 12% more biomass than D1:2-PSII (p ϭ 0.0474) in the light-limiting conditions of this experiment.
To test for a growth advantage between the D1:1-PSII and D1:2-PSII cells at various light intensities, an equal number of cells from each strain was inoculated into a mixed culture. The relative genotype of the mixed culture was monitored over time as described under "Experimental Procedures." Standardization of the method established a linear correlation (R 2 ϭ 0.9777, see supplemental Fig. S2). In cells grown heterotrophically in darkness (TAP medium), no discernable difference between strains was observed ( Fig. 2A). Mixed cultures grown photoautotrophically (HS medium ϩ 5 mM HCO 3 Ϫ ) at 3 E m Ϫ2 s Ϫ1 contained ϳ30% more D1:1-PSII cells than D1:2-PSII cells after 10 days (Fig. 2B). By contrast, at high light fluxes (290 E m Ϫ2 s Ϫ1 ), D1:2-PSII cells were in 500% excess after 8 days compared with D1:1-PSII cells (Fig. 2C).
RT-PCR results of psbA transcript levels are listed in Table 3. When normalized to total Chl concentration and the C. reinhardtii-PSII strain, D1:1-PSII and D1:2 strains accumulate ϳ30 -45% less psbA transcript. These data are in agreement with D1 protein quantification (Table 3 and supplemental Fig.  S3) in which an approximate 40% decrease in D1 accumulation was observed for D1:1-PSII and D1:2-PSII compared with C. reinhardtii-PSII.
Characterization of Thylakoid Membrane Fragments-PSII concentration in thylakoid membrane fragments prepared from C. reinhardtii-PSII, D1:1-PSII, and D1:2-PSII was estimated through spin quantification of tyrosine Y D ⅐ detected using EPR. Light minus dark difference spectra were obtained that allow separation of P700 ϩ and the fast decaying tyrosine Y Z ⅐ from the slow decaying Y D ⅐ (43). Spectra of samples that were illuminated for 30 s at room temperature and then darkadapted for 30 min at 0°C (supplemental Fig. S4) were integrated (supplemental Fig. S5), and the results are listed in Table  3. Y D ⅐ decays very slowly in the dark at pH Ͻ 7.2 with biphasic kinetics. At 21°C, Vass and Styring (56) measured these kinetics in PSII-enriched spinach grana membranes as t1 ⁄ 2 f ϭ 41 min (21%) and t1 ⁄ 2 s ϭ 510 min (79%). The faster component represents the oxidation of S 0 by Y D ⅐ and has an activation energy of 30 kJ/mol (55). Using these values, we can estimate that at 0°C, less than 7% of Y D ⅐ has been reduced during the 30-min dark adaptation. D1:1-PSII and D1:2-PSII thylakoid membrane fragments have 2-fold higher Chl content per Y D ⅐ than those from C. reinhardtii-PSII. We emphasize that these preparations also contain significant amounts of PSI as evidenced by the lightadapted EPR spectra (supplemental Fig. S4) and lithium dode-cyl sulfate-polyacrylamide gel (supplemental Fig. S6). This feature arises naturally due to the lack of well ordered grana in the Chlamydomonas chloroplast.   Knowledge of the Chl/Y D ratios and the O 2 evolution rates normalized to Chl (Table 3) allowed for the determination of PSII turnover frequencies (TOFs). With DMBQ, D1:2-PSII had a slightly faster TOF (31.1 s Ϫ1 ) followed by D1:1-PSII (26.9 s Ϫ1 ) and then C. reinhardtii-PSII (21.9 s Ϫ1 ). The use of an alternative electron acceptor (2,5-dichloro-p-benzoquinone) did not affect this trend, and the values closely agree with the DMBQ results (supplemental Table S2). These TOFs are similar to rates reported for Chlamydomonas PSII-enriched thylakoid membrane fragments (45 s Ϫ1 ) (39) and core particles (27 s Ϫ1 ) (57).
In Vitro Flash O 2 Yield-Flash O 2 yield from thylakoid membrane fragments was measured for 30 STFs at repetition frequencies from 0.1 to 1 Hz. Excluding the dark adaptation period, these STF frequencies correspond to constant illumination light intensities of 0.02-0.2 E m Ϫ2 s Ϫ1 . A representative oscillation trace at 0.5 Hz is shown in Fig. 3A. Full data are available in supplemental Fig. S7. The O 2 yield from 30 STFs, Y O2 , was averaged for each trace as shown in Fig. 4. When normalized to Y D -PSII concentration D1:2-PSII produces the most O 2 per flash followed by D1:1-PSII. C. reinhardtii-PSII has significantly lower Y O2 over the STF frequencies tested.
In Vitro WOC Cycling Efficiency-The rate of damping of oscillations in Y O2 was monitored by Fourier transform (model independent, Fig. 5A) or fitting to our extended Kok model (Fig.  5B). In general, both period Ϫ1 and average Kok cycle hits increase with increasing STF frequencies. Both methods indicate that oscillations are of higher quality (slower rate of damping) in D1:1-PSII followed by D1:2-PSII. C. reinhardtii-PSII has significantly lower period Ϫ1 and Kok model hits (␥ avg ).
Variable Chl-a Fluorescence Yield-FRR fluorometry analysis allows in vivo measurements of whole cells over a wider range of frequencies than O 2 detection using a Clark-type electrode (45). The variable Chl-a fluorescence yield (F v /F m ) for whole cells was measured for 50 STFs at repetition frequencies from 0.2 to 100 Hz. Excluding the dark adaptation period, these STF frequencies correspond to constant illumination light intensities of 0.16 -80 E m Ϫ2 s Ϫ1 . A representative oscillation trace at 0.5 Hz is shown in Fig. 3B. Full data are available in supplemental Fig. S8. The average F v /F m value from the first 50 STFs, denoted ͗F v /F m ͘, is shown in Fig. 6. ͗F v /F m ͘ decreases with increasing STF frequency for all strains. Whole cells containing C. reinhardtii-PSII have the highest average F v /F m values followed by D1:2-PSII and then D1:1-PSII.

Quantification of psbA transcript and D1 protein in whole cells and Y D in thylakoid membrane fragments
Transcript and protein data are normalized to total Chl-aϩb and values of the C. reinhardtii-PSII strain and represent means of five (qPCR) or three (Western blot) biological replicates with standard error. Thylakoid membrane fragments were prepared from cells growth at 100 E m Ϫ2 s Ϫ1 (detailed under "Experimental Procedures"). For EPR quantification of Y D in thylakoid membrane fragments, dark-adapted spectra (supplemental Fig. S3) were integrated twice to calculate the area under the absorbance spectra. Spin quantification was performed using Fremy's salt (supplemental Fig. S4). O 2 evolution was measured at 25°C under saturating light intensity in the presence of 0.1 mM DMBQ and 2 mM ferricyanide. increase in the ratio of Chl:PSII and higher Chl-a/b ratios ( Table 3). The higher Chl:PSII ratio likely results from the lower steady-state pool of psbA transcripts in these strains compared with the C. reinhardtii-PSII strain ( Table 3). Given that all three constructs utilize the same native psbA promoter and thylakoid-targeting sequence from Chlamydomonas, we hypothesize that the heterologous transcripts are either less stable in vivo or that cells adapt to the presence of the photochemically more efficient hybrid PSII enzyme by making less psbA product. However, the modified codons in the synthetic genes might also play a role in decreasing translation efficiency (59,60). The higher Chl-a/b ratios seen for D1:1-PSII and D1:2-PSII compared with C. reinhardtii-PSII (Tables 1 and 3) may result from either an increased competition between Chl-a and Chl-b for binding to LHC antenna proteins or the smaller total number of PSII centers (which bind only Chl-a). Alternatively, the PSI content in D1:1-PSII and D1:2-PSII thylakoid membrane frag-ments may significantly differ compared with C. reinhardtii-PSII. Our present data do not allow distinction between these options.

Normalization of O 2 Rates to PSII Concentration Reveals That D1:2-PSII Has a Higher Intrinsic TOF-Given the measured differences in D1 accumulation in whole cells, accumulation of tyrosine-D (Y D ) radical in thylakoid membranes and
Chl content, O 2 measurements were normalized to the absolute concentration of PSII centers instead of total Chl. The trend in D1 protein content in whole cells quantitatively tracks with the Y D radical content in thylakoid membranes, which further corroborates the normalization to PSII. The TOF in thylakoid membranes is proportional to the intrinsic efficiency of PSII and is independent of downstream electron acceptors due to use of an exogenous quinone electron acceptor.
D1:1-PSII Has Higher WOC Cycling Efficiency at Low STF Frequencies-The inverse period and ␥ avg , representing intrinsic WOC cycling efficiency, are substantially higher for D1:1-PSII than for D1:2-PSII at low STF frequencies, and both mutants are much higher than C. reinhardtii-PSII at all measured frequencies (Figs. 5 and 7). These data reveal the significant operational advantage of the D1:1 isoform over D1:2 at low light intensities that was not previously recognized. One way to interpret this observation is that cytochrome (cyt) b 559 is more readily photo-reduced in D1:2-PSII (and C. reinhardtii-PSII) than D1:1-PSII as originally hypothesized by Sander et al. (26). Spectroscopically, contributions from such side pathway reactions were found to have a lower quantum efficiency in isolated Thermosynechococcus PSII core complexes containing the D1:1 isoform (11). In addition to Y Z donation and cyclic electron flow around PSII through cyt b 559 , the hole in P 680 ϩ may be filled by the nonradiative recombination of {P 680 ϩ Q A Ϫ } (61, 62). The thermodynamics and kinetics of this process are discussed below.
F v /F m Does Not Predict Y O2 and WOC Cycling Efficiency-In vitro, Y O2 increases with increasing STF frequency (0.1-1 Hz) (Fig. 4A), whereas in vivo, the average ͗F v /F m ͘ decreases with increasing STF frequency (0.2-100 Hz) (Fig. 6). In both cases,   FEBRUARY 22, 2013 • VOLUME 288 • NUMBER 8 the quality of transient oscillations improves with increasing STF frequency (Figs. 5 and 7). This longer coherence of oscillations is quantified both by the (model-independent) period Ϫ1 and the (model-dependent) Kok hit parameter (␥ avg ). Both reflect the cycling efficiency of the WOC. F v /F m is proportional to the quotient of quantum yield (QY) of PSII charge separation divided by the total Chl emission yield (2). Accordingly, it does not depend solely on PSII and is not the intrinsic QY of PSII photochemistry. However, Y O2 reports on the net conversion as follows: 2H 2 O 3 O 2 ϩ 4H ϩ ϩ 4e Ϫ , and thus measures the intrinsic photochemical QY of the WOC. We therefore draw attention to TOF, Y O2 , period Ϫ1 , and ␥ avg as the more relevant measures of PSII function.

Natural Variants of PSII Subunit D1 Tune Photochemical Fitness
Discrepancies between F v /F m and Y O2 have been observed previously at low light intensities (analogous to low STF frequencies) and were attributed to cyclic electron flow around PSII through cyt b 559 (63). Following an actinic pulse, cyt b 559 is reduced in chloroplasts with a t1 ⁄ 2 of ϳ100 ms (64). In vitro experiments in spinach have indicated that cyt b 559 can then reduce oxidized Chl Z ϩ , which fills the hole in P 680 ϩ with a t1 ⁄ 2 of ϳ500 ms (65,66). These relatively slow kinetics are faster than the dark time between STFs in the Y O2 experiments in this work. If such a pathway is active, P 680 ϩ would be competitively reduced by Y Z (which in turn oxidizes the WOC releasing O 2 ) and cyt b 559 /Chl Z . Thus, F v /F m does not scale directly with Y O2 (WOC photochemical QY), period Ϫ1 , and ␥ avg (WOC cycling efficiency) and should be interpreted accordingly.
At all STF frequencies, F v /F m is higher in C. reinhardtii-PSII followed by D1:2-PSII and least in D1:1-PSII (Fig. 6). Because of the large increase in Chl content (and hence F 0 fluorescence) in the mutants, this sequence is not surprising but also not particularly revealing. However, the change in F v /F m values over the 500-fold STF frequency range, i.e. slope, decreases at different rates and should reflect the F v contribution (photochemical charge separation within PSII). The average slope obtained by least squares fitting to a straight line reveals the following trend: C. reinhardtii-PSII (Ϫ0.065) Ͼ D1:2-PSII (Ϫ0.047) Ͼ D1.1-PSII (Ϫ0.034). We suggest the larger slope for C. reinhardtii-PSII reflects greater regulation of PSII charge separation within the native strain than the mutants. Although the cyanobacterial D1 isoforms outperform the algal D1 isoform at all flash frequencies as revealed both by WOC cycling efficiency (Fig. 5) and by WOC photochemical QY (Fig. 6), stronger regulation of PSII charge separation in the native D1 isoform may be an advantage for overall survival fitness (i.e. photoprotection).
Structure-Function Analyses of D1 Isoforms and Evolutionary Implications-The midpoint potential (E m ) of Pheo/Pheo Ϫ is regulated by the presence of a glutamine (D1:1-PSII) or glutamate (D1:2-PSII and C. reinhardtii-PSII) at D1 position 130. Recently, E m values for Pheo/Pheo Ϫ in Thermosynechococcus BP-1 were determined at Ϫ522 mV for D1:1 (24) and Ϫ505 mV for D1:2 (13). This difference was found to be more subtle than previous experiments in Synechocystis 6803 in which substitution of Gln-130 with Glu-130 resulted in a Ϫ33 mV difference in the E m of Pheo/Pheo Ϫ (23). The less negative E m of Pheo/ Pheo Ϫ in D1:2 supports enhanced charge separation efficiency by increasing the free energy gap between P 680 * and Pheo (23,24,61,62).
Given the results from these previous studies, it is reasonable to expect E m (Pheo/Pheo Ϫ ) to be more negative (higher energy) in D1:1-PSII (containing Gln-130) compared with D1:2-PSII and C. reinhardtii-PSII (both containing Glu-130). The E m of Q A /Q A Ϫ also varies with the D1 isoform. In Thermosynechococcus, E m (Q A /Q A Ϫ ) was found to be Ϫ140 mV in D1:1 (15) and Ϫ103 mV in D1:2 (14). Variations in the E m of Q A /Q A Ϫ control the kinetics of {P 680 ϩ Q A Ϫ } recombination as discussed by Vass and Cser (62) as follows. Because the free energy gap between Q A Ϫ and P 680 ϩ is very large (Ͼ1 V), the kinetics of this tunneling    (61) that D1:2 is specifically tuned for enhanced photoprotection at high light intensities, whereas D1:1 is more prone to photoinhibition, which has been repeatedly demonstrated in the literature (14, 20, 26 -29).
Why then do cyanobacteria such as Synechococcus 7942, Thermosynechococcus, and others predominantly express D1:1? Our data reveal a significant advantage in terms of WOC cycling efficiency at low light intensities for PSII centers containing D1:1 over D1:2 and an algal D1 isoform at low light intensities. In both in vitro flash O 2 (Fig. 5) and in vivo FRR fluorometry (Fig. 7) studies, D1:1-PSII has more efficient WOC cycling at low STF frequencies. The difference is striking at 0.2 Hz (Fig. 7) where Fourier transform analysis estimates D1:1-PSII has a flash cycle period of 4.68 Ϯ 0.04, whereas D1:2-PSII and C. reinhardtii-PSII have significantly poorer periods (farther from idealized four) of 5.03 Ϯ 0.10 and 4.95 Ϯ 0.06, respectively. A similar conclusion is reached from Kok modeling of the average hit probability, ␥ avg to be 0.590 Ϯ 0.010 in D1:1-PSII versus 0.499 Ϯ 0.016 in D1:2-PSII and 0.510 Ϯ 0.020 in C. reinhardtii-PSII.
The observation of improved oscillations in Y O2 at low STF in D1:1-PSII is consistent with whole cells measurements of Synechocystis 6803 mutants by Tichy et al. (29) that expressed Synechococcus 7942 D1 isoforms. At one flash frequency (1 Hz), the strain expressing D1:1 has peak Y O2 on flashes three and eight, and the strain expressing D1:2 has peak Y O2 on flashes four and nine. Fitting to the classic Kok model (47) gave miss parameters (␣) of 0.241 and 0.271 for D1:1 and D1:2, respectively (29).
The enhanced quality of oscillations in D1:1-PSII at low STF frequencies is the result of a more stable S 3 Kok cycle intermediate. We find that the fast component of S 3 decay kinetics in vitro is 40% faster in D1:2-PSII and C. reinhardtii-PSII compared with D1:1-PSII (Fig. 8). These experiments, which were performed at 24°C, produce half-times consistent with previous reports in spinach PSII membranes. Using a similar Y O2 method, Messinger et al. (55) calculated the kinetics of S 3 decay at 20°C to be t f 1 ⁄ 2 ϭ 7 s (28%) and t s 1 ⁄ 2 ϭ 94 s (72%), where t f 1 ⁄ 2 is the fast half-time and t s 1 ⁄ 2 is the slow half-time. More recently, these kinetics were directly measured at 20°C via EPR by Chen et al. (68) and found to be t f 1 ⁄ 2 ϭ 10 Ϯ 6 s (11%) and t s 1 ⁄ 2 ϭ 124 Ϯ 6 s (89%). In the time scales studied here, S 3 is reduced by the acceptor side of PSII and not Y D (68 FEBRUARY 22, 2013 • VOLUME 288 • NUMBER 8 environments, or in the lower mixed layer of the oceanic water column must subside on very low incident light intensities (2,69,70). If the rate of PSII-WOC cycling becomes limited by the incident photon flux (and subsequent charge separation), the presence of the D1:1 isoform uniquely enables the WOC to extend the lifetime of higher and less stable S-states. Structurally, this is accomplished by minimizing the recombination kinetics of {P 680 ϩ Q A Ϫ }. This effect is observed directly in the lightlimiting biomass accumulation and competition experiments described in this work; subtle changes in acceptor side midpoint potentials do have a real impact on cell growth and competition in vivo.

Natural Variants of PSII Subunit D1 Tune Photochemical Fitness
We summarize the structure-function relationship of D1 isoforms in the context of our model in Fig. 9. In D1:1-PSII, glutamine is present at position D1-130, causing the reduction potential (E m ) of Pheo/Pheo Ϫ to be more negative (Fig. 9, left panel) and resulting in less charge separation and a lower O 2 QY. The E m of Q A /Q A Ϫ is proportionally more negative, which decreases the {P 680 ϩ Q A Ϫ } recombination rate. At low light intensities, this is an advantage. S 3 is stabilized, and WOC cycling efficiency improves. However, at high light intensities, inhibition of the nonradiative {P 680 ϩ Q A Ϫ } recombination results in photoinhibition as flux is directed to alternative pathways (e.g. 3 {Pheo Ϫ P 680 ϩ }) that generate singlet O 2 . In D1:2-PSII and C. reinhardtii-PSII, glutamate is present at position D1-130, causing the E m of Pheo/Pheo Ϫ to be less negative and resulting in a higher probability of charge separation and a higher O 2 QY. The E m of Q A /Q A Ϫ is proportionally less negative, which increases the {P 680 ϩ Q A Ϫ } recombination rate. At low light intensities, this is a disadvantage. S 3 is destabilized, and WOC cycling efficiency decreases. However, at high light intensities, the enhancement of the nonradiative {P 680 ϩ Q A Ϫ } recombination results in photoprotection as flux is directly away from pathways (e.g. 3 {Pheo Ϫ P 680 ϩ }) that generate singlet O 2 .
Whereas this model physically satisfies the observation of an enhanced WOC cycling efficiency phenotype in D1:1-PSII, it does not fully explain the differences between F v /F m , Y O2 , period Ϫ1 , and ␥ avg . We postulate that in addition to thermodynamic differences of Pheo and Q A , cyclic electron flow around PSII involving cyt b 559 may also contribute more in D1:2-PSII than D1:1-PSII, as described above. The functional role of PSII cyclic electron flow, if such exists, could serve to dissipate light energy, or to pump protons into the lumen to support the pH gradient/electromotive force. No evidence to support the latter has emerged to date.
As previously mentioned, the changes in E m values for Pheo/ Pheo Ϫ and Q A /Q A Ϫ measured in Thermosynechococcus between D1:1 and D1:2 are more subtle than when the single point mutant Q130E was prepared in Synechocystis 6803 (23). Given that ϳ25 out of 360 amino acids vary between D1:1 and D1:2 in both Synechococcus 7942 and Thermosynechococcus, we concur with the previous hypothesis (24,71) that other amino acids besides D1-130 partially compensate for changes in the Pheo hydrogen bonding environment. Current work is underway to uncover these contributions using our versatile Chlamydomonas model system.
In conclusion, through the heterologous expression of Synechococcus 7942 D1:1 and D1:2 in the Chlamydomonas chloroplast, we have directly compared the efficiency of the cyanobacteria D1 isoforms and the algal D1 isoform. At low light intensities (low STF frequencies), D1:1-PSII has significantly better WOC cycling than D1:2-PSII and C. reinhardtii-PSII due to a more stable S 3 intermediate. This difference in WOC cycling efficiency is explained structurally through alterations of the E m values of Pheo/Pheo Ϫ and Q A /Q A Ϫ that functionally control the efficiency of charge separation and {P 680 ϩ Q A Ϫ } recombination. The phenotypes exhibited by D1:1-PSII are consistent with improved fitness for cyanobacteria living at very low light fluxes in natural environments. Ϫ E m (P 680 /P 680 ϩ )). Midpoint potentials are based on measurements in T. elongatus (12)(13)(14)16).