Regulation of Photosynthetic Light Harvesting Involves Intrathylakoid Lumen pH Sensing by the PsbS Protein*

The biochemical, biophysical, and physiological properties of the PsbS protein were studied in relation to mutations of two symmetry-related, lumen-exposed glutamate residues, Glu-122 and Glu-226. These two glutamates are targets for protonation during lumen acidification in excess light. Mutation of PsbS did not affect xanthophyll cycle pigment conversion or pool size. Plants containing PsbS mutations of both glutamates did not have any rapidly inducible nonphotochemical quenching (qE) and had similar chlorophyll fluorescence lifetime components as npq4-1, a psbS deletion mutant. The double mutant also lacked a characteristic leaf absorbance change at 535 nm (DeltaA535), and PsbS from these plants did not bind dicyclohexylcarbodiimide (DCCD), a known inhibitor of qE. Mutation of only one of the glutamates had intermediate effects on qE, chlorophyll fluorescence lifetime component amplitudes, DCCD binding, and DeltaA535. Little if any differences were observed comparing the two single mutants, suggesting that the glutamates are chemically and functionally equivalent. Based on these results a bifacial model for the functional interaction of PsbS with photosystem II is proposed. Furthermore, based on the extent of qE inhibition in the mutants, photochemical and nonphotochemical quenching processes of photosystem II were associated with distinct chlorophyll fluorescence life-time distribution components.


INTRODUCTION
In conditions of excess light, photosynthetic light harvesting is regulated by a feedback de-excitation mechanism termed energy-dependent quenching (qE), which increases thermal dissipation of excess absorbed light energy in photosystem II (PSII). The qE mechanism is triggered by conditions that limit photosynthetic carbon fixation and result in increased acidification of the chloroplast thylakoid lumen (1)(2)(3)(4). The thermal dissipation of excess excitation energy is most commonly measured and referred to as nonphotochemical quenching (NPQ) of PSII chlorophyll (Chl) a fluorescence. Although there are several components of NPQ, in higher plants qE can account for the major part of NPQ and is characterized by its relatively fast induction and relaxation kinetics, on a physiological time scale of seconds to minutes. The decrease in the intensity of Chl fluorescence is the result of the decrease in the electronic excited state lifetime of Chl caused by an increased thermal dissipation rate constant (5).
The rapid response of the qE process is chemically associated with changes in the transthylakoid membrane pH gradient (∆pH). The ∆pH change has at least two functions in qE. First, it activates the violaxanthin de-epoxidase (VDE) that converts violaxanthin (V) to antheraxanthin (A) and zeaxanthin (Z) (6). A and/or Z are essential elements of qE (7-9). 5 qE (14-16) is a carboxylate-modifying agent (17) that binds to PsbS (18). Although it was suggested that the DCCD binding site is in the lumenal loops of PsbS, the exact binding site has not been determined. Importantly, site-directed mutagenesis experiments indicated that two of the PsbS glutamates, E122 and E226, are necessary for the function of PsbS (13).
In this paper we used single and double mutations of PsbS (E122Q/E226Q) to make a detailed biochemical and biophysical analysis of the role of these two glutamates in pH sensing and DCCD binding. We probed the role of the E122 and E226 residues by monitoring the changes in the PSII Chl a fluorescence lifetime distributions, intensities, and photochemical efficiencies. Additionally, we also monitored the absorbance change at 535 nm (∆A 535 ) (19) that is obligatorily associated with qE (10,20,21). The ∆A 535 has been suggested to reflect a red shift in the absorption spectrum of Z that occurs upon binding to PsbS (22,23). From these measurements we formulated a model of the influence of the glutamate mutants on the fractional quenching of the populations of PSII (24) that is consistent with the symmetrical structure of the PsbS protein, the DCCD binding stoichiometry, and the stoichiometry of the xanthophyll components on a per PSII unit basis. 7 additional 10 min. The crosslinked thylakoids were centrifuged for 2 min at 15300 g and resuspended with protein solubilization buffer without reductant for analysis by SDS-PAGE. For immunoblotting of crosslinked samples, 5 nmol Chl was loaded in each lane.
Xanthophyll cycle pigment analysis-For zeaxanthin formation experiments, leaf discs were sampled from overnight dark-adapted plants as controls, sampled at noon in the growth chamber (low light conditions), and sampled at noon, floated on water, and then treated at 1600 µmol photons m -2 s -1 for 30 min (high light-treated). All samples were rapidly frozen in liquid nitrogen and stored at -80°C until analysis. Pigments were measured by HPLC (25), and the mean ± SD was calculated (n=6-9).
Leaf absorbance measurements-Leaf absorbance changes were measured using the Non-Focusing Optics Spectrophotometer (NoFOSpec) (26), but modified as described (27) to allow semi-simultaneous measurements of absorbance changes at four different wavelengths.
This was accomplished by aiming four separate banks of light-emitting diodes (LEDs, HLMP-CM15, Agilent Technologies, Santa Clara, CA), each filtered through a separate 5 nm bandpass interference filter (Omega Optical, Brattleboro, VT), into the entrance of a compound parabolic concentrator. Each bank of LEDs was filtered with a separate interference filter, at 500, 520, 535 and 545 nm each with a 5 nm bandpass (full width at half height). The photodiode detector was protected from direct actinic light by a Schott BG-18 filter. Current from the photodiode was converted to a voltage by an operational amplifier, and the resulting signal was AC-filtered to remove background signals, and sampled by a 16-bit analog-to-digital converter on a personal 8 probe pulses was set at 10 µs. Actinic illumination was provided by a set of 12 red LEDs (HLMP-EG08-X1000, Agilent Technologies, Santa Clara, CA) and controlled by the timing circuitry. Measuring pulses at each wavelength were given in sequence at 1-100 ms intervals, depending upon experiment.
Chl a fluorescence changes were also measured with the NoFOSpec instrument using the 525 nm measuring pulse to excite Chls, while protecting the detector with a Schott RG-9 filter (28) Leaves were cut from plants dark-adapted overnight, and the leaf petiole was wrapped in a small piece of moist cotton. The F m was first recorded by giving 800 ms white light of 30,000 µmol photons m -2 s -1 . After 30 s in the dark, a red actinic light of 1300 µmol photons m -2 s -1 was switched on for 215 s, and the leaf absorption at 535 nm was recorded. After allowing relaxation of qE in the dark for 110 s, leaf absorption at 535 nm was recorded again, and the ∆A 535 was calculated. After finishing the 535 nm measurement, the same actinic light was switched on again for 2 min, and then a white light pulse (800 ms, 30,000 µmol photons m -2 s -1 ) was applied to measure the F m '. DCCD binding analysis-14 C-DCCD labeling was conducted on thylakoids prepared as described previously (29) at two different pH conditions, pH 7.8 and pH 5, for 3 h at room temperature. For labeling at pH 7.8, thylakoid samples containing 30 µg of Chl were labeled in buffer T2 (5 mM Tricine pH 7.8, 50 mM sorbitol, 10 mM EDTA). For labeling at pH 5, thylakoid samples were labeled in citrate buffer (30 mM NaCitrate pH 5, 50 mM sorbitol, 10 mM EDTA). After mixing thylakoids with buffer, 7.5 µl of 100 µM 14 C-DCCD (0.5 µCi) in ethanol was added to make a final total volume of 100 µl. For gel analysis, a first dimension separation using a 12% acrylamide Tris-Sulfate gel with 6 M urea was used (30). After staining 9 the gel with Coomassie blue, the region corresponding to PsbS and light-harvesting complex proteins was cut out and equilibrated in Tris-HCl buffer pH 6.8 with 2% (w/v) SDS for 25 min, then transferred to the same buffer containing 60% glycerol for another 10 min. The gel slice was then loaded for second dimension separation by SDS-PAGE on a 10-16% gel without urea (31). The labeling intensities of PsbS mutants with respect to the wild type were acquired from the dried gels with the Instant Imager (Packard) instruments and calculated from the average of three independent measurements after correcting for the quantity of PsbS on the SDS-PAGE. µmol photons m -2 s -1 , Walz DT-Cyan filter) was applied to induce photosynthetic electron transport, thylakoid lumen acidification, V de-epoxidation, and PsbS protonation. Saturating pulses were applied every 60 s to measure F m '. After 10 min the sample temperature was reduced to 3°C (requiring ~3 min after switching water baths), while the actinic light was continued for a total of 15 min. All illumination was extinguished, and the sample was then rapidly turned to face the fiber-optic probe of the fluorescence lifetime instrument to illuminate the sample (140 µmol photons m -2 s -1 , 635 nm) for 5 min before initiating the fluorescence lifetime determination.
Fluorescence lifetime determinations were made under conditions of light-and low temperature (3° C)-saturated lumen acidification (32) to maintain maximal levels of V de-epoxidation and PsbS protonation while the PSII redox state was approximately 60-80% reduced, depending on the level of energy dissipation (33). The fluorescence intensity conditions during the lifetime acquisition corresponded to those defined as F s (34).
The PSII Chl a fluorescence lifetimes were determined using a multifrequency phasemodulation fluorimeter (Model K2-004, ISS Instruments, Urbana, IL, USA) using a red laser diode (peak emission 635 nm) for excitation and a red-sensitive microchannelplate photomultiplier tube (Hamamatsu Photonics, Hamamatsu, Japan R3809U-50) for emission detection. Excitation and emission were applied and measured, respectively, from the front surface of the sample using a bifurcated quartz fiber-optic probe with the single terminus facing the sample. The excitation diode-laser intensity was attenuated to 140 µmol photons m -2 s -1 at 30 MHz to avoid photobleaching during measurements, which normally required 10 min to complete. The reference and sample signals were detected at each modulation frequency using high-transmittance (>80%) narrow waveband (12 nm halfwidth) interference filters (Corion Inc, Franklin MA USA) centered at 645 nm and 689 nm, respectively. An automatic rotating filterwheel exchanged the sample and reference positions such that the phase and modulation signals were repeatedly determined using 4-6 cycles at each modulation frequency until standard errors were reduced below 0.04 degrees for the phase angle shift and 0.001 for the demodulation ratios.

Effect of glutamate mutations on PsbS function and qE-Glutamate residues in the
lumenal loops of PsbS were changed to glutamines by site-directed mutagenesis to eliminate H +binding capacity while minimizing alteration of the protein structure (13). The E122Q and E226Q mutations were made individually and pairwise and expressed in npq4-1 Arabidopsis plants that lack endogenous wild-type PsbS. As a control, npq4-1 plants were transformed with the wild-type psbS gene. For each single mutant (npq4-E122Q and npq4-E226Q), the double mutant (npq4-E122QE226Q), and the control (npq4-1 + psbS), two to three independent lines carrying single insertions were selected (Fig. 1A).
Immunoblot analysis showed that PsbS protein levels in all transgenic lines were 3-to 5fold higher than in wild-type plants (Fig. 1). Immunoblotting was conducted on washed thylakoid membranes, demonstrating that the PsbS protein was inserted in the membrane in all cases. Furthermore, protein crosslinking experiments with isolated thylakoids showed no differences in PsbS crosslinking patterns (data not shown), indicating that the glutamate mutations of PsbS and enhanced protein expression level do not affect the physical association of PsbS within the thylakoid membranes or its interaction with other proteins. Figure 2 shows the NPQ induction traces in the transgenic lines, the wild type, and npq4-1. Compared with the npq4-1 + psbS lines, which had higher levels of NPQ than the wild type due to overexpression of PsbS (33), the single mutant lines had about one-third as much NPQ.
Most of the NPQ was rapidly reversible (qE). There was no significant difference between the single mutant lines npq4-E122Q and npq4-E226Q. The NPQ level in the npq4-E122QE226Q double mutant lines was indistinguishable from that of npq4-1, which lacks qE, indicating that the double mutant PsbS was nonfunctional.
Xanthophyll cycle pigment composition and ∆A 535 -There were no significant differences in either the total xanthophyll cycle pigment pool sizes or V de-epoxidation states in high light (1700 µmol photons m -2 s -1 ) in any of the different lines sampled at the same conditions (data not shown). The xanthophyll cycle pool size was 26.5±0.8, 28.2±0.9, and 29.6±1.2 mmol/mol Chl a in the three sample conditions, and the respective de-epoxidation states were 0.053±0.010 in overnight dark-adapted leaves, 0.068±0.010 in leaves sampled at noon under growth light conditions, and 0.649±0.010 in leaves treated for 30 min with light that was ten times higher than the growth light. Thus, we conclude that neither the amount of PsbS in the thylakoid membrane, as shown before (33,36), nor the mutations of the PsbS protein influence the xanthophyll cycle pool size or de-epoxidation in high light. Figure 3A shows that ∆A 535 was linearly correlated with the extent of NPQ in each sample. There was a significant x-intercept value of about 0.36, indicating that a small component of the total NPQ is independent of ∆ A535 and PsbS. This component is attributable to qI, a component of NPQ that is related to photoinhibitory damage to PSII and associated with a slowly reversing component of F m quenching that is observed in the absence of PsbS, as discussed in more detail below. We further analyzed the linear relationship between ∆A 535 and NPQ by consolidating the variations of the sampling in each measurement and plotting the lines for the mean values of NPQ and ∆A 535 from the wild type, npq4-1, npq4-1 + psbS, npq4-E122Q, npq4-E226Q, and npq4-E122QE226Q lines (Fig. 3B). This analysis increased the significance of the slope and intercept values (Table 1). There were no significant differences between npq4-1 and the double mutant or between the npq4-E122Q and npq4-E226Q single mutants.
DCCD binding-DCCD binds to carboxylate residues in hydrophobic environments (17,37). It is known for its ability to inhibit qE (14-16), which suggests that carboxylate residues are involved in the process. PsbS purified from spinach or Arabidopsis and PsbS expressed in E.
coli bind DCCD at pH 7.5. Three acidic residues in each of the two lumen-exposed loops were suggested to be the possible binding sites (18). Fig. 4 shows that, in thylakoids isolated from wild-type Arabidopsis, DCCD bound to PsbS at pH 5. However, no binding of DCCD was detected at pH 7.8 (data not shown). DCCD binding was only 45±10% and 50±12% of the control in the npq4-E122Q and npq4-E226Q single mutants, respectively, and binding was undetectable in the npq4-E122QE226Q double mutant (Fig. 4). These results indicate that DCCD can equally bind to E122 and E226 and that mutation of one of the glutamates does not affect the chemical characteristics of the other in a cooperative manner.
Chl a fluorescence lifetimes-We further analyzed the Chl a fluorescence lifetime distributions in these mutants. Figure 5A shows the phase shift and demodulation ratio data as a function of modulation frequency (symbols) and the globally fit model (lines), with the residual errors in the lower panel. The patterns of the phase angle shifts and higher demodulation ratios indicate that the npq4-1 + psbS overexpression lines have the fastest decay times (lowest phase angles and highest demodulation ratios), followed by the wild type, the two single mutants npq4-E122Q and npq4-E226Q, the double mutant, and finally the npq4-1. The residual error plot indicates the model exhibits no remarkable systematic deviations and a random distribution around the mean = 0 for both the phase angle shifts and demodulation ratios.  Figure   5A. The npq4-1 + psbS exhibited less than 3% of the c6 level in the double mutant, indicating the PsbS functionality may have been nearly saturated. It was clear, however, that the npq4-1 + psbS line had slightly less PsbS activity than the wild type + psbS lines described previously (33). The wild type + psbS lines with higher PsbS protein levels exhibited larger relative c1:c3 ratios than the npq4-1 + psbS line analyzed here, leading to about a 56 ps (16%) faster average lifetime in the former samples.

DISCUSSION
PsbS senses the ∆pH through two symmetrically arranged, lumen-exposed glutamates, E122 and E226-Unlike typical light-harvesting complex proteins, PsbS has four rather than three transmembrane helices. Protein sequence analysis showed high similarity between helix I and helix III and also between helix II and helix IV (38,39), reflecting the symmetrical topology of the PsbS protein (Fig. 7). The two lumen-exposed loops (40) are also highly similar, with E122 and E226 located in the middle of each loop. Mutating one or the other of these two glutamates only partly inhibits the PsbS function in qE, whereas the npq4-E122QE226Q double mutant totally disrupts the PsbS function (Fig. 2). These results are consistent with the suggestion that glutamates E122 and E226 of PsbS serve as the pH sensors for qE (13).
The function of E122 and E226 was further confirmed by investigation of DCCD binding. DCCD is a protein modifying reagent which covalently binds to carboxylate residues involved in reversible protonation in hydrophobic environments (17). DCCD has previously been shown to act as a specific inhibitor of qE (14-16), thus leading to the search for target sites using 14 C-DCCD. Binding sites have been identified in Lhcb proteins (41), namely CP26 (42) and CP29 (43), but evidence for these proteins being the functional binding site(s) of this qE inhibitor was lacking. Our results show that the single mutations E122Q and E226Q each reduce the level of DCCD binding to PsbS by approximately 50% and in the same way decrease the amplitude of qE. These finding imply that PsbS, rather than CP26 or CP29, is the target site for qE inhibition by DCCD. Moreover, since the double mutant is not labeled, E122 and E226 are the only DCCD binding sites in PsbS, implying that they have an essential role in PsbS function. It should be noticed that purified PsbS, either extracted from thylakoids or recombinant from E. coli, binds DCCD at pH 7.5, but low pH is required for labeling of PsbS in thylakoid membranes, suggesting that either the conformation of PsbS is different in the two environments or that interactions with neighbor subunits in PSII supercomplexes prevents exposure of lumenal glutamate residues until the pH is decreased.
It is interesting to note that the two glutamate residues are responsible each for 50% of both DCCD binding and control of qE activity. This suggests that, whatever the mechanism is for the activity of PsbS in qE, the two halves of the two-fold symmetrical molecule are acting independently. Recent work (12) showed that in the presence of PsbS, a 10 ps spectral component is present that can be interpreted as a Chl a to Z energy transfer, and Z binding to PsbS has also been reported (23). On this basis, the qE quenching can be tentatively understood in terms of the protonation of each glutamate leading to binding of two Z molecules, which are capable of accepting and dissipating energy from singlet excited Chl a. Thus, PsbS can be viewed as a thylakoid sensor of excess light. In the presence of a low thylakoid lumen pH, protonation of PsbS results in thermal dissipation of excess absorbed light energy, i.e., a feedback downregulation of photosynthetic light harvesting by the ∆pH.
In addition to influencing qE, the E122Q and E226Q mutations also inhibit the ∆A 535 leaf absorbance change (Fig. 3). It was found early on that ∆A 535 is invariably associated with qE (20,21). Here we report a single linear relationship exists between NPQ and ∆A 535 even when taking into consideration two different factors, namely different levels of PsbS protein and different amino acid substitutions in the PsbS protein (Fig. 3). It was hypothesized that ∆A 535 is due to a red absorption shift of Z upon binding to PsbS (22). The linear relationship between ∆A 535 and NPQ suggests that one of two Z-binding sites is affected in each single mutant (E122Q or E226Q), whereas Z binding might be eliminated completely in the double mutant (E122QE226Q). Figure 8A, our current biochemical model of PsbS function postulates that PsbS associated with PSII may exist in one of three states depending on the lumen acidity and the concentration of Z (and/or A). The W state is defined as a PsbS with its critical glutamate residues (E122 and E226) in an unprotonated state. W converts reversibly to X upon protonation of the glutamates as determined by their pKa value. The protonated state X is hypothesized to be an activated state for potential binding of Z. Upon binding of Z, the PsbS switches from the X to the Y state, which is the fully functional energy dissipation mode.

Interpreting the fluorescence lifetime distribution components with respect to the glutamate mutants, PsbS protein level, and PSII function-As illustrated in
Consistent with our recent papers (33,44), we have found it necessary to expand our model to six states to account for the opening/closing (oxidation/reduction) of the PSII reaction centers. In order to simplify the model parameterization and interpretation, our experiments were designed to yield both saturating light-induced lumen pH and Z concentration conditions, thereby reducing to a minimum both the W (by saturated protonation) and X states (by saturated xanthophyll binding). Practically, we have also concluded that we are unable to resolve the W and X components in either the open or closed PSII states, most likely because they are likely to differ by less than 20% between the open and closed W or X components. Therefore, our empirical global model linking scheme assumes the W c +X c and W o +X o components to be indistinguishable, yielding a simplified four component scheme (Fig. 8B). As mentioned in the results, we postulate the rapid c1 and negative c2 components to be more strongly related to the PsbS protein level than PsbS function per se, because these components were resolved in the double and both single mutants. Hence, we do not consider them to be associated directly with a particular state of PSII in the above scheme. We propose that extremely high PsbS concentrations may have indirect effects on the photochemical activity of the PSII core antenna and possibly alter energy transfer pathways among Chl complexes, perhaps between the antenna and core. With respect to Figure 8B, the (1100 ps) c6 component of our lifetime distribution model would be assigned to the combined [W c +X c ] states. We emphasize that the assignment of c6 to the W state is supported by the fact that the c6 contribution is maximal in the npq4-1, very similar to npq4-1 in the double mutant, reduced by ~50% in both single mutants (E122Q and E226Q) compared to the double mutant, reduced by >80% in the wild type and down to less than 3% in the npq4-1 + psbS. The c4 (300 ps) component appears consistent with emission from the W o +X o state PsbS conformations with open PSII traps. However, as indicated above it is unlikely that the W o state would be a significant component because the key PsbS glutamate residues were likely titrated to virtually complete protonation. Likewise, we feel it is logical to assign the c3 (220 ps) and c5 (525 ps) states to the open Y o and closed Y c PSII centers, respectively. A main tenet of the model outlined in Figure 8B, that is consistent with many steady-state spectroscopy experiments, is that increasing the population of the Y state acts synergistically to following the illumination treatments in these leaf experiments (1250 µmol photons m -2 s -1 for 20 min) as opposed to 20-25% observed in isolated thylakoids which were treated with lower (500 µmol photons m -2 s -1 ) light intensities (36). We note as mentioned before that in thylakoids we measured an additional lumen pH-induced 16% decrease in the lifetime center of the W mode tentatively attributed to protonation and/or conformational changes of other PSII proteins besides PsbS, which was absent in the npq4-1 mutant. Previously, we reported that the W mode center was decreased during aerobic photoinhibition in thylakoids largely independent of the width or fraction of the distribution (45). So both things considered, we conclude that the ~50% decrease in the W mode in the F m ' condition compared to W in the F m state in these leaf experiments is attributed to both photoinhibitory damage caused by the strong light and protonation/conformational changes in other PSII proteins besides PsbS. Further we consider that all lifetime modes may be partially attenuated by PSII photochemical activity through kinetic processes that are beyond the resolution of our instrument (i.e., 5-10 ps).

Concluding Remarks
One of the key points of the model described above is the implications of the lifetime distribution fractional intensities on the structure-function of PsbS. In our view the simplest explanation for the similarities of the two single mutants (E122Q and E226Q) is that they have equivalent effects because of the bilateral symmetry of the PsbS protein structure. The data indicate, as suggested in Figure 9, that each single mutation elicits a ~50% inhibition effect that becomes 100% in the double mutant. Hence, we propose that PsbS has two equivalent functional sites, one on each side, that are likely associated with pH-activated binding of the xanthophylls. It is thus suggested that PsbS associates with its binding site to attach itself to the PSII holocomplex (likely in or near the core antenna proteins) in one of two facial orientations with a 50% probability. And further, according to this model, two Z per PsbS will be the saturation amount for qE.  In this case, [W] would be 0.5. For npq4-1 + psbS (and in the wild type, not shown), either possible interaction orientation of PsbS with PSII is functional.