Deprotonation of the 33-kDa, Extrinsic, Manganese-stabilizing Subunit Accompanies Photooxidation of Manganese in Photosystem II*

Photosystem II catalyzes photosynthetic water oxidation. The oxidation of water to molecular oxygen requires four sequential oxidations; the sequentially oxidized forms of the catalytic site are called the S states. An extrinsic subunit, the manganese-stabilizing protein (MSP), promotes the efficient turnover of the S states. MSP can be removed and rebound to the reaction center; removal and reconstitution is associated with a decrease in and then a restoration of enzymatic activity. We have isotopically edited MSP by uniform 13C labeling of the Escherichia coli-expressed protein and have obtained the Fourier transform infrared spectrum associated with the S1 to S2transition in the presence either of reconstituted 12C or13C MSP. 13C labeling of MSP is shown to cause 30–60 cm−1 shifts in a subset of vibrational lines. The derived, isotope-edited vibrational spectrum is consistent with a deprotonation of glutamic/aspartic acid residues on MSP during the S1 to S2 transition; the base, which accepts this proton(s), is not located on MSP. This finding suggests that this subunit plays a role as a stabilizer of a charged transition state and, perhaps, as a general acid/base catalyst of oxygen evolution. These results provide a molecular explanation for known MSP effects on oxygen evolution.

Photosystem II catalyzes photosynthetic water oxidation. The oxidation of water to molecular oxygen requires four sequential oxidations; the sequentially oxidized forms of the catalytic site are called the S states. An extrinsic subunit, the manganese-stabilizing protein (MSP), promotes the efficient turnover of the S states. MSP can be removed and rebound to the reaction center; removal and reconstitution is associated with a decrease in and then a restoration of enzymatic activity. We have isotopically edited MSP by uniform 13 C labeling of the Escherichia coli-expressed protein and have obtained the Fourier transform infrared spectrum associated with the S 1 to S 2 transition in the presence either of reconstituted 12 C or 13 C MSP. 13 C labeling of MSP is shown to cause 30 -60 cm ؊1 shifts in a subset of vibrational lines. The derived, isotope-edited vibrational spectrum is consistent with a deprotonation of glutamic/ aspartic acid residues on MSP during the S 1 to S 2 transition; the base, which accepts this proton(s), is not located on MSP. This finding suggests that this subunit plays a role as a stabilizer of a charged transition state and, perhaps, as a general acid/base catalyst of oxygen evolution. These results provide a molecular explanation for known MSP effects on oxygen evolution.
In oxygenic photosynthesis, the multi-subunit protein complex, photosystem II (PSII), 1 uses light energy to oxidize water and to form molecular oxygen. Hydrophobic subunits, such as the D1 and D2 proteins, bind most of the prosthetic groups involved in charge separation. Water oxidation occurs at a catalytic site containing four manganese atoms. The catalytic site accumulates the four oxidizing equivalents required for water oxidation. The five sequentially oxidized states of the catalytic site are called the S states. Each oxidation of the catalytic site is associated with the reduction of quinone accep-tor molecules, Q A , a single electron acceptor, and then Q B , a two-electron, two-proton acceptor (reviewed in Ref. 1).
The 33-kDa, manganese-stabilizing protein (MSP) of PSII plays an important role in water oxidation (reviewed in Ref. 2). As an extrinsic subunit, MSP can be removed from the plant reaction center by several different types of biochemical treatments (3)(4)(5)(6). MSP has also been removed by mutagenesis in the cyanobacterium, Synechocystis sp. PCC 6803 (7), and in a green algae, Chlamydomonas reinhardtii (8). In the presence of low concentrations of calcium and chloride, plant MSP was found to be required for photosynthetic oxygen evolution and for maintaining the stability of the manganese cluster (9,10). In the presence of high concentrations of calcium and chloride, oxygen evolution occurs, but the steady state rate of enzymatic activity is impaired upon removal of MSP (7,9,(11)(12)(13)(14). In addition, removal of MSP and replacement with calcium and chloride result in kinetic inhibition of the S state transitions (12,(15)(16)(17)(18)(19). MSP can be rebound to PSII (20), and this reconstitution reverses the effects described above.
In this report, we investigate the molecular mechanism by which MSP exerts its effects on one step of the oxygen evolution cycle, the S 1 to S 2 transition. S 1 is a dark-stable state; the S 2 state can be produced by illumination at 200 K (see Ref. 21, for example). At this temperature, all the other S state transitions are blocked (22). To identify microscopic structural changes, occurring on MSP during the S 1 to S 2 transition, we have employed a combination of difference FT-IR spectroscopy and isotope editing of MSP. FT-IR spectroscopy can be used to obtain information about structure/function relationships in PSII. For light-induced enzymatic processes, this is a powerful technique with which to follow mechanistically important structural changes (for review, see Refs. 23 and 24). Isotope editing, or complete 13 C labeling of just the MSP subunit, allows us to focus on structural changes in MSP, even on the unlabeled background of the PSII reaction center (25). These data provide new evidence concerning the role of MSP in photosynthetic water oxidation.

EXPERIMENTAL PROCEDURES
Purification of MSP-Escherichia coli, containing a spinach psbO gene in an expression plasmid, was used to produce MSP (25). Normally, MSP is produced with a transit sequence at the amino terminus; this transit sequence targets the protein to the thylakoid lumen and is cleaved before MSP binds to PSII (reviewed in Ref. 2). For isotope editing and in vitro reconstitution, expression of MSP in E. coli is necessary in order to ensure the adequate production of labeled, processed protein (25). The gene, encoding E. coli-expressed MSP, differs from the gene, encoding native spinach MSP, in three respects. First, the transit sequence has been deleted from the gene. Second, this construct contains an amino-terminal methionine, which is essential for expression. The presence of this methionine was verified by amino acid sequencing (25 mutation, V235A, which was found to be introduced into the gene, probably as a result of propagation of the plasmid in E. coli. Val-235 is not a conserved residue in MSP. The V235A mutation has been characterized previously and has been shown to have no effect on the binding of MSP at room temperature or on the activity and stability of MSP-reconstituted PSII (26,27). To express and label MSP for spectroscopic studies and for double-stranded DNA sequencing, 1-liter cultures were inoculated from frozen glycerol bacterial stocks of E. coli, carrying the psbO plasmid. The V235A mutation was discovered by doublestranded sequencing of the psbO expression plasmid; this experiment was carried out at the completion of these studies. Because these glycerol stocks were also employed in our earlier work, this finding suggests that the V235A mutation was also present in the 12 C-and 13 C-labeled MSP employed in our previous FT-IR studies (25). This finding does not change the conclusions derived from those earlier experiments.
Production and purification of 12 C MSP or 13 C MSP was performed as described (25). The purity (approximately 90%) and isotopic enrichment (greater than 95%) were verified (25).
PSII Preparations-PSII membranes (28) were isolated from spinach and were treated so as to remove the native MSP, while retaining manganese (25). 12 C or 13 C MSP was rebound to PSII (25). PSII samples, to which MSP has been rebound, lack the extrinsic 24-and 18-kDa subunits of PSII and require calcium and chloride for activity (reviewed in Ref. 2). PSII complexes were purified from this material through the use of ion exchange chromatography (29) and were concentrated for FT-IR studies. This procedure, carried out in the presence of excess detergent, ensures that 12 C and 13 C MSP are specifically bound. Sodium dodecyl sulfate-polyacrylamide gel electrophoresis and Western analysis verified the removal and subsequent reconstitution of 12 C or 13 C MSP (25). Oxygen evolution assays were performed as described (30) on a Clark-type O 2 electrode (YSI 5300, YSI Inc., Yellow Springs, OH) using 10 g/ml Chl, 0.4 mM recrystallized 2,6-dichlorobenzoquinone, and 1 mM potassium ferricyanide. Rates of oxygen evolution for the 12 C MSP and 13 C MSP rebound samples were indistinguishable and were between 800 and 1000 mol O 2 (mg Chl-h) Ϫ1 . For some experiments, PSII complexes were purified directly from PSII membranes (29). Rates of oxygen evolution in these preparations were similar to the rates given above.
EPR Analysis-EPR spectra were obtained on a Bruker EMX spectrometer as described (21). The samples contained one molar eq of potassium ferricyanide per mol of PSII reaction center. In order to generate the multiline signal and Fe 2ϩ Q A Ϫ signals, samples were illuminated at 200 K with saturating light for 10 min and flash-frozen in liquid N 2 . Multiline spectra were recorded at 10 K with the following instrument settings: microwave frequency, 9.435 GHz; microwave power, 20 milliwatts; modulation amplitude, 32 G; time constant, 0.16 s. Four scans of 2.8 min were recorded and averaged before and after illumination. The Fe 2ϩ Q A Ϫ EPR signal was recorded at 4 K with the following instrument settings: microwave frequency, 9.435 GHz; microwave power, 40 milliwatts; modulation amplitude, 32 G; time constant, 0.33 s. Four scans of 5.6 min were recorded and averaged before and after illumination. The signal amplitudes were corrected for any difference in chlorophyll concentration and gain.
FT-IR Spectroscopy on PSII-Infrared spectra were recorded on a Nicolet 60-SXR FT-IR spectrometer equipped with a liquid nitrogencooled MCT/B detector and a KBr beamsplitter. The samples contained one molar eq of potassium ferricyanide per mol of PSII reaction center. In experiments conducted under 10 min of illumination, the mirror velocity was 1.57 cm s Ϫ1 , a Happ-Genzel apodization function was used, the spectral resolution was 8 cm Ϫ1 , and 2500 mirror scans were coadded for each double-sided interferogram. In experiments conducted after 1 min of illumination, spectral parameters were the same, except that 250 mirror scans were co-added. The glycerol-containing sample (6 l) was placed on a 25-mm germanium window, concentrated under N 2 gas for 20 min at 4°C, and then sandwiched with a CaF 2 window (Harrick Scientific, Ossining, NY). The temperature control apparatus and illumination conditions, employing continuous illumination and a Dolan-Jenner annular illuminator, were described previously (21,31,32). The cryostat was equipped with two temperature sensors; the temperature sensor on the sample was continuously monitored through the use of LabView software. In experiments performed at 200 K, the light intensity was 200 -300 E s Ϫ1 m Ϫ2 ; a red filter and a heat filter were employed. The temperature during data acquisition was 200 Ϯ 0.3 K. The amide I absorbance for all PSII samples was between 0.5 and 0.3, and all spectra were normalized to an amide II absorbance of 0.5.
In some experiments, temperature was maintained at 264 Ϯ 1 K through the use of a Harrick variable temperature cell (33,34). The samples contained 1 molar eq of potassium ferricyanide per mol of PSII reaction center. PSII samples contained less than 1% glycerol and were concentrated and dehydrated under a stream of cold N 2 gas, as described above. A single pulse from the 532-nm second harmonic of a pulsed YAG laser was used as the excitation source. Data acquisition parameters were the same as described above except that 180 mirror scans were coadded for every single-sided interferogram.
FT-IR Spectra on Model Compounds-DL-Glutamic acid and DL-aspartic acid were purchased from Sigma. 13 C-Labeled DL-glutamic acid (99% 13 C at C␦) and DL-aspartic acid (99% 13 C at C␥) were purchased from Isotech (Miamisburg, OH). FT-IR spectra on these compounds in solution, at a concentration of 40 mM, were obtained through the use of a Microcircle cell equipped with a ZnSe crystal (Spectra-Tech, Stamford, CT). A Nicolet Magna 550 FT-IR spectrometer, equipped with a MCT/A detector and a KBr beamsplitter, was employed. The mirror velocity was 2.53 cm s Ϫ1 , a Happ-Genzel apodization function was used, the spectral resolution was 8 cm Ϫ1 , and 6000 mirror scans were coadded for each double-sided interferogram. The difference spectrum, associated with the deprotonation of the carboxylic acid side chain, was derived by measuring the FT-IR spectrum at pH 3.8 and 4.8 (for glutamic acid) and at pH 3.4 and 4.4 (for aspartic acid). Data were directly ratioed to give the difference spectra associated with the deprotonation of the side chain. These pH values were chosen to give a minimal contribution to the spectrum from the titration of other functional groups in the amino acid.
Analysis of the Isotope-edited Spectrum-To test our assignments of the isotope-edited spectrum, Grams software (Galactic Industries, Salem, NH) was used to reconstruct the experimental data from the assigned spectral components. The spectral components were modeled as Gaussian lines. The initial frequencies and numbers of spectral components were chosen based on inflection points or peaks in the isotope-edited spectrum. The amplitudes, frequencies, and linewidths were varied until the best fit to the experimental data was obtained.

RESULTS
EPR Spectroscopy-Illumination at 200 K generates the multiline form of the S 2 state in PSII complexes, as assessed by EPR spectroscopy (Fig. 1A). The PSII sample employed in Fig.  1A was suspended in aqueous buffer containing 25% glycerol. Because very concentrated samples are required for difference FT-IR spectroscopy, FT-IR PSII samples were concentrated by a factor of approximately 3 with a stream of cold nitrogen. To show that this procedure has no effect on the yield of the S 2 state, we present EPR data obtained on a PSII sample, which was concentrated by a factor of 3 with nitrogen gas in the EPR tube ( Fig. 1B). Except for a change in the background, caused by the smaller sample volume, there is no significant change in the amplitude or yield of the S 2 state.
EPR controls were performed to show that PSII complexes, containing either 12 C or 13 C MSP PSII, generate an S 2 Q A Ϫ state. Fig. 1 (C and D) shows the lineshapes and intensities of the signals are similar to each other ( Fig. 1, C and D) and to the signal generated in control PSII (see Fig. 1A and Ref. 21). This experiment indicates that both samples are capable of S 1 to S 2 advancement at 200 K and that, as expected, isotopic labeling of the MSP does not have a significant effect on the yield of the S 2 state or the magnitude of the hyperfine splittings to manganese.
As an additional control, the EPR signal, assigned to the Fe 2ϩ Q A Ϫ , was also measured in both 12 C and 13 C MSP PSII samples (Fig. 1, E and F). The lineshapes and intensities of the signals are similar (Fig. 1, E and F) and are also similar to signals obtained from control PSII samples (21). These results confirm that the amounts of charge separation are indistinguishable, when 12 C and 13 C MSP reconstituted PSII samples are compared. Fig.  2A, we present a difference or light-minus-dark FT-IR spectrum obtained by continuous illumination at 200 K. As shown by the EPR control experiments (Fig. 1, described above) and by our previous studies (21, 31, 32, 35), these data are a

Controls for FT-IR Studies of the S 1 and S 2 States-In
obtained under 10 min of continuous illumination ( Fig. 2A, dashed line) exhibits broad spectral features between 1500 and 1200 cm Ϫ1 . When data were obtained after 1 min of red-filtered and heat-filtered illumination and from the same PSII preparation, a similar spectrum was observed ( Fig. 2A, solid line). In agreement with our published work (35), alteration in illumination time from 10 min to 1 min had no effect on the spectrum obtained. The inset in Fig. 2 shows the output of the sample temperature sensor; the data in the inset reveal that there were no significant temperature changes at the sample under illumination ( Fig. 2, inset, arrows).
At 200 K, all S state transitions are blocked except the S 1 to S 2 transition (22) and, therefore, the S 2 state can be generated by continuous illumination. At 264 K, a laser flash to a darkadapted sample was used to advance from the S 1 to S 2 state. To achieve a stable base line at this temperature, the sample contained no significant concentration of cryoprotectant or water. Under these conditions, we obtain a difference FT-IR spectrum (Fig. 2B) similar to that previously reported (36 -39); this spectrum has been attributed to the S 1 to S 2 transition (36 -39). This apparent temperature, cryoprotectant, and/or water dependence is under investigation.
FT-IR Spectra of the S 1 to S 2 Transition in 12 C and 13 C MSP-containing PSII Samples-The S 1 to S 2 spectrum obtained at 200 K exhibits minor intensity variations from preparation to preparation. For example, when Fig. 2A is compared with Fig. 2C, a variation in the ratio of the 1478 cm Ϫ1 line to broad spectral features between 1500 and 1200 cm Ϫ1 , previously assigned to S 2 -minus-S 1 (21,31,32,35), is observed. When PSII complexes in glycerol are illuminated at 200 K, the S 2 state is formed in the majority of centers, but Chl ϩ is oxidized in 33 Ϯ 9% of PSII centers (21). Chl ϩ contributes intensity at 1478 cm Ϫ1 (40). Therefore, small spectral variations may be attributable to the variation in the yield of Chl ϩ , measured previously (21). To average adequately over these effects in the isotope experiments described below, multiple FT-IR samples were employed, the 12 C and 13 C MSP reconstituted samples were purified during the same 6-month time period, using the same reagents, and the PSII samples were evaluated to ascertain that the oxygen evolution rates of the preparations were in the range from 800 to 1000 mol of O 2 (mg Chl-h) Ϫ1 .
A comparison of spectra obtained from these PSII complexes, containing either rebound 12 C-or 13 C-labeled MSP, is shown in Fig. 2 (C and D). The difference FT-IR spectrum of the S 1 to S 2 transition is dominated by contributions from carboxylic acid and carboxylate groups close to or ligating to the manganese cluster (21,31,32). This assignment was based on group frequency arguments and on the lack of significant 15 N shifts upon global 15 N labeling (21). In agreement with this previous assignment, significant spectral differences, attributable to isotopic shifts, are observed when 12 C and 13 C data are compared; these differences are particularly evident in the 1750 -1700 cm Ϫ1 and 1500 -1300 cm Ϫ1 regions. The spectra shown in Fig.  2 (C and D) are light-minus-dark spectra. The control, darkminus-dark difference spectra, obtained on the same samples, gave the expected result, i.e. no significant spectral features are observed (Fig. 2, E and F). Fig. 3 (A and B) shows an expansion of the 1850 -1550 cm Ϫ1 region of difference spectra, obtained from 12 C and 13 C MSP PSII complexes. Fig. 4 (A and B) shows an expansion of the 1600 -1150 cm Ϫ1 region. In Fig. 3C, we present the direct one-to-one subtraction of are normalized based on total protein concentration, the double difference spectra (Figs. 3C and 4C) are isotope-edited difference spectra, which will reflect only structural changes involving amino acid residues on MSP. Also presented are control 12 C-minus-12 C (Figs. 3D and 4D) and 13 C-minus-13 C subtractions (Figs. 3E and 4E), which, as expected, show no well defined spectral features. The base-line deviation, observable at frequencies greater than 1775 cm Ϫ1 , is caused by interference fringes, which are a consequence of performing FT-IR spectroscopy in transmission mode. Interference fringes cannot be eliminated, but contribute to the spectrum only as a sinusoidal roll in the base line.

FIG. 1. Light-minus-dark, difference EPR spectra, reflecting S 2 Q A
؊ -minus-S 1 Q A . Panels A and B show the S 2 multiline signal, obtained from PSII complexes before (A) and after (B) concentration by a factor of approximately 3 with a stream of cold nitrogen. Panels C and D show the S 2 multiline signal, obtained from 12 C and 13 C MSP PSII complexes, respectively. Panels E and F show the Fe 2ϩ Q A Ϫ signal, obtained from 12 C and 13 C MSP PSII complexes, respectively. Spectra were generated by illumination with red-and heat-filtered light at 200 K.
Assignment of the Isotope-edited Spectrum-The isotope-edited spectrum is a double difference spectrum, so 13 C-induced isotope shifts will cause the appearance of bands with the opposite sign. The isotope-edited spectrum will reflect structural changes occurring only on MSP during the S 1 to S 2 transition. The discussion of the isotope-edited 12 C-13 C, double difference spectrum (Figs. 3C and 4C) will be divided into three parts: 1) the amide I, II, and III regions; 2) negative vibrational lines, arising from MSP in the S 1 state; and 3) positive vibrational lines, arising from MSP in the S 2 state.
Amide I, II, and III Assignments-In the spectral region between 1800 and 1200 cm Ϫ1 , the FT-IR absorption spectrum of PSII is expected to exhibit three lines arising from vibrations of the peptide backbone (41). These lines are referred to as amide I, II, and III (41). The amide I band corresponds primarily to the CϭO stretch of the peptide backbone, and the amide II band is the out-of-phase combination of CN stretching and NH in-plane bending modes (41). The amide III band is the in-phase combination of the CN stretch and NH bend, with CC stretching and CO bending character as well (41). The frequencies of amide I, II, and III bands are expected to vary (as much as Ϯ20 -30 cm Ϫ1 ) as a result of hydrogen bonding and other perturbations of the environment of the peptide bond (42,43).
In the FT-IR spectrum of PSII, the amide I and II bands have maxima at 1657 and 1551 cm Ϫ1 , respectively (25). Amide III is expected at approximately 1300 cm Ϫ1 (41), and the FT-IR spectrum of PSII contains a band at 1334 cm Ϫ1 (data not shown). 13 C labeling of MSP will downshift the fraction of the amide I, II, and III bands that arise from this subunit. For example, 13 C labeling was shown to downshift the MSP amide I band by approximately 50 cm Ϫ1 in our earlier work (25).
The isotope-edited double difference spectrum (Figs. 3C and 4C) contains three spectral features, at 1658, 1557, and 1336 cm Ϫ1 , which we ascribe to amide I, II, and III vibrational lines, respectively. These features are observed in the isotope-edited spectrum; therefore, these lines must arise from MSP. The assignment of spectral features, in the double difference spectrum, at 1658 and 1557 cm Ϫ1 to the amide I and II bands is consistent with our previous assignments, which were based on 15 N labeling (21). The amide II and III lines in the isotopeedited spectrum have a negative/positive/negative lineshape, which is consistent with a 13 C-induced downshift of a derivative-shaped spectral feature (see Ref. 21, for an example). Because the amide II and III features have this lineshape, we assume that the amide I component, which is located in a more congested spectral region, has this lineshape as well. The magnitude of the 13 C shift is difficult to estimate precisely from the negative/positive/negative lineshape, but the magnitude of the shift must be less than the total width of the negative/positive/ negative feature. These linewidths are in the range from 30 to 70 cm Ϫ1 . The observation of contributions from amide I, II, and III in the isotope-edited spectrum is consistent with a small conformational change occurring on MSP during the S 1 to S 2

FIG. 2. Light-minus-dark, difference FT-IR spectra obtained at 200 K (A and C-F) or at 264 K (B).
Spectra at 200 K were obtained through the use of red-filtered and heat-filtered illumination; samples contained glycerol and water. Spectra at 264 K were obtained through the use of a 532-nm laser flash; samples contained less than 1% glycerol and were dehydrated. In A (dashed line), the spectrum was obtained under 10 min of illumination at 200 K. In A (solid line), the spectrum was obtained after 1 min of illumination at 200 K. In B, the spectrum was obtained at 264 K. The 200 K data shown in C and D reflect the S 1 Q A to S 2 Q A Ϫ transition in 12 C and 13 C MSP PSII complexes, respectively. The 200 K data shown in E and F are control, dark-minus-dark difference spectra obtained from the 12 C and 13 C MSP PSII samples, respectively. For A, difference spectra are averages of four data sets. For B, the difference spectrum is an average of three data sets. For C-F, difference spectra are averages of 12 data sets. The tickmarks on the y axis represent 0.002 absorbance units. The inset shows the output from a temperature sensor mounted on the sample holder, and the arrows denote times when the sample was subjected to illumination.
transition. This conformational change could be due to a change in hydrogen bonding of the peptide backbone.
Negative Bands, Arising from MSP in the S 1 State-Two negative features at 1740 and 1711 cm Ϫ1 are observed in the isotope-edited spectrum; two positive lines at 1696 and 1680 cm Ϫ1 are also observed, which we assign to 13 C-downshifted vibrations (Fig. 3C, solid fill). Vibrational modes from the carbonyl groups of chlorophyll could make a contribution in this spectral region (31,40). However, 13 C labeling of MSP will not shift vibrational modes of pigments; this result supports the assignment to amino acid residues of MSP. The only amino acid side chains, with expected vibrational features in this region, are glutamic and aspartic acid residues (44 -47). Therefore, the 1740 and 1711 cm Ϫ1 spectral features are assigned to carbonyl stretching vibrations, arising from glutamic or aspartic acid residues in MSP. 13 C labeling causes a downshift of these lines to 1696 and 1680 cm Ϫ1 ; note the appearance of positive lines with these frequencies (Fig. 3C, solid fill). The spectral shifts are 44 and 31 cm Ϫ1 , respectively, in good agreement with the expected downshift to 1701 and 1673 cm Ϫ1 for a harmonic CϭO stretching mode upon 13 C labeling. Thus, the magnitude of the 13 C shift supports the assignment of these lines to the CϭO vibration (Fig. 3C, solid fill).
The observation of two different lines with well separated frequencies suggests that at least two different glutamic or aspartic acid residues, located in MSP, are perturbed upon oxidation of manganese to give the S 2 state. The frequencies of carboxylic acid CϭO stretches reflect changes in double bond character in the CϭO bond as well as changes in the basicity of the oxygen (45). A typical range of frequencies is from 1750 to 1720 cm Ϫ1 (45,48). There is currently no experimental evidence that MSP binds manganese (reviewed in Ref. 2). Therefore, we assume that frequencies of 1740 and 1711 cm Ϫ1 arise from uncoordinated carboxylic acid groups (48).
The assignment of these lines to carboxylic acid residues is strengthened by the observation of broad, negative spectral features at 1288 and 1192 cm Ϫ1 in the isotope-edited spectrum (Fig. 4C, solid fill). We assign these lines to the expected, coupled C-O and O-H in-plane vibrations, originating from glutamic/aspartic acid residues (45)(46)(47). Any downshifted lines are not observable, perhaps because of the breadth of these spectral features and because of the spectral cut-off from the calcium fluoride windows.
Comparison to Model Compound Data-The observation of negative lines, arising from carboxylic acid contributions to the isotope-edited spectrum, leads to the conclusion that the S 1 to S 2 transition perturbs glutamic and aspartic acid residues on MSP. There could be two types of structural changes, which would result in a carboxylic acid contribution to the isotopeedited spectrum. The first is a frequency shift of the CϭO stretching vibration, perhaps due to a perturbation in the pK a or hydrogen bonding of these groups (see Ref. 31, and references therein). This first type of structural change would give

FIG. 3. Light-minus-dark, difference FT-IR spectra, reflecting S 2 Q A
؊ -minus-S 1 Q A . Spectra A and B were obtained from the 12 C and 13 C MSP PSII complexes, respectively, and are reproduced from Fig. 2  (C and D). The double-difference, isotope-edited spectrum, obtained as a direct one-to-one subtraction of A and B, is shown in C, where it is multiplied by a factor of 2. Spectra D and E are control 12 C-minus-12 C spectra and 13 C-minus-13 C spectra, respectively. The spectra in A and B are each an average of 12 data sets; the spectra in D and E are an average of only 6 data sets and are thus divided by ͌ 2. The tickmarks on the y axis represent 0.002 absorbance units. The lines in solid fill are assigned to the COOH group; lines in cross-hatched fill are assigned to the CO 2 Ϫ group.

FIG. 4. Light-minus-dark, difference FT-IR spectra, reflecting S 2 Q A
؊ -minus-S 1 Q A . Spectra A and B were obtained from the 12 C and 13 C MSP PSII complexes, respectively, and are reproduced from Fig. 2  (C and D). The double-difference, isotope-edited spectrum, obtained as a direct one-to-one subtraction of A and B, is shown in C. Spectra D and E are control 12 C-minus-12 C spectra and 13 C minus-13 C spectra, respectively. The spectra in A and B are each an average of 12 data sets; the spectra in D and E are an average of only 6 data sets and are thus divided by ͌ 2. The tickmarks on the y axis represent 0.002 absorbance units. The lines in solid fill are assigned to the COOH group; lines in cross-hatched fill are assigned to the CO 2 Ϫ group.
rise to derivative-shaped spectral features in the isotope-edited spectrum, each of which would be accompanied by a 13 C-induced, downshifted, derivative-shaped line. The isotope-edited spectrum obtained (Figs. 3C and 4C) is not consistent with this explanation. The second type of structural change, which could give rise to such a spectral contribution, is a protonation change. A protonation change, converting a carboxylic acid residue to a carboxylate, has a characteristic vibrational signature (45)(46)(47). The CϭO and C-O stretches of the carboxylic acid moiety disappear and are replaced by the asymmetric and symmetric stretching vibrations of the RCO 2 Ϫ anion. Because the electron is delocalized over both oxygens in the carboxylate anion, the asymmetric and symmetric stretching vibrations are expected to be intermediate in frequency between the CϭO and C-O stretching vibrations of the carboxylic acid (45,48).
This effect is illustrated in Fig. 5 (A and D), in which the deprotonation spectra of the carboxylic acid side chains in aspartic and glutamic acid, respectively, are shown. When the carboxylic acid side chain of aspartic acid deprotonates, lines at 1721 and 1211 cm Ϫ1 , assigned to the CϭO and C-O/OH vibrational modes, are replaced with intensity at 1574 and 1391 cm Ϫ1 , assigned to the asymmetric and symmetric stretching modes of the carboxylate anion (Fig. 5A). A similar spectrum is obtained upon deprotonation of the carboxylic acid side chain of glutamic acid, with slight frequency shifts to 1718, 1221, 1555, and 1399 cm Ϫ1 (Fig. 5D).
Spectra were also obtained on 13 C-labeled aspartic and glutamic acid; only one carbon atom was labeled and that was the C␦ or C␥ atom of the side chain (R 13 COOH). These data are shown in Fig. 5B and E. As expected, all features assigned to CO vibrations of the side chain exhibit 13 C shifts. Because the concentrations of the solutions were identical, spectra could be subtracted on a direct one-to-one basis to give an isotope-edited spectrum (Fig. 5, C and F). Note that the frequencies obtained from positive and negative lines in the double difference spectrum can vary slightly (Ͻ12 cm Ϫ1 ), when compared with predictions made solely from the difference spectra. This is due to the fact that peak positions in the double difference spectrum are influenced by linewidth and lineshape as well as by frequency.
The isotope-edited spectra derived for aspartic and glutamic acid show that all vibrational features, arising from the amino acid R group, are altered in frequency upon 13 C labeling (Fig. 5,  C and F). For example, the asymmetric stretching vibrations (1579 and 1556 cm Ϫ1 ) downshift 43-40 cm Ϫ1 and the symmetric stretching vibrations (1392 and 1400 cm Ϫ1 ) downshift 32-25 cm Ϫ1 (Fig. 5, C and F, cross-hatched lines). The CϭO vibrations of the carboxylic acid are affected by 13 C labeling, as expected, but the position of the downshifted line is not obvious (Fig. 5, C and F, solid fill lines). This is caused by the breadth of the spectral features and the fact that a Fermi resonance may complicate this spectral region in carboxylic acids (49,50). The expected Fermi resonance is a combination of the CϭO FIG. 5. Difference FT-IR spectra associated with the deprotonation of the carboxylate side chain in aspartic and glutamic acid. Spectra were obtained on 40 mM solutions; alterations in pH were used to titrate the amino acid's COOH group. A and B show difference spectra obtained from 12 C aspartic acid and 13 C aspartic acid, respectively. C represents the double, 12 C-minus-13 C, difference spectrum derived from these data (A minus B). D and E show difference spectra obtained from 12 C glutamic acid and 13 C glutamic acid, respectively. F represents the double, 12 C-minus-13 C, difference spectrum derived from these data (D minus E). The lines in solid fill are assigned to the COOH group; lines in cross-hatched fill are assigned to the CO 2 Ϫ group. See "Experimental Procedures" for additional information. Tick marks on the y axis are 0.02 absorbance units. stretching and OH bending modes (49,50). For example, this expected Fermi resonance may be responsible for the fact that intensity is observed at 1733 cm Ϫ1 upon 13 C labeling of glutamic acid (Fig. 5, E and F). Note that the effect of the putative Fermi resonance is not as noticeable in aspartic acid, as compared with glutamic acid (compare Fig. 5, C and F). This work provides a characteristic vibrational signature, i.e. frequencies, isotope shifts, and patterns of negative/positive lines, for the isotope-edited spectrum associated with the deprotonation of a carboxylic acid residue.
Positive Bands, Arising from MSP in the S 2 State-Comparison of model compound data, associated with the deprotonation of carboxylic acid side chains in vitro (Fig. 5), with the isotope-edited spectrum, obtained from PSII (Figs. 3C and 4C), supports the conclusion that a aspartic/glutamic acid residue(s) deprotonates upon the S 1 to S 2 transition.
In the isotope-edited spectrum of PSII (Figs. 3C and 4C), two broad, positive lines are observed at 1643 and 1630 cm Ϫ1 ; these features downshift to 1594 and 1582 cm Ϫ1 upon 13 C labeling (Fig. 3C, cross-hatched). Some cancellation of intensity is observed, with the negative lines substantially less intense when compared with the positive lines. The 13 C isotope shifts are 49 and 48 cm Ϫ1 , respectively, assuming that the 1643 cm Ϫ1 line shifts to 1594 cm Ϫ1 and that the 1630 cm Ϫ1 line shifts to 1582 cm Ϫ1 .
Two additional positive lines at 1489 and 1455 cm Ϫ1 are also observed in the isotope-edited PSII spectrum (Fig. 4C, crosshatched). These lines exhibit 13 C shifts, giving rise to one observed, negative line at 1396 cm Ϫ1 (Fig. 4C, cross-hatched). Cancellation of intensity at approximately 1430 cm Ϫ1 may explain the observation of only one downshifted spectral feature. Thus, the observed 13 C isotope shift is 59 cm Ϫ1 , assuming that the 1455 cm Ϫ1 line shifts to 1396 cm Ϫ1 .
From the comparison to model compound data (Fig. 5) and the lack of significant 15 N shifts in the 1643-1630 and 1489 -1455 cm Ϫ1 regions (21,46,47), these spectral features are assignable to the asymmetric and symmetric C-O stretching frequencies, arising from deprotonated glutamate and aspartate residues on MSP. In general, the magnitude of the 13 C shifts support this assignment for the asymmetric stretching vibrations. The 1643 and 1630 cm Ϫ1 lines show 49 -48 cm Ϫ1 13 C downshifts (Fig. 3C), in reasonable agreement with the 43 and 40 cm Ϫ1 13 C shifts observed upon C␦ and C␥ labeling in glutamate and aspartate (Fig. 5, C and F). On the other hand, the symmetric stretching vibration at 1455 cm Ϫ1 shows a much larger shift (59 cm Ϫ1 ) when compared with the 32-25 cm Ϫ1 13 C isotope shift observed upon C␦ and C␥ labeling in model compounds (compare Fig. 4C with Fig. 5, C and F). The larger magnitude of the 13 C induced shift in the isotope-edited PSII spectrum is explainable because 13 C labeling of MSP labels all carbon atoms, not just the C␦ and C␥ positions, and because the symmetric stretching vibration most probably involves significant displacement of other carbon atoms in the amino acid side chain (46,47).
The frequencies of the asymmetric and stretching vibrations of glutamate and aspartate residues in MSP ( Fig. 3C and 4C) are substantially upshifted from the frequencies observed for model compounds in water (Fig. 5, C and F). Frequency shifts are to be expected in the protein environment (for example, see Ref. 51). The fact that the asymmetric and symmetric stretching frequencies are both upshifted suggests that the amino acid side chains are not coordinated to metals (48). The observation of four different lines with well separated frequencies, assignable to carboxylate groups, indicates that at least two different glutamate or aspartate residues contribute to the spectrum.
Reconstruction of the Isotope-edited Spectrum-To test whether the assignments and interpretations described above can account for the isotope-edited spectrum in a quantitative way, the isotope-edited spectrum was fit and then reconstructed from these spectral components. The results of this procedure are summarized in Fig. 6. Fig. 6A (dashed line) shows the isotope-edited spectrum, derived by 13 C labeling of MSP. Fig. 6A (solid line) shows the result of simulation of the spectrum, using the spectral components shown in Fig. 6 (B-D). As shown, Fig. 6B is the spectral component, which we have assigned to the amide I, II, and III vibrational contributions. Fig. 6C represents negative spectral components, which we have assigned to glutamic and/or aspartic acid residues in MSP. Fig. 6D represents positive spectral components, assigned to glutamate and/or aspartate residues in MSP. Addition of these individual 12 C-13 C spectral components results in an adequate reconstruction of the isotope-edited spectrum. This can be ascertained by a direct comparison of the reconstructed and experimental spectrum (Fig. 6A) or from the residual (Fig. 6E). Therefore, we conclude that we can quantitatively account for all the lines in the isotope-edited spectrum with the assignments given above. Summary of Vibrational Assignments-Taken together, model compound data on 12 C and 13 C glutamic and aspartic acid and the isotope-edited PSII spectrum provide evidence for the full or partial deprotonation of two aspartic/glutamic acid residues in MSP upon the S 1 to S 2 transition (Fig. 7). The location of the base, which accepts these protons, will be discussed below. The observation of spectral features in the amide I, II, and III regions is consistent with a small conformational change in the peptide backbone of MSP during the S 1 to S 2 transition (Fig. 7). DISCUSSION Upon illumination of PSII at 200 K, the S 1 to S 2 transition can be studied independently of other transitions (22). Illumination at this temperature has been used to acquire the S 2 Q A Ϫminus-S 1 Q A vibrational spectrum (21,31,32,35). The spectroscopic results that we have presented here are consistent with a deprotonation reaction, accompanying the S 1 to S 2 transition. This deprotonation reaction converts glutamic/aspartic acid residue(s) to the corresponding anionic species. The use of isotope editing allows us to pinpoint the location of these amino acid residues and conclude definitively that this deprotonation reaction occurs on MSP during the S 1 to S 2 transition (Fig. 7). We have also obtained evidence for a MSP conformational change, which occurs during the S 1 to S 2 transition. Structural changes on MSP contribute to the S 2 Q A Ϫ -minus-S 1 Q A spectrum, but do not account for all spectral features observed in this difference spectrum.
The question of the number of deprotonation reactions is of interest, but cannot be ascertained quantitatively by the intensity of vibrational lines in this isotope-edited spectrum. This is caused by the fact that infrared molar absorptivity can vary slightly when amino acid residues are in different solvents and, therefore, in different protein environments. Because at least two CϭO frequencies are observed, we deduce that two (or more) amino acid residues are affected by the S 1 to S 2 transition. However, the reactions need not be full deprotonation reactions of each site, but might consist of partial deprotonation reactions of interacting residues (see Ref. 52, for an example).
Another interesting aspect of the isotope-edited spectrum is the observed breadth of some of the spectral features. This spectral breadth has been discussed previously as possible evidence for carboxylate rearrangements, which occur on the time scale of the S 1 to S 2 transition (21, 32).
At pH 6.0, glutamic/aspartic acid residue(s), with a typical pK a of approximately 4, would be expected to be deprotonated and in the carboxylate form. However, the glutamic/aspartic acid residue(s), which is affected by the S 1 to S 2 transition, has a perturbed pK a , because this residue(s) is initially in the protonated form. Unusual pK a values for carboxylate groups in proteins have been observed previously (see Ref. 50, and references therein). Oxidation of manganese is known to occur on this S state transition (reviewed in Ref. 53). The oxidation of a metal would be expected to stabilize the anionic form of the carboxylate (31). Thus, we conclude that the photo-oxidation reactions provide the driving force for MSP structural changes and deprotonation reactions.
While we cannot comment on the protonation state of these aspartates/glutamates in the transition state, our observation demonstrates that deprotonation occurs somewhere on the S 1 to S 2 reaction coordinate. Deprotonation could be coupled to the oxidation reaction, with a single transition state for both reactions. However, the reported solvent deuterium isotope effects on this step are in the range from 1.2 to 2.9 (for example, see Refs. 54 -56). Although some range of values has been reported, these effects are much smaller than the values expected for a chemical reaction that is coupled to proton transfer (57). Possible explanations of the low solvent isotope effect are that the oxidation reaction constitutes the rate limitation or that deprotonation/oxidation are indeed coupled, but the transition state is highly asymmetric for the proton transfer reaction (57).
To explain our results and previous results in the literature concerning the role of MSP on the S 1 to S 2 transition, we propose that this deprotonation reaction is the molecular mechanism by which MSP exerts its kinetic effects on the S 1 to S 2 transition. A complete or partial proton transfer reaction can stabilize a charged transition state and thus can lower the activation barrier for the reaction. Lowering of the activation barrier will increase the rate of the reaction, if the transition state is charged (58). The presentation of negative charge, in close proximity to the manganese cluster and at an appropriate time in enzymatic turnover, may be responsible for tuning the kinetic properties of the S 2 state and optimizing enzymatic efficiency.
As stated above, this hypothesis explains previous results in the literature. For example, in the absence of MSP, evidence for a slowing of the S 1 to S 2 /S 2 to S 3 oxidation reactions has been obtained (15-17, 19, 59). Removal of MSP also results in a slowing of the reverse, recombination reactions, by which S 2 combines with Q A Ϫ or Q B Ϫ and converts back to the S 1 state (15)(16)(17). Such a deceleration of both forward and back reactions is consistent with an increase in the activation barrier for the S 1 to S 2 transition, when MSP is removed. Thermodynamic effects on the oxidation potential of the S 2 state are also expected as the result of generation of anionic species in the environment of the manganese cluster. The energetic effect of such negative charges will depend on the properties of the catalytic site.
Removal of MSP also decreases the rate of the S 2 to S 3 transition and the rate of recombination of S 3 with Q A Ϫ and Q B Ϫ (15-17, 19, 59). We propose that the generation of negative charge on MSP, in response to the S 1 to S 2 transition, may also be important in adjusting the activation barrier for the next reaction, the S 2 to S 3 transition. In fact, the subsequent generation of the S 3 state, through photooxidation events in the PSII reaction center, may involve a similar deprotonation event on MSP and these deprotonations may be reversed during the S 4 to S 0 transition, when water oxidation is completed and molecular oxygen is released. The conformational changes occurring on MSP may be linked to and may be responsible for this reversal. In this model, MSP acts as an accumulator of negative charges on the early S state transitions, undergoing successive conformational changes, and then reprotonates and conformationally relaxes during the last enzymatic step. Further, we hypothesize that these sequential changes in charge and structure are important in facilitating enzymatic turnover and photosynthetic water oxidation. This proposed mechanistic role for MSP incorporates the idea that the photo-induced events in the reaction center prepare the conformational state (i.e. deprotonated MSP), which is important in facilitating the next step in the reaction. Our results demonstrate that MSP loses a proton(s) on the S 1 to S 2 transition. The location of the base, which picks up these protons, is of interest. Our analysis of the isotope-edited spectrum suggests that the proton(s) is lost to a site, the protonation of which is not detectable in the isotope-edited FT-IR spectrum. Because only sites on MSP will contribute to the isotope-edited double difference spectrum, we conclude that the proton acceptor is not located on MSP. One possibility is that amino acid residues on the hydrophobic subunits, such as CP47, D1, and D2, act as the base. Two additional possibilities for the proton acceptor, given the spectral range employed, are the solvent or the manganese cluster itself, perhaps due to protonation of a -oxo-bridge. Protonation of -oxo linkages in the manganese cluster has been proposed as a method by which the reactivity of the cluster can be controlled (60). A protonation of a -oxo bridge or solvent would not be detectable in the 1800 -1200 cm Ϫ1 region (48). We do not favor the explanation that solvent is acting as the proton acceptor, because a recent study of PSII core particles in glycerol has shown that the S 1 to S 2 transition does not lead to a stoichiometric amount of proton release (61).
Our work has led us to propose that MSP acts as an accu- mulator of negative charges and a proton donor during the early S transitions and that this behavior is important in the catalytic mechanism. We have also provided a rationalization of the inhibitory effects, observed in the absence of MSP, on the kinetics of the enzyme. Although MSP is not absolutely required for oxygen evolution, this subunit plays an important role as a catalyst of the reaction, i.e. in the acceleration of the water oxidation process. Our observations lead to the intriguing hypothesis that MSP may serve as a general acid/base catalyst of photosynthetic water oxidation (for recent examples of general acid/base catalysts in other systems, see Refs. [62][63][64][65][66][67][68]. In future experiments, we will use site-directed mutagenesis to determine the location of these amino acid residues. It should be noted that at the present time, there is conflicting evidence from chemical modification and site-directed mutagenesis experiments concerning possible functional roles of carboxylates on MSP (69 -72).