Characterization of the light-induced cross-linking of the alpha-subunit of cytochrome b559 and the D1 protein in isolated photosystem II reaction centers.

Illumination of the isolated reaction center of photosystem II generates a protein of 41 kDa molecular mass. Using immunoblotting, it is confirmed that the protein is an adduct of the D1 protein and the α-subunit of cytochrome b559. Its formation seems to be photochemically induced, being independent of temperature between 4 and 20°C and unaffected by a mixture of protease inhibitors. The maximum levels are detected when the pH is in the region 6.5-8.5 and when illumination intensities are moderate. Although higher light intensities induce a higher rate of formation, the accumulation of elevated levels of the 41-kDa protein does not occur due to light-induced degradation. This degradation is also unaffected by the presence of protease inhibitors. Proteolytic mapping and N-terminal sequencing indicates that the cross-linking process involves the N-terminal serine of the α-subunit of cytochrome b559 and D1 residues in the 239-244 FGQEEE motif close to the QB binding site. In conclusion, the results indicate that the N terminus of the α-subunit is exposed on the stromal side of photosystem II in such a way as to undergo light-induced cross-linking in the QB region of the D1 protein. They also suggest that the 41-kDa adduct may be an intermediate before the light-induced cleavage of the D1 protein in the FGQEEE region.

Illumination of the isolated reaction center of photosystem II generates a protein of 41 kDa molecular mass. Using immunoblotting, it is confirmed that the protein is an adduct of the D1 protein and the ␣-subunit of cytochrome b 559 . Its formation seems to be photochemically induced, being independent of temperature between 4 and 20°C and unaffected by a mixture of protease inhibitors. The maximum levels are detected when the pH is in the region 6.5-8.5 and when illumination intensities are moderate. Although higher light intensities induce a higher rate of formation, the accumulation of elevated levels of the 41-kDa protein does not occur due to light-induced degradation. This degradation is also unaffected by the presence of protease inhibitors. Proteolytic mapping and N-terminal sequencing indicates that the cross-linking process involves the N-terminal serine of the ␣-subunit of cytochrome b 559 and D1 residues in the 239 -244 FGQEEE motif close to the Q B binding site. In conclusion, the results indicate that the N terminus of the ␣-subunit is exposed on the stromal side of photosystem II in such a way as to undergo lightinduced cross-linking in the Q B region of the D1 protein.
They also suggest that the 41-kDa adduct may be an intermediate before the light-induced cleavage of the D1 protein in the FGQEEE region.
Photosystem II (PSII) 1 is the oxygen-evolving system of photosynthesis. It is a membrane-located multicomponent, pigment protein complex that functions as a light-driven water/ plastoquinone oxidoreductase (1,2). The heart of the PSII complex is the reaction center, in which the initial primary and secondary charge separation occurs. In several respects, the PSII reaction center shares features similar to those of the purple photosynthetic bacteria (3). In the bacterial reaction center, the cofactors that catalyze charge separation are bound to a heterodimer consisting of the L and M subunits. Determi-nation of the structure of the reaction center of Rhodopseudomonas viridis gave information at the level of atomic resolution of the organization of the cofactors and the polypeptide chains of the L and M subunits (4,5). The amino acid sequences of the reaction center proteins, D1 and D2, share sufficient homologies with the L and M subunits to allow a reasonable model to be proposed, describing how these proteins fold and how the redox-active cofactors are bound (6 -8).
From the comparative approach and from direct experimental evidence (9,10), it is quite clear that the D1 protein is analogous to the L-subunit of the bacterial system. Nevertheless, D1 protein shows the unusual characteristic of rapid turnover that is greater than in any other PSII protein and indeed faster than most other proteins in the photosynthetic membrane (11). In contrast, the L-subunit does not turn over rapidly. The phenomenon of rapid turnover of D1 protein is even more amazing since, like the L-subunit, it forms the active branch of the PSII reaction center, being involved in both donor and acceptor side redox reactions.
There is good reason to believe that the rapid turnover of D1 protein is a consequence of the vulnerability of PSII to photodamage and that this effect is the basis of the physiological phenomenon of photoinhibition (12,13).
In studying the rapid turnover of D1 protein in vivo, Greenberg et al. (14) concluded that the degradation involved a cleavage that yielded a 23.5-kDa fragment containing the N terminus of the protein. A similar photoinduced fragment has also been generated using in vitro systems (15)(16)(17). In the case of isolated PSII reaction centers, it was shown that a 23-kDa N-terminal fragment of the D1 protein is observed when they are illuminated under aerobic conditions in the absence of exogenous electron acceptor (17). Under such conditions, recombination occurs between the radical pair state consisting of the oxidized primary electron donor (P680 ϩ ) and reduced primary electron acceptor pheophytin a (pheo Ϫ ). This recombination forms a triplet state of P680 (18), which is able to generate singlet oxygen (19,20). It is likely, therefore, that this highly reactive form of oxygen causes damage to the D1 protein. The damage itself does not seem to be directly involved in the cleavage process but triggers the D1 protein for degradation (21) possibly via a conformational change (22). Because this isolated complex consists only of the D1 and D2 proteins, the ␣and ␤-subunits of cytochrome b 559 (Cyt b 559 ), and the product of the psbI gene (9,23), it has been concluded that the mechanism of cleavage is contained within the reaction center and that no external proteases are involved. For this reason, we have undertaken a study to elucidate the nature of the autocatalytic cleavage, which generates the 23-kDa N-terminal fragments. Previous studies have indicated that the cleavage occurs on the C-terminal side of residue 238 in the D1 sequence (14,24). As emphasized by Greenberg et al. (14), this cleavage site is lo-cated in the loop joining putative transmembrane segments IV and V, which is in the vicinity of the Q B binding region (6).
The investigation that we report here was stimulated by the observation that in addition to the appearance of the lightinduced 23-kDa N-terminal fragment, a band was induced at about 41 kDa when isolated reaction centers were exposed to strong illumination. This band was reported by Shipton and Barber (25) and subsequently shown, by immunological blotting, to consist of D1 protein and the ␣-subunit of Cyt b 559 (26). Here, we further characterize the properties of this D1 protein/ ␣-subunit Cyt b 559 adduct and consider how the formation of a covalent linkage between these two reaction center subunits could be involved in events leading to D1 protein cleavage and the generation of the 23-kDa N-terminal fragment.

Isolation of PSII Reaction Centers and Irradiation
Conditions-Photosystem II reaction center (RCII) complexes were isolated from pea and wheat chloroplast membranes as described by Nanba and Satoh (9) with modifications described by Chapman et al. (27). Irradiation of RCII complexes was performed, if not otherwise stated, in a thermostatic cuvette at 4°C at a chlorophyll concentration of 50 g ml Ϫ1 , using a Flexilux 650 lamp able to give fluence rates up to 2000 mol m Ϫ2 s Ϫ1 . Irradiation was carried out in 50 mM Tris-HCl, pH 8.0, containing 2 mM n-dodecyl ␤-D-maltoside. For pH-dependence experiments, Bis-Tris propane, MES, and Tris were used as buffer systems.
SDS-PAGE and Limited Proteolysis-SDS-PAGE in the presence of 6 M urea was carried out as described by Barbato et al. (28) using a 12-18% linear acrylamide gradient. The stacking gel did not contain urea. Gels were stained with acid-free Coomassie Brilliant Blue R-250 (0.5% in 40% methanol) and destained in 3 mM Na 2 CO 3 , 10 mM NaHCO 3 containing 20% methanol. For proteolysis, the method of Cleveland et al. (29) was used. The stained bands were cut out from the gel and loaded onto the stacker of a second identical gel and overlaid with the desired amount of protease (dissolved in 62.5 mM Tris-HCl, pH 6.8, 0.1% SDS, and trace amounts of bromphenol blue). When the tracking dye was near to the bottom of the stacking gel, the current was switched off for 3 h to allow proteolysis to proceed, and then electrophoresis was carried out.
Immunoblotting-Proteins separated on SDS-PAGE intended for immunoblotting were transferred onto nitrocellulose membranes (sartorius, 0.45 m) using the transfer buffer described by Dunn (30). Filters were stained with 0.2% Red Ponceau S in 3% trichloroacetic acid and destained in Tris-buffered saline. For immunodetection, filters were incubated in rabbit primary antibody and then with goat-anti-rabbit biotinylated IgG and streptavidine alkaline phosphatase-conjugate (Sigma). Visualization of immunocomplexes was obtained by adding suitable chromogenic substrates.
For N-terminal sequencing, proteins separated on gels were transferred onto polyvinylidene difluoride membranes (ProBlot, Applied Biosystems) according to Matsuidara (31), using 10 mM CAPS, pH 11. Gel solutions were stirred overnight on Amberlite MB1, and 1 mM 2-dimethylamino-ethanethiol hydrochloride was added to all buffers. To locate bands, filters were stained with 0.4% Coomassie Brilliant Blue R-250 in 40% methanol and 1% acetic acid and destained in 40% methanol. N-terminal sequencing was performed by automated Edman degradation using an Applied Biosystems protein/peptide sequencer (model 477A).
Source of Antibodies-Properties of polyclonal antisera used in this study have been previously described. Anti-D1N was raised to the N-terminal fragment produced by Lys-C digestion of wheat D1 protein (32); anti-D1C was raised against a synthetic peptide corresponding to the 19 C-terminal residue of the spinach protein and was a kind gift from Dr. Peter Nixon; anti-Cyt ␣ was raised against the ␣-subunit of Cyt b 559 isolated by preparative SDS-PAGE (26).
Chlorophyll concentration was measured according to Arnon (33). Densitometric analyses of immunostained gels were performed using a Hirchmann densitometer. Fig. 1 shows a Western blot using an anti-D1N antibody of isolated PSII reaction center proteins separated by SDS-PAGE. Lane 0 is the dark control, while the subsequent lanes show the effect of preilluminating the iso-lated PSII reaction centers with 2000 mol quanta m Ϫ2 s Ϫ1 of white light at pH 6.5 for the periods of time indicated. As a consequence of this treatment, a breakdown fragment of the D1 protein is detected at about 23 kDa, and a band appears at 41 kDa. It has been previously shown, and is confirmed here, that the 23-kDa fragment contains the N terminus part of D1 protein and therefore results from a cleavage in the loop joining the putative transmembrane segments IV and V (17). In contrast, the 41-kDa fragment has been shown to be an adduct of the D1 protein and the ␣-subunit of Cyt b 559 (26,34). To obtain more information about the nature of this adduct, proteolytic mapping was conducted.

Generation of a 41-kDa Adduct of the D1 Protein and the ␣-Subunit of Cyt b 559 -
Lys-C Proteolysis Indicates that the Site for Light-induced Cross-linking between the D1 Protein and the ␣-Subunit of Cyt b 559 Is on the C-terminal Side of the D1 Residue 238-The 41-kDa adduct was isolated from PSII reaction centers prepared from wheat using SDS-PAGE after illumination (Fig. 2). Wheat reaction centers were used because in this species the D1 protein contains a single lysine residue at position 238, and the ␣-subunit of Cyt b 559 contains no lysine residues (35). Given this situation, we were able to use the endoprotease Lys-C to specifically cut the D1 protein at residue 238.
Panel A of Fig. 2 shows an immunoblot with anti-D1N antiserum of the 41-kDa adduct (lanes 1 and 3) and isolated D1 protein (lanes 2 and 4), either undigested (lanes 1 and 2) or digested (lanes 3 and 4) with Lys-C protease. As can be seen in lane 4, a 19-kDa fragment of the D1 protein is detected by anti-D1N, which corresponds to the 2-238 segment of this protein (32,36). The same 19-kDa fragment was observed when the 41-kDa band was partially digested by Lys-C (arrow on lane 3A). When an identical blot was probed with an antiserum to the C-terminal region of the D1 protein (anti-D1C), the results presented in panel B were obtained. In this case, a 10-kDa fragment of the D1 protein was detected (arrow on lane 4B), which is the 239 -344 C-terminal segment. Indeed, Nterminal sequencing confirmed that the first 10 residues of this fragment corresponded to residues 239 -248 (see Table I). This 10-kDa C-terminal fragment, however, was not detected as a consequence of partial digestion of the 41-kDa adduct; instead, the anti-D1C detected a 21-kDa band (arrow on lane 3B), which is thought to contain the 10-kDa C-terminal portion of the D1 protein and the ␣-subunit of Cyt b 559 . Indeed, this was confirmed by blotting with anti-Cyt ␣ antiserum (arrow on lane 3 of panel C in Fig. 2).
As the ␣-subunit of Cyt b 559 does not contain any lysine residues (35), Lys-C proteolysis had no effect on this isolated subunit (compare lanes 2C and 4C in Fig. 2). Therefore, the above experiments show unequivocally that the site for lightinduced cross-linkage is located at an amino acid residue lying on the C-terminal side of lysine-238 of the D1 protein.
V8 Proteolysis Indicates That the Cross-linking Site Is Located on the N-terminal Side of Residue 244-A second series of experiments were conducted using Staphylococcus aureus V8 protease. Fig. 3 shows the effect of treatment of the ␣-subunit of Cyt b 559 with two concentrations of this enzyme. In both cases, a main digestion band (SaCyt ␣ ) was detected, by either Coomassie staining (panel A) or immunoblotting with anti-Cyt ␣ (panel B), which had a molecular mass of about 1 kDa lower than the native band. N-terminal sequencing of both bands (see Table I) gave the same sequence, indicating that the V8-induced proteolytic cleavage had occurred toward the C-terminal end of the ␣-subunit.
The effect of V8 proteolysis of the isolated D1 protein and the 41-kDa adduct was also investigated, as shown in Fig. 4. Immunoblotting with anti-D1C antiserum (panel A) showed that the digestion of D1 with V8 (lane 2) resulted in the appearance of two or, more often, three bands in the range of 8 -10 kDa. These bands probably correspond to Sa8 and Sa10 doublet described by Marder et al. (37) and Greenberg et al. (14). As these bands were recognized by the C-terminal specific antibody (raised to the last 19 amino acid residues of the protein), they must contain the C terminus of D1. In particular, it was found by N-terminal sequencing (see Table I) that the Sa8 fragment is derived from a cleavage at glutamate 244, as suggested previously by Marder et al. (37). In the case of V8 proteolysis of the 41-kDa adduct (lane 4 of panel A), the anti-D1C antibody recognizes three fragments in the range of 15-20 kDa and a 8-kDa fragment. As this last fragment has the same electrophoretic mobility and immunological reactivity as the Sa8 fragment from D1, it is reasonable to assume that they are the same proteolytic fragment (i.e. from 245 to the C terminus of the D1 protein (see Fig. 9)).
When anti-Cyt ␣ was used to probe V8 proteolytic digestion products of the 41-kDa adduct (panel B, lane 1), a number of bands were detected in the 15-20-kDa range. Of note is that two of them are the same as those detected by the anti-D1C antibody (marked by circles in lane 4 of panel A and lane 1 of  panel B). Since the Sa8 fragment was detected after treatment of the 41-kDa adduct with V8, it seems likely that cross-linking of the ␣-subunit and the D1 protein occurs on the N-terminal side of residue 244. This conclusion, taken together with the result derived from the Lys-C experiments (Fig. 2), suggests that the linkage takes place between 239 and 244 of D1 protein, i.e. in the FGQEEE motif. Detection of the V8-induced doublet of the 41-kDa band by both anti-D1C and anti-Cyt ␣ indicates that these contain the branched peptide formed by the ␣-sub-  1A, 1B, and 1C), undigested D1 (lanes 2A and  2B), undigested cyto ␣ (lane 2C), Lys-C-digested 41-kDa band (lanes 3A, 3B, and 3C), Lys-C digested D1 (lanes 4A and 4B), and Lys-C-digested cyto ␣ (lane 4C). Wheat RCII complexes were irradiated at a chlorophyll concentration of 50 g/ml for 15 min with 600 mol m Ϫ2 s Ϫ1 white light to induce the formation of the 41-kDa adduct. After SDS-PAGE and acid-free Coomassie staining, the corresponding band was cut out from the gel and loaded on the stacker of a second identical gel. For proteolysis, acrylamide bands were overlaid with 20 l of Lys-C (10 units/ml), and electrophoresis was started. When the tracking dye reached the bottom of the stacking gel, power was switched off for 3 h to allow proteolysis; gels were then run as usual. See Table I for N-terminal sequence of the C-terminal fragment of D1 induced by Lys-C treatment (arrow on lane 4B). Bands marked with a circle in lanes 2C and 4C are aggregation products of cyto ␣ . Arrows on lanes 3B and 3C emphasize the identity of the fragments detected by anti-D1C and anti-cyto ␣ from Lys-C proteolysis of the 41 kDa. Arrow on lane 3A marks the N-terminal fragment from Lys-C proteolysis of the 41-kDa adduct and D1, which represents the 2-238 segment of the protein.   Fig. 5, A  and B. In Fig. 5A, it can be seen that the initial increase in the level of the D1/Cyt b 559␣ adduct depends on the light intensity being enhanced at higher intensities. However, at higher fluence rates, the final level after 30 min of illumination is lower than observed when the light intensities are more moderate. This finding suggests that the actual level of the 41-kDa band depends on the relative rates of its formation and degradation. In Fig. 5B, an immunoblot with anti-Cyt ␣ antiserum is shown after irradiating isolated PSII RC complexes with 400 mol quanta m Ϫ2 s Ϫ1 for different periods of time. As can be seen, the level of the 41-kDa adduct increases with time, with the parallel appearance of a new weak band at about 20 kDa.
When irradiation was performed in the presence of a mixture of protease inhibitors, no effect on the 41-kDa band was observed, neither on its appearance nor disappearance (data not shown). Similarly, as Fig. 6 shows, formation of the 41-kDa adduct was independent of temperature over the range 4 -20°C.
In further experiments, the effect on the presence/absence of oxygen and pH were investigated. Very low levels of oxygen (less than 1 M) were obtained by flushing the RCII suspension with oxygen-free nitrogen and by including a chemical trap consisting of 10 mM glucose, 0.1 mg ml Ϫ1 catalase, and 0.1 mg ml Ϫ1 glucose oxidase. The results of this experiment are shown in Fig. 7, where it can be seen that only those RCII samples illuminated under aerobic conditions (lanes 6 -8) produced the 41-kDa band due to cross-linking of D1 protein and the ␣-subunit of Cyt b 559 . For studying the pH sensitivity, the RCII preparation was suspended in buffers ranging in pH from 5.5 to 9.5. As Fig. 8 shows, irradiation at pH 5.5 does not result in the appearance of the 41-kDa adduct. At pH 9.5, the formation of the adduct is also severely inhibited. In contrast, the light- induced 41-kDa band is readily detected at pH values 6.5, 7.5, and 8.5. When a sample that had been irradiated at pH 7.5 (i.e. containing the 41-kDa band) was incubated in the dark at pH 5.5 or 9.5, no effect on the level of the adduct was noted, an observation that rules out the possibility that the band is not observed in Fig. 7 because of its instability toward pH 5.5 or 9.5.

DISCUSSION
The immunological blotting data presented above are consistent with the previous conclusion that the 41-kDa band generated by illuminating isolated PSII RCs is a cross-linked adduct of the D1 protein and the ␣-subunit of Cyt b 559 . Treatment of this adduct, isolated from illuminated RCIIs of wheat with Lys-C, generated a 21-kDa fragment that consisted of the C-terminal portion of the D1 protein and the ␣-subunit of Cyt b 559 . Thus, the cross-linking occurs on the C-terminal side of residue 238 of the D1 protein. The C-terminal portion of the D1 protein (from 239 to 344) was found to have an apparent molecular mass of about 10 kDa, while the ␣-subunit of Cyt b 559 has about the same mass, thus accounting for the observed 21-kDa value for the Lys-C-induced fragment of the adduct.
Experiments using S. aureus V8 protease further suggest that the cross-linking site is on the N-terminal side of the tyrosine residue 245 of the D1 protein. This conclusion is based on the fact that the Sa8 fragment formed due to V8 cleavage at glutamate 244 on the D1 protein is also generated when the 41-kDa adduct is digested with this enzyme (e.g. Fig. 3, lane A4). Thus, these results lead to the overall conclusion that the adduct is formed by a cross-linking of the ␣-subunit of Cyt b 559 with a residue in the 239 -244 motif FGQEEE (see Fig. 9).
The N terminus of the ␣-subunit of Cyt b 559 is not blocked and can be sequenced by Edman degradation as indicated in Table I. In contrast, the N terminus of the D1 protein is blocked and cannot be readily sequenced by the Edman procedure. We have also found that the 41-kDa adduct is not amenable to N-terminal sequencing using Edman degradation. The reason for this is unclear, but it could be that the N terminus of the ␣-subunit is not available in the adduct as it is in the free protein. This therefore suggests that the cross-linking to the FGQEEE motif involves the N-terminal residue of the ␣-subunit, which is a serine. The action of V8 on the 41-kDa adduct is also consistent with this since the products of this proteolysis did not indicate an attachment of the C-terminal end of the ␣-subunit to the D1 protein where the Cyt subunit has a V8 cleavage site (e.g. Fig. 4C) The identified cross-linking site on the D1 protein has the unusual feature of having three glutamic acid residues adjacent to each other. Such a motif is not typical and occurs in only a few proteins (24). The pK a value for a single glutamic acid residue is 4.3 and thus at neutral pH would be unprotonated. However, the existence of three adjacent carboxyl groups of glutamic acid would be expected to significantly shift the pK a to a higher value so that protonation would occur at neutral pH. This shift of pK a could aid a condensation reaction with the N-terminal serine of the ␣-subunit, thus forming a peptide linkage. Such a reaction is endothermic and would not be expected to happen spontaneously. Indeed, the cross-linking is induced by illumination. The mechanism involved does not seem to be enzymic (lack of effect of protease inhibitors and temperature) but involves a physically induced reaction driven by the absorption of light energy. Moreover, we have also shown that oxygen needs to be present for the adduct to appear. It is difficult to perceive how the proposed condensation reaction is driven. Under anaerobic conditions the heme of Cyt b 559 becomes photoreduced, but no such reaction occurs when oxygen is present (38). It is conceivable that the formation of the adduct is dependent on the redox state of the heme, which affects the conformation of the ␣-subunit. Alternatively, lightinduced oxidation processes could aid the cross-linking.
Although the precise mechanism for the formation of this adduct is not clear, the identification of the linkage site has at least two important implications. The first is that the N terminus of the ␣-subunit must be exposed on the stromal side of PSII, as suggested by Tae et al. (39) and Vallon et al. (40). Moreover, our data indicate that Cyt b 559 may be more closely located at the D1 rather than the D2 protein and that the N terminus of the ␣-subunit is positioned near to the Q B site. This conclusion would mean that the heme, which is probably ligated to the histidine residues 22 and 17 on the ␣and ␤-subunit, respectively (39), is more intimate with Q B than Q A . No cross-linking product was observed with the ␤-subunit, which, like the psbI gene product, is blocked at its N terminus (23).
The second important implication is related to our finding that the formation of a D1 protein/Cyt b 559 ␣-subunit adduct occurs via cross-linking in the 239 -244 motif of the D1 protein.
Studies of D1 protein degradation in both in vivo and in vitro systems identify the 239 -244 region as the primary site of cleavage when acceptor side photoinhibition occurs, generating a 23-kDa N-terminal and a 10-kDa C-terminal fragment of the D1 protein. Studies with isolated reaction centers have shown that the 41-kDa adduct is formed prior to the appearance of the 23-kDa N-terminal D1 cleavage product (e.g. see Fig. 1 and Ref. 17). It is possible, therefore, that the cross-linking of the ␣-subunit to the D1 protein is the first step in the sequence of events leading to the primary cleavage of this protein. The formation, for example, of a branch peptide linkage could possibly aid a hydrolysis reaction, resulting in cleavage of the D1 protein in the FGQEEE motif without causing any degradation of the ␣-subunit itself in line with other observations (41). However, further experiments are required to give credibility to this idea. Among the experiments to be carried out will be those that identify whether the linkage and cleavage processes involve the same specific residues.