Light-harvesting complex II binds to several small subunits of photosystem I.

Mobile light-harvesting complex II (LHCII) is implicated in the regulation of excitation energy distribution between Photosystem I (PSI) and Photosystem II (PSII) during state transitions. To investigate how LHCII interacts with PSI during state transitions, PSI was isolated from Arabidopsis thaliana plants treated with PSII or PSI light. The PSI preparations were made using digitonin. Chemical cross-linking using dithio-bis(succinimidylpropionate) followed by diagonal electrophoresis and immunoblotting showed that the docking site of LHCII (Lhcb1) on PSI is comprised of the PSI-H, -L, and -I subunits. This was confirmed by the lack of energy transfer from LHCII to PSI in the digitonin-PSI isolated from plants lacking PSI-H and -L. Digitonin-PSI was purified further to obtain an LHCII.PSI complex, and two to three times more LHCII was associated with PSI in the wild type in State 2 than in State 1. Lhcb1 was also associated with PSI from plants lacking PSI-K, but PSI from PSI-H, -L, or -O mutants contained only about 30% of Lhcb1 compared with the wild type. Surprisingly, a significant fraction of the LHCII bound to PSI in State 2 was not phosphorylated. Cross-linking prior to sucrose gradient purification resulted in copurification of phosphorylated LHCII in the wild type, but not with PSI from the PSI-H, -L, and -O mutants. The data suggest that migration of LHCII during state transitions cannot be explained sufficiently by different affinity of phosphorylated and unphosphorylated LHCII for PSI but is likely to involve structural changes in thylakoid organization.

In oxygenic photosynthesis two photosystems, PSI 1 and PSII, work in series to convert light energy into chemical energy. PSI is also involved in cyclic electron transport without the participation of PSII, and this process serves to produce additional ATP and to regulate the transthylakoidal proton gradient (1), but in plants this is a minor part of the electron transport. In linear electron transport, PSI and PSII must operate with the same rate, but natural environmental conditions, such as the quality and quantity of light, are constantly fluctuating, and this may alter the balance between the two photosystems. The two photosystems have different absorption spectra, and therefore a change in light quality may favor one photosystem over the other. However, plants can balance the excitation energy distribution between the two photosystems via a mechanism known as state transitions, which was discovered more than 30 years ago (2)(3)(4). If PSII is overexcited relative to PSI, the plastoquinone pool becomes overreduced, and this will activate a kinase that phosphorylates a mobile pool of light-harvesting complex II (LHCII), leading to the lateral movement of LHCII in favor of PSI. This is the so-called "State 2," in which the PSII antenna is smaller and the PSI antenna is larger than in State 1 (5,6). State 1 is obtained when PSI is preferentially excited, which leads to oxidation of the plastoquinone pool and inactivation of the LHCII kinase. The phospho-LHCII is then dephosphorylated by a redox-independent phosphatase and moves back to PSII. Although the phenomenon of state transitions has been recognized for a long time, there is still considerable uncertainty about the mechanism. Two models have been proposed to explain the movement of LHCII. According to one model, alteration in the surface charge upon phosphorylation leads to structural changes of the thylakoid membrane and results in the movement of phospho-LHCII away from grana stacks (7)(8)(9). According to another model, the net movement of LHCII toward PSI in State 2 is caused by PSII with higher affinity for unphosphorylated LH-CII and PSI with higher affinity for phospho-LHCII, therefore movement of phospho-LHCII is a question of molecular recognition (6). Although the models differ in the way of explaining state transitions, they both involve phosphorylation of LHCII as a prerequisite for the initiation of State 1-State 2 transitions. A search for kinases involved in the phosphorylation of thylakoid proteins has been carried out by many workers for more than 20 years. A family of proteins, thylakoid-associated kinases, was identified as good candidates for LHCII kinases (10,11). The antisense Arabidopsis plants with low amounts of thylakoid-associated kinase 1 showed a lower level of LHCII phosphorylation and were deficient in state transitions, but the phosphorylation of LHCII was distributed equally between PSII and PSI under white light, which was also the case in wild type plants (11). This indicates that the correlation between LHCII phosphorylation and state transition is complex. Recently, Depege et al. (12) reported a novel kinase in Chlamydomonas reinhardtii, thylakoid-associated serine-threonine protein kinase, and demonstrated that it is required for the phosphorylation of LHCII and for state transitions.
LHCII consists of three different proteins, Lhcb1, Lhcb2, and Lhcb3. Lhcb1 and Lhcb2 are the most abundant and can form Lhcb1 homotrimers and Lhcb1/2 heterotrimers, which are believed to be a mobile complex (13,14). Both proteins usually * This work was supported by the Nordic Joint Committee for Agricultural Research and by the Danish National Research Foundation. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
‡ To whom correspondence should be addressed. exist in several very similar isoforms, but specific isoforms are not well conserved between species, indicating that they probably are redundant rather than having specific functions. However, this is another unclear point because it is not known whether there are biochemical differences between the mobile and nonmobile LHCII apart from the reversible phosphorylation described above. Electron microscopy studies have revealed that PSII core complexes are found as dimers surrounded by LHCII trimers (15)(16)(17)(18). There are specific binding sites for the trimers, but the number of trimers/PSII complex differs depending on species and growth conditions. Generally, strongly bound, intermediately bound, and loosely bound trimers can be recognized. In contrast to the situation with PSII, the binding site of LHCII on PSI is not known.
PSI is composed of a multisubunit core complex (PSI core) and outer antenna, the light-harvesting complex I (LHCI). The PSI core complex in higher plants consists of 14 different subunits (PSI-A to PSI-L, PSI-N) (19,20) and a recently discovered subunit PSI-O (21). LHCI is composed of four different subunits of about 21-24 kDa, Lhca1 to Lhca4.
The association between LHCII and PSI is relatively weak, and a stable complex is difficult to purify. Early studies (22) suggested that LHCII was bound to LHCI rather than directly to the PSI core, but this conclusion was based on the inability to reconstitute a LHCII⅐PSI complex using PSI devoid of LHCI-680 (composed of Lhca2 and Lhca3). We now know that preparation of a PSI complex devoid of LHCI is likely to cause the loss of additional small core subunits that were unknown at the time. Electron microscopy studies have revealed that all LHCI subunits bind at the side of PSI-F and PSI-J subunits of the PSI core complex (23). Arabidopsis plants lacking PSI-H were highly deficient in state transitions and have identical PSI antenna size in both States 1 and 2, whereas in wild type the antenna size of PSI was found to increase about 33% during transition to State 2 (5). Therefore we suggested that LHCII binds directly to the PSI core and that PSI-H is part of a docking site (5). Based on cross-linking and x-ray crystallography data, the PSI-H subunit is positioned close to PSI-L and PSI-I on the opposite side of the PSI core complex compared with PSI-F, PSI-J, and LHCI (24,25). However, until now, no direct biochemical evidence has ever been found for the association of LHCII to PSI during state transitions.
To investigate how LHCII interacts with PSI during state transitions, we purified LHCII⅐PSI complex from wild type Arabidopsis in State 1 and State 2, and we also used mutants lacking the PSI-H, -L, -O, or -K subunits to determine the docking sites for LHCII. The results indicate that PSI-H, -L, -O, and -I all participate in forming the docking site. Furthermore, the results show that both phosphorylated and unphosphorylated LHCII are associated with PSI in State 2.

EXPERIMENTAL PROCEDURES
Plant Materials-Arabidopsis thaliana (L.) Heyhn ecotype Col-0 was used for all experiments. Plants were grown in peat in a controlled environment Arabidopsis chamber (Percival AR-60L, Boone, IA) at a photosynthetic flux of 100 -120 mol photons m Ϫ2 s Ϫ1 , 20°C, 70% relative humidity. The photoperiod was 8 h. Transformants lacking specific PSI subunits were obtained by antisense or cosuppression and have been reported before (5, 26 -28, PSI-O mutant). 2 Light Treatment-Colored filters were used in the experiments to provide PSII light (L2) and PSI light (L1) essentially as described by Pfannschmidt et al. (29). PSII light was obtained with an orange filter (Rosco, 105 orange, Teadon Aps, Stenløse, Denmark), and PSI light was obtained with a red filter (HT 027 medium red, LEE Filters, Andover, UK). Gray filter (209 neutral density, LEE Filters) was used to adjust the light coming through the orange filter to a level similar to that through the red filter (50 -70 mol photons m Ϫ2 s Ϫ1 ). The filters were mounted in a controlled environment chamber equipped with 400-watt Powertone HPI-T Plus lamps (Philips). Six-week-old wild type plants were exposed to PSI or PSII light for 1 h prior to harvesting of leaves. Plants lacking PSI-H, -L, -O, or -K subunits were treated with PSII light for 1 h.
State Transitions in Leaves-State transitions were measured with a pulse amplitude modulation 101-103 fluorometer (Walz, Effeltrich, Germany) in the growth chamber equipped with filters as described above. Plants were dark-adapted for 30 min before the measurements. A detached leaf from a wild type plant was fixed to the light fiber, which was positioned so the leaf was horizontal and received the same irradiation as the plants used for preparation of thylakoids. Maximum fluorescence yield (F m ) was determined by exposing the leaf to a saturating flash (0.8 s, 6000 mol photons m Ϫ2 s Ϫ1 ). The leaf was then exposed to orange light (which favors PSII, L2) for about 20 min, and the maximum fluorescence yield in State 2 (F m2 ) was determined. The fiber with the leaf was then transferred rapidly to the red light (which favors PSI, L1), and after 20 min the maximum fluorescence yield (F m1 ) was measured. The leaf was then exposed to orange light for another 20 min and finally kept in darkness. The difference of maximal fluorescence in State 1 and State 2 was calculated as F m1 /F m2 . The relative change in fluorescence was calculated as in Equation 1 (5,30), where F i and F ii designate fluorescence in the presence of PSI light in State 1 and State 2, respectively, and F i Ј and F ii Ј designate fluorescence in the absence of PSI light in State 1 and State 2, respectively (see Fig.  1).

Isolation of DM-PSI (␤-Dodecyl Maltoside-solubilized PSI) and
Digitonin-PSI-After plants were treated with light as described above, rosette leaves were rapidly harvested and frozen in liquid nitrogen. Thylakoids were prepared as described previously (31) except that all solutions contained 10 mM NaF to inhibit phosphatase. DM-PSI was isolated according to Jensen et al. (27). The sucrose gradients were prepared by freezing and thawing at 4°C of 11 ml of 0.6 M sucrose, 20 mM Tricine-NaOH, pH 7.5, 0.06% DM, and 10 mM NaF. Digitonin-PSI was prepared by solubilizing thylakoids (0.5 mg of Chl/ml) with digitonin (final concentration 0.5% (w/v)) for 30 min at 4°C with stirring, followed by centrifugation at 48,000 ϫ g for 30 min. The supernatant was centrifuged at 257,000 ϫ g for 1 h at 4°C. The pellet was resuspended in 20 mM Tricine, pH 7.5, 20% glycerol, 10 mM NaF, and the samples (digitonin-PSI) were frozen in liquid nitrogen and stored at Ϫ80°C until use. The Chl concentration and Chl a/b ratio were determined in 80% acetone (32).
Preparation of LHCII⅐PSI Complex-For preparing pure LHCII⅐PSI complex, the digitonin-PSI preparations were diluted to 0.3 mg of Chl/ml in 20 mM Tricine, pH 7.5, 0.3% (w/v) DM, 10 mM NaF. Half of the preparation was treated with chemical cross-linker dithio-bis(succinimidylpropionate) (DTSP) on a shaker for 30 min at 22°C in darkness as described previously (24). After incubation with 0.015% (w/v) DTSP at room temperature for 30 min in darkness, the reaction was stopped by the addition of 1/7 volume of 10 mM Tris, pH 7.4, 1 mM EDTA. The DTSP-treated and the control samples were applied to sucrose gradients containing 0.06% (w/v) DM and centrifuged as described above. The LHCII bands and LHCII⅐PSI complex bands were collected with syringes, frozen in liquid nitrogen, and stored at Ϫ80°C.
Analysis of Phosphate Content-Phosphate content of LHCII in State 2 was analyzed according to Martensen (33). First the purified LHCII preparations were ashed to convert protein-bound phosphate to inorganic phosphate, then the inorganic phosphate was measured by absorbance at 660 nm in a microtiter plate reader after complexation of ammonium molybdate with malachite green. The standard curve was made with KH 2 PO 4 (0 -1 nmol). The content of phosphate in LHCII⅐PSI in State 2 was determined by immunoblotting using the phosphothreonine antibody and comparing with a dilution series of LHCII with known content of phosphate.
SDS-PAGE and Immunoblotting-All the gels used were home-made 8 -25% gradient gels prepared according to Fling and Gregerson (34), except that the first dimension of diagonal electrophoresis was on 12% homogeneous gels. All samples were loaded on Chl basis. Immunoblotting was carried out by transferring the electrophoretically separated proteins to nitrocellulose membranes followed by incubation with different polyclonal antibodies and detection with the ECL system (Amersham Biosciences). To determine the composition of cross-linked products on one-dimensional gels, the nitrocellulose strip for each lane was cut into two sections and incubated with two different antibodies.
In this way the two sections could be aligned exactly. To determine the relative content of Lhcb1, Lhcb2, and phosphothreonine in the LHCII⅐PSI complexes each sample was electrophoresed together with a dilution series of known amounts of LHCII to ensure that the response on the immunoblots was in a linear range. Quantitation was carried out by scanning the x-ray films and analyzing them using the ImageQuant software (Molecular Dynamics).
Polyclonal antibodies against PSI proteins were prepared in rabbits and have been described before (21,35). Antibodies against Lhcb1 and Lhcb2 were kind gifts of S. Jansson, Umeå University, Sweden. Antibodies against phosphothreonine were obtained from Cell Signaling Technology, Inc. (Beverly, MA).
Diagonal Electrophoresis-The digitonin-PSI preparations (treated with PSII light) were diluted to 0.3 mg of Chl/ml in 20 mM Tricine, pH 7.5, 0.3% (w/v) DM, and 10 mM NaF. Chemical cross-linking with DTSP was carried out as described above. The cross-linked samples were mixed with 1 volume of nonreducing sample buffer (50 mM Na 2 CO 3 , 15% w/v sucrose, 2.5% (w/v) SDS), the solution was incubated for 20 min at room temperature, and loaded on the first dimension 12% gel. After electrophoresis, the lanes were cut out and incubated for 30 min in reducing sample buffer (50 mM Na 2 CO 3 , 15% sucrose, 2.5% SDS, and 50 mM dithiothreitol) to obtain complete reductive cleavage of the crosslinked products. The gel slice was placed on top of a second 8 -25% gradient gel and reelectrophoresed.
P700 Photooxidation Measurements-The photooxidation kinetics of P700 were monitored with the absorbance changes at 810 nm with 860 nm as reference (⌬A 810 -860 ) using the dual wavelength unit ED-P700DW of a PAM 101-103 fluorometer (Walz) as described by Bukhov et al. (36) with some modifications. Actinic light from a KL1500 halogen lamp (Schott, Mainz, Germany) with different light intensities was controlled by an electronic shutter, which opened for 1 s with 40-s intervals. Digitonin-PSI preparations (15 g of Chl) were diluted to a total volume of 500 l in 50 mM Tricine buffer, pH 7.5, 0.01% (w/v) digitonin, and 0.5 mM ascorbate and 0.05 mM methyl viologen were added prior to measurements. Ascorbate was used to ensure complete reduction of P700 in the dark, and methyl viologen to prevent the back-reaction from PSI acceptors. The output signal from the fluorometer was passed to an oscilloscope where averaging of signals was performed. The half-life (t1 ⁄2 ) was calculated after fitting of exponential curves to the absorption curves.

RESULTS
State Transitions in the Growth Chamber-To make sure that we chose the right light conditions to do the following experiments, we first tested whether the plants performed state transitions in the growth chamber. To induce state transitions, we equipped a growth chamber with orange and red filters for preferential excitation of PSII and PSI, respectively. The light intensity passing through the filters was about 50 -70 mol photons m Ϫ2 s Ϫ1 , The plants were dark-adapted for 30 min before measuring the maximum fluorescence signal (F m ), and subsequently the leaves were exposed to orange light (L2, which favors PSII) for 20 min (Fig. 1). After the steady-state fluorescence level was reached, the maximal fluorescence in State 2 (F m2 ) was determined. Then State 1 was induced by illuminating with red light (L1, which favors PSI) for 20 min, and the maximal fluorescence in State 1 (F m1 ) was determined.
To quantify state transitions we determined F m1 and F r (30). For wild type Arabidopsis plants, the value of F m1 /F m2 was determined as 1.19 Ϯ 0.03, and F r as 0.86 Ϯ 0.05 (Ϯ S.D., n ϭ 4). The values are in good agreement with Haldrup et al. (30), who got the numbers 1.051 Ϯ 0.005 and 0.851 Ϯ 0.034 at a light intensity of 80 mol photons m Ϫ2 s Ϫ1 using a conventional setup. This indicates that the right experimental conditions were established for performing state transitions in our growth chamber.
Detergent Solubilization and Purification of PSI-From plants exposed to different light conditions we first used DM to solubilize thylakoids and then isolate PSI from sucrose gradients. As shown in Fig. 2A (lanes 1 and 2) the isolated PSI did contain all known higher plant PSI subunits but did not have any detectable LHCII attached both in State 1 and in State 2. The absence of LHCII in DM-PSI was confirmed by immunoblotting (data not shown). This demonstrates that the association between LHCII and PSI is weak and could be disrupted during solubilization with DM. Therefore, we used a milder solubilization with the nonionic detergent digitonin instead of DM and without running sucrose gradients. The resulting preparations were highly enriched in PSI and contained large amounts of LHCII ( Fig. 2A, lanes 3 and 4). However, the preparations are obviously not pure PSI but also contain some PSII. Thus, we could not at this point determine how large a fraction of the LHCII was actually associated with PSI. However, it is clear that the LHCII isolated together with PSI was much more phosphorylated in State 2 than in State 1 (Fig. 2B). This result further confirmed that the reversible phosphorylation was functional under the chosen light conditions and that state transitions were induced.
Chemical Cross-linking and Diagonal Electrophoresis-To determine whether LHCII in the digitonin-PSI preparation was associated with PSI and to investigate the possible docking sites of LHCII on PSI, we employed chemical cross-linking, diagonal electrophoresis, and immunoblotting. Lunde et al. (5) suggested that LHCII may bind to PSI-H and PSI-L. To test this possibility, digitonin-PSI was cross-linked with DTSP, subjected to SDS-PAGE under nonreducing conditions, and electroblotted to nitrocellulose membranes. Two cross-linked products could be clearly identified with an antibody against Lhcb1 (Fig. 3A, lane 1). One cross-linked product (Fig. 3, A and  B, ⌬) with an apparent molecular mass of ϳ30 kDa, was most probably a product of PSI-I and Lhcb1 because an antibody against PSI-I identified a product of the same size. Previous work has shown that cross-linking products migrate with apparent molecular masses corresponding to the combined mo- FIG. 1. State transitions in wild type plants. Dark-adapted leaves were exposed to light favoring PSII or PSI as indicated. Each light condition lasted 20 min. Fluorescence was detected during the illumination, and saturating flashes were given as indicated (Fm, Fm2, Fm1). The figure is a representative trace of one leaf. Similar results were obtained with leaves from different wild type plants. lecular masses of the individual proteins (24), and 30 kDa is in excellent agreement with the combined molecular masses of Lhcb1 (25 kDa) and PSI-I (4 kDa). The other cross-linked product (Fig. 3A, *) had an apparent molecular mass of 55-60 kDa. Antibodies against PSI-I, PSI-L, and PSI-H reacted with a band of exactly the same size. The combined molecular mass of Lhcb1, PSI-I, PSI-H, and PSI-L is 57 kDa, and the data therefore indicate that the immunoreactive cross-linking product is a product of these four proteins. The previously described (24) cross-linking products involving PSI-LϩH (28 kDa), PSI-LϩI (22 kDa), and PSI-HϩI (14 kDa) are also clearly seen on the blot. With an antibody against Lhcb2 a cross-linking product with apparent molecular mass of around 50 kDa (ƒ) was observed (Fig. 3C). However, this product is clearly smaller than the cross-linking product involving PSI-I, -L, and -H, and it may correspond to an Lhcb2 dimer.
A well known problem with using antibodies against crosslinked proteins is that epitopes may be lost because of the cross-linking (24). However, with a cleavable cross-linker such as DTSP this problem can largely be overcome by electrophoresing the cleaved cross-linked products in a second dimension. To confirm the above conclusions further, the gel slices after the first dimension were therefore cut out, incubated with reducing sample buffer to cleave the cross-linked products completely, and reelectrophoresed. In the first dimension, crosslinked products will migrate approximately according to the combined apparent molecular mass, but after cleavage the previously linked proteins will migrate separately according to their normal apparent molecular masses and therefore form vertically aligned spots in the second dimension (24). Proteins that are not cross-linked will migrate to the same position in both dimensions and thus form a diagonal in the final gel. The blot in Fig. 4A confirms that Lhcb1 formed two cross-linked products of 30 and 55-60 kDa. The 55-60-kDa product is clearly resolved into Lhcb1 (25 kDa) and PSI-L (14 kDa; note that the protein migrates faster than the actual molecular mass of 18 kDa). PSI-H (10 kDa) is also vertically positioned under Lhcb1 and PSI-L corresponding to 55-60 kDa in the first dimension (Fig. 4B), and the spots of PSI-L and PSI-H can be resolved into two cross-linking products. One might be Lhcb1/ PSI-L/PSI-H, and another might be Lhcb1/PSI-L/PSI-H plus PSI-I (4 kDa). The presence of PSI-I in the 60-kDa product could not be confirmed in the second dimension, probably because of the lower sensitivity of the PSI-I antibody. The product of ϳ30 kDa is seen to consist of large amounts of Lhcb1 together with PSI-I (Fig. 4C). The presence of PSI-L and PSI-H at slightly higher position is evidence of a different product of those two proteins, which was described previously (24). These results clearly demonstrate that Lhcb1 binds to the PSI core in the digitonin-PSI preparation in State 2, and the docking site involves PSI-L, PSI-H, and PSI-I. Similar diagonal electrophoresis using antibody against Lhcb2 did not reveal the nature of the 50-kDa cross-linking product (data not shown). Thus, the product is either a dimer of Lhcb2 or a product between Lhcb2 and a non-PSI protein.
Kinetics of P700 Photooxidation-To see whether the LHCII was functionally attached to PSI in the digitonin-solubilized PSI, energy transfer was determined directly as time course of photooxidation of P700 in vivo. Fig. 5 shows the curves of P700 photooxidation of digitonin-PSI from the wild type and PSI-H and -L mutants. It can be clearly seen that the rate of photooxidation was much slower in the mutants than in the wild type. The t1 ⁄2 for saturation and the relative antenna size of these different samples were calculated from such curves ( Table I).
The results indicate that in the absence of PSI-H, PSI-L, or PSI-O, the antenna size in State 2 was about 15-16% smaller than in the wild type.
Purification and Characterization of a LHCII⅐PSI Complex-To get more direct evidence about the docking site of LHCII on PSI, we further purified the digitonin-PSI preparations from both wild type and mutant plants on sucrose gradients containing DM. This procedure is milder than direct solubilization of thylakoid membranes with relatively high concentrations of DM. PSI was purified in this manner both directly using the digitonin-PSI preparations and after crosslinking of the digitonin-PSI with DTSP. Three main bands were obtained from the sucrose gradients and characterized by Chl a/b ratio and SDS-PAGE. The upper band was highly enriched in LHCII but did not contain any PSI, the middle band was LHCII⅐PSI complex, and the lower band was aggre- gates containing several different components including PSI, LHCII, and PSII. The polypeptide composition of the LHCII⅐PSI complex from wild type (State 2) is shown in Fig. 6. LHCII was clearly bound to PSI when isolated after crosslinking with DTSP, but it is difficult to determine the amount of LHCII in the non-cross-linked samples from Coomassiestained gels. Fig. 7 shows immunoblotting analysis of the LHCII⅐PSI complex. Using an antibody against Lhcb1 (Fig. 7A), LHCII was also detected when DTSP was not used. Furthermore, the quantitation showed that Lhcb1 was present in State 1 at a level of about 40% compared with State 2 (Table II). LHCII⅐PSI was prepared in the same way from plants lacking PSI-H, -L, -O, or -K, in all cases after treatment of the plants with PSII light. Compared with the wild type, less than half the amount of Lhcb1 was bound to PSI lacking PSI-H, -L, or -O, whereas the content of Lhcb1 was higher in the absence of PSI-K than in the wild type ( Fig. 7A and Table II). When the preparations were cross-linked with DTSP prior to sucrose gradient centrifugation, the resulting LHCII⅐PSI preparations contained about 5-fold more Lhcb1 (Fig. 7B and Table II). Even with crosslinking the PSI preparations were essentially devoid of Lhcb1 for the PSI-O mutant. In the PSI-H and PSI-L less plants, some Lhcb1 was purified with the PSI, although about 60 -65% less than in the wild type in State 2 ( Fig. 7B and Table II). Using an antibody against Lhcb2, the results were quite different from those found with the Lhcb1 antibody both without cross-linking (Fig. 7C) and with cross-linking (Fig. 7D). The quantitation showed that the content of Lhcb2 in the PSI-H or -L less PSI was about 35-45% higher than that in the wild type when DTSP was used (Table II). Without cross-linking, PSI from the PSI-H or -O mutants contained about 30 -40% less Lhcb2 than the wild type, whereas in the PSI-L less PSI, the content of Lhcb2 was almost the same as in the wild type (Table II).
The content of Lhcb1 and Lhcb2 as determined above is relative. An estimate of the absolute amounts was made by comparing immunoblots of the purified LHCII⅐PSI and known amounts of the isolated LHCII from the sucrose gradients. Assuming that the isolated LHCII was composed of Lhcb1 and Lhcb2 in a 1:1 ratio we could estimate the molar ratio of Lhcb1 to P700 to 0.24 and 1.5 in the non-cross-linked and the crosslinked LHCII⅐PSI preparations, respectively. For Lhcb2 the corresponding numbers were 0.19 and 0.45. These numbers give a rough indication of the amount of LHCII in the preparations. An average content of two Lhcb monomers/P700 corresponds to an antenna size of about 15% of the LHCI⅐PSI complex. This value corresponds well to the difference between the wild type and the state transitions deficient mutants ( Fig.  5 and Table I). Hence, the cross-linked samples contained an amount of LHCII in excellent agreement with kinetic measure-

TABLE I
Time course of P700 photooxidation in digitonin-PSI preparations P700 absorbance changes were monitored as shown in Fig. 5 under three different light intensities (37, 13, or 3 mol photons m Ϫ2 s Ϫ1 ). At each light intensity, two to four independent measurements were made. The half times (t 1/2 ) (ϮS.D.) shown were determined at 13 mol photons m Ϫ2 s Ϫ1 . The relative antenna size and the statistical analysis (comparison of each mutant to the wild type by two-factor analysis of variance) were based on data from all light intensities.  (C and D). The samples in A and C were prepared without DTSP, and the samples in B and D were treated with DTSP prior to sucrose gradient centrifugation. The wild type (WT) material was obtained from plants treated with PSII light (L2) or PSI light (L1). The material from mutants lacking PSI subunits was from plants treated with PSII light. All samples were loaded on Chl basis (1 g of Chl). ments, whereas the non-cross-linked PSI had obviously lost a large fraction of the associated LHCII.
Phosphorylation of LHCII is believed to be a prerequisite for the initiation of state transitions, so we also analyzed the phosphorylation level in the LHCII⅐PSI complexes. Fig. 8 shows the immunoblotting analysis of LHCII⅐PSI complexes using a phosphothreonine antibody. With DTSP cross-linking, phosphorylated LHCII could be detected in the wild type but was not detectable in PSI-H, -L, or -O mutant plants in State 2 ( Fig. 8A and Table II). Without DTSP, a very small amount of phosphorylated LHCII could be detected in the wild type and in PSI prepared from the PSI-K mutant in State 2 (Fig. 8B). Notably, only a very small fraction of LHCII was phosphorylated in the wild type in State 2 even though both Lhcb1 and Lhcb2 were present in the preparation (Table II). Phospho-LHCII has been shown to be more resistant to thylakoid proteases than unphosphorylated LHCII (37). Therefore, it was important to ensure that the use of DTSP did not differentially affect proteolysis of LHCII or phospho-LHCII. We therefore compared the content of phospho-LHCII relative to total Lhcb1 after incubation of the digitonin preparations with and without DTSP. During the incubation no degradation of LHCII was detected, and the relative content of phospho-LHCII was the same with and without DTSP (data not shown).
Chemical phosphate determination of the purified LHCII showed the presence of 0.9 phosphate group/Lhcb monomer in State 2 (data not shown). Using immunoblotting with the phosphothreonine antibody and comparing with dilution series of the LHCII, we estimated that the LHCII⅐PSI preparation in State 2, which was isolated after cross-linking with DTSP, contained about 3 phosphate groups in LHCII/P700. This number is somewhat higher than the about two monomers/P700 determined with the antibodies against Lhcb1 and Lhcb2. The difference may reflect multiple phosphorylations or inaccuracy in quantitation from immunoblots but indicate that in the cross-linked sample, most of the LHCII bound to PSI in State 2 was phosphorylated.

DISCUSSION
The majority of LHCII trimers are bound to PSII (15), but a fraction of mobile LHCII is able to dissociate from PSII. The reversible dissociation of LHCII is responsible for state transitions, which serve to balance excitation energy distribution between PSI and PSII. Plants without PSI-H do not perform state transitions, the PSI antenna does not increase in State 2, and LHCII remains with PSII (5). Therefore, PSI-H was proposed as docking site of LHCII on PSI (5). However, the mutants with less state transitions also show changes in thylakoid structure (fewer stacks in grana) (38), and indirect effects could not be excluded. Until now, there still are some controversies about whether the mobile LHCII actually docks to PSI. No direct biochemical evidence has been provided yet about the docking site of LHCII to PSI. We therefore wanted to investigate the direct association between LHCII and PSI.
Several workers have made efforts to obtain a purified LHCII⅐PSI complex, but so far the existence of a LHCII⅐PSI complex was mostly based on indirect evidence. Bassi and Simpson (22) purified a significant amount of LHCII together with PSI from barley thylakoids, and Williams et al. (39) obtained similar preparations from pea. Although they showed that energy transfer from LHCII to PSI seemed to take place, it is questionable whether the large amount of LHCII in the PSI preparations reflected the situation in stacked thylakoids. Recently, Pesaresi et al. (40) reported a stable LHCII⅐PSI aggregate present in low light-adapted thylakoids of an Arabidopsis mutant knocked out in one of the psaE genes. However, because state transitions in the mutant are almost completely suppressed, this may be an abnormal LHCII⅐PSI complex, and no evidence for its functionality in energy transfer has been obtained.
The big obstacle to isolation of an LHCII⅐PSI complex seems to be a weak association between LHCII and PSI. We used both DM and digitonin as detergents to solubilize thylakoids and isolate PSI. With DM used in the initial solubilization, no LHCII was associated with the purified PSI. In contrast, very mild solubilization with digitonin allowed copurification of LH-CII and PSI. The amount of LHCII in the preparations appeared to be much higher than what would be expected to be bound to PSI, but nevertheless the LHCII⅐PSI preparations could be used to investigate the docking site for LHCII on PSI. Chemical cross-linking showed that dominant cross-linking products were seen with subunits PSI-H, -L, and -I, either as binary products or more complex products involving several subunits. The three PSI subunits are all located together on one side of PSI, opposite to the side where LHCI proteins are bound (20,24,25). Furthermore, deficiency in state transitions has been observed in mutants lacking the same subunits, except for PSI-I which has not yet been investigated in a knock out mutant. Thus, the data strongly suggest that the three subunits form a docking site for LHCII on eukaryotic PSI.
PSI-O is a recently discovered subunit that has been suggested to be located in contact with PSI-H and PSI-L (21). Plants lacking PSI-O were also highly deficient in performing state transitions, 2 so we thought that the PSI-O subunit might also be involved in docking LHCII. However, we did not find any cross-linking products of LHCII and PSI-O (data not shown). Possibly, PSI-O does not make direct contact with Lhcb1 or Lhcb2, but the absence of a cross-linking product cannot be used to conclude that two proteins are not in contact.
The cross-linking studies proved that LHCII interacts with PSI in the digitonin preparations but did not address whether LHCII was functionally attached to PSI. The kinetics of P700 photooxidation demonstrated that the energy transfer from LHCII to PSI was more efficient in wild type than the PSI-H, PSI-L, and PSI-O mutants in State 2, further confirming that the three subunits could be the docking sites for LHCII on PSI.
To obtain additional biochemical evidence for the direct association between LHCII and PSI, we further purified the digitonin-PSI on sucrose gradients using low amounts of DM. The PSI complex was very pure, and although a large fraction of the bound LHCII may have been lost during the purification, the purified PSI did contain significant amounts of LHCII. In the isolated PSI, LHCII was bound to PSI in the wild type in State 2 and in PSI from plants lacking PSI-K, but much less in the wild type in State 1 or in PSI from plants lacking PSI-H, -L, or -O. Thus, the observations further substantiate that PSI-H, -L, and -O subunits form the docking site for LHCII. Bassi and Simpson (22) suggested that LHCII is bound to LHCI because reconstitution of an LHCII⅐PSI complex could not be achieved in the absence of LHCI-680. We now know that the dissociation of LHCI from PSI is likely to cause the loss of additional small PSI core subunits that were not known at that time.
All of the data presented here indicate a specific function of PSI-H, -L, -O, and -I in binding LHCII. In cyanobacteria, the subunits PSI-L and -I are present but not -O and -H. PSI-L in cyanobacteria has a role in trimerization of PSI (41), and PSI-I has a stabilizing role on PSI-L (42). Hence, in cyanobacteria the two subunits are also involved in mediating supramolecular associations. Apparently, PSI in eukaryotes does not form trimers, and the PSI-L and PSI-I subunits have instead acquired the role of mediating the interaction between PSI and LHCII. Two new subunits, PSI-H and PSI-O, have then been further added to the docking site during the evolution of LHCII and state transitions.
Although the correlation between phosphorylation of LHCII and state transitions has been shown in many studies (6 -9, 43, 44), a demonstration that LHCII phosphorylation is directly involved in state transitions has not been presented. It is also not clear whether phosphorylation of LHCII causes a change in the thylakoid structure, which promotes LHCII migration, or causes a lower affinity for PSII and/or a higher affinity for PSI. Furthermore, phosphorylation of LHCII might have a direct role in regulation LHCII turnover because phospho-LHCII is more resistant to proteolysis by proteases in the thylakoid membrane (37). Hence, the significance of phosphorylation of LHCII during state transitions could be in the protection of exposed LHCII in the stroma membranes against proteases. A few studies have indicated a lack of correlation between phosphorylation and state transitions. Snyders and Kohorn (11) found that LHCII associated with PSI in State 2 was not phosphorylated. Furthermore, LHCII phosphorylation is very low at high light intensities (45), whereas at least some studies have suggested significant state transitions at higher light intensities (31,46). On the other hand, state transitions are normally determined by indirect methods, and it is possible that they do not really take place at high light intensity (47). The data presented here show that the LHCII⅐PSI complex isolated from wild type plants contains unphosphorylated LHCII in State 2. The LHCII⅐PSI complex isolated without chemical cross-linker contained about 24% LHCII compared with the preparation with cross-linker, but the content of phosphothreonine was less than 5% (Table II), indicating that most of the LHCII was unphosphorylated in the non-cross-linked samples. Based on the content of phosphate we can estimate that after cross-linking about 80% of the LHCII in the PSI preparations was phosphorylated in State 2. In State 1 phosphorylated LHCII was not detectable even after cross-linking. This indicates that phosphorylated LHCII is bound to PSI in State 2 but more weakly than the unphosphorylated LHCII. The antibodies used to determine phosphorylation levels have limited sensitivity, and they cannot distinguish between different phosphorylation sites. Additional studies in the future are needed to determine more accurately the ratio between phosphorylated and unphosphorylated LHCII bound to PSI under different conditions. According to the molecular recognition model, LHCII would move from PSII to PSI during transition to State 2 because the phosphorylated LHCII has a higher affinity for PSI and/or a lower affinity for PSII. However, the data reported here indicate that phosphorylated LHCII does not have a higher affinity for PSI. Maybe phospho-LHCII is more loosely bound to PSII than LHCII without phosphorylation, but if this were the case why would unphosphorylated LHCII also move away to PSI? The data presented here are actually more compatible with a variant of the surface charge model. We propose that LHCII moves in state transitions not because the phospho-LHCII has higher affinity for PSI but because the phosphorylation causes structural changes in the thylakoid membranes, which promote movement of LHCII, i.e. both in its phosphorylated and unphosphorylated form. However, we find it unlikely that the charges of the phosphate group contribute significantly to an electrostatic driving force and find it more likely that simple diffusion describes the actual movement. The LHCII appears to bind to a specific site on PSI, and if this site is nonfunctional (as in mutants lacking PSI-H) LHCII tends to remain associated with PSII. In this sense both the molecular recognition and the surface charge models are needed to explain the migration of LHCII.