Pigment Binding of Photosystem I Light-harvesting Proteins*

Light-harvesting complexes (LHC) of higher plants are composed of at least 10 different proteins. Despite their pronounced amino acid sequence homology, the LHC of photosystem II show differences in pigment binding that are interpreted in terms of partly different functions. By contrast, there is only scarce knowledge about the pigment composition of LHC of photosystem I, and consequently no concept of potentially different functions of the various LHCI exists. For better insight into this issue, we isolated native LHCI-730 and LHCI-680. Pigment analyses revealed that LHCI-730 binds more chlorophyll and violaxanthin than LHCI-680. For the first time all LHCI complexes are now available in their recombinant form; their analysis allowed further dissection of pigment binding by individual LHCI proteins and analysis of pigment requirements for LHCI formation. By these different approaches a correlation between the requirement of a single chlorophyll species for LHC formation and the chlorophylla/b ratio of LHCs could be detected, and indications regarding occupation of carotenoid-binding sites were obtained. Additionally the reconstitution approach allowed assignment of spectral features observed in native LHCI-680 to its components Lhca2 and Lhca3. It is suggested that excitation energy migrates from chlorophyll(s) fluorescing at 680 (Lhca3) via those fluorescing at 686/702 nm (Lhca2) or 720 nm (Lhca3) to the photosystem I core chlorophylls.

The main function of plant light-harvesting complexes (LHC) 1 is the absorption of solar radiation and the efficient transmittance of excitation energy toward reaction center chlorophylls (Chl). LHC are composed of a protein moiety to which Chls and carotenoids are noncovalently attached. In higher plants 10 distinct light-harvesting apoproteins (Lhc) can be distinguished. Four of them are exclusively associated with photosystem (PS) I (Lhca1-4), another four with PS II (Lhcb3-6) and two (Lhcb1 and 2) are preferentially but not exclusively associated with PS II (1,2). Up to now there is only limited insight in distinct functions apart from light harvesting of the various LHC. A proportion of LHCIIb (composed of Lhcb1 and 2) is involved in redistribution of excitation energy between PS II and PS I by state transitions that occur upon selective overexcitation of one PS (3). CP29 (Lhcb4), CP26 (Lhcb5), and CP24 (Lhcb6) have particularly high contents of violaxanthin (vio) that can be converted to zeaxanthin, which may exert a special role in regulation of light harvesting as revealed by nonphotochemical quenching (4). Therefore, it was suggested that these LHCs regulate energy transmittance to the PS II core (5). As a consequence of the interrelation of pigment composition and function in LHCIIs, it is important to have detailed knowledge about the pigment composition of the various LHCs for getting insight into their potentially different functions.
During the past decade the pigment composition and spectral properties of LHCs belonging to PS II were studied in great detail (6 -11). These investigations demonstrated that despite the large similarity of the protein sequences of all Lhc proteins (12), differences exist with regard to pigment binding and spectroscopic properties. The LHC studied in the most detail is LHCIIb. Crystallographic data and biochemical characterization of LHCIIb revealed the binding of at least 12 Chls (seven Chl a and five Chl b), approximately two luteins (lut), one neoxanthin (neo), and a substoichiometric amount of vio (6,13,14). By structure analysis, amino acid sequence comparison, and site-specific mutagenesis, nine Chl binding sites, two lutbinding sites (L1 and L2) at helices 1 and 3 (12,13), and one neo-binding site at helix 2 (15) could be localized. It is still under debate whether the detected vio replaces lut (8,11) or whether there is an additional peripheral vio-binding site (9). Replacement of one xanthophyll species (e.g. lut) by an other (e.g. vio) could be demonstrated by LHC reconstitution experiments with Lhcb1 (8,11). The pigment composition of the minor Chl-binding proteins of PS II (CP29, CP26, and CP24) differs from that of LHCIIb. They ligate fewer Chls (eight to ten), lut (approximately one), and neo (approximately 0.5) but more vio (one-half to one) and consequently have a lower total carotenoid content than LHCIIb (6,7,9,16,17). Interestingly, an interrelation between the location of the LHC within PS II and its preference for one Chl species exists. CP29 and CP26 located adjacent to the PS II core preferentially bind Chl a, whereas the peripheral LHCIIb contains almost equal numbers of Chl a and b (6,7,9). In addition there seems to be a correlation between the Chl a/b ratio of a LHC and the requirement for one Chl species for LHC formation. By omission of an individual Chl species in reconstitution mixtures, the requirement of Chl a for CP29 formation (Chl a/b ϭ 3; Ref. 18) and of Chl b for formation of CP26 (Chl a/b ϭ 2) and LHCIIb (Chl a/b ϭ 1.4) could be demonstrated (19 -21). These results mimic very well the situation found in mutants defective in synthesis of Chl b (22,23) and demonstrate the usefulness of the reconstitution approach for analyzing the structural role of individual pigment species for LHC formation.
By contrast, little attention was paid to the pigment compo-sition of LHCI, which is a consequence of its difficult purification. LHCI can be subfractionated into two populations called LHCI-680 and LHCI-730 according to their 77 K (Kelvin) fluorescence emission maximum (24). The former is composed of polypeptides with 23 and 24 kDa (Lhca2 and Lhca3, respectively), and the latter is composed of polypeptides with 21 and 20.5 kDa (Lhca1 and Lhca4, respectively) (25)(26)(27). Use of other detergent mixtures for PS I solubilization in combination with improved separation techniques allowed splitting of LHCI-680 into two different fractions, one enriched in Lhca2 (LHCI-680B) and the other enriched in Lhca3 (LHCI-680A) (26,27). Despite this progress in correlating fluorescence properties with individual Lhca polypeptides, analyses of the pigment composition are very limited. There are only detailed analyses about LHCI-730 of barley (28) and tomato (29), a maize LHCI holocomplex of unknown polypeptide composition (30), and a red algal LHCI (31). In some studies only Chl a/b ratios were determined, which range from 1 to 3 for LHCI-680 and from 2.2 to 3.6 for LHCI-730 (24,25,(32)(33)(34)(35)(36). In other reports the values are simply based on calculations of the difference in pigment content of PS I holocomplex versus PS I core complex lacking LHCI (37,38) and deviate strongly from that obtained by HPLC analyses of isolated LHCI(-730), which demonstrated the presence of seven to ten Chls, approximately one lut, and substoichiometric amounts of vio and ␤-car per Lhca protein (29,30). This indicates that the pigment composition of LHCI(-730) is similar to that of minor Chl-binding proteins of PS II with the exception that ␤-car is present and neo is absent (28 -30, 34). However, there are no analyses about pigment composition of LHCI-680 available yet that allow assessment of common features and differences of the two LHCI subpopulations that would be essential for a better insight into potentially different functions of the various LHCI. This lack extends to the pigment binding properties of individual Lhca proteins with the exception of Lhca1 and Lhca4 (29,39,40) and to studies regarding the significance of individual pigments for LHCI formation, where only preliminary results are available up to now for Lhca1 and Lhca4 (41).
To gain insight into pigment binding by LHCI subfractions, we isolated LHCI-680 and LHCI-730 from tomato leaves and determined their pigment composition. For further dissection of pigment binding by these LHCI subpopulations, we constructed expression plasmids of lhca2 and lhca3, so that now, for the first time, we could overexpress and reconstitute the full set of LHCI apoproteins. This allowed differentiation of pigment binding by individual Lhca proteins, which are present in twos in LHCI-680 or LHCI-730. Finally, the role of individual pigments for LHCI formation was tested by reconstituting each Lhca protein in the absence of an individual pigment species. The obtained results provide a better insight into differential pigment binding of the various LHCI and demonstrate that distinct pigment requirements exist for formation of the individual LHCIs. The presented data fill a gap regarding knowledge of pigment composition of LHCs in higher plants and form the basis for future analyses aiming to elucidate potentially different functions of the various LHCIs.

EXPERIMENTAL PROCEDURES
Isolation of Native LHCI-680 and LHCI-730 -Isolation of PS I from tomato thylakoids was as described in Ref. 29. Following ultracentrifugation (24 h, 112,500 ϫ g) the PS I-containing band was collected, diluted with 5 volumes of cold distilled water, and centrifuged overnight at 258,000 ϫ g. The resulting PS I pellet was suspended in 10 mM Tricine/NaOH (pH 7.8), 1 mM EDTA/NaOH (pH 7.8), and 30% sucrose and stored at Ϫ70°C at a Chl concentration of 1.5-2 mg/ml. For isolation of LHCI thawed PS I preparations were diluted with 4 volumes of cold distilled water and centrifuged (42,000 ϫ g, 30 min). The pellet was suspended in 5 mM Tris/HCl (pH 7.5) to a Chl concentration of 0.5 mg/ml and solubilized by adding 0.2% Zwittergent 3-16, 1% n-dodecyl ␤-D-maltoside and 1% n-octyl ␤-D-glucopyranoside and mixing for 60 min at 4°C. Aliquots of this solution equivalent to 0.5 mg of Chl were loaded onto sucrose gradients (0.06 -0.8 M sucrose, 5 mM Tricine/NaOH (pH 7.8), and 0.1% n-dodecyl ␤-D-maltoside). Centrifugation was performed for 25 h at 246,000 ϫ g and 4°C. The two bands containing LHCI (second and third from the top) were collected, concentrated with centricons (cut off 10 kDa; Millipore, Eschborn, Germany), and used immediately for characterization or stored at Ϫ70°C. Alternatively, density gradient bands containing LHCI-680 and LHCI-730 were diluted 20-fold with 2 mM Tricine-NaOH (pH 7.8) and centrifuged overnight at 602,000 ϫ g to remove sucrose and n-dodecyl ␤-D-maltoside. LHCI pellets were suspended in a small volume of the supernatant and used for determination of the Chl/protein stoichiometry as described below.
Construction of lhca2 and lhca3 Expression Plasmids-cDNAs of tomato lhca2 and lhca3, kindly provided by E. Pichersky (University of Michigan, Ann Arbor, MI), were used for PCR with Pfu polymerase to produce restriction sites suitable for cloning into expression plasmids. Primers were designed to generate restrictions sites for NdeI and BamHI (lhca2) and BamHI and SalI (lhca3). Amplified DNA was precipitated and subsequently digested with the respective restriction enzymes. Following a further precipitation lhca2 was ligated into the pet3a vector (Novagen, Bad Soden, Germany), and lhca3 was ligated into the pDS expression plasmid (42). Molecular biological work was performed according to standard procedures (43). For cloning of lhca2 into the vector, an internal BamHI site had to be removed first by introduction of a silent mutation following the PCR-based mutation protocol of Ref. 44. To achieve overexpression, the first base triplet GTT coding for Val had to be replaced by the triplet TCA coding for serine and silent mutations had to be introduced in the third (GCT instead of GCA), fourth (GAT instead of GAC), and fifth (CCA instead of CCT) triplets, which was achieved by using an appropriate forward primer for PCR. The cloning strategy resulted in an additional vector derived Met at the N terminus of Lhca2. Use of the pDS expression plasmid for lhca3 resulted in four additional amino acids (Met, Arg, Gly, and Ser) at the N terminus of the protein. Correct DNA amplification and ligation was examined by DNA sequencing of the entire coding region. In addition the pDS expression plasmids harboring lhca1 and lhca4 described in Ref. 45 were used.
Protein Overexpression and Inclusion Body Isolation-For overexpression of Lhca1, Lhca3, and Lhca4, the corresponding pDS expression plasmids were transformed into Escherichia coli strain JM 101 and overnight cultures in Luria Bertrani medium supplemented with 100 g/ml ampicillin (LB-Amp) were grown. These cultures were used as inoculum for new LB-Amp cultures, which were grown to an optical density of approximately 0.6 under agitation (175 rpm, 37°C). Following addition of isopropyl-1-thio-␤-D-galactopyranoside to a final concentration of 1 mM incubation was continued for another 4 to 5 h under the same conditions. Afterward cells were collected by centrifugation (5 min, 10,000 ϫ g) for inclusion body preparation. The pet vector with lhca2 was transformed into E. coli Bl 21 cells. These cells were grown overnight in LB-Amp supplemented with 2% glucose. Then cells were collected by centrifugation (5 min, 10,000 ϫ g), suspended in glucose depleted LB-Amp and grown to an optical density of approximately 0.5 (225 rpm, 37°C). Following the addition of isopropyl-1-thio-␤-D-galactopyranoside to 1 mM (final concentration), cultures were kept at 40°C for 8 h under agitation (225 rpm). Finally, the cells were harvested by centrifugation (10,000 ϫ g, 5 min). The inclusion body protein was isolated as described (46). The protein concentrations of inclusion body preparations were determined by a dye binding assay (47), and accumulation of recombinant Lhca proteins was checked by fully denaturing PAGE according to Ref. 48 with subsequent Coomassie staining.
LHC Reconstitution-Reconstitutions were done with either total pigment extracts or mixtures of individual pigments by the detergent exchange method (29). Total pigment extract as well as individual pigments were isolated from tomato thylakoids as described in Ref. 46. The reconstitution mixtures contained either 30 g of inclusion body protein and pigments equivalent to 40 g of Chl aϩb for subsequent partially denaturing gel electrophoresis (49) or 150 g of protein and pigments equivalent to 200 g of Chl aϩb for density gradient ultracentrifugation (29). For analysis of differences in pigment binding, the different Lhca proteins were reconstituted with the same total pigment extract. The molar ratio of neo:vio:lut:Chl b:Chl a:␤-car in total pigment extract was 0.2:0.2:1:2.9:8.5:0.1.To test the significance of individual pigment species for LHC formation, mixtures of individual pigments were used; for Lhca1 and Lhca4 reconstitution mixtures contained the pigments in the stoichiometry that was found for native LHCI-730, those for Lhca2 and Lhca3 had the same composition as native LHCI-680 (cf. Table I). In these analyses one Chl or carotenoid species was omitted, and the amount of the other Chl or carotenoids was increased correspondingly to maintain the original Chl/carotenoid stoichiometry.
Miscellaneous-Photometric Chl quantification was performed in 80% acetone using the equations of Porra et al. (50). Protein quantification of LHCIs by the BCA assay (51) was performed with samples adjusted to equal Chl amounts. Absorption was corrected for Chl by parallel measurement of the samples in BCA solution lacking CuSO 4 . Bovine serum albumin was used as a reference protein. Chl/protein stoichiometries of native LHCI-680 and LHCI-730 were calculated by using a molecular mass of 23.5 kDa for LHCI-680 (average of Lhca2 and Lhca3) and 21.5 kDa for LHCI-730 (average of Lhca1 and Lhca4). For analysis of pigment composition by HPLC, pigments of density gradient bands were extracted by secondary-butanol (52), diluted with acetone, and loaded onto a Chromolith SpeedROD RP-18e column (Merck), which was developed with an acetone gradient from 70 to 100%. Eluted pigments were detected by a MD-1515 multiwavelength detector (Jasco, Gross-Umstadt, Germany) and quantified on basis of calibration curves obtained for individual pigments. 77 K fluorescence emission spectra were recorded with a Fluoromax 2 (ISA Jobin Yvon-Spex, Grasbrunn, Germany). Samples from the density gradients were adjusted to 60% glycerol, 5 mM Tricine/NaOH (pH 7.8), 0.05% n-dodecyl ␤-D-maltoside, and 2 g Chl/ml. The measurements were done in 1-nm steps with a slit width of 2 nm for excitation and emission light. Excitation was at 410 nm.

RESULTS
Composition of Native LHCI Preparations-Fractionation of PS I by density gradient ultracentrifugation resulted in the resolution of two bands in the upper part of the centrifuge tube (Fig. 1A), which were strongly enriched in LHCI proteins. The lower density band had a fluorescence emission around 680 nm with peaks at 680 and 686 nm (Fig. 1B, dashed line). The higher density band exhibited a strong fluorescence at 734 nm (Fig. 1B, solid line). Thus, the spectra allowed assignment of the density gradient bands to LHCI-680 and LHCI-730, which was confirmed by analysis of their polypeptide composition (Fig. 1C). LHCI-680 migrates like a monomer in a partially denaturing gel (not shown), is strongly enriched in Lhca2 and Lhca3, and is only slightly contaminated by other Lhc proteins. The LHCI-730 band exhibiting migration of a dimer in partially denaturing gels (not shown) possesses dominating bands of Lhca1 and Lhca4 and is almost free of contaminating proteins. Interestingly, an additional protein migrating between Lhca1 and Lhca2 was found throughout the experiments in PS I preparations and in LHCI-730 but not in LHCI-680 (Fig. 1C, asterisks). A protein with comparable electrophoretic properties was also described for PS I of barley (27), but its identity has not yet been elucidated.
Overexpression of Lhca Proteins-Because of the presence of two different Lhca proteins in the LHCI density gradient bands, no insight into differences in pigment binding of the individual Lhca proteins can be achieved by characterization of native LHCIs. Therefore the reconstitution technique was employed. For Lhca1 and Lhca4 overexpression and reconstitution of native-like proteins was demonstrated earlier (29). In the course of this work we constructed expression plasmids with tomato genes lhca2 and lhca3 and overexpressed these proteins. These were isolated as inclusion bodies and exhibited apparent molecular masses of 23 (Lhca2) or 23.5 kDa (Lhca3) as shown in Fig. 2. For comparison overexpressed Lhca1 and Lhca4 were also used for reconstitution analyses; these proteins migrate in the gel at 22 kDa (Lhca1) and as a double band around 21.5 kDa (Lhca4).
Reconstitution and Pigment Binding of Recombinant Lhca Proteins-To test differential pigment affinities of the various Lhca proteins, we reconstituted the individual proteins with the same total pigment extract and isolated the reconstituted LHCI (r-Lhca) by density gradient ultracentrifugation, followed by pigment quantification by HPLC. Calculation of pigment/protein stoichiometries is impaired in this case by constitutively expressed bacterial proteins still present as contaminants in the density gradient bands. Because of the interference of these proteins with protein quantification, we used lut as a reference for comparing pigment compositions, because it has been demonstrated that approximately one lut is present per Lhca protein (Table I and Refs. 29 and 30). As is obvious from Table II, no pigment species is discriminated by the different Lhca proteins. Thus, although not present in native LHCI, neo can be ligated by all Lhca proteins. Although binding of neo, vio, and ␤-car is comparable for all proteins with the exception of the neo content of Lhca1 and Lhca3 and the ␤-car content of Lhca3, pronounced differences were found for Chl binding properties.
r-Lhca1, r-Lhca2, and r-Lhca4 bind approximately seven Chl molecules on the basis of one lutein molecule, whereas r-Lhca3 binds approximately one-third less Chl (Table II). r-Lhca3 also deviates most strongly with regard to preferential binding of one Chl species as is reflected by the high Chl a/b ratio of 6.1 as compared with the lower ratios obtained for Lhca2 (2.3), Lhca1 (3.5), and Lhca4 (2.6). In addition to these agreements in carotenoid binding and differences in Chl binding of the four Lhca proteins, it must be emphasized that for all proteins pigment binding is rather specific because pigment compositions of the reconstituted LHC deviate pronouncedly from that of the pigment mixture used for reconstitution (Table II). To assure proper folding of reconstituted Lhca, we recorded 77 K fluorescence emission spectra (Fig. 3). The emission spectrum of r-Lhca2 (dashed line) exhibits a broad peak with a maximum at 687 nm and a shoulder at approximately 702 nm (arrow). By contrast, r-Lhca3 (solid line) has two distinct fluorescence emission peaks. One is narrow and has its maximum at 680 nm, and the other is broad and peaks around 720 nm. r-Lhca1 and r-Lhca4 have fluorescence peaks at 684 and 728 nm as reported earlier (29).
Pigment Requirements for LHCI Formation-To test the significance of individual pigments for LHCI formation, the reconstitutions were performed with pigment mixtures, with each depleted in one pigment (Fig. 4). The omission of ␤-car had the smallest effect on LHC formation. For Lhca1, Lhca2, and Lhca4, the LHC yield was not reduced when compared with reconstitutions with all pigments present. Only for Lhca3 a slight decrease in the amount of reconstituted LHC occurred in most experiments. Neither did the lack of vio prevent LHC formation of the Lhca proteins, as is obvious from Fig. 4. Only in Lhca1 did the absence of vio result in a fainter band reflecting a decreased formation or lower stability of this LHC. By contrast, omission of lut impaired LHC formation strongly. The most pronounced effect was present for Lhca4 and the smallest one was for Lhca2, whereas Lhca1 and Lhca3 were intermediary. With regard to Chl similar requirements were observed for Lhca1 and Lhca3 on the one hand and for Lhca2 and Lhca4 on the other hand. Omission of Chl a resulted in loss of LHC formation in Lhca1 and Lhca3. By contrast, reconstitution of these proteins in the absence of Chl b yielded a weak LHC band, indicating reduced formation and/or stability of LHC. For Lhca2 and Lhca4, no LHCs were formed in the absence of Chl b that are stable enough to endure electrophoretic separation. Lack of Chl a resulted in formation of stable LHC with Lhca2 and Lhca4. Lhca2 was the only Lhca protein, which showed no effect on LHC formation/stability upon reconstitution in absence of a Chl species (Chl a). Finally, it is interesting to note that we could not observe a dimer band of either Lhca2 or Lhca3 under conditions where the LHCI-730 heterodimer is readily formed. Currently we are analyzing the potential dimerization behavior of Lhca2 and Lhca3 under less stringent conditions. DISCUSSION LHCI-730 ligates more Chls than LHCI-680; the number of Chls is approximately 23 for LHCI-730 and 18 for LHCI-680. This gives an average of approximately 10 Chls for each of the four different LHCI proteins, which agrees with the value of 10 Chls per Lhca protein described for the LHCI holocomplex of maize (30) and is higher than the value obtained earlier for LHCI-730 (29). Interestingly, in the present study we found approximately 8.5 Chl a per LHCI-730 protein. A corresponding number of Chl a was found in the LHCI holocomplex (30) and a red algal LHCI, which binds Chl a as the only Chl (31). By contrast, LHCI-680 proteins bind approximately two Chl molecules less on the basis of one apoprotein, and mainly Chl a is affected by this reduction.
Because all of the potential Chl-binding amino acids described for LHCIIb (13) are conserved in all Lhca proteins (12), there is no obvious reason for the reduced Chl content of LHCI in comparison with LHCIIb. Because of higher Chl a/b ratios of LHCIs, one explanation could be the lack of Chl b1 and b2 (and a7) that are probably stabilized by other Chls (e.g. b5) in LHCIIb (13,14). Point mutation of glutamate at position 102 in Lhca4, which corresponds to the Chl b5 binding site in LHCIIb (13), supports this idea because the mutant r-Lhca4 contains only approximately one Chl less than the wild type Lhca4 (53). Assuming that Lhca2 and Lhca3 occur in vivo as dimers as was suggested (30,54,55), the lower Chl content of LHCI-680 in comparison with LHCI-730 could be caused by monomerization during isolation, which may result in the release of peripheral Chls possibly located at the interface of the two proteins and stabilized by both subunits. This would be in agreement with data about heterodimerization of LHCI-730 indicating the presence of such Chl (29). However, analysis of recombinant monomeric Lhca1-4 complexes indicates that in fact LHCI-680 apoproteins taken together have a lower Chl binding capacity than the LHCI-730 proteins (see below).
The number of Chl b bound by LHCI-730 and LHCI-680 is similar. However, LHCI-730 binds more Chl a than LHCI-680, which is also reflected by the higher Chl a/b ratio of LHCI-730. This corresponds with results of earlier analyses in which Chl a/b ratios of LHCI preparations were compared, and a preference for Chl a was observed in LHCI-730 (24,25,(33)(34)(35)(36). In comparison with the LHCI holocomplex of maize, which contains two Chl b per protein (30), we found more Chl b in LHCI-730 (2.9) and LHCI-680 (2.6). The higher values appear more reasonable because the PS I holocomplex contains ϳ200 Chls at a Chl a/b ratio of six (33) and eight Lhca proteins (2,54). Therefore an even higher Chl b content would be expected than the one reported here.
By achieving reconstitution of Lhca2 and Lhca3, it was possible to compare the Chl binding properties of all individual Lhca proteins for the first time. From a comparison of Tables I and II it becomes obvious that all r-Lhca possess on the average approximately three to four Chls less on the basis of one lut than LHCI-730 and LHCI-680. Although partial reduction could be explained by binding of additional Chls as a consequence of dimerization, the significant differences are surprising and raise the question of whether lut is a suitable reference. In this context it is remarkable that the Chl/Car ratios of 5 and 6.2 for r-Lhca1 and r-Lhca4 and of 5.8 and 3.9 for Lhca2 and Lhca3, respectively, were on the average comparable with those of native LHCI-730 (5.9) and LHCI-680 (5.1; cf. Tables I  and II). Assuming corresponding Chl/Car ratios of native and reconstituted LHCI, which was shown for various LHCII (11,14,17), the lut content of the r-Lhca would be underestimated, and the Chl amounts would be consequently 39% (Lhca1), 43% (Lhca2), 27% (Lhca3), and 55% (Lhca4) higher than those in Table II. The average of the corrected values of Lhca1 and Lhca4 as well as of Lhca2 and Lhca3 result in a total Chl content per apoprotein of 10.7 (LHCI-730 proteins) and 8.6 (LHCI-680 proteins). These values correspond quite well with those obtained for the native LHCIs and indicate that in the reconstituted proteins lut is partially bound instead of other carotenoids.
Regardless of this correction, the LHCI-680 proteins Lhca2 and Lhca3 on the average bind less Chl than LHCI-730 proteins (Table II), specifically less Chl a. Thus, the lower Chl (a) content of native LHCI-680 is an intrinsic feature of the constituent proteins and is not only caused by Chl loss from putative monomerization. The reduced Chl content is mainly caused by Lhca3, which on the basis of one lut binds at least two Chls less than the other Lhca proteins. This effect is predominantly caused by reduced Chl b binding, and consequently the Chl a/b ratio is pronouncedly higher than that of the other Lhca proteins (Table II). Interestingly the other LHCI-680 protein Lhca2 forms the LHCI with the lowest Chl a/b ratio, demonstrating significantly different Chl preferences by the LHCI-680 proteins.
A comparison of Chl binding in LHC of PS I and PS II shows that LHCI-730 and LHCI-680 resemble most closely CP24 and CP26, which bind ten and nine Chls, respectively (16,17). With regard to their preference for one Chl species, the closest relationship exists between LHCI-730 and CP29, which binds eight Chl at a Chl a/b ratio of 3 (18), and between LHCI-680 and CP26, which possesses a Chl a/b ratio of 2.2 (17). Interestingly, LHCI-730 binds most Chl a per apoprotein among all LHC.
LHCI proteins differ not only in Chl binding but also in the vio content. LHCI-730 has a higher content of vio in comparison with LHCI-680. By contrast, the lut content is about the same for both LHCIs. In agreement with the majority of earlier studies (29,30), approximately one lut per apoprotein was found in LHCIs. ␤-car is also present in equal amounts of  approximately 0.4 molecules per apoprotein in both LHCI forms. neo, a component of all LHC belonging to PS II (6,7,9) was not detected in LHCI-680 and LHCI-730. Therefore, our results confirm more recent pigment analyses of LHCI-730 (29) and LHCI holocomplex (30). With a content of approximately one lut and altogether two carotenoid molecules per apoprotein, the native LHCIs also in this respect have their closest relatives in the minor LHCIIs CP29, CP26, and CP24 (9,16,17,56). Another similarity of these LHCs of PS I and PS II is the higher vio content compared with that of LHCIIb (9,14,17,56). This might be of importance for the regulation of light harvesting via the violaxanthin cycle, which operates also in PS I (57,58). Because of a stronger enrichment of vio in LHCI-730 as compared with LHCI-680, this property might be especially pronounced in the former LHCI.
The nonstoichiometric presence of single carotenoids observed for LHCIs extends to all LHC of PS II, where less than one molecule of vio and/or neo are regularly found for the different LHCIIs (6,9,14,17,56). The reason for this feature is not clear yet. The fact that all three amino acid motifs involved in the binding of two lut (12) and one neo (15) are present in all Lhca proteins brings about the question of whether all three binding sites are occupied in LHCI proteins or if one is vacant. Because of the presence of one lut in all LHCIs and the requirement of lut for LHCI formation/stabilization by reconstitution (Fig. 4), there seems to be one lut-binding site, probably the L1 site, whose occupation by the ␤,⑀-carotenoid lut is needed for the formation of stable LHCI in all Lhca proteins with the exception of Lhca2 (Fig. 4). Support for this assignment comes from the observation that Lhca1 and Lhca4 with deletions of the entire extrinsic N-terminal region including the amino acids involved in formation of the L2 site are still able to form monomeric LHCI (45). Because the amount of the ␤,␤-carotenoids vio and ␤-car sums up to almost one molecule, two possibilities for their binding exist. First, both could bind to the same binding site, which has a low specificity regarding the bound ␤,␤-carotenoid species. In this case in all LHCI molecules of a population this site would be filled. For CP29 and CP26, it was suggested that the second central lut-binding site L2 is such a site, which accommodates either vio or neo (9,17,56). Secondly, vio and ␤-car could bind to different peripheral sites, where binding is not tight and part of the pigments becomes released during LHCI isolation. However, this scenario seems to be rather improbable because of the relatively fixed vio/␤-car ratio found repeatedly in LHCI isolations (Table  I). Because the second possibility would also require the existence of two peripheral binding sites with loosely ligated pigments, which is unlikely according to recent knowledge (8,9,11), we favor the idea that vio and ␤-car are bound to the same binding site. Reconstitution studies with recombinant mutated Lhca proteins will be useful to identify this/these binding site(s).
In accord with the result of the native LHCIs, a slight enrichment of vio was found in the LHCI-730 component Lhca1 as compared with the LHCI-680 subunits and Lhca4 independently of using one lut or the total carotenoid content of native LHCI as reference. Whether this distinct vio binding among the Lhca proteins is related to the amino acid sequence, which is particularly conserved for Lhca2, Lhca3, and Lhca4 as opposed to Lhca1 (1,59), cannot be decided yet. Interestingly, Lhca1 has its closest relative among other Lhcs in CP29 (59). Because this LHC is thought to play a major role in nonphotochemical quenching (56), a similar function of Lhca1 in PS I is possible. Other conspicuous features observed for all r-Lhcas are the ligation of neo that was present in the reconstitution mixture and the very low content of ␤-car in r-Lhca as com-pared with native LHCIs. Possibly neo can be bound to the rather unspecific ␤,␤-carotenoid-binding site proposed above, where it could preferably be attached in comparison with ␤-car. This finding may be of interest with regard to the in vivo situation because it indicates that the pigment composition of LHCIs may depend not only on the protein structure but also on the pigments available during LHCI formation. This was proposed for Chl binding by LHCI in barley (60), and results obtained with maize seedlings exposed to intermittent light also revealed flexibility in pigment binding in vivo (61). Additionally, reconstitution studies demonstrated the interchangeability of LHC proteins and carotenoids from alga and higher plants (21,62,63).
With regard to LHC biogenesis the availability of single pigment species is of differential importance for the various LHCs as revealed by altered stoichiometries of the LHCs in the thylakoid membrane as a consequence of different light treatments (22, 61, 64 -66) or disruption of pigment synthesis by mutations (23,60,67,68). Our experiments show that the absence of the ␤,␤-carotenoids vio and ␤-car did not impair LHCI formation/stability. By contrast, lut is an important structural element for Lhca1, Lhca3, and Lhca4 but not for Lhca2. In LHCII, CP26 and CP24 lut can be substituted by other carotenoids (8,11,16,19), which is not the case for Lhca1, Lhca3, and Lhca4 as is obvious from Fig. 4, where increased amounts of vio and ␤-car in the reconstitution mixtures could not prevent strong reduction or absence of LHCI bands. In Chlamydomonas reinhardtii that has somewhat different Lhca proteins compared with higher plants (69), the absence of lut, vio, and neo did not impair assembly of LHCI proteins, resulting in the conclusion that zeaxanthin can replace these pigments in functional LHCI (68). This is in line with reconstitution analyses of higher plant Lhcb1, where zeaxanthin could be bound to the same extent as lut (11). It will be interesting to test whether Lhca proteins also can form stable r-Lhcas with Chls and only zea as xanthophyll as has been demonstrated for the red algal LhcaR1 (31).
Perhaps the most interesting pigment requirement for r-Lhca assembly is that of either Chl a or Chl b. One protein of both LHCI-730 (Lhca1) and LHCI-680 (Lhca3) folded to a stable LHCI in the absence of Chl b, and the other two proteins Lhca2 and Lhca4 assembled stable LHCIs in the absence of Chl a (Fig. 4). This is in agreement with studies about Chl b-free chlorina f2 barley plants, in which some alleles did not accumulate Lhca4 and had reduced amounts of Lhca2 (23), whereas in other chlorina f2 alleles, no loss or reduced amount of Lhca proteins was found (70). That in both LHCI-680 and LHCI-730 either one apoprotein needs Chl a for assembly and the other one needs Chl b may be of special importance with regard to changing conditions during in vivo assembly, where at least part of both LHCI subpopulations can be formed and serve as antenna as was shown for Lhca1 in Lhca4-depleted plants (71) or various chlorina f2 mutants where either of these two proteins is absent (23). Interestingly this different Chl requirement in LHCI assembly seems to be manifested in the Chl a/b ratios, which are higher for r-Lhca1 and r-Lhca3 than for r-Lhca2 and r-Lhca4. This correlation is also valid for LHC of PS II because CP29 that has the highest Chl a/b ratio of 3 could be folded to stable LHC in the absence of Chl b (18) and Lhcb1 with the low Chl a/b ratio of approximately 1.3 could be reconstituted to stable LHCII when Chl a was omitted (39).
Most LHCI-680 preparations described up to now show either a sharp fluorescence emission peak at 680 nm and a minor broad peak with a maximum between 700 nm and 740 nm (25-27, 72) or a single broad peak with a maximum at 690 nm, which has a red flank extending into the far red region (33).
LHCI-680 obtained in this work by mild detergent treatment and sucrose density gradient ultracentrifugation exhibits features of the latter type but additionally shows a splitting of the main peak into two peaks with maxima at 680 and 686 nm (Fig.  1B). Reconstitution of the individual Lhca proteins allowed assignment of the 680-nm peak to r-Lhca3 and assignment of the 686-nm peak to r-Lhca2 (Fig. 3). In addition a 702-nm fluorescence component could be attributed to r-Lhca2 and an additional broad peak with a maximum around 720 nm to r-Lhca3. Thus, r-Lhca3 resembles to some extent r-Lhca4, although the long wavelength peak of the latter is more pronounced and at a longer wavelength (29,39). Interestingly, Lhca3 and Lhca4 differ with regard to Chl-binding sites from all other Lhc proteins by having asparagine instead of histidine at the position of the Chl a5 binding site (1). This difference might be involved in establishing long wavelength properties. It is assumed that this feature is caused by Chl dimer or trimer formation in LHCI(-730) (39,73,74). Because Chl a5 and b5 are in close contact in LHCIIb (13) on the one hand and removal of b5 results in abolition of long wavelength fluorescence in Lhca4 (53) on the other hand, it is conceivable that a5 and b5 are involved in Chl dimer formation.
Analysis of leaves of Lhca2/Lhca3 antisense plants indicated the presence of low energy Chls associated with the presence of Lhca2 and Lhca3 that fluoresce at 735 and 702 nm (55). The latter fluorescence component was also detected in an LHCI holocomplex and was attributed to Lhca2 and Lhca3 (74). It was suggested that this long wavelength fluorescence develops when Lhca2 and Lhca3 are in a dimeric state (30,55). Possibly this F735 is present in the broad long wavelength peak in r-Lhca3 and becomes more prominent upon association of Lhca2 and Lhca3. Alternatively, a new spectral component could arise as a consequence of dimerization as is the case in LHCI-730 (29,75). Because of the availability of recombinant Lhca2 and Lhca3, it will be possible now to analyze dimerization of Lhca2 and Lhca3 in detail. It will be very interesting to see whether dimers of these proteins adopt long wavelength characteristics comparable with that observed for the subunits of LHCI-730 as a consequence of heterodimerization as was proposed for LHCI-680 proteins (30,55).
Because of the detection of additional spectral details in LHCI-680 and monomeric Lhca2 and Lhca3 in this work, an excitation energy migration pathway for LHCI-680 can be suggested in which energy migrates from F680 via F686 or F702 to F720 and finally to Chls of the inner antenna or reaction center. Following the suggestion that the long wavelength Chls in the peripheral antenna function in concentrating excitation energy close to low energy Chls of the core complex (76), two different routes for the excitation energy from the peripheral antenna to the center of PS I may be used via long wavelength Chls of either Lhca4 in LHCI-730 or Lhca3 in LHCI-680.