Configuration and Dynamics of Xanthophylls in Light-harvesting Antennae of Higher Plants

Resonance Raman excitation spectroscopy combined with ultra low temperature absorption spectral analysis of the major xanthophylls of higher plants in isolated antenna and intact thylakoid membranes was used to identify carotenoid absorption regions and study their molecular configuration. The major electronic transitions of the light-harvesting complex of photosystem II (LHCIIb) xanthophylls have been identified for both the monomeric and trimeric states of the complex. One long wavelength state of lutein with a 0-0 transition at 510 nm was detected in LHCIIb trimers. The short wavelength 0-0 transitions of lutein and neoxanthin were located at 495 and 486 nm, respectively. In monomeric LHCIIb, both luteins absorb around 495 nm, but slight differences in their protein environments give rise to a broadening of this band. The resonance Raman spectra of violaxanthin and zeaxanthin in intact thylakoid membranes was determined. The broad 0-0 absorption transition for zeaxanthin was found to be located in the 503–511 nm region. Violaxanthin exhibited heterogeneity, having two populations with one absorbing at 497 nm (0-0), 460 nm (0-1), and 429 nm (0-2), and the other major pool absorbing at 488 nm (0-0), 452 nm (0-1), and 423 nm (0-2). The origin of this heterogeneity is discussed. The configuration of zeaxanthin and violaxanthin in thylakoid membranes was different from that of free pigments, and both xanthophylls (notably, zeaxanthin) were found to be well coordinated within the antenna proteins in vivo, arguing against the possibility of their free diffusion in the membrane and supporting our recent biochemical evidence of their association with intact oligomeric light-harvesting complexes (Ruban, A. V., Lee, P. J., Wentworth, M., Young, A. J., and Horton, P. (1999) J. Biol. Chem. 274, 10458–10465).

The light-harvesting antenna (LHA) 1 of higher plants binds five types of xanthophylls: lutein, neoxanthin, violaxanthin, zeaxanthin, and antheraxanthin. The last three constitute the xanthophyll cycle, which has been suggested to participate in the process of dissipation of excess excitation energy, giving rise to nonphotochemical fluorescence quenching (1)(2)(3). Several important questions concerning these xanthophylls remain unanswered: why is there such a variety of xanthophyll types in antenna; what is the exact molecular mechanism of zeaxanthin action in nonphotochemical fluorescence quenching; where are xanthophylls located; and what is the nature of their interaction with the protein and other pigments? LHA consists of a number of pigment-protein complexes accommodating different types and amounts of xanthophylls (5)(6)(7)(8). The major and most characterized LHA complex, the trimeric LHCIIb, binds 2 luteins, 1 neoxanthin, and between 0.1 and 1 violaxanthin per monomer (7,9). The amount of bound violaxanthin was found to depend on the treatment during purification (9), as well as on plant growth conditions (10). The two luteins of LHCIIb are thought to correspond to the two carotenoid molecules located near the transmembrane helixes A and B in the inner core of the complex (11), thus being tightly associated with it and probably having a structural role. Site-directed mutagenesis experiments have suggested that neoxanthin is associated with helix C (12). Neoxanthin was found to have the highest affinity of binding to the complex (9). In contrast, the binding affinity of violaxanthin to the complex is the lowest, and it can be easily removed by various treatments. This biochemical work has therefore established that a large population of xanthophyll cycle carotenoids is peripherally bound to LHA complexes.
The structure of the minor PSII antenna complexes as well as all LHCI complexes is not known; therefore, it is not clear where xanthophyll molecules are bound to them, what the nature of this binding is, or what the carotenoid configuration is involved. The data on xanthophyll stoichiometry in the minor PSII antenna are controversial (4,6,8) and may well again reflect natural variation of xanthophyll ratios. However, it is clear that unlike LHCIIb, the minor antenna PSII complexes, CP24, CP26, and particularly CP29, contain at least one strongly bound xanthophyll cycle carotenoid. The efficiency of violaxanthin de-epoxidation-a prerequisite for the photoprotective energy dissipation state in thylakoid-was found to be the reciprocal of the relative affinity of binding of this xanthophyll to various antenna components, suggesting that only loosely bound molecules are accessible to the violaxanthin deepoxidase (9). A similar tendency was observed for the PSI antenna, which contains at least 50% tightly bound violaxanthin that was not converted into zeaxanthin even under condi-tions favoring maximum de-epoxidation (9,13). Therefore, it may be argued that the "structural" violaxanthin molecules, which are strongly coordinated within the minor LHA complexes, do not play a role in the xanthophyll cycle.
It is clear that new methodologies are needed to identify, locate, and analyze xanthophylls and their configuration and function in LHA complexes. One powerful approach has been to reconstitute light-harvesting complexes from Lhcb polypeptides and pigment mixtures of varying composition. Site-directed mutagenesis of these polypeptides has been used to determine the location and specificity of carotenoid binding sites (12). The effects of carotenoid binding to peripheral sites in LHA complexes has allowed investigation of their role in energy dissipation (14 -17). Along with the development of these biochemical techniques there has been progress toward the development of instrumental methods to analyze antenna xanthophylls. Absorption, linear and circular dichroism, and triplet state spectroscopies have been applied to identify their electronic absorption bands and energy transfer pathways to and from chlorophyll molecules (18 -22). However, it remains difficult to make unambiguous assignments for even simpler systems, such as those containing only three xanthophylls, and impossible to make reasonable spectral assignments in the Soret absorption region of whole thylakoid membranes. Recently, we have combined optical absorption spectroscopy with selective excitation resonance Raman spectroscopy in order to identify the absorption transitions of lutein and neoxanthin in LHCIIb trimers (23). Unlike previous work (18), which used LHCIIb containing three types of carotenoids, we have prepared LHCII, which contained only lutein and neoxanthin, a more simple system. The characteristic Raman feature for neoxanthin and lutein, the 1 maximum position, was identified and used to build the Raman excitation profiles for the isolated complex. This allowed not only an estimation of the energies for the three absorption bands (0-0, 0-1, and 0-2) but also an observation of the dynamics of carotenoid configuration upon oligomerization of the complex.
The approaches mentioned above are susceptible to various artifacts and limitations, such as the removal of lipids and pigments by the detergents used in the isolation and reconstitution procedures and the resulting alteration in protein conformation, xanthophyll binding affinity, and xanthophyll environment. One way forward would be the isolation of a more integrated LHA, in which pigments are less perturbed by detergent treatment. For example, we have recently isolated an oligomeric LHCIIb-enriched antenna complex using very mild detergent treatment of PSII particles and unstacked thylakoid membranes (9), and this preparation contained a higher proportion of violaxanthin and zeaxanthin. However, the most ideal approach would be to analyze carotenoid structure and dynamics in situ in intact thylakoid membranes.
The aim of the work described in this paper is to further develop the new spectroscopic methodologies to analyze xanthophylls, and most importantly to make progress toward their application to more complex systems including the whole thylakoid membrane. This approach not only enables the determination of the electronic transitions of all xanthophylls in vivo but also allows monitoring of their conformational and configurational dynamics and binding within the native LHA. To achieve this aim, we have first undertaken a systematic search for the characteristic features of isolated carotenoids that are found in the photosynthetic membrane, in order to use them for constructing the resonance Raman profiles. This allowed absorption band assignments not only for xanthophylls of isolated antenna complexes but also violaxanthin and zeaxanthin of intact antennae in thylakoid membranes. We have obtained for the first time the resonance Raman spectra of violaxanthin and zeaxanthin in vivo. It is concluded that zeaxanthin adopts a configuration that is likely to reflect its well defined binding within the antenna, rather than a free location in the membrane. This work establishes a new approach to the study of complex carotenoid-containing systems and offers a broad range of applications, from identification and assessment of xanthophyll configuration in reconstituted/isolated complexes to in vivo investigation of xanthophyll cycle carotenoids in order to establish their role in photoprotective mechanisms.

MATERIALS AND METHODS
Carotenoid samples were prepared as described in Ref. 24 by Dr. Denise Phillip (John Moores University, Liverpool, United Kingdom). LHCIIb was prepared from dark-adapted spinach leaves using isoelecrofocusing of PSII-enriched particles, as described in Ref. 8. Purification of LHCIIb trimers and removal of violaxanthin was carried out on a sucrose gradient (9). LHCIIb monomers were prepared by phospholipase A2 treatment of LHCIIb trimers for 36 h in the presence of 20 mM of CaCl 2 at a chlorophyll concentration of 0.5 mM, followed by purification on a sucrose gradient as described previously (9). Intact thylakoid membranes were obtained by the procedure described in Ref. 8. To induce maximum violaxanthin de-epoxidation, thylakoids were incubated at room temperature at a chlorophyll concentration of 200 M for 2 h in a medium containing 5 mM D-isoascorbate, 10 mM HEPES, and 10 mM sodium citrate at pH 5.5 with or without 5 mM Mg 2ϩ .
Absorption spectra were recorded on a Varian Cary E5 double-beam scanning spectrophotometer; measurements at 4 K were performed using a helium bath cryostat (Utreks). Low temperature resonance Raman spectra were obtained in a helium flow cryostat (Air Liquide, Paris, France) using a Jobin-Yvon U1000 Raman spectrophotometer equipped with a liquid nitrogen-cooled charge-coupled devices detector (Spectrum One, Jobin-Yvon, Paris, France) as described in Ref. 25. Excitation was provided by Coherent Argon (Innova 100) and Krypton (Innova 90) lasers (at 457.9, 476.5, 496.5, 488.0, 501.7, and 514.5 nm and at 528.7 and 413.1 nm, respectively) and a Liconix helium-cadmium laser (at 441.6 nm). The choice of this wavelength range was determined by the absorption profiles of the xanthophylls used. Fig. 1 displays absorption spectra of the four major xanthophylls that have been studied. The number of laser excitation lines (indicated in Fig. 1 by dotted arrows) covers all of the 0-0 transitions and also part of the electron-vibrational bands (0-1 and 0-2). It is likely that carotenoid absorption maxima in vivo will be somewhat shifted from those in organic solvent. However, from previous experience with LHCIIb xanthophylls, we have found that pyridine provides an environment with polarizability (n ϭ 1.5092) much closer to that of lipid and membrane protein environments than such commonly used solvents as n-hexane (n ϭ 1.3750) or ethanol (1.3611). The amplitude of the resonance effect is proportional to the square of the absorption probability. This offers an efficient line-narrowing effect in the Raman excitation profiles compared with absorption spectra, and the increased selectivity makes it possible to resolve complex bandshapes.

RESULTS
Wavelength-selective Resonance Raman Spectroscopy of Isolated Carotenoids-Carotenoids are very efficient Raman scatterers and exhibit a very strong resonance enhancement (26). Four main frequency regions have been observed, calculated, and assigned as follows: 1 -C ϭ C stretching vibrations; 2 -C-C stretches coupled either to C-H in-plane bending or C-CH 3 stretching; 3 -CH 3 in-plane rocking vibrations; 4 -C-H out-of-plane bending modes. In RR spectra of carotenoid molecules, the position of 1 varies according to the number of conjugated CϭC bands that these molecules possess, being higher in the case of shorter conjugated chains (27). cis-Carotenoids also exhibit higher 1 frequencies than all-trans-carotenoids (28,29). We have found this frequency to be in general unaffected by the solvent type (pyridine, n-hexane, ethanol, and cyclohexane; data not shown), in agreement with the published data for ␤-carotene (30). Fig. 2 shows 488 nm excited RR spectra of the four main LHC antenna xanthophylls and ␤-carotene, all dissolved in pyridine. The different numbers of conjugated double bonds in these carotenoids (from 9 to 11) and the 9-cis conformation of neoxanthin give rise to a large 1 variability from zeaxanthin at 1524 cm Ϫ1 to neoxanthin at 1533 cm Ϫ1 . Other regions of the RR spectra are also specific for certain xanthophylls. The arrows in Fig. 2 show the characteristic frequencies for neoxanthin in the 2 region at 1120, 1132, and 1203 cm Ϫ1 , probably due to its 9-cis conformation, and for violaxanthin in the 3 band at 1007 cm Ϫ1 . The 4 region was very low in intensity for all carotenoids, out-of-plane modes being formally resonance-forbidden for fully planar molecules. However, they can become significant under conditions in which the carotenoid undergoes configurational rearrangements leading to twisting of the molecule, due, for instance, to interaction with its environment. This situation is not frequently found in solvent or detergent media, but it has been observed in certain cases for carotenoids attached to antenna complexes (26,23). Molecular distortion of this kind requires energy, which can be gained in the close contact with protein environments such as a hydrophobic helix. Thus, the 4 region can be used as a marker for pigment-protein interactions and environmental perturbations involving the carotenoid molecule.
The position of the 1 band for each carotenoid was determined for all excitation lines used (Fig. 3A). Each additional double bond produces an ϳ3 cm Ϫ1 downshift of 1 . Comparing lutein and neoxanthin, there is a difference of 8 cm Ϫ1 in the 1 position. This parameter can therefore be used to identify xanthophyll absorption transitions in LHCII complexes containing these two types of xanthophylls (see below). Features in the 2 and 3 regions can be used to construct additional excitation profiles to confirm band assignments. For example, the 1 frequencies for lutein and violaxanthin for some excitation lines can be very close (only 2 cm Ϫ1 apart) (Fig. 3A). However, additional analysis of the 3 region can be used to make unambiguous assignments, because the corresponding frequency difference between lutein and violaxanthin is at least 4 cm Ϫ1 (Fig. 3B).
Organization of Xanthophylls in LHCIIb: Lutein and Neoxanthin-Neoxanthin and lutein have very strong binding affinities to LHCIIb compared with violaxanthin, which can be removed by detergent (9). Therefore, it was possible to prepare LHCIIb free from violaxanthin, containing only lutein and neoxanthin. The RR spectra of this LHCIIb sample, after monomerization, were obtained, and the 1 maximum position was analyzed as a function of the excitation wavelength. Whereas this parameter was only slightly variable with the resonance wavelength for isolated xanthophylls (Fig. 3), for LHCIIb monomers, it was strongly dependent upon it (Fig. 4A, open circles). For 457.9 and 488.0 nm excitations, the 1 position was close to that of isolated neoxanthin, whereas for other excitation wavelengths, it was similar to that of lutein. This indicates that neoxanthin contributions dominate the RR spectra obtained with 457.9 and 488.0 nm excitations. Fig. 4A also shows the second derivative of the absorption spectrum of the LHCIIb monomer, which shows two maxima, at 457 and 486 nm. The 29-nm shift between these bands is in a good agreement with that expected between 0-0 and 0-1 transitions of xanthophylls. An additional diagnostic parameter was the relative amplitudes of the neoxanthin-specific bands at 1203 cm Ϫ1 in the 2 region and at 1006 cm Ϫ1 in the 3 band (see Figs. 2 and 3). Excitation profiles for both of these closely matched the 1 excitation profile, confirming the assignment of the 485 nm band to neoxanthin. We suggest that the 495 and 466 nm bands observed in the absorption spectrum most likely originate from lutein (the 476 nm band arises from chlorophyll b).
The absorption spectrum of the LHCIIb trimer is known to contain a new transition around 510 nm not found in monomers (18, 23) (see Fig. 4B). Although it was suggested to belong to violaxanthin (18), it was still present in preparations free from this carotenoid (23). Analysis of RR excitation profiles for 1 in LHCIIb trimers has been used as evidence that the 510 band belonged to lutein, because the 1 frequency for both 501.7 and 514.5 nm excitations was very close to that of lutein (23). In the second derivative spectrum (Fig. 4B), it is significantly smaller than the 485 and 495 nm bands, which are suggested to arise from neoxanthin and lutein, respectively. Because there are two luteins bound to each LHCIIb, if the 495 nm band corresponds to the 0-1 absorption band of one lutein molecule, then the 510 nm band should belong to that of the other one. Why then are their absorption amplitudes in the second derivative spectrum so different? A possible answer to this question is that the 510 nm band may have a smaller extinction coefficient. However, a more likely explanation lies in the observation that the 510 nm band is at least 70% broader than that at 495 nm. As the amplitude of peaks in a second derivative spectrum is reciprocal to their bandwidth, this results in the relative amplitude of the 510 nm component being lower. In-deed, in the trimer minus monomer difference spectrum, the bandwidth of the band around 510 nm reaches 18 nm (Fig. 4B, solid line). Resonance Raman Spectroscopy of Thylakoid Membranes-The Soret absorption band of the whole thylakoid membrane is more complex than that of LHCIIb in the carotenoid region (Fig. 5); this complexity arises from, among other factors, the presence of extra carotenoids, violaxanthin, and ␤-carotene. However, it is possible to do comparative spectroscopic studies on the membranes by replacing violaxanthin with zeaxanthin by activation of the xanthophyll cycle. Activation of de-epoxidation in thylakoids yielded about 80% replacement of violaxanthin with zeaxanthin. The 4 K absorption spectrum changes significantly after de-epoxidation of violaxanthin. In Fig. 5, arrows indicate a decrease in 488 and 460 nm regions and the appearance of new bands at 505 and 476 nm. This is consistent with the room temperature difference spectra-illuminated-minus-dark, measured on leaves (31,32). However, the spectrum recorded at 4 K reveals more structure. The second derivative of the difference spectrum (ϩZea)-(ϩVio) resolves a complex picture (Fig. 5B). The three characteristic negative bands show a doublet structure at 4 K: 488/497, 452/460, and 429/423 nm. The doublet structure may arise from the two populations of violaxanthin (i.e. integrally bound or peripheral), both of which were de-epoxidized. There is also complexity in the positive bands, and the 503 and 511 nm positive bands may correspond to zeaxanthin 0-0 transitions. However, the de-epoxidation process could have caused changes in antenna conformation, which could then affect other xanthophylls, altering their absorption parameters and giving rise to the complexity of the spectra shown in Fig. 5. Therefore, more evidence is required to identify the origin of the absorption changes observed upon violaxanthin de-epoxidation.
RR spectra were measured for the thylakoid membranes containing only violaxanthin and those enriched with zeaxanthin. The spectra were clearly different in the 1 region, the position of 1 maximum being 3 cm Ϫ1 downshifted in thylakoids containing zeaxanthin (Fig. 6, circles). It was found that the RR spectra of thylakoids was much wider than the spectrum of LHCIIb (Fig. 6A, broken line). This is most likely to be due to a more complex structure of thylakoid RR because of additional contributions of violaxanthin, zeaxanthin, ␤-carotene, and small amounts of antheraxanthin. A deconvolution of this region using the 1 spectra of isolated carotenoids (Fig. 6A, solid lines) provides evidence to support this view. The spectrum for violaxanthin-containing thylakoids can be explained as the sum of lutein, violaxanthin, neoxanthin, and ␤-carotene (Fig. 6A). The 1 regions for ␤-carotene and zeaxanthin are almost identical and are added together in Fig. 6B; again, a good fit to the thylakoid spectrum was obtained. Thus, the downshift in the 1 position after de-epoxidation can be explained by an increase in the low-frequency zeaxanthin/␤-carotene signal (around 1522 cm Ϫ1 ) and a decrease in the violaxanthin band at 1528 cm Ϫ1 . Lutein (1526 cm Ϫ1 ) and neoxanthin (1534 cm Ϫ1 ) bands remained almost unchanged after de-epoxi-dation. The domination of the spectrum by lutein is explained by neoxanthin being further out of resonance from 501.7 nm excitation compared with lutein, because it absorbs at 485 nm, whereas lutein absorbs at 495 nm (see Fig. 1 and discussion above).
An alternative approach to deconvolution of the spectrum is to calculate difference RR spectra (de-epoxidized-minus-epoxidized) following normalization at the 1540 cm Ϫ1 region, where contribution of the zeaxanthin and violaxanthin signals is very low (see Fig. 6, A and B). Fig. 6C displays a number of such difference spectra for the 1 region. The vertical dotted line indicates the normalization point, where the difference was always zero. It was found that the shape of the spectrum was strongly dependent on excitation wavelength. With 528.7, 514.5, and 501.7 nm excitation, an almost symmetrical positive band was found, but with 496.5 and particularly 488 nm exci-FIG. 5. A, 4 K absorption spectra of thylakoid membranes containing violaxanthin (ϩVio) and following 80% conversion to zeaxanthin (ϩZea). (ϩZea)-(ϩVio) represents the corresponding difference spectrum. Vertical arrows indicate characteristic negative and positive bands in the difference spectrum. B, second derivative of the (ϩZea)-(ϩVio) difference spectrum, revealing the doublet nature of the spectrum (note that the spectrum has been inverted).
FIG. 6. 1 region of the resonance Raman spectra of thylakoid membranes containing violaxanthin (A) and upon 80% conversion to zeaxanthin (B), excited at 501.7 nm. Open circles represent the original spectra. Broken line, spectrum for LHCIIb, excited at 488.0 nm. L, V, N, Z, and ␤ indicate spectra of isolated lutein, violaxanthin, neoxanthin, zeqaxanthin, and ␤-carotene, respectively, used to fit the thylakoid spectra. Only the amplitudes have been varied to obtain the best fit using SigmaPlot software. As the spectra in this region are very similar for zeaxanthin and ␤-carotene (see Figs. 2 and 3), only one trace was used to represent both carotenoids for zeaxanthin-containing membranes (B). The modeled spectra matched the empirical data very closely and can be seen occasionally as dotted curves. C, (ϩZea)-(ϩVio) difference spectra for thylakoids at various excitation wavelengths. Original spectra were normalized at 1540 cm Ϫ1 (indicated by a vertical dotted line).
tation, there was a rather asymmetric decrease in the high frequency region of 1 due to a decrease at 1530 cm Ϫ1 . These data may be explained by the positive band arising from zeaxanthin and the negative one from violaxanthin, the wavelength dependence of the spectrum arising from differential excitation of these two pigments. Indeed, when the positions of the maxima and minima were plotted as a function of excitation wavelength (Fig. 7A), it was found that all components gave a 1 value either close to that for violaxanthin (around 1530 cm Ϫ1 ) or zeaxanthin (1520 cm Ϫ1 ); even the wavelength dependence of each matched that of the isolated pigments in vitro (Fig. 7A, full lines; taken from Fig. 3A). Furthermore, the relative extent of increase in 1 intensity around 1520 cm Ϫ1 matched the two positive bands in the low temperature difference spectrum de-epoxidized minus epoxidized (Fig. 7B). Therefore, it is concluded that the transitions at 505-510 and 476 nm belong to the 0-0 and 0-1 bands of zeaxanthin. On the other hand, the wavelength dependence for the intensity of the negative band around 1530 cm Ϫ1 indicates that 488 and 460 nm absorption bands arise from violaxanthin. Even the 497 nm second derivative band shown in Fig. 5 may also belong to violaxanthin, because excitation at 496.5 nm produced a relative decrease in the 1 intensity in the violaxanthin region. Unfortunately, the limited number of excitation lines available did not allow resolution of the 452 and 460 nm bands, but their distance from the 488 and 497 nm transitions suggests that they arise from 0-1 vibrational satellites.
An attempt was made to compare absolute RR amplitudes for violaxanthin-and zeaxanthin-containing thylakoids on the same scale without using this normalization procedure. First, exactly the same amount of chlorophyll was used in each sample, and an average was calculated from several replicates. Second, the small but reproducible signals from chlorophyll at 1437, 1354, and 1327 cm Ϫ1 were used for normalization. Both methods gave very similar results. As expected, the intensity of the 1 band was dependent upon excitation wavelength (Fig.  7C). The replacement of violaxanthin by zeaxanthin caused significant changes in the excitation profiles, particularly the decrease in intensity with 488 nm excitation and increases above 496.5 nm in the de-epoxidized thylakoids. The calculated difference spectrum (ϩZea)-(ϩVio) clearly showed the positive changes at 476.5 and 496.5 nm and the negative band at 488 nm, again similar to the shape of the corresponding absorption difference spectrum. This confirms that the band at 488 nm absorption is in a good resonance with 488.0 excitation. The formation of zeaxanthin enhances the RR signal above 500 nm, where zeaxanthin 0-0 absorption resonates with the 501.7, 514.5, and 528.7 nm lines.
The 3 region of the RR spectrum was also investigated in the thylakoid samples. The 3 for zeaxanthin is located at 1003 cm Ϫ1 , whereas for violaxanthin it is shifted down to 1007 cm Ϫ1 (see Fig. 3B and Fig. 7D, inset). Therefore, the 1003/1007 cm Ϫ1 amplitude ratio monitors the relative contribution of these xanthophylls. The dependence of this ratio upon excitation wavelength was determined (Fig. 7D). The lowest value of the ratio was found to be at 488 nm excitation, consistent with this resonance arising from violaxanthin. The ratio increased upon de-epoxidation for all excitations used, consistent with an increase in zeaxanthin content, whereas the larger differences between de-epoxidized and epoxidized samples observed for 514.5 and 528.7 nm excitation result from selective excitation of the zeaxanthin present.
Thus, excitation at 488.0 nm is selective for violaxanthin, whereas excitation above 500 nm (e.g. at 528.7 nm) is selective for zeaxanthin. With this information, it is possible to explore the state of violaxanthin and zeaxanthin in the thylakoid mem- brane, compared with pigments dissolved in detergent/lipid micelles or in organic solvent. In Figs. 8 and 9, RR difference spectra were obtained for xanthophylls dissolved in pyridine (spectrum 1), in the free pigment fraction following detergent treatment of thylakoid membranes (spectrum 2), and for thylakoid de-epoxidation treatment (spectrum 3) excited at 488.0 nm (Fig. 8) and 528.7 nm (Fig. 9). The spectra are (ϩVio)-(ϩZea) and (ϩZea)-(ϩVio) for 488.0 and 528.7 nm excitation, respectively. For 488.0 nm excitation, the thylakoid spectra in the 1 , 2 and 3 regions matches very closely that for isolated violaxanthin in pyridine and for the free pigment fraction. This clearly identifies the RR difference spectrum for thylakoids as the violaxanthin spectrum with characteristic violaxanthin features at 1529, 1184, 1213, and 1006 cm Ϫ1 and adds strength to the assertion that the 488 nm band originates from violaxanthin. The 4 region for the thylakoid spectrum exhibits one sharp transition at 949 cm Ϫ1 and another at 962 cm Ϫ1 , whereas the structure of this region for the spectrum of the isolated pigment, either in solvent or in detergent/lipid micelles, is almost absent, and intensity is reduced. The presence of features in the 4 region suggests that violaxanthin in vivo is distorted, most likely due to its binding to antenna complexes.
The thylakoid spectra for 528.7 nm excitation are again similar to the spectra of isolated zeaxanthin, either in solvent or in detergent-lipid mycelles (Fig. 9). The 1 position around 1522 cm Ϫ1 and 3 maximum at 1003 cm Ϫ1 are identical in all three spectra. The 2 regions of zeaxanthin in pyridine and the thylakoid spectrum are also similar apart from the downshift of the 1190 cm Ϫ1 band to 1185 cm Ϫ1 . This downshift is also present in the spectrum of the free pigment fraction. The 4 region for the RR difference spectra of thylakoid membranes is strongly enhanced and clearly structured compared with that for isolated zeaxanthin. For the free pigment fraction, this region is slightly enhanced but less structured in comparison to the thylakoid membrane spectrum. This suggests that in vivo zeaxanthin is in a well defined environment. DISCUSSION In this paper, we have demonstrated a new approach to the characterization of higher plant xanthophylls using comparison of absorption band structure and resonance Raman excitation profiles. The approach proved to be an effective methodology for identification and monitoring of the molecular conformation and configuration of carotenoids both in isolated pigment-protein complexes and, most significantly, in intact thylakoid membranes. This method is based on the identification of a number of characteristic fingerprints in the resonance Raman spectrum for each carotenoid involved. Analysis of iso- FIG. 8. Four main regions of the xanthophyll resonance Raman spectrum, excited at 488 nm. 1 is the isolated violaxanthin spectrum. 2 is a difference between Raman spectrum of the free pigment fraction, containing violaxanthin, and the spectrum for the fraction containing zeaxanthin (see text). 3 is the difference between the Raman spectrum of thylakoid membranes enriched in violaxanthin and the spectrum for membranes enriched in zeaxanthin. Original spectra were normalized to the chlorophyll bands at 1437, 1354, and 1327 cm Ϫ1 .
FIG. 9. Four main regions of the xanthophyll resonance Raman spectrum excited at 528.7 nm. 1 is the isolated zeaxanthin spectrum. 2 is a difference between Raman spectrum of the free pigment fraction, containing zeaxanthin, and the spectrum for the fraction containing violaxanthin (see text). 3 is the difference between the Raman spectrum of thylakoid membranes enriched in zeaxanthin and the spectrum for membranes enriched in violaxanthin. Original spectra were normalized to the chlorophyll bands at 1437, 1354, and 1327 cm Ϫ1 . lated LHCIIb monomers and trimers containing only two types of xanthophylls, lutein and neoxanthin, allowed the identification of their corresponding absorption bands and molecular configuration.
In the monomeric state of LHCIIb, the configuration and environment of the two LHCIIb luteins are very similar, both absorbing at 495 nm. The broader full width at half maximum for the band at 495 nm in the monomer spectrum compared with that of the trimer (Fig. 4) may suggest that their maxima positions differ within this region and therefore that their environment is also slightly different. In the trimer there is a 510 nm band not found in the monomer; this band was assigned to a red-shifted lutein in the former. Its intensity was significantly reduced in the second derivative spectrum because of the much broader FWHM compared with the short wavelength form of the pigment. In the trimer, the 510 nm lutein undergoes significant twisting, as a result either of an influence of the protein or of interaction with the other pigments or lipids present. It is not clear from our data whether this lutein plays a specific role in, for example, stabilization of the trimer and/or photoprotection of chlorophylls. However, these large differences in environment of the two luteins are important from structural and spectroscopic viewpoints (11, 18 -21). This approach has revealed the molecular dynamics of a bound xanthophyll as a function of the state of oligomerization of the complex, and this for the first time. These factors may (at least in part) be behind the existence of monomeric (minor antenna), trimeric (major LHCIIb), and oligomeric states of antenna proteins in vivo. In addition, the methodology described here could be also used in the investigation of the LHC reconstitution process, using various types of xanthophylls (33,34).
The application of RR spectroscopy to thylakoid membranes has revealed important new information about the xanthophyll cycle carotenoids, zeaxanthin and violaxanthin. The physiological role of these carotenoids is unclear, as is their location in the thylakoid membrane. Indirect measurements (correlating nonphotochemical fluorescence quenching with de-epoxidation state) have suggested that xanthophyll cycle carotenoids may be completely free in the thylakoid membrane and can only move to a very small number of quenching centers in LHCII proteins upon ⌬pH formation (35). Some data on isolated LH-CII have suggested that only the minor LHCII can bind violaxathin and zeaxanthin tightly and that therefore these must be the functional proteins in nonphotochemical fluorescence quenching (36). In contrast, we have suggested that all of these carotenoids are bound, mostly to peripheral sites on each of the proteins. The strongly associated violaxanthin molecules located within minor complexes were found to be inaccessible to the de-epoxidase enzyme and therefore unlikely to be involved in photoprotection (9,13). Therefore, we have suggested that the peripheral violaxanthin and zeaxanthin molecules play the key role in photoprotective energy dissipation. In this study, we have been able to obtain the first data on the state of the xanthophyll cycle carotenoids in the thylakoid membrane. The more structured RR spectrum for both violaxanthin and zeaxanthin in the thylakoid membrane compared with those free in solution indicate that both of these carotenoids are in fact in well coordinated environments in the thylakoid membrane, almost certainly bound to protein and not free to move. Zeaxanthin appeared to be more distorted than violaxanthin. Indeed, the 4 region is more structured and intense for zeaxanthin than violaxanthin, suggesting that the former is in a tight association with the protein, consistent with estimations of binding affinities to LHCII proteins (9).
Ultra-low temperature absoprtion spectroscopy has also revealed new information about these carotenoids. The zeaxan-thin-minus-violaxanthin absorption spectrum of thylakoids had a complex structure. The heterogeneity within this spectrum suggested that violaxanthin exists in two different populations. This suggestion is consistent with the observation that violaxanthin was found to be differently bound to the minor and major LHCII and LHCI complexes. It is possible that one population corresponds to violaxanthin associated with the minor LHCII or LHCI, whereas the other is peripherally bound to LHCIIb. An alternative explanation is that the dual band structure of the difference spectrum zeaxanthin minus violaxanthin originates from excitonically coupled violaxanthin. In this case, the coupling energy (V) will be within 180 cm Ϫ1 taking into account the Davidov's splitting (E) of approximately 360 cm Ϫ1 (E ϭ 2⅐V). In the case of a moderate transition dipole moment and a parallel orientation of two interacting violaxanthin molecules, they should be located within 0.5-0.7 nm of each other (37,38).
It is interesting to mention that acidification did not cause alteration in violaxanthin maximum positions, which one would anticipate in the case of ⌬pH-induced detachment of violaxanthin from antenna to undergo de-epoxidation. Resonance Raman spectra were also found to be unaffected (data not shown). This indicates that acidification has no immediate effect on the state of violaxanthin.
Clearly, the application of spectroscopic methods to the analysis of xanthophylls in isolated complexes and intact thylakoids provides a new approach to understanding the structure and dynamics of these molecules. In particular, the selectivity of the RR technique and its freedom from the artifacts and problems normally associated with absorption and fluorescence measurements on complex intact systems have provided new information about violaxanthin and zeaxanthin in vivo. In the future, we hope to apply a similar methodology to whole leaves.