Mechanisms Underlying Carotenoid Absorption in Oxygenic Photosynthetic Proteins*

Background: Tuning of carotenoid function, through modulation of their electronic properties, is seen throughout Nature. Results: Two photosynthetic proteins are able to modulate the effective conjugation length of bound carotenoid cofactors. Conclusion: Altering the conformation of conjugated end cycles via steric hindrance provides a means of tuning the electronic properties of carotenoids. Significance: A novel mechanism tuning the functional properties of carotenoids is revealed. The electronic properties of carotenoid molecules underlie their multiple functions throughout biology, and tuning of these properties by their in vivo locus is of vital importance in a number of cases. This is exemplified by photosynthetic carotenoids, which perform both light-harvesting and photoprotective roles essential to the photosynthetic process. However, despite a large number of scientific studies performed in this field, the mechanism(s) used to modulate the electronic properties of carotenoids remain elusive. We have chosen two specific cases, the two β-carotene molecules in photosystem II reaction centers and the two luteins in the major photosystem II light-harvesting complex, to investigate how such a tuning of their electronic structure may occur. Indeed, in each case, identical molecular species in the same protein are seen to exhibit different electronic properties (most notably, shifted absorption peaks). We assess which molecular parameters are responsible for this in vivo tuning process and attempt to assign it to specific molecular events imposed by their binding pockets.


Carotenoids (Cars)
are essential cofactors in the first steps of the photosynthetic process. They play a role as light harvesters, complementing chlorophyll (Chl) absorption in the bluegreen range of the spectrum (see e.g. Refs. [1][2][3][4]. They also act as photoprotective molecules through a number of different mechanisms. By quenching Chl triplet states ( 3 Chl), they prevent the energy transfer-mediated formation of singlet oxygen ( 1 O 2 ), one of the most harmful reactive oxygen species (5). 3 Chl species are inevitably formed, with a low but significant yield, during excitation energy transfer in light-harvesting proteins and/or after charge recombination in reaction centers (RCs; 6 -8). Cars can additionally quench any 1 O 2 that may nevertheless be formed. More recently it was shown that, in both plants and cyanobacteria, Cars play an essential role in regulating the amount of excitation energy reaching the RC in high light environments, thus preventing damage due to overexcitation of these proteins (9 -11).
Cars achieve these functions through their electronic properties, which arise from their linear conjugated polyene chain. They exhibit a fairly simple structure, built from the assembly of isoprenoid units (see Fig. 1), and a number of their electronic properties have been successfully predicted. Their main absorption transition, which corresponds to a transition from the ground state to the second excited singlet state (S 0 3 S 2 ), tightly depends on the number of conjugated carbon-carbon double bonds present in this chain (12)(13)(14)(15)(16) and on the refractive index of their local environment (17)(18)(19)(20)(21)(22). However, predicting the full electronic structure of Car molecules has turned out to be extremely complex. Despite the intense level of research on Car properties over the last 40 years, several new, low energy excited states have been proposed for these molecules in the last decade alone (23,24), and precise calculations of their electronic and vibrational properties still remain a challenge (25). In vivo, protein binding sites provide a highly anisotropic environment to Car molecules. In these conditions, it is extremely difficult to characterize the most important parameters that govern their electronic properties. However, determining these parameters would represent an important approach in our understanding of the role of the protein matrix in tuning the first steps of the photosynthetic process. Given that the role of carotenoids in other biological systems also generally involves their electronic properties (e.g. signaling functions, making specific use of their color), such an understanding should also prove more widely applicable throughout carotenoid research.
The combined use of electronic absorption and resonance Raman (RR) spectroscopies may help in determining the molecular parameters underlying the tuning of Car electronic transitions. In RR spectra, the frequency of the CϭC stretching mode of these molecules (the 1 band) gives direct access to the structure of the alternated system of their electronic ground state. As we recently showed (26) and as illustrated in Fig. 2, a different relationship is observed between the frequency of the 1 band and the position of the S 0 3 S 2 transition, depending on the molecular mechanism tuning this transition. The frequency of the 1 band may thus yield direct indications on these mechanisms.
We have studied the absorption properties of ␤-carotene and lutein ( Fig. 1) bound to two photosynthetic proteins isolated from photosystem II (PSII): the RC and the major light-harvesting complex, LHCII, respectively. The PSII-RC binds two ␤-carotene molecules, which at low temperature have their main absorption transition at 489 and 507 nm (27)(28)(29), the former being perpendicular and the latter parallel to the membrane plane (27,30,31). These Car molecules exhibit only limited singlet-singlet energy transfer to Chls and essentially no quenching of Chl triplets, but are both able to transfer an electron to the oxidized primary electron donor, albeit with very low efficiency (27,(32)(33)(34)(35). LHCII is assembled into a trimeric form in the photosynthetic membrane, with each monomer containing two lutein molecules whose binding sites are related by pseudo-symmetry. Whereas in LHCII monomers these luteins both absorb at 495 nm, in LHCII trimers one of them (lut2) has its absorption shifted to 510 nm (lut1 absorption remains at 495 nm; 3, 36). It has been proposed that one of these two luteins (lut1) is involved in quenching Chl singlet excited states during the pH-dependent phase of nonphotochemical quenching (9), the central mechanism that regulates excitation transfer in PSII.
In this work, we determine the factors that underlie differences in the electronic properties of these Car molecules. In addition, we attempt to relate these differences to the threedimensional structures of these proteins, as determined by x-ray crystallographic studies (31,(37)(38)(39). We propose a mechanism by which their binding pockets may impose specific conformations on carotenoids to modulate their electronic structure and thus tune their absorption spectra.
Sample Preparation-LHCII complexes were prepared from Spinacia oleracea PSII-enriched particles by isoelectric focusing, followed by sucrose gradient centrifugation (41). These trimeric complexes contain two lutein molecules, one neoxanthin, and only negligible amounts of violaxanthin per monomer (41). PSII-RCs with two bound ␤-carotenes were isolated following the method described by Telfer et al. (35). Prior to use, LHCII trimers and PSII-RC were concentrated using Centricon 30 K and 100 K cut-off concentrating tubes, respectively, to a final optical density Ͼ5 at 675 nm.
Spectroscopy-Absorption spectra were collected using a Varian Cary E5 double beam scanning spectrophotometer. The S 0 3 S 2 electronic transition of Car molecules displays three vibrational sublevels; the 0 -0 sublevel (the red-most band) was used to determine the position of this transition. RR spectra were recorded with 90°signal collection using a two-stage monochromator (U1000; Jobin Yvon, Longjumeau, France) equipped with a front-illuminated, deep-depleted CCD detector (Jobin Yvon). Excitation wavelengths were provided by a 24-W Sabre laser (Coherent, Palo Alto, CA); typically, Ͻ20 mW reached the sample, and sample integrity was verified by following RR spectral evolution during the experiment. Measurements at low temperature (77 K) were performed using a nitrogen-flow cryostat (Air Liquide, Sassenage, France).

RESULTS AND DISCUSSION
Isolated ␤-Carotene and Lutein-The position of the S 0 3 S 2 transition of Car molecules tightly depends on their molecular structure and in particular on the length of the -electronconjugated system. Increasing this length induces a progressive  loss of the double bond character of the CϭC bonds. This double bond character can be directly probed by RR spectroscopy, as it influences the frequency of the 1 Raman band. As a result, for a given solvent, a linear correlation between this frequency and the position of the S 0 3 S 2 transition exists, when expressed as a function of the length of the Car-conjugated chain (Fig. 2, black line). The frequency of the 1 Raman band for lutein and ␤-carotene does not correspond to that expected for Cars with 10 and 11 double bonds, respectively, whereas they nevertheless show the same relationship between 1 frequency and absorption position ( Fig. 2 and Ref. 26). These cyclic Cars, when isolated in solvents, actually display effective conjugation lengths of 9.3 and 9.6, respectively. This result was attributed to rotation of the conjugated end cycles out of the plane, such that the ring CϭCs are only partially conjugated (26,42). The position of the main absorption transition of Car molecules also tightly depends on the properties of the solvent in which they are dissolved, and particularly on its refractive index (17)(18)(19)(20)(21)(22). Again, there is a correlation between the position of the S 0 3 S 2 absorption transition and the frequency of the 1 Raman band for a given Car molecule, when expressed as a function of solvent polarizability. This correlation is plotted in Fig. 2 for ␤-carotene (blue) and lutein (red), and displays a different slope to that for different Cars in the same solvent (black), as already observed and discussed for several Car molecules (26).
␤-Carotene in PSII Reaction Centers-At low temperature, PSII-RCs exhibit two main peaks in the Car absorption region, at 489 and 507 nm (Fig. 3). Linear dichroism experiments showed that these peaks correspond to distinct Car molecules, with different orientations relative to the membrane plane (27), and considering their position, they must be attributed to the 0 -0 sublevel of the absorption transition of each molecule. Although the resolution between these peaks becomes much lower at room temperature, the overall position of the Car absorption transition appears not to shift by more than a few nanometers between low temperature and room temperature (see Fig. 3). This is not specific to PSII and was already observed in other photosynthetic complexes. For instance, in light-har-vesting complexes from purple bacteria it was shown that, in general, the absorption transitions of the bound Car molecules are very poorly dependent on temperature (43). Fig. 4A displays RR spectra of PSII-RCs obtained at room temperature with 514.5-and 488.0-nm excitation, chosen to be selective for the 507-and 489-nm-absorbing ␤-carotene, respectively. They contain four main groups of bands, denoted 1 to 4 , typical of Car molecules. The 1 band around 1520 cm Ϫ1 arises from stretching vibrations of CϭC double bonds. As mentioned above, its frequency depends on the length of the -electron-conjugated chain and on the Car conformation. The 2 band at 1160 cm Ϫ1 arises from stretching vibrations of C-C single bonds coupled with C-H in-plane bending modes. This region may be used as a fingerprint for the assignment of Car configurations (trans-cis). The 3 band at ϳ1000 cm Ϫ1 arises from in-plane rocking vibrations of the methyl groups attached to the conjugated chain, coupled with in-plane bending modes of the adjacent C-Hs. Finally, the 4 band around 960 cm Ϫ1 arises from C-H out-of-plane wagging motions coupled with CϭC torsional modes (out-of-plane twists of the carbon backbone). When the conjugated system of the Car is symmetrical and the molecule is planar, these out-of-plane modes will not be coupled with the electronic transition. As a result, these bands will not be resonance-enhanced upon excitation and will exhibit very low intensity in RR spectra. However, distortions around C-C single bonds will increase the coupling of these modes with the electronic transition, resulting in an increase in resonance enhancement, i.e. 4 gains intensity. Although both spectra are globally similar and typical for alltrans-␤-carotene, at 514.5 nm the 1 frequency is down-shifted by 6 cm Ϫ1 compared with 488.0 nm whereas 3 and 4 are both slightly narrower. These differences are also observed at 77 K ( Fig. 4B; see also Ref. 35), indicating that the detailed structure of each ␤-carotene molecule is the same at both temperatures. It is worth noting that the increase at 488.0 nm in the width of 3 at room temperature is translated into a splitting of this band into two components at 77 K. In Fig. 5 we plot the 1 frequencies and absorption positions of the two ␤-carotenes from PSII-RCs at room temperature  (red triangles) compared with the frequencies and positions of different length carotenoid molecules (black line) and of ␤-carotene in various solvents (blue line) taken from Fig. 2. The values obtained for the blue-absorbing carotene fit on the blue slope, suggesting that the position of this absorption transition is mainly governed by the local polarizability of its binding site. It may be noted that the binding site of this molecule exhibits a rather low local polarizability (equivalent to that found in chloroform, blue open symbol labeled f, with a polarizability value, R n , of 0.266). On the other hand, the difference in 1 frequency of the two ␤-carotene molecules accompanying the shift in their S 0 3 S 2 transition (see dashed red lines in Fig. 5) is much too large to be accounted for by changes in the local polarizability of their protein binding sites. In this case, considering the difference in the position of their electronic transition, a change of 1 frequency of approximately 2 cm Ϫ1 should be expected, i.e. three times smaller than actually observed. The 6-cm Ϫ1 change seen here corresponds to a sizeable change of the apparent length of the conjugated chain, probably through perturbation of its alternated system. Such a change in the alternated system (observed upon increasing the length of the conjugated chain) would indeed up-shift the position of the Car absorption by approximately 17 nm, i.e. it is sufficient to explain the difference in absorption between the two PSII-RC-bound ␤-carotene molecules. Thus the dashed line connecting these two points is parallel to the black line, which relates Cars of different chain length in the same solvent (Fig. 5).
The same relationship between absorption position and 1 frequency is also shown for measurements of the PSII-RC at 77 K (Fig. 5, inverted red triangles). The dashed line between these two points exactly parallels that obtained at room temperature, but at the lower temperature there is a shift of about 5 cm Ϫ1 in the 1 frequency for both ␤-carotene molecules. Exactly the same shift for this Raman band, between room and low temperature, has recently been described in RR spectra of Cars in solution and was explained by an intrinsic sensitivity of the Raman frequency to temperature (for details, see Ref. 44).
Lutein Molecules in LHCII-The absorption spectrum of LHCII trimers in the Car region is rather complicated because these complexes bind up to four carotenoid pigments per monomer, namely two luteins, one 9-cis-neoxanthin and usually one violaxanthin/zeaxanthin (present in negligible amounts in our sample). Indeed, the position of the individual absorption transitions could only be determined through second derivative analyses of absorption spectra obtained at 4 K (3). However, as in PSII, comparing the absorption spectra of LHCII at low temperature and room temperature suggests that the main band positions of the LHCII-bound Car molecules are very poorly sensitive to temperature (Fig. 6).
RR spectra of the LHCII-bound luteins at 77 K are displayed in Fig. 7B. As documented extensively in the literature (3,43,44), at this temperature 496.5-and 514.5-nm excitations yield RR spectra dominated by contributions from lut1 (absorbing at 495 nm) and lut2 (absorbing at 510 nm), respectively. The frequency of the 1 band, which arises from the CϭC stretching modes of lutein molecules, is observed at 1531 and 1527 cm Ϫ1 for lut1 and lut2, respectively. In the 3 region (around 1000 cm Ϫ1 ), lut1 (at 496.5 nm) exhibits two overlapping components of similar amplitude, at 1003 and 1007 cm Ϫ1 . The same two bands are seen for lut2 (514.5-nm excitation), but the intensity of the higher frequency component is less than one third that of the lower frequency one. The 4 region exhibits higher intensity and structure for lut2 than for lut1 (514.5-and 496.5-nm excitation, respectively), indicating a higher degree of distortion for lut2 in its LHCII binding site (as previously concluded; 3, 45). At room temperature, the broadening of the Car electronic transitions must, at least in part, impair selective excitation of each Car. In room temperature spectra obtained using excitation at 514.5 nm (Fig. 7A, upper trace), the 1 is quite narrow (full width at half-maximum ϳ16 cm Ϫ1 ), and the structure of the 4 region is very similar to that observed using the same excitation at 77 K. We may thus conclude that this wavelength still ensures selective excitation of the lut2 molecule at the higher temperature and that the distortion of this molecule, up to now only FIGURE 5. Correlation between the S 0 3 S 2 electronic transition and the frequency of the 1 Raman band for the two ␤-carotene molecules in PSII-RC at room temperature (RT; red triangles) and 77 K (red inverted triangles). For comparison, the relationship between Cars of different conjugation length in the same solvent (n-hexane) and the relationship expressed as a function of solvent polarizability for ␤-carotene are also shown (cf. Fig. 2). FIGURE 6. Absorption spectra of LHCII trimers at room temperature (RT; solid line) and 77 K (dashed line).

Tuning the Absorption of Photosynthetic Carotenoids
observed at low temperature, is also present in LHCII at room temperature. Thus this distortion does not result from temperature-induced reorganization of the binding site. In this spectrum the position of the 1 band is at 1522 cm Ϫ1 . As in PSII-RC, we thus observe a 5-cm Ϫ1 shift between experiments conducted at 77 K and room temperature. In contrast, the spectrum at 496.5 nm exhibits significant broadening of 1 when the measurement is taken at room temperature (Fig. 7A, lower trace; full width at half-maximum ϳ18 cm Ϫ1 ). The contributions of more than one carotenoid are thus present in this spectrum, due to a reduction in resonance selectivity as a result of broadening of the Car absorption transitions at the higher temperature. Indeed, in the 2 region a small but significant increase in intensity is observed for the shoulder at ϳ1130 cm Ϫ1 . This is consistent with an increase in neoxanthin contributions to the spectrum, as bands on the low frequency side of 2 are quite typical for 9-cis-carotenoids. Contributions of lut1 thus cannot be selectively observed in RR spectra at room temperature. Indeed, this was found to be difficult even at 77 K, where a "pure" lut1 spectrum was only obtained after removing the Neo contribution by differential analysis (46).
In Fig. 8, the absorption position and 1 frequency of luteins in LHCII are compared with those obtained for lutein in various solvents (red line) and with the frequencies and positions of different length carotenoid molecules (black line). A large shift is observed in 1 frequency between the two luteins at low temperature (4 cm Ϫ1 ; blue inverted triangles). As for ␤-carotene in PSII-RC, this shift, which reflects a change in the alternation of the conjugated CϭC chain, is enough to account for the energy difference between their S 0 3 S 2 absorption transitions. As in PSII-RC, the 1 frequency of lut2 at room temperature (the only one we could safely extract from the RR spectra) is shifted by approximately 5 cm Ϫ1 compared with its low temperature value. By analogy, we can deduce the expected 1 frequency of lut1 at room temperature, by shifting its low temperature value by 5 cm Ϫ1 . The resulting value fits perfectly with the in vitro relationship between the lutein 1 frequency and the position of its S 0 3 S 2 transition according to the polarizability of the sol-vent (Fig. 8, red line), tending to validate this approach. As in PSII-RC, we may conclude that the position of the electronic transition of this molecule is mainly governed by the local polarizability of its binding site. However, the deduced polarizability for this lutein is much higher than that calculated for the blue ␤-carotene in PSII-RC, as it corresponds to a value slightly higher than that for nitrobenzene (j, red open symbol; Fig. 8) with a refractive index, R n , of 0.319.
Mechanisms Tuning Carotenoid Absorption in PSII-RC and LHCII-In both PSII-RC and LHCII, RR spectroscopy unambiguously shows that the position of the absorption transition of the blue-absorbing Car molecule is governed mainly by the polarizability provided by the protein environment. Indeed, the position of this transition and the frequency of the 1 mode of these molecules strictly obey the correlation obtained for both ␤-carotene and lutein according to solvent refractive index. The deduced average value of the environment polarizability of the blue-absorbing ␤-carotene in PSII-RC is lower than that of the blue-absorbing lutein in LHCII. This is consistent with the environment deduced by analysis of x-ray crystallographic structures (30,31,37). In PSII-RCs, Cars are mainly surrounded by amino acids and are quite distant from other cofactors; they exhibit only low rates of energy transfer to the bound Chl molecules (27,33,34). On the other hand, the luteins in LHCII are in very close contact with the LHCII-bound Chl molecules, both at the levels of their end cycles and of the conjugated CϭC chain (37). Some of these Chls, such as Chl a 603 , are nearly in van der Waals contact with lut1 (closest distance 3.83 Å) and are likely to provide an environment of higher polarizability.
By contrast, it is quite clear that the energy shifts between the blue-and the red-absorbing Car molecules in both studied complexes are not induced by a variation in polarizability of their binding sites. If so, the position of the absorption transition of these Cars and their 1 Raman frequency would obey a correlation similar to the blue/red lines in Fig. 2, whereas it is clear that they deviate from these lines (see Figs. 5 and 8). Again,  this conclusion is consistent with the description of the environment of these molecules provided by the crystallographic structures of the two pigment-protein complexes: the blue and red luteins in LHCII, as well as the blue and red ␤-carotene molecules in PSII-RC, are embedded in quite similar protein environments, which are unlikely to display large changes in average polarizability (indeed, in LHCII, the two binding pockets are related by the local 2-fold symmetry of the complex). Instead, the absorption transition of these molecules and the frequency of their 1 Raman band behave as if the conjugated chain of the Car molecules was increased by nearly one CϭC double bond at constant polarizability (Figs. 5 and 8). Note that for both red-absorbing Cars, the main 2 band is also seen to shift to lower frequency in parallel with the downshift in 1 (Figs. 4 and 7); this is exactly as expected for an increase in conjugation length (see e.g. 47). Thus the apparent lengths of the conjugated chain of the red-absorbing lutein and ␤-carotene in LHCII and PSII-RC (at room temperature) become 10 and 10.2, respectively (Figs. 5 and 8). The external parameters susceptible to induce such changes are not documented in the literature, and again, there is no dramatic change in the environment of these pigments which could be at the origin of such a change. It was shown that the luteins of LHCII and the ␤-carotenes in PSII-RC experience different distortions at low temperature (3,35), and we show in this work that these distortions also exist at room temperature. However, small distortions around C-C bonds are expected to have little influence on the structure of the CϭC conjugated chain, and, whereas in LHCII the red-absorbing lutein is distorted (48), in PSII it is the blueabsorbing Car that exhibits the larger distortion (35).
However, lutein and ␤-carotene both exhibit shorter conjugation length in solvents than expected from their chemical structure (9.3 and 9.6, respectively; 26). This was explained in terms of out-of-plane rotations of the conjugated end cycles, resulting in a loss of conjugation. In the crystal structure of ␤-carotene, the ␤-ionone rings are indeed twisted out of the conjugated plane (dihedral angles ϳ42°; 49). Although no solution structure currently exists for either carotenoid, density functional theory (DFT) simulations performed on ␤-carotene predict dihedral angles for the end cycles of ϳ47° (42,50,51;see Fig. 9), and this has been calculated to shorten the conjugation length by the exact amount predicted from the Raman 1 position (42). Bringing one of these end cycles back into the plane of the CϭC conjugated chain should accordingly result in a net increase of the effective conjugation length of these molecules of approximately 0.6 -0.7, exactly as observed here for the redabsorbing lutein and ␤-carotene in LHCII and PSII-RC. We thus propose that these proteins are able to tune the absorption of their red-absorbing carotenoid via the rotation of conjugated end cycles toward the conjugated plane of the molecule, this rotation being imposed by their binding pocket through steric hindrance. As lutein only has one conjugated end cycle it must be this ␤-ring that is implicated in LHCII, whereas for ␤-carotene in PSII-RC this rotation may involve one or both end cycles.
The initial crystallographic structure of LHCII led to the conclusion that the lutein end rings were all in a conformation perpendicular to the CϭC chain (37), a conformation that would induce a further shortening of the CϭC chain of the lutein molecules. This is at variance with the present vibrational analysis of these pigments and with the position of their electronic transitions. However, a more recent analysis led to the conclusion that these molecules display different conformations, with lut2 being more distorted than lut1 (39), fully in agreement with our previous conclusions (3). The progressive distortion of lut2 occurring from C9 to C13, as observed in the latter analysis, twists one of the end rings to orient it parallel to residue TRP46, ϳ3.4 Å away, to optimize van der Waals interactions with it (see Fig. 9E; note, however, that this figure was drawn using the earlier structure). As a result, this ring is brought back into the plane of the CϭC chain, assuming that this ring contains the (partially) conjugated CϭC bond (i.e. that it is the ␤-ring) then it would become more conjugated as a result of this distortion, entirely consistent with our proposed mechanism. It would also explain why the red shift of lut2 absorption only occurs when this molecule is distorted; in LHCII monomers, where the lut2 conformation is relaxed, its absorption transition coincides with that of lut1 (495 nm; 3).
Similarly, in the most recent crystallographic structure of photosystem II, a clear difference appears between the cycle geometries of the two ␤-carotenes bound to the PSII reaction center (31). This is illustrated in Fig. 9, A-D, where the four end rings are compared with that of the DFT-calculated in vitro structure (42). Both cycles of the (blue-absorbing) ␤-carotene perpendicular to the membrane plane (Bcr645 in Protein Data Bank structure 3ARC) make a large angle with the conjugated CϭC chain (dihedral angles 59º and 68°; Fig. 9, C and D). This is quite different for the (red-absorbing) ␤-carotene parallel to the membrane (Bcr651), where one of the cycles makes a dihedral angle of only 12°with the plane of the CϭC chain, as seen in Fig. 9A. Once again, this twisting back into the plane allows the end ring to lie more or less parallel to an overlapping aromatic FIGURE 9. Structural details of the carotenoid end rings in PSII-RC and LHCII, drawn using PyMOL from Protein Data Bank entries 3ARC and 1RWT, respectively. Ball-and-stick representations colored by atom except for carbons are shown for: protein-bound carotenoids (carbons in yellow); aromatic residues discussed in the text (gray); and other cofactors present (white). For PSII-RC, the DFT-calculated ␤-carotene structure (42; cyan) has been fitted to Bcr651 (A and B) and Bcr645 (C and D). For LHCII, the latest structure containing the twist in lut2 is not yet available; in the structure shown here (E), whereas the ring orientation relative to TRP46 is correct, the position of carbon atoms immediately preceding the ring (C7-9) is not. residue, this time PHE113 of the PsbD polypeptide (ϳ3.9 Å away; Fig. 9A). Note that the second end ring, although lying out of the plane, also makes a smaller angle than those measured for the blue-absorbing Car (48°; Fig. 9B). These structural differences are perfectly in line with the conclusions of this study and would account for the difference in conjugation length measured by Raman between these two molecules.
Thus in both LHCII and PSII-RC, steric hindrance from a nearby aromatic residue forces an end ring of the red-absorbing Car back toward the conjugated plane of the molecule. This presents the possibility of testing our conclusions by site-directed mutagenesis; replacing this residue with a smaller, nonaromatic one (e.g. Ala) should allow the Car end ring to take up its relaxed conformation. As a result its conjugation length, absorption position and Raman 1 frequency would all be similar to that of the blue-absorbing Car and in vitro. We are currently designing mutagenesis experiments to carry out this work in photosystem II of the cyanobacterium Synechocystis sp. PCC6803. Note that, in the absence of well established isolation protocols for cyanobacterial PSII-RC, the analysis will have to be performed on core preparations of photosystem II, containing approximately 50 pigment cofactors including Ͼ10 ␤-carotenes. Extracting the absorption and Raman signatures of two individual Cars from this large population, although substantially more difficult, should nevertheless be possible.
It is worth noting that, in the RR spectra of the red-absorbing ␤-carotene bound to PSII, only a single, narrow contribution is seen in the 3 region, whereas for the blue-absorbing ␤-carotene 3 is relatively broad at room temperature and even splits into two components at 77 K (Fig. 4). Although this tendency is not so clear for LHCII, it is nevertheless seen; the red-absorbing lutein exhibits one major component with a satellite at higher frequency, whereas this higher frequency component is more prominent for the blue-absorbing lutein (Fig. 7). 3 arises from in-plane rocking vibrations of the methyl groups attached to the conjugated chain, coupled with in-plane bending modes of the adjacent C-Hs. In Raman spectra of fucoxanthin the 3 band is similarly composed of two components, which was attributed to differences in the methyl nearest neighbors in the chemical structure of the carotenoid (46). Along the same lines, the splitting of this band observed in the blue-absorbing Car may reflect the out-of-plane rotation of the end cycles, as this rotation is likely to perturb the rocking frequency of the neighboring methyl group. It is striking that in lutein, where only one rocking mode should be concerned, the intensity of the additional component is weaker than in ␤-carotene (where potentially two methyl groups may be concerned). In solvents, where it was concluded that the end cycles are at least partially out of the plane, the 3 is also observed to be broader in Raman spectra at room temperature (26). The structure of this band could thus be a direct indication of the conformation of the end cycles in both ␤-carotene and lutein, representing a probe for this mechanism of conjugation-length modulation.
Finally, it is striking that Nature has generally used Cars with conjugated end cycles in oxygenic photosynthesis. Our results show that by playing on the conformation of these cycles, the position of the absorption transitions of these cyclic Car molecules may be tuned by up to 15 nm/ring. Note that for a Car without end rings it would be highly unfavorable, energetically, to rotate the ends such that a conjugated CϭC became (partially) unconjugated. This is not the case when conjugated end rings are present; steric factors have already determined that these rings are poised in a partially conjugated conformation. It may be that this property explains the recruitment of Car molecules with conjugated end cycles in the photosynthetic process, allowing for an optimization of the excitation energy cascade in these complex light-transducing structures.