A proteoliposome-based system reveals how lipids control photosynthetic light harvesting

Integral membrane proteins are exposed to a complex and dynamic lipid environment modulated by nonbilayer lipids that can influence protein functions by lipid-protein interactions. The nonbilayer lipid monogalactosyldiacylglycerol (MGDG) is the most abundant lipid in plant photosynthetic thylakoid membranes, but its impact on the functionality of energy-converting membrane protein complexes is unknown. Here, we optimized a detergent-based reconstitution protocol to develop a proteoliposome technique that incorporates the major light-harvesting complex II (LHCII) into compositionally well-defined large unilamellar lipid bilayer vesicles to study the impact of MGDG on light harvesting by LHCII. Using steady-state fluorescence spectroscopy, CD spectroscopy, and time-correlated single-photon counting, we found that both chlorophyll fluorescence quantum yields and fluorescence lifetimes clearly indicate that the presence of MGDG in lipid bilayers switches LHCII from a light-harvesting to a more energy-quenching mode that dissipates harvested light into heat. It is hypothesized that in the in vitro system developed here, MGDG controls light harvesting of LHCII by modulating the hydrostatic lateral membrane pressure profile in the lipid bilayer sensed by LHCII-bound peripheral pigments.

Integral membrane proteins are exposed to a complex and dynamic lipid environment modulated by nonbilayer lipids that can influence protein functions by lipid-protein interactions. The nonbilayer lipid monogalactosyldiacylglycerol (MGDG) is the most abundant lipid in plant photosynthetic thylakoid membranes, but its impact on the functionality of energy-converting membrane protein complexes is unknown. Here, we optimized a detergent-based reconstitution protocol to develop a proteoliposome technique that incorporates the major lightharvesting complex II (LHCII) into compositionally well-defined large unilamellar lipid bilayer vesicles to study the impact of MGDG on light harvesting by LHCII. Using steady-state fluorescence spectroscopy, CD spectroscopy, and time-correlated single-photon counting, we found that both chlorophyll fluorescence quantum yields and fluorescence lifetimes clearly indicate that the presence of MGDG in lipid bilayers switches LHCII from a light-harvesting to a more energy-quenching mode that dissipates harvested light into heat. It is hypothesized that in the in vitro system developed here, MGDG controls light harvesting of LHCII by modulating the hydrostatic lateral membrane pressure profile in the lipid bilayer sensed by LHCII-bound peripheral pigments.
Photosynthetic energy transformation starts with the harnessing of solar photons and the spatial transfer of the collected light energy to photochemically active reaction centers local-ized within photosystems (PS) 4 to ignite electron transport. The processes of light harvesting and energy transfer are realized by specialized light-harvesting pigment protein complexes (LHC, 1). LHCs were tuned by evolution for an almost loss-free ultrafast electronic excitation energy transfer between proteinanchored pigments (1,2). Some LHCs are not only perfect light harvesters, but in addition have the remarkable built-in capacity to switch from a light-harvesting mode to an efficient energy-quenching mode under light stress (3)(4)(5). This energyquenching mode (qE) is one of the most important photoprotective mechanisms in photosynthetic organisms ensuring survival and fitness in a highly dynamic environment (6). In plants and algae, LHCs fulfill their dual role as light harvester and energy quencher as transmembrane integral protein complexes buried in the amphiphilic thylakoid membrane system. Excellent structural data for the major plant LHCII exists that makes it an interesting candidate to study light harvesting in thylakoid membranes (7,8). LHCII serves as the main light-harvesting antenna for PSII and under some conditions for PSI. Detailed structural information combined with data from ultra-fast and steady-state spectroscopy leads to an in-depth understanding of the functionality of the isolated LHCII and its dynamic switch between light harvesting and qE (e.g. 9 -11). In native membranes, however, LHCII is embedded in a lipid bilayer. A substantial gap in our knowledge base exists on how the lipid matrix in thylakoid membranes interacts with LHCII (and other proteins) and modulates light harvesting. For nonphotosynthetic membrane protein complexes, like mechanosensitive channels and membrane transporters, strong evidence exists that the composition and physicochemical properties of the lipid bilayer have significant impact on the conformation and functionality of these proteins (12)(13)(14)(15)(16)(17). A concept that bridges physicochemical lipid bilayer properties with structural and functional alterations of membrane proteins is the lateral membrane pressure (LMP) hypothesis (12,18) also known as forcefrom-lipid principle (17). The so-called nonbilayer lipids play a central role in the LMP hypothesis. In contrast to bilayer-forming lipids that have an overall cylindrical shape, nonbilayer lip-ids adopt a conical shape because their hydrophilic head group is smaller than the hydrophobic fatty acid tail. According to the lipid shape-structure concept (19) isolated conical-shaped nonbilayer lipids adopt nonlamellar structures such as inverted hexagonal (HII) phases in aqueous environments (rods of aligned lipids with the smaller head groups facing to the center of the rod and the fatty acids moiety facing outward). Many biomembranes contain a high percentage of nonbilayer lipids. Thylakoid membranes are dominated by the nonbilayer lipid monogalactosyldiacylglycerol (MGDG) that makes up about half of the total lipids (20,21). The role of MGDG in photosynthetic energy transformation and its regulation is not established and remains elusive (21). Recently, it was shown that the presence of MGDG in the lipid matrix increases the structural stability of LHCII. This observation was interpreted as a consequence of modulation of LMP by this nonbilayer lipid (22). Nonbilayer lipids generate a higher lateral membrane pressure in the hydrocarbon region of the lipid bilayer because of their bulkier fatty acid part requiring more space (12,13,18,23). The higher membrane pressure in the hydrophobic membrane region is sensed by membrane proteins that can modulate their conformation and functionality (24). For photosynthetic thylakoid membranes, molecular dynamics simulation confirmed that the presence of MGDG leads to a significant increase of the physical pressure in the hydrophobic fatty acid part (25). In this study, the impact of the nonbilayer lipid MGDG for LHCII functionality was tested by reconstituting isolated trimeric LHCII from spinach into lipid liposomes with different mol % of MGDG leading to proteoliposomes. Proteoliposomes are a versatile tool for studying energy-transducing membrane proteins (26) and have frequently been used in photosynthesis research (e.g. [27][28][29][30][31][32]. Here, a detergent-based reconstitution protocol (26,32) was optimized which generated LHCII-proteoliposomes with a very low protein density, allowing specifically the study of lipid-protein interactions.

LHCII proteoliposomes with low protein density
Most studies on LHCII proteoliposomes used relatively high protein-to-lipid ratios leading to LHCII aggregation. LHCII aggregation is known to cause energy quenching by changes of the conformation of the pigment-protein complex (e.g. [33][34][35]. LHCII aggregation interferes with lipid-only-induced alterations of the LHCII structure and function. To avoid proteinprotein interactions by LHCII aggregation, we refined a detergent-based reconstitution protocol leading to very low protein densities in the final LHCII-proteoliposomes (Fig. 1). Addition of detergent to preformed large unilamellar lipid vesicles (LUV) (made of isolate thylakoid lipids) destabilizes the LUV bilayer necessary to allow LHCII incorporation (step II in Fig. 1A). LHCII was isolated from spinach. We used the detergent Triton X-100 for the LUV destabilization because it has a distinct absorption peak around 275 nm (Fig. 1A, bottom) that is missing for other detergents. The 275-nm Triton X-100 absorption peak is used to monitor detergent removal upon Bio-Bead treatment (Fig. 1A, bottom). Thus, it is ensured that the final LHCII proteoliposomes are detergent free. The successful incorporation of LHCII into the liposomes is validated by colocalization of fluorescent dyes (Fig. S1) that stain the lumen of the proteoliposomes (pyranine), the lipid bilayer (BODIPY), and LHCII (chlorophyll autofluorescence). The final proteoliposomes have a mean diameter of about 200 nm (Fig. 1B), determined by dynamic light scattering and a molar lipid-totrimeric LHCII ratio of about 60,000 (Fig. 1C). Assuming that lipids are organized in a bilayer and occupy on average a molecular area of 0.66 nm 2 (36) and assuming that each LHCII trimer binds 42 chl (7,8), it follows that each proteoliposome contains on average six LHCII trimers. This translates to a protein area fraction of smaller than 0.2% (area of trimeric LHCII is 33.2 nm 2 ), i.e. the LHCII concentration in proteoliposomes is very diluted. Two types of proteoliposomes were prepared to study the impact of MGDG on the LHCII functionality. One type contained MGDG and the other type contained almost no MGDG (small traces come as contamination from other thylakoid lipids). For the latter, MGDG was replaced by the chargeneutral DGDG. The lipid and fatty acid analysis measured for the final LHCII-proteoliposomes preparation is given in Fig.  1D. The fatty acid profile for the four thylakoid lipid classes is in accordance with the literature (37). Native thylakoid membranes contain about 50 mol % MGDG (20,21). We decided to produce proteoliposomes with no more than 25 mol % MGDG (Fig. 1D) because some reports mentioned that higher MGDG abundances in liposomes can lead to HII formation (22,38). Although it is debated whether high MGDG concentration leads to HII phase in liposomes we want to avoid nonbilayer HII formation in LHCII proteoliposomes, and as a consequence the MGDG concentration was reduced to a "safe" value.

Structural integrity of LHCII in proteoliposomes
The structural integrity and protein aggregation level of LHCII in proteoliposomes was probed by fluorescence spectroscopy and circular dichroism (CD). Room temperature chlorophyll (chl) fluorescence spectra preferentially exciting chl a at 420 nm or chl b at 475 nm (39) reveal similar values for the maximal fluorescence emission wavelengths (Fig. S2, A and B). This indicates efficient energy transfer from chl b to chl a in LHCII proteoliposomes. This is further supported by almost indistinguishable normalized chl fluorescence emission spectra if exited at 420 nm or 475 nm for both MGDG-containing and MGDG-depleted proteoliposomes (Fig. S2, C and D). The presence of excitonically disconnected chl b would lead to a blueshifted shoulder for 475 nm excitation compared with 420 nm excitation. The lack of this shoulder gives clear evidence that all chl b is energetically well-connected to the LHCII pigment system. Low temperature (77 K) fluorescence spectra ( Fig. 2A) provide information about the unbinding of chls and LHCII aggregation. Diagnostic for chl b unbinding is an emission at around 655 nm (preferential chl b excitation at 475 nm). 77 K emission spectra of both proteoliposomes with and without MGDG reveal a very small 655 nm signal, indicating that the vast majority of LHCII in proteoliposomes is intact ( Fig. 2A), which is in agreement with the data in Fig. S2. Additionally, there is no difference in the 655 nm (F655) peak between MGDG-containing and MGDG-depleted LHCII proteoliposomes (see error bars in Fig. 2A). Furthermore, the emission

Lipid-induced energy quenching
around 700 nm (F700) is a signature for LHCII aggregation (33,40,41). F700 emission for detergent-solubilized LHCII trimers is about 10% relative to the maximal emission at around 681 nm (F681) (41). An increase in the F700/F681 ratio reflects LHCIItrimer aggregation. In LHCII proteoliposomes the F700/F681 ratio is 15-16% ( Fig. 2A). This low F700/F681 ratio in LHCII proteoliposomes suggests a very low level of LHCII-trimer aggregation, as expected for highly protein-diluted membranes. A low level of LHCII aggregation is also supported by fluorescence lifetime measurements (see below). As for the 655 nm emission, no statistically significant difference of the F700/F681 ratio exists between proteoliposomes with and without MGDG (see error bars in Fig. 2A). These conclusions are supported and complemented by CD spectra of LHCII-proteoliposomes. The CD spectra in the Soret region show a pronounced negative peak at 472 nm ((Ϫ)472 nm) that is characteristic for trimeric LHCII and is absent in monomeric LHCII (42)(43)(44). In detail, the (Ϫ)472 nm to (Ϫ)490 nm ratio in trimeric LHCII is between 0.7 Figure 1. Establishing an LHCII-proteoliposome system. A, scheme of producing LHCII proteoliposomes from LUVs by the detergent-mediated incorporation approach (see text for details). Graphs at the bottom shows absorption spectra at the different preparation steps. Note the Triton X-100 peak at 275 nm that disappears after Bio-Beads treatment. B, proteoliposomes size determination by dynamic light scattering. The Ϯ gives half of the full width at half-maximum (n ϭ 5, biological repetitions). No difference was apparent between ϩMGDG and ϪMGDG proteoliposomes. C, molar lipid to trimeric LHCII ratio in proteoliposomes. Lipid content was determined from 2D TLC, LHCII trimer content from chl determination assuming 42 chls per trimer. Error bars represent 95% confidence interval (n ϭ 5, biological repetitions). D, lipid and fatty acid analysis. Upper panel shows the mol % for the four different lipid classes in ϩMGDG and ϪMGDG proteoliposomes measured by 2D TLC. Error bars represent S.D. (n ϭ 3, biological repetitions for both ϩMGDG and ϪMGDG samples). Fatty acids for the individual lipid classes separated by TLC were quantified by GC. Error bars represent S.D. (n ϭ 4 for ϩMGDG and n ϭ 3 for ϪMGDG proteoliposomes).

Figure 2. Spectroscopic characterization of LHCII-proteoliposomes. A,
representative chl fluorescence spectra of ϩMGDG and ϪMGDG proteoliposomes. Note the low emissions at 655 nm (indicative for chl b) and 700 nm (indicative for LHCII aggregates). Error bars at both wavelengths represent S.D. from 13 biological repetitions. B, room temperature CD spectra. The spectra show the mean of two biological replicates per sample type. A CD spectrum for the pure buffer was subtracted. Characteristic wavelength for LHCII trimer and protein aggregation are indicated (see text for details).

Lipid-induced energy quenching
and 0.75 (43,44), that is very similar to the ratio of 0.72 in our LHCII proteoliposomes (Fig. 2B). Further support for a trimeric LHCII organization in proteoliposomes is given by the broad (ϩ)412 nm signal (Fig. 2B) that is absent in LHCII monomers (42). Clear signatures for LHCII aggregation are the occurrence of a (Ϫ)438 nm peak and a significant reduced (Ϫ)676 nm peak (45). Absence of these signatures in CD spectra of the LHCII proteoliposomes ( Fig. 2B) indicate that LHCII is in a nonaggregated state, which is in line with the low F700/F681 ratio (see above). There are also no significant differences in CD spectra in MGDG-containing and MGDG-depleted proteoliposomes. Overall, the fluorescence and CD spectra demonstrate the structural integrity of LHCII in our proteoliposomes and its organization as a nonaggregated trimer.

Impact of MGDG on the LHCII fluorescence yield
A straightforward method for detecting energy quenching in isolated LHCII is the measurement of the relative chl fluorescence yield (⌽ fluor ) because a decrease in ⌽ fluor gives strong indication for an increased dissipative deactivation of electronic excited states into heat. ⌽ fluor of the various samples can be compared by normalizing the maximal fluorescence intensity by the absorbance of each sample at the excitation wavelength. This fluorescence-to-absorption ratio (⌽ fluor ) quantifies the amount of light emitted from LHCII per light absorbed by the pigment protein complex. Excitation wavelengths probing either preferentially chl a (420 nm) or chl b (475 nm) were used (Fig. S2). To determine the absorption values at 420 nm (Abs 420 ) and 475 nm (Abs 475 ), the spectra were baseline corrected for an unspecific light-scattering background (continuous increase from red to blue spectral regions) caused by the proteoliposomes (Fig. S2E). Room temperature fluorescence excited at either 420 nm (Fig. S2A) or 475 nm (Fig. S2B) was measured from the same samples used for absorption spectroscopy allowing ⌽ fluor calculation. To compare MGDG-containing and MGDG-depleted proteoliposomes the ⌽ fluor for ϪMGDG proteoliposomes were set to 1 (Fig. 3). It turns out that ⌽ fluor for LHCII in MGDG-containing proteoliposomes is about 25% lower compared with their MGDG-depleted counterparts. This result provides strong evidence that the presence of MGDG in lipid bilayers causes energy quenching in trimeric LHCII. To study the impact of MGDG on energy quenching in LHCII in more detail a third MGDG concentration was added at around 10 mol % MGDG (Fig. 3B). ⌽ fluor in LHCII proteoliposomes with about 10 mol % MGDG is very similar to ⌽ fluor in MGDG-depleted samples.

Chl fluorescence lifetime analysis of LHCII proteoliposomes
A more in-depth characterization of MGDG-dependent energy quenching in LHCII was performed by measuring chl fluorescence lifetimes in ϩMGDG and ϪMGDG proteoliposomes by time-correlated single photon counting (TCSPC) (Fig. 4). Fig. 4A gives examples for chl fluorescence relaxation kinetics for ϩMGDG and ϪMGDG LHCII proteoliposomes. Kinetics for ϩMGDG proteoliposomes are clearly faster than in MGDG-depleted samples, suggesting that MGDG causes shortening of the excited state lifetime in LHCII. For both types of proteoliposomes the fluorescence relaxation kinetics can be fitted by bi-exponential decay curves as is typical for isolated trimeric LHCII (28,(46)(47)(48). It was confirmed that the bi-exponential parameters derived from TCSPC analysis are independent on the laser power used for this experiment (Fig. S3), indicating absence of exciton-exciton annihilation events. The average fluorescence lifetime ( av ) for MGDG-depleted proteoliposomes calculated from a bi-exponential fit is 3.66 ns (Fig.  4B). This lifetime is in agreement with literature values for nonaggregated trimeric LHCII in detergent shells that are around 3.6 ns (28,35,40,47). The good correlation of average fluorescence lifetimes between LHCII in detergent and LHCII in MGDG-depleted proteoliposomes indicates that LHCII is in a non-energy-quenched light-harvesting state in the latter. Closer inspection of the LHCII fluorescence kinetics in MGDG-depleted proteoliposomes reveals that the faster exponential component has a time constant of fast ϭ 0.46 ns with a relative amplitude (A) of A fast ϭ 0.15 (Fig. 4). For the slower

Lipid-induced energy quenching
dominating component (A slow ϭ 0.85) slow is 4.24 ns. It is noteworthy that the fast (ϳ0.5 ns) component is most likely not caused by LHCII aggregation (48). Rather it was hypothesized that the slow and fast fluorescence lifetimes reflect two conformations of nonaggregated LHCII trimers (48). No indication was found for a component with very long lifetimes that would indicate free chls, which have fluorescence lifetimes of around 6 ns (46,48,49). For example, chls that unbind from LHCII trimers by high Triton X-100 concentrations show lifetimes of about 5.8 ns (46,48). The absence of these very long-living decay components supports the conclusion that LHCII in proteoliposomes is structurally intact (see above). Addition of ϳ25 mol % MGDG to LHCII proteoliposomes induces significant changes in all four bi-exponential fluorescence relaxation parameters (Fig. 4, C and D). In detail, A fast increases slightly by ϳ4% at the expense of A slow . At the same time slow (3.4 ns) is ϳ20% faster and fast (0.63 ns) ϳ37% slower in MGDG-containing LHCII-proteoliposomes. As a consequence of these alterations in the two exponential fluorescence decay phases in ϩMGDG proteoliposomes the average lifetime ( av ) of LHCII decreased by 17% (3.05 ns) relative to their ϪMGDG counterparts (Fig. 4B). If the decrease of the steady state chl fluorescence yield (⌽ fluor ) (Fig. 3A) is exclusively caused by shortening of fluorescence lifetimes (activation of energy dissipative pathways in LHCII) then it is expected that the MGDG-induced decrease in av and the corresponding decreases in ⌽ fluor have the same magnitude. The MGDG-induced decrease in ⌽ fluor seems slightly higher with ϳ25%. However, it must be pointed out that this difference between av and ⌽ fluor is statistically not significant (see error bar in Fig. 3A). It follows that the decline in fluorescence yield of LHCII in ϩMGDG liposomes could be entirely explained by a shortening of fluorescence lifetimes.

Discussion
For nonphotosynthetic membranes, there is good evidence that generic physicochemical properties of lipid bilayers modulated by nonbilayer lipids exert control over membrane protein conformation and function. Until today the role of the nonbilayer thylakoid lipid MGDG on photosynthetic light harvesting was unknown. This study employed proteoliposomes to study the role of MGDG on the structure and function of plant LHCII. The benefit of proteoliposomes is that they are compositionally and structurally well-defined, i.e. they offer a significant advantage over complex structured intact thylakoid membranes for identification of cause and effect relationships. A prerequisite to examine lipid-induced alterations on LHCII structure and function with proteoliposomes is avoidance of LHCII aggregation by a drastic dilution in protein density. To our knowledge, only two publications published studies on LHCII on highly protein-diluted proteoliposomes. However, in both publications severe problems were reported for highly protein-diluted samples. In Ref. 30, LHCII trimers monomerized, and in Ref. 31, proteoliposomes without MGDG form mixtures of lipid vesicles and planar sheets, making the direct comparison with their MGDG-containing counterparts difficult. Here we established a LHCII proteoliposome reconstitution protocol with very low protein densities (molar lipid/LH-CII trimer ratio ϳ60,000) embedded in large (ϳ200 nm diameter) LUVs with different mol % of MGDG. The large proteoliposome diameter ensures minimized membrane bending forces for LHCII in the lipid bilayer. In detail, the curvature of a proteoliposome with 200 nm diameter leads to height difference of the lipid bilayer on two opposite sides of the LHCII trimer (distance 7.5 nm) of less than three Å, i.e. LHCII experience an almost flat lipid-membrane that resembles the situation in native thylakoid membranes. Steady state fluorescence, CD spectroscopic, and fluorescence lifetime analyses of LHCII for both MGDG-depleted and MGDG-containing proteoliposomes (Fig. 2) reveal the structural integrity of the complex (no chl unbinding) as well as its organization as a nonaggregated trimer, laying the foundation to study lipid-LHCII interactions specifically. The key outcome of this study is that presence of MGDG in proteoliposomes leads to significant energy quenching of LHCII (⌽ fluor is lower and av is shorter) (Figs. 3 and 4). Because nothing else changed in the two types of proteoliposomes except the MGDG content (i.e. both ϪMGDG and ϩMGDG proteoliposomes contain the same amount of charged lipids, Fig. 1D) this result gives strong evidence for the specific role of the nonbilayer MGDG for modulating light harvesting by LHCII, i.e. MGDG in lipid bilayers induces energy quenching in LHCII. However, it is noteworthy that the MGDG-induced change in fluorescence lifetimes from 3.66 to 3.05 ns is a magnitude higher than the photochemical trapping time in PSII (300 to 500 ps), i.e. the impact on competing with photochemistry is small. On the other hand, we want to point out that native thylakoid membranes contain about ϳ50 mol % MGDG thus the fluorescence quenching observed here with ϳ20 mol % could be significantly higher in native thylakoid membranes.
An intriguing question is the molecular mechanism of how nonbilayer MGDG determines the light-harvesting efficiency of LHCII. One possibility is that modulation of the LMP by MGDG leads to conformational changes in LHCII that trigger dissipative pathways for excited pigment states. To explore this possibility in more detail we generated an in-scale model of the LHCII trimer together with LMP changes induced by MGDG derived from molecular dynamics simulations (25) (Fig. 5). Fig.  5 gives the false color-coded structural flexibility of LHCII that is derived from the crystallographic temperature (or B) factor (8). The color code varies from very rigid regions (blue) to highly flexible regions (red). As expected for a membrane integral protein complex the hydrophobic part made of rigid ␣-helices embedded in the lipid bilayer is mostly very stiff (8,50,51). However, an important exception is the xanthophyll neoxanthin (neo) that protrudes from the protein surface into the fatty acid region of the lipid bilayer. This neo is highly flexible (8,51). Comparing MGDG-induced lateral pressure changes along the z axis of the lipid bilayer (red curve in Fig. 5, bottom) with the LHCII flexibility reveals a significant (several 10 MPa) increase in lipid-bilayer pressure on the level of neo (orange arrow). Thus, it is likely that the presence of MGDG in lipid bilayers leads to bending of LHCII-bound neo as was reported for LHCII crystals where adjacent LHCII proteins cause neo bending (34). Interestingly, based on the fact that neo distortion correlates with energy dissipation (51-53), it was hypothesized that the bending of LHCII-bound neo switches the protein Lipid-induced energy quenching from a light harvester to an energy dissipater (52,54). Indeed, molecular dynamics simulation identified a strong correlation between neo bending, pigment rearrangement in LHCII, and energy quenching (51). However, it is also possible that neo distortion is not the trigger but only a reporter for an energy dissipative state that is established by other pigment-pigment rearrangements (e.g. Chl a603-lutein2) (51). Also, chls a611/612 are potential candidates for LMP-induced quenching because they are localized at the LHCII periphery (PDB ID: 1RWT). We speculate that the lateral pressure increase at a certain z-position in the lipid bilayer, caused by MGDG, leads to distortion of neo or other pigment-pigment rearrangements and switches LHCII to an energy-quenching state (Fig. 5, bottom). A direct impact of physical pressure on light harvesting by LHCII was observed by studying the functionality of the isolated detergent-solubilized protein at different isotropic hydrostatic pressures measured with a high-pressure cell (48). In accordance with our study, av and ⌽ fluor decrease with an increase in hydrostatic pressure in the pressure cell, i.e. LHCII switches to an energy-quenched mode (48). The similarities in both studies goes further: In Ref. 48, the amplitudes of the two fluorescence lifetime components change with increasing pressure, i.e. A slow /A fast decreases. The same tendency is observed for ϩMGDG liposomes. A slow /A fast decreases from 5.6 in MGDGdepleted proteoliposomes to 4.4 in MGDG-containing samples (Fig. 4C). Furthermore, in the isotropic pressure cell experiment slow accelerates, whereas fast decelerates with higher pressure (48). We observed the same changes from ϪMGDG to ϩMGDG proteoliposomes (Fig. 4D). These similarities between isotropic hydrostatic pressure experiments on detergent-solubilized LHCII trimers and the impact of MGDG in LHCII proteoliposomes support the model that the MGDG-dependent pressure increases in hydrophobic regions of the lipid bilayer are sensed by flexible LHCII parts leading to energy quenching (Fig. 5). An intriguing outcome of the study of Ref. 48 is that pressure-induced energy quenching is caused by very localized and small conformational changes and not by largescale compression of the overall protein. These localized structural changes explicitly include the possibility of neo bending (48) in support of our model presented in Fig. 5.
The physiological relevance for a LMP-dependent mechanism to switch LHCII structure and function as proposed here is given if the LMP is variable, i.e. the LMP can be changed by environmental factors. Three observations indeed indicate that LMP in thylakoid is variable. (i) Under certain conditions MGDG separates to nonbilayer HII phases in intact thylakoid membranes (e.g. [55][56][57][58]. This lipid-phase separation of MGDG would decrease the abundance of MGDG in the bilayer phase. As a consequence, the lateral membrane pressure in hydrophobic membrane bilayer regions decreases that according to the model in Fig. 5 would switch LHCII to a light-harvesting state. Interestingly HII formation was found stimulated by low light treatment (58), i.e. at situations where efficient light harvesting is required and not photoprotection. (ii) The results summarized in (i) measured large-scale HII formation that are visible in electron micrographs. However, recent molecular dynamics simulations indicated that nonbilayer HII formation could be more frequent on a smaller nanometer-length scale that is not detectable by conventional EM (36) as was proposed earlier (56). The molecular dynamics study reveals that the propensity for HII formation is strongly dependent on environmental conditions (e.g. water content). (iii) The LMP profile is a generic physicochemical property of the lipid bilayer and is therefore dependent not only on nonbilayer lipids but on all lipophilic membrane components. In this respect, the xanthophyll zeaxanthin (zea) is interesting. The presence of zea in the lipid bilayer increases membrane order and rigidity (59,60) that is known to enhance the lateral membrane pressure in hydrophobic membrane regions (13). Zea is converted from violaxanthin by the xanthophyll cycle and promotes qE formation (61,62). Thus, modulation of generic physicochemical membrane properties by free zea in the membrane could be another way to

Lipid-induced energy quenching
control LMP and therefore light harvesting by LHCII. It follows that the xanthophyll cycle has two implications. First, zea would activate qE directly by the well-established binding to LHCIIs (61). Second, zea could modulate light harvesting indirectly via changes in LMP. The three examples presented above demonstrate that the lipid matrix in thylakoid membranes is dynamic and responsive to environmental conditions that could modulate the LMP that in turn could control light-harvesting efficiency by LHCII (Fig. 5). Furthermore, evidence exists that the lipid/fatty acid content in plant thylakoid membranes undergo diurnal alterations (63,64) and change under stress like heat (65), cold (66), hypoxia (67), drought (68), and phosphate starvation (69). However, future studies have to reveal the dynamics of lipid/fatty acid compositions and its implications on light harvesting.

Isolation of trimeric LHCII
Trimeric LHCII was isolated from dark-adapted spinach from a local market according to Ref. 41. This protocol goes back to the original Triton X-100 -based isolation procedure described in Refs. 70 and 71. The final LHCII preparation was solubilized in 0.37% Triton X-100 at a chl concentration of 1.65 mM.

Preparation of LHCII proteoliposomes
For liposomes, the isolated plant lipids MGDG (C36:6 and C34:6), DGDG (C36:6 and C34:3), PG (C34:4 and C34:3), and SQDG (C34:3) were used (Lipid Products). Approximately 5 mol of isolated lipids in chloroform were mixed and evaporated off by nitrogen gas to form a thin lipid layer. The lipid film was rehydrated in a 10 mM HEPES buffer (pH 7.6, KOH) and vortexed thoroughly. The resulting multilamellar liposomes were passed sequentially through a 0.4 m and 0.2 m nucleopore membrane using a high-pressure extruder (Lipex TM ) at ϳ3.5 bar for the 0.4 m extrusion, and 10 bar for the 0.2 m extrusion, leading to LUVs. The lipid concentration in LUVs was determined by 2D thin layer chromatography (TLC) as described in Ref. 72. The lipid concentration in LUV stock was adjusted to 0.8 mM in 1.6 ml. Triton X-100 was added to a final concentration of 1 mM to destabilize the liposomes. After 5 min of incubation under slow stirring, 200 l of isolated LHCII at a chl concentration of 25 M (presolubilized in 600 M Triton X-100/10 mM HEPES, pH 7.6) was added and incubated for 30 min at room temperature (incorporation of the protein into LUVs), the Triton X-100 was removed with 25 mg Bio-Beads TM SM-2 Resin overnight at 4°C, followed by two 100-mg Bio-Beads TM addition steps for an hour and 45 min, respectively. The Triton removal was tracked using the 275 nm absorbance. Finally, proteoliposomes were further purified by gel filtration with a PD-10 Sephadex filtration column. Proteoliposomes were stored on ice in the dark and used freshly. The pigment composition of proteoliposomes with and without MGDG are (relative to total Chl): neoxanthin, 0.9 Ϯ 0.1 and 1.0 Ϯ 0.1; violaxanthin, 0.2 Ϯ 0.1 and 0.3 Ϯ 0.1; lutein 2.4 Ϯ 0.1 and 2.3 Ϯ 0.1, respectively. No zeaxanthin was detected.

Biochemical characterization
The lipid classes (MGDG, DGDG SQDG, PG) of proteoliposomes were quantified from lipid extracts by 2D TLC (73). From the same lipid extract fatty acids were converted to fatty acid methyl esters and quantified with gas chromatography (GC) by comparison of the fatty acid GC peaks with an internal fatty acid standard (2.613 nmol l Ϫ1 1,2,3-tripentadecanoylsn-glycerol, Nu-Chek Prep) as described in Ref. 65.

Steady-state fluorescence spectroscopy
For ⌽ fluor measurements anaerobic conditions with undiluted proteoliposomes (A 420 nm ϭ 0.1) were established with a glucose/glucose-oxidase/catalase system. Oxygen measurements with a polarographic oxygen electrode reveal anaerobiosis after about 1 min with this enzyme system for about 1 h. Anerobiosis was required to minimize damage to LHCII because of the measuring lights (the lack of damage was indicated by unchanged signal amplitudes in repetitive measured fluorescence spectra). The same anaerobic sample was used for both absorption and fluorescence spectra to determine Abs 420 and Abs 475 and the maximum fluorescence intensities at these excitation wavelengths. Optical absorption spectra were recorded with a Hitachi U3900 spectrometer (2-nm slit width, 200 -750 nm, optical path length 10 mm). Chl fluorescence spectra were measured with a FluoroMax 4 (Horiba Yvon) spectrofluorometer. Emission spectra between 640 to 800 nm (slit width 4 nm) were recorded for 420 or 475 nm excitation wavelength (slit width 2 nm). Fluorescence spectra at 77 K were measured with samples shock-frozen in liquid nitrogen. The excitation wavelength was set to 475 nm (2 nm slit width).

CD spectroscopy
VIS-CD (CD between 400 to 700 nm) spectroscopy measurements were performed with an AVIV 202SF CD spectrometer in a 0.5 cm cuvette in 10 mM HEPES/KOH buffer with concentrated proteoliposomes. The proteoliposomes concentration was performed with Amicon Ultra (30 kDa exclusion size). Redilution of concentrated proteoliposomes leads to no alteration in 77 K fluorescence spectra, indicating that the concentration did not changed LHCII integrity. Spectra recorded with HEPES buffer alone were subtracted. Measurements were done under anaerobic conditions (glucose/glucose-oxidase/catalase system).

Time correlated single photon counting
TCSPC measurements of fluorescence lifetime were acquired using Becker-Hickl module SPC-150 in conjunction with Becker-Hickl SPCM software. The 420-nm excitation pulses with a repetition rate of 3.8 MHz were generated by a Coherent Verdi G10 532 nm diode pumped laser which pumps a Coherent Mira 900f Ti:Sapphire Oscillator set to 840 nm. The resultant pulsed beam was then frequency doubled using a ␤-barium borate crystal to obtain 420-nm pulses at a repetition rate of 76 MHz. A pulse picker composed of a Harris SiO 2 crystal, and a Coherent 7200 cavity dumper in combination with an ENI voltage amplifier (model 403LA) to drive the acoustic waves was used to reduce the repetition rate to the desired 3.8 MHz for these experiments. Samples were prepared to have an optical Lipid-induced energy quenching density of ϳ0.1 and measured in a Starna cell cuvette at room temperature with Spectrosil far UV Quartz windows, a 1-mm path length (front face detection), and a 0.4 ml volume. To obtain sufficient fluorescence counts from the sample, a long pass filter (polarizer set to magic angle) was used to detect wavelengths longer than 650 nm. 420-nm light was used to excite the chlorophyll a Soret band. Fluorescence was detected using a Hamamatsu R3809U microchannel plate photomultiplier tube (MCP PMT) and the IRF had a full width half-max of ϳ60 ps. Measurements were taken at several powers and it was determined that there was no power dependence for measurements taken at average powers less than 100 W (Fig. S3). Therefore, all measurements taken at powers below 100 W were averaged to obtain a larger sample size. The full width at half-maximum of the laser spot was about 650 m. Under our conditions the fluorescence lifetimes were unchanged for at least 60 min indicating stability of the samples over the time of measurements.

Dynamic light scattering
The diameter of the LUVs and proteoliposomes was measured in a 10-mm cuvette using a Delsa TM Nano Zeta Potential and submicron particle size analyzer (BD Biosciences). Latex beads (BD Biosciences) between 100 nm and 300 nm in diameter were used to create a size standard curve. The liposome and proteoliposome diameters were determined from this standard curve.