The light reactions of photosynthesis are hosted and regulated by the chloroplast thylakoid membrane (TM) — the central structural component of the photosynthetic apparatus of plants and algae. The two-dimensional and three-dimensional arrangement of the lipid–protein assemblies, aka macroorganisation, and its dynamic responses to the fluctuating physiological environment, aka flexibility, are the subject of this review. An emphasis is given on the information obtainable by spectroscopic approaches, especially circular dichroism (CD). We briefly summarise the current knowledge of the composition and three-dimensional architecture of the granal TMs in plants and the supramolecular organisation of Photosystem II and light-harvesting complex II therein. We next acquaint the non-specialist reader with the fundamentals of CD spectroscopy, recent advances such as anisotropic CD, and applications for studying the structure and macroorganisation of photosynthetic complexes and membranes. Special attention is given to the structural and functional flexibility of light-harvesting complex II in vitro as revealed by CD and fluorescence spectroscopy. We give an account of the dynamic changes in membrane macroorganisation associated with the light-adaptation of the photosynthetic apparatus and the regulation of the excitation energy flow by state transitions and non-photochemical quenching.
The chloroplast thylakoid membrane (TM) is the central structural component of the photosynthetic apparatus of plants and algae housing all major protein complexes that carry out the photoinduced electron and proton transfer. The highly specialised role of the TM in oxygenic photosynthesis, which emerged at least 2.4 billion years ago in cyanobacteria, according to the geological and fossil records [1,2], is reflected in their biochemical composition and structural organisation, which is distinct from any other biomembranes. The TM solves critical design problems in constructing and operating the photosynthetic machinery. On one hand, it is a sound framework embedding the protein complexes, enabling their lateral and vertical separation, defined spatial arrangement of the electron transport (ET) chain components and vectorial proton and ET, insulating the two aqueous phases — stroma and lumen — to allow and withstand the formation of transmembrane electrochemical potential. The dense packing of proteins in the membrane and the stacking of granal thylakoids provide a large absorption cross-section per unit volume. We term these features, i.e. the two-dimensional and three-dimensional arrangement of protein complexes in the TM, as macroorganisation. On the other hand, the TM is necessarily a flexible and dynamic structure, which needs to accommodate a number of mobile carriers as part of the normal operation of the photosynthetic ET chain but more importantly to provide for dynamic regulation of the photosynthetic reactions under a constantly fluctuating environment and for the removal, repair and replacement of damaged or expired units — be it proteins or cofactors. We refer to these functions and the related structural and macrostructural changes as the flexibility of the TM.
The remarkable three-dimensional lamellar organisation of the TM has sparked interest and has been under the microscope, literally and figuratively, since the introduction of electron microscopy (EM) (for historical review see ref. ). New details have been continuously emerging with technological advancement — cryo-EM, for instance, has recently brought near-atomic resolution structures of the major photosynthetic supercomplexes Photosystem I (PSI) and Photosystem II (PSII) [4–6] in the membrane. Although the number of books and review articles devoted to the topic is too large for all to be listed here, we must recommend several outstanding reviews on the TM architecture [7–10], biogenesis [11–13], dynamics [14–16] and methodology for visualising the TM structure .
Optical and EM are two families of methods comprising the principal tools for studying the chloroplast ultrastructure. In addition, various spectroscopic approaches have been successfully applied to probe the structural dynamics of the TM — including optical electronic and vibrational spectroscopies, NMR and electron spin resonance, as well as X-ray and neutron spectroscopy techniques. In general, spectroscopy has several potential advantages, for example: (1) it can be non-invasive, requiring minimal preparation of the sample or applied in vivo and in situ; (2) it can follow real-time dynamics in the functioning, living system; (3) both single particles or ensembles can be probed with inherent statistical averaging. In this review, we briefly summarise the current knowledge on the architecture, the lateral and multilamellar macroorganisation and the dynamic flexibility of the TM with a special emphasis on the information gathered by spectroscopic means and particularly by circular dichroism (CD) spectroscopy. The review is organised as follows: in section 1, we present the composition, general architecture and macroorganisation of the TM on different levels of hierarchy; in section 2, we introduce CD spectroscopy and recent advances and applications for studying the structure and macroorganisation of the photosynthetic complexes; section 3 is devoted to the structural flexibility of LHCII, as a main factor in the TM macroorganisation and flexibility; section 4 discusses the structural dynamic of the TM primarily associated with light acclimation and adaptation.
Architecture of the TM
The granal TM
In the compartmentalised cells of eukaryotic algae and plants, the TM are found in the chloroplast inner volume, surrounded by the chloroplast stroma. In developing chloroplasts or proplastids, the TM can form a continuum with the inner chloroplast envelope but in mature chloroplast the two are separated . Higher-plant TM have a complex three-dimensional architecture generally consisting of stacked thylakoids forming grana that are interconnected by stroma lamellae. The duality makes for discrete membrane regions — the non-appressed stromal regions, the appressed membranes in the granum core and the grana margins, each with its distinct protein composition . Thylakoids are not isolated vesicles, as might be implied by their name (sac-like), but form a continuous network enclosing a single aqueous volume, the thylakoid lumen [20–22].
The three-dimensional structure of the granum has been a debatable subject for decades and several models have been put forward and used in the scientific and educational literature (for reviews see [7,10]). One of the earliest, the ‘helical model', was proposed first by Paolillo , based on EM of serial thin sections, and was later slightly refined and revised by Mustárdy and co-workers [24,25]. This model is supported by most high-resolution electron tomography data available [21,22,26,27]. According to the helical model, the granum has a roughly cylindrical shape comprising a variable number of stacked granal thylakoids with a flat discoid shape and typical diameters in the range of 400–600 nm. The grana are connected by a fretwork of parallel stromal thylakoids spiralling around them and connecting with the grana margins via narrow junctions in a right-handed helical arrangement (Figure 1).
Organisation of the granal TM revealed by cryo-EM.
While largely confirming the helical model, 3D reconstructions from electron tomography data [21,22] revealed details and irregularities not observable by serial sectioning, e.g. in the granal thylakoid diameters and the angles with the adjoining stroma lamellae. The granum-stroma junctions resemble slits whose length vary considerably, suggesting a possible mechanism for regulating the movement of protons and membrane proteins between the TM domains .
The average repeat distances (RD) between the membrane layers in TM vary between taxonomic groups of organisms and depending on the environmental and experimental conditions in vitro and in vivo. In stacked regions of the grana, the space between adjacent thylakoids, the interthylakoidal space, is narrower than the lumenal space, 3.2 nm and 4.5 nm, respectively ; this has been corroborated by small-angle neutron scattering . In such a crowded space, movement of proteins would be highly restricted without partial or local unstacking of grana .
Although EM remains the most powerful experimental tool to visualise the ultrastructure of the TM, extracting the finest details, there are fundamental drawbacks and limitations mainly associated with the sample preparation and the inherently frozen (literally and figuratively) structure. Recent advances of confocal laser scanning microscopy (CLSM) can distinguish ultrastructural details in the chloroplasts of live cells  providing a way to follow structural changes under different conditions, environments and in time. For example, 3D structured illumination microscopy (3D-SIM) doubles the resolution compared with conventional wide-field CSLM [31,32]. Iwai et al.  demonstrated the capability of 3D-SIM to visualise and measure the diameters of individual grana in the chloroplasts of Arabidopsis leaves. The chloroplast kinase double mutant stn7 stn8 and the phosphatase mutant tap38 have a phenotype with larger and smaller-diameter grana, respectively, compared with the wild type , which could be quantified in situ with great precision from the 3D-SIM images. Wood et al. [35,36] compared chloroplasts from dark- and light-adapted leaves, showing the changes in granal diameter and number of stacked thylakoids in response to the light conditions. (Figure 2)
Subdiffraction-resolution live cell imaging analysis of the grana size of
Arabidopsis thaliana chloroplasts (reproduced from ref. [ 33]).
Composition and heterogeneity
Like other biological membranes, the thylakoids contain glycerolipids organised in ion-impermeable bilayers, enabling the generation of proton-motive force . The lipid composition of the TM is, however, rather different from other cell membranes. In most plants, algae and cyanobacteria, TM predominantly consist of four lipid classes: two neutral galactolipids — monogalactosyldiacylglycerol (MGDG) and digalactosyl diacylglycerol (DGDG), and two anionic lipids — the glycolipid sulfoquinovosyl diacylglycerol (SQDG) and the phospholipid phosphatidylglycerol (PG) [38,39]. MGDG and DGDG are the major constituents — ∼50% and 25%, respectively  and are almost exclusively found in the TM and the inner chloroplast envelope under normal growth conditions. DGDG can replace phospholipids in other plant cell membranes under phosphate deprivation . The TM lipids are also unusual in that they have highly unsaturated acyl chains. Polyunsaturated MGDG is a non-bilayer forming lipid adopting inverted hexagonal (HII) phase on hydration (see [41,42] for reviews). In lipid mixtures resembling the native thylakoid lipid composition, the coexistence and reversible transformation of HII and lamellar phase has been observed depending on the hydration .
The different lipids are asymmetrically distributed in the stroma- and lumen-exposed layers of the membrane . Up to 40% of the thylakoid lipids are bound to proteins [44,45]; lipids are found as structural components of the major photosynthetic protein complexes — PSI [46,47], PSII [4,48] and light-harvesting complex II (LHCII)  — and are involved in the photosynthetic reactions .
The TM are densely packed with proteins — the total integral proteins and pigment–protein complexes can take 70% to 80% of the area in appressed granal membranes [51,52]. The chloroplast genome is relatively small and most of the chloroplast proteins are nuclear-encoded and targeted to the chloroplast via a small N-terminal transit peptide cTP predicted to be present in 4255 proteins in the Arabidopsis thaliana genome . Large-scale proteomic analyses of plant TM identified 400 proteins in total [54–56]. Nearly half are transmembrane proteins, the majority of which constitute the complexes of the photoinduced linear and cyclic ET chains and chlororespiration — PSII, LHCII, PSI, cytochrome (cyt) b6f complex, the ATP synthase, NDH/PGR5, the plastid terminal oxidase (PTOX). The next largest groups consist of ‘maintenance' proteins — chaperones, translocators, proteases and isomerases function in the targeting, maturation, folding and assembly, repair and degradation of proteins — and of oxidative stress proteins — SOD, ascorbate peroxidase, thioredoxins, fibrillins. At least 40 and up to 80 proteins are specific to the lumen [57,58] — the most known of which are the oxygen-evolving complex proteins, plastocyanin and the violaxanthin deepoxidase (VDE) but also, among others, cyt c6A, the lipocalin CHL, and the proteases Deg1, 5 and 8 involved in PSII assembly and degradation .
Perhaps the most characteristic feature of the granal TM is the lateral heterogeneity or sorting of the membrane proteins in different domains — PSII and LHCII are found predominantly in the appressed regions in the granum core, PSI and ATP-ase are located exclusively in the stromal thylakoids and grana margins [19,60], and the cyt b6f complex is evenly distributed throughout the TM [7,61]. The lateral separation of PSII and LHCII, displaying a low level of the protrusion to the stroma, and PSI and the ATP-synthase, with their extensive stroma-exposed structures, optimises the packing density of the membrane system. Sorting of the stroma-side flat and protruding particles, followed by stacking of the flat regions, has been proposed to stabilise the granal membrane ultrastructure [25,62]. How sorting and stacking are carried out during the formation of grana is not fully understood and will be briefly discussed in the following paragraphs.
Supramolecular organisation of PSII
In the presence of cations, attractive molecular interactions between PSII and LHCII lead to formation of supercomplexes, comprised of dimeric PSII core complexes and peripherally attached LHCs (Figure 3) The PSII–LHCII supercomplex is the primary functional unit of PSII in higher plants, binds the majority of chlorophylls (Chls) and is found exclusively in the grana, whereas dimeric and monomeric core complexes with few or no LHCs are abundant in the stroma thylakoids [63,64]. The PSII–LHCII macrodomains are themselves heterogenous in the sense that they include a mixture of PSII–LHCII supercomplexes of different sizes, depending on the physiological conditions [61,65]. In plant TM a dimeric core, denoted C2, accommodates typically two copies of the so-called minor LHCII, (CP29, CP26 and CP24) and two LHCII trimers (trimer S) strongly bound to the complex and two more trimers, moderately bound (trimer M) to form a C2S2M2 supercomplex [61,65,66]. These supercomplexes can bind a third type of L-trimer, to form C2S2M2L1–2 supercomplexes but such complexes are rare .
Structure of the PSII–LHCII supercomplex determined by cryo-EM [
Within the PSII–LHCII supercomplex the innermost LHCII S-trimer is attached to a dimeric PSII via CP29, binding to one PSII core monomer and CP26 to the other. CP29 is essential for the formation of PSII–LHCII supercomplexes .
PSII–LHCII supercomplexes are capable of assembling into higher-order structures in the grana of higher plants such as megacomplexes, consisting of two or more supercomplexes or extended highly ordered semi-crystalline arrays [61,68]. The role of these arrays and even their presence was debated until Kirchhoff showed that their appearance depends on the physiological conditions [69,70] — the semi-crystalline domains are enhanced at low light or low temperatures and can cover the entire grana disc [71,72]. Kouřil et al.  confirmed that the semi-crystalline domain formation depends on the growth light conditions and different packing of PSII–LHCII supercomplexes under different circumstances such as the presence or the absence of minor PSII antenna complexes and the PsbS protein [73,74] was shown.
Biogenesis of TM and formation of grana
The development and organisation of TM involve many factors – environmental conditions (light) and several membrane proteins. The membrane material for the development and maintenance of TM probably originates from the inner plastid envelope . The traffic of material from the inner membrane to the TM network has been thought to be organised by vesicles sometimes observed accumulating in the stroma near the chloroplast envelope . A recent study has systematically uncovered the presence of such vesicles in mature chloroplasts, proplastids, and etioplasts (Chl-less plastids in the leaves of plants grown in darkness) and under a variety of conditions, thus supporting the hypothesis of vesicle-mediated transport .
The prolamellar body found in etioplasts contains mainly the protochlorophyllide oxidoreductase (PORA) and protochlorophilide, which can be rapidly transformed to Chl upon the onset of light . However, various proteins specific to the TM have been identified in the prolamellar body, including photosystem subunits, cytochromes, ATP-synthase, thylakoid-bound proteases and ascorbate peroxidase, which is a good evidence to the significance of the prolamellar body in the TM development . Prothylakoids are the precursors of mature thylakoids but their photochemical function  and protein composition are different. They are devoid of Chl–protein complexes and contain a relatively large amount of cF1F0 ATPase. They have asymmetrical distribution of galactolipids (MGDG and DGDG) and phospholipids similar to TM indicating that Chl and Chl–protein complexes may not be involved in the origin of the asymmetric transmembrane distribution of these lipids . One of the proteins shown to play role in TM biogenesis and maintenance is the intrinsic membrane protein IM30. The exact physiological function of IM30 is not yet clear, however its localisation in the inner chloroplast envelope and TM is a clue and studies of IM30-less mutants [80,81] suggest its role in TM biogenesis. Hennig and co-workers  provided the evidence that IM30 can trigger membrane fusion. More recently, the same group has shown that Mg2+ binds IM30 stabilising its structure and concomitantly induce structural rearrangements in the membrane leading to increased exposure of hydrophobic surfaces .
The stacking of thylakoids to form grana depends on several different forces and interactions and is not completely understood. Thylakoids carry a net negative surface charge and will electrostatically repel one another, hence it has been postulated that stacking is modulated by the distribution of protein complexes with less negative surface charge, on one hand, and on the charge screening by cations, on the other [84,85]. Grana stacks reversibly unstack by incubation of TM in a solution without cations , leading to mixing with the stroma membranes and losing of the lateral heterogeneity of the protein complexes. Upon addition of 5 mM Mg2+ to the suspension, the heterogeneity is restored and the typical organisation of stacked thylakoids connected by stroma lamellae is observed [87–89]. These kinds of experiments have suggested the earlier view that lateral segregation of photosynthetic complexes is a consequence of thylakoid stacking  and that it could be the main reason for the existence of grana [90,91]; however, the reverse causality can also be argued for, namely, that protein sorting is a prerequisite for stacking (see below).
For stacking to occur, the electrostatic repulsion forces must be overcome by attractive counterforces. Chow et al.  raised the hypothesis that grana stacking is driven by an overall increase in the chloroplast entropy because it frees space for free diffusion of soluble components in the stroma. In support of this hypothesis, it was demonstrated that adding water-soluble macromolecules to the medium causes spontaneously restacking of thylakoids .
Since LHCII is the most abundant protein in the granal TM, its role in mediating stacking has been suggested by several authors and supported by an abundance of experimental evidence (reviewed in ). Examining the charge distribution on the stromal surface of LHCII, Standfuss et al. proposed a model of stacking where positive and negative charges on LHCII in opposing membrane sheets attract in a ‘velcro-like' fashion . Nevertheless, the idea that LHCII alone mediates grana stacking may also be too simplistic. Pribil et al.  argue against the role of LHCII, citing the observation of apparently normal granal architecture in Arabidopsis mutants devoid of the major trimer-forming Lhcb1 and Lhcb2 proteins . However, in these LHCII-less mutants, the role of LHCII is taken up by CP26, which is synthesised in large amounts and organised into trimers . As CP26 is also a protein of the LHCII family (Lhcb5), it is possible that it can substitute Lhcb1/2 for the putative stacking function, so the question then boils down to terminology. Markedly large grana that extend almost throughout the chloroplast were seen in PSII-less mutants that still accumulate LHCII, for example, hcf136. Also plant seedlings grown in the presence of lincomycin, which inhibits the synthesis of the chloroplast-encoded photosystem core protein subunits, tend to form very large thylakoid stacks enriched in LHCs . Based on series of CD experiments (see below), Garab and co-workers proposed that it is the propensity of PSII and LHCII to self-assemble into densely packed macrodomains that is the primary driving force for lateral sorting of the complexes, which then facilitates stacking, and not vice versa .
A rotational model of grana stacking  examines the apparent existence of circular aggregates of PSII–LHCII supercomplexes within the membrane plane. The author postulates that transverse interaction of these aggregates generates torque and the adjacent thylakoids rotate forming a helical arrangement. Hence, this model also points out that stacking is a consequence of the sorting of PSI and PSII and not the cause. However, it must be said that experimental evidence for such circular motion is not shown and, moreover, the rotation would cause rupturing of interconnected stromal thylakoids, which has not been observed in any microscopic studies either .
Role of lipids and non-bilayer phases
Various interactions with lipids, specifically MGDG and DGDG, have been proposed to contribute to the granal TM architecture, for example by hydrogen bonds across bilayers [43,99–101]; however, firm experimental data and a consistent picture have not yet emerged. An interesting point is the apparent lipid polymorphism of the TM, mainly driven by the high amounts of the non-bilayer lipid MGDG [42,102]. The coexistence and dynamic exchange between non-bilayer phases and the bilayer was proposed as a mechanism for self-regulation and maintenance of an optimal lipid : protein ratio of the TM . Signatures of non-bilayer phases in TM were later found by 31P-NMR  and the fluorescence of the lipid- phase-sensitive dye merocyanine-540 . More recently, higher-resolution NMR data have identified the coexistence of lamellar and several non-lamellar phases, tentatively assigned to inverted hexagonal HII phase and isotropic phases on the stromal and lumenal side of the TM, respectively . The changes in the relative abundance of these phases in response to lipid saturation level and the physicochemical conditions (pH, temperature), hint at their role in the structural dynamics of the TM.
The grana margins
Earlier grana margins were thought to be essentially free of proteins, and traces of large complexes detected in the grana-enriched fraction of TM were considered to be some contamination from the near-by grana . It has been recently shown that TM curvature at grana margins is controlled by thylakoid curvature protein family (CURT1) which have conserved homologues in cyanobacteria . The CURT1 proteins are small polypeptides with two transmembrane regions, a tentative N-terminal amphipathic helix, capable of forming oligomers and are localised in grana margins. The amount of CURT1 proteins and the number of membrane layers in grana stacks and the area of grana margins are directly correlated. In the absence of CURT1 proteins, grana consist of fewer but broader layers of membrane, whereas overexpression of CURT1A lead to higher stacks of smaller grana discs and tubular stroma lamellae and increased oligomerization of CURT1 proteins. The observed negative correlation between the diameter and height of grana was attributed to the fact that a fixed proportion of TM is incorporated into the grana stacks . Therefore, an increase in the diameter of grana must be compensated for by a decrease in the number of stacks and vice versa. It is worth noting that grana margins are a structural feature of higher-plant TM that form ‘true' grana and are not observed in, e.g. green algae .
Functional roles of grana
Earlier it was postulated that the principal advantage in the granal organisation of the TM is separating the two photosystems to prevent energy spillover, that is, quenching of excitations in PSII by PSI and to allow for controlled distribution of excitation energy between the photosystems via state transitions [109,110]. However, cyanobacteria and many algae containing LHCII-like antenna proteins (e.g. red algae and diatoms, described below) possess no grana without apparent detrimental consequences to the light reactions. It is now accepted that the evolutionary pressure to develop grana was probably not a single critical advantage but a combination of factors [111,112]. One obvious advantage of thylakoid stacking is that it tremendously increases the number of photosynthetic complexes and the absorption cross-section per unit volume [25,61,84,113]. The other principal role of grana is to provide a dynamic and flexible platform that can adjust various subcomponents of the photosynthetic apparatus in response to the environmental conditions — regulating thermal dissipation of excess energy or non-photochemical quenching (NPQ) [114,115], facilitating state transitions [116–118] and the repair and reassembly of damaged PSII [15,119], balancing linear and cyclic ET [60,120] and ATP synthesis [8,111,112]. These regulatory roles require structural and functional plasticity, or flexibility, of the TM as a whole and involve concerted, cooperative or cascade rearrangements of the membrane macroorganisation. The price to pay for the ability to densely pack large numbers of LHCII–PSII units in the granal TM, and especially in ordered semi-crystalline domains, is the restricted movement of molecules through it — from the mobile electron carriers plastoquinone and plastocyanine to the movement of lipids, proteins and protein complexes for the maintenance and repair of the photosynthetic machinery [15,16,121].
TMs in diatoms
The photosynthetic apparatus of diatoms displays peculiar properties with respect to membrane topology, pigment composition, and organization of the protein-bound pigments compared with those of higher plants. Diatoms possess a chloroplast envelope with four membranes, which results from a secondary endosymbiotic event, where a heterotrophic eukaryotic host engulfed a photosynthetic eukaryote, related to red algae . Diatom cells contain either a few small, or one large chloroplast . Like cyanobacteria and red algae, the TM of diatoms do not exhibit grana and are not differentiated into granal and stromal lamellae. The TM span through the whole length of the chloroplast in bands of three  and have been shown to be connected via anastomoses [125,126]. Despite the lack of grana per se, the association of the TM into bands confers some of the properties of stacked membranes (see below).
The light-harvesting antenna systems of diatoms are related to the plant LHCs but use Chl c as accessory Chl besides Chl a and fucoxanthin as the main carotenoid, therefore named fucoxanthin-Chl binding proteins, FCPs . FCPs can be categorised as three major types, the main antenna proteins Lhcf, PSI related Lhcr, and Lhcx, which were proven to play a role in the photoprotection (NPQ) in addition to light harvesting .
The lateral distribution of the pigment–protein complexes in diatoms is not as strict as is in the grana-containing plant chloroplasts. The FCPs are homogenously distributed in all membranes; however, PSI is slightly more abundant in peripheral membrane leaflets . This idea was further elaborated by the domain formation working model, which introduced the concept that PSII and its antenna could be more enriched in the core membranes . A recent study by Flori et al.  supports this unequal distribution. Regarding the lipid composition, the peripheral membrane are more enriched in SQDG  compared with plant TM, that could be the reason for the uneven arrangement of pigment–protein complexes.
Probing the structure and macroorganisation by CD
Theoretical basis of CD spectroscopy
In this section, we will give a brief introduction to the basis of CD spectroscopy of photosynthetic materials, based primarily on the books of Cantor and Schimmel  and van Amerongen et al. .
This scalar triple product is often the dominant term for excitonically coupled molecules. The ± sign denotes that the dimer has two CD bands of the opposite sign but equal rotational strength (spectrally integrated intensity), i.e. a conservative spectrum. Because the excitonic transition energies are split apart by a gap equal to twice the interaction energy, the oppositely signed CD bands do not (completely) cancel each other and the excitonic CD spectrum of many photosynthetic antenna complexes is at least an order of magnitude stronger than the CD arising from the chirality of the photosynthetic pigments (termed intrinsic CD).
The excitonic CD is extremely sensitive to the interaction energy and to the mutual geometry of the coupled chromophores. Thus, it is a useful structural probe for multi-chromophore aggregates such as light-harvesting complexes. However, obtaining direct structural information about real-life photosynthetic complexes is challenging, to say the least. The shape of the CD spectrum depends simultaneously on the geometry of the dimers, the transition energies of the uncoupled chromophores and their fluctuations (bandwidths) and the strength of coupling (the interaction energy), different combinations between these parameters can result in nearly indistinguishable CD spectra, making it difficult to derive the structure from the spectra. Even if the basic structure is known, as is the case for the major plant photosynthetic complexes, because of the delocalised nature of the excitonic states, the spectral bands have contributions from multiple coupled chromophores . Finally, each chromophore contributes with multiple excited states with mixed electronic and vibrational nature — some of which may lie outside the measured wavelength region, tremendously complicating the energy landscape . For instance, the CD spectrum of LHCII (and related protein complexes) in the Chl Qy region is non-conservative , which is attributed to mixing of the Qy transitions with higher-energy electronic levels of both Chls and carotenoids (Car) . Nevertheless, CD and related spectroscopy techniques like circularly polarised luminescence are highly valuable when used e.g. in combination with theoretical modelling of the excitonic states of photosynthetic complexes [138,142–144].
ACD allows for independent measurement of the excitonic CD bands (Figure 4b), which brings valuable information, otherwise lost by rotational averaging, such as the exact peak positions and hence the excitonic split (rotational averaging). More importantly, ACD ‘labels' the excitonic CD bands with orientations of the underlying transitions, which dramatically reduces the possible geometries that can fit the spectra and the degrees of freedom in model fitting. A simple and clear example of using ACD to separately probe exciton transitions is the CsmA baseplate protein found in the chlorosome antenna of green-sulfur photosynthetic bacteria, where bacteriochlorophyll a effectively forms exciton dimers . The ACD spectra of the CsmA protein were theoretically calculated following the molecular exciton approach [148,150]. Based on these data the NMR structure of the protein could be validated . In the ACD calculations, an earlier theory by Hansen  was used that includes the full spatial variation of the electromagnetic field and is referred to as fully retarded description. Recently, Lindorfer and Renger  have presented simplified formalism to calculate the ACD spectra of chlorosome baseplate that can be applied to describe the ACD of molecular aggregates.
The UV CD of peptides and proteins is widely used to study their structure, folding, and conformational dynamics; membrane insertion; and effects of ligand binding [135,136,154]. The far-UV CD spectra of α-helices, aromatic peptides and β-sheets has specific characteristics below 240 nm that have made CD spectroscopy a standard tool for monitoring the secondary structures of proteins. Below 240 nm the spectra are dominated by n–π* and π–π* transitions of the peptide backbone [155,156]. The n−π* transition gives rise to a negative peak in the CD spectrum at ∼224 nm. According to Moffitt's theory  excitonic interactions in α-helices result in three π−π* transitions with split energies: one gives rise to the negative peak at ∼210 nm and the other two are degenerate, at ∼190 nm, with amplitudes strongly depending on the probing direction. The ACD of oriented polypeptides can provide additional information regarding the angle of orientation of the secondary structures [158–161]. Helical peptides inserted into lipid bilayers exhibit a specific ACD spectrum. As the 210 nm transition is polarised parallel to the helix principal axis, a distinguishing feature of the ACD of α-helices is the presence or absence of 210 nm negative band, being indicative of surface or transmembrane helix alignment, respectively . For polytopic transmembrane proteins, the average inclination angle of the transmembrane helices can be extrapolated.
Long-range chiral order and membrane macroorganisation
The dominant CD of pigment–protein complexes in the visible wavelength region is induced by short-range (up to a few nm) excitonic couplings between the chromophores as outlined above. The same holds for suspensions of small aggregates, in the absence of any sizeable long-range interactions between the chromophores. The situation is different in large 3D assemblies, which contain densely packed and strongly interacting chromophores, assembled in large arrays with long-range chiral order and sizes commensurate with the wavelength of the visible light. Various biological assemblies of such kind — DNA condensates, viruses, and especially TM — can exhibit intense CD bands, far exceeding in magnitude both the intrinsic CD of the chromophores and the short-range excitonic CD of the pigment–protein complexes [162,163]. As with excitonic CD, this type of induced CD, termed ‘giant' or ‘psi-type' (polymer- or salt-induced), is an emergent property of the supramolecular organisation of the assembly and cannot be represented by merely the sum CD of its constituent parts.
Intact leaves, chloroplasts and isolated stacked TM contain all three types of CD (intrinsic excitonic and psi-type) superimposed on each other, with characteristic psi-type CD bands in the Chl Qy and in the blue-green wavelength region (Figure 5). The psi-type CD is attributed to the long-range dipole-dipole interactions of Chls and Car in the PSII and LHCII embedded in the membrane in ordered three-dimensional arrays called chiral macrodomains [163,164]. Although the psi-type CD contains some contribution from circular intensity differential scattering (CIDS), especially giving rise to characteristic long tails flanking the absorption bands [165,166], it is not merely an optical artefact but an intrinsic spectroscopic signature uniquely depending on the three-dimensional macroorganisation and therefore carries meaningful physical information . The high degree of similarity of the CD of intact leaves and chloroplasts isolated from the same leaves testify that the psi-type CD is also not a side effect of TM isolation . Conversely, it also means that CD spectroscopy can be used to probe the macroorganisation in vivo, and in cases where isolation of TM is not practical.
Psi-type CD spectra of pea chloroplasts (
Pisum sativum) and of lamellar macroaggregates of LHCII isolated from pea as described in ref. [ 168].
The psi-type CD of TM is tightly correlated with the presence of grana [169,170]; it is diminished in Chl-b-less mutants and is abolished in chloroplasts of wild-type plants upon unstacking by removing the ions from the medium . On the other hand, psi-type CD is observed also in some algae, which do not have typical granal organisation, like diatoms . Signals resembling the psi-type CD of TM can be observed in simpler in vitro assemblies, for example Pearlstein  reported giant CD from chlorophyllide a — myoglobin aggregates. Psi-type CD can also be observed in some cases from isolated LHCII aggregates, termed lamellar macroaggregates [168,173]. In these samples, the presence of long-range ordered LHCII arrays can always be seen, for example by their ability to orient in external magnetic fields . Also electron micrographs show that they are composed of stacked lamellae with lateral order — hexagonal arrangement of LHCII  not dissimilar to stacked 2D membrane crystals [94,176]. In contrast, small, disordered aggregates of LHCII do not display psi-type CD. The psi-type CD intensity in lamellar macroaggregates is proportional to their size and more specifically to the size of the ordered domain . Thus, psi-type CD reflects the extent of the long-range chiral order of the chromophores, following the theoretical predictions . Interestingly, the CD spectrum is distinct from that of native TM with the main bands in the Qy region having opposite signs. It could be speculated that this is related to the different lateral arrangement of LHCII complexes in crystalline aggregates — with alternative up-down orientation [94,176].
The psi-type CD of semi-crystalline LHCII aggregates (Figure 5) may suggest a possible connection between the ordered semi-crystalline domains of PSII–LHCII [61,177] and the psi-type CD of TM; however, this has not been unequivocally established. While ordered arrays are not always observed, the psi-type CD is persistent for granal TM. Conversely, negatively stained electron micrographs of ordered arrays are obtained from TM mildly treated with detergent, but even low detergent concentrations destroy the psi-type CD [73,178]. On the other hand, if the membranes were cross-linked with glutaraldehyde before the treatment, the CD was preserved , showing that the CD changes were indeed reflecting the detergent-induced disassembly of the macrodomains.
The main spectral features of the psi-type CD of thylakoids are a positive band at ∼686–690 nm, a negative band with a maximum at 672–676 nm and a broad positive band peaking at 505–510 nm (Figure 5). The remaining peaks in the CD spectrum originate from short-range excitonic interactions. Garab et al.  established that these bands originate from distinct structural entities in the TM as they responded differently to various treatments. The intensity of the 672–676 nm band decreased faster upon osmotic or mechanical shock and was associated with thylakoid stacking. It was also found to be more sensitive to high temperature and strong light compared with the positive band at 686–690 nm . The opposite response was observed when resuspending TM in a hyperosmotic medium (containing 2 M sucrose) — the 506 nm band completely disappeared, the intensity of the positive band at 690 nm was reduced by half, while the negative band at 672 nm was unchanged .
The tight connection between the long-range macroorganisation of the TM and the psi-type CD was neatly demonstrated by Kovács et al. , comparing wild-type Arabidopsis plants and antisense or knock-out mutants of Lhcb6 (CP24) — one of the monomeric LHCII complexes. As CP24 is a minor antenna subunit of the PSII supercomplex, its absence does not alter the Chl, Car or protein composition of the TM to a significant extent but the assembly of the PSII–LHCII supercomplexes was drastically impaired and only C2S2 supercomplexes were accumulated. Consequently, these smaller supercomplexes formed macrodomains in a densely packed arrangement (Figure 6), distinct from that of wild type plants, containing predominantly C2S2M2 type supercomplexes (Figure 3). Remarkably, the main positive psi-type band at 690 nm decreased proportionally with the Lhcb6 content and completely vanished in the knock-out mutant, whereas the 505-nm band was unaffected or even increased in intensity. From these results, the authors proposed that the 690-nm CD band does not depend on the granal stacking per se but on the lateral arrangement of the PSII–LHCII supercomplexes in the granal TM.
EM and CD spectroscopy of TM from wild type and CP24-deficient Arabidopsis plants.
Tóth et al.  systematically compared the CD spectra of various plant species and mutants in effort to obtain more information on the structures and interactions responsible for the psi-type bands. They found that the 690 nm band varies among species, as do the ratios of the different bands, corroborating their association with distinct structural moieties. No obvious relationship between the granal ultrastructure and the CD was noticed. On the other hand, the 506 nm and 690 nm bands depend on the amount of LHCII in the TM, in agreement with earlier observations during greening of bean leaves .
The CD spectra of LHCII-enriched leaves, i.e. from plants grown in the presence of lincomycin, which blocks the synthesis of core proteins, and several mutants with altered Car composition (aba4, npq1, npq2 lut2) were compared to determine the role of different pigment–protein complexes in the formation of the chiral macrodomains that give rise to the psi-type CD bands. In LHCII-enriched leaves, the amplitude of the (+) 690 nm band was about half of that in control plants and the (+) 506 nm band was not visible. The 506 nm band is also absent from the Chl b-less mutant ch1 (and in the analogous barley chlorina-f2 mutant) and in lamellar aggregates of isolated LHCII . On the other hand, all psi-type CD bands, including the 506 nm band, were observed in mutants with altered xanthophyll composition (aba4, npq1, npq2), except for the lut2 mutant, where the intensity of both 506 nm as well as the 690 nm band was reduced. From all these, it becomes clear that the 506-nm band originates from the long-range chiral order of β-carotene bound to the PSII core complexes in the PSII–LHCII domains of the grana. Another clue that the psi-type CD bands originate from PSII–LHCII supercomplexes is the reduced intensity of the CD bands in PsbW-deficient plants, as the absence of the PsbW protein subunit destabilises the PSII–LHCII supercomplexes .
The psi-type CD, as a sensitive and accessible probe for the macroorganisation and flexibility of the TM, has been used to assess the response to physiological conditions and abiotic stress — more recent examples include the effects of temperature and osmolarity [180,182], thermotolerance-inducing effects of isoprenes  or selenium toxicity .
Structure and flexibility of LHCII
The molecular and excitonic structure of LHCII
The trimeric LHCII consisting of the Lhcb1–3 protein isoforms, is the major light-harvesting antenna complex in plants , comprising the S and M trimers in the PSII–LHCII in Figure 3. Prior to the cryo-electronic microscopy structures of the PSII supercomplex [4,186,187], the X-ray crystallographic structure of the trimer was resolved independently by the groups of W. Kühlbrandt [94,188] and W. Chang [49,189]. Each monomeric subunit noncovalently binds eight Chl a, six Chl b and four xanthophylls. The transmembrane part of the protein complex has high structural rigidity [94,190] ensuring strict geometry and electrostatic environment of the pigments, which in turn determine the excited-state properties of the complex. The Chls are arranged in two layers — eight Chls near the N-terminus on the stromal side of the TM and six on the lumenal side (Figure 7). The Chls are closely spaced together and there are several pigment groups with mutual distances of 9–12 Å — Chls a 602/603, a 610/611/612, b 601′/608/609 on the stromal side and Chls a 613/614, b606/607 on the lumenal side. Excitonic couplings between these closely spaced Chls create at least partially delocalised states and should give rise to CD. In addition, the Cars, especially the two central luteins (Lut) and neoxanthin (Neo) strongly interact with various Chls.
Arrangement of Chls in the LHCII monomer — the positions of the Mg and N atoms of the 14 Chls in the LHCII monomer are overlaid on a schematic drawing of the apoprotein.
Several groups have used the published structures of LHCII to calculate the Chl exciton states using quantum-chemical theoretical methods and fitting to spectroscopy data [142,191–193]. The general approach is to construct an exciton Hamiltonian with diagonal elements corresponding to the unperturbed transition energies (site energies) of all Chls, accounting for any shifts caused by the environment (protein, lipids, water, etc.), and off-diagonal elements corresponding to the Coulomb interaction energies (excitonic couplings) between Chls. The energy eigenstates (exciton states) of the complex are obtained by diagonalising the Hamiltonian and can be used to calculate optical spectra, including CD. All published models are in good agreement with regard to the interaction energies . The strongest couplings, between Chls a611 and a612, have magnitude ∼100 cm−1. However, because of uncertainties in the calculated (or fitted) site energies, it is difficult to assign the dominant contributions of specific Chls to the exciton eigenstates, respectively absorption/CD bands. The model of Zucchelli et al.  assigns the lowest eigenstate to Chl a612, in agreement with earlier mutagenesis and spectroscopy data , whereas other models point to a610 [192,195] and the possibility of a temperature-dependent swap has been put forward [192,196]. The identities of the upper energy levels are also debatable. This poses difficulties in modelling and interpreting the spectroscopic data, including the CD spectra. Another challenge is to include vibronic and higher-energy transitions in the Hamiltonian for a realistic depiction of the optical spectra .
Finally, it is worth noting that exciton calculations use fixed atom co-ordinates from the crystallographic structure with the implicit assumption that the structure is maintained under the various physiological and experimental conditions of interest. Structural and, consequently, the energetic disorder is often included in simulations as random fluctuations in the site energies . The dramatic effect of static and dynamic disorder (basically slow and fast conformational fluctuations) on the excitonic landscape and spectroscopic properties of LHCII have been shown by a variety of methods, including single-molecule spectroscopy , molecular dynamics simulations  and inelastic neutron scattering . Hybrid approaches combining quantum mechanical and all-atom molecular dynamics calculations [200,201]; may eventually provide a deeper theoretical insight of the photophysical functions of the complexes in different physicochemical environments.
Flexibility of LHCII — excitonic signatures of aggregation
Being a major component of the grana, LHCII is also a key player in several of the regulatory functions that granal membranes perform, especially balancing the excitation energy flow through state transitions and NPQ. These processes are understood to involve some controlled changes in the conformation and macroorganisation of LHCII [114,115]. The sensitivity of CD to both conformation and macroorganisation has inspired extensive studies of the structural and functional flexibility of LHCII using CD spectroscopy. These revealed a surprising variability under different conditions that emphasise the fragile nature of the excitonic states and prompt the question of what the native, physiologically relevant state really is.
The CD spectrum of LHCII solubilised in mild detergents such as dodecyl maltoside (DDM) has a characteristic non-conservative negative–positive–negative shape in the Chl Qy region [202,203] and prominent negative bands in the Soret region (c.f. Figure 8a, LHCIIDDM). The CD spectra easily distinguish between different oligomeric states of the complex. Because LHCII trimers are stabilised by PG molecules, monomers can be obtained by treating trimeric LHCII with phospholipase a2  or by refolding recombinant LHCII in the absence of PG . The CD spectra of trimers and monomers show specific differences in the Soret and Qy regions — notably the negative band at 473 nm is lost, as well as a negative band (or shoulder) at 648 nm, and a small positive feature at 412 nm [203,204]. These signatures have been exploited to detect LHCII trimer dissociation, e.g. under light treatment [205,206] or in artificial membranes .
CD spectra of LHCII and thylakoid membranes.
As any membrane protein, LHCII is not water-soluble and in the absence of surfactants readily aggregates in aqueous solutions. The CD spectra of aggregates are markedly different from the detergent-solubilised trimers (Figure 8a) — with characteristic deviations in the Car and Chl regions . This is a strong evidence of changes in different exciton states having energies throughout the visible wavelength region including in the lowest energy states. Comparison of aggregated and solubilised LHCII has revealed other tell-tale signs of structural changes — Resonance Raman spectroscopy indicates conformational changes of Car as well as Chls [209–211]. If aggregated and solubilised LHCII have measurable structural differences, it is logical to ask which of these, if any, corresponds to the native physiologically relevant state. Comparing the CD spectra of TM and isolated LHCII, although admittedly crude, is warranted by the fact that the CD of TM and especially of isolated grana or BBY particles is dominated by LHCII . Interestingly, the CD spectra of thylakoids washed in hypotonic buffer, to eliminate the psi-type CD (Figure 8b), were more similar to the spectra of LHCII aggregates than to solubilised LHCII [208,212]. Even closer similarity can be found between lamellar LHCII aggregates and native LHCII-enriched TMs, isolated from plants treated with lincomycin . Solubilisation of TM and of LHCII-enriched TM with detergent also induced virtually the same changes in the CD spectra as solubilisation of LHCII aggregates (compare the yellow difference spectra in Figure 8).
It could be concluded that the CD spectra reflect excitonic couplings resulting from the supramolecular organisation of LHCII, which are retained in the lamellar aggregates as in the native TM. On the other hand, solubilisation of the complexes alters the excitonic landscape presumably by destroying specific protein–protein interactions. Another possibility is that the detergent environment alters the structure or exciton interactions as compared with the native lipid membrane. The fact that the LHCII CD spectra differ between detergents or even enantiomers of the same molecule (α- or β-maltoside) is a testimony for the latter.
To determine whether ‘protein interactions' or ‘detergent interactions’ are the main reason for the observed spectral differences between aggregated and solubilised LHCII, a gel dialysis system was applied first by Ilioaia et al.  and later by Akhtar et al. . Solubilised LHCII trimers are trapped in a polymer gel, which is then dialysed to extract the detergent without aggregation. The specific effect of aggregation, i.e. of the protein interactions in aggregates, is seen by comparing the LHCII spectra in dialysed gel and in detergent-free solution. The difference CD spectrum shows changes at 437, 484 and 681 nm (Figure 8). In contrast, after re-addition of β-DDM to the dialysed gel, the negative CD band at 492 nm is enhanced. Thus, aggregation and detergent removal have distinct spectral signatures, indicating that different pigment groups in the complex are affected. In addition, the spectra of solubilised LHCII vary with detergents. These variations can be explained with conformational changes in the complex upon interaction with the detergent molecules. Because the 492-nm CD band is diminished in Neo-less LHCII [215,216] and TM from Neo-deficient plants , the difference at 492 nm probably results from a change in the conformation of the LHCII-bound Neo, which is also evident from resonance Raman spectra . On the other hand, the negative band at 437 nm induced by aggregation is not affected by the presence of Neo; it could reflect new excitonic interactions formed between neighbouring LHCII trimers.
Conformational changes and excited-state dynamics in LHCII aggregates
The absorption and fluorescence maxima of LHCII aggregates are red-shifted by 1–3 nm, which indicates that the transition energies of certain Chl a molecules, respectively exciton states, are lower  — presumably because of a conformational change. Aggregates of LHCII exhibit strong fluorescence quenching and a proportional decrease in the Chl excited-state lifetime [219–222]. Aggregation-induced quenching is also observed in other LHCs, for example FCP from diatoms  and is thought to be involved in the photoprotective NPQ in vivo (see section 4.3 below). Although the exact mechanism of quenching in aggregates is still debated, there is ample evidence that it involves conformational changes in the pigment–protein complex .
The low-temperature fluorescence spectra of quenched LHCII aggregates exhibit a characteristic band in the far-red region, at 700 nm and above [219,224]. The far-red emission was detected also at room temperature by time-resolved fluorescence and attributed to a Chl–Chl charge transfer (CT) state formed upon aggregation, mixed with an exciton state . The formation of Chl CT states was also confirmed by Stark spectroscopy [226,227] and single molecule spectroscopy showed that they can intermittently be formed in solubilised LHCII as well but their prevalence is higher under low pH and detergent concentrations [228,229]. Analysing the quantum beatings in two-dimensional electronic spectra of LHCII, Ramanan et al.  detected coherent mixing between a low-energy (700 nm) CT state and an exciton state and assigned them to the Chl pairs a603/b609 and a602/a603, respectively.
While the presence of far-red-emitting CT states in LHCII is now established, their relationship with fluorescence quenching is debated. Holzwarth and co-workers postulated that the far-red emission of LHCII aggregates in vitro and of intact leaves under NPQ conditions signify the same mechanism of energy dissipation involving Chl CT states [225,231,232]. Enhanced far-red emission was associated with NPQ in plants  and desiccated lichens . Control of the CT character by changes in the pigment local environment is thought of as an effective mechanism to regulate energy dissipation . Valkunas and co-workers, however, argue that the Chl–Chl CT states in LHCII aggregates are not an intrinsic part of the quenching mechanism [236,237]. Also, the far-red signature observed in leaves under NPQ could be attributed to the PSII reaction centre (RC) rather than the antenna .
Aggregation affects not only the overall Chl excitation lifetime but also the pathways and dynamics of energy transfer. Rutkauskas et al.  determined that the singlet–singlet annihilation rate in aggregated LHCII is higher than in the solubilised state and attributed this change to a faster inter-monomer energy transfer. Two-dimensional spectroscopy also revealed faster kinetics of energy transfer between Chl b and a through intermediate-energy excitons .
From these studies, it becomes clear that the molecular environment has a significant impact on the excitonic interactions and consequently the excited-state dynamics of LHCII. The activation of fluorescence quenching is not strictly dependent on aggregation of protein–protein contacts  — a change in the local physicochemical environment is sufficient. It may be reasonable to expect finding a specific change in the excitonic CD spectra that is correlated with fluorescence quenching. However, a simple comparison of the CD spectra of LHCII in quenched and unquenched states does not reveal any such candidate. The aggregation signatures at 437 and 484 nm can be ruled out because they are absent from ‘quenched' gels and present in ‘unquenched' LHCII-enriched membranes. The same applies to the Neo-specific band at 494 nm — the CD amplitude is reduced if β-DDM is replaced by α-DDM but the fluorescence lifetime is unchanged. Further studies with more precise and sensitive approaches may help to identify the formation of the elusive quencher.
Macroorganisation and fluorescence quenching of LHCII in reconstituted membranes
Analysis of the CD and fluorescence spectroscopy shows that neither aggregates nor solubilised LHCII fully retain the excitonic structure found in the native TM. Reconstituted lipid–protein membranes, or proteoliposomes, are commonly used to study LHCII in a native-like lipid bilayer environment [241–245]. The CD spectra of such artificial LHCII membranes, however, [207,213,246–248] more closely resemble that of detergent-solubilised LHCII than the CD of aggregates or native TM. Interestingly, the Neo-associated negative CD band at 492 nm is just as characteristic for LHCII in reconstituted lipid membranes as in β-DDM micelles (Figure 9a), while the same band is not observed in native TM. Conversely, the aggregation-specific band at 437 nm is not as pronounced in reconstituted membranes as in the native membrane. It appears that LHCII in lipid bilayers also displays a high degree of variability of the Chl and Car excitonic interactions. This cannot be attributed to the lipid composition but to the structural organisation of the membranes.
Dependence of the CD spectra and fluorescence lifetimes on the lipid : protein ratio in reconstituted LHCII membranes.
One obvious origin of the variability among the native and reconstituted membranes is the protein density, or lipid : protein (L/P) ratio. Reconstituted membranes contain considerably more lipids per protein (typically 100 : 1 to 1000 : 1 in our experiments) than granal TM, which are composed of 70–80% protein or ∼1.5 lipids per Chl [51,250]. This may well explain the ‘aggregation' CD signature of native membranes.
The fluorescence yield of LHCII liposomes is lower than solubilised LHCII trimers  and the induced quenching depends on the lipid composition . However, single LHCII trimers embedded in nanodiscs show the same fluorescence lifetimes (3–4 ns) as detergent-solubilised complexes , indicating that protein–protein interactions, and not the lipid environment, are responsible for quenching in proteoliposomes [207,245,248]. The fluorescence lifetimes of LHCII proteoliposomes are in the range of 1–1.7 ns for L/P ratios of 100 : 1 to 400 : 1 [207,213] — considerably shorter than the 4-ns lifetime in detergent micelles. On the other hand, it is not certain that the fluorescence lifetime of detergent-solubilised LHCII represents the native state. Belgio et al.  concluded that in native TM the excitation lifetime of LHCII (if not connected to the RC) is ∼2 ns, based on measurements on LHCII-enriched membranes from lincomycin-treated Arabidopsis plants. In LHCII-enriched membranes from pea, average lifetimes in the range of 1–1.5 ns have been observed . We must stress that although we refer to the TM from lincomycin-treated plants as ‘native', they have composition and macroorganisation radically different from the physiologically active TM. The same effects — clustering of LHCII in densely packed macrodomains — can be responsible for a faster nonradiative excitation decay in native LHCII-enriched membranes as in reconstituted membranes and aggregates. One common feature of these samples is the large functional domain size, i.e. the ability of excitations to hope between LHCII units, which makes them accessible to quenchers [213,252]. Very short fluorescence lifetimes (<100 ps) can also be observed in PSII-enriched membrane fragments (BBY), depending on their lipid content (P. Lambrev, Y. Miloslavina, A. Holzwarth, H. Kirchhoff, unpublished data). For these reasons, it is very difficult to make conclusions about the ‘native' excitation lifetime of LHCII from these types of samples.
At a first glance, it appears trivial to control the L/P ratio of reconstituted membranes and test its influence. However, reconstituted membranes are highly heterogeneous with respect to vesicle sizes and L/P. To determine the distribution of these parameters, LHCII membranes were first reconstituted with fluorescent lipid dyes, separated into fractions by discontinuous density gradient ultracentrifugation and then analysed by single-vesicle fluorescence microscopy . The results showed that even after separating the reconstituted mixture into several fractions of uniform density (buoyancy), they still displayed a wide distribution of vesicle sizes and L/P ratios and even the presence of non-reconstituted protein and empty vesicles. We could expect that because of this apparent structural heterogeneity, there exist subpopulations of reconstituted membranes with considerably different excitonic landscapes. By quantifying the lipid and protein content of individual vesicles, it is possible to correlate the number of LHCII per liposome or the surface density concentration (number of LHCII per unit area) with the vesicle size. Surprisingly, there was a striking dependence of the surface density concentration on vesicle size — smaller liposomes had higher protein densities and vice versa. In other words, LHCII was preferably assembled in protein-dense macrodomains in smaller vesicles and no vesicles were found to contain single LHCII trimers, contrary to some earlier reports .
The combination of lipid dyes for spectrophotometric quantification of the lipid content and the separation of proteoliposomes by density provided a platform to determine the relationship between Chl fluorescence and the L/P ratio in absolute terms over a wide range of L/P densities . The results confirmed the strong relationship between protein density and fluorescence quenching — as the L/P was lowered, the average fluorescence lifetime progressively shortened, spanning the entire range of values between solubilised LHCII and strongly quenched aggregates. A strong correlation was also observed between the L/P ratio, the fluorescence lifetime (Figure 9b) and the appearance of far-red fluorescence emission at 700 nm, confirming the experiments of Natali et al. . In contrast, no correlation was found between the fluorescence quenching and the aggregation-specific CD bands at 437 at 484 nm and even for strongly quenched proteoliposomes, the CD spectra were markedly different from those of lipid-free LHCII aggregates, corroborating the previous finding that these excitonic signatures are not connected to the generation of quenchers.
From the studies of LHCII in different physicochemical environments, a general scheme emerges, according to which LHCII plays a role in grana formation not only by electrostatic interactions between adjacent layers but mainly by lateral hydrophobic interactions in the membrane plane, which give it an affinity to assemble into dense clusters or macrodomains. It can be envisioned that this will facilitate sorting and segregation of LHCII–PSII in turn allowing membrane layers to come into close range before stacking is stabilised by electrostatic interactions or hydrogen bonds. The intrinsic capacity of LHCII for NPQ simply by way of clustering or aggregating in the membrane without the absolute requirement for external quenchers or activators (zeaxanthin, PsbS) is also of note and is in line with a general allosteric model of NPQ activation , according to which LHCII is intrinsically predisposed to quenching and the role of external activators is to shift the thermodynamic balance toward the quenched state. Studies with reconstituted membranes suggest that left unchecked LHCII will spontaneously switch into energy-dissipating state. Uncontrolled quenching evidently does not happen in vivo and the reasons for this discrepancy are not yet completely understood.
ACD of TM and LHCII
The capability of ACD spectroscopy to separate excitonic transitions based on the orientation of the transition dipole moments can prove useful for disentangling the spectral signatures of photosynthetic complexes and membranes. Following up on earlier studies of oriented chloroplasts , Miloslavina et al.  performed ACD measurements on stacked and unstacked TM, lamellar LHCII aggregates and PSII-enriched membranes and showed that both the psi-type CD and the excitonic CD is dramatically affected by the sample orientation. Stacked TM and lamellar LHCII aggregates can be aligned in a moderate magnetic field (0.5–1 T) due to their large diamagnetic susceptibility anisotropy, related to the long-range order of the protein complexes in the chiral macrodomains . In both lamellar aggregates as well as stacked TM, the ACD spectra showed a sign inversion of the main psi-type CD band around 680 nm, that could at least partially be due to the angular dependence of CIDS. The experimental isotropic CD could be approximately recovered by the linear combination of the face- and edge-aligned ACD spectra according equation 11, showing that the measured ACD is not affected by polarization cross-talk artefacts.
The excitonic ACD spectra of washed TM and PSII-enriched membranes oriented by polyacrylamide gel compression also showed specific well reproducible differences with the isotropic CD, especially in the Soret region, where two characteristic positive bands, at 447 nm and 484 nm, were observed for both sample types. Following the reasoning in section 2.2, bands with enhanced intensity in the face-aligned spectra originate from transitions oriented preferentially in the membrane plane. In general, the ACD spectra showed several intense narrow bands resolved more clearly than in the isotropic spectra, demonstrating the potential of ACD to identify specific pigments and pigment interactions.
Isolated LHCII is difficult to align macroscopically but can be oriented if embedded into lipid membranes. Reconstituted LHCII membranes were oriented by gel compression and as dehydrated films on fused silica plates and ACD spectra were measured in the face-aligned position. The ACD spectra were strikingly different in comparison with the isotropic CD spectra, featuring a larger number of well-resolved bands (Figure 10). The spectra were independent from the method of orientation and qualitatively similar spectra were observed from oriented lamellar LHCII aggregates that did not possess psi-type CD. Additional controls confirmed that the ACD spectra are free from artefacts and present useful information about the molecular structure of the complex.
CD and ACD spectra of LHCII in oriented lipid bilayers.
The most prominent features in the ACD spectra, e.g. the positive bands at 445 and 482 nm, are present in the isotropic spectra as well but better resolved and much more intense in ACD. Other bands, especially in the wavelength region 500–650 nm, are only resolved in ACD, whereas the isotropic CD is flat and featureless. These bands seem to originate from an excitonic coupling between optically weak transitions Chl and Car transitions of intermediate energy between the Chl B and Q band, that have not been documented before. Especially striking is the sign inversion in the red-most part of the spectrum, i.e. the region of the lowest-energy Chl a sinks — the negative peak at 682 nm in the CD spectrum becomes a positive peak at 679 nm in the face-aligned ACD.
The face- and edge-aligned ACD spectra essentially sort the excitonic transitions in LHCII separating those that are oriented preferentially in the membrane plane from those that are preferentially perpendicular to the membrane plane. For a deeper analysis and assignment of the CD bands, the basic theoretical formalism, outlined in section 3.5 above and in more detail , can be applied in combination with quantum mechanical modelling. Using this theory and the exciton Hamiltonian of Müh et al. , the ACD spectra of LHCII in the Chl Qy region were simulated. The characteristic features of the ACD spectra, including the sign inversion of the long-wavelength band, were reproduced by the calculation without changing a single model parameter, further corroborating the exciton model. Particularly, the agreement between the simulations and experimental data in the red spectral edge give confidence in the assignment of the lowest-energy states — Chl a610 and Chl a603 on the stromal side and Chl a604 and a613 on the lumenal side. Removal of these Chls or using a different set of site energies resulted in worse fit. On the other hand, the deviations from the experimental spectra show that there is room for refinement of the model.
UV-ACD has been applied to study the membrane insertion of various small polypeptides, such as toxins and antibiotics  but its ability to monitor the structure and macroorganisation of photosynthetic proteins has not been explored until now . The UV-CD spectrum of TM shows (Figure 11) the characteristic shape for α-helical polypeptides: it has two negative bands, at 210 nm and 224 nm, crossing the zero at ∼201 nm, and a strong positive band near 193 nm. This is not surprising considering that TM contain primarily intrinsic membrane proteins containing multiple transmembrane α-helices (Figure 3). The ACD spectra of oriented membranes showed an enhanced positive band at 193 nm and a negative band at 224 nm, whereas the 210 nm band was largely missing, entirely consistent with the prevailing transmembrane orientation of the protein α-helices. Very similar spectra were recorded from PSII-enriched granal membranes. These results show that ACD, in general, can probe the embedding and orientation of photosynthetic complexes in the membrane. However, the ACD spectrum of isolated LHCII deviated from the predicted spectrum calculated by applying Moffitt's theory on the protein backbone of LHCII . The differences suggest that the UV-CD spectra of LHCII and of pigment–protein complexes, in general, reflect not only the protein secondary structure but might also include contributions from the pigment cofactors and pigment–protein couplings — an area that evidently needs further exploration.
UV SRCD and ACD spectra of thylakoid membranes (a) and reconstituted LHCII membranes (b).
Dynamic flexibility of TMs
Light-induced changes in macroorganisation
Granal TM are highly flexible adjusting their structure, composition and supramolecular organisation at different hierarchical levels — on microscopic as well as mesoscopic scale — and on timescales spanning seconds to days. Long-term acclimation to the physiological environment typically involves changes in the lipid, protein and pigment composition and chloroplast ultrastructure. For example, sun and shade leaves differ in their Chl a/b ratios, Car content, shade leaves accumulate more LHCs per RC, have larger grana diameter and number of layers per granum and decreased number of grana per chloroplast  to maximise the efficiency of light use. Long-term acclimation brings about the reorganisation of the PSII–LHCII supercomplexes, modulating the antenna size and the content of specific Lhcb proteins . Here we will briefly summarise some of the short-term changes in the membrane macroorganisation in response to environmental factors, particularly to light intensity and colour, while the topic is more comprehensively covered in recent reviews [15,16,29].
Exposure to high light (HL) reduces the grana diameter along with increased mobility of grana-hosted protein complexes  and partial unstacking of grana discs in the leaves of Arabidopsis  and isolated TM of spinach . The reduction in grana would be compensated with an increased number of grana per chloroplast under HL, to create a larger contact area between grana and stromal lamellae. The thylakoid lumen increases its size in the presence of light and shrinks in the dark . This swelling and shrinkage is controlled by regulated ion fluxes across TM through ion channels (for review see ref. ). Influx of Cl− ions is one of the main contributors to lumen expansion. The Cl− influx is driven by light-induced generation of transmembrane proton-motive force . Beside chloride ion transport, the ΔpH-dependent accumulation of Ca2+ in the lumen has been documented and is thought to be mediated by a Ca2+/H+ antiporter . The increased ion concentration and respectively, increased osmotic potential of the lumen would lead to more swelling. Mg2+ efflux partially compensates for the increased osmotic potential created by Cl− and Ca2+ influx . The reduction in grana diameter, modification of stacking and lumen swelling under HL conditions are thought to facilitate PSII repair and diffusion of mobile electron carriers plastoquinone and plastocyanin enhancing the rate of ET between the two photosystems . FRAP (fluorescence recovery after photobleaching) analysis of digitonin-solubilised grana membranes revealed significantly increased mobility of proteins after HL treatment, however, after 30 min of HL protein mobility level off.
Small-angle neutron scattering (SANS) measurements — which give information about the periodic arrangement of TM further elaborate the capability of TM to undergo structural changes affecting their lamellar order and RD [265,266]. SANS measurements on isolated spinach TM showed fast reversible light-induced changes in the RD on timescales of seconds. The light-induced changes were inhibited in the presence of proton uncouplers and enhanced by cyclic ET around PSI, confirming that they are driven by the photoinduced proton and ET. Rapid small but clear light-induced changes in the membrane organization and RD were observed in cyanobacteria [265,267] green algae , diatom cells [265,266,269] as well as intact leaves of various species  — for a more extensive treatment see refs. [28,62]. It must be noted that while the RD in cyanobacteria and diatoms increased upon illumination (i.e. swelling), the data showed shrinkage or reduced RD in plant TM — the source of this discrepancy is not clear and highlights the complex nature of the membrane dynamics and the insufficient current level of understanding.
Dynamic reversible changes in the membrane macroorganisation of isolated TM upon illumination can be followed by changes in the psi-type CD . These were originally thought to be solely dependent on the photoinduced electron and ion transport because of their sensitivity to ΔpH, uncouplers and inhibitors. Later results, however, showed that the magnitude of the changes increases with light intensities beyond the saturation of photochemical reactions and hence must involve additional mechanisms [272,273]. Light-induced reversible changes were also detected in the psi-type CD of lamellar LHCII aggregates, associated with changes in their chiral macroorganisation . Prolonged illumination also causes monomerization of the LHCII trimers [205,275]. The observations were explained in terms of a so-called thermo-optic mechanism — according to it, the trigger for the structural transition is a local thermal jump caused by the dissipation of photon energy [179,205,276,277]. Reversible light-induced unstacking and disordering of stacked lamellar LHCII aggregates was also confirmed by EM and attributed to the thermo-optic mechanism . The transient local heating upon excitation-energy dissipation could be followed by the broadening of the transient absorption spectra of LHCII . It is believed that local conformational changes in the vicinity of the highly flexible N-terminus of the complex  result in the release of specifically bound cations, especially Mg2+, which then temporarily shift the balance of electrostatic forces between the stacked membrane sheets and cause unstacking . The remarkable flexibility of LHCII aggregates and dynamic modulation of their multilamellar structure represents another evidence for the active role of LHCII in promoting and regulating grana macroorganisation.
Like higher-plant grana TM, diatoms cells exhibit large psi-type CD signals, indicating the presence of higher ordered complexes . The light-induced TM reorganisations detected by SANS were correlated with light-induced changes in the chiral macroorganisation of the pigment system detected by psi-type CD . In the case of P. tricornutum the changes in the membrane RD were closely correlated with the psi-type CD [171,279] thus further elaborating the high structural flexibility of the TM. Comparison of CD spectra of P. tricornutum cells with SANS spectra revealed that loss of CD amplitude under heat treatment was accompanied by an increase in membrane RD, confirming that the psi-type CD signal not only originates from the lateral macroorganisation, but vertical appression of TM also contributes to it . A recent study of investigating the effect of different ions on the TM architecture of T. pseudonana  led the understanding forward. The different responses of the positive and negative psi-type CD bands to salt concentration suggest their different origin as is the case in plants (see 2.3 above). The optimal range of concentrations at which the membrane structure appeared to be unchanged was also very similar to higher plants which led the authors to propose that the inner four of the six diatom TM are equivalent to the grana of plants.
Macroorganisation changes associated with state transitions and protein phosphorylation
Protein phosphorylation is a key regulatory and signalling mechanism of the photosynthetic apparatus and is intricately linked with the dynamic macroorganisation of the TM [281,282]. In plant TM, phosphorylation is controlled by the cooperative action of at least two kinases, STN7 and STN8, and two cognate phosphatases, PPH1/TAP38 and PBCP ; several other kinases and phosphatases are present in chloroplasts . The phosphorylation state of LHCII triggers state transitions as a buffering system for dynamic low-light acclimation under fluctuating light conditions, balancing the excitation energy flux between the two photosystems. During state transitions a subpopulation of the LHCII proteins, normally associated with PSII (in state 1), becomes phosphorylated by the STN7 kinase , and migrates from the PSII-enriched stacked region to the PSI-containing unstacked stromal region – interacting with PSI and forming PSI–LHCII supercomplexes (in state 2) [116,282,285–287]. Despite the lack of mature grana as observed in higher-plant chloroplasts, the TM of green algae (Chlorophyta) also contain well distinguished appressed and non-appressed regions and are capable of undergoing state transitions — up to 80% of the LHCII are thought to migrate between PSII and PSI in C. reinhardtii . However, there has been some debate over how much of the LHCIIs are mobile and whether they all transfer energy to PSI in state 2 [268,289–291]. In contrast with green algae, diatoms do not exhibit differentiated granal and stromal TM regions and lack state transitions .
State transitions as such are dynamic changes in TM macroorganisation and presumably require remodelling of the photosystem supercomplexes. EM studies have revealed PSI–LHCI–LHCII supercomplexes in plant TM in state 2 [293–295] and a high-resolution cryo-EM structure has been recently presented , showing the specific interaction of the phosphorylated Thr in Lhcb2 with the PsaL subunit of PSI and the involvement of other PSI subunits (PsaO in PsaH) known to be essential for LHCII docking. Interestingly, no dissociation of PSII–LHCII supercomplexes was detected in state 2, indicating that the mobile fraction of LHCII belongs to the loosely bound L-trimer population [295,297]. On the other hand, disassembly or destabilisation of PSII supercomplexes, by light treatment, promote state transitions . Grieco et al.  proposed that the L-type LHCII trimers function as the antenna of both photosystems and phosphorylation simply increases the excitation energy transfer efficiency to PSI. It is likely that LHCII in the stromal lamellae is energetically connected to PSI, but its physical interactions are very susceptible even to mild detergents and intact complexes cannot be isolated. We have shown that LHCII isolated from dark-adapted plants can efficiently transfer energy to PSI in vitro in reconstituted membranes . LHCII could substitute LHCI as a PSI antenna in an LHCI-deficient Arabidopsis mutant . Spectroscopic studies of wild-type plants estimate that PSI binds significantly more LHCII in vivo than can be accounted for in the PSI–LHCI–LHCII supercomplex [302–304].
LHCII phosphorylation and state transitions are associated with changes in the grana/stroma macroassembly on a mesoscopic level, for example, changes in grana stacking [30,305]. The fine causal interrelationship between phosphorylation, protein mobility and membrane macroorganisation changes is only beginning to be unravelled. The surface charge density model of grana stacking  would predict that phosphorylation will perturb the equilibrium of forces maintaining stacking by the addition of more negative charges to the thylakoid surface . The grana membrane diameter and the number of layers per stack decreased, while the number of grana per chloroplast increased upon phosphorylation of LHCII [35,36,306]. Also the STN7 and STN8 kinase mutants of Arabidopsis had constitutively larger grana; conversely, phosphatase mutants (either of the PSII-specific PCBP or the LHCII-specific PPH1/TAP38) have reduced grana size compared with wild-type Arabidopsis [33,34,258,259,307], as was seen in Figure 2. The smaller grana size is thought to increase the contact area between the granal and stromal thylakoids and facilitate interaction and movement of proteins and electron carriers between the appressed and non-appressed regions . For example, the increased rate of linear ET under low light conditions could be because of faster plastoquinone movement between PSII and cyt b6f  — in granal TM this is the rate-limiting step for the photoinduced ET .
Phosphorylation of LHCII leads to an increase in the TM flexibility — enhanced heat- and light-induced reorganizations in TM upon phosphorylation in vitro were detected by CD spectroscopy . The role of light-controlled phosphorylation of LHCII in regulating the granal TM macroorganisation was also elucidated by Janik et al. , assembling LHCII isolated from dark-adapted or HL-treated leaves (partly phosphorylated and containing zeaxanthin) with thylakoid lipids. In both cases multilamellar structures were formed, as previously shown , but dark-adapted LHCII assembled into lamellar stacks stabilised by trans-layer rivet-like structures, not observed in the more disordered lamellae from HL–LHCII.
Both CD spectroscopy and SANS measurements revealed changes in the membrane macroorganisation and RD (stacking) upon state transitions in C. reinhardtii . In state 1, the membranes were more tightly stacked with well-defined periodicity, whereas in state 2 the periodic structure of the TM was practically eliminated, revealing a more profound reorganisation than moderate unstacking or increase in RD (Figure 12).
Hypothetical model for the chloroplast remodeling during state transitions in
If TM reorganisations related to state transitions are primarily driven by electrostatic interactions of the phosphorylated LHCII, then similar effects could be expected upon PSII phosphorylation. Light-dependent phosphorylation of the PSII core proteins by the STN8 kinase facilitates lateral diffusion of damaged PSII through densely packed grana to the unstacked stroma exposed membranes where the D1 protein is replaced [259,310]. The STN8-dependent reduction in the grana diameter and stacking in protoplasts and thylakoids upon HL treatment suggests that these changes are triggered by increased PSII phosphorylation levels . Similar to LHCII phosphorylation in low light, decreased grana size upon HL stress would provide increased surface area contact between the grana and stromal lamellae, in turn increasing the accessibility of damaged D1 to the enzymes of repair. It has been shown that that dephosphorylation and subsequent degradation of the D1 protein are concentrated in grana thylakoids; after the repair of D1 protein, monomeric PSII core complex assemble in the stroma lamellae  and migrate back to the stacked grana membranes and assemble into dimers and PSII supercomplexes .
Rearrangements associated with NPQ
NPQ — the protective mechanism of thermal dissipation of excess excitations — down-regulates the quantum efficiency of PSII photochemistry by 60–80% . Although state transitions fall under the definition of NPQ, the main NPQ component is the so-called energy-dependent quenching that dissipates excess absorbed light energy safely into heat and is activated by the acidification of the thylakoid lumen and the consequent de-epoxidation of the LHCII-bound xanthophyll violaxanthin to zeaxanthin and protonation of the PSII-associated protein PsbS [114,115,312]. The formation of quenching sites in the antenna effectively reduces the excitation diffusion length (the distance over which excitations can migrate before they are lost via radiative or nonradiative decay) and hence the fraction of excitations reaching the RCs . The molecular mechanism behind NPQ has so far not been unequivocally established and there may in fact be several mechanisms acting in parallel, such as decay via the S1 state of Lut [209,314] or to another dark electronic state , via a mixed Chl–Car exciton state , or a Chl–Chl CT state [225,231,232], or a Chl–Car CT state [317,318]. There is much debate over which of these mechanisms take place in vivo — some authors rule out the relevance of Chl–Chl CT states [237,238] while other studies point to the possibility of Chl–Car mechanisms to be an artefact of the laboratory measurements . There is, however, a general consensus that the activation of NPQ is associated with membrane macroorganisation changes . The macroorganisation itself is crucial for the effectiveness of NPQ to remove excitations from the RC. Excitations migrate between pigment–protein complexes over long distances, characterised by diffusion length of ∼50 nm [320,321] or functional domain size of several tens of antenna complexes [252,322]. Because of the connectivity, it is sufficient to switch several antenna complexes into a quenched state in order to effectively lower the excitation pressure on the RC [320,323].
The aggregation model of LHCII was postulated by Horton and co-workers [324,325] primarily based on the spectroscopic similarities of NPQ in vivo and in isolated LHCII aggregates (for review see [114,115]). Aggregation of LHCII was thought to be promoted by either zeaxanthin  or protonated PsbS [231,326]. Holzwarth and co-workers  proposed a four-state, two-site model for NPQ, separating the site of action of zeaxanthin and PsbS — the latter invoking detachment of LHCII from the PSII supercomplexes followed by aggregation. This model, although not widely accepted, explained the different fluorescence kinetics, steady-state and time-resolved fluorescence spectroscopy signatures of NPQ in zeaxanthin- and PsbS-deficient Arabidopsis mutants. The appearance of a far-red-enhanced fluorescence component attributed to aggregated LHCII and at the expense of PSII–LHCII emission was critically PsbS-dependent [231,233] and suggested a different quenching mechanism (via Chl–Chl CT state) in the detached LHCII compared with the zeaxanthin-dependent quenching in the PSII-attached antenna. A recent study of an Arabidopsis mutant devoid of minor antenna complexes (NoM) confirmed the formation of far-red states under NPQ in the trimeric LHCII and showed the dependence of quenching on Lut in the minor but not in major LHCII . However, the NoM mutant also showed that quenching in the major LHCII strongly depends on zeaxanthin.
Biochemical evidence emerged of the PsbS-dependent detachment of LHCII from the PSII supercomplex in the form of a complex composed of minor and trimeric LHCII with a corresponding shortening of the average PSII–PSII distances . Probably the most convincing evidence for the existence of large-scale macroorganisation changes during NPQ activation came from freeze-fracture EM , which confirmed the shorter distances between PSII and revealed clustering of LHCII on the periphery of the grana (Figure 13). The PSII–LHCII rearrangements were found to be dependent on PsbS, which suppressed the formation of ordered semicrystalline arrays and increased the mobility of the pigment–protein complexes in the grana [72,330]. In TM from lincomycin-treated plants, PsbS was found to induce the formation of large LHCII aggregates . The localisation of PsbS in the grana was revealed by immunogold labelling with a specific antibody . In dark-adapted conditions, PsbS was found to be associated with the PSII supercomplexes, whereas in HL conditions it interacted with the trimeric LHCII providing further support to the proposed PsbS-dependent detachment of LHCII from PSII.
Structural model of NPQ-related reorganization of thylakoid grana membranes.
The PsbS protein was found to influence the flexibility of the TM and the Mg-dependent unstacking/restacking in vitro by using CD spectroscopy . Generally, if NPQ is associated with large-scale lateral macroorganisation changes, it is logical to expect also changes in membrane stacking and the three-dimensional macrostructure. While there are indications for NPQ-related changes in grana stacking, multilamellar order and RD in plants and algae from CD and SANS experiments (for review see ref. ), a precise mechanistic link or the consequence of elementary events has not yet been established.
Diatoms exhibit a stronger NPQ compared with land plants that is influenced by both the pH changes as well as the xanthophyll cycle [227,333,334]. Spectroscopic studies point towards multiple quenching sites/processes in FCPs [334–339]. In vitro aggregation of FCPa results in lowering of the fluorescence yield  and the appearance of far-red emission with CT character . Diatoms do not possess PsbS, but Lhcx proteins have been demonstrated to play an essential role in NPQ . Lhcx proteins were found to be associated with PSI in pennate diatom P. tricornutum  only in crude FCP fractions but not in isolated trimeric FCP complexes so far [341,342]. To examine the influence of Lhcx1/Fcp6, CD spectra of cells, isolated TM and FCP complexes of antisense mutants of Fcp6 were analysed . Depletion of Fcp6 led to visible changes in the CD of cells and TM and to a complete loss of the ability to undergo light-induced CD changes, whereas no differences were detected in isolated FCPs, indicating that Fcp6 acts by altering the macroorganisation of FCPs in the membrane, in a way reminiscent of the function of PsbS in plants.
Conclusions and perspectives
The knowledge of the structure and macroorganisation of the TM, from near-atomic structures of the membrane-embedded protein complexes to the overall three-dimensional architecture has advanced significantly in the past years, and much of this is owing to EM and particularly cryo-EM techniques. Cryo-EM tomography largely confirmed the helical model of the granum while providing additional details and refinements. We can expect further advancements in uncovering three-dimensional structures that can be transiently formed during the biogenesis of the TM or under different physiological states. An ideal method that preserves both the 3D architecture and resolution is cryo-focused ion beam (cryo-FIB) milling. Cryo-FIB combined with cryo-EM tomography has been used to reveal high-resolution 3D structures of chloroplasts in intact vitreous Chlamydomonas cells [344,345] and is a promising approach enabling accurate measurements of the thylakoid architecture. With this method it is now possible to visualise an entire stack of TM and create a topographic map of both stroma- and lumen-exposed surfaces with resolution of all major integral protein complexes . Undoubtedly, studies of this kind will elevate the understanding of not only the membrane architecture and topology but also the forces and causal relationships that maintain and regulate the TM structure.
The unparalleled spatial resolution of EM and X-ray diffraction makes them invaluable structural analysis tools but the extreme dynamic flexibility of the TM and its components require complementary spectroscopy and microscopy techniques that can probe structural and macrostructural changes in real-time and in vivo. CD spectroscopy provides sensitive and accessible spectroscopic fingerprints of different levels of organisation of the photosynthetic apparatus but is hampered by the poor understanding of the origin of the CD features of TM and the inability to relate the spectral CD changes to particular changes in the molecular structure and macroorganisation. The ACD technique can potentially alleviate this problem by adding a molecular co-ordinate label to the CD bands that can help in identifying the underlying excited states in combination with structure-based quantum modelling . However, at present the model calculations are restricted to the lowest-energy Qy transitions of Chls, whereas higher-energy Chl and Car excited states are omitted as this exponentially increases the model complexity and computational cost. Extending the Hamiltonian will not only allow for a better understanding of the red region, whose non-conservative shape suggests mixing with higher states , but also allow for a deeper analysis of the spectral region below 600 nm, where the most striking features of ACD are observed. ACD spectroscopy is yet to be applied to the rest of the photosynthetic pigment–protein complexes and supercomplexes, for which detailed structural and quantum modelling data are available. Applying quantum model fitting to combined CD/ACD and ultrafast 2D electronic spectroscopy data  can provide the most accurate and detailed understanding of the excitonic structure and dynamics of photosynthetic complexes. So far, ACD has been limited to measurements of uniaxially oriented samples because of the difficulty in disentangling chiral from achiral contribution. This limitation can in principle be overcome by complete Stokes–Mueller polarimetry, where all elements, chiral and achiral, of the Mueller matrix are obtained separately .
The psi-type CD has been an extremely sensitive indicator of changes in membrane macroorganisation, detectable under a wide variety of conditions and organisms — dark/light adaptation, state transitions, NPQ, the antenna composition, physicochemical environment and various abiotic stresses alter the psi-type CD of TM in a reproducible way. While these changes are in many cases relatable to independent measurements of structural parameters by other techniques, the theoretical description and understanding of the psi-type CD is at present glaringly inadequate. The differential spectral responses of psi-type CD could be used, in principle, to detect different physiological stress responses and membrane rearrangements, but systematic studies are scarce. A comparative circular spectropolarimetric study of a variety of plant and algal species has uncovered surprising spectral variability, evidently related to the differences in their membrane macroorganisation . An exciting new prospect is the ability to measure CD spectra in the field and follow the dynamic flexibility of the photosynthetic apparatus in situ in the natural fluctuating environment of plants .
3D structured illumination microscopy
circular intensity differential scattering
confocal laser scanning microscopy
fluorescence recovery after photobleaching
plastid terminal oxidase
Small-angle neutron scattering
The authors thank Dr. Gyo˝zo˝ Garab for ideas, fruitful discussion and critical reading of the manuscript. Support from Diamond Light Source Ltd, UK and the B23 SRCD beamline staff (Dr. Giuliano Siligardi, Dr. Tamás Jávorfi, Dr. Rohanah Hussain) for UV SRCD measurements is appreciated. P.H.L. acknowledges financial support from the Hungarian Ministry of Finance (GINOP-2.3.2-15-2016-00001) and the National Research, Development and Innovation Office (NN-124904, 2018-1.2.1-NKP-2018-00009).
The Authors declare that there are no competing interests associated with the manuscript.