By continuing to browse this site you agree to us using cookies as described in About Cookies
Notice: Wiley Online Library will be unavailable on Saturday 7th Oct from 03.00 EDT / 08:00 BST / 12:30 IST / 15.00 SGT to 08.00 EDT / 13.00 BST / 17:30 IST / 20.00 SGT and Sunday 8th Oct from 03.00 EDT / 08:00 BST / 12:30 IST / 15.00 SGT to 06.00 EDT / 11.00 BST / 15:30 IST / 18.00 SGT for essential maintenance. Apologies for the inconvenience.
The process of oxygenic photosynthesis enabled and still sustains aerobic life on Earth. The most elaborate form of the apparatus that carries out the primary steps of this vital process is the one present in higher plants. Here, we review the overall composition and supramolecular organization of this apparatus, as well as the complex architecture of the lamellar system within which it is harbored. Along the way, we refer to the genetic, biochemical, spectroscopic and, in particular, microscopic studies that have been employed to elucidate the structure and working of this remarkable molecular energy conversion device. As an example of the highly dynamic nature of the apparatus, we discuss the molecular and structural events that enable it to maintain high photosynthetic yields under fluctuating light conditions. We conclude the review with a summary of the hypotheses made over the years about the driving forces that underlie the partition of the lamellar system of higher plants and certain green algae into appressed and non-appressed membrane domains and the segregation of the photosynthetic protein complexes within these domains.
Oxygenic photosynthesis produces most of the organic matter on Earth, as well as almost all of its oxygen. The primary steps of this process – the conversion of light energy to usable chemical energy – are carried out by four multisubunit membrane–protein complexes (Figure 1). Two of the complexes, photosystem I (PSI) and photosystem II (PSII), function as molecular photovoltaics – emitting electrons upon the absorbance of light energy. The third complex, cytochrome b6f (cyt. b6f), mediates the transport of electrons between the two PSs and further contributes to the formation of a proton-motive force (pmf). The last complex, an F-type ATPase (CF1CF0-ATP synthase), converts the pmf into ready-to-use ATP molecules by rotary catalysis. Supplementing these complexes are a quinone molecule, called plastoquinone (PQ), and a small, water-soluble copper-binding protein, called plastocyanin (PC), which mediate the transport of electrons from PSII to cyt. b6f (PQ), and from the latter to PSI (PC).
All of the aforementioned components reside within flattened vesicles called thylakoids, which also provide a medium for energy transduction. The most complex form of thylakoid membranes is the one found in the chloroplasts of higher plants. In these species, the thylakoids form elaborate, highly interconnected three-dimensional (3D) lamellar networks hallmarked by the differentiation into two distinct morphological domains: cylindrical stacks, comprised of multiple, tightly appressed layers, called grana, and unstacked membrane regions that interconnect the grana, called stroma lamellae (Figure 2A). Underlying the structural differentiation is segregation of the photosynthetic protein complexes within the membranes: PSII and its peripheral major and minor light-harvesting complexes are localized primarily to the granal membranes, whereas PSI, its antennae, and ATP synthase are concentrated in non-appressed regions of the network, namely the stroma lamellae, the grana end-membranes and the grana margins. This unique morphologically and compositionally segregated organization of the thylakoid membranes has been proposed to play different roles in the function and light adaptation of the photosynthetic apparatus.
In this review we discuss the macroscopic organization of the photosynthetic apparatus of higher plants, as derived from high-resolution microscopic examinations, primarily electron microscopy techniques, as well as from genetic, biochemical, and spectroscopic analyses. As an example of the ability of the photosynthetic machinery to dynamically adjust to changes in its light environment, we discuss the structural rearrangements and underlying biochemical modifications that occur during short-term chromatic adaptation responses, called state transitions. A detailed description of the composition and structure of the individual protein complexes that constitute the apparatus is generally not included. For these topics, the reader is referred to the excellent reviews available in the literature.
Architecture of the higher-plant thylakoid membrane network
Higher-plant chloroplasts have been studied by microscopic examinations since the mid-1800s. They were first described as ‘chlorophyll granules’ by von Mohl (1837) and ‘pigment-bound structures’ by Unger in 1848 based on observations using the light microscope. A bit later, minuscule granulations – dots seen in the comparatively transparent surrounding of the chloroplast – were visualized and termed ‘grana’ and ‘stroma’, respectively (Meyer, 1883; Schimper, 1885). Using shorter wavelengths (UV) and higher magnification powers, the morphological heterogeneity of the thylakoid membranes became apparent: dark grana structures surrounded by lighter colored stroma lamellae (Doutreligne, 1935; Heitz, 1936; Menke, 1940). The next major advancements were made possible with the use of the electron microscope, which was first built in 1931 by E. Ruska and M. Knoll, and made commercially available by Siemens in 1939. Very soon afterwards, the first thin-section transmission electron microscope (TEM) images of chloroplasts (Kausche and Ruska, 1940) and leaf preparations (Menke, 1940) were published. However, the complexity of the thylakoid membrane structure was revealed only several years later, following developments in sample (chemical) fixation methods (Algera et al., 1947; Granick and Porter, 1947).
Models describing the spatial organization of higher-plant thylakoid membranes have been around since the mid-1950s (Hodge et al., 1955; Steinmann and Sjostrand, 1955). The first prominent one, by Menke (1960), proposed that ordered arrays of grana stacks are traversed by equally spaced stroma lamellar sheets. He also coined the term ‘thylakoid’ (‘sac-like’ in Greek; Menke, 1962). The double membranal nature of the thylakoid lamellae, along with its enclosed luminal space, was established the following year (Heslop-Harrison, 1963; Weier et al., 1963). These findings, along with the notion that the lumen is continuous throughout the entire lamellar system, resulted in the emergence of different views for the arrangement of the grana and stroma thylakoids within the network. Weier et al. (1963) proposed a tubular fretwork arrangement in which the grana are transversely intersected at multiple planes along their long axis by tubular- or bridge-shaped stroma lamellae; although this was later proven to be a fixation artifact (Falk and Sitte, 1963). Alternatively, Heslop-Harrison (1963) proposed that the stroma lamellae, described as perforated sheets, intersect the grana at an angle to their axis and thus a single stroma lamellar sheet can join several layers within the body of the granum. Around the same time, another model was introduced by Wehrmeyer (1964), suggesting a spiral arrangement of the stroma lamellar frets around the grana. Modification of this model, called the ‘helical fretwork model’, came about following serial section TEM analyses by Paolillo and co-workers (Figure 3A; Paolillo and Falk, 1966; Paolillo et al., 1969; Paolillo, 1970). In this technique, the objects of interest are traced in successive thin sections and the volume is interpolated from the contours. The proposed helical arrangement provided an explanation for how the stroma lamellae can be connected to a granum stack at multiple sites yet appear to intersect it horizontally in grana mid-sections. Later studies employing, again, serial-section TEM or scanning electron microscopy (SEM) provided support for the helical model (Brangeon and Mustárdy, 1979; Staehelin, 1986). A modified version of this model was subsequently proposed by Mustárdy and Garab (2003). A different model, called the folded-membrane model, was put forth to illustrate the segregation of PSI and PSII into stroma and grana lamellar domains, respectively (Andersson and Anderson, 1980; Anderson et al., 1988). Later on, it was extended by Arvidsson and Sundby (1999; Figure 3B), who proposed that grana are formed by a single continuous membrane that folds on itself and is stabilized solely by surface charge-mediated electrostatic interactions. A notable virtue of this model is that it can readily account for reversible unstacking of the membranes in the appressed regions of the network (see below).
Even though conventional cross-section electron microscopy has provided valuable information about the organization of the thylakoid membranes, this method can provide only two-dimensional (2D) information. Given the complexity and pleiomorphic nature of the thylakoid lamellar system, deducing its 3D structure from 2D data is precarious. In recent years, application of electron microscope tomography (EMT) has become increasingly common. In EMT, a series of projection images of a sample rotated about an axis perpendicular to the electron beam are collected. The individual projections are then aligned in a common frame of reference in silico and the 3D object is reconstructed, usually by weighted back-projection algorithms, to a resolution of several nanometers (Penczek et al., 1995; Mastronarde et al., 1997; Lučićet al., 2005; Robinson et al., 2007; Barcena and Koster, 2009; Bartesaghi and Subramaniam, 2009; Leis et al., 2009; Nevo et al., 2009).
While some of the discrepancies between the different models may arise from interpretation of the electron tomographic data, which is not a trivial task for a pleiomorphic and intricate system such as the thylakoid network, others may be due to the nature of the samples used in the different studies. The works of Daum and Mustárdy and their colleagues (Mustárdy et al., 2008a; Daum et al., 2010) were performed on isolated chloroplasts and isolated solubilized thylakoids, respectively, whereas the works of Shimoni et al. (2005) and part of the work of Austin and Staehelin (2011) were performed on leaf specimens. The isolation of chloroplasts and, even more so, of thylakoids probably disrupts the structure of the thylakoid network (Kouřil et al., 2011). Another factor relates to the preparation techniques employed in the studies. Ideally, to observe the cellular milieu in its most native state, one would prefer to employ cryo-immobilization techniques and to examine the samples under cryogenic conditions. For thick biological samples, such as plant leaf tissues, vitrification –‘freezing’ of water in the amorphous state – can be achieved by high-pressure freezing (HPF; Moor and Hoechli, 1970; Dubochet and Mcdowall, 1981; Moor, 1987). However, for electron microscope examination, the thickness of the samples cannot exceed ∼ 0.3–1 μm, depending on the microscope’s accelerating voltage. This necessitates following HPF either by freeze-substitution and resin embedding, which allow conventional sample sectioning, staining and visualization, or by cryo-sectioning and examination by cryo-TEM. With both alternatives possessing their pros and cons, they share the fact that samples are initially rapidly immobilized. Application of the second approach, termed cryo-electron microscopy of vitreous sections (CEMOVIS), has not yet become routine due to the difficulty in consistently obtaining high-quality sections (Al-Amoudi et al., 2005; Bouchet-Marquis and Hoenger, 2011). In fact, to date, it has been employed on intact leaf tissues only twice (Michel et al., 1991; Kirchhoff et al., 2011). In these two works, organelle ultrastructure and dimensions were shown to be comparable when prepared either for CEMOVIS or by freeze-substitution. While Daum et al. (2010) did employ CEMOVIS in their tomographic study, their starting material for HPF was isolated chloroplast suspensions. On the other hand, in the work of Shimoni et al. (2005) and in part of the work by Austin and Staehelin (2011), leaf tissues underwent HPF (and subsequently freeze-substitution), thus having the advantage that the material immobilized was in its native context.
The distribution of photosynthetic protein complexes within the thylakoid membranes
Efforts aimed at determining the distribution of the protein complexes that conduct the light-driven reactions of photosynthesis between the appressed and non-appressed regions of the thylakoid membrane began in the mid-1960s by J. Anderson and co-workers, who studied the PSI/II and chlorophyll (a/b) content of detergent-fractionated spinach thylakoids (reviewed in Anderson, 1999). At the same time, Branton (1966) published a paper entitled ‘Fracture faces of frozen membranes’, giving a new interpretation for membrane fractured surfaces revealed by an emerging technique called freeze-fracture, which allows visualization of membrane proteins by electron microscopy. In this technique, a rapidly frozen tissue or cell/membrane suspension is mechanically fractured while maintained at low (−160 to −100°C) temperature. This process tends to result in splitting of the membrane bilayers along their central hydrophobic plane, giving rise to two fracture faces that are complementary to each other. As membrane-integral protein complexes remain intact during this procedure, their size and organization within the (fractured) membranes can be determined (Figure 2C). To visualize peripheral protein complexes, the membrane surface is exposed by a process known as ‘etching’, in which frozen water present on the surface is sublimated by controlled elevation of the temperature.
The distribution of the fourth protein complex, cyt. b6f, in the thylakoid membranes of higher plants is somewhat less clear. Most studies indicate an almost even distribution between granal and stroma lamellar domains (Cox and Andersson, 1981; Anderson, 1982b; Allred and Staehelin, 1985, 1986; Olive et al., 1986). Using membrane fractionation and immuno-electron microscopy, the distribution of cyt. b6f was also shown to change during state transitions in Chlamydomonas, with a more significant fraction of the complex being localized to PSI-enriched regions of the network under light conditions that preferentially excite PSII (Vallon et al., 1991). In a recent study performed in collaboration with H. Kirchhoff, we analyzed the redox-equilibration kinetics within the high-potential chain (the part of the electron transport chain between cytochrome f and P700) and concluded that 30–50% of the cyt. b6f population resides within grana cores; the remaining fraction is likely distributed between the grana margins and the stroma lamellae (Kirchhoff et al., 2011). Contrary to this, some studies suggest preferential localization of cyt. b6f in the non-stacked regions of the network (Dunahay et al., 1984; van Roon et al., 2000). Accordingly, it has been suggested by Dekker and Boekema (2005) that one of the extrinsic loops of cyt. b6f (as seen in the crystal structures), which protrudes into the stromal space between the stacked layers, may prevent the complex from being localized to the stacked regions, at least under conditions of increased stacking. For more quantitative discussions of the distribution of photosynthetic protein complexes within the different thylakoid membrane domains (including the grana margins and grana end-membranes) and its impact on electron transport modes and efficacy, we refer the reader to the following reviews by Albertsson (1995, 2000, 2001).
Many years after the pioneering freeze-fracture studies were conducted, atomic force microscopy (AFM) was employed to probe the distribution of PSI and ATP synthase within the thylakoid membranes, as well as the surface architecture of the latter (Figure 2D; Kaftan et al., 2002). More recently, Kirchhoff et al. (2008) used AFM to visualize PSII complexes within isolated grana preparations. This work revealed that complexes associated with the two opposing membranes that surround the thylakoid lumen are laterally shifted with respect to each other, preventing steric clashes between their oxygen-evolving complexes, which protrude deeply into the luminal compartment. Atomic force microscopy (Binnig et al., 1986), also known as scanning force microscopy (SFM), is a proximal probe microscopy technique used to map surface topography. Unlike the electron microscope, the atomic force microscope is not just a high-resolution imaging tool, but can also be used to manipulate matter at forces and length scales that control molecular structures and interactions. Another advantage of AFM over the electron microscope is that it can also operate in fluids, offering the possibility of extracting dynamic information (reviewed in Reich et al., 2001). While AFM has been extensively applied to the study of the chromatophore membranes of anaerobic phototrophs, providing spectacular views of their reaction centers and antenna complexes (reviewed in Scheuring, 2006; Scheuring and Sturgis, 2009; Sturgis et al., 2009), its application to the study of the supramolecular organization of photosynthetic complexes within the thylakoid membranes is regretfully rather limited. A detailed summary of the application of AFM to the investigation of photosynthetic membranes, as well as of isolated photosynthetic protein complexes, is given in Vacha et al. (2005). That review also discusses other microscopic techniques employed in photosynthetic research, including optical, electron, and scanning tunneling microscopy.
Association of the two photosystems with their antenna complexes
Both PSI and PSII are laterally associated with extensive peripheral antenna systems that increase their spectral and spatial absorption cross-sections. In addition, the major antenna complex of PSII, light-harvesting complex II (LHCII), plays vital roles in the stacking of the thylakoid membranes, in the distribution of absorbed light energy between the two PSs, and in the protection of the photosynthetic apparatus against photooxidative damage.
The LHC of plant PSI consists of four, closely related antenna proteins, designated Lhca1-4, which share an LHCII-type fold and bind to the reaction center with varying stoichiometries, depending on light conditions and other environmental factors. Each of these (LHCI) proteins possesses approximately 15 chlorophyll (a and b) molecules (Amunts et al., 2007, 2010) and, compared with other chlorophyll a/b-binding proteins, is typified by a red-shifted absorption profile (Croce et al., 2002). The first insights into the organization of the PSI–LHCI supercomplex came from single-particle electron microscopy analyses, which indicated that binding of LHCI to PSI occurs at its PsaF/J side (Boekema et al., 2001; Germano et al., 2002; Kargul et al., 2003). Subsequently, the details of the assembly were unraveled by X-ray crystallography studies of PSI–LHCI supercomplexes isolated from pea thylakoids (Figure 4A; Ben-Shem et al., 2003; Amunts et al., 2007, 2010). The structures revealed that the four Lhca antenna proteins are assembled onto the PSI core complex as two heterodimers, Lhca1–Lhca4 and Lhca2–Lhca3, as previously suggested by biochemical and mutagenic analyses (Jansson et al., 1996; Schmid et al., 1997, 2002; Ganeteg et al., 2001; Ihalainen et al., 2002). The two dimers associate in series to form a ‘half-moon-shaped belt’ (Ben-Shem et al., 2003) at the PsaF side of the reaction center, consistent with electron microscopy observations, with Lhca4 and Lhca2 lying near each other at the central part of the belt. Efficient excitonic coupling between the LHCI antennae and the reaction center, which are separated by a deep cleft, is enabled by peripheral chlorophylls, termed ‘gap’ chlorophylls, which are clustered in several regions along the core–antenna interface (Figure 4A). Chlorophyll molecules positioned at the contact regions between the LHCI monomers similarly ensure rapid energy transfer along the antenna belt itself. These ‘linker’ chlorophylls may also play a role in dimer formation. Association between the LHCI cluster and the PSI core, which is apparently weak, is achieved primarily through interactions between helix C of Lhca1 and two tilted helices of the PSI PsaG subunit. Notably, this subunit is absent from PSI complexes of cyanobacteria and red algae, which do not utilize Lhc proteins for light harvesting, but rather make use of large, water-soluble molecular assemblies, coined phycobilisomes. The apparently loose coupling amongst the Lhca proteins and between them and the reaction center is likely to facilitate changes in the composition of the antenna system in response to alterations in photon fluence (Bailey et al., 2001). Interestingly, a recent study showed that even though the Lhca1/4 and Lhca2/3 dimers have essentially indistinguishable absorption and emission characteristics, they nevertheless differ in their response to high light (Wientjes and Croce, 2011).
The association of PSII with its LHCs is comparatively more versatile. The PSII core dimer of higher plants and green algae is associated with two types of chlorophyll a/b-binding antenna complexes: monomeric, minor Lhcb proteins (CP24, CP26, and CP29), and major trimeric LHCII complexes. The latter consist of different combinations of the Lhcb1–3 proteins, which possess a highly similar amino acid sequence, and of which Lhcb1 and -2 are the predominantly (∼ 90%) expressed isoforms (Jansson, 1994). Each LHCII trimer binds 42 chlorophyll molecules (24 Chl a and 18 Chl b) and 12 carotenoids of three different types, with the latter playing a mostly photoprotective role (Liu et al., 2004; Standfuss et al., 2005, and references therein). Together, the three major LHCII proteins constitute about 30% of the total protein content of the chloroplast thylakoid membranes, making LHCII the most abundant membrane protein in nature.
That PSII can bind a variable number of LHCII complexes had already been observed in the freeze-fracture studies conducted from the early 1970s to mid-1980s (Armond et al., 1977; Wollman et al., 1980; and see Staehelin and Arntzen, 1983). However, most of our knowledge about the association of PSII with its antenna proteins comes from electron microscope image- and single-particle analyses of negatively stained or cryo-immobilized preparations, obtained from detergent-solubilized membranes (reviewed in detail in Dekker and Boekema, 2005). A frequently isolated PSII–LHCII assembly (particularly from spinach thylakoids) is the so-called C2S2 supercomplex (Figure 4B) – composed of a dimeric PSII core (C2) and two ‘strongly bound’ LHCII trimers (S2). In addition, the complex contains two copies of each of the minor antenna proteins, CP26 and CP29, which mediate the association of the two LHCII trimers to the complex. Altogether, the C2S2 complex contains a total of approximately 200 chlorophyll molecules (Boekema et al., 1995; Hankamer et al., 1997, 2001; Nield et al., 2000a,b; Harrer, 2003; Nield and Barber, 2006).
The forces that govern granum formation and partitioning of the thylakoid network
In 1966, Izawa and Good reported that isolated spinach thylakoid membranes suspended in low-salt buffer solutions become completely unstacked and that this can be reversed by the addition of mono- or divalent cations. Subsequently, other groups made similar observations, which strongly hinted at electrostatic forces as an important modulator of thylakoid membrane appression (e.g. Staehelin, 1976; Rubin et al., 1981; Telfer et al., 1984; Kaftan et al., 2002). This notion was elaborated by Barber, who proposed a model in which surface charges play a key role in membrane stacking/unstacking (Barber et al., 1977; Barber, 1980, 1982, 1983; Rubin and Barber, 1980; Rubin et al., 1981). According to the model, electrostatic repulsion prevents close appression of the granal thylakoid membranes unless their (negatively charged) surfaces are effectively screened by counterions, whereupon van der Waals and other attractive forces take over and shift the equilibrium towards the stacked state.
Several years before the surface charge model was published, it was shown that membrane unstacking and restacking are respectively accompanied by randomization and re-segregation of the photosynthetic protein complexes within the membranes (Ojakian and Satir, 1974; Staehelin, 1976), suggesting that the two processes – membrane appression and lateral segregation of the photosynthetic protein complexes – are intimately linked. Given the extremely high density of photosynthetic protein complexes in the membranes (with the overall area fraction occupied by them being about 0.7; Kirchhoff et al., 2002; Haferkamp et al., 2010), the absence of such a linkage would indeed be surprising. We have already mentioned that, by virtue of their stroma-protruding domains, ATP synthase and PSI cannot fit into the appressed regions of the thylakoid membrane. The tendency of PSII–LHCII complexes to self-associate into densely packed macrodomains likewise contributes to their segregation into the granum loci and further promotes exclusion of the sterically incompatible ATP synthase and PSI complexes from these regions (Dekker and Boekema, 2005; Nevo et al., 2009; Adam et al., 2011).
Consistent with being the major occupant of grana membranes, LHCII turned out to also have an essential role in their appression. The first evidence for this came from an electron microscope study in which a barley mutant deficient in LHCII (due to lack of chlorophyll b) was shown to have poorly stacked membrane domains (Goodchild et al., 1966). Later on, it was shown that unstacked thylakoid membranes, in which the stromal-exposed N-terminal segment of LHCII was removed by mild proteolysis (with trypsin), required conditions of higher ionic strength to be restacked compared with untreated membranes (Steinback et al., 1979; Carter and Staehelin, 1980). A more direct demonstration for the role of LHCII in membrane adhesion came from studies showing that liposomes into which purified LHCII was incorporated underwent cation-induced aggregation, which was largely impeded when the vesicles were pre-treated with trypsin (Mcdonnel and Staehelin, 1980; Mullet and Arntzen, 1980; Ryrie et al., 1980). Likewise, supplementing thylakoid membranes deficient in LHCII with the purified complex restored the stacking characteristics observed in native membranes (Day et al., 1984). More recently, it was shown that constitutive production of pea Lhcb1 in transgenic tobacco plants led to (amongst other things) increased grana stacking (Labate et al., 2004). Similarly, a mutant of Chlamydomonas, in which some Lhcb proteins were over-expressed, had more, and more tightly appressed, membrane regions compared with the wild type (Mussgnug et al., 2005).
But how exactly does LHCII promote membrane stacking? The results obtained from the experiments in which the stromal-exposed N-terminal segment of LHCII was removed strongly suggested that this part of the protein plays an essential role in the process. This segment contains four positively charged residues, which contrast with the negatively charged nature of the thylakoid membrane surface as well as with the bulk of the stromal surface of the LHCII itself. It also possesses a threonine residue whose phosphorylation leads to a decrease in membrane appression (see section ‘Light-induced rearrangements in supramolecular organization and thylakoid network architecture during state transitions’ below). It was thus quite clear that the contribution of this segment to membrane adhesion should have an electrostatic origin. A structure-based model for the function of this segment in adhesion was put forth by Standfuss et al. (2005), who determined the structure of pea LHCII within stacked 2D membrane crystals. In this model, the positively charged N-terminal segments of the LHCII trimers present on one membrane layer interact with the negatively charged stromal surface of opposing LHCII complexes present in the adjacent membrane across the partition gap, and vice versa. The remaining negative charge on the LHCII and thylakoid stromal surface was proposed to be mitigated by counterion screening, as proposed by Barber (Barber et al., 1977; Barber, 1980, 1982, 1983; Rubin and Barber, 1980; Rubin et al., 1981). The importance of neutralizing the excess charge is highlighted by the recent finding that over-expression of a plastidial transglutaminase, which attaches polyamines to glutamine residues, including those present in LHCII and other antenna proteins, results in a marked increase in the number of stacked layers in the grana (Ioannidis et al., 2009). It remains to be seen whether the arrangement proposed by Standfuss et al. (2005) for LHCII complexes within the opposing membrane layers can be realized in the context of PSII–LHCII supercomplexes, which govern the protein landscape of the grana membranes.
In addition to those discussed so far, other forces, of both enthalpic and entropic origin, are likely to be involved in the stacking of the thylakoid membranes in the grana. These include hydration forces and lipid–lipid and lipid–protein interactions, as well as depletion interactions (reviewed in Nevo et al., 2009). It is now clear that PSII also plays a critical, though probably indirect, role in grana membrane stacking and organization. This was established in recent studies of Arabidopsis mutants deficient in the thylakoid-associated protein kinase STN8 (St11 in Chlamydomonas), which is responsible for the phosphorylation of (amongst other proteins) several PSII subunits (Bonardi et al., 2005; Vainonen et al., 2005; Reiland et al., 2011). The studies revealed that stn8 plants have fewer grana consisting of significantly longer and more tightly appressed thylakoids compared to the wild type (Fristedt et al., 2009, 2010). These observations suggest that PSII phosphorylation can dynamically regulate grana activity under varying light conditions by controlling the stacking and folding of the membranes. Indeed, it was demonstrated in the above studies that the increase in length and stacking of the grana membranes in Stn8 mutants largely inhibited the lateral migration of photodamaged PSII subunits as well as of the metalloprotease FtsH between the grana the stroma lamellar membranes.
We conclude this section with a note on the density of PSII and LHCII within the grana membranes. Not having sufficient amounts of PSII and LHCII in the membranes would be detrimental for photon capture and energy transfer efficiency. It would also lead to destabilization of the planar organization of the lipid bilayers themselves. This is because the major lipid (50% of the total) of the thylakoid membranes, monogalactosyldiacylglycerol (MGDG; Block et al., 1983; Kobayashi et al., 2009; Adam et al., 2011), has a poor tendency to form bilayers in aqueous solutions. Instead, it packs into inverted hexagonal (HII) structures, which are more accommodating of its conical shape (Epand, 1998; Lee, 2000): It is only in the presence of (certain) proteins that these bilayer-defying lipid molecules are induced to form canonical, planar membrane structures. Notably, one protein that was shown to induce such a transition is LHCII (Simidjiev et al., 2000). Thus, too low a density of LHCII (and, potentially, PSII) would leave large domains organized into inverted hexagonal phases. On the other hand, too high a density of LHCII and PSII would compromise their mobility in the membranes, as well as that of PQ, hindering processes (including repair of photodamaged PSII complexes) that require reorganizations in PSII–LHCII macrostructure, and obstructing PQ-mediated electron transport (which is restricted to begin with; Joliot et al., 1992; Lavergne et al., 1992; Yu et al., 1993; Blackwell et al., 1994; Kirchhoff et al., 2000; Kirchhoff, 2008). Excessive packing of the LHCII antennae may also lead to non-radiative energy dissipation, similarly to that elicited during the ΔpH-dependent component (qE) of the photoprotective response associated with non-photochemical quenching (NPQ). Thus, it is imperative that the density of PSII and LHCII be maintained at the desired level. A way that this might be achieved was proposed by Garab et al. (2000), who suggested that fluctuations in the protein/lipid ratio in the grana membranes are damped by transitions between the planar and inverted hexagonal (HII) forms of the membranes. Such transitions, which are likely to occur at the highly curved regions of the grana margins, can maintain the desired protein/lipid ratio in the membranes by dynamically controlling the area fraction available for LHCII and PSII.
Light-induced rearrangements in supramolecular organization and thylakoid network architecture during state transitions
Ambient light changes continuously, both in intensity and in spectral composition. These changes can be gradual, like those occurring along the course of a day or year, or can be abrupt, like those associated with alterations in shading or cloud coverage. The photosynthetic apparatus must be able to adapt to these changes, such that it can effectively utilize light when it is scarce, minimize photoinduced damage under intense light, and rapidly repair damaged components when the protective capacity is exceeded. This is particularly relevant for terrestrial plants, in which acclimation by means of movement towards a more hospitable environment is impossible, dictating sole reliance on internal adaptations. In the following paragraphs, we discuss a short-term acclimation process that enables plants (as well as other oxygenic phototrophs) to maintain a high photosynthetic yield under a fluctuating light environment. As detailed below, this process involves changes in the PSII–LHCII assembly and the distribution of LHCII in the thylakoid membranes, which in turn lead to macroscopic reorganization of the lamellar network. Due to space limitations, we do not discuss long-term adaptation processes, or the vital processes involved in photoprotection (through NPQ) of the vulnerable PSII complex and its repair when photooxidative damage inevitably occurs.
To achieve a high quantum yield of linear electron transport, the two photosystems need to be excitonically balanced, as they are electrically serially wired. Yet, they have different absorption and exciton trapping characteristics. Fluctuations in light quality and/or intensity can therefore readily lead to preferential excitation of one PS over the other, compromising the efficiency of energy conversion. Such unwanted situations are prevented by a rapid, redox-controlled process, called ‘state transitions’, which adjusts the amount of excitation energy delivered to the reaction centers of the two PSs. This process may be followed by a longer-term adaptation response that ultimately leads to changes in the stoichiometry of the photosystems (reviewed in Anderson, 1999; Pfannschmidt, 2003; Dietzel et al., 2008).
A breakthrough in our understanding of the mechanism of state transitions in higher plants came from the studies of Bennett and co-workers, who showed that state transitions involve phosphorylation and dephosphorylation of (as it turned out) a threonine residue located within the stromal-exposed N-terminal segment of the LHCII proteins Lhcb1 and Lhcb2 (Bennett, 1977, 1979a,b, 1980). Subsequently, Allen, Horton and colleagues showed that LHCII phosphorylation, as well as dephosphorylation, are governed by the redox state of the PQ pool (Allen et al., 1981; Horton and Black, 1982; for a review see Aro and Ohad, 2003). Over-reduction of the PQ pool, which occurs when PSII is preferentially stimulated, leads to activation of thylakoid-associated kinase(s) that phosphorylate LHCII. Oxidation of the PQ pool, due to an over-stimulated PSI, deactivates the kinase, leading to dephosphorylation of LHCII by constitutively active thylakoid-bound phosphatase(s). Identifying the LHCII kinase(s) and phosphatase(s) turned out to be a difficult task, with the first likely candidates for the kinase (coined TAKs) detected only in 1999 (Snyders and Kohorn, 1999 and see Snyders and Kohorn, 2001). A more directly involved kinase (paralogous to STN8), called STN7 in plants and Stt7 in Chlamydomonas, was found a few years later, and is currently considered to be the primary effector of LHCII phosphorylation. However, it is not yet clear if this function is exerted directly or through downstream-acting kinase(s) (Depege et al., 2003; see also Bellafiore et al., 2005; Bonardi et al., 2005; Frenkel et al., 2007; Wagner et al., 2008; Lemeille et al., 2009; Pesaresi et al., 2009; Lemeille et al., 2010; for the strategy employed to screen for Chlamydomonas state transitions mutants, which eventually led to the discovery of Stt7/STN7, see Fleischmann et al., 1999; Kruse et al., 1999). Other studies revealed that activation of the LHCII kinase is initiated by binding of PQ(H2) to the Qo site of cyt. b6f, implicating the latter as the direct kinase effector (Lemaire et al., 1987; Wollman and Lemaire, 1988; Gal et al., 1990; Vener et al., 1997; Zito et al., 1999; Wollman, 2001). Reviews by Wollman (2001) and Lemeille and Rochaix (2010) elaborate on these studies as well as on the manner by which information received at the Qo site, which is located at the luminal side of cyt. b6f, may be transmitted to the catalytic domain of the kinase at the stromal face of the thylakoid membrane. Importantly, the aforementioned processes occur only at low light intensities: as the irradiance level increases, phosphorylation of LHCII decreases and eventually ceases completely (Rintamaki et al., 2000; Hou et al., 2003), potentially due to thioredoxin-mediated inactivation of the LHCII kinase (Depege et al., 2003; see also Rochaix, 2007; Lemeille and Rochaix, 2010). The arrest of state transitions at high light may signify the takeover of NPQ, which becomes progressively more dominant as the irradiance level increases. The identity of the LHCII phosphatase, called TAP38/PPH1, was revealed only last year (Pribil et al., 2010; Shapiguzov et al., 2010).
But how does phosphorylation (or dephosphorylation) of LHCII, which resides with PSII in the grana membranes, affect the amount of light energy delivered to PSI in the stroma lamellae? Studies conducted in the early 1980s suggested that the answer to this is as follows: upon phosphorylation, a fraction (∼ 20% in higher plants and ∼ 80% in Chlamydomonas) of the LHCII population detach from their cognate PSII complexes in the grana and migrate to non-appressed regions of the thylakoid membrane. Once there, the LHCII complexes functionally couple with PSI, increasing its absorption cross-section at the expense of that of PSII. Dephosphorylation of the complexes leads to their return to the grana stacks and re-association with PSII, reversing the distribution of excitation energy between the two PSs (reviewed in Staehelin and Arntzen, 1983). This scheme was verified by electron microscopy, spectroscopic, and biochemical studies (Kyle et al., 1983; Larsson et al., 1983; Black et al., 1986; Vallon et al., 1991; Delosme et al., 1996). A more recent study, employing mechanical fractionation, suggested that the association between (phosphorylated) LHCII and PSI during the transition to state II occurs primarily at the interface between the stacked and unstacked regions of the network, namely at the grana margins (Tikkanen et al., 2008). The authors also proposed that this association also relies on the movement of PSI–LHCI complexes from the stroma lamellae to these regions. The presence of PSI at or near the grana margins is supported by previous fragmentation and separation studies (reviewed in Albertsson, 2001) and by an AFM analysis, which revealed dense clusters of this complex at the granum–stroma interface (Kaftan et al., 2002).
Two models have been proposed to account for the migration of phosphorylated LHCII from the grana to the stroma lamellae. The first model rationalized the migration in the framework of the surface charge model: the extra charge added by the phosphate group to the stromal-exposed segment of LHCII repels negative charges present on PSII and decreases membrane stacking, leading to the detachment of LHCII from PSII and its diffusion to non-appressed regions of the network (Barber, 1982). Such a mechanism, however, is at odds with the observation that phosphorylated LHCII complexes remain associated with PSII in the stacked membrane domains in a Chlamydomonas mutant lacking PSI centers (Delosme et al., 1996), indicating that changes in surface charge density per se are insufficient to cause the migration of (phosphorylated) LHCII out of the grana. On the other hand, the above finding fits well with an alternative model, put forth by Allen (1990, 1992a,b), who proposed that the forces driving LHCII migration are of a more subtle and specific nature. According to this model, called the ‘molecular recognition’ model, LHCII is able to bind specifically not just to PSII but also to PSI. Phosphorylation of LHCII was presumed to lead to a conformational change that lowers its affinity to PSII and at the same time increases its affinity to PSI. Consistent with this, it was subsequently shown that a conformational change is induced at the N-terminal segment of LHCII following phosphorylation (Nilsson et al., 1997). Thus, LHCII has two bound states, with PSI or with PSII, which are in dynamic equilibrium governed by its phosphorylation.
The next question was, naturally, where does LHCII bind to on the PSI complex? This was clarified in an elegant study by Lunde et al. (2000), who investigated the ability of transgenic Arabidopsis plants lacking specific PSI subunits to perform state transitions. Their results indicated that binding of LHCII to PSI is mediated by the PsaH subunit. (Plants deficient in the PSI PsaL subunit were also impaired in state transitions, but the effect was proven to be secondary, since PsaL connects PsaH to the PSI core.) Consistent with this, the PsaH, -I and -L subunits were implicated in binding of LHCII to PSI by cross-linking experiments (Zhang and Scheller, 2004). As evident from the PSI–LHCI crystal structure shown in Figure 4(A), the region containing these subunits is found at the tip of the core complex, which is located opposite the region facing the LHCI cluster, rationalizing how PSI can simultaneously associate with both types of antenna complex. Such LHCI–PSI–LHCII assemblies were visualized a year later by a single-particle electron microscopy analysis of complexes derived from Arabidopsis (Kouřil et al., 2005), which showed an extra density at the PsaH/L/A/K side of the PSI core complex, corresponding to a bound LHCII trimer. This is consistent with the model proposed by Amunts and Nelson based on the superposition of LHCII on the 3.4-Å crystal structure of pea PSI (Amunts et al., 2007). In Chlamydomonas, however, binding of trimeric LHCII to LHCI–PSI in state II has not been ascertained (Kouřil et al., 2011). Analysis of LHCI–PSI complexes in this state revealed a density which could potentially accommodate three LHCII monomers, but due to its shape, not an LHCII trimer (Germano et al., 2002). In another study, it was shown (biochemically) that a hyper-phosphorylated form of the minor monomeric Lhcb protein CP29 binds to LHCI–PSI in state II (Kargul et al., 2005). The authors proposed that the extra density near PsaH, as observed in the electron microscope projection maps of these complexes in state II, corresponds to phospho-CP29, which could act in the coupling of LHCII to PSI. Subsequently, it was revealed that CP26 and an LHCII type II protein called LhcbM5 also associate with PSI in state II-adapted Chlamydomonas cells (Takahashi et al., 2006). The model proposed in this study holds that, following phosphorylation, CP26, CP29, and LhcbM5 migrate from PSII to PSI and serve as docking sites for the binding of two LHCII trimers (rather than one, as suggested in plants). Utilizing RNA interference (RNAi), it was more recently shown that CP29 (and not CP26) is essential for association of LHCII with PSI in Chlamydomonas cells adapted to state II (Tokutsu et al., 2009).
Notably, PsaH is absent from cyanobacteria and red algae, which, as mentioned above, do not rely on LHC-like complexes for peripheral light harvesting. In these organisms, PSI assembles into trimeric complexes. Ben-Shem and Nelson showed that, by virtue of its location, PsaH prevents similar assembly of plant PSI monomers (Ben-Shem et al., 2003). They also found that the C-terminus of the PsaL subunit, which facilitates trimer formation in cyanobacteria, is truncated in plants. Based on these observations, they proposed that PSI trimerization was lost in plants to enable binding of LHCII.
As discussed above, in addition to phosphorylation of the LHCII antenna complex, phosphorylation of PSII subunits also takes place (Rintamaki et al., 1997; Vener, 2007), and has been shown to control stacking and folding of the thylakoid membranes (Fristedt et al., 2009, 2010). In a recent study, the connection between phosphorylation, PSII–LHCII supercomplex assembly and state transitions was investigated (Dietzel et al., 2011). It was shown that phosphorylation of the PSII CP43 core subunit (which is the most prominently phosphorylated protein of PSII) results in disassembly of PSII complexes, probably due to the introduction of negative charges. These events were found to precede and to be independent of state transitions. The latter was concluded from the fact that wild-type-like remodeling of PSII was observed in a mutant deficient in the LHCII kinase STN7 and, therefore, is permanently locked in state I. In addition, state transitions were observed to be significantly (three-fold) accelerated in a mutant lacking the PSII subunit Psb27, which is required for formation of the PSII supercomplex. From these findings, the authors deduced that release of PSII–LHCII supercomplexes facilitates accessibility of the kinase to the granal-localized LHCII antenna complexes, and is thus essential for their phosphorylation and subsequent movement out of the granum core (Figure 5). CP43 is one of the PSII proteins (along with D1, D2, and PsbH) whose phosphorylation depends on STN8 (Bonardi et al., 2005; Vainonen et al., 2005; Reiland et al., 2011). However, in stn8, CP43 phosphorylation persists (Dietzel et al., 2011), perhaps by STN7, which is known to have some substrate overlap, or by another as yet unidentified kinase. Interestingly, in the stn7 mutant, light (quality)-dependent, STN8-mediated phosphorylation of PSII is abolished. The activity of the Chlamydomonas ortholog of STN8 likewise depends on Stt7 (Lemeille et al., 2010). It thus seems that the regulation of photosynthetic electron transport, in both plants and green algae, involves a complex network of phosphorylation and de-phosphorylation events, which are tightly linked to the redox state of the PQ pool and cyt. b6f activity, and are subservient to additional control by the ferredoxin–thioredoxin system. For a detailed discussion of this topic, we refer the reader to the following reviews: Rochaix (2007, 2011), Lemeille and Rochaix (2010) and Vener (2007).
Recently, Iwai et al. (2010b) employed (chlorophyll) fluorescence lifetime imaging microscopy (FLIM) to follow the dissociation of LHCII from PSII during the transition to state II in live Chlamydomonas cells. During this transition, a shift of the mean fluorescence lifetime from 170 ps, which is attributed to PSII-bound LHCII, to 250 ps was observed. The latter was ascribed to detached LHCII complexes, as it predominated in a mutant lacking both PSs, and was absent from the stt7 mutant. The rise of the 250 ps component was also shown to correlate with the rise of phospho-LHCII. Notably, the dissociated LHCII complexes spread out and formed spotted regions in state II, suggesting that they form aggregates, an observation that was supported by biochemical analyses. Based on these and previous findings obtained from Chlamydomonas, the authors proposed a model for the remodeling of PSI and PSII during state transitions (Iwai et al., 2010b; Minagawa, 2011). In this model, phosphorylation of major LHCII proteins triggers division of PSII megacomplexes into supercomplexes; thereafter, phosphorylation of the minor LHCII proteins CP26 and CP29, as well as of the PSII core subunits D2 and CP43, induces displacement of LHCII from the PSII core complex. Aggregation of some of the detached phospho-LHCII complexes are suggested to serve in energy dissipation, whereas others attach to PSI–LHCI, increasing its absorption cross-section. Thus, state transitions may also have a preparatory role in non-photochemical quenching, as proposed earlier (Tikkanen et al., 2008). In Chlamydomonas, transition to state II is accompanied by a switch to cyclic electron transport (Finazzi et al., 1999, 2002), which was proposed to restore ATP levels (Bulte et al., 1990), implicating state transitions as an important controller of ATP homeostasis in this organism. [Such control is probably also present in higher plants (see next section ‘What are grana good for?’), but there it is apparently not coupled to state transitions.] Recently, Minagawa’s group, which performed the aforementioned FLIM measurements, was able to perform the formidable task of isolating the supercomplex that drives cyclic electron transport in Chlamydomonas (Iwai et al., 2010a). In another study, the involvement of STN8 in the regulation of the transition from cyclic to linear electron flow in dicots upon the activation of photosynthesis was revealed following comparative phosphoproteome profiling of wild-type and stn8 Arabidopsis (Reiland et al., 2011).
The changes in surface charge density, macroscopic organization, and distribution of PSII and LHCII during state transitions are expected to lead to structural alterations of the thylakoid network. An early freeze-fracture electron microscopy study suggested that the changes in membrane architecture associated with the state I → state II transition in higher plants are limited to about a 20% decrease in the number of stacked membranes, which was proposed to be achieved by membrane flow between grana and stroma thylakoids (Kyle et al., 1983). However, in a recent study, combining atomic force, scanning electron, and transmission electron microscopy, we observed that the transition to state II is accompanied by other, more significant structural rearrangements (Chuartzman et al., 2008). These included, in addition to membrane unstacking, displacement of layers in the stacks, resulting in a staggered appearance, as well as swelling of the thylakoids, which also became wavy – suggesting a large increase in membrane deformability. We proposed that the rearrangements are initiated at the highly curved regions of the grana margins and are then transmitted to the non-appressed regions of the network, making it less ordered and more expanded. In the work of Dietzel et al. (2011), rearrangements were observed that included loosely appressed, deformed granum structures and more stroma lamellae in state II, which were notably absent in the stn7 mutant. The structural rearrangements of the network in this state result in an increase of the accessible surface area of the granal thylakoids, and thus may facilitate phosphorylation of LHCII and its migration out of the grana. We also proposed that the alterations in the structure of the granum–stroma assembly necessitate breakage and re-formation of some of the membrane bridges that adjoin neighboring layers in the stacks (Chuartzman et al., 2008). Remarking on the latter, Austin and Staehelin (2011) recently noted that no ruptured thylakoids were observed in an early freeze-fracture study performed by Kyle et al. (1983). We completely agree with these authors that the transition to state II does not lead to rupturing of the membrane system. Such ruptures, which would of course be detrimental, were not observed in our images either. Exactly because of this, we proposed that the aforementioned alterations in membrane connectivity reflect fission and fusion events, which are general phenomena of membranous systems in the cell. These may be mediated by an auxiliary membrane remodeling system, like the ones operating in the mitochondria, endoplasmic reticulum (ER) or the Golgi apparatus. Alternatively, they may occur spontaneously due to changes in the structure of the grana margins upon shuttling of LHCII and PSI complexes (see previous section ‘The forces that govern granum formation and partitioning of the thylakoid network’).
What are grana good for?
Grana had probably emerged as part of the changes that ultimately enabled plants to inhabit various terrestrial niches, an endeavor that exposed them to new, highly variable light environments (Anderson, 1999; Mullineaux and Emlyn-Jones, 2005; Nelson and Ben-Shem, 2005; Amunts and Nelson, 2009; Nevo et al., 2009). Of these changes, the most important was probably the substitution of the bulky, water-soluble phycobilisomes, which sterically obstruct membrane appression, with LHCII, which both allows for and facilitates tight stacking of the membranes (Horton, 1999; Chow et al., 2005; Nevo et al., 2009). Fine-tuning of the size (diameter) and appression of the membranes in the stacks and, therefore, of their activity, was subsequently acquired through the introduction of multiple phosphorylation sites into LHCII and PSII. But what advantages do grana actually confer that make them so ubiquitous in vascular plants?
The first proposal was made by Andersson and Anderson (1980) in their seminal paper, which decisively demonstrated that PSII/LHCII and PSI are segregated into appressed and non-appressed regions of the thylakoid membrane. They noted that this segregation could act to minimize the spillover of excitation energy from PSII to PSI, which constitutes a deeper energy trap (see also Anderson, 1981, 1982a). A related hypothesis, made by Trissl and Wilhelm (1993), was that physical separation of the two PSs would negate asynchronies in photon processing by PSII and the more vigorous PSI, in which exciton trapping proceeds about three times faster. As already discussed, balanced activity of the two PSs is further warranted by the shuttling of LHCII complexes between appressed and non-appressed membrane domains during state transitions. The dense packing of PSII and LHCII in the appressed grana domains would also maximize their connectivity (and, therefore, energy trapping by PSII) along the membranes and, possibly, across the partition gap between neighboring stacked layers (Trissl et al., 1987; Trissl and Wilhelm, 1993; but see Kirchhoff et al., 2004). The additional pathway for exciton transfer offered by the latter process could offer two advantages: (i) it may reduce the protein density required for maximal harvesting of light energy, thereby relieving constraints on the diffusion of PQ molecules within the grana membranes (Mullineaux, 2005) and (ii) it may compensate for the loss of dimensionality associated with the substitution of phycobilisomes by LHCII as the primary accessory antenna, maximizing the efficiency of capture and transfer of light energy (Armond et al., 1976; Boardman et al., 1978; Horton, 1999). This may be particularly important for the ability of land plants to thrive in a shaded environment (Boardman et al., 1978; Mullineaux and Emlyn-Jones, 2005), a need that was probably not encountered by their immediate progenitors – presumably a kind of alga that inhabited shallow waters. Consistent with this, grana in plants grown in the shade or at low light contain more (sometimes, considerably more) thylakoid layers than those found in plants grown at moderate or high light intensities (Anderson, 1999, and references therein).
The appression of membranes in the grana also enables tight control over the accessibility of PSII and LHCII to stromal or membrane-associated enzymes and, by virtue of this, control of their macroscopic organization and activity. Examples of such proteins are the thylakoid-associated kinases and phosphatases, proteases involved in the degradation of photodamaged PSII subunits, and enzymes involved in the xanthophyll cycle, which plays an important role in the establishment of the quenched (energy dissipative) state of LHCII during NPQ (Jahns et al., 2009). A role for grana in the control of this process has been proposed (Horton, 1999). Likewise, it was suggested that confinement of PSII to the grana may prevent overloading of its repair system under conditions of excessive photoinhibition by limiting the accessibility of damaged complexes to the proteases that degrade the D1 protein (Anderson and Aro, 1994). Due to their unique location and properties, the grana margins probably play an important role in the above processes. These regions probably also serve as the business end of the molecular and morphological reorganizations that occur during state transitions (Chuartzman et al., 2008; Tikkanen et al., 2008), and may accommodate those involved in the transitions between linear and cyclic electron transport (Tikkanen et al., 2008). They should therefore be regarded as a distinct entity of the thylakoid membrane (Albertsson et al., 1991; Wollenberger et al., 1994; Albertsson, 2001), which enables rearrangements in PSII–LHCII macrostructure to be readily transformed into changes in the activity and connectivity of the two PSs and, therefore, in electron transport rates and modes.
It is quite impossible (and, perhaps, unnecessary) to determine which of the aforementioned potential benefits had been the primary driving force behind grana formation. Along the course of evolution, various functions were gradually gained, concomitant with the acquisition of more refined control mechanisms. The segregation of the thylakoid membranes into granal and stromal lamellar domains could thus be looked upon as yet another manifestation of the tendency of evolution to utilize compartmentalization as a means of avoiding spurious interactions, and of achieving better and more dynamic control over cellular functions and responses (Anderson et al., 1988; Anderson, 1999).
We are indebted to Egbert J. Boekema (University of Groningen), Nathan Nelson (Tel-Aviv University), and Alexander Ruban (Queen Mary University of London) for kindly providing figures used in the review. We also wish to thank Ruti Kapon, Eyal Shimoni and Shlomi Dagan (Weizmann Institute of Science) for critically reading the manuscript. Financial support to ZR, from the Minerva Foundation, the F.I.R.S.T. Program of the Israel Science Foundation (1282/09), and the United State–Israel Binational Agricultural Research and Development (US-4334-10) Fund, is greatly acknowledged.