Dimers of light-harvesting complex 2 from Rhodobacter sphaeroides characterized in reconstituted 2D crystals with atomic force microscopy Lu-Ning Liu 1,2 , Thijs J. Aartsma 1 and Raoul N. Frese 1 1 Huygens Laboratory, Department of Biophysics, Leiden University, The Netherlands 2 State Key Lab of Microbial Technology, Shandong University, Jinan, China Photosynthetic bacteria use a large part of their internal volume for functionalized invaginations of the intracyto- plasmic membrane containing the photosynthetic machinery. The most abundant protein complexes in the intracytoplasmic membrane are light-harvesting (LH) complexes, responsible for the absorption of sun- light and subsequent excited state energy transfer, and reaction centers (RCs), to which the energy is directed to initiate the primary charge transfer reactions [1,2]. There are various photosynthetic purple bacterial spe- cies that display high similarity between the molecular structures of the individual photosynthetic protein complexes [3–9]. Nevertheless, differences exist: purple bacteria assemble a variety of photosynthetic unit architectures, the simplest being an array of RC–LH complex 1 (LH1) core complexes, as seen in Blasto- chloris viridis and Rhodospirillum rubrum [10], whereas other species, e.g. Rhodobacter sphaeroides, synthesize Keywords 2D crystallization; atomic force microscopy; light-harvesting complex 2; polymorphism; Rhodobacter sphaeroides Correspondence R. N. Frese, Huygens Laboratory, Biophysics Department, Leiden University, 2333CA Leiden, The Netherlands Fax: +31 (0)71 527 5936 Tel: +31 (0)71 527 5970 E-mail: frese@physics.leidenuniv.nl (Received 26 February 2008, revised 28 March 2008, accepted 16 April 2008) doi:10.1111/j.1742-4658.2008.06469.x Microscopic and light spectroscopic investigations on the supramolecular architecture of bacterial photosynthetic membranes have revealed the pho- tosynthetic protein complexes to be arranged in a densely packed energy- transducing network. Protein packing may play a determining role in the formation of functional photosynthetic domains and membrane curvature. To further investigate in detail the packing effects of like-protein photosyn- thetic complexes, we report an atomic force microscopy investigation on artificially created 2D crystals of the peripheral photosynthetic light-har- vesting complexes 2 (LH2’s) from the bacterium Rhodobacter sphaeroides. Instead of the usually observed one or two different crystallization lattices for one specific preparation protocol, we find seven different packing lat- tices. The most abundant crystal types all show a tilting of LH2. Most sur- prisingly, although LH2 is a monomeric protein complex in vivo, we find an LH2 dimer packing motif. We further characterize two different dimer configurations: in type 1, the LH2’s are tilted inwards, and in type 2, they are titlted outwards. Closer inspection of the lattices surrounding the LH2 dimers indicates their close resemblance to those LH2’s that constitute a lattice of zig-zagging LH2’s. In addition, analyses of the tilt of the LH2’s within the zig-zag lattice and that observed within the dimers corroborate their similar packing motifs. The type 2 dimer configuration exhibits a tilt that, in the absence of up-down packing, could bend the lipid bilayer, leading to the strong curvature of the LH2 domains as observed in Rhodobacter sphaeroides photosynthetic membranes in vivo. Abbreviations AFM, atomic force microscopy; DDM, dodecyl-b- D-maltoside; DMPC, dimyristoyl phosphatidylcholine; DOPC, dioleoyl phosphatidylcholine; DOTM, dodecyl-b- D-thiomaltoside; DPPC, dipalmitoyl-phosphatidylcholine; LDAO, N,N-dimethyldodecylamine-N-oxide; LH, light-harvesting; LH1, light-harvesting complex 1; LH2, light-harvesting complex 2; LPR, lipid ⁄ protein ratio; OG, octyl-b-glucopyranoside; OTG, n-octyl-b- D- thioglucopyranoside; PC, phosphatidylcholine; RC, reaction center. FEBS Journal 275 (2008) 3157–3166 ª 2008 The Authors Journal compilation ª 2008 FEBS 3157 peripheral LH complexes, LH complexes 2 (LH2’s), or configure RC–LH1 complexes into dimeric supercom- plexes [11]. Moreover, the structures of specific photo- synthetic protein complexes may be variable. This is exemplified by the 3D crystallographic structures of LH2’s from the species Rhodopseudomonas acidophila and Rhodospirillum molischianum [6,7]. Both species synthesize LH2’s from repeating a-helical protein units that are cylindrically arranged in a ring-like structure, but Rhodop. acidophila complexes display nine-fold symmetry and Rhodos. molischianum complexes display eight-fold symmetry. Also, the exact arrangement of the light-interacting chromophores within the protein scaffold differs between these species. Finally, the shape of the photosynthetic membranes is highly spe- cies-dependent: B. viridis membranes are large, flat sheets, Rhodos. molischianum membranes are stacked thylakoids, and Rhodob. sphaeroides membranes con- tain bud-like chromatophores [12]. The photosynthetic bacterium Rhodob. sphaeroides is one of the few purple bacterial species that is amenable to genetic manipulation, which facilitates a study of the interdependence of membrane organization, membrane shape, protein structure and protein composition. Dif- ferent types of LH antenna complexes and RCs from Rhodob. sphaeroides have been structurally analyzed by X-ray crystallography, cryoelectron microscopy, and atomic force microscopy (AFM) [3,13–16]. Recent advances in AFM imaging and polarized spectroscopy, utilizing these structural models, has revealed the molecular architecture of native Rhodob. sphaeroides membranes [17,18]. In all cases, images revealed close proximity of the photosynthetic components, which thus comprise a densely packed energy-transferring net- work. Polarized light-spectroscopic measurements on intact membranes revealed remarkable homology in supramolecular organization in Rhodob. sphaeroides membranes [19]. Monte Carlo simulations, assessing the effect of the differences in size and shape of the protein complexes, showed the importance of protein- packing effects on the formation of like-protein domains, membrane curvature, and the creation of dif- fusive pathways within crowded membranes [18]. Packing effects of like proteins can be most clearly observed in artificially created 2D crystals [20]. Such crystals are formed from detergent-solubilized proteins mixed with lipids by gradually removing the detergent. There are essentially two main methods for removing the detergent for 2D crystallization: flow dialysis and bio-beads. Other variables involve the types of lipids or mixtures of lipids used and the lipid ⁄ protein ratio (LPR). Two-dimensional crystals of the photosynthetic bacterial LH2 have been extensively studied by means of AFM [13,14,21–24]. Table 1 summarizes the method used and the crystal lattices from different species observed with AFM. It was found that the morphol- ogy of the 2D crystals and the LH2 arrangement are highly dependent on the crystallization conditions, including the lipid, detergent scavenger, protein con- centration, LPR, and species used. Creating 2D crys- tals of LH2 by flow dialysis in the presence of dioleoyl phosphatidylcholine (DOPC) as lipid has been shown to produce highly structured tubular crystals, although these may contain different packing lattices [13]. With Table 1. Summary of AFM data from 2D crystals of LH2. DMPC, dimyristoyl phosphatidylcholine; OG, octyl-b-glucopyranoside; DPPC, dipal- mitoyl-phosphatidylcholine; DDM, dodecyl-b- D-maltoside; DOTM, dodecyl-b-D-thiomaltoside. Species Lipid(s) Detergent Method Lattice type Rubrivivax gelatinosus [21] Egg PC OTG Bio-beads Square Rhodobacter sphaeroides [14] DOPC OTG Bio-beads Square 1 (90°) Rhodopseudomonas acidophila [22] Egg PC LDAO Dialysis Disordered Rhodobacter sphaeroides [13] DOPC OG Dialysis Square 1 (90°) Zig-zag 2 (90°) Disordered Rhodopseudomonas acidophila [23] DMPC OTG Bio-beads Square 1 (90°) Zig-zag 2 (90°) Disordered Rhodobacter sphaeroides [24] DOPC ⁄ DPPC OG ⁄ DDM ⁄ DOTM Rinse Disordered Rhodobacter sphaeroides (current) Egg PC OTG Bio-beads Square 1 (90°) Square 2 (60°) Zig-zag 1 (120°) Zig-zag 2 (90°) Dimer type 1 Dimer type 2 Disordered AFM study on LH2 2D crystal L N. Liu et al. 3158 FEBS Journal 275 (2008) 3157–3166 ª 2008 The Authors Journal compilation ª 2008 FEBS the use of bio-beads and DOPC, the more common membrane vesicles are obtained, which have been reported to contain only one type of packing motif [14]. Chami et al. [25] found that n-octyl- b-d- thioglucopyranoside (OTG) can greatly increase the crystal size. Detergent removal using bio-beads can be achieved at a much higher rate than with dialysis [26]. On the basis of AFM imaging of both the periplasmic and cytoplasmic sides of LH2’s after reconstitution, ellipticity and tilt of the LH2’s have been reported [14,21,23]. A comparison between the different packing configurations found in 2D crystals of LH2’s from Rhodop. acidophila showed for one specific LPR a square-packing lattice consisting of LH2’s that were tilted relative to the lipid-membrane plane, and for another LPR a lattice of zig-zagging nontilted LH2’s [23]. In that study, it was concluded that the observed tilt of LH2’s within the square lattice was not due to an intrinsic property of LH2’s of Rhodop. acidophila, but was caused by specific interactions induced by packing. In contrast, tilted LH2’s and RC–LH1’s have been observed in native membranes of Rhodob. sphaeroides [17]. Moreover, in the aforementioned model of the photosynthetic membrane of this particu- lar species, we showed the importance of protein pack- ing for the full appearance of the membrane [18]. More specifically, we showed that the formation of LH2 domains has a strong influence on the curvature of the membrane. In the absence of a 3D crystallo- graphic structure, we assigned an intrinsic curvature to the LH2 of Rhodob. sphaeroides that could originate from a slight conical shape or specific binding of a curved lipid. In any case, a tilted configuration in packed conditions could ultimately lead to membrane curvature. Here we further investigate in detail the packing effects in artificially created 2D crystals of the LH2 from Rhodob. sphaeroides . Our results show that with egg phosphatidylcholine (PC) as lipid and bio-beads as detergent scavenger, a multitude of different packing configurations can be obtained within one preparation. AFM images allow the differences in protein packing and interaction within the membrane to be visualized. Within the packed lattices, we find a new conforma- tion of LH2’s, namely a dimeric configuration. Detailed investigations of the observed LH2 packing patterns reveal that the different packing lattices con- sist of similarly interacting LH2’s. The dimeric LH2 configuration is very likely to exist as an intermediate packing configuration. Furthermore, we find that LH-2, in all cases, show a tilted conformation. One of the two types of LH2 dimer configurations is shown to possess a tilt that could bend the lipid bilayer, leading to the bud-like membrane curvature as observed for Rhodob. sphaeroides in vivo. Finally, on the basis of these images, schematic models of possible LH2 arrangements and protein–protein contacts in the reconstituted 2D crystal membranes are proposed. Results Two-dimensional crystals of Rhodob. sphaeroides LH-2s were prepared according to the protocol of Rigaud et al. [26], involving the addition of OTG for LH2 solubilization and detergent removal by bio-beads. This method has been applied successfully for the 2D crystallization of LH2’s from a range of photosynthetic bacterial species, as shown in Table 1. In contrast to other studies, here we utilized egg PC (Sigma, St Louis, MO, USA) as lipid in combination with Rhodob. sphaeroides LH2’s. A range of prepara- tions were examined with different LPRs (0.35, 0.4, 0.45, 0.5, 0.55, 0.6). All preparations were shown to form large vesicles as determined by electron microscopy (data not shown). For the AFM imaging, all samples were prepared with LPR = 0.5. Our imaging method has been described before [13,17]. In essence, ultrasoft tapping-mode AFM was applied in combination with the choice of appropriate buffer for electrostatic balancing of the AFM tip and the substrate. Electrostatically balanced tapping-mode AFM was shown to produce the highest-resolution images of naturally curved membranes [17], and was applied here to obtain detailed information on possible tilts of the protein complexes above the membranes. Figure 1 shows typical 2D crystals of densely packed LH2’s as observed with AFM. The protruding mass appears most bright; the mica surface is dark. Even at this low magnification, the LH2’s were visible as rings, about 7 nm wide. Crystals had a diameter of up to 2 lm and a height of 6.5 ± 1.0 nm (n = 20). Areas of empty lipid bilayer without incorporated LH2’s had an average height of 4.0 ± 0.7 nm (n = 20). Already at the low magnification of Fig. 1, the regular arrangement of LH2’s and the varieties thereof could be observed. It is noteworthy that several different LH2 arrangements could coexist in the same membrane fragment. We found no less than seven different packing arrangements in terms of the different lateral arrangements and tilts of LH2’s. There were two rectangular arrays, two zig-zag lattices, two types of dimeric organization, and a disordered arrangement. The majority of crystalline lattices were occupied by so-called ‘zig-zag’ rows of LH2’s ( 80%). The other 20% contained crystallized arrangements of LH2’s in a square pattern and with a L N. Liu et al. AFM study on LH2 2D crystal FEBS Journal 275 (2008) 3157–3166 ª 2008 The Authors Journal compilation ª 2008 FEBS 3159 disordered distribution. Most surprisingly, a dimeric LH2 configuration could be observed, as shown in Fig. 1 (areas 4 and 6). Regardless of the protein-packing arrangement, the LH-2s were clearly resolved in all these types of periodicities, with outer and inner diameters of 6.5 ± 0.5 nm and 3.2 ± 0.3 nm (n = 50), respectively, in agreement with previous descriptions [7,13,14,21]. High-resolution AFM topographs showing the ‘zig- zag’ lattices in close detail are shown in Fig. 2. Differ- ent angles between adjacent strongly protruding LH2 rings, approximately 120° (Fig. 2A) and 90° (Fig. 2B), indicated two types of zig-zag lattice. The inset of Fig. 2A shows a high-magnification image of LH2 rings embedded in the 2D crystals. Within this zig-zag lattice, we were able to visualize the weakly protruding LH2’s. These lower LH2’s had an average height of  4.5 A ˚ (n = 20) above the lipid bilayer, whereas the strongly protruding LH2’s had an average height of  9.9 A ˚ (n = 30). On the basis of the spacing between LH2 rows measured in this work and previous estima- tions [27], the arrangements of up-LH2’s and down- LH2’s in the zig-zag lattices are schematically illustrated as green and turquoise circles. Twenty per cent of crystal lattices in the preparation were found to occupy the square arrangement of LH-2s, as shown in Fig. 3. We recorded two different square lattices with a variation of angles between the progressing lines of protruding LH2’s: approximately 90° (Fig. 3A) and 60° (Fig. 3B). The former pattern has been well described earlier in Rhodob. sphaeroides [14], whereas the latter array is observed for the first time. The 60° square lattice consists of four strongly protruding LH2’s arranged in rhombic periodicity. The spacing within this domain could fit for two weakly protruding LH2’s, which could be occasionally recog- nized (Fig. 3B, arrows). These down-LH2’s could form a hexagonal motif surrounding a central up-LH-2. These available results enabled the arrangements of up-LH2’s and down-LH2’s in these lattices to be sche- matically depicted in Fig. 3. We also found a novel arrangement of LH2 crystal packing, termed a dimer lattice. Figure 4 shows high- magnification images of the areas containing dimers; the larger lattice of origin is indicated in Fig. 1 (area 4). LH2’s were found to be aligned, forming rows along the long axis of the dimers within a period- ical lattice. Figure 4A shows an area where two LH2’s contact closely, separated from neighboring LH2–LH2 dimers. Adjacent rows of dimers are separated such that the dimeric LH2 in the neighboring row faces the empty central space. The distance between adjacent rows of dimers was found to be 5.3 ± 0.4 nm (n = 20), less than the size of the LH2. Figure 4B,C presents two different dimer lattices (type 1 and type 2) with opposing lines of progress. The directions of the progressing lines were related to that of the surround- ing zig-zag lattices. This will be discussed below. The inset of Fig. 4A shows a scheme of the packing config- uration of dimers. Such a specific arrangement is exemplified in the inset of Fig. 4B, in which the up and down configuration of LH2’s can be viewed. Here, the less protruding LH2’s are visualized, showing their location to be precisely within the gap between two adjacent dimers. Discussion Dimeric LH2 Surprisingly, we observed dimeric configurations of LH-2. Native membranes from four different LH2- containing photosynthetic bacteria, including Rhodob. sphaeroides, have been imaged by AFM before [17,28– 32]. In all cases, no sign of LH2 dimers has been reported, although LH1–RC complexes are mainly arranged in rows of dimers in Rhodob. sphaeroides.An AB C Fig. 1. Overview of typical crystals imaged with AFM (raw images). Areas of different crystal lattices are indicated by dashed lines. (A) Two types of zig-zag (areas 1 and 2), disordered (area 3) and dimer (area 4) lattices. (B) Zig-zag lattice (area 2), square lattice (area 5) and dimers (area 6). (C) Zig-zag lattice (area 1) and dimers (area 6). Scale bars: (A) 100 nm; (B) 50 nm; (C) 200 nm. AFM study on LH2 2D crystal L N. Liu et al. 3160 FEBS Journal 275 (2008) 3157–3166 ª 2008 The Authors Journal compilation ª 2008 FEBS important clue about the origin of the dimers can be found upon close inspection of the crystal lattices. LH2 dimers are always surrounded by zig-zag lattices (Fig. 4). With respect to these surrounding zig-zag lattices, we found two different dimer morphologies (Fig. 5A,B). When we represented a zig-zag line as progressing Vs (or VVVV), where the corners of the V are occupied by LH2’s, we found dimers progressing along both sides of the V. For instance, in Fig. 4B, the V-shaped zig-zag lines appear to correlate with dimers progressing along the \-side of the V (or zig-side), whereas in Fig. 4C, the dimers progress along the ⁄ -side of the V (or zag-side). It thus seems that the LH-2s associated with either the zig-side or the zag-side of zig-zag lines actually originate from LH2 dimers. More evidence for the dimeric origin of zig-zag lattices can be obtained from high-resolution AFM images of dimers and zig-zag LH2’s, as shown in Fig. 5. In dimeric LH2’s, there were two tilted types found with different protruding heights, type 1 dimers (Fig. 5A) and type 2 dimers (Fig. 5B). Type 1 dimers have high contacting sides and low peripheral sides (Fig. 4A), whereas type 2 dimers show an opposite tilt- ing orientation. The tilt of LH2 dimers was further studied by measuring the protruding profiles of each LH2 within the dimer on the basis of the height differences of both sides and the size of LH2’s (Fig. 5C,D). This allowed the tilting angles to be calcu- lated as 5.0 ± 0.5° and 3.5 ± 0.3° (n = 20), respec- tively. These two tilting angles bear a striking resemblance to that of the zig and zag LH2’s as shown in Fig. 5F,G. Depending on the directions along which LH2’s were measured, we found the same angles as for A B Fig. 2. Raw AFM images of the two different zig-zag lattice types. (A) Zig-zag lattice (type 1) with 120° angle: zoomed-in image of the areas indicated by the dashed line in Fig. 1A (area 1). Inset: 3D enhanced close view showing weakly protruding LH2’s (black arrows) in between the strongly protruding zig-zag lines (white arrows). (B) Zig-zag lattice (type 2) with 90° angle: zoomed-in image of the area indicated by the dashed line in Fig. 1B (area 2). Strongly and weakly protruding LH2’s are schematically illustrated as green and turquoise circles, respectively. Scale bars: (A) 10 nm; (B) 20 nm. A B Fig. 3. Raw AFM images of the two types of square-packing lat- tices observed. Strongly and weakly protruding LH2’s are schemati- cally illustrated as green and turquoise circles, respectively. (A) Square lattice (type 1) with 90° angle. Note that the schematic arrangement of LH2’s is according to earlier descriptions [14,27]. (B) Square lattice (type 2) with 60° angle (see text). Arrows indicate weakly protruding complexes. The rhombic domain formed by up-LH2’s and the potential hexagonal motif of down-LH2’s are shown. Scale bars: (A) 30 nm; (B) 20 nm. L N. Liu et al. AFM study on LH2 2D crystal FEBS Journal 275 (2008) 3157–3166 ª 2008 The Authors Journal compilation ª 2008 FEBS 3161 the two dimer types. In the absence of up-down pack- ing, such as in native membranes, the observed tilt could bend the lipid bilayer. Remarkably, the direction and degree of tilt of the type 2 dimer were consistent with the parameters that were used in modeling the packing-induced membrane curvature of native Rhod- ob. sphaeroides membrane buds [18]. To conclude, we found that the dimers within the two dimer lattices tilt similarly to the zig-zag LH2’s. Furthermore, the alignment of the dimers along their long axis coincides with their zig or zag LH2 counter- parts within the surrounding zig-zag lattices. These two observations strongly suggest that the zig-zag lattice is composed of LH2 dimers. Why LH2 is dimerized within this particular prepa- ration, whereas it is absent in native membranes, is unknown. The presence of dimers in these crystals may reflect a more dense packing condition. In mutant Rhodob. sphaeroides membranes where RC–LH1 dimerization was inhibited, we also observed a particu- lar RC–LH1 dimer effect [18]. There, we found the RC–LH1 monomers to be rotationally locked within one unique orientation in half of the cases. Rotational locking had been observed before, but only for RC–LH1 dimers, and not for monomers [33]. We could relate this effect to an increased packing strain within the mutant membranes induced by dense pack- ing. As the protein helices constituting LH2 and LH1 are similar, the dimerization that we observed here might just reflect another packing effect acting on these similar LH proteins. Packing lattices By means of AFM, we observed a multitude of differ- ent packing arrangements within one 2D crystal prepa- ration. In Table 1, we summarize our findings and those of previous published studies. Within the five separate AFM studies on 2D crystals of LH2’s published to date (two of which were on Rhodob. sphaeroides LH2), only one type of square (square 1 with 90° angle) and one type of zig-zag lattice (zig-zag 2 with 90° angle) have been reported. Here we find two types of square lattices, two types of zig-zag lattices and two dimer lattices, including the commonly found disordered ‘lattice’; no less than seven different packing configurations for LH2 (see Table 1). In any AFM study, a certain amount of selectivity cannot be avoided, as only those areas that reveal a protruding mass can be discussed. In this respect, possible rear- rangements due to the adhesion to the mica surface represent an intrinsic variability that is largely beyond the control of an experimenter. Nevertheless, unlike other studies on Rhodob. sphaeroides LH2, this study combined egg PC as lipid and bio-beads as detergent scavenger. Egg PC is a mixture of very similar lipids containing all different fatty acid side-chains. In contrast, the commonly used DOPC contains only one type of lipid. Here we speculate that the differences between the lipids in egg PC, although small, may induce different LH2–LH2 interactions. This effect may be further enhanced by the use of bio-beads, instead of the dialysis method, which does not uniformly remove detergent throughout a preparation. On the other hand, 2D crystals of LH2’s from Rubrivivax gelatinosus have been prepared using bio-beads and egg PC as well, and there only one type of crystal lattice was found [21]. The differences and variations in packing lattices might therefore possibly originate from the structural differences between the LH2’s from different species. Similarly, it has been documented that the LPR is a critical factor in packing LH2 from Rhodop. acidophila in either a square or a zig-zag lattice [23]. Our observation that different patterns of LH2 packing coexist within the same crystal preparation suggests that this is not the case for the Rhodob. sphaeroides LH2. On the basis of the observations shown in Figs 1–4, we represent five schematic models of LH2 arrange- ments in either up or down (and in one case unknown) configuration, without considering the tilt of LH2 in Fig. 6. The strongly protruding LH2’s, or up-LH2’s, protrude about 1.0 nm, in good agreement with the previous data obtained on Rhodob. sphaeroides and Ru. gelatinosus 2D crystals [13,14,21]. In the latter AB C Fig. 4. High-resolution AFM images of the dimer areas surrounded by zig-zag lattices. Images are 3D-enhanced for clarity. (A) Zoomed- in image of Fig. 1A (area 4). Inset: closer view of dimeric LH2’s. Strongly and weakly protruding LH2’s are schematically illustrated as green and turquoise circles, respectively. (B) Type 1 dimer con- figurations. Inset: zoomed-in images show strongly and weakly (indi- cated by white arrows) protruding LH2’s. ‘Dimer’ and ‘zig-zag’ LH2 lattices are indicated by white lines. (C) Type 2 dimer configurations. ‘Dimer’ and ‘zig-zag’ LH2 lattices are again indicated by white lines. Note the opposing lines of progress as compared to type 1. See text for details. Scale bars: (A) 30 nm; (B) 50 nm; (C) 30 nm. AFM study on LH2 2D crystal L N. Liu et al. 3162 FEBS Journal 275 (2008) 3157–3166 ª 2008 The Authors Journal compilation ª 2008 FEBS study, it was shown that the strongly protruding side was the periplasmic side of the complex by means of thermolysin digestion that only affected this face [21]. Also in Rhodob. sphaeroides native membranes, which preferentially showed their periplasmic face up, similar heights of LH2 protrusions above the membrane were measured [13,14]. We thus conclude that the strongly protruding LH2’s expose their periplasmic face. Here, we occasionally visualized configurations of lower-lying LH2’s. We found that these protruded 4.5 A ˚ above the membrane, in good agreement with the earlier reports of the protrusions of the down-LH2’s from Rhodob. sphaeroides, which represents the cytoplasmic face [13,14]. In contrast to the previously reported arrangements (Fig. 6A,D), we visualized a new square lattice (Fig. 6B), a new zig-zag lattice (Fig. 6C), and a novel organization, a dimer lattice (Fig. 6E). Actually, we observed zig-zag lines in  80% of the lattices, indicat- ing that there is a strong preference for LH2’s to constitute the zig-zag lines. The 90° lines of zig-zagging up-LH2’s and down-LH2’s (Fig. 6D) have been characterized before for both Rhodob. sphaeroides and Rhodop. acidophila [13,14,23]. The 120° zig-zag lattice could represent a more dense packing motif accom- plished by a translation of alternating rows of LH2’s (Fig. 6C). On the other hand, this lattice could also represent a hexagonal lattice of all up-LH2’s, with alternating lines of zig-zagging LH2’s significantly lowered due to adhesion to the mica surface. Such a lattice has been reported before to exist in LH2-only Rhodob. sphaeroides membranes, which contains all A B E 6.3 9.8 10.0 7.9 6.2 6.1 7.6 8.5 10.6 6.7 10.0 11.0 6.4 8.8 6.4 C D F G Fig. 5. Comparison between two types of dimer configuration (A, B) and the zig-zag lattice (E). (A) LH2 dimer type 1 isolated from the image shown in Fig. 4B. (B) LH2 dimer type 2 originating from Fig. 4C. (C, D) The two distinct height profiles of LH2 dimers, indicating the tilt (n = 20). (E) Zig-zag LH2 lattice isolated from Fig. 2B. (F, G) The height profiles of zig-zag LH2 with respect to two vertical directions (n = 20), indicated as E1 (\) and E2 ( ⁄ ), respectively in (E) (height in A ˚ ). Scale bars: (A, B) 5 nm; (E) 10 nm. L N. Liu et al. AFM study on LH2 2D crystal FEBS Journal 275 (2008) 3157–3166 ª 2008 The Authors Journal compilation ª 2008 FEBS 3163 up-LH2’s [27]. In that case, the significant lowering of alternating rows of LH2 implies the existence of specific interactions between the LH2’s constituting a zig-zag row. This study provides further evidence for this, as we found similar interactions between the LH2’s within the zig-zag lattices and those within the LH2 dimer configuration. In addition, we observed non-zig-zagging LH2’s within two different square- packing lattices (Fig. 6A,B). As seen in Fig. 6F, as compared to the zig-zag, 90° square-packing lattices represent a less packed configuration, indicating that dense packing may induce the specific interactions leading to zig-zagging LH2’s. The intermediate density of the dimer lattice, on the other hand, indicates that the dimer organization might be a transient configura- tion between 90° square lattice and others. Tilting of LH2 Our images also allow us to characterize in detail the heights and tilts of the LH2’s for all different crystal lattices. In contrast to the many differences regarding packing lattices and configurations discussed in the previous section, we found all LH2’s to be tilted on the periplasmic side similarly in all lattices. Similar tilts have been observed for Rhodob. sphaeroides LH2 before, packed within a square lattice [14]. In contrast, Gonc¸ al- ves et al. [23] reported that no tilt of LH2 from Rhodop. acidophila was observed in type 2 crystals (zig-zag), but a 4° tilt was observed in type 1 crystals (square). The similar tilts of Rhodob. sphaeroides LH2 in all lattices indicate that the packing density and the induced interactions among neighboring LH2’s are the predominant factors that drive the arrangements of LH2. In addition, it has been observed that tilting of LH2’s shows some dependency upon the packing den- sities of different lattices. The least packed, disordered organization presents the smallest tilt 1.3 ± 0.2° (n = 20), whereas the largest tilt, 6.5 ± 0.5° (n = 20), is observed in the zig-zag lattice, which also represents the most densely packed configuration. In conclusion, crystalline packing in a high number of configurations of LH2’s in 2D crystals has been resolved by AFM. We characterized no less than seven different LH2 lattices in only one specific preparation. All individual LH2’s of Rhodob. sphaeroides are tilted, depending upon the packing densities of LH2’s in the crystal lattices. We found a novel dimeric organization of LH2, and showed the close resemblance to the LH2’s that form the zig-zag lattice. Such a lattice has also been observed in LH2-only domains of adhered Rhodob. sphaeroides membrane patches [27], which in native conditions are spherically shaped [34]. One type of the dimers observed here displays a tilt capable of curving the membrane in such a manner, leading to the spherical domains as observed in intact LH2-con- taining Rhodob. sphaeroides membranes [17,18,34]. Although long-range curvature is strongly reduced in 2D crystals, the similarity in configuration of this dimer and that of the LH2 complexes forming the 11.4 10.8 12.8 7.0 6.0 12.9 7.0 7.0 7.15.3 10.9 Strongly protruding LH2 (up) Weakly protruding LH2 (down) 12.0 15.0 22 000 F A B C D E 20 000 21 000 18 367 ABCDE 21 208 21 224 20 408 19 047 19 000 18 000 LH2 density (µm –2 ) Fig. 6. Schematic models for the different lattice types of LH2 observed by AFM with measured distances in nanometers and angles (n = 20). (A, B) Square-packing lattices corresponding to the images from Fig. 3A,B. (C, D) Zig-zag lattices correspond- ing to Fig. 2A,B. (E) The dimer lattices from Fig. 4 (note: packing configurations for both lattices in Fig. 4 are the same). (F) Total LH-2 densities of the different lattices. AFM study on LH2 2D crystal L N. Liu et al. 3164 FEBS Journal 275 (2008) 3157–3166 ª 2008 The Authors Journal compilation ª 2008 FEBS curved domains in native Rhodob. sphaeroides mem- branes indicates the existence of specific, packing- induced interactions between LH2 and the lipid mem- brane of this species in vivo. Experimental procedures LH2 purification After 4 days of growth, Rhodob. sphaeroides wild-type cells were harvested and disrupted by sonication. Unbroken cells and cell debris were removed by centrifugation (10 000 g for 20 min). The supernatant containing chromatophores was pelleted by centrifugation at 265 000 g for 90 min at 4 °C, and then resuspended in 10 mm Tris buffer (pH 7.5) to A 850 nm = 200Æcm )1 . Mem- branes were solubilized with 1% (w ⁄ v) N,N-dimethyldode- cylamine-N-oxide (LDAO) (Fluka, Buchs, Switzerland) for 40 min. Insoluble material was removed by centrifugation for 10 min at 10 000 g. The supernatant was then incubated in 10 mm Tris buffer and 0.5% Triton, and centrifuged for 5 min at 10 000 g. The pellet was resus- pended in 10 mm Tris buffer (pH 7.5) and incubated with 1% LDAO for 20 min at room temperature. LH2’s were purified using a discontinuous sucrose density gradient of 1.2 m and 0.6 m sucrose in 10 mm Tris buffer, and centrifuged for 2 h at 300 000 g. Two-dimensional crystallization Purified LH2’s (20 mgÆmL )1 ) were mixed with a lipid buffer composed of 2.5 mgÆmL )1 egg PC (Sigma) and 20 m m OTG (Sigma) [26] to a final LPR of 0.5 (w ⁄ w). Detergent removal was performed through three additions of 5 mg of SM2 Bio-Beads (Bio-Rad, Hercules, CA, USA) [35,36]. After 2.5 h of stirring at room temperature, the reconsti- tuted material was stored at 4 °C for AFM analysis. AFM The AFM sample of LH2 crystals was prepared by adsorbing 2 lL of sample solution onto the surface of freshly cleaved mica covered with absorption buffer (10 mm Tris ⁄ HCl, pH 7.5, 150 mm KCl, 25 mm MgCl 2 ) for 60 min, and then carefully rinsing with recording buffer (10 mm Tris ⁄ HCl, pH 7.5, 150 mm KCl) in order to remove weakly bound crys- tal patches. Imaging was performed with a commercial AFM instrument (NanoscopeIII; Digital Instruments, Santa Bar- bara, CA, USA) and standard silicon nitride cantilevers with a length of 85 lm, a force constant of 0.5 NÆm )1 and operat- ing frequencies of 25–35 kHz (in liquid) (Veeco NanoProbe Tips, Santa Barbara, CA, USA) were used. High-resolution AFM images were obtained using tapping mode in liquid and with amplitude setpoint adjusted to minimal forces. Acknowledgements The authors thank Dre ´ de Wit for growing the Rhodob. sphaeroides cells and purifying the LH2. This research was sponsored by the Dutch Science Foundation [Netherlands Organization for Scientific Research (NWO)]. This project is part of the research programme ‘From Molecule to Cell’, funded by the NWO and the Foundation for Earth and Life Sciences (ALW). R. N. Frese gratefully acknowledges the NWO for a veni-grant. Lu-Ning Liu acknowledges financial support from a PhD Study-Abroad Scholar- ship of Shandong University (China). References 1 Cogdell RJ, Gall A & Ko ¨ hler J (2006) The architecture and function of the light-harvesting apparatus of purple bacteria: from single molecules to in vivo membranes. Q Rev Biophys 39, 227–324. 2 Fleming GR & van Grondelle R (1997) Femtosecond spectroscopy of photosynthetic light-harvesting systems. Curr Opin Struct Biol 7, 738–748. 3 Allen JP, Feher G, Yeates TO, Komiya H & Rees DC (1987) Structure of the reaction center from Rhodobact- er sphaeroides R-26: the protein subunits. Proc Natl Acad Sci USA 84, 6162–6166. 4 Cogdell RJ, Fyfe PK, Barrett SJ, Prince SM, Freer AA, Isaacs NW, McGlynn P & Hunter CN (1996) The purple bacterial photosynthetic unit. Photosynth Res 48, 55–63. 5 Deisenhofer J, Epp O, Miki K, Huber R & Michel H (1985) Structure of the protein subunits in the photo- synthetic reaction centre of Rhodopseudomonas viridis at 3A ˚ resolution. Nature 318, 618–624. 6 Koepke J, Hu X, Muenke C, Schulten K & Michel H (1996) The crystal structure of the light-harvesting com- plex II (B800–850) from Rhodospirillum molischianum. Structure 4, 581–597. 7 McDermott G, Prince SM, Freer AA, Hawthornthwa- ite-Lawless AM, Papiz MZ, Cogdell RJ & Isaacs NW (1995) Crystal structure of an integral membrane light- harvesting complex from photosynthetic bacteria. Nature 374, 517–521. 8 McLuskey K, Prince SM, Cogdell RJ & Isaacs NW (2001) The crystallographic structure of the B800–820 LH3 light-harvesting complex from the purple bacteria Rhodopseudomonas acidophila strain 7050. Biochemistry 40, 8783–8789. 9 Roszak AW, Howard TD, Southall J, Gardiner AT, Law CJ, Isaacs NW & Cogdell RJ (2003) Crystal struc- ture of the RC–LH1 core complex from Rhodopseudo- monas palustris. Science 302, 1969–1972. 10 Miller KR (1982) Three-dimensional structure of a photosynthetic membrane. Nature 300, 53–55. L N. Liu et al. AFM study on LH2 2D crystal FEBS Journal 275 (2008) 3157–3166 ª 2008 The Authors Journal compilation ª 2008 FEBS 3165 11 Jungas C, Ranck J-L, Rigaud J-L, Joliot P & Verme ´ glio A (1999) Supramolecular organization of the photosyn- thetic apparatus of Rhodobacter sphaeroides. EMBO J 18, 534–542. 12 Drews G & Imhoff JF (1991) Phototrophic purple bac- teria. In Variations in Autotrophic Life (Shively JM & Barton LL, eds), pp. 51–97. Academic Press, London. 13 Bahatyrova S, Frese RN, van der Werf KO, Otto C, Hunter CN & Olsen JD (2004) Flexibility and size hetero- geneity of the LH1 light harvesting complex revealed by atomic force microscopy: functional significance for bacterial photosynthesis. J Biol Chem 279, 21327–21333. 14 Scheuring S, Seguin J, Marco S, Le ´ vy D, Breyton C, Robert B & Rigaud J-L (2003) AFM characterization of tilt and intrinsic flexibility of Rhodobacter sphaeroides light harvesting complex 2 (LH2). J Mol Biol 325, 569–580. 15 Scheuring S, Francia F, Busselez J, Melandri BA, Rigaud J-L & Le ´ vy D (2004) Structural role of PufX in the dimerization of the photosynthetic core complex of Rhodobacter sphaeroides. J Biol Chem 279, 3620–3626. 16 Walz T, Jamieson SJ, Bowers CM, Bullough PA & Hunter CN (1998) Projection structures of three photo- synthetic complexes from Rhodobacter sphaeroides: LH2 at 6 A ˚ , LH1 and RC–LH1 at 25 A ˚ . J Mol Biol 282, 833–845. 17 Bahatyrova S, Frese RN, Siebert CA, Olsen JD, van der Werf KO, van Grondelle R, Niederman RA, Bullough PA, Otto C & Hunter CN (2004) The native architecture of a photosynthetic membrane. Nature 430, 1058–1062. 18 Frese RN, Pa ` mies JC, Olsen JD, Bahatyrova S, van der Weij-de Wit CD, Aartsma TJ, Otto C, Hunter CN, Frenkel D & van Grondelle R (2008) Protein shape and crowding drive domain formation and curvature in biological membranes. Biophys J 94, 640–647. 19 Frese RN, Siebert CA, Niederman RA, Hunter CN, Otto C & van Grondelle R (2004) The long-range organization of a native photosynthetic membrane. Proc Natl Acad Sci USA 101, 17994–17999. 20 Stahlberg H, Fotiadis D, Scheuring S, Re ´ migy H, Braun T, Mitsuoka K, Fujiyoshi Y & Engel A (2001) Two-dimensional crystals: a powerful approach to assess structure, function and dynamics of membrane proteins. FEBS Lett 504, 166–172. 21 Scheuring S, Reiss-Husson F, Engel A, Rigaud J-L & Ranck J-L (2001) High-resolution AFM topographs of Rubrivivax gelatinosus light-harvesting complex LH2. EMBO J 20, 3029–3035. 22 Stamouli A, Kafi S, Klein DC, Oosterkamp TH, Frenken JW, Cogdell RJ & Aartsma TJ (2003) The ring structure and organization of light harvesting 2 complexes in a reconstituted lipid bilayer, resolved by atomic force microscopy. Biophys J 84, 2483–2491. 23 Gonc¸ alves RP, Busselez J, Le ´ vy D, Seguin J & Scheur- ing S (2005) Membrane insertion of Rhodopseudomonas acidophila light harvesting complex 2 investigated by high resolution AFM. J Struct Biol 149 , 79–86. 24 Milhiet P-E, Gubellini F, Berquand A, Dosset P, Ri- gaud J-L, Le Grimellec C & Le ´ vy D (2006) High-resolu- tion AFM of membrane proteins directly incorporated at high density in planar lipid bilayer. Biophys J 91, 3268–3275. 25 Chami M, Pehau-Arnaudet G, Lambert O, Ranck J-L, Le ´ vy D & Rigaud J-L (2001) Use of octyl beta-thiog- lucopyranoside in two-dimensional crystallization of membrane proteins. J Struct Biol 133, 64–74. 26 Rigaud J-L, Chami M, Lambert O, Le ´ vy D & Ranck J-L (2000) Use of detergents in two-dimensional crystallization of membrane proteins. Biochim Biophys Acta 1508, 112–128. 27 Bahatyrova S (2005) Atomic force microscopy of bacterial photosynthetic systems: a new model for native membrane organization. PhD Thesis, University of Twente, Enschede. 28 Gonc¸ alves RP, Bernadac A, Sturgis JN & Scheuring S (2005) Architecture of the native photosynthetic appara- tus of Phaeospirillum molischianum. J Struct Biol 152, 221–228. 29 Scheuring S, Busselez J & Le ´ vy D (2005) Structure of the dimeric PufX-containing core complex of Rhodobacter blasticus by in situ atomic force microscopy. J Biol Chem 280, 1426–1431. 30 Scheuring S, Gonc¸ alves RP, Prima V & Sturgis JN (2006) The photosynthetic apparatus of Rhodopseudo- monas palustris: structures and organization. J Mol Biol 358, 83–96. 31 Scheuring S, Rigaud J-L & Sturgis JN (2004) Variable LH2 stoichiometry and core clustering in native mem- branes of Rhodospirillum photometricum. EMBO J 23, 4127–4133. 32 Scheuring S & Sturgis JN (2005) Chromatic adaptation of photosynthetic membranes. Science 309, 484–487. 33 Frese RN, Olsen JD, Branvall R, Westerhuis WH, Hunter CN & van Grondelle R (2000) The long-range supraorganization of the bacterial photosynthetic unit: a key role for PufX. Proc Natl Acad Sci USA 97, 5197–5202. 34 Hunter CN, Pennoyer JD, Sturgis JN, Farrelly D & Niederman RA (1988) Oligomerization states and associations of light-harvesting pigment–protein complexes of Rhodobacter sphaeroides as analyzed by lithium dodecyl sulfate-polyacrylamide gel electrophoresis. Biochemistry 27, 3459–3467. 35 Rigaud J-L, Le ´ vy D, Mosser G & Lambert O (1998) Detergent removal by non-polar polystyrene beads. Eur Biophys J 27, 305–319. 36 Rigaud J-L, Mosser G, Lacapere J-J, Olofsson A, Le ´ vy D & Ranck J-L (1997) Bio-beads: an efficient strategy for two-dimensional crystallization of membrane proteins. J Struct Biol 118, 226–235. AFM study on LH2 2D crystal L N. Liu et al. 3166 FEBS Journal 275 (2008) 3157–3166 ª 2008 The Authors Journal compilation ª 2008 FEBS . Dimers of light-harvesting complex 2 from Rhodobacter sphaeroides characterized in reconstituted 2D crystals with atomic force microscopy Lu-Ning Liu 1 ,2 ,. protruding LH2 (up) Weakly protruding LH2 (down) 12. 0 15.0 22 000 F A B C D E 20 000 21 000 18 367 ABCDE 21 20 8 21 22 4 20 408 19 047 19 000 18 000 LH2 density