Structural role of PufX in the dimerization of the photosynthetic core complex of Rhodobacter sphaeroides.

Monomeric and dimeric PufX-containing core complexes have been purified from membranes of wild-type Rhodobacter sphaeroides. Reconstitution of both samples by detergent removal in the presence of lipids leads to the formation of two-dimensional crystals constituted of dimeric core complexes. Two-dimensional crystals were further analyzed by cryoelectron microscopy and atomic force microscopy. A projection map at 26-A resolution reveals that core complexes assemble in an "S"-shaped dimeric complex. Each core complex is composed of one reaction center, 12 light-harvesting 1 alpha/beta-heterodimers, and one PufX protein. The light-harvesting 1 assemblies are open with a gap of density of approximately 30-A width and surround oriented reaction centers. A maximum density is found at the dimer junction. Based on the projection map, a model is proposed, in which the two PufX proteins are located at the dimer junction, consistent with the finding of dimerization of monomeric core complexes upon reconstitution. This localization of PufX in the core complex implies that PufX is the structural key for the dimer complex formation rather than a channel-forming protein for the exchange of ubiquinone/ubiquinol between the reaction center and the cytochrome bc1 complex.

In purple photosynthetic bacteria, highly organized transmembrane pigment-protein complexes perform absorption of light and its conversion into chemical energy. Two light-harvesting complexes (LH), 1 LH2 and LH1, ensure the collection of light. Then the excitation energy is funneled toward the special pair of bacteriochlorophylls in the reaction center (RC), followed by an electron transfer from the special pair of bacteriochlorophylls to the ubiquinone (Q) acceptors Q A and Q B . After two photoreactions and proton captures, ubiquinol (QH 2 ) is formed at the Q B site that dissociates from the RC into the membrane. The bc 1 complex utilizes QH 2 and oxidized cyto-chrome c 2 as reductant and oxidant, respectively. The net result is a cyclic electron transfer that promotes the formation of a proton gradient across the membrane, which is utilized for ATP synthesis by F 1 F 0 -ATP synthase (for a review, see Ref. 1).
The mechanisms of excitonic energy migration and photoinduced electron transfer have been studied in detail using biophysical approaches and analyzed at the molecular level following the resolution of the atomic structures of the RC (2), LH2 (3,4), and the bc 1 complex (5). However, despite the wealth of information available on the individual proteins, their supramolecular organization in the membrane remains undetermined. A crucial issue is the spatial organization between LH1 and RC, forming the so-called core complex, in which the transformation of light energy into charge separation occurs. So far, no atomic model of the core complex is available. Early studies by various electron microscopy (EM) techniques of native membranes of Rhodopseudomonas viridis and Ectothiorhodospira halochloris have shown a single RC surrounded by a closed circle of light-harvesting molecules (6 -8). Organization of LH1 complexes in a closed ring of 16 subunits large enough to house a RC was observed on a cryo-EM projection map of LH1 purified from Rhodospirillum rubrum (9). Later, two-dimensional crystals were grown from purified entire core complexes of R. rubrum (10 -12), pufXdeleted Rhodobacter sphaeroides (13), and R. viridis (14) and analyzed by cryo-EM or negative stain EM, all consistent with a closed LH1 ring of 16 subunits around the RC. More recently, a high resolution atomic force microscopy (AFM) analysis of single molecules in native membranes of R. viridis showed LH1 to form a closed ellipse of 16 subunits around a specifically oriented RC (15).
A closed assembly of LH1 raises the question of QH 2 /Q exchange between the RC and the bc 1 complex (16). At this point, a protein encoded by the pufX gene found in Rhodobacter sp. (sphaeroides and capsulatus) has been shown to be necessary for photosynthetic growth (17,18) and more specifically to facilitate the transport of reducing equivalent from the Q B site of the RC to the bc 1 complex (17,19). In R. sphaeroides, this additional peptide PufX is a small protein of 82 amino acids with one membrane span (20 -22). Much data led to the hypothesis that the PufX protein is located in the inner ␣ ring of LH1 (20 -22), forming a specific channel and/or interrupting the LH1 continuity and thus allowing a facilitated diffusion of QH 2 /Q between RC and bc 1 complex. Furthermore, biochemical studies have shown that PufX-containing core complexes could be isolated from R. sphaeroides membranes as dimeric structures with a 1:1 PufX/RC stoichiometry and a decreased number of LH1-associated Bchls per RC (23).
The only structural information available about a PufX-containing core complex has been derived from a low resolution negative stain EM analysis of native two-dimensional crystals of a LH2-deleted R. sphaeroides strain (24). This projection map revealed two C-shaped incomplete antenna rings each surrounding one RC. Finally, polarized spectroscopy of R. sphaeroides membranes containing or lacking PufX provided evidence that PufX induced a specific orientation of the RC in the LH1 ring as well as the formation of a long range order of LH1-RC core complexes (16,25). Taking together all of these considerations, PufX is thought to play a key role in the core complex. One challenge now being addressed is the localization of PufX in the core complex to get a deeper insight into its implication in the structure-function relationship critical for bacterial photosynthesis.
In order to understand the role of PufX in the core complex, we have performed a structural study by two-dimensional crystallization of monomeric and dimeric PufX-containing core complexes isolated from wild-type R. sphaeroides. The purified complexes were analyzed by sedimentation and absorption spectroscopy. Two-dimensional crystals were produced by detergent removal from protein/lipid/detergent mixtures and analyzed by cryo-EM and AFM. Both monomeric and dimeric core complexes reassemble in dimeric PufX-LH1-RC complexes when reconstituted in lipid membranes. The cryo-EM projection map, revealing protein densities of trans-membrane segments, showed an "S"-shaped core complex dimer. The LH1 assemblies are open with a lack of protein density within the gaps, the dimer junction has a stronger density than the rest of the "S", and the RC reveals an asymmetric density maximum within the open LH1. Taking all information together, we propose a model in which PufX is located at the dimer junction and directly involved in the formation of the supramolecular assembly of core complexes rather than providing a direct pathway for the QH 2 /Q exchange.

EXPERIMENTAL PROCEDURES
Materials-All phospholipids and detergents were of highest purity and were purchased from Avanti Polar Lipids and from Fluka or Anatrace, respectively. Bio-Beads SM2 (25-50 mesh) from Bio-Rad were extensively washed with methanol and water before use as described (26). All other reagents were of analytical grade.
Isolation and Purification of the Monomeric and Dimeric Complex-Core complexes from semiaerobically grown R. sphaeroides strains with and without PufX were isolated as described previously (23), except that after the washing step in 2 M NaBr, 0.2 M sucrose, membranes were directly resuspended at 10 mg/ml protein and extracted with 3% n-octyl ␤-D-glucopyranoside (OG) and 0.5% sodium cholate. After extraction, the supernatant was run on a sucrose gradient (10 -40%) in buffer glycyl-glycine (50 mM), pH 7.8, 0.6% OG, and 0.2% sodium cholate. As previously shown (22,23), three major bands were isolated, including an LH2-containing band and two PufX-containing bands corresponding to the monomeric or dimeric core complexes.
Reconstitution and Two-dimensional Crystallization-Purified LH1-RC-PufX complexes at 0.4 mg/ml in a buffer containing 0.1% DOTM, 200 mM NaCl, 50 mM gly-gly, pH 7.8, were mixed with egg phosphatidylcholine/egg phosphatidic acid (9:1, mol/mol) at lipid/protein ratios between 0.75 and 1 (w/w). The lipid/protein/detergent mixture was allowed to equilibrate overnight in the dark. Detergent removal was performed through two successive additions of 5 mg of SM2 Bio-Beads/ 50-l sample, for 12 h each, at 4°C (28). Crystals were then withdrawn and kept at 4°C for EM and AFM analysis.
Electron Microscopy-Negative stain and cryo-EM were performed using Philipps CM120 and Philipps FEG200 electron microscopes operated at 120 and 200 kV, respectively. Two-dimensional crystals were adsorbed for 1 min on glow-discharged carbon-coated 300-mesh grids. Two-dimensional crystals were either negatively stained with 1% uranyl acetate, or quick frozen in liquid ethane for cryo-EM analysis. The latter were transferred into the electron microscope using a Gatan cryo transfer system. Images were directly assessed using low electron dose procedures at a magnification of ϫ40,000 on a 14-m 2048 ϫ 2048-pixel Gatan slow scan CCD camera corresponding to 3.5 Å/pixel.
Atomic Force Microscopy-Mica prepared as described (29) was used as support and freshly cleaved before every experiment using Scotch tape. 3 l of two-dimensional crystal solution (0.1 mg/ml) was injected into the adsorption buffer drop (10 mM Tris-HCl, pH 7.6, 200 mM KCl, 5 mM MgCl 2 ) on the mica surface. After 2 h, the sample was carefully rinsed with recording buffer (10 mM Tris-HCl, pH 7.6, 200 mM KCl). Imaging was performed with a commercial Nanoscope III AFM (from Digital Instruments, Santa Barbara, CA) equipped with a 150-m scanner (J-scanner) and oxide-sharpened Si 3 N 4 cantilevers with a length of 100 m (k ϭ 0.09 N/m; Olympus Ltd., Tokyo, Japan). The AFM was operated in contact mode applying forces of Ͻ0.2 nN at scan frequencies around 3 Hz.
Image Processing-Cryo-EM images of two-dimensional crystals recorded on the CCD camera were treated using the MRC image processing package (30,31). Briefly, images were unbent running two cycles of CCUNBEND, and the symmetry space group was defined using ALL-SPACE (32). The final map was calculated by merging images using ORIGTILTK. Projection maps were calculated using programs of the CCP4 package (33). An isotropic temperature factor (B ϭ Ϫ500) was applied to compensate for the degradation of image amplitudes. Density contouring is at 0.25 root mean square with densities above the mean contoured by solid lines.

RESULTS
LH1-RC-PufX core complexes were extracted from membranes of semiaerobically grown wild type R. sphaeroides using an OG/sodium cholate mixture (22,23). As previously shown, two major bands corresponding to monomeric and dimeric forms of LH1-RC-PufX complexes are found on a sucrose gradient (22,23). Crystallization trials of both monomeric and dimeric PufX-containing core complexes led to small (ϳ50-nm diameter) noncrystalline vesicles, whatever the rate of detergent removal (dialysis or Bio-Beads) or lipid composition (data not shown).
Since the nature of the detergent is a crucial parameter for the result of two-dimensional crystallization trials (34,35), other detergents than the OG/cholate mixture were tried. For this purpose, the monomeric and dimeric LH1-RC-PufX bands purified in the OG/cholate sucrose gradient were dialyzed against buffers containing different detergents. After detergent exchange, the core complexes in different detergents were reconstituted, and the results were analyzed by EM. The best reconstitutions, in terms of crystallinity and size of the crystals, were obtained from PufX-containing core complexes in DOTM. The integrity of monomeric and dimeric core complexes in DOTM was confirmed by sedimentation on a sucrose gradient, absorption spectroscopy, and gel electrophoresis. As shown in Fig. 1 (left), sucrose gradient patterns similar to those previously described for the core complex in OG/cholate were obtained with core complexes in DOTM (23). The monomeric PufX-containing core complex (tube 2) ran at a density similar to that of the PufX-deleted core complex (tube 3), whereas the dimeric PufX-containing core complex ran at higher density (tube 1). The composition of core complexes in DOTM was analyzed by SDS-PAGE electrophoresis and revealed the presence of PufX (lanes 1 and 2 in Fig. 1A) in contrast to the pufX-deleted strain (lane 3 in Fig. 1A), as well as the subunits L, M, and H of RC and the subunits ␣/␤ of LH1 from wild-type R. sphaeroides. Finally, the spectra of core complexes in DOTM showed typical absorption peaks of pigments in native membranes at 872 and 804 nm, without detectable bands at 820 nm (disassembled LH1) or at 540 -560 nm (cytochrome bc 1 complex contamination). All of these data indicate that the core complexes purified in DOTM are identical to the previously well characterized core complexes in OG/cholate (23).
For two-dimensional crystallization, monomeric or dimeric core complexes in DOTM were mixed with egg phosphatidyl-choline/egg phosphatidic acid (9:1, mol/mol) at lipid-to-protein ratios between 0.75 and 1.0 (w/w). Complete detergent removal at 4°C using Bio-Beads resulted in the formation of vesicular structures whose crystallinity could be directly observed in low magnification images of negatively stained samples (Fig. 2, A  and B). Strikingly, reconstitutions from both isolated monomeric and dimeric core complexes result in two-dimensional crystals constituted of dimers. Indeed, Fourier transforms (top insets in Fig. 2, A and B) revealed identical vector lengths for both two-dimensional crystals, and corresponding filtered images revealed the same packing arrangement of dimers in the lipid bilayer (bottom insets in Fig. 2, A and B). It has to be recalled that in the absence of PufX (i.e. with core complex from pufX-deleted strain), dimerization of core complexes does not occur upon reconstitution (data not shown; see also Ref. 13).
Better crystals, as judged from size (Ͼ1 ϫ 1 m) and packing coherence, were obtained starting from the dimeric core complex sample ( Fig. 2A). The less coherent crystal packing obtained with the monomeric core complex was probably due to defects introduced by an incomplete dimerization of the complexes upon reconstitution.
Two-dimensional crystals formed from the dimeric preparation were embedded in vitreous ice and analyzed by cryo-EM. The two-dimensional crystals had p22 1 2 1 symmetry and unit cell dimensions of a ϭ 381.5 Å, b ϭ 130.0 Å, and ␥ ϭ 89.5°, housing two dimeric core complexes in up-and-down orientation as clearly visible by the "S" and "Z" appearance of dimers in neighboring rows (Fig. 3, Table I). As judged from phase residual analysis, a projection map could be calculated to 26-Å resolution ( Table I).
The cryo-EM projection map reveals that the core complexes assemble in a particular "S"-shaped arrangement with densities enclosed within each cavity of the "S." In agreement with general models and structural data of core complexes, the "S"shaped densities are interpreted as LH1 assembly, and the two enclosed densities are interpreted as two RCs (1,12,15). However, the LH1 complex is open, forming a bow density over ϳ285°(see double-headed arrow in Fig. 3B) with inner and outer diameters of ϳ75 and ϳ110 Å, respectively. In the projection map of the dimer, three remarkable features can be stressed. First, there is no detectable density in the ϳ30-Å gap of the LH1 assembly, indicating the absence of any transmembrane segment at this position (1 in Fig. 3B). Second, a maximal density is found at the dimer junction (2 in Fig. 3B), whereas the rest of the projection density along the "S" is homogeneous. Third, the enclosed density originating from the RC is asymmetric, with the strongest density at the side approximately opposed to the LH1 gap (3 in Fig. 3B). Since random orientation would have resulted in an averaged centered, ill defined, circular blob, the asymmetric density of the RC indicates a specific orientation of the RC within the core complex. This is in agreement with Frese et al. (25), who have shown such a unique orientation of the RC with respect to PufX to maximize the diffusion rate of the QH 2 /Q out of and into the Q B site. All of these features are consistent with the nonsymmetrized projection map, in which the unit cell contains four unique core complexes (i.e. two unique dimer complexes (1, 2, and 3 in Fig. 3C).
Two-dimensional crystals and densely packed membranes were further analyzed by AFM to acquire additional topographical information on the LH1-RC-PufX assembly (Fig. 4). Twodimensional crystals produced from the purified dimeric core complexes were adsorbed to the mica surface. Overview topographs revealed strongly corrugated ripples (Fig. 4, A and B). At higher magnification, the ripples revealed substructure corresponding to the dimeric organization of the complexes. Obviously, the incorporation of dimeric core complexes in up-anddown orientation bends the lipid membrane in opposite directions (Fig. 4C, line 1, as shown by the height profile (Fig.   4D, trace 1). Such wave shape of the membranes has hampered imaging at higher resolution by AFM and explains the strong uneven staining effect observed in negative stain electron microscopy (see bottom insets in Fig. 2, A and B). However, the ripples reveal a periodicity of 385 Å, consistent with the long axis of the unit cell as observed in EM (Table I) and an amplitude of 38 Å (Fig. 4D, trace 1). Along the top of the ripple axis (Fig. 4C, line 2) a periodicity of 125 Å, consistent with the short axis of the unit cell as observed in EM (Table I), and an amplitude of 8 Å are found in section analysis (Fig. 4D, trace 2).
Additional information was provided by imaging membranes, in which the dimeric core complexes were reconstituted at high density but not crystalline (Fig. 4, E and F). Two different types of surfaces can be clearly identified on the membrane, corresponding to dimers inserted in up-and-down orientations in the lipid bilayer. In one orientation, the protein surfaces can be imaged and show lines of protein in dimeric arrangement (arrow 1), whereas in the opposite orientation, the surfaces reveal a blurry topography also arranged in lines with a similar width of about 220 Å (arrow 2). Such a blurry topography, covering the whole dimer surface, could be attributed to the presence of extramembranous protein domains with a high flexibility (36). This is in contrast with previous AFM topographies of different antenna complexes without PufX (15,37,38) and is probably related to the extramembranous domain of PufX proteins. However, the finding of dimers in a noncrystalline membrane represents additional strong evidence that dimerization is specific to PufX rather than resulting from crystallographic constraints. DISCUSSION A single trans-membrane protein, termed PufX in R. sphaeroides, has been evidenced to play a key role in the initial steps of photosynthesis. Functional studies have shown that the PufX protein is essential for photosynthetic growth and is required for a fast QH 2 /Q exchange between the Q B site in the RC and the Q o site of the cytochrome bc 1 complex (19,23,25). From a structural point of view, PufX-containing core complexes have been shown to be dimeric in native membranes of R. sphaeroides (24). However, this finding raised strong criti- a Number of lattices in the merged image (Fig. 3).  Table I) projection map of the dimer core complex two-dimensional crystals at 26-Å resolution (scale bar, 50 Å). C, nonsymmetrized (p1) projection map of the dimer core complex two-dimensional crystals at 26-Å resolution (scale bar, 50 Å). In both the symmetrized and the nonsymmetrized projection maps, two lines of up-and-down oriented complexes are clearly visible. Important features are consistently found in both maps as indicated by 1, 2, and 3. The LH1 assembly reveals a continuous density over ϳ285°(double-headed arrow in B) and a ϳ30-Å gap, where no density is found (1), the density at the dimer junction is stronger than the rest of the "S" density (10 density contour lines as compared with 3-6 density contour lines) (2), and the RC reveals an asymmetric density (3). cism, since the study was performed on tubular crystalline membranes gained from a LH2 deletion mutant grown under partial aerobic nonphotosynthetic conditions (16,39,40). Moreover, this projection map was calculated from negative stained sample, hence contrasting extramembranous domains rather than trans-membrane parts. Finally, the presence of bc 1 complex was proposed within the complex, but little structural evidence documented this hypothesis (24).
In order to obtain more insight into the role of PufX in the dimeric core complex, we have purified and two-dimensionally crystallized the dimeric PufX-containing core complexes from R. sphaeroides. The native core complexes of R. sphaeroides have been isolated by gentle solubilization with a specific mixture of OG and cholate (23), followed by a detergent exchange with DOTM, a low critical micellar concentration detergent that allowed stabilization of the dimeric core complex and improved two-dimensional crystallization. Biochemical and spectroscopic analyses (Fig. 1) were consistent with previous works on membranes or isolated complexes, indicating a dimeric core complex with a PufX/RC stoichiometry of 1:1 and a decreased number of LH1-associated Bchls per RC (23,24,41). Two-dimensional crystals were analyzed by cryo-EM in order to acquire information about the protein densities of the transmembrane segments of the core complex. Despite their size and their apparent long range coherence ( Fig. 2A), the two-dimensional crystals reconstituted from the purified dimeric core complexes provided information only to a medium resolution of 26 Å in cryo-EM (Fig. 3A, Table I). This unusual low resolution for an analysis by cryo-EM indicates a poor order of the twodimensional crystals. In this context, AFM analysis of the two-dimensional crystals revealed that the up-and-down incorporation of core complexes induced a peculiar waving of the two-dimensional crystals (Fig. 4). Such bending of the membrane could be related to the tubular morphology of native membranes, in which all core complexes have the same orientation in the membrane (24,42). It is likely that such waving hampered order at higher resolution in the two-dimensional crystals.
However, the cryo-EM map limited to 26-Å resolution revealed a dimeric assembly of the core complex with open LH1 rings. The "S"-shaped densities are interpreted as LH1 assembly, and the two central densities are interpreted as two RCs, in agreement with general models and structural data of core complexes (1,12,15). The inner and outer diameters of ϳ75 and ϳ110 Å, respectively, of the LH1 complex are large enough to house ␣/␤-heterodimers with the ␣-polypeptide facing the ring inside (3,4,12) and are in agreement with the reported values of inner and outer diameters of ϳ70 and ϳ120 Å, respectively (12,15). However, most importantly, projection density maps of the dimeric LH1-RC-PufX complexes show unequivocally gaps in the LH1 rings surrounding the RC (Fig. 3, B and C). The LH1 complexes are open, covering only slightly more than three-quarters (ϳ285°) of a full circle, and no projection density is found within the gap (1 in Fig. 3B). Since closed LH1 assemblies with similar diameters are constituted of 16 LH1 subunits (12,15), we propose that our LH1 density contains 12 Ϯ 1 LH1 ␣/␤-heterodimers plus one PufX. This number of LH1 subunits is furthermore consistent with biochemical and spectroscopic measurements that have reported ϳ24 Bchl molecules per RC, corresponding to ϳ12 LH1 subunits in R. sphaeroides (24). In addition, the resemblance of our cryo-EM projection map with the negative stain map from the native membranes (24) indicates that no modification occurred through the purification and reconstitution procedures and that our map represents the native state of the LH1-RC-PufX assembly.
PufX is a small protein of 82 amino acids in R. sphaeroides. Circular dichroism, structure prediction, and proteolytic digestion have shown that PufX forms one membrane span and interacts strongly with the ␣-polypeptide and not with the ␤-polypeptide of LH1 (20,21,43). These data evidenced that PufX is associated in the inner ␣-ring of LH1 and interrupts the LH1 assembly.
In the symmetrized as well as in the nonsymmetrized projection maps, a stronger density is found (2 in Fig. 3, B and C) at the dimer junction. This increased density is attributed to the two PufXs, whereas identical ␣/␤-heterodimers constitute the rest of the "S" density ( Fig. 5). Indeed, a dimer of PufXs is larger (164 amino acids) than an ␣/␤-heterodimer (104 amino acids). In particular, the C-terminal domain of PufX, of about 28 amino acids, has been predicted to form an amphipathic ␣-helix (44), which might increase the projection density. Reinforcing this localization, we have shown that monomeric PufX-containing core complexes dimerize upon reconstitution (Fig. 2), whereas core complexes from PufX-deleted strain never dimerize (data not shown) (13). It is likely that homologous trans-membrane interactions between two PufXs lead to dimerization of core complexes in lipid membranes. In addition, this localization solves the problem caused by the inversion of ␣/␤-heterodimers on both sides of the "S" density (Fig. 5). Indeed, the formation of circular assembly with inner ␣and outer ␤-rings results from homologous interaction between ␣/␤heterodimers. If ␣/␤-heterodimers could assemble in both orientations, there would be no reason for the formation of circular antennas rather than random waves of LH subunits (3,4,9,12,15).
The generally proposed localization of PufX within the gap, closing the LH1 assembly, can be rejected (16,25). Such a localization would agree with the role of PufX to interrupt the LH1 assembly but would involve the formation of two ␣/␤heterodimers in opposite orientation at the dimer junction. In addition, these ␣/␤-heterodimers at the dimer junction should also have the property to induce dimerization, since the two PufXs would be separated from each other, excluding their direct role in dimer formation. Furthermore, the presence of a highly tilted trans-membrane helix in the gap that would produce weak densities in the projection map is also unlikely, since to fill a gap of ϳ30-Å length in a lipid membrane of ϳ40-Å thickness, a trans-membrane helix of ϳ50-Å length would be required. Such a long helix is in disagreement with hydropathy analysis of PufX (21) and with data reporting a minimal core portion of 25 amino acids of PufX to be sufficient to inhibit the LH1 assembly (21,22). Finally, a possible localization of PufX at the peripheral ends of the "S" can be rejected for the same reasons.
However, if PufX is located at the dimer junction and not within the gap, why are there no additional LH1 subunits that close the three-quarter circle, and what is the role of the gap? A first proposal would be to interpret the gap as a channel for the QH 2 /Q exchange between the RC and the bc 1 complex. However, in addition to the fact that a 30-Å-wide gap is exces-sively large for a QH 2 /Q channel, it would still be unclear how the assembly of the core complex could know when to stop the assembly of additional LH1 around RC (42). A second proposal would be to consider the gap as a docking site for the bc 1 complex and more specifically for its subunit IV. In this context, several results have suggested that this subunit (14 kDa), found so far only in the isolated bc 1 from R. sphaeroides, was essential for photosynthetic electron transfer activity (24). Indeed, it has been shown that subunit IV is absent in aerobically grown R. sphaeroides, although the bc 1 complex is fully functional for respiration (45). Interestingly, a study of the timedependent assembly of the photosynthetic unit has shown that subunit IV of bc 1 is, together with PufX and the H subunit of the RC, the first protein to be present, followed by the synthesis of LH1 ␣/␤ polypeptides, leading to the encircling of the RC (43). Thus, it could be proposed that, at an early stage of assembly, subunit IV is used as a placeholder to hamper further LH1 subunit addition at the position where we have now observed a gap in the purified LH1-RC-PufX complexes.
In conclusion, we have been able, based on reconstitution experiments and cryo-EM analysis of two-dimensional crystals, to gain much evidence for a localization of PufX at the dimer junction and for its role as a structural organizer within the core complex of R. sphaeroides. We suggest that PufX could act as a symmetry breaker to prevent the clustering of RC-LH1 complexes into large hexagonal arrangements, as observed for several species of purple bacteria without pufX (6,8,15,46). Due to this suggestion, PufX-induced core dimerization would facilitate indirectly the functional interplay of RC-LH1 core complexes with other components of the photosynthetic unit like LH2 and bc 1 complex. Such a hypothesis will be tested by future AFM analysis of native membranes, consisting of all components of the photosynthetic chain, LH2, LH1, PufX, RC, and bc 1 complexes.