To examine the morphogenesis of the ICM membrane in dependence of LH2 assembly, LH2 wild type (wt) or mutant complexes are expressed in a deletion strain of R. sphaeroides (LH2^- LH1^- RC^-). This strain is devoid of endogenous BChl-binding proteins 1 but capable of BChl synthesis (Jones et al., 1992). Changes in ICM development may thus be directly correlated with changes in expression and/or assembly of LH2 owing to the absence of additional BChl-binding complexes (LH1 and RC). The first mutant of which we studied the membrane topology is LH2 αAL_16-4S/βAL₁₂ (Table 1). We have previously shown that the stable assembly of the LH2-like complex is severely affected by massive mutagenesis of the chromphor binding site (Kwa et al., 2004). The LH2 absorption spectra with red most absorption bands of the BChl-B8502 and BChl-B800 are at ∼849 and 800 nm (Fig. 1A) and thus typical for the spectra of the antenna complex from R. sphaeroides (see, e.g. Cogdell and Scheer, 1985; van Grondelle, 1985; Braun and Scherz, 1991). The structural stability of LH2 αAL_16-4S/βAL₁₂, however, is significantly impaired as compared with the one of LH2 wt (Fig. 1B). In addition, the LH2 complex concentration relative to total protein concentration in membranes of LH2 αAL_16-4S/βAL₁₂ is approximately six times lower than that of membranes containing LH2 wt (Table 2).

The ultrastructures of ICM of R. sphaeroides DD13 cells not expressing LH2, expressing LH2 wt and model LH2 αAL_16-4S/βAL₁₂, are shown in Fig. 2. In the absence of LH2 complexes, vesicular ICM invaginations are not observed, while in the presence of LH2 wt complexes mature ICM is observed with vesicle-like invaginations ranging from 35 to 55 nm in diameter. The size and shape of the vesicular invaginations are similar to those reported previously for R. sphaeroides 2.4.1 cells grown photosynthetically and containing LH2, LH1 and RC (Gibson, 1965; Fraker and Kaplan, 1972; Kiley and Kaplan, 1988; Feniouk et al., 2002). In contrast, the radii of the ICM invaginations containing αAL_16-4S/βAL₁₂ are significantly enlarged relative to the invaginations of ICM containing LH2 wt, and moreover they are changed from normal spherical to tubular shaped membranes of up to a few hundred nanometres in length (Fig. 2C). Apparently, LH2 properties and expression level are correlated with the size and morphology of ICM in deletion strain DD13 devoid of core complexes. Previously, it has been shown that the absence of LH2 but presence of LH1 and RC results in tubular ICM in R. sphaeroides. Here, it is shown that tubular ICM, albeit of smaller size, also develop in the presence of massively mutated LH2 complexes.

To determine whether the altered membrane morphology is accompanied by changes in the membrane lipids, the phospholipid compositions of ICM containing LH2 wt or αAL_16-4S/βAL₁₂ complex were determined by ESI-MS/MS as described in Brugger et al. (1997; 2006). The phospholipid composition of cellular membrane, i.e. of intact cells, and of isolated chromatophores did generally not significantly differ (data not shown), which is similar to previous findings obtained from studies with R. sphaeroides (Onishi and Niederman, 1982). Largely, in line with previous studies (Russell and Harwood, 1979; Onishi and Niederman, 1982) PE, PC and PG make up the major phospholipids in R. sphaeroides (Table 3), while others are present only in trace amounts. PE is the most abundant phospholipid in the cellular membranes of R. sphaeroides strain DD13 (Table 3). In R. sphaeroides DD13 expressing LH2 wt, PE is slightly increased as compared with cells not expressing LH2 (∼61 ± 2.2% versus ∼58 ± 0.8%) whereas PC is reduced (from 27 ± 0.7% to ∼23 ± 0.6%). The relative amount of PG makes up less than 20% in all cases and thus does not change remarkably in mutant or wt membranes. This differs from previous studies, which reported on a significant increase in PG upon photosynthetic growth in R. sphaeroides 2.4.1 (Russell and Harwood, 1979). In a very recent MS analysis of the cell lipids from Rhodopseudomonas acidophila, PG has been shown to even decrease upon induction of photosynthetic growth to a value as low as 11–16% (Russell et al., 2002). Remarkably, the phospholipid composition of cells expressing αAL_16-4S/βAL₁₂ is dramatically changed. The relative PE content has dropped by ∼13.5% while the PC and PG content rose by ∼25.6% and ∼13.8% respectively. The PC+PG/PE ratio is thus increased from 0.65 in R. sphaeroides strain DD13 expressing LH2 wt to 0.90 in cells expressing αAL_16-4S/βAL₁₂. Thus, mutagenesis of the LH2 proteins not only results in a significant reduction in the complexes' expression level and a change in membrane morphology but also in a change of the overall phospholipid composition, in particular, the bilayer to non-bilayer lipid ratio.

Both the ratio of bilayer/non-bilayer lipids and fatty acyl chain length have been implied to partake in regulation of membrane curvature energy (e.g. Vikstrom et al., 2000). The fatty acyls of R. sphaeroides DD13 devoid of LH2, expressing LH2 wt or αAL_16-4S/βAL₁₂, were not analysed in detail but only characterized by the total number of carbons (ΣC) and double bonds (Σ). In all three samples, the major fatty acyl chain composition is 36:2, which is likely to be composed of two fatty acids and one double bond each (C18:1). Vaccenic acid, 18:1Δ11, makes up > 80% of the lipid tails in photosynthetically grown R. sphaeroides 2.4.1 and R. capsulatus (Russell and Harwood, 1979), and 45% in Rps. acidophila (Russell et al., 2002). The acyl-tails of the lipids of R. sphaeroides DD13 show only minor variations upon ICM development; neither the amount nor the degree of saturation varied significantly (data not shown). Similarly, the acyl-chains of PC which consist primarily of C18:2 are not altered upon expression of LH2. On the other hand, the composition of the PE acyl chains is significantly altered in cells expressing αAL_16-4S/βAL₁₂ as compared with cells expressing LH2 wt (Fig. 3). The major PE acyl-chain 36:2 is reduced from ∼79% in DD13 cells expressing LH2 wt to ∼61% in cells expressing αAL_16-4S/βAL₁₂. Concurrently, saturated lipids with less carbons (ΣC:ΣΔ = 29:0, 30:0, 31:0, 32:1, 32:0 and 33:1) are increased (Fig. 3). These shorter acyl-chains were either untraceable and/or present in only scarce amounts in cells expressing LH2 wt. The hydrocarbon composition of PC shows only minor variations (not shown), indicating that exclusively the acyl composition of PE is modulated upon alterations of the LH2 expression level and ICM morphology in R. sphaeroides DD13.

Taken together, our findings indicate that interactions between the lipids and LH2 influence the properties of the ICM. Pronounced changes in morphology of the ICM containing modified LH2 are accompanied by changes in the phospolipid composition. The relative content of bilayer lipids is increased as well as the content of shorter, fully saturated fatty acyl chains of these lipids. Both changes concur with the more planar morphology of the ICM in cells expressing αAL_16-4S/βAL₁₂. Membranes of cells not expressing LH2 and cells expressing LH2 wt, however, have a similar hydrocarbon tail composition, indicating that the lipid tail length and saturation do not play a critical role in formation and maintenance of ICM in the presence of native LH2.

Recently, evidence is accumulating that conserved lipid-binding motifs exist (Wakeham et al., 2001; Palsdottir et al., 2003; Palsdottir and Hunte, 2004). To identify aa residues of the LH2 polypeptides which potentially partake in specific LH2–lipid interactions, a number of different strategies have been pursued. In the TM stretches of the α- and β-polypeptides of bacterial light harvesting complexes, relatively few residues are found to be highly conserved (Zuber, 1986; Braun et al., 2002). In 34 different β-subunits only the residues at position 0, −4 and −8 are noted to be strictly conserved. Conspicuously, at the N-terminal domain of the β-subunit the glutamate residues at position −20 and −23 are also conserved in both LH2 and LH1 β-subunits even from such relatively remote genera as Erythrobacter and Chromatium. In the high-resolution structure of LH2 from Rps. acidophila (McDermott et al., 1995; Papiz et al., 2003), βGlu-20 is neither in close contacts with BChls pigments nor with residues from neighbouring subunits, and is thus not obviously involved in protein–protein or BChl–protein interactions. It is in close contact, however, with the Car, in particular, with atoms of its polar head groups. βGlu-20 is located at the cytoplasmatic end of the β-TMH, at the outer surface of the cylindrical structure of the LH2. It is therefore likely in contact with the membrane lipids, particularly, with the headgroups (Prince et al., 2003). Thus, βGlu-20 has been replaced by glutamine, alanine or lysine which did not affect the functional assembly of LH2 (see below), and the mutant complexes have been analysed for their interactions with the immediate lipid environment. To that end the LH2 wt and mutant lipid compositions of cellular membranes and of the boundary lipids, i.e. the lipids which remain closely attached to the complex upon purification from the membrane have been analysed (Table 4). Remarkably, the composition of LH2 boundary lipids is distinctively different from the composition of bulk lipids as present in the cellular membranes. PE is clearly crowding at the LH2–lipid interface, constituting as much as 88% of the total phospholipids (as compared with 61% in bulk phase). PC makes up only 12%, and PG is below detection level. It should be noted that the composition of the LH2 boundary lipids as determined here may depend on the experimental conditions, for example, the detergent used for extraction from the ICM. We have roughly estimated the molar ratios of lipids to LH2: in isolated complexes five to six lipid molecules are found per LH2 complex. Thus, of the nine potential PE sites, nearly about 60% appear occupied by a lipid molecule even after detergent treatment. Out of the lipids that are still attached ∼90% are PE molecules, thus less than one of the potential PE sites may have a PC attached (PC content is ∼10% of the 5–6 phospholipids found attached to isolated LH2). Upon replacement of Glu-20 with Ala, PE remains clearly the major phospholipid closely associated with LH2 αWT/βWT_−20A; however, the relative amount of PE is reduced by ∼17% in comparison to wt LH2 (from 88 ± 2.1% to 74 ± 3.1%) (Fig. 4; Table 4). PC is increased from 12% to 26%. In the LH2 αWT/βWT_−20A mutant, the total amount of lipids found attached remains roughly constant but the PE content is reduced by ∼17%, i.e. only three to four of the lipid binding sites still bind a PE. The number of sites occupied by PC molecules has apparently increased to two out of the six to seven sites which are occupied by lipid molecules.

In contrast to the boundary lipid composition, the bulk lipid composition of membranes containing αWT/βWT_−20A is not significantly affected by the substitution of glutamate with alanine (Table 4). The hydrocarbon chains of the PE boundary lipids with 36:2 acyl chains have conspicuously increased in the mutant by ∼20% relative to wt (Table 5). The tail composition of PE in close vicinity of αWT/βWT_−20A is hence similar to the tail composition of PE in bulk phase. The lipids that are attached to isolated αWT/βWT_−20A have thus been found altered as compared with the lipids attached to LH2 wt. The specificity of lipid–LH2 interaction is diminished by the replacement of glutamate −20 with alanine. It can be estimated that while in LH2 wt nearly all of the potential lipid binding sites have a PE molecule attached, in αWT/βWT_−20A one-third of the sites have a PC molecule attached. This indicates that the conserved residue glutamate −20 contributes to the crowding of PE at the LH2 lipid interface. The fully protonated ammonium group of PE is suited to electrostatically interact with the negatively charged side-chain of glutamate −20 contrary to the N⁺ of PC which is shielded by the three methyl groups. In addition, the ammonium group of PE has the capability of forming hydrogen bonds with the carboxy group of Glu, a property not shared with PC. The presence of a combined positively charged and polar residue, namely Lys-17 and Gln-16, in the immediate vicinity of Glu-20 further supports the notion that this stretch of the β-TMH constitutes a lipid-binding motif. Positively charged residues are present in lipid-binding motifs primarily at the n-side of the membrane (Palsdottir and Hunte, 2004). In case of spectrin the sequence stretch, IAEWKDGL appears to be essential for binding to PE enriched membranes as shown by deletion mutation studies (Hryniewicz-Jankowska et al., 2004). The lipid-binding motif comprises two negatively charged, one aromatic and one positively charged residue (Fig. 5A). However, single residue substitutions within the proposed motif have not been carried out and the significance of the single residues is not known. Similarly, the aa sequence of the N-terminal edge of the LH2 β-TMH contains residues with negatively and positively charged side-chain. In addition, a lysine residue in combination with a polar residue is frequently found in lipid-binding motifs (Wakeham et al., 2001; Palsdottir et al., 2003; Palsdottir and Hunte, 2004). The putative PE binding site is illustrated in the X-ray structure of Rps. acidophila (Fig. 5). Lys-17 and Gln-16 as well as Glu-20 are clearly making up one interface of the β-TMH (Fig. 5B) and thus are likely to constitute a lipid-binding surface within the proposed binding motif. PE was previously suggested to be selectively accumulated on the cytoplasmic face of the ICM in R. sphaeroides (Marinetti and Cattieu, 1981), which further supports the notion that PE specifically binds to the putative lipid-binding motif at the cytoplasmic edge of the β-subunit. Furthermore, LH complexes from R. capsulatus have been shown to be stably inserted into the membrane in vitro only when associated with PE (Pucheu et al., 1999). These findings also suggest a close interaction between LH2 and PE molecules.

As apparent from the LH2 high-resolution structure from Rps. acidophila (McDermott et al., 1995), βGlu-20 interacts with the Car molecule, rhodopin glucoside, particularly the glucoside moiety (Fig. 5). In R. sphaeroides, the major Car is not rhodopin glucoside but spheroidenone. These carotenoids have similar polyene chains but differ in their headgroups; in rhodopin glucoside it is the glucose group, in sphaeroidenone a methoxy and a keto group (see Fig. S1). An inspection of the X-ray structure indicates that these functional groups may be close to the conserved Glu-20 in the LH2 from R. sphaeroides (Fig. 5). To examine the effect of the Car polar end groups on LH2–lipid interactions, we set out to replace the native sphaeroidenone with neurosporene. In the R. sphaeroides mutant strain, DG2, the last step of the sphaeroidene biosynthesis is disrupted (Hunter et al., 1994). This strain has as major Car the biosynthetic precursor neurosporene, which lacks the polar end groups of spheroidenone, the methoxy and keto groups (see Fig. S1). To study the effect of the Car on the membrane properties, LH2 wt and mutant LH2 αWT/βWT_−20Q, αWT/βWT_−20A and αWT/βWT_−20K were expressed in R. sphaeroides strain DG2 with neurosporene as major Car. Wt and mutant LH2 complexes have closely similar spectral properties indicating that the functional assembly of the LH2 mutant complexes is retained also in the presence of neurosporene instead of sphaeroidenone (Fig. S1). Some minor modifications are observed in the absorption spectra in the blue region of αWT/βWT_−20A owing to increased light scattering (Kwa et al., 2004). In addition, the fluorescence excitation spectra of αWT/βWT_−20A are slightly altered in comparison to the spectra of wt LH2. The energy transfer from Cars to BChls is somewhat reduced, particularly in αWT/βWT_−20A, and there are minor alterations in the shape of the excitation spectrum which may be due to the increased scattering. However, none of the differences indicates that the functional assembly of the mutant complexes is significantly impaired by the combined alterations, the replacement of glutamate −20 with alanine, and of sphaeroidenone with neurosporene. The structural stabilities of wt LH2 and αWT/βWT_−20Q, αWT/βWT_−20A and αWT/βWT_−20K containing neurosporene as assessed by heat denaturation (see Braun et al., 2003; Kwa et al., 2004) within the native membrane are very similar (see Fig. S2). The similarity, however, no longer holds for the isolated complexes as shown for solubilized neurosporene containing LH2 wt and αWT/βWT_−20A (see Fig. S2). The exchange of sphaeroidenone with neurosporene and the replacement of glutamate −20 with alanine both result individually in destabilization of LH2 complex. If the two alterations are combined, the effect is much larger than the sum of the individual effects, indicating that the βGlu-20 and Car's polar moiety synergistically contribute to LH2 stability. However, this becomes obvious only in detergent, when most of the lipids are removed from the complex. Obviously, the surrounding lipids notably contribute in the mutant LH2 to the stability of the LH2 structure. It has not been possible to obtain sufficient amounts for the ESI-MS/MS analysis of isolated LH2 αWT/βWT_−20A from DG2 strain by detergent extraction from the membrane owing to the mutants' peculiar membrane properties. A possible role of the Car in LH2–lipid interactions has been suggested previously (Olivera and Niederman, 1993). Perhaps in LH2 wt as opposed to αWT/βWT_−20A, the surrounding lipids are bound by Glu-20 and the polar moiety of the sphaeroidenone, and therefore not removed by the relatively mild βOG treatment. Alternatively, the alterations result in conformational changes which are less stable, in particular, upon delipidation.

To further explore the relation between membrane morphology and LH2, the membranes of LH2 wt and αWT/βWT_−20A are compared in R. sphaeroides DD13 and DG2 strains. Independent of the major Car, the vesicular ICM containing LH2 wt are of oval shape with sizes ranging from 35 to 55 nm (Figs 2B and 6A). The ICM containing αWT/βWT_−20A have also a very similar morphology, indicating that the changes at the LH2–protein–lipid interface upon mutation of glutamate −20 are not resulting in altered membrane morphology in sphaeroidenone-producing cells. On the contrary, the ICM of LH2 αWT/βWT_−20A in the neurosporene-producing DG2 strain has a clearly distinct morphology. There are membrane invaginations of the normal oval shape, but there are also substantial numbers of ‘abnormal’ cytoplasmic invaginations and of enlarged vesicular structures ranging from 35 to 90 nm in diameter (Fig. 6D). Thus, the Glu-20Ala mutation and change in Car also affect synergistically the membrane structure. In a previous work it has been shown that decreasing the LH2 content results in enlargement of the chromatophores isolated from R. sphaeroides (Sturgis and Niederman, 1996). Duplication of the number of LH2 complexes in these membrane results in a decrease of few nanometers of the average chromatophore diameter (from 37 to 41 nm). The impaired stability of DG2 αWT/βWT_−20A (Fig. 6) may result in a reduction of assembled LH2 in the membrane. In order to determine the expression levels of LH2 αWT/βWT_−20A both in DD13 and DG2 strains, the content of LH2 complex in the membranes has been compared with the respective total protein content (Table 1). In the membranes of LH2 αWT/βWT_−20A containing neurosporene, the total protein content relative to LH2 content is slightly increased (1.5×), indicating a minor reduction in the LH2 expression level as compared with LH2 wt. However, the total protein relative to LH2 is almost as much increased (1.3×) in the membranes of LH2 αWT/βWT_−20A containing sphaeroidenone (Fig. 6C). The significant enlargement in diamenter (by ∼40 nm), observed for the vesicular invaginations of DG2 αWT/βWT_−20A, is thus apparently not due to the slight reduction in LH2 level in the membrane. Interestingly, the morphology of enlarged vesicular ICM invaginations is also observed in the Car-less R. sphaeroides R26.1 strain (Fig. 6B) possibly supporting the membrane shaping effect of the Car. The effect on the membrane may be related to the reduced affinity for PE, which may be even further reduced by the alteration in the Car end group in immediate vicinity of βGlu-20. The data are yet insufficient to determine whether the PE–LH2 interaction is altered merely because of the change in carotenoids' chemical structure and aa side-chain or by a conformational change of the N-terminal domain of the LH2. Taken together, these findings suggest that the N-terminal edge of the LH2 complex, comprising glutamate −20 of the β-subunit and the carotenoids' polar head group synergistically contribute to the morphogenesis of the vesicular intracytoplasmatic membranes by specific interactions with adjacent lipids.

What are the molecular features of the LH2 complex that determine the vesicular membrane curvature? Monounsaturated PE molecules are postulated to favour non-bilayer curvature owing to their physico-chemical properties and relative geometries. The postulated PE binding site of the LH2 could function to concentrate PE selectively in the cytoplasmic leaflet. However, induction of positive curvature by only lipid distribution effects would require PE accumulation at the periplasmic bilayer leaflet because of its small head group volume in relation to the acyl tail volume (Chernomordik and Zimmerberg, 1995; Farsad and De Camilli, 2003). In addition, mere lipid distribution effects appear to induce buds of much larger sizes than the typical invagination of the ICMs in R. sphaeroides (Sarasij et al., 2007).

According to the ‘bilayer-couple’ hypothesis, protein-mediated membrane deformation could be driven by insertion of protein into one leaflet only (Farsad and De Camilli, 2003; Antonny et al., 2005). It has been proposed that membrane insertion by an amphipathic helix into one leaflet is sufficient per se for facilitating budding events on lipid monolayers (Ford et al., 2002). Extending the ‘bilayer-couple’ hypothesis, a possible notion for the lipid dependent membrane shaping effect of the LH2 complex may be envisioned as follows: binding of PE selectively to the N-terminal edge of LH2 should result in the formation of a stably bound lipid rim at the cytoplasmic edge of LH2. Our data suggest that about six PE molecules bind to the isolated LH2 complex. This may even underestimate the actual number of PE bound as some may be removed by the detergent treatment. It has been estimated that 50–100 lipid molecules per LH2 are found in the membrane (Bustamante and Loach, 1994). The PE rim at the N-terminal edge should thus effectively enlarge the surface area of the LH2 complex in the cytoplasmic bilayer leaflet. Thus, the selective binding of PE should induce the bilayer surface imbalance as required for membrane deformation, specifically, the membrane curvature for biogenesis and maintenance of the ICM invagination.

An alternative model for the mechanism of membrane deformation has recently been introduced based on the induction of chirality and tilt of specific lipid molecules in membrane domains (Sarasij et al., 2007). Clustering of proteins and lipids into distinct domains has been discussed as a requirement for many cellular budding events. Distinct membrane domains, so-called CM sites, have been proposed to be the site of the assembly of photosynthetic units (Reilly and Niederman, 1985; Koblizek et al., 2005). During the stage of LH2 accumulation, CM invagination is stimulated at these sites (Koblizek et al., 2005). Accumulation of PE at distinct membrane sites would result in loosely packed lipid head group areas which may promote recognition and partitioning of LH2 (Antonny et al., 2005; Mesmin et al., 2007). The distinct acyl chain composition of the LH2 associated PE lipids, i.e. more saturated acyl side-chains, also points at distinct lipid domains in which LH2 is enriched. Longer and saturated acyl tails in combination with cholesterol have previously been found to accumulate in membrane microdomains. In prokaryotes, which do not synthesize sterols, however, cholesterol or related compounds are absent. Because of somewhat similar chemical properties, carotenoids may replace cholesterol in their role of formation of lipid domains and preferentially interact with PE molecules that have saturated fatty acyl chains (Wisniewska et al., 2003). Interestingly, carotenoids have recently been found, just as cholesterol, in association with long-chain saturated lipids (Wisniewska et al., 2003). Accumulation of PE in domains may thus be achieved by specific interactions with Car molecules.

In turn, the specific interactions between the LH2 and the boundary PE lipids may induce ordered arrays of PE with distinct chain tilt or head group orientation at the CM sites. Such ordering has been proposed to drive membrane invaginations of particular size and shape (Sarasij et al., 2007). It could effectively be disturbed by replacement of some of the PE by PC molecules as shown to be the case in the mutant LH2 in which glutamate −20 is replaced by alanine. Light inducible chiral macromolecular organization of higher plant LHC-lipid assemblies have previously been reported on in the membrane system of chloroplasts (Garab et al., 2000; Simidjiev et al., 2004). This long range order of photosynthetic proteins is dependent on specific non-bilayer lipids. The thylakoid membrane is characterized by a high structural stability required for optimal function and at the same time structural flexibility allowing for rapid diffusion of mobile components. This apparent duality has been hypothesized to depend on an optimum protein/lipid ratio which may be effectively regulated by incorporation of high amounts of non-bilayer lipid (Garab et al., 2000).

In conclusion, the data presented in this work show that massive mutagenesis of LH2 complex results in changes of the morphology and lipid composition of the intracytoplasmatic membrane of R. sphaeroides. These findings extend previous studies which have shown that in LH2 deletion mutants (LH2^- LH1⁺ RC⁺) ICM shape is significantly altered; and that onset of photosynthetic growth is accompanied by changes in membrane lipid composition. However, we also discovered that not only the absence of LH2 but changing its protein sequence results in altered membrane properties. The membranes containing such mutants are of tubular shape and contain elevated levels of PC complemented by reduced levels of PE. We revealed that PE is selectively enriched at the LH2–lipid interface of R. sphaeroides. Conserved residue, β-glutamate-20, which is located at the N-terminal edge of the TMH, and the adjacent keto carbonyl groups of the complexes' carotenoids make up a potential PE binding site in LH2. Modification of the glutamate and the Car at this site significantly impairs the selective binding of PE. However, the most striking result from this study is that the modifications of the LH2 PE binding site also result in significant changes of the vesicular membrane shape leading to the conclusion that selective binding and crowding of PE at the LH2 lipid interface is essential for maintaining the exact morphology of the surrounding lipid membrane.

The bacterial strains used in this work include Escherichia coli strain S17-1 [(thi pro hsdR_ hsdM_ recA RP4-2 (Tc::mu Kan::Tn7)] and R. sphaeroides strain DD13 and DG2 (genomic deletion of both pucBA and pufBALMX; insertion of SmR and KanR genes respectively) (Jones et al., 1992). The mobilizable plasmids used were based on pRKCBC1 (TcR, derivative of pRK415; insertion of a 4.4 kb fragment encompassing pucBAC); briefly, this expression vector contains the pucBA genes as a 420 bp KpnI–BamHI insert (Jones et al., 1992). Growth conditions for E. coli and R. sphaeroides were as described in (Fowler et al., 1995). For E. coli, tetracycline was used at concentrations of 10 mg ml⁻¹. For R. sphaeroides, the antibiotics were tetracycline (1 mg ml⁻¹) and neomycin (10 or 20 mg ml⁻¹). Conjugative transfer of plasmid from E. coli S17-1 to R. sphaeroides was performed as described (Fowler et al., 1995).

The construction of LH2 αAL_{16 4S}/βAL₁₂ has been carried out as described previously (Kwa et al., 2004). LH2 αWT/βWT_−20Q, αWT/βWT_−20A and αWT/βWT_−20K was constructed by site-directed mutagenesis (QuikChange II, Stratagene) by directly mutating pucB in pRKCBC1 as described in Garcia-Martin et al. (2006).

Rhodobacter sphaeroides membranes were prepared from cells grown semi-aerobically in the dark by disruption in a French pressure cell and subsequent centrifugation on a sucrose step gradient (Braun et al., 2002).

Denaturation was carried out as described in Kwa et al. (2004). Purified LH2 membranes were adjusted to A850 = ∼4 cm⁻¹ in TE buffer (10 mM Tris, 1 mM EDTA, pH 8.0) and measured in thermostated quartz cuvette. Circular dichroism spectra were recorded with 1 nm s⁻¹ scan rate under temperature control. During temperature experiments, the heating rate is 2°C min⁻¹, recorded from 15°C to 95°C with integration time of 0.2 s at 845 nm. Data acquisition is done by spectra manager software, analysed, plotted or smoothed by data analysis software Origin 7.0 (OriginLab Cooperation, Northampton, MA, USA).

Protein concentrations were determined either by protein assay kits from Fluka advance (Seelze, Germany), Roche ESL (Basel, Switzerland) and Pierce BCATM (Rockford, USA) or from the absorption at 280 nm (absorption coefficient at 280 nm calculated from the amino acid composition, . All the samples were prepared according to the manufacturer's protocol; and repeated more than three times each. For the estimation of the LH2 protein content, the extinction coefficient of B850 BChl was taken as 120 mM⁻¹ cm⁻¹ (Clayton and Clayton, 1981).

Modelling of the interaction between the β-TMH and Car was carried out by use of the high-resolution data of Rps. acidophila (McDermott et al., 1995). Replacements of the Car atoms (Rps. acidophila > R. sphaeroides) and subsequent energy minimizations were done using WebLab Viewer 3.7.

Cells were fixed immediately after collection with 2.5% (v/v) glutardialdehyde (Fisher Scientific Co., Fair Lawn, NJ, USA) in 75 mM sodium cacodylate, 2 mM MgCl2, pH 7.0, for 1 h at room temperature; rinsed several times in fixative buffer and post-fixed for 1 h with 1% osmium tetroxide in fixative buffer at room temperature. After two washing steps in distilled water, the cells were stained en bloc with 1% uranyl acetate in 20% acetone for 30 min Dehydration was performed with a graded acetone series. Samples were then infiltrated and embedded in Spurr's low-viscosity resin (Spurr, 1969). After polymerization, ultra thin sections with thickness between 50 and 70 nm were cut with a diamond knife and mounted on uncoated copper grids. The sections were post-stained with aqueous lead citrate (100 mM, pH 13.0). All micrographs were taken with an EM 912 electron microscope (Zeiss, Oberkochen, Germany) equipped with an integrated OMEGA energy filter operated in the zero loss modes.

For mass spectrometry analyses, R. sphaeroides are grown semiaerobically in the dark at 28°C, and cultures were harvested in their mid-logarithm phase when the absorbance at 650 nm reached 1.2–1.5. To extract total phosholipids, 0.5 ml of sample + 1 ml of methanol was vortexed throughout and left on ice for at least 5 min. Three millilitres of chloroform and 3 ml of water were added and vortexed. Mixtures were centrifuged (3000 g for 15 min at 4°C) and the bottom phase was collected and transferred to a clean tube and dried under stream of nitrogen. Lipid profiling of R. sphaerodies LH2 samples was carried out by electrospray ionization mass spectroscopy (ESI-MS/MS) essentially as described in Brugger et al. (1997). Small sample aliquots (1–10 μl) of cells, chromatophores or isolated LH2 were added to 110 μl of a ammonium acetate (5 mM) in methanol spiked with a mixture of lipid standards (PE/PC/PG). Without prior mixing, the samples were sonicated for 5 min at RT. After mixing precipitated proteins were pelleted at 16000 g in a table top centrifuge at 4°C. The supernatant was transferred to another microtube and subjected to mass spectroscopic analysis. Microflow-ESI-MS/MS analysis was performed on a Micromass QII triple-stage quadrupole tandem mass spectrometer equipped with a microflow-ESI source (Z spray) from Micromass (Manchester, UK). Argon was used as collision gas at a nominal pressure of 2.5 × 10⁻³ millibar. The cone voltage was set to 50 V for PC, 45 V for PG and to a cone ramp of 40–65 V in a mass range of mass/charge (m/z) 600–1000 for PE respectively. Resolution of Q1 and Q3 was set to achieve isotope resolution. Quantification of PE was performed by neutral loss scanning, selecting for a neutral loss of 141 (positive ion mode) at a collision energy of 27 eV (1 eV = 1.602 × 10–19 J). PC quantification was performed by precursor ion scanning for fragment ion m/z 184 (positive ion mode, collision energy of 32 eV). PG quantification was performed by precursor ion scanning for fragment ion m/z 171 (negative ion mode) with a collision energy of 37 eV. Unsaturated PE and PG standards were synthesized and purified via HPLC as described (Koivusalo et al., 2001). Quantitative analyses were performed as described (Brugger et al., 2000; 2004). Phosphate determination was performed according to (Rouser et al., 1970). The significance of data was tested by analysis of variance with repeated measures.

UV-visible absorbance spectra were recorded on a Lamda 25 spectrophotometer (PerkinElmer Life Sciences) or Shimadzu UV-2401PC. Circular dichroism measurements were performed on a Dichrograph CD6 (Jobin Yvon, Division Instruments, USA) in 1 mm or 1 cm cylindrical Quartz cuvette or Jasco J715 spectropolorimeter in 1 mm rectangular Quartz cuvette (Hellma, Mühlheim, Germany). The fluorescence excitation spectra of R. sphaeroides chromatophores were recorded by Spex FluoroLOG spectrofluorometer (NJ, USA). Excitations were scanned from 300 to 850 nm with the emission wavelength of 880 nm.