The differing composition of LH2 (peripheral light-harvesting) complexes present in Rhodopseudomonas palustris 2.1.6 have been investigated when cells are grown under progressively decreasing light intensity. Detailed analysis of their absorption spectra reveals that there must be more than two types of LH2 complexes present. Purified HL (high-light) and LL (low-light) LH2 complexes have mixed apoprotein compositions. The HL complexes contain PucABa and PucABb apoproteins. The LL complexes contain PucABa, PucABd and PucBb-only apoproteins. This mixed apoprotein composition can explain their resonance Raman spectra. Crystallographic studies and molecular sieve chromatography suggest that both the HL and the LL complexes are nonameric. Furthermore, the electron-density maps do not support the existence of an additional Bchl (bacteriochlorophyll) molecule; rather the density is attributed to the N-termini of the α-polypeptide.
- light-harvesting complex
- puc gene
- purple photosynthetic bacterium
- Rhodopseudomonas palustris
- X-ray crystallography
Purple bacterial photosynthesis begins when solar energy is absorbed by the LH (light-harvesting) complexes. The resulting excitation energy is then funnelled to RCs (reaction centres), where charge separation occurs . Most species of purple bacteria have two types of LH complex: core (LH1) complexes and peripheral (LH2) complexes. These complexes are modular with each individual module constructed from pairs of short (50–60 amino acids) very hydrophobic polypeptides called α and β. The LH1 complex from Rhodopseudomonas palustris, for example, consists of 15 such α/β-polypeptide modules that are oligomerized to surround the RC . Each module binds two Bchl (bacteriochlorophyll) a molecules and one Car (carotenoid) molecule. The Bchl a molecules are strongly coupled, giving rise to the intense Qy absorption band at ~875 nm. In the case of the LH2 complex from Rhodopseudomonas acidophila strain 10050, each module binds three Bchl a molecules and one Car [3–5]. Nine such modules are arranged circularly to form a single LH2 complex. The α-polypeptide is located inside the ring, and the β-polypeptide is on the outside. Nine monomeric Bchl a molecules have their bacteriochlorin macrocycles oriented parallel to the plane of the membrane and absorb the light with a NIR (near-IR) absorption maximum at ~800 nm. These are called the B800 Bchl a molecules. These B800 Bchls are separated, centre-to-centre, by 2.1 nm and are considered monomeric. A further 18 Bchl a molecules have their bacteriochlorin macrocycles oriented perpendicular to the membrane plane. They are responsible for the absorption band at ~850 nm. The macrocycles of the B850 Bchl a molecules sit very close to each other, ~0.9 nm, and are strongly exciton-coupled. In the case of Phaeospirillum molischianum, the LH2 complexes are octamers .
In some species of purple bacteria, such as Rps. acidophila 7050 and 7750 or Rps. palustris, the Qy absorption bands of the Bchl in the LH2 complexes can vary depending on the growth conditions. When Rps. acidophila strains 7750 and 7050 are grown under LL (low-light) conditions, a different type of LH2 complex is formed with the Qy absorption bands at 800 and 820 nm rather than at 800 and 850 nm. This different type of LH2 complex has been called both LH3 complex and B800–820 LH2 complex [5,7,8]. The ability to change the type of LH2 complex in response to growth under different light intensities is related to the presence of multiple α/β-polypeptides, which are encoded (in the case of Rps. acidophila) by at least four different α/β-apoprotein gene pairs (called the pucBA genes) [9,10]. When cells of Rps. palustris strain 2.1.6 are grown under HL (high-light) intensity, they synthesize a standard LH2 complex with the Qy absorption bands of Bchl a at 800 and 850 nm. Under LL intensity, Rps. palustris strain 2.1.6 replaces the standard HL LH2 (B800–850) complex with LL LH2 (B800–low-850) complex [11–15]. Again, this ability to adapt and to synthesize LH2 complexes with different NIR absorption spectra is related to the presence of multiple genes encoding LH2 complex α/β-polypeptides [16,17]. The complete genome of Rps. palustris has been sequenced . There are five different pucBA genes present in the Rps. palustris genome (pucBAa, b, c, d and e) and their expression is regulated by both light intensity and light colour [18,19].
In a previous study, Hartigan et al.  described the isolation of yet a further different type of LH2 complex from the LL-grown Rps. palustris 2.1.6. This complex, called the LH4 complex by the authors, has an unusual absorption spectrum in the NIR showing only a broad band at ~800 nm. The LH4 complex has polypeptides that are encoded by pucBd and pucAd . The LH4 complex has been crystallized, and low-resolution diffraction data (7.5 Å; 1 Å=0.1 nm) have been collected for these crystals, allowing a 7.5 Å model based on these data to be described . This model suggests that the complex is octameric and that each α/β-apoprotein pair binds an extra Bchl a molecule relative to three Bchl a molecules found in the LH2 complex from Rps. acidophila. An AFM (atomic force microscopy) study of LL membranes from Rps. palustris also suggested that this LH2 complex is predominantly octameric .
The structures of the HL and LL LH2 complexes could potentially be more complicated than has been generally assumed [12,18,23]. Given the profusion of pucBA genes, the question arises as to whether the LH2 complexes in Rps. palustris consist of rings where each ring is homogeneous, containing only one type of α/β-apoprotein, with the spectroscopic variation due to mixtures of different types of rings, or alternatively consist of rings that can be heterogeneous, having mixtures of α/β-apoprotein types within a single LH2 complex ring. In the later case, the resultant spectroscopic properties will then depend on the compositions of these mixtures. This has been partially addressed using a range of spectroscopic techniques, which have included electronic absorption, CD, fluorescence and resonance Raman studies on ensembles [11–15,24]. It was hypothesized that in the LL LH2 complexes, owing to the unique excitation wavelength dependency on the Raman spectrum in the high-frequency Bchl a carbonyl region, and the inability to isolate a native B800–820 complex, rings containing multiple α-apoproteins are present . Quite recently, we have investigated the properties of the LL LH2 complex using single-molecule spectroscopy . The single-molecule spectra for the LL LH2 complex could only be satisfactorily explained by assuming the presence of a nonameric LH2 complex ring with a mixture of both B850-like and B820-like site energies . As the B850-like and B820-like Bchls are associated with different apoproteins, the spectra could only be explained by assuming the presence of rings with a heterogeneous apoprotein composition. The heterogeneity concept from single-molecule spectroscopy was also supported by femtosecond transient absorption spectroscopy, where high-energy exciton states were found in the regions of both 820 and 850 nm . In the present paper, we describe a detailed biochemical characterization of the LH2 complexes from Rps. palustris. We set out to investigate whether there are, in fact, more than two types of LH2 complex and, if so, what are their quaternary structures.
Cells of Rps. palustris strain 2.1.6. were grown anaerobically in C-succinate medium  at 30°C between rows of incandescent bulbs . In the present study, cultures were also placed at different distances away from a single bulb so that the intensity of light illuminating the cell culture could be varied from 220 to 90, 20, 10, 6.5 and 5.5 lx. In the Results section, these light intensities have been called HL, LL1 (low-light intermediate 1), LL2 (low-light intermediate 2), LL, FLL1 (far low-light) and FLL2 (extreme low-light) respectively. To minimize any self-shading, cells were regularly transferred into fresh medium to ensure a continual low culture attenuance (D850=0.5 cm−1). Cells were harvested by centrifugation (relative centrifugal field, rav=1248 g, 30 min) and resuspended in 20 mM Mes buffer (pH 6.8) (Sigma–Aldrich) containing 100 mM KCl. The harvested cells were ruptured by three passages through a French Press (950 psi; 1 psi=6.9 kPa) in the presence of a little DNase and MgCl2 and the membrane fraction was pelleted by centrifugation at rav=184000 g for 2 h. The membrane pellet was resuspended and homogenized in 20 mM Tris/HCl buffer (pH 8.0) (Fisher Scientific) and the concentration was adjusted to give an absorption at 850 nm of 70 cm−1. The membranes were then solubilized by the slow dropwise addition of LDAO (N,N-dimethyldodecylamine-N-oxide) (Fluka) to 1% (v/v) final concentration. After stirring for 30 min in the dark at room temperature (21°C), any unsolubilized material was removed by centrifugation at rav=16000 g for 10 min. The supernatant was then layered on to a sucrose step gradient. The gradient consisted of 0.8, 0.6, 0.4 and 0.2 M sucrose prepared in 20 mM Tris/HCl buffer (pH 8.0) containing 0.1% LDAO. The gradients were then centrifuged for 16 h at rav=149000 g at 4°C. The upper-pigmented band contained the LH2 complexes and the lower pigmented band contained the LH1–RC complexes. The band containing LH2 complexes was collected and purified on a DE-52 cellulose column (Whatman). After a desalting step using a PD-10 column (GE Healthcare), the LH2 complexes were purified further by chromatography on a Resource-Q column (1 ml, GE Healthcare) and subsequently passed through a Superdex-200 gel-filtration column [1.6 cm i.d. (internal diameter)×100 cm long, GE Healthcare]. To ensure high purity, only fractions of the LH2 complex with a ratio between the absorption maximum at ~800 nm and the maximum at the protein absorption (A280) above 3.0 were collected for both spectroscopic analysis and crystallization trials.
Before size-exclusion chromatography analysis, the HL and LL LH2 complexes were detergent-exchanged into DDM (dodecyl maltoside), accomplished by washing the LH2 complex sample five or six times with 20 mM Tris/HCl buffer (pH 8.0) containing 0.02% DDM using a Vivaspin2 centricon [50000 Da MWCO (molecular-mass cut-off), Sartorius Stedim Biotech]. The DDM exchange was needed because one of the standards, LH1–RC complex from Rhodospirillum rubrum S1, is only stable in DDM. The intact size of the HL and LL LH2 complexes were determined by passing the purified LH2 complexes through a Superdex-200 gel-filtration column (1.6 cm i.d.×100 cm long) using 0.02% DDM in 20 mM Tris/HCl buffer (pH 8.0) and comparing their elution profile with that of known standards also in 0.02% DDM. The flow rate was 0.5 ml/min.
The purity and the size of the LH2 complex apoproteins were assessed by running on 12% NuPAGE Novex Bis-Tris mini gels (Invitrogen) using Novex Sharp protein standard (Invitrogen) molecular-mass markers. The protein (10 μl, A850=40 cm−1) samples were mixed with 5 μl of 300 mM DTT (dithiothreitol) and 5 μl of NuPAGE SDS sample buffer. After heating in a water bath at 70°C for 20 min, the mixtures were loaded on to the gel. NuPAGE SDS 1×running buffer (Invitrogen) in deionized water was used for the running buffer. The electrophoresis conditions were constant at 200 V for 40 min. The gels were stained with SimplyBlue SafeStain (Invitrogen) and destained with deionized water overnight.
The purified HL and LL LH2 complexes were sent to the FingerPrint Proteomic Facility, University of Dundee, Dundee, U.K., for polypeptide identification. Each sample was first digested in-gel with trypsin before analysis by one-dimensional nano-LC coupled to ESI (electrospray ionization)–MS/MS (tandem MS) using a 4000 QTRAP (Applied Biosystems) MS/MS system. The molecular mass of the LH2 complex peptide fragments were compared with hypothetical fragment masses predicted from their gene sequences, published online by UniProtKB/Swiss-Prot (based on the complete genomic sequence of Rps. palustris reported by Larimer et al. ).
Room temperature absorption spectra of cells, membranes and solubilized complexes were measured in a Shimadzu UV-1700 spectrophotometer scanning from 250 to 950 nm. The 10K absorption spectra of LH2 complexes were measured using a Varian Cary E5 double-beam scanning spectrophotometer. The LH2 complex samples were prepared for the low-temperature spectra by the addition of 60% (v/v) glycerol in 20 mM Tris/HCl buffer (pH 8.0) containing 0.1% LDAO. The temperature of these samples was maintained by a helium bath cryostat (Maico Metriks).
CD, fluorescence and fluorescence excitation spectroscopy were carried out following the method of Collins et al. . The CD spectra were recorded with a Jasco J-815 spectropolarimeter with a bandwidth of 4 nm. Samples were measured in a 1 mm demountable quartz cuvette (Starna Scientific). Fluorescence emission and fluorescence excitation spectra were recorded using a Photon Technology International fluorimeter equipped with a red-sensitive avalanche photodiode detector (Advanced Photonics). Concentrated LH2 complex samples were prepared in 66% (v/v) glycerol in 20 mM Tris/HCl buffer (pH 8.0) containing 0.1% LDAO. The low-temperature environment was created using a liquid nitrogen cryostat (OptistatDN, Oxford Instruments).
The resonance Raman spectra were recorded with a Jobin-Yvon U1000 spectrometer equipped with a back-thinned CCD (charge-coupled device) camera. The spectra were recorded with 90° geometry with the samples maintained at 77 K in a SMC-TBT flow cryostat (Air Liquide) cooled with liquid nitrogen. The samples were excited at 363.8 nm with a Coherent Inova 100 Ar+ laser with an incident intensity of less than 10 mW at the sample surface.
X-ray crystallography was used to attempt to determine the three-dimensional structures of both the HL and the LL forms of the LH2 complex from Rps. palustris 2.1.6. Initial crystallization conditions for these proteins were obtained using commercial HT sparse matrix screens available from Molecular Dimension using sitting-drop vapour diffusion. Robotic equipment, including the Cartesian Honeybee 8+1 nanodrop robot (Digilab), the Microlab Star liquid dispensing robot (Hamilton Robotics) and the Rhombix Imager (Thermo Fisher Scientific), was used in this process. The best conditions from the screens were then optimized in macroscale using a 24-well EasyXtall Tool tray (Qiagen). These best conditions were as follows. The purified LH2 complex proteins were first washed three to four times with 20 mM Tris/HCl buffer (pH 8.0) containing 0.1% LDAO, then concentrated using a Vivaspin2 centricon (50000 Da MWCO). The final concentrations of LH2 complex proteins were adjusted to give an A850 of 80 cm−1 for HL LH2 complex and A800 of 80 cm−1 for LL LH2 complex respectively. A 10 μl volume of the LH2 complex protein in 20 mM Tris/HCl buffer (pH 8.0) containing 0.1% LDAO was pipetted on to a sitting drop bridge, while the reservoir chamber was filled with 1 ml of the solution containing 0.1 M Tris/HCl (pH 9.5), 35% PEG [poly(ethylene glycol)] 400, 0.1 M NaCl and 0.1 M MgCl2 in the case of HL LH2 complex, and 0.1 M Tris/HCl (pH 9.5), 37% PEG 400, 0.1 M NaCl and 0.1 M MgCl2 in the case of LL LH2 complex. A 10 μl volume of the reservoir solution was then mixed gently into the LH2 complex drop on the bridge. The crystallization wells were subsequently sealed, and the tray was placed in a temperature-controlled incubator (at 10 and/or 20°C). The 10°C temperature was particularly important for the HL LH2 complex crystallizations as it reduced rapid precipitation. Crystals longer than 300 μm grew within a few weeks. Useful LL LH2 complex crystals were also obtained from the following conditions containing 50 mM glycine (pH 10), 33% PEG 1000 and 50 mM NaCl in the reservoir. X-ray diffraction data from several of these crystals was collected at the European Synchrotron Radiation Facility in Grenoble, France, and at the Diamond Light Source, Oxfordshire, U.K. All datasets were processed and scaled using the d*TREK program . Further calculations were performed using the following programs: PHASER  for MR (molecular replacement), MOLREP  for self-rotation function, and REFMAC  for structure refinement, which are all parts of the CCP4 suite of programs . Electron-density maps were examined and Figure 10 was generated in the program COOT .
Growing the cells
Absorption spectra of whole cells grown under different light intensities were recorded and are presented in Figure 1. The HL-grown cells (Figure 1, –––– line) show an absorption spectrum in the NIR region, rather similar to those from Rps. acidophila and Rhodobacter sphaeroides, with two strong absorption maxima at 805 and 862 nm that originate from the HL LH2 complexes . The B850 band is the most intense and has a shoulder at around 875 nm, indicating the presence of LH1–RC complexes (Figure 1). As the light intensity is decreased, the NIR absorption spectra of the cells become markedly different. The 800 nm absorption band is more intense in the LL-grown cells, and the ~850 nm band becomes correspondingly weaker and broader. As a result, the core 875 nm band is more apparent in the LL-grown cells (Figure 1, ·–·–· line). This situation reverses when the Rps. palustris cells are grown under the extreme lowest-light intensities, FLL1 and FLL2 (Figure 1, ··–··–·· and ····· lines respectively). Under such extreme conditions, the intensity of the ~850 nm absorption band recovers towards its intensity seen in the HL spectrum and narrows correspondingly. The relative molar ratio of LH2 complex to LH1 complex also varies depending on the light intensity at which the cells were grown. This ratio varies from approximately 3:1 under the highest light intensity used to 4.8:1 under the lowest light intensity used (Figure 2). Figure 2 also shows how the ratio of the absorbance at 862 nm to that at 805 nm varies with the light intensity under which the cells are grown. This ratio reflects the shift between the HL LH2 complex (high ratio) and the LL LH2 complex (low ratio). The value of this ratio reaches a minimum at 10 lx and then rises again under the lowest light intensities. This ‘dip’ in the ‘850/800’ ratio as Rps. palustris cells are grown under progressively decreasing light intensities is consistently observed. The precise physiological reason for the return of a B800–850-‘like’ complex (often alongside the B800–820 complex) under very low light intensities has not yet been studied in any detail, but a similar effect has also been reported previously for Rps. acidophila .
Purification and protein characterization
The HL and LL LH2 complexes were isolated and purified as described in the Experimental section. Their respective room temperature absorption spectra are presented in Figure 3(A). The HL LH2 complexes have Bchl a Qy absorption bands at 803 and 857 nm (Figure 3A, continuous line), whereas the LL LH2 complexes have absorption bands at 802 and 850 nm. The differences (Figure 3A, inset) in the intensities of the B800 and B850 absorption bands in the two types of purified LH2 complexes are consistent with those described above in the case of whole cells. If there were only two types of LH2 complexes possible, i.e. HL LH2 complex and LL LH2 complex, then overlaying the NIR absorption spectra of the LH2 complex fractions prepared from cells grown under the different light intensities should reveal an isosbestic point. If, however, there are more than two types of LH2 complex possible, then a single isosbestic point would not be expected with the overlaid absorption spectra. In order to make this measurement as clear as possible, the absorption spectra used for these comparisons were recorded at 10 K (Figure 4). These LH complexes are fully intact at cryogenic temperatures, and their energy transfer reactions are essentially unchanged even down to 7 K . At this temperature, the B800 absorption band sharpens and the B850 band both sharpens and red-shifts. The expanded potential isosbestic region at ~845 nm is shown in the inset in Figure 4. These overlaid spectra were very carefully normalized in the Qx band region at ~595 nm, and it can be seen that there is no single well-defined isosbestic point.
The polypeptide composition of the purified HL and LL LH2 complexes was investigated using NuPAGE Bis-Tris gel electrophoresis. Typical results are presented in Figure 5. Both the HL and LL LH2 complexes have four bands, two at ~10 kDa (H1 and L1) and two overlapping at ~7 kDa (H2 and L2). Owing to the similarity between the different possible Puc polypeptides, it is not possible to determine from such gels which of these bands correspond to which polypeptide. The identity of these polypeptides was therefore investigated using nano-LC–ESI–MS/MS. The bands seen on the gels were individually excised before MS analysis. The results of this analysis are shown in Figure 6. In the case of the HL LH2 complexes, the two weak bands (H1) can be assigned as the PucAa and PucAb apoproteins and the two overlapping bands (H2) can be assigned as the PucBa and PucBb apoproteins. In the case of the LL LH2 complexes, the two weak bands (L1) can be assigned to the PucAa and PucAd apoproteins and the two overlapping L2 bands can be assigned as the PucBa, PucBd and PucBb apoproteins.
Size-exclusion chromatography of the HL and LL LH2 complexes in the presence of DDM was used to estimate their relative size. Their elution profiles (Figure 7) were compared with the LH2 complex from Rps. acidophila 10050 (nonameric), the LH2 complex from Phs. molischianum (octameric) and the LH1–RC complex from Rsp. rubrum S1 (16-meric) [3,3,37]. Both the HL and LL LH2 complexes have similar sizes. Moreover, both of these complexes appear to be slightly bigger than the LH2 complex from Rps. acidophila.
CD spectroscopy can be used to investigate the organization of both Bchl a molecules and carotenoids in purple bacterial LH complexes [7,38,39]. The CD spectra of the HL and LL LH2 complexes in the NIR (700–950 nm) region are presented in Figure 8. There are common features shown in the spectra of both HL and LL LH2 complexes. There is a negative band at around B800 and an S-shaped band, with zero-crossing close to the absorption maximum of the B850 band. A detailed inspection of the CD spectrum of the HL LH2 complexes shows the presence of a negative band (–) at 794 nm and an S-shaped band [846 (+) and 873 (–) nm] with a zero-crossing at 857 nm. In the LL LH2 complexes, the CD spectrum shows that the negative B800 band is located at 789 nm. The magnitude of the S-shaped band at B850 decreases and its zero-crossing is red-shifted to 849 nm. There is an additional CD band [805 (+) and 819 (–) nm] in the spectrum of the LL LH2 complexes (Figure 8). Thus the LL LH2 complex CD spectrum shows a 789 (–) nm band and two S-shape bands [804 (+) and 819 (–) nm, and 839 (+) and 869 (–) nm]. The CD spectrum profiles in the UV–visible (350–700 nm) region of the HL and LL LH2 complexes are very similar. There are positive bands in the Soret region (350-400 nm) and the Qx region (589 nm) of Bchl a molecules, and a broad S-shaped profile in the carotenoid region (400–550 nm).
Figure 9 shows the fluorescence emission (broken line) and excitation spectra (continuous line) of the HL and LL LH2 complexes. Upon excitation at 800 nm, both HL and LL LH2 complexes show a single broad fluorescence emission band with the maximum intensity peaking at 897 and 892 nm, for HL and LL LH2 complexes respectively. The excitation spectra (emission at 900 nm) of the HL and LL LH2 complexes exactly mirror their absorption spectra. This clearly shows that energy transfer reactions between the pigments within these isolated LH2 complexes are fully intact.
To elucidate further the interactions of the Bchl a molecules within these complexes, Raman spectra were measured at 77 K under resonance conditions with the Soret electronic transition (363.8 nm). Under these conditions, all of the Bchl a molecules of the LH2 complexes are expected to contribute more or less equally . Figure 10 displays the high-frequency spectra (1590–1725 cm−1) for the HL (black line), LL1 (mid-grey line), LL2 (light-grey line) and LL (dark-grey line) LH2 complexes. All of the spectra are dominated by a band at 1609 cm−1 attributed to the methane-bridge-stretching modes of Bchl a molecules having their central Mg2+ ion in its 5-co-ordination state . The contributions from the carbonyl-stretching modes (1620–1710 cm−1) are relatively small in comparison and, for this reason, Figure 10(B) shows an enlarged view of most of this region (1640–1710 cm−1). The bands centred at 1627, 1658 and 1667 cm−1 are attributed to the C3-acetyl and C131-keto Bchl a carbonyl-stretching modes . The band at 1697 cm−1 is attributed to the free-from-interaction C131-keto carbonyl group of the B800 Bchl a . It is evident that, going from HL to LL via LL2 and LL1, there is a progressive reduction of the ~1620–1630 cm−1 region, which is accompanied by an increased intensity in the vibrational modes located between ~1650 and ~1680 cm−1. From HL to LL, there is also a small rearrangement in the spectral region of the keto carbonyl-stretching mode of the B800 Bchl molecule (1697 cm−1), but this is at the limit of detection (Figure 10B).
Reasonably sized (largest dimension >300 μm) and quality HL and LL LH2 complex crystals were found to diffract X-rays at the synchrotron up to a resolution of approximately 4.5 and 4.7 Å respectively. The absorption spectra of the redissolved crystals are shown in Figure 3(B), confirming that both HL and LL LH2 complexes were successfully crystallized. These crystals show monoclinic lattice symmetry, P21 or C2. Diffraction was, unfortunately, relatively anisotropic for all crystals tested, i.e. the diffraction limits were different in different spatial directions. Subsequently, a satisfactory merging of the diffraction data was only possible to a resolution of 6.3 Å in the case of the HL LH2 complex P21 crystal, 6.2 Å for the LL LH2 complex P21 crystal and 6.5 Å for the LL LH2 complex C2 crystal. This ‘standard’ data processing has excluded some of the highest-resolution diffraction spots. Therefore a special anisotropic scaling and ellipsoidal truncation with the use of the Diffraction Anisotropic Server  was applied to both datasets for the LL LH2 complex crystals in an attempt to correct for these effects. This ‘anisotropic’ processing produced a 5.0 Å dataset for the C2 crystal of LL LH2 complex and a 5.6 Å dataset for the P21 crystal. The summary of crystal data, including space groups, cell dimensions and data indexing and processing statistics for selected best HL and LL LH2 complex crystals is presented in Table 1. The unit cell volume and lattice symmetry together with the expected solvent contents of ~70% (crystals of Rps. acidophila 10050 LH2 complex [3,4] have solvent contents of 69–73%, those of Rps. acidophila 7050 crystals  are 74%, and those of Phs. molischianum LH2 complex crystals  are 71%) revealed a presence of one whole ring of LH2 complex in the crystal asymmetric unit.
In order to extract the structural information from the intensity data (both ‘standard’ and ‘anisotropic’ processing data were used), SRF (self-rotation function) maps were calculated first for all of the datasets to examine their internal symmetries . These maps show clearly that the diffraction data were produced by the ring structure, with a unique direction of the ring axis in the crystal lattice. For all datasets, the SRF maximum value for nine-fold symmetry, which coincides with the ring axis, is always slightly higher than the one for eight-fold symmetry. The difference, however, between them is within the one rms (root mean square) value of these SRF maps, therefore the ring axis symmetry cannot be determined uniquely. Consistently, the belt of the ring two-fold symmetries (perpendicular to the ring axis) is rather featureless, and does also not indicate the axial symmetry of the LH2 complex. For reference, the SRF map for our 1.85 Å LH2 complex structure from Rps. acidophila 10050 crystallized with the detergent LDAO (A. W. Roszak, A. T. Gardiner and R. J. Cogdell, unpublished work) clearly shows nine pseudo-2-fold axes perpendicular to the ring axis and the SRF maximum value for the axial nine-fold symmetry is at 10.3×map-rms level. Interestingly, the same data exhibit eight-fold and ten-fold symmetry components along the ring axis, but at much lower levels, 4.3 and 5.6×map-rms respectively, but still above the 3×map-rms value. The SRF maximum values for the eight-fold and nine-fold symmetries for our low-resolution Rps. palustris LH2 complex data are quite similar, and generally are not much higher than the values for the seven-fold or ten-fold, or even higher, symmetries. The native Patterson maps were also calculated for these datasets, but they were similarly not conclusive in distinguishing ring size. It is therefore suggested that there is rotational disorder in these LH2 complexes about their ring axes. This disorder could be due to rotation about the ring axis that is different from the basic angle between α/β-heterodimers (40° in case of Rps. acidophila LH2 complex) for the LH2 complexes in different crystal lattice cells, or due to some degree of heterogeneity of the α/β-polypeptides.
Subsequently, an MR method was applied to investigate possible solutions for all of the Rps. palustris LH2 complex datasets. The nonameric structure of the LH2 complex from Rps. acidophila 10050, the octameric one from Phs. molischianum, and the artificial/theoretical octameric model built from the Rps. acidophila 10050 α/β-heterodimer modules and their associated three Bchls, were used as MR search models. Both Rps. acidophila-based LH2 complex search models gave some satisfactory solutions in terms of the crystal lattice packing, whereas the molischianum LH2 complex solutions often produced clashes. Solutions using all three search models became possible, and often produced even better results, after some pruning of the N- and C-termini of α/β-polypeptides and/or removal of the protruding portions of the cofactors. However, the values of the associated figures of merit (the RFZ for rotational freedom, the TFZ for translational freedom and the LLG parameter, indicating the overall fit of the model in the crystal lattice ) showed that the solutions obtained with the Rps. acidophila 10050 nonameric LH2 complex model are clearly better than solutions obtained with an octameric Phs. molischianum LH2 complex model or with theoretical ‘octameric Rps. acidophila LH2 complex’ model. Resulting figures of merit for all these cases are presented in Supplementary Table S1 (at http://www.BiochemJ.org/bj/440/bj4400051add.htm). It is noticeable that the MR program  produces nine equivalent top solutions (with the highest similar LLG value) for the nonameric Rps. acidophila search model that all occupy the same space in the lattice with a rotation about the ring axis by 40°, suggestive of nine-fold symmetry in the LH2 complexes from Rps. palustris. If any lower LLG solutions were found, usually the next nine of them showed the ring rotated by 20°, effectively placing these solutions in between the top ones. This confirms the observation found in the SRF maps that the LH2 complex ring position about the axis in our Rps. palustris LH2 complex crystals is not ordered. MR solutions obtained using the theoretical octameric ‘Rps. acidophila LH2 complex model’ also occupy the same position in the crystal lattice as the nonameric Rps. acidophila model, but the LLG values indicate that the agreement between the model and the experimental data is lower. This is most likely to be due to less satisfactory crystal packing, as the diameter of the octamer is smaller than of the nonamer. In contrast, the Phs. molischianum search model gives many more top-LLG solutions (with the LLG values much lower than for the Rps. acidophila model) located in several different positions in the crystal lattice. These observations suggest that the overall shape of the Rps. acidophila LH2 complex, which is cylindrical and slightly conical, agrees better with the experimental data than the cylindrical, but much less conical, Phs. molischianum LH2 complex. The MR results also suggest that the nonameric ring architecture is preferred.
The best way to judge the quality of the MR solutions is through examination of the electron-density maps. Map quality gives an indication of the quality of phasing of the diffraction data by the used search model and therefore about the topological similarity of the search model to the real structure from which the experimental data came. The electron-density maps generated by Phaser  for the top MR solutions present a similar picture to the one described in the present paper. Figure 11 shows examples of such maps for the LL LH2 complex data in the P21 space group for the nonameric Rps. acidophila (Figures 11A and 11B) and octameric Phs. molischianum (Figures 11C and 11D) LH2 complex search models. Both models produced reasonable electron density for α-apoprotein helices, but less complete ones for the β-apoprotein helices, although β-apoprotein helices calculated using the nonameric Rps. acidophila model were slightly more complete than those calculated using the other two models. Poor density for the β-apoprotein helices possibly results from a different topology of the Rps. palustris LH2 complex in comparison with the MR models. More important differences were observed for the Bchl macrocycles as the maps generated by the Phs. molischianum model do not reveal electron density that would correspond to the bacteriochlorin macrocycles of the tightly coupled B850 Bchl a molecules. In the corresponding region of the structure calculated from the Rps. acidophila model, there is a ring of electron density for these macrocycles. Moreover, the Rps. acidophila model also produces electron density for most of the bacteriochlorin macrocycles of the monomeric B800 Bchls, also not visible in the Phs. molischianum electron density. Electron-density maps generated by the Rps. acidophila LH2 complex-based theoretical octameric complex produced some unsatisfactory features similar to the Phs. molischianum LH2 complex model maps. The electron density obtained for the best solutions do not support the existence of an additional Bchl molecule. This has been suggested by Hartigan et al.  to be present below the B800 Bchl macrocycle in the LL LH2 complex structure from Rps. palustris. The lowest bands of electron density visible in Figure 11 are associated with the N-termini of the α-polypeptides and not with Bchl macrocycles.
Attempts were also made to refine some of the better solutions by the rigid-body refinement with the α-apoprotein helices, β-apoprotein helices and B800/B850 Bchl molecules in separate rigid-body groups, but the results were rather unsatisfactory. Instead, a different mode of refinement was used that relies upon ‘jelly-body’ and local NCS restraints implemented in the latest version of Refmac . The values of R and Rfree for various data are presented in Supplementary Table S1. These refinements have produced significant improvement of the R factors in comparison with rigid-body refinement. The best R factors, 0.358 for the HL LH2 complex P21 data, and 0.382 for the LL LH2 complex P21 data were obtained for the Rps. acidophila nonameric model, whereas the Phs. molischianum octameric model refined against the same data produced values of 0.513 and 0.479 respectively. The theoretical ‘Rps. acidophila octameric LH2 complex model’ refined against the same data produced values of 0.575 and 0.494 respectively. The differences between these R factors again suggest the preference of the nonameric LH2 complex model for the Rps. palustris HL and LL LH2 complex data.
Maps from both the MR and the ‘jelly-body’ refinement at such low resolution are, however, heavily biased by whichever model is used. In order to test the model-dependence on the phasing, omit maps were calculated for the Rps. acidophila and Phs. molischianum models by the removal of one α/β-polypeptide pair from each model. Figure 12 presents omit maps calculated for both HL (Figures 12A and 12B) and LL (Figures 12C and 12D) P21 data. It is clear that density for the missing helices appears in maps for the Rps. acidophila nonameric model (in the 3-o'clock position). In contrast, the Phs. molischianum octameric model became slightly deformed and the maps reveal no density for the missing helices.
The results of two previous studies on the polypeptide composition of the LH2 complexes from Rps. palustris 2.1.6 are compared with those obtained in the present study (Table 2). Tadros and Waterkamp  proposed that there are two major polypeptides (PucABa and PucABb) and two minor polypeptides (PucABc and PucABd) present in the HL LH2 complex. It is not clear from their paper whether the two minor polypeptides were contaminants or not. Tharia et al.  found that three peptide pairs (PucABa, PucABb and PucABd) were present in the HL LH2 complex, whereas the LL LH2 complex only contained two pairs of polypeptides (PucABa and PucABd) . Interestingly, the amount of β-polypeptide present from the pucABd gene pair in the HL LH2 complex was much greater than that of the α-polypeptide. No explanation was offered for this result. In the present study, we have demonstrated that there are more than two types of LH2 complexes from Rps. palustris strain 2.1.6 synthesized under different light intensities. This is clear from the lack of an isosbestic point in Figure 4. It is therefore not surprising that our study shows that the LH2 complexes from Rps. palustris 2.1.6 are composed of a rather complicated mixture of α/β-polypeptides. In our hands, two peptide pairs (PucABa and PucABb) were identified in the HL LH2 complex, whereas PucABa, PucABd and PucBb-only were found in the LL LH2 complex (Figure 6). The exact polypeptide composition varies depending on the precise growth conditions employed and with such closely related LH2 complexes, it may well be almost impossible to separate them from each other.
As with the LH2 complex α-polypeptide from Rps. acidophila 10050, the PucAa and PucAb apoproteins from Rps. palustris strain 2.1.6 have tyrosine and tryptophan at positions 44 and 45 respectively. In the LH2 complex from Rps. acidophila 10050, the phenol side chain of α-Tyr44 and the indole side chain of α-Trp45 play important roles in formation of hydrogen bonds with the acetyl groups of the α-B850 Bchl and the β-B850 Bchl respectively. These residues are replaced by phenylalanine and methionine in the PucAd apoprotein of LL LH2 complex from Rps. palustris 2.1.6. In the B800–820 LH2 complex from Rps. acidophila 7050, the replacement of Tyr44 and Trp45 with Phe44 and Leu45 respectively results in a loss of hydrogen-bond donors for the hydrogen bonds to the Bchl acetyl groups as found in the B800–850 LH2 complex from Rps. acidophila 10050 . However, in the B800–820 LH2 complex, the acetyl group of the α-B850 Bchl forms a new hydrogen bond instead, to a hydroxy group of α-Tyr41, which was a phenylalanine residue in the B800–850 LH2 complex. The acetyl group of the β-B850 Bchl is not hydrogen-bonded in the crystal structure of the B800–820 LH2 complex. In the B800–820 LH2 complex, the Bchl C3-acetyl groups are rotated out-of-plane with respect to the bacteriochlorin plane [3,5]. This effect reduces the extent of π-conjugation in the Bchl and results in a blue shift of its site energy [45,46], reflected in the shift of the absorption band from 850 to 820 nm.
In each α/β-polypeptide, three Bchl molecules contribute to the six vibrational modes present in the high-frequency region (1620–1720 cm−1) of a Raman spectrum. These modes correspond to the C3-acetyl and C131-keto Bchl a carbonyl-stretching modes  and were distinguishable from one another in the following systematic study using the B800–850 LH2 complex from Rba. sphaeroides. The replacement of α-Tyr44 and α-Tyr45, to Phe-Tyr (single mutant) or to Phe-Leu (double mutant) produced B800–839 and B800–826 complexes respectively . Resonance Raman studies of these mutants identified the breakage of one or two hydrogen bonds respectively between the protein residues and the respective C3-acetyl carbonyl groups of the B850 Bchls . The removal of a hydrogen bond to the acetyl carbonyl group was signalled by a shift of the Raman peak from the wild-type LH2 complex at 1635 cm−1 to a position expected for a non-hydrogen-bonded free acetyl carbonyl, i.e. 1659 cm−1. It was demonstrated further, using wild-type and mutant antenna complexes, that there is a consistent linear relationship between the downshift in the Bchl a C3 acetyl-stretching mode and the red shift in the Qy absorption maximum . This is because, in these complexes, the α/β-polypeptide dimers were considered to contain only one type of α-polypeptide and one type of β-polypeptide (the existence of the second β-polypeptide in the LH2 complex from Rba. sphearoides  was not known at this time, but its homology with the puc1B-encoded-polypeptide is such that it does contribute to any spectral contamination).
In a homogenous LH complex sample, all of the Bchl a molecules are expected to contribute more or less equally to the Raman signal when probed in the Soret transition, but under pre-resonant conditions with the Qy absorption band, the relative signal from the red-most chromophore will be enhanced [40,51]. Thus, when compared with Soret excitation, the positions of the vibrational bands will be the same, albeit with different intensities. However, if a mixed population of LH complexes is present, the intensities and positions of the carbonyl vibrational modes observed will not necessarily be the same owing to the presence of subpopulations having different Bchl a Qy-transitions. This is exactly what happens in Rps. palustris when LL LH2 complex is compared with its HL variant . This previous work proposed that LL LH2 complex contains rings containing multiple α-polypeptides since no native-like B800–820 LH2 complex (also called LH3 complex) was observed . However, this previous study did not analyse the polypeptide composition, but rather relied solely on the comparison of the electronic absorption properties together with other studies where the polypeptide composition had been established . In the present study, where the absorption (Figure 3) and Raman spectra (Figure 10) mirror the previous Raman study, and where the polypeptide composition has been established, it can now be unambiguously concluded that rings with mixtures of α-polypeptides (PucAa and PucAd; see Figure 6) are indeed present in LL LH2 complexes from Rps. palustris. The PucAd protein is not able to from hydrogen bonds with Bchl a molecules to produce a B800–850 LH2 complex. Furthermore, going from HL to LL, via LL2 and LL1, there is a progressive reduction in the apparent strength of the Bchl a hydrogen-bond network (Figure 10), which, when combined with the evolution of the blue shift in the B850 Bchl a Qy-transition (Figure 3), suggests that the population with mixed ‘rings’ containing PucAd increases as the growth conditions are pushed further towards lower light intensities.
The low-resolution X-ray diffraction obtained for the HL and LL LH2 complex crystals from Rps. palustris 2.1.6 is a result of the limited crystalline order and that feature of these crystals could be, in part, due to the LH2 complex ring heterogeneity. Such heterogeneity could also contribute to the rotational disorder about the LH2 complex ring axis apparent in the self-rotation functions. The MR solutions obtained with the X-ray diffraction data suggest that the most probable structure for both of the HL and LL LH2 complexes from Rps. palustris 2.1.6 is a nonamer and that the overall shape of the complex is more similar to the Rps. acidophila LH2 complex than to the Phs. molischianum LH2 complex. Molecular sieve chromatography also suggests that both these HL and LL LH2 complexes from Rps. palustris 2.1.6 are nonameric. Higher-resolution diffracting crystals, if they can be produced, are now required in order to try to obtain a better picture of the exact details of the LH2 complex ring heterogeneity. However, since there are more than two types of LH2 complexes present and since it is not clear that any of these complexes can be individually purified, this may well prove to be a very difficult challenge. A possible way out of this impasse may be through the recent advances in MS that now allow large detergent-solubilized multiprotein complexes to ‘fly’ as intact complexes. Experiments are now underway to try to develop the use of this technique with the various LH2 complexes from Rps. palustris (and indeed on many other interesting photosynthetic LH complexes in our laboratory), and it is hoped that these results will prove rewarding in the future.
Tatas Brotosudarmo prepared samples, grew crystals and prepared the paper. Aaron Collins camied out absorption, CD and fluorescence spectroscopy. Andrew Gall carried out resonance Raman spectroscopy. Aleksander Roszak carried out crystallography. Alastair Gardiner prepared samples and grew crystals. Robert Blankenship and Richard Cogdell were principal investigators.
This work was supported as part of the Photosynthetic Antenna Research Center (PARC), an Energy Frontier Research Center funded by the U.S. Department of Energy, Office of Science, Office of Basic Energy Sciences [grant number DE-SC 0001035]. A.W.R., A.T.G. and R.J.C. acknowledge financial support from the Engineering and Physical Sciences Research Council. A.G. acknowledges financial support from the Agence Nationale de la Recherche (ANR), France [grant number ANR-07-CEXC-009].
We gratefully acknowledge Dr James Sturgis for providing us with the LH2 complexes from Phs. molischianum. We acknowledge the European Synchrotron Radiation Facility for provision of synchrotron radiation facilities, and we thank Dr S. McSweeney, Dr G.A. Leonard, Dr L. Terradot, Dr H. Belrhali and Dr S. Brockhauser, for their assistance in using beamlines ID29 (twice), ID14-4, BM14 and ID23-2. We also acknowledge the support of the Diamond Light Source for provision of beamtime at station I04.
Abbreviations: Bchl, bacteriochlorophyll; Car, carotenoid; DDM, dodecyl maltoside; ESI, electrospray ionization; FLL1, far low-light; FLL2, extreme low-light; HL, high-light; i.d., internal diameter; LDAO, N,N-dimethyldodecylamine-N-oxide; LH, light-harvesting; LH1, core LH; LH2, peripheral LH; LL, low-light; LL1, low-light intermediate 1; LL2, low-light intermediate 2; MR, molecular replacement; MS/MS, tandem MS; MWCO, molecular-mass cut-off; NIR, near-IR; PEG, poly(ethylene glycol); RC, reaction centre; rms, root mean square; SRF, self-rotation function
- © The Authors Journal compilation © 2011 Biochemical Society