Article No. mb982050 J. Mol. Biol. (1998) 282, 833±845

Projection Structures of Three PhotosyntheticComplexes from Rhodobacter sphaeroides: LH2 at6 AÊ , LH1 and RC-LH1 at 25 AÊ

Thomas Walz*, Stuart J. Jamieson, Claire M. Bowers, Per A. Bulloughand C. Neil Hunter

Krebs Institute for BiomolecularResearch, Department ofMolecular Biology andBiotechnology, University ofShef®eld, Firth CourtWestern Bank, Shef®eldS10 2TN, UK

E-mail address of the correspondT.Walz@Shef®eld.ac.uk

Abbreviations used: LH, light-hacentre; 2-D, two-dimensional; 3-D,OBG, octyl-b,D-glucoside; DHPC, dphosphatidylcholine; bchl, bacteriodimyristoyl phosphatidylcholine; Dphosphatidylcholine; CTF, contrast

0022±2836/98/390833±13 $30.00/0

Three photosynthetic complexes, light-harvesting complex 2 (LH2), light-harvesting complex 1 (LH1), and the reaction centre-light-harvestingcomplex 1 photounit (RC-LH1), were puri®ed from a single species of apurple bacterium, Rhodobacter sphaeroides, and reconstituted into two-dimensional (2-D) crystals. Vesicular 2-D crystals of LH1 and RC-LH1were imaged in negative stain and projection maps at 25 AÊ resolutionwere produced. The rings formed by LH1 have approximately the samemean diameter as the LH1 rings from Rhodospirillum rubrum (�90 AÊ ) andtherefore are likely to be composed of 15 to 17 ab subunits. In the projec-tion map calculated from the RC-LH1 2-D crystals, the reaction centre isrepresented by an additional density in the centre of the ring formed bythe LH1 subunits. The marked improvement of shape and ®ne structureafter a rotational pre-alignment of the RC-LH1 unit cells before averagingstrongly suggests that the RC is not in a unique orientation within theLH1 rings. Tubular crystals of LH2 showed a high degree of order andallowed calculation of a projection map at 6 AÊ resolution from glucose-embedded specimens. The projection structure shows a ring of nineab subunits. Variation of the a-helical projection densities suggests thatthe 9-fold symmetry axis is tilted with respect to the membrane normal.

# 1998 Academic Press

Keywords: photosynthesis; electron crystallography; 2-D crystals; light-harvesting complexes; membrane protein

*Corresponding author

Introduction

Photosynthesis is a biological reaction of funda-mental importance, because it represents a highlyef®cient and productive means of converting solarenergy into chemical energy. In many cases, theinitial stages of conversion of light energy into use-ful chemical energy involve the participation ofseveral membrane-bound protein complexes. In thelast 15 years, structural details of many of theseindividual protein components have emerged, toadd to the large array of chemical, spectroscopic,cell biological and genetic data that have been

ing author:

rvesting; RC, reactionthree-dimensional;iheptanoylchlorophyll; DMPC,OPC, dioleoyltransfer function.

obtained on photosynthetic systems in plants andbacteria. One of the next challenges is to describethe assembly and organisation of an intact photo-system. The purple photosynthetic bacteria providean ideal experimental system for such work,because of the relative simplicity of their photosys-tem. Most of these bacteria synthesise two types oflight-harvesting (LH) complexes; LH2, present invariable amounts, ef®ciently collects light energyand transfers it to the LH1 complex, which is inconstant proportion to and in contact with thephotosynthetic reaction centre (RC). The RC is thesite of the initial charge separation and subsequentelectron transport across the membrane.

A major breakthrough in the understanding ofmembrane proteins came with the elucidation ofstructures of the bacterial RCs from Rhodopseudo-monas viridis (Deisenhofer et al., 1985) and Rhodo-bacter sphaeroides (Chang et al., 1986; Allen et al.,1987) by X-ray crystallography. In recent years,electron microscopy has proved a powerful

# 1998 Academic Press

834 Projection Structures of LH2, LH1 and RC-LH1

alternative to X-ray crystallography to investigatethe structure and organization of membraneproteins even to near-atomic resolution if suitabletwo-dimensional (2-D) crystals are available(Henderson et al., 1990; KuÈ hlbrandt et al., 1994;Grigorieff et al., 1996; Kimura et al., 1997). In gener-al, it is much easier and faster to grow 2-D crystalsfrom membrane proteins than to grow largewell-ordered 3-D crystals. Moreover, electronmicroscopy can be used to extract important infor-mation from 2-D crystals that do not show a highdegree of crystallinity. Hence, electron microscopyis very suited to the rapid determination of oligo-meric states of LH complexes, and for visualizingthe RC-LH1 interactions. Indeed, the ®rst directview of the structure of a bacterial antenna com-plex was obtained by electron crystallography of2-D crystals of the LH1 complex from Rhodospiril-lum rubrum yielding a projection map at 8.5 AÊ res-olution (Karrasch et al., 1995). This map revealedthat the complex consists of a ring of 16 ab hetero-dimers with a diameter large enough to accommo-date the reaction centre within. Subsequently, thestructures of two LH2 complexes were solved toatomic resolution by X-ray crystallography toreveal a nonameric arrangement of ab subunits inRps. acidophila (McDermott et al., 1995) and an octa-meric arrangement in Rhodospirillum molischianum(Koepke et al., 1996). In addition, an electroncrystallographic projection map of LH2 fromRhodovulum (Rhv.) sul®dophilum was calculatedusing 2-D crystals, revealing a nonameric arrange-ment (Montoya et al., 1995; Savage et al., 1996).

Despite the wealth of information on individualproteins, the organisation of the different com-plexes within the bacterial photosynthetic mem-brane is poorly understood, although severalmodels have been proposed, based upon the spec-troscopic properties of the complexes (Monger &Parson, 1977; Vos et al., 1986; Hunter et al., 1989).Electron microscopic studies on in vivo crystalline,intact photosynthetic membranes of Rps. viridis(Miller, 1982; Stark et al., 1984) and Ectothiorhodos-pira halochloris (Engelhardt et al., 1986) and onarti®cial 2-D crystals of photosynthetic core(RC-LH1) complexes from R. rubrum (Walz &Ghosh, 1997; Stahlberg et al., 1998) and Rps. viridis(Ikeda-Yamasaki et al., 1998) showed that the bac-terial RC is surrounded by a ring of LH1 subunits.However, the closed LH1 ring structure raises thequestion of how the quinone molecules can partici-pate in electron transfer between the reactioncentre within the ring and the cytochrome bc1 com-plex lying outside the ring. In Rb. sphaeroides, asmall polypeptide, PufX (Lilburn et al., 1992; Barzet al., 1995a,b; Cogdell et al., 1996), has been impli-cated as being essential for the ef®cient transfer ofquinone between the reaction centres and the cyto-chrome bc1 complexes, since in its absence nophotosynthetic growth is observed (Farchaus et al.,1990, 1992; McGlynn et al., 1994, 1996). The -pep-tide identi®ed in R. rubrum has been suggested toplay a similar role (Ghosh et al., 1994).

Although a large amount of structural infor-mation on bacterial photosynthetic proteins hasemerged over the last few years, it is still dif®cultto directly correlate all this information to producea coherent model of a bacterial photosystem,because individual structures have been obtainedfrom proteins puri®ed from different organisms.In order to construct an integrated view ofbacterial photosynthesis, it would be most helpfulto know the structures of all the complexes fromjust a single species; ideally, a species easily acces-sible to genetic manipulation. Here, we present the®rst results of our attempts to use Rb. sphaeroides asa model system for an integrated structural studyof bacterial photosynthesis. We present a projectionmap of LH2 at 6 AÊ resolution and lower-resolutionprojection maps of LH1 as well as of the RC-LH1photounit determined by electron microscopy of2-D crystals.

Results

Purification and spectroscopy

Following solubilization of the photosyntheticmembranes with octyl-b,D-glucoside (OBG) forLH2 preparations (or diheptanoyl phosphatidyl-choline (DHPC) for LH1 and RC-LH1 prep-arations), the complexes were puri®ed in a two-step procedure as described in Materials andMethods. Figure 1a shows a SchaÈgger gel of mem-branes, puri®ed complexes and two-dimensional(2-D) crystals for LH2. One noteworthy aspect ofthe detergent DHPC is its property of stabilisingthe RC-LH1 interactions, which are normally dis-rupted by OBG. This was ®rst noted for R. rubrumby Kessi et al. (1995), and it appears that DHPCis similarly suitable for RC-LH1 cores fromRb. sphaeroides.

The native state and quality of the puri®ed com-plexes as well as of the reconstituted 2-D crystalswere assessed by absorbance spectroscopy(Figure 1b). In comparison with native membranes(continuous line in Figure 1b), the puri®ed LH2complexes in 1% OBG (broken line) display aslightly broadened absorption peak at 790 nmindicating some increased heterogeneity in thebacteriochlorophyll (bchl) environment. Afterreconstitution into 2-D crystals this heterogeneity isdiminished, as evidenced by the sharpening of thepeak in this absorption region (bold line). Interest-ingly, the 850 nm absorption peak from nativemembranes shifted to 854 nm in the 2-D crystals.In the case of LH1 and RC-LH1, the spectra of themembranes, puri®ed complexes and 2-D crystalsshowed only negligible differences.

LH2 2-D crystals in negative stain

The 2-D crystals of LH2 complexes were foundover a wide range of conditions. Addition of mag-nesium ions to the dialysis buffer did not improvethe quality of the 2-D crystals and was omitted in

Figure 1. SchaÈgger gel and absorbance spectroscopy of LH2 complexes. a, In lane a of the SchaÈgger gel, a mem-brane preparation from an LH2-only Rb. sphaeroides mutant strain was run. Lane b shows the LH2 complexes afterpuri®cation in octyl-b,D-glucoside (OBG) and lane c shows the LH2 2-D crystals. The small shift of the b-polypeptidein lane b compared to lanes a and c is most likely a mobility effect due to the OBG present in the solubilized LH2complexes. The markers are: A, ovalbumin (41.8 kDa); B, carbonic anhydrase (28.8 kDa); C, b lactoglobulin(19.3 kDa); D, lysozyme (15.4 kDa); E, bovine trypsin inhibitor (6.4 kDa); and F, insulin, a and b chain (3.0 kDa).b, Absorption spectra of native LH2-only photosynthetic membranes (continuous line), puri®ed LH2 complexes in 1%OBG (broken line), and reconstituted LH2 2-D crystals (bold line).

Projection Structures of LH2, LH1 and RC-LH1 835

further reconstitution experiments. The concen-tration of sodium chloride (varied between 50 and200 mM) did not show signi®cant in¯uence on theoutcome of crystallization trials. All three lipidsused for reconstitution experiments (DMPC, DOPCand plant PC) yielded ordered arrays of LH2; how-ever, the quality was variable and the best crystalswere produced using DOPC.

Successful dialysis experiments yielded largenumbers of tubular crystals (Figure 2a). Collapsedtubes varied considerably in length (from 2 to over5 mm) but displayed a relatively constant width(650(�50) nm, n � 25). When inspected in a laserdiffractometer, two sets of sharp and strong dif-fraction spots lying on square lattices and originat-ing from the two layers of the collapsed tube couldbe detected in almost every crystal imaged.Unusually for negatively stained crystals, opticaldiffraction spots could often be seen beyond 20 AÊ

resolution (Figure 2c). Only images where the twolattices could clearly be indexed and separatedfrom each other were used in order to calculateprojection maps of both layers.

The unit cell of the tubular crystals(a � b � 149 AÊ ) houses four LH2 rings. Asobserved previously with collapsed sacculi ofOmpF-overexpressing Escherichia coli, the projec-tion maps of the two layers appear slightly differ-ent due to differences in the effect of staining andadsorption on the two layers (Hoenger et al., 1993).While in both layers the LH2 rings can clearly bedepicted, in the projection map from the lessevenly stained layer two of the four rings display a

lower contrast and less ®ne structure (Figure 3a),suggesting that there are two orientations of thecomplexes within the membrane. In the projectionmap of the more evenly stained layer (Figure 3b)all rings have approximately the same contrast andalready indicate a 9-fold symmetry. Using the pro-gram ALLSPACE (Valpuesta et al., 1994) on imagesof glucose-embedded 2-D crystals (see later), acrystallographic p22121 symmetry was revealed(see Table 1) and was applied to the projectionmaps (Figure 3c and d). In Figure 3e, a contourplot of the projection map in Figure 3d is presenteddenoting the 2-fold screw axes and 2-fold rotationaxes that relate the four rings in the unit cell toeach other and de®ne the p22121 symmetry of theLH2 2-D crystal.

LH2 2-D crystals in glucose

Six micrographs (12 crystalline layers) of glu-cose-embedded tubular LH2 2-D crystals wereselected by optical diffraction for further proces-sing. In the selected images, diffraction spotsof both lattices were visible by eye beyond 8 AÊ res-olution. A contrast transfer function (CTF) plot of agood image after unbending and CTF correction isshown in Figure 4a. The Fourier data are almostcomplete up to 6 AÊ and a few strong diffractionspots are present to a much higher resolution.Occasionally, diffraction spots of IQ2 (IQ asde®ned by Henderson et al., 1986) could even beseen at 3.5 AÊ resolution. Figure 4b shows a p1 pro-jection map calculated to 6 AÊ resolution from the

Figure 2. Negative stain electron microscopy of tubular LH2 2-D crystals. a, The low-magni®cation image of a suc-cessful dialysis experiment shows a large number of tubular 2-D crystals with only a small fraction of vesicular 2-Dcrystals. b, Low-dose images at a nominal magni®cation of 50,000� were recorded close to focus and, accordingly,the square lattice formed by the LH2 complexes is hardly visible. c, The calculated diffraction pattern from the areashown in b reveals two sets of diffraction spots originating from the upper and the lower layer of the collapsed tubu-lar 2-D crystal. The scale bar represents 2 mm in a, 100 nm in b, and (5 nm)ÿ1 in c.

836 Projection Structures of LH2, LH1 and RC-LH1

image yielding the CTF plot shown in Figure 4a.The seven best data sets were merged, phases andamplitudes averaged according to p22121 sym-metry and a projection map calculated (Figure 4c).The phase residuals in resolution shells are sum-marised in Table 2. Sharpening of Fourier ampli-tudes by the application of an arti®cial negativetemperature factor (as applied in the case of theprojection maps of LH1 from R. rubrum (Karraschet al., 1995) and of LH2 from Rhv. sul®dophilum(Savage et al., 1996) was not needed to reveal thedensity at the level of secondary structure.

The unit cell outlined in black in Figure 4cshows again four individual LH2 complexes,which are now resolved into two concentric ringsof nine densities each with average diameters of35 AÊ and 62 AÊ , respectively. By comparison withthe atomic model of the LH2 from Rps. acidophila(McDermott et al., 1995), the inner ring ofroughly circular densities is interpreted as theprojection densities of the a-helices from the a-subunits running almost perpendicular to themembrane plane. The outer ring of moreelongated densities represents the projections of

the a-helices from the b-subunits, which are moretilted with respect to the membrane normal. Lessdensity is seen between the two concentric ringswhen compared to the published projection mapsof R. rubrum LH1 (Karrasch et al., 1995) andRhv. sul®dophilum LH2 complexes (Savage et al.,1996). Moreover, the expected non-crystallo-graphic 9-fold symmetry is not evident in thisprojection map. The inner helical densities showa marked variation of contrast around the ring(Figure 4c), which is evident also in maps fromindividual crystals (see Figure 4b). Although notas marked, a similar variation of contrast can beseen in the outer ring of densities.

Individual LH2 rings appear to be packed verytightly in the membrane and form two differenttypes of crystal contacts via their b-subunits.Eight of the nine b-subunits of each LH2 ring arein close contact with two b-subunits from each ofthe four adjacent LH2 rings. The four a-helicesinvolved in each crystal contact are too close toallow for intercalating lipid molecules. Hence,crystals are held together by direct helix-helixinteractions.

Figure 3. Projection maps of LH2 in negative stain. a, Two rings in the unit cell (a � b � 14.9 nm) calculated fromthe less evenly stained layer show lower contrast than the other two rings, indicating that the LH2 rings are incorpor-ated into the membrane in opposite orientations. b, After applying a p22121 symmetry, all rings display the same con-trast and reveal a pseudo-9-fold symmetry. c, In the unit cell calculated from the more evenly stained layer, all LH2rings show approximately the same contrast and the non-crystallographic pseudo-9-fold symmetry is already distinctin the unsymmetrised projection map. d, However, the non-crystallographic pseudo-9-fold symmetry becomes moreevident after applying the crystallographic p22121 symmetry. e, In the contour plot of the projection map shown in d,the unit cell is outlined in black. The crystallographic symmetry elements (2-fold axes and 2-fold screw axes) thatrelate the four LH2 rings in a unit cell with each other are also drawn.

Projection Structures of LH2, LH1 and RC-LH1 837

2-D Crystals of LH1 and RC-LH1 complexes

Crystals of LH1 and RC-LH1 complexesformed either as vesicular or sheet-like mem-

Table 1. Internal phase residuals of all possibshown in Figure 4

Two-sidedplane group

Phase residualversus other spots

(90� random)

p1 30.7a

p2 39.4b

p12 a 84.3p12 b 74.5p121 a 22.6c

p121 b 23.4c

c12 a 84.3c12 b 74.5p222 81.7p2221 a 67.7p2221 b 61.0p22121 30.1b

c222 81.7p4 60.8p422 83.5p4212 54.8

Internal phase residuals were determined from sa Note that in space group p1 no phase compar

theoretical phase residuals based on the signal-to-nb Within 20% of target residual.c Within 30% of target residual.

branes at lower lipid to protein ratios than thosefor the tubular LH2 crystals (Figure 6a and c).To date, only relatively small and poorly ordered2-D arrays of these complexes have been

le two-sided plane groups from the image

Number ofcomparisons

Target residualbased on statistics

taking Friedel weightinto account

706353 45.6220 31.6213 31.2220 31.6213 31.2220 31.6213 31.2786 37.4786 37.4786 37.4786 37.4786 37.4665 38.6

1406 34.41406 34.4

pots of IQ1 to IQ5 to 6 AÊ resolution.ison is possible, so the numbers given here areoise ratio of the observed diffraction spots.

838 Projection Structures of LH2, LH1 and RC-LH1

obtained. Again, DOPC yielded the best crystals.The crystallinity of the best 2-D arrays was suf®-cient to calculate projection maps of the LH1 andRC-LH1 complexes in negative stain to a resol-ution of 25 AÊ . Unlike the 2-D crystals formed byLH1 and RC-LH1 complexes from R. rubrum,which display a p22121 symmetry (Karrasch et al.,1995; Walz & Ghosh, 1997), both complexes fromRb. sphaeroides form crystals with a trigonal-likelattice (a � b � 120 AÊ , g � 120�) and no diffractionspots indicating an orthorhombic lattice could beobserved in the power spectrum. Similar trigonal2-D crystals were recently reported for RC-LH1complexes from Rps. viridis (Ikeda-Yamasaki et al.,1998). The LH1 rings (Figure 6b) have an averagediameter of about 90 AÊ , similar to that of LH1rings from R. rubrum (Karrasch et al., 1995). Thesame average diameter of LH1 rings wasmeasured in the RC-LH1 projection map(Figure 6d). The stain exclusion from the RCappears rather featureless in our projection map,similar to the situation found in 2-D crystals ofRC-LH1 complexes from R. rubrum (Walz &Ghosh, 1997). However, the projection structureof the RC could be substantially enhanced bytreating the individual unit cells of the RC-LH12-D crystal as single particles and introducing arotational alignment step before averaging theunit cells (see Materials and Methods). Figure 7ashows the Fourier peak ®ltered image of an areaof the RC-LH1 2-D crystal shown in Figure 6c.The crosses indicate the 850 unit cells that wereused to calculate the correlation averages. Theintroduction of a rotational alignment step beforeaveraging the unit cells led to a more distinctlyfeatured envelope for the RC (Figure 7c) whencompared to the featureless correlation averagecalculated without rotational alignment of theunit cells (Figure 7c). At this resolution there wasno evidence for any preferred orientations of theRCs.

Figure 4. Electron cryo-microscopy of glucose-embedded LH2 2-D crystals. a, The contrast transferfunction (CTF) plot of a Fourier transform of a typicalimage of an LH2 crystal embedded in glucose showsalmost complete data to a resolution of 6 AÊ . However,strong diffraction spots can be seen up to the Nyquistfrequency of 3.6 AÊ . The circles represent the zero tran-sitions of the CTF and the boxed numbers depict the IQvalues of the individual re¯ections as de®ned byHenderson et al. (1986). b, The displayed projection mapin p1 was calculated from the image that yielded theCTF plot shown in a. Even without applying any sym-metry, the two concentric rings of membrane-spanninga-helices are clearly seen, while the non-crystallographic9-fold symmetry of the individual rings is clearly miss-ing. c, The ®nal projection structure was calculated bymerging seven unbent and CTF-corrected lattices andapplying a p22121 symmetry. The map was calculatedto 6 AÊ resolution from a 100 % complete data set.

Table 2. Phase residuals and data completeness in resolution shells

Resolutionshell (AÊ )

Number ofphases

Mean value of�ac (deg.)

Standard error(deg.)

Mean figureof merit

Completeness of dataup to IQ5 (IQ3)a (%)

1±6.0 507 26.9 1.2 0.78 92.3 (52.5)1±20.0 51 21.3 3.4 0.83 94.1 (62.7)20.0±12.0 83 18.1 2.5 0.86 94.0 (71.1)12.0±8.0 157 23.8 2.0 0.82 95.5 (68.2)8.0±6.0 216 33.9 1.8 0.70 78.0 (31.5)

�ac, Difference between the symmetry-imposed phase of 0� and 180� and the observed combined phase (Bullough & Tulloch,1990).

a Completeness of data up to IQ7 spots was 100% to a resolution of 6 AÊ .

Projection Structures of LH2, LH1 and RC-LH1 839

Discussion

We report the crystallisation of all the majorprotein complexes of the photosynthetic machineryof a single species of purple bacterium. The resol-ution is not yet adequate to determine the numberof subunits in the LH1 complex but the averagediameter of the LH1 ring of approximately 90 AÊ issimilar to that of LH1 from R. rubrum (Karraschet al., 1995) and therefore it is likely that the ringconsists of 15 to 17 ab subunits. In addition, thesestudies provide further con®rmation that as inR. rubrum (Walz & Ghosh, 1997; Stahlberg et al.,1998), the RC of Rb. sphaeroides is enclosed withinthe LH1 ring, at least in the absence of the PufXpolypeptide. For further detailed analysis, weawait improvements in crystal quality. Contrary tothe results for LH1 and RC-LH1, the LH2 crystalshave provided us with the opportunity of collect-ing high-resolution data.

Projection structures from negatively stainedand glucose-embedded LH2 crystals

All well-ordered 2-D crystals of LH2 display atubular morphology and a p22121 symmetry inwhich adjacent LH2 rings are related to eachother by screw axes within the membrane plane.Projection maps of LH2 from Rb. sphaeroides innegative stain as well as embedded in glucoseclearly show the nonameric organisation of thering. This is the same as observed for LH2 fromRps. acidophila (McDermott et al., 1995) andRhv. sul®dophilum (Savage et al., 1996), but differ-ent from the octameric ring formed by LH2 fromR. molischianum (Koepke et al., 1996). While adja-cent LH2 rings are clearly separated in the pro-jection map calculated from the negativelystained specimens (see Figure 3), in the projectionmap from the glucose-embedded specimens theouter domains of adjacent LH2 rings come intovery close contact (Figure 4c). The negative stainonly contrasts extramembranous protein domainsand the highest contrast from sugar-embeddedpreparations is likely to derive from the trans-membrane domains, so it appears that crystalcontacts are formed mainly within the lipidbilayer. Compared to the 2-D crystals obtainedfrom Rhv. sul®dophilum LH2, which also show ap22121 symmetry (Savage et al., 1996), the rings in

the 2-D crystals from Rb. sphaeroides LH2 arepacked in a different and more compact way.Only a few b-subunits of a given ring appear tomake direct contact to b-subunits of adjacentrings in the Rhv. sul®dophilum 2-D crystals, whilein the 2-D crystal of Rb. sphaeroides LH2, eightout of nine b-subunits are involved in direct crys-tal contacts. This might explain the better order-ing found in our 2-D crystals.

The missing 9-foldnon-crystallographic symmetry

The atomic structures of the LH2 complexesfrom Rps. acidophila (McDermott et al., 1995) andR. molischianum (Koepke et al., 1996) show near-perfect 9-fold and 8-fold symmetry of the rings,respectively. These structures were determinedfrom 3-D crystals of detergent-solubilized com-plexes that do not give any information on howthe rings are incorporated into a lipid bilayer.Our projection map indicates a marked deviationfrom 9-fold symmetry. This can be seen to a lesserextent in the projection map of LH2 fromRhv. sul®dophilum (W. KuÈ hlbrandt, personal com-munication and see Figure 3b of Savage et al.,1996).

To demonstrate that the deviation from the9-fold projection symmetry is signi®cant, we per-formed several tests. All individual projectionmaps that were merged to give the ®nal aver-aged p22121 map (Figure 5a) were also calculatedwithout any symmetry averaging, aligned to acommon phase origin and the amplitudes andphases merged into one data set. The individualp1 maps as well as the merged and averaged p1projection (Figure 5c), all showed the same vari-ations in density at the same positions, albeit thatthe maps were more noisy. In addition, two inde-pendent p22121 maps were calculated, one from aset of three images and another from an indepen-dent set of four images. In both cases there wasa consistent deviation from 9-fold projection sym-metry. Finally, a map of the estimated noise wascalculated for the averaged p22121 map asdescribed (Bullough & Henderson, 1990). Themap reveals a noise level of �1 contour lines(Figure 5b), which is signi®cantly less thanthe observed maximum deviation in densities

Figure 5. Quality assessment of the LH2 projection map. a, The p22121 projection map was contoured in such away that each contour line represents two-®fths of the root-mean-square of the mean density. In this representation,the maximum deviation in densities between inner helices in one ring are nine contour levels. b, The noise in thep22121 projection map was determined as described by Bullough & Henderson (1990) and contoured in the same wayas in the LH2 projection map in a. This map demonstrates that the noise level in the projection map (�1 contourlevel) cannot be responsible for the loss of the non-crystallographic 9-fold symmetry. c, The seven lattices used to cal-culate the p22121 projection map were also merged without applying a symmetry, illustrating that the deviation fromthe 9-fold symmetry is not random but identical in all images. d, Simulated p22121 lattice of LH2 from Rps. acidophila(McDermott et al., 1995) having a tilt of 5� relative to the membrane plane. The resemblance of this simulated map tothe actual projection map of LH2 from Rb. sphaeroides is striking and con®rms a slight tilt of the LH2 rings to beresponsible for the deviation of the individual rings from the perfect non-crystallographic 9-fold symmetry.

840 Projection Structures of LH2, LH1 and RC-LH1

between inner helices in one ring (difference �nine contour lines, see Figure 5a).

There could be several explanations for thedeviation of the rings from the 9-fold symmetry.For example, the subunits might be more disor-dered in one part of the ring. This disorder couldbe spatial or temporal. However, it is unlikely that

such a disorder is always con®ned to the samepart of the LH2 ring, especially since no correlationbetween the strong densities and the crystal con-tacts could be found. It might also be that some ofthe subunits are lost, leading to a lower density inthis region, but again, it is dif®cult to explain whysubunits should be lost only in a speci®c domain

Figure 6. Negative stain electron microscopy of vesicular LH1 and RC-LH1 2-D crystals. a, Typical LH1 2-D crystalobtained by reconstitution of DHPC-solubilised LH1 with OBG-solubilised DOPC. b, The projection map was calcu-lated to a resolution of 25 AÊ from a selected area of the micrograph shown in a and no symmetry was applied.Although the resolution is too poor to visualise the individual ab subunits, the average diameter of 90 AÊ , which issimilar to that determined for R. rubrum LH1, indicates that the ring might also be composed of approximately 16 absubunits. c, Typical image of an RC-LH1 2-D crystal negatively stained with uranyl formate. d, In the unsymmetrisedprojection map of RC-LH1 at 25 AÊ resolution determined from a selected area in the image shown in c, the reactioncentre is represented by a rather featureless stain exclusion in the centre of the LH1 ring. The scale bars in a and crepresent 200 nm and the side-length of b and d is 24 nm.

Projection Structures of LH2, LH1 and RC-LH1 841

of the ring. The most likely explanation is a slighttilt of the individual LH2 rings relative to themembrane plane. To test this hypothesis, we calcu-lated a projection map of LH2 from Rps. acidophilausing the atomic coordinates determined byMcDermott et al. (1995) and applying small tiltangles and a suitable temperature factor. The pro-jection of a ring tilted by 5� was then aligned ontoa p22121 lattice (Figure 5d). The similarity of theresulting projection map to the projection structuredetermined from our 2-D crystals of Rb. sphaeroidesLH2 is so striking that we propose that thedeviation from the perfect 9-fold symmetry arisesfrom a tilt of the rings of approximately 5� relativeto the membrane normal.

The photosynthetic unit of Rb. sphaeroides;relationship between structural, spectroscopicand mutagenesis data

Although a great deal is known about thelight-harvesting apparatus of Rb. sphaeroides interms of spectroscopy and mutagenesis, it hasnot been possible to place this information withina structural context, except by extrapolation fromdata obtained from other bacteria such asRps. acidophila and R. rubrum. The present workis an attempt to redress this imbalance, and laysthe foundations for further work on geneticallymodi®ed systems, given that Rb. sphaeroides iswell established as a system for genetic manipu-

Figure 7. Single-particle treatment of an RC-LH1 2-D crystal. a, Fourier peak ®ltered area from the RC-LH1 2-Dcrystal shown in Figure 6c. The white crosses mark the unit cells that were used to calculate the correlation averages.b, Averaging without prior rotational alignment of the unit cells leads to a projection structure of the RC-LH1 com-plex, where the RC is represented by an almost featureless density. c, After rotational alignment of the individualunit cells, the RC in the resulting average shows a distinct shape, demonstrating that the RC is not ®xed in a uniqueorientation within the crystal lattice.

842 Projection Structures of LH2, LH1 and RC-LH1

lation of LH2 (Fowler et al., 1997), LH1 (Olsenet al., 1997; Sturgis et al., 1997) and the RC(Woodbury & Allen, 1995).

The quality of our LH1 and RC-LH1 crystals issuf®cient to show the circular arrangement of theLH1 observed by Miller (1982), Stark et al. (1984),Engelhardt et al. (1986) and Karrasch et al. (1995).In these ordered arrays, the RC-LH1 complexes aresimilar to LH1, but exhibit a region of stain exclu-sion in the centre of the LH1 rings that we attributeto the RC. The RC is apparently not in a singleorientation with respect to the crystal lattice, whichis demonstrated by the dramatic improvement inenvelope de®nition and ®ne structure after intro-ducing a rotational alignment step (on the RC)prior to averaging of the individual unit cells (seeFigure 7). Therefore, the RC can clearly not beinvolved in speci®c crystal contacts of the RC-LH12-D arrays, which would necessitate a single orien-tation of the RCs in the lattice.

It is important to note that the LH1 and RC-LH1 crystals were prepared from LH1-only andRC-LH1-only mutants, respectively, and that ineach case the PufX polypeptide was absent dueto deletion of the PufX gene from the expressionplasmid (McGlynn et al., 1994, 1996). It is known

that the absence of PufX from the RC-LH1 coresresults in a loss of photosynthetic growth, andthat it has been suggested that PufX preventsLH1 from completely encircling the RC, whichwould otherwise prevent reduced quinones frommoving from the RC QB site to the cytochromebc1 complex (Lilburn et al., 1992; Barz et al.,1995a,b; Cogdell et al., 1996). It might be signi®-cant that in the case of the RC-LH1 cores witheither normal or reduced levels of LH1, theabsence of PufX generally increases the numberof LH1 bchls per RC by 1 to 2. One of the nextsteps is to obtain good-quality 2-D crystals fromRC-LH1 cores containing PufX. A substantialimprovement is required in the quality of the 2-Dcrystals prepared from LH1 or RC-LH1 com-plexes, suf®cient to provide the level of structuralinformation that can be gained from the LH2crystals reported here. Such an improvementwould allow the information obtained frommutagenesis, which has provided detailed infor-mation on the H-bonding arrangement of thebchls in LH1 (Olsen et al., 1994, 1997; Sturgiset al., 1997), to converge with crystallographicinformation to provide a useful structural modelfor LH1.

Projection Structures of LH2, LH1 and RC-LH1 843

Materials and Methods

Chemicals

Octyl-b,D-glucoside (OBG), dimyristoyl phosphatidyl-choline (DMPC) and dioleoyl phosphatidylcholine(DOPC) were from SIGMA (SIGMA-Aldrich CompanyLtd., Dorset, England; diheptanoyl phosphatidylcholine(DHPC) and plant phosphatidylcholine extract werefrom Avanti Polar Lipids (Avanti Polar Lipids, Inc.,Alabaster, AL, USA). All other chemicals used wereeither from SIGMA (SIGMA-Aldrich Company Ltd.,Dorset, England) or from Merck (Merck Ltd., Dorset,England).

Purification of LH2, LH1 and RC-LH1 complexes

Membranes were isolated from LH2-only, LH1-onlyand RC-LH1-only strains of R. sphaeroides (Jones et al.,1992) as described (Olsen et al., 1994). Neither the LH1-only nor the RC-LH1-only strain contained PufX. Thisarose as a result of deleting the cognate gene(McGlynn et al., 1994, 1996). They were solubilised ineither 100 mM OBG (LH2) or 40 mM DHPC (LH1 andRC-LH1; Kessi et al., 1995). For LH2, the buffer was10 mM Hepes (pH 7.5), 50 mM NaCl; and for LH1 andRC-LH1, 10 mM Tris (pH 7.5), 150 mM NaCl, 2 mMMgCl2 was used. In each case, the ®nal absorbance(850 nm for LH2; 875 nm for LH1) was �20 (1 cmpath-length). Following solubilisation in the dark over20 minutes, each mixture was centrifuged at 150,000 gfor 25 minutes and the clear supernatant applied to aDEAE-Sepharose column. The complexes were elutedwith the appropriate buffer, containing either 30 mMOBG or 2 mM DHPC, in the range of 300 to 400 mMNaCl. After further puri®cation by size-exclusion chro-matography on Superdex-200, the complexes wereready for 2-D crystallisation.

Protein concentrations were determined by the BCAassay method (Pierce & Warriner Ltd., Chester,England). Gel electrophoresis was performed accordingto SchaÈgger & von Jagow (1987) with a Pharmacia MiniGel System (Pharmacia Biotech, St. Albans, England).Absorption spectra of native membranes, solubilizedprotein and reconstituted 2-D crystals were recordedusing a Beckman scanning spectrophotometer and aGuided Wave model 260 ®bre optic spectrophotometer(Guided Wave Inc., El Dorado Hills, CA, USA).

2-D crystallization

LH2 in 10 mM Hepes (pH 7.5), 50 mM NaCl and1% (w/v) OBG was mixed with various volumes of4 mg/ml DMPC, DOPC or plant PC in 2% (w/v)OBG to give ®nal lipid-to-protein weight ratios ran-ging from 0.4 to 0.9. The samples were diluted withthe OBG-containing buffer to give a ®nal protein con-centration in the samples of 0.5 mg/ml. The detergentwas slowly removed using a continuous-¯ow dialysismachine (Jap et al., 1992). During the ®rst two hoursof the dialysis, the temperature was increased from20�C to 35�C, where it was held for six hours.Within the following two hours, the temperature wasdecreased again to 20�C, where it was held for sixhours. This temperature cycle was repeated fourtimes to give a total dialysis time of 64 hours.Samples were then collected and stored at 4�C in thedark. LH1 and RC-LH1 complexes were reconstituted

into 2-D crystals as described for the LH2 complexesusing OBG-solubilized lipids.

Electron microscopy and image processing

The results of reconstitution experiments werechecked by adsorbing samples on a carbon-coated plasticgrid and negative staining with 0.75% (w/v) uranylformate. Micrographs were recorded at a nominalmagni®cation of 50,000� on a Philips CM100 electronmicroscope.

For cryo-electron microscopy, LH2 2-D crystals wereprepared in glucose following the back injection methoddescribed by Nogales et al. (1995). Brie¯y, a small squareof carbon evaporated on mica was ¯oated onto a 2%(w/v) glucose solution and picked up with a 200-meshcopper grid. Then 3 ml of crystal suspension was appliedonto the side where the carbon had been in contact withthe mica, and vigorously mixed with the remaining dropof glucose on the carbon ®lm. Subsequently, the gridwas blotted, air-dried and used immediately for cryo-electron microscopy. Micrographs were recorded with aPhilips CM200 electron microscope equipped with a®eld emission gun and operated at 200 kV. Low-doseimages at a nominal magni®cation of 38,000� wererecorded at electron doses between 5 and 10 eÿ/AÊ 2 atthe specimen and developed in Kodak D-19 developerfor 12 minutes.

Suitable micrograph areas were selected by opticaldiffraction and digitised with a Zeiss SCAI scanner usinga 28 mm step-size for micrographs of negatively stainedcrystals and 7 mm for those of glucose-embeddedcrystals. Images were processed using the MRC imageprocessing package (Henderson et al., 1986; Crowtheret al., 1996) following procedures described by Amos et al.(1982). For single-particle averaging of the RC-LH1crystals, a digitised image was unbent using the MRCsoftware and further processed using the SEMPERimage-processing package. A reference patch was chosenfrom the Fourier peak ®ltered image, used to calculate across-correlation map, and 850 small images (32 � 32pixels) were extracted from the raw image at the pos-itions of the cross-correlation peaks. A single unit cell(LH1 and RC) was chosen from the ®ltered image, theLH1 ring masked out, and the modi®ed density used fortranslational and rotational alignment of the individualunit cells extracted from the raw image. The averagewas subsequently used as new reference, and this cyclewas repeated until the averaging converged. This pro-cess was carried out with three initial references takenfrom different areas of the Fourier-®ltered image, andthe three resulting averages all exhibited the same shapeof the RC.

Acknowledgements

We thank Rachel Pugh for running the SchaÈgger gelshown in Figure 1a and Henning Stahlberg for help withthe single particle work on the RC-LH1 2-D crystals. Weacknowledge generous ®nancial support from the Bio-technology and Biological Sciences Research Council(UK), the Wellcome Trust, the Royal Society, PhilipsElectron Optics, Astra Charnwood, Proctor and GamblePharmaceuticals, Merck Sharp and Dohme, the WolfsonFoundation and the Human Frontier Science Program.T.W. thanks EMBO for a longterm fellowship.

844 Projection Structures of LH2, LH1 and RC-LH1

References

Amos, L. A., Henderson, R. & Unwin, P. N. T. (1982).Three-dimensional structure determination byelectron microscopy of two-dimensional crystals.Prog. Biophys. Mol. Biol. 39, 183±231.

Allen, J. P., Feher, G., Yeates, T. O., Komya, H. & Rees,D. C. (1987). Structure of the reaction centre fromRhodobacter sphaeroides R-26: the cofactors. Proc. NatlAcad. Sci. USA, 84, 5730±5734.

Barz, W. P., Francia, F., Venturoli, G., Melandri, B. A.,Vermigelio, A. & Oesterhelt, D. (1995a). Role ofPufX protein in photosynthetic growth of Rhodobac-ter sphaeroides. 1. PufX is required for ef®cient light-driven electron transfer and photophosphorylationunder anaerobic conditions. Biochemistry, 34, 15235±15247.

Barz, W. P., Francia, F., Venturoli, G., Melandri, B. A.,Vermigelio, A. & Oesterhelt, D. (1995b). Role ofPufX protein in photosynthetic growth of Rhodobac-ter sphaeroides. 2. PufX is required for ef®cientubiquinone/ubiquinol exchange between the reac-tion centre QB site and the cytochrome bc1 complex.Biochemistry, 34, 15248±15258.

Bullough, P. A. & Henderson, R. (1990). Phase accuracyin high-resolution electron-microscopy of trigonaland orthorhombic purple membrane. Biophys. J. 58,705±711.

Bullough, P. A. & Tulloch, P. A. (1990). High-resolutionspot-scan electron-microscopy of microcrystals of analpha-helical coiled-coil protein. J. Mol. Biol. 215,161±173.

Chang, C.-H., Tiede, D., Tang, J., Smith, U., Norris, J. &Schiffer, M. (1986). Structure of Rhodopseudomonassphaeroides R-26 reaction centre. FEBS Letters, 205,82±86.

Cogdell, R. J., Fyfe, P. K., Barrett, S. J., Prince, S., Freer,A. A., Isaacs, N. W. & Hunter, C. N. (1996). Thepurple bacterial photosynthetic unit. Photosynth.Res. 48, 55±63.

Crowther, R. A., Henderson, R. & Smith, J. M. (1996).MRC image processing programs. J. Struct. Biol.116, 9±16.

Deisenhofer, J., Epp, O., Miki, K., Huber, R. & Michel,H. (1985). X-ray structure analysis at 3 AÊ resolutionof a membrane protein complex: folding of theprotein subunits in the photosynthetic reactioncentre from Rhodopseudomonas viridis. Nature, 318,618±624.

Engelhardt, H., Engel, A. & Baumeister, W. (1986). Stoi-chiometric model of the photosynthetic unit ofEctothiorhodospira halochloris. Proc. Natl Acad. Sci.USA, 83, 8972±8976.

Farchaus, J. W., GruÈ nberg, H. & Oesterhelt, D. (1990).Complementation of a reaction centre-de®cient Rho-dobacter sphaeroides pufLMX deletion strain in transwith pufBALM does not restore the photo-synthetic-positive phenotype. J. Bacteriol. 172,977±985.

Farchaus, J. W., Barz, W. P., GruÈ nberg, H. & Oesterhelt,D. (1992). Studies on the expression of the pufXpolypeptide and its requirement for photohetero-trophic growth in Rhodobacter sphaeroides. EMBO J.11, 2779±2788.

Fowler, G. J. S., Hess, S., SundstroÈm, V. & Hunter, C. N.(1997). The role of bArg-10 in the B800 bacterio-chlorophyll and carotenoid pigment environmentwithin the light harvesting LH2 complex of Rhodo-bacter sphaeroides. Biochemistry, 36, 11282±11291.

Ghosh, R., Ghosh-Eicher, S., Di Berardino, M. &Bachofen, R. (1994). Isolation and characterizationof a B875 protein kinase from Rhodospirillumrubrum. Biochim. Biophys. Acta, 1184, 28±36.

Grigorieff, N., Ceska, T. A., Downing, K. H., Baldwin,J. M. & Henderson, R. (1996). Electron-crystallo-graphic re®nement of the structure of bacteriorho-dopsin. J. Mol. Biol. 259, 393±421.

Henderson, R., Baldwin, J. M., Downing, K. H., Lepault,J. & Zemlin, F. (1986). Structure of purple mem-brane from Halobacterium halobium: recording,measurement and evaluation of electron micro-graphs at 3.5 AÊ resolution. Ultramicroscopy, 19, 147±178.

Henderson, R., Baldwin, J. M., Ceska, T. A., Zemlin, F.,Beckmann, E. & Downing, K. H. (1990). Model forthe structure of bacteriorhodopsin based on high-resolution electron cryo-microscopy. J. Mol. Biol.213, 899±929.

Hoenger, A., PageÁs, J.-M., Fourel, D. & Engel, A. (1993).The orientation of porin OmpF in the outer mem-brane of Escherichia coli. J. Mol. Biol. 233, 400±413.

Hunter, C. N., Olsen, J. & van Grondelle, R. (1989).Photosynthetic antenna proteins: 100 ps beforephotosynthesis starts. Trends Biochem. Sci. 14, 72±76.

Ikeda-Yamasaki, I., Odahara, T., Mitsuoka, K., Fujiyoshi,Y. & Murata, K. (1998). Projection map of thereaction center-light harvesting 1 complex fromRhodopseudomonas viridis at 10 AÊ resolution. FEBSLetters, 425, 505±508.

Jap, B. K., Zulauf, M., Scheybani, T., Hefti, A.,Baumeister, W., Aebi, U. & Engel, A. (1992). 2Dcrystallization: from art to science. Ultramicroscopy,46, 45±84.

Jones, M. R., Fowler, J. G. S., Gibson, L. C. D., Grief,G. G., Olsen, J. D., Crielaard, W. & Hunter, C. N.(1992). Construction of mutants of Rhodobactersphaeroides lacking one or more pigment-proteincomplexes and complementation with reactioncentre, LH1 and LH2 genes. Mol. Microbiol. 6, 1173±1184.

Karrasch, S., Bullough, P. A. & Ghosh, R. (1995). The8.5 AÊ projection map of the light-harvesting com-plex I from Rhodospirillum rubrum reveals a ringcomposed of 16 subunits. EMBO J. 14, 631±638.

Kessi, J., Ghosh, R. & Bachofen, R. (1995). Puri®cation ofan LHI-RC-complex of Rhodospirillum rubrum bysolubilization of chromatophores with a short-chainlecithin. Photosynth. Res. 46, 353±362.

Kimura, Y., Vassylyev, D. G., Miyazawa, A., Kidera, A.,Matsushima, M., Mitsuoka, K., Murata, K., Hirai, T.& Fujiyoshi, Y. (1997). Surface structure of bacterio-rhodopsin revealed by high resolution electroncrystallography. Nature, 389, 206±211.

Koepke, J., Hu, X., Muenke, C., Schulten, K. & Michel,H. (1996). The crystal structure of the light-harvest-ing complex II (B800-850) from Rhodospirillummolischianum. Structure, 4, 581±597.

KuÈ hlbrandt, W., Wang, D. N. & Fujiyoshi, Y. (1994).Atomic model of plant light-harvesting complex byelectron crystallography. Nature, 367, 614±621.

Lilburn, T. G., Haith, C. E., Prince, R. C. & Beatty, J. T.(1992). Pleiotropic effects of pufX gene deletion onthe structure and function of the photosyntheticapparatus of Rhodobacter capsulatus. Biochim. Biophys.Acta, 110, 160±170.

McDermott, G., Prince, S. M., Freer, A. A.,Hawthornthwaite-Lawlee, A. M., Papiz, M. Z.,Cogdell, R. J. & Isaacs, N. W. (1995). Crystal struc-

Projection Structures of LH2, LH1 and RC-LH1 845

ture of of an integral membrane light-harvestingcomplex from photosynthetic bacteria. Nature, 374,517±525.

McGlynn, P., Hunter, C. N. & Jones, M. R. (1994). TheRhodobacter sphaeroides Puf X protein is not requiredfor photosynthetic competence in the absence of alight harvesting system. FEBS Letters, 349, 349±353.

McGlynn, P., Westerhuis, W. H. J., Jones, M. R. &Hunter, C. N. (1996). Consequences for the organis-ation of reaction centre light harvesting antenna 1(LH1) core complexes of Rhodobacter sphaeroides aris-ing from the deletion of amino acid residues fromthe C terminus of the LH1 a polypeptide. J. Biol.Chem. 271, 3285±3292.

Miller, K. R. (1982). Three-dimensional structure of aphotosynthetic membrane. Nature, 300, 53±55.

Monger, T. & Parson, W. W. (1977). Singlet-triplet fusionin Rhodopseudomonas sphaeroides chromatophores.Biochim. Biophys. Acta, 460, 393±407.

Montoya, G., Cyrklaff, M. & Sinning, I. (1995). Two-dimensional crystallization and preliminarystructure analysis of light harvesting II (B800-850)complex from the purple bacterium Rhodovulumsul®dophilum. J. Mol. Biol. 250, 1±10.

Nogales, E., Wolf, S. G., Zhang, S. X. & Downing, K. H.(1995). Preservation of 2±D crystals of tubulinfor electron crystallography. J. Struct. Biol. 115,199±208.

Olsen, J. D., Sockalingum, G. D., Robert, B. & Hunter,C. N. (1994). Modi®cation of a hydrogen bond to abacteriochlorophyll a molecule in the light-harvest-ing 1 antenna of Rhodobacter sphaeroides. Proc. NatlAcad. Sci. USA, 91, 7124±7128.

Olsen, J. D., Sturgis, J., Hunter, C. N. & Robert, B.(1997). Site-directed modi®cation of ligands to thebacteriochlorophylls of the light harvesting LH1and LH2 complexes of Rhodobacter sphaeroides.Biochemistry, 36, 12625±12632.

Savage, H., Cyrklaff, M., Montoya, G., KuÈ hlbrandt, W.& Sinning, I. (1996). Two-dimensional structure of

light harvesting complex II (LHII) from the purplebacterium Rhodovulum sul®dophilum and comparisonwith LHII from Rhodopseudomonas acidophila.Structure, 4, 243±252.

SchaÈgger, H. & von Jagow, G. (1987). Tricine-sodiumdodecyl sulfate-polyacrylamide gel electrophoresisfor the separation of proteins in the range from 1 to100 kDa. Anal. Biochem. 166, 368±379.

Stahlberg, H., Dubochet, J., Vogel, H. & Ghosh, R.(1998). Are light-harvesting I complexes fromRhodospirillum rubrum arranged around the reactioncentre in a square geometry? J. Mol. Biol. 282, 819±831.

Stark, W., KuÈ hlbrandt, W., Wildhaber, H., Wehrli, E. &MuÈ hlethaler, K. (1984). The structure of the photo-receptor unit of Rhodopseudomonas viridis. EMBO J.3, 777±783.

Sturgis, J., Olsen, C. N., Robert, B. & Hunter, C. N.(1997). The functions of conserved tryptophanresidues of the core light harvesting complex ofRhodobacter sphaeroides. Biochemistry, 36, 2772±2778.

Valpuesta, J. M., Carrascosa, J. L. & Henderson, R.(1994). Analysis of electron microscope images andelectron diffraction patterns of thin crystals of �29connectors in ice. J. Mol. Biol. 240, 281±287.

Vos, M., van Grondelle, R., van der Kooi, F. W., van derPoll, D., Amesz, J. & Duysens, L. N. M. (1986).Singlet-singlet annihilation at low teperatures in theantenna of purple bacteria. Biochim. Biophys. Acta,850, 501±512.

Walz, T. & Ghosh, R. (1997). Two-dimensional crystalli-zation of the light-harvestingI- reaction centrephotounit from Rhodospirillum rubrum. J. Mol. Biol.265, 107±111.

Woodbury, N. W. & Allen, J. P. (1995). The pathway,kinetics and thermodynamics of electron transfer inwild type and mutant reaction centers of purplenonsulfur bacteria. In The Anoxygenic PhotosyntheticBacteria (Blankenship, R. E., Madigan, M. T. &Bauer, C. E., eds), Kluwer, Dordrecht.

Edited by K. Nagai

(Received 27 February 1998; received in revised form 23 June 1998; accepted 7 July 1998)