AFM Characterization of Tilt and IntrinsicFlexibility of Rhodobacter sphaeroidesLight Harvesting Complex 2 (LH2)

Simon Scheuring1*, Jerome Seguin2, Sergio Marco1, Daniel Levy1

Cecile Breyton3, Bruno Robert2 and Jean-Louis Rigaud1

1Institut Curie, UMR-CNRS168 and LRC-CEA 34V11 rue Pierre et Marie Curie75231 Paris, Cedex 05, France

2Service de Biophysique desFonctions MembranairesDepartement de BiologieJoliot-Curie, CEA and URA2096, CNRS CEA-Saclay91191 Gif sur Yvette, France

3Institut de BiologiePhysico-ChimiqueCNRS-UMR709913 rue Pierre et Marie Curie75005 Paris, France

Atomic force microscopy (AFM) has developed into a powerful tool toinvestigate membrane protein surfaces in a close-to-native environment.Here we report on the surface topography of Rhodobacter sphaeroides lightharvesting complex 2 (LH2) reconstituted into two-dimensional crystals.These photosynthetic trans-membrane proteins formed cylindrical oligo-meric complexes, which inserted tilted into the lipid membrane. Thispeculiar packing of an integral membrane protein allowed us to deter-mine oligomerization and tilt of the LH2 complexes, but also protrusionheight and intrinsic flexibility of their individual subunits. Furthermorethe surface contouring reliability and limits of the atomic forcemicroscopy could be studied.

The two-dimensional crystals examined had sizes of up to 5 mm and, asrevealed by a 10 A cryo electron microscopy projection map, p22121 crystalsymmetry. The unit cell had dimensions of a ¼ b ¼ 150 A and g ¼ 908, andhoused four nonameric complexes, two pointing up and two pointingdown. AFM topographs of these 2D crystals had a lateral resolution of10 A. Further, the high vertical resolution of ,1 A, allowed the protrusionheight of the cylindrical LH2 complexes over the membrane to be deter-mined. This was maximally 13.1 A on one side and 3.8 A on the other.Interestingly, the protrusion height varied across the LH2 complexes,showing the complexes to be inserted with a 6.28 tilt with respect to themembrane plane. A detailed analysis of the individual subunits showedthe intrinsic flexibility of the membrane protruding peptide stretches tobe equal and independent of their protrusion height. Furthermore, ouranalysis of membrane proteins within this peculiar packing confirmedthe high vertical resolution of the atomic force microscopy on biologicalsamples, and led us to conclude that the image acquisition function wasequally accurate for contouring protrusions with heights up to ,15 A.

q 2003 Elsevier Science Ltd. All rights reserved

Keywords: membrane protein; photosynthetic protein; 2D-crystallization;cryo electron microscopy; atomic force microscopy*Corresponding author

Introduction

The atomic force microscopy (AFM)1 has becomea powerful tool in membrane protein research,2

allowing information to be acquired at submolecu-lar resolution on the membrane protruding struc-tures of single proteins.3 – 5 Such information isdifficult to obtain by cryo electron microscopy(cryo EM6) where the membrane-embeddedregions are preserved, but connecting loops andprotruding termini are often unresolved.7,8 Usingthe AFM, the heights of membranes and pro-truding structures can be measured accurately

0022-2836/03/$ - see front matter q 2003 Elsevier Science Ltd. All rights reserved

E-mail address of the corresponding author:simon.scheuring@curie.fr

Abbreviations used: AFM, atomic force microscopy;Bchl, bacteriochlorophyll; CTF, contrast transfer function;CV, coefficient of variance; DMPC, 1,2-dimyristoyl-sn-glycero-3-phosphocholine; DOPC, 1,2-dioleoyl-sn-glycero-3-phosphocholine; EM, electron microscopy;LDAO, N,N-dimethyldodecylamine N-oxide; LH1, lightharvesting complex 1; LH2, light harvesting complex 2;OTG, n-octyl-b-D thioglucopyranoside; RMS, root meansquare; SD, standard deviation.

doi:10.1016/S0022-2836(02)01241-X J. Mol. Biol. (2003) 325, 569–580

with a vertical resolution of ,1 A.9,10 Further thehigh signal-to-noise ratio of this instrument allowsthe oligomeric state and sidedness of membraneproteins directly to be assessed in raw dataimages,11 – 13 and poorly ordered single particles tobe recognized and imaged at high resolution(,10 A).14,15 However, interpretation of AFM topo-graphs requires an understanding of AFM surfacecontouring reliability and limits. Crucial para-meters are related to the physics of the surface con-touring mechanism employed, i.e. the feedbackloop, but also on the nature of the sample imaged,i.e. the mobility of any membrane protrudingprotein structures. For both reasons, one generallyexpects the contouring precision to decrease withincreasing height of surface protruding structures.Indeed, on one hand the feedback loop faces strongheight differences within short lateral distances,and, on the other hand, the peptide mobilityincreases with increasing length of protrudingpeptide. To probe the accuracy and limitations ofAFM contouring experimentally, we have per-formed a detailed analysis of two-dimensional(2D) crystals of the light-harvesting complex 2(LH2) from Rhodobacter sphaeroides.

In bacterial photosynthesis, absorption of lightinitiates a cyclic electron transfer coupled to protontranslocation across the cytoplasmic membrane.Two types of trans-membrane protein–pigmentcomplexes, designated light harvesting complexes1 (LH1) and 2 (LH2), absorb light. The LH1 com-plexes are closely associated with the reactioncenter, together forming the so-called core-com-plex. The LH2 complexes are peripheral to thiscore. The energy of absorbed light is transferredfrom LH2 to LH1, and finally to the reaction center,where a redox-reaction causes charge separationacross the membrane.16 – 18 Three-dimensionalcrystal structures of LH2 from Rps. acidophila andRps. molischianum have revealed that the complexesare composed of nine or eight ab-heterodimers,respectively, forming a double ring of proteinaround a hollow cylinder.19,20 Interestingly, a pro-jection map from tubular 2D crystals ofRb. sphaeroides LH2 was found incompatible with aperfect ninefold symmetry of the complex.21 Thedifferent intensities of the projected subunits wereinterpreted assuming the whole LH2 cylinder tobe tilted with respect to the membrane plane.21

This peculiar packing of identical polypeptidesimplies that they protrude by different heightsfrom the membrane plane, offering the oppor-tunity of probing the height and flexibility ofthese membrane protruding protein regions byAFM.

To measure the surface topography of LH2 fromRb. sphaeroides by AFM, we have grown 2D crystalsfrom protein/lipid/detergent mixtures supple-mented with n-octyl-b-D thioglucopyranoside(OTG), a detergent that has been reported toincrease significantly the size of reconstituted 2Dcrystals.22 The use of this additional detergent ledto the production of large crystalline sheets up to

several microns, well suited for AFM analysis, asopposed to the 600 nm width tubular 2D crystalsobtained by Walz et al. The crystals were wellordered with p22121 symmetry and a unit cell ofa ¼ b ¼ 150 A, g ¼ 908, housing four nonamericLH2 rings in up and down orientations. As foundfor tubular crystals, the cryo EM analysis at 10 Aresolution from these large crystalline sheets wasincompatible with a perfect ninefold symmetry ofthe projected LH2 cylinder complex. AFM topogra-phy maps of the 2D crystals at 10 A resolutionrevealed unequivocally that the whole LH2 com-plexes were tilted with respect to the membraneplane. The height distribution of the LH2 subunitsperpendicular to the cylinder tilt axis was foundto vary linearly between 13 A and 6 A on the peri-plasmic side, corresponding to a tilt of 6.28. Thispeculiar packing allowed the intrinsic flexibility ofthe membrane protruding protein domains to beanalyzed by AFM, since the complexes exposedidentical polypeptides, which protrude differentlyfrom the membrane plane. The standard deviations(SDs) of the differently protruding subunitsincreased also linearly with increasing protrusionheight over the membrane. However, the coef-ficient of variance (CV) was found constant for allnine subunits of the complex. This led us to con-clude that the intrinsic flexibility of the peptideswas identical, independent of their protrusionheights. Finally, the data show the AFM imageacquisition mechanism to be equally precisewhen contouring protrusions up to ,15 A inheight.

Results

Purification and topology model

The LH2 complexes of Rb. sphaeroides were puri-fied from photosynthetic membranes throughsolubilization with LDAO, followed by centrifu-gations, and anion exchange and gel filtrationchromatographies.23 LH2 complexes are composedof a and b-polypeptides consisting of 54 (calcu-lated Mr 5599 Da) and 51 (calculated Mr 5317 Da)amino acid residues, respectively. The a andb-polypeptides run separately at apparent molecu-lar masses of ,4.5 kDa and ,3.5 kDa in 16.5%SDS-PAGE (Figure 1(a)). The difference betweenthe apparent and the calculated Mr is probablydue to the stability of the trans-membranea-helices in SDS, their compactness making themolecules appear smaller.

On the basis of a sequence alignment with theLH2 of Rps. acidophila (Figure 1(b)), for which theatomic coordinates are available,19 a topologicalmodel of the LH2 of Rb. sphaeroides can be built(Figure 1(c)). Both, a and b-polypeptides cross themembrane once as trans-membrane a-helices.The a-polypeptide forms a second short hydro-philic a-helix on the periplasmic surface.

570 AFM Study of Rb. sphaeroides LH2

2D-crystallization

When LDAO-purified LH2 complexes are sup-plemented with phospholipids, and detergent isremoved in two hours by successive addition ofBio-Beads,24,25 small vesicular 2D-crystals ofabout 0.1–0.2 mm in diameter are formed (datanot shown). Addition of 20 mM OTG to the initialmicellar lipid–protein–LDAO solution, prior todetergent removal, leads to the formation ofnumerous large vesicular 2D crystals with sizesup to 5 mm.22 Unilamellar sheets were also present,likely originating from vesicular crystals brokenopen. The size of reconstituted 2D crystals waschecked by negative stain electron microscopy(Figure 2(b)). Crystallinity and unilamellarity weretested by calculating Fourier transforms fromimages recorded on a CCD camera (Figure 2(c)).

The use of Bio-Beads for controlled detergentremoval allows reconstitution of 2D crystals in afew hours, avoiding denaturation of membraneproteins. For LH2, we have followed, throughoutthe crystallization process, the integrity of thecomplexes by monitoring the absorption spectraof the protein–pigment complexes. In a nativeenvironment, LH2 complexes exhibit a set ofcharacteristic absorption peaks originating frombacteriochlorophyll (Bchl) and carotenoidmolecules18 (Figure 2(a)). Absorption spectra ofLH2 protein in 0.1% LDAO indicate no shift of thecarotenoid bands, however, significant shifts of theQy transitions of Bchl a molecules from 802 m to800 nm and from 852.8 nm to 848.2 nm are detectedupon detergent solubilization (Figure 2(a); spec-trum 1). The addition of phospholipids to thesolubilized LH2 protein has no influence on thespectra, indicating that protein–lipid contacts internary micelles cannot, alone, restore the native-like absorption bands (Figure 2(a); spectrum 2; see

also inset). A further addition of 20 mM OTG tothe sample does not influence the positions of theLH2 absorption transitions either, indicating nodenaturing effect of this thio-glycosylated deter-gent (Figure 2(a); spectrum 3). However, thebaseline intensity is seen to increase upon OTGaddition, that could be related to the formation oflarger micellar structures which might, inturn, favor the formation of large reconstitutedmembranes.22 Following addition of Bio-Beads, thebaseline progressively decreases and importantly,the Qy transitions of Bchl a molecules shift back,progressively, from 800 nm to 802 nm and from848.2 nm to 852.8 nm (Figure 2(a), spectra 4 and 5;inset). These last observations suggest that com-plete detergent removal is necessary for restoringthe membrane pressure and the native-like absorp-tion peaks.

EM analysis

For EM analysis, the large 2D crystals of LH2obtained in the presence of OTG were quick-frozenin vitreous ice and images acquired on a slow scanCCD camera. The Fourier transform computed ofan image of a crystal after correction of the latticedistortion is shown in Figure 3(a). Significantreflections appear up to 10 A resolution. Indeed,analysis of the overall phase residual after mergingsix lattices showed the data to be reliable up to10 A (Table 1). The crystal symmetry is found tobe p22121 (ALLSPACE26). The unit cell(a ¼ b ¼ 150(^ 0.5) A, g ¼ 90(^0.3)8; n ¼ 6), com-prises four individual circular LH2 complexes,two pointing up, and the other two pointingdown (Figure 3(b) and (c)). In both non-symme-trized and symmetrized maps, individual LH2complexes are resolved into two concentric ringsof nine densities with diameters of ,40 A and,62 A. Comparisons with the atomic modelsof the LH2’s from Rps. acidophila andRps. molischianum indicate, that the inner andouter rings of densities represent the projection

Figure 1. Topology of the Rb. sphaeroides LH2. (a) SDS-PAGE of Rb. sphaeroides LH2. Left lane, marker proteins, withmolecular weights as indicated; right lane, purified LH2 complex. (b) Sequence alignment (ClustalW†) of the a andthe b-polypeptide of (1) Rps. acidophila, (2) Rvi. gelatinosus and (3) Rb. sphaeroides LH2 (NCBI Pubmed‡). The yellowand the blue boxed regions correspond to trans-membrane a-helices in the a-polypeptide and the b-polypeptide ofRps. acidophila, respectively. The red boxed region corresponds to the periplasmic a-helix in the a-polypeptide. Thegray box in the Rvi. gelatinosus a-polypeptide sequence indicates the proteolyzed C terminus used for sidednessassignment by AFM.11 (c) Rb. sphaeroides topology model derived from the sequence alignment with Rps. acidophilashown in (b). Arrow points towards the hollow cylinder center of a complex.

† http://www.ch.embnet.org/software/ClustalW.html

‡ http://www.ncbi.nlm.nih.gov/

AFM Study of Rb. sphaeroides LH2 571

densities of the trans-membrane helices of the aand the b-polypeptide subunits, respectively.19

Interestingly, the ninefold symmetry of the com-plexes is not respected in projection normal to themembrane plane. This observation agrees with

the reported 6 A projection map calculated fromtubular crystals of Rb. sphaeroides LH2 complexes,21

and can be interpreted as resulting from defor-mation of the cylinder or tilt of the entire cylindri-cal molecule with respect to the membrane plane.

Figure 2. Absorption spectroscopy during 2D-crystallization of the Rb. sphaeroides LH2. (a) Absorption spectra show-ing the LH2 characteristic transitions at 375 nm (BChl Soret transition), 460 nm, 490 nm and 512 nm (carotenoid S0 to S2

transition), 590 nm (BChl Qx transition), ,800 nm and ,850 nm (BChl Qy transitions). Spectrum 1: purified LH2 in0.1% LDAO; spectrum 2: sample 1 supplemented and mixed 30 minutes with DOPC (lipid-to-protein: 0.5 (w/w));spectrum 3: sample 2 supplemented and mixed 30 minutes with 20 mM OTG; spectrum 4: early reconstitution after30 minutes stirring with 1 mg Bio-Beads/10 ml sample 3; spectrum 5: 2D crystals after two hours stirring with 2 mgBio-Beads/10 ml sample. Inset: Shift of the Qy peak from 848.2 nm (sample 1 and sample 2) over 850.4 nm (sample 4)to 852.8 nm (sample 5). (b) Negative stain overview electron micrograph of Rb. sphaeroides LH2 2D crystals grownusing OTG (full image size: 7.5 mm). The crystals are vesicular and frequently break open upon adsorption on thegrid. (c) Fourier transform of negatively stained unilamellar 2D crystal shown in (b).

Figure 3. Cryo electron microscopy analysis of the Rb. sphaeroides LH2 2D-crystals. (a) Calculated Fourier transformof a cryo embedded LH2 2D crystal. The size of the boxes indicates quality of the spots.38 The circles represent thezero passes of the contrast transfer function (CTF). The broken circle represents 10 A resolution, which was chosen asresolution cutoff as judged by phase residual analysis. (b) One unit cell (a ¼ b ¼ 150 A; g ¼ 908) of the non-sym-metrized ( p1) projection map of one LH2 2D crystal at 10 A resolution. (c) One unit cell (a ¼ b ¼ 150 A; g ¼ 908) ofthe p22121 symmetrized projection map from six merged images at 10 A resolution.

572 AFM Study of Rb. sphaeroides LH2

AFM analysis

To interpret the marked deviation from the nine-fold symmetry in projection maps, we have per-formed an AFM analysis of the 2D crystals. The2D crystals are adsorbed onto freshly cleavedmica using an adsorption buffer containing 50 mMMgCl2, and topographs are acquired using thesame buffer without divalent ions (see Materialsand Methods). Height measurements in buffersolution of the 2D crystals (69.4(^1.1) A; n ¼ 24),and of DOPC bilayers without protein incorpo-rated (40.4(^1.5) A; n ¼ 22) indicate that hydro-philic protein domains protrude significantly

from the membrane, in agreement with previousreports.11,27

A raw data AFM topograph of a Rb. sphaeroidesLH2 2D crystal is shown in Figure 4(a). The unitcell dimension is similar to that determined byEM (Table 1), housing four individual LH2 rings,,53 A in diameter. Calculated power spectrashow complete diffraction patterns to ,12 A reso-lution (Figure 4(b)). Phase residual and signal-to-(4·noise) analyses of the unit cells extracted andtreated as single particles reveal a resolution upto 10 A (Figure 4(c)). It should be pointed outthat although the crystals imaged are the same,the two techniques employed reveal different

Figure 4. Atomic force microscopy analysis of LH2 2D-crystals from Rb. sphaeroides. (a) The unit cell (a ¼ b ¼ 150 A;g ¼ 908) is delineated including all p2 symmetry centers. White circles delineate the four rings within a unit cell (H,high side, periplasmic surface; L, low side, cytoplasmic surface). The small squares in the left corner of the unit cellout-line in (a) (and inset) indicate the lipid region, which was used to define height 0.0 A. Raw data AFM topographof a Rb. sphaeroides LH2 2D crystal. The tilted integration of the LH2 cylinders is clearly visible by the uneven heightdistribution of the ring surfaces (scale bar represents 200 A; full gray scale: 15 A). Inset: Average topography of theRb. sphaeroides LH2 2D crystal. One side of the protein rings (diameter 53 A) protrudes far more out of the lipid bilayer(high edge: 13.1 A; low edge 6.3 A) than the other (high edge: 3.8 A; low edge: invisible). All four rings within the unitcell are tilted (full gray scale 13.1 A). (b) Calculated Fourier transform of image (a). The diffraction pattern is completeto ,12 A resolution. The broken gray circle corresponds to 10 A resolution. (c) Phase residual and spectral signal-to-(4·noise) analysis plots from single particle analysis of 321 unit cells extracted from image (a) and its correspondingback-trace image. The Nyquist frequency corresponds to 7.4 A. The (4·noise) line crosses the signal at 9.9 A, the phaseresidual is close to the 458 at 10.3 A resolution, and crosses 458 at 9.5 A.

Table 1. Characterization of Rb. sphaeroides LH2 2D crystals using cryo EM

Dimension Error n

Unit cell dimension (housing four nonameric rings) a ¼ b ¼ 150 A 0.5 A 6a

g ¼ 908 0.38 6a

Defocus range 1200–2500 ASymmetry space group p22121

Resolution range nb Phase res. (458 is random)1 160.0–27.7 36 26.62 27.7–19.2 40 26.53 19.2–15.6 24 38.64 15.6–13.6 21 34.45 13.6–12.1 17 28.06 12.1–11.0 25 37.17 11.0–10.2 21 35.7c

8 10.2–9.6 19 43.69 9.6–9.0 18 48.7

Overall 160.0–10.2 184 31.4

a Number of lattices.b Number of unique reflections (IQ # 7)c Resolution cutoff was set at 10.0 A (see also Figure 3).

AFM Study of Rb. sphaeroides LH2 573

structural aspects: EM allows the calculation of aprojection of the whole structure whereas AFMtopographs represent structural information ofonly one surface. Thus the EM projection mapreveals the p22121 symmetry of the crystal, whereasin AFM topographs, p2 only can be applied, sincethe molecules related by the screw axes offer adifferent surface.

When lipid regions are set to 0.0 A height (whitesquare in Figure 4(a) and inset), two of the LH2rings in a unit cell strongly protrude from themembrane with a maximal height of 13.1 A, whilethe two others weakly protrude from the mem-brane plane with a maximal height of 3.8 A (Figure4(a), inset). This difference in height of the individ-ual LH2 rings observed by AFM is related to theup and down orientation of the complexes, and isconsistent with the p22121 symmetry determinedin the EM projection maps (Figure 3(b) and (c)).

Interestingly, in raw images, an uneven heightdistribution of the subunits of the LH2 rings isclearly visible, indicating a tilt of the rings with

respect to the membrane plane (Figure 4(a)).Averages with applied p2 symmetry clarify thetilted integration of the LH2 molecules in the crys-tal (Figure 4(a) inset). On the strongly protrudingLH2 surface, nine subunits can be identified,which exhibit a maximal protrusion height of13.1 A and a lowest protrusion height of 6.3 A. Incontrast, the weakly protruding LH2 moleculesexhibit a maximal protrusion height of 3.8 A, andonly four subunits can be unambiguously identi-fied in the averaged AFM topography (Figure 5;Table 2; see also Discussion). The peak heights ofthe subunits of the strongly protruding surface inthe average topography of the Rb. sphaeroides LH2were used to characterize the tilt of the cylinder-shaped complexes with respect to the membraneplane (Figure 5(a) and (c); Table 2). They variedbetween 13.1 A and 6.3 A, and were linearly fitted(R ¼ 0.97) perpendicular to the cylinder tilt axis(white line in Figure 5(a)) with a steepness of20.109 corresponding to a tilt of 6.2(^0.8)8( p ¼ 0.95) (Figure 5(c)).

Figure 5. Tilt, SD, CV, and volume analysis of the periplasmic surface of the Rb. sphaeroides LH2 rings. (a) AverageAFM topography. On the high (periplasmic) side, all nine subunits are clearly visible (H1–H9), on the low (cyto-plasmic) side only four (L1–L4) are distinguishable. The lipid region was used to calibrate height 0.0 A. The whiteline is perpendicular to the cylinder tilt axis, starting from the most protruding subunit (full image size: 230 A; fullgray scale: 13.1 A). (b) SD map corresponding to (a). Measuring positions were defined by the topography (a) (fullimage size: 230 A; full gray scale: 1.2 A , SD , 3.2 A). The lipid region exhibits minimal SD of 1.2 A. (c) Height, SD,CV, and volume distribution of the strongly protruding (H1-H9) subunits along the white line in (a) and (b). The plotof the individual subunit protrusion measurements (circles; line: height) corresponds to a tilt of the LH2 cylinder of6.28. The gray star indicates the average protrusion height (9.5 A) of the periplasmic surface. The SD (squares; line:SD) increases by 0.28 A when the polypeptide protrudes 1.0 A more out of the membrane. The coefficient of variance(diamonds; line: CV) is constant, indicating identical elastic properties of the polypeptide between 6.3 A and 13.1 Aof protrusion height over the membrane. The lipid regions (indicated in (a) and (b)) were used to define height 0.0 Aand show a SD (broken line: SD lipid) of 1.2 A. Volumes of the nine subunits are plotted in gray (full circles; rightsided Y-axis).

574 AFM Study of Rb. sphaeroides LH2

Deflection differences of the cantilever on identi-cal positions on the protein surface are representedby standard deviation (SD) maps. Hence, the samepositions were investigated on the simultaneouslycalculated SD map to assess the corresponding SDvalues (Figure 5(b) and (c); Table 2). The SD valuesof the subunits corresponding to the heightmeasurements (highest subunit: 3.2 A SD; lowestsubunit: 2.0 A SD; Figure 5(b) and (c); Table 1) arealso linearly fitted (R ¼ 0.94). Minimal SD valuesof 1.2 A are found in the lipid regions (indicatedin Figure 5(b)).

Finally, the calculation of coefficients of variance(CV) of the nine subunits defined a flat line(steepness: 0.001; R ¼ 0.92) reflecting a constantflexibility of the polypeptide, independent of pro-trusion height (Figure 5(c)). It may thus be calcu-lated that the SD linearly correlates as D0.28 ASD/D1.0 A height (n ¼ 9) in the region of the poly-peptide protrusion, from 6.3 A to 13.1 A in height.The result of a constant CV of 0.28(^0.01) on thenine differently protruding polypeptides indicatesthat, under the scanning conditions stated, thefeedback loop driving the piezo contouring of theprotein surface behaves reliably and does notincrease noise when scanning peptides protrudingto greater height. In addition, the volumes of thenine subunits were calculated (,3100 A3 for thehighest and ,1250 A3 for the lowest protruding

subunit, Figure 5(c); Table 2). The tip convolutioneffect is reflected by the volumes calculated fromthe measurement of identical domains protrudingdifferently from the membrane plane (see alsoDiscussion).

The four LH2 cylinders within the unit cellpresent different center-to-center distances, con-tacts and angular arrangements between eachother, allowing only one alignment of the cryoEM projection (Figure 6; bottom) and the AFMtopography (Figure 6; top) of 2D crystals withidentical LH2 packing. Taking together the p22121

crystal symmetry and the unit cell dimensionsderived from the cryo EM analysis, with thesubunit peak positions and height, and the lipidbilayer thickness from AFM measurements, thetilted packing arrangement of the four LH2 ringswithin one unit cell can be represented (Figure6(b)).

Discussion

We have performed a detailed AFM analysis oftilted inserted trans-membrane proteins, as foundin 2D crystals of light-harvesting complex 2 (LH2)of Rb. sphaeroides. The peculiar packing of thiscomplex in the lipid membrane allowed theinfluence of protrusion height on protein flexibility

Table 2. Characterization of Rb. sphaeroides LH2 2D crystals using AFM

Dimensiona Errora n

Unit cell dimension (housing four nonameric rings) a ¼ b ¼ 150 A 3 A 4b

g ¼ 908 18 4b

Thickness DOPC bilayer 40.4 A 1.5 A 22Thickness LH2 2D crystal 69.4 A 1.1 A 24Symmetry space group p2

Heightc (A) SDd (A) Vol.e (A3) SDf (A3) ng

Lipid region Set: 0.0 A 1.2 A 321High side (extracellular)Subunit 1 13.1 3.2 3100 700 321Subunit 2 11.4 2.9 2450 550 321Subunit 3 10.7 2.8 1900 400 321Subunit 4 8.4 2.5 1650 300 321Subunit 5 7.1 2.2 1450 250 321Subunit 6 6.3 2.0 1250 200 321Subunit 7 7.9 2.3 1900 250 321Subunit 8 9.4 2.5 2100 400 321Subunit 9 11.5 3.2 2800 650 321

Low side (cytoplasmic)Subunit 1 3.8 2.1 321Subunit 2 3.4 1.9 321Subunit 3 2.7 1.8 321Subunit 4 2.2 1.6 321Subunit 5, 6, 7, 8, 9 Not visibleh Not visibleh

a Gained from images taken at different magnifications.b Number of lattices.c The height of the lipid region (as indicated in Figures 4 and 6) was set as 0.0 A.d Gained through single-particle averaging (see Figure 5(b)).e Volumes were calculated by integrating pixel number and gray value using unit cell dimensions and heights determined (see

Materials and Methods).f Deduced from SD values of height measurement.g Number of unit cells taken from trace and retrace images merged to yield the average height and the SD map.h See Discussion.

AFM Study of Rb. sphaeroides LH2 575

to be examined. Further, the system offered aunique opportunity to assess the contouringreliability of the AFM.

Although the production of well-ordered 2Dcrystals of LH2 from Rb. sphaeroides has beenreported and lead to a cryo EM projection map at6 A resolution,21 the tubular shape and the rela-tively limited size of the reconstituted crystalsprecluded analysis by AFM. We have producedlarge 2D arrays of LH2 complexes by detergentremoval from a lipid–protein–detergent micellarsolution, supplemented with octyl-thioglucosideto increase the size of the reconstituted structures.22

Using this additional thio-glycosylated detergent,large unilamellar sheets with sizes of up to 5 mmwere formed, allowing the acquisition of structuralinformation at 10 A resolution. Although not aswell ordered as the tubular 2D crystals reportedby Walz et al.,21 the large planar 2D crystals pro-duced were perfectly suited to AFM, a techniquefor which the size of reconstituted membranes ismore important than a high crystallinity. Analysisof the spectroscopic properties of the LH2 complexduring the crystallization process indicated thatthe reconstitution of the protein into 2D crystalsdid not modify the integrity of all the pigmentbinding sites nor the molecular assembly of thecomplex.

When analyzed by cryo EM, these 2D crystalsshowed p22121 symmetry. The projection map at10 A resolution displayed a ring-like LH2 proteinunit, consisting of nine a/b heterodimers resolved

into two concentric rings with average diametersof ,40 A and ,62 A (Figure 3). However, one ofthe main characteristics of the projection map wasthat the densities of the individual subunits werenot equal within one projected nonamer. The innerdensities and, to a lesser extent, the outer densitiesshowed a marked variation in contrast around thering. This deviation from ninefold symmetry inour projection map agreed with the 6 A projec-tion map calculated from tubular crystals ofRb. sphaeroides LH2.21 By calculating back-projections from the 3D structure of Rps. acidophiolaLH2, these authors suggested a tilt of thewhole cylindrical LH2 complex with respect to themembrane plane of ,58 rather than a deformationof the LH2 cylinders induced by the crystal pack-ing. This interpretation is experimentally corrobo-rated by our detailed AFM analysis of LH2 2Dcrystals.

Firstly, AFM topographs revealed strongly andweakly protruding ring-like structures, corre-sponding to cylindrical LH2 complexes reconsti-tuted in opposite orientations. Nine subunits withan average height of 9.5 A can be identified on thesurface of the strongly protruding cylinders, whileonly four subunits can be identified on the weaklyprotruding surface and the maximal protrusionheight is 3.8 A. For a comparison, we have pre-viously found that native LH2 from Rvi. gelatinosusprotruded out of the membrane by 14 A on oneside and by 5 A on the other side.11 After proteo-lytic cleavage of the periplasmic C terminus of

Figure 6. Packing arrangement of tilted Rb. sphaeroides LH2 complexes in 2D-crystals. (a) Alignment of AFM topogra-phy (top) and cryo EM projection (bottom) maps. The white lines delineate three successive unit cells. (b) Packingmodel derived from the subunit peak positions of the four nonameric cylinders within one unit cell (a ¼ b ¼ 150 A,g ¼ 908, thickness z ¼ 66 A). The subunit peak positions of the surface facing the viewer are plotted with filled circles,the open circles delineate the positions on the other side of the membrane. All axes are in A and in identical scaling.The lines at 20 A and 220 A height correspond to the putative lipid bilayer (DOPC) surfaces. The unit cell is displayedfrom four different angles, as indicated and by the black arrows pointing on the unit cell insets.

576 AFM Study of Rb. sphaeroides LH2

the a-polypeptide, these LH2 protruded 9 A and5 A, respectively, which allowed a sidednessassignment.11The high sequence homologybetween the a and the b-subunit fromRvi. gelatinosus and Rb. sphaeroides (Figure 1(a)),strongly suggested an identical sidedness forRb. Sphaeroides LH2, with the strongly and theweakly protruding surfaces corresponding tothe periplasmic and the cytoplasmic sides,respectively.

Secondly, the subunit peak positions of the peri-plasmic surface (maximal protrusion height of13.1 A) in the average topography of theRb. sphaeroides LH2 were used to analyze in detailthe tilt of the cylinder-shaped complexes withrespect to the membrane plane. This AFM topogra-phy analysis demonstrates that the whole LH2complex is tilted 6.28 with respect to the membraneplane (Figures 5(c) and 6). For comparison Walzet al.21suggested a tilt of ,58 from comparing anEM projection map with calculated projections ofthe Rp. acidophila X-ray structure. However, theAFM surface topography accuracy had been testedand approved by comparison with X-ray and EMstructures,2,28 confirming that the AFM tilt analysisof the LH2 cylinders is precise. Concerning theweakly protruding cytoplasmic surface, it exhibi-ted a maximal protrusion height of 3.8 A, andonly four subunits can be unambiguously identi-fied in the average AFM topography. Given theheight difference between the highest and the low-est subunit on the periplasmic surface (7 A), wecould calculate that, due to the tilt, the lower edgeof the cytoplasmic surface was below the 0.0 Alevel defined by the lipid areas. In this regard, thealignment of the AFM topography and the cryoEM projection (Figure 6(a)) indicated that the leastprotruding cytoplasmic subunits were in closecontact or possibly even covered by the stronglyprotruding periplasmic subunits, and consequentlynot accessible for to the AFM tip. In addition,considering the alignment of cryo EM projectionand AFM topography maps (Figure 6(a)), thesimplest explanation is that the tilt of the cylindersis caused by the protein–protein contacts betweenLH2 complexes in the 2D crystal (Figure 6(b)).In detail, the tilt might be induced by the tilteda-helices of the peripheral b-polypetides as evi-denced by Walz et al.19,21 Another interpretationwould be that the tilt reflects the native state ofthe complex, a hypothesis, which will be testedby AFM analysis of Rps. sphaeroides native photo-synthetic membranes.

Thirdly, the SD map simultaneously calculatedwith the average AFM topography represents amap of deflection differences of the cantilever onidentical positions on the protein surfaces, andhas been interpreted to directly reflect proteinflexibility.29 Protrusion peak analysis of proteinsurfaces allowed the calculation of position proba-bility maps and the assessment of the lateraldisplacement of protein domains.30 Structural fea-tures having the strongest mobility are enhanced

in SD maps and position probability maps allow-ing global variability of distinct regions of proteinsto be identified.29,30 On the LH2 2D crystals, theSD values increased linearly with increasing heightof the subunits (Figure 5(b) and (c)). To testwhether this increase in SD values correspondedto an increased intrinsic flexibility of the peptides,the corresponding CV values were calculated(Figure 5(c)). For all nine molecules a constantSD/height ratio was found, which resulted in aCV of 0.28 (Figure 5(c)). Such normalization of theSD values with the corresponding topographydata reflects adequately protein flexibility, andmight be used in the future to locate function-related flexible domains on membrane protein sur-faces. Indeed, since high resolution AFM on pro-teins is restricted to a narrow range of forcesapplied to the cantilever,10 the SD/height ratio willbe a comparable measure for the intrinsic flexibilityof protruding protein domains. The finding of aconstant SD/height ratio not only showed that theintrinsic flexibility of all subunits was equal, what-ever their height over the membrane, but also thatthe AFM surface contouring is equally reliable forscanning stronger or weaker protruding structures.One might have expected the SD to become moreimportant with increasing height of surface protru-sion as a function of the height, i.e. a less precisecontouring as a function of height. This was notthe case; the feedback loop did not add additionalnoise to the measurement of higher protrudingstructures (in the range up to ,15 A). Finally,although the subunits reveal linearity in protrusionheight due to the tilt of the cylinder, estimates ofthe volumes of contoured protrusions11,14,31,32 donot linearly coincide. The values of the strongestprotruding domains indicate a significant increasein volume due to convolution with the sphericaltip geometry (Figure 5(c); Table 2).

In conclusion, structural information on thesurface-exposed domains of the LH2 complex ofRb. sphaeroides was acquired by AFM. The datawere complementary to those obtained by EM,which mainly provides information on the mem-brane-embedded part of the protein. For the firsttime, the high vertical resolution (,1 A) of theAFM was exploited to experimentally measure thepacking tilt of membrane proteins inserted in alipid bilayer. Furthermore, the data presentedillustrate the power of the AFM to investigatenative photosynthetic membranes, where a highsignal-to-noise ratio will be a prerequisite for theidentification of individual complexes workingtogether within the photosynthetic supercomplex.

Materials and Methods

Materials

Phospholipids of the highest purity were purchasedfrom Avanti Polar Lipids. N,N-dimethyldodecylamineN-oxide (LDAO, 30% solution) was from Fluka and

AFM Study of Rb. sphaeroides LH2 577

n-octyl-b-D thioglucopyranoside (OTG) was from Sigma.Bio-Beads SM2 (25–50 mesh) from Bio-Rad were washedwith methanol and extensively with water before use asdescribed.24 All other reagents were of analytical grade.

Isolation and purification of LH2

Rb. sphaeroides, strain 2.4.1, was grown photohetero-trophically in Bose medium33 at 28 8C in one liter bottles.Cells were harvested and membranes prepared byFrench Press (1000 PSI), followed by low-speed centrifu-gation (20,000g, ten minutes) and ultracentrifugation(200,000g, 60 minutes). LH2 complexes were preparedby diluting the membranes to an A of 50 cm21 at 850 nmwith 20 mM Tris–HCl (pH 8.0), and solubilized for 30minutes at room temperature with 0.7% LDAO. Sampleswere diluted to a final LDAO concentration of 0.15%,and ultracentrifuged to remove non-solubilized material(200,000g, 60 minutes). The solubilized fraction was thenloaded onto an anion-exchange chromatography column(Resource Q, 6 ml, Amersham BioSciences) and elutedwith a linear 0–500 mM NaCl gradient buffer containing0.1% LDAO, 20 mM Tris–HCl (pH 8.0). Final LH2 purifi-cation was achieved by gel-filtration (Superose 12, 25 mlcolumn, Amersham Biosciences) in a 0.1% LDAO,50 mM NaCl, 20 mM Tris–HCl (pH 8.0) buffer. Thepolypeptide composition of the final fraction was con-trolled by SDS-polyacrylamide gel electrophoresis, andCoomassie staining.

2D-crystallization

Purified LH2 complexes at 0.3 mg/ml in a buffer con-taining 0.1% LDAO, 200 mM NaCl, 10 mM Tris–HCl(pH 8.0) were mixed with DOPC at lipid-to-proteinratios between 0.45 and 0.55 (w/w). After 30 minutesstirring in the dark, 20 mM OTG was added. Again, thesample was allowed to equilibrate for 30 minutes in thedark. Detergent removal was performed through twosuccessive additions of 100 mg SM2 Bio-Beads/1 mlsample, for one hour each, at room temperature.24,34

Crystals were then withdrawn and kept at 4 8C for EMand AFM analysis.

Spectroscopic analysis of 2D-crystallization

Electronic absorption spectra were recorded using aUNICAM UV/Vis Spectrometer (UV2) at 200 nm/minute and 0.2 nm resolution. Crystallization conditionswere up-scaled to a total volume of 250 ml, so thatmeasurements during the crystallization process couldbe made in a 1 mm path-length cuvette. After measure-ment, the sample was transferred back to the crystalliza-tion vial.

Electron microscopy

Negative stain and cryo EM6 were performed using aPhilipps CM120 electron microscope operating at120 kV. LH2 2D crystals were adsorbed for one minuteon glow-discharged carbon-coated 300-mesh grids.Crystals were either negatively stained with 1% (w/v)uranyl-acetate, or quick-frozen in liquid ethane for cryoEM analysis. The latter were transferred into the electronmicroscope using a Gatan 626 cryo transfer system.Images were directly assessed on a Gatan slow scanCCD camera. The pixel sampling of the ssCCD camera

was calibrated by cryo EM analysis of purple membrane(unit cell dimension: a ¼ b ¼ 62.5 A, g ¼ 1208).35

Atomic force microscopy

Mica prepared as described36 was freshly cleavedbefore each experiment and used as support. To checkthe cleavage quality the mica was first imaged in,50 ml of an adsorption buffer containing 10 mM Tris–HCl (pH 7.3), 150 mM KCl, 50 mM MgCl2. Subsequently3 ml of protein crystal solution (0.3 mg/ml) was injectedinto the adsorption buffer drop on the mica surface.After 30 minutes the sample was rinsed with tenvolumes of a recording buffer containing 10 mM Tris–HCl (pH 7.3), 150 mM KCl. Imaging was performedwith a commercial Nanoscope III multimode AFM(from Digital Instruments, Santa Barbara, CA, USA)equipped with a 160 mm scanner (J-scanner) usingoxide-sharpened Si3N4 cantilevers with a length of100 mm (k ¼ 0.09 N/m; Olympus Ltd., Tokyo, Japan).The AFM was operated in contact mode applyingforces of ,100 pN. High magnification images wererecorded at scan frequencies of 3–4 Hz (1000–2000 nm/second).

Image processing

Cryo EM images of LH2 2D crystals recorded on theCCD camera were treated using the MRC image proces-sing package.35,37,38 Briefly, images were unbent runningtwo cycles of CCUNBEND, and the symmetry spacegroup was defined using ALLSPACE.26 The final mapwas calculated by merging six images using ORIGTILT.Projection maps were calculated using programs of theCCP4 package,39 and contoured in steps of 0.25 RMS.An isotropic temperature factor (B ¼ 2500) was appliedto compensate for the resolution-dependent degradationof image amplitudes.40

Average AFM topographs of 2D crystals of LH2complexes were calculated by cross-correlation and by asingle-particle averaging (lateral and rotational align-ment) using the SEMPER image processing system41,42

and the MRC image processing package.35,38

AFM topographs were calculated as tilted surfacerepresentations using Image SXM (Steve Barrett)†.

Analysis of the tilt, SD, CV and volume of theLH2 complexes

The membrane protrusion peak height of the LH2 sub-units was directly measured on the average topographyafter subunit peak searching. The tilt angle a of the LH2cylinder is directly obtained from the fit of the ninesubunit height peaks perpendicular to the cylinder axisof the averaged complex with a steepness of mx ¼ tan a.These subunit peak positions were equally used tomeasure the corresponding SD values on the simul-taneously calculated SD map. The CV was defined asCVðnÞ ¼ SDðnÞ=heightðnÞ; where n corresponds to peri-plasmic subunits 1–9. Subunit volumes in A3 weredirectly calculated from the x, y and z dimensions ofAFM average topography of the periplasmic surfacewith the SD value on the lipid bilayer as base heightcutoff, and height minima between subunits as lateralcutoffs. The topography ring diameter d is d ¼ 2

P9n¼1 �

† http://reg.ssci.liv.ac.uk/

578 AFM Study of Rb. sphaeroides LH2

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi{ðxn 2 XÞ2 þ ðyn 2 YÞ2 þ ðzn 2 ZÞ2};

qwhere n are sub-

units 1–9, xn, yn and zn are the coordinates of their peakpositions, and X, Y and Z are the coordinates of the ringcenter calculated by X ¼

P9n¼1 {xn=9} for X, Y, and Z.

Acknowledgements

We thank Drs S. A. Muller, V. Arluison, andA. Buguin for fruitful discussions, and M. Paris ofVeeco France for technical support. This studywas supported by the Institut Curie, the CEA andthe CNRS. S.S. was recipient of a fellowship fromthe French Research Ministry.

References

1. Binnig, G., Quate, C. F. & Gerber, C. (1986). Atomicforce microscope. Phys. Rev. Lett. 56, 930–933.

2. Schabert, F. A., Henn, C. & Engel, A. (1995). NativeEscherichia coli OmpF porin surfaces probed byatomic force microscopy. Science, 268, 92–94.

3. Engel, A., Gaub, H. & Muller, D. J. (1999). Atomicforce microscope: A forceful way with singlemolecules. Curr. Biol. 9, R133–R136.

4. Engel, A. & Muller, D. J. (2000). Observing single bio-molecules at work with the atomic force microscope.Nature Struct. Biol. 7, 715–718.

5. Scheuring, S., Fotiadis, D., Moller, C., Muller, S. A.,Engel, A. & Muller, D. J. (2001). Single proteinsobserved by atomic force microscopy. SingleMolecules, 2, 59–67.

6. Dubochet, J., Adrian, M., Chang, J.-J., Homo, J.-C.,Lepault, J., McDowall, A. W. & Schultz, P. (1988).Cryo-electron microscopy of vitrified specimens.Quart. Rev. Biophys. 21, 129–228.

7. Walz, T., Hirai, T., Murata, K., Heyman, J., Mitsuoka,K., Fujiyoshi, Y. et al. (1997). The three-dimensionalstructure of aquaporin-1. Nature, 387, 624–627.

8. Breyton, C., Haase, W., Rapoport, T., Kuhlbrandt, W.& Collinson, I. (2002). Three-dimensional structureof the bacterial protein-translocation complexSecYEG. Nature, 418, 662–665.

9. Muller, D. J. & Engel, A. (1997). The height of bio-molecules measured with the atomic force micro-scope depends on electrostatic interactions. Biophys.J. 73, 1633–1644.

10. Muller, D. J., Fotiadis, D., Scheuring, S., Muller, S. A.& Engel, A. (1999). Electrostatically balanced sub-nanometer imaging of biological specimens byatomic force microscopy. Biophys. J. 76, 1101–1111.

11. Scheuring, S., Freiss-Husson, F., Engel, A., Rigaud, J.& Ranck, J. (2001). High resolution topographs ofthe Rubrivivax gelatinosus light-harvesting complex 2.EMBO J. 20, 3029–3035.

12. Seelert, H., Poetsch, A., Dencher, N. A., Engel, A.,Stahlberg, H. & Muller, D. J. (2000). Proton poweredturbine of a plant motor. Nature, 405, 418–419.

13. Stahlberg, H., Muller, D. J., Suda, K., Fotiadis, D.,Engel, A., Matthey, U. et al. (2001). Bacterial ATPsynthase has an undacemeric rotor. EMBO Rep. 2,229–235.

14. Scheuring, S., Ringler, P., Borgina, M., Stahlberg, H.,Muller, D. J., Agre, P. & Engel, A. (1999). High reso-

lution topographs of the Escherichia coli waterchannelaquaporin Z. EMBO J. 18, 4981–4987.

15. Fotiadis, D., Hasler, L., Muller, D. J., Stahlberg, H.,Kistler, J. & Engel, A. (2000). Surface tongue-and-groove contours on lens MIP facilitate cell-to-celladherence. J. Mol. Biol. 300, 779–789.

16. Hu, X., Ritz, T., Damjanovic, A., Autenrieth, F. &Schulten, K. (2002). Photosynthetic apparatus ofpurple bacteria. Quart. Rev. Biophys. 35, 1–62.

17. Hu, X., Damjanovic, A., Ritz, T. & Schulten, K. (1998).Architecture and mechanism of the light-harvestingapparatus of purple bacteria. Proc. Natl Acad. Sci.USA, 95, 5935–5941.

18. Sundstrom, V., Pullerits, T. & van Grondelle, R.(1999). Photosynthetic light-harvesting: reconcilingdynamics and structure of purple bacterial LH2reveals function of photosynthetic unit. J. Phys.Chem. B, 103, 2327–2346.

19. McDermott, G., Prince, S. M., Freer, A. A.,Hawthornthwaite-Lawless, A. M., Papiz, M. Z.,Gogdell, R. J. & Isaacs, N. W. (1995). Crystal structureof an integral membrane light-harvesting complexfrom photosynthetic bacteria. Nature, 374, 517–521.

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

21. Walz, T., Jamieson, S. J., Bowers, C. M., Bullough,P. A. & Hunter, C. N. (1998). Projection structuresof three photosynthetic complexes from Rhodobactersphaeroides: LH2 at 6 A, LH1 and RC-LH1 at 25 A.J. Mol. Biol. 282, 833–845.

22. Chami, M., Pehau-Arnaudet, G., Lambert, O., Ranck,J., Levy, D. & Rigaud, J. (2001). Use of octyl beta-thio-glucopyranoside in two-dimensional crystallizationof membrane proteins. J. Struct. Biol. 133, 64–74.

23. Evans, M. B., Hawthornthwaite, A. & Cogdell, R. J.(1990). Isolation and characterisation of the differentB800–850 light-harvesting complexes from low-and high-light grown cells of Rhodopseudomonaspalustris, strain 2.1.6. Biochim. Biophy. Acta, 1016,71–76.

24. Rigaud, J., Levy, D., Mosser, G. & Lambert, O. (1998).Detergent removal by Bio-Beads. Eur. Biophys. J. 27,305–319.

25. Rigaud, J., Chami, M., Lambert, O., Levy, D. &Ranck, J. (2000). Use of detergents in two-dimen-sional crystallization of membrane proteins. Biochim.Biophys. Acta, 23, 112–128.

26. Valpuesta, J. M., Carrascosa, J. L. & Henderson, R.(1994). Analysis of electron microscope images andelectron diffraction patterns of thin crystals of ø29connectors in ice. J. Mol. Biol. 240, 281–287.

27. Tristram-Nagle, S., Petrache, H. I. & Nagle, J. F.(1998). Structure and interactions of fully hydrateddioleoylphosphatidylcholine bilayers. Biophys. J. 75,917–925.

28. Muller, D. J., Heymann, J. B., Oesterhelt, F., Moller,C., Gaub, H., Buldt, G. & Engel, A. (2000). Atomicforce microscopy of native purple membrane.Biochim. Biophys. Acta, 1460, 27–38.

29. Muller, D. J., Fotiadis, D. & Engel, A. (1998). Map-ping flexible protein domains at subnanometerresolution with the AFM. FEBS Letters, 430, 105–111.

30. Scheuring, S., Muller, D. J., Stahlberg, H., Engel,H.-A. & Engel, A. (2002). Sampling the confor-mational space of membrane protein surfaces withthe AFM. Eur. Biophys. J., 31, 172–178.

AFM Study of Rb. sphaeroides LH2 579

31. Fritzsche, W. & Henderson, E. (1996). Volumedetermination of human metaphase chromosomesby scanning force microscopy. Scan. Micr. 10,103–108.

32. Schneider, S. W., Larmer, J., Henderson, R. M. &Oberleithner, H. (1998). Molecular weights of indi-vidual proteins correlate with molecular volumesmeasured by atomic force microscopy. Pflugers Arch.435, 362–367.

33. Bose, S. K. (1963). Bacterial photosynthesis. InBacterial Photosynthesis (Gest, H., ed.), pp. 501–511,The Antioch Press, Yellow Springs, OH.

34. Rigaud, J. L., Mosser, G., Lacapere, J. J., Olofsson, A.,Levy, D. & Ranck, J. L. (1997). Bio-Beads: an efficientstrategy for 2D crystallization of membrane proteins.J. Struct. Biol. 118, 226–235.

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

36. Schabert, F. A. & Engel, A. (1994). Reproducibleacquisition of Escherichia coli porin surface topo-

graphs by atomic force microscopy. Biophys. J. 67,2394–2403.

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

38. Henderson, R., Baldwin, J. M., Downing, K. H.,Lepault, J. & Zemlin, F. (1986). Structure of purplemembrane from halobacterium halobium: recordingmeasurement and evaluation of electron micro-graphs at 3.5 A resolution. Ultramicroscopy, 19,147–178.

39. Collaborative Computational Project (1994). TheCCP4 Suite: programs for protein crystallography.Acta Crystallog. 4, 760–763.

40. Unger, V. M., Hargrave, P. A., Baldwin, J. M. &Schertler, G. F. X. (1997). Arrangment of rhodopsintransmembrane a-helices. Nature, 389, 203–206.

41. Saxton, W. O., Pitt, T. J. & Horner, M. (1979). Digitalimage processing: the semper system. Ultramicro-scopy, 4, 343–354.

42. Saxton, W. O. & Baumeister, W. (1982). The corre-lation averaging of a regularly arranged bacterialcell envelope protein. J. Microsc. 127, 127–138.

Edited by W. Baumeister

(Received 11 July 2002; received in revised form 22 October 2002; accepted 31 October 2002)

580 AFM Study of Rb. sphaeroides LH2