highly ordered two-dimensional crystals of photosystem i reaction center fromsynechococcussp.:...

J. Mol. Biol. (1996) 262, 336–348

Highly Ordered Two-dimensional Crystals ofPhotosystem I Reaction Center from Synechococcussp.: Functional and Structural Analyses

Simone Karrasch 1, Dieter Typke 2, Thomas Walz 1, Mette Miller 3

Georgios Tsiotis 1 and Andreas Engel 1*

The photosystem I reaction center complex from the thermophilic cyano-1M. E. Muller Institute forbacterium Synechococcus sp. was isolated by Triton X-100 solubilization andMicroscopic Structuralfractional precipitation with polyethylene glycol. As shown by gelBiology, Biozentrum of theelectrophoresis, the isolated complex was composed of the 83 kDa subunitsUniversity of Basel

Klingelbergstrasse 70, 4056 A and B, and at least six other subunits with molecular mass below 20 kDa.Basel, Switzerland Electron transfer from the primary electron donor P700 to the FA/FB centers

was demonstrated by flash-induced absorption change of the isolated2Max-Planck-Institut fur complex, while electron paramagnetic resonance (EPR) spectroscopyBiochemie, Am Klopferspitz showed that the complex contained a full set of Fe–S clusters. Isolated18, D-82152 Martinsried bei complexes were reconstituted into two-dimensional crystals in theMunchen, Germany presence of phospholipids and different cations. The crystals were found

to be active by flash-induced separation and EPR spectroscopy. Electron3Institute of Biochemistrymicroscopy and digital image processing of negatively stained andOdense Universityfrozen-hydrated specimens revealed orthorhombic crystals with unit cellCampusvej 55, 5230 Odensedimensions a = 138 Å, b = 145 Å and p121 symmetry. A three-dimensionalDenmarkmap was calculated for negatively stained crystals to 19 Å resolution,whereas the projection map of frozen-hydrated crystals exhibited 8 Åresolution.

7 1996 Academic Press Limited

Keywords: Photosystem I reaction center; two-dimensional crystals;*Corresponding author electron crystallography

Introduction

A fundamental process of life on earth is theconversion of the light energy absorbed bylight-harvesting complexes into chemical energy(Barber & Andersson, 1994). In higher plants andcyanobacteria water is split to liberate oxygen, aproton gradient is established, which drives theATP synthesis, and NADP+ is reduced. Theseprocesses are catalyzed by membrane-boundprotein complexes, the photosystem II (PS II), thecytochrome b6/f complex, and the photosystem I(PS I; Golbeck, 1992). The products, NADPH andATP, are used in the light-independent darkreactions to reduce CO2 to carbohydrates. Thereaction center of PS II is structurally related to the

reaction center of purple bacteria (Deisenhofer &Michel, 1989; Trebst, 1986), whereas PS I showssimilarities to the reaction center of green sulfurbacteria (Mathis, 1990; Nitschke et al., 1987).

The initial steps in the photosynthetic electrontransport in PS I have been analyzed extensively(Golbeck, 1992). Photodissociation of P700, achlorophyll dimer on the lumenal side, leads to thereduction of the primary electron acceptor A0, achlorophyll monomer. The electron then movesfurther down the redox potential gradient to A1, avitamin K derivative, to FX, a 4Fe–4S cluster, and tothe terminal acceptors, the FA/FB centers (Golbeck,1987). From these 4Fe–4S clusters the electronpasses to ferredoxin, the soluble carrier that finallyshuttles the electron to the NADP+ reductase.Reduction of the radical P700+ is achieved by thesoluble electron carriers plastocyanin or cyto-chrome c553.

The PS I from the thermophilic cyanobacteriumSynechococcus sp. consists of 11 subunits (Krausset al., 1993; Golbeck, 1994). Subunits PsaA and PsaB,

Abbreviations used: DMPC, dimyristoylphosphatidylcholine; OBTG, ocytl-b-D-thiogluco-pyranoside; 3D, 2D, three- and two-dimensional; EPR,electron paramagnetic resonance; LPR, lipid-to-proteinratio; Chl a, chlorophyll a.

0022–2836/96/380336–13 $18.00/0 7 1996 Academic Press Limited

Functional and Structural Analyses of PS I 337

Figure 1. a, Silver-stained SDS-polyacrylamide gel of solubilized (lane A) and crystallized (lane B) photosystem Ireaction center from Synechococcus sp. b, Absorption spectra of solubilized (1) and crystallized (2) PS I. c, Flash-inducedcharge separation in solubilized PS I. d, Flash-induced charge separation in crystallized PS I.

83 kDa each, form a heterodimer that houses theprimary electron donor P700, and the electronacceptors A0, A1 and Fx. The 4Fe–4S cluster Fx isformed between PsaA and PsaB (Krauss et al.,1993), each one providing two cysteine residues.Three subunits of molecular mass 8 to 16 kDa,PsaC, PsaD and PsaE, have been localized on thestromal side of the membrane (Ford et al., 1990;Krauss et al., 1993). The terminal electron acceptorsFA and FB are connected to PsaC via eight highlyconserved cysteine residues, while PsaD facilitatesthe binding of PsaC to the PsaA/PsaB heterodimer(Zhao et al., 1990). PsaE is thought to play a role inthe cyclic electron transport (Zhao et al., 1993). PsaFon the lumenal side of the photosynthetic mem-brane mediates the docking of plastocyanin orcytochrome c553 (Farah et al., 1995). Since psaLmutants exhibited normal charge separation, butdid not trimerize upon detergent solubilization,PsaL is thought to form the structure about the3-fold axis of the PS I trimers (see Golbeck, 1994).Four other membrane intrinsic subunits (I, J, K, M)

with molecular masses of 3 to 15 kDa have beenidentified and sequenced, but their functions arestill unknown (Golbeck, 1994). The PS I complexhas a total mass of 340 kDa including 60 to 100antenna chlorophylls (Krauss et al., 1993).

PS I complexes from thermophilic cyanobacteriahave been purified and crystallized (Witt et al.,1987). A recent report describes the 3D structure ofnegatively stained 2D crystals at 18 A resolution(Bottcher et al., 1992), whereas X-ray crystallogra-phy has provided a 6 A map of the PS I complexfrom the thermophilic cyanobacterium Synechococ-cus sp. (Krauss et al., 1993), and most recently a mapat 4.5 A (Schubert et al., 1995). In spite of thisprogress, the 2D crystallization of the PS I complexfrom the cyanobacterium Synechococcus sp. OD 24in the presence of lipids has been systematicallyimproved because the complex is reconstituted inits native environment, and precise information onits integration in the bilayer can be obtained. Theactivity of solubilized as well as crystallized PS Icomplexes was assessed by optical and EPR

Functional and Structural Analyses of PS I338

Figure 2. EPR spectra of solubilized (1) and crystallized(2) PS I after illumination at 200 K. Conditions: 6.3 mW,9.44 GHz.

fractional precipitation yielded a protein complexconsisting of about ten subunits. SDS-polyacryl-amide gel electrophoresis revealed a fuzzy band at60 kDa that is characteristic of the reaction centerproteins PsaA and PsaB kDa (Figure 1a, lane A). Atapproximately 16 to 18 kDa there are two strongand two weak protein bands, the lowest bandcorresponding to subunit PsaL (Kruip et al., 1993).One strong and three weak bands can be seenbelow 14 kDa. The composition of PS I afterreconstitution in a typical two-dimensional crystalis shown in Figure 1a, lane B. Although developedto a higher density, this SDS-PAGE gel exhibits asimilar pattern as that of the solubilized complex,with the characteristic band at 60 kDa representingPsaA and PsaB. There are three bands in the 16to 18 kDa range instead of four, with the bandrepresenting PsaL missing. In Figure 1b theabsorption spectra of solubilized (1) and crystal-lized (2) PS I are shown with absorption maxima at679 nm and 437 nm. The 437 nm Soret band and the679 nm band are typical of Chl a in PS I.

Absorption changes at 700 nm due to photodisso-ciation of the primary electron donor P700 (Kok,1956) allowed the oxidation of P700 to be monitoredduring excitation. As soon as the illuminationstopped, P700+ returned to the reduced state witha half-life time between 30 and 50 ms for both thesolubilized (Figure 1c) and the crystallized PS I(Figure 1d).

EPR spectroscopy was performed at 4 K (datanot shown) and 200 K to identify the compoundsinvolved in electron transport. Photoaccumulationat 200 K in the presence of sodium dithionite causes

spectroscopy. A 3D map of negatively stained 2Dcrystals revealed new structural features of thelumenal surface while low-dose electron mi-croscopy of frozen-hydrated samples yielded aprojection map at a resolution of 8 A.

Results

Biochemical and biophysical characterizationof photosystem I

The purification of photosystem I (PS I) from thethermophilic cyanobacterium Synechococcus sp. by

Table 1. PS I reconstitution conditionsBuffer pH Salt LPR Result

10 mM Hepes 7.0 50 mM MgCl2 0.25 Mosaic7.0 50 mM CaCl2 0.25 Densely packed (Figure 3a)7.0 100 mM NaCl, 10 mM MgCl2 0.25 Mosaic, partially ordered7.0 1 mM EDTA 1.00 Mosaic, partially ordered7.0 100 mM NaCl 1.00 Mosaic, partially ordered

10 mM Tris 8.0 25 mM ammonium ferric citrate 1.00 Mosaic, partially ordered8.0 100 mM LiCl 1.00 Densely packed8.0 100 mM NaCl, 10 mM MgCl2 1.00 Densely packed8.0 100 mM NH4Cl, 10 mM MgCl2 1.00 Empty or disordered8.0 25 to 50 mM NH4Cl 1.00 Partially ordered8.0 100 mM NH4Cl 1.00 Densely packed8.0 100 mM KCl 1.00 Densely packed

10 mM Mes 5.0 100 mM NaCl 1.00 Mosaic, partially ordered5.0 1 mM EDTA 1.00 Densely packed5.5 100 mM NaCl, 1 mM EDTA 1.00 Mosaic5.5 100 mM NaCl, 1 to 5 mM MgCl2 1.00 Mosaic6.0 100 mM NaCl, 1 to 10 mM MgCl2 1.00 Large vesicles, mosaic (Figure 3b)6.0 1 mM EDTA 1.00 Empty vesicles, few mosaic6.0 100 mM NaCl, 1 mM EDTA 1.00 Empty vesicles, few mosaic6.0 25 mM NH4Cl 1.00 Partially ordered6.0 50 to 100 mM NH4Cl 1.00 Densely packed6.0 100 mM NH4Cl, 10 mM MgCl2 1.00 Densely packed6.0 100 mM KCl 1.00 Densely packed6.5 100 mM NaCl, 1 to 10 mM MgCl2 1.00 Densely packed6.5 100 mM NaCl, 1 mM EDTA 1.00 Not ordered

10 mM Mes 5.0 10 to 50 mM ammonium ferric citrate 1.00 Empty or disordered vesicles6.0 10 mM ammonium ferric citrate 1.00 Vesicles, crystalline areas6.0 25 mM ammonium ferric citrate 0.5–1.0 Crystalline, well-ordered (Figure 3c)6.0 50 mM ammonium ferric citrate 1.00 Crystalline areas


Figure 3. Negatively stained vesicles obtained byreconstitution of PS I into 2D arrays using different bufferconditions. a, Reconstitution in CaCl2-containing Hepesbuffer (pH 7.0). b, Reconstitution in Mes buffer (pH 6.0),containing NaCl and MgCl2. c, Reconstitution in Mesbuffer (pH 6.0), containing ammonium ferric citrate. Scalebar represents 50 nm.

apparatus as described in Materials and Methods.The various dialysis buffers differed in saltcomposition and pH and are summarized inTable 1.

Reconstitution experiments carried out in 10 mMHepes buffer (pH 7.0), with different salt compo-sitions and lipid-to-protein ratios (LPR) yieldedsmall well-ordered, partially ordered or mosaiclattices with NaCl and/or MgCl2. In the presence ofCaCl2 the protein was incorporated and denselypacked, but not ordered (Figure 3a). When 10 mMTris buffer (pH 8.0) was used as dialysis buffer, theprotein was mainly reconstituted into disorderedvesicles. A few mosaic lattices were obtained withNH4Cl at concentrations of 25 and 50 mM, as wellas with 25 mM ammonium ferric citrate. Dialysisagainst buffer containing LiCl or KCl resulted in adisordered incorporation of PS I.

At low pH (5.0 to 6.5) reconstitution experimentswith different salt compositions yielded a variety ofvesicles with some partially ordered or mosaiclattices when the dialysis buffer contained 100 mMNaCl and MgCl2 in various amounts (1, 5 or 10 mM;Figure 3b), but a large fraction of empty vesicleswith 1 mM EDTA. Lower concentration (25 mM) ofNH4Cl resulted in vesicles with partially well-or-dered areas, whereas higher concentrations (50 and100 mM) of NH4Cl yielded vesicles that were notwell ordered. The protein was incorporated, but notordered when the dialysis buffer contained KCl.

Dialysis against 10 mM Mes buffer containingvarious concentrations of ammonium ferric citrateproduced disordered vesicles at pH 5.0. However,at pH 6.0 and various LPRs photosystem I wasmainly incorporated into well-ordered crystals. Thebest crystals were reproducibly obtained under thefollowing conditions: 10 mM Mes buffer (pH 6.0),25 mM ammonium ferric citrate, LPR 0.8 to 1.0.Figure 3c shows a typical crystal reconstitutedunder these conditions.

Electron microscopy and image processing ofnegatively stained two-dimensional crystals ofphotosystem I

Electron micrographs of negatively stained 2Dcrystals of the PS I reaction center exhibit a patternof diamond-shaped units arranged on an or-thorhombic lattice with unit cell dimensionsa = 138 A, b = 145 A (Figure 4a). Although thecrystals appear to be well ordered, a row ofprotrusions is sometimes missing (arrows inFigure 4a). The 7,2 diffraction order (arrowhead inFigure 4b) in the power spectrum corresponds to aresolution of 19 A. The weak spots of the oddreflections on the k axis suggest that the crystalpossesses a p121 symmetry, although differentialstaining is breaking the symmetry. This iscorroborated by the correlation average displayedin Figure 4c: one row of the units appears to be themirror image of the other row shifted by half a unitcell, but exhibiting less contrast.

Averages from 12 selected crystals shown in

reduction of all iron–sulfur clusters FX, FA and FB

(Kamlowski et al., 1995). Indeed, EPR spectra ofsolubilized (1) and crystallized (2) PS I recorded at200 K (Figure 2) show the signals of the primaryelectron donor P700 (g = 2.0025) the signals ofthe reduced iron–sulfur clusters FA− (gxx = 2.05,gyy = 1.95, gzz = 1.88), and FB− (gyy = 1.92), as well asthe signal of the reduced FX− cluster (gzz = 1.76; seeRutherford & Heathcote, 1985). These resultsdocument that all three iron–sulfur clusters FA, FB

and FX were present and active in the solubilized aswell as in the crystallized PS I complex.

Reconstitution of photosystem I intotwo-dimensional arrays

For reconstitution experiments solubilized PS I(0.5% OBTG) was mixed with various amounts ofthe phospholipid DMPC, which was solubilized inthe same OBTG-containing buffer. This mixturewas dialyzed against detergent-free buffer in atemperature-controlled continuous flow dialysis


Figure 4. Negatively stained 2D crystal (a) from PS I reaction center reconstituted in the presence of DMPC, 10 mMMes (pH 6.0), and 25 mM ammonium ferric citrate; its diffraction pattern (b); and the correlation average (c). Arrowsin a, mark a row where protrusions are missing. The arrowhead in b marks the 7,2 order at 19 A−1. Scale bars represent0.1 mm (a), 50 A−1 (b) and 75 A (c).

Figure 5 a to m all reveal these features. The crystalsexhibit rows of elongated monomers that arearranged in an alternating up and down orientationin the membrane. The main density is concentratedin a bean-like structure, which can be separatedfurther into two density peaks. Additional densityis found around these ‘‘beans’’ Crystals a to k havein common that they suffer from incompletestaining of one membrane surface, thus explainingdifferent morphologies of adjacent PS I reactioncenter complexes. In the case of crystals l and m thestain is quite evenly distributed on both sides of themembrane. Therefore, individual projections wereaveraged separately. The two best images ofunevenly stained crystals (Figure 5a and b) wereaveraged to give the motif in Figure 5n. Anotheraverage, Figure 5o, which was calculated fromcrystals a to k shows a slightly lower resolutionthan Figure 5n. The images of the evenly stainedcrystals l and m were averaged resulting in themotif displayed in Figure 5p, which exhibits anexcellent conservation of the p121 symmetry.

Automated electron microscopical tomography(Dierksen et al., 1993) was applied to collect tiltseries with up to 34 projections, over a tilt range of−60° to 45°. The electron dose deposited foracquiring a complete series was 60 electrons/A2.Individual reconstructions were calculated fromeach series to judge the quality of staining. The besttwo series were then combined to determine the

lattice lines by a least-squares fit (Figure 6), and tocalculate the 3D map shown in Figure 7. As twodifferent series were combined, the amplitudevalues exhibit some scatter, but the phase data areof good quality even for lattice line (4,−5), to betterthan 20 A resolution. Individual 7 A thick sectionsdisplayed in Figure 7a possess a quite differentstandard deviation (i.e. contrast), reflecting thelimited accessibility of the staining solution to thestructure within the bilayer (Figure 7b). On one sideof the membrane the PS I complex exhibits atrilobed domain protruding 21 to 28 A from thebilayer (marked by S in Figure 7c), which has beenidentified as stromal (Ford et al., 1990; Krauss et al.,1993; Kruip et al., 1993). The other side has a distinctindentation in the center of an elliptical protrusion(marked by L in Figure 7c). This is the site wherethe plastocyanin is supposed to dock. The twomembrane surfaces are distinctly different. Thesurface adsorbed to the carbon film is slightly betterstained, but it is deformed. Both the stromalextrinsic subunits and the lumenal surface arecompressed. The surface facing the vacuumappears to be better preserved, with the stromaldomain protruding by 7 to 14 A further out than theelliptical luminal structure, as estimated from thecontrast variation of 7 A thick sections. A distinctsmall mass within the indentation of the luminalside (marked with an arrowhead in Figure 7a) is afeature that has not been described. The surface


Figure 5. Correlation averages of different negatively stained PS I crystals, and global averages thereof. a to m,Micrographs selected by their optical diffraction pattern yielded averages that document stain variations. In n anaverage was calculated from the best unevenly stained images a and b. The average of images a to k is displayed ino. In p the average of the evenly stained images l and m is shown. The width of every image is 280 A.

rendered perspective view of the better preservedside is shown in Figure 7c, with the putativemembrane surfaces implemented in the sectionsthat possess the highest standard deviation.

Electron microscopy and image processing offrozen-hydrated two-dimensional crystals ofphotosystem I

As documented by Figure 8a, electron micro-graphs of frozen-hydrated 2D crystals of thephotosystem I reaction center yielded diffractionspots that extended frequently to a resolution ofbetter than 10 A. As in well-stained membranes, theodd reflections on the k axis are missing, indicatingthe screw axis along lattice vector b in the plane ofthe crystal. The strong 17,−2 spot (arrowhead inFigure 8a) corresponds to a resolution of 8 A. Theprojection map calculated from this image with nosymmetry imposed is displayed in Figure 8b.Phases were corrected for the phase-contrasttransfer function whereas amplitudes were notmodified. Even with no symmetry applied thescrew axis in the crystal plane is distinct Themonomers exhibit an elongated shape as is the casefor the negatively stained crystals. Although thetotal protein mass is contrasted in vitrified samples,the bean-like structure is still rather prominent.This major, sharp central density peak is flanked by

areas of low protein density. Figure 9 displays theprojection calculated from averaged amplitudesand phases extracted from 12 images with p121

symmetry. The map was calculated to 8 Aresolution, because the phase residual within theresolution interval 7 to 10 A was 70° (52° forIQ = 3 spots). One unit cell (a = 138.2(20.7) A;b = 144.9(22.1) A) is outlined as well as the screwaxes along b. The superimposed boundary of thecomplex (white contour) is taken from the 6 Adensity map of the 28 A thick membrane-embed-ded section (Krauss et al., 1993; and Figure 1a),whereas the black contour represents the boundaryof the stromal protrusion. In this average signifi-cant, but not yet interpretable, fine structure isvisible and the bean-like structure seen in thenegatively stained PS I crystals is distinct.

Discussion

We have assembled highly ordered 2D crystals ofan active PS I reaction center complex. Accordingto SDS-PAGE the composition of the PS I reactioncenter in the reconstituted 2D crystals is similar tothat in the solubilized protein except for subunitPsaL, which is not present in the crystals. Althoughthe isolation protocol requires trimeric complexesto be harvested, detergent exchange is a prerequi-site for reconstituting the 2D crystals described


Figure 6. Plots for selected lattice lines interpolated with the EM program system (Hegerl, 1992). Amplitudes arein arbitrary units, while the z* axis is in A−1.

here. Most likely, the trimers are disrupted by thistreatment into monomers that subsequently packinto the p121 crystals. This is consistent with the lossof subunit PsaL, which is involved in the formationof trimers (Golbeck, 1994). Absorption spectra ofcrystallized PS I and solubilized PS I are verysimilar, and reveal the absorption maxima at679 nm and 437 nm (Soret band) characteristic ofChl a. The activity measurements by flash-inducedcharge separation document that the primaryelectron donor P700 of both the solubilized and thecrystallized PS I reaction center is in a functionalstate. The half-life time for the back reaction fromthe reduced iron-sulfur clusters to P700+ of 30to 50 ms suggests the presence of FA and FB

sulfur–iron cluster: in the presence of the FX clusteralone, the half-life time would be 250 ms (Golbeck &Bryant, 1991). This finding is corroborated by theEPR measurements, which show that all threeclusters, FA, FB and FX, are present and in a

functional state in the solubilized as well as in thecrystallized PS I complex. We conclude that thereconstitution of the solubilized complex into a 2Dcrystal in the presence of lipids has no influence onits biological activity.

Reconstitution experiments carried out with PS Icomplexes in the presence of DMPC show thatduring removal of the detergent by dialysis,incorporation of the protein into sheets or vesiclesoccurs under most of the conditions tested. Whilethe presence of KCl, LiCl or CaCl2 resulted indensely packed membranes (Figure 3a), a strongpropensity for crystallization was observed whenNaCl and MgCl2 were used independently or incombination, leading to the formation of mosaiclattices (Figure 3b). Our goal was therefore to finda salt that would promote a slower crystallization.By far the best crystals, concerning lattice order,size of homogeneous arrays, flatness and repro-ducibility, were obtained when the dialysis buffer


Figure 7. 3D map from negatively stained PS I crystals. a, Horizontal sections spaced by 7 A. The two sections withthe largest standard deviation are marked with an asterix. The most prominent lumenal protrusion (arrowhead) couldbe PsaF, whereas the fine bridge in the lumenal cavity (arrow) cannot be interpreted. b, The standard deviation profileof horizontal sections indicates the boundaries of the lipid bilayer. c, A perspective view has been calculated from thesections shown in a with a marching cube algorithm (Henn et al., 1996) after implementing the membrane as a planarsurface in the two sections exhibiting the highest standard deviation. The stromal and lumenal sides are marked withS and L, respectively. The horizontal sections in a and the perspective view in c have a width of 210 A.

contained 25 to 50 mM ammonium ferric citrate atpH 6.0 (Figure 3c; Table 1).

Electron microscopy and image processing ofsuch negatively stained crystals revealed rows ofmonomers that exhibited an elongated shape, withthe main density concentrated in a bean-likestructure (Figure 4), similar to those reportedby Ford et al. (1990) and Bottcher et al. (1992).Stain-filled regions that separate the monomersappear to mark the cross-section of the lipid moiety.However, different data have been presentedconcerning the vertical position of the complexwithin the bilayer. Ford et al. (1990) published amodel based on surface relief reconstructions frommetal-shadowed freeze-dried samples that suggestsa stromal protrusion of 25 A, and a lumenalprotrusion of 10 A with a central pit. Accordingly,the PS I complex would have a height of 75 A alongthe normal to the membrane plane, assuming thatthe bilayer is 40 A thick. Bottcher et al. (1992) havecalculated a 3D reconstruction from tilted nega-tively stained images. In their map the stromalprotrusion extends by 36 A above the bilayer of50 A thickness, while the lumenal side with itsprominent pit protrudes by 8 A, thus yielding atotal height of 94 A. This height value agrees withthe total height from the 6 A X-ray structure, whichsuggests a stromal protrusion of 35 A, a bilayer of40 A and a lumenal protrusion of 15 A (Krausset al., 1993). The total height of the complex and theheight of the stromal protrusion have also been

estimated from side views of trimer stacks as 92 Aand 25 to 33 A, respectively (Kruip et al., 1993).

The problem with the interpretation of theprevious 3D map of negatively stained 2D crystals(Bottcher et al., 1992) is the shape of the moleculewithin the putative bilayer boundaries, whichimplies an asymmetric disposition of the apposedlipid leaflets. The vertical position of the PS Icomplex within the bilayer can be estimated fromthe horizontal sections presented in Figure 7a.Their standard deviation (SD), a measure of theircontrast, exhibits two maxima separated by 45 A(Figure 7b) that are related to the border of thestain-impermeable bilayer. The overall thickness ofthe structure is approximately 80 A, taking twicethe SD of the background as threshold in Figure 7b.This corresponds to a 20% smaller extension of thePS I complex normal to the membrane plane thanthat determined by X-ray crystallography (Krausset al., 1993), and is probably the result of surfacetension and staining artifacts. Another artifactarises from the interaction of the crystal surfacewith the hydrophilic carbon film. Stromal protru-sions are compressed, or even displaced, assuggested by the missing rows of protrusionsoccasionally observed (arrows in Figure 4a). Ac-cording to the 7 A sections displayed in Figure 7athe stromal protrusion extends between 21 and28 A out from the membrane surface, whereas thelumenal protrusion has a height above the bilayerof approximately 14 to 21 A, values that need to be


Figure 8. Frozen-hydrated 2D crystal of PS I reaction center. a, Fourier components of a highly ordered area afterunbending. Zero values of the fitted phase-contrast transfer function are indicated by ellipses. Odd 0,k spots aremissing. The 17, −2 order at 8 A−1 is marked with an arrowhead. b, The projection map was calculated to 8 A resolutionusing the Fourier components from a with phases and amplitudes corrected for the change in the phase contrast transferfunction. One unit cell (a = 138.5 A, b = 147.0 A) is outlined with the a axis horizontal and the b axis vertical. Contoursare in steps of 0.25 × r.m.s. (root-mean-square) density.

corrected by 20%. Although prone to staining andsurface tension artifacts, our data suggest that thecomplex is more symmetrically incorporated in thelipid bilayer than proposed by Bottcher et al. (1992).

A major difficulty in the 3D structural analysis ofthe p121 PS I crystals presented here was theirlow crystallographic symmetry. The possibility ofuneven staining demonstrated in Figures 4 and 5,and also discussed by Ford et al. (1990) and Bottcheret al. (1992), prevented the enforcement of the p121

symmetry to calculate 3D maps. This required theacquisition of tilt series containing many projec-tions. As beam damage induces a further shrinkageof the structure (Berriman, 1986), the application ofan automated electron tomography system was asignificant advantage (Dierksen et al., 1993). Up to30 projections could easily be recorded at a totaldose of 60 electrons/A2. From such series clear 3Dmaps could be calculated and compared before thedata were merged. A wealth of structural detailscan be recognized on both sides of the complex.The stromal protrusion exhibits two well-resolveddomains at the top, approximately 30 A above thebilayer, whereas a third domain is seen closer to themembrane surface. The indentation on the left ofthe stromal protrusion close to the cleft (Figure 7a,top row) is the site where flavodoxin (Muhlenhoffet al., 1996) or ferredoxin (Lelong et al., 1996) bind,as has been predicted by modeling (Fromme et al.,1994). A complex disposition of domains lining apronounced indentation is resolved on the lumenalside. These domains protrude approximately 15 A

out of the membrane, while the indentation exhibitsa depth of about 5 to 10 A with respect to themembrane surface. As proposed by modeling theindentation is the site of plastocyanin docking(Fromme et al., 1994). This enables the plastocyaninto dock very close to the P700+, facilitating anefficient reduction of the radical. The mostprominent domain (arrowhead in Figure 7a, topright) could be PsaF, the protein involved inplastocyanin docking (Farah et al., 1995). Further, asmall bridge is recognized within the indentation(arrow, Figure 7a, top right). This is a significant,yet uninterpretable, feature. Although distinct onboth sides of the membrane, the bridge is severelydistorted on the side facing the carbon film. Nopublished information is available that would allowthe fine structure of the lumenal PS I complexsurface to be validated. However, the docking sitefor the electron donors plastocyanin and cyto-chrome c553 is likely to possess a complex structurethat optimizes the flow of electrons to P700+.

There is a distinct difference in the lengthvariability of the two axes of the unit cell. Thestandard deviation for the longer axis wasdetermined to be twice as high as the standarddeviation for the shorter axis (Bottcher et al., 1992).We observed the same phenomenon, even morepronounced, for the unit cell size of the frozen-hydrated crystals. The standard deviation forthe length of the b axis was three times thestandard deviation of the length of the a axis(a = 138.2(20.7) A; b = 144.9(22.1) A; average from


Figure 9. A projection map with p121 symmetry calcul-ated from amplitudes and phases averaged from 12images of frozen-hydrated PS I crystals. One unit cell(a = 138.2 A, b = 144.9 A) is outlined with the a axishorizontal and the b axis vertical, and the screw axesalong b are indicated. White arrowheads mark the twoprotein–protein contacts that define the 2D crystal. Theoutermost contour of the membrane resident 28 A thicksection of the 6 A map from X-ray analyses (Krauss et al.,1993) is superimposed on the central complex in white.Major discrepancies are observed at the domain thatforms the interface at the 3-fold axes of crystallizedtrimers (asterix). The black contour marks the boundaryof the stromal protrusion.

helices and helices parallel to the membrane plane(Krauss et al., 1993), prevents such an interpret-ation.

The results presented here demonstrate thathighly ordered 2D crystals of the PS I complex canbe grown in the presence of phospholipids. The 3Dmap from negatively stained crystals indicates thatthe complex is rather symmetrically incorporated inthe bilayer with stromal and lumenal protrusions ofapproximately 30 A and 15 A height, respectively.The projection map at 8 A resolution of frozenhydrated crystals is in excellent agreement withdata from X-ray crystallography, suggesting thatelectron microscopy may provide useful infor-mation on the disposition of the extrinsic stromalsubunits, as well as the interaction of the reactioncenter with the lipid bilayer.

Materials and Methods

Chemicals

Dimyristoyl phosphatidylcholine (DMPC) was pur-chased from the Sigma Chemical Co. (St Louis, MO).Octyl-b-D-thioglucopyranoside (OBTG) was from Cal-biochem Co. (La Jolla, CA) and Triton X-100 was fromFluka Chemie AG (Buchs, Switzerland). All otherchemicals used for membrane preparation, purificationand reconstitution of photosystem I were from eitherFluka Chemie AG (Buchs, Switzerland) or E. Merck(Darmstadt, Germany).

Isolation of photosystem I

Crude and PS I-enriched membranes were prepared asdescribed by Schatz & Witt (1984). Enriched membraneswere resuspended at 2 mg Chl a/ml in a buffer con-taining 10 mM Hepes, 5 mM K2HPO4, 10 mM MgCl2 and25% glycerol (pH 7.5). Triton X-100 was added slowly toa concentration of 2.22% (w/v) from a 20% (w/v) stocksolution. This mixture was incubated for five minutes at37°C. After centrifugation for two minutes at 14,000 g thesupernatant was subjected to fractional precipitation withpolyethylene glycol (PEG) 6000 at concentrations of 1 to10% in the presence of 70 mM MgCl2. The precipitatedmaterial was pelleted for 20 minutes at 22,000 g. Selectedfractions (using SDS-PAGE and absorption spectra) wereresuspended in the same buffer and transferred to aSepharose-CL6B column (70 cm long, 3.0 cm diameter)for elution at 2 ml/h with 10 mM Hepes (pH 7.5), 5 mMK2HPO4, 150 mM NaCl, and 0.03% Triton X-100. Thegreen fractions were pooled and concentrated byprecipitation with PEG 6000 and MgCl2 according to themethod of Ford et al. (1990).

Determination of protein, chlorophyll andP700 concentration

The protein concentration was measured using theBCA protein assay of Pierce (Rockford, IL) according tothe method of Schatz & Witt (1984). Acetone extractionof the membranes and measurement of the absorbance at663 nm was performed to determine the chlorophyllconcentration (Arnon, 1949). The primary electron donor

12 crystals). Nevertheless, frozen-hydrated crystalsexhibited an excellent order yielding diffractionspots to a resolution of 8 A.

The outermost contour of the 28 A thick central,bilayer-resident section of the complex according tothe 6 A X-ray data (Krauss et al., 1993) drawn in the8 A projection map (Figure 9) documents anexcellent agreement of the two data sets. It impliesthat the lipid moiety is restricted to about 20% ofthe crystal surface. The major discrepancy is closeto the 3-fold center of the PS I trimer (marked withan asterix in Figure 9), where a conformationalchange during trimerization could occur. Thisobservation is in agreement with the loss of subunitPsaL concomitant with the detergent exchangerequired for 2D crystallization (see Figure 1). Twotypes of protein contacts (marked with arrowheadsin Figure 9) appear to drive the assembly of the 2Dcrystals. Interestingly, the large variation of the blattice vector could be related to a flexibility of thedomain at the 3-fold axis of the trimer. At aresolution of 8 A single transmembrane a-helicesare expected to be identified. However, the largenumber a-helices (130), including long, kinked


P700 was assessed photochemically according to themethod of Lockau (1979).

SDS-PAGE

For SDS-polyacrylamide gel electrophoresis thesamples were incubated in sample buffer containing0.76% SDS and 1.3% b-mercaptoethanol for one hourat room temperature. They were run on 10% to 23%polyacrylamide gels (Laemmli, 1970) and silver stained.

Flash-induced charge separation

Flash-induced charge separation was measured in ahome-built arrangement in the laboratory of Dr Mantele,Universitat Freiburg, Germany (Bauscher et al., 1990). Amonochromator (model H25, Fa. I. S. A. Jobin Yvon)selected the wavelength of the measuring beam(tungsten–halogen lamp). To avoid sample pre-excitationa shutter was operated between monochromator andsample. Absorption changes were monitored with asilicon photodiode covered by an interference filter(707 nm). A xenon flash lamp excited the sample; spectralregions complementary to sample excitation wereeliminated with a filter. The signal was transferred to acomputer and analyzed using MSPEK software.

EPR spectroscopy

EPR spectra were recorded with a Bruker 300-X bandspectrometer fitted with an Oxford Instruments cryostatand temperature control system. The EPR cavity wasilluminated with a 200 W tungsten projector. The whitelight was filtered through 2 cm of water to removeinfrared radiation. Measurements were performed at9.44 GHz microwave excitation and a temperature of 4 Kor 200 K.

Reconstitution

For the reconstitution experiments the detergent TritonX-100 used for purification of the PS I reaction centerswas exchanged with 0.5% OBTG by PEG 6000/MgCl2

precipitation (Ford et al., 1990) followed by resuspensionin the new detergent-containing buffer (10 mM Hepes,pH 7.0). The precipitation/resuspension procedure wasrepeated three times. Thereafter the PS I protein wasmixed with various amounts of DMPC solubilized in thesame OBTG-containing buffer to achieve lipid-to-proteinratios (LPR; w/w) between 0.5 and 1.0. The final proteinconcentration was adjusted to 1 mg/ml in all exper-iments. The reconstitution mixture (250 ml) was dialyzedagainst detergent-free buffer in a temperature-controlledcontinuous flow dialysis apparatus (Jap et al., 1992).Various dialysis buffers differing in salt composition andpH were explored. The dialysis cell temperature wasadjusted to 26°C for the first 24 hours and increasedlinearly to 37°C during the next 12 hours. After 24 hoursthe temperature was decreased to 26°C within ten hours,resulting in a total dialysis time of 70 hours.

Electron microscopy and image processing ofnegatively stained specimens

Reconstituted membranes were adsorbed to glow-discharged carbon-coated parlodion films that weremounted on electron microscope grids. The grids were

washed on three drops of distilled water, stained with0.75% uranyl formate (pH 4.25), and blotted with filterpaper. Specimens were examined in a Hitachi H-7000transmission electron microscope operated at 100 kVand recorded on Kodak SO-163 plates at a nominalmagnification of 50,000 × . Alternatively, a Philips CM200FEG equipped with a CCD camera and an on-linecomputer control system (Tietz GmbH, Munich) pro-grammed for automated serial recordings (Dierksen et al.,1993) was used to collect low-dose tilt series at a nominalmagnification of 38,000 × . Tilt series that were composedof 27 to 34 projections each were recorded over a tiltrange of − 60° to 45°. The electron dose deposited foracquiring a complete series was between 60 and 150electrons/A2. Image processing was carried out asdescribed by Walz et al. (1994) using the Semper imageprocessing system (Saxton et al., 1979). To interpolate thelattice line data, a linear interpolation, a sin(x)/xinterpolation (Smith, 1981), or a constrained least-squaresfitting algorithm (Shaw, 1984) implemented in the EMprogram system (Hegerl, 1992), was used. The interp-olation algorithm was found to have no significantinfluence on the reconstructed sections. The 3D densitydistribution has been isocontoured with a marching cubealgorithm to calculate perspective views (Henn et al.,1996).

Electron microscopy and image processing offrozen-hydrated specimens

The samples for electron cryo-microscopy wereadsorbed for two minutes onto carbon-coated grids thathad been glow-discharged in amyl amine vapor. Afterblotting for 30 seconds at 4°C and high humidity thegrids were plunged into liquid ethane at −180°C(Subramaniam et al., 1993). Grids were transferred to aGatan 626 cold stage and examined with a Philips CM12microscope in the laboratory of Dr R. Henderson,Cambridge, UK. Low dose images of nominally untiltedcrystals were recorded at 120 kV at a magnification of35,000 × . The exposure of 1.0 second and the selectedillumination conditions resulted in an electron dose of 5to 10 electrons/A2. Images were selected for processingby optical diffraction. Selected areas were digitized into2000 × 2000 pixel areas in steps of 10 mm using amodified Joyce Loebl Mk 4 densitometer. Processing ofimages of frozen-hydrated specimens and merging ofimage amplitudes and phases followed proceduresdescribed by Henderson et al. (1986) and Havelka et al.(1993).

AcknowledgementsThe authors thank Dr Richard Henderson at the MRC

in Cambridge for providing facilities to accomplish theelectron cryo-microscopy, and for his continuoussupport. Dr Mantele at the University of Freiburgperformed the measurements of flash induced chargeseparation and Dr Riedel at the University of Regensburgcarried out the EPR spectroscopy experiments. We areindebted to Dr R. Hegerl at the Max-Planck-Institute forBiochemistry in Martinsried for the interpolation of thelattice lines, and to D. Fotiadis for his help with the gelsdisplayed in Figure 1.


References

Arnon, D. I. (1949). Copper enzymes in isolatedchloroplasts. Polyphenoloxidase in Beta vulgaris.Plant Physiol. 24, 1–15.

Barber, J. & Andersson, B. (1994). Revealing the blueprintof photosynthesis. Nature, 370, 31–34.

Bauscher, M. E. N., Bagley, K., Breton, J. & Mantele, W.(1990). Investigation of models for photosyntheticelectron acceptors. Infrared spectroelectrochemistryof ubiquinone and its anions. FEBS Letters, 261,191–195.

Berriman, J. L. K. R. (1986). Methods for specimenthickness determination in electron microscopy. II.Changes in thickness with dose. Ultramicroscopy, 19,349–366.

Bottcher, B., Graber, P. & Boekema, E. J. (1992). Thestructure of photosystem I from the thermopiliccyanobacterium Synechococcus sp. determined byelectron microscopy of two-dimensional crystals.Biochim. Biophys. Acta, 1100, 125–136.

Deisenhofer, J. & Michel, H. (1989). The photosyntheticreaction centre from the purple bacterium Rhodopseu-domonas viridis. EMBO J. 8, 2149–2170.

Dierksen, K., Typke, D., Hegerl, R. & Baumeister, W.(1993). Towards automatic electron tomography II.Implementation of autofocus and low-dose pro-cedures. Ultramicorscopy, 49, 109–120.

Farah, J., Rappaport, F., Choquet, Y., Joliot, P. & Rochaix,J. D. (1995). Isolation of a psaF-deficient mutant ofChlamydomonas reinhardtii: efficient interaction ofplastocyanin with the photosystem I reaction centeris mediated by the PsaF subunit. EMBO J. 14,4976–4984.

Ford, R. C., Hefti, A. & Engel, A. (1990). Ordered arraysof the photosystem I reaction centre after reconstitu-tion: projections and surface reliefs of the complex at2 nm resolution. EMBO J. 9, 3067–3075.

Fromme, P., Schubert, W. D. & Krauss, N. (1994).Structure of photpsystem I: suggetions on thedocking sites for plastocyanin, ferredoxin and thecoordination of P700. Biochim. Biophys. Acta, 1187,99–105.

Golbeck, J. H. (1987). Structure, function and organiz-ation of the Photosystem I reaction center complex.Biochim. Biophys. Acta, 895, 167–204.

Golbeck, J. H. (1992). Structure and function ofphotosystem I. Annu. Rev. Plant Mol. Biol. 43,293–324.

Golbeck, J. H. (1994). Photosystem I in Cyanobacteria,Kluwer, Dordrecht.

Golbeck, J. H. & Bryant, D. A. (1991). Photosystem I,Academic Press, New York.

Havelka, W. A., Henderson, R., Heymann, J. A. W. &Oesterhelt, D. (1993). Projection structure ofhalorhodopsin from Halobacterium halobium at 6 Aresolution by electron cryo-microscopy. J. Mol. Biol.234, 837–846.

Hegerl, R. (1992). A brief survey of software packages forimage processing in biological electron microscopy.Ultramicroscopy, 46, 417–423.

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

Henn, C., Teschner, M., Engel, A. & Aebi, U. (1996).Real-time isocontouring and texture mapping meet

new challenges in interactive molecular graphicsapplication. J. Struct. Biol. 116, 86–92.

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.

Kamlowski, A., van der Est, A., Fromme, P. & Stehlik, D.(1995). Structural organization of the acceptors A1,Fx, Fa and Fb in photosystem I from EPR in solutionand single crystals. Photosynthesis: from Light toBiosphere. pp. 29–34, Kluwer, Dordrecht, Boston andLondon.

Kok, B. (1956). On the reversible absorption change at705 nm in photosynthetic organisms. Biochim. Bio-phys. Acta, 22, 399–401.

Krauss, N., Hinrichs, W., Witt, I., Fromme, P., Pritzkow,W., Dauter, Z., Betzel, C., Wilson, K. S., Witt, H. T.& Saenger, W. (1993). 3-Dimensional structure ofsystem-I of photosynthesis at 6 angstrom resolution.Nature, 361, 326–331.

Kruip, J., Boekema, E. J., Bald, D., Boonstra, A. F. &Rogner, M. (1993). Isolation and structural character-ization of monomeric and trimeric photosystem-Icomplexes (P700 center dot F(A)/F(B) and P700center dot F(X)) from the Cyanobacterium synechocys-tis PCC-6803. J. Biol. Chem. 268, 23353–23360.

Laemmli, U. K. (1970). Cleavage of structural proteinsduring the assembly of the head of bacteriophage T4.Nature, 227, 680–685.

Lelong, C., Boekema, E. J., Kruip, J., Bottin, H., Rogner,M. & Setif, P. (1996). Characterization of a redoxactive cross-linked complex between cyanobacterialphotosystem I and soluble ferredoxin. EMBO J. 15,2160–2168.

Lockau, W. (1979). The inhibition of photosyntheticelectron transport in spinach chloroplasts by lowosmolarity. Eur. J. Biochem. 94, 365–373.

Mathis, P. (1990). Compared structure of plant andbacterial photosynthetic reaction centers. Evolution-ary implications. Biochim. Biophys. Acta, 1018,163–167.

Muhlenhoff, U., Kruip, J., Bryant, D. A., Rogner, M.,Setif, P. & Boekema, E. (1996). Characterization of aredox active cross-linked complex betweencyanobacterial photosystem I and its physiologicalacceptor flavodoxin. EMBO J. 15, 488–497.

Nitschke, W., Feiler, U., Lockau, W. & Hauska, G. (1987).The photosystem of the green sulphur bacteriumChlorobium limicola contains two early electronacceptors similar to photosystem I. FEBS Letters, 218,283–286.

Rutherford, A. E. & Heathcote, P. (1985). Primaryphotochemistry in photosystem I. Photosyn. Res. 6,295–316.

Saxton, W. O., Pitt, T. J. & Horner, M. (1979). Digitalimage processing: The semper system. Ultrami-croscopy, 4, 343–354.

Schatz, G. H. & Witt, H. T. (1984). Extraction andcharacterization of oxygen-evolving photosystem IIcomplexes from a thermophilic cyanobacteriumSynechococcus sp. Photobiochem. Photobiophys. 7, 1–14.

Schubert, W. D., Klukas, Krauss, N., Saenger, W.,Fromme, P. & Witt, H. T. (1995). Present state of thecrystal structure of photosystem I at 4.5 A resolution.Photosynthesis: from Light to Biosphere. pp. 3–10,Kluwer, Dordrecht, Boston and London.

Shaw, P. J. (1984). Constrained least-squares fitting of thelattice lines in three-dimensional reconstruction ofmonolayer crystals. Ultramicrocsopy, 14, 363–366.


Smith, P. R. (1981). The interpolation of Fourier transformdata in tilted-view three-dimensional reconstruction.Ultramicroscopy, 7, 155–160.

Subramaniam, S., Gerstein, M., Oesterhelt, D. &Henderson, R. (1993). Electron diffraction, analysisof structural changes in the photocycle of bacteri-orhodopsin. EMBO J. 12, 1–8.

Trebst, A. (1986). The topology of the plastoquinone andherbicide binding peptides of photosystem II in thethylakoid membrane. Z. Naturforsch. 41c, 240–245.

Walz, T., Smith, B. L., Agre, P. & Engel, A. (1994). Thethree-dimensional structure of human erythrocyteaquaporin CHIP. EMBO J. 13, 2985–2993.

Witt, I., Witt, H. T., Gerken, S., Saenger, W., Dekker, J. P.& Rogner, M. (1987). Crystallization of reaction

center I of photosynthesis. Low-concentration crys-tallization of photoactive protein complexes from theCyanobacterium synechoccus sp. FEBS Letters, 221,260–264.

Zhao, J. D., Warren, P. V., Li, N., Bryant, D. A. & Golbeck,J. H. (1990). Reconstitution of elec transport inphotosystem I with PsaC and PsaD proteinsexpressed in Escherichia coli. FEBS Letters, 276,180.

Zhao, J. D., Snyder, W. B., Muhlenhoff, U., Rhiel, E.,Warren, P. V., Golbeck, J. H. & Bryant, D. A. (1993).Cloning and characterization of the psaE gene of thecyanobacterium Synechococcus sp.: characterizationof a psaE mutant and overproduction of the proteinin E. coli. Mol. Microbiol. 9, 183–194.

Edited by A. Klug

(Received 12 April 1996; received in revised form 21 June 1996; accepted 8 July 1996)

highly ordered two-dimensional crystals of photosystem i reaction center fromsynechococcussp.:...

Documents