phdthesis scheuring screen

Genehmigt von der Philosophisch-Naturwissenschaftlichen Fakultät

Auf Antrag von

Prof. Dr. Andreas Engel & Prof. Dr. Jean-Louis Rigaud

Basel, den 19. Dezember 2000

Prof. Dr. Andreas D. Zuberbühler

Dekan der PhilosophischNaturwissenschaftlichen Fakultät

4

To my parents,my brothers,and my friends.

5

Index

7

1. AFM and EM in structural biology........................................................................15

1.1. General introduction..................................................................................................151.1.1. The driving force to do science...........................................................................151.1.2. Membrane proteins............................................................................................151.1.3. 2D crystals allow the acquisition of structural information on membrane proteins

in a native-like environment................................................................................161.1.4. Atomic force and electron microscopy cover large resolution ranges and together

provide both surface and volume information.....................................................171.1.5. Results and Perspectives....................................................................................181.1.6. References..........................................................................................................19

1.2. Atomic force microscopy: A powerful tool to observe the assembly and functionof native proteins........................................................................................................211.2.1. Abstract..............................................................................................................211.2.2. Introduction........................................................................................................211.2.3. Conditions for single molecule imaging.............................................................221.2.4. Imaging the ion-driven rotor of the ATP synthase..............................................231.2.5. Conformational flexibility of proteins................................................................251.2.6. The tongue-and-groove interaction of MIP tetramers.........................................261.2.7. Imaging the subcomplexes of the GroE chaperonin system: GroEL and GroES271.2.8. Observing the assembly of membrane proteins..................................................281.2.9. Outlook..............................................................................................................291.2.10.Acknowledgement..............................................................................................301.2.11.References..........................................................................................................30

1.3. Imaging streptavidin 2D-crystals on biotinylated lipid monolayers at highresolution with the atomic force microscope...........................................................351.3.1. Summary............................................................................................................351.3.2. Introduction........................................................................................................351.3.3. Materials and Methods.......................................................................................36

1.3.3.1. Materials................................................................................................36

1.3.3.2. Hydrophobicity measurement......................................................................36

1.3.3.3. Crystallization of streptavidin on biotin-lipid monolayer..................................36

1.3.3.4. Atomic force microscopy (AFM).................................................................37

1.3.3.5. Transmission electron microscopy (TEM).....................................................37

1.3.4. Results...............................................................................................................371.3.4.1. Hydrophobicity and topography of HOPG.....................................................37

1.3.4.2. Crystallization of streptavidin on biotin-lipid monolayer..................................38

1.3.4.3. AFM of streptavidin crystals......................................................................39

1.3.4.4. TEM of streptavidin crystals.......................................................................40

1.3.5. Discussion.........................................................................................................401.3.6. Acknowledgment................................................................................................421.3.7. References..........................................................................................................43

9

2. Application of high resolution AFM.......................................................................47

2.1. High resolution AFM topographs of the Escherichia coli waterchannelaquaporin Z................................................................................................................472.1.1. Abstract..............................................................................................................472.1.2. Introduction.......................................................................................................472.1.3. Results...............................................................................................................492.1.4. Discussion.........................................................................................................502.1.5. Materials and methods.......................................................................................54

2.1.5.1. Reconstitution..........................................................................................54

2.1.5.2. Trypsin digestion......................................................................................54

2.1.5.3. Atomic force microscopy............................................................................54

2.1.5.4. Image processing......................................................................................55

2.1.6. Acknowledgment...............................................................................................552.1.7. References.........................................................................................................55

2.2. High resolution AFM topographs of Rubrivivax gelatinosus light-harvestingcomplex LH2...............................................................................................................592.2.1. Abstract..............................................................................................................592.2.2. Introduction.......................................................................................................592.2.3. Results...............................................................................................................612.2.4. Discussion.........................................................................................................652.2.5. Materials and methods.......................................................................................67

2.2.5.1. Materials.................................................................................................67

2.2.5.2. Isolation, purification and proteolysis of LH2 complex....................................67

2.2.5.3. Biochemical and biophysical techniques.........................................................67

2.2.5.4. Reconstitution and 2D crystallization............................................................67


2.2.5.6. Image processing......................................................................................68


3. Combining surface and projection techniques....................................................73

3.1. The aquaporin sidedness revisited...........................................................................733.1.1. Summary...........................................................................................................733.1.2. Introduction.......................................................................................................733.1.3. Results...............................................................................................................743.1.4. Discussion.........................................................................................................783.1.5. Materials and Methods......................................................................................80

3.1.5.1. 2D crystallization......................................................................................80

3.1.5.2. Trypsin digestion......................................................................................80


3.1.5.4. Freeze-drying & metal-shadowing.................................................................80

3.1.5.5. Cryo electron microscopy...........................................................................80


10

4. Structural studies of a membrane transporter...................................................87

4.1. The functional Escherichia coli lactose permease LacY/Cytb562/6His formstrimers: A 2.8 nm 3D reconstruction and preliminary electron crystallographicdata...............................................................................................................................874.1.1. Summary............................................................................................................874.1.2. Introduction........................................................................................................874.1.3. Results and discussion.......................................................................................89

4.1.3.1. Protein purification...................................................................................89

4.1.3.2. Single particle analysis..............................................................................90

4.1.3.3. Reconstitution and 2D crystallization...........................................................91

4.1.4. Perspectives........................................................................................................934.1.5. Material and Methods........................................................................................94

4.1.5.1. Materials................................................................................................94

4.1.5.2. Protein Expression and Purification.............................................................94

4.1.5.3. Reconstitution.........................................................................................94

4.1.5.4. Electron microscopy and image processing....................................................95

4.1.6. References..........................................................................................................95

5. General discussion and conclusions.......................................................................101

6. Acknowledgment..........................................................................................................107

7. Curriculum vitae..........................................................................................................113

7.1. Education...................................................................................................................113

7.2. Teaching.....................................................................................................................113

7.3. Publications...............................................................................................................113

7.4. Meetings.....................................................................................................................114

11

1. AFM and EM instructural biology

13

1. AFM and EM in structural biology

1.1. General introduction

1.1.1. The driving force to do science

What does it look like? What does it do? Itis curiosity, the wish to know, that inducesthese questions. It is human nature - and it isthe driving force to do fundamental researchin any field ranging from astronomy tophysics of smallest matter. The primaryquestions of molecular biologists are: What isthe structure, what is the function, of abiomolecule? Once the structure and functionof a biomolecule are known, its physiologicalrole and interactive mechanisms can bebrought into context within the framework ofthe cell and the whole organism. It is ourultimate goal to understand the molecularmechanisms of all processes in each cell ofour body, ranging from neurons working inour brain, muscle cells allowing us to move, tothe cells of our skin ultimately defining theborderlines of ourselves and our environment.

1.1.2. Membrane proteins

Intrinsic or integral membrane proteins aredefined as proteins that penetrate into and,most often, traverse the lipid bilayer of abiological membrane. Protein structures,which partition into lipid rather than remain inaqueous solution have specific chemicalproperties. They are rich in exposedhydrophobic amino acids and are restricted intheir secondary structure. A consequence ofthese physico-chemical properties is that anintegral membrane protein can only bebrought into aqueous solution whensolubilized in the presence of detergents.

The challenge of understanding membraneproteins and transporters has attracted ourinterest. Figure 1 represents an interesting

result of a genome study. Genomes ofdifferent organisms (E. coli, M. jannaschii, H.sapiens) were screened for their proteincoding open reading frames and these openreading frames were translated into aminoacid chains. The peptides were discriminatedby their hydrophobicities. This resulted intwo families of gene products: Thehydrophilic cytoplasmic proteins, and themembrane proteins, which contain largehydrophobic stretches (e.g., representingtransmembrane helices). By this relativelysimple approach, it has been demonstratedthat 20 - 30 % of genes (the higher theorganism, the larger the percentage) code forstrongly hydrophobic proteins which aremost probably integrated into cellmembranes.

This intriguing result emphasizes theextreme importance of membrane proteins forliving organisms. Membrane proteins connectthe cytoplasm with the extracellular space ofeach living cell, form junctions between livingcells or play an important role in theintracellular compartments. Hence, in bacteria,such molecules work in transport, secretionand bioenergetic processes. Multicellularorganisms even require active communicationbetween their cells. Consequently a largenumber of membrane proteins have evolved,working as receptors for intercellulartrafficking or cellular adhesion andrecognition. Evolution has also created highlyspecific channels and transporters, which areessential for the survival of biologicalsystems; the deletion of many membraneproteins is lethal or leads to severe disease.

The study of membrane protein structureis a difficult challenge: Membrane proteinsremain only folded in their active state when

15

their hydrophobic transmembrane domainsare embedded in a hydrophobic environment -e.g., a lipid bilayer or a detergent micelle. Thisprerequisite makes the growth of 3D crystalsfor structure determination by X-raydiffraction difficult. Consequently, ~2000atomic structures of water-soluble proteinsare available but only ~20 atomic structuresof integral membrane proteins.

The structures so far determined, dividemembrane proteins into two categories: α-helical and β-barrel membrane proteins. Themajority of the structures were determinedusing 3D crystallization and X-raydiffraction. However, three structures havebeen solved by electron crystallography: Plantlight-harvesting complex II,bacteriorhodopsin, and human aquaporin 1.The electron crystallography was carried outusing 2D crystals of protein integrated inlipid bilayers.

1.1.3. 2D crystals allow the acquisitionof structural information onmembrane proteins in a native-like environment

In order to acquire biologically validinformation, it is important to study thestructure of membrane proteins underconditions where they remain functional. Tothis end, membrane proteins are reconstitutedinto 2D crystals in the presence of lipidswhich mimic their native membraneenvironment within a cell (chapters 2.1, 2.2,3.1, 4.1). Although only 3 membrane proteinstructures have been solved to atomicresolution (below 4Å) using electroncrystallography (Kühlbrandt et al., 1994;Henderson et al., 1990; Kimura et al., 1997;Murata et al., 2000), numerous proteins havebeen solved to medium resolution (4Å-10Å)in 2D projection or 3D density maps fromelectron micrographs. Such mediumresolution maps revealed helix arrangements

Figure 1. Number of genes of different organisms as a function of the hydrophobicity of their gene product.Integration over the two peaks corresponding to cytoplasmic and membrane proteins shows that 20 - 30 % of allgenes code for membrane proteins.

16

or/and structural similarities within theaquaporin protein family (Stahlberg et al.,2001).

1.1.4. Atomic force and electronmicroscopy cover large resolutionranges and together provide bothsurface and volume information

Structural biology encompasses a range oftechniques, to elucidate structures andinteractions of biomolecules. The table belowsummarizes some of the advantages anddisadvantages of the various approaches.

The atomic force and the electron

microscope are our tools to investigate thefascinating microcosm of membrane proteins.As listed above, a combination of these twotechniques covers a resolution range frommicrometers to atomic scale, and yields bothsurface and volume information of proteins inthe close to native environment of 2Dcrystals.

Electron microscopy of negatively stainedsamples allows the efficient acquisition ofstructural information. Despite the limitedresolution of ~15Å, the technique provides ahigh signal-to-noise ratio (chapters 1.3, 4.1).By these means, samples of 2D crystals ofmembrane proteins can be checked and a first

Table 1. Advantages and disadvantages of techniques applied in structural biology

technique advantages disadvantages

X-ray crystallography -atomic resolution -well ordered 3Dcrystals are required

-absence of phases

nuclear magnetic resonance -atomic resolution-information about

protein dynamics

-requires large amountsof protein

-requires proteinlabeling

-problems with largemolecules (>40 kDa)

electron crystallography -resolution range fromnanometers to atomicresolution

-requires well ordered2D crystals

-technically difficult

electron microscopy -study of largecomplexes

-study of interactionsbetween single particles

-resolution limit ~5Å-problems with smallmolecules (<100 kDa)

atomic force microscopy -investigations undernative-like conditions

-single molecules canbe addressed

-resolution range fromµm to atomic scale

-time resolvedinformation can begained

-only surface information-protein must beimmobilized on asubstrate

light microscopy -whole living cells canbe studied

-resolution range frommm to nm scale

-resolution limit λ/2 =~200 nm

17

impression of the crystal quality obtained.Furthermore, single solubilized particles canbe visualized, and the averaging of untiltedand tilted images can lead to low resolution3D reconstructions.

Imaging unstained samples under cryoconditions, allows the features of freeze-driedpreparations to be resolved to ~ 9 Å (chapter3.1), and those of vitruous ice embeddedsamples to be resolved to atomic resolution inideal cases (see above), but projection and 3Dmaps can be acquired at any intermediateresolution (chapter 3.1).

Atomic force microscopy allows proteinsurfaces to be contoured at subnanometerresolution in buffer solution mimicking aphysiological environment (chapters 1.2, 2.1,2.2, 3.1) (Müller et al.1999). This offers thepossibility of monitoring function relatedstructural conformational changes of singleproteins or of proteins within complexfunctional assemblies (chapters 1.2, 2.1, 2.2).Loops and termini protruding out of thetransmembrane α-helices, can directly beassessed by the AFM. Consequently,proteolytic cleavage of termini can bemonitored (chapters 2.1, 2.2), leading tounambiguous sidedness assignments.

1.1.5. Results and Perspectives

AFM of membrane proteins demandsmost careful sample preparation in order toachieve high resolution topographs. In thepresent work, preparation procedures for bothhydrophilic (chapter 1.2) and hydrophobic(chapter 1.3) samples were established. Thetwo supports reported (chapter 1.2, 1.3) allowthe physisorption of a wide range ofbiological samples without further fixation,and therefore allow the immobilization andvisualization of proteins in a close to nativeenvironment.

Using atomic force and electronmicroscopy, we have investigated membraneproteins responsible for very different

functions; a water channel, a light-harvestingprotein, and a sugar transporter.

The family of aquaporins, are abundantchannel forming transmembrane proteinsresponsible for selective water transport. Theyare found in bacteria, plants, fungi andmammals (Agre et al., 1993). We havereconstituted purified tetrameric AqpZ, thewater channel of E. coli responsible formaintenance of cell turgor during the volumeexpansion of cell division, into denselypacked vesicles and 2D crystals. Highresolution topographs revealed two distinctprotein surfaces. To assign the sidedness, 2Dcrystals of AqpZ with an N-terminal His-tagwere digested using trypsin. The cleavageresults in a dramatic change of surfaceappearance of one side, allowing thesidedness to be unambiguously assigned(chapter 2.1). Imaging surfaces using loadingforces (50 - 200 pN) on the AFM tip,flexibility mapping, and volume calculationsled to an assignment of the large loop C(chapter 2.1). By combining AFM, metalshadowing electron microscopy, freeze-drying cryo-electron microscopy, andtrehalose-embedded cryo-electronmicroscopy, the sidedness assignment couldbe applied to the whole aquaporin family byan elegant experiment (chapter 3.1): Imagesof 2D crystals were taken, where parts of thecrystal were metal-shadowed, while otherparts were prevented from metal deposition.From such images a direct link betweensurface and density projection could begained. This experimental approach can beapplied to any membrane proteinreconstituted into 2D crystals (chapter 3.1).

Photosynthetic bacteria efficiently convertlight energy into biochemical energy using aset of membrane proteins. In an initial stepphotons are trapped by light-harvestingcomplex (LH2) oligomers, which arecomposed of 9 αβ heterodimers and containa total of 27 bacteriochlorophyls and 9carotenoids. These ring-shaped oligomers

18

were integrated into 2D crystals in thepresence of lipids. Spectra showed thereconstituted proteins to still containchromophores in their native conformation(chapter 2.2). High resolution topographsbefore and after thermolysin cleavage of theC-terminus of the α-subunit defined theposition of this terminus and the sidedness ofthe protein (chapter 2.2). In less purepreparations larger rings were also imagedand assigned as light-harvesting complex 1oligomers. The two protein complexesrepresent a functional assembly inphotosynthetic membranes (Kühlbrandt,1995). The capability of the AFM to imageand identify two different protein species inone membrane opens future perspectives todirectly image native membranes undernative-like conditions (chapter 2.2).

Lactose permease is the best studiedmembrane transporter. It catalyzes thecoupled stoichiometric translocation of β-galactosides and H+ across the cytoplasmicmembrane. To fulfill their physiological roleof transport transporters like this mustundergo strong conformational changes. Thisconformational variability poses difficulties tointegrate LacY into 2D crystals (Zhuang et al.1999). Well diffracting crystals weremultilayered (chapter 4.1). Negative stainelectron microscopy of solubilized particles indifferent orientations with respect to theelectron beam resulted in a low resolution 3Dmap (chapter 4.1). This map shows that thecytochrome b562 engineered into loop 6induces trimerization by hydrophilicinteractions (chapter 4.1). Future experimentsshould increase crystal quality and result in a3D density map from unstained proteins,which can then be combined with theknowledge acquired by biochemical andbiophysical analyses.

1.1.6. References

Agre, P., Preston, G., Smith, B., Jung, J.,Raina, S., Moon, C., Guggino, W. &Nielsen, S. (1993). Aquaporin CHIP: thearchetypal molecular water channel.American Journal of Physiology 265,F436-476.

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

Kimura, Y., Vassylev, D. G., Miyazawa, A.,Kidera, A., Matsushima, M., Mitsuoka, K.,Murata, K., Hirai, T. & Fujiyoshi, Y.(1997). Surface of bacteriorhodopsinrevealed by high-resolution electronmicroscopy. Nature 389, 206-211.

Kühlbrandt, W. (1995). Many wheels makelight work. Nature 374, 497-498.

Kühlbrandt, W., Wang, D. N. & Fujiyoshi, Y.(1994). Atomic model of plant light-harvesting complex by electroncrystallography. Nature 367, 614-621.

Müller, D. J., Fotiadis, D., Scheuring, S.,Müller, S. A. & Engel, A. (1999).Electrostatically balanced subnanometerimaging of biological specimens by atomicforce microscopy. Biophys. J. 76, 1101-1111.

Murata, K., Mitsuoka, K., Hirai, T., Walz, T.,Agre, P., Heymann, J. B., Engel, A. &Fujiyoshi, Y. (2000). Structuraldeterminants of water permeation throughaquaporin-1. Nature 407, 599-605.

Stahlberg, H., Braun, T., Philippsen, A.,Borgnia, M., Agre, P., Kühlbrandt, W. &Engel, A. (2001). The 6.9 Å structure ofGlpF: a basis for homology modeling ofthe glycerol channel from E.coli. J. Struct.Biol., in press

Zhuang, J., Prive, G. G., Werner, G. E.,Ringler, P., Kaback, R. H. & Engel, A.(1999). Two-dimensional crystallization ofthe Escherichia coli lactose permease. J.Struct. Biol. 125, 63-75.

19

1.2. Atomic force microscopy: A powerful tool to observe the assembly andfunction of native proteins

Simon Scheuringa, Dimitrios Fotiadisa, Clemens Möllera,b,c, Andreas Engela and DanielJ. Müllera,c

a M.E.Müller Institute for Structural Biology, Biozentrum, University of Basel, CH-4056 Basel,Switzerlandb Forschungszentrum Jülich, IBI-2: Structural Biology, D-52425 Jülich, Germanyc Max-Planck-Institute of Molecular Cell Biology and Genetics, D-01307 Dresden, Germany

1.2.1. Abstract

Here we discuss the experimentalapproaches that have allowed high resolutionatomic force microscopy (AFM) imaging,and review results that show AFM to be ofgreat interest for biologists, AFM allowssingle proteins to be imaged underphysiologically relevant conditions. Theexceptional signal-to-noise ratio andresolution of AFM topographs enables theoligomerization state and characteristicsubstructures of individual proteins to beresolved. Several examples demonstrate thecapabilities of AFM to directly observe singleproteins, and their conformational changes, tostudy protein-protein interactions and tofollow the assembly of membrane proteins.Here we consider the AFM techniques thathave allowed high resolution imaging, andreview results that show AFM to be apowerful method to analyze biologicalprocesses at the level of single molecules.

1.2.2. Introduction

Atomic force microscopy (AFM; (Binniget al., 1986)) has become a complementarytechnique to X-ray crystallography andelectron microscopy (EM). Of great interestfor biologists, fragile biological samples canbe observed under physiologically relevantconditions. As a result of the AFM'sexceptionally high signal-to-noise ratio,

topographs of single proteins allowsubmolecular details to be resolved to a lateralresolution of ~ 6 Å and a vertical resolutionof 1 Å (Fotiadis et al., 2000; Müller et al.,1995; Schabert et al., 1995; Scheuring et al.,1999b;

Czajkowsky & Shao, 1998). Thisresolution can be attained no matter whetherthe imaged proteins are randomly packed orassembled into two-dimensional (2D)crystalline arrays, which are usually difficultto grow (Czajkowsky & Shao, 1998; Fotiadiset al., 2000; Karrasch et al., 1994; Mou et al.,1996; Mou et al., 1995; Müller & Engel,1999; Scheuring et al., 1999b; Seelert et al.,2000). Standard image averaging techniqueshave proved useful both to interpret thestructural appearance of a protein and toassess the resolution attained (Schabert &Engel, 1994). In such averages commonstructural details of the individual proteins areenhanced while variable features aresuppressed. However, the variable structuralregions of the protein are clearly identified bythe simultaneously calculated standarddeviation map (Müller et al., 1998; Schabert& Engel, 1994). Interestingly, such flexiblestructural regions can undergo functionalrelated conformational changes such asrecently observed by AFM for porin OmpFfrom Escherichia coli (Müller & Engel,1999). In addition, conformational changeshave been visualized in the bacterial surfacelayer of Deinococcus radiodurans (Müller et

21

al., 1996) and measured on bacterio-rhodopsin from Halobacterium salinarum(Haupts et al., 1999; Subramaniam et al.,1999). The surface loop connecting helices Eand F involved in functional conformations ofbacteriorhodopsin has also beendemonstrated by AFM to reversibly changeits conformation upon changing the forceapplied to the stylus (Müller et al., 1995;Heymann et al., 1999).

These results demonstrate how well suitedAFM is for the observation single proteinsand their native substructures. The fact thatintrinsic mechanical properties of thesestructural elements can also be mapped andtheir functional related conformationalchanges directly observed allows conclusionsto be drawn about the function of theseregions (Engel & Müller, 2000). In future,single molecule AFM imaging will beroutinely combined with methods providingcomplementary data for the single proteins ona molecular scale (Engel & Müller, 2000;Schmitt et al., 2000; Weiss, 1999; Weiss,2000). Recently the combination of single-molecule AFM imaging and single-moleculeforce spectroscopy has allowed informationon surface structure, protein folding (Mülleret al., 1999a; Oesterhelt et al., 2000) andprotein-protein interactions to be acquiredsimultaneously (Raab et al., 1999; Ros et al.,1998).

Here we consider AFM techniques thathave provided high-resolution AFMtopographs of native non-crystalline proteins,and demonstrate the importance of singlemolecule imaging as a technique to providenovel insights into fundamental mechanismsof molecular biological processes.

1.2.3. Conditions for single moleculeimaging

Although suitable AFM instrumentationhas been commercially available for manyyears, significant progress has recently beenachieved in several laboratories by optimizing

sample preparation and imaging conditions ofsoft biological structures. Operating themicroscope in buffer solution not only allowsstructural analysis to be carried out undernative conditions (Drake et al., 1989), butalso provides important possibilities tocontrol the tip-sample interactions (Müller etal., 1999b). Stable sample supports whichallow to achieve Angstroem resolution and toprotect the piezo from contact with the buffersolution have been described for bothhydrophilic (Hoh et al., 1991; Schabert &Engel, 1994) and hydrophobic supports(Scheuring et al., 1999a). To adsorbhydrophilic biological samples onto freshlycleaved mica (negative surface charge at pH ≥6) sufficient concentrations of monovalentand/or divalent cations are required tocompensate repulsive electrostatic forces(Müller et al., 1997). In contrast tohydrophilic samples, hydrophobic surfacesreadily adhere to the hydrophobic surface ofhighly oriented pyrolytic graphite (HOPG)(Scheuring et al., 1999a).

Most importantly, fragile biologicalsamples necessitate the adjustment of theforce applied to the cantilever to < 100 pico-Newtons (pN). Otherwise the AFM styluswill reversibly or irreversibly deform the softbiological material. Steric, electrostatic andvan der Waals interactions are the majorforces acting between AFM stylus andbiological sample when imaging in buffersolution (Butt et al., 1995; Müller & Engel,1997; Rotsch & Radmacher, 1997). Force-distance curves, acquired by decreasing theseparation between the sample and the AFMstylus while measuring the cantileverdeflection, reveal the nature of these forces(Butt et al., 1995). The long-rangeelectrostatic double-layer forces (several tensof nm) depend on the charge density of bothinteracting surfaces, the pH and theelectrolyte composition of the buffer solution.Consequently the electrostatic forces can beregulated by adjusting pH and ionic strengthof the imaging buffer solution. Since only

22

short-range interactions between tip andsample (≤ 1 nm) allow acquisition of thestructural details evident on high-resolutiontopographs, the long-range electrostaticforces can be used to minimize the localinteraction forces (Fig. 1, Müller et al.,1999b).

After buffer adjustment, topographs ofprotein surfaces revealing details with a lateralresolution between 5 - 8 Å and a verticalresolution of ~ 1Å can be recordedreproducibly (Fotiadis et al., 2000; Müller etal., 1999c; Scheuring et al., 1999b; Seelert etal., 2000).

To avoid organic contamination of bothsample and tip, buffer solutions are bestprepared using nanopure water (10-18

MΩ/cm). The biological samples should bechecked for their purity by SDS gelelectrophoresis and, if possible, by electronmicroscopy. The fluid cell of the AFM mustbe thoroughly cleaned by repeated ultra-sonication in the presence of ethanol and innanopure water (Müller & Engel, 1997).

To achieve high-resolution the scanningspeed and feedback parameters of the

microscope must be carefully adjusted, scanfrequencies between 4 - 7 Hz and a nominalmagnification giving a pixel sampling of 2 – 3Å / pixel have proved satisfactory.Unfortunately, molecules from the laboratoryenvironment can contaminate the AFM tip.Contaminated tips can be cleaned using SDS-solution and nanopure water (Scheuring et al.,2001), Helmanex (Hoerber et al., 1995), orultraviolet radiation (Thomson et al., 1996)making their repetitive use possible.

1.2.4. Imaging the ion-driven rotor of theATP synthase

FoF1-ATP synthases use the energy of atransmembrane proton (or Na+) gradient tosynthesize the biological energy currencyATP. The flow of cations appear to drive thetransmembrane rotor of the integralmembrane complex FO which is coupled to amolecular shaft (Kato-Yamada et al., 1998;Noji et al., 1997; Sabbert et al., 1996) thatactivates the catalytic F1 complex. In currentmodels of the FoF1-ATP synthases, thenumber of subunits forming the ion driven

Figure 1. Interaction forces between the AFM stylus and a biological specimen . The effectiveinteraction force is the sum of the applied force, the electrostatic repulsion (long range forces) and the van der Waalsattraction (short range forces). The electrostatic interactions are mainly due to the large global tip, while the shortrange forces are mediated by a small local probe, which allows high resolution imaging. Since the electrostaticrepulsion can be regulated by adjusting the ionic strength of the buffer solution the applied force can beelectrostatically 'damped': |Feff| = |Fappl + Fel + FvdW| < |Fappl| (for further details see Müller et al., 1999b).

23

rotor has direct implications for the H+(Na+) /ATP stoichiometry and for the molecularmechanism of ATP synthesis.

The high signal-to-noise ratio andresolution of the AFM has been exploited toimage subunits of isolated Fo rotors fromchloroplast ATP synthase (Seelert et al.,2000). The number of subunits per rotorcould be directly counted in the unprocessedAFM topograph (Fig. 2a). In addition, thereference free (Penczek et al., 1992) averageof the 220 rotors showed a clear 14-foldsymmetry (Fig. 2b). The angular powerspectra of these 220 single rotors weremerged revealing a clear peak for 14-foldrotational symmetry (Fig. 2c; Seelert et al.,2000). Consequently the average (Fig. 2b)was 14-fold symmetrized to reveal the

handedness of the 14 subunits more clearly(Fig. 2d). This stoichiometry is in contrast tothat reported for the E.coli Fo complex whichwas postulated to comprise twelve c-subunits(equal to a subunit III oligomer of chloroplastATP synthase), mainly based on cross-linking experiments (Jones & Fillingame,1998), genetic engineering (Jones &Fillingame, 1998), model building (Dmitrievet al., 1999; Groth & Walker, 1997; Rastogi& Girvin, 1999) and biochemical data thatsuggest that four protons are required for thesynthesis of one ATP. Interestingly, X-rayanalyses of the yeast FoF1-ATP synthasehave yielded a decameric rotor (Stock et al.,1999) while the Na+ driven Fo rotor of theATP synthase from Ilyobacter tartaricus wasfound to be undecameric (Stahlberg et al.,

Figure 2 . Assembly and stoichiometry of the proton driven rotor from chloroplast ATPsynthase. a) Reconstituted, densely packed Fo rotors. The topograph was acquired in contact mode AFM at appliedforces of ~ 50 - 100 pN and in buffer solution (10 mM Tris-HCl, pH 7.8, 25 mM MgCl2). Wide and narrow ringsrepresent the two aqueous surfaces of the membrane spanning rotor. b) Non-symmetrized average after reference freerotational alignment of 220 wide rotor ends imaged as in a). c) Merging of the angular power spectra of the 220large rings clearly elucidates a 14-fold rotational symmetry. d) 14-fold symmetrized average of the average displayedin b).All topographs exhibit a full gray scale of 20Å and are displayed as reliefs tilted by 5°. Scale bars: 100Å (a) and 20Å(b and d).

24

2001). The occurrence of differentstoichiometries implies that there is abiological significance which is not as yetunderstood (Ferguson, 2000).

1.2.5. Conformational flexibility ofproteins

Aquaporin Z (AqpZ) is the E.coli waterchannel responsible for the maintenance ofcell turgor during the volume expansion ofcell division (Borgnia et al., 1999; Calamita etal., 1998). After overexpression andpurification, tetrameric AqpZ oligomers havebeen reconstituted in the presence of lipidsinto two-dimensional crystals and denselypacked vesicles (Ringler et al., 1999). A

projection map of the tetramers was calculatedto 8 Å resolution from cryo electronmicrographs of ice embedded samples(Ringler et al., 1999). Topographs with aresolution of 8 Å resolution have since beenacquired from densely packed vesiclescontaining the oligomers and a high amountof lipid (Fig. 3a; Scheuring et al., 1999b).The AqpZ tetramers are clearly laterallystabilized by their close packing, butrotationally misaligned (Fig. 3a). Alltetramers share the crown-like appearance,although careful comparison of theirpolypeptide loops protruding from the bilayersurface reveals their structural individuality(Fig. 3a). The aligned averaged topographshows to possess twelve surface protrusions

Figure 3. Modulation of individual AqpZ tetramers by the AFM probe. a) Topograph of a vesicledensely packed with AqpZ tetramers exposing their extracellular surface. The topograph was recorded in buffersolution (17 mM Tris-HCl, pH 7.2, 150 mM KCl) using contact mode AFM at an applied force of ~ 80 pN. b)Four-fold symmetrized average of 289 aligned particles as displayed in a) after translational and angular alignment.The large peripheral protrusion has been identified as polypeptide loop C connecting two transmembrane α-helices.c) Minimal force image of ~ 20 individual tetramers. d) The region displayed in c) imaged with a loading force of +80 pN. e) Computed reconstruction illustrating the force induced conformational change on the extracellular surfaceof AqpZ. As the imaging force (left) is increased the large C-loop is progressively displaced from the peripherallocation it otherwise occupies (right; minimal force) (scale bar: 100 Å; full gray scale: 7 Å; Scheuring et al., 1999b).All topographs exhibit a full gray scale of 7 Å and are displayed as reliefs tilted by 5°. Scale bars: 100 Å (a, c, d ande) and 20 Å (b).

25

per tetramer (Fig. 3b). AqpZ has a large C-loop (~ 26 aminoacids), which may exhibitpronounced flexibility. Comparison ofimages of AqpZ tetramers recorded atminimal forces (~ 80 pN, Fig. 3c) with thoseacquired after increasing the stylus loadingforce by + 80 pN (Fig. 3d), shows the fourelongated peripheral protrusions to disappearunder higher loading forces. The computedreconstruction of the force inducedconformational change of the extracellularsurface of AqpZ demonstrates the effect (Fig.3e). This flexibility and the volume of theunperturbed elongated peripheral protrusionsidentify them as the C-loops connecting twotransmembrane α-helices. The use of AFMimages recorded while increasing the appliedforce to assign flexible protein domains hasbeen demonstrated previously. For

bacteriorhodopsin (Müller et al., 1999c) andthe phi29 connector (Müller et al., 1998) theobserved flexibility was of functionalrelevance (Brown et al., 1995; Subramanian etal., 1999; Simpson et al., 2000). In case ofAqpZ, however, the functional relevance of thestructural flexibility remains to be elucidated.

1.2.6. The tongue-and-groove interactionof MIP tetramers

The major intrinsic protein (MIP)expressed in eye lens fiber cells is thefounding member of the aquaporin family(Gorin et al., 1984). Tetrameric MIPextracted and purified from sheep lens fibercells has been reconstituted into doublelayered 2D crystals by dialysis of a protein-lipid-detergent mixture (Hasler et al., 1998).

Figure 4. Structure and interaction of the major intrinsic protein (MIP). a) Densely packed vesicleshowing the extracellular surface of MIP. The topograph was recorded in buffer solution (20 mM Na-acetate-HCl, pH5.0, 50 mM NaCl) using contact mode AFM and applied forces of < 100 pN. b) 4-fold symmetrized average of 445particles imaged as in (a). c) Computer reconstruction of two stacked MIP membranes. To visualize the tongue-to-groove interaction of their extracellular surfaces the top layer has been displaced by one unit cell to the left and to theback (Fotiadis et al., 2000).All topographs exhibit a full gray scale of 14 Å and are displayed as reliefs tilted by 5° (a and b) and by 60° (c). Scalebars: 100 Å (a), 20 Å (b) and 50 Å (c).

26

AFM investigations of crystalline sheets anddensely packed vesicles elucidated theextracellular (Fig. 4a) and the cytoplasmicsurface (Fig. 4c, top) at ~ 6 Å resolution(Fotiadis et al., 2000). Using high forcesapplied to the AFM stylus the double layeredcrystals were disected and access to theinteracting inner surfaces was gained. Theresults demonstrate that the extracellulardomains of MIP tetramers from two adjacentlayers interact with each other through a'tongue-and-groove' fit (Fig. 4c). This findingsupports both the proposed dual function ofMIP, namely to channel water and to act incell-cell adhesion (Benedetti et al., 2000), andthe result of a recent study of lens fiber cellarchitecture in mice expressing mutated MIP:lenses that do not integrate functional MIP inthe plasma membranes have disorganizedfiber cells (Shiels et al. 2000).

1.2.7. Imaging the subcomplexes of theGroE chaperonin system: GroELand GroES

Imaging the subcomplexes of the GroEchaperonin system: GroEL and GroESGroEL is found in the cytoplasm of E. coliwhere it acts as an ATP dependent molecularchaperone that ensures the correct folding ofsoluble proteins (Houry et al., 1999). GroELconsists of 14 identical subunits assembledinto a double ring barrel with a diameter of ~14 nm. The apical domains of GroEL,exposed at both ends of the cylinder, househydrophobic aminoacid residues which trapunfolded or misfolded proteins via hydro-phobic interactions (Braig et al., 1994). Mouet al. (1996) have immobilized GroELcylinders head-on onto a mica support andimaged the opening of the cylinder cavity atsubmolecular resolution by contact modeAFM (Fig. 5a, b).

GroES acts as a co-chaperonin togetherwith GroEL in the folding cycle. The modelpostulates that the interaction between GroES

Figure 5. Observing individual chaperonins, GroEL and GroES. a) Topograph of GroEL cylindersadsorbed to mica. The apical domain of the heptameric GroEL rings positioned at the opening of the barrel are clearlyresolved. b) Seven-fold symmetrized average (n=26) of the particles displayed in (a). c) Topograph of GroES discsadsorbed to mica. d) Seven-fold symmetrized average (n=54) of the particles displayed in (c). The topographs wererecorded in deionized water after fixation of the sample with 2% glutaraldehyde using contact mode AFM (Mou et al.,1996). Topographs exhibit a full gray scale of 140 Å (c and b) or of 80 Å (c and d) and are displayed as reliefs tiltedby 5°. Scale bars: 200 Å (a and c) and 50 Å (b and d). (Images by courtesy of Zhifeng Shao, Virginia).

27

and the apical domain of GroEL is strongerthan the affinity of an unfolded protein to thelatter. Consequently the unfolded protein ispushed into the cavity (Hartl & Martin,1995). To be observed with the AFM, GroESheptamers were adsorbed onto a mica surfacewhere they formed a densely packed layer(Mou et al., 1996). After this, the biologicalsample was chemically fixed withglutaraldehyde and subsequently imaged byAFM (Fig. 5c). The calculated averageelucidates the disc-like molecular organizationof GroES most clearly: a wide ring with adiameter of ~ 84 Å which is thought tointeract with GroEL is topped by a narrowercrown of ~ 45 Å in diameter (Fig. 5d). Acombination of the two experiments by Mouet al. (1996) was made possible due to thedevelopment of a fast scanning AFM (Vianiet al., 1999). With such an instrument it was

possible to directly visualize the binding andunbinding of GroES to GroEL in real time, animportant event in the chaperonin mediatedprotein folding process (Viani et al., 2000).

1.2.8. Observing the assembly ofmembrane proteins

Bacteriorhodopsin, a light-activated protonpump of the archae Halobacteriumsalinarum, shares structural and functionalsimilarities to rhodopsin, a G-protein coupledreceptor in eukaryotes. In addition to theirseven transmembrane α-helices, both proteinshave a retinal as photoactive chromophore.Both structures have been solved to atomicresolution (Belrhali et al., 1999; Luecke et al.,1999; Pebay-Peyroula et al., 1997; Mitsuokaet al., 1999; Kimura et al., 1997; Palczewskiet al., 2000). Although the organization of

Figure 6. Observing the disassembly of purple membranes. a) Photobleached purple membrane imagedin buffer solution (10 mM Tris-HCl, pH 7.8, 150 mM KCl) using contact mode AFM at applied forces of ~ 100pN. Upon exposure to light in the presence of hydroxylamine purple membranes lose their crystallinity. Duringdisassembly of the purple membrane lattice the bacteriorhodopsin molecules remain assembled into trimers. b)Three-fold symmetrized average calculated by single particle alignment of 172 disordered trimers as displayed in (a).c) Adsorption spectra showing the progress of photobleaching in the presence of hydroxylamine. The topographsexhibit a full gray scale of 8 Å and are displayed as reliefs tilted by 5°. Scale bars: 100 Å (a) and 20 Å (b).

28

bacteriorhodopsin in native membranes hasbeen well studied, little is known forrhodopsin (Körschen et al., 1999).

During biosynthesis the characteristicfunctional and structural properties of thepurple membrane are only observed after theformation of bacteriorhodopsin frombacterioopsin and retinal (Sumper &Herrmann, 1976a; Sumper & Herrmann,1976b). The reversal of this process has beenmonitored by AFM after cleaving the Schiffbase bond between retinal andbacteriorhodopsin in the presence ofhydroxylamine and light (Fig. 6; Möller et al.,2000). While individual bacterioopsinsremained stably assembled into trimers, thesetrimers lost their orientation relative to theother protein trimers in the membrane (Fig.6a). This disorder of the bacterioopsin trimersincreased with the completeness of the Schiffbase bond cleavage as can be assessed by theabsorption spectra shown in Fig. 6c.Interestingly, after washing away theretinaloxime and adding retinal, the Schiffbase bond was re-established. With the re-formation of bacteriorhodopsin, the trimersspontaneously re-assembled into membranepatches structurally and spectroscopicallyindistinguishable from native purplemembrane.

1.2.9. Outlook

Proteins are soft and flexible nanomachines, which exist in different functionrelated conformational states. In naturedifferent proteins interact with each other toform functionally relevant complexes. Suchinteractions often induce changes of theprotein structure (Müller et al., 1999c;Kellenberger, 1968). The understanding ofthese molecular interactions together with thefunction related conformational states is ofsignificant importance for biologicalprocesses. If such interactions aresynchronized among a class of proteins theycan be observed using techniques which

require averaging processes. However,biological processes involve action of singleproteins whose structure cannot be assessedby structural techniques applying averagingprocesses. Insight into molecular interactionsand the individual functional states ofproteins can only be acquired by the directobservation of single proteins (Weiss, 1999;Weiss, 2000). The AFM opens a new way ofassessing the function-related conformationalstates of individual proteins at submolecularresolution (Engel & Müller, 2000). Theunique ability of this instrument to directlyvisualize single proteins in their nativeenvironment will provide novel insights intothe interactions between biomolecules and theformation of functional assemblies within anative membrane.

Driven by their dynamic clustering, lipids,other amphiphilic molecules and membraneproteins have been found to form rafts thatmove within the fluid bilayer (Simons &Ikonen, 1997). Such rafts are proposed toserve as platforms for the attachment ofproteins which control protein sorting andtrafficking through secretory and endocyticpathways (Brown & London, 1998; Simons& Ikonen, 1997). The assembly of membraneproteins into rafts appears to be of significantimportance during signal transduction. It willbe a challenge to understand the molecularmechanisms driving the assembly of rafts. Inthis context, the observation of the de- and re-assembly of bacteriorhodopsin into fullyactive purple membrane can be seen as a steptowards studying the formation of functionalmembrane protein assemblies (Möller et al.,2000). In the future, the ability of AFM toimage individual proteins will allow morecomplex biological systems to be studied andwill deliver a novel insight into the biogenesisof various native membranes, into theinteractions between similar or differentmembrane proteins within a membrane, andinto the formation of supramolecularcomplexes.

29

1.2.10. Acknowledgement

This work was supported by the SwissNational Foundation for Scientific Research(grant 4036 - 44062 to A. E.), the SwissPriority Project for Micro and Nano SystemTechnology (MINAST), the EC project forresearch on water channels (grant BIO4-CT98-0024 to A. E.), and the Maurice E.Müller Foundation of Switzerland.

1.2.11. References

Benedetti, E. L., Dunia, I., Recouvreur, M.,Nicolas, P., Kumar, N. M. & Bloemendal,H. (2000). Structural organization of gapjunctions as revealed by freeze-fractureand SDS fracture-labeling. Eur. J. CellBiol. 79, 575-582.

Belrhali, H., Nollert, P., Royant, A., Menzel,C., Rosenbusch, J. P., Landau, E. M. &Pebay-Peyroula, E. (1999). Protein, lipidand water organization inbacteriorhodopsin crystals: a molecularview of the purple membrane at 1.9 Åresolution. Struct. Fold. Des. 7(8), 909-17.

Binnig, G., Quate, C. F. & Gerber, C. (1986).Atomic force microscope. Phys. Rev. Lett.56(9), 930-933.

Borgnia, M., Kozono, D., Calamita, G.,Maloney, P. C. & Agre, P. (1999).Functional reconstitution andcharacterization of AqpZ, the E. coli waterchannel protein. J. Mol. Biol. 29(5), 1169-1179.

Braig, K., Otwinowski, Z., Hedge, R.,Boisvert, D. C., Joachimiak, A., Horwich,A. L. & Sigler, P. B. (1994). The crystalstructure of the bacterial chaperoninGroEL at 2.8Å. Nature 371, 578-585.

Brown, D. A. & London, E. (1998).Functions of lipid rafts in biologicalmembranes. Annu. Rev. Cell. Dev. Biol.14, 111-136.

Brown, L. S., Varo, G., Needleman, R. &Lanyi, J. K. (1995). Functionalsignificance of a protein conformationchange at the cytoplasmic end of helix Fduring the bacteriorhodopsin photocycle.

Biophys. J. 69(5), 2103-11.

Butt, H.-J., Jaschke, M. & Ducker, W.(1995). Measuring surface forces inaqueous solution with the atomic forcemicroscope. Bioelect. Bioenerg. 38, 191-201.

Calamita, G., Kempf, B., Bonhivers, M.,Bishai, W. R., Bremer, E. & Agre, P.(1998). Regulation of the Escherichia coliwater channel gene AqpZ. Proc. Natl.Acad. Sci. USA 95, 3627-3631.

Czajkowsky, D. M. & Shao, Z. (1998).Submolecular resolution of singlemacromolecules with atomic forcemicroscopy. FEBS Lett. 430(1-2), 51-54.

Dmitriev, O. Y., Jones, P. C. & Fillingame, R.H. (1999). Structure of the subunit coligomer in the F1Fo ATP synthase:Model derived from solution structure ofthe monomer and cross linking in thenative enzyme. Proc. Natl. Acad. Sci. USA96, 7785-7790.

Drake, B., Prater, C. B., Weisenhorn, A. L.,Gould, S. A. C., Albrecht, T. R., Quate, C.F., Cannell, D. S., Hansma, H. G. &Hansma, P. K. (1989). Imaging crystals,polymers, and processes in water with theatomic force microscope. Science 243,1586-1588.

Engel, A. & Müller, D. J. (2000). Observingsingle biomolecules at work with theatomic force microscope. Nat. Struct. Biol.7(9), 715-718.

Ferguson, S. J. (2000). ATP synthase: Whatdetermines the size of a ring? Curr. Biol.10, R804-R808.

Fotiadis, D., Hasler, L., Müller, D. J.,Stahlberg, H., Kistler, J. & Engel, A.(2000). Surface tongue-and-groovecontours on lens MIP facilitate cell-to-celladherence. J. Mol. Biol. 300(4), 779-789.

Gorin, M. B., Yancey, S. B., Cline, J., Revel,J.-P. & Horwitz, J. (1984). The majorintrinsic protein (MIP) of the bovine lensfiber membrane: Characterization andstructure based on cDNA cloning. Cell 39,49-59.

Groth, G. & Walker, J. E. (1997). Model ofthe c-subunit oligomer in the membrane

30

domain of F-ATPases. FEBS Lett. 410,117-123.

Hartl, F. U. (1996). Molecular chaperones incellular protein folding. Nature 381, 571-580.

Hartl, F.-U. & Martin, J. (1995). Molecularchaperones in cellular protein folding.Curr. Opin. Struct. Biol. 5, 92-102.

Hasler, L., Heymann, J. B., Engel, A., Kistler,J. & Walz, T. (1998). 2D crystallizationof membrane proteins: rationales andexamples. J. Struct. Biol. 121, 162-171.

Hasler, L., Walz, T., Tittmann, P., H., G.,Kistler, J. & Engel, A. (1998). Purifiedlens major intrinsic protein (MIP) formshighly ordered tetragonal two-dimensionalarrays by reconstitution. J. Mol. Biol. 279,855-864.

Haupts, U., Tittor, J. & Oesterhelt, D. (1999).Closing in on bacteriorhodopsin: progressin understanding the molecule. Annu. Rev.Biophys. Biomol. Struct. 28, 367-399.

Heymann, J. B., Müller, D. J., Landau, E.,Rosenbusch, J., Pebay-Peroulla, E., Büldt,G. & Engel, A. (1999). Charting thesurfaces of purple membrane. J. Struct.Biol. 128(3), 243-249.

Hoh, J. H., Lal, R., John, S. A., Revel, J.-P. &Arnsdorf, M. F. (1991). Atomic forcemicroscopy and dissection of gapjunctions. Science 253, 1405-1408.

Hörber, J. K. H., Mosbacher, J., Haberle, W.,Ruppersberg, J. P., Sakmann, B. (1995).A look at membrane patches with thescanning force microscope. Biophys. J. 68(5), 1687-1693.

Houry, W. A., Frishman, D., Eckerskorn, C.,Lottspeich, F. & Hartl, F. U. (1999).Identification of in vivo substrates of thechaperonin GroEL. Nature 402(6758),147-154.

Jones, P. C. & Fillingame, R. H. (1998).Genetic fusions of subunit c in the Fosector of H+-transporting ATP synthase.J. Biol. Chem. 273(45), 29701-29705.

Karrasch, S., Hegerl, R., Hoh, J., Baumeister,W. & Engel, A. (1994). Atomic forcemicroscopy produces faithful high-

resolution images of protein surfaces in anaqueous environment. Proc. Natl. Acad.Sci. USA 91, 836-838.

Kato-Yamada, Y., Noji, H., Yasuda, R.,Kinosita, K. J. & Yoshida, M. (1998).Direct observation of the rotation ofepsilon subunit in F1-ATPase. J. Biol.Chem . 273, 19375-19377.

Kellenberger, E. (1968). Studies on themorphopoiesis of the head of phage T-even. V. The components of the T4 capsidand of other, capsid-related structures.Virology 34(3), 549-561.

Kimura, Y., Vassylev, D. G., Miyazawa, A.,Kidera, A., Matsushima, M., Mitsuoka, K.,Murata, K., Hirai, T. & Fujiyoshi, Y.(1997). Surface of bacteriorhodopsinrevealed by high-resolution electronmicroscopy. Nature 389(6647), 206-211.

Körschen, H. G., Beyermann, M., Müller, F.,Heck, M., Vantler, M., Koch, K.-H.,Kellner, R., Wolfrum, U., Bode, C.,Hofmann, K. P. & Kaupp, U. B. (1999).Interaction of glutamic-acid-rich proteinswith the cGMP signaling pathway in rodphotoreceptors. Nature 400, 761-766.

Luecke, H., Schobert B., Richter, H.T.,Cartailler, J.P., Lanyi, J.K. (1999).Structure of bacteriorhodopsin at 1.55 Åresolution. J. Mol. Biol. 291(4), 899-911.

Mitsuoka, K., Hirai, T., Murata, K.,Miyazawa, A., Kidera, A., Kimura, Y. &Fujiyoshi, Y. (1999). The structure ofbacteriorhodopsin at 3.0 Å resolutionbased on electron crystallography:implication of the charge distribution. J.Mol. Biol. 286(3), 861-82.

Möller, C., Büldt, G., Dencher, N., Engel, A.& Müller, D. J. (2000). Reversible loss ofcrystallinity on photobleaching purplemembrane in presence of hydroxylamine.J. Mol. Biol. 301(4), 869-879.

Mou, J., Czajkowsky, D. M., Sheng, S., Ho,R. & Shao, Z. (1996a). High resolutionsurface structure of E. coli GroESoligomer by atomic force microscopy.FEBS Lett. 381, 161-164.

Mou, J., Sheng, S., Ho, R. & Shao, Z.(1996b). Chaperonins GroEL and GroES:views from atomic fore microscopy.

31

Biophys. J. 71, 2213-2221.

Mou, J. X., Yang, J. & Shao, Z. F. (1995).Atomic force microscopy of cholera toxinB-oligomers bound to bilayers ofbiologically relevant lipids. J. Mol. Biol.248(3), 507-512.

Müller, D. J., Amrein, M. & Engel, A.(1997). Adsorption of biologicalmolecules to a solid support for scanningprobe microscopy. J. Struct. Biol. 119(2),172-188.

Müller, D. J., Baumeister, W. & Engel, A.(1996). Conformational change of thehexagonally packed intermediate layer ofDeinococcus radiodurans imaged byatomic force microscopy. J. Bacteriol.178(11), 3025-3030.

Müller, D. J., Baumeister, W. & Engel, A.(1999a). Controlled unzipping of abacterial surface layer with atomic forcemicroscopy. Proc. Natl. Acad. Sci. USA96(23), 13170-13174.

Müller, D. J., Büldt, G. & Engel, A. (1995).Force-induced conformational change ofbacteriorhodopsin. J. Mol. Biol. 249(2),239-243.

Müller, D. J. & Engel, A. (1997). The heightof biomolecules measured with the atomicforce microscope depends on electrostaticinteractions. Biophys. J. 73, 1633-1644.

Müller, D. J. & Engel, A. (1999). pH andvoltage induced structural changes ofporin OmpF explain channel closure. J.Mol. Biol. 285(4), 1347-1351.

Müller, D. J., Engel, A., Carrascosa, J. &Veléz, M. (1997). The bacteriophage ø29head-tail connector imaged at highresolution with atomic force microscopyin buffer solution. EMBO J. 16, 101-107.

Müller, D. J., Fotiadis, D. & Engel, A.(1998). Mapping flexible protein domainsat subnanometer resolution with the AFM.FEBS Lett. 430(1,2), 105-111.

Müller, D. J., Fotiadis, D., Scheuring, S.,Müller, S. A. & Engel, A. (1999b).Electrostatically balanced subnanometerimaging of biological specimens byatomic force microscopy. Biophys. J. 76,1101-1111.

Müller, D. J., Sass, H.-J., Müller, S., Büldt,G. & Engel, A. (1999c). Surfacestructures of native bacteriorhodopsindepend on the molecular packingarrangement in the membrane. J. Mol.Biol. 285(5), 1903-1909.

Müller, D. J., Schabert, F. A., Büldt, G. &Engel, A. (1995). Imaging purplemembranes in aqueous solutions atsubnanometer resolution by atomic forcemicroscopy. Biophys. J. 68, 1681-1686.

Noji, H., Yasuda, R., Yoshida, M. & Kinosita,K. (1997). Direct observation of therotation of F1-ATPase. Nature 386, 299-302.

Oesterhelt, F., Oesterhelt, D., Pfeiffer, M.,Engel, A., Gaub, H. E. & Müller, D. J.(2000). Unfolding pathways of individualbacteriorhodopsins. Science 288(5463),143-146.

Palczewski, K., Kumasaka, T., Hori, T.,Behnke, C. A., Motoshima, H., Fox, B. A.,Le Trong, I., Teller, D. C., Okada, T.,Stenkamp, R. E., Yamamoto, M. &Miyano, M. (2000). Crystal structure ofrhodopsin: A G protein-coupled receptor.Science 289, 739-745.

Pebay-Peyroula, E., Rummel, G.,Rosenbusch, J. P. & Landau, E. M.(1997). X-ray structure ofbacteriorhodopsin at 2.5 angstroms frommicrocrystals grown in lipidic cubicphases. Science 277(5332), 1676-1681.

Palczewski, K., Kumasaka, T., Hori, T.,Behnke, C. A., Motoshima, H., Fox, B. A.,Le Trong, I., Teller, D. C., Okada, T.,Stenkamp, R. E., Yamamoto, M. &Miyano, M. (2000). Crystal structure ofrhodopsin: A G protein-coupled receptor.Science 289(5480), 739-45.

Penczek, P., Radermacher, M. & Frank, J.(1992). Three-dimensional reconstructionof single particles embedded in ice.Ultramicroscopy 40, 33-53.

Raab, A., Han, W., Badt, D., Smith-Gill, S.,Lindsay, S., Schindler, H. & Hinterdorfer,P. (1999). Antibody recognition imagingby atomic force microscopy. NatureBiotech. 17, 902-905.

32

Rastogi, V. K. & Girvin, M. E. (1999).Structural changes linked to protontranslocation by subunit c of the ATPsynthase. Nature 402(6759), 263-8.

Ringler, P., Borgnia, M., Stahlberg, H., Agre,P. & Engel, A. (1999). Structure of thewater channel AqpZ from Escherichia colirevealed by electron crystallography.Journal Molecular Biology 291, 1181-1190.

Ros, R., Schwesinger, F., Anselmetti, D.,Kubon, M., Schäfer, R., Plückthun, A. &Tiefenauer, L. (1998). Antigen bindingforces of individually addressed single-chain Fv antibody molecules. Proc. Natl.Acad. Sci. 95, 7402-7405.

Rotsch, C. & Radmacher, M. (1997).Mapping local electrostatic forces with theatomic force microscope. Langmuir13(10), 2825-2832.

Sabbert, D., Engelbrecht, S. & Junge, W.(1996). Intersubunit rotation in active F-ATPase. Nature 381, 623-625.

Schabert, F. A. & Engel, A. (1994).Reproducible acquisition of Escherichiacoli porin surface topographs by atomicforce microscopy. Biophys. J. 67(6),2394-2403.

Schabert, F. A., Henn, C. & Engel, A. (1995).Native Escherichia coli OmpF porinsurfaces probed by atomic forcemicroscopy. Science 268, 92-94.

Scheuring, S., Müller, D. J., Ringler, P.,Heymann, J. B. & Engel, A. (1999a).Imaging streptavidin 2D crystals onbiotinylated lipid monolayers at highresolution with the atomic forcemicroscopy. J. Microsc. 193, 28-35.

Scheuring, S., Reiss-Husson, F., Engel, A.,Rigaud, J. L. & Ranck, J. L. (2001). Highresolution AFM topographs of Rubrivivaxgelatinosus light-harvesting complexLH2. EMBO J. in press.

Scheuring, S., Ringler, P., Borgnia, M.,Stahlberg, H., Müller, D. J., Agre, P. &Engel, A. (1999b). High resolutiontopographs of the Escherichia coliwaterchannel aquaporin Z. EMBO J. 18,4981-4987.

Schmitt, L., Ludwig, M., Gaub, H. E. &Tampe, R. (2000). A metal-chelatingmicroscopy tip as a new toolbox forsingle-molecule experiments by atomicforce microscopy. Biophys. J. 78(6),3275-85.

Seelert, H., Poetsch, A., Dencher, N. A.,Engel, A., Stahlberg, H. & Müller, D. J.(2000). Proton powered turbine of a plantmotor. Nature 405, 418-419.

Shiels, A., Mackay, D., Bassnett, S., Al-Ghoul, K &. Kuszak, J. (2000).Disruption of lens fiber cell architecture inmice expressing a chimeric AQP0-LTRprotein. FASEB 14, 2207-2212

Simons, K. & Ikonen, E. (1997). Functionalrafts in cell membranes. Nature 387, 569-572.

Simpson, A. A., Tao, Y., Leiman, P. G.,Badasso, M. O., He, Y., Jardine, P. J.,Olson, N. H., Morais, M. C., Grimes, S.,Anderson, D. L., Baker, T. S. &Rossmann, M. G. (2000). Structure of thebacteriophage phi29 DNA packagingmotor. Nature 408(6813), 745-50.

Simons, K. & Ikonen, E. (1997). Functionalrafts in cell membranes. Nature 387, 569-572.

Stahlberg, H., Müller, D. J., Suda, K.,Fotiadis, D., Engel, A., Meier, T., Matthey,U. & Dimroth, P. (2001). Bacterialsodium ATPase has an undecameric rotor.EMBO Reports , in press.

Stock, D., Leslie, A. G. & Walker, J. E.(1999). Molecular architecture of therotary motor in ATP synthase. Science286(5445), 1700-5.

Subramaniam, S. (1999). The structure ofbacteriorhodopsin: an emergingconsensus. Curr Opin Struct Biol 9(4),462-8.

Sumper, M. & Herrmann, G. (1976a).Biogenesis of purple membrane:Regulation of bacteriorhodopsin. FEBSLett. 69(1), 149-152.

Sumper, M. & Herrmann, G. (1976b).Biosynthesis of purple membrane:Control of retinal synthesis by bacterio-opsin. FEBS Lett. 71(2), 333-336.

33

Thomson, N. H., Fritz, M., Radmacher, M.,Cleveland, J. P., Schmidt, C. F. &Hansma, P. K. (1996). Protein trackingand detection of protein motion usingatomic force microscopy. Biophys. J.70(5), 2421-2431.

Viani, M. B., Pietrasanta, L. I., Thompson, J.B., Chand, A., Gebeshuber, I. C., Kindt, J.H., Richter, M., Hansma, H. G. &Hansma, P. K. (2000). Probing protein-protein interactions in real time.Nat.Struct. Biol. 7(8), 644-647.

Viani, M. B., Schäfer, T. E., Chand, A., Rief,M., Gaub, H. & Hansma, P. K. (1999).Small cantilevers for force spectroscopyof single molecules. J. Appl. Phys. 86(4),2258-2262.

Weiss, S. (1999). Fluorescence spectroscopyof single biomolecules. Science283(5408), 1676-83.

Weiss, S. (2000). Measuring conformationaldynamics of biomolecules by singlemolecule fluorescence spectroscopy. Nat.Struct. Biol. 7(9), 724-9.

34

1.3. Imaging streptavidin 2D-crystals on biotinylated lipid monolayers athigh resolution with the atomic force microscope

Simon Scheuring, Daniel J. Müller, Philippe Ringler, J. Bernard Heymann, AndreasEngel

M.E.Müller Institute for Structural Biology, Biozentrum, University of Basel, CH-4056 Basel,Switzerland

1.3.1. Summary

Streptavidin crystals were grown onbiotinylated lipid monolayers at an air/waterinterface and transferred onto highly orientedpyrolytic graphite (HOPG). These arrayscould be imaged to a resolution below 1 nmwith the atomic force microscope. The surfacetopographs obtained were compared withnegative stain electron microscopy imagesand the atomic model determined by X-raycrystallography. The streptavidin tetramer (60kD) exposes two free biotin binding sites tothe buffer solution, while two are occupied bythe linkage to the lipid monolayer. Thereforethe streptavidin 2D crystals can be used asnanoscale matrices for binding biotinylatedcompounds. Furthermore, this HOPG-basedpreparation method provides a general novelapproach to study the structure of proteinarrays assembled on lipid monolayers withthe AFM.

1.3.2. Introduction

The atomic force microscope (AFM)(Binnig et al., 1986) has become a powerfultool in structural biology, because topographsof biomolecules can be acquired underphysiological conditions at subnanometerresolution. High resolution AFM requiresatomically flat surfaces (Müller et al., 1995).Frequently used supports to immobilizebiological objects for imaging with the AFMare mica having a polar surface (Müller et al.,1997) and functionalized gold (Wagner et al.,

1996) or silanized glass (Karrasch et al.,1993) allowing covalent crosslinking to beachieved. These supports are suitable for theadsorption of membrane proteinsreconstituted into lipid bilayers (Jap et al.,1992) and of single particles such as solubleproteins (Mou et al., 1996).

In this report atomically flat HOPG(highly oriented pyrolytic graphite) was usedas a hydrophobic support to acquire highresolution topographs of streptavidin 2Dcrystals on biotinylated lipid monolayersusing the AFM. HOPG is produced bydeposition of carbon at high temperature (upto 3000°C) and under pressure from the gasphase (Moore, 1973). The material thusobtained consists of crystallites that are welloriented perpendicular to the basal graphiteplanes (Ohler et al., 1997), and can easily becleaved with scotch tape. On a macroscaleHOPG is not flat, but the surface is separatedinto atomically flat terraces. The flatness ofthese plateaus assures that the surfacefeatures are specimen specific and not due toirregularities of the substrate itself. Thegraphite, consisting of hexagonally orderedcarbon atoms, presents a nonpolar stronglyhydrophobic surface.

Streptavidin is a tetrameric protein withfour biotin binding sites (Green, 1975). Theease with which streptavidin crystallizes in 2Darrays on a biotinylated lipid monolayer(Darst et al., 1991; Avila-Sakar & Chiu,1996) makes it an ideal model system toinvestigate two-dimensional crystals grownon lipid monolayers that are transferred to

35

HOPG. In addition, a comparison can bemade with the transmission electronmicroscope (TEM) of the same crystals aswell as the available atomic structure from X-ray crystallography (Avila-Sakar & Chiu,1996; Hendrickson et al., 1989). Topographsrecorded with the AFM compare favorablywith negatively stained samples and theatomic structure of streptavidin.

1.3.3. Materials and Methods

1.3.3.1. Materials

Streptavidin and biotin-LC-DPPE (biotin-longchain-dipalmitoyl phosphatidyl ethanol-amine) were obtained from Pierce Ltd. (Rock-ford, USA), DOPC (dioleoyl phosphatidylcholine) from Avanti Polar Lipids (Birming-ham, AL), mica from Mica New York (VarickStreet, N. Y. 10013), HOPG from AdvancedCeramics Corporation (Cleveland, USA), andAraldit from Ciba-Ceigy (Basel, Switzerland).

1.3.3.2. Hydrophobicity measurement

A 10 µl drop of millipore filtered H2O wasdeposited on the substrates mica (Bailey,

1984), washed glass (Karrasch et al., 1993),untreated glass, HOPG (Moore, 1973), andteflon. Right after deposition of the drops thediameters were photographed at 6xmagnification using a binocular microscopewith a mounted camera, and the diametersmeasured.

1.3.3.3. Crystallization of streptavidin onbiotin-lipid monolayer

Highly ordered streptavidin arrays wereproduced by depositing 15 µl of streptavidinsolution (10 mM Tris-HCl, pH 7.5, 150NaCl) at a concentration of 0.1 mg/ml in aTeflon well 0.5 mm deep and 4 mm indiameter (Fig. 1a). A 0.5 µl drop of the lipidmixture (0.5 mg/ml Biotin-LC-DPPE :DOPC, 1:4 (mol : mol), in chloroform :hexane, 1:1 (vol : vol)) was then deposited ontop of the protein solution with a Hamiltonsyringe (Fig. 1a). Incubation overnight atroom temperature allowed the adsorption ofprotein to the lipid monolayer and subsequentgrowing of 2D-crystals (Fig. 1b, c).

Figure 1 . Crystallization of streptavidin on a biotin-lipid containing monolayer andadsorption to HOPG. a) A 15 µl drop of streptavidin solution (0.1 mg/ml in 10 mM Tris-HCl, pH 7.5, 150NaCl) was deposited into a teflon well and a 0.5 µl drop of lipid mixture (Biotin-LC-DPPE : DOPC, 1:4) wasspread on the drop to form a monolayer. b) Overnight incubation at room temperature allowed streptavidin bindingto the lipid monolayer and the formation of 2D-streptavidin crystals. c) The monolayer was adsorbed to freshlycleaved HOPG and d) mounted in the AFM in a drop of scanning buffer (10 mM Tris-HCl, pH 7.5, 150 KCl)

36

1.3.3.4. Atomic force microscopy (AFM)

Highly ordered pyrolytic graphite(HOPG) with dimensions of 3 mm x 3 mm x1 mm was glued with water-insoluble Aralditepoxy onto a teflon disc (diameter: 11 mm).The teflon disc was glued to a steel disc(diameter: 10 mm) which was mounted in theAFM (Fig. 1d). Subsequently, the HOPGwas cleaved with scotch tape, ensuring that thesurfaces separate over the whole area. Thesurface was scanned in buffer solution (10mM Tris-HCl, pH 7.5, 150 mM KCl) with aNanoscope III AFM (Digital instruments,Santa Barbara, Calif.) equipped with a 120µm scanner (J-scanner). A 120 µm longcantilever from Digital instruments (k = 0.38N/m) was used. For recording HOPGtopographs the AFM was operated at minimalforce (<0.2 nN) and 4 Hz scan speed.

To adsorb 2D streptavidin crystals, thefreshly cleaved HOPG was brought intocontact with the monolayer on the surface ofthe drop in the teflon well (Fig. 1). Thesample was kept wet throughout thepreparation procedure. A drop of 30 µlscanning buffer (10 mM Tris-HCl, pH 7.75,200 mM KCl) was immediately added andthe specimen was mounted in the atomic forcemicroscope (AFM). The 120 µm scanner (J-scanner) was used together with oxidesharpened Si3N4 cantilevers from Digitalinstruments with a length of 200 µm (k =0.06 N/m). For imaging the AFM wasoperated at constant force mode applyingminimal forces (<0.2 nN) at a scanning speedof 4-6 Hz. The images were correlationaveraged with the SEMPER image processingsystem (Saxton et al., 1979).

1.3.3.5. Transmission electron microscopy(TEM)

For transmission electron microscopy,samples were prepared on a copper gridcovered by a parlodion film and a carbonlayer. A 10 day old hydrophobic grid was

deposited on the lipid monolayer covering theprotein solution in a teflon well for 1 min.The grid was removed and blotted with filterpaper, and subsequently washed three timeswith double distilled water. The specimen wasnegatively stained twice for 15 sec with0.75% uranyl formate, blotted, and dried in anair stream. Micrographs were taken in aHitachi H7000 TEM at low dose conditionsat 50000 x magnification. The negatives weredigitized with a Leafscan-45 (Leaf SystemsInc., Cupertino, CA) at a stepsize of 20 µm(~0.4 nm at the specimen) and selected areaswere correlation averaged (Saxton &Baumeister, 1982).

1.3.4. Results

1.3.4.1. Hydrophobicity and topography ofHOPG

The hydrophobicity of HOPG wascompared with other surfaces using the sittingdrop diameter method (Karrasch et al., 1993).In comparison with the well-used substrates,mica and glass, HOPG exhibits an increasedhydrophobicity (Table 1). As indicated by thedifference in drop diameter, the difference in

Table 1 . Determination of the hydrophobicity ofvarious substrates by measuring the diameter of 10 µlnanopure water droplets deposited on each substrate.

Substrate Diameter (mm)

mica † 12.3± 1.9 (n=41)

glass (etched)* 7.5 ± 0.7 (n=50)

glass (washed)† 7.4 ± 1.3 (n=66)

glass (washed)* 7.4 ± 0.7 (n=50)

glass (untreated)† 4.9 ± 0.1 (n=60)

glass (silanized)* 4.1 ± 0.2 (n=40)

HOPG † 3.7 ± 0.2 (n=54)

teflon † 3.1 ± 0.1 (n=70)

spherical drop 2.67 (V = 4/3πr3)

* (Karrasch et al., 1993)† This work

37

hydrophobicity between HOPG and silanizedglass is significant. Indeed, streptavidincrystals on a biotinylated monolayer couldnot be successfully adsorbed to the silanizedcoverslip, whereas they could be transferredto HOPG reproducibly.

The surface of freshly cleaved HOPGimaged in buffer solution (10 mM Tris-HCl,pH 7.5, 150 mM KCl) revealed large smoothterraces (Fig. 2 a). The atomically flat terraceshad dimensions up to 2 µm, providing largeareas over which adsorbed specimens can beimaged. Scanning a flat area of 200 nm,which is about the scan range for imagingbiomolecules at high resolution, a roughness(rms) of 0.02 nm was measured. The carbonatoms of the HOPG could be seen ashexagonally ordered arrays with a latticeconstant of 0.23 ± 0.03 nm (Fig. 2 a, inset).The HOPG substrate mounting protocoldescribed above (Materials and Methods,Fig.1 d) and used for all measurements in thiswork, was thus good enough to achieveatomic resolution.

1.3.4.2. Crystallization of streptavidin onbiotin-lipid monolayer

Of many soluble proteins crystallized onlipid monolayers (Brisson et al., 1994),streptavidin is the best-studied simple systemto test the suitability of HOPG as a substrateand develop a protocol for the preparation ofsuch crystals for AFM. A lipid monolayercontaining biotin lipids was formed on a dropof streptavidin solution and incubated to allowthe adsorption and crystallization of theprotein (Fig. 1). The appearance of 2D-crystals was inspected by TEM of negativelystained samples and AFM. Many crystals ofvarying size (80-1500 nm) and shape werefound by TEM. Crystals imaged by AFMwere in general smaller (50-400 nm),suggesting that crystals adsorbed to theHOPG surface may be disrupted by theunevenness of the substrate. By lowering thespeed in approaching the HOPG substrate tothe monolayer surface (Fig. 1c), and bycleaving the HOPG with the scotch tape at anangle >120˚, bigger crystals (up to 3 µm)

Figure 2. a) Height image of HOPG recorded in buffer solution (10 mM Tris-HCl, pH 7.5, 150 mM KCl) atminimal force and a scan speed of 4 Hz, showing terraces with atomically flat surfaces (scale bar: 400 nm; full grayscale: 5 nm); Inset: Fourier filtered image of HOPG atoms scanned at minimal force and a scan speed of 12 Hz (scalebar: 5 Å; full gray scale: 2Å). b) Deflection image of a) (scale bar: 400 nm; full gray scale: 1 nm). c) Height imageof streptavidin 2D crystals on biotinylated lipid monolayers scanned in buffer solution (10 mM Tris-HCl, pH 7.6,150 mM KCl) using minimal force and 4.3 Hz scan speed, showing the HOPG terraces (arrows) overlayed bycrystalline patches of irregular shape (scale bar: 300 nm; full gray scale: 6 nm); asterisk indicates high plateau. d)Deflection image of c) (scale bar: 300 nm; full gray scale: 1 nm).

38

were found apparently adsorbed to theatomically flat surfaces of the terraces (Fig.2c, d). Selection of high terraces for scanningcontributed to the quality of high resolutionimaging (see Fig. 2c).

1.3.4.3. AFM of streptavidin crystals

The overview AFM micrograph (Fig. 2c,d) shows many crystalline patches ofstreptavidin 2D-crystals over a scan range of2.3 µm. The arrows indicate the edges ofHOPG terraces that are clearly visible asdiscontinuities between areas of crystallinestreptavidin. Further, breaks between crystalpatches on the terraces themselves may

indicate that the transfer was influenced bythe unevenness of the substrate.

The different crystalline patches typicallyexhibit different lattice orientations. Betweenthe crystalline patches the protein was notordered and not well resolved. Imagingstreptavidin crystals at medium magnification(Fig. 3a, b) allows single proteins missingwithin the crystal lattice to be seen (Fig. 3a,arrow 1) and proteins floating away from thecrystal edges on the lipid monolayer (Fig. 3a,arrow 2). The crystal in the center and the leftof Fig. 3a had continuous lattice lines,although there were big defects (~100 * ~30nm) within the crystal. The crystalline patchon the right top of Fig. 3a reveals a different

Figure 3. a) Height image of streptavidin 2D crystals showing defects of single missing proteins (arrow 1) andsquare shaped edges with loosened proteins (arrow 2) scanned in buffer solution (10 Tris-HCl, pH 7.2, 20 mM KCl)(scale bar: 100 nm; full gray scale: 6 nm). b) Deflection image of a) (scale bar: 100 nm; full gray scale: 0.3 nm). c)Height image of the streptavidin 2D crystal of the central region of a) at higher magnification (scale bar: 50 nm; fullgray scale: 6 nm), arrows correspond to those in a); Inset: Power spectrum of c) (circle marks spot of 7th order;resolution: 0.83 nm). d) Average over 7 different AFM images using different tips and scan angles (square indicatesthe unit cell with dimensions of a = b = 8.2 ± 0.2 nm, g = 88 ± 2˚; indexed according to Avila-Sakar & Chiu, 1996;full gray scale: 1 nm).

39

orientation in comparison to the crystal in thecenter of the image. Since the appearance ofthese two crystals was the same tipasymmetries can be excluded. At highermagnification (Fig. 3c) the central region ofthe big streptavidin crystal is imaged at ~0.8nm lateral resolution (see diffraction pattern,inset of Fig. 3c). The high signal to noiseratio of the AFM allows every single proteinand missing proteins to be resolved (Fig. 3c,arrow 1).

As it is known from other studies (Avila-Sakar & Chiu, 1996; Hendrickson et al.,1989), the streptavidin crystals showed aC222 point group symmetry with unit celldimensions of a = b = 82 Å and g = 90 ˚ withtwo tetramers per unit cell (Fig. 3d). In AFMimages we found unit cell dimensions of a =b = 84 ± 2 Å and g = 88 ± 2 ˚ (Fig 3c (inset),d). The height of the crystalline patches abovethe lipid monolayer were measured as 4.65 ±0.3 nm (n=20).

1.3.4.4. TEM of streptavidin crystals

To examine distortions and other effectsthe adsorption of the streptavidin crystals toHOPG introduce, the AFM images were

compared with low dose electron micrographsof negatively stained streptavidin crystals(Fig. 4). Features up to ~2 nm were resolved(see diffraction pattern, inset Fig. 4a). Theunit cell containing two tetramers were foundto have dimensions of a = b = 82 ± 2 Å and g= 90 ± 2 ˚. In the average from EM (Fig. 4b)each tetramer shows four densities of aboutequal intensity. These correspond to the twolarge and two small protrusions in the AFMimage (Fig. 3d).

1.3.5. Discussion

In this work we introduce a novelpreparation method for high resolution AFMof 2D crystals grown by the lipid monolayermethod (Uzgiris & Kornberg, 1983). Thestreptavidin-biotin lipid system was used astest sample. The fact that we only takeadvantage of hydrophobic interactionsbetween the fatty acyl chains of the lipid andthe pure carbon of the HOPG, indicates thegenerality of this approach. Interestingly, thedifference in hydrophobicity (Table 1)between silanized glass (drop diameter: 4.1 ±0.2 nm) and HOPG (drop diameter: 3.7 ± 0.2nm) is crucial for the adsorption of a lipid

Figure 4. a) Low dose electron micrograph of a negatively stained streptavidin 2D-crystal (scale bar: 100 nm);Inset: Power spectrum of a) (circle marks spot of 3rd order; resolution: 2.59 nm). b) Average of picture a) (squareindicates the unit cell with dimensions a = b = 8.2 ± 0.2 nm, g = 90 ± 2˚).

40

monolayer prepared as described in this work(Fig. 1). HOPG is a layered material breakinginto different layers on cleavage. This can beseen in the AFM images as terraces ofvarying height and size (up to 2 µm in width)(Fig. 2a, b). Thus, HOPG appears uneven atlow magnification, but each terrace isatomically flat, as evidenced by the imaging ofindividual carbon atoms (Fig. 2a; inset). Thisalso indicates that the substrate is stableenough for high resolution AFM. Feedbackloop gains of the instrument can therefore beset high enough for imaging atoms as well asbiological samples adsorbed to the substrate.The surface features on the adsorbedstreptavidin crystals are not influenced byabnormalities in the underlying substratesupport (Fig. 2, 3). Streptavidin 2D crystalsare found to preferentially adsorb to the highplateaus of the HOPG substrate (Fig. 2c, d),which is probably a result of how the contactbetween HOPG and the monolayer isestablished during the transfer procedure.

Since streptavidin-biotin is a well knownsystem, different attempts have been made toachieve molecular resolution in the AFM toinvestigate the biotin binding with the protein.A lipid bilayer was transferred to mica usingthe Langmuir-Blodgett (LB) and theLangmuir-Schaefer technique. While the firstlayer consisted only of DPPE (dipalmitoylphosphatidyl ethanolamine), the second layerwas a lipid mixture of biotin-DPPE andDMPE (dimyristoyl phosphatidylethanolamine). Streptavidin was added, butthe protein did not form highly ordered arraysand molecular resolution could not beachieved (Weisenhorn et al., 1992). Anotherwork reports on the adsorption of streptavidinto a support that was biotinylated afterphotoactivation of well defined regions.Although the biotinylated sites showed highsurface corrugations, only granular featureswith a diameter of 30 nm could be seen inhigh magnification images (Mazzola &Fodor, 1995). Furthermore, the biotin-streptavidin interaction was studied by

measuring the rupture force of a single biotin-streptavidin bond. To this end, the AFM tipwas biotinylated and subsequentlystreptavidin was adsorbed to it. This tip wasapproached to a biotinylated agarose bead andretracted. Free biotin binding sites on thestreptavidin attached to the tip bound thebiotinylated bead and the measured forcecurve exhibit multiple peaks separated by 160± 20 pN, the break force of a single bond(Florin et al., 1994).

With this new preparation methodstreptavidin can be imaged with the AFM atsubmolecular resolution. Proteins within the2D crystals have a high lateral stability asthey are supported in the crystallographicpacking arrangement. As a consequence theyare better resolved than those tetramersfloating away from a crystal edge. The slightdeviation of the lattice parameters measuredon the 2D crystal (a = b = 84 ± 2 Å and g =88 ± 2 ˚) from the literature data may eitherbe the result of drift and distortion of theraster scan, or may be caused by the transferof the sample. The height measured on thestreptavidin crystals (4.65 ± 0.3 nm) underappropriate ion conditions (Müller & Engel,1997) compares favorably to the thickness ofthe molecule derived from the atomiccoordinates (4.3 nm; (Hendrickson et al.,1989)). It is reasonable to assume that theLC-part (CH2)6 of the biotin-LC-DPPE,which reduces the steric hindrance forstreptavidin binding, is the reason of thisdifference, indicating a good correspondencebetween the X-ray data and the AFM heightanalysis.

The unit cell parameters (a = b = 82 ± 2 Åand g = 90 ± 2 ˚) determined by negativestain electron microscopy are similar to thosefound by cryo-EM (Avila-Sakar & Chiu,1996). The small difference between the EM-and AFM-derived unit cell dimensions (seeprevious paragraph) may be due to the strongadsorption to the HOPG, or to the commonlyobserved slight drift in the AFM. The formeris also manifested in the breakup of the

41

crystal patches (Fig. 2c), compared to electronmicrographs where the contiguous crystalsare bigger (Fig. 4a). In contrast to the fourequal density peaks in the negative stainelectron microscopy average representing aprojection through the negative stain envelope(Fig. 4b), the AFM data shows a cleardifference in height and in shape of thesubunits (Fig. 3d). This difference in theappearance of the streptavidin tetramer couldbe correlated with a surface model derivedfrom the X-ray data. Two subunits (right topand left bottom of the molecule) facing thebiotinylated lipid monolayer are resolved assmall protrusions, while the two subunitsfacing towards the tip and exposing the freebiotin binding sites (left top and right bottomof the molecule) appear as high protrusions(Fig 3 d). This interesting feature that can beseen in AFM topographs is outlined in Fig. 5.The two free biotin pockets of each tetrameron the top surface of the streptavidin crystalcan be used to bind other biotinylatedproteins, providing a nanoscale matrix forimmobilizing proteins. The flatness of theHOPG surface is essential to minimizeundulations in the crystal, which would alsoaffect the imaging of, bound proteins. Thenature of a protein layer bound to the crystalsurface is in principle determined by the sizeand shape of the protein. Proteins of similar

size to that of the streptavidin tetramer orsmaller may form a regular packing followingthe lattice of the streptavidin crystal. Largerproteins would likely arrange in differentorientations, but because of the high signal-to-noise ratio of the AFM, may be studied assingle particles.

About fifteen proteins have beencrystallized into two-dimensional arrays onplanar lipid films (Brisson et al., 1994).While some proteins crystallize on a lipidmonolayer of charged lipids, a specificinteraction is an advantage for binding. Theuse of Nickel-chelating lipids to bind proteinswith a histidine-tag is a promising step toestablish a general procedure to crystallizewater soluble proteins on lipid monolayers, asdemonstrated with a his-tagged reversetranscriptase (Kubalek et al., 1994).Combining a general procedure forcrystallization of water soluble proteins on alipid monolayer and a general preparationmethod for these specimen for the AFM, apromising avenue is now available for furtherstudies on proteins under native conditions.

1.3.6. Acknowledgment

The work was supported by the MauriceE. Müller foundation of Switzerland, theSwiss National Foundation for Scientific

Figure 5. Composite of the TEM average (left side), the AFM average (right side) and a surface contour modelderived from the atomic coordinates (lstp.pdb, Protein Databank) (middle). Because of the C222 symmetry of thetetrameric streptavidin molecule, subunits on left top and right bottom are exposed to the AFM tip, while subunitsright top and left bottom are more concealed. The EM projection map resolves four equal density peaks, while theAFM surface reveals the differences in access to the subunits from one side of the crystal.

42

Research (grant 4036-44062 to A. Engel), theSwiss Priority Project for Micro and NanoSystem Technology, and the FrenchINSERM.

1.3.7. References

Avila-Sakar, A. J. & Chiu, W. (1996).Visualization of ß-sheets and side-chainclusters in two-dimensional periodicarrays of streptavidin on phospholipidmonolayers by electron crystallography.Biophys. J. 70, 57-68.

Bailey, S. W. (1984). Classification andstructures of the micas. MineralogicalSociety of America.

Binnig, G., Quate, C. F. & Gerber, C. (1986).Atomic force microscope. Phys. Rev. Lett.56, 930-933.

Brisson, A., Olofsson, A., Ringler, P.,Schmutz, M. & Stoylova, S. (1994). Two-dimensional crystallization of proteins onplanar lipid films and structuredetermination by electron crystallography.Biol. Cell 80, 221-228.

Darst, S. A., Ahlers, M., Meller, P. H.,Kubalek, E. W., Blankenburg, R., Ribi, H.O., Ringsdorf, H. & Kornberg, R. D.(1991). Two-dimensional crystals ofstreptavidin on biotinylated lipid layersand their interactions with biotinylatedmacromolecules. Biophys. J. 59, 387-396.

Engel, A., Schabert, F., Müller, D. J. & Henn,C. (1995). Imaging membrane proteins intheir native environment with the atomicforce microscope. Kluwer, Germany.

Florin, E.-L., Moy, V. T. & Gaub, H. E.(1994). Adhesion forces betweenindividual ligand-receptor pairs. Science264, 415-418.

Green, M. (1975). Avidin. Adv. Protein chem.29, 85-143.

Hendrickson, W. A., Pähler, A., Smith, J. L.,Satow, Y., Merritt, E. A. & Phizackerley,R. P. (1989). Crystal structure of corestreptavidin determined frommultiwavelength anomalous diffraction ofsynchrotron radiation. Proceedings of theNational Academy of Sciences, U.S.A. 86,

2190-2194.

Jap, B. K., Zulauf, M., Scheybani, T., Hefti,A., Baumeister, W., Aebi, U., Engel, A.(1992). 2D crystallization: from art toscience. Ultramicroscopy 46, 45-84.

Karrasch, S., Dolder, M., Schabert, F.,Ramsden, J. & Engel, A. (1993). covalentbinding of biological samples to solidsupports for scanning probe microscopyin buffer solution. Biophys. J . 65, 2437-2446.

Kubalek, E. W., Grice, S. F. J. & Brown, P.O. (1994). Two-dimensionalcrystallization of histidine-tagged, HIV-1reverse transcriptase promoted by a novelnickel-chelating lipid. Journal of StructuralBiology 113, 117-123.

Mazzola, L. T. & Fodor, S. P. A. (1995).Imaging biomolecule arrays by atomicforce microscopy. Biophys. J. 68, 1653-1660.

Moore, A. W. (1973). Chemistry and physicsof carbon. Marcel Dekker Inc., New York.

Mou, J., Sheng, S., Ho, R. & Shao, Z. (1996).Chaperonins GroEL and GroES: viewsfrom atomic fore microscopy. Biophys. J.71, 2213-2221.

Müller, D. J., Amrein, M. & Engel, A. (1997).Adsorption of biological molecules to asolid support for scanning probemicroscopy. Journal of Structural Biology119, 172-188.


Müller, D. J., Schabert, F. A., Büldt, G. &Engel, A. (1995). Imaging purplemembranes in aqueous solution atsubnanometer resolution by atomic forcemicroscopy. Biophys. J. 68, 1681-1686.

Ohler, M., Baruchel, J., Moore, A. W., Galez,P. & Freund, A. (1997). Directobservation of mosaic blocks in highlyoriented pyrolytic graphite. NuclearInstruments and Methods in PhysicsResearch B 129, 257-260.

Saxton, W. O. & Baumeister, W. (1982). The

43

correlation averaging of a regularlyarranged bacterial cell envelope protein. J.Microsc. 127, 127-138.

Saxton, W. O., Pitt, T. J. & Horner, M.(1979). Digital image processing: Sempersystem. Ultramicroscopy 4, 343-354.

Uzgiris, E. E. & Kornberg, R. D. (1983).Two-dimensional crystallization techniquefor imaging macromolecules, with anapplication to antigen-antibody-complement complexes. Nature 301, 125-129.

Wagner, P., Hegner, M., Kernen, P., Zaugg,F. & Semeza, G. (1996). Covalentimmobilization of native biomoleculesonto Au(111) via N-hydroxysuccinimideester functionalized selfassembledmonolayers for scanning probemicroscopy. Biophys. J. 70, 2052-2066.

Weisenhorn, A. L., Schmitt, F.-J., Knoll, W.& Hansma, P. K. (1992). Streptavidinbinding observed with an atomic forcemicroscope. Ultramicroscopy 42-44,1125-1132.

44

2. Application of highresolution AFM

45

2. Application of high resolution AFM

2.1. High resolution AFM topographs of the Escherichia coli waterchannelaquaporin Z

Simon Scheuring1, Philippe Ringler1, Mario Borgnia2, Henning Stahlberg1, Daniel J.Müller1, Peter Agre2, Andreas Engel1,3

1M.E.Müller Institute for Structural Biology, Biozentrum, University of Basel, CH-4056 Basel,Switzerland2Dept. of Biological Chemistry, John Hopkins University School of Medicine, 725 N Wolfe St.,Baltimore, MD 21205, USA

2.1.1. Abstract

Aquaporins form a large family ofmembrane channels involved inosmoregulation. Electron crystallography hasshown monomers to consist of six membranespanning a-helices confirming sequencebased predictions. Surface exposed loops arethe least conserved regions, allowingdifferentiation of aquaporins. Atomic forcemicroscopy was used to image the surface ofaquaporin Z, the water channel of Escherichiacoli. Recombinant protein with an N-terminalfragment including 10 histidines was isolatedas tetramer by Ni affinity chromatography,and reconstituted into two-dimensionalcrystals with p4212 symmetry. Smallcrystalline areas with p4 symmetry werefound as well. Imaging both crystal typesbefore and after cleavage of the N-termini,allowed the cytoplasmic surface to beidentified; a drastic change of the cytoplasmicsurface accompanied proteolytic cleavage,while the extracellular surface morphologydid not change. Flexibility mapping andvolume calculations identified the longestloop at the extracellular surface. This loopexhibited a reversible force inducedconformational change.

2.1.2. Introduction

Aquaporins are ubiquitous membranechannels in bacteria, fungi, plants andanimals. They are highly specific for water orsmall uncharged hydrophilic solutes and areinvolved in osmoregulation. Hydropathyanalysis of the first sequenced members ofthis family indicated six membrane spans andtwo unusually long loops (Gorin et al., 1984;Preston & Agre, 1991). Meanwhile more thanone hundred and sixty genes have beensequenced, almost all having the highlyconserved NPA motifs within these loops(Heymann & Engel, 1999). The role of theunique triplets remains to be established.Approximately half of these channel proteinsare exclusively water selective and do notallow the permeation of small or chargedsolutes. Other channels facilitate the passageof small hydrophilic molecules such asglycerol or urea (Ishibashi et al., 1997a).Passage of ions through waterchannelproteins have been reported, and recent datasuggest this to be regulated by pH (Yasui etal., 1999).

The best characterized waterchannel isaquaporin-1 (AQP1) from human red bloodcells (Agre et al., 1993). Three dimensional(3D) density maps of this protein have beenestablished to a resolution of 6 Å by three

47

groups (Walz et al. 1997; Cheng et al., 1997;Li et al., 1997). These maps show a bundle ofsix highly tilted transmembrane helices thatsurround a central density formed by the longloops. These results confirm the hourglassmodel of Jung et al. (1994), who proposed a

frame of six a-helices that houses the twoNPA motif carrying loops which fold backinto the membrane to form a highly specificchannel. The water flow per channel wasfound to be the same in two dimensional (2D)crystals of AQP1 as in erythrocyte ghosts, 3x 109 water molecules per channel andsecond (Zeidel et al., 1992; Walz et al.,1994b).

In Escherichia coli a waterchannel hasbeen identified by homology cloning(Calamita et al., 1995). Sequence analysis ofthis bacterial channel, AqpZ, revealed asignificant homology to AQP1. RecombinantAqpZ bearing a histidine tag has beenproduced and shown to be active (Borgnia etal., 1999). 2D crystals with sizes ranging upto 5µm have been assembled from thisrecombinant AqpZ tetramers by dialysis of aprotein-lipid-detergent mixture (Ringler et al.,1999). The 3D map of negatively stainedpreparations revealed the same packingarrangement as in AQP1 2D crystals, p4212,while the 8 Å projection map from vitrifiedunstained preparations showed a strikingsimilarity to the AQP1 and the major intrinsicprotein (MIP) projection maps (Ringler et al.,1999; Walz et al., 1995; Hasler et al., 1998).

We have used the atomic force microscope(AFM) (Binnig et al., 1986) to measure thesurface topography of AqpZ crystals in anative environment. As previously reported,this method allows protein surfaces to beimaged at subnanometer resolution (Schabertet al., 1995; Müller et al., 1995b; Mou et al.,1996; Scheuring et al., 1999). Topographs ofAqpZ crystals bearing an N-terminal polyhistidine tag to allow rapid isolation exhibiteda floppy protrusion related to the N-terminusthat could be eliminated by proteolysis. Thisallowed the sidedness of AqpZ to beidentified. After cleavage of the histidine tag,topographs had a lateral resolution of 7 Å anda vertical resolution of 1 Å. The surfacetopography could be related to the loopspredicted from the sequence by hydropathyanalysis.

Figure 1 . a) AFM topograph of AqpZ 2Dcrystals adsorbed to mica (recorded in buffersolution: 17 mM Tris-HCl, pH 7.2, 150 mMKCl). Double and multi layered areas can clearly bedistinguished from single layer crystals by theirhigher appearance (scale bar: 2 µm; full gray scale:30 nm). b) Section analysis along the white line inimage a). The 2D crystals show a uniform height of57 ± 4 Å (n = 45). Double layered areas appear asplateaus twice as high as the single layered crystalsheets (vertical scale bar: 50 Å). c) AFM topographof reconstituted AqpZ (recorded in buffer solution:17 mM Tris-HCl, pH 7.2, 150 mM KCl,). Adensely packed vesicle containing crystalline areaswith p4 symmetry adsorbed onto a crystal sheetwith p4212 symmetry (scale bar: 100 nm; full grayscale: 20 nm). Top inset: average of the sheet withp4212 symmetry (scale bar: 50 Å);. Bottom inset:average of the crystalline areas with p4 symmetrywithin the densely packed vesicle (scale bar: 50 Å).

48

2.1.3. Results

In the presence of 25 mM MgCl2 the 2Dcrystals of AqpZ adsorbed to mica withoutwrinkles and were thus suitable for high-resolution imaging. To this end, theadsorption buffer was exchanged with therecording buffer that was adjusted toelectrostatically balance the van der Waalsforces (Müller, 1999a). The overview in Fig.1a demonstrates the flatness of the 2Dcrystals whose thickness was found to be 57± 4 Å (n = 45; Fig. 1b).

At higher magnification the square latticebecame distinct (Fig. 1c, main frame, top).The unit cell dimensions were a = b = 95 ± 2Å, in excellent agreement with results fromelectron microscopy (Ringler et al., 1999).Correlation averaging revealed a unit cellhousing two tetramers in oppositeorientations with respect to the membraneplane (p4212 packing; Fig. 1c, top right). Thehigh signal-to-noise ratio of the AFM alsoallowed high resolution imaging on denselypacked vesicles which only exhibited small

crystalline areas comprising about 30tetramers arranged with p4 symmetry (Fig.1c, main frame, bottom) The p4 crystals hadunit cell dimensions of a = b = 72 ± 2 Å andhoused a single tetramer (Fig. 1c, bottomright).

The recombinant AqpZ crystallized has anN-terminal fragment of 26 aminoacids,containing a trypsin cleavage site at arg26 anda 10 his-tag at aminoacid positions 2 to 12(Fig. 2a; Borgnia et al., 1999). As aconsequence, a total of 104 aminoacids,including 40 histidines, protruded from thecytoplasmic side of each tetramer. Thesepeptides produced a strong signal in theAFM, resulting in a 20 ± 2 Å high peak (Fig.1c, 3a), the exact position and appearance ofwhich depended critically on the force appliedto the stylus, the scan speed and the directionof the scan (compare Fig. 1c and 3a). Thisextreme flexibility prevented the resolution ofsubstructure. To prove that the largeprotrusion observed indeed arose from the N-terminal domain, crystals were treated withtrypsin (see materials & methods) to cleave

Figure 2 . a) Aminoacid sequence model of AqpZ showing the six membrane spanning helices derived fromhydropathy analysis. Trypsin cleavage sites are located on arg26 and arg230. b) Silver stained SDS - polyacrylamide10% (w/v) gel. Columns from left to right: (M) marker: 97.4 kDa, 66.2 kDa, 42.7 kDa, 31.0 kDa; (1) AqpZsolubilized in 2% OG; (2) AqpZ crystals after overnight trypsin treatment. The AqpZ band is broadened indicatingthat a minor part of the protein remained undigested. The two diffuse bands in the low molecular weight region(below 30 kDa) correspond to trypsin and the cleaved N-terminal fragments; (3) AqpZ-10his crystals. The faint bandsat high molecular weight (~ 200 kDa) in all three lanes arise from specific aggregates.

49

off this flexible end domain as documented inFig. 2b. The digested crystals exhibited astriking change; instead of an ill definedprotrusion of 20 Å height, the cytoplasmicside now showed four distinct protrusionseach with a height of 3.5 ± 0.4 Å above thelipid bilayer (Fig. 3b). In contrast, theextracellular side was neither changed inshape nor height by the trypsin treatment(Fig. 3c; see Table 1).

The extracellular side was not sensitive totrypsin digestion, however it underwent areversible conformational change when theforce applied to the tip was increased duringimaging. At minimal force (~ 80 pN) eachAqpZ subunit showed three majorprotrusions, probably related to the loopsconnecting the membrane spanning helices onthe extracellular side. (Fig. 4a, c). Recording asecond image of the same areas with a forceincreased by + 80 pN, a drasticconformational change was observed(compare figures 4a and b, and 4c and d). Theextracellular AqpZ surface reversibly changedits rather circular appearance into a left-handed windmill, which still protruded out ofthe membrane by 7 Å (Fig. 4e). This forceinduced conformational change was notinfluenced by the trypsin treatment, asillustrated by comparing the topographs ofdigested (Fig. 4c and d) and undigested AqpZcrystals (Fig. 4a and b). The digestedcytoplasmic surface did not show the sameforce dependence, the minor force inducedconformational change was barely noticeable(Figs. 4c and d, insets). The standarddeviation (SD) map of 289 densely packedsingle tetramers (such as shown in Fig. 4a)recorded at minimal force was calculated to

identify the flexible regions of theextracellular surface (Müller et al., 1998). Asdisplayed in Fig. 4f, one region exhibited apronounced SD, while the rest yielded highlyreproducible heights (SD ≤ 0.2 Å).Interestingly, this flexible region alsoexhibited the major force inducedconformational change (compare left andright tetramer in Fig. 4f).

Sequence based structure predictionpostulates AqpZ to be a protein consisting ofsix transmembrane helices connected by threeloops on the extracellular side and two loopson the cytoplasmic side, as well as twocytoplasmic termini (see Fig. 2a). Inagreement with this, on imaging at minimalforce, three protrusions were found on theextracellular surface of the AqpZ monomer,one close to the 4-fold symmetry center, andone small and one elongated protrusion at theperiphery (Fig. 5a). Volume calculations (seematerials and methods) on the protrusionclose to the 4-fold symmetry center resultedin 1278 ± 150 Å3. The small peripheralprotrusion yielded a volume of 984± 134 Å3,while the elongated protrusion had a volumeof 3187 ± 528 Å3, the larger SD reflectingthe flexibility of this region. The singleprotrusion observed per monomer on thedigested cytoplasmic surface had a volume of1222 ± 144 Å3 (Fig. 5b). Since the terminiare removed (Fig. 2a, b), this protrusion isexpected to house loops b and d.

2.1.4. Discussion

Significant progress in the understandingof imaging conditions and the interpretationof topographs recorded with the atomic force

Table 1. Heights of surface protrusions of AqpZ before and after trypsin treatment.

undigestedp4 crystal

undigestedp4212 crystal

digestedp4212 crystal

Extracellularprotrusion 7.3 ± 0.9 Å 6.7 ± 1.0 Å 7.0 ± 0.9 Å

Cytoplasmicprotrusion 18.6 ± 1.8 Å 20 ± 2.0 Å 3.5 ± 0.4 Å

50

microscope (AFM) has allowed the surface

microscope (AFM) has allowed the surfacetopography of bacteriorhodopsin to becorrelated with the helix connecting loops to alateral resolution of 5 Å (Müller et al.,1999b). Here we have used this technology tostudy the surface of AqpZ, the first bacterialwaterchannel identified (Calamita et al.,1995). Its overexpression, isolation and 2Dcrystallization have recently been described(Borgnia et al., 1999; Ringler et al., 1999).

2D crystals adsorbed firmly and withoutfolds or wrinkles to freshly cleaved mica in ahigh ionic strength buffer (Müller et al.,1997). Subsequent change to a bufferadjusted to compensate for van der Waalsinteractions allowed their height to bemeasured accurately (Müller & Engel 97).The result, 57 ± 4 Å, compares favorably withthe height previously reported for AQP1, 58± 3 Å (Walz et al., 1996).

The p4212 crystals of AqpZ with unit celldimensions of a = b = 95 ± 2 Å have similarlattice parameters to those found for AQP1(unit cell dimensions: 96 ± 2 Å) (Walz et al.,1996). However, the p4 crystals of AqpZ(unit cell dimensions of a = b = 72 ± 2 Å) aremore loosely packed than 2D crystals of MIPwhich also exhibit p4 symmetry (unit celldimensions: a = b = 64 ± 1 Å; Hasler et al.,1998). This suggests that more lipidmolecules are interspersed between the AqpZtetramers within the small crystalline areas inthe densely packed vesicles than in the highlyordered MIP crystals.

In experiments with undigested AqpZcrystals the N-terminal tail of 26 aminoacidsprevented the visualization of substructureson the cytoplasmic surface. The four weaklyordered protruding peptides, each containing10 histidines, have a total mass of ~12kDa,and appeared to interact strongly with thesilicon nitride tip. Consequently thecytoplasmic surfaces exhibited peaks ofapproximately 20 Å height that wereinfluenced by the scan direction (Fig. 1c, 3a).In spite of these large protrusions, theintervening extracellular sides could be

Figure 3 . a) AFM topograph (recorded in buffersolution: 17 mM Tris-HCl, pH 7.2, 150 mM KCl) ofan AqpZ-10his 2D crystal with p4212 symmetryrecorded using minimal force (scale bar: 500 Å; fullgray scale: 30 Å). Inset: relief view (tilt: 85°) of a 300Å square, raw data; extracellular surfaces are marked bycircles. b) AFM topograph (recorded in buffer solution:17 mM Tris-HCl, pH 7.2, 150 mM KCl) of an AqpZ2D crystal with p4212 symmetry after trypsintreatment recorded using minimal force (scale bar: 500Å; full gray scale: 10 Å). Inset: relief view (tilt: 85°) ofa 300 Å square, raw data; extracellular surfaces aremarked by circles. c) 3D reconstruction of the trypsincleavage process (see Fig. 2a) observed on thecytoplasmic surface. On digestion this surface changesshape and height drastically in the location of the N-terminal his-tags (right top: undigested state, leftbottom: digested state).

51

resolved as tetramers with a central hole andfour major protrusions.

After trypsin treatment both theextracellular and the cytoplasmic surfaces

Figure 4. a) AFM topograph of AqpZ-10his densely packed in a vesicle, scanned in buffer solution (17 mM Tris-HCl, pH 7.2, 150 mM KCl) at minimal force (~ 80 pN) (scale bar: 250 Å; full gray scale: 18 Å). A smallcrystalline area with p4 symmetry is outlined. Inset: average of a). The white square indicates the unit cell (a = b =72 ± 2 Å) which houses 1 tetramer. b) AFM topograph of the same area as a) recorded in buffer solution (17 mMTris-HCl, pH 7.2, 150 mM KCl) applying an additional force of +80 pN to the tip during scanning. The outlinedarea corresponds to the area marked in a) (scale bar: 250 Å; full gray scale: 18 Å). Inset: average of c). The whitesquare indicates the unit cell (a = b = 72 ± 2 Å) which houses 1 tetramer. c) AFM topograph of trypsin treated AqpZ2D crystals with p4212 symmetry recorded in buffer solution (10 mM Tris-HCl, pH 7.5, 150 mM KCl) usingminimal force (~ 80 pN) (scale bar: 250 Å; full gray scale: 20 Å). The outlined area shows a pronounced latticedistortion. Inset: average of c). The white square indicates the unit cell (a = b = 95 ± 2 Å) which houses 2 tetramers.d) AFM topograph of the same area as c) recorded in buffer solution (10 mM Tris-HCl, pH 7.5, 150 mM KCl)applying an additional force of +80 pN to the tip during scanning. Note the same lattice irregularity as in c) (scalebar: 250 Å; full gray scale: 20 Å). Inset: average of d). The white square indicates the unit cell (a = b = 95 ± 2 Å)which houses 2 tetramers. e) 3D reconstruction illustrating the effect observed on the extracellular surface when theimaging force is increased by + 80 pN during scanning. (right top: minimal force, left bottom: + 80 pN). f)Comparison of the averages of the extracellular surface at minimal force (left), the standard deviation (SD) map(middle), and the extracellular surface average at a additional force of + 80 pN (right) (full image sizes: 72 Å). Theoutlined regions in the in the middle image represent a SD ~ 0.7 Å. These regions correspond to the four elongatedperipheral protrusions in the minimal force average which are strongly displaced in the average gained from imagesrecorded at + 80 pN.

52

could be imaged at high resolution on p4212crystals (Fig. 3b and 4c). Not only did theextracellular side show a consistent surfaceappearance before and after trypsin treatment(compare central particles in the insets of Fig.4a and c), but also its height over the lipidbilayer was not affected by the proteolyticcleavage (Table 1). In contrast, only a smallcytoplasmic protrusion remained after trypsindigestion; the surface appearance changedfrom a single large protrusion to a welldefined tetramer protruding by only 3.5 Åover the lipid bilayer (Fig. 3c). Thus removalof the large flexible cytoplasmic domains bytrypsin allowed the unambiguous assignmentof the AqpZ sidedness. This elegant methodwill be useful to determine the sidedness ofother recombinant membrane proteins withthe AFM.

The visualization and identification ofdistinct loops on a protein surface with theAFM has been shown for bacteriorhodopsin(Müller et al., 1995b; Müller et al., 1999b).

The reliability of topographic data acquiredhas been documented by direct comparisonwith results from electron microscopy(Karrasch et al., 1994) and X-raycrystallography (Schabert et al., 1995). Inaddition, a force induced reversibleconformational change observed on thecytoplasmic surface of bacteriorhodopsin hasindicated the location of the longestcytoplasmic loop (Müller et al., 1995a).Finally, structural features having thestrongest variability in the raw data areenhanced in standard deviation maps allowingflexible regions of proteins to be identified(Müller et al., 1998). All three methods wereapplied to assess the surface topography ofAqpZ. Firstly, we have identified twoapproximately equal, small protrusions (984Å3, 1278 Å3), and one large, elongatedprotrusion (3187 Å3) at the extracellularsurface of AqpZ (Fig. 5). Secondly byapplying an additional force of ~ 80 pN to thetip during scanning, the large protrusion was

Figure 5. Proposed assignment and borderlines between adjacent loops of the AqpZ tetramer.The x, y and z scalings used for three dimensional integration of the protruding volumes are derived from highresolution AFM topographs (full image sizes: 95 Å). a) Extracellular surface exposing protrusions 1, 2, and C withvolumes of 984 ± 134 Å3, 1278 ± 150 Å3, and 3187 ± 528 Å3 respectively. Protrusions 1 and 2 correspond toloops A and E. The similarity of their volumes prevents an unambiguous assignment. b) Cytoplasmic surfacehaving only one defined protrusion housing loops B and D with a volume of 1222 ± 144 Å3.

53

found to undergo a drastic conformationalchange (Fig. 4). Because this change wascompletely reversible, the same area could bescanned many times, and the changemonitored repeatedly. Thirdly, the SD mapcalculated from 289 AqpZ tetramers indicatedthat the large protrusion was the most flexibleone (Fig. 4). Taken together, this suggeststhis large surface domain located at theperiphery of the tetramer with a total volumeof about 3200 Å3 to be related to loop C ofAqpZ, predicted to comprise 26 aminoacids(3700 Å3) (Fig. 2a).

In spite of the limitations related to the tipgeometry, an estimate of the volume ofcontoured protrusions is possible (Fritzsche& Henderson, 1996; Schneider S. W. et al.,1998). We have tested the algorithms used todelineate protrusions and calculate theirvolumes on surface topographs ofbacteriorhodopsin (Müller et al., 1999b) andporin (Schabert et al., 1995, Müller & Engel,1999), and found them to produce correctvolumes within an experimental error of < 20%. The volumes accordingly calculated forthe AqpZ extracellular protrusions are closeto those expected from the sequence predictedloops, although the small loops A (predictedas 6 aminoacids, 900 Å3) and E (9aminoacids, 1300Å3) cannot beunambiguously assigned. However, theposition of the two loops remaining on thecytoplasmic surface after digestion could bedefined. Importantly, the unambiguousdetermination of sidedness achieved using theAFM together with digestion experiments, isessential for the interpretation of structuraldata obtained from AqpZ crystals by cryoelectron microscopy.

In conclusion, structural information onthe surface exposed loops of a membranechannel has been acquired with the AFM.These data are complementary to thoseobtained by electron crystallography whichprovides mainly the 3D density map of themembrane resident part of the protein.Because the AFM is operated under

physiological conditions, function relatedstructural changes can be directly assessed(Müller & Engel, 1999). Such experimentswill be relevant to study the recentlydiscovered regulation of water channels bypH (see Engel et al., accompanying paper).

2.1.5. Materials and methods

2.1.5.1. Reconstitution

Large 2D crystals were produced bydialysis as described by Ringler et al. (1999).Recombinant AqpZ isolated by Ni-affinitychromatography (Borgnia et al., 1999) wassolubilized in 2% n-octyl-b-D-glucoside(OG) at a concentration of 0.5 mg/ml andmixed with dimyristoyl phosphatidylcholine/palmitoyl oleoyl phosphatidyl choline(DMPC/POPC) (1/1) solubilized in 2% OGto a final lipid-to-protein ratio of 0.3. Themixture was dialyzed against a detergent freebuffer (20 mM citric acid, pH 6.0, 200 mMNaCl, 100 mM MgCl2, 3 mM NaN3, 10%glycerol) for three days. 2D crystals werewashed by centrifugation and resuspended inadsorption buffer (see below).

2.1.5.2. Trypsin digestion

For cleavage of the N-terminal fragment,AqpZ-10his crystals were incubated overnightat 4oC with trypsin (1 mg/ml). The crystalswere then washed twice throughcentrifugation in a table centrifuge (HeraeusBiofuge A) at 5000 rpm for 3 minutes withsubsequent removal of the supernatant andaddition of fresh buffer solution. Aftertrypsin treatment samples were investigatedby sodium dodecyl sulfate-polyacrylamidegel electrophoresis (SDS-PAGE) using 10%(w/v) acrylamide gels.

2.1.5.3. Atomic force microscopy

Mica prepared as described (Schabert &Engel, 1994) was used as support. For each

54

experiment the mica was freshly cleaved withscotch tape and imaged in 30-50µl ofadsorption buffer (10 mM Tris-HCl, pH 7.5,150 mM KCl, 25 mM MgCl2) to check thecleavage quality, 3µl of protein crystalsolution (0.1 mg/ml) were then injected intothe adsorption buffer drop on the micasurface. After 2h the sample was carefullyrinsed with recording buffer (10 mM Tris-HCl, pH 7.5, 150 mM KCl). The recordingbuffer was optimized to achieve highresolution as described (Müller et al., 1999a).Imaging was performed with a commercialNanoscope III AFM (from DigitalInstruments, Santa Barbara, CA, USA)equipped with a 120 µm scanner (J-scanner)and oxide-sharpened Si3N4 cantilevers with alength of 120µm (k=0.1 N/m) (fromOlympus Ltd., Tokyo, Japan). The AFM wasoperated in contact mode applying constantlyminimal forces (<0.2 nN) at a scan frequencyof 4-6 Hz. The instrument was calibratedusing layered crystals of MoTe2 as describedpreviously (Müller & Engel, 1997).

2.1.5.4. Image processing

AFM images of crystalline areas wereprocessed by correlation averaging using theSEMPER image processing system (Saxtonet al., 1979). Tetramers of disordered regionswere aligned and averaged by a single particleaveraging protocol, and standard deviationmaps were calculated to assess thereproducibility of height measurements and toidentify flexible regions (Müller et al., 1998).To calculate the volumes of distinctprotrusions in such an average, a SEMPERroutine was used to determine the borderlinesbetween adjacent elevations. To this end, a 3 x3 pixel box was scanned over the image andall the central pixels having a lower gray valuethan any of the 4 pairs of adjacent pixels wereidentified as borderline. Outlined areas wereevaluated by integrating all the pixelsdetermined within this boundary using unitcell dimensions and heights determined with

the AFM.


The authors would like to thank Dr. S. A.Müller for proofreading and discussing themanuscript. We also thank D. Fotiadis and L.Hasler for inspiring discussions. The workwas supported by the Maurice E. Müllerfoundation of Switzerland, the Swiss NationalFoundation for Scientific Research (grant4036-44062 to A. Engel), the Swiss PriorityProject for Micro and Nano SystemTechnology, and the French INSERM (to P.R.).

2.1.7. References

Agre, P., Preston, G., Smith, B., Jung, J.,Raina, S., Moon, C., Guggino, W. &Nielsen, S. (1993). Aquaporin CHIP: thearchetypal molecular water channel. Am. J.Physiol. 265, F436-476.

Binnig, G., Quate, C. F. & Gerber, C. (1986).Atomic force microscope. Phys. Rev. Let.56, 930-933.

Borgnia, M., Kozono, D., Calamita, G.,Nielsen, S., Maloney, P. C. & Agre, P.(1999). Functional reconstitution andcharacterization of E. coli Aquaporin-Z. J.Mol. Biol., in press.

Calamita, G., Bishai, W., Preston, G.,Guggino, W. & Agre, P. (1995).Molecular cloning and characterization ofAqpZ, a waterchannel from Escherichiacoli. J. Biol. Chem. 270, 29063-29066.

Cheng, A., van Hoek, A. N., Yeager, M.,Verkman, A. S. & Mitra, A. K. (1997).Three-dimensional organization of ahuman water channel. Nature 387, 627-630.

Engel, A., Fujiyoshi, Y. & Agre, P. (1999).The importance of aquaporin waterchannel protein structures. The EMBOJournal, this issue

Fritzsche, W. & Henderson, E. (1996).Volume determination of humanmetaphase chromosomes by scanningforce microscopy. Scan. Micr. 10, 103-108

55

Gorin, M. B., Yancey, S. B., Cline, J., Revel,J.-P. & Horwitz, J. (1984). The majorintrinsic protein (MIP) of the bovine lensfiber membrane: Characterization andstructure based on cDNA cloning. Cell39, 49-59.

Hasler, L., Walz, T., Tittmann, P., Gross, H.,Kistler, J. & Engel, A. (1998). Purifiedlens major intrinsic protein (MIP) formshighly ordered tetragonal two-dimensionalarrays by reconstitution. J. Mol. Biol. 297,855-864.

Heymann, J. B. & Engel, A. (1999).Aquaporins: phylogeny, structure andphysiology of water channels. News inPhysiological Sciences, in press .

Ishibashi, K., Kuwahara, M., Gu, Y.,Kageyama, Y., Tohsaka, A., Suzuki, F.,Maruma, F. & Sasaki, S. (1997a). Cloningand functional expression of a new waterchannel abundantly expressed in the testispermeable to water, glycerol and urea. J.Biol. Chem. 272, 20782-20786.

Jung, J., Preston, G., Smith, B., Guggino, W.& Agre, P. (1994). Molecular structure ofthe water channel through aquaporinCHIP. The hourglass model. J. Biol.Chem. 269, 14648-14654.

Karrasch, S., Hegerl, R., Hoh, J. H.,Baumeister, W. & Engel, A. (1994).Atomic Force Microscopy ProducesFaithful High-Resolution Images ofProtein Surfaces in an AqueousEnvironment. Proc Natl. Cad Sci USA 91,836-838.

Li, H., Lee, S. & Jap, B. K. (1997). Moleculardesign of aquaporin-1 water channel asrevealed by electron crystallography. Nat.Struct. Biol. 4, 263-265.

Mou, J., Czajkowsky, D. M., Sheng, S., Ho,R. & Shao, Z. (1996). High resolutionsurface structure of E. coli GroESoligomer by atomic force microscopy.FEBS Lett. 381, 161-164.

Müller, D. J., Amrein, M. & Engel, A. (1997).Adsorption of biological molecules to asolid support for scanning probemicroscopy. J. Struct. Biol. 119, 172-188.

Müller, D. J., Büldt, G. & Engel, A. (1995a).Force-induced conformational change of

bacteriorhodopsin. J. Mol. Biol. 249, 239-243.

Müller, D. J., Schabert, F. A., Büldt, G. &Engel, A. (1995b). Imaging PurpleMembranes in aqueous solution atsubnanometer resolution by atomic forcemicroscopy. Biophys. J. 68, 1681-1686.


Müller, D. J. & Engel, A. (1999). pH andvoltage induced structural changes ofporin OmpF explain channel closure. J.Mol. Biol. 285, 1347-1351.

Müller, D. J., Fotiadis, D. & Engel, A.(1998). Mapping flexible protein domainsat subnanometer resolution with the AFM.FEBS Lett. 430, 105-111.


Müller, D. J., Sass, H.-J., Müller, S., Büldt, G.& Engel, A. (1999). Surface structures ofnative bacteriorhodopsin depend on themolecular packing arrangement in themembrane. J. Mol. Biol. 285, 1903-1909.

Preston, G. M. & Agre, P. (1991). Isolationof the cDNA for erythrocyte integralmembrane protein of 28 kilodaltons:Member of an ancient channel family.Proc. Natl. Acad. Sci. USA 88, 11110-11114.

Ringler, P., Borgnia, M., Stahlberg, H., Agre,P. & Engel, A. (1999). Structure of thewater channel AqpZ from Escherichia colirevealed by electron crystallography. J.Mol. Biol., in press.

Saxton, W. O., Pitt, T. J. & Horner, M.(1979). Digital Image Processing: SemperSystem. Ultramicr. 4, 343-354.

Schabert, F. A. & Engel, A. (1994).Reproducible acquisition of Escherichiacoli porin surface topographs by atomicforce microscopy. Biophys. J. 67, 2394-2403.

56


Scheuring, S., Müller, D. J., Ringler, P.,Heymann, J. B. & Engel, A. (1999).Imaging streptavidin 2D crystals onbiotinylated lipid monolayer at highresolution with the atomic forcemicroscope. J. Microsc. 193, 28-35.

Schneider, S. W., Larmer, J., Henderson, R.M. & Oberleithner, H. (1998). Molecularweights of individual proteins correlatewith molecular volumes measured byatomic force microscopy. Pflugers Arch.435, 362-367.

Walz, T., Hirai, T., Murata, K., Heymann, J.B., Mitsuoka, A., Fujiyoshi, Y., Smith, B.L., Agre, P. & Engel, A. (1997). The 6Åthree-dimensional structure of aquaporin-1. Nature 387, 624-627.

Walz, T., Smith, B., Zeidel, M., Engel, A. &Agre, P. (1994b). Biologically active two-

dimensional crystals of aquaporin CHIP.J. Biol. Chem. 269, 1583-1586.

Walz, T., Tittmann, P., Fuchs, K. H., Müller,D. J., Smith, B. L., Agre, P., Gross, H. &Engel, A. (1996). Surface topographies atsubnanometer resolution reveal asymmetryand sidedness of aquaporin-1. J. Mol.Biol. 264, 907-918.

Walz, T., Typke, D., Smith, B. L., Agre, P. &Engel, A. (1995). Projection map ofaquaporin-1 determined by electroncrystallography. Nat. Struct. Biol. 2, 730-732.

Yasui, M., Kwon, T. H., Knepper, M. A.,Nielsen, S. and Agre, P. (1999).Aquaporin-6: An intracellular vesicle waterchannel protein in renal epithelia. Proc.Nat. Ac. Sc., U.S.A. 96, 5808-5813.

Zeidel, M. L., Ambudkar, S. V., Smith, B. L.& Agre, P. (1992). Reconstitution offunctional water channels in liposomescontaining purified red cell CHIP28protein. Biochem. 31, 7436-7440.

57

2.2. High resolution AFM topographs of Rubrivivax gelatinosus light-harvesting complex LH2

Simon Scheuring1, Francoise Reiss-Husson2, Andreas Engel1, Jean-Louis Rigaud3 andJean-Luc Ranck3

1M.E. Müller Institute for Microscopy at the Biozentrum, University of Basel, Klingelbergstrasse70, CH-4056 Basel, Switzerland.2Centre génétique Moléculaire, UPR-CNRS 2167, 91198 Gif-sur-Yvette Cedex, France3Institut Curie, UMR-CNRS 168 and LRC-CEA 8, 11 rue Pierre et Marie Curie, 75231 ParisCedex 05, France

2.2.1. Abstract

Light harvesting complexes 2 (LH2) areantenna proteins in the bacterialphotosynthetic apparatus and are built of αβ-heterodimers containing 3bacteriochlorophylls and 1 carotenoid each.We have used atomic force microscopy(AFM) to investigate reconstituted LH2 fromRubrivivax (Rvi.) gelatinosus, which has a C-terminal hydrophobic extension of 21 aminoacids residues on the α-subunit. Highresolution topographs revealed a nonamericorganization of the regularly packedcomplexes incorporated into the membrane inboth orientations. Native LH2 showed onesurface which protruded by ~ 14 Å and onewhich protruded by ~ 5 Å from themembrane. Thermolysin cleaved protein witha shorter C-terminus of the α-subunits had aheight of ~ 9 Å and a different appearance ofprotruding structures, allowing theassignment of the periplasmic surface. Minorcontaminants were imaged as rare large rings(~ 120 Å diameter) in interaction with LH2.Their diameter and rotational power spectrasuggested these rings to be hexadecamericLH1 complexes.

2.2.2. Introduction

Purple photosynthetic bacteria provide anideal experimental system for describing the

assembly and organization of photosyntheticsystems. Absorption of light and itsconversion into chemical energy is performedby highly organized transmembrane pigment-protein complexes: the light harvestingcomplexes 2 (LH2) and 1 (LH1), and thereaction center (RC). In addition to the wealthof information from biochemistry, molecularbiology and spectroscopy, structural data haveadvanced our understanding of the singlecomponents of the bacterial photosyntheticapparatus and have provided a model for itsfunctional mechanism. According to thismodel, the light energy is initially trapped bythe peripheral antenna LH2 complexes andthe excitation energy is transferred to LH1complexes which are closely associated withthe RC, forming the so-called core complex.The subsequent photon induced redoxreaction in the RC causes charge separationacross the membrane which in turn initiates acyclic electron transport and the formation ofan electrochemical proton gradient across themembrane (Papiz et al., 1996).

Different types of LH antenna complexeshave been isolated from various species ofpurple bacteria and their structures solved athigh resolution. The basic motif is aheterodimer consisting of two smallpolypeptides α and β, which both span themembrane once as transmembrane α-helices.Together they form a heterodimer to whichthree bacteriochlorophyll a pigments

59

molecules (Bchl) and one carotenoid are noncovalently bound. Ring-like associations ofthe αβ-heterodimers have been reported fromstructural analysis of three-dimensional (3D)and two-dimensional (2D) crystals.McDermott et al. (1995) first solved thestructure of LH2 from Rhodopseudomonas(Rps.) acidophila by X-ray crystallographyshowing an αβ-nonamer shaped as a hollowcylinder, formed by the membrane spanninghelices of the 2 subunits, with the α-subunitinside and β outside. Atomic structures from3D crystals of Rhodospirillum (Rp.)molischianum LH2 revealed an octamericarrangement (Koepke et al., 1996). Electroncrystallographic data were also acquired on2D crystals of LH2's from Rhodovulum(Rhv.) sulfidophilum and Rhodobacter (Rhb.)

sphaeroides, both exhibiting a nonamericorganization (Montoya et al., 1995; Walz etal., 1998). The structure of the LH1 complex,however, is still unknown at atomicresolution. An electron crystallographicprojection structure of Rhodospirillum (Rs.)Rubrum LH1 at 8.5 Å showed a similar ring-like structure, in this case consisting of 16αβ-heterodimers (Karrasch et al., 1995). Thediameter of the LH1 complex suffices toaccommodate a reaction center within thering, a model confirmed by analyses of 2Dcrystals of LH1-RC core complexes fromRhb. sphaeroides (Stahlberg et al., 1998;Walz & Ghosh, 1997).

In spite of the wealth of informationavailable on the individual components of thephotosynthetic apparatus of photosynthetic

Figure 1. a) Topology model of Rvi. gelatinosus LH2 (strain S1; gene sequence accession number AF312921).The α- (consisting of 71 aminoacids) and the β-polypeptide (consisting of 51 aminoacids) cross the membrane onceeach. The boxed regions in the sequence correspond to a-helical stretches in the Rps acidophila LH2 structure,aminoacids surrounded by circles in dark gray are predicted to be transmembrane (using TMpred5), those surroundedby circles in light gray are predicted to be α-helical (using GOR46). The triangle indicates the thermolysin cleavagesite on the C-terminus of the a-polypeptide. The arrow points towards the center of the nonamer in the membraneplane. b) Silver stained SDS-polyacrylamide 10 % (w/v) gel. Columns from left to right: (1) marker: 97.4 kDa,66.2 kDa, 42.7 kDa, 31.0 kDa; (2) Thermolysin (37 kDa), (3) Crystals of native LH2, band at 115 kDa (4),Crystals of thermolysin treated LH2, band at 82 kDa. c) Absorption spectra of native and thermolysin cleaved LH2reconstituted into lipid bilayers. The bacteriochlorophyll and carotenoid absorption spectra do not change uponcleavage of the C-terminus of the a-subunit (native LH2: black line, digested LH2: gray line). The absorption spectradocument the native state of the protein: arrows 1, 2, 3: carotenoid peaks at 460 nm, 488 nm and 517 nmrespectively; arrow 4: Qx peak at 595 nm; arrow 5: Qy peak at 802 nm; arrow 6: Qy peak at 859 nm.

60

bacteria, their supramolecular organization inthe membrane is poorly understood. Modelsof interaction between LH2, LH1 and the RChave been established to explain the highlyefficient energy transfer of a photon trappedby an LH2 antenna complex to LH1 andfinally the RC as reviewed by Kuhlbrandt(1995). The structural model forunderstanding light capture and transfer is aclose packing of ring-like structures withLH2 complexes surrounding the LH1/RCcore complex. However, the size of the ringsof the different LH’s complexes as well as anopen organization of LH1 complexes aroundRCs allowing an efficient quinone/quinoltransfer is still under debate (Jungas et al.,1999; Francia et al., 1999; Loach, 2000; Freseet al., 2000).

Here we present a structural study of theLH2 complex from Rvi. gelatinosus. Ascompared to LH2's from other species, themost distinct feature is related to the sequenceof the α-polypeptide chain, which has a C-terminal extension that is 21 amino acidslonger than the C-terminus of α-subunitsfrom Rps. acidophila. As proposed on thebasis of its hydrophobicity, this extensioncould be folded into a second transmembranehelix, leading to an α-subunit spanning themembrane twice in a hair pin structure(Brunisholz et al., 1994). Recent electronmicroscopy studies of Rvi. gelatinosus LH2reconstituted in 2D crystals providedprojection maps of negatively stained and inice embedded samples revealing a cylindricalring-like assembly of Rvi. gelatinosus LH2complexes with a 9-fold symmetry with innerand outer diameters similar to those reportedfrom X-ray models of the nonameric Rps.acidophila LH2 (Ranck et al., in preparation;McDermott et al., 1995). Comparison of theprojection maps from 2D crystals of nativeand truncated LH2, in which the C-terminalextension has been digested by thermolysin,did not allow any extra density to be detectedinside or outside the nonameric ring.Therefore, the location of the C-terminal

extension could not be identified.We have used the atomic force microscope

(AFM) (Binnig et al., 1986) to measure thesurface topography of LH2 of Rvi.gelatinosus in a native environment and tolocate the C-terminal extension. As previouslyreported, this method allows protein surfacesto be imaged at subnanometer resolution(Schabert et al., 1995; Müller et al., 1995;Scheuring et al., 1999; Fotiadis et al., 2000;Engel & Müller, 2000). Topographs of LH2complexes had a lateral resolution of ~ 8 Åand a vertical resolution of ~ 1 Å. Anonameric organization of the αβ-heterodimers with one strongly (~ 14 Å) andone weakly (~ 5 Å) membrane protrudingsurface was found. Imaging membranesreconstituted with native and digested LH2allowed localization of the C-terminalaminoacids of the α-subunit of Rvi.gelatinosus and consequently the assignmentof the periplasmic surface. In addition,together with LH2 nonamers (diameter ~ 50Å), occasional large rings (diameter ~ 120 Å)were imaged. Their dimensions and rotationalpower spectra suggest these rings to be aminor LH1 contaminant. These resultsdemonstrate the potential of AFM for theassessment of the in vivo photosyntheticsystem.

2.2.3. Results

Rvi. gelatinosus LH2 protein is built bythe α- and the β-polypeptides consisting of71 and 51 aminoacids. Sequence alignmentwith LH2 of Rps. acidophila (usingClustalW4) and hydropathy analysis led tothe topology model displayed in figure 1a.The thermolysin cleavage site has beendetermined by HPLC and mass spectroscopy(Ranck et al., in preparation) and is indicatedby the triangle. The limited proteolysis isillustrated by the silver stained gel in Figure1b which was obtained after a mildsolubilization of 2D crystals reconstitutedfrom native and from digested LH2. Native

61

LH2 complexes run with an apparent Mw of115 kDa, while digested complexes run at82 kDa, reflecting the removal of 20aminoacids from each α-subunit of thecomplex. Sharp bands document the specificand effective cleavage of the protein. Fig. 1cdisplays the absorption spectra of the nativeLH2 (black line) and truncated LH2 (grayline) after reconstitution into lipid bilayers.Both spectra show the characteristicabsorption bands of carotenoids at 460 nm,488 nm and 517 nm, the Qx Bchl band at595 nm, and the Qy bands at 802 nm and859 nm, indicating native Bchls andcarotenoid binding sites (see also Ranck et

al., in preparation).

4link: http://www.ch.embnet.org/software/ClustalW.html

5link: http://www.ch.embnet.org/software/TMPRED_form.html

6link: http://pbil.ibcp.fr/cgi-bin/npsa_automat.pl/page=npsa_gor4.html

Large 2D arrays of LH2 complexes wereproduced by detergent removal from amicellar solution containing the LDAO-purified protein supplemented with eggphosphatidyl choline and octylthioglucoside,a detergent which has been recently reportedto significantly increase the size of thereconstituted 2D crystals (Chami et al., in

Figure 2 . Height measurements of LH2 2D crystals adsorbed to mica in buffer (10 mM Tris-HCl, pH 7.2,150 mM KCl, 25 mM MgCl2) and imaged under physiological conditions (10 mM Tris-HCl, pH 7.2, 150 mMKCl). a) AFM topograph of double and multi layered areas can clearly be distinguished from single layer crystals bytheir higher appearance (full image size: 4 µm; full gray scale: 30 nm) b) Section analysis along the white line inimage a). The 2D crystals show a uniform height of 64.5 ± 2.8 Å (n = 46) (vertical scale bar: 100 Å). c) AFMtopograph of a LH2 sheet containing particles in up-and-down crystalline packing in the center and crystalline areasexposing only the lower surface on the edges (full image size: 700 nm; full gray scale: 10 nm). d) Section analysisalong the white line in image c). The two different surface types are clearly visible: While the central region shows acharacteristic section analysis for up-and-down packing with strong surface corrugation, the edge areas appear smoothand show a height of only 57 Å above the mica support (vertical scale bar: 50 Å). e) Medium magnification imageof a crystalline sheet of LH2 rings. At this magnification the ring-structure of the complexes is already clearlyvisible. The crystals show coherence only over small regions and lattice displacements of half a unit cell are frequent.While the ring-structure of the high side is well visible in the crystalline areas, the lower circles can better be seenwithin the crystal defects (bottom right) (full image size: 400 nm; full gray scale: 4 nm). f) Section analysis alongthe white line in image e). The height of LH2 rings can directly be measured from such section analysis of raw data(vertical scale bar: 20 Å).

62

preparation; see also material and methods).The use of bio-beads as detergent removingagent allowed production of the reconstitutedmembranes in few hours, avoiding anydenaturation of the LH2 complexes.

Using a high ionic strength buffer (10 mMTris-HCl, pH 7.2, 150 mM KCl, 25 mMMgCl2) 2D-crystals of native and predigestedRvi. gelatinosus LH2 could be firmly attachedto the mica AFM support (Schabert & Engel,1994). To acquire high resolution topographs,the buffer was carefully exchanged (10 mM

Tris-HCl, pH 7.2, 150 mM KCl) to achieveslightly repulsive force curves (data notshown) on the reconstituted 2D crystals(Müller et al., 1999). Under such conditionsthe 2D-crystals of both sample typesremained attached to the mica for hours. Theheight measured for the lipid bilayer, 41.9Å (native sample), 41.2 Å (digested sample)and for the crystals, 64.5 Å (native sample),60.7 Å (digested sample) indicated the C-terminus protrudes from the membrane (seeTable 1, Fig. 2a and 2b).

Table 1. LH2 dimensions measured with the AFM under physiological conditions

native LH2 sample digested LH2 sample

average SDa nb average SDa nb

Overall thickness

lipid bilayer: 41.9 Å ± 2.5 Å 37 41.2 ± 2.5 Å 23

up-and-down 2D-crystal: 64.5 Å ± 2.8 Å 46 60.7 ± 1.6 Å 33

one sided 2D-crystal: 60.4 Å ± 1.4 Å 11 57.4 ± 2.5 Å 81

dimensions of large protrusions:

height (lipid-top)c 13.9 Å ± 1.7 Å 72 8.9 ± 1.4 Å 110

height (ring center-top)d 7.6 Å ± 2.0 Å 73 5.3 ± 0.9 Å 33

diameter of tope 49.0 Å ± 0.5 Å 3* 49.0 ± 0.5 Å 3*

volume / subunit 3297Å3 ± 517 Å3 # 1765 Å3 ± 353 Å3 #

dimensions of small protrusions:

height (lipid-top)c 5.6 Å ± 0.8 Å 152 5.4 ± 0.5 Å 34

height (ring center-top)d 5.6 Å ± 1.0 Å 24 5.3 ± 0.8 Å 137

diameter of tope 54.0 Å ± 0.5 Å 3* 54.0 ± 0.5 Å 3*

volume / subunit 1001 Å3 ± 161 Å3 # 952 Å3 ± 182 Å3 #

a standard deviation.b number of measurements.c height difference between the top of the ring and the lipid surface.d height difference between the top and the center of the ring.e ring diameter measured at its top after rotational averaging.* average of 3 independent AFM topographs at different nominal magnifications scaled with respect to the unitcell dimensions found in cryo electron microscopy.# volumes and deviations calculated taking standard errors in height measurements and standard deviation signal trough single particle averaging into account.

63

At higher nominal magnification the ring-structure of LH2 became distinct, revealingthe two different sides of the transmembranecomplex (Fig. 2e), which are regularly packedin alternative orientations. Unit cellscontaining 2 rings have dimensions ofa = 82 Å, b = 133 Å, and γ = 90°. Arrows infigure 2e indicate defects in the crystal latticewhere lower rings were more clearly visiblethan within the up-and-down packing betweenthe strongly protruding surfaces. Such areaswere also identified at lower magnification(see arrows in Fig. 2c, Fig. 2d) and allowedtheir height to be measured, 60.4 Å (nativesample), 57.4 Å (digested sample), assummarized in Table 1.

In agreement with electron microscopystudies (Ranck et al., in preparation) highresolution topographs (Fig. 3a, c, e, g)revealed that Rvi. gelatinosus shares thecircular nonameric organization of LH2subunits with Rps. acidophila (McDermott et

al., 1995), Rhb. sphaeroides (Walz et al.,1998) and Rhv. sulfidophilum (Montoya etal., 1995). The reconstituted LH2 complexes(Fig. 3a, b, c, d) protruded by 13.9 ± 1.7 Åfrom one side of the membrane (Table 1, Fig.3a, b), but only by 5.6 ± 0.8 Å on the otherside (Table 1; Fig. 3c, d). These lowerprotrusions could be best imaged close to theedges of sheets (see arrows in Fig. 2c), whereunidirectionally inserted complexes werefound to assemble into a hexagonal closestpacking with unit cell dimensions ofa = b = 76 Å, and γ = 60°. The membranesreconstituted in presence of digested LH2revealed a very similar overall organization ofthe protein (Fig. 3e, f, g, h). The weaklyprotruding (5.4 ± 0.5 Å) surface showed anidentical surface appearance compared to thenative sample (compare Fig. 3g, h with Fig.3c, d). Relatively large areas with complexesexposing their lower surface to the tip allowedthe right handed twist of the protruding

Figure 3. High resolution raw data AFM topographs and corresponding averages (all in 15° tilted representationand corresponding heights). Averages were gained through a single particle alignment routine an 9-fold symmetrized.a) Raw data topography of the strongly (~ 14 Å) protruding surface of the native LH2 complex. The 9-foldsymmetry is in the raw data visible (scale bar: 100 Å; full gray scale: 20 Å). b) Average of image a) (scale bar:20 Å; full gray scale: 14 Å). c) Raw data topography of the weakly (~ 5 Å) protruding surface of the native LH2complex (scale bar: 100 Å; full gray scale: 20 Å). d) Average of image c) (scale bar: 20 Å; full gray scale: 6 Å).e) Raw data topography of the strongly (~ 9 Å) protruding surface of the digested LH2 complex (scale bar: 100 Å;full gray scale: 20 Å). f) Average of image e) (scale bar: 20 Å; full gray scale: 9 Å). The loss of protrudingstructure compared to b) localizes the C-terminal position of the a-subunit. g) Raw data topography of the weakly(~ 5 Å) protruding surface of the digested LH2 complex (scale bar: 100 Å; full gray scale: 20 Å). h) Average ofimage g). The proteolysis had no influence on the topography of the weakly protruding surface (scale bar: 20 Å; fullgray scale: 6 Å).

64

structure to be resolved (Fig. 3g, h). The bandshift in the gel (Fig. 1b) documents acomplete cleavage of the C-terminus of the α-subunit. Accordingly, a change in the stronglyprotruding surface was observed in thedigested sample. Its height was reduced to8.9 Å over the lipid bilayer (Fig. 3e, f), andthe cleaved protrusions appeared broader thanthe uncleaved protrusions in figure 3a. Thus,the volume integral over the protrudingstructure changed from ~ 3300 Å3 to~ 1800 Å3 (Table 1), compatible with the lossof the C-terminus.

As displayed in figure 3, the lower ringshad a larger diameter than the higher rings.Taking unit cell dimensions determined bycryo-electron microscopy for scaling the

AFM data, the respective diameters were54.0 Å for the lower ring and 49.0 Å for thehigher ring (see Table 1).

In membranes reconstituted from a LH2preparation in which the hydroxyapatitecolumn purification step was omitted (seematerials and methods), a few giant rings(diameter of ~ 120 Å) were imaged atsubmolecular resolution surrounded bysmaller LH2 rings (diameter of ~ 50 Å) (Fig.4a). Rotational power spectra of 68 LH2rings and 7 large rings from the same imagesare displayed in figure 4b. The merged powerspectra of the LH2 rings showed a strongsignal corresponding to the 9-fold symmetry(gray line in Fig. 4b). The large rings,however, showed a much stronger intensityfor 8-fold symmetry than for 9-foldsymmetry, together with weak signalscorresponding to 12 and 16-fold symmetries.

2.2.4. Discussion

During the last few years AFM hasbecome a powerful tool in membrane proteinresearch. This progress is the result ofimproved instrumentation as well asoptimized recording conditions (Engel &Müller, 2000). The AFM allows informationto be acquired on the membrane protrudingstructure of single proteins. Such informationis difficult to obtain with cryo-electronmicroscopy where the membrane embeddedparts are preserved, but connecting loops andprotruding termini are often distorted.Furthermore heights of membranes and loopscan be measured accurately in buffer solutionwith the AFM (Müller et al., 1999a), poorlyordered single particles can be recognized andimaged at high resolution (< 10 Å)(Scheuring et al., 1999; Fotiadis et al., 2000;Seelert et al., 2000), and sidednessassignments can directly be obtained fromraw data (Scheuring et al., 1999).

LH2 complexes are rings of αβ-heterodimers, each subunit crossing theplasma membrane of the photosynthetic

Figure 4. a) Raw data AFM topograph showing a~ 120 Å ring surrounded by LH2 nonamers revealingdistances between the complexes of ~ 10 Å (seeDiscussion; scale bar: 100 Å; full gray scale: 20 Å) b)Rotational power spectra of 7 large rings (black line)and 68 small rings (gray line) from images as displayedin figure a). The small LH2 rings display a clear 9-foldsymmetry. The large rings show a strong amplitude onthe 8-fold and a weak on the 16-fold symmetry, whilethe 9-fold and 18-fold symmetry is less pronounced,supporting the assignment of the large rings with thehexadecameric LH1 rings.

65

bacteria once. For Rps. acidophila LH2 theα-polypeptide contains a perpendiculartransmembrane alpha-helical segment, whichconstitutes the inner surface of the ring, whilethe β-polypeptide is a transmembrane helixthat is tilted by ~ 15° and forms the peripheryof the cylinder (McDermott et al., 1995). TheLH2 of Rvi. gelatinosus carries a large C-terminus on the α-polypeptide of about 35amino acids (half of the polypeptide), with ahydrophobic extension of 21 amino acids,protruding out of the membrane on theperiplasmic side (Fig. 1).

High resolution AFM topographs ofreconstituted arrays of LH2 complexesrevealed the nonameric organization of theLH2 rings (Fig. 3), in agreement with anelectron crystallographic study of the same2D crystals, which showed the ring-likenonameric organization of the αβ-heterodimers, with outer and inner diametersof about 66 Å and 30 Å respectively (Rancket al., in preparation). However, theseprojection maps did not show majordifferences between digested and undigestedLH2 complexes, preventing the identificationof the C-terminal extension. Topographs ofmembranes packed with native and digestedLH2, revealing a thermolysin induced changeof height and surface appearance of thestrongly protruding surface, allowedidentification of the C-terminus of the a-polypeptide (Fig. 3). Thus, the surfaceprotruding ~ 14 Å from the membranerepresent the periplasmic side, while thecytoplasmic surface housing the N-terminiprotrudes by only ~ 5 Å. The volume changeof the periplasmic protrusions from~ 3300 Å3 to ~ 1800 Å3 (Fig. 3, Table 1)induced by thermolysin corresponds to acalculated change of 10 aminoacids. This ismuch less than the 20 aminoacids cleaved bythermolysin (Fig. 1a, b), but can be explainedby the flexibility of the uncleaved C-terminusprotruding far out of the membrane surface(14 Å).

The strongly protruding periplasmic

surface which is thought to comprise the 35C-terminal aminoacids of the α-polypeptideand 6 C-terminal aminoacids of the β-polypeptide (Fig.1) has a smaller diameter(~ 49 Å) than the cytoplasmic protrusions(~ 54 Å) housing 11 N-terminal aminoacidsof the α-polypeptide and 14 N-terminalaminoacids of the β-polypeptide (Table 1).Since the α-subunits face the inner diameterof the LH2 rings in Rps. acidophila(McDermott et al., 1995), the moreprotruding structure is expected to form anarrower cylinder, corroborating thesidedness assignment based on proteolyticcleavage.

The high signal-to-noise ratio of AFMtopographs allows single non orderedproteins to be imaged at submolecularresolution (Scheuring et al., 1999; Seelert etal., 2000; Fotiadis et al., 2000). Figure 4ashows a ring with a diameter of ~ 120 Åwhich is in surrounded with LH2 rings withdiameters of ~ 50 Å. Contact distancesbetween large and LH2 rings are identical tothose between LH2 rings, ~ 10 Å.Identification of the large ring as minor LH1contaminants is based on the agreement of itsdiameter with that of LH1 (116 Å, Karrasch etal., 1995), and the merged rotational powerspectrum of seven large rings, revealing amajor peak at 8-fold and minor peaks at 12-and 16-fold symmetry. Fusions of two LH2rings can be excluded, as such rings shouldpossess a strong peak at 9-fold symmetry andshould occur more frequently.

In conclusion, we have imaged the surfacesof the nonameric complexes from Rvi.gelatinosus LH2 with the AFM and identifiedthe periplasmic surface with the C-terminusof the a-polypeptide. The high signal-to-noiseratio of images acquired with the AFMallowed recognition of different proteincomplexes that work together in the nativemembrane. As a next step in understandingthe photosynthetic apparatus nativemembranes will be directly imaged with theAFM under physiological conditions. This

66

will resolve questions related to theoligomerization states of different LHcomplexes in situ and to the open/closedconfiguration of the LH1 rings.

2.2.5. Materials and methods

2.2.5.1. Materials

All phospholipids were of the highestpurity and were purchased from Avanti PolarLipids. N,N- dimethyldodecylamine N-oxide(LDAO, 30 % solution) was from Fluka andn-Octyl-b-D thioglucopyranoside (OTG) wasfrom Sigma. Thermolysin was purchasedfrom Boehringer. Bio-Beads SM2 (25 - 50mesh) from Bio-Rad were extensively washedwith methanol and water before use asdescribed (Levy et al., 1990). All otherreagents were of analytical grade.

2.2.5.2. Isolation, purification andproteolysis of LH2 complex

The light-harvesting LH2 complex wasisolated from photosynthetically grown Rvi.gelatinosus cells (Strain S1) essentially asdescribed previously (Jirsakova et al., 1996),with slight modifications (Ranck et al., inpreparation). Briefly, solubilization of thebroken cells with LDAO was followed bytwo successive chromatographic purificationon DEAE-Sepharose FF and Sepharose CL-6B (Pharmacia) columns. Further purificationwas achieved by chromatography on ahydroxyapatite (Biosepra) column eluted in abuffer containing 10 mM Tris-HCl, pH 8.0,1 mM EDTA and 0.1 % LDAO.

A limited digestion by thermolysin (4 hrincubation time at 22°C, enzyme/LH2 molarratio = 20) was performed on purified LH2 inLDAO solution, and its effect was analyzedby denaturing SDS-PAGE and by MALDI-TOF mass spectroscopy as describedelsewhere (Ranck et al., in preparation).

2.2.5.3. Biochemical and biophysicaltechniques

Protein concentration was determined fromthe absorption at 854 nm using ε = mM-1cm-1 = 382 (Sturgis et al., 1995) and values of12530 Daltons and 10933 Daltons for themolecular weights of the native and thethermolysin cleaved LH2 respectively (Rancket al., in preparation). Absorption spectra ofthe vesicles were recorded on a Cary 2300spectrophotometer.

Vesicles containing native or cleaved LH2were analyzed by SDS gel electrophoresis onsilver stain 10 % acrylamide gels underoxidizing conditions.

2.2.5.4. Reconstitution and 2Dcrystallization

2D crystallization of native and truncatedLH2 complexes were performed as describedpreviously (Ranck et al., in preparation;Chami et al., in preparation). Briefly, purifiedLH2 complexes were diluted to about 0.5 mg/ml in a buffer containing 10 mM Tris-HCl,pH 8.0, 400 mM NaCl, 0.1 % LDAO andsupplemented with 20 mM Octylthioglucoside. Then egg phosphatidyl cholinewas added at lipid-to-protein ratios rangingfrom 0.3 to 0.5 w/w and the micellar solutionallowed to equilibrate for about one hour inthe dark at room temperature. Detergentremoval was performed through threesuccessive additions of 5 mg SM2 Bio-Beadsfor 1 hour each, according to the batchprocedure previously described by Rigaud etal. (1998). After 3 hours of stirring inpresence of polystyrene beads at roomtemperature, the reconstituted material waskept at 4° C for AFM analysis.


Mica prepared as described by Schabert &Engel (1994) was used as support and freshlycleaved before every experiment using scotch

67

tape. To check the cleavage quality the micawas imaged in 30 - 50 µl of adsorption buffer(10 mM Tris-HCl, pH 7.2, 150 mM KCl,25 mM MgCl2). Subsequently 1 µl ofprotein crystal solution (0.1 mg/ml) wasinjected into the adsorption buffer drop on themica surface. After 2 hours the sample wascarefully rinsed with recording buffer(10 mM Tris-HCl, pH 7.2, 150 mM KCl).The recording buffer was optimized toachieve high resolution as described (Mülleret al., 1999a). Imaging was performed with acommercial Nanoscope III AFM (fromDigital Instruments, Santa Barbara, CA, USA)equipped with a 120 µm scanner (J-scanner)and oxide-sharpened Si3N4 cantilevers with alength of 120 µm (k = 0.1 N/m; OlympusLtd., Tokyo, Japan). Sharpest tips adequatefor high resolution imaging were cleaned in1 % SDS and subsequently washed 3 timesin distilled water. The AFM was operated incontact mode applying forces below < 0.2 nNat a scan frequency of 4 - 6 Hz. Theinstrument was calibrated using layeredcrystals of MoTe2 as described previously(Müller & Engel, 1997).

2.2.5.6. Image processing

AFM images were calculated as 15° tiltedsurface representations using Image SXM(Fig. 3a, c, e, g, 4a). Rings were aligned andaveraged by a single particle averagingprotocol using the SEMPER imageprocessing system (Fig. 3b, d, f, h; Saxton etal., 1979).

Volumes in Å3 were directly calculatedusing x, y and z dimensions of AFMtopograph averages, aminoacid numbers werededuced using 1.35 Å3/Da for averageprotein density in secondary protein structure,and 110 Da as average molecular weight ofone aminoacid.


We thank Kitaru Suda for making the gelshown in figure 1b, C. Möller and D.J.Müller for fruitful discussions on the AFMtechnique, and C. Leiniger for her help withthe manuscript. This work was supported bythe Swiss National Foundation for ScientificResearch (grant 4036 - 44062 to A. E.), theMaurice E. Müller Foundation of Switzerlandand the Centre National de la RechercheScientifique (programme PCV to FRH).

2.2.7. References

Binnig, G., Quate, C. F. & Gerber, C. (1986).Atomic force microscope. Phys. Rev. Lett.56, 930-933.

Brunisholz, R. A., Suter, F. & Zuber, H.(1994). Structural and spectralcharacterization of the antenna complexesof Rhodocyclus gelatinosus. Indication ofa hair-pin-like arranged antennaapoprotein with an unusually high alaninecontent. Eur. J . Biochem. 222, 667-675.

Chami, M., Pehau-Arudet, G., Lambert,O.,Ranck, J.L., Levy, D. and Rigaud ,J.L.(2000). Use of octyl-beta-thiogluco-pyranoside in 2D crystallization ofmembrane proteins. J. Struct. Biol. inpreparation

Engel, A. & Müller, D. J. (2000). Observingsingle biomolecules at work with theatomic force microscopy. Nat. struct. biol.7, 715-718.

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

Francia, F., Wang, J., Venturoli, G., Melandri,B., Barz, W. & Oesterheld, D. (1999). Thereaction center-LH1 antenna complex ofRhodobacter sphaeroides contains onePufX molecule, which is involved indimerization of this complex. Biochem. 38,6834-6845.

Frese, R. N., Olsen, J. D., Branvall, R.,Westerhuis, W. H., Hunter, C. N. & vanGrondelle, R. (2000). The long-range

68

supraorganization of the bacterialphotosynthetic unit: A key role for PufX[see comments]. PNAS 97, 5197-202.

Jirsakova, V. & Reiss-Husson, F. (1994). Aspecific carotenoid is required forreconstitution of the Rubrivivaxgelatinosus B875 light harvesting complexfrom its subunit form B820. FEBS Lett.353, 151-4.

Jungas, C., Ranck, J. L., Rigaud, J. L., Joliot,P. & Vermeglio, A. (1999).Supramolecular organization of thephotosynthetic apparatus of Rhodobactersphaeroides. EMBO J. 18, 534-42.

Karrasch, S., Bullough, P. A. & Ghosh, R.(1995). The 8.5 A projection map of thelight-harvesting complex I fromRhodospirillum rubrum reveals a ringcomposed of 16 subunits. EMBO J. 14,631-8.

Koepke, J., Hu, X., Muenke, C., Schulten, K.& Michel, H. (1996). The crystal structureof the light-harvesting complex II (B800-850) from Rhodospirillum molischianum.Structure 4, 581-97.

Kuhlbrandt, W. (1995). Many wheels makelight work. Nature 374, 497-498.

Levy, D., Bluzat, A., Seigneuret, M. &Rigaud, J.-L. (1990). A systematic studyof liposome and proteoliposomereconstitution involving bio-bead-mediatedtriton-X removal. BBA 1025, 179-190.

Loach, P. A. (2000). Supramolecularcomplexes in photosynthetic bacteria.PNAS 97, 5016-5018.

McDermott, G., Prince, S.M., Freer, A.A.,Hawthornthwaite-Lawless, A.M., Papiz,M.Z., Cogdell, R.J., Isaacs, N.W. (1995).Crystal structure of an integral membranelight-harvesting complex fromphotosynthetic bacteria. Nature 374, 517-521.

Montoya, G., Cyrklaff, M. & Sinning, I.(1995). Two-dimensional crystallizationand preliminary structure analysis of lightharvesting II (B800-850) complex fromthe purple bacterium Rhodovulumsulfidophilum. J. Mol. Biol. 250, 1-10.

Müller, D. J. & Engel, A. (1997). The height

of biomolecules measured with the atomicforce microscope depends on electrostaticinteractions. Biophys. J. 73, 1633-1644.

Müller, D. J., Fotiadis, D., Scheuring, S.,Müller, S. A. & Engel, A. (1999a).Electrostatically balanced subnanometerimaging of biological specimens by atomicforce microscopy. Biophys. J. 76, 1101-1111.

Müller, D. J., Sass, H.-J., Mülller, S., Büldt,G. & Engel, A. (1999b). Surfacestructures of native bacteriorhodopsindepend on the molecular packingarrangement in the membrane. J. Mol.Biol. 285, 1903-1909.

Müller, D. J., Schabert, F. A., Büldt, G. &Engel, A. (1995). Imaging purplemembranes in aqueous solutions atsubnanometer resolution by atomic forcemicroscopy. Biophys. J. 68, 1681-1686.

Papiz, M.Z., Prince, S.M., Hawthornthwaite-Lawless, A.M., McDermott, G., Freer,A.A. , Isaacs, N.W., Cogdell, R.J., (1996).A model for the photosynthetic apparatusof purple bacteria. Trends in PlantScience, 1, 198-206

Ranck, J.L., Ruiz, T., Pehaut-Arnaudet, G.,Arnoux,B. and Reiss-Husson, F., (2000).2D structure of the light-harvestingcomplex LH2 from Rv. Gelatinosus:comparison of the native complex and of atruncated form obtained by limitedproteolysis. J. Biol. Chem., inpreparation.

Rigaud, J.L., Lévy, D., Mosser, G., andLambert, O., (1998). Detergent removal byBio-Beads. Eur. Biophys.J., 27, 305-319.

Saxton, W. O., Pitt, T. J. & Horner, M.(1979). Digital image processing: TheSemper system. Ultramicroscopy 4, 343-354.

Schabert, F. A. & Engel, A. (1994).Reproducible acquisition of Escherichiacoli porin surface topographs by atomicforce microscopy. Biophys. J. 67, 2394-2403.

Schabert, F. A., Henn, C. & Engel, A. (1995).Native Escherichia coli OmpF porinsurfaces probed by atomic force

69

microscopy. Science 268, 92-94.

Scheuring, S., Ringler, P., Borgnia, M.,Müller, D. J., Agre, P. & Engel, A. (1999).High resolution topographs of theEscherichia coli waterchannel AqpZ.EMBO J. 18, 4981-4987

Seelert, H., Poetsch, A., Dencher, N., Engel,A., Stahlberg, H. & Müller, D. J. (2000).Proton-powered turbine of a plant motor.Nature 405, 418-419.

Stahlberg, H., Dubochet, J., Vogel, H. &Ghosh, R. (1998). Are the light-harvestingI complexes from Rhodospirillum rubrumarranged around the reaction center in asquare geometry? J. Mol. Biol. 282, 819-31.

Sturgis, J., Jirsakova, V., Reiss-Husson, F.,Cogdell, R. & Robert, B. (1995). Structureand property of bacteriochlorophyllbinding site in peripheral light-harvestingcomplexes of purple bacteria. Biochem.34, 517-523.

Walz, T. & Ghosh, R. (1997). Two-dimensional crystallization of the light-harvesting I-reaction centre photounit fromRhodospirillum rubrum. J. Mol. Biol. 265,107-11.

Walz, T., Jamieson, S. J., Bowers, C. M.,Bullough, P. A. & Hunter, C. N. (1998).Projection structures of threephotosynthetic complexes fromRhodobacter sphaeroides: LH2 at 6 Å,LH1 and RC-LH1 at 25 A. J. Mol. Biol.282, 833-45.

70

3. Combining surface andprojection techniques

71

3. Combining surface and projection techniques

3.1. The aquaporin sidedness revisited

Simon Scheuring1, Peter Tittmann2, Henning Stahlberg1, Philippe Ringler1, MarioBorgnia3, Peter Agre3, Heinz Gross2, and Andreas Engel1

1M.E. Müller Institute for Microscopy at the Biozentrum, University of Basel, Klingelbergstr. 70,CH-4056 Basel, Switzerland2Institute of Applied Physics, ETH Zürich, CH-8093 Zürich, Switzerland3Department of Biological Chemistry and Medicine, Johns Hopkins University School ofMedicine, Baltimore MD 21205-21850, USA

3.1.1. Summary

Aquaporins are transmembrane waterchannel proteins, which play importantfunctions in the osmoregulation and waterbalance of micro-organisms, plants, andanimal tissues. All aquaporins studied to dateare thought to be tetrameric assemblies offour subunits each containing its ownaqueous pore. Moreover, the subunits containan internal sequence repeat forming twoobversely symmetric hemichannels predictedto resemble an hourglass. This uniquearrangement of two highly related proteindomains oriented at 180° to each other posesa significant challenge in the determination ofsidedness. Aquaporin Z (AqpZ) fromEscherichia Coli was reconstituted into highlyordered two-dimensional crystals. They werefreeze-dried and metal-shadowed to establishthe relationship between surface structure andunderlying protein density by electronmicroscopy. The shadowing of some surfaceswas prevented by protruding aggregates.Thus, images collected from freeze-driedcrystals that exhibited both metal-coated anduncoated regions allowed surface reliefreconstructions and projection maps to beobtained from the same crystal. Crosscorrelation peak searches along latticescrossing metal-coated and uncoated regions

allowed an unambiguous alignment of thesurface reliefs to the underlying densitymaps. AqpZ topographs previouslydetermined by AFM could then be alignedwith projection maps of AqpZ, and finallywith human erythrocyte aquaporin-1 (AQP1).Thereby features of the AqpZ topographycould be interpreted by direct comparison tothe 6Å three-dimensional structure of AQP1.We conclude that the sidedness we originallyproposed for aquaporin density maps wasinverted (Walz et al., 1996).

3.1.2. Introduction

To maintain metabolic processes watermolecules must efficiently permeate theplasma membranes of cells in all livingorganisms. Since the diffusion of watermolecules through lipid bilayers has anactivation energy > 10 kcal/mol (Chandy etal., 1997), the existence of specific waterpores was postulated more than 4 decadesago (Sidel & Solomon, 1957). The firstmember of this family termed the aquaporins(Chrispeels & Agre, 1994) and designed byevolution to facilitate water transport, wasidentified by Preston et al. in 1992.Aquaporin sequences share six hydrophobicstretches, which correspond to trans-membrane helices. Two long conserved loops,

73

B and E, connect helices 2 and 3, and 5 and 6,respectively, and accommodate the highlyconserved NPA motifs (Gorin et al., 1984;Preston & Agre, 1991). These loops foldback into the membrane, to form the structureof the pore (Jung et al., 1994). Permeabilitystudies by stopped flow measurementsindicate flow rates in the range of 109 watermolecules per channel and per second, and anactivation energy < 5 kcal/mol (Walz et al.,1994b; Zeidel et al., 1992).

Aquaporin-1 (AQP1) of humanerythrocytes (Agre et al., 1993) is structurallythe best studied aquaporin. Two-dimensional(2D) crystals with two tetramers packed inopposite orientation into a unit cell withdimensions of a = b = 96 Å and g = 90° havebeen reconstituted in the presence of lipids(Walz et al., 1994a). These highly ordered 2Dcrystals diffracted a 300 kV electron beam toat least 3.5 Å resolution (Mitsuoka et al.,1999). A 3D density map at 6 Å resolutioncalculated from projections of samples tiltedwith respect to the electron beam, revealed aright-handed bundle consisting of sixtransmembrane a-helices surrounding acentral density (Walz et al., 1997; Cheng etal., 1997), in agreement with sequence basedstructure prediction. The central densityformed by the long loops B and E hasrecently been resolved as two short helicesprojecting outwards from the center of themonomer which are connected to adjacenthelices by loop regions (Mitsuoka et al.,1999), thus confirming the hour glass model(Jung et al., 1994).

The E. coli waterchannel AqpZ has beenidentified by expression cloning (Calamita etal., 1995) and overexpressed in its nativesystem (Borgnia et al., 1999). This bacterialwaterchannel maintains cell turgor during thevolume expansion of cell division (Calamita etal., 1998). Highly ordered 2D crystals havebeen grown by dialysis of protein-lipid-detergent mixtures (Ringler et al., 1999). Thesquare unit cells with dimensions a = b = 95Å, g = 90°, contained eight monomers and

had p4212 symmetry (Ringler et al., 1999).Cryo electron microscopy provided aprojection map at 7 Å resolution exhibitingthe characteristic features of AQP1 consistentwith the high sequence homology of theaquaporins (Ringler et al., 1999). 2D crystalsassembled from AqpZ bearing an N-terminalfragment of 26 aminoacids containing 10histidines had the same symmetry and unitcell dimensions. AFM studies before andafter removal of this N-terminal fragment withtrypsin, allowed the sidedness of AqpZsurfaces to be unambiguously assigned(Scheuring et al., 1999). The crown-likeextracellular side possesses three protrusionsof 7 Å height per monomer, of which thelargest one was identified as loop C, whichcomprises 26 aminoacids. One protrusion permonomer was visible on the cytoplasmicsurface, probably resulting from loops Band/or D (Scheuring et al., 1999).

In contrast, the sidedness of AQP1 hasbeen determined using surface reliefreconstruction of metal-shadowed AQP1 2Dcrystals before and after digestion withcarboxy peptidase Y (Walz et al., 1996).Although these results promoted a straight-forward interpretation, the recent 4.5 Å 3Ddensity map by Mitsuoka et al., (1999)suggested a different assignment of thesidedness. Therefore, we analyzed freeze-dried AqpZ 2D crystals that were partiallymetal-shadowed to calculate both surfacereliefs and projection maps of one and thesame crystal. In this way the sidedness of theAqpZ projection map could be identified andrelated to that of AQP1. The results presentedhere suggest that the sidedness of AQP1previously reported by Walz et al. (1996) hasto be revised.

3.1.3. Results

Trypsin digested AqpZ 2D crystalsadsorbed to glow discharged carbon filmswere washed in distilled water, and quicklyfrozen in liquid nitrogen. After freeze-drying

74

and deposition of a 5 Å thick heavy-metalfilm (see Walz et al. 1996), the crystals wereimaged at a temperature of - 180° C at dosesbelow 5 electrons/Å2. Overview images(Figure 1a) were taken with the Gatan slowscan CCD camera at a magnification ofx4'000 and directly used to positioncrystalline areas for low dose imaging at highmagnification. To this end, tightly adsorbed,

flat single layered sheets were selected (seeFigure 1a).

Overview images of metal-shadowedsamples (Figure 1a; arrow top right indicatesthe shadowing direction) were carefullysearched for the borderlines of metal-coating(indicated by the black and the white arrowsin Figure 1a). Unshadowed areas resultedwhen aggregates (indicated by the asterisk in

Figure 1. a) Overview image of a freeze-dried and subsequently metal-shadowed crystalline sheet of AqpZ adsorbedto a carbon-coated electron microscopy grid. The black and white arrows pointing towards each other indicate theborderline of metal-coated and uncoated areas. The asterisk indicates the aggregate, which prevented parts of thecrystal from metal deposition (compare aggregate and shadow borderline). The white frame defines the position of thefollowing higher magnification image. The shadowing azimuth is indicated in the top right (scale bar: 1µm). b)Low dose image of the area outlined in Figure 1a. The metal-shadowed side (top left) is darker than to the uncoatedregion (bottom right). The squares (n = 971) and crosses (n = 1444) indicate unit cell positions on the metal-coatedside and on the uncoated side, respectively, which fitted with a displacement tolerance smaller than 0.05 (4.75 Å) tothe square lattice of 95 Å. The shadowing azimuth is given in the top right (scale bar: 100 nm). c) Correlationaverage of the 971 unit cells found on the unidirectionally metal-shadowed area. The four-fold symmetry of thetetramers is lost due to the unidirectional metal deposition. The shadowing azimuth is displayed in the top right (thewhite frame indicates the unit cell: 95 x 95 Å). d) Four-fold symmetrized surface reconstruction of the crosscorrelation average displayed in Figure 1c (the white frame indicates the unit cell: 95 x 95 Å). Rhombi surroundingthe higher, extracellular surface form a right-handed windmill-like structure. e) Correlation average of the 1444 unitcells found on the uncoated freeze-dried area. The four-fold symmetry of the tetramers is preserved. Adjacent tetramersappear with different brightness probably due to crystal - carbon film interaction (white frame indicates unit cell: 95x 95 Å). f) p4212 symmetrized projection map of the cross correlation average displayed in Figure 1e (the whiteframe indicates the unit cell: 95 x 95 Å). The central tetramer is the view from the extracellular side (correspondingto the central particle in Figure 1d) and is surrounded by rhombi forming a right-handed windmill-like structure.

75

Figure 1a) blocked the metal beam. Areassuch as that outlined by the square in Figure1a were then recorded at a magnification ofx45'000 (5 Å/pixel on the CCD camera,Figure 1b), yielding a large number of unitcells from the freeze-dried, metal-shadowedarea and from the uncoated area. Raw data ofshadowed areas appear dark (Figure 1b, topleft), while uncoated areas are bright (Figure1b, bottom right). Such images were Fourierpeak filtered to produce a first reference for across correlation peak search on both themetal-coated and the uncoated areas (Figure1b, the cross correlation peaks on the metal-shadowed area are indicated by squares, onthe uncoated area by crosses). A latticeyielding the perpendicular vectors of 95 Ålength was fitted to the correlation peaks witha tolerance of 0.05 (4.75 Å). Figure 1cdisplays the cross correlation average of the971 unit cells (white frame indicates the unitcell) found on the metal-coated area in Figure1b. Adjacent tetramers have a differentappearance due to the up and down packingof the particles within the crystal. As aconsequence of unidirectional shadowing(arrow top right indicates the shadowingdirection) the four fold symmetry is lost.Figure 1d (white frame indicates the unit cell)

shows the four fold symmetrized surfacerelief reconstruction of Figure 1c. The pixelsampling of 5 Å prevented resolution of finesubstructures in this average, but theorientation of the lipid filled rhombusbetween adjacent tetramers, and the heightdifference of the tetramers with respect toeach other is distinct. The projection averageof 1444 unit cells of the uncoated freeze-driedarea (Figure 1e, white frame indicates the unitcell) reveals the tetramer organization andorientation within the crystal lattice. Probablyas result of interactions between the crystaland the carbon film or residual metalshadowing neighboring particles differslightly in brightness. Nevertheless, theappearance of adjacent tetramers is verysimilar and the four-fold symmetry isessentially preserved. After p4212symmetrization (Figure 1f, white frameindicates the unit cell) the rotation of theparticles with respect to each other and theorientation of the rhombus-shaped lipidinterspace is clearly visible. From Figure 1 weconclude that the surface protruding mostfrom the lipid bilayer (central tetramer inFigure 1d) corresponds to the tetramersimilarly surrounded by rhombi in right-handed orientation in the projection map

Figure 2. Correlation between surface topography and electron density projection map (panel sidelenghts 190 Å).a) Average of high resolution AFM topographies (10° tilted surface representation) of trypsin digested AqpZ 2Dcrystals imaged in buffer solution. The crown-like extracellular surface protrudes 7 Å out of the lipid bilayer, whilethe cytoplasmic surface only protrudes by 3.5 Å (Scheuring et al. 1999). b) Surface reconstruction (10° tilted surfacerepresentation) of freeze-dried unidirectionally metal-shadowed AqpZ 2D crystals. As indicated, right-hand orientedrhombus-shaped lipid interspaces surround the extracellular surface. The asterisk indicates the position of protrudingstructure in the metal-shadowing surface reconstruction map, which is not present in the topography recorded by theAFM. c) Average density map of freeze-dried AqpZ 2D crystals. As indicated, right-hand oriented rhombus-shapedlipid interspaces surround the projection viewed from the extracellular side. The asterisk indicates protein densitywhich induces a protrusion signal in the surface reconstruction (indicated by asterisk in Figure 2b, see Discussion).d) Average density map of cryo electron micrographs from trehalose embedded AqpZ 2D crystals (Ringler et al.1999).

76

(central tetramer in Figure 1f). As previouslyreported, the most protruding surface isextracellular (Scheuring et al., 1999).Therefore, this central tetramer represents theprojection from the extracellular side.

To obtain higher resolution structuralinformation images of both metal-coated anduncoated crystals were recorded at amagnification of x77'000 (3.1Å/pixel on theCCD camera). The unit cell dimensions werefound to be a = b = 95 ± 1 Å and g = 90 ± 1°(n = 25). Surface relief reconstructions fromsuch images of unidirectionally metal-shadowed AqpZ crystals had a resolution of12 Å (Figure 2b, Fuchs et al., 1995; Kistler etal., 1977; Guckenberger, 1985). The averagesurface relief obtained from the metal-shadowed specimen (Figure 2b) is consistentwith the average resulting from highresolution AFM topographs (Figure 2a;Scheuring et al., 1999) in that both exhibitone strongly and one weakly protrudingtetramer. The two averages also correlatefavorably in the topology along the peripheryof the higher extracellular surface, the strongindentation in the center of the cytoplasmicsurface, and the overall particle organization.However the inner ring of protrudingstructures on the extracellular surface isoriented differently in the AFM topographand the outer ring protrusions of thecytoplasmic surface of the reconstructed reliefare more pronounced than the featuresdetermined with the AFM.

A projection map was calculated to 12 Åresolution from the unshadowed freeze-driedcrystal areas (Figure 2c). This map is verysimilar to the cryo electron microscopyprojection map with a resolution of 7 Å(Figure 2d, Ringler et al. 1999). The tetramersshow an inner ring of densities close to theirfour-fold symmetry center and an outer ringof densities along their periphery. Themonomers can be distinguished and theopposite rotation of adjacent tetramers withrespect to the lattice lines leading to rhombus-shaped lipid interspaces is evident. These features allow the unambiguous alignment of

Figure 3 . Overlay of AqpZ AFM topographyrecorded in buffer solution on projection maps ofAqpZ and AQP1, both rendered at 7Å resolution (fullframe sizes 95 Å). a) AFM topography of theextracellular surface of AqpZ with outlined andannoted protrusions and overall shape (seeDiscussion; Scheuring et al. 1999). b) AFMtopograph of the cytoplasmic surface of AqpZ withoutlined protrusions housing non membrane buriedparts of loop B (see Discussion; Scheuring et al.1999). c) Projection map of AqpZ (Ringler et al.1999) viewed from the extracellular side. Outlines ofcorresponding surface protrusions are overlaid: LoopC is at the periphery and spans a major part of amonomer. d) Projection map of AqpZ (Ringler et al.1999) viewed from the cytoplasmic side. Outlines ofcorresponding surface protrusions are overlaid: non-membrane-buried parts of loop B (comprising theNPA motif, which participates in the water pore)cross the center of one monomer. e) Projection mapof AQP1 (Walz et al. 1995) viewed from theextracellular side after an applied clockwise rotation of15°. Protrusions are outlined as in Figure 3c. f)Projection map of AQP1 (Walz et al. 1995) viewedfrom the cytoplasmic side after an appliedcounterclockwise rotation of 15°. Protrusions areoutlined as in Figure 3d.The maximum rotational alignment cross correlationcoefficient was obtained when the lowest densitybetween monomers (indicated by the asterisk) and ofasymmetric densities within the monomer (indicatedby 1 and 2) were superimposed.

77

the two maps (Figure 2c and 2d).The extracellular and cytoplasmic

topologies revealed by AFM experiments(Scheuring et al 1999) are shown in Figures3a and 3b, respectively. Correspondingregions of the projection map of AqpZ areshown aligned in Figures 3c and 3d,respectively. Finally, Figures 3e and 3f showthe projection map of AQP1 aligned withrespect to that of AqpZ (Figures 3c and 3d).This alignment is compatible with thatreported by Ringler et al. (1999), but it hasbeen improved by considering the 1 %difference of the unit cell size. Thisimprovement yielded a correlation coefficientof 78 % after a clockwise rotation of theAQP1 tetramer (Figure 3e) by 15°. If,however the AQP1 tetramer shown in Figure3f is rotationally aligned with AqpZ in Figure3c, the correlation coefficient is 65 %,allowing the unambiguous assignment of theAQP1 projection map in Figure 3e asextracellular. To relate protrusions observedby AFM with loops or protruding helices, theprojection maps are overlaid by the

topography contours.Figures 4a and 4b display perspective

views of the 3D map of AQP1 at 6Åresolution from the extracellular and thecytoplasmic side, respectively. a-helices arerepresented as gray sausages, the bright partsof which display the ends facing the viewer.The map is overlaid by contours ofprotrusions imaged by AFM under nativeconditions (see Figures 3a and 3b). Flexibleparts of proteins, such as loops, are mostlyaveraged out, hence the AFM topographyimplements additional information on theorganization of the non-membrane-stabilizedprotein structure to the 3D density map.

3.1.4. Discussion

AFM investigations of densely packed orregularly arranged membrane protein layerscan provide information on their sidedness inconjunction with either proteolytic cleavage ofa terminal domain (Scheuring et al., 1999), orspecific binding of antibodies (Müller et al.,1996). The sidedness of AqpZ surfaces has

Figure 4. Surface rendered 3D electron density map of AQP1 at 6 Å resolution (Walz et al. 1997). The righthanded bundle of the 6 transmembrane spanning a-helices is aligned according to the projection maps in Figures 3eand f. Superposed contours indicate the topographical features of AqpZ 2D crystals measured by AFM in buffersolution. Helix assignment as proposed by Heymann & Engel (2000). a) View from the extracellular side, withassigned loops indicated. The small protrusion on the periphery might be part of loop C which connects the end ofhelix 3 with the beginning of helix 4 (see Discussion). b) View from the cytoplasmic side. The AFM topographshows one protrusion corresponding to parts of loop B which spans the water pore (see Discussion).

78

been unambiguously defined by imagingcrystals before and after proteolytic cleavageof the cytoplasmic N-terminus identifying the7 Å high crown-like surface as extracellular(Figure 2a; Scheuring et al., 1999).

To link topographical data obtained byAFM with the projection structure acquiredby electron crystallography, we have analyzedfreeze-dried crystals that were partially metal-shadowed. As illustrated in Figure 1, surfacereliefs calculated from metal coated areaswere thus in register with projection mapsfrom uncoated areas of the same crystal,allowing the unambiguous assignment oftopography and projection map. This novelapproach is of particular advantage foraquaporins whose quasi two-fold symmetrymakes the assignment of the sidednessdifficult (Mitsuoka et al., 1999).

The experiment described here can beapplied in general provided that large,coherent 2D crystals are available. It has beendesigned to solve the discrepancy between thesidedness assignment of AQP1 by Walz et al.(1996) and the 4.5 Å 3D density map byMitsuoka et al., (1999). The regions assignedto loop B and E in the latter map suggest anopposite sidedness than that proposed byWalz et al. (1996). Since AFM experimentsallowed the sidedness of AqpZ topographiesto be firmly assigned (Scheuring et al. 1999),and because AqpZ projection maps could bealigned with those of AQP1 (Figure 3;Ringler et al. 1999), establishing the linkbetween relief reconstruction and projectionmap appeared to be a straight-forwardapproach to settle this pertinent question. Theexperimental results documented in Figures 1and 2 provide a solid basis to align thesurface topography of AqpZ recorded byAFM to 7 Å resolution with the projectionmap of AQP1 (Figure 3) and hence with the3D map of AQP1 (Figure 4).

A recent sequence alignment study of 160aquaporin sequences revealed highlyconserved residues within all helical segments(Heymann & Engel 2000). Together with the

general similarity of projection maps fromdifferent aquaporins (Engel et al. 2000;Daniels et al. 1999) this suggests a conservedhelical packing arrangement. From helicalperiodicity analysis Heymann & Engel(2000) proposed a helix assignment that isindicated in Figure 4. This assignment isconsistent with the findings from fittinghelical stretches to elongated structures in the4.5 Å 3D density map (de Groot et al. 2000).Therefore, the contoured protrusions of AqpZoverlayed on the 3D map of AQP1 are highlyrelevant. They provide a solid basis to selectthe most likely helix assignment from the twopossibilities given by de Groot et al. (2000)and Heymann & Engel (2000). Figure 4aindicates that the peripheral protrusion C islikely to connect helices 3 and 4. This iscompatible with the assignment of thisprotrusion to loop C based on volumecalculations and flexibility mapping by AFM(Scheuring et al. 1999). The previouslyunassigned peripheral protrusion on theextracellular surface appears to be the C-terminal end of helix 3. Thus, the otherprotrusion, now labeled A, must representloop A, in agreement with the helicalassignment by Heymann & Engel (2000) andde Groot et al. (2000). Since loop D is shortin all aquaporins and the N- and C-termini arealmost completely removed by trypsindigestion of AqpZ 2D crystals (Scheuring etal. 1999), the protrusion contoured in Figure4b is likely to present the surface of loop Bthat folds back into the membrane andconnects helices 2 and 3. Position and shapeof this protrusions further confirm the helicalassignment shown in Figure 4.

Taken together, compelling evidence hasbeen accumulated to justify revision of thesidedness assignment proposed by Walz etal. (1996). The pertinent question arises as tohow apparently solid data could have beenmisinterpreted. A possible explanation isrelated to the observation that carboxypeptidase Y treatment in solution tends toproduce disordered crystals and aggregates

79

(B. Heymann, D. Fotiadis, D. J. Müller,unpublished results). Therefore, the fewcrystals found by Walz et al. (1996) afterdecarboxylation may not have been properlydigested. The structural differences however,may have resulted from lattice disorder andsurface contamination by proteolyticfragments.

The alignment of the AqpZ topographywith the AQP1 3D map corroborates thetopology of AQP1 derived from fitting helicalstretches using the program ROTTRANS (deGroot et al. 2000) to the 4.5 Å 3D densitymap (Mitsuoka et al. 1999) and the helixassignment from sequence analysis(Heymann & Engel 2000). With thesidedness issue resolved, the overallarchitecture of AQP1 is now established andwill help tracing the polypeptide in higherresolution electron crystallographic analysis.

3.1.5. Materials and Methods

3.1.5.1. 2D crystallization

2D crystals were produced by dialysis ofsolubilized AqpZ (1 mg / ml, in 2% octyl-b-D-glucopyranoside (OG)) mixed with 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine (POPC) / 1,2-dimiristoyl-sn-glycero-3-phosphocholine (DMPC) 1/1(from Avanti Polar Lipids, Inc., USA) in 2%OG (from Anatrace, Inc., USA) at a LPR of0.4 against a detergent free buffer (10 mM -citrate, pH 6, 200 mM NaCl, 100 mM MgCl2,10% glycerol, 0.005% NaN3) (Ringler et al.,1999). Crystals were harvested after threedays.

3.1.5.2. Trypsin digestion

For cleavage of the N-terminal fragment,AqpZ-10his crystals were incubated overnightat 4°C with trypsin (1 mg/ml). After trypsintreatment samples were investigated bysodium dodecyl sulfate-polyacrylamide gelelectrophoresis (SDS-PAGE) using 10%

(w/v) acrylamide gels (data not shown, fordetails see Scheuring et al. 1999).


AqpZ 2D crystals were deposited ontofreshly cleaved muscovite mica (from MicaNew York Corp., New York, USA) andimaged in buffer solution (10 mM Tris-HCl,pH 7.2, 150 mM NaCl) at high resolution(for details see Scheuring et al. 1999).

3.1.5.4. Freeze-drying & metal-shadowing

AqpZ crystals were adsorbed (2 min.) toglow-discharged (1 min.) carbon-coated 400mesh grids. These were washed twice withdouble-distilled water, blotted and plungedinto liquid nitrogen. The grids were thenfreeze-dried and metal-shadowed in theMIDILAB (Gross et al., 1990), as detailed inWalz et al. (1996). After shadowing, gridswere transferred to a specially designed Gatancryo holder, and examined. Images weredigitally recorded with a Gatan-694 slow scanCCD camera with a maximal image size of10242 pixels. Images were correlationaveraged using the SEMPER imageprocessing package (Saxton et al., 1979),while surface reconstructions of metal-shadowed crystals were calculated inSEMPER or MILAN (Fuchs et al., 1995)image processing packages. The handednessdetermined by the sample orientation in themicroscope, the image acquisition with theCCD camera, and the data transfer to variousprocessing systems was carefully controlledby coadsorption of amyloid fibrils.

3.1.5.5. Cryo electron microscopy

AqpZ 2D crystals mixed with 3 - 10%trehalose were adsorbed to carbon-coatedcopper electron microscopy grids. The gridswere blotted and plunged into liquid ethane.Vitrified specimens were recorded in aHitachi H8000 electron microscope with a

80

LaB6 filament, operated at 200 kV under lowdose conditions (~ 5 e- / Å2) and processedusing the MRC image processing package(for details see Ringler at al 1999).


We thank Dr. Claire Goldsbury whoprovided the amylin fibers coadsorbed withthe AqpZ crystals to overcome sidednessproblems. We acknowledge fruitfuldiscussions with Dr. S. A. Müller who helpedin assembling the manuscript. This work wassupported by the Swiss National Foundationfor Scientific Research (grant 4036 - 44062to A. E.), the Swiss Priority Project for Microand Nano System Technology (MINAST),the Maurice E. Müller Foundation ofSwitzerland and the National Institutes ofHealth (to P. A.).

3.1.7. References

Agre, P., Preston, G., Smith, B., Jung, J.,Raina, S., Moon, C., Guggino, W. &Nielsen, S. (1993). Aquaporin CHIP: thearchetypal molecular water channel. Am. J.Physiol. 265, F436-476.

Borgnia, M., Kozono, D., Calamita, G.,Maloney, P. C. & Agre, P. (1999).Functional reconstitution andcharacterization of AqpZ, the E. coli waterchannel protein. J. Mol. Biol. 29, 1169-1179.

Calamita, G., Bishai, W., Preston, G.,Guggino, W. & Agre, P. (1995).Molecular cloning and characterization ofAqpZ, a waterchannel from Escherichiacoli. J. Biol. Chem. 270, 29063-29066.

Calamita, G., Kempf, B., Bonhivers, M.,Bishai, W. R., Bremer, E. & Agre, P.(1998). Regulation of the Escherichia coliwater channel gene AqpZ. Proc. Natl.Acad. Sci. USA 95, 3627-3631.

Chandy, G., Zampighi, G. A., Kreman, M. &Hall, J. E. (1997). Comparison of thewater transporting properties of MIP andAQP1. J. Membr. Biol. 159, 29-39.

Cheng, A., van Hoek, A. N., Yeager, M.,Verkman, A. S. & Mitra, A. K. (1997).Three-dimensional organization of ahuman water channel. Nature 387, 627-630.

Chrispeels, M. & Agre, P. (1994).Aquaporins: water channel proteins ofplant and animal cells. Tr. Biochemic. Sci.19, 421-425.

Daniels, M. J., Chrispeels, M. J. & Yeager,M. (1999). Projection structure of a plantvacuole membrane aquaporin by electroncryo-crystallography. J. Mol. Biol. 294,1337-1349

De Groot, B. L., Heymann, J. B., Engel, A.,Mitsuoka, K., Fujiyoshi, Y. & Grubmüller,H. (2000). The fold of human aquaporin1. . J. Mol. Biol. submitted.

Fotiadis, D., Hasler, L., Müller, D.J.,Stahlberg, H., Kistler, J. & Engel A.(2000). The surface topography of lensMIP supports dual functions. . J. Mol.Biol. submitted

Fuchs, K., Tittmann, P., Krusche, K. &Gross, H. (1995). Reconstruction andrepresentation of surface data from two-dimensional crystalline, biologicalmacromolecules. Bioimaging 3, 12-24.

Gorin, M. B., Yancey, S. B., Cline, J., Revel,J.-P. & Horwitz, J. (1984). The majorintrinsic protein (MIP) of the bovine lensfiber membrane: Characterization andstructure based on cDNA cloning. Cell 39,49-59.

Guckenberger, R. (1985). Surface reliefderived from heavy-metal-shadowedspecimen-Fourier space techniquesapplied to periodic objects.Ultramicroscopy 16, 357-370.

Heymann, J. B. & Engel, A. (2000).Structural clues in the sequences of theaquaporins. J. Mol. Biol. 295, 1039-1053.

Heymann, J. B., Müller, D. J., Landau, E. M.,Rosenbusch, J. P., Pebay-Peyroula, E.,Büldt, G. & Engel, A. (1999). Charting thesurfaces of purple membrane. J. Struct.Biol. 128, 243-249.

Jung, J., Preston, G., Smith, B., Guggino, W.& Agre, P. (1994). Molecular structure ofthe water channel through aquaporin

81

CHIP. The hourglass model. J. Biol.Chem. 269, 14648-14654.

Kistler, J., Aebi, U. & Kellenberger, E.(1977). Freeze-drying and shadowing atwo-dimensional periodic specimen. J.Ultrastr. Res. 59, 76-86.

Li, H., Lee, S. & Jap, B. K. (1997). Moleculardesign of aquaporin-1 water channel asrevealed by electron crystallography.Nature Struct. Biol. 4, 263-265.

Mitsuoka, K., Murata, K., Walz, T., Hirai, T.,Agre, P., Heymann, J. B., Engel, A. &Fujiyoshi, Y. (1999). The structure ofaquaporin-1 at 4.5-Å resolution revealsshort a-helices in the center of themonomer. J. Struct. Biol. 128, 34-43.

Mulders, S., Preston, G., Deen, P., Guggino,W., van, O. C. & Agre, P. (1995). Waterchannel properties of major intrinsicprotein of lens. J. Biol. Chem. 270, 9010-9016.

Müller, D. J. & Engel, A (1999). pH andvoltage induced structural changes ofporin OmpF explain channel closure. J.Mol. Biol. 285, 1347-1451.

Müller, D. J., Schoenenberger, C. A., Büldt,G. & Engel, A. (1996). Immuno-atomicforce microscopy of purple membrane.Biophys. J. 70, 1796-1802.

Preston, G. M. & Agre, P. (1991). Isolationof the cDNA for erythrocyte integralmembrane protein of 28 kilodaltons:Member of an ancient channel family.Proc. Natl. Acad. Sci. USA 88, 11110-11114.

Preston, G. M., Carroll, T. P., Guggino, W.B. & Agre, P. (1992). Appearance ofwater channels in Xenopus Oocytesexpressing red cell CHIP28 protein.Science 256, 385-387.

Ringler, P., Borgnia, M., Stahlberg, H., Agre,P. & Engel, A. (1999). Structure of thewater channel AqpZ from Escherichia colirevealed by electron crystallography. J.Mol. Biol. 291, 1181-1190.


Scheuring, S., Ringler, P., Borgnia, M.,Stahlberg, H., Agre, P. & Engel, A.(1999). High resolution AFM topographsof the Escherichia Coli waterchannelaquaporin Z. EMBO J. 18, 4981-4987.

Sidel, V. W. & Solomon, A. K. (1957).Entrance of water into human red cellsunder an osmotic pressure gradient. J.Gen. Phys. 41, 243-257.

Walz, T., Hirai, T., Murata, K., Heymann, J.B., Mitsuoka, A., Fujiyoshi, Y., Smith, B.L., Agre, P. & Engel, A. (1997). The 6 Åthree-dimensional structure of aquaporin-1. Nature 387, 624-627.

Walz, T., Smith, B., Agre, P. & Engel, A.(1994a). The three-dimensional structureof human erythrocyte aquaporin CHIP.EMBO J. 13, 2985-2993.

Walz, T., Smith, B., Zeidel, M., Engel, A. &Agre, P. (1994b). Biologically active two-dimensional crystals of aquaporin CHIP.J. Biol. Chem. 269, 1583-1586.


Walz, T., Typke, D., Smith, B. L., Agre, P., &Engel, A. (1995). Projection map ofaquaporin-1 determined by electroncrystallography. Nature Struct. Biol. 2,730-732.

Zeidel, M. L., Ambudkar, S. V., Smith, B. L.& Agre, P. (1992). Reconstitution offunctional water channels in liposomescontaining purified red cell CHIP28protein. Biochem. 31, 7436-7440.

82

4. Structural studies of amembrane transporter

85

4. Structural studies of a membrane transporter

4.1. The functional Escherichia coli lactose permease LacY/Cytb562/6Hisforms trimers: A 2.8 nm 3D reconstruction and preliminary electroncrystallographic data

4.1.1. Summary

A chimeric protein consisting of lactosepermease with cytochrome b562 in the middlecytoplasmic loop and 6 His residues at the Cterminus (LacY/L6cytb562/417H6; “redpermease”) was overexpressed inEscherichia coli, and isolated by nickelaffinity chromatography after solubilizationwith dodecyl-ß,D-maltopyranoside or decyl-ß,D-maltopyranoside. Red permease was theneither reconstituted in the presence ofphospholipids yielding densely packedvesicles and well-ordered two-dimensional(2D) crystals, or examined as single particles.Single particle analysis of 1188 trimericparticles uniformly adsorbed to the electronmicroscopy grid reveals a ~ 5.5 nm x ~ 7.7nm bean-shaped protein with a central stain-filled indentation, at a resolution of ~ 2 nm. Athree dimensional (3D) reconstruction couldbe calculated at ~ 2.8 nm resolution fromtilted images, showing that the cytochromeb562 induces trimerization. Reconstitutedsamples yielded densely packed vesicles ortrigonal crystals of trimeric LacY. The bestpacked crystals showed diffraction spots upto a resolution of 8 Å, however these crystalswere multilayered and separation of the layersby Fourier peak filtering was not possible.

4.1.2. Introduction

The lactose (lac) permease of Escherichiacoli catalyzes the coupled stoichiometrictranslocation of ß-galactosides and H+ acrossthe cytoplasmic membrane. Encoded by thelacY gene, the permease is a paradigm for a

large family of secondary transport proteinsfrom archeae to the mammalian centralnervous system that convert free energystored in an electrochemical ion gradient intoa concentration gradient (reviewed in Kaback,1976; Kaback, 1983; Poolman & Konings,1993). The polytopic, hydrophobic membraneprotein, LacY, has been solubilized, purified tohomogeneity, reconstituted intoproteoliposomes (Viitanen et al., 1986), andfound to be solely responsible in itsmonomeric form for ß-galactoside transport(Sahin-Tóth et al., 1994). All availableevidence indicates that the protein iscomposed of 12 α-helical membrane-spanning segments connected by hydrophilicloops with the N and C termini on thecytoplasmic face (reviewed in (Kaback, 1996;Kaback & Wu, 1997). Hydropathy profilingsuggests that many membrane transportershave similar secondary structures, indicatingthat both the basic tertiary structure andmechanism of action of these enzymes haveprobably been conserved. Therefore, studieson bacterial transport proteins, which areconsiderably easier to manipulate than theireukaryotic counterparts, have importantrelevance to transporters in higher ordersystems.

Site-directed and Cys-scanningmutagenesis (Frillingos et al., 1998a;Frillingos et al., 1998b) demonstrate that only6 out of the 417 amino acid residues inlactose permease are irreplaceable withrespect to active transport. On the other hand,site-directed fluorescence and chemicallabeling indicate that widespreadconformational changes occur during enzyme

87

turnover (reviewed in (Frillingos et al., 1998b;Kaback & Wu, 1997). About 10% of theactive Cys-replacement mutants are altered byalkylation, and these Cys residues cluster onhelical faces. Moreover, monoclonal antibody(mAb) 4B1 which binds to the periplasmicloop connecting helices VII and VIII (Sun etal., 1996) specifically blocks all reactionscatalyzed by the permease that involve net H+

translocation (Carrasco et al., 1984; Frillingos& Kaback, 1996). Taken together, the resultssuggest that the permease is comprised of abundle of loosely packed rigid bodies (i.e.,transmembrane helices) with few residues thatare essential for the mechanism and specificsurface contours that permit sliding motionsto occur during transport. Based on a varietyof site-directed approaches which includesecond-site suppressor analysis andsite-directed mutagenesis, excimerfluorescence, engineered divalent metalbinding sites, chemicalcleavage, electronparamagnetic resonance,thiol cross linking andidentification ofdiscontinuous mAbepitopes, a helixpacking model has beenformulated (reviewed in(Frillingos et al., 1998b;Kaback et al., 1997;Kaback & Wu, 1997).

Although site-directed structuralapproaches and ex-tensive mutational ana-lysis have led to aproposed helix packingmodel for LacY andmechanism for couplingbetween substrate andH+ translocation (Fril-lingos et al., 1998b;Kaback, 1997), arelatively low-resolutionstructure obtained by

direct means (i.e., helix packing) will provideimportant additional insight. Little is knownabout the helix tilts, for example. The mostsevere limitation to obtaining structuralinformation on LacY is the lack of three-dimensional (3D) crystals. Attempts toproduce highly ordered two-dimensional(2D) crystals suitable for electroncrystallography have been reported for LacY(Li & Tooth, 1987) and for melibiosepermease (Rigaud et al., 1997), but thesecrystals were not of sufficient quality forhigh-resolution structural analysis. BecauseLacY is extremely hydrophobic and probablyhighly flexible outside of the bilayer (LeCoutre et al., 1997), it is notoriously difficultto crystallize. To generate a higher proportionof polar surface area, a fusion between LacYand cytochrome b562 (red permease) hasbeen constructed and shown to have nativelactose transport activity (Privé et al., 1994).

Figure 1 . a) Negative stain electron micrograph of LacY trimers uniformlyadsorbed to the electron microscopy grid (scale bar: 50nm). b) Average of 1188rotationally aligned, end-on views (scale bar: 5nm).

88

Zhuang et al. (1999)found conditions forpurification,reconstitution and 2Dcrystallization of redpermease obtained byoverexpression in E.coli. Single particleanalysis of denselypacked LacY mono-mers and reconstitutedtrimers were reportedto have an appro-ximately trapeziformshape 5.4 nm long and4.1 nm wide on oneside and 5.1 nm wideon the other withperipheral protrusionsand a central stain-filled depression(Zhuang et al., 1998).

Here we present a3D reconstructioncalculated from tiltedimages of solubilizedLacY (red permease) trimers at 2.8 nmresolution. The protein has a bean-like shapewith two peripheral density maxima.Sideviews showed that the cytochrome b562emerging from loop 6 of each monomerinduces trimerization. Furthermore, denselypacked vesicles and 2D crystals were grown,also revealing trimeric organization of theparticles. However, the best crystals had astrong tendency to stack. Diffraction patternsof trehalose embedded crystals showed spotsto a resolution < 8 Å, but the layers could notbe separated by Fourier peak filtering,implying that the stacking of the crystallinelayers is in register.

4.1.3. Results and discussion

4.1.3.1. Protein purification

After induction with 0.3 mM IPTG, cells

were harvested at an OD600 of 2. Crudemembranes containing overexpressed redpermease were collected after rupturing thecells in a French press and washed with 6%sodium cholate to remove adsorbedimpurities. LacY/L6cyt b562/417H6 wasefficiently extracted by solubilizing thecholate-washed membranes with 3% dodecyl-β,D-maltopyranoside (DDM) or 3% decyl-β,D-maltopyranoside (DM) at pH 7.5. Theprotein was subsequently purified to nearhomogeneity in a single step by metal chelateaffinity chromatography on a Ni-NTAcolumn. During this step, the DDM or DMconcentration was adjusted to 0.01% or 0.1%respectively. The fractions containing redpermease were directly used for singleparticle analysis or reconstitutionexperiments.

Figure 2. a) Micrograph of negatively stained of LacY trimers and b) micrographof the same region after tilting the grid to 60°. The particles numbered 1, 2 and 3 inboth projections document how the tilt parameters were determined.

89

4.1.3.2. Single particle analysis

Red permease solubilized with DDM weredirectly investigated using negative stainelectron microscopy. An image at 50'000 xmagnification taken from an untiltedspecimen is displayed in Fig. 1a. The circleindicates a lacY trimer adsorbed to the gridend-on. Such particles were automaticallyselected (see Materials and Methods) andaveraged after alignment, resulting in aprojection map at ~ 2nm resolution (Fig. 1b).The trimer clearly consists of three bean-shaped proteins with dimensions of ~ 5.5 nmx ~ 7.7 nm each and 2 density maximatowards the two ends of the monomers. Themap suggests an indentation in the center ofeach molecule (Fig. 1b).

To calculate a 3D density map thespecimen was imaged at 0° and 60° tilt with

respect to the electron beam (Fig 2a and b).Images were recorded at 40'000 xmagnification to acquire large areas forparticle alignment. Particles numbered 1, 2and 3 indicate 3 corresponding LacY trimersin the untilted and the tilted image. Therotational orientation of the particles wasdefined in the non-tilted images andsubsequently used for the tilted projection.

Figure 3 shows the 3D reconstruction at ~2.8nm resolution comprising 850 projectionpairs from 8 image pairs as displayed infigure 2. The two end-on views (images 4 and10) are markedly different. One side iscapped at the center of the trimer (4), the otherside shows a strong indentation in the centerof the trimer (10). This suggests that thecapped side corresponds to the cytoplasmicsurface where the three large loops containingcytochromes b562 interact to induce trimer

1 2 3 4

5 6 7 8

9 10 11 12

Figure 3 . Surface rendered 3D reconstruction of LacY trimers at a resolution of ~2.8nm calculated from imagepairs like that shown in figure 2. The image top left is a sideview, the following 11 images show the trimer rotatedby 30° around the axis indicated. The cap structure on one side is supposed to arise from the three cytochromes b562in loop 6 which induce trimerization through hydrophilic interactions (scale bar: 5nm).

90

formation. Correspondingly, the other surfacerepresents the periplasmic side where onlylittle surface protruding loops are to beexpected.

4.1.3.3. Reconstitution and 2Dcrystallization

Different reconstitution strategies weretested. No matter whether dodecyl ß,D-maltopyranoside (DDM; critical micelleconcentration, cmc = 0.007%) or decyl ß,D-maltopyranoside (DM; cmc = 0.07%) wasused to solubilize the protein, the lipid wasprepared in octyl ß,D-glucopyranoside (OG;cmc = 0.6%) or a mixture of OG and DM forthe dialyses which were carried out at a lipid-to-protein ratio (LPR) ranging from 0.5 to 1.5(w/w).

As previously described, the temperatureprofile for one dialysis cycle profile wasused: 12 °C (12 h), 12-37 °C (24 h), 37 °C (24h), 37 °C-12 °C (12 h).While DM wasremoved in one dialysiscycle, two cycles wererequired to removeDDM. Various divalentcations were added tothe dialysis buffer, butin most cases this led toprotein aggregation.Trying to trap LacY inone conformation bythe addition of lactoseor beta – D – galacto-pyranosyl – 1 – thio –beta – D – galacto –pyranoside (TDG) tothe dialysis buffer didnot improve crystal-linity. Similarly, neitherglycerol nor DDThelped to improvecrystallinity or to pre-vent stacking of the

crystalline layers. The best reconstitutionbuffer solution contained 10 mM Tris (pH7.4), 150 mM NaCl, 0.05% NaN3 and 40mM Mg2+, corroborating conditionsestablished by Zhuang et al. (1999). Highprotein concentrations (at least 1 mg/ml) werea prerequisite for assembling densely packedvesicles (Fig. 4). In addition, the presence of1-palmitoyl 2-oleolyl phosphatidylcholine(POPC) (transition temperature 3 °C) wasessential for efficient integration of theprotein into the bilayer because of the lowtemperature (12 °C) during the initial stagesof the dialysis. The best crystals were grownusing a mixture of POPC and DMPC of 1/1(w/w).

Densely packed vesicles were foundcontaining trimeric LacY in pseudocrystallineorder. The appearance of the particles in theraw data micrograph (Fig. 4a) and the averageproduced after single particle alignment (Fig.

Figure 4 . a) Negative stain micrograph of densely packed, pseudocrystallineLacY. The trimeric organization is directly visible (scale bar: 50nm). b) Average of571 rotationally aligned particles from image a) (scale bar: 5nm).

91

4b) are in very good agreement with thestructure of the solubilized protein (Fig. 1).Again, the monomers are bean-shaped withtwo density maxima at either end. Themonomer reveals a handedness showing astronger density in the clockwise directionwith respect to the trimer symmetry axis,consistent with the density organizationrevealed by single particle analysis of thesolubilized proteins (Fig. 1b). However, theparticle dimensions, ~ 4.5 nm x ~ 6.5 nm aresmaller than found for the solubilized protein.There are two explanations for thisdiscrepancy: Firstly, the solubilized particlesare surrounded by detergent molecules, whichincrease their apparent size. Secondly, onlymembrane protruding domains of trimersincorporated in a membrane are stained.

Well ordered crystals were imaged usingnegative stain (Fig. 5) or trehalose embeddedcryo electron microscopy (Fig. 6). Both

methods document that the crystalline areasare fairly well ordered. In negatively stainedsamples trigonal lattices with dimensions of a= b = 9.8 nm, similar to those reported byZhuang et al. (1999) (a = b = 10.4 nm), werefound (Fig. 5). Calculated diffraction patternsrevealed spots up to the 3rd order (Fig. 5b).The average (Fig. 5c) calculated from theimage displayed in figure 5a again showsclearly a trimeric organization of the protein.However no handedness can be detected insuch maps. This is probably a consequenceof stacking of crystalline layers with proteinsintegrated in opposite orientations withrespect to the membrane planes. As a result,the contrast of such crystals is strong, buthandedness information is lost and theparticles appear small ~ 4.0 nm x ~ 5.2 nm.This explanation is supported by the fact thatthe contrast of the array displayed in figure 5aseems to fade towards the left side of the

Figure 5. a) Electron micrograph of negatively stained LacY 2D crystals. The stacking of the crystalline layersleads to a contrast variation depending on whether the layers are in register (center) or out of register (left edge),respectively (scale bar: 50nm). b) Calculated diffraction pattern of image a). Diffraction spots are sharp and 3rd orderspots are visible corresponding to a resolution of ~2nm. c) Cross correlation average of the right part of the crystalin image a). The trimeric organization is resolved, handedness information is lost, probably due to overlay of trimersin both orientations (scale bar: 5nm).

92

image; where the stacking of the layers is notin perfect register, inducing a contrastmodulation.

Analogous to the negative stain electronmicroscopy studies, cryo-electronmicrographs of trehalose embedded LacYcrystals showed strong long rangemodulations (Moiré fringes), due to theirstacking arrangement (Fig. 6a). Calculateddiffraction patterns (Fig. 6b) show spots at ~8Å resolution (indicated by the circle).Because of the in-register-stacking of thecrystalline layers, information from theindividual crystal layers could not beseparated by Fourier filtering. However, thediffraction pattern proves that LacY can beintegrated into highly ordered crystalssuitable for high resolution cryo-electronmicroscopy.

4.1.4. Perspectives

To obtain further structural information onthe membrane transporter LacY, highlyordered 2D crystals must be grown. Theresults of the present study support the ideathat the functional LacY/cyt b562/417H6forms stable trimers in buffers containing

Mg2+ and DDM at a concentration close tothe cmc or into lipid bilayers. At a LPR closeto 1, these trimers associate into highlyordered 2D crystal stacks.

The recently found stability of LacY inDM allows detergent to be removed bothefficiently and rapidly by dialysis. Thispossibility allows a new dialysis conditions tobe explored using the dialysis machine orbuttons.

The already established crystallizationprotocol (Zhuang et al., 1998) using DDMsolubilizing LacY offers furtherimprovements. In particular, dialysis bufferswhich mediate destacking of the crystallinemulti layers (Fig 5, 6) should be screened.

Other 2D crystallization techniques mustbe explored as well, such as controlleddilution of protein-lipid-detergent mixtures(Remigy, personal communication),crystallization at a functionallized lipidmonolayer (Levy et al. 1999) or thewithdrawal of detergents using biobeads asdescribed in Rigaud et al 1998.

Figure 6. a) Cryo electron micrograph of trehalose embedded LacY 2D crystals. The stacking of crystalline layersinduce large distance modulations (scale bar: 50nm) b) Calculated diffraction pattern of image a). Diffraction spotscorresponding to a resolution of ~8Å are visible (indicated by circle).

93

4.1.5. Material and Methods

4.1.5.1. Materials

Isopropyl-ß,D-thiogalactopyranoside(IPTG) was from Boehringer MannheimGmbH, Germany. Sodium cholate was fromFluka, Switzerland., dodecyl-ß,D-malto-pyranoside (DDM), decyl-ß,D-malto-pyranoside (DM) and octyl-ß,D-gluco-pyranoside (OG) were from Anatrace,Switzerland. 1,2-dimyristoyl-sn-glycero-3-phosphocholine (DMPC) and 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine (POPC)were from Avanti Polar Lipids, Inc., USA.Nickel-nitrilotriacetic acid (Ni-NTA) resinwas from QIAGEN AG, Switzerland. Allother chemicals used were of analytical gradeand purchased from Merck (Schweiz) AG,Dietikon, Switzerland.

4.1.5.2. Protein Expression andPurification

Expression of the fusion protein betweenlactose permease and cytochrome b562 hasbeen described (Privé et al., 1994). Briefly,the cytochrome b562 gene was inserted intothe XhoI restriction site in the regionencoding the middle cytoplasmic loop (L6).The cassette version of the lacY gene usedwas under control of the lacpromoter/operator. To facilitate purification, 6consecutive His residues were engineered tothe C-terminus yielding theLacY/L6cytb562/417H6, referred to as "redpermease". E. coli XL-Blue cells transformedwith a plasmid encoding the fusion proteinwere cultivated in LB broth at 37 °C and thegrowth rate was monitored by optical density(OD) measurements at 600 nm. When anOD600 of 0.8 was achieved, the cells wereinduced with 0.3 mM IPTG. Cells wereharvested by centrifugation when OD600 ofabout 2 was reached, and were then rupturedin a French press cell. Membranes werecollected by centrifugation at 250,000 g for

4h, washed in 6% sodium cholate (Viitanen etal., 1986), and solubilized with 2% dodecyl-ß,D-maltopyranoside (DDM) or 2% decyl-ß,D-maltopyranoside (DM). Thesolubilization buffer contained 1M NaCland/or 1mM imidazol or 1mM histidine toprevent the unspecific binding ofcontaminants to the columns. Solubilized redpermease was purified by nickel affinitychromatography (Loddenkotter et al., 1993).Briefly, the 2% DDM extract from 10 g wetweight of cells was added to 1.5 ml of the Ni-NTA-agarose slurry in the column, androtated at 4 °C overnight. The column wasthen washed with 50 mM potassiumphosphate (pH 7.5), 200 mM NaCl, 0.01%DDM or 0.1% DM respectively until theabsorption baseline was reached. Then thecolumn was washed with 50 mM potassiumphosphate (pH 7.5), 200 mM NaCl, 0.01%DDM or 0.1% DM, 10mM Imidazol or10mM histidine. Once the baseline had againbeen reached, the protein was eluted with thesame buffer containing either 300mM ofimidazol or 150mM histidine. Purifiedsamples were analyzed by sodium dodecylsulfate/polyacrylamide gel electrophoresis(SDS-PAGE) and visualized by Coomassieblue and silver staining. Protein fractionscontained 2-5 mg/ml.

4.1.5.3. Reconstitution

Purified red permease solubilized in0.01% DDM or 0.1% DM was mixed withvarious amounts of detergent solubilizedphospholipids. The lipid:protein ratios (LPR,w/w) varied between 0.5 and 1.5. The lipidsused were 1:1 (w/w) mixtures ofDMPC:POPC. They were solubilized in 2%OG or a 1:1 (v/v) mixture of 2% OG and0.2% DM. The protein-lipid mixtures wereincubated for 6 hours at room temperature,and then dialyzed against detergent-freebuffer in a temperature controlled dialysisapparatus (Jap et al., 1992). The best resultswere obtained with 10 mM Tris (pH 7.5), 150

94

mM NaCl, 0.005% NaN3 and 40 mMMgCl2. In a typical dialysis experiment, thesample was dialyzed for 12 h at 13 °C. Thetemperature was then increased linearly to 37°C over the next 24 h, and kept at 37 °C for24 h. It was subsequently continuouslylowered over a 12 h period to 13 °C. Largecrystalline sheets were observed after furtherdialysis for 3 days at 13 °C. The reconstitutedmembranes were stored at 4 °C.

4.1.5.4. Electron microscopy and imageprocessing

Solubilized particles and crystalline sheetswere adsorbed for 1 min to parlodion carbon-coated grids rendered hydrophilic by glow-discharge in a low pressure of air, washedwith 3 drops of water, and stained with 0.75%uranyl formate. Images were recorded with aHitachi H-7000 electron microscope operatedat 100 kV.

LacY 2D crystals mixed with 3 - 10%trehalose were adsorbed to carbon-coatedcopper electron microscopy grids. The gridswere blotted and plunged into liquid ethane.Vitrified specimens were recorded in aHitachi H8000 electron microscope with aLaB6 filament, operated at 200 kV under lowdose conditions (~ 5 e- / Å2).

Electron micrographs of densely packedvesicles selected for image processing weredigitized using a Leafscan-45 scanner (Leafsystems, Inc., Westborough, MA, USA).

All image processing steps describedbelow were performed using the SEMPERimage processing system (Saxton et al.,1979). For single-particle analysis a referencewas established by selecting a particularlywell preserved triangularly shaped particles(such as that indicated by a circle in Fig. 1a).A 20-fold rotational symmetry was applied tothis reference resulting in a circle withdimensions of the LacY triangles (Thuman-Commike & Chiu, 1996) . The cross-correlation function of such a reference with afield of triangle-shaped particles revealed

correlation peaks at the particle positions,irrespective of their angular orientation. Peakcoordinate lists served to extract subframesfor subsequent alignment.

Micrographs of 2D crystals were selectedwith an optical diffractometer based on thealignment of the microscope and the crystalorder. Suitable areas were digitized and theircrystal quality and unit cell morphologyassessed by Fourier peak filtration (Aebi etal., 1973).

4.1.6. References

Aebi, U., Smith, P. R., Dubochet, J., Henry, C.& Kellenberger, E. (1973). A study of thestructure of the T-layer of bacillus brevis.J. Supramol. Struct. 1, 498-522.

Carrasco, N., Viitanen, P., Herzlinger, D. &Kaback, H. R. (1984). Monoclonalantibodies against the lac carrier proteinfrom Escherichia coli. 1. Functionalstudies. Biochemistry 23, 3681-3687.

Frillingos, S., Gonzalez, A. & Kaback, H.(1998a). Cysteine-scanning mutagenesisof helix IV and the adjoining loops in thelactose permease of Escherichia coli: Glu126 and Arg 144 are essential.Biochemistry 47, 14284-14290.

Frillingos, S. & Kaback, H. R. (1996).Monoclonal antibody 4B1 alters the pKaof a carboxylic acid at position 325 (helixX) of the lactose permease of Escherichiacoli. Biochemistry 35, 10166-71.

Frillingos, S., Sahin-Toth, M., Wu, J. &Kaback, H. R. (1998b). Cys-scanningmutagenesis: a novel approach to struc-ture-function relationships in polytopicmembrane proteins. FASEB J in press.

Jap, B. K., Zulauf, M., Scheybani, T., Hefti,A., Baumeister, W., Aebi, U. & Engel, A.(1992). 2D crystallization: from art toscience. Ultramicroscopy 46, 45-84.

Kaback, H. R. (1976). Molecular biology andenergetics of membrane transport. J. CellPhysiol. 89, 575-593.

Kaback, H. R. (1983). The lac carrier proteinin Escherichia coli. J Membr Biol 76, 95-112.

95

Kaback, H. R. (1996). The lactose permeaseof Escherichia coli: past, present andfuture. In Handbook of BiologicalPhysics: Transport processes inEukaryotic and prokaryotic Organisms.Vol II. Edited by Konings 2n, KabackHR, Lolkema JS. Amsterdam: Elsevier,203-227.

Kaback, H. R. (1997). A molecularmechanism for energy coupling in amembrane transport protein, the lactosepermease of Escherichia coli. Proc Natl.Acad. Sci U S A 94, 5539-43.

Kaback, H. R., Voss, J. & Wu, J. (1997).Helix packing in polytopic membraneproteins: the lactose permease ofEscherichia coli. Curr Opin Struct Biol 7,537-42.

Kaback, H. R. & Wu, J. (1997). Frommembrane to molecule to the third aminoacid from the left with a membranetransport protein. Q Rev Biophys 30, 333-64.

Le Coutre, J., Narasimhan, L. R., Kumar, C.,Patel, N. & Kaback, H. R. (1997). Thelipid bilayer determines helical tilt angleand function in lactose permease ofEscherichia coli. Proc. Natl. Acad. Sci. 94,10167-10171.

Li, J. & Tooth, P. (1987). Size and shape ofthe Escherichia coli lactose permeasemeasured in filamentous arrays.Biochemistry 26, 4816-4823.

Loddenkotter, B., Kammerer, B., Fischer, K.& Flügge, U.-I. (1993). Expression of thefunctional mature chloroplast triosephosphate translocator in yeast internalmembrane and purification of thehistidine-tagged protein by a single metal-affinity chromatography step. Proc. Natl.Acad. Sci. U S A 90, 2155-2159.

Poolman, B. & Konings, W. N. (1993).Secondary solute transport in bacteria.Biochem. Biophys. Acta 1183, 5-39.

Privé, G. G., Verner, G. E., Weitzman, C.,Zen, K. H., Eisenberg, D. & Kaback, H.R. (1994). Fusion proteins as tools forcrystallization: The lactose permease fromEscherichia coli. Acta Cryst. 50, 375-379.

Rigaud, J. L., Mosser, G., Lacapere, J. J.,Olofsson, A., Levy, D. & Ranck, J. L.(1997). Bio-Beads: An efficient strategyfor two-dimensional crystallization ofmembrane proteins. J. Struct. Biol. 118,226-235.

Sahin-Tóth, M., Lawrence, M. C. & Kaback,H. R. (1994). Properties of permeasedimer, a fusion protein containing twolactose permease molecules fromEscherichia coli. Proc Natl. Acad. Sci U SA 91, 5421-5.

Saxton, W. O., Pitt, J. T. & Horner, M.(1979). Digital image processing: theSemper system. Ultramicroscopy 4, 343-354.

Sun, J., Wu, J., Carrasco, N. & Kaback, H. R.(1996). Identification of the epitope formonoclonal antibody 4B1 whichuncouples lactose and proton translocationin the lactose permease of Escherichia coli.Biochemistry 35, 990-8.

Thuman-Commike, P. A. & Chiu, W. (1996).PTOOL: a software package for theselection of particles from electroncryomicroscopy spot-scan images. JStruct Biol 116, 41-7.

Viitanen, P., Newman, M. J., Foster, D. L.,Wilson, T. H. & Kaback, H. R. (1986).Purification, reconstitution, andcharacterization of the lac permease ofEscherichia coli. Methods Enzymol. 125,429-52.


96

5. General discussionand conclusions

99

5. General discussion and conclusions

This thesis presents a new preparationmethod for 2D crystals grown on afunctionalized lipid monolayer for highresolution atomic force microscopy and newstructural data on several physiologicallyrelevant membrane proteins derived fromatomic force and electron microscopyinvestigations.

Chapter 1 presents a technical developmentfor high resolution AFM. In order to developa preparation technique which allows thestudy of 2D crystals of water soluble proteinsgrown on functionalized lipid monolayerswith the AFM, 2D crystals of streptavidinwere assembled on biotinylated lipidmonolayers and transferred into the AFMusing highly oriented pyrolytic graphite(HOPG) as sample support. Parallelhydrophobicity measurements revealed thatHOPG surfaces are more hydrophobic thansilanized glass, allowing protein crystalsgrown on lipid monolayers to be stablyattached and imaged using AFM. Therecently developed technique for 2Dcrystallization of membrane proteins by Levyet al. (1999) using functionalized lipidmonolayers together with the novelpreparation technique presented in this thesisopens a wide range of membrane proteins tobe investigated: All his-tagged proteins maybe immobilized on NiNTA functionalizedlipid monolayers and investigated underphysiological conditions. This will giveinsight into the structure of membraneproteins, which have resisted crystallizationby other techniques.

The use of defined scanning buffers whichcompensate the surface charges of the AFMtip and the sample (Müller et al., 1999) hasallowed protein surfaces to be contoured at alateral resolution of ~7Å and a far bettervertical resolution. The extremely high signal-to-noise ratio of the AFM makes the

achievement of such high resolution alsopossible for poorly ordered proteins asdocumented for three membrane and twowater soluble proteins (Scheuring et al., 1999;Fotiadis et al., 2000; Seelert et al., 2000; Mouet al., 1996). Single particle averaging of themolecules yields a representative surface ofthe protein species revealing theoligomerization and loop organization.

In Chapter 2, the use of high resolutionAFM to investigate the Escherichia coli waterchannel AqpZ and the Rubrivivax gelatinosuslight-harvesting complex LH2 is presented.This lead to new insights into the organizationof their loops and termini and assignment oftheir sidedness.

AqpZ was imaged before and aftercleavage of the N-terminal his-tag withtrypsin. Proteolysis induced a drastic changeof surface appearance and height over thelipid membrane on one side of the protein,which consequently could be assigned ascytoplasmic. The extracellular surface,however showed strong changes oftopographical features when imaged with aloading force of ~100pN applied to the tip:the largest protrusion was strongly displacedby the tip. Together with flexibility mappingand volume calculations this protrusion wasassigned as the loop C.

Similarly, LH2 was imaged underphysiological conditions before and afterprotease treatment. In this case, thermolysinwas used to specifically cleave the C-terminusof the α-subunit. Absorption spectra showedthat both the native and the cleaved proteinscontained all the chromophores. The moststrongly protruding surface changed both it'sheight over the lipid membrane and it'ssurface appearance upon protease treatment.This determined the periplasmic surface andthe position of the C-terminus of the α-subunit. In reconstitutions from less

101

stringently purified samples, large rings,assigned as light-harvesting complexes 1were imaged in interaction with LH2 rings,supporting the proposed model of the actionmechanism of the photosynthetic apparatus(Kühlbrandt, 1995).

Because of it's high signal-to-noise ratiohigh resolution AFM allows loops andtermini of proteins to be identified and thedifferent protein species within a membraneto be distinguished on the raw data imageswithout further processing. This opens theperspectives of imaging native membranesunder close to native conditions in futurestudies.

Chapter 3 presents the combination ofsurface and projection techniques to elucidatethe sidedness of aquaporins. Atomic forcemicroscopy studies combined withproteolysis experiments lead to anunambiguous assignment of surfacesidedness of membrane proteins (see Chapter2). Cryo-electron microscopy density mapselucidate the intramembraneous parts of suchmolecules (Walz et al., 1997). In order toassign the sidedness information from AFMexperiments to a density map, large 2Dcrystals of AqpZ were freeze dried and metalshadowed (Gross et al., 1990). Micrographswere taken where aggregates had blocked thedeposition of metal on parts of a crystal,hence surface and density information couldbe obtained from one crystal. Datasets fromsuch preparations were used to align the highresolution AFM topography to the highresolution cryo-EM density maps of AqpZand AQP1. It was shown that the sidednessassignment previously published (Walz et al.,1996) had to be revised. The methoddescribed can be applied to any membraneprotein reconstituted into 2D crystalsallowing a unambiguous surface sidednessassignment to a density map.

Chapter 4 reports on structural studies ofthe Escherichia coli lactose permease (LacY)using electron microscopy techniques. Thefusion protein LacY/cytb562/C417His6 (red

permease) was overexpressed in it's nativesystem and purified using Ni-affinitychromatography. The purified protein wasimaged in it's solubilized state revealing atrimeric organization. A low resolution 3Dreconstruction showed that trimerization isinduced by hydrophilic interactions betweenthe cytb562 of each monomer. Such trimerscould be reconstituted into lipid bilayersresulting in densely packed vesicles and 2Dcrystals (Zhuang et al., 1999). Best crystalspack the trimers in a hexagonal lattice anddiffract to ~8Å, but stacking of crystallinelayers prevented the calculation of a highresolution projection map. The observedcytochrome b562 induced trimerization of redpermease together with the finding that theprotein is stable in decyl-maltoside increasesthe range of crystallization conditions, whichcan now be investigated to obtain highresolution crystals.

Atomic force and electron microscopycover a resolution range from micrometers toangstroms and provide both surface andvolume information that is acquired underclose to native conditions. Membrane proteinswere reconstituted in lipid membranes, whichrepresent their native environment. Therefore,the results presented in this thesis meet thegoal to acquire structural information onbiological molecules under conditions, whichensure that the unseparable relationshipbetween structure and function is intact.

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

Gross, H., Krusche, K., Tittmann, P. (1990).Recent progress in high-resolutionshadowing for biological transmissionelectron microscopy. Proceedings of XIIthInternational Congress for ElectronMicroscopy, 510-511

Kuhlbrandt, W. (1995). Many wheels makelight work. Nature 374, 497-498.

102

Lévy, D., Mosser, G., Lambert, O., Moeck, G.S., Bald, D. & Rigaud, J. L. (1999). Two-dimensional crystallization on lipid layer: asuccessful approach for membraneproteins. J. Struct. Biol. 127, 44-52.

Mou, J., Sheng, S., Ho, R. & Shao, Z. (1996).Chaperonins GroEL and GroES: viewsfrom atomic fore microscopy. Biophys. J.71, 2213-2221.


Scheuring, S., Ringler, P., Borgnia, M.,Müller, D. J., Agre, P. & Engel, A. (1999).High resolution topographs of theEscherichia coli waterchannel AqpZ.EMBO Journal 18, 4981-4987

Seelert, H., Poetsch, A., Dencher, N., Engel,A., Stahlberg, H. & Müller, D. J. (2000).Proton-powered turbine of a plant motor.Nature 405, 418-419.


Walz, T., Hirai, T., Murata, K., Heymann, J.B., Mitsuoka, A., Fujiyoshi, Y., Smith, B.L., Agre, P. & Engel, A. (1997). The 6Åthree-dimensional structure of aquaporin-1. Nature 387, 624-627.


103

6. Acknowledgment

105

6. Acknowledgment

I thank Prof. Dr. Andreas Engel. Youalways showed up with an extremely goodfeeling for the situations I was in during myphD. After my diploma, when I thought I hadto stop there with science, you pushed mealmost against the wall, telling me that youbelieved I had a certain talent. Then,sometimes you put me under more pressure,sometimes less - sometimes you just smiledor went running with me. Now, looking backI have the impression it was always exactlywhat I needed. Fortunately, you thought meafter a certain time to remain in a,scientifically spoken, stable conformation andour work got better and more efficient.Maybe I remember mostly your reaction afterthe AFM slipped from this carriage I wassteering. You examined the surface of thecarriage and said: "It's slippery, it could havehappened to anyone.", and it was never a topicagain. Scientifically you are full of knowledgeand full of ideas. I grabbed as much of yourknowledge as I could, and you made mecombine your ideas and mine. I wish, and Iam sure, that everything you take into yourhands in your personal and professional lifewill be conducted by this force I have seenyou can give!

I thank Daniel J. Müller. It was you whomotivated me to join the lab, and it was youwho trained me in AFM. During the nextyears, we had a lot of ups and downs, whichmight be because we are very different or verysimilar in our minds, something, which I havenot yet figured out. Now, for some timeeverything is superb for work as for going tothe bars together. Besides the AFM I havelearned from you how to make beautifulfigures, how to talk to people, and how to putcreativity into my work. The next thingshould be, how to become extremelyorganized. I am convinced that you will do

well, when you face the responsibility forleading your own group: may the force bewith you!

I thank Henning Stahlberg. You enteredthe lab just about two years ago and I can nottell how much knowledge and character youbrought into the group. You are a greatscientist and in addition you take the time tocommunicate what you know. Most of what Iknow about computers and image processingI learned from you, during the time youhelped me writing these routines (for volumeand probability calculations). Recently youalso showed me how to do cryo-EM. Yourmodest ways and integrity should be honored,and I hope for you that you will succeed inyour personal and professional life!

I thank Shirley Müller. You shaped upsurely every manuscript which ever left mydesk. Aware of the importance oftransmission of data into a understandablewritten form, I owe you a lot. Unfortunately Inever had a STEM project, so we neverworked on a subject together, neverthelessyour clear and scientific brain helped me andeveryone in the group when you statedsomething, which was probably true, in agroup seminar. It was also you, who listenedcarefully when I ran around with myaquaporin sidedness theory. Thanks for that.You are extremely important for this group,scientifically exact, and humanly warmhearted, keeping all us boys around you onthe ground.

I thank Philippe Ringler. It was greatworking with you on the AqpZ project. I wishfor you and your family all the best. You are agood microscopist, it's a pity that you do notuse the microscope at the moment. But maybeyou will come back to microscopy and make

107

the beautiful Goldringler images.

I thank Bernard Heymann. You supportedstrongly the writing of the streptavidin story.Later we mostly disagreed which taught me tohold on to my opinions.

I thank Dimitrios Fotiadis. Hey Dimiman,as you know, we are not the same. But themore time we spent together (and now it'smore than 8 years!!!), the better we do. Wedid our diploma and phD thesis together andhad some good laughs together, and I assumethey won't be the last. You are an extremelyexact scientist. If you put something in yourhead, you surely get there. I wish you all thebest, and if we ever might be in labs far apart,we'll meet for a beer in Rio-Bar, or a coffee inthe Biozentrum cafeteria, and see how timepasses by.

I thank Lorenz Hasler. I always liked yourways, although we do not have "s'Heu uf drgliche Bühni". Everyone around you can besure, that if you say something, you mean itexactly that way (...and I have there someexperience). You were always straight, helpfuland supported me with biochemical problems.I wish you all the best for your future inAmerica and hope to see you sooner or later.Backgammon rocks, and sorry I took yourpipette so often!

I thank Thomas Braun. You are a honestand modest person. Your computerknowledge is way beyond my understanding(I can't promise you here that I will learn'Latex'). You are daily asking the group forlunch and by this social being together, youcreate a great atmosphere in the group. I wishyou all the best for the rest of your phD!

I thank Herve Remigy. You have a greatsense of humor and made me smile a lot oftimes during the last years. Now you have puta lot of energy into the dilution machine, and Ihope it will help to crystallize difficult

proteins. You have a lot of skills, hopefullyyou will use them all to solve the problemsahead. Good luck.

I thank Clemens Möller. You are asensible colleague, something like thebarometer of the group, able to spread a calmand peaceful atmosphere. I hope you will getmore self confidence and stronger pushingyourself and your opinions through. Yousurely bring a great knowledge of physics tothe group and this in the context of structuralbiology makes a great combination. We havea very different music taste, which made itvery calm measuring AFM together, sinceyou could not stand my music, similarly inthe other direction. I wish you all the best onyour way.

I wish Gabriel Fedrigo all the best for hisphD. You just started a few weeks ago, andput yourself already deeply into the blackboxof 2D crystallization. Everyone, who workswith electron crystallography would like toknow more what's going on in there,hopefully your work, supported by thebiophysical background you bring with you,will elucidate some of the secrets. All thebest!

I wish Andreas Schenk and ThomasKaufmann all the best for their diplomathesis. I think you two are extremelyintelligent and interested in this research andwill become specialists soon. For Thomas Ihope also that we will have a good timetogether and work successfully. I amconvinced we can make large LacY crystals.

I thank Christine Widmer. You supportedthe difficult LacY project a lot, and I hope wewill do more in the near future. After somedifficulties in the beginning, we have bothlearned how to calm the waves. I hope wesoon get good crystals, consequently usingmore the microscope than the french press. Iwish you all the best for your personal life

108

with family and hobbies.

I thank Kitaru Suda. Like a magician youare doing biochemistry. And like a magicianyou play the guitar. Your wisdom is broad inscience, history and culture. You even knowthe Swiss folk songs better than anyone ofus. You help everybody. What can I say?! -We are all lucky having you around. Thankyou again for the millions of little and bighelps.

I thank Ueli Aebi. It is an extraordinaryinstitute of which I could use all the facilitiesand infrastructure. I know, you are stronglyinvolved in the background organization ofeverything here, work I do not see directly,but which leads to this effective andstimulating atmosphere. Thanks too, fortaking out your credit card and saving my lifein this hotel in Atlanta.

I thank Cora-Ann Schönenberger, ClaireGoldsbury and Martin Stolz. You amused mealways a lot when passing by on the 3rd

floor. Claire, thanks again for the amyloidfibrils used during the sidedness assignmentexperiments.

I thank Robert Häring, Robert Wyss andRoland Bürki. Thanks to you, the MIH isrunning. El-Röbi and Rolli take surely 90%of all the computer problems away from us,meanwhile Mech-Röbi fixes the centrifuges.Taking this together I assume, if you were notaround I would just be writing the first line.Thanks!

I thank Barbara Merz. Your work on the3rd floor makes these lab so efficient. Also,you always prepare beautiful plates for the3rd floor parties.

I thank Urs Fürstenberger. Fortunately,you are floormanager and not complicated atthe same time. Thanks to you, I got thechemicals ordered quickly , the packets weresent immediately, the money reimbursed intime. Good luck with your new job, nowtaking care of the whole Biozentrum!

I thank Beat Schumacher. For a fewmonths now you have been the floormanager,and you do your job extremely well.Everything goes quick, go on so!

I thank Peter Tittmann and Heinz Gross. Ihad a great time with you two in Zürich doingthe sidedness experiments on AqpZ. I hopeyou will use the advantage of your machineagain to get surface and projectioninformation on one crystal. Working withyou is a pleasant thing, there was sciencework and good discussions during the coffeetimes. Peter, thanks for the hours you spenton the microscope with me.

Merci, Jean-Louis Rigaud. Depuis qu'on acommence a collaborer, j'avais deja deux foisde temps exelente à Paris, grâce a toi. On avaittravaile efficace et avec des idées, et on avaitbien rigoler ensemble les soirées apres leboulot. En plus tu as un humeur uncroyable.Je serais heureux, si on continue a travaillerensemble. J'ai bien trouvé avec toi quelqu'un,duquel je peux beaucoup apprendre pour letravail et la vie. Merci pour tous.

Merci, Daniel Levy. Tu m'avais appris defaire la technique de la crystallisationmonocouche. En plus tu es un très bonscientifique, qui prend son temps a expliquercomment faire les manipes. J'espere que lesgens vont remarquer que la technique que tuavais introduit est forte. Tout le bon pour tavie avec ta jeune famille et la science.

109

7. Curriculum vitae

111

7. Curriculum vitae

Simon Andreas Scheuringborn in Basel, Switzerland, on March 8,1973

7.1. Education

1979-1984 Primary school (Basel)

1984-1992 High school (Basel), Matura(Typus B)

1992-1997 Diploma in molecular biology,Biozentrum, University Basel.

Thesis topic: "Structure and function ofGroEL and its complexes investigated withthe EM and the AFM" (in the group of Prof.A. Engel.)

1997-2000 PhD thesis in structuralbiology, Biozentrum, University Basel.

Thesis topic: "atomic force and electronmicroscopic analysis of membrane channelsand transporters" (in the group of Prof. A.Engel.)

7.2. Teaching

1996 Instructor in the "Blockkurs inBiophysik und Strukturbiologie"

1997 Instructor in the "EMBO PracticalCourse in Scanning Probe Microscopy" atthe Biozentrum, Basel, Switzerland.

Instructor in the "Blockkurs in Biophysikund Strukturbiologie"

1998 Teacher in the "Practical Course inAtomic Force Microscopy in Biology" at theBiozentrum, Basel, Switzerland.


1999 Tutoring for students of Biology atthe Biozentrum at the University of Basel


2000 Teacher in the "EMBO PracticalCourse of Current Methods in MembraneProtein Research" at the EMBL, Heidelberg,Germany.


7.3. Publications

Single proteins observed by atomic forcemicroscopy.

Single Molecules, IN PRESSSimon Scheuring, Dimitrios Fotiadis,

Clemens Möller, Andreas Engel and Daniel J.Müller

Imaging and manipulation of biologicalstructures with the AFM.

Micron, SUBMITTEDDimitrios Fotiadis, Simon Scheuring,

Andreas Engel and Daniel J. Müller

High resolution AFM topographs of theRubrivivax gelatinosus light-harvestingcomplex LH2.

The EMBO Journal, IN PRESSSimon Scheuring, Francoise Reiss-

Husson, Andreas Engel, Jean-Louis Rigaudand Jean-Luc Ranck

Conformational changes, flexibilities andintramolecular forces observed on individualproteins using AFM.

Single Molecule, 2000, 1: 115-118Daniel J. Müller, Dimitrios Fotiadis,

Clemens Möller, Simon Scheuring andAndreas Engel

The aquaporin sidedness revisited.Journal of Molecular Biology, 2000, 299

(5):1271-1278

113

Simon Scheuring, Peter Tittmann,Henning Stahlberg, Philippe Ringler, MarioBorgnia, Peter Agre, Heinz Gross andAndreas Engel

Direct observation of postadsorptionaggregation of antifreeze glycoproteins onsilicates.

Langmuir, 2000, 16 (13):5785-5789Ph. Lavalle, A. L. DeVries, C.-C. C.

Cheng, S. Scheuring and J. J. Ramsden

High resolution AFM topographs of theEscherichia coli water channel aquaporin Z.

The EMBO Journal, 1999, 18 (18):4981-4987

Simon Scheuring, Philippe Ringler, MarioBorgnia, Henning Stahlberg, Daniel J. Müller,Peter Agre and Andreas Engel

Imaging streptavidin 2D-crystals onbiotinylated lipid monolayers at highresolution with the atomic force microscope.

Journal of Microscopy, 1999, 193:28-35.Simon Scheuring, Daniel J. Müller,

Philippe Ringler, J. Bernard Heymann andAndreas Engel

Electrostatically balanced subnanometerimaging of biological specimens by atomicforce microscopy.

Biophysical Journal, 1999, 78:1101-1111Daniel J. Müller, Dimitrios Fotiadis,

Simon Scheuring, Shirley A. Müller andAndreas Engel

A novel preparation method for highresolution AFM introduced with 2D-streptavidin crystals grown on biotinlipidmonolayer.

Microscopy and Microanalysis MeetingProceedings, 1998, 4, (2), 312 - 313.

Simon Scheuring, Daniel J. Müller,Philippe Ringler, J. Bernard Heymann andAndreas Engel

7.4. Meetings

Raster-Sonden-Mikroskopien undOrganische Materialien V. Diskussions-tagung, 7 – 9 Oktober 1996, Münster,Deutschland.

Präsentation: "Die E. coli ChaperoninGroEL untersucht mit dem EM und demAFM". Simon Scheuring, Daniel J. Müllerand Andreas Engel.

Raster-Sonden-Mikroskopien undOrganische Materialien VI. Diskussions-tagung, 8 - 10 Oktober 1997, Tübingen,Deutschland.

Präsentation: "A novel preparation methodfor high resolution AFM introduced with 2D-streptavidin crystals grown on biotinlipidmonolayer". Simon Scheuring, Daniel J.Müller and Andreas Engel.

Microscopy and Microanalysis '98, July12-16, Atlanta, Georgia, USA.

Presentation: "A novel preparation methodfor high resolution AFM introduced with 2D-streptavidin crystals grown on biotinlipidmonolayer". Simon Scheuring, Daniel J.Müller, Philippe Ringler, J. BernardHeymann and Andreas Engel.

Raster-Sonden-Mikroskopien undOrganische Materialien VII. Diskussions-tagung, 7 - 9 Oktober 1998, Berlin,Deutschland.

Präsentation: "High resolution AFMtopographs identify sides, handedness andloops of the E. coli waterchannel AqpZ".Simon Scheuring, Mario Borgnia, Daniel J.Müller, Peter Agre and Andreas Engel.

Meeting of the EU-Biotech Program:"Water and glycerol channels from the MIPfamily: structure, function, regulation andexploitation". 15 - 17 September, 1999,Hamburg, Deutschland.

Presentation: "The sidedness and loops ofAqpZ". Simon Scheuring, Philippe Ringler,

114

Mario Borgnia, Henning Stahlberg, Daniel J.Müller, Peter Agre and Andreas Engel.

Raster-Sonden-Mikroskopien und Orga-nische Materialien VIII. Diskussionstagung,4 - 6 Oktober 1999, Basel, Schweiz.

Meeting of the EU-Biotech Program:"Water and glycerol channels from the MIPfamily: structure, function, regulation andexploitation". 15 - 17 March, 2000, Haute-Nendaz, Schweiz.

Presentation: "The surface of AQP2investigated with the atomic forcemicroscope". Simon Scheuring, LorenzHasler, Paul Werten, Peter Deen and AndreasEngel.

3rd International Conference on"Molecular biology and physiology of waterand solute transport". July 1 - 5, 2000,Göteborg, Sweden.

Poster: "The Sidedness of Aquaporinsdetermined by atomic force and transmissionelectron microscopy". Simon Scheuring,Peter Tittmann, Henning Stahlberg, PhilippeRingler, Mario Borgnia, Peter Agre, HeinzGross and Andreas Engel.

Rastersondenmikroskopie in Forschungund Industrie, 10. Oktober 2000, Wuppertal,Deutschland.

Präsentation: "Neue biologische Appli-kationen in der Rasterkraftmikroskopie".Simon Scheuring and Andreas Engel.

115

phdthesis scheuring screen

Documents