Three-Dimensional Reconstruction ofAxonemal Outer Dynein Arms In Situ by

Electron Tomography

Pietro Lupetti,1 Salvatore Lanzavecchia,2 David Mercati,1 Francesca Cantele,2

Romano Dallai,1 and Caterina Mencarelli1*

1Laboratory of Cryotechniques for Electron Microscopy,Dipartimento di Biologia Evolutiva, Universita di Siena, I-53100 Siena, Italy

2Dipartimento di Chimica Strutturale e Stereochimica Inorganica,Universita di Milano, I-20133 Milan, Italy

We present here for the first time a 3D reconstruction of in situ axonemal outerdynein arms. This reconstruction has been obtained by electron tomographyapplied to a series of tilted images collected from metal replicas of rapidly frozen,cryofractured, and metal-replicated sperm axonemes of the cecidomid dipteranMonarthropalpus flavus. This peculiar axonemal model consists of several micro-tubular laminae that proved to be particularly suitable for this type of analysis.These laminae are sufficiently planar to allow the visualization of many dyneinmolecules within the same fracture face, allowing us to recover a significant num-ber of equivalent objects and to improve the signal-to-noise ratio of the recon-struction by applying advanced averaging protocols. The 3D model we obtainedshowed the following interesting structural features: First, each dynein arm hastwo head domains that are almost parallel and are obliquely oriented with respectto the longitudinal axis of microtubules. The two heads are therefore positioned atdifferent distances from the surface of the A-tubule. Second, each head domainconsists of a series of globular subdomains that are positioned on the same plane.Third, a stalk domain originates as a conical region from the proximal head andends with a small globular domain that contacts the B-tubule. Fourth, the stemregion comprises several globular subdomains and presents two distinct points ofanchorage to the surface of the A-tubule. Finally, and most importantly, contraryto what has been observed in isolated dynein molecules adsorbed to flat surfaces,the stalk and the stem domains are not in the same plane as the head. Cell Motil.Cytoskeleton 62:69–83, 2005. ' 2005 Wiley-Liss, Inc.

Key words: dynein; electron microscopy; microtubule-based motors; quick-freeze; deep-etch

INTRODUCTION

Axonemal and cytoplasmic dyneins are minus end-directed microtubule-based molecular motors. They arelarge molecular assemblies, with a mass of 0.6–2 milliondaltons, and are composed of 1–3 heavy chains (MW�520 kDa) as well as a number of intermediate chains(MW 60–135 kDa) and light chains (MW <25 kDa)[Mitchell, 1994; Porter, 1996]. The heavy chain behavesas a molecular motor, possesses a hydrolytic site forATP, and is able to bind to microtubules in an ATP-dependent manner.

*Correspondence to: Caterina Mencarelli, Dipartimento di Biologia

Evolutiva, Universita di Siena, via Aldo Moro 2, I-53100 Siena, Italy.

E-mail: mencarelli@unisi.it

Contract grant sponsors: University of Siena (PAR) to C.M. and D.M.

and MIUR (PRIN) to P.L. and S.L.

P. Lupetti and S. Lanzavecchia contributed equally to this work.

Received 2 March 2005; Accepted 9 June 2005

Published online in Wiley InterScience (www.interscience.wiley.

com).

DOI: 10.1002/cm.20084

' 2005 Wiley-Liss, Inc.

Cell Motility and the Cytoskeleton 62:69–83 (2005)

In cilia and flagella, axonemal movement resultsfrom the sliding of adjacent doublet microtubules, pow-ered by the coordinated activity of different dyneinisoforms that are arranged on the microtubules in twolongitudinal rows of outer (ODAs) and inner (IDAs)dynein arms. A detailed knowledge of dynein arm archi-tecture is a fundamental requirement for understandingthe process that converts dynein conformational changesand ATP hydrolysis into microtubule doublet sliding.

Data are available on the structure of isolateddynein molecules that have been adsorbed onto mica andvisualized by quick-freeze, deep-etch (QF/DE) micro-scopy [Goodenough and Heuser, 1982, 1984; Sale et al.,1985]. More recently, a detailed structural–functionalanalysis has been performed on dynein isoforms purifiedfrom the inner arm; in this study, electron micrographsof negatively stained molecules were analyzed by com-puter-aided image-processing techniques [Burgess et al.,2003, 2004]. These studies have shown that each dyneincomplex consists of a head domain carrying the ATPasesite, and two other domains emerging from the head: aslender stalk, which is able to bind to the B-tubule in anATP-dependent manner, and an elongated stem domain,which binds to the A-tubule.

Information on dynein arm organization in situ hascome principally from the analysis of metal replicas ofquick-frozen, deep-etched axonemes of Chlamydomonasflagella [Goodenough and Heuser, 1982, 1984]. The insitu 3D appearance of the dynein arm is well preservedby this procedure. More recently, Burgess et al. [1991]introduced the use of computer-aided image analyses ofmetal replicas to reduce the noise generated by localminor structural variations along the axoneme and thesignal disturbances inherent in metal replica procedures.These authors used a combination of digital and humanimage processing, in which the computer was used toenhance the signal-to-noise ratio (S/N) of linear arrays ofarms, images of which were then combined by eye toachieve a 3D model.

At present, one of the best ways to obtain 3D infor-mation on protein complexes is to reconstruct models by3D electron microscopy [Frank, 1996]. This techniquehas been used when negative staining was the only avail-able technique for sample preparation, and was later im-proved by cryo-electron microscopy of frozen hydratedsamples [for a review, see Baumeister and Steven,2000]. When metal replicas are used, 3D models havebeen computed for molecules packed in 2D crystals[Aquaporin 1: Walz et al., 1996], in helices (actin fila-ments: Morris et al., 1994], or for isolated molecules[VacA: Lanzavecchia et al., 1998; Aquaporin 0: Zampighiet al., 2003].

For the study of Aquaporin crystals, a surface reliefreconstruction strategy was adopted [Baumeister and

Lembcke, 1992]; however, this technique cannot beapplied in the case of bent surfaces. In all other cases, 3Dreconstruction from projections was applied according tostandard 3D electron microscopy techniques. Heliceshave been reconstructed from single untilted exposures,while isolated particles were reconstructed from pairs ofmicrographs (tilted and untilted) by random conical tiltstrategy [Radermacher et al., 1987]. The model obtainedby 3D reconstruction, however, represents the metalmould rather than the actual structure under study. Never-theless, the outer envelope of the structure can be inferredif an imprint is computationally extracted from the mold[Lanzavecchia et al., 1998; Zampighi et al., 2003].

A 3D model of in situ flagellar dynein outer armcan best be obtained by EM tomography, which is basedon taking a series of micrographs as the specimen istilted around one or two axes [Baumeister et al., 1999].In the last few years, electron tomography has providednew opportunities for describing the structure of subcel-lular assemblies and has become a powerful tool forstructural analysis of molecular complexes and organ-elles in situ [Subramaniam and Milne, 2004; Baumeister,2005]. The most common starting materials for suchanalyses are 100–200-nm-thick sections from plastic-embedded samples [see, e.g., He et al., 2003]. Alterna-tively, frozen hydrated samples embedded in vitrified icehave also be used [Medalia et al., 2002; Steven and Aebi,2003]. Electron tomography of dynein has been appliedprimarily to Epon-embedded and thick-sectioned axo-nemes and basal bodies, which have been viewed byhigh-voltage EM [O’Toole et al., 2003], or has beendone on negatively stained samples. Additionally, EMtomography has been used to analyze frozen hydratedaxonemes in the cryoEM [McEwen et al., 2002]. How-ever, the observation of frozen hydrated samples requiressophisticated and expensive instrumentation.

In this study we have employed, for the first time,various techniques of EM tomography to analyze dyneinarms in situ from quick-frozen, deep-etched flagellaraxonemes. This strategy allows one to analyze structuraldetails that are conserved in metal replicas of rapidly fro-zen material. Such replicas are very stable over time andcan be observed at room temperature with a conventionaltransmission electron microscope. Moreover, the effi-cacy of preserving and visualizing the fine structuraldetails of axonemal complexes by rapid freezing, fol-lowed by cryofracture and metal replication is nowwell established [Goodenough and Heuser, 1982, 1984;Burgess et al., 1991; Lupetti et al., 1998; Mencarelliet al., 2001]. Nevertheless, the suitability of metal re-plicas for tomographic reconstructions has until nowremained unknown.

We have used the unusual sperm flagellar axo-nemes of the cecidomid dipteran Monarthropalpus

70 Lupetti et al.

flavus, which has been extensively characterized by con-ventional fixation, embedding and thin-sectioning proce-dures, as well as by previous QF/DE preparations[Lupetti et al., 1998]. These axonemes have two highlyuseful features for EM tomography: (1) the absence ofstructures, such as the nexin links, radial spokes, and theIDAs, all of which in the usual 9þ2 axoneme mightinterfere with the visualization of dynein arms and (2)the considerable diameter of the axoneme, which allowsone to obtain almost ‘‘planar’’ images of several side-by-side outer doublet microtubules with their aligned dyneinarms, all visualized from the same viewing angle. Suchimages constitute excellent material for computerizedanalysis and 3D reconstruction of dynein arms.

MATERIALS AND METHODS

Sample Preparation and Electron Microscopy

For thin sectioning, intact deferent ducts were dis-sected from adult males of M. flavus in a buffer consist-ing of 70 mM K-acetate, 40 mM Hepes, 5 mM MgSO4,1 mM EDTA, and 0.2 M sucrose, pH 7.9, and fixed for1 h at 48C in 2.5% glutaraldehyde diluted in the samebuffer. Deferent ducts were then washed in buffer over-night and postfixed in 1% osmium tetroxide for 1 h at48C. They were then dehydrated with a graded series ofethanols and embedded in Epon-Araldite. Thin sectionswere routinely stained with uranyl acetate and lead citrate.

For deep-etch electron microscopy, spermatozoaobtained by opening the deferent ducts were demem-branated by addition of 0.1% Triton X-100 to the suspen-sion, followed by a 15-min incubation at 48C. Demem-branated spermatozoa were rinsed twice by resuspensionin the above buffer without Triton X-100 and subsequentcentrifugation at 15,000g for 10 min at 48C. The finalpellet was processed for electron microscopy by thequick-freeze, deep-etch (QF/DE), rotary replication pro-cedure [Heuser, 1981] as described by Lupetti et al.[1998]. Tilt series images from replicas of the best pre-served axonemes were collected at steps of 1.58 from�458 to þ458 for a total of 61 images per series. Thesample was manually oriented so that the tilt axis laysapproximately along the longitudinal axis of the axo-neme.

Electron micrographs were recorded at a magnifi-cation of 39,0003 and digitized with an Umax scannerPowerLook 3000 at 21 lm per sample (0.53 nm/pixel).

Image Processing

Analysis of the periodicity and 2D averaging.After a visual selection of good areas of the replicas, andcollection of the tilt series, the quality of the untiltedimages was first evaluated. The axonemal regions thatwere analyzed present two types of periodicities, orthog-

onal to each other. One periodicity ran along the axone-mal axis, corresponding to dynein arms (24 nm), withharmonics up to 6 nm. The 8-nm line was enforced alsoin this axis because it corresponded to the periodicity ofthe tubulin dimers. The second, orthogonal periodicitywas generated by parallelism of the microtubule doubletsin the axoneme. Corresponding spectral spots werebroadened because of small bending of the doublets ineach sheet. The area chosen for subsequent recons-truction was selected by analyzing the regularity of theperiodic pattern. To do this, a portion of the untilted pro-jection was Fourier-filtered in one direction, imposing a24-nm periodicity. This filtering processed each row ofdoublets individually, so that only dynein arms belong-ing to the same row were averaged, while no averagewas performed between arms belonging to differentrows. The most detailed averages were those fromdynein rows originally displaying the most regular pitch.

Alignment of the series. The series of images waspreliminarily aligned by correlation [Frank and McEwen, 1992]. Each image was aligned with the one atnearest tilt angle. Thus, the images at �1.58 and þ1.58were aligned with the untilted one, the images at �38and þ38 were aligned with the previously aligned �1.58and þ1.58 ones, and so on until the end of the series.Subsequently, the alignment was refined using referencepoints [see Lawrence, 1992]. Because of the high con-trast characteristic of metal replicas, the reference pointswere selected using a number of small structural detailsthat were easily detectable in all images. Once the coor-dinates of each reference point were measured in allimages, the alignment parameters, i.e. rotations andshifts, were obtained as described by Jing and Sachs[1991].

3D reconstruction. 3D models were reconstructedby weighted back projection (WBP), under the assump-tion that the tilt angles were correctly determined duringthe image recording. Large maps of 512 3 512 3 128voxels, corresponding to the center of the replica, werecomputed. The results were visualized by using Amira(TGS).

Restoration and filtering. The design of goniome-ters currently used in EM tomography is such that it isimpossible to collect complete tilt series. Tomographicreconstructions are therefore negatively influenced bythe well-known problem of missing regions. In our case,the missing wedge in Fourier space is quite large,because of the limited tilt-angle of the goniometric stagebeing used. The effect of the missing wedge consists ofan elongation of the reconstruction along the Z-axis[Frank and Radermacher, 1986]. This effect can be parti-ally corrected for by the technique of ‘‘projection onto

3D Modeling of Outer Dynein Arm In Situ 71

convex sets’’ [POCS: Carazo and Carrascosa, 1987;Carazo, 1992]. A POCS filter with 50 iterations, basedon the positivity constraint, was applied to fill the miss-ing volume in Fourier space. This filter also improvedthe contrast of the reconstruction.

In addition, the S/N of the reconstruction could beincreased by averaging identical structures along thesame tubule. Averaging could be performed in bothdirect and Fourier space. For both strategies, it was con-venient to compute the reconstruction with the tubulesparallel to one edge of the array, e.g..y. Since the tilt axisof the goniometer, originally brought in coincidence toy, was not exactly parallel to the tubules, an ad hocreconstruction was computed according to this newrequirement. After that, the reconstruction was Fourier-filtered in one direction only, as was previously done for2D untilted images.

Finally, a second group of small maps wasobtained by averaging the segments of each tubule indirect space. These segments were aligned with eachother by using 3D correlation, to match best superposi-tion. After this, segments belonging to the same tubulewere averaged. The results were very similar to those ofthe Fourier filtering. This approach offered the opportu-nity to measure standard deviation of the repeating pitch:24 6 0.3 nm.

RESULTS

The Organization of the MonarthopalpusSperm Axoneme

The structure of the sperm axoneme of M. flavushas been previously described [Baccetti et al., 1974;Lupetti et al., 1998]. The unusual axoneme of thiscecidomid dipteran consists of 170–200 microtubulardoublets organized in laminae with a cartwheel-likearrangement (Fig. 1a); when viewed from the headtoward the tail tip, all dynein arms are oriented clock-wise. Doublet microtubular laminae present differentcurvatures and, as is clearly evident in Figure 1a, someof these rows of doublets exhibit a quasi-planar arrange-ment. Each doublet is endowed only with ODAs. Miss-ing are IDAs, any radial spokes, nexin links, or centralpair microtubules and their accompanying structures(Fig. 1b).

The ultrastructural organization of microtubularlaminae and of in situ ODAs can best be analyzed insamples prepared by the quick-freeze, deep-etching pro-cedure (QF/DE). In this study, all images were obtainedfrom sperm axonemes demembranated in the absence ofadded ATP and were therefore considered to be in therigor conformation. All the images shown are arrangedso that the distal tip of the axoneme is to the left.

Optimal preparations were those fractured nearlyparallel to the longitudinal axis of the demembranatedaxonemes. In these images (Figs. 1c and 2a), microtubu-lar laminae appear almost planar, so that a great numberof dynein arms can be visualized in the same field.Dynein arms are intercalated among parallel microtubulerows with the usual repeat of 24 nm; they consist of threemain domains: (1) a basal region (stem) contacting theA-tubule, which sometimes can be resolved into twosmaller units, (2) a prominent globular region, compris-ing two head domains which are located at different lev-els with respect to the microtubule, and (3) a distal stalkwhich originates from the more proximal head (pH) andcontacts the B-tubule.

The quasi planarity of these microtubular laminaeallows one to recover a significant number of equivalentdynein arms within the same fracture plane, thus increas-ing the S/N in models obtained by averaging protocols.

2D Analysis

The dynein arm repeat was first analyzed by proc-essing images recorded with no tilt, i.e. 2D. The aver-aged area in Figure 2b shows ten rows of ODAs reveal-ing the same overall pattern of organization. A certaindegree of variation is, however, displayed by some rows.This can be attributed to the presence of mechanical dis-tortions along microtubular laminae that may haveoccurred during sample processing before freezing, or tothe presence of debris deposited onto the fractured sur-face of the sample before metal replication. The effect ofsuch contamination is evident in rows 3 and 10 (countingrows from top to bottom) where it results in a severe lossof definition in the averaged image. Additionally, inrows 8 and 9 the dynein arm region contacting theA-microtubule surface is less clear than in other rows,probably because of the mechanical distortions duringfreeze-fracturing. Consequently, we selected rows4–6 for further tomographic analyses, since in these rowsthe averaged structures clearly showed the best defini-tion. Still it should also be noted that the dynein stalks inrow 6 are less evident than in the other two rows; thiscould be a consequence of a reduced metal depositiononto them, as could occur if these doublets were turnedslightly inward.

3D Reconstruction and Resolution

An area surrounding the region previously exam-ined was reconstructed in 3D using the collected tilt ser-ies. The reconstructed volume comprised about 10 dou-blets, each one for a length of 24–25 arms; the array hada size of 512 3 512 3 128 voxels with a voxel edgeabout 1 nm long. The amount of information conveyed

72 Lupetti et al.

in this reconstruction was quite large. Though the vol-ume comprising the replica should basically consist of abinary object, in the reconstruction we observed a wideand continuous range of density values. This phenom-enon, intrinsic to this tomographic approach and conse-quent to the finite number of projections, is furtherenhanced by the limited tilt angle and by the averagingprocesses. The continuous density variation can be easilyobserved by showing the signal on plane sections in ascan of the volume (Fig. 3a).

For a better visualization of the overall shape ofthe replica, the reconstructed volume was thresholded

and partitioned into a region corresponding to the metallayer and a surrounding zone with no structural features.The featureless region was masked out and the metalwas visualized excluding the rest (Fig. 3b). This strategyrequires the selection of the threshold used to partitionthe data between metal and nonmetal regions, whichwas done on the basis of personal experience and bycomparing the resulting images with the original micro-graphs. We also took advantage of the restoration POCSfilter (see Methods) which improved significantly thecontrast of the maps and facilitated the segmentation ofthe data.

Fig. 1. (a) Spermatozoa within the deferent duct,

cross-sectioned at different levels: proximal (p) and dis-

tal (d) sections of the axoneme. (b) Higher magnifica-

tion of a cross section of an axoneme; here, microtubu-

lar laminae are arranged in a quasi-planar fashion.

Dynein outer arms are visible on each doublet (arrow-

heads). (c) Replica of a demembranated rapidly frozen,

cryofractured sperm axoneme in the rigor state. This

axoneme was fractured nearly parallel to the longitudi-

nal axis of the sperm. In this condition microtubular

laminae appear almost planar, and a great number of

dynein arms can be visualized intercalated among paral-

lel microtubular rows.

3D Modeling of Outer Dynein Arm In Situ 73

The reconstructed model shows the metal replicaof the fractured axoneme. Measurements previouslymade on samples processed with the same metal-deposi-tion protocol (electron beam gun vacuum evaporation)but reconstructed by single particle strategy have shownthe thickness of the metal to be about 2–3 nm [Lanzavec-chia et al., 1998]. However, there is no way to know howdeeply the metal penetrates into the innermost regions ofthe sample, or exactly how it condenses or coalesces dur-ing metal deposition onto samples embedded in ice.Clearly, the thickness of the metal replica is not homoge-neous. A greater amount of metal is deposited on themicrotubules, and a lesser amount on the dynein arms.This is a consequence of the low-angle rotary shadowingused for metal evaporation. During this process, largestructures protruding the most from the plane of thefractured surface, e.g. the doublet microtubules, preventthe metal from reaching the innermost regions of thesample.

For a sample of this type, which has to be regardedas a single structure, resolution cannot be assessed byconventional criteria such as Fourier Shell Correlation ofindependent models [van Heel, 1987]. Therefore, resolu-tion can be estimated by looking at the finest details thatcan be perceived. For specimens displaying obviousstructural periodicities, resolution can also be estimatedby the clarity of this periodicity. In our case, both the 3Dreconstructions and the original micrographs show peri-

odic signals up to 6 nm (data not shown), while the 4 nmlayer lines were missing in both types of data. This can-not be ascribed to the tomographic procedure and isprobably due to the metal-coating process.

Therefore, we can estimate resolution by the metalthickness, taking into account, however, that the resolu-tion in the XY plane and along the Z-direction that can beobtained in an incomplete tilt series is not the same. Inspite of the thickness measured by instrumental monitor-ing during the evaporation process (2–3 nm), the thick-ness of the metal that can be inferred in the reconstruc-tion is about 3–4 nm in the XY plane and is broadened toabout 4–6 nm along Z.

Averaging of Repeats

After reconstruction of a metal replica, the 3D vol-ume can be rotated and visualized from different vantagepoints to observe both its upper and lower surfaces. Thelower surface would of course be in contact with theactual surface of the microtubules and hence would cor-respond to its mould. The situation with respect to thearms is different (see below) and can be better under-stood by observing averaged images. To do this noise fil-tration, we averaged the reconstructed volume alongeach individual doublet using the 24-nm-repeat corre-sponding to the structural periodicity of dynein armsalong microtubules. The results of averaging are shownin Figure 4; the doublets shown here correspond to those

Fig. 2. (a) Higher magnification of part of the field shown in Figure 1c. (b). One-dimensional Fourier

filtering of portion (a) using the 24-nm-repeat of the arm as basic periodicity. To facilitate the comparison

between this figure and Figure 4, rows of dynein arms are numbered here from top to bottom.

74 Lupetti et al.

numbered as 4–6 in Figure 2. This procedure leads to areinforcement of all those structural features that main-tain the same spatial orientation in different dynein arms,while it strongly reduces signals from those structuresdisplaying local variation along the microtubule.

The lower part of the replica (Fig. 4b) shows themetal in contact with the microtubules. As explained ear-

lier, the metal is able to cover only the upper surface ofthe cylinders; therefore, the mould is concave to accom-modate the tubule inside, but is truncated where thebending of the cylinder prevents the metal from coveringthe surface. On the other hand, it appears that the metalis able to completely surround the dynein arms; thiswould be possible because these structures are much

Fig. 3. 3D reconstruction of the portion of the axo-

neme presented in Figure 2a. The reconstruction is a 3D

array of density values. (a) Three orthogonal sections

are cut in the 3D array and displayed; white regions cor-

respond to the metal mould. The horizontal section is

cut at mid level and shows the dynein arms and the

tubules. Vertical sections show the vanishing of den-

sities in the upper and lower region of the reconstructed

volume. (b) The voxels corresponding to the metal have

been segmented from the volume and displayed with a

volume-rendering technique.

3D Modeling of Outer Dynein Arm In Situ 75

Fig. 4. Average of the recon-

structed volume along each individ-

ual doublet using the 24-nm-repeat

corresponding to the structural

periodicity of dynein arms along

microtubules. (a) Front view and

(b) back view of the replica. The

dynein arms shown in this figure

correspond to those numbered 4, 5,

and 6 in Figure 2. See the text for

further comments on this figure.

smaller than the microtubules, are isolated, and do notlie on any substrate. As a consequence, no concavity isobserved on bottom views of ODAs. In these views, aconsistent prolongment is observed projecting downwardfrom the distal dynein head (Fig. 4b).

At the resolution we were able to achieve, it wasnot possible to detect any empty volume inside the metalthat might correspond to the real organic structure, sincethe metal thickness was not resolved on the small arms.However, taking into account the molecular mass of thearm and its relative volume, we imagine it would betotally included within the 3D model we are presenting.

The Structure of the Arm

A few arms from doublet number 4 in Figures 2and 4 are shown in Figure 5 in different views. This rowhas been selected because it looks better resolved withrespect to the others, by taking into account both visualimpression and spectral analysis. Images a, b, and c cor-respond respectively to –158, 08, and þ158 tilting of the3D model, as viewed from outside of the axoneme.Images d, e, and f represent instead the –158, 08, andþ158 inside views of the model. To facilitate thedescription of structural details of in situ ODAs, we havelabeled in Figure 6 all the different molecular domainswe describe below.

ODAs are well separated from each other and donot show any obvious overlapping with neighboring

structures (Fig. 4). Each dynein arm displays a deep fur-row that is clearly visible in both outside and insideviews; it is oriented at about 1258 with respect to thelongitudinal axis of the A-tubule, splitting the majorglobular region into two clearly discernible heads, whichwe term proximal head (pH) and distal head (dH) inFigure 6. As a consequence, the two heads are not posi-tioned at the same distance from the surface of the A-tubule, the more distal head lying closer to it (Fig. 5).The differential positioning of the heads is better visual-ized in tilted views of the 3D model (Figs. 5a, 5c, 5d,and 5f). In both front and posterior views, each headappears to have two relatively distinct subdomains thatare positioned on the same plane (Figs. 5 and 6). Thisplane is almost orthogonal to the longitudinal microtubu-lar axis in the Z-direction but clearly oblique in the X–Ydirection (almost 458). A broad central contact regionbetween these two bipartite heads is present in the fur-row between them, independent of the threshold selectedfor the 3D rendering of the model.

A lateral conical region from which the stalk origi-nates is visible on one side of the pH (asterisk in Fig. 6).The stalk (Sk in Fig. 6) contacts the B-tubule with anangle of about 1208 with respect to the longitudinal axisof this tubule. A globular tip is visible at the distal end ofthe stalk.

The stem is formed by several different domains(Fig. 6). An ovoid region (St1) adheres to the surface of

Fig. 5. A few arms from doublet number 4 in Figures 2 and 4 are shown under different viewpoints.

Images a–c are views of the 3D model tilted respectively –158, 08, and þ158, as seen from the outside of

the axoneme. Images d–f represent instead the –158, 08, and þ158 inside views of the model.

3D Modeling of Outer Dynein Arm In Situ 77

the A-tubule and contacts another globular domain (St2)that is located underneath the pH and does not directlycontact the A-tubule. An elongated domain (St3) lies inthe space between the dH and St1. Finally, both St1 andSt3 appear to contact a globular structure (St4) that islocated on the surface of the A-tubule in the inter-armregion (Fig. 6a). Therefore, the whole dynein arm con-tacts the A tubule surface at two distinct attachmentpoints (arrows in fig. 6a). The back view of this model isshown in Figure 6b. In this view, the two heads, thestalk, the St2, and St4 domains are easily recognizable.In addition, a further globular domain, St5, is visiblebeneath and in continuity with the dH.

DISCUSSION

Current data available on the structure of isolatedaxonemal dynein molecules have provided valuableinformation on the molecular organization of this proteincomplex. Purified ODAs are dimeric or trimeric struc-tures (depending on the species) that possess a bouquet-like structure, with two or three globular heads con-nected to a common base by slender stems [Johnson andWall, 1983; Goodenough and Heuser, 1984]. A thin stalkprojects out from each head. Negative staining of iso-lated dynein isoforms, followed by single particle imageprocessing, has shown that the dynein head is a ring-likearrangement of seven similarly sized subdomains organ-ized around a central region of dense staining, either achannel or a cavity [Samso et al., 1998; Asai andKoonce, 2001]. This ring-like structure is shared by allmembers of the so-called AAA (ATPases Associatedwith diverse cellular Activities) protein superfamily[Hanson et al., 1997; Ogura and Wilkinson, 2001].Sequence data have indicated that dynein heavy chainsindeed belong to this group of mechanoenzymes [King,2000]. In the dynein molecules studied so far [Burgesset al., 2003, 2004], the two sides of the ring are asym-metric, with the stem and the stalk emerging from it atdifferent sites and with a different reciprocal orientationdepending on the functional state of the molecule. Thestem is organized into different domains, i.e., a linkerproximal to the head, an intermediate elongated neck,and a thicker globular base. The linker has been pro-posed to lie across the head forming a sharp bend withthe adjacent neck of the stem. A further sharp bend con-

Fig. 6. High-magnification 3D model of a single ODA in front

(a) and back (b) view. pH, proximal head; dH, distal head; Sk, stalk;

St1, St2, St3, St4, and St5 indicate the different stem domains. Arrow-

heads indicate the contact region between the two heads; arrows indi-

cate instead the contact regions between ODA and the A-tubule.

78 Lupetti et al.

nects the neck of the stem with the base of the molecule.The latter appears to be composed of two distinctregions. The images obtained so far demonstrate a highdegree of both planar and torsional flexibility for thestem region of dynein arms.

Structural analyses performed on in situ dyneinarms demonstrated that the bouquet structure describedin isolated molecules condenses as it binds to the A-tubule, giving rise to a compact array with a major glob-ular head from which a single slender stalk extends tocontact the B-tubule of the adjacent doublet [Goode-nough and Heuser, 1984; Sale et al., 1985; Burgess et al,1991; Burgess, 1995].

Data produced so far, however, have not com-pletely clarified how the structural domains of isolateddynein molecules are arranged within in situ arms. Issuesrequiring further investigation include the following: (1)the arrangement of the different head domains within thein situ dynein arm, (2) the organization of the arm baseregion contacting the A-tubule, (3) the degree of overlapbetween adjacent arms, and (4) the actual disposition ofthe two or three stalks that can be seen so clearly in theisolated arms but not in situ.

In an effort to determine how the structural data onisolated arms relate to the arms in situ, we describe herethe first detailed analysis of the structure of in situ ODAs,based on tomographic reconstruction of micrographsobtained by preparing axonemes by QF/DE proceduresfor electron microscopy. For this analysis, new imageanalysis protocols had to be developed. The 3D model wepresent complements and extends previously publisheddata, providing new fine structural details that were notvisualized in previous studies performed by 2D averagingof images from metal replicas from other axonemal mod-els [Burgess et al., 1991]. Our results were obtained froman unusual type of flagellar axoneme in which only theouter arms are present along doublet microtubules,thereby circumventing the usual interference from otheraxonemal structures such as IDAs and radial spokes.

Owing to the compact structure that dynein armsdisplay in situ, it is difficult to define precisely the local-ization and spatial arrangement of some parts of the mol-ecule as described by the negative staining procedures.Although the head and the stalk seen in negative stainingare clearly identifiable by QF/DE, the stem and themolecular domain interacting with the A-tubule are moredifficult to discern by QF/DE.

The subdomains we were able to visualize in thedynein heads by QF/DE are strongly reminiscent of theAAA structure previously evidenced in negative stainsof isolated dynein complexes. The resolution weobtained so far, however, does not permit us to establisha precise correspondence between the subdomainsobserved in situ by QF/DE and those described in puri-

fied molecules by negative staining. To make the inter-pretation of our results easier, Figure 7 compares thedynein AAA ring as seen in negatively stained isolateddynein molecules (panels a and c) [Burgess et al., 2004]with side views of our 3D model (panels b and d).

In situ the two AAA rings appear planar, parallel toeach other and obliquely oriented with respect to the lon-gitudinal axis of the microtubules. The central hole visi-ble in isolated dynein complexes in rigor condition[Burgess et al., 2004] is not visible in either the front orback views of our model. However, a concavity is visiblein lateral views of the distal side of the dH. This prob-ably corresponds to the central hole seen in negativestain. Our inability to visualize this feature is probablydue to both the oblique orientation of the ring and themetallization process. No concavity is visible on the prox-imal side of the pH either, probably due to the presence ofthe linker domain positioned adjacent to the head.

Previous studies performed by QF/DE on demem-branated axonemes revealed the occurrence of a singlestalk [Goodenough and Heuser, 1984; Sale et al., 1985;

Fig. 7. Comparison between lateral views of our in situ 3D model

and the side views of isolated molecules reported by Burgess et al.

[2004]. (a) View of the ‘‘left’’ side of purified molecules after negative

stain. (b) Lateral view of the proximal head as observed from the

proximal end of the axoneme. (c) View of the ‘‘right’’ side of purified

molecules after negative stain. (d) Lateral view of the distal head as

seen from the distal tip of the axoneme. False colors were used to evi-

dence the AAA ring in the proximal (blue) and distal (red) head.

Panels 7a and 7c are mirrored views of Figure 1D in the work of

Burgess et al. [2004].

3D Modeling of Outer Dynein Arm In Situ 79

Burgess et al., 1991; Gee et al., 1997; Lupetti et al.,1998; Mencarelli et al., 2001]. Analyses of isolateddynein complexes revealed, on the other hand, the pres-ence of a stalk emerging from each one of the dyneinheads. When studying in situ ODAs in Monarthropalpusaxonemes, we were able to observe two distinct stalksemerging from the same dynein arm only in very fewinstances [Lupetti et al., 1998]. Consequently, all tracesof it were lost during the averaging protocol used to pro-duce the 3D reconstructions shown here, and thereforeevidences only one stalk. It remains an intriguing prob-lem to determine the detailed composite construction ofthe stalks that connect dynein arms to the B-tubule. Inany case, in isolated dynein complexes the stalks appearas slender lateral projections emerging from the AAArings, while in our model the base of the stalk appears asa conical region adjoined to the proximal face of the pH.It can be reasonably hypothesized that this conical regioncorresponds to the so-called linker domain of the stem,which has been described as associated with the head[Burgess et al., 2004]. It is known that the associationbetween the head and the linker is more extended inrigor molecules and that, in this configuration, the pointsof emergence for the stem and the stalk are closer to eachother [Burgess et al., 2003, 2004]. Therefore, it is likelythat in rigor the region of the stalk in proximity of thehead is masked by the linker domain. Alternatively, anunnatural pooling of metal around the point where thestalk connects with the head might be responsible for theconical appearance of this region. Our in situ observa-tions also indicate that the stalk of the dynein armemerges almost perpendicular from the plane of the ring-like head. In contrast, negative staining of purifieddyneins in rigor suggests that the stalk emerges from thesame plane as the ring [Burgess et al., 2004], undoubt-edly because of the flattening that molecules inevitablyundergo during negative staining. The stalk region endswith a globular domain contacting the B-tubule. Thisstructural feature was demonstrated in high resolutionstudies performed on isolated molecules [Goodenoughand Heuser, 1984; Gee et al., 1997; Burgess et al., 2003,2004] and is visualized in situ for the first time in thisstudy.

In this study, the two head domains show a point ofcontact with each other in an area localized within thecentral furrow. Assuming that the two heads possess asimilar overall orientation, with the linker region of thestem located on the proximal face of the AAA ring, thenthe linker domain of the dH could contribute to the con-tact between the two heads. This structural feature couldbe functionally relevant. It has been reported that theoverall activity of the outer arm depends upon specificinteractions between the different dynein heavy chains,thus suggesting that, in vivo, the functional state of each

dynein heavy chain subunit may be influenced by con-formational changes occurring in the adjacent subunit inthe same arm complex [Nakamura et al., 1997]. Thus,the linker domain could mediate the transmission of con-formational information from the proximal to the distalhead. We also note that in situ the distal and the proxi-mal face of each head are not equivalent, and the occur-rence of a structural dissimilarity between the two facesof the head has already been suggested by the analysis ofisolated dynein molecules [Burgess et al., 2004].

Regarding other parts of the stem, no slender fila-mentous domain comparable to the neck region previ-ously described in negative stains of isolated dynein cmolecules [Burgess et al., 2003, 2004] is evident; thisregion is therefore likely to assume a different conforma-tion in situ or is likely to be hidden among the othermore bulky domains. The stem rather appears as a seriesof globular subdomains. Two of these, indicated as St2and St5 here, are more internally located and form sharpbends with the adjacent head and with the more basalSt1 domain; they can be reasonably related to the pH anddH, respectively. St5 is clearly visualized only in backviews of the arm, in which it appears to extend into adrop-like structure. The organization of this domain, atthis stage, has to be considered with caution. In frontview St5 is masked by the presence of an elongated,transversal domain (St3) that appears in the hollowformed by the dH and St1.

The attachment of the arm to the surface of theA-tubule appears to be mediated primarily by the twodomains St1 and St4. It has not been possible to establishif these regions are actual components of the dynein armor instead if they are contributed by the two other molec-ular complexes that are known to be necessary for thecorrect attachment of ODAs onto the A-tubule surface,i.e. the docking complex [Takada et al., 2002] and theOda5p/Oda8p/Oda10p complex [Wirschell et al., 2004].The available molecular evidence on the presence of twodistinct anchorage complexes for ODAs is in accordancewith our structural observations. This finding is particu-larly interesting for its functional implications in theODA mechanochemical cycle, as previously hypothe-sized by Lindemann and Hunt [2003]. In fact, in a theo-retical analysis of the power stroke process, these authorsnoted that, given the flexibility of the stem, the mecha-nism of force generation would be possible only if theglobular head is stabilized by a second point of attach-ment on the A-tubule. Our results provide the first evi-dence for this structural feature; further analyses per-formed at higher resolution are necessary to confirm ourobservations.

The current structural model for the dynein powerstroke proposes that a rotation of the AAA ring coulddisplace the stalk laterally and apply force to the adja-

80 Lupetti et al.

cent doublet, thus inducing a longitudinal sliding [Geeet al., 1997; Burgess et al., 2003, 2004]. This model isbased on comparative electron micrographs of negativelystained in rigor and relaxed isolated dyneins. Becausethese images were obtained from molecules that wereforced to lie flat on a substrate during preparation, theproposed model assumes that the net movement of thestalk occurs on the same plane of the AAA ring. How-ever, the results in this report show that the AAA ring ispositioned on a plane oblique to the longitudinal axis ofmicrotubules and, in addition, that the stalk is obliquelyoriented with respect to the AAA ring. This implies that,if the power stroke really depends principally on therotation of the AAA ring, the force generated would onlyact along the longitudinal axis of microtubules if a sub-stantial rearrangement occurred in the stalk-head orienta-tion. In particular, the rotation of the head would imposea significant torsion to the stalk, thus generating an elas-tic energy that could be released during recovery afterpower stroke. In this regard it is also interesting to notethat some dynein isoforms have been reported to possessthe ability to both translocate and rotate microtubules invitro [Vale and Toyoshima, 1988; Kagami and Kamiya,1992].

The elucidation of the mechanism of dynein powerstroke will ultimately require the determination of thestructure of the arm complex in situ both in active andrigor states with a proper comparison of the pre- andpostpower stroke conformations. Accomplishing this viathe higher resolution methods introduced here shouldpermit a proper definition of how the various subdomainsof the arm change during arm movement.

Comments on the Tomographic Analysis

Electron tomographic procedures have been usedfor the first time in this study to analyze metal replicas ofnative biological specimens to produce high-resolution3D models. Therefore, some methodological aspects alsodeserve specific comments.

Our model was obtained by collecting a series ofimages of the replica while tilting the stage from �458 toþ458 (single axis tomography). A generic limitation ofthis geometry consists of the missing wedge in Fourierspace, since we lacked a significant amount of informa-tion along the reciprocal Z-axis. As a consequence, thereconstructions are affected by an anisotropic resolutionthat is higher in the XY plane and lower along Z [see,e.g., Turner and Valdre, 1992]. The direction of the tilt-ing axis also affects the anisotropy with regard to obliquefrequency components of the signal. When the tiltingaxis is oriented far away from the direction of thetubules, the latter appear different in the reconstruction.The most efficient way to improve the quality of thereconstruction would be to increase the tilting angle of

the specimen up to about 708. This, however, requiresthe use of a sophisticated goniometric stage fitted on adedicated electron microscope with high electrons accel-erating voltage, because, at high tilt-angles, images maybe affected by radiation damage consequent to theincreased path of electrons across the sample. A furtherimprovement to partly overcome the limitation conse-quent to the missing wedge consists in the collection oftwo series of images, with orthogonal tilt axes (dual axistomography). In this case the missing wedge becomes amissing pyramid with a large reduction of its volume. Itshould be noticed that the total number of images couldbe the same, because the tilting step of each series couldbe increased, without a significant loss of information[Penczek et al., 1995].

A typical problem of electron tomography consistsof sample deterioration during the quite long timeneeded to collect images. Our replicas were imaged 61times in a manually controlled microscope, and the col-lection of one series required almost 2 h of exposure tothe electron beam. Surprisingly, appreciable differenceswere noticed neither in direct space nor in the spectrum,when comparing the untilted view before and after thisdata collection. This indicates that replicas can be con-sidered a good sample for tomographic experiments, asalso suggested by Lanzavecchia et al. [2005]. However,the use of metal replicas for electron tomographic recon-struction has been criticized because of the resolutionlimit intrinsic to the protocol of metal replication. Reduc-ing the metal thickness would probably improve the reso-lution, although this might still increase radiation damageduring the collection of images. Hopefully, a better under-standing of the metal deposition process itself wouldallow us to define a deposition protocol that will be moresuitable for producing replicas optimized for such image-analyses protocols. Such work is in progress.

ACKNOWLEDGMENTS

Special thanks to Dr. Joel Rosenbaum (Departmentof Molecular, Cellular, and Developmental Biology, YaleUniversity) for his many stimulating discussions and sug-gestions and for his critical reading of the manuscript.

REFERENCES

Asai DJ, Koonce MP. 2001. The dynein heavy chain: structure,

mechanics and evolution. Trends Cell Biol 11:196–202.

Baccetti B, Dallai R, Giusti F, Bernini F. 1974. The spermatozoon of

arthropoda XXV. A new model of tail having up to 170 dou-

blets:Monarthropalpus buxi. Tissue Cell 6:269–278.Baumeister W. 2005. From proteomic inventory to architecture. FEBS

Lett 579:933–937.

Baumeister W, Lembcke G. 1992. Structural features of archaebacte-

rial cell envelopes. J Bioenerg Biomembr 24:567–575.

3D Modeling of Outer Dynein Arm In Situ 81

Baumeister W, Steven CA. 2000. Macromolecular electron micro-

scopy in the area of structural genomics. Trends Biochem Sci

25:624–631.

Baumeister W, Grimm R, Walz J. 1999. Electron tomography of mol-

ecules and cells. Trends Cell Biol 9:81–85.

Burgess SA. 1995. Rigor and relaxed outer dynein arms in replicas of

cryofixed motile flagella. J Mol Biol 250:52–63.

Burgess SA, Dover SD, Woolley DM. 1991. Architecture of the outer

arm dynein ATPase in an avian sperm flagellum, with further

evidence for the B-link. J Cell Sci 98:17–26.

Burgess SA, Walker ML, Sakakibara H, Knight PJ, Oiwa K. 2003.

Dynein structure and power stroke. Nature 421:715–718.

Burgess SA, Walker ML, Sakakibara H, Oiwa K, Knight PJ. 2004.

The structure of dynein-c by negative stain electron micro-

scopy. J Struct Biol 146:205–216.

Carazo JM. 1992. The fidelity of 3D reconstructions from incomplete

data and the use of restoration methods. In: Frank J, editor.

Electron Tomography. NY: Plenum. p 117–164.

Carazo JM, Carrascosa JL. 1987. Information recovery in missing

angular cases: an approach by convex projections method in

three dimensions. J Microsc 145:23–43.

Frank J. 1996. Three-Dimensional Electron Microscopy of Macromo-

lecular Assemblies. NY: Academic Press. 342 p.

Frank J, McEwen BF. 1992. Alignment by cross-correlation. In: Frank J,

editor. Electron Tomography. NY: Plenum. p 205–213.

Frank J, Radermacher M. 1986. Three-dimensional reconstruction of

nonperiodic macromolecular assemblies from electron micro-

graphs. In: Koehler J, editor. Advanced Techniques in Biologi-

cal Electron Microscopy. Berlin: Springer-Verlag. p 1–72.

Gee MA, Heuser JE, Vallee RB. 1997. An extended microtubule-bind-

ing structure within the dynein motor domain. Nature 390:636–

639.

Goodenough UW, Heuser JE. 1982. Structure of the outer dynein arm.

J Cell Biol 95:798–815.

Goodenough UW, Heuser JE. 1984. Structural comparison of purified

dynein proteins with in situ dynein arms. J Mol Biol 180:1083–

1118.

Hanson PI, Roth R, Morisaki H, Jahn R, Heuser JE. 1997. Structure

and conformational changes in NSF and its membrane receptor

complexes visualized by quick-freeze/deep-etch electron micro-

scopy. Cell 90:523–535.

He W, Cowin P, Stokes D. 2003. Untangling desmosomal knots with

electron tomography. Science 302:109–113.

Heuser J. 1981. Preparing biological samples for stereomicroscopy by

the quick-freeze, deep-etch, rotary-replication technique. Meth

Cell Biol 22:97–122.

Jing ZQ, Sachs F. 1991. Alignment of tomographic projections using

an incomplete set of fiducial markers. Ultramicroscopy 35:37–43.

Johnson KA, Wall JS. 1983. Structure and molecular weight of the

dynein ATPase. J Cell Biol 96:669–678.

Kagami O, Kamiya R. 1992. Translocation and rotation of microtu-

bules caused by multiple species of Chlamydomonas inner-armdynein. J Cell Sci 103:653–664.

King SM. 2000. AAA domains and organization of the dynein motor

unit. J Cell Sci 113:2521–2526.

Lanzavecchia S, Lupetti P, Bellon PL, Dallai R, Rappuoli R, Telford JL.

1998. Three-dimensional reconstruction of metal replicas of

the Helicobacter pylori vacuolating cytotoxin. J Struct Biol

121:9–18.

Lanzavecchia S, Cantele F, Bellon PL, Zampighi L, Kreman M,

Wright E, Zampighi GA. 2005. Conical tomography of freeze-

fracture replicas: a method for the study of integral membrane

proteins inserted in phospholipid bilayers. J Struct Biol 149:

87–98.

Lawrence MC. 1992. Least-squares methods of alignment using

markers. In: Frank J, editor. Electron Tomography. NY: Plenum.

p 197–204.

Lawrence MC, Jaffer MA, Sewell BT. 1989. The application of the

maximum entropy method to electron microscopic tomogra-

phy. Ultramicroscopy 31:285–302.

Lindemann CB, Hunt AJ. 2003. Does axonemal dynein push, pull, or

oscillate? Cell Motil Cytoskeleton 56:237–244.

Lupetti P, Mencarelli C, Rosetto M, Heuser JE, Dallai R. 1998. Struc-

tural and molecular characterization of dynein in a gall-midge

insect having motile sperm with only the outer arm. Cell Motil

Cytoskeleton 39:303–317.

McEwen BF, Marko M, Hsieh CE, Mannella C. 2002. Use of frozen-

hydrated axonemes to assess imaging parameters and resolu-

tion limits in cryoelectron tomography. J Struct Biol 138:47–

57.

Medalia O, Weber I, Frangakis AS, Nicastro D, Gerish G, Baumeister

W. 2002. Macromolecular architecture in eukaryotic cells

visualized by cryoelectron tomography. Science 298:1209–

1213.

Mencarelli C, Lupetti P, Rosetto M, Mercati D, Heuser JE, Dallai R.

2001. Molecular structure of dynein and motility of a giant

sperm axoneme provided with only the outer dynein arm. Cell

Motil Cytoskeleton 50:129–146.

Mitchell DR. 1994. Cell and molecular biology of flagellar dyneins.

Int Rev Cytol 155:141–180.

Morris EP, Katayama E, Squire JM. 1994. Evaluation of high-resolu-

tion shadowing applied to freeze-fractured, deep-etched par-

ticles: 3D helical reconstruction of shadowed actin filaments.

J Struct Biol 113:47–55.

Nakamura K, Wilkerson CG, Witman GB. 1997. Functional interac-

tion between Chlamydomonas outer arm dynein subunits: the gsubunit suppresses the ATPase activity of the ab dimer. Cell

Motil Cytoskeleton 37:338–345.

Ogura T, Wilkinson AJ. 2001. AAAþ superfamily ATPases: common

structure-diverse function. Genes Cells 6:575–597.

O’Toole ET, Giddings TH, McIntosh JR, Dutcher SK. 2003. Three-

dimensional organization of basal bodies from wild-type and d-tubulin deletion strains of Chlamydomonas reinhardtii. Mol

Biol Cell 14:2999–3012.

Penczek P, Marko M, Buttle K, Frank J. 1995. Double-tilt electron

tomography. Ultramicroscopy 60:393–410.

Porter ME. 1996. Axonemal dyneins: assembly, organization, and reg-

ulation. Curr Opin Cell Biol 8:10–17.

Radermacher M, Wagenknecht T, Verschoor A, Frank J. 1987. Three-

dimensional reconstruction from a single-exposure, random

conical tilt series applied to the 50S ribosomal subunit of

Escherichia coli. J Microsc 146:113–136.

Sale WS, Goodenough UW, Heuser JE. 1985. The substructure of iso-

lated and in situ outer dynein arms of sea urchin sperm flagella.

J Cell Biol 101:1400–1412.

Samso ,M., Radermaker M, Frank J, Koonce MP. 1998. Structural

characterization of a dynein motor domain. J Mol Biol 276:

927–937.

Steven AC, Aebi U. 2003. The next ice age: cryo-electron tomography

of intact cells. Trends Cell Biol 13:107–110.

Subramaniam S, Milne JLS. 2004. Three-dimensional electron micro-

scopy at molecular resolution. Annu Rev Biophys Biomol

Struct 33:141–155.

Takada S, Wilkerson CG, Wakabaiashi K, Kamiya R, Witman GB.

2002. The outer dynein arm-docking complex: composition

and characterization of a subunit (Oda1) necessary for outer

arm assembly. Mol Biol Cell 13:1115–1129.

82 Lupetti et al.

Turner JN, Valdre U. 1992. Tilting stage for biological applica-

tions. In: Frank J, editor. Electron Tomography. NY: Plenum.

p 167–196.

Vale RD, Toyoshima YY. 1988. Rotation and translocation of micro-

tubules in vitro induced by dyneins from Tetrahymena cilia.

Cell 52:459–469.

van Heel M. 1987. Similarity measures between images. Ultramicro-

scopy 21:95–100.

Walz T, Tittmann P, Fuchs KH, Muller DJ, Smith BL, Agre P, Gross H,

Engel A. 1996. Surface topographies at subnanometer-resolution

reveal asymmetry and sidedness of aquaporin-1. J Mol Biol

264:907–918.

Wirschell M, Pazour G, Yoda A, Hirono M, Kamiya R, Witman GB.

2004. Oda5p, a novel axonemal protein required for assembly

of the outer dynein arm and an associated adenylate kinase.

Mol Biol Cell 15:2729–2741.

Zampighi GA, Kreman M, Lanzavecchia S, Turk E, Eskandari S,

Zampighi L, Wright EM. 2003. Structure of functional single

AQP0 tetramers inserted in phospholipid membranes. J Mol

Biol 325:201–210.

3D Modeling of Outer Dynein Arm In Situ 83