Previous Article |
Table of Contents
| Next Article 
BIOPHYSICS
Three-dimensional reconstruction of the dynactin complex by single-particle image analysis
, ¶
*Department of Biomedical Sciences, Imperial College London, London SW3 6LY, United Kingdom; Institute of Cell Biology and Biosystems Technology, University of Rostock, D-18059 Rostock, Germany; and Randall Centre, King's College London, London SE1 1UL, United Kingdom
Edited by Thomas D. Pollard, Yale University, New Haven, CT and approved January 20, 2005 (received for review December 20, 2004)
 |
Abstract |
|---|
 Top Abstract Materials and MethodsResultsDiscussionReferences |
|---|
CapZ | dynein | image reconstruction | molecular motors
Dynactin is an 
1.1-MDa complex of at least nine distinct proteins, most of which are unique to the dynactin complex. These include p150Glued (8) and a filament forming actin-related protein (Arp1) (9). The structural arrangement of the complex is not fully understood, in large part because of its complexity. Most information has come from electron microscopic studies of rotary-shadowed dynactin complexes (10) and from a combination of rotary shadowing and complex fragmentation and antibody labeling studies (11). The dynactin complex appears as a rod-like structure of 37 nm (10) of which the Arp1 forms the major component; i.e., 9–11 subunits of Arp1 are thought to form the backbone on which the other proteins are arranged. Among the family of actin-related proteins, Arp1 is the most similar in terms of sequence identity to actin (54% to cytoplasmic actin), and it is the only one known to form filaments (12). The interaction of Arp1 with spectrin, and thereby the cargo, is important for cargo transport (3). Arp1 also interacts with a whole host of other proteins found within the dynactin complex. One of the other major components of the complex with which it interacts is the actin-capping protein, a heterodimer of 
- and 
-subunits (37 and 32 kDa). This capping protein was first discovered in muscle Z-lines as a filamentous (F)-actin barbed-end capping protein (13), and the muscle isoform of this protein has been named CapZ. The capping protein (CaP) is also found in non-muscle cells (14) and is thought to occupy the same position on the Arp1 backbone of the dynactin complex, probably as a backbone stabilizer. A major subcomplex within dynactin containing three distinct proteins takes the form of a projecting shoulder/sidearm structure (11). The long, flexible sidearm, thought to consist mainly of p150Glued, is capable of interacting with dynein intermediate chain (15, 16) and microtubules (8). The shoulder, which forms the attachment point of p150Glued to the Arp1 backbone, consists of at least four dynamitin (p50) molecules and two copies of p24. At the pointed end of the Arp1 backbone lies the so-called pointed-end complex (PEC), which contains at least four proteins, including a single subunit of a unique actin-related protein, Arp11 (11). Three other PEC components, p62, p27, and p25, contain predicted cargo binding motifs. P62 has been shown to interact strongly with Arp11 (11) and probably also with Arp1 (17).
To understand the function of dynactin, it is essential to determine the structure of the complex. Until now, the detailed 3D structure of dynactin has not been visualized. We have used single-particle image analysis methods to investigate the 3D structure of dynactin. Dynactin cannot be treated as a helical structure, because of its capping proteins; therefore, conventional helical reconstruction methods cannot be used. In the present study, we have applied single-particle methods using the IMAGIC-5 package (Image Science, Berlin) to produce the 3D reconstruction of the dynactin complex. We report here structural details of the dynactin backbone and make some preliminary observations on how the capping protein is likely to interact with the barbed-end of the Arp1 filament backbone.
|
Materials and Methods |
|---|
|
TopAbstractMaterials and MethodsResultsDiscussionReferences |
|---|
Electron Microscopy. Negative staining was carried out as described in ref. 20, using 1% uranyl acetate and a protein concentration of 25 µg/ml in 30 mM KCl/5 mM Pipes (pH 7.2). Grids were examined in a Philips CM100 transmission electron microscope, and micrographs were taken under low-dose conditions (Kodak SO163 film) at a magnification of 50,850 with an underfocus value of 0.6–1.3 µm. Tilted data sets were collected separately at a stage-tilt value of +40°.
Image Processing. Electron micrographs of dynactin complexes were scanned in a Leafscan-45 (Leaf Systems, New York) at a step size of 10 µm and subsequently coarsened to 20 µm for processing. Particle picking and image analysis by single-particle methods was carried out by using the IMAGIC-5 suite of programs (21, 22); 7,168 complexes (particles) were used in this study. Picking of particles was carried out manually, and selected images were carefully screened. Those images containing particles that were touching each other were discarded. Image densities were normalized and bandpass-filtered to remove low frequencies to a value commensurate with the diameter of the particle.
Alignment of particles was carried out first to a single, manually rotated to vertical and centered particle (which had been loosely masked). Subsequent alignment was performed by multireference alignment to representative masked, aligned, centered class averages. To improve the signal-to-noise ratio in individual images, aligned particles were subjected to multivariate statistical analysis and classified into 350 classes containing like views. Class sums (averages) were produced for each class. Class-selection, centering, and multireference alignment was repeated until no discernible improvement in alignment was seen.
Euler-Angle Assignment and 3D Reconstruction. The angular reconstitution method was used to find the relative orientation between class averages. For initial assignment of Euler angles, an anchor-set made up of 2D projections from a 3-nm-resolution, 10-subunit, Arp1 model (see below) projected out at 12° intervals over the Euler sphere was used. For all subsequent steps, an anchor-set created from the data 3D was used (see Results and Discussion for more details). Three-dimensional reconstructions were masked by using a binary mask, care being taken not to cut off any data. Classes that failed to obtain Euler-angle assignments (as judged from the similarity of 2D reprojections of the 3D reconstruction along Euler angles assigned to the input class) were discarded before further rounds of Euler assignment.
The best classes, as judged by the above criteria, were used to calculate a 3D reconstruction by the back projection method. IMAGIC uses an "exact filter" subroutine to weight the class-averages before back-projection. The filter assumes that the reconstructed mass occupies most of the cubic volume. It has been shown that overrepresentation of low frequencies within the 3D volume for objects of unequal dimension, such as filaments and rods, can cause degradation in final 3D reconstructions (23). To allow the correct weighting of the low-frequency components, a modification of the IMAGIC "exact filter," designed by these authors, was used. Once the best 3D reconstruction was obtained from a given round of alignment, this was reprojected in 2D to use as a reference-set for another round of alignment before proceeding again to Euler-angle assignment and the production of another 3D map. This process of refinement (cycling through multi reference alignment, Euler-angle assignment, and 3D reconstruction) was then repeated until no further improvement of the output 3D reconstruction could be observed.
Modeling the Arp1 Filament Backbone. General sequence comparison of the Arp1 revealed it to be very similar to cytoskeletal actin (9). The structure of the Arp1 backbone has not been determined in detail, although preliminary observations of Bingham and Schroer (12) indicated that the Arp1 filament is helical and that the helicity is similar to that of F-actin. We generated a tertiary structure for the Arp1 subunit (Swiss-Prot protein sequence accession no. P42024 [SwissProt] ) by using the program SWISS-MODEL (24, 25). Model Arp1 filaments of 9, 10, and 11 subunits were built by using the globular (G)-actin orientation and helical parameters of the Holmes model actin filament (26) with an axial rise of 2.75 nm and an angular rise of 167.5°. Models were built by using the program QHLX of SITUS 1.4 (http://situs.biomachina.org). Three-dimensional maps were produced from the models and were bandpass-filtered to a lower resolution of 3 nm.
Fitting of crystallographic structures of the muscle isoform of the capping protein (CapZ) (27) (PDB ID code 1IZN [PDB] ) and Arp1 filament model to data were carried out by using O (28). Three-dimensional models were rendered by using PYMOL 0.95 (http://pymol.sourceforge.net).
|
Results |
|---|
|
TopAbstractMaterials and MethodsResultsDiscussionReferences |
|---|
|
Alignment and Assessment of 2D Structure from Class Averages. It is known that Arp1 has a high sequence identity (54%) with actin (9) and predicted structural similarity (Fig. 4 A and B). The structure we observe in our class averages is consistent with observations that Arp1 is capable of forming filaments like actin (12). Visual assessment of backbone details for each class average strongly suggested polar structure. Subunit shape and inner and outer domains could be distinguished in many class averages (Fig. 2B, arrows). Based on structural similarity of the Arp1 subunit to G-actin and of the backbone seen in class averages to F-actin, we adopt actin terminology (defined in Fig. 4 A and B) to describe the structure of the Arp1 subunit and the backbone.
|
|
Euler-Angle Assignment and 3D Reconstruction. Our initial 3D reconstruction of the dynactin complex indicated that the views were limited to those around the long axis. This is due to the fact that the particle is a rod and therefore lies, naturally, on its long axis on the carbon support film. We also found that the backbone had a preferential orientation around its long axis, possibly due to the large shoulder/sidearm component projecting from its surface. A reconstruction by angular reconstitution methods without a starting anchor-set was not possible from this data set. The common-line theorem used in this method generally relies on there being at least three views of the molecule rotated around three distinct common tilt axes (29). Our direct measurements, from class averages (see above), show a high degree of similarity of the backbone to F-actin structure. To produce a 3D reconstruction from our data set, we therefore had a potential model, an Arp1 filament, that we could use as an anchor-set for an initial round of Euler-angle assignment. For the very first round of Euler-angle assignment, we built a low-resolution, 10-subunit Arp1 model (3-nm resolution) with F-actin helical parameters. Two-dimensional reprojections of the model were used as an anchor-set. All subsequent refinement steps were carried out by using only anchor-sets generated from data. There was a good representation of orientations (every 10°) over 200° around the long axis. To attempt to obtain more views and a more complete 3D structure, a separate tilted data set (<20% of total data set) was included in the final data set. Inclusion of these data (which included particles tilted around the short axis, , 77–110°) improved the range of orientations around the long axis by a further 15°. Only data where structure in 2D reprojections from the 3D closely matched that of the input class averages was allowed to contribute to the final 3D. The resolution of the final map was 3 nm as determined by Fourier shell correlation (0.5 correlation value). All maps were low-passfiltered to this value before display here.
The 3D Structure of Dynactin and Fitting of Crystallographic Data and Models. In surface-rendered views (Figs. 2C and 3), the helical nature of the backbone subunits of the complex is clear. In addition to the periodic "zig-zag" structure of the Arp1 backbone, the extra mass contributing a halo or a spike of density is clearly visible, with a cleft in the density at the center of the halo (Fig. 3, arrow). This mass gives the complex a distinctly non-Factin-like appearance (see Fig. 2 A3 and A4, respectively). Views tilted away from the viewer (Fig. 2 B1 and C1) show that this central mass is elongated and crosses the barbed end of the filament.
|
The fitting of the Arp backbone highlights the presence of additional non-helically arranged mass (i.e., not contributed by the backbone) at the barbed end. Additional Arp subunits did not fit well into this extra mass, which is most likely attributable to CaP (69 kDa), the only large globular protein thought to bind to at this end of the complex. A preliminary fitting of the elongated CapZ crystal structure (27) (Fig. 5 A and B) into this elongated mass was carried out by eye avoiding major steric clashes with the Arp subunits in the backbone. By this method, a good fit was obtained to the density and there was close agreement with the appearance of the distinctive "spike" or "halo" observed at the barbed end in original class average views when the fitted map was rotated to comparable orientations. The relative contributions of Arp1 and CaP to the distinctive "halo" structure of the barbed end begin to become clear (Fig. 5C). The gap between the ultimate Arp subunit and CaP gives rise to the cleft seen in the barbed end density. There is an additional shield-like mass in the region of the barbed end located across the outer face of the ultimate Arp subunit and projecting outward from the dynactin backbone (see Fig. 5C) that could not be accounted for by CaP density.
|
-helices running along the underside of the -sheets (these helices are shown in Fig. 5D). Additional mass not contributed by the backbone Arp subunits is also observed at the pointed end. This must be contributed by pointed end complex proteins (see Discussion for more detail).
|
|
Discussion |
|---|
|
TopAbstractMaterials and MethodsResultsDiscussionReferences |
|---|
The PEC. The best-characterized component of the PEC is Arp11. Arp11 is thought to bind at the pointed end of the dynactin complex in a manner similar to Arp1, because residues involved in backbone formation are conserved at its barbed end. The tenth Arp subunit at the pointed end of the backbone of our structure is therefore likely to be the Arp11 subunit. Residues at the pointed end of Arp11 are not conserved with respect to other Arps, indicating that it performs some other function. This function is likely to be interaction with other PEC proteins, most likely with p62, which is thought to interact with both Arp1 and Arp11 (36) and probably accounts for most of the additional mass at the PEC in our reconstruction. We do not observe sufficient mass to account for the additional small proteins (p25 and p27) reportedly located at this end of the complex. This lack could be due to that fact that there is certain flexibility in the pointed end complex or that some of the extra mass is highly elongated or penetrated by stain and therefore not well resolved in our reconstruction.
The Barbed End. In our fit, CaP lies across the barbed end of the dynactin backbone with equivalent positions on each subunit closely approaching equivalent positions on the two barbed-end Arp1 subunits (Fig. 6). The potential for interaction with two Arp subunits is consistent with the fact that CaP binds to filamentous actin (or Arp1) but does not bind to monomers (37). The closest approach of CaP to Arp subunits is via the long -helices running along the underside of the -sheets of each subunit. These helices, and the C-terminal residues crucial for binding of the CaP subunits to actin/Arp, have been proposed as likely interaction points with the actin/Arp1 backbone (27, 37, 38). Yamashita et al. (27) suggested that C-terminal residues may change conformation to bind to Arp1. The closest approaches of CaP to Arp1 are to subdomain 3 of the two barbed-end Arp1 subunits, around residues associated with the longitudinal subunit–subunit interface of the Arp1 backbone. In our model, equivalent positions on CaP interact with equivalent positions on each of the two barbed-end Arp1 subunits. CaP possesses a twofold symmetrical architecture (residues crucial for maintaining architecture are conserved between the two subunits), which would be advantageous to this type of interaction (27). The CaP, positioned to directly block subunit–subunit interaction sites, would be optimally placed to perform its proposed role as a stabilizer (37, 39) of the filament backbone. It has been speculated that CaP might bind to the barbed end in a manner similar to gelsolin. However, the cleft between subdomains 1 and 3, to which gelsolin binds (PDB ID code 1C0F
[PDB]
) (40), does not appear to be directly obstructed in our fitted map (Fig. 6B). A distinct mode of binding from that of gelsolin is therefore indicated. The mode of CaP binding has broader significance because it is also a key component in the organization of the Z-lines in muscle.
The Shoulder/Sidearm Complex. What we have described so far accounts for 60% of the mass of the dynactin complex; what of the rest? This is contained in the shoulder/sidearm complex. We do observe some mass located over, and projecting out from, the outer surface of the ultimate Arp at the barbed end (Fig. 5C). This mass is most likely contributed by the attachment point of the p150Glued/dynamitin "shoulder complex" thought to bind to the Arp1 backbone in this region (11). There is insufficient mass to account for the whole of p150Glued (2 x 150 kDa), dynamitin (4 x 50 kDa), and smaller globular proteins (2 x 24 kDa). In the case of the p150Glued, the most likely reason for this is that it is highly elongated, being made up of a large proportion of -helical coiled-coil with it globular component placed at the end furthest from the Arp1 backbone (8). It is also very flexible, based on our own and previous observations (Fig. 1C) (11). These two factors mean that its structure will not be clearly observed in class averages. The issue with the dynamitin (shoulder complex) can be explained in the same manner because sequence prediction revealed a high potential for forming elongated -helical coiled-coil (41) rather than a compact globular structure. It has been suggested that the dynamitin may lie along the backbone in a manner similar to tropomyosin on actin filaments; this is supported by "dotblot" comparison of dynamitin and tropomyosin sequence. This comparison shows that dynamitin has periodically arranged repeats very similar in spacing to those of tropomyosin that are regularly spaced at intervals of an integral number of actins (P. Bennett, personal communication). There would appear to be a discontinuous strand of density running along the backbone in our reconstruction (see Fig. 3) that may be attributed to the presence of dynamitin.
In conclusion, we have visualized the detailed structure of the dynactin complex and the way in which several of its key components fit together to perform their functions. The backbone closely resembles that of F-actin, and the capping protein appears to be bound in a unique way distinct from that of gelsolin but blocking subunit–subunit interaction sites. Fig. 7 shows a model of how the dynactin complex might be arranged based on our results and incorporating those of others (10, 11). One intriguing question that we would like to answer is whether dynamitin really does lie along the Arp1 backbone in a manner similar to tropomyosin in muscle filaments. Further biochemical characterization of the interactions between PEC proteins and their crystallization will aid us in further determination of the detailed structure of the PEC.
|
|
Acknowledgements |
|---|
|
Footnotes |
|---|
This paper was submitted directly (Track II) to the PNAS office.
Abbreviations: CaP, capping protein; PEC, pointed-end complex.
Present address: School of Crystallography, Birkbeck College, University of London, Malet Street, London WC1E 7HX, United Kingdom. 
¶ To whom correspondence should be addressed. E-mail: walter.steffen{at}kcl.ac.uk.
© 2005 by The National Academy of Sciences of the USA
|
References |
|---|
|
TopAbstractMaterials and MethodsResultsDiscussionReferences |
|---|
CiteULike 
Complore 
Connotea 
Del.icio.us 
Digg What's this?
This article has been cited by other articles in HighWire Press-hosted journals:
|
|
 |
|
  K. A. Melkonian, K. C. Maier, J. E. Godfrey, M. Rodgers, and T. A. Schroer Mechanism of Dynamitin-mediated Disruption of Dynactin J. Biol. Chem., July 6, 2007; 282(27): 19355 - 19364. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 S. W. Clark and M. D. Rose Alanine Scanning of Arp1 Delineates a Putative Binding Site for Jnm1/Dynamitin and Nip100/p150Glued Mol. Biol. Cell, September 1, 2005; 16(9): 3999 - 4012. [Abstract] [Full Text] [PDF]
|
|
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
