Structure of a bacterial dynamin-like protein lipid tube provides a mechanism for assembly and membrane curving

Abstract

Proteins of the dynamin superfamily mediate membrane fission, fusion, and restructuring events by polymerizing upon lipid bilayers and forcing regions of high curvature. In this work, we show the electron cryomicroscopy reconstruction of a bacterial dynamin-like protein (BDLP) helical filament decorating a lipid tube at approximately 11 A resolution. We fitted the BDLP crystal structure and produced a molecular model for the entire filament. The BDLP GTPase domain dimerizes and forms the tube surface, the GTPase effector domain (GED) mediates self-assembly, and the paddle region contacts the lipids and promotes curvature. Association of BDLP with GMPPNP and lipid induces radical, large-scale conformational changes affecting polymerization. Nucleotide hydrolysis seems therefore to be coupled to polymer disassembly and dissociation from lipid, rather than membrane restructuring. Observed structural similarities with rat dynamin 1 suggest that our results have broad implication for other dynamin family members.

Figure 1

Fourier-Bessel Reconstruction of Native BDLP-GMPPNP Lipid Tubes at ∼26 Å Resolution (A) Annotated BDLP-GDP crystal structure (PDB 2J68, Low and Löwe [2006]) showing paddle surface mutants (for E). (B) BDLP tubes in amorphous ice (left). Fourier transform taken from a typical single BDLP tube showing good diffraction to ∼26 Å. Annotation shows lattice and assigned Bessel orders for each of the nine layer line pairs (right). (C) Fourier-Bessel reconstruction of the native BDLP tube at ∼26 Å resolution. A 90° slice of the helix in cross-section is shown. Its architecture agrees well with an ∼11 Å resolution reconstruction (see below) obtained through a single particle helical method (Sachse et al., 2007). (D) As in (C), but showing a surface view of the helix. Note the zigzag arrangement of the asymmetric units. Longitudinal contacts (red arrows) induce curvature while lateral contacts (yellow arrows) run almost in parallel with the tube axis. (E) BDLP liposome binding spin assays using paddle mutations that abrogate lipid binding (mutant positions shown in A).

Figure 2

Native and SAM-Labeled Helical Reconstruction of BDLP-GMPPNP Lipid Tubes by Single-Particle Methods at 11.0 Å and 16.9 Å Resolution, Respectively (A) Density surface overview of the native BDLP tube reconstruction. Red dumbbells show zigzag arrangement of the dimeric asymmetric unit. (B) As in (A), but sliced along the tube axis exposing the globular outer layer, inner radial spokes, and lipid tube core (red). (C) As in (A), but showing the tube in cross-section to the helix axis. The lipid core is in red. (D) Close-up view of region outlined in yellow in (C), showing surface detailing of the asymmetric unit and two-fold symmetry. (E) As in (A), but a close-up view of the asymmetric unit showing surface detail. (F) Density surface overview of the BDLP tube reconstruction incorporating the human p73α SAM-domain as a label, fused between neck and trunk. (G) Close-up view of the region outlined in yellow in (F). (H) Superposition of native (blue) and labeled (orange) reconstructions filtered to a resolution of 16 Å. Note the additional bridge of orange density between radial spokes attributed to the label. (I) As in (H) but showing region enclosed by dotted lines. The unexpected thinness of the orange density bridge is thought to be due to label flexibility.

Figure 3

Electron Density Details and the Central Lumen (A) Grayscale representation of the 3D reconstruction intensity values to show dynamic range. The inner ring has a thickness of 5 nm and shows a strong intensity band at a radius of 5 nm, probably representing the lipid head groups. No such ring is visible at a radius of 10 nm, where the outer leaflet head groups would be expected for a standard membrane bilayer of 5 nm thickness. (B) Same grayscale representation showing detailing of the spokes in radial cross-section (right top and bottom). The strongest intensities most likely correspond to the centers of BDLP alpha helices. (C) Analysis of the BDLP tube in cross-section shows little evidence for the presence of a normal, ordered bilayer outer leaflet. 1, Nontubulated, uncoated liposome bilayers are ∼5 nm in width. 2, Cryo-transmission electron microscopy (cryo-TEM) image of a BDLP tube stub. Strong radial density is observed at ∼5 nm and ∼7 nm which may represent a highly compressed bilayer. Alternatively, the band at 7 nm radius is from BDLP. 3 and 4, projections of the combined radial sections from (A) (the high-resolution reconstruction) and their radial averages also show sharp peaks both at ∼5 nm and ∼7 nm, thus agreeing with the TEM image. (D) Same as in (C4), line plot showing the magnitudes of the peaks and bands in the rotationally averaged sections of the high-resolution reconstruction (see A).

Figure 4

Fitting of the BDLP-GDP Crystal Structure into Native and Labeled Reconstruction Requires Substantial Domain Rearrangement (A) Model of the helical BDLP filament fitted into the native tube reconstruction shown in cross-section. (B) Close-up stereo image of region outlined in yellow in (A), showing the fit of two BDLP molecules that form the dimeric asymmetric unit of the reconstruction. (C) Close-up view of indicated region from (B), showing fit of the neck helices. (D) Close-up view of indicated region from (B), showing fit of trunk helices. (E) Surface view showing the fit of the GTPase domain homodimer within the density. (F) Left: GDP-containing dimer of the BDLP GTPase domains as crystallized. Right: In order to fit the BDLP lipid tube density accurately, a GTP-form of the dimer was generated by superimposing the two halves of the BDLP GTPase dimer onto the dimer of hGBP1 (PDB code 2B92, Ghosh et al. [2006]). This results in a rotation of 15° as is shown in the figure and the two domains move slightly closer. (G) Modeled fit of two dimeric BDLP molecules each with a fused p73α SAM domain between amino acids 498 and 499. The label acts as an anchor to orient the fitted BDLP molecule. (H) Reconstructed electron density with the threshold greatly increased to reveal the strongest details only. In the radial spokes, for example, the quality of the density is sufficient for dimeric barrels representing α helices to be observed. The handedness of the helices winding around each other in this region fits the crystal structure. (I) Stereo plot showing the fit of the atomic model main chain within the density of the GTPase domain. The density is slightly more sharpened than in the other figures (B factor of 800 Ų, compared to 400 Ų previously) to emphasize the secondary structure elements on the inside. To our eyes, the resolution corresponds to around 11 Å as indicated by the FSC (Figure S2I). The GTPase domain consists of one large central β sheet, and this feature and its twist are clearly resolved. The GTPase domain structure of BDLP is fitted here as a rigid body, so slight movements indicated by the new density have not been adjusted. The large opening at the bottom of the figure contains the junction between the neck and GTPase domains and for this figure the atomic model in this region has been removed for clarity reasons. (J) Averaged power spectrum of 8150 in-plane rotated tube segments used in the native single particle reconstruction. Layer lines could be resolved to a resolution of 11.5 Å, confirming all other estimates of resolution. (K) Platinum rotary shadowing after fixation with 1%–2% uranyl acetate. Under these conditions, the tubes slightly unwind (along the weaker lateral contact mediated by helix H4) and expose the seam between left-handed 11-start helices. (L) Longitudinal sections through a cryotomogram of a BDLP tube. The surface section clearly shows the left-handed striations of the 11-start rise. The middle section shows the strong outer density generated by the GTPase domains. The lipid tube clearly runs along the length of the filament. No bilayer is apparent here, although resolution may be limiting. (M) Molecular interpretation of the images in (K) and (L), showing the longitudinal 11-start left-handed rise.

Figure 5

The GTP and Lipid-Induced Conformation Changes A three-step morph between the BDLP-GDP crystal structure (Low and Löwe, 2006) and the fitted BDLP-GMPPNP model (this study). The sequence of domain rearrangements is unknown and shown arbitrarily. Please also consult Movies S1 and S2, showing the same data in motion, from two different angles. Most changes can be accommodated with two-hinged movements, and both rotations are in plane. Note that in panel 3, helix 12 (H12, colored orange) is likely to follow the distinct bridge of density (dotted red line) that connects to the top of the GTPase domain.

Figure 6

Model of the BDLP-GMPPNP Helical Filament Shows Protein-Protein Contacts and Mechanism of Lipid Curvature (A) Model of the helical BDLP filament in cross-section to the helix axis showing a fitted lipid bilayer. Only the inner ring of the lipid head groups (and hence lipids) is clearly observed in the 3D density, although the averaged density profile (Figure 3D) shows two sharp peaks that agree with direct end-on views (Figure 3C). The outer leaflet may not stand out in 3D because the ring of head groups is disrupted by the BDLP trunk tips and/or the bilayer is compressed to about half its natural thickness. A standard outer leaflet (5 nm bilayer thickness) is modeled for size comparison only. Shown close up are protein-protein contacts between a pair of asymmetric units. The focus is on interaction between the central neighboring neck and trunk helices. (B) Surface view of the BDLP filament model. Shown close up is the arrangement of three dimeric asymmetric units within the helix. Polymerization arises through longitudinal back-to-back contacts between GTPase domains, between neck and trunk helices, plus lateral association of H4 helices. The disordered switch 2 region is represented by a dashed orange line. Note that the lateral contact is smaller, probably leading to unwinding in Figure 4K. (C) Side and top view superposition of BDLP-GDP (Low and Löwe [2006], residues 68–348, colored cyan although helix 13 colored blue for clarity), BDLP-GMPPNP model (this study, lipid-bound form, residues 68–348, mainly colored green although helix 12 is colored orange), and rat dynamin 1 (nucleotide free, residues 33–304, colored red). Note how BDLP helix 12 and rat helix α5 are almost identically positioned and run in phase. The kink in helix α5 corroborates the BDLP-GMPPNP tube reconstruction, which suggests that helices 12 and 13 separate and act as a hinge (in the BDLP crystal structure H12 and H13 are almost continuous). (D) Side view of the rat GTPase domain (residues 2–304) compared to the BDLP-GMPPNP GTPase domain with the kink between helices 12 and 13 modeled. The bending between helices 12 and 13 is observed in the equivalent rat helix α5. Also note how the N termini of the GTPase domains in both rat (2–33) and BDLP (2–68) contribute to the formation of a hydrophobic groove that seats the GED in BDLP. Rat Pro 32 is situated in the equivalent position to BDLP Gly 68 (Hinge 2a), suggesting the rat GTPase domain may also show flexibility around this region and the kink in helix α5.

Figure 7

A “Passive” Polymerization/Depolymerization Model for Fusion and Fission (A) Schematic drawing showing the different stages of BDLP/dynamin-induced fission and fusion. Polymerization is induced by GTP binding and induces high curvature. Hydrolysis to GDP causes catastrophic disassembly and produces a transition state that can either go back (gray arrow) or resolve through the rearrangement of the membrane linkage (blue connections). If the two membranes belong to the same surface, this results in fission. If they belong to two different surfaces (two vesicles, for example), the process results in fusion. (B) More detailed drawing of the same model as in (A) (bottom), shown from the side. Tubulation causes high curvature through the insertion of the paddle into the outer leaflet by pure displacement of lipids and/or compression of the lipid tails. After disassembly, this leaves the bilayer in an unstable state that can be relieved through the combination of two (or more) into one, producing less curvature.