If you don't remember your password, you can reset it by entering your email address and clicking the Reset Password button. You will then receive an email that contains a secure link for resetting your password
If the address matches a valid account an email will be sent to __email__ with instructions for resetting your password
Université de Lille, CNRS, Inserm, CHU Lille, Institut Pasteur de Lille, U1019-UMR9017-CIIL-Centre d'Infection et d'Immunité de Lille, F-59000 Lille, France
Weill Cornell Medicine, Department of Anesthesiology, 1300 York Avenue, New York, NY 10065, USAWeill Cornell Medicine, Department of Physiology and Biophysics, 1300 York Avenue, New York, NY 10065, USAKavli Institute at Cornell for Nanoscale Science, Cornell University, Ithaca, New York, NY 14853, USA
Endosomal Sorting Complex Required for Transport (ESCRT) proteins assemble on the cytoplasmic leaflet of membranes and remodel them. ESCRT is involved in biological processes where membranes are bent away from the cytosol, constricted, and finally severed, such as in multi-vesicular body formation (in the endosomal pathway for protein sorting) or abscission during cell division. The ESCRT system is hijacked by enveloped viruses to allow buds of nascent virions to be constricted, severed and released. ESCRT-III proteins, the most downstream components of the ESCRT system, are monomeric and cytosolic in their autoinhibited conformation. They share a common architecture, a four-helix bundle with a fifth helix that interacts with this bundle to prevent polymerizing. Upon binding to negatively charged membranes, the ESCRT-III components adopt an activated state that allows them to polymerize into filaments and spirals, and to interact with the AAA-ATPase Vps4 for polymer remodeling. ESCRT-III has been studied with electron microscopy (EM) and fluorescence microscopy (FM); these methods provided invaluable information about ESCRT assembly structures or their dynamics, respectively, but neither approach provides detailed insights into both aspects simultaneously. High-speed atomic force microscopy (HS-AFM) has overcome this shortcoming, providing movies at high spatio-temporal resolution of biomolecular processes, significantly increasing our understanding of ESCRT-III structure and dynamics. Here, we review the contributions of HS-AFM in the analysis of ESCRT-III, focusing on recent developments of non-planar and deformable HS-AFM supports. We divide the HS-AFM observations into four sequential steps in the ESCRT-III lifecycle: 1) polymerization, 2) morphology, 3) dynamics, and 4) depolymerization.
INTRODUCTION
Research into the Endosomal Sorting Complex Required for Transport (ESCRT) family of proteins is tightly linked to two questions that arose regarding the mechanism by which proteins are sorted in the endosomal and biosynthetic pathways, namely: i) what is the signal that determines the fate of cargo proteins which enter these pathways, and ii) what constitutes the machinery that mediates this sorting? Initial genetic and biochemical investigations showed that mutations in a set of budding yeast genes led to aberrations in the endosomal protein sorting pathway (
). Thus, a subclass of protein mutants that resulted in defects in the endosomal pathway was discovered to lead to the formation of pre-vacuolar compartments in yeast cells that accumulated cargo proteins (
). The abrogated sorting of cargo proteins towards their endpoint - the yeast vacuole, which serves to hydrolyze cargo proteins, and is thus functionally similar to metazoan lysosomes – indicated that these proteins might either recognize the sorting signals in the endosomal pathway and/or carry out sorting themselves.
Biochemical and structural investigations later demonstrated structural and functional heterogeneity within this set of proteins. While all these proteins are cytosolic, some were shown to co-assemble into soluble complexes at fixed stoichiometries before interacting with lipid membranes. Others were shown to shift their distribution towards membranes, forming membrane-bound complexes with other ESCRT proteins of indeterminate sizes and stoichiometries (
). Soon, a clear distinction could be made between the various subcomplexes that composed the system, as they could be separated into those that function in recognition and binding of ubiquitinated cargo proteins (ESCRT-0 and ESCRT-I), those that serve as downstream acceptors of the cargo proteins and initiate the assembly of the sorting machinery (ESCRT-II), and the most downstream components that are necessary for multivesicular body (MVB) formation (ESCRT-III) – a crucial step in endosomal sorting, as cargo proteins must be internalized within late endosomes to become accessible to lysosomal hydrolases. The most abundant component of the ESCRT-III system is Snf7 (aka Vps32, CHMP4 in human), which forms filaments that assemble in large-scale supramolecular structures such as spirals that are at the basis of functional characteristics of the ESCRT-III system in membrane deformation.
An additional component, the AAA-ATPase Vps4, was shown to interact with ESCRT-III subunits, allowing them to be recycled to the cytosol and reused (
). Thus, transient recruitment of ESCRT proteins to the membrane of endosomal compartments, ESCRT-mediated membrane remodeling away from the cytosol, and subsequent separation of the nascent intralumenal vesicles (ILVs) from the endosomal membrane, are central to protein sorting in the endosomal system (Figure 1a). This understanding inspired numerous studies that confirmed the importance of ESCRT proteins to enveloped virus budding (
). For a detailed overview of the numerous cellular processes involving ESCRT proteins, we direct the interested reader to a comprehensive recent review (
Figure 1Cellular functions of the ESCRT-III system and models of ESCRT-III mediated membrane deformation mechanisms. a. Overview of processes where ESCRT is involved. Schematic of a cell illustrating where ESCRT proteins assemble and mediate membrane deformation processes. b. Schematic of membrane and ESCRT-III during abscission in cytokinesis and neuronal pruning. c. Mono-component spiral spring model d. Multi-component sequential polymerization model.
The identification of ESCRT proteins as an evolutionarily conserved and physiologically important membrane-remodeling machinery gave rise to numerous proposals on the molecular mechanism that governed ESCRT assembly, ESCRT-mediated membrane remodeling, Vps4-mediated ESCRT disassembly and ESCRT-mediated membrane scission. Initial models suggested that homopolymeric assemblies of ESCRT-III components acted as loaded spiral springs that release accumulated energy in an out-of-plane buckling deformation of the membrane-bound protein assembly (Figure 1c)(
). These models were inspired by findings that the intracellular plasma membrane leaflets of cells overexpressing Snf7 were decorated either by planar, spiral-like structures, or by helical structures found in membrane protrusions pointing away from the cytosol formed in presence of hydrolytically inactive Vps4, as demonstrated by electron microscopy (
). Experimentally, filaments were observed to be single- or multi-stranded, elastic and with a preferred radius of curvature that would experience stress both in the innermost, highly curved parts of the assembly and in the outermost low-curvature filament stretches (
). The stress accumulated in the filaments was estimated in terms of energy and compared to the energy needed for possible bending and buckling of the membrane. The estimation was on the order of ∼170 kBT (kBT being the product of the Boltzmann constant and temperature, used in biophysics as a scalable factor to express energy in relevant values at the molecular scale (1 kBT = 4.11×10-21 J = 4.11 pN·nm)). As the bending energy required to shape a hemispherical membrane dome from a flat membrane is on the order of ∼160 kBT, the energy accumulated in the filament would be sufficient to bend the membrane (
). However, it remained unclear whether the loaded spiral springs model of membrane deformation was applicable in a physiological context and how it could account for the variety of membrane geometries (both in terms of curvature as well as length scales) that ESCRT-III must adapt to. Thus, the limitations of the monocomponent loaded spiral spring model led to the elaboration of an expanded model, in which a sequential, Vps4-mediated, ATP-fueled polymerization-depolymerization sequence of different ESCRT-III components leads to gradual changes in the composition and mechanical properties of ESCRT-III assemblies, concomitant with gradual changes in the geometry of the underlying membrane (Figure 1d)(
). Both models are based on the concept that the membrane-bound protein assemblies store energy by stress accumulation and release it during cell membrane remodeling. This concept is appealing as membrane deformation implicates energy costs that exceed single molecule actions.
In this review, we give a brief description of recent developments in high-speed atomic force microscopy (HS-AFM) that provided novel insights into the structural, dynamical and biophysical properties of the most abundant ESCRT-III component, Snf7 (aka Vps32, CHMP4 in humans). We extend our discussion regarding the unique possibilities offered by HS-AFM to characterize structure and dynamics of biomolecular assemblies.
Understanding the molecular mechanism of the ESCRT system
The biggest hurdle in understanding the molecular mechanism of the ESCRT system is the study of both its structure and its dynamics, parameters that are in this case inextricable from each other, because the ESCRT system has no fixed structure but is in a continuous turnover and rearrangement throughout its functional cycle. The simplest approach, that also provides the most protein-specific results, is the bottom-up reconstitution of the system with a limited number of purified components (
). Such isolated versions of the system allow to verify whether purified components behave as predicted, and how they relate to ESCRT-III structural assemblies in cells.
Electron microscopy (EM) provided information about the structure of ESCRT components in isolation or in combination with other components (
). This technique identified the formation of various supramolecular structures, planar as well as three-dimensional. Yet, electron micrographs represent snapshots of the system frozen in time and do not provide information about the path the system took to arrive at a given state nor how it will develop from that state. Crucially, questions regarding the transition from a single-component planar configuration to a non-planar structure by addition of other components remain unanswered – in other words, using EM, the dynamics of protein assemblies can only be inferred. Shortly after the first description of the ESCRT system (
), the ability of several ESCRT-III components to homo- or hetero-polymerize was first assessed by EM, resulting in observations of diverse higher-order structures. Recombinant Saccharomyces cerevisiae Vps24 was found to polymerize into helical structures, while Snf7 formed sheet-, ring- and string-like structures (
). However, similar EM studies resulted in contrasting outcomes, requiring either point mutations or large-scale deletions of the autoinhibitory C-terminal region in ESCRT-III subunits to facilitate polymerization (
). Fluorescence microscopy takes advantage of fluorescently labeled membrane and protein components to observe the dynamics of different ESCRT proteins polymerizing or co-polymerizing on the membrane. In this configuration, the action of the protein is observed using a lipid vesicle or a planar lipid bilayer as substrate on which the ESCRT proteins can assemble and interact. The information derived from this approach depends on the choice of the geometry of the lipid substrate: using intact liposomes in combination with purified ESCRT components, confocal fluorescence microscopy provided information on how the protein and membrane colocalize over time and how the shape of the membrane changed with addition of proteins to the system (
). Using planar supported lipid bilayers, no information about changes in the membrane geometry could be obtained, but observation of the colocalization of individual ESCRT components could be used to derive information about the interaction of ESCRT proteins (
). The benefit of obtaining such dynamic information comes at a cost – the loss of information about the structure of ESCRT-III assemblies on lipid bilayers. Rigorously, no statement regarding the structure and assembly of the ESCRT proteins can be made based on the measurement of fluorescence signals emerging from attached fluorescence tags. Also, the typical spatial resolution of these fluorescence microscopy experiments is in the hundreds of nanometer range, larger than the size of an individual ESCRT-III spiral and orders of magnitude larger than the thickness of a filament.
Thus, neither electron microscopy nor fluorescence microscopy allow to fully appreciate the ESCRT-III system’s interaction between its constituents and with lipid bilayers. Covering this blind spot to understand the structure-dynamics duality of ESCRT assemblies, HS-AFM provides movies with nanometer lateral and subsecond temporal resolution.
HS-AFM to the rescue
AFM makes use of a nanometer sharp tip attached to a flexible micro-cantilever to scan the surface of a sample (Figure 2a). Recording the signal of the laser reflected from the cantilever informs about the tip displacement and is used to control the distance between sample and tip to reconstruct a topographic image of the sample. HS-AFM is the result of a series of developments that greatly enhanced the scanning speed of AFM for the study of the dynamics of biomolecules (
). HS-AFM imaging has helped elucidate dynamic processes of several biological systems at the molecular level on model membranes. A breakthrough was the study of the structural dynamics of canonical molecular motors (
Listeriolysin O Membrane Damaging Activity Involves Arc Formation and Lineaction -- Implication for Listeria monocytogenes Escape from Phagocytic Vacuole.
), to observations of peripheral membrane proteins and their interaction with supported lipid bilayers (SLBs), like oscillations of MinDE, the major components of the Min protein system that defines the site of bacterial cell division (
Figure 2HS-AFM for the study of the structure and dynamics of membrane-associated ESCRT-III assemblies. a. Schematic of the HS-AFM setup (main components labeled). b. HS-AFM supports. A planar stiff support (top, brown), a non-planar stiff support (middle, pink), and a planar soft support (bottom, blue) mounted on top of a glass cylinder. c. Left: On all supports the lipids spread forming a SLB. For the non-planar support and the soft support, oxygen plasma treatment was needed to render the surface hydrophilic to facilitate lipid spreading. The two halves of each panel show either the support before and after lipid spreading (planar stiff support (c, top left), and planar soft support (c, bottom left)) or the SLB-covered support with and without the surface topology subtracted (c, middle left). Right: The SLBs follow the support topography but display a local roughness of only some hundreds of picometers on all the supports. d. Examples of ESCRT-III assembly morphologies on each sample supports.
HS-AFM typically operates at between 1 and 20 frames per second, yet simply providing the acquisition frame rate is an imperfect indicator of HS-AFM performance. The possible speed of image acquisition is crucially limited by the speed of the feedback operation, which is in turn challenged by the travel velocity of the tip with respect to the sample and the corrugation of the sample. In other words, a completely flat sample allows for faster image acquisition than a highly corrugated sample. An image of a sample area of defined size can be acquired faster and more frequently by reducing the number of lines sampled; the physical principles behind the image acquisition process remain the same, thus one can trade acquisition speed for pixel resolution. Yet, many biological phenomena proceed faster than can be temporally resolved by acquiring between 1 and 20 frames per second. Therefore, new HS-AFM approaches were designed to allow observations of the action of single molecules, increasing the temporal resolution for the characterization of conformational dynamics at least 1000-fold. First, in HS-AFM line scanning (HS-AFM-LS) the y-scanning, i.e. the slow scan axis, is disabled, and instead of raster imaging, the central scan line is repeatedly recorded. This mode has been applied to transporter proteins such as glutamate transporters (GltpH, (
)) where a lateral displacement of the E-F loop opens the cytoplasmic gate for proton uptake. A second modality to characterize rapid conformational transitions is HS-AFM height spectroscopy (HS-AFM-HS)(
), where the AFM tip remains fixed at a defined position. In this mode, with a temporal resolution of 10 μs, the diffusion of molecules under the tip can be assessed, the diffusion coefficient calculated, and oligomer size and molecular surface concentrations determined (
). One of the most attractive advantages of these techniques is that they are applied to unlabeled proteins.
Unfortunately, the HS-AFM field is still relatively small, and it is perceived that only few groups in the world apply the technique successfully. Indeed, HS-AFM is a rather new technique (∼10 years) which was mainly implemented in home-built HS-AFM setups, often relying on in-house expertise of researchers in a specific lab. However, recent reviews and protocols (
), summer schools providing hands-on training and live demonstrations of HS-AFM operation, and a growing number of commercially available microscopes paired with user-friendly software and automated setups have now allowed many groups with little previous expertise in HS-AFM to utilize this technique, with promising results (
). Indeed, as relatively easy-to-use commercial HS-AFMs become available, it is expected that an increasing number of researchers take advantage of the technique’s possibilities to characterize the dynamics of the biological system of their interest.
Highlighted by all the different applications mentioned, HS-AFM imaging is a powerful and versatile technique to study the ESCRT-III system. Most recently, new developments in AFM surface supports have opened new possibilities to study membrane-bound protein systems unconstrained by the planarity and rigidity of traditional supports. Traditional HS-AFM supports such as muscovite mica provided atomically flat surfaces for the structural and dynamical investigation of the ESCRT-III system (Figure 2b, top) (
). Practically, the use of such a planar and rigid support prevents the study and observation of ESCRT-III 2D to 3D conformational changes and its mechanism in membrane deformation. The design of new non-planar (Figure 2b, middle) and soft polydimethylsiloxane-based (PDMS-based) (Figure 2b, bottom) supports has provided information about the topological preference of Snf7 spirals and allowed to observe spiral buckling (
In contrast to SLBs on atomically flat mica (Figure 2c, top), the non-planar sample support consisted of undulated bumps with ∼150 nm periodicity, ∼35 nm height and ∼75 nm radius (Figure 2c, middle). The soft support, which served as a compressible substrate under supported lipid bilayers, was made of PDMS and had a Young’s modulus of ∼270 kPa (
). The PDMS was ∼10 μm thick and flat as a freshly cleaved mica sheet was used as a mold (Figure 2c, bottom). For the formation of the SLBs on either the nanopatterned undulated support or on the PDMS, the surfaces were treated with oxygen plasma to increase their hydrophilicity and facilitate liposome adsorption and SLB spreading.
The potential of HS-AFM for dynamic observations of reconstituted systems at high spatial resolutions was harnessed to research the ESCRT system as early as 2011 (
Association of the Endosomal Sorting Complex ESCRT-II with the Vps20 Subunit of ESCRT-III Generates a Curvature-sensitive Complex Capable of Nucleating ESCRT-III Filaments.
Journal of Biological Chemistry.2011; 286: 34262-34270
Association of the Endosomal Sorting Complex ESCRT-II with the Vps20 Subunit of ESCRT-III Generates a Curvature-sensitive Complex Capable of Nucleating ESCRT-III Filaments.
Journal of Biological Chemistry.2011; 286: 34262-34270
), these works provided important insights into ESCRT structural dynamics. We hope that this review will inspire many more researchers to apply AFM technology to study the ESCRT system, as the current studies have readily allowed to provide novel insights into polymerization, morphology, dynamics and depolymerization of ESCRT-III on supported lipid bilayers, as detailed below.
Polymerization
A key assumption about the membrane-remodeling mechanism of the ESCRT system is the proposed ability of ESCRT-III to form membrane-bound higher-order structures. For example, the ability of Snf7 to oligomerize was found to be important in order to trap cargo molecules internalized in the endosomal pathway at endosomal membranes. Mutations that inhibited Snf7 oligomerization resulted in cargo molecules being mis-sorted to the plasma membrane, whereas mutations that caused excess Snf7 oligomerization resulted in cargo molecules being accumulated in pre-vacuolar structures (
). Thus, fine-tuned ESCRT-III polymerization on membranes is thought to be crucial to ESCRT-mediated membrane remodeling processes. Understanding the way in which these proteins come together to form higher-order structures became a crucial point of interest in ESCRT studies. As described above, the interpretation of the earliest EM observations of ESCRT-III homo- or heteropolymers was complicated by the polymers being formed in solution in absence of membranes - it was unclear to what extent such structures were representative of membrane-bound ESCRT-III structures found in cells (
), establishing how individual ESCRT subunits interacted with lipid bilayers to form higher-order structures required a technique that allowed dynamic observation of these interactions in a liquid environment and at ambient conditions.
The first HS-AFM-based observation of ESCRT-III polymerization (
) utilized a SLB deposited on atomically flat mica supports immersed in a liquid chamber. The open nature of the HS-AFM liquid chamber allows manipulation of the buffer during imaging, and thus allowed instantaneous changes in the concentration of Snf7 in solution. Upon addition of Snf7 to the imaging chamber, the formation of Archimedean spiral-like structures was observed, with some Snf7 spirals forming by branching from pre-existing Snf7 spirals (Figure 1a, Figure 3a). The Snf7 filaments were shown to be multi-stranded, and the dynamics of the individual filaments could be tracked due to the high spatio-temporal resolution of HS-AFM. The homo-polymerization of Snf7 into spiral-like structures on mica-SLBs represented the first unambiguous observation of ESCRT-III polymerization dynamics at the molecular scale. The approach used in this study, involving a single ESCRT-III component, remains useful to investigate the biophysical properties of individual ESCRT components in isolation.
Figure 3Structural dynamics of membrane-associated ESCRT-III assemblies. a) to d) Several aspects of the polymerization of ESCRT-III. a) Addition of Snf7 monomers to the suspension in the fluid cell led to the in-situ formation of spiral-like assemblies on planar, rigid-supported lipid bilayers, which branch and split to form new turns or new assemblies. b) Curved rigid substrates displaying three different geometries (convex protrusions, shallow concave interstices (green) and saddle-shaped areas (red) are used to support a lipid bilayer on which the polymerization of Snf7 was followed depending on membrane curvature. The images show Snf7 assemblies populating all the geometries adjusting their localization and morphology to the geometry of the underlying bilayer. c) Polymerization of Snf7 spirals on flat, soft PDMS surfaces. d). Spiral morphology before and after addition of Vps2 and Vps24. Subsequent addition of Vps2/Vps24 to Snf7-only assemblies on flat rigid supports induced spiral densification, seen as a compact, disk-like morphology. e) to g) Morphology of the ESCRT-III assemblies. e) High-resolution imaging on flat, rigid supports allowed visualization of the connection between filaments, split filaments connecting spirals and filament splits within spirals. Crowded surface conditions lead to deformation and polygonality of the filaments. f) Low density of protein on flat, rigid supported membranes led to isolated CHMP4B-ΔC spirals. g) Injecting lower concentrations of Snf7 than in f) to the fluid cell allowed observation of Snf7 spiral doublets, consisting of two Snf7 spirals connected by a single Snf7 filament. h) and i) ESCRT-III dynamics. h) HS-AFM imaging on flat-soft deformable substrates revealed the out-of-plane transition of the inner turns of Snf7 spirals; without the constraint of a rigid support, Snf7 shows ability to deform the membrane. i) Liposomes adsorbed on solid substrates as a platform for CHMP4-ΔC polymerization, deformation of the ensemble after addition of CHMP2-ΔC. j) to l) ESCRT-III depolymerization. j) Planar spiral structures on SLBs on rigid supports: Washing out soluble ESCRT-III components from the HS-AFM chamber and adding Vps4 and ATP to the imaging solution led to depolymerization of ESCRT-III assemblies consisting of Snf7, Vps2 and Vps24. The innermost, over-curved part of the assembly seemed refractory to depolymerization. A similar experiment is shown in k) in the presence of soluble Snf7 in the fluid cell: Dynamic reorganization where pre-formed spirals shrink due to depolymerization, and the membrane areas freed allow nucleation and growth of newly formed spirals. l) Membrane-free tubes of CHMP2A-ΔC-CHMP3 formed in suspension and adsorbed on rigid, SLB-covered supports. In the presence of Vps4 and ATP, remodeling of the tube can be followed by HS-AFM. Local constriction of the tubes preceded their disassembly. Panels a) and e) are reproduced from (
While the experimental setup, using a planar, rigid substrate supporting a lipid bilayer, allowed for real-time observations of Snf7 polymerization, it lacked the geometric heterogeneity of biological membranes. Thus, these initial studies could not provide information on the curvature sensitivity of the proteins, which could serve as an additional cue for nucleation of ESCRT-III assemblies, independent of or in combination with the established biochemical cues for nucleation (such as upstream ESCRT components – ESCRT-II)(
). Previous investigations into the curvature dependence of ESCRT polymerization were limited to observations using conventional AFM, which lack temporal resolution as compared to HS-AFM observations (
Association of the Endosomal Sorting Complex ESCRT-II with the Vps20 Subunit of ESCRT-III Generates a Curvature-sensitive Complex Capable of Nucleating ESCRT-III Filaments.
Journal of Biological Chemistry.2011; 286: 34262-34270
). Early attempts to reconstitute ESCRT complexes for conventional AFM were performed by pre-incubating ESCRT components with liposomes and subsequent deposition of the sample on mica. Such an approach resulted in relatively sparse ESCRT interaction with an SLB, ostensibly with differential affinity for the planar regions of the SLB versus edges of SLB defects depending on ESCRT composition. Therefore, the results presented in ref. (
Association of the Endosomal Sorting Complex ESCRT-II with the Vps20 Subunit of ESCRT-III Generates a Curvature-sensitive Complex Capable of Nucleating ESCRT-III Filaments.
Journal of Biological Chemistry.2011; 286: 34262-34270
) were inconclusive, as protein aggregation on edges of SLB defects could be the result of the protein interacting not only with the SLB edge, but also with the surface of the hydrophilic mica support; nor do they appear to be protein specific, as it is quite commonly observed that proteins adsorb in membrane areas of lower energy cost, such as edges and defects. Another challenge of such an approach is that it is prone to overinterpretations regarding the composition of observed structures as all ESCRT components are co-incubated with liposomes prior to sample application on mica. Finally, this approach did not provide conclusive evidence of the formation of higher-order structures, as opposed to approaches where the formation of spiral structures was readily trackable through prolonged time-lapse observation (
Association of the Endosomal Sorting Complex ESCRT-II with the Vps20 Subunit of ESCRT-III Generates a Curvature-sensitive Complex Capable of Nucleating ESCRT-III Filaments.
Journal of Biological Chemistry.2011; 286: 34262-34270
The limitation of surface planarity was recently overcome by using a HS-AFM calibration grid as a nanopatterned non-flat substrate for SLB formation. By exposing the calibration grid to oxygen plasma, its surface was rendered hydrophilic, allowing for efficient adsorption and spreading of small unilamellar vesicles injected directly into the imaging chamber. The result was a rigidly supported lipid bilayer with three distinct geometries present throughout the sample: convex protrusions, shallow concave interstices nested between protrusions, and saddle-shaped areas separating laterally neighboring protrusions (Figure 2b,c,d, middle, Figure 3b). This approach allowed dynamic observations of Snf7 polymerization on curved lipid bilayers at the molecular level and showed that Snf7 monomers were curvature-insensitive, populating membrane areas of different geometries, while higher-order structures were curvature-sensitive, adjusting their localization and morphology to the geometry of the underlying bilayer (
). These observations further substantiated prior suggestions that the high torsional rigidity of the Snf7 filaments is the driving factor for membrane curvature sensing and adaptation (
Another step forward was the design of soft sample supports. Using a soft PDMS cushion underneath the SLB, the ability of Snf7 spirals to deform the membrane without the constrain of the rigid support could be tested (Figure 2b, bottom panel, Figure 3c). This approach allowed the observation of Snf7 polymerization and of mature Snf7 spirals that revealed visible differences to the spirals grown on SLBs of the same composition on mica. On the PDMS-SLB support, Snf7 spirals were smaller and had a more compact filament packing.
Besides revealing how an individual ESCRT-III component, Snf7, polymerizes to form higher-order structures on supported lipid bilayers, HS-AFM also allowed observations of sequential addition of other ESCRT-III components and the subsequent changes in the morphology of Snf7 spirals (
). When adding Vps2/Vps24 to the imaging chamber in case of both the mica-SLB and the PDMS-SLB, HS-AFM time-lapse imaging showed compaction of the Snf7/Vps2/Vps24 spirals with single unpaired filaments and filament pairs co-existing (Figure 3d), indicating a second round of polymerization of ESCRT-III subunits alongside Snf7 subunits (
). These HS-AFM observations, in combination with reconstitution of ESCRT-III polymerization by fluorescence microscopy on SLBs, proved crucial to promote considerations of the molecular mechanism of ESCRT-III mediated membrane deformation, resulting in the proposal of a Vps4-mediated, ATP-fueled, sequential polymerization-depolymerization sequence of ESCRT-III components. In such a framework, changes in the composition of ESCRT-III assemblies lead to gradual changes in their biophysical properties and to changes in the geometry of the membrane-bound protein complexes (
Due to its ability to resolve fine structural details of biological samples, HS-AFM allows the characterization of the morphology of individual ESCRT assemblies. ESCRT morphology could be tightly linked to ESCRT function, as the shape of ESCRT assemblies is presumed to change as a function of assembly composition (
). The morphologies of ESCRT assemblies as observed by HS-AFM were correlated with the specific protocols used to prepare samples. As described before, early investigations using conventional AFM relied on incubation of proteins with liposomes prior to deposition on the sample surface. Thus, the only ESCRT proteins present in these samples would be those that effectively interacted with the liposomes in suspension. The modular bottom-up approach developed in later works (
), where SLB deposition on the sample stage is separated from protein polymerization, provided a chemical gradient between the protein in suspension and the initially protein-free lipid bilayer, which promoted nucleation and polymerization of ESCRT-III assemblies on the SLBs. Such an approach resulted in crowded homo-polymeric budding yeast Snf7 spiral structures being observed (Figure 3e)(
). In crowded conditions, the outer turns of Snf7 spirals deviated from their Archimedean morphology, resulting in a polygonal structure, with Snf7 filaments adopting a linear shape, including kink-discontinuities along the outer turns (Figure 3e). The observation of the Snf7 spiral morphology and their packing characteristics allowed also for the calculation of the persistence length and the parametrization of the loaded spiral spring model that assumes a preferred radius of curvature of the individual Snf7 filament (
), though the irregular structure does not seem to be due to lateral compression. Similarly, addition of yeast Vps2 and Vps24 to pre-formed Snf7 spirals resulted in densification of the spirals without a change in the spiral area, indicating the formation of filament bundles and an increase in the filament content of ESCRT-III assemblies (
) allowed the observation of a distinct Snf7 assembly morphology, with two Snf7 spirals being formed from and connected by a single Snf7 filament, thus forming a doublet of paired Snf7 spirals (Figure 3g)(
). The structure of budding yeast Snf7 has been solved by X-ray crystallography in its open conformation suggesting that helices 1 - 4 allow the interaction of two Snf7 molecules during polymerization (
). The distinct paired spiral morphology indicated that, even though a Snf7 filament had a preserved handedness, it could form spiral pairs with both handedness in the two sub-spirals. This recently observed morphology of Saccharomyces cerevisiae Snf7 is very similar to that previously observed in its Caenorhabditis elegans homolog Vps32 using transmission electron microscopy (
), lending additional credibility to the proposal that Snf7 filaments form spiral structures that can store potential energy by overgrowing their preferred bending radius. In short, the doublet spiral structure served to minimize this energy penalty by effectively preventing formation of extremely large low-curvature spiral structures. Thus, the detailed observation of the morphological properties of ESCRT-III assemblies is a useful approach to obtain information about the biophysical properties of these assemblies.
Dynamics
HS-AFM allows observations of biological samples with high spatial resolution. In fact, when using recently developed averaging and localization techniques in HS-AFM image processing, it is possible to resolve fine structural details of membrane proteins, approaching the spatial resolution characteristic for electron microscopy (
). However, structural detail on the level of individual molecules can be obtained even without compromising the dynamic information that is obtained in HS-AFM imaging, making HS-AFM indispensable as the foremost biophysical technique in structural dynamics studies of biological samples. This was demonstrated by the first observations of budding yeast Snf7 polymerizing on supported lipid bilayers (
), where lateral rearrangements of Snf7 filaments during the polymerization stage were visible in exquisite detail. Similar dynamic observations were made for the minimal ESCRT-III system comprising Snf7, Vps2 and Vps24 in (
). While all processes that are described in this review (polymerization, depolymerization and scission in particular) are dynamic in nature, a major advantage emerged from the development of HS-AFM supports that allowed observation of three-dimensional rearrangements of ESCRT-III assemblies, made possible by the development of deformable substrates for SLBs (
). This revealed the dynamic nature of the inner turns of Snf7 spirals that underwent an out-of-plane transition, presumably due to the potential energy accumulated by Snf7 polymerization (Figure 3h). These developments opened new avenues that can be utilized to explore the out-of-plane deformations expected to be observed in ESCRT-mediated membrane budding. Another improvement over previous HS-AFM experiments on planar, mica-SLBs is the development of deformable, rupture-resistant liposomes that remodel when incubated with ESCRT proteins. While the structural detail of ESCRT-III assemblies is not visible using this approach, it allowed imaging the bulk rearrangements of the liposomes as a function of ESCRT addition to the sample (Figure 3i). These developments are crucial to expanding HS-AFM measurements to more complex biological processes that result in large-scale three-dimensional deformations, which were previously hindered by the constraints of a rigid SLB.
Depolymerization
Depolymerization of ESCRT-III assemblies and their remodeling by the AAA-ATPase Vps4 is thought to be crucial to ESCRT-mediated membrane remodeling processes (
). Depending on the choice of the sample preparation protocol used, HS-AFM provided novel information about depolymerization and remodeling processes. Injecting ESCRT-III proteins into the imaging solution of the HS-AFM chamber, obtaining planar spiral structures on rigid SLBs, washing out soluble ESCRT-III components from the chamber, and finally adding Vps4 and ATP to the imaging solution, the depolymerization of ESCRT-III assemblies containing budding yeast Snf7, Vps2 and Vps24 was observed (Figure 3j)(
). In contrast, without washing out the ESCRT components from the imaging chamber, a similar depolymerization of assemblies was observed, but the membrane areas that were freed in this process were rapidly repopulated by nascent spirals (Figure 3k). While this experiment allowed direct visualization of Vps4-mediated depolymerization at the molecular scale and a turnover of the ESCRT-III components on the membrane, the rigid support prevented any out-of-plane deformation. A different approach was made possible by protocols that allow formation of tubular ESCRT-III structures, which were amenable to investigations into membrane scission (
). CHMP2AΔC-CHMP3 helical tubes (Vps2-Vps24 homologs, respectively) were incubated in solution by removing the maltose binding protein fused to CHMP2AΔC using a protease. Once assembled, the tubes were added to a HS-AFM rigid support covered with a SLB and imaged (
). Vps4 and ATP were then added to the solution, constriction of the tube diameter was visualized as a decrease in tube height, and disassembly of the tubes was observed (Figure 3l). These observations were then interpreted to propose a model by which CHMP2A-CHMP3 helices disassembled on the inside of cytokinetic necks, leading to constriction of the surrounding membrane tube and scission. This study, like most HS-AFM studies on the ESCRT system, did not consider the full set of ESCRT-III components, and thus only represented a partially reconstituted ESCRT-III system. In general, while HS-AFM-based experiments allow deciphering the mechanistic contributions of individual ESCRT-III components and their biophysical properties, they often cannot provide insights into the complex spatiotemporal organization of the ESCRT system as a whole. In this respect, recent advances by fluorescence microscopy-based approaches notably yielded a detailed description of the ESCRT-III depolymerization sequence (
). Further efforts in the use of HS-AFM will have to be invested to observe a more complete polymerization and depolymerization sequence of the ESCRT system at the molecular scale, a major obstacle to overcome being the inability of HS-AFM to distinguish between molecules of very similar sizes and shapes.
Outlook and perspective
The diversity of membrane protein-mediated processes that has been explored by HS-AFM is a testament to the power of this method to observe dynamic processes at the molecular level. It has so far proven most useful in relatively simple reconstituted systems; in the case of the ESCRT system, it provides information about the mechanical properties of filaments and supramolecular structures formed by only a handful of protein components. It does not replace other types of microscopy, but should rather serve as a complementary technique that allows a more complete understanding of the molecular mechanisms of processes taking place at membrane surfaces.
An inherent constraint in HS-AFM is the fact that it is incapable of differentiating between individual proteins of similar size and shape in multicomponent assemblies. This is particularly important in studies of reconstituted ESCRT systems, as the mechanistic contributions of different ESCRT components might be correlated to their localization within the multicomponent ESCRT assembly (
). In fluorescence microscopy studies, colocalization assays of individual protein species allow observations of dynamics and localization of individual components in complex protein systems, although not at molecular resolution. Attempts to label ESCRT-III components for HS-AFM studies and distinguish between them based on the difference in volume have not yielded the desired results so far, and using antibodies to specific ESCRT components to decorate assemblies have similarly failed because of the high mobility of antibodies in the imaging area. However, using a very small admixture of an ESCRT-III component conjugated to another non-interacting protein, with most of the sample remaining non-conjugated, could yield stable samples with only a small number of higher-volume (and thus more prominent) conjugated monomers incorporated into the assemblies. Analysis of such samples would rely on the ability of HS-AFM to acquire large amounts of data in the same imaging area without damaging the sample. The position of each prominent monomer in a spiral filament is a function of its angular displacement and radial distance from the spiral center, and a large number of observations would permit statistical reconstruction of filament composition. If filaments of assemblies formed in a multicomponent system were indeed compositionally different, such an approach would result in distinct bands occupied by individual ESCRT-III proteins in the polar occupancy plots. However, this approach relies on ESCRT-III assemblies being homogenous with very little deviation of the spiral path from an Archimedean spiral; this is unfortunately not the case for Snf7 spirals in crowded conditions, as these assemblies show significant morphological differences, including polygonality. This could be overcome by HS-AFM observations at lower membrane-bound protein densities, as observed for human CHMP4B (
). In general, acquiring dynamic, structural and compositional information at the same time would increase the value of HS-AFM as a biophysical method.
The technical developments in HS-AFM support design have opened new avenues for observations of reconstituted systems, not just of ESCRT, but also membrane proteins in general. The nanopatterned support coated with supported lipid bilayers can theoretically be used for studies of any membrane-curvature sensing peripheral or transmembrane protein or protein system. By varying the conditions used to prepare the HS-AFM calibration grid that serves as the nanopatterned sample support, the geometry of the substrate could be fine-tuned (curvature, periodicity) to match the specific needs of the experimenter. Similarly, HS-AFM calibration grids of other geometries (linear arrays of interchanging troughs and valleys, hexagonal arrays of pits of different sizes etc.) are commercially available and represent a valuable resource of nanopatterned supports that can be explored in a similar fashion. While the exact parameters of preparation steps needed to achieve a desired topography from an intact calibration grid are still poorly understood, adoption of this method by more experimental groups could lead to a better understanding of this issue, ultimately developing a process that would yield inexpensive and reproducible nanopatterned supports for HS-AFM experiments. This would surely prove invaluable to reconstitution studies of membrane curvature sensing and/or inducing proteins, such as the thylakoid-membrane remodeling protein Vipp1 (
) allowed the observation of membrane deformation by the ESCRT system on the nanoscale, and as such should become an appealing support for HS-AFM studies of similar processes. These supports are relatively easy and inexpensive to prepare, requiring readily available chemicals (PDMS) and equipment that is often already present in labs, such as a plasma etcher. Though the preparation of these PDMS supports is relatively simple, there was, to the best of the authors’ knowledge, no attempt to use them for reconstitution studies of membrane proteins prior to the findings reported in (
). While the PDMS layer could ultimately still be too rigid to be compressed by membrane protein assemblies, thus allowing only very faint deformations to be observed, the authors found that the PDMS-SLB support allowed experiments and insights into biological processes that were previously unobservable on rigid supports, such as the dynamic out-of-plane transitions of the inner turns of Snf7 spirals. Use of other polymers, or improvements in the protocol used to form substrates, could prove beneficial to developing novel HS-AFM supports that might allow observations of large-scale, protein-mediated deformations of the underlying lipid bilayer. The stiffness of elastomer substrates such as PDMS can be tuned from the MPa to the kPa range (
). The effects of such variations on ESCRT-III polymerization remain unexplored. Nevertheless, the PDMS-SLBs constitute a major advancement in HS-AFM studies of biological samples where deformation in the vertical direction was previously hindered by the rigidity of the traditional mica supports.
REFERENCES
Rothman JH
Stevens TH
Protein sorting in yeast: Mutants defective in vacuole biogenesis mislocalize vacuolar proteins into the late secretory pathway.
Listeriolysin O Membrane Damaging Activity Involves Arc Formation and Lineaction -- Implication for Listeria monocytogenes Escape from Phagocytic Vacuole.
Association of the Endosomal Sorting Complex ESCRT-II with the Vps20 Subunit of ESCRT-III Generates a Curvature-sensitive Complex Capable of Nucleating ESCRT-III Filaments.
Journal of Biological Chemistry.2011; 286: 34262-34270
00217-X&id=gr1.jpg){kind=link}
00217-X&id=gr2.jpg){kind=link}
00217-X&id=gr3.jpg){kind=link}