GTP-stimulated membrane fission by the N-BAR protein AMPH-1

SUMMARY

Membrane-enclosed transport carriers sort biological molecules between stations in the cell in a dynamic process that is fundamental to the physiology of eukaryotic organisms. While much is known about the formation and release of carriers from specific intracellular membranes, the mechanism of carrier formation from the recycling endosome, a compartment central to cellular signaling, remains to be resolved. In C. elegans, formation of transport carriers from the recycling endosome requires the dynamin-like, Eps15-homology domain (EHD) protein, RME-1, functioning with the Bin/Amphiphysin/Rvs (N-BAR) domain protein, AMPH-1. Here we show, using a free-solution single-particle technique known as burst analysis spectroscopy (BAS), that AMPH-1 alone creates small, tubular-vesicular products from large, unilamellar vesicles by membrane fission. Membrane fission requires the amphipathic H0 helix of AMPH-1 and is slowed in the presence of RME-1. Unexpectedly, AMPH-1-induced membrane fission is stimulated in the presence of GTP. Furthermore, the GTP-stimulated membrane fission activity seen for AMPH-1 is recapitulated by the heterodimeric N-BAR amphiphysin protein from yeast, Rvs161/167p, strongly suggesting that GTP-stimulated membrane fission is a general property of this important class of N-BAR proteins.

Graphical Abstract

GTP-stimulated membrane fission by the N-BAR protein AMPH-1

The amphiphysin protein, AMPH-1, and its binding partner, the EHD protein, RME-1, are required for formation of membrane-bound carriers from recycling endosomes in C. elegans. Here, we show that AMPH-1 mediates membrane fission and is activated in the presence of GTP and inhibited in the presence of its binding partner, RME-1. The population-resolved kinetics of fission are quantified using Burst Analysis Spectroscopy (BAS), in which fluorescence bursts (red peaks) are proportional to the size and abundance of fluorescently-labelled liposomes. A BAS heat map shows the precursor-product relationship between the starting liposomes and fission products.

INTRODUCTION

Vital cellular processes, from cell growth to synaptic transmission, rely on membrane-bound carriers to transport molecular cargo to and from specific intracellular compartments throughout the cell. Compartment-specific proteins are required to release transport carriers from the intracellular compartment by a process known as membrane fission [1,2]. The role of these proteins in intracellular membrane fission has been informed by the mechanochemical mechanism of dynamin-1 at the synapse [3]. However, research in the past two decades has revealed additional membrane fission mechanisms, which must also be considered when determining how tubular vesicular carriers are formed from intracellular compartments ([4–6] and references therein).

The movement of material from early endosomal compartments back to the plasma membrane, a process known as recycling, involves a regulated membrane fission process that is not well understood. While recycling has been examined in several model systems [7], studies of the basolateral recycling endosome of C. elegans have yielded several key observations. In this model system, formation of tubular-vesicular carriers from the recycling endosome returns receptors back to the plasma membrane and requires the activities of at least two interacting proteins, receptor mediated endocytosis-1 (RME-1) and amphiphysin 1 (AMPH-1) [8]. The discovery of RME-1 [7], a dynamin-like, Eps15 Homology Domain (EHD) ortholog [9], initially suggested a dynamin-like, mechanochemical mechanism [10,11]. RME-1, the sole EHD protein ortholog in C. elegans, is a nucleoside triphosphate hydrolase [12] that binds and functions with AMPH-1 (the only amphiphysin ortholog in that organism) in a functional partnership that is also observed at the recycling endosome in mammals [8]. In contrast to the dynamin family, studies of EHD proteins from lamprey synapses and mammalian cells suggest that not all EHD proteins induce membrane fission; rather, some stabilize membrane tubules and others inhibit membrane fission [13,14], leaving the function of RME-1 unresolved. Furthermore, the well-characterized role of amphiphysin proteins in vesicle formation during endocytosis [13] may have obscured a role in carrier formation at the recycling endosome [8].

Recently, models for membrane fission mechanisms have expanded to include protein crowding and amphipathic helix-insertion, raising the possibility that AMPH-1 could play a more direct role in the formation of tubular-vesicular carriers from the recycling endosome [6,15]. Amphiphysins are members of the Bin/Amphiphysin/Rvs (N-BAR) domain family of membrane-curvature sensing and/or inducing proteins [16]. In general, BAR domain proteins form an arc-shaped dimer with a positively charged, concave face. The positive charges and arc shape facilitate interactions with highly curved, negatively charged phospholipid bilayers. Membrane curvature can also be induced through a wedge-like mechanism, such as insertion of the N-terminal amphipathic H0 helices into one leaflet of the lipid bilayer [4,13,17].

In order to test the roles of RME-1 and AMPH-1 in membrane fission, we reconstituted a fission reaction in vitro, using Burst Analysis Spectroscopy (BAS; [18,19]). This single particle fluorescence-based method was successfully used to study the aggressive membrane fission activity of the ENTH domain of epsin, a well-characterized membrane fission protein that uses an amphipathic helix-insertion mechanism for membrane fission [19–21]. Using the same methodology, we now identify AMPH-1 as a membrane fission protein. Surprisingly, AMPH-1 membrane fission activity is stimulated in the presence of GTP, an unexpected result, because AMPH-1 has no known nucleotide binding domain. Furthermore, the presence of RME-1 slows AMPH-1 mediated membrane fission, in the absence or presence of nucleotides, suggesting a regulatory role for this EHD ortholog. Because membrane fission activity requires the AMPH-1 N-terminal, amphipathic H0 helix, our results are consistent with a model in which tubular-vesicular products are formed at the basolateral recycling endosome, at least in part, through a GTP-stimulated, amphipathic helix-insertion mechanism.

EXPERIMENTAL METHODS

Protein Expression and Purification

Both wild type C. elegans Amphiphysin 1 (AMPH-1) and variants (AMPH-1 L9Q, AMPH-1-eGFP) were produced as previously described [8], with minor modifications (see Online Methods for details). Receptor mediated endocytosis 1 (RME-1) from C. elegans was also produced essentially as previously described [22] with modifications (see Online Methods for details). The coding sequences of Reduced Viability upon Starvation 161p and 167p (Rvs167p and Rvs161p) from S. cerevisiae were cloned from yeast genomic DNA into a single pET-Duet 1 vector. Rvs161 was fused to a cleavable, N-terminal polyHis-GST-TEV affinity tag and Rvs167 was fused to a cleavable, N-terminal polyHis-MBP-TEV affinity tag. The Rvs161/167p heterodimer was expressed in E. coli BL21 (DE3) T5R at 18 °C. See Online Methods for details of the expression conditions used and purification strategy. The ENTH domain of epsin was purified as previously described [19]. For the “mock purification”, the plasmid pGEX-N-His-GST was used to express GST alone, which was then carried through the purification protocol described above for AMPH-1. In all cases, purified proteins were stored at −80°C and were removed from cryo-storage and quick-thawed immediately prior to use. In the case of the AMPH-1 proteins, an additional centrifugation step (200,000 × g for 30 min at 4 °C) was applied immediately after thawing in order to remove a small amount of aggregated protein that formed upon cryo-storage. The concentrations listed herein for AMPH-1 (and variants), RME-1 and Rvs161/167p are calculated for the protein dimers.

Liposome Preparation

Liposomes were prepared as previously described with minor modifications [8,19]. See Online Methods for details.

Membrane fission assays using BAS and FCS

Fluorescence burst measurements were taken with a custom-built, multi-channel BAS microscope [18,23]. Membrane fission assays were performed as previously described [19], with some modifications. The measurement variability of each BAS data set was examined using a Monte Carlo-based uncertainty estimator. See Online Methods for additional details.

Membrane binding assays using MC-BAS

Binding of fluorescent AMPH-1 and RME-1 to liposomes was examined using a multi-color variant of BAS (MC-BAS; [23]). For these experiments, 200 nm PS liposomes labeled with Vybrant DiD (0.03%) were prepared by manifold extrusion as described [8,19]. Fluorescent liposomes (20–30 pM) were mixed with eGFP fusion variants of AMPH-1 and RME-1, either in the presence or absence of nucleotides, in standard fission assay buffer. When nucleotide was present, proteins were pre-incubated with the nucleotide for 1 min prior to mixing with liposomes. Basic sample handling, temperature and BAS data collection parameters were the same as descried for fission assays [19], except that dual excitation with co-aligned 488 and 642 nm lasers (each 50 μW at the sample) was employed.

Electron Microscopy

Fission reactions examined by electron microscopy (EM) were carried out using the same sample handling protocol and lipid:protein ratios employed for BAS. However, to improve particle occupancy on the EM grids, the total protein and lipid concentrations were increased: 3 μM AMPH-1 or 3 μM RME-1 and 60 pM, 200 nm PS liposomes. Samples were incubated either with or without 0.5 mM nucleotide for different amounts of time at 23 °C prior to grid application and staining. All samples were applied to formvar/carbon coated 400 mesh copper grids (Electron Microscopy Services) and stained with 2% uranyl acetate. Grids were imaged with a JEOL 1200EX transmission electron microscope at 100 kV. For the analysis of membrane tubule morphology (Figure 5D), grids were imaged with a FEI Tecnai F20 TEM and image acquisition was carried out using K2 camera at a nominal magnification of 7.8 kX, yielding 4.8Å/pixel. A total of 10 and 12 micrographs were collected for AMPH-1 alone and AMPH-1 supplement with GTP, respectively. For automated analysis of membrane tubule diameter heterogeneity (Figure S9), grids were imaged with a FEI Tecnai F20 TEM and image acquisition was carried out using K2 camera at a nominal magnification of 19 kX, yielding 1.87Å/pixel. For AMPH-1 supplemented with GTP, a total of 40 micrographs were collected, while for AMPH-1 alone, 42 micrographs were acquired.

Membrane tubule morphologies and diameters were analyzed with two different approaches. In the first approach, membrane tubules were manually picked using RELION-3.1.0 manual picker in helical mode to select isolated/non-overlapping tubules. The selected tubules were then extracted into segments yielding 194 and 297 segments for AMPH-1 with and without GTP, respectively. The diameters of each segment were then manually measured across the tubule segments using the relion_display function while adhering to the following two criteria: (1) measurements were carried out at the most visible and thickest part at the middle of the segments and (2) if an expanded “bleb” was observed on the tubule, the measurement was carried out at the bleb. The measurements obtained from relion_display in pixels were then converted to angstroms and used to construct the distribution plot in Figure 5D.

In the second approach, micrographs were processed using cryoSPARC v3.3.2 and particles were picked using the blob_picker tool along the tubular long axis. Initial 2D classification was carried out by classifying all particles into 50 classes. Good quality 2D classes were then selected for the measurement of tubular diameters. From the selected 2D classes, AMPH-1 with and without GTP yielded 7,567 particles and 6,817 particles, respectively. For the AMPH-1 sample supplemented with GTP, the total number of particles was then randomly split into 3 subsets. Each subset was then used to generate a 2D class-average image. The diameter of the tubular liposome was measured with the tape measure function in UCSF ChimeraX v1.3 from each class-average 2D image. For the AMPH-1 sample alone, a total of 6,817 particles were 2D-classified to separate smaller and larger diameter tubular structures, yielding 3,508 and 3,309 particles, respectively after selection. The particles were then subjected to the same procedures as described above for the measurement of tubular diameter.

MANT-GTP binding assay

(2’-(or-3’)-O-(N-methylanthraniloyl) guanosine 5’-triphosphate (MANT-GTP) was obtained from Thermo-Fisher as a frozen 5 mM stock solution. Immediately prior to use, samples of AMPH-1 were loaded on a Sephadex S200 gel filtration column equilibrated in sample buffer (25 mM Tris, pH 7.5, 100 mM KCl, 5 mM MgCl₂). The AMPH-1 dimer peak was then collcted and concentrated. Both AMPH-1 and MANT-GTP were separately diluted into sample buffer at twice the final target concentration, mixed 1:1 and then incubated for 5 min at 23 °C prior to measurement. In all cases, the final MANT-GTP concentration was 2 μM and the final AMPH-1 concentration was varied from 3–100 μM (dimer). For competition experiments, the final GTP concentration was 250 μM. Emission spectra were acquired using a photon-counting, T-format steady state fluorometer (PTI) using excitation at 356 nm with a ± 2 nm bandpass. Emission was collected from 380–500 nm using a ± 2 nm bandpass. All measurements were taken at 23 °C using a thermally jacketed cuvette holder connected to a high-precision water bath.

ATPase and GTPase assays

Prior to GTPase measurements, trace amounts of co-purifying ATPase activity was eliminated from the AMPH-1 samples: a 200 μl slurry of ATP agarose (adenine C-8 linked to solid matrix via 9-atom spacer; Milipore-Sigma) equilibrated in buffer (25 mM HEPES, pH 7.5, 150 mM KCl, 10 mM MgCl₂, 1 mM DTT) was incubated with 1 ml of 20 μM AMPH-1 (dimer) in the same buffer with gentle agitation for 1 hr at 4 °C. AMPH-1 was recovered following removal of the ATP-agarose resin by centrifugation. ATP and GTP turnover were measured in real-time using a colorimetric free phosphate assay (EnzCheck; Thermo-Fisher). See the Online Methods for additional details.

RESULTS

Liposomes are clustered by RME-1 and vesiculated by AMPH-1

In order to test the functions of RME-1 and AMPH-1 in formation of tubular-vesicular products, we used BAS to examine changes in liposome size after incubation in the absence, or presence of AMPH-1 and RME-1. BAS determines the size and concentration distributions of complex nanoparticle samples, like mixtures of fluorescent liposomes, based on their fluorescence burst distribution (Figure S1). Importantly, BAS can quantify particle size distributions in real time over a 100-fold size range using a single, free solution sample [18]. If a protein stimulates membrane fission, larger bursts, which represent starting liposomes, are expected to disappear and a population of more abundant, smaller bursts, representing fission products, are expected to appear. Starting liposomes were made from rehydrated PS mixed with a fluorescent lipid dye (0.3% TopFluor PE), extruded to an average diameter of 200 nm, and diluted to concentrations appropriate for single particle measurements (~ 30–50 pM; Figure 1A).

Figure 1. AMPH-1, but not RME-1, induces liposome membrane fission.

Fluorescence bursts from 30–40 pM, 200 nm diameter PS liposomes labeled with 0.3% TopFluor PE and incubated for 30 min at 23 °C in the absence (A) or presence (B-C) of protein and nucleotide and observed with burst analysis spectroscopy (BAS; [18]). (B) Fewer burst events with much higher intensity are observed upon addition of 1 μM RME-1 and 0.5 mM ATP. (C) A large increase in the abundance of bursts with much lower intensity are observed upon addition of 1 μM AMPH-1 and 0.5 mM GTP. (D) BAS population distribution analysis of starting liposomes (blue); liposomes incubated in the presence of 1 μM RME-1 and 0.5 mM ATP (green); and 1 μM AMPH-1 with 0.5 mM GTP (purple). A 5-fold dilution was required to adequately characterize the higher concentration, lower-intensity objects produced in the presence of AMPH-1 (inset). (E) BAS population distributions from the same conditions as (D) are plotted as heat maps, including liposome and nucleotide controls; each BAS plot is a combination of three, independent experimental replicates. Liposome morphology is examined by negative-stain electron microscopy in the absence of protein (F); with 1 μM RME-1 (G); with 1 μM RME-1 and 0.5 mM ATP (H); with 1 μM AMPH-1 (I); with 1 μM AMPH-1 and 0.5 mM GTP (J); and 1 μM AMPH-1 in the absence of liposomes (K). Sample preparation for electron microscopy was conducted at the same lipid-to-protein ratios used for BAS, but at three-fold higher total protein and lipid concentrations in order to yield optimal sample density on the grids. Scale bar indicates 0.5 microns for (G) and (H) and 1 μm for (F) and (I-K).

In the presence of 1 μM RME-1, the burst distribution of the starting 200 nm liposomes changed dramatically, with a decrease in burst frequency and large increase in mean burst intensity (Figure 1B). This behavior is not consistent with the expectations of protein-induced membrane fission. By contrast, the disappearance of the starting 200 nm liposomes observed in the presence of 1 μM AMPH-1 and GTP was accompanied by a large increase in the total burst frequency and decrease in mean burst amplitude (Figure 1C). Application of BAS revealed that the total population of liposome objects (Figure 1D, blue) was dramatically depleted in the presence of RME-1 (Figure 1D, green). However, in the presence of AMPH-1 a large increase in the number of lower-intensity objects was observed (Figure 1D, purple), behavior that is consistent with protein-stimulated membrane fission.

When the BAS results are presented as heat maps (Figure 1E), the impact of RME-1 and AMPH-1 on the starting liposome population is striking. Instead of an increase in number of small bursts, expected for small liposome fission products, the presence of RME-1 resulted in a shift of the initial liposome population to a lower concentration of much larger (right side of the map, greater intensity) membrane products in the absence (−) and presence of the nucleotides, ATP and GTP. This result is consistent with RME-1 inducing large-scale clustering, or clumping, of the starting 200 nm liposomes, without significant formation of a population of smaller liposome objects that would be consistent with fission. Similar results were obtained using liposomes of more complex lipid compositions, including (1) Folch, plus 5% PI(4,5)P₂; and (2) Folch, plus PI(4)P and PI(4,5)P₂ (Figure S2). Additionally, when the RME-1 concentration was reduced to 25 nM to lower the extent of liposome clustering from the protein alone, the addition of ATP was observed to increase this clustering effect (Figure S3).

In the presence of AMPH-1, the initial liposome population shows a significant shift to smaller (lower intensity) products in the presence of GTP, in comparison to a minor shift in the absence of nucleotide, or in the presence of ATP (Figure 1E). Morphological analysis by negative stain electron microscopy revealed that a uniform field of 200 nm starting PS liposomes (Figure 1F) appears as clumps of liposomes and tubules in the presence of RME-1 (Figure 1G), as clustered tubules in the presence of RME-1 with ATP (Figure 1H), as narrow tubules in the presence of AMPH-1 (Figure 1I), and as narrow tubules and smaller tubular-vesicular products, in the presence of AMPH-1 with GTP (Figure 1J). Importantly, in the absence of liposomes, the morphology of AMPH-1 is particulate, and much smaller than the membrane products (Figure 1K). In total, these observations suggest that AMPH-1 mediates a membrane fission reaction that is modulated by GTP.

AMPH-1 induces membrane fission and the rate of fission is stimulated by GTP

To further define the function of AMPH-1, we used BAS to measure the rate that the large, starting liposomes disappeared and smaller membrane products appeared. While the formation of narrow tubules by AMPH-1 has been described previously [8], and small, vesicular-tubular morphologies have been observed in liposome preparations with high concentrations of amphiphysin from D. melanogaster [24], a fission function for amphiphysin has also been disputed, due to a restricting effect of the BAR domains [21]. Moreover, the stimulation of small membrane product formation by GTP was unexpected for this family of proteins, which are not predicted to bind nucleotides. To rule out an artifactual, co-purifying, GTP-stimulated fission-active contaminant derived from the E. coli expression host, we carried out a mock protein purification, in which the same procedure that was employed to prepare AMPH-1 was carried out in the absence of AMPH-1 expression. When this mock sample was tested, no membrane fission activity was observed (Figure S4). To address the possibility that PS liposomes have an artificial propensity for GTP-stimulated AMPH-1 induced membrane fission, we repeated the fission assay and obtained similar results using liposomes made with a lipid composition found enriched in late endosomes, including PC, PS, PA, PE and PI(4,5)P₂ (Figure S5). Furthermore, dose-dependent formation of small membrane products with increasing AMPH-1 concentrations supports attribution of membrane-fission activity to purified AMPH-1 (Figure S6). A slower and nucleotide-independent fission activity is also detectable, as the protein:lipid ratio increases (Figure S6).

The small size and increased abundance of the fission products suggest a specific precursor-product relationship between these objects and the larger starting liposomes. We therefore examined the rate of 200 nm liposome conversion to smaller membrane products in the absence and presence of GTP (Figure 2A–B). For these experiments, a protein:lipid ratio was chosen that permitted the slower, nucleotide-independent fission activity to also be quantified. By coarse-binning the observed object sizes into small (S), medium (M), or large (L; Figure 2C–E) ranges, we confirmed the expected precursor-product relationship between the three sizes. Specifically, the time-dependent decrease in the concentration of the large and medium sized liposomes (Figure 2D and E) is accompanied by an increase in the concentration of much smaller fission products (Figure 2C). Addition of GTP accelerates the rate of membrane fission at this protein:lipid ratio by at least three-fold, so that after 60 min at 25 °C, the starting liposomes are mostly converted to small membrane products. These limit fission products possess approximately 10-fold less membrane than the starting 200 nm liposomes and are produced, in the presence of GTP, at a rate of about one fission event every 20 sec, per large liposome (see Supplemental Information).

Figure 2. GTP stimulates AMPH-1 induced fission to small, tubular-vesicular products.

Population-resolved kinetics of membrane fission by AMPH-1 (1 μM) in the absence (A), or presence (B) of GTP. Data from (A) and (B) are coarse binned by size and plotted as a function of time: small (C) medium (D), or large (E). Error bars reflect a Monte Carlo estimate of BAS uncertainty based on the measured photon counting histogram noise (see Methods). Solid lines in C, D and E show fits to a single exponential rate law, except for the medium bin in the absence of GTP (D). In this case, the data was fit to a third-order polynomial as a visual guide. The size of the fission products after a 1 hr incubation at 25°C with 1 μM AMPH-1 + GTP was compared to liposomes extruded to the diameters indicated, using both BAS (F) and fluorescence correlation spectroscopy (FCS; G). The morphology of fission products was observed by negative stained electron microscopy (H–J). Scale bar indicates 1 micron for (H), 0.5 micron for (I) and 0.2 μm for (J). Liposomes were prepared as in Figure 1. Each BAS plot is a combination of three, independent experimental replicates.

GTP also changes how fission proceeds. In the presence of GTP, the decrease in both large and medium sized objects, as well as the appearance of the smallest fission products, all display approximately single exponential kinetic behavior (Figure 2C–E). By contrast, in the absence of GTP, the medium sized population displays distinctly non-exponential behavior (Figure 2D), even though the largest and smallest populations continue to display slower, single exponential kinetics (Figure 2C and E). Analysis of the smallest end products by (1) BAS in comparison with 30, 50, 100 and 200 nm diameter liposome standards (Figure 2F), (2) fluorescence correlation spectroscopy (FCS; Figure 2G), and (3) morphological analysis by negative stain EM (Figure 2H–J), indicates that the fission products produced by AMPH-1 are tubular-vesicular particles that have membrane content similar to 20–40 nm diameter vesicles.

A slow intrinsic GTPase activity of AMPH-1 is not stimulated by liposomes

The surprising observation of nucleotide-stimulated fission led us to examine whether AMPH-1 binds GTP directly. For this experiment, we employed a well-established fluorescent analog of GTP (MANT-GTP). When bound in the nucleotide pocket of GTP-binding proteins, MANT-GTP typically displays enhanced fluorescence (i.e., an elevated quantum yield; [25–28]). Strikingly, the addition of AMPH-1 to MANT-GTP results in a robust fluorescence enhancement that is fully reversed in the presence of excess GTP (Figure 3A). Titration of a fixed amount of MANT-GTP (2 μM) with increasing amounts of AMPH-1 yields a hyperbolic dose-response curve that is consistent with a simple ligand binding event and an apparent dissociation constant of ~ 75 μM (Figure 3B).

Figure 3. AMPH-1 binds and slowly hydrolyzes GTP.

(A) GTP binding by AMPH-1 observed with the fluorescent GTP analog MANT-GTP. The fluorescence emission spectra of MANT-GTP alone (2 μM; green), plus AMPH-1 (40 μM dimer; black) or plus AMPH −1 (40 μM) and GTP (250 μM; blue) are shown. The increase in MANT-GTP fluorescence in the presence of AMPH-1, which is competitively reversed by excess GTP, is a strong indicator of nucleotide binding by the AMPH-1 dimer. (B) Titration of 2 μM MANT-GTP with AMPH-1. The fluorescence enhancement of the MANT-GTP as a function of protein concentration is shown. The data was fit to simple hyperbolic binding isotherm, yielding an apparent binding constant (K_1/2) of 75 μM. Error bars display the standard deviation of n = 3 technical replicates. (C) The initial rate (v₀) of GTP hydrolysis by AMPH-1 (2.5 μM dimer) as a function of GTP concentration. Eadie-Hofstee plot of the steady state turnover data is shown in the inset. The observed K_m and k_cat are 200 μM and 0.049 min⁻¹, respectively. (D) Steady state kinetics of GTP turnover by AMPH-1 examined in the presence of PS liposomes. Samples were prepared in the same manner as panel (C) but supplemented with 50 μg/ml PS lipids as 200 nm extruded unilamellar liposomes prior to measurement. The observed K_m and k_cat are 220 μM and 0.050 min⁻¹, respectively. (E) The non-hydrolyzable GTP analog GMPPNP competitively inhibits GTP turnover by AMPH-1. Assay conditions were identical to panel (B) and samples had either no addition (0 μM) or added GMPPNP (100 μM, 200 μM). Data is displayed as double-reciprocal (Lineweaver-Burke) plots. (F) GMPPNP also competitively inhibits GTP hydrolysis by AMPH-1 in the presence of PS liposomes.

We next examined whether AMPH-1 possesses detectable GTPase activity. We first compared the GTP hydrolysis activity of purified AMPH-1 before and after an additional gel filtration step. While no substantial improvement in purity (based on SDS-PAGE) was apparent following gel filtration, these AMPH-1 samples did display decreases in both GTP turnover and membrane fission activity. Surprisingly, however, gel filtration also caused AMPH-1 to become more resistant to protease digestion and less capable of binding to PS liposomes (Figure S7 F–G), even though MANT-GTP binding was unaffected (Figure S7E). Overall, these observations suggest that gel filtration treatment perturbs the structure of the AMPH-1 dimer without altering its ability to bind GTP. As an alternative, we treated AMPH-1 with ATP-agarose to remove cross-reactivity from any co-purifying ATPase contaminants. This step eliminates a minor, ~ 70 kDa contaminant (most likely E. coli DnaK; [29,30]) and results in an ~ 3x decrease in observed rate of ATP turnover (Figure S8A–B). However, ATP-agarose treatment has no impact on either AMPH-1 protein levels or GTP turnover (Figure S8A and C). We characterized the steady state hydrolysis of GTP in the presence of ATP-agarose-treated AMPH-1, both with and without 200 nm PS liposomes. In both cases, AMPH-1 displays saturation kinetics typical of a GTPase enzyme (Figure 3, C–D). The apparent K_m for AMPH-1 in the absence of membrane is ~ 200 μM, with an observed k_cat of 0.049 min⁻¹. Addition of PS liposomes results in no apparent change in the observed hydrolysis kinetics. Importantly, the non-hydrolyzable GTP analog, GMPPNP, acts as an effective competitive inhibitor, both in the presence and absence of liposomes (Figure 3, E–F). While it remains formally possible this GTPse activity comes from a trace, co-purifying and highly aggressive bacterial enzyme, the similarity of the nucleotide half-response for (1) fission (see Figure 4 below), (2) MANT-GTP binding and (3) GTP hydrolysis (i.e., the K_m) strongly suggest these observations are linked to the same event and originate with AMPH-1.

Figure 4. Membrane fission by AMPH-1 is nucleotide dependent but membrane binding is not.

The dependence of stimulated AMPH-1 fission activity on nucleotide concentration (A), and nucleotide identity (B), was examined by BAS. Liposome particle distributions were examined following a 30 min incubation of 200 nm PS liposomes with 1 μM AMPH-1 in the absence, or presence of (A) increasing GTP concentration, and (B) 0.5 mM GTP, 0.5 mM GDP, 0.5 mM GTPγS, or 0.5 mM GMPPNP. Each BAS plot is a combination of three, independent experimental replicates. (C) Schematic of an experiment for measuring binding of AMPH-1-eGFP (120 nM; blue ovals) to liposomes labelled with the lipophilic dye, Vybrant DiD (0.06%; red circles). (D) Example photon histories from two, paired detection channels of a multi-color BAS microscope (MC-BAS; [23]): eGFP (525 ± 18 nm) and DiD (705 ± 36 nm). (E-G) MC-BAS heat maps of liposomes bound to AMPH-1-eGFP in the absence (E), or presence of 0.5 mM GTP (F), or 0.5 mM GMPPNP (G). The spread of the distributions along the positive diagonals of the plot measures the population size distribution at a given AMPH-1-eGFP:DiD stoichiometry, while the extent of spread along the negative diagonals is proportional to the range of binding stoichiometries. Each MC-BAS plot is a combination of three, independent experimental replicates. The color scale, shown above each MC-BAS plot, displays the normalized fractional population level.

GTP analogs suggest a role for nucleotide state in AMPH-1 activity on liposomes

We next determined whether GTP stimulation of AMPH-1 membrane fission is dose-dependent. Using BAS, we measured the rate of fission as a function of GTP concentration (Figure 4A). At higher GTP concentrations, increased membrane fission is detected as an increase in the disappearance of the starting liposomes and enhanced appearance of small fission products. In order to further examine the role of GTP in fission activation, we compared the impact of GTP to that of slowly hydrolyzable and non-hydrolyzable nucleotide analogs, as well as the hydrolysis product, GDP (Figure 4B). Surprisingly, GDP stimulates fission to almost the same extent as seen in the presence of GTP. Fission was detected, but not activated in the presence of the slowly hydrolyzable analog, GTPγS, and fission appeared to be completely inhibited in the presence of the non-hydrolyzable GTP analog, GMPPNP (Figure 4B).

Inhibition of membrane fission in the presence of GMPPNP may mean that this nucleotide analog interferes with AMPH-1 binding to liposomes. To test this possibility, we examined membrane binding by measuring coincidence of AMPH-1-eGFP and liposomes labelled with a far-red lipophilic dye, Vibrant DiD, using a multi-color version of BAS (MC-BAS; Figure 4C) [23]. In this type of experiment, the burst amplitude of particles that carry two spectrally separable dyes are measured in two different detection channels using a pair of co-aligned excitation lasers (Figure 4D). MC-BAS enables both the size and relative stoichiometric distributions of the double-labeled particles to be determined [23]. AMPH-1-eGFP displays comparable MC-BAS binding distributions on PS liposomes in the absence and presence of either GTP, or GMPPNP (Figure 4 E–G). In all cases, the position and overall shape of the main MC-BAS distribution are very similar. While small differences are apparent at lower contour levels, these low amplitude variations are difficult to resolve from noise and are therefore likely not significant [23]. In total, these observations are not consistent with the notion that inhibition of membrane fission by GMPPNP is caused by a defect in AMPH-1 binding to membranes.

In order to test whether tubulation is affected by nucleotides, we examined the morphology of AMPH-1 bound liposomes, in the absence (Figure 5A) and presence of GTP (Figure 5B), or GMPPNP (Figure 5C). Negative stain electron microscopy revealed that AMPH-1 bound liposomes appeared to be more loosely wrapped by the protein in the absence of nucleotide than in the presence of GTP. Surprisingly, in the presence of GMPPNP, no membrane tubules were observed, despite robust liposome binding by the protein under these conditions (Figure 4G). Quantitative analysis by both manual (Figure 5D) and automated (Figure S9) segment identification confirmed that, in the presence of GTP, AMPH-1 forms a more uniform distribution of regular membrane tubule structures.

Figure 5. Membrane tubule morphology by AMPH-1 is nucleotide dependent.

Micrographs of negatively stained, 1 μm PS liposomes incubated with (A) AMPH-1 alone, (B) AMPH-1 plus 0.5 mM GTP and (C) AMPH-1 plus 0.5 mM GMPPNP. Scale bar represents 1 μm. (D) Quantitation of AMPH-1 membrane tubules in the presence and absence of GTP. Membrane tubules were manually picked using the RELION-3.1.0 manual picker in helical mode to identify isolated/non-overlapping tubules. The selected tubules were extracted into segments yielding 194 and 297 segments for AMPH-1 with and without GTP, respectively. Membrane tubules diameters among the isolated segments were then measured (see Methods) and the frequency of the observed tubule morphologies (shown in insets I – V) were binned by diameter. Scale bar of the insets represents 0.5 μm.

Yeast amphyphysin, Rvs161/167p, also displays GTP-stimulated fission activity

If the GTP-stimulated membrane fission activity we observed with C. elegans AMPH-1 is a general property of N-BAR proteins, then we expect an orthologous N-BAR protein to exhibit similar activity in the membrane fission assay. To test this prediction, we repeated the BAS assay using the yeast amphiphysin, Rvs161/167p. Like C. elegans, yeast only has one amphiphysin protein, Rvs161/167p [31]. In contrast to the worm amphiphysin, Rvs161/167p is a heterodimeric protein. While the Rvs167p domain structure is similar to AMPH-1, Rvs161p is shorter: it is missing most of the middle domain and the C-terminal SH3 domain [32]. Strikingly, we observe a large stimulation of Rvs161/167p fission activity in the presence of GTP (Figure 6). We also observe a modest enhancement of fission in the presence of GDP, but little effect on basal fission activity with other nucleotides.

Figure 6. GTP stimulates membrane fission by the yeast amphyphysin, Rvs161/167p.

BAS heat map of liposomes incubated for 10 min with 200 nM Rvs161/167p in the absence, or presence of 0.5 mM GTP, 0.5 mM ATP, 0.5 mM GMPPNP, or 0.5 mM GDP. Liposomes were prepared as in Fig. 1. Each BAS plot is a combination of three, independent experimental replicates. Abbreviations: Lip, liposomes; liposome sizes: S, small; M, medium; L, large.

Fission by AMPH-1 requires a functional, amphipathic H0 helix

If AMPH-1 induced fission employs the amphipathic H0 helix mechanism for membrane fission [6], then we expect a mutation that changes the amphipathicity of this helix to result in a defect in membrane fission activity either with, or without, disrupting membrane binding. In order to test this prediction, we introduced a mutation, L9Q, into the N-terminal, H0 helix sequence, which changes the hydrophobic leucine to the polar, uncharged glutamine, on the non-polar face of the H0 helix (Figure 7A). As predicted, the L9Q mutation interferes with membrane fission activity by AMPH-1 (Figure 7B). However, the L9Q mutation has only a small impact on membrane binding by AMPH-1. MC-BAS measurements demonstrate that AMPH-1-(L9Q)-eGFP binds to DiD-labeled liposomes similarly to wild type AMPH-1-eGFP, with only a modest shift of the binding distribution to lower protein:lipid ratios (Figure 7C and D). This shift suggests that, at the same total protein concentration, the L9Q mutant does not achieve the same protein density at the membrane surface as wild type AMPH-1. Nonetheless, in contrast to the behavior of H0 helix mutations in other N-BAR proteins [20,21], membrane binding by the L9Q AMPH-1 mutant remains significant.

Figure 7. AMPH-1 fission activity is disrupted by a mutation in the H0 helix.

(A) Helical wheel projection of the H0 helix of AMPH-1. Circled, single-letter amino acid abbreviations are colored by polarity of the side chain functional groups: basic (blue); acidic (red); polar imidazole (cyan); polar amido (pink); polar hydroxyl (purple); small, aliphatic (grey); large, aliphatic (yellow). Red arrow indicates the position of L9. (B) BAS heat map of liposomes (prepared as in Fig. 1) incubated at 23 °C for 1 hr in the absence or presence of 1 μM AMPH-1, or 1 μM AMPH-1 L9Q. (C, D) Examination of the binding of AMPH-1-eGFP (C) and AMPH-1L9Q-eGFP (D) to liposomes (prepared as in Fig. 4) by MC-BAS. Protein concentration in both experiments was 120 nM and each MC-BAS plot is a combination of three, independent experimental replicates. The color scale, shown above each MC-BAS plot, displays the normalized fractional population level. Abbreviations: Lip, liposomes; liposome sizes: S, small; M, medium; L, large.

A role for RME-1 in regulation of membrane fission by AMPH-1

If AMPH-1 is a membrane fission factor, what is the role of its binding partner, RME-1? We examined the effect of incubating RME-1 with AMPH-1 in order-of-addition mixing experiments (Figure 8A) and found in all cases, the presence of RME-1 inhibited the membrane fission activity of AMPH-1. Inhibition by RME-1 is increased, as the RME-1: AMPH-1 stoichiometry is increased (Figure 8B). Interestingly, the propensity of RME-1 to cluster liposomes (Figure 1) is ameliorated by the presence of AMPH-1, regardless of the order of mixing, or the stoichiometry, up to a 1:1 ratio of RME-1:AMPH-1. Similar results were obtained with liposomes made from more complex lipid compositions that included PI(4)P and PI(4,5)P₂ (Figure S2). These results suggest that the role of RME-1 is to regulate the membrane fission activity of AMPH-1.

Figure 8. AMPH-1 fission activity is reduced in the presence of RME-1.

(A) BAS heat maps of liposomes incubated with 0.25 mM GTP + 0.25 mM ATP in the absence or presence of 1 μM AMPH-1 alone, or 1 μM AMPH-1 and 1 μM RME-1. Mixing order is indicated by the arrow: before the arrow, the pre-mixed pair is incubated at 23 °C for 1 min, followed addition of the third component, with which the mixture is incubated for a further 30 min at 23 °C prior to the collection of BAS data. (B) BAS heat maps of liposomes incubated for 30 min at 23 °C in the absence or presence of 1 μM AMPH-1 alone, or pre-mixed with 1 μM RME-1, or decreased concentrations of RME-1, to achieve the stoichiometries indicated. Liposomes were prepared as in Fig. 1. Each MC-BAS plot is a combination of three, independent experimental replicates. Abbreviations: A, AMPH-1; R, RME-1; Lip, liposomes; liposome sizes: S, small; M, medium; L, large.

DISCUSSION

Amphiphysin proteins, with their N-terminal, H0 amphipathic helices and banana-shaped structures, promote tight curvature on membranes and have been observed to cause membrane fission in the absence of dynamin-like proteins [6,24]. Nonetheless, amphiphysin is generally known for its role in endocytosis, where it sculpts the neck of plasma membrane-derived endocytic vesicles into highly curved domains and recruits additional factors, including dynamin [1,17]. Due to this prominent involvement in endocytosis, a role for amphiphysin in endocytic recycling may have been under appreciated, although a recycling defect is observed upon depletion of the mammalian BIN1/Amph2 isoform [8], which, like AMPH-1, lacks the clathrin/AP-2 (CLAP) binding domain that is attributed to the endocytic function of amphyphysin.

Our results suggest a general model, where amphiphysin proteins play a direct role in membrane fission using an active, nucleotide-driven mechanism. Specifically, we demonstrate that C. elegans AMPH-1 and S. cerevisiae Rvs161/167p have membrane fission activity that is stimulated by GTP. We further demonstrate that the amphipathicity of the AMPH-1 H0 helix is necessary for fission, as a mutation in the hydrophobic face of the H0 helix results in a defect in membrane fission, while only modestly affecting membrane binding. Together, these results are consistent with an amphipathic helix-insertion model for membrane fission, as suggested for the ENTH domain of epsin [19–21], the amphipathic helix insertion by Sar1p [33], as well as other proteins with amphipathic alpha helices [6].

The observation that GTP stimulates amphiphysin-mediated fission is unexpected, as neither AMPH-1, nor Rvs161/167p, possess a canonical GTP-binding site, based on sequence predictions. Nonetheless, the AMPH-1 dimer in solution binds GTP and displays a robust, if slow, intrinsic GTPase activity (Figure 3). Intriguingly, GTP hydrolysis is neither stimulated, nor inhibited, by the presence of membranes, and AMPH-1 binds just as robustly to membranes in the absence as in the presence of nucleotides (Figure 4). Instead, nucleotide binding alters the morphology of AMPH-1 bound membrane tubules (Figure 5), suggesting that the nucleotide binding is linked to a conformational change in AMPH-1 on membranes.

While the role of nucleotide binding in stimulating a conformational change in AMPH-1 will require further study, our findings suggest that, rather than regulating membrane binding, GTP acts to change the protein conformation through a molecular switch mechanism, as seen for the small, regulatory GTPases [34]. In the case of AMPH-1, this model predicts that nucleotide state shifts the protein conformation from an inactive, or “off” state, to an active, or “on” state. If the interconversion between “on” and “off” states is reversible, then we would expect, even in the absence of nucleotide, higher concentrations of total protein would increase the abundance of the “on” state, and overall fission activity. As expected, higher concentrations of AMPH-1 result in membrane fission activity that is less sensitive to nucleotide stimulation (Figure S6). Furthermore, GDP also stimulates AMPH-1 fission activity (Figure 4), suggesting that hydrolysis of GTP stabilizes the “on” conformation. Conversely, the non-hydrolyzable GTP analog, GMPPNP, inhibits fission activity, suggesting that this GTP analog stabilizes the “off” conformation of AMPH-1. Because membrane binding appears to play no role in stimulating the conformational switch, it is likely that other factors, like RME-1, act to modulate the nucleotide state and conformation of AMPH-1.

Prevailing mechanistic models for EHD proteins like RME-1 deviate from the canonical, mechanochemical model for the neuronal dynamin-1 [11,35]. Beyond the preference for ATP (not GTP) and very slow hydrolysis rates, EHD proteins (and non-neuronal dynamin-2; [36]) appear to work with N-BAR protein partners on elongated tubules [8,13,14,37]. As reported previously [22], we show that the EHD homolog, RME-1, causes membrane tubulation, but does not exhibit membrane fission activity. Nor does RME-1 stimulate membrane fission with AMPH-1, as has been reported for dynamin-2 and amphiphysin [36]. Instead, our findings strongly support a model in which RME-1 acts to slow membrane fission by AMPH-1. Notably, it has been previously suggested that other EHD homologs behave as regulators, rather than fission factors [38,39]. For example, the lamprey EH domain protein (1-EHD) has been reported to inhibit extended tubule formation, a state which is not able to mediate membrane fission [40] but rather is thought to act as a “ruler” to prevent excessive tubulation. Similarly, previous work indicated that tubules formed with RME-1 and AMPH-1 are significantly shorter than those formed by either protein alone [8].

Although the role for EHD proteins appears disparate [11,35], there is some evidence linking ATP hydrolysis and membrane fission for the closest RME-1 homolog in humans, EHD-1 [38]. Furthermore, rme-1 mutants in C. elegans (and loss of EHD-1 in mammalian cells) block receptor release from enlarged recycling endosomes, suggesting a positive role in carrier formation, and mutations that disrupt the interaction between RME-1 and AMPH-1 are likewise defective in recycling [8,41–43]. Why then does RME-1 reduce AMPH-1 mediated fission, in vitro? One possibility is that RME-1 slows fission in order to produce carriers that are larger and more tubular in shape, such as those that recycle receptors, in vivo [7,8,35,44], rather than the small products formed by AMPH-1 alone, in vitro.

Additional studies are required to determine the mechanism behind regulation of AMPH-1-mediated membrane fission activity. For example, the ATP binding and hydrolysis cycle of RME-1 may regulate the timing of AMPH-1 fission activity and/or GTP turnover. Another enticing idea incorporates the elongated amphiphysin structure with its ability to change lattice shape and tubule diameter, suggesting the possibility that RME-1 binding to AMPH-1 regulates carrier formation by altering protein lattice distribution on tubules, as observed for other N-BARs [45], other EHD proteins [38] and dynamin [11]. A regulatory role for RME-1 could explain the morphological differences observed between long, narrow tubules wrapped by an AMPH-1 lattice in the absence of RME-1, compared with in the wide, shorter tubules formed by an AMPH-1/RME-1 lattice in the presence of ATPγS [8]. In this scenario, until ATP hydrolysis releases RME-1 from the membrane (by analogy to EHDs; [38,46]), RME-1-bound AMPH-1 is unable to shift to the lattice conformation that allows for the formation and fission of narrow tubules.

Future studies focused on the structures of these different lattices and the role of nucleotide binding and hydrolysis, change in conformation, fission, and regulation are expected to shed light on the mechanism for formation of tubular-vesicular carriers from the recycling endosomes, as well as provide a precedent to consider for formation of carriers from other compartments in the cell.