Asymmetric Tethering of Flat and Curved Lipid Membranes by a Golgin

See allHide authors and affiliations

Science  02 May 2008:
Vol. 320, Issue 5876, pp. 670-673
DOI: 10.1126/science.1155821


Golgins, long stringlike proteins, tether cisternae and transport vesicles at the Golgi apparatus. We examined the attachment of golgin GMAP-210 to lipid membranes. GMAP-210 connected highly curved liposomes to flatter ones. This asymmetric tethering relied on motifs that sensed membrane curvature both in the N terminus of GMAP-210 and in ArfGAP1, which controlled the interaction of the C terminus of GMAP-210 with the small guanine nucleotide–binding protein Arf1. Because membrane curvature constantly changes during vesicular trafficking, this mode of tethering suggests a way to maintain the Golgi architecture without compromising membrane flow.

The Golgi apparatus has a stable architecture despite the intense flux of membrane that passes through it (1, 2). Golgins, which probably correspond to molecular strings observed by electron microscopy (3), contribute to this architecture by tethering membranes thanks to their coiled-coil structure, up to 200 nm in length (4). Some golgins link specific transport vesicles to cisternae and may restrict their diffusion (5). But because vesicles bud and fuse within minutes there must be regulatory mechanisms to promote and disrupt these links within the same time scale. We studied human golgin GMAP-210. This protein is located at the cis Golgi, and its overexpression induces the formation of clusters of small vesicles (radius, R ≈ 30 nm) at the expense of the Golgi (6, 7).

The first 38 residues of GMAP-210 form an ALPS (amphipathic lipid-packing sensor) motif, a lipid-binding module that is remarkably sensitive to membrane curvature (8). Soluble and unfolded in the presence of weakly curved liposomes (R > 100 nm), ALPS motifs form an amphipathic α helix at the surface of small liposomes (R < 50 nm) (8, 9). Through this motif, GMAP-210 could trap small transport vesicles. On its C terminus, GMAP-210 is predicted to contain a GRAB (GRIP-related Arf binding) domain that may interact with the small guanine nucleotide–binding protein Arf1 in the guanosine 5′-triphosphate (GTP) state (10). We reasoned that this putative interaction may be stable only on flat membranes because ArfGAP1, a guanosine triphosphatase (GTPase) activating protein for Arf1 at the Golgi, contains two ALPS motifs, making its activity exquisitely dependent on membrane curvature (9, 11, 12). Thus, the presence of ALPS motifs both in GMAP-210 and in ArfGAP1 suggests an asymmetric mode of tethering between flat and curved membranes (Fig. 1A). In vivo, this tethering would be ideally suited to confine vesicles in the vicinity of flat cisternae. Moreover tethering should be disrupted as soon as one of the two membranes lost its identity: that is, when the vesicle becomes flat by membrane fusion or when the cisternae becomes curved by budding. Thus, GMAP-210 could readily recycle such as to be always properly orientated.

Fig. 1.

Membrane attachment of the two ends of GMAP-210. (A) With its N-terminal ALPS motif and its C-terminal GRAB domain, GMAP-210 could connect small vesicles to membranes containing Arf1GTP. The former interaction is stable only on curved membranes (8), whereas the latter should be stable only on flat membranes because ArfGAP1 contains ALPS motifs and is very active on curved membranes. (B) Domain organization of GMAP-210 and scheme of the constructs (24). (C) GMAPC-short or GMAPC-long (0.5 μM) was incubated with liposomes (0.5 mM lipids, R = 115 ± 51 nm), with Arf1 (1 μM), and with GDP or GTP (66 μM) at 1 μM free Mg2+. The liposomes were recovered by flotation and bound proteins were analyzed by SDS–polyacrylamide gel electrophoresis. (D) NBD fluorescence assay. The cuvette contained liposomes (0.2 mM lipids, R = 42 ± 8 nm), NBDGMAPC-long (0.125 μM), and Arf1GDP (0.75 μM). When indicated, GTP (10 μM) was added, and nucleotide exchange was promoted by lowering temporally the concentration of free Mg2+. Thereafter ArfGAP1 or ArfGAP3 (0.1 μM) was added. (E) Effect of membrane curvature. (Top) Same as in (D) with liposomes of different radius. (Bottom) Rate of NBDGMAPC-long dissociation versus liposome radius as determined from two independent experiments. (F) Opposite effect of membrane curvature on the N and C termini of GMAP-210. GMAPC-long was incubated with the liposomes and with Arf1 and GTP as in (C) followed by 3-min incubation with 0.1 μM ArfGAP1. GMAPN was simply mixed with the liposomes.

Interaction between the GRAB domain of GMAP-210 and a soluble form of Arf1GTP has not been observed (10). Because Arf1 interacts with its partners at the surface of lipid membranes, we reassessed its interaction with GMAP-210 in the presence of liposomes. Two truncated forms of GMAP-210 (GMAPC-short contained amino acids 1597 to 1830 and GMAPC-long included amino acids 1597 to 1843) were used (Fig. 1B and table S1). Both included part of the coiled-coil region and the GRAB domain, but the longer construct contained a short predicted amphipathic helix downstream from the GRAB domain. These constructs were incubated with liposomes in the presence of Arf1, and liposome-bound proteins were recovered by flotation. Arf1GTP but not Arf1GDP efficiently recruited GMAPC-long to the liposomes (Fig. 1C). No recruitment by Arf1GTP was observed for GMAPC-short or when GMAPC-long carried the Leu1783 → Ala1783 (L1783A) mutation (Fig. 1C and fig. S1). This mutation eliminates the Golgi localization of the C-terminal region of GMAP-210, and the cognate residue in the GRIP domain is critical for the interaction with Arl1GTP (10). Thus, in the same way as the GRIP domain (13, 14), the GRAB domain may interact simultaneously with the lipid membrane through its amphipathic helix and with Arf1GTP (Fig. 1A). To test this, we attached a membrane-sensitive fluorescent probe (NBD) to this helix via a cysteine mutation (fig. S2, A to C). Guanosine diphosphate (GDP)–to–GTP exchange on Arf1 promoted an increase and a blue shift in the fluorescence of NBDGMAPC-long, suggesting that the amphipathic helix of the GRAB domain contacted the liposome surface (Fig. 1D and fig. S3, A to D). The kinetics of the NBD fluorescence change matched the time course of Arf1 activation, and its amplitude varied with Arf1 concentration in a manner suggesting a stoichiometric interaction between the two proteins (fig. S3B).

Two GTPase activating proteins for Arf1, ArfGAP1, and ArfGAP3 (Gcs1p and Glo3p in yeast) are localized at the cis Golgi (15) and are candidates for promoting GTP hydrolysis in the Arf1GTP-GRAB domain complex. To test this, we took advantage of the fluorescence signal associated with the translocation of NBDGMAPC-long. ArfGAP1 reversed the NBD signal within minutes, whereas ArfGAP3 had almost no effect (Fig. 1D). This observation fits with the idea that ArfGAP1/Gcs1p can act on several Arf/Arl-effector complexes at the Golgi, including complexes with golgins, whereas ArfGAP3/Glo3p is more specific to the Arf1-coatomer complex (1618). Because the activity of ArfGAP1 is strongly dependent on membrane curvature owing to its ALPS motifs (9, 12), we asked whether ArfGAP1 retained this feature when it promoted GTP hydrolysis in the Arf1GTP-GMAPC-long complex. Reducing liposome size from 100 to 35 nm accelerated 100-fold the dissociation of the complex triggered by ArfGAP1 (Fig. 1E). This response resembles that observed on liposomes covered solely by Arf1GTP or by Arf1GTP in complex with coatomer (11). Thus, if the Arf1GTP-GRAB domain complex is not sensitive to membrane curvature per se, the control of its attachment by ArfGAP1 creates sensitivity to membrane curvature opposite to that of the N terminus. In liposome binding experiments, the N-terminal region of GMAP-210 (GMAPN included amino acids 1 to 375) bound preferentially to small liposomes, whereas the complex between the C-terminal region (GMAPC-long) and Arf1GTP was stable, in the presence of ArfGAP1, only on large liposomes (Fig. 1F).

Having defined simple rules for the membrane attachment of the two ends of GMAP-210, we wished to reconstitute its tethering activity on liposomes. Because full-length GMAP-210 was difficult to express in E. coli and poorly soluble, we used a shorter construct, miniGMAP, made by fusing GMAPN and GMAPC-long (Fig. 1B). MiniGMAP contained the ALPS motif, one-third of the coiled-coil region, and the GRAB domain with its amphipathic helix. When overexpressed in HeLa cells, this construct affected the Golgi morphology in a manner similar to that of the full-length form (figs. S4 and S5). We assessed the effect of miniGMAP on a homogeneous population of liposomes. Dynamic light scattering(DLS) and electron microscopy (EM) were used to detect liposome aggregation. MiniGMAP caused small liposomes (R = 32 ± 9 nm) covered with Arf1GTP to assemble in large aggregates (0.5 to 1 μm) within minutes (Fig. 2, A and B, and fig. S6). In contrast, almost no aggregation was observed for liposomes devoid of Arf1GTP or when miniGMAP lacked the ALPS motif (Fig. 2A and fig. S7). Liposome aggregation also diminished when vesicle size increased (see below). Thus, the tethering activity of GMAP-210 relies both on its ALPS motif and on the interaction of its GRAB domain with Arf1GTP. Tethering was very efficient: Liposomes aggregated with only 25 nM miniGMAP (fig. S6), a concentration 10- to 100-fold lower than those in other tethering reactions reconstituted with proteins and liposomes (19, 20). This corresponds to 10 to 15 copies of miniGMAP per liposome (table S2), a density similar to what is used to artificially dock liposomes through complementary DNA molecules (21).

Fig. 2.

Liposome tethering induced by miniGMAP. (A) Small liposomes (R = 32 ± 9 nm, 50 μM accessible lipids) covered or not with Arf1GTP (0.25 μM) were mixed with miniGMAP, miniGMAPΔALPS, or GMAPN (0.125 μM). Liposome aggregation was followed by DLS. (Left) Mean radius and polydispersity (shaded area) over time. (Right) Size distribution before (yellow bars) and after (blue bars) the reaction. (B) Electron micrographs of negatively stained small liposomes (R = 38 ± 9 nm) covered with Arf1GTP and incubated or not with miniGMAP as in (A). (C) Two populations of liposomes (small, R = 36 ± 7 nm; large, R = 143 ± 45 nm; 25 μM accessible lipids each) and either covered or not with Arf1GTP (0.25 μM) were mixed. Then 0.125 μM miniGMAP was quickly added, and liposome aggregation was followed by DLS. Thereafter, ArfGAP1 or the 4Ki mutant was added at 0.25 μM. Yellow bars indicate initial size distribution; blue bars, size distribution after aggregation; and red or green bars, final size distribution after ArfGAP1 or 4Ki, respectively, addition. (D) Typical assembly of large (L) liposomes (R = 144 ± 55 nm; 25 μM accessible lipids) covered with 0.125 μM Arf1GTP after incubation with 62.5 nM miniGMAP and with small (s) naked liposomes (R = 41 ± 16 nm; 50 μM accessible lipids). See figs. S6 to S9 for supplementary data and gallery of electron micrographs.

Next, we conducted experiments in which two populations of liposomes of defined size were mixed: one covered with Arf1GTP and one devoid of Arf1 (Fig. 2C). Both were used at the same concentration of accessible lipids. Shortly after liposome mixing, miniGMAP was added, and aggregation was followed. Lastly, ArfGAP1 was added to test the resistance of the liposome aggregates. Strong aggregation was observed for two mixtures: those containing small naked liposomes and small liposomes covered with Arf1GTP and those containing small naked liposomes and large liposomes covered with Arf1GTP (Fig. 2C). In contrast, the aggregation signal was much weaker when both liposome populations were of large size (fig. S8). Thus for membrane aggregation to occur, the presence of highly curved membranes and of membrane-bound Arf1GTP is required (Fig. 2A), but these two determinants do not need to be on the same liposome.

The asymmetric and the symmetric tethering geometries differed in their sensitivity to ArfGAP1. The majority of the aggregates formed in mixture containing only small liposomes disassembled within minutes upon ArfGAP1 addition, whereas those formed by large liposomes covered with Arf1GTP and small naked liposomes were resistant (Fig. 2C). In the latter case, we observed disassembly when we used a mutant of ArfGAP1 (4Ki) that displays more avidity to flat membranes than the wild-type form owing to specific mutations in its ALPS motif (8). Thus, ArfGAP1 through its ALPS motif is capable of selectively disrupting miniGMAP-induced tethering according to the curvature of the membrane on which Arf1GTP anchors the GRAB domain. The fact that ArfGAP1 or the 4Ki mutant can reverse liposome aggregation suggests that no substantial membrane fusion occurred during tethering. EM analysis of incubations conducted with small naked liposomes and large liposomes supplemented with Arf1GTP revealed the formation of clusters containing a few large liposomes and many small liposomes, the latter forming a kind of cement around the larger ones, suggesting that large liposomes contact preferentially small liposomes and vice versa (Fig. 2D and fig. S9). If membrane tethering was random, we should have observed direct contacts between large liposomes as well as clusters of small liposomes such as those visualized previously (Fig. 2B). Thus, in the presence of ArfGAP1, GMAP-210 forms an asymmetric tether that can stably connect a curved membrane to a flat one displaying Arf1GTP but not other geometrical combinations (Fig. 1A).

Next we established a system suitable for light microscopy. We mixed giant liposomes (tens of micrometers in size) labeled with a green fluorophore with small liposomes labeled with a red fluorophore. The former were visible under the microscope, whereas the latter gave a red fluorescence background. Upon attachment of small liposomes, the contour of the giant liposomes became red (fig. S10). Arf1 was allowed to undergo cycles of GTP binding and hydrolysis on the two populations of liposomes, caused by the addition of ArfGAP1 and of the phosphatidylinositol 4,5-bisphosphate (PIP2)–dependent guanine nucleotide exchange factor Arno (22). Because both large and small liposomes contained PIP2, Arno-catalyzed GDP/GTP exchange on Arf1 occurred on the two populations of liposomes, whereas ArfGAP1-catalyzed GTP hydrolysis in Arf1 occurred preferentially on the small liposomes owing to their strong curvature. Thus, with both Arno and ArfGAP1 present, Arf1GTP should be found at steady state preferentially on the giant liposomes. Under these conditions, miniGMAP caused the large liposomes to be surrounded by red fluorescence (Fig. 3, A to C, and fig. S11). In contrast, when ArfGAP1 was absent, we observed numerous red spots of various size and rarely connected to the giant liposomes (Fig. 3D). Thus, ArfGAP1 helps to organize GMAP-210 tethering by preventing the formation of symmetric assemblies between small liposomes, thereby favoring asymmetric assemblies between large and small liposomes.

Fig. 3.

Self-organization of membrane tethering by miniGMAP when the GTPase cycle of Arf1 is controlled by Arno and ArfGAP1. The sample contains giant liposomes (green) and small liposomes (red; R = 42 ± 10 nm), all with the same composition (Golgi mix plus 4% PIP2). Arrows indicate typical liposome assemblies. (A and B) The liposome mixture was supplemented with Arf1 (380 nM), GTP (150 μM), Arno (20 nM), ArfGAP1 (185 nM), and miniGMAP (62.5 nM). a.u., arbitrary units. (C and D) MiniGMAP or ArfGAP1 was omitted. The liposomes were visualized by epifluorescence microscopy [(A), (C), and (D)] or by confocal microscopy (B). Other confocal images are shown in fig. S11.

Multiple tethering events involving different membranes and several long coiled-coil proteins (e.g., p115) occur at the interface between the endoplasmic reticulum and the cis Golgi (4). The minimal model presented here seems adapted to the capture of small transport vesicles at this region, but additional interactions with protein coats, Rabs, and the cytoskeleton may impose a more specific role to GMAP-210 (4, 23). Nevertheless, a reversible tethering mechanism based on the detection of membrane curvature is straightforward because curvature is a good index for the completion of budding and fusion events. By permitting transient interactions between membranes that are continuously remodeled, GMAP-210 may contribute to the self-organizing properties of the Golgi.

Supporting Online Material

Materials and Methods

Figs. S1 to S10

Tables S1 and S2

References and Notes

View Abstract

Navigate This Article