Dynamin self-assembly stimulates its GTPase activity.

GTP hydrolysis by dynamin is required to drive coated vesicle budding at the plasma membrane. A diverse set of molecules including microtubules, grb2, and acidic phospholipids stimulate dynamin GTPase activity in vitro, although the physiological relevance of these effectors remains to be determined. Dynamin has been shown to assemble around microtubules, the most potent stimulatory molecule, into structures indistinguishable by electron microscopy from collars captured in vivo at the necks of endocytic coated pits. Under low ionic strength conditions purified dynamin self-assembles into rings and helical stacks of rings. Here we show that dynamin self-assembly stimulates its GTPase activity as much as 10-fold. Thus, we identify dynamin, itself, as the first effector of dynamin GTPase activity known to be physiologically relevant. Assembled dynamin's stimulated GTPase activity is not dependent on the direct interaction of high affinity GTP binding sites since a mutant defective in GTP binding and hydrolysis can coassemble with and stimulate GTP hydrolysis by wild-type dynamin. Finally, we find that GTP destabilizes assembled dynamin structures, suggesting that the activated rates of GTP hydrolysis reflect a continuing cycle of assembly, GTP hydrolysis, and disassembly.

been shown to assemble around microtubules, the most potent stimulatory molecule, into structures indistinguishable by electron microscopy from collars captured in vivo at the necks of endocytic coated pits. Under low ionic strength conditions purified dynamin self-assembles into rings and helical stacks of rings. Here we show that dynamin self-assembly stimulates its GTPase activity as much as 10-fold. Thus, we identify dynamin, itself, as the first effector of dynamin GTPase activity known to be physiologically relevant. Assembled dynamin's stimulated GTPase activity is not dependent on the direct interaction of high affinity GTP binding sites since a mutant defective in GTP binding and hydrolysis can coassemble with and stimulate GTP hydrolysis by wildtype dynamin. Finally, we find that GTP destabilizes assembled dynamin structures, suggesting that the activated rates of GTP hydrolysis reflect a continuing cycle of assembly, GTP hydrolysis, and disassembly.
Dynamin, a 100-kDa GTPase, is specifically required for endocytic clathrin coated vesicle formation (1,2). Mutations in shibire, the Drosophila homologue of dynamin, cause a pleiotropic defect in endocytosis (3,4). Similarly, transient overexpression of GTPase-defective dynamin mutants blocks receptor-mediated endocytosis in mammalian cells (5,6). More detailed phenotypic analysis of stable transformants overexpressing the K44A dynamin mutant defective in GTP binding and hydrolysis revealed that clathrin-coated vesicle budding was blocked at a stage following coated pit assembly and invagination, but preceding the formation of constricted coated pits (7). Receptor-bound ligands that accumulate in these late intermediates in coated vesicle budding are sequestered from exogenously added macromolecular probes but remain accessible to small molecules (8).
Recombinant human dynamin-1, a neuron-specific isoform, purified from baculovirus-infected Sf9 cells, was found to selfassemble under low ionic strength conditions into rings and stacks of rings (9). The dynamin rings were identical in dimension to electron dense "collars" observed around the necks of the endocytic profiles accumulating at nerve terminals in the Drosophila shibire ts mutant (10). The suggestion that these collars might correspond to dynamin was confirmed by electron microscopy-immunolocalization of dynamin to electron dense helical bands that assembled on membrane invaginations in permeabilized synaptosome preparations incubated with GTP␥S, 1 a nonhydrolyzable analogue of GTP (11). These results, together with the immunolocalization of endogenous and overexpressed dynamin molecules specifically to clathrin coated pits (1, 7) and the GTP-requirements for coated vesicle formation in vitro (12), suggested an early working model (1,9) for dynamin function. We have proposed that dynamin is targeted to and evenly distributed on clathrin lattices in its unoccupied or GDP-bound form. GTP-GDP exchange then triggers dynamin to redistribute from the clathrin-lattice and to assemble into collars at the necks of now constricted coated pits. Finally, GTP hydrolysis by dynamin is required, perhaps to trigger a concerted conformational change by the assembled dynamin, for vesicle detachment and also for recycling of dynamin for reutilization. Thus both the self-assembly and GTPase properties of dynamin are integral to its function. Compared to other members of the GTPase superfamily, dynamin has a relatively high intrinsic rate of GTP hydrolysis (13)(14)(15)(16). Dynamin's GTPase activity can be stimulated in vitro by a diverse group of effectors which include microtubules (13), glutathione S-transferase fusions of SH3 domain-containing proteins (17,18), and acidic phospholipids (15). Each of these effectors binds to dynamin through its ϳ100-amino acid C-terminal proline-arginine-rich domain (PRD), and each is multivalent. The importance of multivalent interactions in stimulating dynamin GTPase activity was established using monoclonal antibodies directed against dynamin's PRD. In these experiments, the stimulatory activity of intact IgGs was enhanced by further cross-linking, while monovalent Fab fragments could not stimulate dynamin GTPase activity unless they were cross-linked (16). The finding that both microtubule-and phospholipid-stimulated dynamin GTPase activity shows strong cooperativity also suggested that dynamin-dynamin interactions were required for stimulated GTPase activity (19).
Here we show that dynamin self-assembly in vitro stimulates GTPase activity in the absence of any effector molecules. GTP binding and hydrolysis, in turn, destabilize the dynamin assemblies, suggesting a cycle of GTP hydrolysis, disassembly, and reassembly. The ability of dynamin to stimulate its own * This work was supported by National Institutes of Health Grant GM42455 (to S. L. S.). This is TSRI manuscript no. 10065-CB. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
‡ GTPase activity is not dependent on a high affinity GTP binding site, since a mutant defective in GTP binding and hydrolysis can coassemble with and stimulate GTP hydrolysis by wild-type dynamin.

EXPERIMENTAL PROCEDURES
Materials-Protease inhibitor mixture tablets were from Boehringer Mannheim and calpain inhibitor I was obtained from Calbiochem. Serum-free medium, EX-CELL 401, was purchased from JRH Biosciences (Lenexa, KS). Fetal calf serum and Fungizone solution were from Irvine Scientific (Santa Ana, CA). The supplier of Macro-Prep High Q strong anion exchange support and Macro-Prep Ceramic Hydroxyapatite was Bio-Rad. [␣-32 P]GTP was from Amersham Corp. GTP and GDP were acquired from Sigma and GTP␥S was from Boehringer Mannheim. Unless otherwise indicated, all other chemicals were reagent grade.
Expression of Recombinant Dynamin in Sf9 Cells and Purification of Dynamin Protein-Human neuronal dynamin (dynamin-1) and K44A dynamin-1 containing a lysine to alanine conversion at amino acid 44 (7) were expressed in recombinant baculovirus-infected Sf9 cells exactly as described previously (16). However, a new purification procedure was used. Cells were infected at 3-5 plaque-forming units/cell with high titer virus stocks and then harvested 65 h later by centrifugation at 500 ϫ g for 10 min, washed once with phosphate-buffered saline and repelleted. The cell pellet obtained from a 0.5-liter culture was resuspended in 25 ml of Hepes column buffer (HCB, 20 mM Hepes, pH 7.2, 2 mM EGTA, 1 mM MgCl 2 , 1 mM DTT) containing 100 mM NaCl (referred to as HCB100), a protease inhibitor mixture tablet, and 1 mM PMSF. All subsequent steps were performed at 4°C. Cells were homogenized by N 2 -cavitation at 500 p.s.i. for 25 min prior to slow release. The homogenate was diluted 2-fold in HCB0 (no NaCl) and then centrifuged at 50,000 rpm for 60 min in a Beckman Ti60 rotor. The supernatant was collected.
The supernatant was brought to 30% ammonium sulfate by slow addition of salt. After 10 min of slow stirring, the solution was centrifuged for 10 min in a Beckman JA-20 rotor at 10,000 ϫ g. The pellet was gently resuspended using a loose fitting Dounce in 10 ml of HCB50 (50 mM NaCl) with protease inhibitors. The resuspended fraction was recentrifuged for 10 min at 10,000 ϫ g to remove aggregated protein. The solubilized 30% NH 2 SO 4 cut was then applied to a High-Q strong anion exchange (Bio-Rad) column (10 ϫ 1.5 cm) preequilibrated in HCB50. The column was washed with 50 ml of HCB50 and then with 50 ml of HCB100. Dynamin was finally step eluted in a volume of about 10 ml of HCB250. Protease inhibitor mixture, 40 M calpain inhibitor 1, and 1 mM PMSF were added to the pooled High Q eluate which was then brought to 5 mM CaCl 2 . The dynamin fraction was loaded onto a 10-ml Macro-Prep Ceramic Hydroxyapatite column (Bio-Rad) preequilibrated with HCB250, 5 mM CaCl 2 . The column was washed with 30 ml of 200 mM KPO 4 pH 7.2 and then eluted in 400 mM KPO 4 (pH 7.2). K44A and wild-type dynamin-1 behaved identically during this procedure and yielded up to 15 mg of dynamin from 1 ϫ 10 9 infected Sf9 cells. Purity was greater than 95% as judged by Coomassie Blue staining following SDS-PAGE. Aliquots were stored at Ϫ80°C in 400 mM KPO 4 containing 1 mM DTT, 40 M calpain inhibitor 1, and 1 mM PMSF.
Dialysis of Dynamin and Assay of GTPase Activity-Dynamin was transferred to HCB150 or GTPase assay buffer referred to as PH buffer (20 mM Pipes, 20 mM Hepes, 2 mM MgCl, 1 mM EGTA, 1 mM DTT, pH 7.0), by overnight dialysis with two buffer changes. Spectra/Por 2 dialysis membrane tubing (Spectrum Medical Industries, Houston, TX) with a molecular weight cut-off of 12,000 -14,000 or a framed dialysis membrane for a microdialyzer system 100 apparatus (Pierce) with a molecular weight cut-off of 8,000 were used with equivalent results.
GTPase assays were performed in PH buffer with 1 mM DTT and 0.1% bovine serum albumin in a final volume of 20 l, essentially as described elsewhere (7,16). The final ionic strength of the GTPase assay buffer was always adjusted to that of PH buffer and control experiments confirmed that addition of small amounts of HCB150 to GTPase assays, while maintaining final ionic strength, had no effect on GTPase activity. Reactions were initiated by the addition of GTP (0.1 Ci of [␣-32 P]GTP) (Amersham). 1.5-l aliquots were removed at each time point and spotted onto cellulose polyethyleneimine thin layer chromatography plates with fluorescent indicator (J. T. Baker, Inc.). Nucleotides were resolved by TLC in 1 M LiCl 2 , 2 M formic acid (1:1). Quantitation of GTP and GDP at each time point was performed on a PhosphorImager (Molecular Dynamics Inc., Sunnyvale, CA). Rates of GTP hydrolysis were calculated from a minimum of five time points and expressed as the percent of GDP per GTP plus GDP.
Dynamin Assembly Assay and SDS-PAGE Analysis-Dynamin selfassembly into oligomeric structures was assayed as described previ-FIG. 1. Dynamin forms sedimentable structures upon dilution into GTPase assay buffer conditions. A sedimentation assay was used to analyze the self-assembly of wild-type (a and b) and K44A (c and d) dynamins into structures that remained soluble (S) or were pelleted (P) during centrifugation at greater than 100,000 ϫ g for 15 min. Dynamins were dialyzed into HCB150 overnight and subsequently diluted 8-fold into either HCB150 (a and c) or PH buffer (b and d). Equal aliquots of soluble and pelleted dynamin were analyzed by SDS-PAGE on 7.5% acrylamide gels. The gel was stained with Coomassie Brilliant Blue, and the relative distribution between fractions was determined by densitometer scanning and analysis using Molecular Dynamics Image-Quant software. The broad higher molecular band in S fractions corresponds to a contaminant in the bovine serum albumin used as carrier protein. ously (9). Dynamin at 2 mg/ml in HCB150 was diluted or dialyzed 8-fold into various buffer conditions on ice. Supernatant and pellet fractions were obtained after centrifugation at 50,000 rpm for 15 min in a TLA100 rotor (Beckman Instruments). Dynamin in each fraction was resolved by SDS-PAGE, on 7.5% acrylamide gels, using standard methods as described previously (16). Quantitation of dynamin was performed by scanning Coomassie Brilliant Blue-stained gels on a Molecular Dynamics personal densitometer and using Molecular Dynamics ImageQuant software to calculate relative intensities of dynamin in supernatant and pellet fractions. Standard curves of dynamin were analyzed to determine the optimal linear protein concentration sensitivity of the assay.

RESULTS
We showed previously that purified dynamin self-assembles into sedimentable structures composed of rings and small stacks of rings upon dilution into low ionic strength buffers (9). Interestingly, the GTPase activity of dynamin is routinely measured under low ionic strength conditions (13)(14)(15). We therefore used a sedimentation assay to determine whether dynamin self-assembly occurs under GTPase assay conditions. Thus, dynamin was diluted into GTPase assay buffer (referred to here as PH buffer, see "Experimental Procedures") and subjected to sedimentation analysis as described previously (9). The SDS-polyacrylamide gel in Fig. 1 (lanes a) shows that Ͼ95% of dynamin is found in the supernatant in the starting buffer, HCB150 which contains 150 mM NaCl. However, upon 8-fold dilution into PH buffer, Ͼ75% becomes sedimentable (Fig. 1b). The structures assembled by dynamin in PH buffer were indistinguishable from those described previously (9) (data not shown). As reported previously (9) self-assembly at low ionic strength does not require either the presence of guanine nucleotides or dynamin's high affinity GTP binding activity since the K44A mutant of dynamin defective in GTP binding and hydrolysis (7) shows comparable self-assembly activity (Fig. 1, c and d).
It had previously been shown that stimulated dynamin GTPase activity was highly cooperative (19). Given dynamin's ability to self-assemble under GTPase assay conditions, we examined the concentration dependence of dynamin's intrinsic GTPase activity. For this experiment stock solutions of dynamin were prepared in HCB150 and then diluted 8-fold into PH buffer at 4°C so that the final concentration of NaCl in the assay remained constant. GTPase assays were transferred to 37°C and initiated by addition of [␣-32 P]GTP. The kinetics of GTP hydrolysis were determined for each concentration of dynamin as described under "Experimental Procedures." The rate of GTP hydrolysis increased in a nonlinear fashion with increasing dynamin concentration ( Fig. 2A). A plot of the specific activity of dynamin's intrinsic GTPase (Fig. 2B) gave a sigmoidal curve showing a sharp concentration dependence and reaching maximal specific activity at 2 M dynamin. These data suggested that dynamin's intrinsic GTPase activity could be stimulated by self-assembly.
While the trends shown in Fig. 2 were consistently obtained, we noted several experimental factors affecting dynamin's GTPase activity. For example, when dialysis was performed in the absence of DTT, higher specific activities were obtained at all concentrations of dynamin (not shown). When dynamin was dialyzed into buffer containing 100 mM NaCl (as opposed to 150 mM) higher specific activities were obtained at high concentrations of dynamin. Finally, preincubation of dynamin (for 30 -90 min) after dilution into GTPase buffer resulted in higher specific activities at lower dynamin concentrations but did not affect specific activities at higher concentrations. This result probably reflects the slower kinetics of self-assembly at lower concentrations of dynamin. It should be noted that the high variation of intrinsic GTPase rates for dynamin reported in the literature, from as low as Ͻ1 min Ϫ1 (14, 18) to as high as 23 min Ϫ1 (15), probably reflect differences in protein handling and assay conditions.
To further test the relationship between dynamin GTPase activity and self-assembly, dynamin was preassembled by dialysis into PH buffer (9) and then assayed for intrinsic GTPase activity upon dilution into PH buffer containing increasing concentrations of NaCl. Other samples were diluted in parallel and subjected to sedimentation analysis. As can be seen, dynamin's intrinsic GTPase activity (Fig. 3A) decreased with increasing salt concentrations in parallel with the disassembly of sedimentable structures (Fig. 3B).
These results suggest that dynamin-dynamin interactions are important for stimulated GTPase activity. Does the stimulation of dynamin GTPase activity in assembled structures require that the neighboring dynamin molecules are themselves able to bind and hydrolyze GTP? To address this question, wild-type dynamin was coassembled with increasing amounts of a mutant dynamin (designated K44A) that is defective in GTP binding and hydrolysis by virtue of a point mutation in the first conserved GTP binding element (7). As seen in Fig. 1 (lanes 7 and 8), K44A dynamin was able to self-assemble into structures indistinguishable from wild-type dynamin (see also Hinshaw and Schmid (9)). However, as shown in Fig. 4, A and B (open circles), the K44A mutant on its own had little or no intrinsic GTPase activity, nor could its and sedimentability (panel B) were determined on dynamin preassembled by dialysis into PH buffer. The GTPase activity of 4 g of dynamin was determined in 1 mM GTP, as described under "Experimental Procedures," while sedimentability was assayed in PH buffer using an equivalent concentration of dynamin but without GTP. GTPase activity be stimulated by microtubules (not shown, but see Damke et al. (7)). Nonetheless, when K44A dynamin was allowed to coassemble with a fixed and small amount of wildtype dynamin during dialysis into PH buffer, these mixtures showed greatly stimulated GTPase activity dependent on the concentration of K44A dynamin (Fig. 4, closed triangles). In fact, the K44A mutant appeared to be even more effective than wild-type dynamin at stimulating specific GTPase activity. The maximum specific activity of wild-type dynamin alone under these assay conditions was 9.4 min Ϫ1 (Fig. 2B), while when coassembled with K44A mutant, wild-type dynamin hydrolyzed GTP at a rate of ϳ20 min Ϫ1 (Fig. 4B).
In an effort to determine why wild-type dynamin showed greater GTPase activity when coassembled with the K44A mutant dynamin than with itself, we examined the effects of GTP on the stability of the assembled dynamin structures. The experiment in Fig. 5 shows that incubation of preassembled wild-type dynamin (lanes a) with either GTP (lanes b-d) or GTP␥S (lanes e and f) destabilized the sedimentable structures, releasing dynamin into the supernatant. In fact, GTP␥S was more potent than GTP. A 58% release of sedimentable dynamin occurred in the presence of 10 M GTP␥S (lanes e), while only a 42% release occurred following incubation in 250 M GTP (lanes c). GDP, however, was considerably less effective at destabilizing assembled dynamin with only 32% released at 1 mM GDP (lanes i).
The observation that nonhydrolyzable GTP analogues were at least as effective as GTP in destabilizing dynamin assemblies allowed us to compare the relative stability of preassembled K44A dynamin to wild-type dynamin in the presence of GTP␥S. In this way, the differential hydrolysis of GTP by wild-type and K44A dynamin would not affect the results. The data in Fig. 6 shows the effect of increasing concentrations of GTP␥S on the stability of preassembled wild-type dynamin (open squares), K44A mutant dynamin (open circles) or a 4:1 (K44A:wild type) mixture of the two (closed triangles) as assessed by sedimentation analysis. As expected, assembled K44A dynamin required ϳ20-fold more GTP␥S to trigger its disassembly than did wild-type dynamin, reflecting its greatly reduced affinity for GTP (7). Moreover, when wild-type dynamin was coassembled with the K44A mutant at the ratio that gave near-maximum stimulation of GTPase activity (see Fig. 4B), these coassembled structures were as resistant to GTP␥S-induced disassembly as the K44A mutant itself. DISCUSSION We have shown that dynamin's intrinsic GTPase activity is significantly enhanced under conditions that favor self-assembly of dynamin into rings and helical stacks of rings. This directly confirms previous suggestions based on the cooperative behavior of dynamin's GTPase activity when stimulated by acidic phospholipids and microtubules (19), that dynamin-dynamin interactions were key to regulating its GTPase activity. Moreover, we have shown that stimulation of dynamin GTPase does not depend on the high affinity binding or hydrolysis of GTP by neighboring dynamin molecules. While the nature of the dynamin-dynamin interactions required for stimulated GTPase activity remain to be determined, we have identified dynamin as the first effector of its GTPase activity known to be physiologically relevant. In this regard it is of interest to note that all of the suppressors of shibire so far identified through genetic analysis are intra-allelic (20), consistent with the suggestion the dynamin-dynamin interactions are critical to its in vivo function.
Interestingly, we also show that GTP binding destabilizes GTP␥S (e and f), and GDP (g-i) were determined. Dynamin was preassembled by dialysis overnight into PH buffer. Eightfold dilutions of dynamin were made into PH buffer containing the varying concentrations of nucleotide. The assays were kept at 4°C for 10 min before separation into soluble (S) and pelletable (P) fractions by ultracentrifugation and analysis as described in Fig. 1. structures assembled by purified dynamin under low ionic strength conditions. This suggests that the activated rates of GTP hydrolysis that occur at higher concentrations of dynamin reflect a continuing cycle of assembly, GTP hydrolysis and disassembly. That the K44A mutant is even more effective at stimulating GTP hydrolysis rates by wild-type dynamin can be explained by the fact that these coassembled structures are resistant to destabilization by GTP binding. In contrast, it is unlikely that intramolecular dynamin interactions are absolutely required for the low basal rates of GTP hydrolysis seen in dilute solutions of dynamin for several reasons. First, if collisions between individual dynamin molecules were required for GTP hydrolysis, then dynamin's GTPase activity would remain highly concentration dependent even under dilute conditions. Instead, we find that the rate of GTP hydrolysis obtained at concentrations below an apparent threshold for dynamin assembly (i.e. Ͻ500 nM) is concentration independent. Second, the basal rate of dynamin GTPase observed under high ionic strength conditions that disfavor self-assembly are also concentration-independent (not shown). Finally, limited proteolysis of dynamin removes the PRD and prevents dynamin assembly (9). Truncated dynamin molecules exhibit the same intrinsic GTPase activity as intact dynamin (15,16), 2 and this "basal" GTPase rate is independent of protein concentration. 2 Our current working model for dynamin function (1,9) suggests that dynamin is targeted to and evenly distributed on clathrin lattices in its unoccupied or GDP-bound form. GTP-GDP exchange then triggers dynamin to redistribute from the clathrin lattice and to assemble into rings forming helical collars at the necks of invaginated coated pits. These collars function to form constricted coated pits (7,8). Finally, GTP hydrolysis by dynamin is required, perhaps to trigger a concerted conformational change by the assembled dynamin, for vesicle detachment. Dynamin disassembles at this stage for recycling and reutilization. Many aspects of this early model need to be tested and confirmed. In particular, consistent with dynamin's membership in the GTPase superfamily of proteins, we assume that conformational changes triggered by GTP binding and hydrolysis affect dynamin's interaction with as yet unidentified partners required for membrane binding, self-as-sembly under physiological conditions and other aspects of coated vesicle budding.
Key to elucidating dynamin's function will be an understanding of its GTPase cycle. In our initial studies, reported here, we have identified properties of dynamin GTPase activity in vitro some of which are consistent, others inconsistent with our current working model. For example, the finding that dynamin GTPase activity is highest when dynamin is assembled into rings or helical stacks of rings is consistent with GTP hydrolysis being triggered after collar assembly. Preassembled dynamin rings can hydrolyze GTP at rates in excess of 10 min Ϫ1 , and if these rings are stabilized, for example by coassembly with the K44A mutant, rates can increase to 20 min Ϫ1 . These rates are sufficient to account for the apparently short-lived nature of the constricted coated pit intermediate. The finding that GTP hydrolysis releases soluble dynamin from assembled structures is also consistent with our proposal that dynamin is recycled for reutilization at this stage.
In contrast, we find that dynamin self-assembly in vitro does not require guanine nucleotides, yet GTP appears to be required in vivo for the redistribution of dynamin from clathrin lattices and its assembly into collars. Unlike unassembled dynamin, dynamin collars are spatially segregated from the clathrin lattice (1,11). We speculate that GTP/GDP exchange would trigger a conformation change in dynamin, as it does for other members of the GTPase superfamily. The GTP-induced conformational change may be necessary in vivo to alter dynamin interactions with partners on the clathrin lattice releasing it for assembly. Assembled dynamin is destabilized by GTP or GTP␥S in vitro, yet GTP␥S stabilizes helical dynamin stacks triggering their exaggerated assembly on synaptosomal membrane fractions (11). These differences suggest that other factors may be required in vivo to regulate both dynamin assembly and disassembly and that these factors might be membrane associated. It will be essential to identify components that interact with dynamin in vivo in order to determine their role in regulating dynamin targeting, GTPase activity, and assembly. The assays developed here will be important in testing the function and determining the mechanism of action of both dynamin and its partners.