Phosphatidylinositol (4,5)-Bisphosphate-dependent Activation of Dynamins I and II Lacking the Proline/Arginine-rich Domains*

Dynamins comprise a family of GTPases that participate in the early stages of endocytosis. The GTPase activity of neuronal specific dynamin I is stimulated by microtubules, negatively charged phospholipid vesicles, and Src homology 3-containing proteins, including Grb2. These activators were previously shown to bind to a proline/arginine-rich domain (PRD) in the carboxyl-terminal region of the enzyme. Dynamin II, which is ubiquitously expressed, had not been purified or characterized previously. In this study, the enzymatic properties of rat dynamin II and of D746, a dynamin II truncation mutant lacking the PRD, have been characterized. Dynamin II has a higher basal activity than dynamin I, but the two types of dynamin are stimulated similarly by microtubules, Grb2, and phospholipids. D746 is not activated by microtubules or Grb2, highlighting the significance of the PRD for these interactions, but it is activated by phospholipid vesicles containing phosphatidylserine or phosphatidylinositol-4,5- bisphosphate. Moreover, in contrast to previous reports, the PRD appears not to be required for phospholipid-stimulated self-assembly of dynamin, which is a key element in the regulation of its activity. Similar results were obtained with bovine brain dynamin I that had been subjected to limited proteolytic digestion to remove the PRD. Our data highlight the potential involvement of dynamin pleckstrin homology domains in the regulation of GTPase activity by phospholipids.

Dynamins constitute a family of GTPases that carry out an essential function in clathrin-dependent endocytosis (for recent reviews, see [1][2][3][4]. Although the precise role of dynamin is still uncertain, it has been proposed to couple the energy released from GTP hydrolysis to the generation of mechanical force for membrane fission during coated vesicle budding (2). Three forms of dynamin have been identified in vertebrates. Dynamin I, found only in neurons, has been implicated in the process of presynaptic vesicle recycling. Dynamin II, which is ubiquitously expressed, is thought to participate in receptor-mediated endocytosis. Dynamin III is most abundant in testes, but its function is unknown. Of these three forms, only dynamin I has been purified and characterized. However, amino acid se-quence analysis reveals that dynamins I, II, and III have a similar organization of functional domains. A GTP binding catalytic domain is located in the amino-terminal region (residues 1-300); a pleckstrin homology (PH) 1 domain spans residues 510 -620; and a proline/arginine-rich domain (PRD) extends over 100 residues from amino acid 750 to the carboxyl terminus (residues 864 in rat dynamin I and 871 in rat dynamin II). Because of its very basic nature (pI ϳ12.5) and the presence of multiple Src homology 3 binding motifs, the PRD can interact with numerous macromolecules. Some of these, including microtubules (5-7), negatively charged phospholipid vesicles (7)(8)(9), and Src homology 3-containing proteins, e.g., Grb2 (10, 11), were found to stimulate dynamin I GTPase activity to various extents.
The mechanism of GTPase activation is poorly understood. However, based on kinetic data showing a dependence of activity on enzyme concentration (8,9,12,13), dynamin self-association appears to be a key element of the activation process. It has been proposed that the two most potent activators of dynamin I, microtubules and phosphatidylserine (PS) vesicles, stimulate activity by facilitating dynamin self-assembly (8). They do so by providing multivalent negatively charged surfaces that bind to the highly basic PRDs of dynamin molecules. The ionic nature of these interactions is evident by their marked sensitivity to ionic strength; both microtubule-and PS-stimulated activities are nearly abolished at physiological salt concentrations (8). Other activators of dynamin GTPase, including Grb2 and a subset of anti-dynamin antibodies (11,13), were also found to interact with the PRD. These proteins stimulate dynamin I GTPase to only about 20% of the maximal activities obtained in the presence of microtubules or PS vesicles. However, it is believed that the underlying mechanism of their activation is similar, i.e. stimulation of GTPase occurs as a consequence of dynamin cross-linking by anti-PRD antibodies or Grb2.
Rat dynamins I and II have an overall sequence identity of ϳ80%, but their PRDs are less than 50% identical (2). In view of this difference and of the apparent significance of the PRD in the regulation of dynamin I, we undertook an investigation of the activation properties of dynamin II. Because dynamin II has not been purified from cells or tissues, we expressed rat dynamin II in Sf9 cells in sufficient quantities for biochemical analysis. We found that dynamin II has a 10-fold higher basal activity than dynamin I. However, despite the differences in amino acid sequence of their PRDs, the two forms of dynamin interact in a similar manner with microtubules, Grb2, and phospholipid vesicles. To reinvestigate the role of the PRD in GTPase activation, a dynamin II truncation mutant (termed D746) which lacks this domain was expressed and purified. As expected from previous studies of proteolyzed dynamin I lacking the PRD (7,11), D746 failed to be activated by microtubules or Grb2. Surprisingly, the PRD was not required for expression of phospholipid-stimulated activity or for cooperative GTPase activation because of dynamin self-association. Moreover, phosphatidylinositol 4,5-bisphosphate (PI(4,5)P 2 )-stimulated GTPase activity was undiminished at physiological ionic strength. This behavior was also observed using papaincleaved dynamin I fragments that lack the PRD, indicating that both dynamins can be activated in the absence of that domain.
The PH domain of dynamin I has been expressed and characterized (14,15). Like many other PH domains, it was found to bind phospholipids, with phosphoinositides having the highest affinity. Moreover, deletion of the PH domain resulted in the loss of PI(4,5)P 2 -stimulated GTPase of mutant expressed dynamin I (16). However, the question of whether the PRD is also required for this activation was not addressed. The results presented in the paper suggest that dynamins are subject to multiple modes of regulation, some involving interactions with the PRD, others with the PH domain.
Construction of Recombinant Dynamins-cDNAs of carboxyl-terminal His 6 -tagged wild-type dynamin II and D746, a dynamin II mutant truncated after residue 746, were constructed by polymerase chain reaction using the rat dynamin II cDNA as a template (pCMV 96-15) (17). The primers were designed such that a unique 5Ј-EcoRV site and a 3Ј-HindIII site were introduced. The amplified DNA was digested with EcoRV and HindIII and ligated to the pBacPak 8 plasmid (CLON-TECH) digested with SmaI and HindIII. The resulting plasmids were cotransfected with BacPak 6 viral DNA, digested with Bsu36I (New England Biolab) into Sf9 insect cells to produce recombinant baculoviruses. Recombinant viruses were plaque purified and amplified using the standard procedure. DNA sequencing was performed to confirm the integrity of all constructs.
Purification of Recombinant Dynamin II-250 ml of Sf9 cells grown in IPL-41 medium supplemented with 10% fetal calf serum and 1% pleuronic acid were infected with recombinant baculoviruses at a multiplicity of infection of 5. After 64 h, the cells were harvested by centrifugation at 1,000 ϫ g for 10 min. Cells were then washed once with 100 ml of phosphate-buffered saline. The following procedures were performed at 4°C. Cells were resuspended in 10 ml of buffer A (20 mM Hepes, pH 8.0, 1 mM MgCl 2 , 150 mM NaCl, 0.5 mM dithiothreitol, and 1 mM phenylmethylsulfonyl fluoride) and homogenized with a Dounce homogenizer. The homogenate was centrifuged at 100,000 ϫ g for 1 h, and the resulting supernatant was passed through a 2-ml Ni 2ϩ -NTA agarose column equilibrated with buffer A. The column was washed with 20 ml of buffer A followed by 20 ml of buffer A containing 30 mM imidazole. Recombinant dynamin was eluted with buffer A containing 100 mM imidazole. Protein yields were 4 mg and 3.5 mg for wild type dynamin and D746, respectively.
Purification of Dynamin I-Bovine brains were homogenized with 0.1 M MES, pH 7.0, 1 mM EGTA, 1 mM MgSO 4 , 1 mM dithiothreitol, 1 mM sodium azide, and a range of protease inhibitors: 0.2 mM phenylmethylsulfonyl fluoride and 10 mg/liter of N ␣ -benzoyl-L-arginine methyl ester, N ␣ -p-tosyl-L-arginine methyl ester, N ␣ -p-tosyl-L-lysine chloromethyl ketone, leupeptin, and pepstatin A (buffer B). The extract was chromatographed on three consecutive ion exchange columns: DE52cellulose, P11 phosphocellulose, and SP-Sepharose. Fractions highly enriched in dynamin were mixed with microtubules, ultracentrifuged, and dynamin cosedimenting with microtubules was released by resuspension in 10 mM GTP. The supernatant was passed through a DE52 column to remove any traces of tubulin. Dynamin was usually about 95% pure after these steps. However, if further purification was necessary, the dynamin preparation was centrifuged on a 5-15% sucrose gradient at 112,000 ϫ g for 16 h.
Other Proteins-Tubulin was purified according to the procedure of Williams and Lee (18), but MES instead of PIPES buffer was used. For GTPase activation experiments tubulin was polymerized with taxol at a 2-fold molar excess to tubulin dimer. Grb2 was expressed in Escherichia coli as a fusion protein with glutathione S-transferase (GST) and purified on glutathione-Sepharose.
GTPase Assays-GTPase activities were measured by the release of 32 P i from [␥-32 P]GTP (19) after incubation at 37°C in buffer B containing additionally 1 mM MgGTP. The reaction time varied from 30 min for low concentrations of dynamin to 2 min at high dynamin concentrations to ensure that less than 15% of added GTP was hydrolyzed.
Proteolysis with Papain-Dynamin (1-2 M) was digested with papain for various times at 30°C at a 1:1,000 enzyme:dynamin ratio. Papain was activated by incubation on ice for 15 min in a solution containing 0.5 M NaCl, 25 mM Tris-HCl, pH 7.5, 2 mM EDTA, and 5 mM dithiothreitol. Digestion was terminated by adding iodoacetic acid to a final concentration of 4 mM, and samples were either used for GTPase assay or boiled with SDS sample buffer for SDS-polyacrylamide gel electrophoresis.
Preparation of Phospholipid Vesicles-Phospholipids were dissolved in chloroform and dried under argon. Dried lipids were dissolved in 0.1 M MES buffer and sonicated for 10 min at maximum power (Bath sonicator model W185; Heat System Ultrasonics, Inc., Farmingdale, NY). PS and PS/PC vesicles gave similar level of GTPase activation, although at different total lipid concentrations. In this study, only vesicles containing 100% PS were used. PI(4,5)P 2 was prepared as a mixture with PC in a 1:9 molar ratio.
Other Methods-Protein concentration was determined as described by Bradford (20) using bovine serum albumin as a standard. SDSpolyacrylamide gel electrophoresis was carried out according to the method of Laemmli (21) as modified by Matsudaira and Burgess (22). Immunoblot analysis was carried out by the method of Towbin et al. (23) as described previously (24).

Characterization of the GTPase Activities of Dynamin II-
Because dynamin II has never been purified, there has been no previous characterization of its properties. To study dynamin II, rat dynamin II with a COOH-terminal His 6 tag was expressed from recombinant baculovirus in Sf9 cells and purified by chromatography on nickel resin. The expressed protein migrates with an apparent molecular weight of 101,000 on SDSpolyacrylamide gel electrophoresis (Fig. 1), consistent with the monomer molecular weight of 98,169 calculated from the amino acid composition. Differences in the PRDs of the two types of dynamin suggested the possibility of differences in their regulation. Therefore, the GTPase activity of the purified enzyme was assayed in the presence or absence of four known activators of dynamin I: phospholipid vesicles consisting of 100% PS or 10% PI(4,5)P 2 and 90% PC, GST-Grb2, and microtubules. Dynamin II has a 10-fold higher basal activity than bovine brain dynamin I (ϳ20 min Ϫ1 versus ϳ2 min Ϫ1 for dynamin I). Nevertheless, the enzymatic properties of the two dynamins are similar (Fig. 2). GTPase activation profiles of both dynamins are biphasic, and maximal activities are reached at similar dynamin:activator ratios. Microtubules and anionic liposomes are the most effective activators, accelerating the GTPase of each to more than 200 min Ϫ1 . Grb2, expressed as a fusion protein with GST, stimulated the GTPase of each to a maximal level of ϳ55 min Ϫ1 . Although specific activities of dynamins were slightly variable from preparation to preparation (see also Ref. 7), we observed dramatic variations of the enzyme activation depending on the nature of the phospholipids, with up to 5-fold differences in activation using different lipid batches but the same dynamin preparation. This may account in part for the lower PI(4,5)P 2 -stimulated GTPase ac-tivities of human dynamin I reported earlier 2 (9). Nevertheless, both forms of dynamin showed nearly identical activation patterns, regardless of the maximal specific activities achieved.
Activation of the GTPase Activity of Truncated Dynamin II, D746 -The PRD contains binding sites for all the currently known stimulators of dynamin GTPase activity, including microtubules, Src homology 3-containing proteins, and negatively charged phospholipids. Hence, it has been proposed that this domain is required for dynamin activation (1), and data supporting this view have been presented (7,11). It has been shown recently, however, that expressed dynamin I PH domains interact with certain negatively charged phospholipids, e.g. PI(4,5)P 2 (K D ϭ 4.4 M) and PS (K D ϭ 47.2 M) (15). In light of these observations we reexamined the question of whether or not the PRD is indeed essential for phospholipid-stimulated GTPase activity. A recombinant protein containing residues 1-746 of rat dynamin II was expressed in Sf9 cells from a recombinant baculovirus and purified on nickel resin (Fig. 3,  inset). Like full-length dynamin II, D746 has a 10-fold higher basal GTPase activity than dynamin I, but, as expected, its ability to be activated by microtubules or GST-Grb2 is nearly abolished (Fig. 3). In contrast, the GTPase activity of D746 is stimulated potently by negatively charged phospholipid vesicles. Vesicles containing PI(4,5)P 2 were especially efficacious. Thus, despite deletion of the PRD, two of the activators retained the ability to activate the truncated protein. Therefore, we conclude that the PRD is not required for all modes of dynamin activation.
Effect of Proteolytic Digestion on the GTPase Activity of Bovine Brain Dynamin I-Previous studies demonstrating the importance of the PRD for phospholipid-stimulated GTPase activity examined proteolytic fragments of dynamin I lacking the COOH-terminal domain (7,11). Our results with D746 prompted us to reevaluate these earlier findings. Bovine brain dynamin I was subjected to limited papain digestion, yielding 2 In Ref. 9, PI(4,5)P 2 -stimulated GTPase activities of up to 25 min Ϫ1 were reported for dynamin I. However, those assays were performed at exceedingly low dynamin concentrations (5 nM) and far from the optimal dynamin:lipid molar ratios of about 1:10 -1:20 (see Fig. 8). This, together with the potential variation caused by differences among lipid batches discussed under "Results," may explain the 10-fold differences in specific activities between this paper and the earlier report.  an 88-kDa fragment that lacks the PRD, as determined by its failure to be recognized by an antibody against residues 837-851 (Fig. 4B) and its inability to bind to GST-Grb2 (25). The 88-kDa fragment is gradually digested to an 85-kDa polypeptide, which is cleaved further to 53-and 32-kDa fragments. Recognition by antibodies raised against the GTP binding domain (residues 45-358) (Fig. 4C) confirms that the 53-kDa fragment contains the catalytic site. The 32-kDa fragment contains the PH domain and is thus recognized by antibodies raised against a peptide corresponding to residues 607-624 (Fig. 4D). The latter two fragments appear to exist in a nonco-valent complex because they comigrate on a Superose 12 gel filtration column (data not shown). Not only the 85-kDa but also the 53-kDa/32-kDa fragments retain GTPase activity that is approximately 10-fold higher than the basal activity of undigested dynamin I (Fig. 5B). In agreement with previous reports (7,11), dynamin I fragments lacking the PRD are not stimulated by microtubules or by GST-Grb2 (Fig. 5A). However, as predicted from analysis of D746, activation by PS or PI(4,5)P 2 vesicles is undiminished.
PI(4,5)P 2 -stimulated GTPase Activity Occurs at Physiological Ionic Strength-Activation of dynamin I GTPase activity in vitro by microtubules and PS vesicles is apparent only under conditions of low ionic strength (7) (Fig. 6A). The same is true for expressed rat dynamin II (data not shown). This salt sensitivity has raised a question about the physiological relevance of these activators (2). In contrast, Grb2-stimulated activity was not reduced by salt (Fig. 6A), and PI(4,5)P 2 -stimulated activities of both intact and truncated dynamins were only slightly reduced at NaCl concentrations higher than 150 mM (Fig. 6B).
Cooperative Activation of D746 and Dynamin I Fragments-In the presence of microtubules or PS vesicles, the specific GTPase activity of dynamin I increases cooperatively as a function of dynamin concentration (8) (Fig. 7). This behavior is characteristic of self-associating enzymes having activities that depend on their state of oligomerization (26). Tuma and Collins (8) proposed that dynamin GTPase activation involves dynamin self-association of dynamin molecules which is facilitated by their alignment on multivalent surfaces such as microtubules or PS vesicles. As discussed above, the interactions between dynamin I and these anionic surfaces are believed to be mediated by the positively charged PRDs. However, our results demonstrate that the PRD is not required for cooperative activation by PI(4,5)P 2 because it occurs not only with intact dynamins (Fig. 8, A and C) but also with D746 (Fig. 8D) and with the 53-kDa/32-kDa proteolytic fragments (Fig. 8B), which lack the PRD. Specific GTPase activities rise with increasing dynamin concentrations until they reach a maximum and thereafter decline. These maxima occur at different dynamin concentrations but at similar dynamin:phospholipid ratios. A plausible explanation for this behavior is that after maximal activities are achieved at a given dynamin:phospholipid ratio, steric blocking prevents further dynamin binding and activation. Because maxima occur at dynamin:phospholipid ratios of about 1:200, each dynamin molecule apparently occupies an area of the liposome surface containing 100 lipid molecules (approximately 6 nm ϫ 6 nm), if only the outer monolayer is considered.
Our results indicate that the COOH-terminal PRDs of dynamins I and II are not required for the activation of dynamin GTPase activity by phospholipids. In view of the well documented interactions between specific phospholipids and dynamin PH domains (15,16), it seems likely that activation occurs as a consequence of phosphoinositide binding to this domain. Significantly, this binding can take place even at physiological ionic strength, unlike the interaction between dynamin and microtubules, another potent stimulator of GTPase activity. Finally, if GTPase activation is contingent upon dynamin self-association on a multivalent matrix, then our data suggest that dynamin molecules align themselves on membranes via phospholipid-PH domain interactions. DISCUSSION Many characteristics of the regulation of dynamin activity have been described. The general features of this regulation, as they are currently understood, include a low basal GTPase of the unassembled enzyme, the ability to self-associate on an appropriate matrix, and cooperative autoactivation as a consequence of suitable alignment on the matrix. The PRD has been thought to account for interactions with GTPase regulators and for mediating self-assembly and autoactivation through these interactions. Instead, our findings indicate that the PRD is not required for all of the regulatory properties listed above: (i) Removal of the dynamin I PRD by limited proteolysis results in a 10-fold increase in its basal GTPase activity; (ii) dynamin lacking the PRD is activated by phospholipids; and (iii) the PRD is not required for the cooperative increase in GTPase activity when activated by PI(4,5)P 2 . If increases in specific activity reflect dynamin self-association, then it appears that the PRD is not required for PIP 2 -driven polymerization. In the presence of PI(4,5)P 2 , stimulation of GTPase activities of intact dynamin and dynamin lacking the PRD is not affected by physiological ionic strength. In contrast, two of the PRD-mediated interactions, i.e. those with microtubules and PS vesicles, occur only at low ionic strength. The interaction between the PRD and Grb2 is resistant to physiological salt concentration, but Grb2-stimulated GTPase activity is relatively low compared with maximal microtubule-or lipid-stimulated activities. Salim et al. (16) were the first to demonstrate that activation of dynamin I GTPase by phosphoinositides involves regions of dynamin outside the PRD. They showed that dynamin I mutants with deleted PH domains expressed Grb2-stimulated GTPase activity but were not activable by PI(4,5)P 2 . Our observation that D746 and papain digests of dynamin I can be activated by phospholipids despite their lack of PRDs suggests that interactions with the PH domains are both necessary and sufficient for enzyme activation. Although the mechanism of GTPase activation remains to be established, it is likely that dynamin self-association, promoted by phospholipid binding to the PH domains, is involved. However, a more direct effect of phospholipids on catalysis cannot be ruled out.
Rat dynamins I and II have an overall sequence identity of 79%, but their PRDs are only 46% identical (2). However, the PRDs of the two forms of dynamin have nearly the same isoelectric point (ϳ12.5) and percentage of prolines (ϳ30%). This similarity of composition most likely accounts for the almost identical interactions of dynamins I and II with their PRDbinding activators, Grb2 and microtubules. The PH domains of the two dynamins have 84% sequence identity, explaining the similar activation properties observed in the presence of phospholipid vesicles, which may interact with both the PH domains and the PRDs. Aside from the approximately 10-fold higher basal activity of dynamin II, the only other physical difference we detected between dynamins I and II is the greater tendency of dynamin II to aggregate in the absence of activators (not shown). However, it should be noted that dynamin I was purified from bovine brain, whereas dynamin II is a recombinant protein with a His 6 tag at the carboxyl terminus. Native dynamin II has not yet been isolated from cells or tissues.
Removal of the dynamin I PRD by papain cleavage results in a 10-fold increase in basal GTPase activity. This finding, which disagrees with an earlier observation (11), suggests that the PRD is inhibitory in native dynamin, perhaps being displaced in the presence of PRD-binding activators like Grb2. We note that the maximal activity obtained by proteolytic cleavage of dynamin I is approximately 25 min Ϫ1 , close to the values of about 50 min Ϫ1 for Grb2-stimulated GTPase and for GTPase stimulated by antibodies raised against the PRD (12). In contrast, activators that promote dynamin alignment (e.g. microtubules or phospholipid vesicles), stimulate GTPase activity to more than 200 min Ϫ1 .
It is widely believed that the authentic in vivo regulators of dynamin GTPase activity remain to be identified (2). As stated above, microtubules and PS vesicles, the first molecules shown to stimulate dynamin activity in vitro, do not interact with dynamin at physiological ionic strength (7). Moreover, dynamin does not colocalize with microtubules in cells (27), nor are the early steps of endocytosis believed to involve microtubule-dependent motors (1)(2)(3)(4). Although Grb2 binds to dynamin at high ionic strength (Fig. 6A), it is a relatively poor activator of GTPase activity, only slightly better than anti-dynamin antibodies (13). In this paper we demonstrate that PI(4,5)P 2 vesicles are potent stimulators of dynamin GTPase activity, even at physiological ionic strength. Because phosphoinositides are subject to multiple modifications by kinases and phosphatases, they represent versatile binding partners for proteins, such as dynamin, which have been implicated in membrane trafficking (28). However, phospholipids alone probably cannot account for specific targeting of dynamin to the coated pits, a function more suitable for dynamin-binding proteins such as Grb2 (29), amphiphysin (30), or AP2 (31). Therefore, we propose that dynamin is recruited to the coated pit by these proteins, many of which interact with the PRD, and that GTPase activity at the coated pit is stimulated by phosphoinositides, primarily via interactions with the PH domain. The validity of this model is currently being investigated in our laboratory.