Src-dependent tyrosine phosphorylation regulates dynamin self-assembly and ligand-induced endocytosis of the epidermal growth factor receptor.

Endocytosis of ligand-activated receptors requires dynamin-mediated GTP hydrolysis, which is regulated by dynamin self-assembly. Here, we demonstrate that phosphorylation of dynamin I by c-Src induces its self-assembly and increases its GTPase activity. Electron microscopic analyses reveal that tyrosine-phosphorylated dynamin I spontaneously self-assembles into large stacks of rings. Tyrosine 597 was identified as being phosphorylated both in vitro and in cultured cells following epidermal growth factor receptor stimulation. The replacement of tyrosine 597 with phenylalanine impairs Src kinase-induced dynamin I self-assembly and GTPase activity in vitro. Expression of Y597F dynamin I in cells attenuates agonist-driven epidermal growth factor receptor internalization. Thus, c-Src-mediated tyrosine phosphorylation is required for the function of dynamin in ligand-induced signaling receptor internalization.

Vital cellular responses including cell metabolism, proliferation, and differentiation are regulated by specific interactions between extracellular ligands and their plasma membraneanchored receptors. Accessibility of ligand to receptor, is therefore, an important facet of biological regulation. Receptor expression on the cell surface is dictated, at least in part, by vesicle trafficking via regulation of receptor internalization. Receptor internalization (also termed endocytosis) is implicated in receptor resensitization, down-regulation, and more recently signal transduction (1,2). Endocytosis of G proteincoupled receptors and receptor tyrosine kinases is dependent on the invagination, constriction, and fission of vesicles (clathrin-coated and caveolae) from the plasma membrane into the cytosol. Dynamin, a large GTPase, plays a crucial role in these steps of receptor-mediated endocytosis (3,4).
The GTPase activity of dynamin is stimulated 5-10-fold over basal levels by self-assembly (5). In the native state, dynamin exists as a homotetramer (6). At low ionic strength conditions (7), or in the presence of GDP and ␥-phosphate analogues at physiological salt conditions (8), concentrated dynamin tetra-mers spontaneously self-assemble into spiral stacks of rings, similar to the structure which appears as an electron dense "collar" around the necks of endocytic pits observed at synaptic nerve terminals (9,10). Purified dynamin also assembles onto phospholipid liposomes to generate dynamin-coated helical tubes that constrict and vesiculate upon GTP addition (11,12). On the other hand, dynamin-decorated phospholipid nanotubes undergo nucleotide-dependent conformational changes causing extension in pitch along the dynamin helix without constriction or fragmentation (13,14).
Recently, the GTPase effector domain of dynamin was shown to play a role in self-assembly and subsequent increase in GTPase activity (6,15). The GTPase effector domain has an intrinsic GTPase activating protein activity and interacts with the N-terminal GTPase domain to stimulate GTP hydrolysis. Although the molecular details governing the action of dynamin in the fission of vesicles from the plasma membrane have been controversial, it is, nonetheless, clear that GTP binding and hydrolysis play critical roles (3,4).
Tyrosine phosphorylation plays an important, if poorly understood, role in the internalization of cell surface receptors. Exposure of cells to tyrosine kinase inhibitors profoundly attenuates B cell receptor (16) and asialoglycoprotein receptor (17) endocytosis. Furthermore, inhibition of the non-receptor tyrosine kinase c-Src attenuates stem cell factor-induced c-Kit internalization in hematopoietic cells (18), antibody-induced internalization of neural cell adhesion molecule L1 in neuroblastoma cells (19), and EGFR 1 endocytosis (20). Conversely, overexpression of c-Src causes an increase in EGFR internalization rate following EGF treatment (21).
Some components of the cellular endocytic machinery have been shown to undergo regulated tyrosine phosphorylation. In the case of EGFR internalization, one such target is clathrin. Stimulation with EGF increases c-Src-mediated tyrosine phosphorylation of clathrin, which regulates clathrin translocation to the plasma membrane (20). Another target is dynamin, which can directly interact with c-Src (22) and becomes tyrosine-phosphorylated in response to insulin (23) and lysophosphatidic acid (24) stimulation. The c-Src-mediated tyrosine phosphorylation of dynamin is required for agonist-induced endocytosis of ␤ 2 -adrenergic (25) and M1 muscarinic acetylcholine (26) receptors. However, unlike clathrin, the mechanisms whereby phosphorylation of dynamin regulates receptor-mediated endocytosis are unclear. Here, we show that tyrosine phosphorylation of dynamin induces its self-assembly and subsequent increase in GTPase activity, and is required for agonist-induced EGFR internalization.

EXPERIMENTAL PROCEDURES
Expression Plasmids-Expression vectors for wild type and K44A dynamin have been described before (25). To generate single tyrosine mutants (Y231F, Y354F, and Y597F), each tyrosine residue of the wild type or K44A dynamin I was mutated to phenylalanine by overlapping polymerase chain reactions (UAC or UAU (Y) 3 UUC or UUU (F)). The double-tyrosine mutant (Y231F/Y597F) dynamin was constructed by recombination of the two single tyrosine mutant constructs. All dynamin I cDNAs were subcloned in pcDNA3 expression vector. Baculovirus transfer vector containing dynamin I cDNA (wild type or Y231F/ Y597F) was constructed in EcoRI/NotI sites of pVL1393 plasmid. The plasmid encoding the FLAG epitope-tagged EGFR was generated by polymerase chain reactions. The FLAG epitope coding sequences were introduced at the 5Ј end of EGFR cDNA in pBK expression vector.
Cell Culture and Transfection-LipofectAMINE and tissue culture reagents were from Invitrogen. COS-7 cells were maintained in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum and 100 g ml Ϫ1 gentamicin at 37°C in a humidified 5% CO 2 atmosphere. 70 -80% confluent COS-7 cells in 100-mm plates were transfected with a total of 7 g of DNA and 35 l of LipofectAMINE according to the manufacturer's instructions. One day after transfection, cells were detached by trypsin and seeded in either two 100-mm plates (for immunoprecipitation) or 24-well dishes (for enzyme-linked immunosorbent assay). All assays were performed ϳ60 h after transfection.
Dynamin Immunoprecipitation and Immunoblotting-Serumstarved (overnight) dynamin-expressing cells in 100-mm plates were pretreated with the appropriate concentrations of chemicals as indicated in the figure legends. The Src kinase inhibitor PP2 and the EGFR kinase inhibitor tyrphostin AG1478 were from Calbiochem. Sodium orthovanadate (Na 3 VO 4 ) was from Sigma. Cells were stimulated with 10 ng ml Ϫ1 EGF (Calbiochem) at 37°C for the indicated times, washed once with ice-cold phosphate-buffered saline, lysed in 1 ml of glycerol lysis buffer (5 mM HEPES, pH 7.4, 250 mM NaCl, 10% (v/v) glycerol, 0.5% Nonidet P-40, 2 mM EDTA, protease inhibitor tablets (Roche Molecular Biochemicals), 1 mM phenylmethylsulfonyl fluoride, 100 M Na 3 VO 4 ) on ice, and clarified by centrifugation. Cell lysates were mixed with 2 g of a monoclonal anti-dynamin antibody (Hudy-1; Upstate Biotechnology) and 70 l of 30% slurry of Protein G plus/Protein Aagarose beads (Oncogene) and rotated for 4 h at 4°C. Immune complexes were washed three times with ice-cold glycerol lysis buffer containing 0.1% (w/v) SDS and denatured in Laemmli sample buffer followed by boiling. Immunoprecipitated proteins were resolved on 4 -20% SDS-polyacrylamide gels (Invitrogen) and transferred to nitrocellulose membranes (Bio-Rad), immunoblotted with a 1:2,000 dilution of a horseradish peroxidase-conjugated anti-phosphotyrosine antibody (PY20H; Transduction Laboratories) to detect tyrosine phosphorylation signals. Before immunoprecipitation, 50 l of the cell lysate was aliquoted to provide samples for dynamin expression determined by immunoblotting with a 1:2,000 dilution of an anti-dynamin antibody (Hudy-1) followed by a 1:5,000 dilution of a horseradish peroxidaseconjugated anti-mouse secondary antibody (Jackson ImmunoResearch). Proteins were visualized using Supersignal chemiluminescence reagent (Pierce) and quantified using Fluor-S MultiImager (Bio-Rad).
Dynamin Isolation Using GST-Grb2 Affinity Binding-Isolation of endogenous dynamin II for analysis of its tyrosine phosphorylation content was performed by affinity purification with full-length GST-Grb2 fusion protein as described previously (24,25). Briefly, serumstarved cells in 100-mm plates were stimulated with 10 ng ml Ϫ1 EGF at 37°C for the indicated times in the presence of 100 M Na 3 VO 4 , washed once with ice-cold phosphate-buffered saline, and lysed in 1 ml of glycerol lysis buffer on ice. Cell lysates were clarified by centrifugation and incubated with the GST-Grb2 fusion protein (10 ng for 1 g of cell lysates), immobilized on glutathione-conjugated agarose beads, for 3 h at 4°C. Protein complexes were extensively washed with ice-cold glycerol lysis buffer containing 0.5 M NaCl and 0.1% (w/v) SDS, resuspended in Laemmli sample buffer, and resolved on SDS-polyacrylamide gels. Tyrosine phosphorylation of dynamin II was detected by immunoblotting as described above. The amount of isolated dynamin II was determined by immunoblotting with a 1:250 dilution of an anti-dynamin II antibody (Transduction Laboratories).
Sequestration of EGF Receptor-Agonist-induced internalization of FLAG epitope-tagged EGFR was determined by enzyme-linked immu-nosorbent assay as described (27). COS-7 cells transiently expressing FLAG-EGFR in 24-well dishes were washed once with phosphate-buffered saline and incubated with Dulbecco's modified Eagle's medium containing EGF (10 ng ml Ϫ1 ) for 30 min at 37°C. All subsequent steps were performed at room temperature. Stimulation was terminated by removing media and fixing the cells with 3.7% formaldehyde in TBS (20 mM Tris, pH 7.5, 150 mM NaCl) for 10 min. Cells were washed three times with TBS and incubated in a blocking solution (TBS containing 1% bovine serum albumin) for 45 min. A M1 anti-FLAG monoclonal antibody (Sigma) was added to the cells at a dilution of 1:1,000 in the blocking solution supplemented with 1 mM CaCl 2 for 1 h. After three washes with TBS, cells were reblocked for 15 min and incubated with an alkaline phosphatase-conjugated goat anti-mouse secondary antibody (Sigma) diluted 1:1,000 in the blocking solution for 1 h. Cells were washed three times with TBS, and 200 l of a colorimetric alkaline phosphatase substrate (Bio-Rad) was added to each well. When adequate color change was achieved, 100-l reaction solutions were transferred to 96-well plates containing 100 l of 0.4 M NaOH, and absorbance was determined using a microplate reader (Bio-Rad) at 405 nm. Background signal was concurrently determined using empty vectortransfected cells and subtracted from each of the values for the receptor-transfected cells. Receptor sequestration was defined as the fraction of total cell surface receptors that were removed from the plasma membrane following agonist treatment and thus were not accessible to antibody from outside of the cell.
Purification of Recombinant Rat Dynamin I-Dynamin expression baculoviruses were generated by co-transfection of recombinant baculovirus transfer vector (pVL1393), containing either wild type or Y231F/ Y597F rat dynamin I cDNA, together with BaculoGold DNA using the BaculoGold Transfection Kit (BD PharMingen). Purification of dynamin was carried out essentially as described (5). Briefly, 3 liters of suspension culture of Sf9 cells (1 ϫ 10 6 cells ml Ϫ1 ) were infected with high titer dynamin I-encoding recombinant virus stocks and harvested 48 h later by centrifugation at 2,500 ϫ g for 15 min. Pellets were washed in ice-cold phosphate-buffered saline and resuspended in 50 ml of icecold HCB (20 mM HEPES, pH 7.0, 2 mM EGTA, 1 mM MgCl 2 , 1 mM dithiothreitol, protease inhibitor tablets (Roche Molecular Biochemicals), and 1 mM phenylmethylsulfonyl fluoride) containing 100 mM NaCl (HCB100). Cells were disrupted by ultrasound sonication (7ϫ 10-s pulses on ice), diluted 1:1 with HCB containing 50 mM NaCl (HCB50) and clarified by centrifugation at 50,000 rpm in TFT 50.38 rotor (Sorvall) for 1 h. Ammonium sulfate ((NH 4 ) 2 SO 4 ) was slowly added to the supernatant fraction up to 30% saturation, and the mixture was stirred for 30 min at 4°C followed by centrifugation for 15 min at 10,000 ϫ g. Pellets were resuspended in 50 ml of HCB50 using a loose fitting Dounce homogenizer and centrifuged again for 10 min at 10,000 ϫ g. Solubilized 30% (NH 4 ) 2 SO 4 was applied to a 5-ml Q-Sepharose anion exchange column (Amersham Biosciences) pre-equilibrated with HCB50. The column was washed with 100 ml of HCB100, and dynamin was then eluted with HCB containing 150 mM NaCl (HCB150). 2-ml fractions were collected at 1 ml min Ϫ1 , and dynamin-containing fractions, identified by immunoblotting, were combined. Purified dynamin in HCB150 was concentrated up to 2 mg ml Ϫ1 using Vivaspin 20 (Vivascience). Protein assays and SDS-PAGE were used to assess protein yield and purity, which was greater than 90% as judged by Coomassie Blue staining.
In Vitro Dynamin Phosphorylation-Purified c-Src (recombinant human c-Src (⌬N85), lacking the first 85 amino acid residues) was the kind gift of Y-C. Ma (28) and M. J. Eck (29). Purified recombinant dynamin I and c-Src were mixed in 50 l of kinase reaction buffer (10 mM Pipes, pH 7.0, 10 mM MnCl 2 , 50 mM NaCl, 100 M ATP, 10 Ci of [␥-32 P]ATP) on ice. Reactions were performed at either 37°C or room temperature for the indicated times and terminated by addition of Laemmli sample buffer followed by 5 min boiling. Phosphorylated dynamin was resolved by SDS-PAGE, visualized by autoradiography, and quantified using a Storm PhosphorImager (Amersham Biosciences) to calculate phosphorylation stoichiometry.
In Vitro Dynamin GTPase Assay-Dynamin GTPase activity was determined by measuring the release of 32 P i from [␥-32 P]GTP-dynamin as described (30). Purified recombinant dynamin I in kinase reaction buffer (routinely 2 g of dynamin in 12.5-l solution) was added to a final volume of 75 l of GTPase assay buffer (20 mM HEPES, pH 7.0, 10 mM MgCl 2 ) on ice. Reactions were initiated by addition of 25 l of 1 mM [␥-32 P]GTP mixture (ϳ200 cpm pmol Ϫ1 ) in GTPase assay buffer followed by incubation at 30°C for the indicated times. The reactions were terminated by addition of 1 ml of isobutyl alcohol/benzene (v/v, 1:1) and 0.25 ml of 4% tungstosilic acid in 3 N H 2 SO 4 followed by brief mixing. Next, 0.3 ml of 10% ammonium molybdate was added followed by vigorous vortexing and brief centrifugation. 500 l of the aqueous phase solution, containing released 32 P i from [␥-32 P]GTP, were counted using a ␤-counter (Packard).
In Vitro Dynamin Assembly Assay-Dynamin self-assembly was monitored by a simple sedimentation assay (5). Purified recombinant dynamin I (4 g in 25 l of kinase reaction buffer) was diluted into 125 l of GTPase assay buffer on ice and mixed with 50 l of 1 mM GTP in GTPase assay buffer, yielding a final concentration of 20 ng ml Ϫ1 dynamin. Mixtures were incubated for 2 min at 30°C and then centrifuged at either 3,000 or 150,000 ϫ g for 10 min. Pellets were dissolved in Laemmli sample buffer and resolved by SDS-PAGE. Dynamin in the pellets was visualized by Coomassie Blue staining and quantified using Fluor-S MultiImager (Bio-Rad).
Negative Staining Electron Microscopy-Purified recombinant dynamin I was resuspended in phosphorylation reaction buffer as described. The concentrations of salt and dynamin in the reaction mixture were 50 M and 0.16 mg ml Ϫ1 , respectively. The mixture was adsorbed to carbon-coated EM grids that had been glow-discharged using a Hummer X glow discharge unit (Technics West Inc.) just prior to use as described (31), and negatively stained with 2% aqueous uranyl acetate before air-drying. Images were obtained using a Philips 301 electron microscope at 80 kV with 50-m objective aperture.

EGFR Stimulation Induces Tyrosine Phosphorylation of Dy-
namin-c-Src-mediated tyrosine phosphorylation of clathrin (20) and dynamin (25) are required for EGFR and ␤ 2 -AR inter-nalization, respectively. However, the involvement of tyrosine phosphorylation of dynamin in the process of EGFR endocytosis has not been reported. To test this possibility, we examined EGF-induced tyrosine phosphorylation of dynamin I transiently expressed in COS-7 cells after pretreatment with the tyrosine phosphatase inhibitor sodium orthovanadate. Stimulation with EGF caused a 5-10-fold increase in the tyrosine phosphorylation of dynamin (Fig. 1A). Interestingly, we found that the tyrosine phosphate content of K44A dynamin was 2-3-fold higher than that of wild type dynamin (Fig. 1A). After targeting to the plasma membrane, K44A dynamin remains there because of its impaired ability to bind and hydrolyze GTP (32,33). This may suggest that tyrosine phosphorylation of dynamin was stabilized at the plasma membrane.
Using K44A dynamin I transiently expressed in COS-7 cells, we determined the time course of its EGF-induced tyrosine phosphorylation. In the presence of sodium orthovanadate, tyrosine phosphorylation of K44A dynamin was significantly increased within 5 min and reached a plateau 30 min after EGF stimulation (Fig. 1B). The signal of K44A dynamin tyrosine phosphorylation was attenuated by the EGFR kinase inhibitor AG1478 as well as the Src kinase inhibitor PP2 (Fig. 1C), indicating that EGF-induced tyrosine phosphorylation of dy-FIG. 1. EGFR-regulated tyrosine phosphorylation of transiently expressed dynamin I. A, EGF promotes phosphorylation of wild type and K44A dynamin. COS-7 cells transiently expressing wild type or K44A dynamin were serum-starved overnight and pretreated with 100 M Na 3 VO 4 for 1 h followed by stimulation with EGF (10 ng ml Ϫ1 ) for an additional 1 h. Dynamin immunoprecipitates (IP) were resolved by SDS-PAGE, transferred to nitrocellulose filters, and immunoblotted (IB) with anti-phosphotyrosine (pTyr) antibody (PY20H). Values shown are expressed as percent of the EGF-induced tyrosine phosphorylation signal of K44A dynamin. B, time-dependent tyrosine phosphorylation of K44A dynamin by EGF. COS-7 cells expressing K44A dynamin were serum-starved, exposed to Na 3 VO 4 for a total of 2 h, and treated with EGF at 37°C for the indicated times. Tyrosine phosphorylation content of K44A dynamin was determined as above. Values are presented as percent of maximal signal obtained after 2 h stimulation. C, EGF-regulated tyrosine phosphorylation of dynamin involves Src kinase. Serum-starved K44A dynaminexpressing cells were treated with Na 3 VO 4 for 30 min before adding either PP2 (5 M) or AG1478 (250 nM). After an additional 30-min incubation, cells were stimulated with EGF for 1 h and analyzed for dynamin tyrosine phosphorylation. Data shown are expressed as percent of EGF-induced tyrosine phosphorylation of K44A dynamin from control cells, treated with the vehicle dimethyl sulfoxide only. A representative immunoblot is shown on the right of each panel. In each box, the lower panel shows total dynamin expression in lysates from the corresponding sample. Each data point represents the mean Ϯ S.E. from at least three independent experiments.
namin was EGFR-and Src kinase activity-dependent.
Because the data shown in Fig. 1 were obtained using transient expression of exogenous dynamin I, we confirmed these results by examining the EGF-induced tyrosine phosphorylation of endogenous dynamin. Endogenous dynamin II was isolated using GST-Grb2 affinity binding (24,25) from COS-7 cells after stimulation with EGF for the indicated times ( Fig. 2A), following pretreatment with sodium orthovanadate. Tyrosine phosphorylation of endogenous dynamin II was significantly increased within 1 min after stimulation, similar to the previously reported isoproterenol-stimulated signal (25). Unlike the situation for K44A dynamin I (Fig. 1B), the EGF-induced tyrosine phosphorylation of endogenous dynamin II was transient; the signal was significantly reduced 30 min after stimulation. Maximal EGF-induced tyrosine phosphorylation of endogenous dynamin II (10 min stimulation) was attenuated by PP2 and AG1478 (Fig. 2B), like that of transiently expressed dynamin I (Fig. 1C). Taken together, these results resemble our previous finding that ␤ 2 -AR stimulation induces c-Src-mediated tyrosine phosphorylation of dynamin (25), and suggest that this event is a general cellular response to receptor activation.
Dynamin Is a Direct Substrate of c-Src Kinase-Available evidence demonstrates that c-Src directly interacts with dynamin (22) and that c-Src kinase activity is required for tyrosine phosphorylation of dynamin (25). To determine whether c-Src can directly phosphorylate dynamin, we employed an in vitro kinase assay using purified proteins. Baculovirus encoding rat dynamin I was generated and used to infect Sf9 cells. The recombinant dynamin protein was purified to homogeneity using ammonium sulfate fractionation and ion-exchange Q-Sepharose chromatography. Purified dynamin was competent in GTP hydrolysis and was not tyrosine phosphorylated (data not shown). As shown in Fig. 3A, dynamin could serve as a direct substrate of c-Src. Maximum stoichiometry of phosphorylation achieved was calculated to be 0.39 mol of P i /mol of dynamin. Because the basic unit of the dynamin oligomer is a homotetramer (6), it is reasonable to assume that only a fraction of dynamin molecules are phosphorylated by c-Src in vitro.
Previously, two dynamin tyrosine residues (Tyr 231 and Tyr 597 ) were identified as being phosphorylated following ␤ 2 -AR stimulation in HEK293 cells (25). To determine whether the tyrosine-phosphorylated residues by c-Src in vitro are the same as those identified in cultured cells, we compared the in vitro phosphorylation pattern of recombinant wild type and Y231F/Y597F dynamin proteins. The mutant protein was impaired by ϳ40 -50% in its tyrosine phosphorylation compared with wild type protein (Fig. 3B). Mass spectrometric analysis of trypsin-digested fragments of in vitro c-Src-phosphorylated dynamin and immunoblotting of the phosphorylated dynamin with an antibody generated specifically against phospho-Tyr 597 revealed phosphorylation of two residues, Tyr 354 and Tyr 597 . Tyr 597 was previously identified to be phosphorylated in response to ␤ 2 -AR activation in cultured cells, whereas Tyr 354 was not detected in the prior in vivo analysis (25) (Fig. 3C). Tyr 231 , another residue that becomes phosphorylated following ␤ 2 -AR stimulation in cultured cells (25), was not found to be phosphorylated by c-Src in vitro, suggesting that Tyr 231 could be phosphorylated by another kinase rather than c-Src. Together, these data suggest the relevance of c-Src-mediated phosphorylation of Tyr 597 in the in vivo function of dynamin.

FIG. 2. EGF-induced tyrosine phosphorylation of endogenous dynamin II.
A, time-dependent tyrosine phosphorylation of endogenous dynamin II by EGF. COS-7 cells were serum-starved overnight, exposed to Na 3 VO 4 for a total of 2 h, and then stimulated with EGF at 37°C for the indicated times. Endogenous dynamin II was isolated using GST-Grb2 affinity beads and the tyrosine phosphorylation content of isolated dynamin II was determined by immunoblotting (IB) as outlined under Fig. 1. Values are presented as EGF-induced fold increases in tyrosine phosphorylation over unstimulated samples. B, EGF-induced tyrosine phosphorylation of endogenous dynamin II requires Src activity. Serum-starved cells were treated with Na 3 VO 4 for a total of 2 h. PP2 (5 M) or AG1478 (250 nM) were added 30 min prior to stimulation with EGF for 10 min. The tyrosine phosphorylation of endogenous dynamin II was determined as described above. Data shown are expressed as EGF-induced fold increases in tyrosine phosphorylation over unstimulated samples. A representative immunoblot is shown on the right of each panel. In each box, the lower panel shows the amount of dynamin II isolated on the GST-Grb2 beads in the corresponding samples. Each data point represents the mean Ϯ S.E. from five independent experiments. pTyr, phosphotyrosine.

EGF-induced Tyrosine Phosphorylation of Dynamin Is
Required for EGFR Internalization-To identify the physiological significance of the tyrosine phosphorylation sites in dynamin following receptor activation, we examined EGF-induced tyrosine phosphorylation content of the single tyrosine mutant dynamins, Y354F and Y597F, as well as the double tyrosine mutant dynamin Y231F/Y597F, transiently expressed in COS-7 cells. In addition, we examined the tyrosine phosphorylation of the K44A dynamin version of each tyrosine mutant, because K44A dynamin showed a higher tyrosine phosphorylation signal compared with wild type dynamin following EGF stimulation (Fig. 1A). EGF stimulation induced tyrosine phosphorylation of both wild type and K44A dynamin (Fig. 4A), consistent with the data shown in Fig. 1A. Replacement of Tyr 354 with Phe showed no substantial effect on the tyrosine phosphorylation level of either wild type or K44A dynamin in response to EGF stimulation (Fig. 4, A and B). In contrast, the Y597F mutation resulted in 50 -70% reduction in the EGFinduced dynamin tyrosine phosphorylation signal. Furthermore, replacement of both Tyr 231 and Tyr 597 with Phe caused a more dramatic attenuation (75-90%) in the tyrosine phosphorylation of dynamin than the Y597F single mutation following EGF stimulation. These results suggest that EGF stimulation induces phosphorylation of both Tyr 597 and Tyr 231 , but not Tyr 354 , in COS-7 cells.
To establish the physiologic relevance of dynamin tyrosine phosphorylation in signaling receptor endocytosis, we tested the effect of the tyrosine mutant proteins on agonist-induced EGFR internalization in COS-7 cells. Expression of Y597F dynamin caused 60% reduction in EGF-induced EGFR internalization, whereas expression of the Y354F dynamin had no effect (Fig. 4C), as would be predicted from the lack of Tyr 354 phosphorylation following EGF stimulation. Furthermore, expression of the Y231F/Y597F dynamin exhibited additional attenuation (80 -90%) of EGF-induced EGFR internalization, which was equivalent to the degree of inhibition observed with GTPase-deficient K44A dynamin. Taken together, these results demonstrate that phosphorylation of both Tyr 597 and Tyr 231 was required for EGFR internalization following EGF stimulation, whereas phosphorylation of Tyr 354 was not involved in EGFR endocytosis.
Tyrosine Phosphorylation of Dynamin Potentiates Its GTPase Activity in Vitro-Tyrosine phosphorylation of dynamin ( . Reactions were carried out at 37°C for 30 min, and products were then visualized by autoradiography. B, c-Src phosphorylates wild type and Y231F/Y597F dynamin proteins in vitro. Purified dynamin proteins were mixed with c-Src (molar ratio of c-Src/dynamin ϭ 1/5) in kinase reaction buffer at room temperature for the indicated times. At the end of each time point, a fixed portion of the reaction mixture was taken and analyzed for phosphorylation content by autoradiography. Control samples were similarly prepared, but without c-Src. Note that purified dynamin was minimally phosphorylated even in the absence of added c-Src (A and B), presumably by a kinase that was co-purified with dynamin. Because this phosphorylation was not c-Src-dependent, this basal phosphorylation signal was subtracted to calculate stoichiometry of the c-Src-dependent tyrosine phosphorylation of dynamin. Stoichiometry of phosphorylation of both wild type and Y231F/Y597F dynamin was obtained from two independent experiments (left panel). A representative autoradiograph is shown on the right. C, c-Src phosphorylates dynamin I on Tyr 354 and Tyr 597 . Schematic summary of dynamin structure and identification of phosphorylated tyrosine residues in vitro by c-Src and in cells following receptor activation are shown. GED, GTPase effector domain; PRD, proline-rich domain. required for ligand-induced endocytosis of signaling receptors. To test whether c-Src-mediated tyrosine phosphorylation regulates dynamin GTPase activity, we compared the rate of GTP hydrolysis by in vitro tyrosine-phosphorylated and unphosphorylated recombinant dynamin proteins. Fig. 5A shows that preincubation of dynamin with c-Src and ATP induced a 4 -5fold increase in steady-state GTP hydrolysis rate compared with the unphosphorylated control protein. Assays performed using dynamin alone, dynamin mixed with ATP or c-Src, or dynamin mixed with non-hydrolyzable ATP analogues (App(NH)p or ATP␥S) and c-Src, all failed to stimulate GTPase activity of dynamin (Fig. 5A), demonstrating that tyrosine phosphorylation of dynamin is required for the enhanced GTP hydrolysis.
Our results demonstrate that the Tyr 597 residue of dynamin is phosphorylated by c-Src in vitro (Fig. 3), and that this phosphorylation regulates ligand-induced EGFR endocytosis (Fig.  4C), albeit by undefined mechanisms. To determine the role of Tyr 597 phosphorylation on the function of dynamin, we compared the GTPase activities of in vitro tyrosine-phosphorylated wild type and Y231F/Y597F recombinant dynamin proteins. Because Tyr 231 was not phosphorylated by c-Src in vitro, mutation of this residue (Y231F) would not be expected to influence the tyrosine phosphorylation-regulated GTPase activity of dynamin. After 2 min incubation with GTP at 30°C, Y231F/ Y597F dynamin showed ϳ40% reduction in its tyrosine phosphorylation-induced initial rate of GTP hydrolysis compared with the wild type protein (Fig. 5B). Unphosphorylated wild type and Y231F/Y597F dynamin had equally low basal GTPase activities. These results demonstrate that in vitro tyrosine phosphorylation of dynamin stimulates its GTPase activity, and that phosphorylation of Tyr 597 significantly controls this process.
Tyrosine-phosphorylated Dynamin Spontaneously Self-assembles to Generate Spiral Stacks of Interconnected Rings-At low ionic strength, purified dynamin self-assembles into structures composed of rings and spiral stacks, which are sedimentable by high-speed centrifugation (Ͼ100,000 ϫ g) (5,7). Because self-assembly of dynamin is known to regulate its GTP hydrolysis rate (5,6,15), we hypothesized that the increased GTPase activity of tyrosine-phosphorylated dynamin involves its enhanced self-assembly. Purified recombinant dynamin protein was phosphorylated, diluted in GTPase reaction buffer, and incubated at 30°C, exactly as described for the GTPase assay. Self-assembled dynamin was collected by sedimentation. Initial results, utilizing high-speed centrifugation (150,000 ϫ g for 10 min), showed no difference in self-assembly of in vitro tyrosine-phosphorylated or unphosphorylated dynamin proteins (Fig. 6A, small gel). However, when centrifugal force was reduced to 3,000 ϫ g, a striking difference was noted; dynamin incubated with c-Src and ATP was sedimentable at 3,000 ϫ g, whereas unphosphorylated proteins precipitated in trace amounts only (Fig. 6A).
To further establish a relationship among dynamin GTPase activity, self-assembly, and tyrosine phosphorylation, we examined the effect of tyrosine phosphorylation of dynamin on its self-assembly using wild type and Y231F/Y597F recombinant proteins that were phosphorylated by c-Src. Fig. 6B shows that the Y231F/Y597F dynamin has ϳ40% diminution in its tyrosine phosphorylation-promoted sedimentation compared with wild type protein, exactly paralleling the effect on GTPase activity (Fig. 5B). Unphosphorylated Y231F/Y597F dynamin showed the same extent of precipitation as wild type control. Taken together, these data demonstrate that tyrosine-phosphorylated dynamin polymerizes into large molecular mass struc- tures. Because dynamin self-assembly enhances its GTPase activity, these results also suggest that c-Src-mediated phosphorylation of dynamin regulates the GTPase activity by controlling self-assembly.
To confirm that tyrosine phosphorylation induces formation of real self-assembled structures of dynamin large enough to be precipitated at relatively low (3,000 ϫ g) centrifugal force, we employed negative-staining electron microscopy. As shown in Fig. 7A, clustered stacks (50 -300 nm length) of tightly interconnected dynamin rings were observed in solutions of tyrosine-phosphorylated samples obtained by incubation with c-Src and ATP. High magnification of a large stack showed a well ordered spiral array of dynamin rings (Fig. 7B). In con-trast, unphosphorylated control dynamin was composed of short curved structures, some of which developed partial rings and globules (Fig. 7C).
Helical stacks of dynamin have been observed upon dilution into low ionic strength buffer (Ͻ50 mM salt) (7) or in the presence of GDP and ␥-phosphate analogues at physiological salt conditions (8). However, those structures were prevalent only at high dynamin protein concentrations (Ͼ1 mg ml Ϫ1 ), whereas the tyrosine phosphorylation-induced assembled structures shown here were generated at much lower protein concentrations (0.16 mg ml Ϫ1 ). Furthermore, the helical stacks appeared tightly ordered (Fig. 7, A and B), similar to spirals obtained using a C terminus truncated 90-kDa fragment of dynamin, which was compactly arranged compared with intact dynamin protein in the presence of GDP and ␥-phosphate analogues at physiological salt conditions (8). Dimensions of the stacks (ϳ50 nm) were comparable not only to the structures obtained at high dynamin protein concentrations in vitro, but also to the collars observed at the neck of invaginated coated pits that accumulate at synaptic nerve terminals (9, 10). It has also been shown that helically coated phospholipid tubes generated from purified dynamin and liposomes in vitro have the same diameter (11). Taken together, these results demonstrate that tyrosine phosphorylation of dynamin induces its spontaneous self-assembly into large helical structures composed of interconnected rings. DISCUSSION Several models exist to describe the role of dynamin in the fission of vesicles from the plasma membrane to the cytosol, including dynamin being a pinchase (7,11,12), a poppase (13,14), or an adapter (15) to recruit downstream effectors, which mediate the fission step. Regardless of which model is correct, it is an indisputable fact that control of dynamin GTPase activity is the key for endocytic vesicle formation (3,4). Similarly, it is well established that dynamin self-assembles to form spiral structures on the neck of invaginated pits during endocytosis (9,10,33), which has been demonstrated to stimulate its GTPase activity (5,6,15). Although a certain conformation of GTP-bound dynamin was proposed to favor self-assembly (8), regulatory mechanisms inducing and stabilizing the assembled dynamin structures remain unclear. The present data show that c-Src-mediated tyrosine phosphorylation promotes significant increases in dynamin self-assembly into helically interconnected rings and also stimulates the rate of GTP hydrolysis, both of which are hallmarks for dynamin execution of vesicle budding from the plasma membrane.
Agonist stimulation triggers recruitment of c-Src and dynamin into the vicinity of activated receptor on the plasma membrane. Some receptors, such as EGFR (34) and ␤ 3 -AR (35), bind c-Src directly, whereas others, such as ␤ 2 -AR (36), form complexes with c-Src via bridging adapter proteins. Dynamin translocates to the plasma membrane by direct interaction with phospholipids (37) and adapter proteins, such as amphyphisin (3,38). Once in close proximity to each other on the plasma membrane, c-Src phosphorylates dynamin. In support of this hypothesis is our finding that K44A dynamin, which is competent to translocate to invaginated pits on the plasma membrane following receptor activation, but which does not recycle to the cytosol because of its impaired ability to bind and hydrolyze GTP (32,33), displays a significant increase in tyrosine phosphorylation compared with wild type dynamin (Fig.  1A). Furthermore, disruption of c-Src recruitment to activated ␤ 2 -AR inhibits tyrosine phosphorylation of dynamin (39). These data suggest that dephosphorylation of dynamin occurs in the cytosol following GTP hydrolysis-induced disassembly.
Phosphoamino acid analysis of in vitro c-Src-phosphorylated FIG. 5. Tyrosine phosphorylation of dynamin potentiates the rate of GTP hydrolysis. A, c-Src-catalyzed phosphorylation of dynamin increases its GTPase activity. Purified recombinant rat dynamin I was co-incubated with c-Src (molar ratio of c-Src/dynamin ϭ 1/5) and ATP for 6 h at room temperature to generate tyrosine-phosphorylated dynamin protein. Controls included incubation of dynamin alone (CN), or together with ATP, c-Src, or c-Src and the non-hydrolyzable ATP analogues App(NH)p and ATP␥S. At the end of reaction, dynamin (0.2 M) from each sample was incubated in GTPase assay buffer at 30°C for 15 min, and GTP hydrolysis rates were then determined as described under "Experimental Procedures." B, GTP hydrolysis by tyrosine-phosphorylated wild type and Y231F/Y597F dynamin proteins. Wild type and Y231F/Y597F-phosphorylated dynamin proteins were generated as outlined in A. Control samples (Ϫc-Src) were prepared by incubating each dynamin protein alone in kinase reaction buffer. The rate of GTP hydrolysis was determined after incubation in GTPase buffer at 30°C for 2 min. The asterisk indicates p Ͻ 0.05 versus tyrosine-phosphorylated wild type dynamin. Data represent mean Ϯ S.E. from five independent experiments done in triplicate (A) or duplicate (B). dynamin revealed two tyrosine residues, Tyr 354 and Tyr 597 , to be phosphorylated (Fig. 3C). However, in vivo analysis shows that Tyr 597 , but not Tyr 354 , becomes phosphorylated in response to EGF stimulation. Furthermore, phosphorylation of dynamin on Tyr 597 , but not Tyr 354 , appears to be required for EGF-induced EGFR internalization (Fig. 4). Tyr 597 in dynamin also needs to be phosphorylated to mediate ␤ 2 -AR internalization (25) and expression of Y231F/Y597F dynamin inhibits M1 muscarinic acetylcholine receptor internalization (26). Replacement of Tyr 597 with Phe reduces the tyrosine phosphorylationinduced GTPase activity (Fig. 5B) and self-assembly of dynamin (Fig. 6B), suggesting that phosphorylation of Tyr 597 regulates dynamin function in endocytosis of ligand-activated receptors by controlling self-assembly and subsequent GTP hydrolysis.
Does tyrosine phosphorylation of dynamin play a role in clathrin-mediated endocytosis of constitutively recycling nutrient receptors, similar to ligand-induced internalization of signaling receptors? Vallis et al. (40) showed that expression of bovine Y596F mutant dynamin I has no effect on transferrin uptake, suggesting that phosphorylation of Tyr 596 , which corresponds to the Tyr 597 residue of rat dynamin I, is not required for constitutive endocytosis of nutrient receptors. It is possible that the function of dynamin is regulated differently for constitutive endocytosis of nutrient receptors, in the absence of tyrosine phosphorylation. In support of this hypothesis, the existence of two functionally and biochemically distinct subpopulations of clathrin-coated pits that mediate agonist-regulated internalization of the ␤ 2 -AR and constitutive endocytosis of the transferrin receptor has been demonstrated (41). Alternatively, the tyrosine phosphorylation-induced assembled structure of dynamin may have an unidentified specific role for endocytosis of activated signaling receptors, which is not required for constitutive endocytosis of recycling receptors. In the case of EGFR endocytosis, one possibility is that tyrosine phosphorylation-induced dynamin self-assembly leads to recruitment of the ligand-activated receptor to the endocytic machinery. It has been suggested that biochemical differences between EGF and transferrin uptake mostly reflect specific requirements for the recruitment of the EGFR to coated pits (42). We also cannot exclude the possibility that the effect of Y597F dynamin on EGFR endocytosis may be indirect, such as by interfering with receptor activation, because expression of K44A dynamin was shown to alter high affinity binding of EGF to the receptor (43). Currently, however, what determines this specificity of tyrosine phosphorylation-regulated function of dynamin in signaling receptor internalization is not known.
How does tyrosine phosphorylation induce dynamin selfassembly? It is well established that phosphotyrosine residues primarily serve as docking sites for proteins containing Src FIG. 6. Tyrosine phosphorylation promotes dynamin self-assembly. A, c-Src-phosphorylated dynamin precipitates at low centrifugal force. Tyrosinephosphorylated dynamin and appropriate controls were prepared as outlined under Fig. 5A, and diluted in GTPase buffer to exactly mimic conditions used in the GTPase assay. After incubating at 30°C for 2 min, each dynamin sample was centrifuged at 3,000 ϫ g for 10 min. Pelleted dynamin was visualized by Coomassie Blue staining and quantified. Values shown are expressed as fold-increase over control (CN) in which dynamin was incubated alone during the in vitro phosphorylation reaction. Additional controls included the high speed centrifugation (150,000 ϫ g for 10 min) of dynamin proteins prepared in the absence or presence of c-Src with ATP. B, tyrosine phosphorylation-promoted self-assembly of wild type and Y231F/Y597F dynamin proteins. Wild-type and Y231F/Y597F dynamin proteins were subjected to in vitro phosphorylation by c-Src kinase, diluted in GTPase buffer, and incubated at 30°C for 2 min, exactly as described under Fig. 5B. In each sample, precipitated dynamin was visualized by Coomassie Blue staining and quantified after centrifugation at 3,000 ϫ g for 10 min. Values are expressed as fold increase over the wild type control (Ϫc-Src). The asterisk indicates p Ͻ 0.05 versus tyrosine-phosphorylated wild type dynamin. A representative gel is shown for each panel. Each data point represents mean Ϯ S.E. from five independent experiments. homology (SH) 2 or phosphotyrosine-binding domains (44,45). Indeed, some dynamin-binding proteins stimulate the GTPase activity of dynamin (3,38). However, our in vitro results using purified dynamin show that tyrosine phosphorylation per se increases the GTPase activity of dynamin. Additionally, dynamin itself does not contain SH2 or phosphotyrosine-binding domains, indicating a lack of potential intramolecular SH2 or phosphotyrosine-binding domain-mediated interactions in tyrosine-phosphorylated dynamin. Several lines of evidence have suggested that tyrosine phosphorylation of proteins may stimulate their enzymatic activity by inducing a local conformational change rather than by creating binding sites for other partners. For example, most protein-tyrosine kinases, including receptor tyrosine kinases and non-receptor tyrosine kinases such as Src, are activated by phosphorylation of tyrosine residues in the activation loop (45). The SH2 domain-containing tyrosine phosphatase SHP-2 is tyrosine-phosphorylated in growth factor-stimulated cells, which leads to its activation (46). Furthermore, it has been reported that c-Src phosphorylates the G␣ s , G␣ i , G␣ o , and G␣ q/11 GTPases (47,48). In in vitro reconstitution assays, tyrosine-phosphorylated G␣ s shows an increased rate of binding to GTP␥S as well as an increased rate of receptor-stimulated GTP hydrolysis (47). Together, the present results suggest the possibility that tyrosine phosphorylation may induce a conformational change of dynamin, which favors self-assembly.
Tyr 597 is well conserved among dynamin isoforms and resides in the PH domain. This domain is required for endocytosis (49,50), presumably to recruit dynamin to the plasma membrane via binding to phospholipids (37). The contribution of the PH domain to dynamin self-assembly is not clear. In vitro, the PH domain was shown either not to participate (51,52) or to even be a negative regulator (6, 53) of dynamindynamin interaction. However, it is not clear whether the dynamin proteins used in these studies were tyrosine phosphorylated. Interaction of the PH domain phosphorylated on Tyr 597 with other domains of dynamin remains to be determined. On the other hand, whereas phosphorylation of Tyr 354 in dynamin appears to be physiologically irrelevant, at least with regard to EGFR endocytosis, the in vitro results demonstrate that phosphorylation of this residue exerts a positive effect on dynamin self-assembly and GTPase activity. Thus, it will be of interest to test whether phosphorylation of Tyr 354 alters other functions of dynamin, such as rapid endocytosis of synaptic vesicles in nerve terminals (3,38), because the Tyr 354 residue is exclusively found in neuronal dynamin I.
In addition to being phosphorylated on tyrosine residues, neuronal dynamin I is also phosphorylated on a serine residue by protein kinase C in intact synaptosomes (54). In vitro, protein kinase C-mediated phosphorylation stimulates dynamin I GTPase activity (55), and increases binding to calcium (56), but blocks binding to phospholipids (57). Upon synaptic membrane depolarization, serine-phosphorylated dynamin I was dephosphorylated by the calcium-dependent phosphatase calcineurin (58), which was required for endocytosis in nerve terminals (59). Calcineurin-mediated dephosphorylation restores the ability of dynamin to bind phospholipids (57) and inhibits its GTPase activity (58) in vitro. Thus, it will be of interest to examine the relationship between tyrosine and serine phosphorylation of dynamin I in nerve terminals and their impact on endocytosis.
The effect of GTP binding on dynamin self-assembly is not entirely clear. In vitro, addition of GDP with ␥-phosphate analogues induces dynamin self-assembly to generate spiral stacks (8), and treatment of isolated nerve terminals with GTP␥S promotes tubular membrane invaginations decorated with electron-dense dynamin rings (10). However, the binding of guanine nucleotides GTP␥S, GTP, or GDP destabilizes assembled dynamin structures (5), and GTP binding-impaired K44A dynamin self-assembles, like wild type protein under low ionic strength conditions (5). Our data show that tyrosine phosphorylation controls dynamin self-assembly ( Fig. 6) even in the absence of guanine nucleotides (Fig. 7). Furthermore, stimulation with EGF promotes tyrosine phosphorylation of GTP binding-deficient K44A dynamin (Fig. 1), suggesting that nucleotide binding was not required for tyrosine phosphorylation. It is not clear at this juncture whether c-Src-regulated tyrosine phosphorylation exerts any effect on the affinity of dynamin to bind these nucleotides. Dynamin can also generate helically assembled structures around phospholipid liposomes (11,12) or nanotubes (13,14) in vitro, which were shown to undergo conformational changes in a nucleotide-dependent manner (11)(12)(13)(14). Furthermore, the recently resolved crystal structure of the dynamin GTPase domain (60) and three-dimensional reconstruction of dynamin in the constricted state (12) provides additional evidence for the requirement of dynamin self-assembly and conformational change in endocytic vesicle formation. Thus, it remains to be determined how tyrosine phosphorylation collaborates with the guanine nucleotide in the context of dynamin self-assembly and conformational change.
Together, our results establish the novel paradigm that posttranslational tyrosine phosphorylation of dynamin serves as a regulator of dynamin function in endocytosis of signaling receptors via control of self-assembly. Furthermore, the agonistdependent tyrosine phosphorylation of dynamin provides a Purified recombinant rat dynamin I was incubated in kinase reaction buffer with (A and B) or without (C) c-Src, as outlined under Fig. 5. The concentrations of dynamin and salt in the reaction buffer were 0.16 mg ml Ϫ1 and 50 mM, respectively. After incubation for 6 h at room temperature, each sample was adsorbed to carbon-coated EM grids and negatively stained as described under "Experimental Procedures." A, tyrosine-phosphorylated dynamin shows clustered large stacks of interconnected rings. B, a high magnification image of a large single spiral stack of assembled dynamin in which each ring was visibly preserved. C, unphosphorylated control dynamin consists of curved structures (large arrow), globules (small arrow), or partial rings (arrowheads). The images in A and C are shown at the same magnification. Scale bars, 100 nm. mechanism by which tyrosine kinase receptors regulate their own accessibility to external stimulation and, as a result, cellular response.