Dynamin 3 Is a Component of the Postsynapse, Where it Interacts with mglur5 and Homer

Transcription

1 Current Biology, Vol. 13, , March 18, 2003, 2003 Elsevier Science Ltd. All rights reserved. DOI /S (03) Dynamin 3 Is a Component of the Postsynapse, Where it Interacts with mglur5 and Homer Noah W. Gray, 1,2 Lawrence Fourgeaud, 3 Bing Huang, 2 Jing Chen, 2 Hong Cao, 2 Barbara J. Oswald, 2 Agnès Hémar, 3 and Mark A. McNiven 1,2, * integral role in synaptic vesicle recycling [4, 9]. We and others identified Dynamin 2 (Dyn2) as a ubiquitously expressed isoform that can mediate multiple endocytic processes [10, 11], liberate vesicles from the Golgi [12], 1 Molecular Neuroscience Program and Graduate and act as a modulator of the actin cytoskeleton [13 15]. School Dynamin 3 (Dyn3) was originally isolated as a testesspecific 2 Center for Basic Research in Digestive Diseases isoform [16] but was later found to be addition- Mayo Clinic ally expressed in brain, lung, and heart [3, 17]. During 200 First Street, S.W. brain development, Dyn3 is rapidly upregulated through- Rochester, Minnesota out synaptogenesis, suggesting a possible role for this 3 CNRS UMR 5091 isoform during synaptic development [17]. But unlike Institut François Magendie other dynamin isoforms, when expressed in epithelial Rue Camille Saint Saens cells, Dyn3 did not colocalize with standard endocytic Bordeaux Cedex markers, nor did it accumulate at clathrin-coated pits France [3]. To date, no study has identified a specific function for Dyn3 in any cell type. To define the roles of different dynamin isoforms in Summary neuronal tissue, we examined the developmental expression and localization of dynamin isoforms in dissociated The dynamins comprise a large family of mechanoenzymes rat hippocampal neurons (Figure S1). We found known to participate in membrane modeling Dyn1 and Dyn3 to be dynamically regulated at the mrna events [1, 2]. All three conventional dynamin genes and protein levels as well as differentially localized over (Dyn1, Dyn2, Dyn3) are expressed in mammalian brain time in culture, paralleling the patterns of known synapand produce more than 27 different dynamin proteins tic markers. These data suggested that dynamin iso- as a result of alternative splicing [3]. Past studies have forms may perform a role during synaptogenesis. Dyn2 suggested that Dyn1 participates in specialized neuus did not exhibit these same changes over time, causing ronal functions such as rapid synaptic vesicle recyhippocampal to focus on the other two isoforms. Using mature cling [4], while Dyn2 may mediate the conventional neurons, we next sought to further define clathrin-mediated uptake of surface receptors [5]. the localization of Dyn1 and Dyn3. All subsequent de- Currently, the distribution, expression, and function of scriptions were derived from neurons that were over 21 Dyn3 in neurons, or in any other cell type, are comantibodies days in vitro (DIV). Neurons were colabeled with Dyn1 pletely undefined. Here, we demonstrate that Dyn1 (Figure 1A) and either synaptotagmin (Figure and Dyn3 localize differentially in the synapse. Dyn1 1A ) or PSD-95 (data not shown). Both synaptic markers concentrates within the presynaptic compartment, labeled axonal varicosities, where the Dyn1 fluorescent while Dyn3 localizes to dendritic spine tips. Within the signal was the highest (arrows, Figure 1A ). To confirm postsynaptic density (PSD), we found Dyn3, but not the presynaptic antibody localization of Dyn1, Dyn1aa- Dyn1, to be part of a biochemically isolated complex GFP was transfected into neurons at 10 DIV, and cells comprised of Homer and metabotropic glutamate re- were processed for immunocytochemistry at 21 DIV. ceptors. Finally, although dominant-negative Dyn3 did Axonal varicosities running alongside dendrites were not seem to inhibit receptor endocytosis, overexpresconcentrations strongly labeled with Dyn1aa-GFP (Figure 1B), and these sion of a specific Dyn3 spliced variant in mature neusynapsin colocalized with the synaptic marker rons caused a marked remodeling of dendritic spines. (arrows, Figure 1B ). We next confirmed that These data suggest that Dyn3 is a postsynaptic dyfunctional the Dyn1aa-GFP in the presynaptic varicosities was namin and, like its binding partner Homer, plays a by challenging the transfected neurons to significant role in dendritic spine morphogenesis and take up the lipophilic dye FM4-64 upon stimulation. Norremodeling. mal synaptic vesicle recycling occurred after depolariza- tion in neurons expressing Dyn1aa-GFP (arrows, Figure Results and Discussion 1C), but cells expressing a mutant Dyn1 deficient in GTP hydrolysis (Dyn1aa(K44A)-GFP, [5]) failed to take up FM4-64 (arrows, Figure 1D). Thus, a GTPase-deficient Dynamin was initially proposed to function during syn- Dyn1 was capable of blocking synaptic vesicle recyaptic vesicle recycling, based on work utilizing the Drocling. Our immunofluorescence and dye uptake data sophila temperature-sensitive shibire protein, the fruitfly suggested that Dyn1 localized and functioned at the homolog of dynamin [6, 7]. In mammalian cells, there presynapse, but to confirm this at the ultrastructural are three distinct dynamin isoforms, each with a specific level, we fixed and processed adult rat forebrain synaptissue distribution. Dynamin 1 (Dyn1), originally isolated tosomes for immunogold electron microscopy. Antibodas a microtubule binding protein in brain [8], plays an ies against Dyn1 mainly labeled the presynaptic terminals of synaptosomes (Figures 1E and 1E ), and gold *Correspondence: mmcniven@mayo.edu particles were found around clusters of synaptic vesi-

2 Brief Communication 511 Figure 1. Dyn1 Is Enriched in Synaptic Boutons (A and A ) (A) Dyn1 antibodies stained the axons and axonal varicosities of mature neurons (see arrows). (A ) These varicosities were shown to be synaptic boutons, as demonstrated by colabeling with synaptotagmin (red, see arrows). (B and B ) Neurons transfected with Dyn1aa-GFP were labeled with antibodies against synapsin (red) and the dendrite-specific marker MAP2b (blue). (B) Dyn1aa-GFP trafficked into the axons and was sequestered in axonal varicosities. (B ) The addition of the synapsin channel confirmed that these varicosities were synaptic boutons, for the Dyn1aa-GFP and synapsin both label the same structures (see arrows). (C) Neurons transfected with Dyn1aa-GFP were challenged to take up FM4-64 in response to a depolarizing stimulus. Arrows indicate Dyn1aa- GFP-positive boutons that are also labeled with the dye (red). (D) Functional Dyn1 is necessary for synaptic vesicle recycling, for neurons transfected with a GTPase-deficient Dyn1 (Dyn1aaKA-GFP) were unable to take up FM4-64 following depolarization. Arrows indicate mutant Dyn1-positive boutons that are not labeled with the dye. The dye was actively endocytosed by surrounding nontransfected axons. The scale bar represents 5 m for (A) (D). (E and E ) Two examples of Dyn1 antibody labeling in synaptosomes reveal a presynaptic localization for Dyn1 at the ultrastructural level. The white arrows indicate clusters of Dyn1 immunolabeling. Mit, mitochondria; DC, dense-core granule; PSD, postsynaptic density. The scale bar represents 100 nm. fungal toxin used to label the filamentous actin in dendritic spines. Figure 2B suggests that Dyn3 does not reside on or in the dendrites, but rather at the tips of dendritic spines. Each Dyn3 puncta was slightly larger in diameter than that of the spine shaft, suggesting that Dyn3 was mostly in the spine head. Indeed, neurons costained for Dyn3 and PSD-95 showed a substantial colocalization between these two proteins, again, at the tips of the dendritic spines (data not shown). These experiments suggested that a majority of Dyn3 resides in the postsynapse. To determine whether exogenously expressed Dyn3 would target to the postsynapse, Dy- n3aaa was tagged with GFP and was transfected into neurons. After allowing transfected neurons to mature and develop spines, we colabeled cultures with rhodamine-phalloidin or with antibodies against PSD-95 or synapsin. Consistent with Dyn3 antibody labeling, the Dyn3aaa-GFP chimera was targeted to the ends of phal- loidin-stained dendritic spines (data not shown) and was cles, but outside of the active zone itself. A blinded quantitation (see the Supplemental Experimental Procedures available with this article online) was completed to determine the number of gold particles localizing to the pre- or postsynapse (see Table S1 in the Supplemental Data). Taken together, these morphological results strongly suggest that a majority of Dyn1 in the neuron is found in the presynaptic terminal and participates in synaptic vesicle recycling. To further define the localization of Dyn3, we compared the distribution of this dynamin isoform with that of various neuronal markers in mature cultures. Interestingly, neurons colabeled with antibodies to Dyn3 and the dendrite/soma-specific marker microtubule-associated protein 2b (MAP2b) suggested that the Dyn3 puncta did not reside directly on the shafts of the dendrites (Figure 2A). To investigate this pattern further, neurons were triple labeled with antibodies against Dyn3 and MAP2b in conjunction with rhodamine-conjugated phalloidin, a

3 Current Biology 512 Figure 2. Dyn3 Resides at the Tips of Dendritic Spines (A) Neurons were colabeled with antibodies against Dyn3 (green) and MAP2b (red). Dyn3 puncta were observed along the dendrites but were adjacent to the dendritic shaft (arrows). (B) Neurons were triple labeled with Dyn3 (green) and MAP2b (blue) antibodies and rhodamine-phalloidin (red) to reveal Dyn3 puncta at the tips of actin-rich dendritic spines (arrows). (C) To confirm this postsynaptic localization, neurons were transfected with Dyn3aaa-GFP and were labeled for PSD-95 (red). The Dyn3aaa- GFP puncta labeled the tips of spines and colocalized with PSD-95 (arrows). (D) Neurons were transfected with Dyn3aaa-GFP and were colabeled for MAP2b (blue) and synapsin (red). The Dyn3 signal was distinct from the presynaptic marker synapsin (arrows). The scale bar represents 5 m for (A) (D). (E and E ) Two examples of Dyn3 antibody labeling in synaptosomes reveal a predominantly postsynaptic localization at the ultrastructural level. The white arrows indicate clusters of Dyn3 immunolabeling. SVs, synaptic vesicles; DC, dense-core granule; PSD, postsynaptic density. The scale bar represents 100 nm. isolated synaptosome and postsynaptic density (PSD) fractions was performed (Figure 3A) to determine in which subcellular neuronal compartments each dynamin isoform was enriched. Synaptosomes (containing both pre- and postsynaptic structures) were initially isolated from whole rat brain. Detergent treatment of the synaptosomes solubilized the membranes and released the postsynaptic density, a detergent-resistant complex of scaffolding and receptor proteins, which was then pelleted (PSDI fraction). The remaining supernatant fol- lowing PSDI pelleting contained presynaptic proteins and postsynaptic proteins that did not interact with the PSD. We found that after this detergent treatment both Dyn1 and synaptophysin (a presynaptic control protein) were extracted into this supernatant. In contrast, a portion of Dyn3 pelleted with the PSDI fraction, although a substantial population of Dyn3 was also extracted, possibly representing a pool that did not strongly associate with the PSD. As a control for PSD isolation, Western incorporated into the postsynapse, as demonstrated by colocalization of Dyn3aaa-GFP and PSD-95 (Figure 2C). In addition, these Dyn3aaa-GFP puncta were adjacent to presynaptic terminals labeled by synapsin (Figure 2D). Other Dyn3aaa-GFP puncta were observed throughout the dendrites and even within the axons, but these Dyn3 populations did not colocalize with synaptic markers (data not shown). Focusing on this postsynaptic localization for Dyn3, we again performed immunogold electron microscopy on isolated adult rat forebrain synaptosomes, this time using Dyn3 antibodies. Most of the Dyn3 gold particles were found to be postsynaptic and within 50 nm of the synaptic cleft (Figures 2E and 2E ). A blinded quantitation of gold particle localization confirmed this notion (Table S1). These morphological results strongly suggest that a majority of Dyn3 is found in the postsynaptic region. To complement the morphological characterization of dynamin isoform localizations, a Western blot analysis of

4 Brief Communication 513 Figure 3. Dyn3 Is Biochemically Sequestered to Postsynaptic Sub- cellular Fractions, Where it Interacts with Homer and mglur5 (A) Equal protein amounts of whole brain (WB), synaptosomes (Syn), and postsynaptic density fractions (PSD I, II, III) were separated on SDS-PAGE gels and blotted for Dyn1, Dyn3, synaptophysin (presynaptic protein), Homer, and PSD-95 (postsynaptic proteins). Neither Dyn1 nor synaptophysin were pelleted with the PSD, but Dyn3 was isolated in the PSDI fraction and was enriched in the PSDII fraction. (B) Sequence comparison of the putative Dyn1 and Dyn3 Homer binding sites. Of these dynamins, only Dyn3 contains the Homer binding sequence. (C) Both Pan-Homer and Homer1-specific (Vesl) antibodies immunoprecipitated Dyn3, but not Dyn1, from whole brain lysate. In addition, mglur5 antibodies also specifically immunoprecipitated Dyn3. Following the gel separation and transfer of the samples to PVDF mem- brane, the membrane was split according to molecular weight con- straints and was blotted for the specific proteins listed. (D) Dyn3 antibodies can also immunoprecipitate mglur5. Nonspe- cific rabbit IgGs, Dyn1, or Dyn3 antibodies were used for immunoprecipitation from adult hippocampi. Only Dyn3 antibodies could coimmunoprecipitate mglur5, as demonstrated by the mglur5 Western blot on the left. The same blot was then stripped and reprobed for dynamin (on the right) by using a pan-dynamin antibody (MC63) to demonstrate that equivalent amounts of each dynamin isoform were immunoprecipitated. confirm the specificity of this Dyn3-Homer interaction, we completed immunoprecipitations of Homer from rat brain homogenate followed by Western blotting for dynamin isoforms. Only Dyn3 was coimmunoprecipitated with either pan-homer antibodies ( -Homer, Figure 3C), Homer1-specific antibodies ( -Vesl, Figure 3C), or Cupidin (Homer2) antibodies (data not shown), while Dyn1 was not isolated in any case. To test whether Dyn3 interacts with other components of the PSD as part of a complex mediated by Homer scaffolds, we immunoprecipitated a type 1 metabotropic glutamate receptor, previously shown to bind to the EVH1 domain of Homer [20], from whole hippocampi under nondenaturing conditions and blotted for dynamin. mglur5 antibodies immunoprecipitated Dyn3, but not Dyn1, in this experiment (Figure 3C). A control antibody ( -GST) did not immunoprecipitate any of the above proteins (data not shown). The reciprocal experiment, which involved immunoprecipitating with a control antibody (nonspecific rabbit IgG), -Dyn1, or -Dyn3 antibodies was also performed, followed by a Western to detect members of this PSD complex. We focused on mglur5 and found that only Dyn3 antibodies, and not the control antibody or Dyn1 antibody, could immunoprecipitate this receptor (Figure 3D). To demonstrate the efficiency of the IP, the same blot was stripped and reprobed with a pan-dynamin antibody (MC63). This blot demonstrates that, although similar amounts of Dyn1 and Dyn3 were immunoprecipitated by their respective antibodies, only Dyn3 could isolate mglur5 (Figure 3D). In addition, Dyn3 antibodies could also immunoprecipitate another metabotropic glutamate receptor isoform, mglur1a (data not shown). These findings suggest that Dyn3 is specifically recruited to the metabotropic glutamate receptors by binding to Homer in the postsynapse. To test the functional implications of the biochemical interactions described above, Dyn3aaa-GFP and mglur5a-n-myc were coexpressed in hippocampal neurons to determine whether Dyn3 participates in mglur5 internalization. The myc epitope on the receptor was exposed to the outside of the cell, allowing us to label the extracellular myc and later visualize only the internalized receptors for qualitative and quantitative measures (see the Supplemental Experimental Proce- dures). Mature hippocampal neurons coexpressing Dyn3aaa-GFP and mglur5a-n-myc demonstrated a ro- bust constitutive internalization of the mglur5a receptor over a 30-min time period (Figures 4A and 4A ). We utilized a dominant-negative version of Dyn3 (Dyn3- aaa(k44a)-gfp), containing a mutation in its GTPase region that was previously shown to block dynamin blots demonstrated that PSD-95 and Homer, integral scaffolding proteins within the PSD, were highly enriched in PSDI. Additionally, after a second round of detergent treatment and pelleting (PSDII fraction), Dyn3 was enriched, suggesting that this subpopulation of Dyn3 was tightly bound to the PSD. Dyn3 was not found in the PSDIII protein pellet, produced by extraction with ionic detergents, suggesting that Dyn3 is part of the PSD but likely does not function as a scaffolding protein. Subsequent to the morphological (Figure 2) and bio- function and endocytosis for Dyn1 and Dyn2 [5, 21]. chemical (Figure 3) observations supporting a postsyn- Cells coexpressing this dominant-negative mutant form aptic localization for Dyn3, we sought to identify any of Dyn3aaa and wild-type mglur5a-myc did not display potential Dyn3 binding partners within the PSD. A previ- any reduction in the number of mglur5a-myc puncta ous study demonstrated that GST-tagged Dyn3 proline- (internalized receptors) over 30 min (Figures 4B and 4B ). rich domain (PRD) could isolate the PSD scaffolding Receptor internalization under each transfection condiprotein Homer from brain homogenate. This interaction tion was quantified and is displayed as a percentage of with the Dyn3-PRD was predicted to be through Homer s each wild-type isoform control (Figure 4C). No signifi- EVH1 domain [18]. The EVH1 domain of Homer is ex- cant difference was found in the number of internalized tremely specific in its binding capabilities, requiring a receptor puncta between these two conditions. Similarly, PPXXFRP motif for proper binding [19]. While Dyn1 and neither dominant-negative Dyn1 nor dominant-negative Dyn3 sequences are highly conserved, only Dyn3 con- Dyn2 inhibited constitutive mglur5a uptake (Figure 4C, tains this specific motif within its PRD (Figure 3B). To L.F. et al., submitted).

5 Current Biology 514 Figure 4. Dyn3 Does Not Mediate mglur5 Internalization in Neurons, but Spliced Variants Regulate Dendritic Spine Morphogenesis (A B ) Mature hippocampal neurons transfected with both (A) Dyn3aaa-GFP and (A ) mglur5a-myc exhibit robust internalization of mglur5 in a constitutive manner, similar to those neurons expressing both a (B) GTPase-deficient Dyn3 (Dyn3aaa(K44A)-GFP) and (B ) mglur5a-myc. There was no difference in the amount of internalized mglur5 puncta in the mutant Dyn3-expressing cells, and this finding suggests that Dyn3 may not play a role in mglur5 receptor endocytosis. The scale bar represents 10 m. (C) Image analysis of transfected cells revealed no difference in the number of internalized puncta per cell area in Dyn3aaa, Dyn2aa, or Dyn1aa mutant-expressing cells. The error bars represent SEM. (D) Dyn3aaa-GFP does not have any effect on the morphogenesis of dendritic spines when expressed in neurons. (E) Although only differing by ten amino acids, expression of the Dyn3 spliced variant Dyn3baa-GFP in mature neurons results in an extensive increase in filopodial growth and a concomitant reduction in mushroom-shaped dendritic spines. The scale bar represents 10 m. splicing amongst the dynamin family members in the neuron. This study provides the first detailed examination of Dyn3 localization and function in neurons and compares this distribution and function to that of Dyn1. Since its identification in 1993, the function and subcellular local- ization of Dyn3 in any of the four tissues that express this isoform have remained undefined. Our data suggest that, in a mature neuron, Dyn3 localizes to the postsynapse and associates with components of the PSD (Figures 2 and 3). This recruitment to Homer and mglur5 could place Dyn3 in a position to participate in the devel- opment and growth of the postsynapse. In addition, with a number of PSD proteins dimerizing and multimerizing, and recent evidence that Dyn2 binds to Shank in the PSD [22], much flexibility exists for recruiting dynamin into many different PSD protein complexes (Figure S2). The surprising data that Dyn3baa-GFP expression can induce morphological changes in the developing neuron is both intriguing and exciting. The growth and mainte- nance of dendritic filopodia and spines are dependent upon the actin cytoskeleton, a network recently found to be dynamically influenced by dynamin. Dyn2 is an important factor in the maintenance of numerous actin- In an attempt to discern a function for Dyn3 at the postsynapse and in the dendrites, we expressed a second common spliced variant of Dyn3, Dyn3baa, in hippocampal neurons to examine a possible role for this spliceoform during receptor endocytosis. Dyn3baa differs from Dyn3aaa by only ten amino acids, a cassette that is spliced into Dyn3baa immediately preceding the pleckstrin homology domain [3]. Remarkably, expression of Dyn3baa-GFP induced a massive outgrowth of filopodia on transfected cells and a marked decrease in the normal distribution of mushroom-shaped dendritic spines (Figure 4E). This phenotype was apparent 2 3 days following transfection and was persistent for the life of the neuron (up to 4 weeks in culture). As with our previous observations, we did not see this effect when neurons expressed Dyn3aaa-GFP for a comparable amount of time (Figure 4D). Taken together, it seems that, although Dyn3 may not play an important role in the internalization of receptors and other components from the postsynaptic membrane, it participates in the maintenance of dendritic morphology, specifically, by regulating the outgrowth of dendritic protrusions and the morphogenesis of dendritic spines. This result also underscores the potential importance of alternative

6 Brief Communication 515 based cellular structures, such as the podosome [23] or of dynamin 2, an isoform ubiquitously expressed in rat tissues. the comet tails found on motile vesicles and invasive Proc. Natl. Acad. Sci. USA 91, Sontag, J.-M., Fykse, E.M., Ushkaryov, Y., Liu, J.-P., Robinson, pathogens [13, 14]. In addition, another study produced P.J., and Sudhof, T.C. (1994). Differential expression and regulaevidence demonstrating a direct effect of Dyn2 on actin tion of multiple dynamins. J. Biol. Chem. 269, filament polymerization and organization [15]. The fact 12. Jones, S.M., Howell, K.E., Henley, J.R., Cao, H., and McNiven, that the filopodial induction induced by Dyn3baa excles from the trans-golgi network. Science 279, M.A. (1998). Role of dynamin in the formation of transport vesi- pression is latrunculin sensitive (N.W.G. and M.A.M., unpublished data) may imply that Dyn3 is acting through 13. Orth, J.D., Krueger, E.W., Cao, H., and McNiven, M.A. (2002). The large GTPase dynamin regulates actin comet formation and an actin-dependent network in order to induce these movement in living cells. Proc. Natl. Acad. Sci. USA 99, morphological changes. Given that neurons perform a 14. Lee, E., and De Camilli, P. (2002). Dynamin at actin tails. Proc. number of unique processes, integrate into more com- Natl. Acad. Sci. USA 99, plex networks, and express a variety of unique genes, 15. Schafer, D.A., Weed, S.A., Binns, D., Karginov, A.V., and Cooper, the need for multiple dynamins to perform specialized J.A. (2002). Dynamin 2 and cortactin regulate actin assembly functions may explain the diversity of this GTPase family and filament organization. Curr. Biol. 12, Nakata, T., Takemura, R., and Hirokawa, N. (1993). A novel memin the mammalian brain. ber of the dynamin family of GTP-binding proteins is expressed specifically in the testis. J. Cell Sci. 105, 1 5. Supplemental Data 17. Cook, T., Mesa, K., and Urrutia, R. (1996). Three dynamin-encod- Supplemental Data including a developmental expression profile of ing genes are differently expressed in developing rat brain. J. dynamin isoforms, a model suggesting a PSD localization for dy- Neurochem. 67, namin isoforms, and complete Experimental Procedures are avail- 18. Tu, J.C., Xiao, B., Yuan, J.P., Lanahan, A.A., Leoffert, K., Li, able at M., Linden, D.J., and Worley, P.F. (1998). Homer binds a novel proline-rich motif and links group 1 metabotropic glutamate receptors with IP3 receptors. Neuron 21, Acknowledgments 19. Beneken, J., Tu, J.C., Xiao, B., Nuriya, M., Yuan, J.P., Worley, P.F., and Leahy, D.J. (2000). Structure of the Homer EVH1 do- We would like to thank Xiaoyun Wang and Mu-Ming Poo for assismain-peptide complex reveals a new twist in polyproline recogtance with the hippocampal neuronal cultures; Vanda Lennon, Doug nition. Neuron 26, Murphy, Teiichi Furuichi, and Peter McPherson for the kind gifts of 20. Brakeman, P.R., Lanahan, A.A., O Brien, R., Roche, K., Barnes, antibodies; Enrique Torre for advice concerning immunocytochem- C.A., Huganir, R.L., and Worley, P.F. (1997). Homer: a protein istry; Dave Zacharias for neuronal transfection protocols; and that selectively binds metabotropic glutamate receptors. Nature Heather Thompson for critically reading this manuscript. N.W.G. is 386, supported by National Institutes of Health Neuroscience Training 21. Henley, J.R., Krueger, E.W., Oswald, B.J., and McNiven, M.A. Grant NS (1998). Dynamin-mediated internalization of caveolae. J. Cell Biol. 141, Received: June 10, Okamoto, P.M., Gamby, C., Wells, D., Fallon, J., and Vallee, R.B. Revised: January 20, 2003 (2001). Dynamin isoform-specific interaction with the shank/ Accepted: January 20, 2003 ProSAP scaffolding proteins of the postsynaptic density and Published: March 18, 2003 actin cytoskeleton. J. Biol. Chem. 276, Ochoa, G.C., Slepnev, V.I., Neff, L., Ringstad, N., Takei, K., Daniell, L., Kim, W., Cao, H., McNiven, M., Baron, R., et al. (2000). References A functional link between dynamin and the actin cytoskeleton 1. McNiven, M.A., Cao, H., Pitts, K.R., and Yoon, Y. (2000). Pinching at podosomes. J. Cell Biol. 150, in new places: multiple functions for the dynamin family. Trends Biochem. Sci. 25, Note Added in Proof 2. Danino, D., and Hinshaw, J.E. (2001). Dynamin family of mechanoenzymes. Curr. Opin. Cell Biol. 13, The data referred to as L.F. et al., submitted are now in press: 3. Cao, H., Garcia, F., and McNiven, M.A. (1998). Differential distri- Fourgeaud, L., Bessis, A.S., Rossignol, F., Pin, J.P., Olivo-Marin, bution of dynamin isoforms in mammalian cells. Mol. Biol. Cell J.C., and Hemar, A. (2003). The metabotropic glutamate receptor 9, mglur5 is endocytosed by a clathrin independent pathway. J. Biol. 4. Takei, K., Mundigl, O., Daniell, L., and De Camilli, P. (1996). The Chem., in press. synaptic vesicle cycle: a single vesicle budding step involving clathrin and dynamin. J. Cell Biol. 133, Damke, H., Baba, T., Warnock, D.E., and Schmid, S.L. (1994). Induction of mutant dynamin specifically blocks endocytic coated vesicle formation. J. Cell Biol. 127, Chen, M.S., Obar, R.A., Schroeder, C.C., Austin, T.W., Poodry, C.A., Wadsworth, S.C., and Vallee, R.B. (1991). Multiple forms of dynamin are encoded by shibire, adrosophila gene involved in endocytosis. Nature 351, van der Bliek, A.M., and Meyerowitz, E.M. (1991). Dynamin-like protein encoded by the Drosophila shibire gene associated with vesicular traffic. Nature 351, Shpetner, H.S., and Vallee, R.B. (1989). Identification of dynamin, a novel mechanochemical enzyme that mediates interactions between microtubules. Cell 59, Shupliakov, O., Low, P., Grabs, D., Gad, H., Chen, H., David, C., Takei, K., De Camilli, P., and Brodin, L. (1997). Synaptic vesicle endocytosis impaired by disruption of dynamin-sh3 domain interactions. Science 276, Cook, T.A., Urrutia, R., and McNiven, M.A. (1994). Identification