Mouse Profilin 2 Regulates Endocytosis and Competes with SH3 Ligand Binding to Dynamin 1*

Mammalian profilins are abundantly expressed actin monomer-binding proteins, highly conserved with respect to their affinities for G-actin, poly-l-proline, and phosphoinositides. Profilins associate with a large number of proline-rich proteins; the physiological significance and regulation of which is poorly understood. Here we show that profilin 2 associates with dynamin 1 via the C-terminal proline-rich domain of dynamin and thereby competes with the binding of SH3 ligands such as endophilin, amphiphysin, and Grb2, thus interfering with the assembly of the endocytic machinery. We also present a novel role for the brain-specific mouse profilin 2 as a regulator of membrane trafficking. Overexpression of profilin 2 inhibits endocytosis, whereas lack of profilin 2 in neurons results in an increase in endocytosis and membrane recycling. Phosphatidylinositol 4,5-bisphosphate releases profilin 2 from the profilin 2-dynamin 1 complex as well as from the profilin 2-actin complex, suggesting that profilin 2 is diverging the phosphoinositide signaling pathway to actin polymerization as well as endocytosis.

Profilin is a 15-kDa protein, abundantly expressed in all eukaryotic cells from yeast to mammals. Most organisms harbor more than one profilin gene, and in mouse and human four genes have been identified up to now (1)(2)(3)(4). Profilin was first isolated in a 1:1 complex with monomeric actin, and extensive characterization of its effect on actin polymerization established its role as a key regulator of actin dynamics (for a review see Ref. 5). The interaction of profilin and actin is regulated by phosphoinositides (6). Phosphatidylinositol 4,5-bisphosphate (PtdInsP 2 ) 2 was shown to disrupt the "profilactin" complex, and it was suggested that this provides a mechanism to release ATP-charged actin monomers, which in turn could promote local actin polymerization (7).
In addition to binding to actin, all profilins identified so far except for the profilin of Vaccinia virus (8) and the mouse profilin IIB splice form (9) have been shown to bind to poly-L-proline, an ability of profilin proven to be essential in yeast (10,11). However, the functional mechanisms and cellular pathways regulated by profilin-poly-L-proline interaction remain unclear. The first physiological ligand described to bind profilin via its poly-L-proline binding site was the focal-adhesion phosphoprotein VASP, which contains multiple proline-rich stretches (12). Meanwhile, a continuously growing number of profilin ligands have been identified (13).
Neurons express profilin 1 as well as profilin 2, and a proteomicsbased screen in mouse brain identified a number of novel profilin-ligand interactions (14). Profilin 2, in particular, interacts with ligands involved in signal transduction, membrane trafficking, and vesicle recycling such as ROCK2, synapsins, POP-130/CyFIP1, and dynamin 1. These findings suggest a not yet recognized additional role for profilin 2 in membrane trafficking. Here we address this hypothesis by examining the functional importance of the profilin 2-dynamin 1 interaction. Dynamin 1 is a neuron-specific GTPase, which assembles around the necks of budding vesicles and is thought to help in the membrane fission process. Early studies on various dynamin mutants have established the essential role of dynamin in endocytosis (15) and synaptic vesicle recycling. To assemble the endocytic machinery, which initiates changes in membrane curvature, dynamin 1 interacts with a number of effector molecules such as amphiphysin (16) and endophilin (17). Although the importance of these interactions is well established, the spatial and temporal regulation of this process has just begun to be unraveled (18). Endocytosis and actin polymerization are tightly linked (19,20), and it has been proposed that actin polymerization might help to constrain or stabilize the membrane curvature during endocytosis. A number of recent reports support the intimate relation of dynamin and the actin cytoskeleton. So, dynamin has been shown to cause actin rearrangements at the immunological synapse in activated T-cells (21). Furthermore, molecules such as Abp1 (22), cortactin (23), and syndapin (24) can recruit dynamin to actin-rich structures like podosomes (25), ruffles (23) and actin comet tails (26). However, our understanding of the nature and regulation of the link of dynamin to the actin cytoskeleton is still far from complete.
Here, we provide evidence that one function of profilin 2 is to control dynamin 1 activity by sequestering the binding site for SH3-containing ligands such as endophilin and amphiphysin. We also show that PtdInsP 2 is a common signal to release profilin 2 from dynamin 1 as well as G-actin, providing a mechanism by which activation of the endocytic machinery and actin polymerization can be orchestrated at PtdInsP 2rich membrane domains.

MATERIALS AND METHODS
Antibodies-Anti-dynamin 1 antiserum KG-43 was generated by immunizing rabbits with full-length GST-dynamin 1 expressed in insect cells. The anti-amphiphysin antibody was provided by P. De Camilli. The anti-endophilin antibody was produced by immunizing rabbits with the full-length GST fusion protein expressed in Escherichia coli (expression vector was a generous gift from H. Söling). Antibodies against Mena and VASP were produced by immunizing rabbits with the respective His-tagged proteins purified from insect cells (baculovirus clones were obtained from F. Gertler). The anti-profilin 2 antibody "P2-T" was generated by immunizing rabbits with fragments produced by partial digest of mouse profilin 2 with V8 protease. Secondary antibodies were purchased from Pierce. A, dynamin 1 and profilin 2 co-immunoprecipitate from total brain lysate. The immunoprecipitate with a profilin 2-specific polyclonal antibody was subjected to SDS-PAGE, and dynamin 1 as well as profilin 2 were detected by Western blotting using the respective antibodies. B, dynamin 1 binds directly and specifically to profilin 2. Recombinant dynamin 1aa (0.3 nmol) was allowed to bind to profilin 1, profilin 2, or DNase 1 beads (each 20 nmol). Almost complete binding of dynamin 1 to profilin 2 can be seen, while only weak binding to profilin 1 and no binding to the DNase 1 control beads was observed. C, poly-L-proline competes for binding of dynamin 1 to profilin 2. Poly-L-proline beads were incubated with dynamin 1aa (ϩdyn 1, 0.2 nmol), or profilin 2 (ϩprofilin 2), or first saturated with profilin 2 and then incubated with dynamin 1aa (ϩdyn 1/ϩprofilin 2). No binding of dynamin 1 to profilin 2 occurs under these conditions when the proline binding site of profilin 2 is occupied by Uptake of FM 1-43 into Cells and Quantitation by Fluorescence Scanning-For comparative kinetic studies, mouse cortical neurons of wild-type and profilin 2 knock-out mouse embryos (embryonic day 14) were prepared the same day under the same conditions. Cells were seeded at 5 ϫ 10 5 per well into 24-well plates and allowed to grow and spread for 5-6 days. Uptake buffer was modified Tyrode's solution (150 mM NaCl, 4 mM KCl, 2 mM MgCl 2 , 2 mM CaCl 2 , 10 mM glucose, 10 mM Hepes, pH 7.4). FM 1-43 was diluted in Tyrode's to a concentration of 10 M, sonicated, spun for 10 min at 7000 rpm, and prewarmed to 37°C. Before the experiment, cells were washed twice with Tyrode's solution, and the solution was aspirated. Uptakes were initiated by injecting 400 l of 10 M FM 1-43 in Tyrode's solution into the wells in a fluorescent plate reader (Fluoroskan Ascent II, Labsystems), exciting FM 1-43 fluorescence at 485 nm and reading emission at 590 nm, measuring FM 1-43 in 3-s intervals. Kinetic data were analyzed with the GraFit program (Erithacus Software Ltd.). Cell numbers were accounted for by incubating the cells with Hoechst 33342 (10 mg/ml, Molecular Probes) 1:30,000 in PBS/0.1% Tween 20 for 20 min at room temperature and reading the fluorescence signal at 460 nm after washing three times with PBS and exciting at 355 nm, and by determining total protein content by Bradford assay.
Isolation of Protein Complexes-Protein complex formation on beads was carried out as described previously (14).
Characterization of Profilin 2 and Endophilin Binding Sites on Dynamin 1-To map the profilin 2 binding sites on the PRD of dynamin 1, 32 12-mer peptides covering the entire PRD of dynamin 1 were synthesized on a SPOT membrane. As a positive control for profilin binding, two peptides from VASP (see Table 1) known to bind profilin 2 were included (51). The membrane was blocked overnight in 5% milk powder, 0.15 M NaCl, 20 mM Tris-HCl, pH 8, 0.5% Tween 20, 0.02% sodium azide, and then incubated for 4 h with 10 g/ml profilin 2 or endophilin. After 4 washes, the membrane was successively incubated with the respective first polyclonal antibodies against profilin 2 or endophilin and the second horseradish peroxidase-labeled anti-rabbit antibody. The signal was developed using ECL reagent (Amersham Biosciences).
Analytical Ultracentrifugation of Profilin 2-Dynamin 1 Complexes-Sedimentation velocity experiments were done at 20°C using a twochannel charcoal centerpiece (3-mm path length) at a speed of 55,000 rpm in a Beckman Optima XL-A centrifuge fitted with a four-hole AN-60 rotor. Sedimentation velocity profiles were collected by monitoring the absorbance signal at 285 nm. To determine the sedimentation properties of profilin-dynamin complexes, profilin 2 and dynamin 1 were mixed at 25 times molar excess of profilin 2 in 20 mM phosphate, pH 7.5, 150 mM NaCl, and 1 mM ␤-mercaptoethanol. Sedimentation coefficient and molecular weight distributions were analyzed by the C(s) method implemented in the Sedfit software package (27).
Uptake of FITC-transferrin into Cells-For microscopic analyses, HeLa cells expressing wild-type DsRed-profilin 2, or the DsRed-profilin 2 mutants S134Y and S138D, deficient for poly-L-proline binding, were seeded on coverslips in antibiotic and serum-free medium and starved for 16 h. Uptake was started by replacing the starving medium by serum-containing medium with 20 g/ml FITC-transferrin (Molecular Probes). The uptake was stopped by dipping the coverslips into ice-cold PBS and transferring them to ice-cold 4% paraformaldehyde/PBS.
Actin Polymerization Assays-Actin was purified from rabbit muscle as described (28). Actin polymerization was assayed in 100-l reactions containing 5 M actin (containing 5% pyrene-actin) plus the proteins of interest (profilin 2 at 5 M and dynamin at 2 M) in G buffer (2 mM Tris, pH 8.0, 0.2 mM CaCl 2 , 0.2 mM ATP). Reactions were started by adding 90 l of premixed components into a well of a black 96-well plate (Greiner) containing 10 l of 10ϫ polymerization buffer (0.1 M imidazole, pH 7.2, 20 mM MgCl 2 , 10 mM EGTA). After brief mixing, the polymerization reaction was monitored at 25°C with a fluorescent plate reader (Fluoroskan Ascent II, Labsystems) in 10-s intervals by exciting pyrene fluorescence at 342 nm and measuring fluorescence emission at 388 nm. The results were plotted as relative fluorescence versus time using the program GraFit (Erithacus Software Ltd.).
Immunoprecipitation-For immunoprecipitations with rabbit polyclonal antisera, 30 l of protein A-Sepharose-slurry (Amersham Biosciences) was washed with PBS and incubated with 25 l of serum in 1 ml of PBS for 1-2 h at 4°C on a rotating wheel. After three washes with lysis buffer, the beads were incubated with brain lysate for 2 h then washed 5 times with ice-cold lysis buffer, and the proteins were eluted by boiling in 1ϫ SDS-sample buffer.
Protein Purification-All recombinant profilin 2 used in the experiments was the mouse profilin 2a spliceform, the recombinant dynamin 1 the spliceform dynamin 1aa. Recombinant mouse profilin 2a was purified as described before (29). Recombinant dynamin 1aa was expressed as a GST fusion protein in Hi5 insect cells after infection with baculovirus and purified using glutathione beads. All assays were performed after cleavage and removal of the GST moiety. Grb2, Src-SH3, amphiphysin, and endophilin were expressed as GST fusion proteins in E. coli.
Competition Binding Assays-Nunc Maxisorp ELISA plates were coated with 0.5 g/ml GST-Grb2, GST-amphiphysin, GST-endophilin, and GST-Src-SH3-domain in 50 mM NaHCO 3 , pH 8.5/0.01% NaN 3 overnight at 4°C. The wells were washed 3 times with 100 l of PBS/ 0.05% Tween 20 at room temperature. Unspecific binding sites were blocked with 1% BSA in PBS/0.05% Tween 20 for 2 h at room temperature. The wells were then incubated with an equal amount of recombinant dynamin 1 and increasing amounts of recombinant profilin 2a (or BSA as a control) in 1ϫ profilin dialysis buffer (29) for 2 h at 4°C. Afterward, the wells were washed and incubated with anti-dynamin 1 antibody Hudy-1 1:500 in 1% BSA/PBS/0.05% Tween 20 for 1 h at room temperature. After washing, 100 l of alkaline phosphatase coupled goat-anti-mouse antibody 1:1000 in 1% BSA/PBS/0.05% Tween was incubated for 1 h at room temperature. After five washes, the wells were incubated with 100 l of a 1 mg/ml p-nitrophenyl phosphate (Pierce) solution in 5 mM MgCl 2 /1 M Na 2 CO 3 /NaHCO 3 , pH 9.5, until a discernable yellow signal appeared. The signal was quantitated in an ELISA plate reader at 405 nm.
poly-L-proline. D, profilin mutants deficient in poly-proline binding still bind to actin. Total brain lysate was passed over beads loaded with profilin 2 or the profilin 2 point mutants S134Y and S138D or glutathione beads as a control. Western blot of bead fractions probed with antibodies against actin and dynamin 1. Both profilin 2 mutants bind to actin but not dynamin 1. E, mapping of the profilin 2 binding site on dynamin 1. Full-length dynamin 1 was treated with serial dilutions of the protease subtilisin and the resulting proteolytic fragments allowed to bind to profilin 2 beads. Bound and unbound fractions were subjected to SDS-PAGE and tested with the dynamin-specific antibody hudy-1. With increasing amounts of subtilisin (1:500) a stable C-terminal 15-kDa fragment (PRD) is generated, which binds to profilin 2. F, mapping binding sites for profilin 2 and endophilin in the PRD of dynamin 1. SPOT array analysis of overlapping peptides representing the whole PRD of dynamin reveal three weaker and two stronger profilin 2 binding sites (peptides 20 -21 and 26). Endophilin shows one weak and two stronger binding sites (peptides 20 and 26 -30). Lower panel, schematic representation of the PRD of dynamin 1 with strong binding sites shown in black and weak ones in gray. Amphiphysin and Grb2 (33,34) binding sites are also indicated. Numbers below the sequence correspond to the spot numbers on the membrane and refer to the starting point of the overlapping 12-mer peptides on the array. CtrlA and -B are peptides derived from VASP, which have been shown to bind profilin (51). G, profilin 2 binds to dynamin 1 in a 2:1 complex. Sedimentation analysis by analytical ultracentrifugation shows three major species: profilin 2 monomers (15 kDa), profilin 2/GST-dynamin 2:1 complexes (155-160 kDa), and GST-dynamin dimers (250 kDa).
Preparation of PtdInsP 2 Micelles-Phosphatidylinositol 4,5-bisphosphate (Sigma) was reconstituted in methanol/chloroform 1:1 to a final concentration of 3 mg/ml. For making micelles, an appropriate amount of stock solution was dried under nitrogen. The lipid film was rehydrated in 20 mM HEPES, pH 7.2/0.1 mM EDTA for at least 30 min at room temperature with occasional mixing to give a concentration of 3 mg/ml. Finally, the mixture was sonicated in a sonicating water bath for 15 min.

RESULTS
Profilin 2 Is a Ligand of Dynamin 1-Dynamin 1 had been shown previously to be part of the profilin 2 complex isolated from mouse brain (14). This finding suggests that the actin-binding protein profilin 2 might play a role in regulating endocytosis. To address this hypothesis in more detail, we focused on the biochemical aspects of the profilin 2-dynamin 1 interaction. First, we showed that the association of dynamin 1 and profilin 2 occurs under physiological conditions. As shown in Fig.  1A, dynamin 1 can be co-immunoprecipitated with profilin 2 from brain lysates using a profilin 2-specific antibody. Association by co-immunoprecipitation could be due to direct binding, or be mediated by a third protein. To test for direct binding, we expressed and purified recombinant dynamin 1 and performed binding studies using beads loaded with profilin 1 and profilin 2 (Fig. 1B). In this two-component binding assay dynamin 1 binds efficiently to profilin 2, whereas binding to profilin 1 is an order of magnitude weaker. The binding of profilin 2 to dynamin 1 was independent of the salt concentration within the physiological range (50 -250 mM) and largely independent of divalent ions such as Ca 2ϩ and Mg 2ϩ . In addition, binding was not affected by ATP, GMP, or GTP and only slightly reduced in the presence of GTP␥S (data not shown). From these results we concluded that dynamin 1 binds directly and specifically to profilin 2.
We next asked which of the three functional domains of profilin 2, the actin binding site, the PtdInsP 2 binding site, or the poly-L-proline binding domain, are employed for dynamin 1 binding. Our results show that the poly-L-proline binding site of profilin is required for dynamin 1 binding, because occupation of this site by a synthetic proline homopolymer abolishes dynamin 1 binding completely (Fig. 1C). Furthermore we generated two mutant forms of profilin 2, which bind actin, but lack poly-L-proline binding. The Ser-138 3 Asp mutant mimics phosphorylation of Ser-138, which abolishes poly-L-proline binding. The second profilin 2 point mutation Ser-134 3 Tyr replicates a mutation that has previously been shown to affect poly-L-proline binding in profilin 1 (30). As shown in Fig. 1D both profilin 2 mutants bind actin comparably to wild-type profilin 2, whereas dynamin 1 binding is completely abrogated.
These findings suggested that dynamin 1 most likely uses a prolinerich stretch of amino acids to bind to profilin 2. Dynamin 1 consists of an N-terminal domain harboring the GTPase activity, a central PH domain, and a coiled-coil stretch followed by a conspicuous prolinerich C-terminal domain (see Fig. 1F). To test if the PRD provides the profilin 2 binding site on dynamin 1, we chose to cleave the dynamin 1 molecule into different domains by partial proteolysis and then test the dynamin 1 fragments for profilin 2 binding. The protease subtilisin has been shown to produce a specific pattern of dynamin 1 fragments, in the process of which a 15-kDa C-terminal peptide containing the prolinearginine-rich domain is generated (31). We therefore cleaved recombinant dynamin 1 with limiting amounts of subtilisin and performed binding assays with the generated fragments using profilin 2 beads. The bound dynamin 1 fragments were then identified using the monoclonal antibody Hudy-1, which recognizes an epitope (amino acids 822-838) in the 15-kDa PRD (32). Fig. 1E shows that, with increasing amounts of protease, a stable C-terminal 15-kDa fragment is generated, which binds to profilin 2. Based on this we concluded that the PRD of dynamin mediates binding to profilin 2.
The PRD of dynamin 1 comprises a number of putative profilin binding sites (Fig. 1F) (14). To identify the exact binding sites of profilin 2 within the PRD and to compare these with the docking sites of other known dynamin 1 ligands we made use of a SPOT peptide array covering the entire PRD. A total of 32 overlapping 12-mers staggered by 3 amino acids were synthesized on a membrane support and tested for binding of profilin 2 (see "Material and Methods"). As shown in Fig. 1F, five profilin 2 binding sites were identified within the PRD, and based on the recovery of signal after extensive washing we could distinguish three weaker and two stronger binding sites (see Table 1). Interestingly, the strong profilin 2 binding sites overlap with the docking sites for Grb2 and amphiphysin (33,34). Endophilin, a recently described SH3 ligand of dynamin 1 is also thought to bind along the PRD (35). Using the same SPOT peptide array and recombinant endophilin we could identify one weak and two strong endophilin binding sites on the PRD. The strong endophilin binding sites stretch over a region of several peptides and overlap with the profilin 2 binding site as well as the Grb2 and amphiphysin docking site.
Although we have identified several potential profilin 2 binding sites along the PRD, the question remains whether all these sites are occupied in the profilin 2-dynamin 1 complex. To determine the stoichiometry of the complex in vitro we performed velocity sedimentation using an analytical ultracentrifuge. As shown in Fig. 1G, a major component with  Strong (ϩϩ) and weak binding (ϩ) was judged from the remaining signals after extensive wash of the blot. Peptides for the positive control of profilin 2 binding (33)(34) were derived from the profilin 2 ligand VASP.

Peptide no. Sequence Profilin 2 binding
Endophilin binding PPPPPGPPPPPG ϩϩ Ϫ molecular mass 15 kDa was identified, which corresponds to the excess monomeric profilin that was used to shift the equilibrium toward complex formation. Apart from monomeric profilin 2, complexes of 155-160 and 250 kDa and some higher molecular mass complexes were identified. The 250-kDa complex represents the GST-dynamin 1 dimer, and the higher molecular mass complexes correspond to oligomers of GST-dynamin 1, whereas the 155-to 160-kDa complex is in good agreement with a 2:1 complex of profilin 2 (15 kDa) and GST-dynamin 1 (126 kDa). These results show that, under the conditions used in the sedimentation experiment, the profilin 2-dynamin 1 complex exists at a 2:1 stoichiometry.
Binding of Actin and Dynamin 1 to Profilin 2 Is Mutually Exclusive-How could the profilin 2-dynamin 1 complex link the actin cytoskeleton to the endocytic machinery? One possibility is a ternary complex of profilin 2, dynamin 1, and actin, which targets actin monomers to the site of endocytosis. We tested this hypothesis of a ternary complex using G-actin immobilized on DNase 1 beads. Profilin 2 alone binds to the G-actin-loaded beads and forms a profilin-actin complex as expected from previous results (e.g. Ref. 36). However, no binding of the profilin 2-dynamin 1 complexes to G-actin-loaded beads was observed ( Fig. 2A), suggesting that profilin 2 can either bind to actin or to dynamin 1 but not both concomitantly. In cells, ϳ50% of profilin is normally found free of actin, and the other half is tightly bound in a stable G-actin-profilin complex. To exclude the possibility that lack of trimeric complex formation was due to the different history of complexes formed by dynamin 1, recombinant mouse profilin 2 and purified rabbit actin, versus dynamin 1 and the authentic "profilactin" found in cells (37), we isolated native profilactin from mouse brain and repeated the binding assay with dynamin 1 and DNase 1 beads. This experiment confirmed again that dynamin 1 does not associate with actin-bound profilin 2 ( Fig. 2A, last two lanes). On the contrary, G-actin tends to destabilize a preformed dynamin 1-profilin 2 complex as shown in Fig.   2B. When profilin 2 beads loaded with dynamin 1 were incubated in the presence of 50 M G-actin, the dynamin 1-profilin 2 complex became instable and dynamin 1 was partially released (Fig. 2B).
These binding assays do not exclude the possibility that the association of profilin 2 and dynamin 1 could still influence actin polymerization by an as yet unrecognized mechanism. We therefore performed actin polymerization studies using fluorescently labeled pyrene-actin in the presence of profilin 2 and dynamin 1. As shown in Fig. 2C, dynamin 1 alone did not have any significant effect on actin nucleation or filament elongation. Even at the highest dynamin 1 concentration used in the assay (2 M dynamin, dynamin-actin ratio 1:2) polymerization kinetics were not affected, and steady-state levels of actin polymerization were comparable to control experiments. Profilin 2 on the other hand slowed down the nucleation and elongation step of actin polymerization, as shown before by others (38). When profilin 2-dynamin 1 complexes were added, no significant change was observed compared with adding profilin 2 alone. Only at the highest dynamin 1 concentration used in our assays (profilin 2-dynamin 1 ratio 2:1), a slight increase in the actin polymerization rate was observed compared with profilin 2 alone (Fig. 2D). This effect of dynamin 1 is most likely due to the equilibrium binding and sequestration of profilin 2 by dynamin 1, essentially increasing the concentration of free actin in the system. One conclusion from these results is that neither dynamin 1 nor the profilin 2-dynamin 1 complex has a direct effect on actin polymerization kinetics.
Profilin 2 Competes with SH3 Ligands for Binding to Dynamin 1-Because the profilin 2-dynamin 1 complex has no direct effect on actin polymerization, we reasoned that the role of profilin 2 binding to dynamin 1 is more likely to regulate the activity of dynamin 1 itself or dynamin 1 assembly at sites of endocytosis. We therefore tested whether profilin 2 binding would affect the enzymatic properties of dynamin 1. In our hands neither the intrinsic GTPase activity nor the oligomerization state of dynamin 1 was altered upon binding of the FIGURE 2. Dynamin 1 and actin binding to profilin 2 is mutually exclusive. A, actin was immobilized on DNase 1 beads, and binding of recombinant profilin 2 (saturating amounts) and dynamin 1 (35 g) was tested. Bound and unbound fractions were separated by SDS-PAGE, and proteins were visualized by Coomassie Blue staining. Dynamin 1 does not interact with actin (lanes 3 and  4). Complexes formed between dynamin 1 and profilin 2 do not bind to rabbit actin (lanes 5 and 6), nor does dynamin 1 bind to the native profilactin complex isolated from mouse brain (lanes 7 and 8). B, actin competes with dynamin 1 for binding to profilin 2. Dynamin 1 was allowed to bind to profilin 2 beads, and the beads were washed and then incubated with either G buffer alone or with 50 M G-actin in G buffer. Note that G-actin partially removes dynamin 1 from profilin 2. C, actin polymerization, monitored by increase of fluorescence of pyrene actin (filled circles, 5 M actin alone). Dynamin 1 (2 M) does not affect actin polymerization kinetics (open triangles), while profilin 2 (5 M) delays polymerization by sequestering actin monomers (open circles). Profilin 2 together with dynamin 1 has a slightly increased polymerization rate compared with profilin 2 alone (filled diamonds). D, polymerization rates deduced from three independent polymerization curves as shown in C. The maximal speed of polymerization is expressed in percent control (actin alone). profilin 2 (data not shown). Therefore binding of profilin 2 to dynamin 1 is neutral with respect to intrinsic activities.
Apart from the oligomerization of dynamin 1, the binding of Grb2 and endophilin to the PRD are important for the late steps in vesicle membrane invagination and scission (39,40). Using the PRD peptide arrays we had shown that the profilin 2 binding sites overlap with the ones of endophilin, Grb2, and amphiphysin (see Fig. 1F). It is therefore conceivable that profilin 2 interferes with binding of these SH3-domain ligands. To address this question, we set up an ELISA-based method to test for competition (see "Material and Methods"). Briefly, Grb2, Src-SH3, endophilin, and amphiphysin were used to coat plastic wells, and dynamin 1 was allowed to bind to the respective ligands. The amount of bound dynamin 1 was then determined using a dynamin 1-specific antibody and an alkaline phosphatase-coupled secondary antibody for substrate development. To test the ability of profilin 2 to compete with the solid phase-bound SH3 ligands, increasing amounts of profilin 2 or BSA as a control were added together with dynamin 1. Fig. 3A shows the summary of at least three competition experiments for the dynamin 1 ligands Grb2, amphiphysin, endophilin, and the Src-SH3 domain. BSA had no effect on the binding of dynamin 1 to its ligands, whereas increasing amounts of profilin 2 competed significantly with the binding of dynamin 1 to the respective ligands.
To test whether such competition also occurred in vivo, we analyzed the composition of the profilin 2-dynamin 1 complexes isolated from mouse brain. Despite the abundance of dynamin 1 bound to profilin 2 beads, dynamin 1 ligands such as amphiphysin and endophilin were excluded from the complex (Fig. 3B), showing an even more stringent competition than observed in the ELISA experiments.
The ELISA-based assay showed ϳ40% competition, and the difference between the two assays is most likely due to sterical constrains on the ELISA support and a higher local profilin 2 concentration seen on the beads. However, both experiments are in agreement with an inhibition of SH3 ligand binding to dynamin 1 by profilin 2.
Profilin 2-Dynamin 1 Complex Is Regulated by Phosphoinositides-Profilin 2 can compete for the binding sites of amphiphysin, endophilin, and Grb2, and possibly other SH3 molecules, on dynamin 1. To remove profilin 2 and allow SH3 ligands to bind and recruit dynamin 1 to the endocytic machinery at the membrane, a signal must exist which allows dissociation of profilin 2 from dynamin 1 under physiological conditions. Phosphoinositides have been shown to regulate binding of profilin to actin (6,41), and we therefore reasoned that such a regulatory mechanism might also be used in the cell to regulate the profilin 2-dynamin 1 interaction. Acidic phospholipids are usually exposed only on the cytosolic side of the plasma membrane, and because of their localization in membrane domains have been discussed as spatial organizers of actin polymerization as well as endocytosis (42). Profilin 2 and dynamin 1 have both been shown to bind phosphoinositides such as phosphatidylinositol 4,5-bisphosphate (PtdInsP 2 ) (29,41,43). To directly test whether PtdInsP 2 regulates the profilin 2-dynamin 1 complex, and to address specificity, we first isolated the whole profilin 2 complex from brain lysate on profilin 2 beads and then incubated the complex with PtdInsP 2 micelles. As shown previously, the complex contains at least 30 different proteins some of which have been identified by mass spectrometry (14) and Western blot. After incubation of the total complex with PtdInsP 2 micelles, the released proteins were analyzed by SDS-PAGE and silver staining as well as Western blot. Under these conditions, mainly two ligands of profilin 2 were released from the complex: actin, which runs at 45 kDa, and a second protein of 100 kDa (Fig. 4A). Using Western blotting, the 100-kDa protein was identified as dynamin 1 (Fig. 4B, upper panel). In our hands, only actin and dynamin 1 are regulated by PtdInsP 2 , whereas other known profilin 2 ligands such as VASP or Mena are not released from the complex, suggesting that a different regulatory mechanism must exist to dissociate these molecules from the profilin 2 complexes (Fig. 4B, lower panel).
Profilin 2 Is an Inhibitor of Membrane Uptake-The suppression of SH3 ligand binding to dynamin 1 implies that in vivo profilin 2 should in fact inhibit dynamin 1 and consequently down-regulate endocytosis. To test this hypothesis we performed uptake experiments in HeLa cells as well as in primary neurons isolated from profilin 2 null mice. HeLa cells FIGURE 3. Profilin 2 competes with binding of SH3 ligands to dynamin 1. A, in an ELISA-based assay, solid-phase bound SH3 ligands (Grb2, endophilin, amphiphysin, and Src-SH3) were incubated with dynamin 1 and binding assessed using the dynamin 1-specific hudy-1 antibody. Increasing amounts of profilin 2 compete significantly with dynamin 1 binding, whereas addition of a control protein (BSA) has no effect on dynamin 1 binding (*, p Ͻ 0.05; **, p Ͻ 0.01). B, profilin 2 beads were incubated with brain lysate and the bound complexes released. Eluted proteins were separated by SDS-PAGE and detected by Western blot using antibodies against amphiphysin, dynamin 1, and endophilin. While dynamin 1 binds efficiently to profilin 2, the dynamin 1 ligands amphiphysin and endophilin are excluded and remain in the unbound fraction. were chosen, because they express dynamin 1 (Fig. 5B) but are devoid of profilin 2, thereby providing an experimental model system for adding back profilin 2. Neurons on the other hand allow us to study profilin 2 function in the authentic cell type, which normally expresses this protein.
To follow the profilin 2 distribution and to monitor expression levels, we expressed a DsRed-profilin 2 fusion in HeLa cells. Cell fractionation experiments, as well as actin and poly-L-proline binding studies, showed that DsRed-profilin 2 behaved like the native profilin 1 from HeLa cells that was used as an internal control (data not shown). Expression of DsRed-profilin 2 had no adverse effect on cell shape, cytoskeletal organization, or cell growth. Most of DsRed-profilin 2 was evenly distributed in the cytoplasm, and some localization to focal adhesion sites was seen as it has been reported before for GFP-profilin (9,44). Using live cell confocal microscopy we could observe DsRed-profilin 2 also on vesicles moving on microtubule tracks (data not shown).
To address the question whether profilin 2 can regulate endocytosis, we performed a receptor-mediated uptake assay using FITC-labeled transferrin in a mixed population of HeLa cells expressing different levels of DsRed-profilin 2. Transferrin is internalized through the classic receptor-mediated endocytosis pathway involving clathrin-coated pits as well as dynamin (45). As shown in Fig. 5A, cells expressing DsRedprofilin 2 show a marked decrease in uptake of transferrin when compared with neighboring cells not expressing DsRed-profilin 2. To show that the inhibition of transferrin uptake is due to the binding of profilin 2 to dynamin 1 and not to actin, we performed the uptake assay on HeLa cells transfected with DsRed fusion proteins of profilin 2 mutants S134Y and S138D, which both bind to actin but are deficient for dynamin 1 binding (see Fig. 1D). In contrast to the wild-type profilin 2, both mutants did not affect transferrin uptake (Fig. 5A).
Clathrin-mediated uptake has been shown to be influenced by alterations in the actin cytoskeleton (20), and expression of profilin 2 could potentially have changed the cytoskeletal organization in HeLa cells. However, overexpression of wild-type profilin 2 did not alter the overall organization of the actin cytoskeleton (Fig. 5A, lower panel), and no changes in focal adhesions and cell-cell contacts were observed using antibodies for VASP and vinculin (data not shown). Furthermore, the profilin 2 point mutants bind to actin but had no inhibitory effect on transferrin uptake.
The results obtained in HeLa cells show that ectopic expression of profilin 2 in non-neuronal cells slows down receptor-mediated endocytosis. However, profilin 2 is normally expressed in neurons, and we therefore wanted to verify these results in mouse neurons. For this we took advantage of profilin 2 knock-out mice, which lack any detectable profilin 2. 3 Using cultured cortical neurons we measured endocytosis and membrane turnover of the amphiphilic dye FM 1-43. This dye becomes fluorescent upon insertion into lipid membranes, whereas it is virtually non-fluorescent in the aqueous phase. FM-dyes have therefore been widely used to examine synaptic vesicle recycling in neurons (46). As shown in Fig. 5C, neurons lacking profilin 2 respond with a 2-fold increase of membrane uptake compared with control neurons. We conclude from this result that lack of profilin 2 leads to a gain of function which further corroborates a role of profilin 2 as a negative regulator of endocytosis.

DISCUSSION
Profilins are actin-associated proteins, regulating the dynamics of the actin cytoskeleton. However, profilins, and profilin 2 in particular, associate with a number of proteins involved in membrane trafficking and endocytosis (14). For example, profilin 1 has been found associated with exosomes in dendritic cells (47), as well as the Golgi compartment (48). In Dictyostelium discoideum and Drosophila profilin null mutants show increased phagocytosis (49,50), and in mutant cells overexpression of the endosomal/lysosomal protein DdLIMP can suppress the profilin null phenotype by a yet unknown mechanism (51). Also in yeast profilin appears to be linked to membrane trafficking as suggested by the genetic interaction between S. cerevisiae profilin and Sec3p (52), and defective fluid phase uptake in profilin mutants (53).
These data show that profilins also play a role in membrane trafficking; the molecular mechanism, however, is not well understood yet. Mice and men have four known profilin genes, and apparently the functions of the different profilins have diverged in mammals to either regulate the cytoskeleton or to regulate membrane trafficking. This is most obvious in neurons, which express the ubiquitous profilin 1 as well as the neuron specific profilin 2 gene. Profilin 2 forms complexes with proteins known to be involved in membrane trafficking such as synapsin and dynamin 1 (14). This finding makes profilin 2 a prime candidate 3 A. Di Nardo and W. Witke, unpublished data. The merged image of the two channels shows that cells expressing profilin 2 take up significantly less transferrin. HeLa cells expressing the profilin 2 mutants S134Y and S138D deficient in dynamin 1 binding have no effect on FITC-transferrin uptake (middle panel). Expression of wild-type DsRed-profilin 2 had no effect on cytoskeletal organization of HeLa cells (lower panel). F-actin was visualized by FITC-phalloidin. B, Western blot showing that HeLa cells express dynamin 1 and that HeLa dynamin 1 is comparable to the brain dynamin 1 with respect to binding to profilin 2. C, cortical neurons were cultured from profilin 2 null mice, and the kinetic of membrane uptake measured by using FM 1-43. Neurons lacking profilin 2 (profilin 2 Ϫ/Ϫ) show an increase in FM 1-43 uptake compared with control neurons cultured from wild-type (wt) mice.
among the mammalian profilins for being a regulator of membrane trafficking. Dynamin 1, on the other hand, is a crucial component of the endocytic machinery and synaptic vesicle recycling in neurons; therefore the interaction with profilin 2 is of particular interest.
One important observation is that dynamin 1 binding to profilin 2 requires the poly-L-proline binding site, because not all profilin ligands seem to bind via the poly-L-proline binding site of profilin (54). We have further shown that the binding of profilin 2 to dynamin 1 can occur on several sites in the PRD, but most likely the two stronger binding sites which overlap with the docking sites for other SH3 ligands (Grb2, endophilin, and amphiphysin) are biologically relevant. This is supported by sedimentation experiments of profilin 2-dynamin 1 complexes suggesting that profilin 2 and dynamin 1 form a 2:1 complex.
Surprisingly, actin binding to profilin 2 inhibits the interaction of profilin 2 with dynamin 1, which at first glance seems contradictory, because the actin binding site and the poly-L-proline binding site on profilin 2 are located in distinct domains (55,56). One explanation is that steric hindrance between the large dynamin 1 molecule and actin is responsible for this competition.
Both the profilin 2-actin interaction and the profilin 2-dynamin 1 binding are regulated by PtdInsP 2 . Although PtdInsP 2 had been shown to be able to disrupt the interaction between profilin 1 and synthetic poly-L-proline (57), this is to our knowledge the first report on a physiological profilin complex other than the actin complex being regulated by PtdInsP 2 . Thereby, profilin 2 represents a common downstream target of the phospholipid messenger pathway, releasing dynamin 1 as well as monomeric actin. This means that at sites of PtdInsP 2 production dynamin 1 is made available to form an endocytic complex and at the same time actin is released for polymerization at the membrane (Fig. 6). This mechanism could explain why local changes in PtdInsP 2 are tightly linked to spatially and temporally coordinated endocytosis and actin polymerization.
Endocytosis is affected by dynamic changes in the actin cytoskeleton (20), which could in principle explain why profilin 2 is having an effect on transferrin uptake. However, our results show that the inhibitory effect of wild-type profilin 2 on transferrin uptake depends on an intact poly-L-proline binding site and binding to dynamin 1. Point mutants of profilin 2, which abolish dynamin 1 binding but leave actin binding intact, have no effect on transferrin uptake. Binding of profilin 2 to dynamin 1 can apparently directly silence the activity of dynamin 1 by regulating the molecular interactions of dynamin 1 with SH3-domain-containing adaptor molecules, such as Grb2, endophilin (SH3P4), amphiphysin, Src-kinase, and possibly others. Binding of SH3 ligands stimulate the GTPase activity of dynamin (58) and are important for assembly and recruitment of the endocytic machinery to the membrane (16). We have mapped the endophilin binding site on dynamin 1 to a region where Grb2 and amphiphysin bind and we have further shown that these sites overlap with the profilin 2 docking sites. Competition of profilin 2 with these SH3 ligands would interfere with the function of dynamin 1 and inhibit endocytosis. Such competition can be observed in vitro. In addition, ectopic expression of profilin 2 results in reduced receptor-mediated endocytosis, and ablation of profilin 2 expression in neurons leads to an increase in membrane uptake. An interesting question is what are the benefits of having a neuronal profilin isoform, which inhibits membrane trafficking? One possible explanation is that the inhibition by profilin 2 provides neurons with a means to decrease random vesicle release and to increase the dynamic range of vesicle turnover. This might not be critical in most cell types, but it is essential in neurons to tightly control neurotransmitter release.
Our results show that in mammalian cells profilin binding to ligands other than actin is physiologically meaningful. This is an important conclusion, because in contrast to actin binding, the poly-L-proline binding activity of profilin is still poorly understood, even though the importance of poly-L-proline binding in general has been well documented (10). To date, more than 30 profilin-binding proteins have been FIGURE 6. Proposed model of profilin 2-dynamin 1 interaction and regulation in cells. Profilin 2 is normally bound to actin and the proline-rich domain of dynamin 1. Binding of profilin 2 inhibits recruitment of SH3 ligands such as endophilin or amphiphysin to dynamin 1. Local increase of PtdInsP 2 at the membrane releases profilin 2 from dynamin 1 and allows assembly of the endocytic machinery. Concomitantly, PtdInsP 2 releases G-actin from the profilactin complexes and makes monomeric actin available for filament elongation.
described belonging to such different groups as nuclear proteins, Rac-Rho effector molecules, cell adhesion molecules, and membrane scaffolding proteins (for review see Ref. 13), and it will be critical to learn more about these novel aspects of profilin function. We suggest that one biological function of profilins is to silence ligands by inhibiting their activity or by competing with binding of other molecules. This is true for actin, it is true for the proline-rich ligand dynamin 1, and it is a likely mechanism for regulating other ligands as well. It has been suggested that poly-L-proline binding molecules show a certain degree of degeneracy in binding to their proline-rich ligands (59). Profilin being very promiscuous in its binding to poly-L-proline ligands and being very abundant in cells (up to 0.5% of total cell protein) thus may act as a "poly-proline buffer" for molecules containing signaling domains such as WW, EVH1, and SH3. Further experiments with a larger set of ligands will be required to generalize this hypothesis, but our data as well as previous data obtained for the binding of profilin to Ena-Vasp-like protein Evl are in good agreement with this idea (60).