Synapsin I-associated phosphatidylinositol 3-kinase mediates synaptic vesicle delivery to the readily releasable pool.

Maintaining synaptic transmission requires replenishment of docked synaptic vesicles within the readily releasable pool (RRP) from synaptic vesicle clusters in the synapsin-bound reserve pool. We show that synapsin forms a complex with phosphatidylinositol 3-kinase (PI 3-kinase) in intact nerve terminals and that synapsin-associated kinase activity increases on depolarization. Disruption of either PI 3-kinase activity or its interaction with synapsin inhibited replenishment of the RRP, but did not affect exocytosis from the RRP. Thus we conclude that a synapsin-associated PI 3-kinase activity plays a role in synaptic vesicle delivery to the RRP. This also suggests that PI 3-kinase contributes to the maintenance of synaptic transmission during periods of high activity, indicating a possible role in synaptic plasticity.

Maintaining synaptic transmission requires replenishment of docked synaptic vesicles within the readily releasable pool (RRP) from synaptic vesicle clusters in the synapsin-bound reserve pool. We show that synapsin forms a complex with phosphatidylinositol 3-kinase (PI 3-kinase) in intact nerve terminals and that synapsinassociated kinase activity increases on depolarization. Disruption of either PI 3-kinase activity or its interaction with synapsin inhibited replenishment of the RRP, but did not affect exocytosis from the RRP. Thus we conclude that a synapsin-associated PI 3-kinase activity plays a role in synaptic vesicle delivery to the RRP. This also suggests that PI 3-kinase contributes to the maintenance of synaptic transmission during periods of high activity, indicating a possible role in synaptic plasticity.
A typical nerve terminal in the central nervous system contains about 200 -250 synaptic vesicles (SVs). 1 These can be functionally divided into a small readily releasable pool (RRP) and a large reserve pool. The RRP contains less than 5% of total nerve terminal SVs and is defined morphologically as predocked SVs at the active zone and functionally as SVs that are primed and immediately available for the initial rapid phase of neurotransmitter release (1). The remainder constitutes the reserve pool, which ensures that a continual supply of SVs is available for delivery to the RRP and is drawn upon to maintain neurotransmission during periods of intense or prolonged stimulation (2)(3)(4). The reserve pool SVs surround the active zone and are clustered together in an actin-based cytoskeletal matrix via an interaction with synapsins (3,5).
There are five synapsin proteins (Ia, Ib, IIa, IIb, and III) from three genes (I, II, and III) (6), which maintain the reserve pool SVs as a cluster and prevent their free dispersal within the terminal. Synapsins I and II control SV replenishment of the RRP, because conditions that disrupt their function result in dispersal or depletion of SV clusters near the active zone and enhance synaptic depression (3,5,7). In contrast, disruption of the much less abundant synapsin III does not affect SV clusters and reduces synaptic depression (8). SV liberation from clusters within the actin cytomatrix is mediated by the multi-site phosphorylation of synapsin I and II, which decreases their affinity for SVs or actin (6). Disruption of synapsin I phosphorylation sites thus reduces the number of SVs undergoing exocytosis (9). It is unknown how liberated SVs translocate to the RRP, and the mechanisms governing entry into the RRP (docking/priming) are incompletely understood (10).
Synapsin I has a C-terminal proline-rich domain that interacts with src homology 3 (SH3) domains present in a large number of proteins, including the p85 subunit of phosphatidylinositol 3-kinase (PI 3-kinase) (6,11). A precise role for PI 3-kinase in nerve terminal function has not been established (12)(13)(14), despite that its lipid products play numerous roles the actin cytoskeleton and in vesicle trafficking (15,16). Synapsin I also connects vesicle traffic with the actin cytoskeleton and the synapsin-p85 in vitro interaction suggests that PI 3-kinase activity might play a role in SV recycling if it occurred in nerve terminals. We examined glutamate release from isolated nerve terminals (synaptosomes) and hippocampal CA1 neurons, and our results implicate synapsin-associated PI 3-kinase activity in the delivery of SVs from the reserve pool to the RRP during periods of high activity or prolonged stimulation.
MALDI-TOF MS-Proteins that bound to the GST-SH3 domains were excised from gels and submitted to in-gel tryptic digestion and analyzed by MALDI-TOF MS (20). Proteins were identified with the search engines Protein Prospector and Mascot, and a MOWSE score above 10 3 was considered significant.
Immunohistochemistry-B104 rat brain neuroblastoma cells (22) were grown (1,000 cells/ml) on poly-D-lysine-coated glass cover slips in a 12-well plate in Dulbecco's modified Eagle's medium containing 10% fetal calf serum. Cells were differentiated with 1% horse serum and 1 mM dibutyryl cAMP for 6 days with fresh media every 2 days. Cells were fixed with 4% paraformaldehyde for 20 min, washed with cold phosphate-buffered saline, permeablized with 0.1% SDS with 1% bovine serum albumin for 15 min, and blocked with 1% bovine serum albumin for 45 min. Synapsin Ia goat polyclonal IgG (Santa Cruz Biotechnology) and PI 3-kinase p85-␣ mouse monoclonal IgG1 (Santa Cruz Biotechnology) were used at 1:250 overnight at 4°C. Wells were washed and incubated for 30 min in the dark with both fluorescein isothiocyanateconjugated affinity purified donkey anti-goat IgG (Jackson Immunology Research) for synapsin and Texas Red-conjugated affinity purified goat anti-mouse IgG (Jackson Immunology Research) for p85. After placing a coverslip over the glass slide cell staining was observed using a Leica DM LB microscope at 100ϫ magnification. Controls revealed no color bleeding between channels.

Synapsin I and PI 3-Kinase Interact in Nerve Terminals-
Synapsin associates with many SH3 domain-containing proteins, including Grb2, amphiphysins I and II, and p85 (11). To identify selective binding partners for PI 3-kinase in brain we performed sequential pull-down experiments using GST fusion proteins containing the SH3 domains of p85, Grb2, amphiphysin I (A1), amphiphysin II (A2) and identified interacting proteins by MALDI-TOF mass spectrometry. This sequential approach was aimed at pre-clearing strongly interacting partners like dynamin I, which may bind all the targets. Using GST-Grb2-SH3 first, the major protein extracted from brain homogenates was the highly abundant dynamin I (Fig. 1A). The homogenate was extracted twice more with Grb2-SH3 until no more dynamin I was recovered, then sequentially extracted with the other recombinant SH3 domains. When p85-SH3 extraction followed Grb2-SH3, synaptojanin, and synapsins Ia and Ib were present (Fig. 1A). Subsequent extraction with A1-SH3 or A2-SH3 recovered more synaptojanin and additional unique proteins (Fig. 1A). When the order of extraction was altered such that p85-SH3 was utilized last, an association with synapsins Ia and Ib was still observed, despite prior extraction with three other SH3 domains. Therefore the p85-SH3-synapsin association is highly selective when compared against other SH3 domain proteins and p85 may be a key synapsin-binding protein in nerve terminals.
To identify PI 3-kinase binding partners in nerve terminals, we performed pull-down experiments in synaptosomes using recombinant p85-SH3 or Grb2-SH3 (Fig. 1B). p85-SH3 primarily extracted synaptojanin, dynamin I, and synapsin Ia/b. Grb2 also extracted synaptojanin and dynamin I, but its association with synapsin was barely detectable, highlighting the relative selectivity of the p85-SH3-synapsin interaction. Protein identification was confirmed by MALDI-MS (not shown) and by Western blotting (Fig. 1C). To determine whether the phosphorylation status of the extracted proteins regulate their binding, pull-down experiments were performed in metabolically labeled, resting (Ctrl), deploarized (S1), repolarized (Repol), or redepolarized (S2) synaptosomes. No difference in the amount of protein bound was observed ( Fig. 1B and data not shown) indicating these interactions are independent of their phosphorylation status (Fig. 1D).
To determine whether synapsin I and PI 3-kinase form a complex in intact nerve terminals, we performed a series of co-IPs. IP of p85 revealed synapsin I, as did IP of the catalytic subunit of PI 3-kinase, p110 ( Fig. 2A). Conversely, synapsin antibodies immunoprecipitated p85. Synapsin and p85 also co-localize to SV clusters in specific puncta in the neurites of B104 neuronal cells differentiated with cAMP (Fig. 2, B and C). Thus PI 3-kinase is found in the correct subcellular location for a role in SV recycling.
To determine whether synapsin associates with an active PI 3-kinase complex, we measured PI 3-kinase activity in synapsin I IPs (Fig. 2D). A high level of activity immunoprecipitated with synapsin I in resting nerve terminals. This activity increased about 2-fold after S1 and was reversed when the stimulus was removed (Fig. 2D, Repol). Thus synapsin and PI 3-kinase form an active complex in nerve terminals that is regulated by stimulation.
PI 3-Kinase Regulates Glutamate Release from the Reserve Pool-Next we examined the functional role of PI 3-kinase activity in neurotransmitter release from synaptosomes. Glutamate is the predominant neurotransmitter in synaptosomes, with greater than 80% of nerve terminals being glutamatergic (23). KCl-evoked Ca 2ϩ -dependent glutamate release was strongly inhibited when synaptosomes were preincubated with either 1 M Wortmannin or 100 M LY294002 (Fig. 3A). Neither inhibitor had any effect on KCl-evoked Ca 2ϩ -independent glutamate release (data not shown), ruling out effects on the plasma membrane glutamate co-transporter (23). The same results were obtained with 100 nM Wortmannin when preincubation times were increased to 60 min (data not shown).
Fast and slow phases of glutamate release have been characterized in synaptosomes and represent release from the RRP or RRP replenishment from reserve pool SVs, respectively (24,25). Remarkably, Wortmannin or LY294002 exclusively inhibited the slow phase of Ca 2ϩ -dependent glutamate release (Fig.  3A). We compared a number of inhibitors with different mechanisms of action for their relative effects on the two pools. We arbitrarily defined the release originating from the reserve pool as glutamate released after 90 s of stimulation, with glutamate release prior to 10 s being from the RRP and calculated a ratio (Ͼ90 s/Ͻ10 s). A ratio of greater than one indicates preferential inhibition of the RRP, whereas a ratio less than one indicates a block of SV delivery from the reserve pool. The only antagonists that affected this ratio were Wortmannin and LY294002, which reduced the ratio value to 0.50 Ϯ 0.02 and 0.36 Ϯ 0.04, respectively (Fig. 3B). This indicates that PI 3-kinase activity may be required for the delivery of reserve pool SVs to the RRP during prolonged depolarization.
To ensure that the requirement for PI 3-kinase activity is upstream from the RRP, SVs from the RRP were selectively released by application of hypertonic sucrose (26). This protocol only releases docked and primed SVs and does not require Ca 2ϩ influx. Application of hypertonic sucrose in the presence or absence of extracellular Ca 2ϩ resulted in the release of 2.5 Ϯ 0.3 or 2.4 Ϯ 0.2 nmol/mg glutamate, respectively, an amount expected for release exclusively from the RRP. When synaptosomes were preincubated with Wortmannin or LY294002, no inhibition of sucrose-evoked glutamate release was observed either in the presence or absence of extracellular Ca 2ϩ (Fig.  3C). Therefore PI 3-kinase activity is not required for the fusion of SVs that are already present in the RRP.
The inhibition of a late phase of glutamate release from synaptosomes correlates with a block of SV delivery from the reserve pool. To determine whether a similar inhibition might occur during periods of high synaptic transmission in situ, we made whole-cell patch-clamp recordings of EPSCs from hippocampal CA1 pyramidal neurons after a prolonged train of repetitive stimulation. When a prolonged repetitive stimula-

FIG. 1. Synapsin specifically associates with the SH3 domain of PI 3-kinase.
A, sequential pull-downs with GST-SH3 domains from Grb2, p85, amphiphysin 1 (A1), and amphiphysin 2 (A2) from whole rat brain. A single brain extract was used to sequentially isolate binding proteins for each GST-SH3 domain in the order indicated. Bands marked by dots were unequivocally identified by MALDI-TOF mass spectrometry. Note for the lower gel, synaptojanin was selectively removed from the extract by A1 at the second step, leaving none to associate with p85. B, pull-downs with GST-Grb2 or GST-p85 from synaptosomal lysates. There are no obvious protein differences between pull-downs from synaptosomes at rest (Ctrl) or deploarized for 10 s with 30 mM KCl (S1). Arrows indicate synaptojanin (SJ), dynamin I (Dyn I), or synapsin I a/b (Syn Ia/b). Results are shown in duplicate. C, immunoblotting of pull-downs from synaptosomal lysates using GST-p85 identifies synaptojanin, dynamin I, and synapsin I. D, the association of synaptojanin, dynamin I, or synapsin I a/b with GST-Grb2-SH3 and GST-p85-SH3 is unaffected by their phosphorylation status. The autoradiograph shows the characteristic stimulus-dependent dephosphorylation of SJ and Dyn I and phosphorylation of Syn I during two consecutive cycles of stimulus (S1 and S2) and intervening recovery (Repol). Results are shown in duplicate. tion at relatively lower frequency is applied, depletion of the RRP is usually faster than replenishment from the reserve pool, leading to a gradual decline of EPSC amplitude (4). Synaptic stimulation of pyramidal neurons at 10 Hz for 60 s resulted in a gradual decay of EPSC amplitude to Ϫ44 Ϯ 3% (n ϭ 5) by the end of stimulation (Fig. 3D). In the same neurons, after a 10 min bath-application of LY294002 (50 M), 10 Hz stimulation caused a more rapid decay of EPSCs and resulted in significantly greater depression (Ϫ60 Ϯ 2%, n ϭ 3, p Ͻ 0.05). In both control and LY294002-treated cells 10 Hz stimulation caused no long-term change in synaptic efficacy. Also there was no significant difference in the size of EPSCs between control (249 pA) and drug-treated (248 pA) cells at the beginning of stimulation, indicating LY294002 has no effect on basal neurotransmission (27,28). This suggests that the RRP in LY294002-treated synapses was not efficiently replenished by the delivery of reserve pool SVs, resulting in a faster depletion of SVs during a prolonged train of repetitive stimulation. Genetic disruption of synapsins results in a similar enhancement of synaptic depression due to loss of reserve pool SV clusters (3,5,29).
The PI 3-Kinase-Synapsin Interaction Is Required for SV Delivery-We next addressed whether the synapsin-PI 3-kinase interaction plays a role in the late phase of glutamate release. We designed a potential inhibitory peptide (SynI 585-600) similar to one previously used to block the synapsin-p85 association in vitro (11). The peptide abolished the interaction of synapsin I with p85-SH3, but not that of dynamin I or synaptojanin (Fig. 4, A and B). This is important because dynamin I is a major ligand of the p85-SH3 domain in nerve terminals (Fig. 1B). Two additional synapsin I peptides that contain potential SH3-binding motifs were without effect (Fig.  4, A and B).
The penetratin-SynI 585-600 peptide inhibited only the late phase of Ca 2ϩ -dependent glutamate release (Fig. 5A). It did not affect Ca 2ϩ -independent glutamate release (data not shown), a reliable indicator that nerve terminal viability was not compromised. The selective inhibition of reserve pool glutamate release by penetratin-SynI 585-600 was clearly indicated when exocytosis was separated into early (release Ͻ10 s, inhibition of 6.1 Ϯ 12.4%, n ϭ 4 Ϯ S.E.) and late components (release Ͼ90 s, inhibition of 40.4 Ϯ 3.8%, n ϭ 4 Ϯ S.E.). This is reflected in the reserve/RRP ratio (Fig. 5B). The peptide was without effect on RRP release evoked by hypertonic sucrose (Fig. 5C). To control for potential side-effects of the penetratin leader sequence, peptides identical to the clathrin-binding site on AP-2 (AP-2 ␤2 624 -644 ) (31) were prepared with the same leader sequence (Fig. 5F). AP-2 ␤2 624 -644 had no effect on glutamate release (Fig. 5D) and blocked endocytosis, as expected (Fig. 5E). Thus the data suggest that the direct interaction of PI 3-kinase with synapsin I is important for the delivery of SVs to the RRP but not for the fusion of SVs present in the RRP.
The first stage of SV delivery is liberation of SVs from the reserve pool (6). Because site-specific phosphorylation of synapsin I controls this process, PI 3-kinase may mediate liberation by altering synapsin phosphorylation, either directly or via a downstream protein kinase such as Akt (6). However, we FIG. 2. Synapsin and p85 form a complex in nerve terminals. A, a series of co-IPs reveal a protein complex between synapsin I and PI 3-kinase in nerve terminals. IPs with antibodies to p85 or p110 recovered synapsin I, whereas an IP of synapsin recovered p85. B, immunofluorescence in B104 neurites using synapsin (green) and p85 (red) antibodies. Yellow in the merged image shows co-localization at puncta of clustered SVs but no specific co-localization in the spaces between these clusters. C, co-localization of p85 and synapsin at contacts (long arrow) between neurites from separate B104 cells as well as at neurite branch points (short arrow). No specific co-localization occurred in the neurites other than the clusters of SVs. D, synapsin I IPs from resting (Ctrl), deploarized (S1), or repolarized (Repol) synaptosomes were assayed for PI 3-kinase activity. Data are a percent of PI 3-kinase activity immunoprecipitated from resting synaptosomes, n ϭ 4 Ϯ S.E.
found that neither Wortmannin nor LY294002 had any effect on the depolarization-dependent phosphorylation of synapsin I in synaptosomes, ruling out this possibility (Fig. 6).

DISCUSSION
Our data reveal a new role for PI 3-kinase in the delivery of SVs from the reserve pool to the RRP for exocytosis. This role is mediated by an association between PI 3-kinase and synapsin I, via p85. This is the first report of a specific locus of action of PI 3-kinase activity in SV exocytosis. Our conclusions are based on a selective inhibition of the late phase of glutamate release by both PI 3-kinase inhibitors and a peptide that blocks the synapsin-p85 interaction. This inhibition is not due to a block in SV recycling because none of these treatments had any effect on SV endocytosis (not shown). The early and late phases of glutamate release are well characterized in synaptosomes and almost all treatments that reduce exocytosis do so equally from both phases (24,25). A similar selective inhibition of KCl- evoked glutamate release is also observed in synaptosomes derived from synapsin I knockout mice (5), demonstrating the synapsin-dependent nature of this late phase. Thus the interaction of PI 3-kinase with synapsin during this phase may be important for targeting PI 3-kinase to phosphorylate inositol lipids either on the SV or on the plasma membrane. The most likely explanation for our results is that production of PI(3,4,5)P 3 is necessary for the delivery of SVs to the RRP after their liberation from the reserve pool. SV delivery has two steps, translocation and entry into the RRP. Translocation from SV clusters in the reserve pool to the RRP may occur by passive diffusion, or mechanisms may exist to direct a subset of SVs to release sites. SV entry into the RRP is a two-stage process, mediated by SV docking at active zones followed by a priming event that confers fusion competence on the SV. It is unlikely that the synapsin-PI 3-kinase association plays a role in SV entry into the RRP as synapsin itself is not required for this step (6). PI 3-kinase mediates the specificity of vesicle delivery at various stages of vesicle recycling pathways in other cells; however, different isoforms play distinct roles. Type III PI 3-kinase mainly produces PI (3)P and confers specificity of docking of vesicles at endosomes (32,33). Type II PI 3-kinase plays a role in clathrin-coated bud formation at the plasma membrane and at the trans-Golgi surface via its association with clathrin (34). Type I PI 3-kinases, as described in this paper, produce PI(3,4,5)P 3 as their major in vivo product. PI(3,4,5)P 3 binds many proteins found in nerve terminals, including phosphoinositide-dependent kinase-1, protein kinase B, protein kinase C isoforms, phospholipase C-␥, dynamin I, synaptotagmin, myosin X, AP-2, and neurogranin (35), none of which have been implicated in SV delivery. However, other possible PI(3,4,5)P 3 effectors exist such as cytohesin/msec7-1, a guanine nucleotide-exchange factor for ADP-ribosylation factor 6 (ARF6). Activation of ARF6 by msec7-1 is PI(3,4,5)P 3 -dependent (36). Overexpression of msec7-1 augments transmitter release in Xenopus neuromuscular junctions by increasing the size of the RRP (37). Msec7-1 also interacts with munc 13-1, which is targeted to active zones and implicated in priming (38). Thus regulation of the activation/inactivation cycle of ARF6 by PI 3-kinase might control the entry of SVs into the RRP at a priming step in peripheral neurons (39).
A role for synapsin-associated PI 3-kinase activity upstream from the RRP may reveal a common cellular role for PI 3-kinase in exocytosis and vesicle delivery. In non-neuronal cells, which lack synapsin, its role in vesicle supply is managed Synapsin I was isolated from lysed synaptosomes using a GST-p85-SH3 attached to GSH beads that had either been resting (Ctrl), stimulated with 30 mM KCl (S1), or repolarized (Repol). No difference in recovery or phosphorylation status of synapsin I was observed. Lanes are in duplicate. through the actin cytoskeleton in a synapsin-independent manner (15,16). PI 3-kinase plays key roles in regulation of the actin cytoskeleton (40 -42). For example it directly associates with and activates p21-activated kinase-1, which is downstream of the small GTPases Rac and Cdc42 (43) and directly binds active Rac or Rho (44,45). PI 3-kinase also interacts with profilin (46) (which also binds PI(3,4,5)P 3 ), increasing the interaction of this complex with actin (47). In chromaffin cells, PI 3-kinase inhibitors block exocytosis by specifically affecting actin assembly (40). Our results link these effects with the role of synapsin I in the actin cytoskeleton and provide a framework for how these proteins may work together at synapses. Thus synapsin serves as an anchor regulating the localization and activation of PI 3-kinase at SV clusters to regulate SV trafficking from the actin cytoskeleton. In support, p110-␣ is highly enriched in the growth cones and filopodia of cultured primary hippocampal neurons where it is coincident with enriched Factin (48).
Our results indicate a new role for synapsin I-associated PI 3-kinase activity in glutamate release from central nerve terminals. PI 3-kinase inhibition also enhanced the run-down of synaptic transmission on prolonged repetitive stimulation. Such run-downs are thought to represent synaptic depression, which is one form of synaptic plasticity that is also enhanced in mice lacking synapsin I. This suggests the synapsin-p85 interaction may be involved in preventing synaptic depression and might have a presynaptic role in synaptic plasticity (13,27,28,48). It will be important in future studies to identify the downstream targets of PI(3,4,5)P 3 to provide more information on the specific locus of action of PI 3-kinase in SV exocytosis.