Phosphoinositides in membrane traffic at the synapse

Ca²⁺-regulated exocytosis of neurotransmitters is the primary mechanism by which neurons communicate with each other and with effector cells. Prior to secretion, neurotransmitters are stored in two distinct classes of vesicle. Non-peptide neurotransmitters, including all fast-acting neurotransmitters (e.g. glutamate, acetylcholine, γ-aminobutyric acid and glycine), accumulate in small vesicles – the synaptic vesicles (SVs) – which represent specialized synaptic organelles. Peptide neurotransmitters, by contrast, are stored in larger vesicles – the dense-core vesicles (DCVs) – which might also contain biogenic amines (e.g. dopamine, norepinephrine, serotonine and histamine). DCVs represent the major secretory organelles of neuroendocrine cells. SVs and DCVs differ in a variety of properties, including stimulus-secretion coupling and biogenesis. However, fundamental mechanisms of exocytosis are common to the two organelles (De Camilli and Jahn, 1990; Sudhof, 1995; Calakos and Scheller, 1996; Hannah et al., 1999).

The availability of procedures to purify SVs, the development of powerful in vitro assays to reconstitute vesicular transport reactions, the identification of targets of clostridial neurotoxins, and advancements in forward and reverse genetics in a variety of organisms have led to the identification of many proteins essential for neuronal secretion. These studies have revealed that the protein machinery involved in exocytosis/endocytosis at the synapse is evolutionary highly conserved and similar to that which participates in exocytosis/endocytosis in all eukaryotic cells (Rothman and Warren, 1994; Schekman, 1994; Wu and Bellen, 1997; Geli and Riezman, 1998; Slepnev and De Camilli, 2000). More generally, membrane fusion and budding mechanisms are fundamentally similar throughout the secretory and endocytic pathways. Thus, the exocytic/endocytic cycle of SVs has become a powerful model system in the field of membrane traffic.

Growing evidence indicates that membrane lipids have not only a structural role but also an important regulatory function in membrane traffic. Phosphoinositides, in particular, are key regulatory molecules (De Camilli et al., 1996; Martin, 1998; Odorizzi et al., 2000). The reversible phosphorylation of their inositol ring generates a series of stereoisomers that can bind to cytosolic and membrane proteins with variable affinities and specificities. Thus, membrane phosphoinositides can nucleate cytoskeletal scaffolds, vesicle coats and signaling complexes. Proteins bind to phosphoinositides by interacting either with protein modules – e.g. PH, SH2, PTB, FYVE and C2 domains – or, less specifically, with clusters of positively charged residues (Janmey, 1994; Bottomley et al., 1998; Martin, 1998; Corvera et al., 1999; Sechi and Wehland, 2000). Here, we focus on the many connections between phosphoinositides and membrane dynamics at the synapse.

Almost fifty years ago, Hokin and Hokin (Hokin and Hokin, 1953) observed that stimulation of secretion from pancreatic acinar cells results in increased phosphorylation of phospholipids. Further experiments demonstrated that this effect is limited to a minor fraction of these lipids, the inositol phospholipids (Larrabee et al., 1963), and that it occurs in most secretory systems, including brain synaptosomes (Durell and Sodd, 1966). These findings led to the hypothesis that phosphoinositide turnover is closely linked to the SV cycle (Hawthorne and Pickard, 1979; Fig. 1). However, the subsequent discovery that products of PtdIns(4,5)P₂ hydrolysis are potent intracellular messengers shifted attention to the role of phosphoinositide turnover in intracellular signaling. Only in the past decade has the role of phosphoinositide metabolism in membrane trafficking, including trafficking at the synapse, returned to center stage, owing to a convergence of information obtained from biochemistry, cell-free assays and yeast genetics (Eberhard et al., 1990; Herman and Emr, 1990; Hay and Martin, 1993; Schu et al., 1993; Hay et al., 1995; McPherson et al., 1996). These studies have shown that some of the effects of phosphoinositides are direct – i.e. not mediated by the their cleavage products – and have revealed important physiological roles of novel phosphoinositide species.

The first piece of evidence for a role for PtdIns(4,5)P₂ in vesicular neuroendocrine secretion, independent of its phospholipase-C-mediated cleavage, came from studies of regulated exocytosis of catecholamines from DCVs of broken chromaffin cells (Holz et al., 1989; Eberhard et al., 1990). Holz et al. showed that an ATP-dependent priming step precedes Ca²⁺-triggered secretion (Holz et al., 1989). Furthermore, PtdIns(4,5)P₂-degrading or -masking agents (phospholipase C and neomycin, respectively) produced the same effects as ATP removal, which suggested that ATP is required, at least in part, for PtdIns(4,5)P₂ synthesis (Eberhard et al., 1990). A search for the cytosolic factors required for the ATP-dependent priming step led to the identification of two enzymes involved in phosphoinositide metabolism: a phosphatidylinositol transfer protein (PITPα) (Hay and Martin, 1993) and a type I PIP kinase, which functions primarily as a PIP(4)P 5-kinase (Hay et al., 1995). These findings, together with the identification of a DCV-associated PtdIns 4-kinase essential for DCV exocytosis (Wiedemann et al., 1996), led to the hypothesis that generation of PtdIns(4,5)P₂ from PtdIns is an important step preceding DCV exocytosis.

PITPα catalyzes the ATP-independent exchange of PtdIns between membrane bilayers in vitro (Wirtz, 1991; Sha and Luo, 1999). Its identification as a factor required for an ATP-dependent step may reflect its reported role in PtdIns(4,5)P₂ synthesis. PITP might replenish substrates for PtdIns(4,5)P₂ generation at specific membrane microdomains at which PtdIns(4,5)P₂ generation is needed for secretion. Alternatively, or in addition, it might play a more direct role in phosphoinositide synthesis by presenting the lipid substrate to lipid kinases (Cunningham et al., 1995; Kearns et al., 1998).

Whereas PI 4-kinase is tightly bound to DCV membranes (Wiedemann et al., 1996; Gasman et al., 1998), PIP 5-kinase activity appears to be primarily associated with the plasma membrane (Wiedemann et al., 1998). Furthermore, recent studies have suggested that the PtdIns(4,5)P₂ pool involved in exocytosis is localized in the plasma membrane (Holz et al., 2000). Thus, it remains to be established whether the PtdIns(4)P generated by the DCV-associated kinase represents the precursor of PtdIns(4,5)P₂ needed for exocytosis. In principle, plasma-membrane-associated PIP 5-kinase could phosphorylate ‘in trans’ DCV-membrane PI(4)P after DCVs have docked at the plasma membrane. Alternatively, another PI(4) kinase, present on the plasma membrane, may come into play. The precise function of PtdIns(4,5)P₂ in DCV docking/fusion reactions remains unknown. Generation of Ins(1,4,5)P₃ and diacylglycerol from PtdIns(4,5)P₂ is not required for exocytosis, since phospholipase C treatment of broken cell preparations even reverses the priming reaction (Eberhard et al., 1990). Furthermore, PI 3-kinase inhibitors do not affect regulated exocytosis, which indicates that PtdIns(3,4,5)P₃ generation from PtdIns(4,5)P₂ is not required for this process (Martin et al., 1997). Hence, PtdIns(4,5)P₂ is likely to act directly on a specific target.

After ATP-dependent priming, progression of DCVs to fusion with the plasma membrane requires Ca²⁺ and cytosolic factors (Walent et al., 1992). In broken PC12 cells (a neuroendocrine cell line), a 145 kDa protein – CAPS (Ca²⁺-dependent Activator Protein for Secretion) – can substitute for cytosol in Ca²⁺-triggered exocytosis (Walent et al., 1992). This protein is abundantly expressed in tissues of neuroectodermic origin. It binds to Ca²⁺ with moderate affinity (Ann et al., 1997) and to PtdIns(4,5)P₂ with high specificity but relatively low affinity in a variety of assays (Loyet et al., 1998). In vivo, a large fraction of CAPS is bound to the plasma membrane and DCVs. CAPS is the orthologue of the product of the nematode unc-31 gene, and its critical role in neurosecretion is supported by defects in serotonin release observed in unc-31 mutants (Avery et al., 1993). These findings make CAPS a potential major effector of PtdIns(4,5)P₂ in regulated secretion of DCVs. However, its precise role in exocytosis is unknown. Microinjection of anti-CAPS antibodies in melanotrophs has shown that CAPS acts at a late stage in the secretory pathway (Rupnik et al., 2000). Binding of CAPS to PtdIns(4,5)P₂ results in the partial penetration of this protein into the lipid bilayer (Loyet et al., 1998). Furthermore, CAPS changes its 3-D structure in response to PtdIns(4,5)P₂ binding, and this effect is reversed by increases in Ca²⁺ concentration to levels found in stimulated secretory cells (Loyet et al., 1998). Altogether, these data suggest that CAPS is part of the Ca²⁺-sensing machinery implicated in the fusion of DCVs with the plasma membrane.

Evidence for a requirement for PtdIns(4,5)P₂ synthesis in SV secretion is not as compelling as in DCV secretion, and conflicting reports have been published. Wiedemann et al. showed that, as in the case of regulated secretion of catecholamines from DCVs, incubation of synaptosomes with inhibitors of inositol phospholipid phosphorylation impairs glutamate release (Wiedemann et al., 1998). However, these drugs might have additional effects, which thus limits the conclusions that can be drawn from their use. Furthermore, similar experiments conducted in another laboratory led to opposite results: norepinephrine secretion was dependent on phosphoinositide phosphorylation, whereas glutamate and GABA secretion was not (Khvotchev and Sudhof, 1998). The CAPS protein described above was reported to be selectively localized on DCVs and to be absent from SVs. Indeed two groups have shown that CAPS is not required for glutamate release from SVs in assays involving semi-lysed synaptosomes (Berwin et al., 1998; Tandon et al., 1998). Studies in nematodes, however, revealed that unc-31 mutants not only accumulate serotonin (Desai et al., 1988), which can be co-stored with neuropeptides in DCVs, but also have striking locomotion defects (Avery et al., 1993), resistance to the acetylcholinesterase inhibitor aldicarb and sensitivity to the acetylcholine agonist levamisole (Miller et al., 1996). These observations may implicate nematode CAPS in regulated secretion from SVs as well.

Several proteins that have been implicated in neurotransmitter release, including release from SVs, bind to phosphoinositides in vitro and thus provide indirect evidence for some role for these lipids in SV exocytosis. C2 domains, which are Ca²⁺-and acidic-phospholipid-binding modules (Rizo and Sudhof, 1998), are present in SV-associated proteins (e.g. synaptotagmin, rabphilin, DOC and Munc13) and in proteins of the cytoskeletal scaffold that anchors SVs to the plasma membrane (e.g. rim) (Wang et al., 1997) and piccolo (Fenster et al., 2000). Two of these proteins, synaptotagmin and rabphilin, have been shown to bind to PtdIns(4,5)P₂ directly (Schiavo et al., 1996; Chung et al., 1998). Synaptotagmin, an intrinsic SV membrane protein, has a putative role in the Ca²⁺ regulation of neurotransmitter release (Geppert and Sudhof, 1998), and one of the effects of Ca²⁺ on synaptotagmin in vitro is to act as a switch regulating its relative preference for PtdIns(4,5)P₂ or PtdIns(3,4,5)P₃ (Schiavo et al., 1996). The functional relevance of this binding for exocytosis is underlined by results of experiments at the squid giant synapse. Microinjection of antibodies directed against the C2B domain of synaptotagmin prevents the inhibition of neurotransmitter release induced by co-injection of inositol high-polyphosphates (IPPs), which function as competitive inhibitors of phosphoinositides (Fukuda et al., 1995; Mochida et al., 1997). Rabphilin is an effector for the SV-associated GTPase Rab3a and is recruited to SVs by Rab3a-GTP. Like synaptotagmin, it binds PtdIns(4,5)P₂ by means of its C2B domain. A peptide from the phosphoinositide-binding region of rabphilin inhibits DCV exocytosis from permeabilized chromaffin cells (Chung et al., 1998). However, rabphilin-knockout mice do not show any obvious defect in neurotransmission (Schluter et al., 1999). Recently, Augustin et al. reported a dramatic reduction of the readily releasable pool of glutamate in mice lacking another C2-domain-containing presynaptic protein, Munc13-1, and have suggested a role for Munc13-1 in SV priming (Augustin et al., 1999). Munc13-1 could therefore be a target for the action of PtdIns(4,5)P₂ in the priming reaction of exocytosis, but it has not yet been shown to bind phosphoinositides.

An interesting class of membrane-associated PtdIns(4,5)P₂-binding protein at the pre-synapse is the Mint family. Mint1 and Mint2 participate in SV exocytosis by interacting with Munc18-1/N-Sec1 (Okamoto and Sudhof, 1997) and in the structural organization of synaptic junctions by interacting with the CASK-neurexin complex (Butz et al., 1998). Munc18-1, a syntaxin-binding protein, has a fundamental function in SNARE-mediated fusion, because SV exocytosis, both spontaneous and evoked, is completely absent in Munc18-1-knockout mice in spite of the absence of major defects in synaptic organization and structure (Verhage et al., 2000).

In spite of evidence supporting a role for phosphoinositides in neurosecretion, genetic studies in Drosophila and yeast caution about concluding that inositol phospholipids play a fundamental, rather than regulatory, role in exocytosis. In Drosophila, disruption of a PIP 5-kinase gene – skittles – produces embryonic lethality and major alterations in the cytoskeletal organization of neurons but no defects in either SV or DCV secretion, as assessed by electrophysiological recordings at the Drosophila larval body-wall neuromuscular junction (Hassan et al., 1998). However, at least one other homologue of mammalian PIP 5-kinases (i.e. type I PIP kinases) is present in Drosophila (gene CG3682 at http://flybase.bio.indiana.edu), which could compensate for loss of skittles.

In yeast, disruption of MSS4, which encodes the only PI(4)P 5-kinase in this organism, produces defects in the actin cytoskeleton but not in secretion (Desrivieres et al., 1998; Homma et al., 1998). However, inactivation of the yeast PIK1, which encodes a PI 4-kinase, is lethal, and temperature-sensitive mutants of this gene have severely impaired secretion and endocytosis at restrictive temperature (Hama et al., 1999; Walch-Solimena and Novick, 1999; Audhya et al., 2000). The yeast orthologue of frequenin, a Ca²⁺-binding regulatory protein present in nerve terminals, is an essential co-factor for Pik1p (Hendricks et al., 1999). Strikingly, frequenin overexpression in Drosophila results in chronic facilitation of transmitter release at the larval neuromuscular junction and multiple firing of action potentials (Pongs et al., 1993; Angaut-Petit et al., 1998).

After exocytosis, SV membranes are rapidly internalized and reused for the generation of new SVs. A major pathway for their reformation involves a specialized form of clathrin-mediated endocytosis (Heuser and Reese, 1973; Gonzales-Gaitan and Jackle, 1997; De Camilli et al., 2000), and phosphoinositides appear to play an important role at several steps of this pathway.

Clathrin-mediated endocytosis is thought to start with the binding of the heterotetrameric clathrin adaptor complex AP-2 and of the accessory clathrin adaptor protein AP180 to both protein and lipids in the plasma membrane. Subsequently, the adaptors recruit clathrin and promote its assembly to form the coat (Cremona and De Camilli, 1997; McMahon, 1999; Brodin et al., 2000; Kirchhausen, 2000).

Studies on Ca²⁺ signaling provided the first evidence for a role for phosphoinositides in clathrin coat recruitment. The identification of the Ins(1,4,5)P₃ receptor and the characterization of biosynthetic pathways that generate other inositol polyphosphates (IPPs) led to a search for novel IPP interactors. Unexpectedly, components of the synaptic clathrin coat, namely AP-2 and AP180, and synaptotagmin were isolated as major IPP-binding proteins (Beck and Keen, 1991; Timerman et al., 1992; Voglmaier et al., 1992; Fukuda et al., 1994; Niinobe et al., 1994; Ye et al., 1995). Subsequent studies demonstrated that these proteins also bind phosphoinositides, suggesting that, at least in the case of adaptors, these interactions may function in coat recruitment to the bilayer (Fig. 2; Gaidarov et al., 1996; Hao et al., 1997; Rapoport et al., 1997). Ye et al. mapped the phosphoinositide-binding site to a cluster of positive residues in the N-terminal ENTH-like domain of AP180 (Ye et al., 1995) and the N-terminal region of the α-adaptin subunit of AP-2 (Gaidarov et al., 1996; Gaidarov et al., 1999). As discussed above, binding of phosphoinositides to synaptotagmin is mediated by its C2B domain, which also acts as a main membrane-docking site for the α-adaptin subunit of AP-2 (Zhang et al., 1994; Haucke and De Camilli, 1999).

The physiological role of these interactions is supported by a variety of studies. PITP is required for clathrin-dependent SV biogenesis in a cell-free assay (Schmidt and Huttner, 1998).

Furthermore, expression in wild-type fibroblastic cells of an α-adaptin that has mutations in its phosphoinositide-binding site results in the assembly of AP-2 complexes that do not bind to the plasma membrane and, as a result, have dominant negative effect on clathrin-mediated transferrin uptake (Gaidarov et al., 1999). Masking of PtdIns(4,5)P₂ by exogenous molecules inhibits early steps of clathrin-mediated endocytosis (Jost et al., 1998). Finally, the critical importance of PtdIns(4,5)P₂ synthesis in the recruitment of AP-2/clathrin coats to a physiological membrane was demonstrated in cell-free studies using endosomal and lysosomal membranes as templates (West et al., 1997; Arneson et al., 1999). Although AP-2 has a higher affinity for PtdIns(3,4,5)P₃ than PtdIns(4,5)P₂ (Gaidarov et al., 1996), there is no evidence so far for a role of PtdIns(3,4,5)P₃ in synaptic vesicle recycling. It has been shown, however, that a PI 3-kinase isoform (class II phosphoinositide 3-kinase C2α) is associated with clathrin-coated vesicles from bovine brain. This raises the possibility that PtdIns(3,4,5)P₃ plays a role in some clathrin-dependent budding reactions (Domin et al., 2000).

Liposomes alone can support the assembly of endocytic clathrin coats, which provides direct evidence for interactions between lipids and coats. In these experimental conditions, acidic phospholipids are strictly required for coat formation, and the presence of phosphoinositides enhances coat recruitment (Takei et al., 1998; Cremona et al., 1999). The interactions of adaptors with membrane proteins or membrane lipids are likely to be synergistic. Examples of positive cooperativity between the two types of interaction are the reported effect of 3′-phosphorylated phosphoinositides on the binding of endocytic motifs to AP-2 and the effect of phospholipase D – possibly mediated by phosphatidic acid or PtdIns(4,5)P₂ – on the recruitment of AP-2 to synaptotagmin (Rapoport et al., 1997; Haucke and De Camilli, 1999). Cooperativity of lipid-and protein-binding sites on the membrane might play a critical role in vivo in the regulation of clathrin coat nucleation in time and space. The generation of phosphoinositides at a specific membrane location could trigger coat assembly at a protein binding site which, by itself, is insufficient to nucleate assembly.

Besides AP-2 and AP-180, other proteins implicated in clathrin-mediated membrane internalization bind to phosphoinositides. Among these factors are the ubiquitous non-visual arrestins (β-arrestins), a family of proteins that mediate G-protein-coupled receptor (GPCR) sequestration to endosomes. β-Arrestins bind to clathrin and AP-2, as well as to IPPs, PtdIns(4,5)P₂ and PtdIns(3,4,5)P₃, with high affinity (Gaidarov et al., 1999). Expression of β-arrestin mutants lacking phosphoinositide-binding properties results in marked inhibition of β-adrenergic receptor internalization (Gaidarov et al., 1999). So far, there is no evidence for a role of β-arrestin in SV recycling. However, these findings suggest that phosphoinositide binding represents a common theme in clathrin-adaptor recruitment to the plasma membrane (Gaidarov and Keen, 1999).

Two well-characterized clathrin accessory proteins that participate in SV recycling at the synapse are amphiphysin and endophilin (Slepnev and De Camilli, 2000). Both proteins bind to lipids, and amphiphysin affinity for membranes is enhanced by the presence of phosphoinositides (Cremona et al., 1999). Amphiphysin also binds to clathrin (McMahon et al., 1997; Ramjaun et al., 1997; Slepnev et al., 1998), AP-2 (Wang et al., 1995; Slepnev et al., 1998), dynamin (David et al., 1996) and another clathrin-accessory protein, synaptojanin 1 (McPherson et al., 1996) (see below). Thus, it might function as an adaptor that coordinates the binding of other endocytic proteins to the membrane. Endophilin is critical for the biogenesis of SVs in a cell-free assay (Schmidt et al., 1999) and plays an essential role in the invagination of clathrin-coated pits in vivo (Ringstad et al., 1999). Like amphiphysin, endophilin binds dynamin and synaptojanin and might be involved in their recruitment to sites of endocytosis (de Heuvel et al., 1997; Ringstad et al., 1997). Endophilin also functions as a lysophosphatidic acid acyl-transferase, mediating the synthesis of phosphatidic acid (Schmidt et al., 1999). Potential roles of this enzymatic activity include a direct effect on membrane curvature – given the different geometry of lysophosphatidic acid and phosphatidic acid in the plane of the membrane (Schmidt et al., 1999) – but also a stimulatory effect on PtdIns(4,5)P₂ production, through stimulation of PIP 5-kinases by phosphatidic acid (Jenkins et al., 1994).

Constriction and fission of deeply invaginated buds to generate free vesicles requires the GTPase dynamin, which is a phosphoinositide-binding protein (the properties and function of this protein have been extensively reviewed; De Camilli and Takei, 1996; Hinshaw, 1999; McNiven et al., 2000; Sever et al., 2000). Dynamin forms a scaffold around the vesicle stalk. This scaffold may function in fission either directly, via a GTP-hydrolysis-dependent conformational change that cleaves the vesicle neck, or indirectly, by recruiting and/or regulating other proteins (Kosaka and Ikeda, 1983; Hinshaw and Schmid, 1995; Takei et al., 1995; Sweitzer and Hinshaw, 1998; Takei et al., 1998). However, the presence of dynamin on growing clathrin buds (Takei et al., 1996) and the absence of clathrin coats in Drosophila mutants of dynamin (Kosaka and Ikeda, 1983) suggest additional roles of this molecule in coat formation and/or stabilization.

The phosphoinositide-binding region of dynamin is localized in its PH domain, which binds several phosphoinositides, in particular PtdIns(4,5)P₂ (Barylko et al., 1998). Phosphoinositides considerably boost GTPase activity of dynamin (Zheng et al., 1996) and show cooperativity in this function with SH3-domain-containing proteins (Barylko et al., 1998) and with conditions that promote dynamin assembly (Klein et al., 1998). However, regulation of GTPase activity is unlikely to be the only or the main effect of phosphoinositides on dynamin. Transfection of dynamin mutants lacking the PH domain or of point mutants that exhibit impaired PtdIns(4,5)P₂ binding inhibits receptor-mediated endocytosis (Barylko et al., 1998; Lee et al., 1999; Vallis et al., 1999). This inhibition can be rescued by deletion of the proline-rich region of the molecule, which plays a critical role in targeting dynamin to its sites of action, possibly by interactions with SH3-domain-containing proteins. Thus, it would appear that binding of dynamin to PtdIns(4,5)P₂ is crucial after its targeting to endocytic sites (Vallis et al., 1999).

Phosphoinositides could stabilize the anchoring of dynamin to membranes after its recruitment by SH3-domain-containing proteins; alternatively, the PH domain of dynamin could recruit phosphoinositides to the neck of invaginated buds to favor membrane bending and/or fission. Evidence suggests that amphiphysin and endophilin are functional partners of dynamin in the fission reaction (Schmidt and Huttner, 1998; Takei et al., 1999). At least some of the actions of these two proteins may also be mediated by their direct interactions with lipids (Takei et al., 1999; Farsad et al., 2000).

The hypothesis that phosphoinositides play an important role in the recruitment and function of endocytic proteins at the synapse has recently been supported by the identification and characterization of the polyphosphoinositide phosphatase synaptojanin 1. Synaptojanin 1 is enriched at nerve terminals and interacts with a variety of proteins of the endocytic machinery (McPherson et al., 1996; Roos, 1998; de Heuvel et al., 1997; Haffner et al., 1997; Ringstad et al., 1997; Sakisaka et al., 1997; Yamabhai et al., 1998; Qualmann et al., 1999). It hydrolyzes several phosphoinositide species, including PtdIns(4,5)P₂ and PtdIns(3,4,5)P₃ (McPherson et al., 1996; Chung et al., 1997; Woscholski et al., 1997; Cremona et al., 1999; Guo et al., 1999). Given its enzymatic activity, synaptojanin 1 is expected to be a negative regulator of coat-membrane interactions during uncoating, and this prediction was confirmed genetically (Cremona et al., 1999).

Nerve terminals of synaptojanin-1-knockout mice (Cremona et al., 1999) contain an increased number of clathrin-coated vesicles in the actin-rich area that surrounds SV clusters. Furthermore, cell-free studies on brain cytosol of wild-type and synaptojanin-1-knockout mice have revealed that the knockout cytosol is more potent in promoting coat recruitment to liposomes under conditions in which phosphoinositides undergo turnover. This effect correlates with an accumulation of PtdIns(4,5)P₂ and PtdIns(3,4,5)P₃ on these artificial membranes. However, the major phosphoinositide species increased in neurons of knockout animals is PtdIns(4,5)P₂ (Cremona et al., 1999). Consistent with these observations, disruption of synaptojanin 1 recruitment and/or function by antibody and peptide microinjection at lamprey giant synapses produces an accumulation of clathrin-coated vesicles and pits (Gad et al., 2000). Finally, an increased number of clathrin-coated structures is observed in nerve terminals of C. elegans unc26 mutants, which harbor mutations in the only synaptojanin-like gene of this organism (Harris et al., 2000). Besides an increase in clathrin coats, a hypertrophy of the actin-based cytoskeleton that surrounds the active zone was very evident in injected lamprey synapses (Gad et al., 2000) (see below). A plausible explanation for the accumulation of clathrin-coated membranes is that, in vivo, synaptojanin 1 plays a role in clathrin uncoating by decreasing the affinity of the adaptors for the plasma membrane. Such a hypothesis fits with the observation that the ATPase Hsc70, which is critically required for clathrin uncoating, is not sufficient to remove the adaptors (Hannan et al., 1998). Other morphological alterations clearly evident in the lamprey and/or C. elegans model systems (e.g. defects in fission and an enhanced actin cytoskeleton) might reflect the pleiotropic roles of PtdIns(4,5)P₂ and its important action in actin nucleation (see below and Sakisaka et al., 1997); alternatively, they might be secondary effects due to the sequestration of coat components and their accessory factors on membranes.

Given the hypothesis that phosphoinositides play a role in the nucleation of clathrin-coated vesicles, these phospholipids might undergo regulated synthesis prior to endocytosis. An attractive possibility is that the same pool of phosphoinositides generated in preparation for exocytosis (see above and Fig. 3) participates in the coating reaction. This pool might then be supplemented by additional phosphoinositide synthesis occurring after exocytosis is completed.

Studies of budding reactions on a variety of intracellular membranes have implicated small GTPases as critical switches for the initiation of coat recruitment; more specifically, ARF family members appear to be involved in the nucleation of clathrin coats (Springer et al., 1999). Some of the effects of ARF, in turn, are mediated by phosphoinositides. Current models predict that coat recruitment starts when a guanine nucleotide exchange factor (GEF) for ARF (the signature for which is the zpresence of a Sec7-homology domain) generates ARF-GTP (Serafini et al., 1991; Goldberg, 1998; Mossessova et al., 1998), ARF-GTP binds to membranes and promotes coating by a series of independent but synergistic mechanisms, including direct binding to coat proteins – as shown for the COP1 coat (Goldberg, 1998) – or the recruitment and/or activation of enzymes that enhance production of phosphoinositides. For example, ARF binds to and activates phospholipase D (Roth et al., 1999), which in turn generates phosphatidic acid, a potent activator of PIP 5-kinases (Jenkins et al., 1994). ARF proteins can also recruit PI 4-and PIP 5-kinases (Godi et al., 1999; Honda et al., 1999; Jones et al., 2000). Phosphoinositides generated by these reactions, in turn, may participate in a positive feed-forward loop to stimulate phospholipase D further (Liscovitch and Cantley, 1995; Singer et al., 1997) or to stimulate ARF itself through additional recruitment of ARF GEFs that contain phosphoinositide-binding PH domains (Jackson and Casanova, 2000). These results and considerations suggest the existence of a phosphoinositide phosphorylation/dephosphorylation cycle that acts downstream of the GTP/GDP cycle of ARF-family GTPases (Fig. 3).

Although ARF6 was proposed to have a role in clathrin-mediated endocytosis (D’Souza-Schorey et al., 1995), it is unclear whether ARF family members function in clathrin-mediated SV recycling. So far, no ARF proteins have been found to be concentrated at the synapse, and brefeldin A, which blocks a subset of ARF-GEFs, does not block clathrin-mediated SV recycling (Mundigl et al., 1993; Shi et al., 1998). However, proteins that act upstream or downstream of ARF are present in nerve terminals, and recently at least a brefeldin-A-independent ARF-GEF, Msec7, has been identified in this compartment (Ashery et al., 1999; Neeb et al., 1999). Msec7-1 is the rat homologue of human cytohesin-1 and belongs to a subset of brefeldin-A-insensitive ARF-GEFs that includes ARNO, GRP1/ARNO3 and GBF1. Notably, these proteins have been shown to bind to phosphoinositides and to be recruited to the plasma membrane upon local production of PtdIns(3,4,5)P₃ (Jackson and Casanova, 2000). Thus, the existence of an ARF-like protein in nerve terminals is plausible. Furthermore, components of the endocytic machinery, namely AP180 and amphiphysin, inhibit phospholipase D (Lee et al., 1997; Lee et al., 2000) and might help to shut down the positive feedback loop depicted in Fig.

Synaptojanin 1, which cleaves PtdIns(4,5)P₂, could also function as a very powerful inhibitor of this feedback loop by acting as a switch-off signal at the end of the endocytic reaction. Finally, the involvement of an ARF-like GTPase in SV endocytosis is supported by the strong stimulatory effect of GTPγS on clathrin coat recruitment (Takei et al., 1995; Gustaffson et al., 1998). However, in these experiments it is difficult to discriminate between a bona fide stimulatory effect of GTPγS on coating mediated by an ARF-like protein or other G proteins and an indirect stabilization of clathrin coats due to inhibition of the fission reaction. An important priority for future studies will be the precise identification of the enzymes responsible for phosphoinositide biosynthesis in nerve terminals and their regulation by GTPases.

Many proteins that function in the regulation of actin dynamics bind to phosphoinositides, and local production of phosphoinositides on membranes is a mechanism to generate focal nucleation of actin (Janmey, 1994; Martin, 1998; Sechi and Wehland, 2000). A major pathway through which phosphoinositides can promote actin nucleation involves their cooperation with Rho-type GTPases in recruiting WASP family members to the plasma membrane (Rohatgi et al., 1999; Zigmond, 2000). A convergence of biochemical and morphological studies have recently suggested that the actin cytoskeleton has an important role in the SV cycle (Gustaffson et al., 1998; De Camilli et al., 2000; Qualmann et al., 2000) and therefore raised the possibility that phosphoinositide turnover plays a critical role in the control of this actin pool.

Strands of actin have been detected within the SV cluster (Hirokawa et al., 1989), which is consistent with the ability of synapsin, a major SV-associated protein (De Camilli et al., 1990), to bind to actin. However, the bulk of presynaptic actin is localized at the periphery of the vesicle cluster. This localization is particularly evident at large synapses, such as the frog neuromuscular junction or the synapses of the giant reticulospinal axon of the lamprey, where filamentous actin can be visualized with a sufficient level of resolution by light microscopy using fluorescent phalloidin (Brodin, 1999; Dunaevsky and Connor, 2000; Gad et al., 2000). Genetic studies in C. elegans and Drosophila have shown that these actin-rich zones, which surround ‘active zones’ of secretion, play an important role in defining the structure of the synapse (Schaefer et al., 2000; Wan et al., 2000; Zhen et al., 2000). Furthermore, they are the sites at which the majority of clathrin-coated vesicles form (Heuser and Reese, 1973; Roos and Kelly, 1999; Gad et al., 2000). It is therefore interesting that some of the proteins involved in SV endocytosis also appear to have a role in actin function (Mundigl et al., 1998; Witke et al., 1998; Ochoa et al., 2000; Qualmann and Kelly, 2000). Significantly, in yeast, all forms of endocytosis are critically dependent on actin (Geli and Riezman, 1998).

Synaptojanin 1 function is strongly linked to actin dynamics. As discussed above, disruption of synaptojanin 1 function at the synapse results not only in an impairment of the SV cycle but also in accumulation of actin around active zones; this suggests that an equilibrium between polymerization and depolymerization of actin has been altered (Sakisaka et al., 1997; Gad et al., 2000). These effects can be explained by a corresponding imbalance between PtdIns(4,5)P₂ synthesis and dephosphorylation. Studies of the giant lamprey synapse under a variety of experimental conditions have shown that this actin pool is highly dynamic and that its polymerization is enhanced by nerve terminal stimulation (Gustaffson et al., 1998; Brodin, 1999). The precise role of actin in the endocytosis reaction is still unclear. One of its roles may be to propel the nascent vesicle away from the membrane back to the vesicle cluster.

In the classical model of clathrin-mediated endocytosis, an endosomal sorting station is the target for endocytic vesicles after uncoating. Phosphoinositide metabolites, in particular PtdIns(3)P, which binds to FYVE domains, play a critical role in endosomal function in a variety of systems (Corvera et al., 1999; Stenmark and Aasland, 1999; Wurmser et al., 1999). However, the involvement of endosomes, and therefore of PtdIns(3)P, in SV recycling is questionable. There is evidence that an endosomal sorting station can be bypassed during SV reformation and that newly uncoated vesicles are targeted directly to the SV cluster (Takei et al., 1996; Murthy and Stevens, 1998). The few bona fide endosomes present at the synapse might mediate housekeeping membrane recycling and sorting – possibly via the AP3-dependent pathway (Shi et al., 1998) – of the fraction of SV proteins that escapes direct recycling. From these endosomes, some proteins are targeted to the axon for retrograde flow in multivescicular bodies. It will be of interest to determine whether homologues of the yeast Fab protein (a PI(3)P 5-kinase) participate in the biogenesis of these organelles, as Emr and co-workers have proposed for the biogenesis of homologous organelles in yeast (Odorizzi et al., 1998).

Endosome-like compartments do accumulate in nerve terminals after massive stimulation, but they are thought to form by bulk endocytosis from the plasma membrane (Takei et al., 1996). Thus, they may function in parallel with the plasma membrane as ‘donors’ of clathrin-coated vesicles rather than as ‘acceptors’ of newly uncoated vesicles. Bulk endocytosis is thought to be an actin-dependent reaction (Schmalzing et al., 1995), and is thus another process that might involve phosphoinositide-mediated actin polymerization.

Several features make phosphoinositides powerful modulators of membrane traffic and membrane-cytoskeleton interactions, in addition to their well-established roles as precursors of intracellular second messengers. First, the reversible phosphorylation of their inositol group can function in the regulated and reversible recruitment of cytosolic proteins to membranes, much like tyrosine phosphorylation of membrane proteins. Second, the phosphorylated state of the entire population of inositol phospholipids in a membrane microdomain could in principle be changed very rapidly owing to positive feedback loops built into phosphoinositide metabolic pathways (such as that involving ARF, phospholipase D and phosphoinositide kinases; see Fig. 3). Third, one can predict from structural properties of phosphoinositide-metabolizing enzymes that at least some of them act processively – that is, they undergo several cycles of catalysis without leaving the membrane (Scott et al., 1990; Scott et al., 1994; Rao et al., 1998). In the case of a vesicle, such a mechanism could allow the modification of the composition of the entire vesicle membrane by one or few enzyme molecules. Fourth, owing to the synergistic roles of phosphoinositides and membrane proteins, generation of specific phosphoinositide species can be used to temporally and spatially regulate the recruitment of cytosolic proteins.

Because of these properties, phosphoinositides can function as very powerful signals to tag a membrane for a given fate. Thus, much like GTPases, phosphoinositides can function in regulating membrane traffic and in defining the vectoriality of transport. Not surprisingly, a reciprocal relationship between these two regulatory mechanisms has been identified.

Here, we have summarized growing evidence linking phosphoinositide metabolism to the control of neurosecretion and the SV cycle. An important role for PtdIns(4,5)P₂ has clearly emerged both in the cascades of reactions leading to exocytosis and in the compensatory endocytic mechanisms. Other phosphoinositides probably play roles in presynaptic function. The dual effects of phosphoinositides on coat formation and actin nucleation strongly suggest that these two processes are linked. Some protein targets for the actions of phosphoinositides have been identified, and genetic disruption of normal phosphoinositide metabolism has been shown to affect the vesicle cycle. It remains to be understood whether phosphoinositides act as essential switches or simply as critical regulatory components. The mild phenotype observed in yeast after disruption of PtdIns(4,5)P₂-synthesizing enzymes speaks against a fundamental role of this phosphoinositide species in the exocytic reaction. However, further use of genetics in higher organisms will be required if we are to answer this question. Given the high cytosolic levels of certain IPP species, it will be important to understand the potential physiological role of competition between IPPs and phosphoinositides in binding to specific proteins and the possible independent roles of IPPs. It will also be important to define how the effects of phosphoinositides on actin and membrane traffic discussed here impact on intracellular signaling pathways controlled by phosphoinositide metabolites. Although, we anticipate that these two actions of phosphoinositides are interrelated, this remains an unexplored area of research.

We thank Drs Markus Wenk, Niels Ringstad, Gilbert Di Paolo and Lorenzo Pellegrini for discussion and critical review of the manuscript. Work reviewed in this commentary and carried out in the lab. of the authors was supported in part by grants NIH NS36251 and CA46128 to PDC and by grants from Telethon (project D61 and D111) and from MURST (COFIN97 and COFIN2000) to O.C.

Recent crystallographic papers have revealed the structural basis for the interaction of phosphoinositides with the N-terminal domain of the clathrin coat protein AP180/CALM, now shown to have the fold of an ENTH domain (Ford et al., 2001; Mao et al., 2001). Furthermore, the ENTH domain of epsin, another clathrin coat accessory protein, was shown to bind PtdIns(4,5)P₂ (Itoh et al., 2001).