Janus-faced trafficking at glutamatergic synapses

Rapid communication in the nervous system requires the release and detection of neurotransmitters at synapses. These two events occur at opposite faces of the synaptic cleft and involve very different cellular machinery in presynaptic and postsynaptic compartments. Such asymmetry, manifested by synaptic vesicle exocytosis and recycling at the presynaptic terminal and receptor activation and generation of ionic currents at the postsynaptic membrane, provides the foundation for unidirectional information propagation in all neural circuitry.

Because of their distinct organization and molecular composition, mechanisms governing the function of presynaptic and postsynaptic compartments have traditionally been the object of independent inquiry. Paradigmatic in this regard is membrane trafficking, the sine qua non of presynaptic function. In presynaptic terminals, the recycling of synaptic vesicles has long been known to require clathrin-mediated endocytosis (1). Much more recently, it has become clear that membrane trafficking is not the sole province of the presynapse. Indeed, at excitatory glutamatergic synapses in the mammalian central nervous, the abundance of postsynaptic receptors is subject to exquisite regulation by membrane trafficking (2). Yet the molecular machinery for postsynaptic trafficking, and the degree to which such machinery mirrors its presynaptic counterparts, remains enigmatic. Now, a paper in this issue of PNAS from the laboratory of Pietro De Camilli (3) reports that the phosphoinositide phosphatase synaptojanin-1, previously shown to regulate synaptic vesicle endocytosis (4–7), performs double duty by controlling the internalization and abundance of AMPA-type glutamate receptors at the postsynaptic membrane.

Gong and De Camilli (3) began by examining excitatory synaptic transmission in cultured hippocampal neurons derived from synaptojanin-1-knockout (SJ1-KO) mice. To measure synaptic transmission, they performed electrophysiological recordings of miniature excitatory synaptic currents (mEPSCs) to detect spontaneous glutamate release events. From this analysis, they found that mEPSCs from SJ1-KO neurons are much larger in amplitude than mEPSCs recorded from wild-type (WT) neurons. Because the recorded currents are mediated by ionotropic AMPA-type glutamate receptors, these results suggest an increase in the abundance of AMPA receptors at the synapse. Consistent with this notion, nonstationary fluctuation analysis of mEPSCs, a statistical method that extracts single-channel information from the fluctuating variation in recorded current, indicated an increase in the absolute number of activated AMPA channels but no change in single-channel conductance. Similarly, whole-cell recordings of glutamate-evoked responses revealed a significant increase in AMPA receptor responses, but no increase in the response mediated by NMDA receptors, a pharmacologically distinct subtype of glutamate-gated ion channel also found at excitatory synapses.

One potential problem with these experiments is that both the postsynaptic and presynaptic neurons in the culture lack synaptojanin-1, and changes in electrophysiological responses recorded postsynaptically could conceivably be due to changes in glutamate release from the presynaptic neuron. Indeed, SJ1-KO neurons have a well-known defect in neurotransmitter release upon prolonged stimulation because of impaired synaptic vesicle recycling (4–7). To address this point, Gong and De Camilli (3) first used a pharmacological strategy to chronically block excitatory synaptic transmission with the AMPA receptor antagonist 6-cyano-7-nitroquinoxaline-2,3-dione (CNQX). This manipulation has been found to increase mEPSC frequency and amplitude and increase the size and turnover rate of the recycling pool of synaptic vesicles (8). The authors reasoned that if the increased mEPSC amplitude measured in SJ1-KO neurons was due to decreased presynaptic release efficiency, this difference should be abolished by prolonged CNQX treatment. However, mEPSC amplitudes were larger upon CNQX blockade in both WT and SJ1-KO neurons, arguing against a presynaptic mechanism. More compelling support for a postsynaptic effect in the SJ1-KO neurons was provided by rescue experiments, where Gong and De Camilli found that re-introduction of a WT synaptojanin-1 cDNA into postsynaptic SJ1-KO neurons reduced mEPSC amplitudes and thus reversed the knockout phenotype.

How does postsynaptic synaptojanin-1 reduce the number of synaptic AMPA receptors? As mentioned above, synaptojanin-1 plays an important role in the clathrin-dependent recycling of synaptic vesicles (4–7). Enzymatically, synaptojanin-1 is a polyphosphoinositide phosphatase that dephosphorylates phosphatidylinositol 4,5-bisphosphate [PI(4,5)P₂] (4, 6) (Fig. 1 A and B). Several clathrin adaptors and endocytic accessory proteins bind PI(4,5)P₂ at the inner leaflet of plasma membrane, events required to nucleate or assemble the clathrin endocytic apparatus (9). After vesicle scission, hydrolysis of PI(4,5)P₂ on the endocytic vesicle allows the release of clathrin adaptors for their reuse and to enable subsequent vesicle trafficking (Fig. 1C). Such dephosphorylation of PI(4,5)P₂ is mediated by the two inositol phosphatase domains of synaptojanin-1. The central 5-phosphatase domain of synaptojanin-1 dephosphorylates PI(4,5)P₂ at the 5 position to produce PI(4)P, whereas the N-terminal SacI domain dephosphorylates PI(4)P to produce PI (6, 10) (Fig. 1 A and B). Taking advantage of point mutations that abolish the phosphatase activity of both the 5-phosphatase and SacI domains of synaptojanin-1, Gong and De Camilli (3) transfected SJ1-KO neurons with either WT or enzymatically inactive mutant synaptojanin-1 and observed no phenotypic rescue by mutant synaptojanin-1. Thus, the phosphatase activity of synaptojanin-1 is required to reduce mEPSC amplitude.

Fig. 1.

Synaptojanin-1 regulates both pre- and postsynaptic endocytosis. (A) Domain organization of synaptojanin-1. From N terminus to C terminus synaptojanin-1 contains a SacI domain that hydrolyzes polyphosphoinositides at the 4 position, a 5-phosphatase domain, and a proline-rich domain (PRD) that associates with endocytic proteins. (B) Synaptojanin-1 catalyzes the removal of both the 4- and 5- position phosphates (yellow circles) from PI(4,5)P₂ to generate PI. (C) Synaptojanin-1 is proposed to facilitate the uncoating of endocytic adaptors from clathrin-coated vesicles by degrading PI(4,5)P₂ (yellow circles). (D) Model for the mirror functions of synaptojanin-1 in both pre- and postsynaptic endocytosis. Within presynaptic terminals (upper), synaptojanin allows recycling of synaptic vesicles. In dendritic spines (lower), NMDA receptor (NMDAR) activation allows Ca²⁺ influx and, through unknown mechanisms, may trigger uncoating of clathrin-coated vesicles (CCVs) by synaptojanin-1 (SJ1), thereby promoting endocytosis of AMPA receptors (AMPARs).

Based on these findings, one attractive possibility is that synaptojanin-1, via its ability to degrade PI(4,5)P₂ and release membrane-associated endocytic proteins, sustains the endocytosis of AMPA receptors postsynaptically, and this in turn reduces synaptic strength (Fig. 1D). Notably, AMPA receptors undergo continuous endocytosis and recycling on a time scale of minutes (11, 12). Such endocytosis is thought to occur at specialized spine endocytic zones (EZs) (13, 14) linked to the postsynaptic density (PSD) via the dynamin isoform dynamin-3 and the PSD adaptor molecule homer (15). Near this EZ is the endocytic adaptor and extended Fer-CIP4 homology (EFC)/FCH–BIN1, amphiphysin, and RVS167 (F-BAR) protein syndapin-1/pacsin-1 (16), as well as endophilin (17), both binding partners of synaptojanin-1. To determine whether the increased mEPSC amplitudes of SJ1-KO neurons are due to impaired endocytosis of AMPA receptors, Gong and De Camilli (3) used a pH-sensitive GFP fusion of the GluR2 AMPA receptor (pH-GluR2) to optically monitor surface AMPA receptors. As expected, NMDA causes a rapid and robust decrease in pH-GluR2 fluorescence in WT neurons because of internalization and acidification-induced quenching of the pH-sensitive GFP. This NMDA-induced internalization was severely blunted in SJ1-KO neurons. These results provide strong evidence that synaptojanin-1 is required for both constitutive and NMDA receptor-induced endocytosis of AMPA receptors, and more broadly implicate postsynaptic PI(4,5)P₂ metabolism in regulating the strength of glutamatergic synapses (Fig. 1).

More study is clearly needed. One of the most striking aspects of the results of Gong and De Camilli (3) is the conservation of synaptojanin-1's role on both sides of the synapse. Yet the functional consequence of augmented or impaired synaptojanin-1 will be quite different and perhaps even complementary depending on the pre- or postsynaptic locus. Whereas SJ1-KO neurons have reduced synaptic vesicle recycling that decreases synaptic efficacy with trains of stimuli, they also exhibit more synaptic AMPA receptors, which increases synaptic strength. Such duality highlights the difficulty in ascribing physiological effects to pre- or postsynaptic changes based on global loss of function. Another key question will be how the action of synaptojanin-1 is regulated postsynaptically, and whether signaling pathways that initiate or sustain synaptic plasticity impinge on synaptojanin-1. In that regard, it is interesting to note that the Ca²⁺-dependent protein phosphatase calcineurin dephosphorylates and activates synaptojanin-1 (18) and likewise stimulates AMPA receptor endocytosis associated with long-term depression (LTD) of synaptic strength (12, 19). Within dendrites and spines, regulation of endocytosis by synaptojanin-1 could control the flux of cargo into endosomal compartments, which serve as reserve pools for AMPA receptors mobilized during long-term potentiation (LTP) (20).

Despite intense study in model mammalian cell systems, yeast, and presynaptic terminals, little is known of the molecular mechanisms that localize, organize, and control the endocytic machinery of dendrites and spines. An emergent principal of synapse modification is that regulation of postsynaptic receptor number and content is central to learning-related plasticity in diverse synaptic circuits (2). Yet a conundrum is how individual synapses rapidly adjust receptor composition to enact learning-related plasticity while maintaining weighted synaptic strengths encoded by receptor content to stably store information. Tight tuning of spatially restricted endocytic machinery provides an attractive substrate for spine-specific modification. By revealing synaptojanin-1 as a key regulator of AMPA receptor endocytosis and AMPA receptor-mediated synaptic transmission, Gong and De Camilli (3) have highlighted the Janus-faced nature of membrane trafficking on both sides of the synaptic divide.