PACSIN1 regulates the dynamics of AMPA receptor trafficking

Dynamic trafficking of AMPA receptors (AMPARs) into and out of synapses plays an important role in synaptic plasticity. We previously reported that the protein kinase C and casein kinase II substrate in neurons (PACSIN) forms a complex with AMPARs through its interaction with the protein interacting with C-kinase 1 (PICK1) to regulate NMDA receptor (NMDAR)-induced AMPAR endocytosis and cerebellar long-term depression. However, the molecular mechanism by which PACSIN regulates the dynamics of AMPAR trafficking remains unclear. Using a pH-sensitive green fluorescent protein, pHluorin, tagged to the extracellular domain of the GluA2 subunit of AMPARs, we demonstrate dual roles for PACSIN1 in controlling the internalization and recycling of GluA2 after NMDAR activation. Structure and function analysis reveals a requirement for the PACSIN1 F-BAR and SH3 domains in controlling these NMDAR-dependent processes. Interestingly, the variable region, which binds to PICK1, is not essential for NMDAR-dependent GluA2 internalization and is required only for the correct recycling of AMPARs. These results indicate that PACSIN is a versatile membrane deformation protein that links the endocytic and recycling machineries essential for dynamic AMPAR trafficking in neurons.


Results and Discussion
PACSIN is required for AMPAR endocytosis and recycling after NMDAR activation. We have previously used the conventional antibody-feeding technique to demonstrate that shRNA-mediated knockdown of the neuron-specific PACSIN1 reduces the number of internalized GluA2-containing AMPARs 15 min post-NMDA stimulation 28 . However, we could not rule out a role of PACSIN1 in AMPAR recycling because the accumulation of intracellular receptors measured using this assay reflects the balance between the amount of receptor internalization from the plasma membrane and the recycling of these receptors back to the cell surface. In order to study PACSIN1 function in controlling the dynamics of AMPAR internalization and recycling, we took advantage of pH-GluA2 to visualize the trafficking of AMPARs with high temporal resolution in living hippocampal neurons 19,30,31 . This reporter was constructed by fusing pHluorin to the extracellular domain of the GluA2 subunit. When endocytosis occurs, the fluorescence of pH-GluA2 initially decreases due to the quenching of pHluorin by the acidic environment inside the lumen of endosomes, but subsequently recovers as pH-GluA2 recycles back to the plasma membrane. In control neurons transfected with pSuper empty vector, perfusion of 20 μ M NMDA (in the presence of 1 μ M tetrodotoxin) for 3 min resulted in a significant quenching of pH-GluA2 fluorescence in the soma and dendrites due to receptor internalization, which fully recovered within 30 min upon NMDA washout (Fig. 1a,b). The decrease of pH-GluA2 fluorescence intensity was not due to the leakage of protein due to neuronal damage as we never observed any change in cytosolic mCherry fluorescence in the recorded neurons (Fig. 1a). However, the amount of internalized pH-GluA2 was significantly reduced in neurons that overexpressed two independent PACSIN1 shRNA sequences compared with control neurons (Fig. 1a-c). In addition to the endocytosis defect, the rate of pH-GluA2 recycling following NMDA washout was significantly accelerated in PACSIN1 knockdown neurons (Fig. 1d,e). The impairment in NMDA-induced pH-GluA2 internalization and recycling caused by PACSIN1 shRNA#1 could be fully rescued by the overexpression of shRNA#1-resistant PACSIN1 in hippocampal neurons ( Fig. 1b-e). Overexpression of PACSIN1 alone did not affect the dynamics of AMPAR trafficking after NMDAR activation ( Fig. S1a-d). Furthermore, to ascertain that the effect of PACSIN1 knockdown was specific for pH-GluA2, we examined the response of cytoplasmic pHluorin following NMDA treatment, which is also known to induce acidification of the intracellular compartment 32 . Unlike pH-GluA2, the kinetics of cytosolic pHluorin were indistinguishable between neurons transfected with vector only, PACSIN1 shRNA#1 or a PACSIN1 overexpression construct (Fig. S2a-c). Moreover, the internalization of pH-GluA2 is effectively blocked by two independent endocytosis inhibitors, dynasore and dynole (Fig. S3). This result suggests that the altered kinetics of pH-GluA2 in PACSIN1 knockdown neurons are likely to be due to the trafficking of AMPARs rather than NMDA-induced intracellular acidification. PACSIN1 therefore plays dual roles in controlling the endocytosis and recycling of AMPARs following NMDA stimulation.
Structure and function analysis of PACSIN in regulating AMPAR dynamics. Next, we sought to understand the underlying mechanism of action by performing structure and function analysis of PACSIN1. Using a molecular replacement strategy, we determined the molecular requirement for each functional domain of PACSIN1 in regulating NMDAR-dependent AMPAR endocytosis and recycling (Fig. 2a). Recent structural studies have identified a unique loop of hydrophobic residues (Ile-122 and Met-122) that protrude from the membrane interaction surface of the PACSIN1 F-BAR domain 33 . This so-called "wedge-loop" is believed to be inserted into a membrane leaflet to promote membrane bending and is thought to contribute to the formation of budding vesicles from the plasma membrane during endocytosis. Replacing these two hydrophobic residues with negatively charged amino acids completely abolishes the ability of the PACSIN1 F-BAR domain to bind lipids and tubulate liposomes in vitro [33][34][35] . In contrast to PACSIN1 wild-type (WT) expression, replacing endogenous PACSIN1 with the I122E,M123E wedge-loop mutant failed to rescue the pH-GluA2 endocytosis and recycling deficit in PACSIN1 knockdown neurons ( Fig. 2b-d). These data demonstrate that the membrane binding and deformation activities of the PACSIN1 F-BAR domain are critical for the activity-dependent trafficking of AMPARs following NMDAR activation in neurons.
The SH3 domain of PACSIN1 is a well-known protein-protein interaction module that binds to the large GTPase dynamin and the actin cytoskeleton regulator, N-WASP (neural Wiskott-Aldrich syndrome protein), both of which play crucial roles in endocytosis 36 . To determine the importance of the PACSIN1 SH3 domain, we examined the effect of the P434L mutant, which disrupts all PACSIN1 SH3-dependent interaction with its proline-rich domain-containing binding partners, including dynamin and N-WASP. We found that substituting endogenous PACSIN1 with the P434L mutant also caused impairment in AMPAR endocytosis and recycling ( Fig. 2b-d). Similarly, deletion of the entire SH3 domain (PACSIN1 Δ SH3) phenocopied the effect of the PACSIN1 P434L mutant (Fig. 2b-d). Given the established involvement of dynamin and the actin cytoskeleton in AMPAR trafficking [37][38][39] , these results indicate a potential role of PACSIN1 as a scaffold protein linking dynamin and actin polymerization to promote vesicle fission and provide mechanical forces to propel these vesicles away from the plasma membrane.
Next, we investigated the effect of the PACSIN1 N364/376D mutation in the NPF motifs within the variable region, which is known to disrupt interaction with Eps15 homology domain (EHD) protein 1, also known as Rme1 40 , a crucial regulator of AMPAR recycling in neurons 41 . Interestingly, the PACSIN1 N364/376D mutant displayed comparable endocytosis to pH-GluA2 but an accelerated rate of reinsertion after NMDA washout ( Fig. 2b-d). This result demonstrates that the PACSIN1− EHD1 interaction is not essential for NMDA-induced GluA2 endocytosis, but plays a pivotal role in GluA2 recycling.
The PICK1−PACSIN interaction is specifically required for the NMDAR-dependent recycling of AMPARs. The variable region of PACSIN1 is also known to interact with PICK1, although the binding is independent of NPF motifs 28 . To assess the role of the PICK1− PACSIN1 interaction in the dynamic trafficking of AMPARs, we replaced endogenous PACSIN1 with the PACSIN1 tmE mutant, in which Ser-343, Ser-345 and Ser-346 are mutated to glutamates, and which displays reduced binding to PICK1 28 (Fig. 3a). The expression of the PACSIN1 tmE mutant did not affect NMDA-induced pH-GluA2 internalization, but produced an increased rate of recycling ( Fig. 3b-d). In contrast, the effect of the PACSIN1 tmA (alanine) mutant was indistinguishable from that of PACSIN1 WT (Fig. 3b-d). These data suggest that the PACSIN1− PICK1 interaction is specifically required to regulate the rate of GluA2 recycling.
To further confirm these results, we assessed the effect of PACSIN1 knockdown in primary neurons derived from either PICK1 WT or knockout mice. PICK1 knockout neurons displayed normal endocytosis of pH-GluA2, but an accelerated rate of recycling upon NMDA washout (Fig. 4a-d), consistent with a role of PICK1 in regulating AMPAR recycling by retaining internalized GluA2 in endosomes [16][17][18][19][20] . We found that PACSIN1 knockdown was able to reduce the amplitude of pH-GluA2 endocytosis in both PICK1 WT and knockout neurons (Fig. 4a,b), suggesting that the role of PACSIN1 in regulating AMPAR endocytosis is independent of PICK1 function. However, loss of PACSIN1 function in PICK1 knockout neurons did not further enhance the rate of pH-GluA2 recycling (Fig. 4a,b), indicating that both PICK1 and PACSIN1 regulate pH-GluA2 recycling through a common pathway.
To confirm that PACSIN1 regulates the NMDA-induced internalization of AMPARs independent of PICK1, we performed an antibody-feeding assay with myc antibodies to measure the degree of intracellular accumulation of endocytosed myc-GluA2 subunits in live transfected hippocampal neurons over time. In control neurons, myc-GluA2 internalization could be observed at 10 min time point after a 5 min treatment with NMDA (50 μ M NMDA + 1 μ M tetrodotoxin), with the remaining myc-GluA2 in the intracellular compartment over the next 10 min (Fig. 5a-d). In PICK1 knockdown neurons, the same proportion of surface myc-GluA2 was internalized 10 min after NMDA treatment (Fig. 5a-d). Consistent with the role of PICK1 in prolonging intracellular retention of AMPARs 16-20 , these internalized receptors fully recycled back to the plasma membrane within the next 10 min (Fig. 5a-d). Importantly, loss of PACSIN1 function caused a significant reduction in myc-GluA2 internalization in PICK1 wild-type and knockdown neurons (Fig. 5a-d). These results provide an independent confirmation to support our previous findings.
In conclusion, we have demonstrated dual roles of PACSIN1 in controlling the NMDAR-dependent dynamics of AMPAR endocytosis and recycling that involves multiple regulatory elements through its distinct functional domains. We propose that, during AMPAR endocytosis, PACSIN1 is recruited to the clathrin-coated pit at the plasma membrane through a SH3-dependent interaction with a component of the endocytic machinery, such as dynamin (Fig. 6). The binding of dynamin to the SH3 domain forces PACSIN1 into an open conformation and releases the intramolecular inhibition of the PACSIN1 F-BAR domain, which is essential for its membrane deformation activity and facilitates vesicle endocytosis and fission. Given that PACSIN1 exists in dimers and oligomers, the other PACSIN1 SH3 domain can interact with the actin nucleator, N-WASP, to provide mechanical forces to propel vesicles away from the plasma membrane through active actin remodeling. Subsequently, PACSIN1 could bind to EHD1 and/or PICK1 to control the intracellular retention and correct recycling of AMPARs to the plasma membrane. The fact that the PACSIN1-PICK1 interaction can be tuned, potentially by phosphorylation of multiple serine residues in the variable region, adds another layer of regulation to ensure that AMPARs are recycled efficiently and in a timely manner to support synaptic transmission and plasticity.

Methods
DNA constructs. The PACSIN1 shRNA targeting sequences and the bicistronic pRK5-H1-shRNA#1-CMV-HA-PACSIN1 rescue construct were generated and characterized previously 24,28 . PACSIN1 point mutants were generated using the overlapping PCR protocol. Plasmids encoding PICK1 shRNA, myc-GluA2, pH-GluA2 and cytosolic pHluorin have been reported previously 18,28,32,42 . Neuronal cultures and transfection. All animal handling procedures were approved by the Animal Care and Use Committee of the Johns Hopkins University School of Medicine and the University of Queensland Animal Ethics Committee, and were conducted in accordance with relevant guidelines and regulations set by the National Institutes of Health in the USA and the Australian National Health and Medical Research Council. Hippocampal neurons from embryonic day 18 rat pups were plated onto poly-L-lysine-coated coverslips in Neurobasal growth medium supplemented with 2% B27, 2 mM Glutamax, 50 U/mL penicillin, 50 μ g/mL streptomycin and 5% fetal bovine serum. Neurons were switched to serum-free Neurobasal medium 24 h after seeding and fed twice a week. Homozygote PICK1 knockout mice 43 were obtained by mating heterozygote parents. High-density cortical neurons were prepared from PICK1 knockout pups and WT littermates at postnatal day 0 as previously described 18 . Neurons were maintained in glial-conditioned Neurobasal medium supplemented with 1% horse serum, 2% B-27 and 2 mM Glutamax, and fed twice a week. Neurons were transfected at days in vitro (DIV) 12-14 using lipofectamine 2000 (Invitrogen). pH-GluA2 recycling assay. pH-GluA2 recycling live-imaging assays were performed 48 h post-transfection as described previously 42 . Briefly, coverslips containing neurons were assembled in a closed perfusion chamber and continuously perfused with artificial cerebrospinal fluid (ACSF; 25 mM HEPES, 120 mM NaCl, 5 mM KCl, 2 mM CaCl 2 , 2 mM MgCl 2 , 30 mM D-glucose, 1 μ M tetrodotoxin, pH 7.4). After 10 min of baseline recording (F 0 ), neurons were perfused with NMDA solution (25 mM HEPES, 120 mM NaCl, 5 mM KCl, 2 mM CaCl 2 , 0.3 mM MgCl 2 , 30 mM D-glucose, 1 μ M tetrodotoxin, 20 μ M NMDA, 10 μ M glycine, pH 7.4) for 3 min before the perfusion medium was switched back to recording buffer for the remainder of the session. All imaging experiments were performed at room temperature using a Zeiss LSM 510 Meta/NLO system. The pHluorin fluorescence was imaged at 488 nm excitation and collected through a 505-550 nm filter, while the mCherry signal was imaged at 561 nm excitation and 575-615 nm emission. Neurons were imaged through a 63X oil objective (N.A. = 1.40) at a 3 μ m single optical section close to the plasma membrane and collected at a rate of 1 image per min. Images were analyzed using ImageJ software (NIH) by calculating the normalized change in average pHluorin over mCherry fluorescence intensities from a somatodendritic area defined manually to compensate for x-y movements of the recorded neurons. The fluorescence intensity change is expressed as Δ F/F o and the amplitude of fluorescence change (Δ F max /F o ) represents the extent of GluA2 endocytosis. The rate of GluA2 recycling can be calculated as the time taken from fluorescence minima to 50% of the fluorescence maxima (t 1/2 ). Antibody-feeding myc-GluA2 internalization assays. Cultured hippocampal neurons were transfected with PICK1 and PACSIN1 knockdown constructs, together with a myc-GluA2 reporter construct. To determine the amount of receptor internalization, surface myc-GluA2 was first labeled with mouse anti-myc antibody in live neurons, followed by 5 min incubation with 50 μ M NMDA + 1 μ M tetrodotoxin, prepared in normal ACSF containing 2 mM Mg 2+ , to induce receptor internalization. Cells were fixed with 4% paraformaldehyde/4% sucrose in PBS 5 min (t = 10′ ) or 15 min (t = 20′ ) after NMDA treatment. The remaining surface myc antibody was stained with a saturating concentration of Alexa-568 anti-mouse secondary antibody under non-permeabilizing conditions (surface), and the internalized myc antibody was labeled with Alexa-488 anti-mouse secondary antibody once neurons were permeabilized (internalized). Images were collected with a 63X oil-immersion objective  In this model, activation of NMDARs triggers the recruitment of PACSIN1 to the clathrin-coated pit at the plasma membrane through an SH3-dependent interaction with a component of the endocytic machinery, such as dynamin or N-WASP. This endocytic complex facilitates membrane deformation, invagination and fission to generate vesicles containing AMPARs, as well as actin nucleation to provide mechanical forces to propel vesicles away from the plasma membrane. Subsequent interaction of PACSIN1 with EHD1 and/or PICK1 controls the intracellular retention and the correct recycling of AMPARs to the plasma membrane. We propose that the activity-dependent phosphorylation of PACSIN1 in the variable region serves as a molecular switch that regulates the efficient recycling of AMPARs to support synaptic transmission and plasticity.