Modulation of synaptic function by VAC14, a protein that regulates the phosphoinositides PI(3,5)P₂ and PI(5)P.

Normal steady-state levels of the signalling lipids PI(3,5)P(2) and PI(5)P require the lipid kinase FAB1/PIKfyve and its regulators, VAC14 and FIG4. Mutations in the PIKfyve/VAC14/FIG4 pathway are associated with Charcot-Marie-Tooth syndrome and amyotrophic lateral sclerosis in humans, and profound neurodegeneration in mice. Hence, tight regulation of this pathway is critical for neural function. Here, we examine the localization and physiological role of VAC14 in neurons. We report that endogenous VAC14 localizes to endocytic organelles in fibroblasts and neurons. Unexpectedly, VAC14 exhibits a pronounced synaptic localization in hippocampal neurons, suggesting a role in regulating synaptic function. Indeed, the amplitude of miniature excitatory postsynaptic currents is enhanced in both Vac14(-/-) and Fig4(-/-) neurons. Re-introduction of VAC14 in postsynaptic Vac14(-/-) cells reverses this effect. These changes in synaptic strength in Vac14(-/-) neurons are associated with enhanced surface levels of the AMPA-type glutamate receptor subunit GluA2, an effect that is due to diminished regulated endocytosis of AMPA receptors. Thus, VAC14, PI(3,5)P(2) and/or PI(5)P play a role in controlling postsynaptic function via regulation of endocytic cycling of AMPA receptors.

Introduction

Phosphorylated phosphoinositide (PI) lipids reside on the cytoplasmic side of eukaryotic membranes and regulate a diverse array of cellular functions. The sn-3, 4 and 5 positions of the inositol head group have the potential to be phosphorylated in all seven combinations and each type of PI plays unique roles in mammals. Each PI recruits a distinct set of protein effectors and regulates multiple cellular events including membrane traffic (Corvera et al, 1999; Roth, 2004), protein sorting (Saksena et al, 2007), growth factor signalling (Cantley, 2002), ion homeostasis (Balla, 2006; Dong et al, 2010), cell survival (Brunet et al, 2001) and cell motility (Yin and Janmey, 2003).

PI lipids are tightly regulated and inter-converted by an array of lipid kinases and phosphatases. A conserved protein complex including PIKfyve/FAB1/PIP5K3 (GenBank accession # NP_035216), VAC14 (NP_666328) and FIG4/SAC3 (NP_598760) is responsible for the biosynthesis and turnover of PI(3,5)P2 (Jin et al, 2008; Ikonomov et al, 2009). PIKfyve is the PI(3)P 5-kinase that phosphorylates PI(3)P to form PI(3,5)P2 (Gary et al, 1998), whereas FIG4 dephosphorylates PI(3,5)P2 back to PI(3)P (Rudge et al, 2004; Duex et al, 2006a). The presence of both a kinase and a phosphatase in the same complex allows for tight regulation of PI(3,5)P2 levels. VAC14, a HEAT repeat protein, forms a scaffold for formation of the complex and also brings in other regulatory factors (Jin et al, 2008). Loss of PIKfyve or VAC14 causes a loss or decrease in PI(3,5)P2 levels, respectively (Duex et al, 2006b; Zhang et al, 2007; Ikonomov et al, 2011). While knockout of FIG4 might be predicted to increase PI(3,5)P2, FIG4 also activates PIKfyve and thus, PI(3,5)P2 is decreased in Fig4 fibroblasts (Chow et al, 2007) and yeast (Duex et al, 2006a). Interestingly, impairment of PIKfyve activity by either loss of VAC14 (Zhang et al, 2007) or pharmacological inhibition with YM201636 (Sbrissa et al, 2012) also causes a decrease in PI(5)P levels in mammalian cells. Note that the yeast S. cerevisiae does not produce PI(5)P. In mammals, the PI(5)P pool could come either from direct phosphorylation of phosphatidylinositol by PIKfyve (Sbrissa et al, 1999), or via dephosphorylation of PI(3,5)P2 by myotubularin family phosphatases (Tronchère et al, 2004); both pathways are dependent on PIKfyve activity.

The PIKfyve/VAC14/FIG4 complex is critical for endomembrane homeostasis and has been implicated in an ever-growing list of processes. In yeast, PI(3,5)P2 regulates vacuole fission, vacuole acidification (Bonangelino et al, 1997, 2002; Gary et al, 1998), retrograde traffic from the vacuole (Bryant et al, 1998; Dove et al, 2004) and is involved in the assembly of transcriptional regulators (Han and Emr, 2011). In metazoans, the PIKfyve/VAC14/FIG4 pathway regulates endosome-to-trans-Golgi network retrograde transport (Rutherford et al, 2006), autophagy (Rusten et al, 2007; de Lartigue et al, 2009; Ferguson et al, 2009), exocytosis (Osborne et al, 2008), calcium channel activation (Shen et al, 2009; Dong et al, 2010) and degradation (Tsuruta et al, 2009). In plants, the PIKfyve/VAC14/FIG4 pathway is involved in endocytosis, vacuole formation, auxin transporter recycling, and pollen development (Hirano and Sato, 2011; Hirano et al, 2011). It remains to be determined which of the mammalian pathways are regulated by PI(3,5)P2 and/or PI(5)P.

Analysis of several mouse mutants point to critical roles for the PIKfyve/VAC14/FIG4 pathway in the central and peripheral nervous systems. Vac14 gene trap (Vac14) mice have half of the normal levels of PI(3,5)P2 and PI(5)P. They develop normally, yet die perinatally with numerous neural defects including spongiform-like lesions and increased apoptosis in the brain, and intracellular vacuolation in peripheral neurons (Zhang et al, 2007). Fig4 (Chow et al, 2007; Ferguson et al, 2009; Lenk et al, 2011) and Vac14 (ingls) (Jin et al, 2008), mouse mutants defective in PI(3,5)P2 and PI(5)P regulation, have similar patterns of neurodegeneration to that observed in the Vac14 mouse.

Importantly, human patients with minor defects in the PIKfyve/VAC14/FIG4 pathway also display severe neurological problems. Mutations in FIG4 are responsible for human Charcot-Marie-Tooth disease type 4J (CMT4J), a recessive disorder affecting the peripheral nervous system (Chow et al, 2007; Nicholson et al, 2011). The most common CMT4J allele, I41T, destabilizes the mutant FIG4 protein and impairs its binding to VAC14 (Ikonomov et al, 2010; Lenk et al, 2011). Heterozygous FIG4 mutations have also been identified in patients with amyotrophic lateral sclerosis (ALS) and primary lateral sclerosis (PLS), two forms of motor neuron disease (Chow et al, 2009). These observations suggest that the PIKfyve/VAC14/FIG4 pathway has specialized functions in the nervous system.

Defects in multiple neural cell types likely contribute to the pathologies observed in VAC14/FIG4-deficient mouse models. Both neurons (Chow et al, 2007; Zhang et al, 2007, 2008; Katona et al, 2011) and astrocytes (Jin et al, 2008; Ferguson et al, 2009) are affected by mutations in VAC14 or FIG4. However, expression of FIG4 in neurons, but not astrocytes, rescues the spongiform-like lesions, gliosis and early lethality in Fig4 mice (Ferguson et al, 2012), emphasizing the importance of the PIKfyve/VAC14/FIG4 pathway in neuronal function.

Neurons are highly polarized cells that process electrochemical signals by extending long specialized processes—axons and dendrites—to facilitate information transfer through neural circuits. Accordingly, the endosomal system in neurons has both general and specialized pathways, such as long-range trafficking along neurites (Ibáñez, 2007), as well as specialized recycling in both the presynaptic and postsynaptic terminals (Kennedy and Ehlers, 2006; Dittman and Ryan, 2009). These membrane events are critical for multiple aspects of neuronal function such as neurite outgrowth, neurotrophic factor signalling and synaptic plasticity (Lasiecka and Winckler, 2011). However, the role of PIKfyve/VAC14/FIG4 in neuronal function remains unknown.

Here, we address the functional significance of the PIKfyve/VAC14/FIG4 pathway in cultured neurons from the hippocampus of wild-type and Vac14 mice. To gain insight into the cellular distribution of PIKfyve/VAC14/FIG4 pathway, we developed an antibody to VAC14, suitable for immunofluorescence microscopy, and found that endogenous VAC14 localizes to multiple organelles, consistent with multiple roles for PI(3,5)P2 in the endomembrane system. VAC14 partially colocalizes with early endosomes, late endosomes, lysosomes and autophagosomes. In neurons, VAC14 is found in both somatodendritic regions and axons. Notably, a substantial amount of endogenous VAC14 is present at synaptic sites, suggesting a role for VAC14 in the regulation of synaptic efficacy. Indeed, we find that synaptic function is altered in neurons cultured from Vac14 and Fig4 mice. Both postsynaptic function and surface expression of AMPA-type glutamate receptors are enhanced in Vac14−/− hippocampal neurons. Expression of VAC14 in Vac14 neurons reverses the synaptic phenotype, indicating a cell-autonomous and post-developmental role for VAC14 in regulating excitatory synaptic strength. We further show that the elevated surface AMPA receptor levels in Vac14 neurons are due to decreased endocytosis at postsynaptic sites. Together, our results identify control of PI(3,5)P2 and/or PI(5)P synthesis as a novel regulatory pathway at synapses that influences surface levels of glutamate receptors and synaptic function.

Results

Vac14−/− hippocampal neurons exhibit vacuolation, but otherwise develop normally in culture

Consistent with its importance in the nervous system, expression of VAC14 is abundant in the brain relative to other tissues (Supplementary Figure S1A and B). In this study, we sought insights into the neuronal-specific functions of the PIKfyve/VAC14/FIG4 complex. We focused on hippocampal neurons because VAC14 expression in the hippocampus is similar to other brain regions (Supplementary Figure S1C) and the hippocampus is largely spared from neurodegeneration, even at the time of death in Vac14 and Fig4 animals (Chow et al, 2007; Zhang et al, 2007). Hippocampal neurons from Vac14 embryos remain viable for several weeks, which enabled us to examine the impact of Vac14 deletion in these cells.

Although no spongiform lesions were observed in hippocampal regions in vivo, cultured Vac14 hippocampal neurons developed small vacuoles in the soma as early as 1 day in vitro (DIV) (Supplementary Figure S2A). Similarly to fibroblasts, the neuronal vacuoles are positive for the late endosome/lysosome marker LAMP1 and negative for the early endosome marker EEA1 (Supplementary Figure S2B). Notably, at the neuron density used (2 × 104 per 1.91 cm2), vacuoles are also observed in neurites by 12 DIV, suggesting that VAC14 functions in both the soma and neurites. Vacuole formation in Vac14 neurons appears to be activity independent; the degree of vacuolation in cells subjected to activity blockade (1 μM TTX, 40 μM CNQX, 20 μM APV) from DIV3 to DIV18 was similar to that in untreated neurons (Supplementary Figure S2C).

The presence of vacuoles in neurons lacking VAC14 did not appear to affect axon and dendrite development in culture, as assessed using the five stage model (Dotti et al, 1988; Supplementary Figure S3A). Vac14 neurons progressed from stage I (lamellipodia) to stage IV/V (complicated networks) at a rate similar to wild-type neurons (Supplementary Figure S3B), implying normal neurite outgrowth and differentiation.

Subcellular localization of VAC14 in cultured fibroblasts

To determine the sites of action of the PIKfyve/VAC14/FIG4 complex in neurons, we determined the localization of endogenous VAC14. Previous attempts to localize components of the PIKfyve complex in non-neuronal cells relied on overexpression of tagged proteins, and have produced divergent results. An earlier study indicated that overexpressed, tagged PIKfyve is confined to late endosome/lysosome compartments (Ikonomov et al, 2001), whereas other analyses found tagged FAB1/PIKfyve primarily localized to early endosomes (Cabezas et al, 2006; Rutherford et al, 2006).

To better understand the endogenous cellular distribution of the PIKfyve/VAC14/FIG4 complex, we raised a rabbit polyclonal antibody against full-length human VAC14 protein. After extensive affinity purification, we obtained a reagent that, in western blot analysis, revealed a major band at the expected molecular weight (88 kD) in wild-type but not in Vac14 brain (Supplementary Figure S1A). In wild-type fibroblasts, permeabilized with saponin prior to fixation, VAC14 was present on punctate organelles distributed throughout the cytoplasm; these structures were absent from Vac14 fibroblast controls (Figure 1A). Nuclear staining was frequently observed in both wild-type and Vac14 cells (Supplementary Figure S4A); thus, the antibody is not suitable to test whether VAC14 is also localized in the nucleus.

Figure 1

Endogenous VAC14 partially colocalizes with multiple endocytic organelles. (A) Polyclonal VAC14 antibody recognizes punctate structures in wild-type cells. Fibroblasts were permeabilized with saponin followed by fixation, then labelled with anti-VAC14 antibody. Bottom panels, DIC images. (B) In fibroblasts, endogenous VAC14 colocalizes with both EEA1 and LAMP1. Wild-type fibroblasts were triple labelled with rabbit anti-VAC14, chicken anti-EEA1 and rat anti-LAMP1. The majority of VAC14 colocalized with either EEA1 (yellow arrows) or LAMP1 (turquoise arrows). Some VAC14 colocalized with both (white arrows) or neither (green arrow) markers. (C) VAC14 partially colocalizes with the late endosome marker LBPA (arrow). Fibroblasts were double labelled with rabbit anti-VAC14 and mouse anti-LBPA. (D) VAC14 partially colocalized with lysosomes (arrow). To label lysosomes, prior to fixation, fibroblasts were pulsed with Texas Red-Dextran (Mw 70 kD) for 1 h and chased in the absence of dextran for 24 h. (E) The limiting membrane of vacuoles in Vac14 cells is positive for LAMP1 while negative for LBPA, suggesting a lysosomal origin. Vac14 fibroblasts were double labelled with rat anti-LAMP1 and mouse anti-LBPA. (D, E) DAPI (blue) used to label nuclei. (F) VAC14 partially colocalizes with LC3-RFP puncta (arrows). Fibroblasts transfected with LC3-RFP were fixed and labelled with anti-VAC14. (A–F) Bar=10 μm.

To determine the relative distribution of VAC14 on endosomal and lysosomal membranes, we performed triple labelling experiments in primary fibroblasts and determined the distribution of VAC14, EEA1 and LAMP1 puncta (Figure 1B; Supplementary Figure S4B and F). Consistent with earlier studies, EEA1 and LAMP1 labelled distinct compartments. The majority of VAC14 puncta colocalized with EEA1 (20±5%), LAMP1 (30±5%) or both markers (19±7%). These triple-labelled puncta likely represent intermediate endosomes. Thus, VAC14, PI(3,5)P2 and potentially PI(5)P, are present in multiple locations within the endomembrane system, including early endosomes, late endosomes and lysosomes (Supplementary Figure S5). Some VAC14 puncta (31±8%) did not colocalize with either EEA1 or LAMP1, suggesting that VAC14 may also function on other compartments.

LAMP1 is present on both late endosomes and lysosomes. To determine whether VAC14 is found on one or both of these compartments, we examined VAC14 localization relative to LBPA (late endosomes) or internalized dextran (lysosomes). Partial colocalization was observed between VAC14 (15±6%) and LBPA (Figure 1C; Supplementary Figure S4C and G), which indicates that some VAC14 resides on late endosomes. To determine whether lysosomes also contain VAC14, cells were incubated with a fluid phase marker, 70 kD Texas Red-dextran, and then chased in the absence of dextran for 24 h to allow it to reach lysosomes. Partial colocalization was observed between VAC14 (23±9%) and lysosomes loaded with dextran (Figure 1D; Supplementary Figure S4D and G), suggesting that some VAC14 is also localized on lysosomes. Interestingly, the limiting membrane of vacuoles in Vac14 fibroblasts is positive for LAMP1, but negative for LBPA (Figure 1E), implying that the large vacuoles derive solely from lysosomes.

In metazoans, the PIKfyve/VAC14/FIG4 pathway is thought to play a role in autophagy, either during fusion of autophagosomes with endosomes/lysosomes, or recycling of lysosomes from autolysosomes (Rusten et al, 2007; de Lartigue et al, 2009; Ferguson et al, 2009). LC3 is a common marker of autophagosomes. We transfected wild-type or Vac14 fibroblasts with LC3-RFP and colabelled transfected cells with anti-VAC14. VAC14 (17±13%) partially colocalized with LC3 (Figure 1F; Supplementary Figure S4E and G), suggesting that autophagosomes may contain PI(3,5)P2 and/or PI(5)P. Alternatively, these PI(3,5)P2 and/or PI(5)P containing regions may represent the interface between autophagosomes and endosomes/lysosomes.

Localization of VAC14 in neurons

To determine the localization of VAC14 in neurons, we first examined its distribution in the soma. In this case, neurons were not permeabilized with saponin prior to fixation; thus, the images indicate both membrane bound and cytosolic pools of VAC14. A significant portion of the VAC14 localized to punctate structures (Supplementary Figure S6). As in fibroblasts, VAC14 puncta colocalized both with the early endosome marker, EEA1 (Supplementary Figure S6A), and with the late endosome/lysosome marker, LAMP2 (Supplementary Figure S6B).

To test whether VAC14 is present in dendrites, hippocampal neurons were labelled with antibodies against VAC14 and against MAP2, a microtubule-associated protein that is highly expressed in dendrites but not in axons. Notably, discrete VAC14 puncta were found in MAP2-positive dendrites (Figure 2A; Supplementary Figure S7A). Moreover, another pool of VAC14 puncta was evident in MAP2-negative neurites, implying an axonal localization. To test this further, we labelled neurons with anti-VAC14 and anti-TAU-1, which preferentially labels axons in younger cultures (Horton et al, 2005). Again, VAC14 puncta were present in TAU-1-labelled axons (Figure 2A; Supplementary Figure S7A), although this axonal VAC14 pool was less prominent than the dendritic pool. In neurites, VAC14 puncta partially colocalized with both EEA1 (23±12%) and LAMP1 (29±9%) (Figure 2B; Supplementary Figure S7B and C), suggesting that VAC14 in neuronal processes functions in pathways that involve early and late endosomes as well as lysosomes.

Figure 2

VAC14 is found in both dendrites and axons, and colocalizes with endocytic and synaptic markers in hippocampal neurons. (A) Wild-type and Vac14 neurons were double labelled with rabbit anti-VAC14 and mouse anti-MAP2 (dendrites) or mouse anti-TAU-1 (axons). Arrows indicate the localization of VAC14 on dendrites (MAP2-positive and TAU-negative neurites). Arrowheads indicate the localization of VAC14 on axons (MAP2-negative and TAU-1-positive neurites). Bar=10 μm. (B) VAC14 partially colocalizes with EEA1 or LAMP1 in the neurites (arrows). Wild-type and Vac14 neurons were triple labelled with rabbit anti-VAC14, chicken anti-EEA1 and rat anti-LAMP1. Bar=5 μm. (C) VAC14 displays significant colocalization with several synaptic markers: synapsin, synaptotagmin and synaptobrevin. Wild-type and Vac14 neurons were double labelled with rabbit anti-VAC14 and guinea pig anti-synapsin, mouse anti-synaptotagmin or mouse anti-synaptobrevin. Bar=5 μm. (D) VAC14 partially localizes at excitatory synapses (labelled with both the synaptic vesicle glutamate transporter vGlut1 and postsynaptic marker PSD95). Wild-type and Vac14 neurons were labelled with rabbit anti-VAC14, mouse anti-PSD95 and guinea pig anti-vGlut1. Arrows indicate examples of colocalization. (C, D) Lower panels show straightened dendrites from corresponding top panels. Bar=5 μm.

Endogenous VAC14 localizes to synapses

Interestingly, the most striking colocalization was observed between VAC14 and synaptic markers. A substantial number of VAC14 puncta colocalized with the presynaptic terminal markers synapsin, synaptotagmin and VAMP/synaptobrevin (Figure 2C; Supplementary Figure S7D). To test whether VAC14 colocalizes with excitatory synapses, we performed triple labelling against VAC14, vGlut1 (the glutamate transporter on presynaptic vesicles), and the postsynaptic scaffolding protein PSD95. VAC14 puncta colocalized extensively with vGlut1/PSD95 double-positive puncta (Figure 2D; Supplementary Figure S7E), suggesting a role for VAC14 in excitatory synapse function.

Altered synaptic function in cultured Vac14−/− neurons

To examine a functional role for the PIKfyve/VAC14/FIG4 pathway at the synapse, we measured miniature excitatory postsynaptic currents (mEPSCs) in pyramidal-like neurons from Vac14 hippocampal cultures and corresponding wild-type controls. mEPSCs represent unitary synaptic currents mediated by the spontaneous fusion of single synaptic vesicles, and are often used to reveal functional changes in synaptic strength. Pyramidal-like neurons with little to no vacuolation were targeted for electrophysiology. Given the neurodegeneration observed in other regions of the brain at the time of birth, one might expect synaptic function to be diminished in Vac14 neurons. Surprisingly, mEPSCs from Vac14 neurons displayed a significant increase (24±6%) in amplitude relative to wild-type mEPSCs (Figure 3A and B), suggesting an inhibitory role for VAC14 in synaptic function. We found no difference in mEPSC frequency or decay time in Vac14 mEPSCs (Figure 3C–E). In a parallel experiment, we found mEPSC amplitude was similarly increased in Fig4 mice (Figure 3F and G), which also have reduced PIKfyve kinase activity. Together, these data suggest that the increase in mEPSC amplitude in both Vac14 and Fig4 neurons results from defects in PI(3,5)P2 and/or PI(5)P synthesis.

Figure 3

Loss of VAC14 or FIG4 leads to an increase in excitatory synaptic function. (A) Representative mEPSC recordings of wild-type (N=32) and Vac14 neurons (N=32). (B) Mean mEPSC amplitude in Vac14 neurons is larger than in wild-type neurons, 20.86±1.03 pA versus 16.83±0.91 pA, respectively. *P=0.0045, t-test. (C) Mean mEPSC frequency is similar in wild-type (1.21±0.26 Hz) and Vac14 (1.16±0.24 Hz) neurons. (D, E) Summary of mEPSC kinetics in wild-type and Vac14 neurons. (D) Individual mEPSCs overlaid. Thick lines show the mean trace. Dashed line is aligned to the mean peak inward current of Vac14 mEPSC. Scaled overlay shows similar kinetics between wild-type and Vac14 mEPSCs. (E) Mean mEPSC decay is similar between wild type (3.80±0.15 ms) and Vac14 (3.69±0.14 ms). (F) Representative mEPSC traces of wild-type (Fig4+/+) (N=12) and Fig4 neurons (N=14). (G) Mean mEPSC amplitude in Fig4 neurons is larger than in Fig4+/+ neurons (14.64±0.39 pA versus 18.73±1.38 pA, respectively, *P=0.0164, t-test). (H) Mean mEPSC frequency is similar in Fig4+/+ and Fig4 (0.99±0.22 Hz versus 0.92±0.28 Hz, respectively, P=0.8542, t-test). Error bars are standard error of the mean (s.e.m.).

Although we found no change in mEPSC frequency in either Vac14 (Figure 3C) or Fig4 neurons (Figure 3H), VAC14 is localized to axons (Figure 2A), and therefore, is well positioned to contribute to presynaptic function. The enlargement of endocytic compartments in Vac14 cells (Zhang et al, 2007) also suggested that the increase in mEPSC amplitude in Vac14 neurons could have resulted from increased glutamate release by enlarged presynaptic vesicles (increased quantal content). To examine this possibility, we performed transmission electron microscopy on thin sections from the hippocampus and hindbrain of wild-type and Vac14 mice at P0. The hindbrain was included in these studies because it is the most vacuolated brain region in the Vac14 animal at the time of death. We found similar synaptic vesicle diameter in wild-type and Vac14 presynaptic terminals of both brain regions (Figure 4A and B).

Figure 4

Presynaptic probability of release is enhanced in Vac14 neurons. (A, B) Synaptic vesicles from Vac14 are not larger than synaptic vesicles observed in brains from wild type. (A) Electron microscopy of excitatory synapses, evident by the thickening of the postsynaptic membrane, in wild-type and Vac14 hippocampus and hindbrain. Bar=100 nm. (B) Quantitation of the diameter of synaptic vesicles. Cumulative probability distribution of synaptic vesicle diameter. No significance difference was found between wild type and Vac14 by a two-sample Kolmogorov-Smirnov test (kstest2, Matlab) (hippocampus, P=0.32; hindbrain, P=0.46). Three wild-type and three Vac14 animals were analysed. Hindbrain: N=567 vesicles from 33 terminals for wild type and 388 vesicles from 29 terminals for Vac14. Hippocampus: N=433 vesicles from 33 terminals for wild type and 66 vesicles from 15 terminals for Vac14. (C, D) The number of synapses is decreased in Vac14 neurons. (C) Wild-type and Vac14 hippocampal neurons were triple labelled with rabbit anti-MAP2 (blue), mouse anti-PSD95 (red) and guinea pig anti-vGlut (green). Examples of straightened dendrites are shown. Bar=5 μm. (D) Quantitation of the number of synapses on the first 100 μm of dendrites starting from the soma. The numbers of synapses were normalized to the average of wild type. Vac14 neurons had fewer synapses (N=91 for wild type and 71 for Vac14) (*P=1.7 × 10−4, t-test). Error bars, s.e.m. (E, F) Presynaptic probability of release is increased in Vac14 neurons. NMDA currents were pharmacologically isolated from AMPA and GABAA mediated currents and recorded at −70 mV (solution contained zero Mg2+). An extracellular stimulating electrode was placed locally and used to stimulate vesicle release in afferent axons. Once a stable response was obtained, 20 μM MK801, an open-channel blocker of NMDA receptors, was added to the bath for 5 min without stimulation. Then, 200 stimulations were delivered and amplitude of the current measured (mean amplitude is shown by filled squares). In the presence of MK-801, the current is progressively blocked. The data for each genotype were fitted with a double exponential curve. The rate of progressive blockade of NMDA current was significantly greater in Vac14 neurons; mean amplitude of the 2–11 stimulations is significantly lower in Vac14 neurons. *P=0.0217, Anova1 (Matlab).

Despite similar mEPSC frequency between Vac14 and wild-type neurons, we observed that the number of excitatory synapses in the first 100 μm of Vac14 dendrites was modestly but significantly decreased (Figure 4C and D). This discrepancy raised the possibility that although Vac14 neurons have fewer presynaptic inputs, the terminals may have elevated neurotransmitter release probability. To further examine presynaptic function, we measured the probability of synaptic vesicle release by recording postsynaptic NMDA receptor mediated currents in the presence of the use-dependent NMDA receptor antagonist, MK801 (Huettner and Bean, 1988). The degree of blockade is proportional to the number of presynaptic vesicles that release glutamate in response to stimulation (Rosenmund et al, 1993). We found that Vac14 neurons showed greater blockade than wild-type neurons (Figure 4E and F) which suggests an enhancement of release probability in Vac14 neurons. Given that the frequency of mEPSCs was similar, the expected change in mEPSC frequency was likely masked by the decrease in synapse number. Together, these results suggest that the PIKfyve/VAC14/FIG4 pathway modulates neurotransmitter release at the presynaptic terminal.

VAC14 levels are higher in dendrites than axons. To test whether increased mEPSC amplitude in Vac14 neurons is due to loss of VAC14 in the postsynaptic neuron, we transfected neurons with plasmids encoding Citrine-tagged human VAC14. For these experiments, we used calcium phosphate-based transfection because the low transfection efficiency (∼1% of cells) ensures that the few neurons that express VAC14 in Vac14 cultures received the excitatory synaptic contacts from neurons that lack VAC14. Thus, mEPSCs recorded from transfected neurons measure the effect of restoring VAC14 to the postsynaptic cell. We found VAC14 expression reversed the increase in mEPSC amplitude observed in Vac14 relative to wild-type neurons, whereas expression of Citrine alone did not (Figure 5). Moreover, even in wild-type neurons, overexpression of Citrine-VAC14 significantly depressed mEPSC amplitude relative to expression of untransfected neighbours, suggesting that synaptic strength is bidirectionally regulated by the levels of VAC14 in postsynaptic neurons. Together, these results strongly implicate VAC14 in the regulation of postsynaptic function.

Figure 5

Restoration of VAC14 eliminates the increase in mEPSC amplitude in Vac14 neurons. (A) Examples of mEPSC recordings of wild-type and Vac14 neurons that were sham-transfected, expressed VAC14-Citrine or Citrine alone. (B) Quantification of mEPSC amplitude normalized to sham-transfected wild-type neurons shows that there is an inverse relationship between level of VAC14 expression and mEPSC amplitude. In wild-type neurons, VAC14 overexpression leads to a decrease in amplitude. The increase in mEPSC amplitude in Vac14 neurons is reduced by re-introduction of VAC14. Neurons transfected with Citrine were similar to untransfected wild type. One-way ANOVA test was used to compare mEPSC amplitudes (F(5, 12.3), P=8.27 × 10−9). *Individual comparisons between untransfected and VAC14-transfected wild-type and untransfected and VAC14-transfected Vac14 neurons were significant using Tukey-Kramer honestly significant difference (HSD) criterion. WT, wild type; KO, Vac14. Error bars, s.e.m.

Surface AMPA receptors are elevated in Vac14−/− hippocampal neurons

mEPSCs are dominated by currents through AMPARs localized on the postsynaptic membrane (Gong and De Camilli, 2008). Thus, the increase in mEPSC amplitude in Vac14 neurons could result from changes in the number of surface AMPA receptors. Under basal conditions, most AMPA receptors in the hippocampus are heterotetramers of the GluA2 and GluA1 subunits (Lu et al, 2009). By western blot, we found similar levels of total GluA2 between wild-type and Vac14 neurons (Supplementary Figure S8). To test if the level of GluA2 at the cell surface was different, intact cultured neurons were incubated with an antibody against an extracellular epitope of GluA2 (Mouse IgG2a, MAB397; Chemicon), followed by fixation and incubation with a fluorescent secondary antibody under non-permeabilizing conditions. Surface GluA2 puncta were quantified using immunofluorescence microscopy (Figure 6A). In both wild-type and Vac14 neurons, there was a wide range in intensities of surface GluA2 puncta. However, in Vac14 neurons, the average and median surface GluA2 intensities were 34 and 17% higher, respectively, relative to wild-type neurons. The medians differed significantly with 95% confidence, indicated by the non-overlapping notches surrounding the medians in the box plot. Moreover, the cumulative distribution of surface GluA2 puncta intensities was right-shifted in Vac14 neurons (Figure 6B). These data indicate that surface GluA2 levels are increased in Vac14 neurons, which likely accounts for the increased amplitude of mEPSCs. In an independent approach, we measured the ratio of surface to total GluA2 in dendrites. Surface GluA2 was labelled as described above. Then neurons were permeabilized and labelled with a C-terminal GluA2 antibody (Rabbit pAb, AB1768; Chemicon) to identify total GluA2. We found that the ratio of surface to total GluA2 was increased in Vac14 dendrites, relative to wild type (Figure 6C and D). Together, these results suggest that GluA2 receptors are expressed in Vac14 neurons at similar levels, but accumulate on the surface.

Figure 6

Surface GluA2 levels increase in Vac14 neurons. (A) Surface GluA2 was labelled by incubation of intact neurons with mouse anti-GluA2 antibody. Arrows highlight the dendrite used for analysis (enlarged in lower panels). Intensity presented in the ‘fire' LUT colour scheme. (B) Quantitation of the intensity of GluA2 puncta, normalized to the wild-type mean. The relative intensity values from wild-type and Vac14 neurons are presented as a cumulative distribution. A Kolmogorov-Smirnov test demonstrates that the data sets differ significantly; P=1.1 × 10−34. The median of each data set also differ significantly (box plot shown in insert). Box, interquartile range; line, median; square, mean; non-overlapping notches indicate that the two medians are statistically different at the 5% significance level; whiskers, minimum and maximum of the data within 1.5 times the length of the box. N=5272 for wild type and 4756 for Vac14. Error bars, s.e.m. (C) Surface GluA2 subunits accumulate on the surface of Vac14 dendrites. Top three panels (wild type) and middle three panels (Vac14) dendrites (left to right) show surface GluA2, total GluA2, and the dendritic marker, MAP2. Yellow arrows highlight dendrites analysed. Bottom panels show the merged image of the straightened dendrite (left) and MAP2 (right). (D) The ratio of surface to total GluA2 is increased in Vac14 dendrites (1.32±0.052) relative to wild type (1.0±0.0364). *P=6.836 × 10−7, two-sample t-test. Scale bar=10 μm. Error bars=s.e.m.

Trafficking of AMPA receptors is altered in Vac14−/− neurons

GluA2 undergoes constant cycling between the surface and internal pools; once internalized, GluA2 may either be degraded in lysosomes or recycled back to the plasma membrane. Perturbations in endocytosis, recycling or degradation could lead to an accumulation of surface receptors in Vac14 neurons. To measure the rate of endocytosis of AMPA receptors, we performed live labelling of surface GluA2 with anti-GluA2 antibody and then stimulated endocytosis via addition of NMDA (Figure 7A). After 10 min, surface bound GluA2 antibodies were stripped by a brief wash in low pH solution, such that only internalized GluA2 antibodies were detected after fixation and permeablization. Leupeptin was present throughout to prevent lysosomal degradation. Notably, the number of internalized GluA2 puncta was decreased by 30% in dendrites in Vac14 neurons (Figure 7B and E). Total levels of internalized GluA2 puncta in the soma and dendrites were reduced to 71 and 56%, respectively, compared to wild type (Figure 7C and D). These results are consistent with an endocytosis defect in Vac14 neurons. To determine whether internalized GluA2 still enters the degradation pathway in Vac14 neurons, we further measured the proportion of internalized GluA2 puncta that colocalized with the late endosomal and lysosomal marker LAMP1 (Supplementary Figure S9). Though fewer internalized GluA2 puncta were observed in Vac14 neurons, a similar proportion exhibited colocalization with LAMP1 (27% in wild type versus 26% in Vac14). This suggests that the transport of AMPA receptors late in the endocytic pathway is normal in Vac14 neurons. In addition, we tested whether the reduction in internalized puncta was also due to enhanced recycling back to the surface. We transfected neurons at DIV12 with plasmids encoding super-ecliptic pHluorin-tagged GluA1 (Kopec et al, 2006), which are strongly fluorescent when exposed to the neutral pH of the extracellular space. At DIV14, we measured the rate of internalization and recycling of pHluorin-GluA1 following NMDA stimulation (Figure 8). Five minutes of NMDA stimulation markedly reduced the intensity of pHluorin-GluA1. Following wash-out, fluorescence recovered to baseline levels as internalized receptors recycled back to the plasma membrane (Figure 8A–D). The magnitude of NMDA-dependent internalization of GluA1 was diminished in Vac14 neurons (Figure 8E), again suggesting reduced receptor endocytosis in these neurons. To measure the rate of recycling, we calculated the time point after NMDA stimulation at which fluorescence intensity recovered to 50% of the pre-NMDA baseline. Whereas GluA1 internalization was reduced in Vac14 neurons, we found no difference in the rate of recycling relative to wild type (Figure 8F). These results suggest that the initial steps in AMPA receptor endocytosis, rather than post-endocytic sorting, represent the most prominent trafficking defect accompanying loss of VAC14. Together, our findings suggest that the PIKfyve/VAC14/FIG4 pathway regulates excitatory synapse function largely via modulation of AMPA receptor endocytosis (Figure 9).

Figure 7

GluA2 endocytosis is reduced in Vac14 hippocampal neurons. (A) Diagram of experiment. Wild-type or Vac14 hippocampal neurons were treated with the lysosomal protease inhibitor leupeptin, then live labelled with mouse GluA2 antibodies to label those receptors exposed to the cell surface. Endocytosis was stimulated with 50 μM NMDA for 10 min. Then any remaining surface-exposed GluA2 antibody was removed with an acid wash. Neurons were then fixed and labelled with Alexa 555 anti-mouse IgG. (B) Example of internalized GluA2. Intensity presented in the ‘fire' LUT colour scheme. Bar=10 μm. (C) Total internalized GluA2 in the soma was decreased in Vac14 neurons (N=55 for wild type and 42 for Vac14, *P=2.73 × 10−5, t-test). (D) Total internalized GluA2 in the dendrites was decreased in Vac14 neurons (N=57 for wild type and 61 for Vac14, *P=8.33 × 10−5, t-test). (E) The number of internalized GluA2 puncta were decreased in Vac14 neurons (N=57 for wild type and 61 for Vac14, *P=5.91 × 10−8, t-test). A fixed length (35 μm from the soma) was used for dendrites in (D) and (E). Error bars, s.e.m.

Figure 8

Endocytosis of AMPA receptors is reduced in Vac14 neurons. Cultured hippocampal neurons were transfected with pH-GluA1and mCherry at DIV12. At DIV14, cells were placed on the microscope stage and perfused with normal extracellular buffer. Images were acquired once per minute for a 10-min baseline, then switched to NMDA stimulation buffer for 5 min to stimulate internalization. Then, following replacement of the NMDA stimulation buffer with normal buffer, monitored for recycling back to the cell surface. The pHluorin fluorescence was imaged at 488 nm excitation, while mCherry fluorescence was imaged at 559 nm excitation, through a × 60 oil objective at a rate of one image per minute. (A) Representative full-frame images of wild-type and Vac14 neurons during baseline (0–10 min), NMDA stimulation (11–15 min) and recovery after wash out (16–55 min). (B) Changes in pH-GluA1 fluorescence were calculated from straightened dendrites isolated from full-frame images. (C) Wild-type and Vac14 neurons exhibited similar recovery to baseline levels. (D) Average time course for percent change in pH-GluA1 fluorescence, normalized to average baseline intensity. (E) Amplitude of change in fluorescence after wash-out of NMDA is decreased in Vac14 neurons compared to wild type (wild type, 66.72±4.93%; Vac14 44.68±7.10%). (F) t1/2 recycling rate after NMDA removal is normal. *P=0.0224, Two-sample t-test, n=11–14. Error bars, s.e.m.

Figure 9

Model of trafficking defects that promote the elevation of surface AMPA receptors in Vac14 neurons. Endocytosis of surface GluA2 is diminished in Vac14 neurons compared with wild type (1), while no defects were detected in recycling of internal GluA2 to the cell surface (2) or membrane transport late in the endocytic pathway (3). Note that in the absence of VAC14, FIG4 is destabilized (Lenk et al, 2011). EE, early endosomes; LE, late endosomes; RE, recycling endosomes.

Discussion

VAC14 modulates synaptic activity in hippocampal neurons

VAC14 is present in neuronal dendrites and axons and exhibits extensive colocalization with synaptic markers. Thus, the PIKfyve/VAC14/FIG4 pathway likely impacts the synapse at multiple levels, including modulation of both presynaptic and postsynaptic function. Although here we focused on postsynaptic VAC14, our results suggest that there are also effects on presynaptic function. Thus, while mEPSC frequency is unaltered in Vac14 neurons, MK801 use-dependent block of NMDA receptor currents is accelerated in these cells, suggesting elevated neurotransmitter release probability. Consistent with a presynaptic role for PI(3,5)P2 and/or PI(5)P, an earlier report identified C. elegans FIG4 at presynaptic sites in a large scale RNAi screen (Sieburth et al, 2005). An increase in probability of presynaptic vesicle fusion with the plasma membrane fits with a previous study demonstrating enhanced granule exocytosis following knockdown of PIKfyve in cultured chromaffin and PC12 cells (Osborne et al, 2008). This pathway may be interacting directly with exocytic machinery. Alternatively, it is possible that voltage-gated calcium channels or other membrane proteins that are important for membrane excitability are more highly expressed on the surface of Vac14 neurons, which causes increased calcium influx in response to depolarization.

The presence of VAC14, and its lipid products, in the postsynaptic terminal is consistent with earlier findings that VAC14 interacts with nNOS (Lemaire and McPherson, 2006), which interacts with the postsynaptic scaffolding protein PSD95 (Tochio et al, 2000). In addition, MTMR2, a phosphatase that acts on PI(3,5)P2 in vitro and likely in vivo, also interacts with PSD95 (Lee et al, 2010). That VAC14 levels in the somatodendritic region are significantly higher than in axons, strongly suggested that the PIKfyve/VAC14/FIG4 pathway controls critical aspects of postsynaptic function. Indeed, we found that genetic deletion of VAC14 is accompanied by enhanced mEPSC amplitude. Importantly, postsynaptic expression of VAC14 rescues this defect, strongly suggesting that postsynaptic loss of VAC14 is responsible for the enhanced mEPSC amplitude. Moreover, Fig4 neurons exhibited a similar increase in mEPSC amplitude, strengthening the argument that PIKfyve/VAC14/FIG4 pathway, and their lipid products PI(3,5)P2 and/or PI(5)P, play a role in postsynaptic function.

AMPA receptor trafficking in Vac14−/− neurons

Our results further indicate that VAC14 regulates postsynaptic function through regulation of AMPA receptor trafficking. Surface levels of AMPA receptors are tightly controlled by trafficking to and from the cell interior. Consistent with increased mEPSC amplitude, steady-state levels of surface-exposed AMPA receptors are elevated in Vac14 neurons. Using two independent methods, we found that AMPA receptor internalization following NMDA receptor activation is defective in Vac14 neurons. The total levels of regulated endocytosed AMPA receptors, quantified from internalized GluA2 puncta or amplitude of change in pHluorin-GluA1 fluorescence, are reduced in Vac14 neurons, which likely accounts for the elevated surface AMPA receptor levels. This endocytosis defect is consistent with the finding that internalization of transfected GluA2 is reduced in cortical neurons after siRNA knockdown of PIKfyve (Tsuruta et al, 2009). Conversely, MTMR2 knockdown, which is predicted to increase PI(3,5)P2 levels, enhances AMPA receptor endocytosis (Lee et al, 2010). Together, these findings suggest a new role for PI(3,5)P2 and/or PI(5)P, and downstream effector(s), early in the endocytic pathway. Though general endocytosis could be regulated by these signalling lipids, Tsuruta et al (2009) found that the internalization of Cav1.2, but not Kv1.2, was affected by PIKfyve knockdown; thus, the PIKfyve/VAC14/FIG4 pathway may play a more selective role in endocytosis of particular membrane proteins.

While VAC14 is localized throughout the endocytic pathway, we found that the recycling and degradative trafficking of AMPA receptors are normal in Vac14 neurons. Similarly, EGF receptors traffic normally to lysosomes in Vac14 fibroblasts (Zhang et al, 2007), and EGF receptor trafficking is also unaffected by siRNA knockdown of PIKfyve (Rutherford et al, 2006), or overexpression of a dominant-negative mutant PIKfyve (Ikonomov et al, 2003), although EGF receptor degradation is slowed when a PIKfyve inhibitor is used (de Lartigue et al, 2009). Importantly, since Vac14 cells have half of the normal levels of PI(3,5)P2 and PI(5)P, it remains possible that different trafficking steps have distinct sensitivities to the extent of loss of PI(3,5)P2 and/or PI(5)P. Future work is needed to determine if AMPA receptor trafficking late in the endocytic pathway is affected if the PIKfyve/VAC14/FIG4 pathway is inhibited further.

Potential roles of VAC14 in learning and memory

AMPA receptor expression at synapses is regulated to modify synaptic efficacy in the context of long-term potentiation (LTP) and long-term depression (LTD) (Song and Huganir, 2002), as well as homeostatic control of synaptic function driven by persistent changes in neuronal activity (Turrigiano, 2008). Rab5, which likely acts in pathways that are also regulated by PIKfyve (Jefferies et al, 2008), is similarly involved in postsynaptic glutamate receptor trafficking and is required for LTD (Brown et al, 2005), raising the possibility that dynamic control of PI(3,5)P2 and/or PI(5)P synthesis contributes to these forms of synaptic plasticity as well. It is now of interest to assess whether, in addition to the more severe neurophysiological outcomes that accompany PIKfyve/VAC14/FIG4 deficiency, perturbations of the PI(3,5)P2 and/or PI(5)P-related signalling pathways underlie defects in learning and memory. The Fig4-I41T Tg705 transgenic line that survives to 3 months of age with a reduced extent of spongiform degeneration could be useful for this purpose (Lenk et al, 2011).

Excitotoxicity in Vac14−/− neurons

Excitotoxicity has been implicated in many acute and chronic neurological diseases such as stroke (Rothman and Olney, 1986) and ALS (Beal, 1992; Rothstein et al, 1992; Martin, 2010). The increased synaptic efficacy in cultured Vac14 neurons also raises the question of whether excitotoxicity contributes to the neurodegeneration phenotypes observed in VAC14/FIG4-deficient mouse models. Consistent with this idea, PIKfyve overexpression has been shown to protect cultured neurons against excitotoxicity (Tsuruta et al, 2009). In this case, it was postulated that an increase in PIKfyve activity led to downregulation of voltage-gated calcium channels and of GluA2, and possibly other as yet undetermined channels and transporters.

Endogenous VAC14 localizes to multiple compartments in the endomembrane system

A common view in recently published reviews is that PI(3,5)P2 is confined to late endosomes, and has little overlap with its precursor, PI(3)P, which is thought to be confined to early endosomes. However, in fibroblasts, using an antibody to endogenous VAC14, we found that VAC14 was equally distributed between early and late endosomes as well as lysosomes, with some localization to autophagosomes. Assuming that the location of VAC14 on membranes reflects the distribution of PI(3,5)P2 and/or PI(5)P, we predict that these lipids may regulate pathways that emanate from each of these organelles. VAC14 also localizes to punctate spots that contain neither EEA1 nor LAMP1, which suggests that there are as yet undetermined organelles that contain PI(3,5)P2 and PI(5)P. We also found that AMPA receptor internalization and evoked presynaptic vesicle release are altered in the absence of VAC14, which suggests novel roles for PIKfyve/VAC14/FIG4, or their downstream effectors, near the plasma membrane. At present, it remains to be determined whether these effects are specific to neurons or whether the PIKfyve/VAC14/FIG4 pathway regulates selected events near the plasma membrane in other cell types.

In summary, our paper describes a critical new role for VAC14 and, by implication, PI(3,5)P2 and/or PI(5)P, in regulating synaptic function in neurons. Future work will elucidate specific molecular pathways controlled by PI(3,5)P2 and/or PI(5)P and may provide insights into the treatment of human neuropathies that can be mitigated via regulation of these lipids. Thus, development of drugs designed to modulate the levels of these lipids might lead to new therapies for several types of neurological disorders.

Materials and methods

Ethics statement

All animal use was performed in compliance with guidelines of the University Committee on Use and Care of Animals of the University of Michigan and National Institutes of Health.

Electrophysiology

Whole-cell patch clamp recordings were performed with an Axopatch 200B amplifier from 13 to 15 DIV cultured hippocampal pyramidal-like neurons bathed in an extracellular solution containing 119 mM NaCl, 5 mM KCl, 2 mM CaCl2, 2 mM MgCl2, 30 mM glucose, 10 mM HEPES (pH 7.4) plus 1 μM TTX and 10 μM bicuculline to isolate glutamatergic mEPSCs. Internal pipette solution contained 100 mM caesium gluconate, 0.2 mM EGTA, 5 mM MgCl2, 40 mM HEPES, 2 mM Mg-ATP, 0.3 mM Li-GTP, 1 mM QX-314 (pH 7.2). Pipette resistance ranged from 3 to 5 MΩ. Neurons with a pyramidal-like morphology were targeted for analysis. For Vac14 neurons, pyramidal-like neurons with few to no vacuoles were targeted for analysis. Neurons were voltage clamped at −70 mV, and series resistance was not compensated. mEPSC amplitude and frequency were analysed offline using Minianalysis (Synaptosoft). Average traces and statistical analysis was performed in Matlab (Mathworks). Statistical differences between control and experimental conditions were determined by ANOVA and Tukey-Kramer honestly significant difference (HSD) test.

In order to measure probability of evoked synaptic vesicle release in presynaptic terminals forming synapses onto wild-type or Vac14 neurons, 24 h after transfection, warm (37°C), 0 Mg2+ solution was applied to the bath: 125 mM NaCl, 2.5 mM KCl, 5 mM HEPES, 2 mM CaCl2, 33 mM D-Glucose, pH 7.4). AMPA receptor antagonist, CNQX (20 μM), and GABAA receptor antagonist, bicuculline (10 μM), were included to isolate NMDA receptor currents. Neurons were voltage clamped at −70 mV and evoked excitatory postsynaptic currents were measured by positioning a stimulating electrode close to the neuron. Once a stable response was obtained, MK801 (20 μM), a use-dependent antagonist for NMDA receptors, was applied to the bath for 5 min without stimulation. Following wash-in, 200 stimulations at 0.33 Hz were delivered in the presence of MK801. The peak of the NMDA receptor current was measured for each and normalized to the first response following MK801 wash-in.

Endocytosis assay

Neurons were treated with 20 μM leupeptin for 30 min before live labelling with mouse GluA2 antibodies diluted in normal medium with leupeptin for 15 min. After washing with neurobasal medium, neurons were incubated in normal medium supplemented with 50 μM NMDA and 20 μM leupeptin for 10 min. Endocytosis was stopped by washing in cold 1 × PBS with 0.1 mM CaCl2 and 1 mM MgCl2. Surface bound GluA2 antibodies were stripped with 0.5 M NaCl/0.2 M acetic acid for 4 min on ice. Neurons were fixed in 2% paraformaldehyde and 2% sucrose for 15 min and permeabilized with 0.1% Triton X-100 for 5 min. After blocking with 2% BSA, neurons were incubated with Alexa 555 anti-mouse secondary antibody for 1 h.

To examine the degradation pathway, neurons were incubated with rat anti-Lamp1 antibodies and then Alexa 488 anti-rat and Alexa 555 anti-mouse secondary antibodies.

Endocytosis and recycling assay

Neurons were transfected at DIV12 by Calcium Phosphate and pHluorin-GluA1 endocytosis and recycling live-imaging assays were performed 36–48 h post transfection. Mattek dishes were placed on the stage of the confocal microscope and perfused with normal extracellular buffer (25 mM HEPES, 120 mM NaCl, 5 mM KCl, 2 mM CaCl2, 2 mM MgCl2, 30 mM D-glucose, 1 mM TTX, pH 7.4). Images were acquired once per minute for (I) a 10 min baseline, (II) after the bath was switched to NMDA stimulation buffer (25 mM HEPES, 120 mM NaCl2, 5 mM KCl, 2 mM CaCl2, 0.2 mM MgCl2, 30 mM D-glucose, 1 mM TTX, 20 mM NMDA, 10 mM glycine, pH 7.4) for 5 min to stimulate internalization, and (III), then following washout of the NMDA stimulation buffer to monitor recycling back to the cell surface. The pHluorin fluorescence was imaged at 488 nm excitation, while mCherry fluorescence was imaged at 559 nm excitation, through a × 60 oil objective at a rate of one image per minute. Images were analysed using ImageJ software (NIH) by straightening the primary dendrite of the neuron, starting from the soma, and calculating the fluorescence intensity relative to the average intensity of the baseline period. The degree of GluA1 endocytosis was determined by analysing the first 1–3 min after NMDA stimulation (max decrease in signal) and the rate of GluA1 recycling was determined by fitting a linear curve to the time after max internalization and calculating the time point at which 50% of fluorescence recovered. pHluorin-GluA1 was a gift of Robert Malinow (Addgene plasmid #24000).