PIKfyve regulates Ca_V1.2 degradation and prevents excitotoxic cell death

There is increasing evidence that Ca²⁺ overload and excitotoxicity are central players in a wide variety of neurological diseases, including stroke and Parkinson's and Alzheimer's diseases (Orrenius et al., 2003; Dong et al., 2006). Glutamate receptors and voltage-gated Ca²⁺ channels (VGCCs) are the main sources of Ca²⁺ influx in neurons and contribute significantly to pathological Ca²⁺ elevations. To prevent Ca²⁺-induced damage, neurons have evolved a variety of feedback pathways that limit Ca²⁺ influx during periods of excessive extracellular glutamate or intense electrical activity (Stotz and Zamponi, 2001; Budde et al., 2002; Davis and Linn, 2003; Green et al., 2007; Jarvis and Zamponi, 2007). Among the most important of these feedback pathways is the internalization of Ca²⁺ channels, which are the main sources of Ca²⁺ in excitable cells. Glutamate receptors are internalized after tetanic stimulation or bath treatment with glutamate (Derkach et al., 2007; Lau and Zukin, 2007), but it is not known whether this type of stimulation also causes internalization of VGCCs.

VGCCs are classified according to their biophysical properties and susceptibility to pharmacological blockers. L-type VGCCs (LTCs) have a large persistent current that makes them a large source of intracellular Ca²⁺ (Tsien et al., 1988; Catterall, 2000). Ca_V1.2 is the most abundant LTC in the mammalian brain and is expressed in the spines, dendritic shafts, and cell bodies of most neurons in the brain. Several functions have been described for Ca_V1.2, including roles in gene expression, dendritic arborization, survival, and synaptic plasticity (Westenbroek et al., 1990; Hell et al., 1993; Catterall, 2000; Calin-Jageman and Lee, 2008). Ca²⁺ influx through Ca_V1.2 also contributes to neuronal death during ischemia, and LTC blockers have been shown to decrease neuronal death in models of stroke (Korenkov et al., 2000; Schurr, 2004).

The number of LTCs on the cell surface is altered under many physiological and pathological conditions. During aging and in Alzheimer's disease, there is an increase in the number of cell surface LTCs that is associated with neuronal dysfunction (Disterhoft et al., 1994; Haase et al., 1996; Thibault and Landfield, 1996; Coon et al., 1999). Studies in animals and humans have reported that LTC blockers can improve performance on memory tasks (for review see Thibault et al., 2007). In addition, a developmentally regulated increase in Ca_V1.2 levels could lead to cell death in Parkinson's disease (Day et al., 2006; Chan et al., 2007). However, despite the importance of LTCs in neuronal pathology, the mechanisms that control their expression and trafficking are not well understood.

Phosphatidylinositol (3,5)-bisphosphate (PtdIns(3,5)P₂) is a rare phospholipid that is found in endosomal membranes and is generated by the activity of a lipid kinase called phosphatidylinositol 3-phosphate 5-kinase (PIKfyve; mammals) or Fab1 (yeast; Michell et al., 2006). PIKfyve/Fab1 is critically important for the regulation of membrane trafficking and for the function of endosomes and lysosomes. Fab1 mutants have enlarged vacuoles and defects in multivesicular body invagination and vacuole acidification (Gary et al., 1998; Odorizzi et al., 1998; Efe et al., 2005). Genetic disruption of Fab1 is lethal in model organisms, and inhibition of PIKfyve leads to the enlargement of endosomes and formation of vacuoles in cells in culture (Ikonomov et al., 2001; Nicot et al., 2006; Rusten et al., 2006; Rutherford et al., 2006).

Recently, it has been shown that mutations that alter the production of PtdIns(3,5)P₂ cause severe neuronal defects in both mice and in humans. Mutations in mouse Fig4 and Vac14, two proteins which regulate PIKfyve (Duex et al., 2006; Sbrissa et al., 2007; Botelho et al., 2008), reduce PtdIns(3,5)P₂ levels and lead to ataxia and neuronal death (Chow et al., 2007; Zhang et al., 2007). In addition, mutations in other proteins that control PtdIns(3,5)P₂ levels cause neuronal and muscular diseases, including myotubular myopathy, amyotrophic lateral sclerosis, and Charcot-Marie-Tooth disease (Nicot and Laporte, 2008). This raises the question of why, despite the ubiquitous production of PtdIns(3,5)P₂ in cells, mutations in this pathway preferentially affect excitable cells like neurons and myocytes.

In this study, we report that sustained glutamate exposure causes internalization and degradation of Ca_V1.2 in cortical neurons. Using a proteomic screen, we identify PIKfyve as a Ca_V1.2-binding protein, and we show that this protein binds to Ca_V1.2 upon activation of glutamate receptors. PIKfyve is required for both internalization and degradation of the channel, and loss of PIKfyve increases the susceptibility of neurons to cell death in response to activation of glutamate receptors. Our study suggests that PIKfyve-dependent channel degradation is essential to prevent Ca²⁺-induced toxicity in neurons. These experiments provide a direct link between Ca²⁺ signaling and PIKfyve and help to explain why defects in PtdIns(3,5)P₂ signaling are deleterious to cells that are susceptible to excitotoxicity such as neurons and myocytes.

Ischemia and stroke cause neuronal death by triggering excess release of glutamate into the extracellular space, which activates Ca²⁺ influx through N-methyl-d-aspartic acid (NMDA) receptors and VGCCs and leads to Ca²⁺-induced toxicity (Schurr, 2004). Although it is known that activation of glutamate receptors causes internalization of AMPA and NMDA receptors (Derkach et al., 2007; Lau and Zukin, 2007), little is known about the effect of excess glutamate on the cell surface expression of VGCCs. To determine whether glutamate receptors regulate the number of Ca_V1.2 channels on the cell surface, we used an assay that we previously developed to measure the surface levels of Ca_V1.2 channels in primary neurons (Green et al., 2007). We introduced Ca_V1.2 channels containing both an intracellular YFP and an extracellular HA epitope into primary cortical neurons and used anti-HA antibodies to label channels on the cell surface. Because the HA antibodies do not cross the cell membrane, the ratio of the anti-HA to YFP fluorescence provides a quantitative measurement of the proportion of channels on the plasma membrane.

Stimulation with 50 µM glutamate led to a 60% decrease in the proportion of Ca_V1.2 channels on the cell surface, indicating that a significant proportion of Ca_V1.2 channels are internalized relatively rapidly by a large dose of glutamate (Fig. 1, A and B). To provide independent evidence that activation of glutamate receptors causes loss of cell surface Ca_V1.2 channels, we used Ca²⁺ imaging to measure Ca²⁺ influx in neurons before and after treatment with high concentrations of glutamate. Treatment with 50 µM glutamate for 10 min caused a significant decrease in the intracellular Ca²⁺ rise induced by depolarization. This decrease could be observed both at the population level and in single cells (Fig. 1 C). Together, these experiments suggest that relatively prolonged activation of glutamate receptors causes loss of Ca_V1.2 channels on the cell membrane.

Because prolonged treatment with glutamate can cause excitotoxicity in neurons, we next investigated whether loss of Ca_V1.2 channels was correlated with cell death in neurons. We treated neurons with 10, 50, or 100 µM glutamate for 3, 10, or 30 min or 10 h and measured cell death 10 h later by counting pyknotic nuclei, measuring MAP2 levels in dendrites, or using TUNEL staining (Fig. S1). Although treatment with 50 or 100 µM glutamate for 30 min or 10 h caused substantial apoptotic cell death, stimulation with 50 µM glutamate for 10 min (conditions which cause channel internalization) did not lead to cell death. This suggests that Ca_V1.2 internalization is not correlated with cell death and therefore is more likely to be a cellular response to a potentially cytotoxic stimulus.

In a previous study, we found that stimulation of neurons using relatively mild depolarizing stimuli to activate Ca_V1.2 channels caused a reversible loss of cell surface Ca_V1.2 channels (Green et al., 2007). Under these conditions, Ca_V1.2 channels are returned to the membrane rapidly after restoration of the resting membrane potential. To investigate whether activation of glutamate receptors also led to reversible loss of Ca_V1.2 channels from the cell membrane, we stimulated neurons with 50 µM glutamate for 10 min and then returned the cells to resting conditions for 20 min before measuring cell surface channels. In contrast to our previous findings (Green et al., 2007), we observed no recovery of Ca_V1.2 channels on the cell membrane 20 min after treatment with 50 µM glutamate (Fig. 1 D), suggesting that activation of glutamate receptors leads to a permanent loss of Ca_V1.2 channels on the cell surface.

To determine whether the loss of Ca_V1.2 from the cell membrane was caused by channel degradation, we measured Ca_V1.2 levels in neurons either by Western blotting or by immunofluorescence. We stimulated neurons with 50 µM glutamate and used SDS-PAGE and anti-Ca_V1.2 antibodies to detect endogenous Ca_V1.2 channels (Fig. S2, A–C). We found that this treatment caused a reproducible decrease in the levels of Ca_V1.2 relative to a glyceraldehyde-3-phosphate dehydrogenase (GAPDH) internal control (Fig. 1 E), suggesting that Ca_V1.2 was degraded after activation of glutamate receptors. To provide further evidence of glutamate-induced Ca_V1.2 degradation, we used immunocytochemistry to detect Ca_V1.2 channels in cultures of neurons before and after treatment with 50 µM glutamate for 10 min. Glutamate caused a 35% loss of Ca_V1.2 immunoreactivity from the dendrites and a redistribution of the channel to a region around the cell nucleus (Fig. 1, F and G; and Fig. S2 D). This decrease in the levels of the endogenous channel was not reversed by a 20-min incubation in glutamate-free media (Fig. 1 H), providing additional evidence for Ca_V1.2 channel degradation.

We next investigated whether Ca_V1.2 degradation depended on the elevation of the intracellular Ca²⁺ concentration ([Ca²⁺]_i) and on the activation of specific types of glutamate receptors. We used Fura-2 to measure [Ca²⁺]_i and found that treatment with 50 µM glutamate, which causes channel degradation, caused a [Ca²⁺]_i rise that was significantly greater than treatment with 10 µM glutamate, 67 mM KCl, or 2 µg/ml ionomycin (Fig. 2 A). To determine whether Ca²⁺ influx was required for glutamate-dependent Ca_V1.2 degradation, we stimulated cells with 50 µM glutamate in media with reduced extracellular Ca²⁺ and found that this completely prevented Ca_V1.2 degradation (Fig. 2 B). This indicates that Ca²⁺ influx across the membrane is required for this process.

Glutamate stimulation elevates [Ca²⁺]_i by a variety of mechanisms, including activation of Ca²⁺-permeable AMPA and NMDA receptors, metabotropic glutamate receptors, and indirect activation of VGCCs. To examine which of these pathways is responsible for Ca_V1.2 degradation, we treated neurons with 50 µM glutamate in the presence of NBQX, MK801, or DL-AP3 to block AMPA, NMDA, or metabotropic glutamate receptors, respectively. Treatment with NBQX and DL-AP3 had no effect on Ca_V1.2 degradation, but treatment with MK801 prevented loss of Ca_V1.2 from dendrites (Fig. 2, C and D), indicating that Ca_V1.2 degradation depends on NMDA receptors. This is in agreement with a previous study of loss of LTC currents in neurons treated with NMDA (Davis and Linn, 2003).

Glutamate-triggered Ca_V1.2 degradation could occur either via the lysosomal or the proteosomal pathways. To determine which pathways regulate channel degradation, we treated cells with the lysosomal inhibitors bafilomycin and chloroquine or with the proteosomal inhibitor lactacystin. Both bafilomycin and chloroquine significantly inhibited glutamate-induced Ca_V1.2 degradation, whereas lactacystin had no significant effect (Fig. 3 A). In addition, NH₄Cl, which prevents lysosomal acidification, caused a large increase in the number of Ca_V1.2 channels on the cell surface (Fig. 3 B) and reversed the glutamate-induced decrease in Ca_V1.2 channels (Fig. 3 C). Together, these data suggest that lysosomes play a central role in Ca_V1.2 channel degradation in neurons.

To provide further evidence for the importance of lysosomes in the degradation of Ca_V1.2, we investigated whether treatment of neurons with glutamate caused relocalization of Ca²⁺ channels to lysosomal compartments. We introduced a Flag-Myc–labeled Ca_V1.2 channel (Flag-Myc-Ca_V1.2) and a GFP fusion protein of the lysosomal-associated membrane protein (LAMP; LAMP-GFP) into neurons (Fig. S3 A). Treatment with glutamate caused a significant increase in the colocalization of Ca_V1.2 and LAMP-GFP in neuronal dendrites (Fig. 3 D), suggesting that the channels are targeted to lysosomes by activation of glutamate receptors.

To provide further insight into the mechanisms that underlie Ca_V1.2 internalization and degradation, we used a proteomic approach to identify Ca_V1.2-binding proteins. We expressed GST fusion proteins of the intracellular domains of Ca_V1.2 in Neuro2A neuroblastoma cells, immunoprecipitated the proteins, and used multidimensional mass spectrometry to identify associated proteins. GST-YFP was immunoprecipitated and analyzed in parallel as control for nonspecific protein binding. We repeated the experiment five times and generated a score for each Ca_V1.2-interacting protein that reflected the number of times that the protein bound to Ca_V1.2 relative to YFP (Table S1). One of the main proteins that associated with the cytoplasmic C terminus of Ca_V1.2 was PIKfyve, a lipid kinase which is localized in early endosomes (Fig. S3, B and C) and controls endosomal and lysosomal trafficking. Because Ca_V1.2 is degraded in lysosomes in response to glutamate stimulation, we investigated whether PIKfyve has a role in regulating this process.

To verify that Ca_V1.2 interacts with PIKfyve in cells, we introduced Myc-PIKfyve and Flag-Ca_V1.2 expression plasmids into HEK 293T cells, lysed the cells, and immunoprecipitated the channel using anti-Flag or anti-Myc antibodies. Immunoprecipitation of the full-length channel (Fig. 4 C) or of the cytoplasmic C terminus of the channel (Fig. 4 A) resulted in coimmunoprecipitation of full-length PIKfyve. Conversely, immunoprecipitation of PIKfyve caused coimmunoprecipitation of the Ca_V1.2 C terminus (Fig. 4 B). We next expressed GFP-PIKfyve and the Ca_V1.2 C terminus in HEK 293T cells. Expression of the Ca_V1.2 C terminus caused GFP-PIKfyve to form large perinuclear aggregates, providing additional evidence that these two proteins interact in cells (Fig. S3 D).

We then used deletion analysis to map the domains of PIKfyve and Ca_V1.2 that interact with each other. We subdivided PIKfyve into five fragments (Fig. 4 D), expressed them in HEK 293T cells, and tested their ability to interact using coimmunoprecipitation assays. The two N-terminal fragments of PIKfyve did not associate with the C terminus of Ca_V1.2, but the three C-terminal fragments all interacted with this channel (Fig. 4, E and F). We also subdivided the C terminus of Ca_V1.2 into five fragments and tested the association of these fragments with full-length PIKfyve (Fig. 4 G). We found that a domain composed of aa 1,709–1,808 of Ca_V1.2 (CT-2) interacted with full-length PIKfyve, whereas other regions did not (Fig. 4 H). Expression of this CT-2 fragment and the fragment immediately terminal to it (aa 1,809–1,928 or CT-3) caused GFP-PIKfyve to form perinuclear aggregates that were identical to those formed with the full C terminus of Ca_V1.2 (Fig. 4 I). Together, these results suggest that aa 1,709–1,808 of Ca_V1.2 and the C-terminal fragment of PIKfyve are required for the interaction of these two proteins. The region from aa 1,709–1,808 is highly conserved in LTCs across various species, suggesting that it is likely to define a motif that is important for channel function. Although the nature of these proteins made it impossible to purify them from bacteria to test for a direct interaction, our data are consistent with a direct interaction between the C-terminal domains of Ca_V1.2 and PIKfyve.

We next investigated whether Ca_V1.2 interacts with PIKfyve in a glutamate-dependent manner. We used anti-PIKfyve antibodies to immunoprecipitate PIKfyve from cultured neurons before and after treatment with glutamate and used anti-Ca_V1.2 to detect the channel by Western blotting. Ca_V1.2 did not associate with PIKfyve in resting cells, but in neurons stimulated with glutamate, Ca_V1.2 (Fig. 4 J) was reproducibly coimmunoprecipitated with PIKfyve. These data argue that endogenous Ca_V1.2 associates with endogenous PIKfyve and, further, that this occurs in response to activation of glutamate receptors.

The inducible interaction of PIKfyve with Ca_V1.2 under conditions of excitotoxic stress suggested that PIKfyve could be activated by glutamate receptors. To study this question, we made an intracellular sensor for PtdIns(3,5)P₂, which is the main product of PIKfyve in cells. We made a fusion protein of GFP and two copies of the GRAM domain from the phosphatase myotubularin (MTM1; Fig. 5 A), which binds to PtdIns(3,5)P₂ (Tsujita et al., 2004). We introduced this protein into neurons and, upon treatment with 50 µM glutamate, observed a marked relocalization from the cytoplasm and the nucleus to a perinuclear region (Fig. 5, B and C). This change in localization was not observed in cells expressing GFP alone, suggesting that it is specific to the GRAM-containing protein. The change in GFP-GRAM localization was only observed in neurons treated with 50 µM glutamate but not with glutamate in the presence of EGTA, elevated KCl, or ionomycin, indicating that this event occurs only under the conditions that lead to Ca_V1.2 degradation (Fig. 5 D).

The perinuclear localization of GFP-GRAM after glutamate treatment was very similar to that of lysosomes and late endosomes, so we explored whether GFP-GRAM relocalizes to these compartments after activation of glutamate receptors. We coexpressed in neurons GFP-GRAM and the lysosome marker LAMP-Cherry and observed a small increase in the colocalization of the two proteins after treatment with glutamate (Fig. S3 E). This indicates that PtdIns(3,5)P₂ is generated in endosomal compartments in response to glutamate receptor activation.

To provide independent evidence that activation of glutamate receptors leads to an increase in PtdIns(3,5)P₂ in neurons, we used a monoclonal antibody to detect PtdIns(3,5)P₂. This antibody is relatively specific for PtdIns(3,5)P₂, as determined by its ability to detect a decrease in PtdIns(3,5)P₂ levels in knockouts of ppk-3, the Caenorhabditis elegans homologue of PIKfyve (Nicot et al., 2006). Treatment of neurons with 50 µM glutamate resulted in an increase in anti-PtdIns(3,5)P₂ fluorescence, which is consistent with our earlier finding (Fig. 5 B) that glutamate leads to relocalization of the GFP-GRAM domain (Fig. 5, E and F). Together, these results provide evidence that activation of glutamate receptors leads to an increase in PtdIns(3,5)P₂ levels in neurons.

We next asked whether the increase of PtdIns(3,5)P₂ triggered by activation of glutamate receptors depends on PIKfyve activity. We made short hairpin RNAs (shRNAs) that target PIKfyve (shPIKfyve) and two PIKfyve constructs that contain silent mutations in the region recognized by the shRNA that could be used to restore PIKfyve expression in the presence of the shRNAs (wild type [PIKfyve WTres] and kinase negative [PIKfyve KEres]). shPIKfyve reduced the expression of PIKfyve in HEK 293T cells but not of the shRNA-resistant proteins, as measured by Western blotting of Myc-tagged proteins (Fig. S4 A). In addition, shPIKfyve caused a dramatic increase in the number and size of vacuoles in NIH3T3 cells that was suppressed by the introduction of PIKfyve WTres but not of PIKfyve KEres (Fig. S4, B and C), indicating that these shRNAs also reduce the expression of the endogenous proteins. To determine whether PIKfyve generates PtdIns(3,5)P₂ in response to glutamate, we introduced shPIKfyve into neurons, stimulated the cells with 50 µM glutamate, and measured PtdIns(3,5)P₂ production using the GFP-GRAM protein. Introduction of shPIKfyve eliminated the relocalization of GFP-GRAM to the perinuclear region, suggesting that it also eliminated the production of PtdIns(3,5)P₂ (Fig. 5 G). Interestingly, we observed relatively few vacuoles in most neurons expressing the shPIKfyve under resting conditions, suggesting that PIKfyve activity is relatively low in resting neurons (Fig. S4 D). Introduction of a PIKfyve WTres along with shPIKfyve restored the relocalization of the GRAM domain to the perinuclear region in response to glutamate, whereas the PIKfyve KEres domain did not rescue the effects of shPIKfyve on the localization of the GFP-GRAM reporter. Together, these results provide evidence that activation of glutamate receptors leads to increases in PtdIns(3,5)P₂ levels in neurons and that this is mediated by PIKfyve.

Because PIKfyve regulates endosome trafficking, we next investigated whether it contributes to the glutamate-dependent internalization of Ca_V1.2 channels. We transfected neurons with shPIKfyve together with YFP-HA-Ca_V1.2 and measured the surface levels of Ca_V1.2 after stimulation with glutamate. Knockdown of PIKfyve significantly suppressed glutamate-dependent Ca_V1.2 internalization (Fig. 6, A and B), whereas a scrambled control shRNA (shScrambled) had no effect, suggesting that PIKfyve controls Ca_V1.2 levels on the cell surface. To determine whether PIKfyve also regulates degradation of endogenous Ca_V1.2, we introduced shPIKfyve into neurons and measured the amount of Ca_V1.2 channel in dendrites using anti-Ca_V1.2 antibodies (Fig. 6, C and D). shPIKfyve significantly decreased the degradation of Ca_V1.2 in neurons treated with glutamate compared with shScrambled. Introduction of PIKfyve WTres restored the glutamate-dependent degradation of Ca_V1.2, whereas PIKfyve KEres had no effect. Thus, PIKfyve seems to be required both for Ca_V1.2 internalization and degradation in response to increases in glutamate. Because the main product of PIKfyve is PtdIns(3,5)P₂, we asked whether PtdInd(3,5)P₂ was necessary for glutamate-induced Ca_V1.2 degradation. We overexpressed the phosphatase MTM1, which dephosphorylates PtdIns(3,5)P₂ at the 3 position, and measured dendritic Ca_V1.2 levels in neurons. Expression of MTM1 reduced glutamate-induced Ca_V1.2 degradation in dendrites (Fig. S5, A and B), providing additional evidence that PIKfyve and PtdInd(3,5)P₂ help mediate Ca_V1.2 degradation.

We next investigated whether the interaction of PIKfyve with the C terminus of Ca_V1.2 was important for glutamate-dependent Ca_V1.2 degradation. We generated a Myc-Ca_V1.2 channel that lacks the C-terminal PIKfyve-binding site (Ca_V1.2 ΔCT), introduced it into neurons, and measured its degradation after activation of glutamate receptors. Although the wild-type channels were efficiently degraded by activation of glutamate receptors, the levels of Ca_V1.2 ΔCT channels were not significantly reduced (Fig. 6, E and F), suggesting that binding of Ca_V1.2 to PIKfyve is required for Ca_V1.2 degradation in response to the activation of glutamate receptors.

Finally, we tested whether PIKfyve regulates the targeting of Ca_V1.2 channels to lysosomes after activation of glutamate receptors. We introduced LAMP-GFP and Myc-Ca_V1.2 into neurons along with shPIKfyve or shScrambled constructs and then treated cells with glutamate. We measured the colocalization of Ca_V1.2 and LAMP by calculating the coefficient of correlation between Ca_V1.2 and LAMP in cells. Treatment with glutamate caused a loss of diffuse Ca_V1.2 staining and the appearance of Myc-Ca_V1.2 puncta, some of which colocalized with LAMP-GFP, giving rise to an increase in the coefficient of correlation between the two images. Knockdown of PIKfyve substantially reduced the colocalization of Ca_V1.2 and LAMP after activation of glutamate receptors (Fig. 6 G), supporting the idea that Ca_V1.2 channels are trafficked to lysosomes after treatment with glutamate in a manner that depends on PIKfyve.

Under conditions of stroke and anoxia, glutamate is a potent excitotoxic agent that acts by increasing intracellular Ca²⁺ levels both through NMDA receptors and through LTCs (Schurr, 2004). We initially characterized the contribution of different Ca²⁺ channels to glutamate-triggered excitotoxicity in our cells by treating neurons for 30 min with 50 µM glutamate in the presence of blockers for NMDA receptors (MK801), AMPA receptors (NBQX), LTCs (nimodipine), P/Q-type VGCCs (agatoxin IVA), N-type VGCCs (ω-conotoxin GVIA), and R-type VGCCs (SNX482) and measuring cell death 10 h later (Fig. 7, A and B). As expected, the NMDA receptor MK801 blocked ∼80% of the cell death, the AMPA receptors and LTC blockers reduced cell death ∼25%, and the other channel blockers had essentially no effect on glutamate-induced toxicity. This indicates that although NMDA receptors are the primary triggers of excitotoxicity, both LTCs and AMPA receptors also play a role in this process.

The observation that Ca²⁺ influx through LTCs can contribute to toxicity led us to ask whether one of the functions of PIKfyve-dependent Ca_V1.2 degradation might be to reduce Ca²⁺-dependent toxicity in neurons. To answer this question, we transfected neurons with either shScrambled or shPIKfyve and treated the cells with 50 µM glutamate for 10 min. 50 µM glutamate for 10 min did not cause cell death in untransfected neurons or neurons containing shScrambled (Fig. 7 C and Fig. S1). In contrast, treatment with glutamate caused a large increase in apoptosis in shPIKfyve-containing neurons (Fig. 7, C and D). The increase in glutamate-induced cell death in cells containing shPIKfyve was prevented by expression of PIKfyve WTres but not by PIKfyve KEres, indicating that the effects of shPIKfyve are specific and that the kinase activity of PIKfyve is required for its effect on survival. Together, these results strongly suggest that PIKfyve is important for protecting neurons from apoptosis under conditions of moderate glutamate receptor activation.

To determine whether PIKfyve protects neurons from excitotoxicity by mediating internalization and degradation of Ca_V1.2 channels, we transfected neurons with shPIKfyve and treated them with 50 µM glutamate for 10 min in the presence or absence of the Ca_V1.2 blocker nimodipine. We found that blocking Ca_V1.2 channels protected ∼30% of the neurons lacking PIKfyve from glutamate-induced toxicity (Fig. 7 E), suggesting that excess Ca²⁺ entry through LTCs contributes to the increased sensitivity to glutamate in the absence of PIKfyve. Because AMPA receptors also contribute to excitotoxic cell death and these receptors are also known to be internalized in response to treatment of cells with glutamate, we investigated whether blocking AMPA receptors alone or in combination with LTCs prevented glutamate-induced cell death in neurons lacking PIKfyve. Blocking AMPA receptors expressing the shPIKfvye shRNA reduced cell death by ∼30%, whereas the combination of both LTC and AMPA blockers almost completely prevented cell death in neurons. This result argues that the susceptibility to glutamate-induced apoptosis in neurons lacking PIKfyve is caused by the activity of both LTCs and AMPA receptors (Fig. 7 F).

This result also raised the interesting possibility that AMPA receptor internalization in response to glutamate might be mediated by PIKfyve. We examined this hypothesis by introducing into neurons one of the subunits of the AMPA receptor, the GluR2 channel, tagged with an intracellular YFP and an extracellular HA tag. This construct was robustly internalized by treatment of neurons with 50 µM glutamate for 10 min. However, introduction of the PIKfyve shRNA completely prevented GluR2 internalization, indicating that this event also depends on PIKfyve (Fig. S5, C–E). In control experiments, we also tested the role of PIKfyve in regulating the internalization of K_V1.2 channels but found that the surface expression of these channels did not depend on PIKfyve (Fig. S5, F and G). Thus, PIKfyve seems to be important for the internalization of at least one other channel in addition to Ca_V1.2 but not for all cell surface channels.

This study provides new insights into the mechanisms by which neurons are protected from excessive Ca²⁺ influx and excitotoxicity after excessive activation of NMDA receptors, such as what occurs during transient ischemic attacks, epileptic seizures, and in the penumbra of strokes. Our results suggest that neurons internalize Ca_V1.2 channels and AMPA receptors in response to sustained activation of glutamate receptors. This is followed by recruitment of PIKfyve to the early endosomes, probably by binding to the C terminus of Ca_V1.2 and to lipids that are enriched in the endosomal membrane. Recruitment of PIKfyve to the endosomes leads to an increase in PtdIns(3,5)P₂ levels in these vesicles, which promotes their maturation and eventual fusion with lysosomes and leads to Ca_V1.2 degradation (Fig. 8). Thus, prolonged activation of NMDA receptors leads to both removal and degradation of Ca_V1.2 channels by a PIKfyve-dependent mechanism.

This PIKfyve-mediated degradation pathway seems to be essential for the ability of neurons to respond to moderate elevations of extracellular glutamate. Reducing PIKfyve or disrupting PIKfyve's association with Ca_V1.2 channels prevents channel degradation and leads to cell death when neurons are stimulated with subtoxic concentrations of glutamate. Blocking LTCs and AMPA receptors blocks the increase in cell death in neurons lacking PIKfyve, indicating that these ion channels can be toxic even at moderate levels of excitability and need to be down-regulated. The increased susceptibility of neurons to excitotoxicity in the absence of PIKfyve may help to explain why diseases that arise from mutations that alter PtdIns(3,5)P₂ levels disproportionately affect excitable cells such as neurons.

The important finding is that intracellular Ca²⁺ elevations in mammalian neurons regulate the levels of PtdIns(3,5)P₂. Circumstantial evidence suggests that Fab1 function may be regulated by Ca²⁺ signals in other organisms as well. In budding yeast, PtdIns(3,5)P₂ levels are increased by hyperosmotic shock, a process which increases the concentration of intracellular Ca²⁺. In addition, Fab1 mutants are hypersensitive to elevated extracellular Ca²⁺ levels, suggesting that this pathway is important for the response to Ca²⁺-induced stress. Finally, in fission yeast, mutations of sst1, a vesicular Ca²⁺ transporter, suppress the phenotype of cells containing mutations of the Fab1 homologue ste12 (Matsumoto et al., 2002; Onishi et al., 2003). Our data and these corroborating studies suggest that Ca²⁺ regulation of PIKfyve activity might be a relatively ancient mechanism for cellular protection, although this hypothesis clearly needs to be investigated further in other systems.

Because PIKfyve seems to play a general role in the degradation of a variety of cell surface proteins, it is interesting to speculate why it interacts with Ca_V1.2 channels. One possibility is that endosomes derived from Ca²⁺-dependent endocytosis in dendrites and synapses often contain Ca_V1.2 channels. By binding to Ca_V1.2, PIKfyve can be recruited to a specific type of activity-dependent endosome and can play a role in the degradation of postsynaptic proteins. This is consistent with the general notion that the recruitment of proteins that contain FYVE domains to endosomal membranes is mediated by binding to both PtdIns(3)P and proteins located on endosomes (Simonsen et al., 1998; Lawe et al., 2000). It will be interesting to determine whether PIKfyve also participates in the Ca²⁺-dependent degradation of other postsynaptic proteins such as NMDA receptors and whether this depends on PIKfyve binding to Ca_V1.2 channels.

Although PIKfyve has been reported to play a role in endosomal trafficking and lysosome function, it has not been shown to control endocytosis directly. We found that PIKfyve knockdown prevents Ca_V1.2 and AMPA receptor internalization in response to glutamate. This finding suggests either that PIKfyve can regulate the endocytic machinery directly or, alternatively, that altering endosome trafficking can prevent the activity or recycling of molecules required for endocytosis. The fact that we only observe PIKfyve in endosomes and not at the cell membrane argues that the effects of PIKfyve on endocytosis are indirect, but this question will need to be studied more carefully in the future.

Our finding that PIKfyve interacts with Ca_V1.2 and regulates its degradation also has important implications for the pathology of several diseases that affect myocytes and neurons. These cells express large numbers of Ca_V1.2 channels, suggesting that some aspects of the pathology observed in these patients could be caused by dysregulation of VGCCs or, more generally, by failure of activity-dependent protein degradation. If this hypothesis is correct, patients with PtdIns(3,5)P₂ phosphatase mutations might benefit from the use of Ca²⁺ channel blockers like gabapentin and pregabalin, which are widely used clinically for other applications.

In summary, our study identifies a novel mechanism for Ca²⁺ channel degradation in neurons that integrates two unrelated fields, Ca²⁺ channel signaling and endosomal trafficking. These experiments provide new insights into channel homeostasis and suggest a rationale for investigating Ca²⁺ channel dysfunction in diseases caused by defects in PtdIns(3,5)P₂ signaling.

Lipofectamine 2000, terminal deoxynucleotidyl transferase, Fura-2 acetoxymethyl ester, and Hoechst 33258 were purchased from Invitrogen; MK801, NBQX, and nimodipine were purchased from Sigma-Aldrich; SNX482 was purchased from Alomone Labs; DL-AP3 was purchased from Tocris Bioscience; bafilomycin A1 and lactacystin were purchased from EMD; chloroquine was purchased from MP Biomedicals; biotin–deoxy-UTP was purchased from TriLink BioTechnologies; and fluorescein-conjugated streptavidin (DTAF-streptavidin) was purchased from Jackson ImmunoResearch Laboratories, Inc. Antibodies to Ca_V1.2 (1:500; AB5156; Millipore), PIKfyve (1:500; 6C7; Abnova), Myc (1:500; 4A6; Millipore), HA (3F10; Roche), Flag (1:500; M2; Sigma-Aldrich), GAPDH (1:1,000; 6C5; Applied Biosystems), or DsRed (1:500; E64-1077; BD) was used for immunoblot analysis. Antibodies to Ca_V1.2 (1:100; AB5156), MAP2 (1:500; MAB378; Millipore), Myc (1:500; 4A6), HA (1:500; 3F10), PtdIns(3,5)P₂ (1:100; Z-P035; Echelon), or GFP (1:500; 598; MBL) was used for immunocytochemistry.

HEK 293T and NIH3T3 cells were cultured in Dulbecco's modified minimum essential medium containing 10% FBS, 100 U penicillin and 100 µg streptomycin (P/S), and 2 mMl-glutamine (Q). Neuro2A cells were cultured in minimum essential medium containing 10% FBS and P/S. Primary neuronal cultures were prepared from embryonic day 17–19 Sprague-Dawley rats by dissociating dissected cortices with enzyme solution (Hanks’ balanced salt solution with 600 U papain and 0.32 mg/ml cystein). Neurons were plated on coverslips coated with 20 µg/ml polyornithine and 1.42 µg/ml laminin and maintained in culture in Basal Medium Eagle with 5% FBS, P/S, Q, and 0.6% glucose. For immunocytochemistry, neurons were stimulated with glutamate in Tyrode's solution without Mg (129 mM NaCl, 5 mM KCl, 3 mM CaCl₂, 30 mM glucose, 0.1% BSA, and 25 mM Hepes, pH 7.4). Cell lines and cortical neurons were transfected with plasmids using Lipofectamine 2000 according to the manufacturer's instructions. The ratio of plasmids used for specific experiments is given as follows: for surface expression assay, α1C/β1b/α2δ = 2:1:1 (Fig. 1, A–C) and α1C/β1b/α2δ/shRNA = 2:1:1:2 (Fig. 6, A and B); for colocalization assay, α1C/β1b/α2δ/LAMP = 2:1:1:1 (Fig. 3 C) and α1C/β1b/α2δ/LAMP/shRNA = 2:1:1:1:1 (Fig. 6 G); for coimmunoprecipitation, PIKfyve/Ca_V1.2 CT = 1:1 (Fig. 4, A and B), PIKfyve/α1C/β1b = 2:2:1 (Fig. 4 C), PIKfyveΔ/Ca_V1.2 CT = 1:1 (Fig. 4, E and F), and PIKfyve/Ca_V1.2 ΔCT = 1:1 (Fig. 4 H); for relocalization of the GRAM domain, GRAM/shRNA/PIKfyve = 1:2:2 (Fig. 5 F); for Ca_V1.2 degradation assay, PIKfyve/shRNA = 1:1 (Fig. 6, C and D) and α1C/β1b/α2δ = 2:1:1 (Fig. 6, E and F); and for cell death assay, shRNA/PIKfyve = 1:1 (Fig. 7, C and D).

GFP-PIKfyve (mouse) and GFP-HA-K_V1.2 constructs were provided by P. Cullen (University of Bristol, Bristol, England, UK; Rutherford et al., 2006) and L.Y. Jan (University of California, San Francisco, San Francisco, CA; Gu et al., 2003), respectively. The cDNA for PIKfyve was amplified by PCR from the GFP-PIKfyve construct and cloned into the XbaI sites of pCS4-Myc and pCS4-HA. The pCS4-GFP, pCS4-Cherry, and pCS4-HA-GFP plasmids were constructed by the insertion of the EGFP and Cherry coding sequence into the XbaI site in pCS4 and pCS4-HA, respectively. The Rab5, LAMP1, and MTM1 cDNAs were amplified by PCR from the human ORFeome cDNA library (Thermo Fisher Scientific) and subcloned into the BamHI sites of pcDNA3-GFP, pCS4-GFP, pCS4-Cherry, and pCS4-Myc to generate three different vectors encoding each protein. The GluR2 cDNA was amplified from the human ORFeome cDNA library and subcloned into the XbaI sites of pCS4-HA-GFP. The dihydropyridine-resistant Ca_V1.2 α1C, β1b, α2δ, and pEYFP-HA-Ca_V1.2 constructs were described previously (Dolmetsch et al., 2001; Gomez-Ospina et al., 2006; Green et al., 2007). The pGW1-Flag-Ca_V1.2, pGW-Flag-α2δ, GST–Ca_V1.2 CT, and pDEST27-GST-YFP were generated using Gateway technology (Invitrogen). The cDNA for Ca_V1.2 CT was amplified by PCR and cloned into the BamHI site of pcDNA3-Flag. Short hairpin oligonucleotides were designed and inserted into the RNAi-Ready pSIREN-DNR-DsRed-Express vector (Takara Bio Inc.) by ligation into the BamHI and EcoRI sites. For specific experiments, DsRed was eliminated from the RNAi-Ready pSIREN-DNR-DsRed-Express vector using the NcoI restriction enzyme. Site-directed mutagenesis was performed using QuikChange (Agilent Technologies) to generate RNAi-resistant kinase-negative PIKfyve (PIKfyve KEres; Sbrissa et al., 2000). PIKfyve KEres consisted of changing the K¹⁸²⁰ to E in the kinase domain. The cDNA from the GRAM domain of myotubularin (residues 8–112) was amplified by PCR from the MTM1 construct and cloned into pcDNA3-GFP.

The primers used were as follows: PIKfyve forward, 5′-ATCTAGAATGGCCACAGATGACAAG-3′; and reverse, 5′-GTCTAGATCAGCAATTCAGATCCAA-3′; PIKfyve FYVE forward, 5′-AGGATCCATGGCCACAGATGACAAG-3′; and reverse, 5′-GGGATCCCTAACTTAAGGCTATTTT-3′; PIKfyve DEP forward, 5′-GGGATCCTATGCTCATTCTACAGAC-3′; and reverse, 5′-AGGATCCTTATGAGAGCTGCTGTCC-3′; PIKfyve Cpn/TCP-1 forward, 5′-GAGATCTATAAGTGATGCCTTCATC-3′; and reverse, 5′-AAGATCTTTATGAGCATCTCATCCC-3′; PIKfyve linker forward, 5′-GAGATCTACTCGAGATTATTTTCCA-3′; and reverse, 5′-AAGATCTTTAGATGAATTCCTCCTC-3′; PIKfyve PIP5K forward, 5′-AGGATCCCGTTCCCTTTCTCACTCA-3′; and reverse, 5′-GGGATCCTCAGCAATTCAGATCCAA-3′; PIKfyve KEres sense, 5′-AGATTCATTCTGGAGCAAATGCCTCGTTTG-3′; and antisense, 5′-CAAACGAGGCATTTGCTCCAGAATGAATCT-3′; Ca_V1.2 CT-1 forward, 5′-GCCACCATGGTCGGCAAGCCCTCGCAGAGG-3′; and reverse, 5′-CTAGTTGCCAAACAGGCCTCCAGCCCTCCTG-3′; Ca_V1.2 CT-2 forward, 5′-GCCACCATGCACGTCAGCTACTACCAGAGTG-3′; and reverse, 5′-CTACTGTACCCGGACAGCAGGGGACAAGGG-3′; Ca_V1.2 CT-3 forward, 5′-GCCACCATGGAGGCAGCATGGAAACTCAGC-3′; and reverse, 5′-CTACTTTAGACATTCCAGATGGAAGGAGGC-3′; Ca_V1.2 CT-4 forward, 5′-GCCACCATGCGACAAAAGGATCAAGGGGGAG-3′; and reverse, 5′-CTAGCCCCCACTACAGGCTGTGGTCTCCTC-3′; Ca_V1.2 CT-5 forward, 5′-GCCACCATGAGCAGCATGGCCCGGAGAGCC-3′; and reverse, 5′-CTACAGGTTGCTGACATAGGACCTG-3′; GFP forward, 5′-AATCTAGAATGGTGAGCAAGGGCGAG-3′; and reverse, 5′-GGACTAGTTTACTTGTACAGCTCGTC-3′; Cherry forward, 5′-CTCTAGAGCCACCATGGTGAGCAAGGGCGAG-3′; and reverse, 5′-GACTAGTTTACTTGTACAGCTCGTCCATGCCGCC-3′; Rab5 forward, 5′-AGGATCCATGGCTAGTCGAGGCGC-3′; and reverse, 5′-GGGATCCTTAGTTACTACAACACTGATTCCTGG-3′; LAMP1 forward, 5′-AGGATCCGCCACCATGGCGGCCCCCGGCAG-3′; and reverse, 5′-GGGATCCGATAGTCTGGTAGCCTGCGTGACTCC-3′; MTM1 forward, 5′-AGGATCCGCCACCATGGCTTCTGCATCAACTTC-3′; and reverse, 5′-AGGATCCTCAGAAGTGAGTTTGCACATGGGG-3′; 2× GRAM forward, 5′-GGGATCCGCCACCATGAAATATAATTCACACTCC-3′; and reverse, 5′-AGAATTCAGTAATATCTAGACCATAGGAATT-3′; 2× GRAM forward, 5′-AGAATTCATGAAATATAATTCACACTCCTTG-3′; and reverse, 5′-AGGATCCAGTAATATCTAGACCATAGGAATTTTCTCC-3′; and GluR2 forward, 5′-GTCTAGAGCCACCATGCAAAAGATTATGCATAT-3′; and reverse, 5′-GTCTAGAAATTTTAACACTTTCGATGCCATA-3′.

The oligonucleotides used were as follows: mouse PIKfyve shRNA sense, 5′-GATCCGGACAGGGTTGGATCTGAATTCAAGAGATTCAGATCCAACCCTGTCCTTTTTTACGCGTG-3′; and antisense, 5′-AATTCACGCGTAAAAAAGGACAGGGTTGGATCTGAATCTCTTGAATTCAGATCCAACCCTGTCCG-3′; rat PIKfyve shRNA sense, 5′-GATCCGGACAGGGTTGGATCTCAATTCAAGAGATTGAGATCCAACCCTGTCCTTTTTTACGCGTG-3′; and antisense, 5′-AATTCACGCGTAAAAAAGGACAGGGTTGGATCTCAATCTCTTGAATTGAGATCCAACCCTGTCCG-3′; and shScrambled shRNA sense, 5′-GATCCTGTAGGTCGAGAGCGTAGATTCAAGAGATCTACGCTCTCGACCTACATTTTTTACGCGTG-3′; and antisense, 5′-AATTCACGCGTAAAAAATGTAGGTCGAGAGCGTAGATCTCTTGAATCTACGCTCTCGACCTACAG-3′. RNAi-resistant PIKfyve cDNA was made by introducing silent mutations using conventional PCR. The PCR primers used were as follows: forward, 5′-TCTAGAATGGCCACAGATGACAAGAGTTCCCCGACACTGGACTCTGCTAATGATTTG-3′; and reverse, 5′-TCTAGATTAGCAATTTAAGTCTAGTCCCGTCCAGTGGTCTGGCACCATCAAGAAATACTT-3′.

Cortical neurons plated on 12- or 15-mm coverslips and grown in 12- or 24-well plates were fixed with 4% paraformaldehyde in PBS for 10 min at room temperature. The coverslips were washed in PBS once and incubated with a blocking solution (5% BSA in PBS) with (for regular immunocytochemistry) or without (for surface expression assay) 0.4% Triton X-100 for 30 min. The samples were incubated with the indicated primary antibodies for 1 h at room temperature or overnight at 4°C and then washed three times in PBS at room temperature before incubation with secondary antibodies (Alexa Fluor 350 anti–mouse IgG [1:1,000], Alexa Fluor 488 anti–mouse/rabbit IgG [1:1,000], and/or Alexa Fluor 594 anti–mouse/rabbit IgG [1:1,000]) for 30 min at room temperature in the blocking solution. For surface expression assays, the samples were incubated with anti-HA antibodies (3F10) for 1 h at room temperature or overnight at 4°C and then incubated with secondary antibodies (Alexa Fluor 594 anti–rat IgG [1:1000]) for 30 min at room temperature in the blocking solution. Cells were imaged using an epifluorescence microscope (TE2000U; Nikon) equipped with 10× S Fluor 0.5 NA (Nikon), 40× S Fluor 1.30 NA (Nikon), and 60× Plan-Apochromat 1.40 NA (Nikon) objective lenses and a cooled charge-coupled device camera (ORCA-ER; Hamamatsu Photonics) controlled by a computer (Macintosh; Apple) running Open Laboratory software (PerkinElmer).

To measure the surface expression levels, the single regions were defined to encompass the entire cell body and calculate the ratio of HA to YFP fluorescence. To measure the relocalization of the GFP–2× GRAM in neurons, fluorescent images were obtained and analyzed using Open Laboratory software. For each cell, five circular regions of interest (ROIs) were defined around the periphery of the cell nucleus, and a single ROI was defined to encompass the entire cell body. The mean fluorescence intensity calculated from the ROIs around the nucleus was divided by the mean intensity of the entire cell body to generate a perinuclear ratio value (P). The percent change in P in stimulated cells was calculated using the following equation: %P = (P_S − P_R)/P_R, where P_S is the ratio after stimulation and P_R is the ratio in resting cells. Stimulation of cells with glutamate resulted in an increase in puncta that could be detected using the anti-PtdIns(3,5)P₂ antibodies. To quantify the number of puncta, the raw fluorescence images were intensity sliced to generate binary images that included only pixels that were part of a puncta. The area covered by the puncta was divided by the total area of the cell to generate a value for the percentage of pixels in a cell that formed part of a puncta. Colocalization of Flag-Myc-Ca_V1.2 and LAMP-GFP was evaluated by using line profiles drawn along neuronal dendrites for each type of staining using Igor Pro software (WaveMetrics). The coefficient of correlation between the Flag-Myc-Ca_V1.2 and LAMP-GFP line profiles was calculated according to the following equation:

Cells were washed with PBS and then lysed in extraction buffer (20 mM Tris-HCl, pH 7.5, 150 mM NaCl, 5 mM EGTA, 5 µM NaF, 1 µM Na₃VO₄, 0.5% Triton X-100, and 1 mM dithiothreitol) containing a protease inhibitor cocktail tablet (Roche). The cell lysates were analyzed by SDS-PAGE, transferred to polyvinylidene fluoride, probed with primary antibodies, and detected with HRP-conjugated secondary antibodies and luminol reagent (SuperSignal West Dura Extended Duration Substrate; Thermo Fisher Scientific).

Transfected HEK 293T cells were washed with PBS and then lysed in extraction buffer containing protease inhibitors (see Immunoblot analysis). The cell lysates were centrifuged at 13,200 rpm for 10 min, and the resulting supernatant was subjected to immunoprecipitation with antibodies to 10 µg/ml PIKfyve (6C7), 2.5 µg/ml Myc (4A6), or anti-Flag M2 agarose beads (Sigma-Aldrich) for 3 h at 4°C. The immunoprecipitates were subjected to immunoblot analysis with antibodies to Ca_V1.2, Myc, Flag, and HA.

The protein complex associated with the C terminus of Ca_V1.2 was purified and analyzed using mass spectrometry three independent times. One immunoprecipitated sample was analyzed in duplicate to generate a total of five mass spectrometry data samples. For each experiment, three 10-cm plates of Neuro2A cells were transfected with GST-YFP or with GST–Ca_V1.2 CT using Lipofectamine 2000 according to the manufacturer's instructions (24 µg DNA/plate in 1.5 ml of Opti-MEM-I and 60 µl of Lipofectamine in 1.5 ml of Opti-MEM-I). 2 d after transfection, the cells were lysed in lysis buffer (50 mM Tris-HCl, 1% Triton X-100, 150 mM NaCl, and 10 mM EDTA) for 30 min at 4°C with shaking. The lysis buffer was supplemented with an inhibitor cocktail tablet (Roche) and calpain I and II inhibitors. The lysates were centrifuged for 20 min at 14,000 rpm and incubated with 100 µl of glutathione Sepharose beads (GE Healthcare) for 2 h at 4°C. The beads were washed five times in lysis buffer and treated with trypsin before analysis. The peptides were separated using capillary electrophoresis and analyzed by tandem mass spectrometry in the Stanford University Mass Spectrometry facility. Finally, all of the peptide sequences were analyzed using MASCOT software (Matrix Science).

Cortical neurons were loaded with 1 µM Fura-2 for 30 min at 37°C in Tyrode's solution (129 mM NaCl, 5 mM KCl, 2 mM CaCl₂, 1 mM MgCl₂, 30 mM glucose, 0.1% BSA, and 25 mM Hepes, pH 7.4), washed with Tyrode's solution, and placed in a perfusion chamber on the stage of an inverted fluorescence microscope (TE2000U; Nikon). The cells were treated with 50 µM glutamate in Tyrode's solution without Mg (129 mM NaCl, 5 mM KCl, 3 mM CaCl₂, 30 mM glucose, 0.1% BSA, and 25 mM Hepe, pH 7.4) for 10 min before imaging in normal Tyrode's solution and then stimulated with high KCl Tyrode's solution (67 mM NaCl, 67 mM KCl, 2 mM CaCl₂, 1 mM MgCl₂, 30 mM glucose, 0.1% BSA, and 25 mM Hepes, pH 7.4). Imaging was performed at room temperature on an epifluorescence microscope (Eclipse TE2000U; Nikon) equipped with an excitation wheel and an automated stage. Open Laboratory software was used to collect and quantify time-lapse excitation ratio images. Fluorescence images were analyzed using Igor Pro software.

Cortical neurons were stimulated with 50 µM glutamate for the indicated times and were then incubated for 10 h in the absence of glutamate. Neurons were then fixed with 4% paraformaldehyde in PBS for 10 min at room temperature. The coverslips were incubated in a solution (10 µM Hoechst 33258 and 0.4% Triton X-100 in PBS) for 10 min to stain the nuclei. Fluorescence images were obtained by epifluorescence microscopy, and pyknotic nuclei were counted. For TUNEL assay, neurons were fixed with 4% paraformaldehyde in PBS for 15 min at room temperature and incubated with 0.1% Triton X-100 in PBS for 10 min. The samples were then incubated with TUNEL reaction solution (15 U terminal deoxynucleotidyl transferase, 40 µM biotin–deoxy-UTP, 0.1 M potassium cacodylate, 2 mM CoCl₂, and 0.2 mM DTT) for 1 h at 37°C and washed three times in PBS at room temperature before incubation with DTAF-streptavidin (1:800) for 45 min at 37°C. The samples were then incubated in a solution (10 µM Hoechst 33258 in PBS) for 10 min to stain the nuclei.

Fig. S1 shows that glutamate stimulation promotes neuronal cell death. Fig. S2 shows that glutamate stimulation promotes Ca_V1.2 relocalization and degradation. Fig. S3 shows localization of LAMP, PIKfyve, and the GRAM domain. Fig. S4 shows that knockdown of PIKfyve by PIKfyve shRNAs promotes vacuole formation. Fig. S5 shows that PIKfyve is important for specific channel internalization and degradation. Table S1 lists Ca_V1.2-interacting proteins.

We thank Drs. Petter Cullen and Lily Yeh Jan for the PIKfyve and K_V1.2 constructs and Drs. Deborah Solymar, Elizabeth Nigh, Ben Barres, and Thomas Clandinin for critical reading of the manuscript.

This work was supported by grants and awards from the National Institutes of Health (RO1 NS048564), McKnight Scholar, Searle Scholar (to R.E. Dolmetsch), American Heart Association, Japan Society for the Promotion of Science, Mochida Memorial Foundation for Medical and Pharmaceutical Research (to F. Tsuruta), National Institutes of Health Medical Scientist Training Program (to E.M. Green), and Foundation Philippe (to M. Rousset).