Acid-sensing ion channel 3 (ASIC3) cell surface expression is modulated by PSD-95 within lipid rafts

Abstract

Acid-sensing ion channel 3 (ASIC3) is a H⁺-gated cation channel primarily found in sensory neurons, where it may function as a pH sensor in response to metabolic disturbances or painful conditions. We previously found that ASIC3 interacts with the postsynaptic density protein PSD-95 through its COOH terminus, which leads to a decrease in ASIC3 cell surface expression and H⁺-gated current. PSD-95 has been implicated in recruiting proteins to lipid rafts, which are membrane microdomains rich in cholesterol and sphingolipids that organize receptor/signaling complexes. We found ASIC3 and PSD-95 coimmunoprecipitated within detergent-resistant membrane fractions. When cells were exposed to methyl-β-cyclodextrin to deplete membrane cholesterol and disrupt lipid rafts, PSD-95 localization to lipid raft fractions was abolished and no longer inhibited ASIC3 current. Likewise, mutation of two cysteine residues in PSD-95 that undergo palmitoylation (a lipid modification that targets PSD-95 to lipid rafts) prevented its inhibition of ASIC3 current and cell surface expression. In addition, we found that cell surface ASIC3 is enriched in the lipid raft fraction. These data suggest that PSD-95 and ASIC3 interact within lipid rafts and that this raft interaction is required for PSD-95 to modulate ASIC3.

the acid-sensing ion channel 3 (ASIC3) is a member of the degenerin/epithelial sodium channel (DEG/ENaC) family of ion channels that is expressed primarily in mammalian sensory neurons (49). It localizes to nerve terminals, where it may function as a transducer of various sensory stimuli (35). Modest drops in pH activate ASIC3, and several lines of evidence support its role as a metabolic or pain sensor during acidic conditions: 1) ASIC3 is highly expressed in sensory nerves of cardiac (2, 41) and skeletal muscle (31, 32), tissues that have high metabolic activity and are thus susceptible to ischemia leading to production and accumulation of lactic acid; 2) lactate anion potentiates the pH sensitivity of ASIC3, providing a molecular explanation for why lactate is a more potent activator of sensory neurons compared with other acids (23); 3) pharmacological blockade of ASICs attenuate acid-induced pain in human skin (24, 48); and 4) mice lacking ASIC3 demonstrate diminished pain hypersensitivity in models of chronic hyperalgesia (25, 40). In addition to its function as a pH sensor, ASIC3 has been implicated in playing a role in mechanosensation; ASIC3 null mice display altered responses to mechanical stimuli in specific populations of mechanosensory neurons (7, 25, 33, 35).

To begin to understand the regulation of ASIC3, it is essential to understand the anatomic and molecular environment in which the channel functions. Like all DEG/ENaC proteins, ASIC3 subunits consist of two membrane-spanning domains, a large extracellular loop with the NH₂ and COOH termini inside the cell where they can interact with cytosolic proteins. It has been recognized that the COOH terminus of ASIC3 shares homology to type 1 PDZ (PSD-95, Drosophila discs-large protein, zonula occludens protein-1)-binding motifs and allows for binding to several PDZ domain-containing proteins. Investigators in our laboratory reported that two of these proteins markedly altered ASIC3 function: Lin-7b increased ASIC3 H⁺-gated current, whereas PSD-95 (postsynaptic density protein-95) inhibited ASIC3 currents. The mechanism appears to involve trafficking of ASIC3: Lin-7b increased and PSD-95 decreased ASIC3 protein expression at the cell surface (20). CIPP (channel-interacting PDZ domain protein), PIST (PDZ protein interacting specifically with TC10), MAGI (membrane-associated guanylate kinase with inverted orientation), and NHERF (Na⁺/H⁺ exchanger regulatory factor-1) are other PDZ domain-containing proteins that have been shown to interact with ASIC3 and modulate its function (1, 11, 20).

The interaction of ASIC3 with PSD-95 is particularly intriguing. Like ASIC3, PSD-95 has been implicated in pain pathways. Knockdown of PSD-95 in rat spinal cord attenuated, and targeted disruption of PSD-95 in mice abolished, hyperalgesia to mechanical and thermal stimuli following nerve injury (15, 43, 44). Previously, work in our laboratory demonstrated that both PSD-95 and ASIC3 are present in dorsal root ganglia and coimmunoprecipitate together in rat spinal cord (20).

PSD-95 is essential for normal synaptic plasticity at postsynaptic sites, where it integrates signaling by localizing and clustering proteins (26). PSD-95 forms multimers, and each subunit contains three PDZ domains. This allows for binding to the COOH termini of multiple proteins, thus forming large submembrane scaffolds to link multiple signaling partners. PSD-95 can also directly associate with membranes via palmitoyl groups attached to specific cysteine residues at its NH₂ terminus (8, 47). Furthermore, PSD-95 localizes to lipid rafts (cholesterol- and sphingolipid-rich microdomains within cytosolic and surface membranes), and evidence suggests that some functions of PSD-95 might occur within the context of these specialized lipid domains (3, 17, 28, 34, 42, 52). Finally, recent evidence suggests that some peripheral pain signaling mechanisms might be organized within lipid rafts (12). To explore the mechanisms of interaction between ASIC3 and PSD-95, we tested the hypothesis that ASIC3 localizes to lipid rafts and that PSD-95 modulates ASIC3 function within the context of these lipid microdomains.

MATERIALS AND METHODS

DNA constructs.

Rat ASIC3 in pMT3 vector was cloned as described previously (20). HA-ASIC3 was generated by insertion of two hemagglutinin (HA) epitopes (YPYDVPDYA-G-YPYDVPDYA) at the NH₂ terminus of ASIC3. Introduction of this epitope did not alter current properties and did not disrupt PSD-95 inhibition of ASIC3 current (data not shown). Myc-tagged rat PSD-95 in cytomegalovirus (CMV) neo vector was a gift from Johannes Hell, and the Myc epitope was deleted using the QuikChange site-directed mutagenesis kit (Stratagene, La Jolla, CA). Cysteines at the third and fifth residues were mutated to serines in PSD-95 (PSD-95_C3,5S) by site-directed mutagenesis. DsRed (Express-C1) was purchased from Clontech, and green fluorescent protein (GFP; pGreen Lantern) was obtained from Life Technologies.

Cell culture and transfection.

Chinese hamster ovarian (CHO) cells were cultured at 37°C, 5% CO₂ in F12 nutrient medium (GIBCO, Carlsbad, CA) supplemented with 10% fetal bovine serum and 1% penicillin-streptomycin. For electrophysiological studies, cells plated at ∼10% confluence were transfected with cDNAs by using lipid transfection reagent TransFast (Promega, Madison, WI) in 35-mm dishes according to the manufacturer's recommendations. ASIC3 cDNA (0.18 μg/1.5 ml) was cotransfected with PSD-95, PSD-95_C3,5S, or DsRed as control (1.82 μg/1.5 ml) at a 1:10 ratio. All groups were cotransfected with GFP (0.33 μg/1.5 ml) to facilitate detection of expressing cells by epifluorescence. Cells used for biochemistry were transfected by electroporation (15 μg of cDNA per 10⁶ cells) using Gene Pulser II (Bio-Rad, Hercules, CA) or Lipofectamine 2000 (GIBCO) with 7.5 μg of HA-ASIC3 cDNA and 7.5 μg of PSD-95, PSD-95_C3,5S, or DsRed cDNA. Our transfection protocol for electrophysiological studies has low efficiency, but transfected cells are easy to patch clamp and have large currents, whereas our transfection protocol for biochemistry is highly efficient to generate a large quantity of protein.

ASIC3 transgenic mice.

Transgenic mice expressing ASIC3 with two HA epitopes inserted at the NH₂ terminus were generated using the same strategy and plasmid vector (containing the synapsin I promotor) as previously described to generate ASIC1a transgenic mice (51). Mouse brain protein lysate was prepared for immunoprecipitation, immunoblotting, and isolation of lipid rafts by homogenizing tissue in lysis buffer [1% Triton X-100, 50 mM Tris·HCl, pH 7.4, 150 mM NaCl, and one EDTA-Free Complete Mini Tab (Roche)] using a Polytron PT-3100 homogenizer at 5,000 rpm at 4°C. The homogenate was incubated on ice for 20 min and then centrifuged at 4°C for 15 min at 50,000 rpm (Beckmann TLS-55 rotor).

Antibodies.

The following primary antibodies were used: anti-HA high-affinity antibody (clone 3F10), horseradish peroxidase-conjugated anti-HA (anti-HA-HRP) and anti-HA affinity matrix (Roche Applied Biosciences, Indianapolis, IN); anti-transferrin receptor (Zymed Laboratories, South San Francisco, CA); and anti-caveolin-1 and anti-PSD-95 (Upstate, Charlottesville, VA). Secondary antibodies used include anti-mouse IgG (Amersham Biosciences, Little Chalfont, UK), goat anti-mouse (Upstate), or anti-rabbit IgG (Santa Cruz Biotechnology, Santa Cruz, CA).

Immunoprecipitation.

Cells were washed at 4°C three times with PBS+ (1 mM PBS, pH 7.4, 1 mM MgCl₂, and 1 mM CaCl₂) and then incubated for 5 min on a rocker at 4°C in 1 ml of lysis buffer [1% TX-100, 50 mM Tris·HCl, pH 7.5, 150 mM NaCl, and 1 Complete Mini protease inhibitor tablet/25–50 ml (Roche Applied Biosciences, Indianapolis, IN)]. Cells were scraped into prechilled 15-ml tubes and drill homogenized. The lysate was then centrifuged at 13,000 g (4°C) for 20 min, and the supernatant was precleared with 50 μl of washed protein A-Sepharose beads (Sigma, St. Louis, MO). Samples were tumbled with either anti-HA affinity matrix (50 μl) or 2 μg of PSD-95 antibody with 50 μl of protein A-Sepharose beads. The immunoprecipitates were collected by centrifugation, washed five times with 1% TX-100 lysis buffer, and resolved by SDS-PAGE.

Immunoblotting.

Samples were separated by SDS-PAGE on a 7.5 or 12% (wt/vol) gel and transferred to a nitrocellulose membrane (PROTERAN; Schleicher & Schuell). The membrane was blocked in 5% bovine serum albumin in TBST [0.05% (wt/vol) Tween 20 in 10 mM Tris and 100 mM NaCl, pH 7.5] and then incubated with the primary antibody (anti-PSD-95, 1:200; anti-HA-HRP, 1:750; anti-caveolin-1, 1:200; or anti-transferrin, 1:1,000). Membranes, except for blots using anti-HA-HRP, were subsequently incubated with HRP-conjugated anti-mouse (1:2,150) or anti-rabbit (1:1,000) Ig and developed using the Visualizer enhanced chemiluminescence (ECL) system (Upstate). In some cases, membranes were stripped after being washed in TBST using Re-Blot (Chemicon, Temecula, CA) per the manufacturer's recommendations. Stripped membranes were washed, blocked, and reblotted as described above.

Isolation of lipid rafts.

Forty-eight hours after transfection, 100-mm dishes of CHO cells were washed three times with PBS+, collected in 1 ml of PBS+, and centrifuged at 2,000 g for 5 min at 4°C. The pellet was resuspended in 100 μl of MES-buffered saline (24 mM MES, pH 6.5, and 0.15 M NaCl) with protease inhibitors (Complete Mini tablets) plus 1% Triton X-100. The solution was then homogenized by 15 strokes of a prechilled 2-ml tight-fitting Dounce homogenizer and then titrated to 40% sucrose. This solution was placed in a 1.5-ml Beckman thick-walled ultracentrifuge tube (Beckman Coulter, Fullerton, CA) with two equal layers of 30 and 5% sucrose layered on top and centrifuged at 54,000 rpm for 24 h at 4°C in a Beckman TLS-55 rotor. Eight or nine 140-μl fractions of the solution were collected; equal volumes of each fraction from the sucrose gradient were either immunoprecipitated or analyzed by SDS-PAGE on a 7.5 or 12% (wt/vol) gel and immunoblotted as described above.

Electrophysiology.

Whole cell patch-clamp recordings (at −70 mV) in CHO cells were performed at room temperature with an Axopatch-200B amplifier (Axon Instruments, Foster City, CA) and were acquired and analyzed with Pulse/Pulsefit 8.30 (HEKA Electronics, Lambrecht, Germany) and Igor Pro 4.04 (WaveMetrics, Lake Oswego, OR) software 48 h after transfection. Recordings were filtered at 5 kHz and sampled at 2 or 0.2 kHz. Series resistance was compensated by at least 50%. Capacitive currents were compensated for and recorded for normalization of peak current amplitudes (reported as current densities). Micropipettes (2–5 MΩ) were filled with internal solution containing (mM) 100 KCl, 10 EGTA, 40 HEPES, and 5 MgCl₂, pH 7.4 with KOH. External solution contained (mM) 120 NaCl, 5 KCl, 1 MgCl₂, 2 CaCl₂, 10 HEPES, and 10 MES; pH was adjusted with tetramethylammonium hydroxide, and osmolarity was adjusted with tetramethylammonium chloride. Extracellular acidic pH solutions were exchanged within 20 ms from a baseline pH 7.4 using a computer-driven solenoid valve system (2). Data are means ± SE. Single-exponential equations were used to characterize kinetics of desensitization. Statistical significance was assessed using unpaired two-tailed Student's t-test.

Surface biotinylation and NeutrAvidin pull down.

Forty-eight hours after transfection, cells were incubated at 4°C for 10 min, followed by three washes at 4°C with PBS+ and then incubation with 0.5 mg/ml sulfo-N-hydroxysuccinimide-biotin (Pierce) at 4°C for 30 min. Unbound biotin was quenched with 100 mM glycine in PBS+ for 20 min at 4°C. The cells were lysed at 4°C using 1% Nonidet P-40, 63 mM EDTA, 58.3 mM Tris·HCl, pH 8, and 290 mM sodium deoxycholate plus protease inhibitors (Sigma) and then centrifuged at 16,100 g for 10 min to remove any insoluble material. Lysate (800 μg) was tumbled with 30 μl of NeutrAvidin-agarose beads (Pierce) for 24 h at 4°C. After multiple washings, biotinylated proteins were eluted using SDS sample buffer (4% SDS, 100 mM DTT, 20% glycerol, and 100 mM Tris·HCl, pH 6.8), separated by SDS-PAGE, and then immunoblotted as described. Biotinylated and total lysate bands were quantitated by densitometry (ImageJ).

Cholesterol depletion.

Raft-like microdomains were chemically disrupted by depleting cholesterol with methyl-β-cyclodextrin (MβCD; Sigma). In all cases, cells were washed twice with PBS and then incubated with MβCD. For electrophysiological experiments, cells were incubated at 37°C for 2 h in 4 mM MβCD; in lipid raft studies, cells were incubated in 20 mM MβCD for 1 h before lysis for sucrose fractionation.

RESULTS

ASIC3 and PSD-95 localize and interact within lipid rafts.

We tested whether ASIC3 localizes to lipid rafts and whether coexpression with PSD-95 alters its localization (10, 52). Lipid rafts (as well as proteins associated within rafts) are characterized by 1) resistance to solubilization by nonionic detergents at cold temperature and 2) buoyancy in density gradients due to their high lipid content (5). Accordingly, we lysed CHO cells expressing ASIC3 with Triton X-100 at 4°C and then separated fractions by density using a sucrose gradient. We identified fractions that contained lipid rafts by immunoblotting for the endogenous raft-specific protein caveolin-1 (50) and found it exclusively in buoyant, low-density fractions (Fig. 1A, fractions 4 and 5). To confirm successful solubilization of non-raft proteins, we identified the endogenous transferrin receptor [a protein that does not associate with rafts (37)] only in the dense fractions (fractions 8 and 9). Whereas most of the ASIC3 protein was found in the dense fractions, a portion of ASIC3 localized to raft fractions. We then coexpressed PSD-95 to test whether it alters ASIC3 distribution. Consistent with previous studies (17, 34, 42), PSD-95 was found in both raft and dense fractions (Fig. 1B). However, PSD-95 did not produce an obvious change in the distribution of ASIC3 in raft compared with dense fractions.

Fig. 1.

Acid-sensing ion channel 3 (ASIC3) and postsynaptic density protein-95 (PSD-95) are found in lipid raft membrane fractions. Chinese hamster ovary (CHO) cells were transfected with hemagglutinin (HA)-ASIC3 alone (A; n = 6) or with PSD-95 (B; n = 5) at a 1:1 cDNA ratio and solubilized with 1% Triton-X at 4°C. Lysates were separated by sucrose gradient centrifugation, and fractions of equal volumes (fractions 1-9, top to bottom) were resolved by SDS-PAGE and blotted for HA-ASIC3, PSD-95, and the following endogenous proteins: 1) caveolin-1, which is specifically isolated to lipid raft fractions, and 2) transferrin receptor (TfR), which does not associate in lipid rafts. Representative immunoblots reveal ASIC3 and PSD-95 localize to both low-density, detergent-resistant (raft) fractions and high-density, soluble fractions (dense).

Since ASIC3 and PSD-95 were found in both raft and non-raft fractions, we asked whether they interact in either or both of these membrane fractions. We first confirmed that ASIC3 and PSD-95 interact in unfractionated CHO cell lysate. Figure 2 A demonstrates that ASIC3 coimmunoprecipitated with PSD-95 and that coexpression did not alter the total expression of either protein. In addition, neither protein was detected in untransfected cells. We then tested the interaction in lysate collected from raft (pooled fractions 4 and 5) and dense (pooled fractions 8 and 9) fractions after lipid raft preparation. Immunoprecipitated PSD-95 coprecipitated ASIC3 in both raft and dense fractions (Fig. 2B), suggesting the two proteins have the capacity to interact in both lipid raft and non-raft membranes.

Fig. 2.

ASIC3 and PSD-95 interact within lipid rafts. A: cells were transfected with HA-ASIC3 and/or PSD-95 as indicated. Lysates were immunoprecipitated (IP) with anti-PSD-95 antibody, resolved by SDS-PAGE, and then immunoblotted for HA-ASIC3. Total lysates from the same cells were also run on SDS-PAGE and immunoblotted for HA-ASIC3 or PSD-95 to confirm expression (n = 5). We also successfully coimmunoprecipitated PSD-95 with ASIC3 by immunoprecipitating with anti-HA affinity matrix and immunoblotting for PSD-95 (data not shown). B: cells were transfected, solubilized, and fractionated as in Fig. 1B. Buoyant fractions (pooled fractions 4 and 5) and dense fractions (pooled fractions 8 and 9) were separately immunoprecipitated with anti-PSD-95 antibody. Immunoprecipitates were resolved by SDS-PAGE and then immunoblotted for HA-ASIC3, with results indicating that ASIC3 and PSD-95 interacted in both raft and non-raft fractions.

Next, we tested whether ASIC3 is localized to lipid rafts within native neurons. To do this, we used a transgenic mouse model that expresses an epitope-tagged ASIC3 driven by a pan-neuronal synapsin I promoter (51). We first tested whether the transgenic protein interacts with native PSD-95 in brain. Figure 3 A shows that HA-ASIC3 coimmunoprecipitated with PSD-95 brain lysate from two transgenic mice but not in that from a wild-type littermate. We then did a lipid raft preparation of brain lysate from transgenic mice and found that HA-ASIC3 localized to buoyant as well as dense fractions (Fig. 3B).

Fig. 3.

Transgenic expressed ASIC3 interacts with native PSD-95 in mouse brain and localizes to buoyant fractions. A: brain lysates from individual ASIC3 transgenic or wild-type mice were immunoprecipitated with anti-PSD-95 antibody, resolved by SDS-PAGE, and then immunoblotted for HA-ASIC3. Total lysates from the same cells were also run on SDS-PAGE and immunoblotted for HA-ASIC3. B: individual ASIC3 transgenic mouse brain lysate was separated by sucrose gradient centrifugation, and fractions of equal volumes (fractions 1-9, top to bottom) were resolved by SDS-PAGE and blotted for HA-ASIC3. Similar results were obtained in 2 transgenic mice, whereas HA-ASIC3 was not detected in wild-type mouse brain lysate.

Disruption of PSD-95 localization to lipid raft fractions by cholesterol depletion abolishes its inhibition of ASIC3.

We previously found that PSD-95 inhibits ASIC3 pH-activated current (20). In the present study, we found that the two proteins can interact in raft-associated membrane fractions. To test whether lipid raft localization is required for PSD-95 to modulate ASIC3 function, cells were treated with the cholesterol-sequestering drug MβCD to disrupt rafts (22). We first determined whether ASIC3 and PSD-95 lipid raft localization is altered by cholesterol depletion. Figure 4 A shows that MβCD treatment removed PSD-95 from the detergent-resistant fractions. Somewhat surprisingly, MβCD had no effect on the distribution of ASIC3 in lipid raft fractions when coexpressed with PSD-95 (Fig. 4A) or alone (Fig. 4B). Likewise, endogenous caveolin-1 still remained in raft fractions, although more of the protein was detected in the more dense fractions (Fig. 4B). These results are consistent with the observation that raft proteins display variable detergent solubility after cholesterol depletion (16). Given these findings, it was not surprising that after MβCD treatment, we found that ASIC3 and PSD-95 coimmunoprecipitated in the soluble fractions but that no interaction was observed within the raft fractions (Fig. 4C).

Fig. 4.

Cholesterol depletion disrupts PSD-95 localization to lipid rafts and its inhibition of ASIC3 current. CHO cells coexpressing HA-ASIC3 with PSD-95 at a 1:1 cDNA ratio (A; n = 2) or HA-ASIC3 transfected alone (B; n = 2) underwent solubilization and fractionation (lipid raft preparation) after 1 h of preincubation with 20 mM methyl-β-cyclodextrin (MβCD). Lysate fractions were blotted as indicated. C: raft fractions (fractions 4 and 5) and dense fractions (fractions 8 and 9) from cells transfected, solubilized, and fractionated as in A were separately immunoprecipitated with anti-PSD-95 antibody and then immunoblotted for HA-ASIC3 (n = 2). D: characteristic currents recorded by whole cell patch clamp evoked by rapidly changing the pH of the perfusion solution from 7.4 to 6 (indicated by bars above currents) in CHO cells transfected with ASIC3 alone or with PSD-95 at a 1:10 cDNA ratio (expression vector encoding dsRed was used as control to keep cDNA ratio constant). A separate group of cells expressing ASIC3 with and without PSD-95 (current not shown) were studied after treatment with 4 mM MβCD for 2 h. E: mean current density evoked by pH 5 for groups of cells in D (n = 9–14). *P < 0.02 compared with ASIC3 alone. Other statistical analysis as indicated.

Because cholesterol depletion abolished the interaction of PSD-95 with ASIC3 in lipid raft fractions, we then tested whether MβCD alters PSD-95 inhibition of ASIC3 current. Figure 4, D and E, shows that application of a rapid pH drop to cells expressing ASIC3 generated large inward, rapidly desensitizing currents. However, when PSD-95 was coexpressed with ASIC3, current was significantly inhibited. Although PSD-95 reduced ASIC3 current amplitude, it did not alter gating properties such as the pH sensitivity, as measured by pH dose-response, or the kinetics of desensitization (data not shown). This finding is consistent with our previous data and, together with our finding that PSD-95 decreased cell surface expression of ASIC3 (20), suggests that PSD-95 inhibits ASIC3 current by reducing channel expression at the cell surface. We then pretreated cells with MβCD to disrupt lipid rafts before recording. MβCD significantly diminished PSD-95 inhibition of ASIC3 current (Fig. 4, D and E). However, MβCD had no effect on ASIC3 current when ASIC3 was expressed alone. Interestingly, the rates of desensitization were slightly slower in both groups of cells after treatment with MβCD (data not shown), perhaps indicating a direct effect of cholesterol membrane content on channel gating (46). Together, these data suggest that cholesterol depletion with MβCD disrupts the capacity of PSD-95 to associate with lipid rafts and disrupts its modulation of ASIC3 channels.

Palmitoylation sites on PSD-95 are required for its modulation of ASIC3.

Cholesterol depletion may have pleiotropic effects on membrane structure or affect other endogenous proteins that could modulate ASIC3. To more specifically test whether lipid raft localization is necessary for PSD-95 to modulate ASIC3, we mutated PSD-95 to disrupt its association with rafts. Palmitoylation of cysteines at the NH₂ terminus of PSD-95 facilitates its interaction with lipid bilayers and has been shown to be required for the localization of PSD-95 to lipid rafts in tsA201 cells (52). By mutating these cysteines to serines (PSD-95_C3,5S) to prevent palmitoylation (47), we found PSD-95 localization to raft fractions was markedly diminished (Fig. 5A). As with wild-type PSD-95, coexpression of mutant PSD-95_C3,5S with ASIC3 did not alter the distribution of ASIC3 to raft fractions. Figure 5B shows that PSD-95_C3,5S coimmunoprecipitated ASIC3 as predicted since the binding domains are intact, but they did so only in the dense fractions.

Fig. 5.

Mutation of palmitoylation sites on PSD-95 disrupts its localization to lipid rafts, as well as its inhibition of ASIC3 current and cell surface expression. A: CHO cells coexpressing HA-ASIC3 and mutant PSD-95 (PSD-95_C3,5S: cysteines at residues 3 and 5 mutated to serines) at a 1:1 cDNA ratio underwent lipid raft preparation, and lysate fractions were blotted as indicated (n = 4). B: raft fractions (fractions 4 and 5) and soluble heavy fractions (fractions 8 and 9) were separately immunoprecipitated with anti-PSD-95 antibody, resolved by SDS-PAGE, and then immunoblotted for HA-ASIC3. C: typical currents recorded in response to pH 6 applications to CHO cells transfected with full-length ASIC3 alone or coexpressed with wild-type PSD-95 or PSD-95_C3,5S at 1:10 cDNA ratios. D: mean current density of currents evoked by pH 5 for groups of cells in C (n = 8–21). *P < 0.01 compared with ASIC3 alone. **P < 0.01 compared with ASIC3 + wild-type PSD-95. E: cells expressing HA-ASIC3 and dsRed (control) or either PSD-95 or PSD-95_C3,5S were exposed to biotin to bind surface protein. Cell lysate (15 μg) was blotted for ASIC3 to analyze total protein, and cell surface ASIC3 was pulled down with NeutrAvidin beads before blotting. Compared with ASIC3 expressed alone, coexpression of PSD-95 caused a >3.5-fold reduction in the surface-to-total ratio of ASIC3 (n = 5; P < 0.04).

Since PSD-95_C3,5S did not interact with ASIC3 in rafts, we predicted it would not modulate ASIC3 current. In fact, mutation of PSD-95 palmitoylation sites completely abolished the ability of PSD-95 to inhibit ASIC3 current (Fig. 5, C and D). Similar results were previously seen when we disrupted binding between the two proteins: deletion of the four COOH-terminal amino acids of ASIC3 to remove the PDZ-binding domain (ASIC3_Δ4) also abolished PSD-95 inhibition of ASIC3 current (20). To confirm the mechanism by which mutation of PSD-95 palmitoylation sites abolished its capacity to inhibit ASIC3 current, we labeled cell surface ASIC3 with biotin and precipitated it with NeutrAvidin. Figure 5E shows that coexpressed wild-type PSD-95 decreased ASIC3 cell surface expression, whereas PSD-95_C3,5S did not. These data mirror our results seen with MβCD treatment: mutation of PSD-95 palmitoylation sites disrupted its interaction with ASIC3 in lipid rafts, which in turn prevented its modulation of ASIC3 current.

Cell surface ASIC3 is enriched in lipid rafts.

Although PSD-95 and ASIC3 interacted in both lipid raft and non-raft membrane fractions, the major quantity of both proteins localized to the non-raft fractions. PSD-95 inhibited ASIC3 current by ∼90%, yet by disrupting the interaction of PSD-95 with ASIC3 only in the raft fraction, ASIC3 current was completely restored. This implies that even though only a small portion of ASIC3 protein was expressed in the raft fraction, this pool of channels generated most of the recorded current. To further test this hypothesis, we biotinylated cell surface ASIC3 and then ran cell lysate through a lipid raft preparation. A portion of each sucrose gradient fraction was blotted with ASIC3 to determine the amount of total ASIC3 in raft compared with dense fractions (Fig. 6, top), and the rest of each fraction was precipitated with NeutrAvidin beads before blotting to assess cell surface ASIC3 (Fig. 6, bottom). For this experiment, we discarded the bottom “pellet” of the sucrose gradient, which was lane 9 in previous experiments. Total ASIC3 protein was relatively equally distributed between the raft fraction (fraction 5) and the dense fraction (fraction 8). However, most of the cell surface ASIC3 was in the raft fraction.

Fig. 6.

Cell surface ASIC3 is enriched in raft membrane fractions. CHO cells coexpressing HA-ASIC3 were biotinylated and then underwent lipid raft preparation. Next, 50 μl of each fraction (1-8) were resolved by SDS-PAGE and blotted for HA to measure total cell ASIC3 as well as caveolin-1 to define raft fractions. The remaining 100 μl of each fraction were pulled down with NeutrAvidin and then blotted to detect cell surface ASIC3 (n = 3).

DISCUSSION

We previously demonstrated that PSD-95 binds to ASIC3 and reduces its acid-evoked current by decreasing the number of channels at the cell surface. In the present study, we found a portion of expressed PSD-95 and ASIC3 localized and coimmunoprecipitated in lipid raft membrane fractions. Interestingly, modulation of ASIC3 by PSD-95 occurred only within lipid rafts; preventing PSD-95 localization in lipid raft fractions, either by disrupting rafts via cholesterol depletion or by mutating the palmitoylation sites of PSD-95, prevented its reduction of ASIC3 current and cell surface expression. In addition, our results suggest that lipid rafts are important for constitutive cell surface expression of ASIC3.

Lipid rafts are specialized membrane platforms that coordinate trafficking and signaling, and increasing evidence suggests that they can sequester ion channels and regulate their function (29). This is the first demonstration of an ASIC channel localizing to lipid rafts, although related DEG/ENaC channels have been reported to do so (18, 36). Cytosolic proteins, such as PSD-95, can also associate with lipid rafts, and the mechanism generally involves modification with fatty acids (30). We found PSD-95 raft association in CHO cells was dependent on two NH₂-terminal cysteines that undergo palmitoylation, consistent with results in tsA201 cells (52). However, others have reported that mutation of these same palmitoylation sites was not sufficient to disrupt PSD-95 localization to raft fractions in COS-7 cells; additional NH₂-terminal regions were also necessary (34). The reason for these differences is unclear. PSD-95 recruits the K⁺ channel subunit Kv1.4 to lipid rafts (52); however, we found PSD-95 did not qualitatively affect the distribution of ASIC3 in lipid raft fractions.

Rather than regulating the localization of ASIC3 into or out of lipid rafts, we found that PSD-95 modulates the function of ASIC3 within rafts. Previously, we found that the PDZ-binding motif at the ASIC3 COOH terminus interacts with PSD-95. When coexpressed, PSD-95 reduces the amplitude of ASIC3 acid-evoked current by decreasing its cell surface expression. In the present study, we demonstrated that PSD-95 modulation of ASIC3 is dependent on the two proteins interacting within lipid rafts. We showed this by two means. First, we disrupted lipid rafts using the cholesterol-sequestering drug MβCD. This resulted in removal of PSD-95 from lipid raft fractions, and PSD-95 no longer inhibited ASIC3 current. In a second set of experiments, mutation of the palmitoylation sites at the NH₂ terminus of PSD-95 similarly prevented its association with lipid rafts, and this also abolished PSD-95 modulation of ASIC3. Interestingly, unlike other types of lipid modifications, palmitoylation is reversible and regulated (13), and the Bredt laboratory (14) has identified a mammalian palmitoyl transferase that specifically palmitoylates PSD-95. Our data suggest palmitoylation of PSD-95 might represent a means to dynamically regulate the function of ASIC3. It should be noted that in both experiments, PSD-95 and ASIC3 were readily able to interact by coimmunoprecipitation in the non-raft fractions, and yet under these conditions, no functional effect of PSD-95 on current was observed. Only when PSD-95 and ASIC3 were able to interact in lipid rafts did PSD-95 reduce ASIC3 current. Our results parallel the effect of neuronal cell adhesion molecule (NCAM) to inhibit the cell surface delivery of inward-rectifying K⁺ channels (Kir3); this effect was only observed when both proteins were localized to lipid rafts (9).

In addition, our results also suggest lipid rafts are important in the functional expression of ASIC3, independent of its modulation by PSD-95. Although only a small portion of ASIC3 protein was expressed in the raft fraction, our data suggest that this pool of channels generated most of the recorded current. This was further supported by our biotinylation data showing that most cell surface ASIC3 was in the lipid raft fraction. It is well understood that lipid rafts function as platforms for vesicular sorting and trafficking (22). This was first appreciated in epithelial cells, where incorporation of newly synthesized proteins into lipid rafts in the Golgi specifically targets their expression at the apical surface (5, 39). A recent study suggests that the ASIC-related ion channel ENaC traffics to the apical membranes of kidney cells via lipid rafts (19). Moreover, proteins that mediate cell membrane fusion, such as syntaxin, synaptosome-associated protein-25 (SNAP-25), and vesicle-associated membrane polypeptide (VAMP), are enriched in lipid rafts, and cholesterol depletion reduces exocytosis rates (6). Although we cannot rule out the possibility that ASIC3 moved laterally into rafts after trafficking to the cell surface, it is tempting to speculate that lipid rafts are necessary for the targeting of ASIC3 to the cell surface. We believe that some ASIC3 is incorporated into lipid rafts, perhaps at the level of the Golgi, and it is this pool of channels that is shuttled to the cell surface. However, when PSD-95 is bound the channel within rafts, these raft-ASIC3-PSD-95 complexes do not traffic to the cell membrane. Perhaps more intriguing, lipid rafts and PSD-95 are known to contribute to the polarized sorting of neuronal proteins to either axonal versus dendritic membranes of central neurons (26, 27). In the case of ASIC3, rafts and/or PSD-95 might facilitate trafficking to sensory nerve terminals in primary afferents neurons or postsynaptic sites in spinal cord dorsal horn neurons.

The localization of ASIC3 within lipid rafts might also facilitate other protein-protein signaling interactions. In addition to their role in trafficking, lipid rafts serve to concentrate or segregate specific proteins together into signaling complexes (38). For example, evidence suggest that lipid rafts coordinate signaling molecules including d,l-α-amino-3-hydroxy-5-methylisoxazole-4-propionic acid (AMPA) receptors, as well as kinases such as Ras/mitogen-activated protein kinase (MAPK) pathway and Ca²⁺/calmodulin-dependent protein kinase II (CaMKII) at postsynaptic sites (42), and disruption of lipid rafts in cultured hippocampal neurons leads to loss of synapses and increased AMPA receptor endocytosis (17). In addition, the binding of ASIC3 to PSD-95 might also contribute to the formation of a signaling complex. PSD-95 has three PDZ domains, each of which can presumably bind to the COOH-terminal PDZ-binding domains of other proteins. In addition, PSD-95 forms multimers with itself (21) and other related PDZ proteins (4), creating even larger scaffolds to cluster signaling proteins.

Our studies demonstrate that PSD-95 modulates the function of ASIC3, and it is dependent on their binding within specific membrane microdomains: lipid rafts. ASIC3 is primarily expressed in sensory neurons and is implicated in pain and mechanosensation. What might be the functional consequences of our findings on sensory pathways? Growing evidence supports a role for PSD-95 in the development and maintenance of chronic pain pathways in the spinal cord (45). We previously showed that ASIC3 and PSD-95 coimmunoprecipitated in spinal cord, implying a coordinated role in sensation. Sensory signaling mechanisms might also be coordinated within lipid rafts. Our data present a cellular mechanism by which PSD-95 and lipid rafts might regulate sensory signaling via ASIC channels.

GRANTS

This work was supported by National Institutes of Health Grant HL-06419 (to C. J. Benson) and Howard Hughes Medical Institute (to M. J. Welsh).