Internalization of the Kv1.4 Potassium Channel Is Suppressed by ...

THE JOURNAL OF BIOLOGICAL CHEMISTRY © 2000 by The American Society for Biochemistry and Molecular Biology, Inc.

Vol. 275, No. 2, Issue of January 14, pp. 1357–1364, 2000 Printed in U.S.A.

Internalization of the Kv1.4 Potassium Channel Is Suppressed by Clustering Interactions with PSD-95* (Received for publication, July 12, 1999, and in revised form, October 7, 1999)

Denis G. M. Jugloff‡§, Rajesh Khanna‡¶, Lyanne C. Schlichter‡¶, and Owen T. Jones‡§储 From the ‡Division of Cellular and Molecular Biology, Toronto Western Research Institute, University Health Network, Toronto, Ontario M5T 2S8, Canada, the §Department of Pharmacology, University of Toronto, Toronto, Ontario M5S 1A8, Canada, and the ¶Department of Physiology, University of Toronto, Toronto, Ontario M5S 1A8, Canada

The contribution of voltage-dependent ion channels to nerve function depends upon their cell-surface distributions. Nevertheless, the mechanisms underlying channel localization are poorly understood. Two phenomena appear particularly important: the clustering of channels by membrane-associated guanylate kinases (MAGUKs), such as PSD-95, and the regional stabilization of cell-surface proteins by differential suppression of endocytosis. Could these phenomena be related? To test this possibility we examined the effect of PSD-95 on the internalization rate of Kv1.4 Kⴙ channels in transfected HEK293 cells using cell-surface biotinylation assays. When expressed alone Kv1.4 was internalized with a half-life of 87 min, but, in the presence of PSD-95, Kv1.4 internalization was completely suppressed. Immunochemistry and electrophysiology showed PSD-95 had little effect on total or cell-surface levels of Kv1.4 or on current amplitude, activation, or inactivation kinetics. Clustering was necessary and sufficient to suppress Kv1.4 internalization since C35S-PSD-95, a mutant reported to bind but not cluster Kv1.4, (confirmed by imaging cells co-expressing a functional, GFP-varianttagged Kv1.4) restored and, surprisingly, enhanced the rate of Kv1.4 internalization (t1⁄2 ⴝ 16 min). These data argue PSD-95-mediated clustering suppresses Kv1.4 internalization and suggest a fundamentally new role for PSD-95, and perhaps other MAGUKs, orchestrating the stabilization of channels at the cell-surface.

In neurons voltage-dependent potassium channels are key determinants of the resting membrane potential and of membrane excitability, conditioning the frequency, and shape of action potentials (1, 2). Molecular genetic studies indicate that K⫹ channels are encoded by numerous genes whose products can be classified into distinct subfamilies of which the most extensively characterized is that of the Kv1 (Shaker) class (3). Such channels consist of four separate ␣-subunit polypeptides, which assemble as either homo- or heterotetramers (4 –7) to generate a pore-forming channel (4). However, in vivo, such tetramers are often complexed to auxiliary Kv␤ subunits, * This work was supported by awards from the Natural Sciences and Engineering Research Council of Canada (to O. T. J), the Medical Research Council of Canada (to L. C. S), a studentship from the Bloorview Epilepsy Program (to D. G. M. J.), and a Santalo scholarship from the University of Toronto (to R. K). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. 储 To whom correspondence should be addressed: Div. of Cellular and Molecular Biology, Toronto Western Research Institute, University Health Network, MC 11-434, 399 Bathurst St., Toronto, Ontario M5T 2S8, Canada. Tel.: 416-603-5039; Fax: 416-603-5745; E-mail: owen@ playfair.utoronto.ca. This paper is available on line at http://www.jbc.org

which modulate their functional properties (8, 9) and/or their surface expression (10). Thus, K⫹ channels exhibit a considerable potential for diversity, the basis for which has only begun to emerge. Important clues to the physiological roles of K⫹ channel diversity have come from observations that discrete K⫹ channel complexes have unique spatio-temporal patterns of expression in discrete brain regions and even within individual neurons (11–14). Thus, it seems that K⫹ channel diversity provides a powerful means for neurons to tailor voltage-dependent K⫹ efflux to the specific requirements of discrete functional compartments. The specific redistribution of axonal Kv1.1 and Kv1.2 channels during re-myelination (15), and the down-regulation of Kv1.4 in response to seizure activity (16) or hormones (17), underscores the pathophysiological importance of such phenomena. How neurons concentrate specific K⫹ channels into select compartments is unclear but it must involve their segregation and maintenance to prevent their free diffusion throughout the lipid bilayer (14). Important clues have come from the recent identification of a family of proteins termed MAGUKS1 (membrane-associated guanylate kinases), that are concentrated and thought to organize proteins, in specific compartments such as tight junctions in epithelia and synapses in neurons (14, 18). Among the most widely studied MAGUKs is the PSD-95 (postsynaptic density-95) subfamily (14) whose members (e.g. PSD-95 and SAP-97) induce the clustering of proteins including Kv1 channels. Clustering appears to require two sets of interactions. First, there is direct binding of PSD-95 via protein-protein interaction modules termed PDZ (14) domains to proteins, such as Kv1, that contain a specific peptide motif (T/S-X-V) at the distal ends of their carboxyl termini (18, 19). Second, there is self-association of PSD-95 (18, 20, 21). Significantly, the membrane protein binding and PSD-95 self-association events can be dissociated since PSD-95 mutants lacking PDZ domains can self-associate. Moreover, membrane protein binding is unaffected by the deletion of a pair of NH2-terminal cysteine residues required for PSD-95 multimerization (21, 22). The recent demonstration that PSD-95 mediates the clustering and immobilization of Kv1.4 in live cells (23) provides further compelling evidence that PSD-95-type interactions are key determinants of the spatial distribution of Kv1 channels. However, alternative or additional modes may exist through which PSD-95 specifies membrane protein segregation. One mode that may be especially pertinent to neurons is via the regulation of membrane protein endocytosis or internalization (which

1 The abbreviations used are: MAGUK, membrane-associated guanylate kinase; PSD-95, postsynaptic density-95; PCR, polymerase chain reaction; EYFP, extended yellow variant of green fluorescent protein; HEK293, human embryonic kidney cells; PBS-CM, phosphate-buffered saline; ECL, enhanced chemiluminescence; BAPTA, 1,2-bis(O-aminophenoxy)ethane-N,N,N⬘,N⬘-tetraacetic acid.

1357

1358

PSD-95 Suppresses Kv1.4 Endocytosis

describes surface channel depletion when both endocytosis and recycling occur (for definitions see Ref. 24)) (25). Differential rates of membrane protein endocytosis between cell regions would be a powerful means for defining their subcellular distribution and indirectly shaping the electrophysiology of the neuron (26, 27). Powerful support for this hypothesis has come from studies showing differences in endocytosis between axons and dendrites (28) and observations that endocytosis can determine, selectively, the expression of ion channels at the cell surface in both epithelia (29) and neurons (30, 31). Indeed there is now direct evidence that the activation of endocytosis is pivotal in synaptic plasticity and may cause the redistribution of proteins to sites of synaptogenesis (27). We have, therefore, hypothesized that an important role for PSD-95, in addition to segregating ion channels through clustering is to selectively stabilize (or destabilize) clustered channels at the cell surface by modulating their rates of endocytosis. To test our hypothesis, we have determined the rate of internalization of Kv1.4 K⫹ channels in the presence and absence of PSD-95. We now show that when Kv1.4 is expressed alone in HEK293 (human embryonic kidney) cells it is internalized from the cell surface with a t1⁄2 of 87 min. However, upon co-transfection with cDNAs encoding PSD-95, internalization of Kv1.4 is ablated. The ability of PSD-95 to impede Kv1.4 internalization is a direct function of its capacity for mediating Kv1.4 clustering since a PSD-95 mutant, C35S-PSD-95, which retains its ability to bind, but not cluster, allows Kv1.4 to be internalized rapidly. Surprisingly, the rate of internalization of Kv1.4 in the presence of C35S-PSD-95 is much faster (t1⁄2 of 16 min) than for Kv1.4 alone, suggesting the presence of undisclosed internalization motifs or mechanisms in PSD-95 (or its complex with Kv1.4). These data reveal an entirely new role for PSD-95 and channel clustering, and add weight to the notion that ion channel targeting may, in part, be conferred by regional differences in the stabilization of ion channel expression at the cell surface. EXPERIMENTAL PROCEDURES

Materials—A clone encoding myc-tagged PSD-95 in the expression vector pGW1-CMV was a generous gift of Dr. M. Sheng (Harvard University, Boston, MA). Anti-PSD-95 monoclonal antibody was purchased from Transduction Laboratories, Lexington, KY. The appropriate secondary antibodies conjugated to horseradish peroxidase were purchased from Amersham Pharmacia Biotech. Oligonucleotide primers were synthesized by the University of Toronto/Hospital for Sick Children Biotechnologies Services (Toronto, ON, Canada) or ACGT Inc. (Toronto, ON, Canada). All other chemicals were of reagent grade or higher. Plasmids and DNA Constructs—The Kv1.4 mammalian expression vectors were constructed via the polymerase chain reaction (PCR) using, as template, first strand cDNA synthesized from rat cortex (Fast Track, Stratagene, San Diego, CA). First, a Kv1.4 “core” cDNA encoding the NH2 terminus to transmembrane spanning segment S6 (amino acids M (1)-P (577)) was PCR-amplified using the following oligonucleotide primers: 5⬘-GCGTCATCATCAGACCGCATC-3⬘ and 5⬘-AAATTTCCCGGGCAAAGCAATGGTTAAGAC-3⬘ which also served to introduce a 3⬘ SmaI site for subcloning. A second cDNA encoding the Kv1.4 “COOH terminus” (amino acids V (578)-V (655)), was generated likewise using the following oligonucleotide primers: 5⬘-AAATTTTACGTA CCGGTGATTGTGTCTAACTTT-3⬘ and 5⬘-AAGTCTCTCTCGGGCTTCAGC-3⬘ which also introduce a 5⬘ SnaBI site. The resulting PCR products were then subcloned into pCRII (Invitrogen, San Diego, CA). The Kv1.4 core was then excised from pCRII using EcoRI and SmaI, and ligated into the corresponding sites in the mammalian expression vector pSVK3 (Amersham Pharmacia Biotech, Baie d’Urfe, PQ, Canada). To complete the Kv1.4 expression vector, the Kv1.4 COOH terminus was excised from pCRII using SnaBI and ApaI, and ligated into the corresponding sites of the Kv1.4core-pSVK3 vector. An EYFP (extended yellow variant of green fluorescent protein)-tagged Kv1.4 cDNA was constructed by PCR using the Kv1.4-pSVK3 vector as template and the following primers: 5⬘-TTTAAGCTTCTATGGAGGTGGCAATG-3⬘ and 5⬘-GGCACAATTCCCAAC-3⬘. The PCR product was then digested with

HindIII and HincII and ligated into the HindIII and SmaI sites of pEYFP-C1 (CLONTECH, Palo Alto, CA). A PSD-95 mutant lacking cysteine residues at positions 3 and 5 was constructed by PCR using as template myc-tagged PSD-95 in the expression vector pGW1-CMV. The mutagenic primer (5⬘-AAAGAATTCATGGACTCTCTCTCTATAGTGACA-3⬘) corresponded to the first 24 nucleotides of PSD-95 coding sequence together with G to C substitutions (underlined) at the second position of the cysteine-3 or cysteine-5 codons and an EcoRI site for subcloning. The reverse primer was 5⬘-TTTCTCGAGTTTTCAGAGTCTCTC-3⬘ and included the stop codon and an XhoI site for cloning. The identity of the constructs was confirmed by restriction mapping and DNA sequencing (U.S. Biochemical Sequenase V2.0 or ACGT Inc., Toronto, ON). Cell Culture and Transfections—HEK293 were obtained from American Type Culture Collection (Manassas, VA) and grown at 37 °C, 95% O2, 5% CO2 in minimal essential medium (Life Technologies, Inc., Oakville, ON), supplemented with 10% fetal bovine serum, penicillin (10 ␮g/ml), and streptomycin (10 units/␮l). Transient transfections were performed via lipofection of HEK293 cultures. Briefly, cells were seeded directly on 35-mm plastic culture dishes (day 0), allowed to grow to confluency (day 2–5), and then transfected with fresh LipofectAMINE (Life Technologies, Inc.) (2 ␮g of DNA, 8 ␮g of LipofectAMINE reagent/ dish), diluted into minimal essential medium (500 ␮l). After standing for 5 h, the transfection medium was replaced with normal medium, and again at 24 h following the start of the transfection. At no time during the experiment were the cells exposed to antimitotic or antibacterial agents. The transfection efficiency was ⬎65%. Labeling of Cell Surface Proteins for the Detection of Internalization—Cell surface labeling of glycoproteins was performed using a version of the periodate-biotinylation assay described by Prince et al. (32). All procedures were conducted at 4 °C except where indicated. Briefly, at 48 h post-transfection, HEK293 cells were chilled to 4 °C, washed with ice-cold phosphate-buffered saline containing 0.1 mM CaCl2 and 1 mM MgCl2 (PBS-CM), and incubated with 20 mM NaIO4 in PBS-CM for 30 min in the dark, to oxidize cell surface carbohydrate residues. After re-washing with PBS-CM, fresh pre-warmed minimal essential medium was added to the cells and they were returned to the incubator to allow internalization to proceed. At specified times (t ⫽ 0 to 320 min) the cells were re-washed with ice-cold PBS-CM and labeled with 2 mM biotin-LC-hydrazide in 100 mM sodium acetate (pH 5.5) for 30 min in the dark. The cells were then re-washed with PBS-CM and lysed in 100 ␮l of lysis buffer (20 mM HEPES, 150 mM NaCl, 1 mM EDTA, 1% (v/v) Nonidet P-40, pH 7.4), supplemented with the protease inhibitors aprotinin (100 ␮g/ml), leupeptin (100 ␮g/ml), and phenylmethylsulfonyl fluoride (2 mM). Membrane Preparations—An HEK293 soluble membrane fraction was made by shearing HEK293 cell lysates 8⫻ with an 18-gauge needle, and centrifuging at 12,000 ⫻ gav for 3 min to remove insoluble material. Rat brain membranes were prepared as described (33). Membranes for biotinylated protein isolation were used immediately, all others were frozen and stored in liquid N2. Membrane protein concentrations were determined via a commercial kit (DC Protein Assay, Bio-Rad). Separation and Isolation of Biotinylated Proteins—Biotinylated proteins were isolated by incubating solubilized HEK293 membranes (40 ␮g of protein) with streptavidin-agarose (160 ␮l of 50% slurry) (Molecular Probes, Eugene, OR) pre-equilibrated in lysis buffer. After rocking the reaction mixtures gently for 12–16 h at 4 °C, biotin-streptavidinagarose complexes were harvested by centrifugation, washed three times with a 20-fold excess of lysis buffer, and once with 10 mM Tris-HCl (pH 8.0), 300 mM NaCl. The beads were then resuspended in 75 ␮l of SDS sample buffer and boiled for 5 min prior to SDS-PAGE. Synthetic Peptide Antibodies—A peptide corresponding to residues N (14)-R (37) of Kv1.4 (34) (GenBank accession number P15385), plus a COOH-terminal cysteine for coupling, was synthesized by Vetrogen (London, ON, Canada). The identity of the peptide was confirmed by amino acid analysis and mass spectroscopy. Kv1.4 peptide was coupled to keyhole limpet hemocyanin with the cross-linker m-maleimidobenzoyl-N-hydroxysulfosuccinimide ester (Pierce) and the conjugate dialyzed against PBS. New Zealand White rabbits were immunized with the conjugate (Division of Comparative Medicine, University of Toronto) and antisera were collected and the IgG fraction was enriched using protein A affinity chromatography (MAPS kit, Bio-Rad, Mississauga, ON, Canada). SDS-PAGE and Immunoblot Analysis—Protein samples were heated in SDS sample buffer and the proteins resolved by electrophoresis on 7.5% SDS-polyacrylamide gels (35). The proteins were transferred electrophoretically to nitrocellulose and immunoblotted as described previ-

PSD-95 Suppresses Kv1.4 Endocytosis ously (33) using the following primary antibody conditions: Kv1.4, 1:250 of 4 ␮g/␮l, 1 h; PSD-95, 1:250 of 0.25 ␮g/␮l, 1 h. Blots were incubated with horseradish peroxidase-conjugated secondary antibodies at 1:2500 – 4000 for 2 h. Immunoreactive proteins were detected by enhanced chemiluminescence (ECL; Amersham Pharmacia Biotech, Oakville, ON, Canada) and quantified with a Bio-Rad Model GS-670 imaging densitometer. For display, exposed films were digitized via a StudioScan IIsi scanner (AGFA, Toronto, ON, Canada) using FotoLook PS2.07.2 software (Adobe Systems, Mountainview, CA). Digitized images were cropped, using Photoshop 4.0 software (Adobe Systems) and displayed without any further processing. Electrophysiology—The functional expression of Kv1.4 and EYFPKv1.4 in transfected HEK293 cells was assessed through whole cell patch clamp recordings (Axopatch 200A, Axon Instruments) (36). Recordings were obtained 48 h after transfection on cells of simple morphology lacking contacts with surrounding cells. The internal pipette solution (300 –310 mOsm) contained (in mM): 110 KCl, 5 BAPTA (tetrapotassium salt), 2 K2-ATP, 2 MgCl2, and 10 HEPES, adjusted to pH 7.2 with KOH. The bathing solution (320 –330 mOsm) contained (in mM): 145 NaCl, 10 glucose, 4 KCl, 1.8 CaCl2, 1 MgCl2, and 10 HEPES, adjusted with NaOH to pH 7.35. Pipettes were pulled from borosilicate glass (World Precision Instruments, Sarasota, FL) and had resistance of 8 –10 megaohm when filled with internal solution. All voltages were corrected to account for the junction potentials between bath, pipette, and agar bridge solutions (37). Voltages were applied and currents recorded using pCLAMP software (Ver. 6.0, Axon Instruments) and a Tecmar Labmaster data acquisition board. Analog subtraction was used to cancel capacitative transients during data acquisition. Currents were filtered at 5 KHz and the series resistance was compensated 50 –70%. The time courses for activation and inactivation were calculated from nonlinear least squares fitting to Hodgkin-Huxley type n4j kinetics as described elsewhere (38). Fluorescence Imaging—Digital EYFP-Kv1.4 images were captured on an inverted scanning confocal microscope (Bio-Rad MRC-600) equipped with an argon ion laser, using fluorescein optics as detailed previously (33). Statistics—Unless otherwise specified all data are given as mean ⫾ S.E. for n trials and analysis was done using Sigma Plot V.4.0 (Jandel, CA) or Origin 5.0 (Microcal, CT) software. RESULTS

Antibody Characterization—To resolve the internalization of Kv1.4 in detail, we first raised specific polyclonal antibodies against an intracellular sequence corresponding to residues 14 –37 of the Kv1.4 amino terminus (34) (Fig. 1A) that has previously been used successfully for this purpose (11). High titer antisera from rabbits immunized with this synthetic peptide recognized a band of 90 kDa in immunoblots of rat brain membranes (Fig. 1B, lane 1) that could be eliminated by pretreatment of the Kv1.4 antiserum with competing antigenic peptide (Fig. 1B, lane 2). Although this band is similar in size to that found previously in brain (11), it is somewhat higher than that anticipated from the published sequence (73 kDa). To obtain further evidence for the specificity of the Kv1.4 antibody we therefore analyzed immunoblots of lysates of HEK293 cells transfected with cDNAs encoding Kv1.4. As shown in Fig. 1C, lane 1, our Kv1.4 antibody recognized a band of 76 kDa that is absent upon preincubation of the antibody with competing antigenic peptide (lane 2) or in lysates from mock-transfected HEK293 cells (lane 3). Absence of Endogenous PSD-95 in HEK293 Cells—The absence of Kv1.4 in untransfected HEK293 cells is not surprising based on previous electrophysiology (23, 39). However, the expression of significant levels of endogenous PSD-95 could seriously complicate the analysis of its role in the endocytosis of Kv1.4. We therefore examined the levels of PSD-95 in HEK293 cells before and after transfection with a mammalian expression vector encoding PSD-95, tagged with a functionally inert c-myc epitope (22). Immunoblots of rat cortical membranes probed with a commercial antibody against PSD-95 (Fig. 1D, lane 1) disclosed a band of ⬃95 kDa as reported (18, 22). While a band of identical size was present in lysates of PSD-95-

1359

FIG. 1. Antibody characterization. A, postulated structure of rat Kv1.4 showing the NH2-terminal amino acid sequence (Asn14-Arg37) used to generate antiserum. Also shown is the single N-linked glycosylation site, Asn354 whose attached sugars are the presumptive site of hydrazide labeling in the two-step periodation/biotinylation assay. B, an antibody against Kv1.4 recognizes a band of ⬃90 kDa (lane 1) in immunoblots of rat hippocampal membranes (2.5 ␮g/lane) which can be eliminated (lane 2) if the antibody is first treated with competing Kv1.4 antigenic peptide (25 ␮g/ml). C, the anti-Kv1.4 antibody recognizes a single band of ⬃76 kDa in lysates of HEK293 cells transfected with Kv1.4-pSVK3 (lane 1) that is eliminated by a competing Kv1.4 antigenic peptide (lane 2). The Kv1.4 signal was absent in non-transfected HEK293 lysates (lane 3). D, an antibody against PSD-95 recognizes a band of ⬃95 kDa (lane 1) in both rat cortical membranes (1 ␮g/lane) and lysates of HEK293 cells transfected with PSD-95-pGW1-CMV (lane 2). The 95-kDa band is absent in non-transfected HEK293 cell lysates (lane 3). The Kv1.4 and PSD-95 blots were analyzed with Monoclonal antibody purification system-purified Kv1.4 antibody (16 ␮g/ml) or monoclonal anti-PSD-95 (1 ␮g/ml), respectively. Blots were developed using a peroxidase-conjugated secondary antibody with chemiluminescent detection (see “Experimental Procedures”). Molecular weights (arrowheads) were derived using molecular weight standards.

transfected HEK293 cells (lane 2), this band was absent from untransfected cells run under identical conditions (lane 3). Specific Cell Surface Labeling of Kv1.4 —To observe the internalization of Kv1.4, we utilized a two-step periodation/biotinylation assay to distinguish proteins present on the cell surface from those within the cytoplasm (32, 40). The basis of this assay lies in the conversion of sugar diols present within cell surface glycoproteins such as Kv1.4 (6), to reactive aldehydes, by the mild, cell impermeant, oxidant sodium periodate. The extent of internalization following oxidation is then deduced from the disappearance of the aldehydic sites available to react with aldehyde-specific, non-permeable labels such as biotin hydrazide. To confirm the specificity of the internalization assay, we examined the cell-surface biotinylation of Kv1.4 expressed in transfected HEK293 cells. Following periodate oxidation and biotin hydrazide labeling, biotinylated proteins were harvested from cell lysates by affinity precipitation with streptavidin-agarose. The precipitates were then analyzed by SDS-PAGE and immunoblotting with Kv1.4 antibody (Fig. 2). In experimental samples two immunoreactive bands were ap-

1360


FIG. 2. Specific detection of cell-surface Kv1.4 via two-step periodation oxidation/biotinylation. A, cell surface sugar residues of HEK293 cells transfected with Kv1.4, were oxidized with periodate, then the cells were washed and immediately labeled with biotin hydrazide. The biotinylated glycoproteins were harvested from cell lysates with streptavidin-agarose and the precipitates analyzed for Kv1.4 by immunoblotting (lane 1) (see “Experimental Procedures”). Specific bands were absent in immunoblots of affinity precipitates from transfected cells not exposed to both periodation and biotinylation steps (lane 2), or in cells exposed to just the sodium periodate (lane 3), or biotinLC-hydrazide (lane 4) steps. B, detection of Kv1.4 in immunoblots is linear with respect to protein concentration between 2 and 20 ␮g. Dashed lines show 95% confidence limits.

parent, the first corresponding to a band of 77 kDa, as anticipated (lane 1) and a second lower band of 71 kDa. Omission of both periodate oxidation and biotinylation (lane 2), just the oxidation (lane 3), or just the biotinylation (lane 4) steps led to the disappearance of the upper, presumptive Kv1.4 band but not the lower band. Together these data suggest that Kv1.4 immunoreactivity is unaffected by the two-step procedure and support the validity of this assay as a way to detect cell surface Kv1.4 in transfected HEK293 cells. Suppression of Kv1.4 Internalization by PSD-95—Using the two-step periodate-biotin hydrazide assay we analyzed the expression of Kv1.4 at the cell surface of HEK293 with or without PSD-95. To preclude variability between transfected dishes the band intensities for cell surface Kv1.4 (obtained from the streptavidin-agarose precipitates) were normalized with respect to the total level of Kv1.4 in each cell lysate (obtained by direct analysis of an aliquot of cell lysate) (32). Such variability arose, primarily, through cell losses during the wash steps rather than differences in transfection efficiency or plated cell density. In the absence of PSD-95 (Fig. 3A, i), the expression of Kv1.4 at the cell surface declined rapidly, and in most samples was undetectable after 5 h. Assuming that Kv1.4 internalization, like other proteins (25), decreased in a single exponential fashion we determined a rate constant of 8.0 ⫾ 1.8 ⫻ 10⫺3 min⫺1 (t1⁄2 ⫽ 87 min) (Fig. 3B). An entirely different picture emerged when Kv1.4 was co-expressed with PSD-95 (Fig. 3A, ii). In this case labeling of cell-surface Kv1.4 did not decrease, but rather, showed a very slight increase characterized by a rate constant of 3.1 ⫾ 0.9 ⫻ 10⫺4 min⫺1 (t1⁄2 ⫽ 37 h) (Fig. 3B). PSD-95-Kv1.4 Interaction Does Not Affect Total or Cell Surface Kv1.4 Expression Levels—The concentration of membrane proteins at the cell surface is determined by both their rates of insertion and their rates of removal (25). It therefore seemed plausible that suppression of Kv1.4 internalization by PSD-95 might cause an overall accumulation of Kv1.4 at the cell surface. To test this possibility we compared the initial expression levels of Kv1.4 at the cell surface, with and without PSD-95, by labeling with biotin hydrazide immediately after periodate ox-

FIG. 3. Suppression of Kv1.4 internalization by PSD-95. A, time course for the internalization of Kv1.4 in the absence (i) and presence (ii) of PSD-95 as determined by the two-step periodate oxidation/biotinylation assay and immunoblotting (see “Experimental Procedures”). The level of Kv1.4 at the cell surface (upper blot set) was corrected for any differences in the total levels of Kv1.4 expression (lower blot set) by simultaneously examining the density of the Kv1.4 bands in the precipitates relative to those in the corresponding cell lysates. Each blot set is a composite of selected cell surface and total partner blots. B, time course for the internalization of Kv1.4 in the presence (filled triangles) and absence (filled circles) of PSD-95. Data were determined from densitometry of multiple blot pairs like those in A. Each data point is expressed as a normalized ratio of cell surface value to total lysate value, independent of the other time points and is expressed as the mean ⫾ S.E. from more than three independent transfection experiments. Lines indicate curves of best fit (nonlinear least squares) assuming internalization proceeds with time (t) according to the equation: A ⫽ Aoe⫺kt, where A and Ao are the calculated and initial band densities, respectively, and k is the corresponding first-order rate constant. In the absence of PSD-95, k ⫽ 8.03 ⫾ 1.7 ⫻ 10⫺3 min⫺1, whereas in the presence of PSD-95, k ⫽ ⫺3.1 ⫾ 0.9 ⫻ 10⫺5 min⫺1.

idation (i.e. at t ⫽ 0 min). Labeling at this time point corresponds to the maximum level of expression of Kv1.4 at the cell surface since internalization of glycoprotein-oxidized Kv1.4 has not yet occurred. Surprisingly, the initial level of Kv1.4 expressed at the cell surface was indistinguishable between cells that had been transfected with Kv1.4 or co-transfected with Kv1.4 and PSD-95 (Fig. 4). To exclude the possibility that a PSD-95-mediated increase in the surface expression of Kv1.4 was masked by some compensatory decrease in Kv1.4 protein synthesis we also examined the effect of PSD-95 on the total cellular levels of Kv1.4. However, the total level of Kv1.4 expression was, if anything, slightly higher in the presence of PSD-95 and this difference was not statistically significant (Fig. 4). Detailed calculations suggest that approximately 50% of the expressed Kv1.4 is located at the cell surface irrespective of the presence of PSD-95. To provide an independent functional confirmation of the

PSD-95 Suppresses Kv1.4 Endocytosis lack of effect of PSD-95 on Kv1.4 cell-surface expression levels, we used whole cell patch clamp electrophysiology (Fig. 5). Expression of Kv1.4 in transfected (but not untransfected) HEK293 cells yields large, inactivating, currents identical to those described previously (23, 34, 39) (Fig. 5A). Essentially identical recordings were obtained from cells co-transfected with PSD-95 and Kv1.4 (Fig. 5B). A more detailed analysis revealed that PSD-95 had no effect on the peak or steady-state Kv1.4 conductances (Fig. 5C) and did not alter the kinetics of activation. At ⫹60 mV the values were Kv1.4, ␶Act ⫽ 7.8 ⫾ 0.5 ms (n ⫽ 6); Kv1.4 ⫹ PSD-95, ␶Act ⫽ 2.7 ⫾ 0.5 ms (n ⫽ 6), and Kv1.4, ␶Inact ⫽ 132 ⫾ 34 ms (n ⫽ 5); Kv1.4 ⫹ PSD-95, ␶Inact ⫽ 111 ⫾ 14 ms (n ⫽ 5) (Fig. 5D) (all p ⬎ 0.05). Suppression of Kv1.4 Internalization by PSD-95 Is Clusteringdependent—A central feature of PSD-95 is its ability to induce ion channel clustering, most likely through interactions medi-

FIG. 4. Steady-state expression of total and cell-surface Kv1.4 is unaltered by co-expression with PSD-95. Total and cell-surface expression of Kv1.4 in HEK293 cells with (black bars) or without (open bars) transfection of PSD-95. Total expression was determined by immunoblotting of cell lysates with Kv1.4 antibodies. Cell-surface expression of Kv1.4 was determined by immunoblotting streptavidin-agarose precipitates of glycoproteins labeled with biotin hydrazide immediately after oxidation as described in the legend to Fig. 2. The data are expressed as the mean ⫾ S.E. (n ⫽ 6, no PSD-95; n ⫽ 5 with PSD-95).

1361

ated by NH2-terminal cysteine residues (21, 22, 41). However, the marked suppression of Kv1.4 internalization by PSD-95 could arise independently of clustering. For example, PSD-95 can act as a scaffold for proteins involved in intracellular signaling (41– 43), whose modulation might indirectly block Kv1.4 internalization. To discern a direct from an indirect role of PSD-95-mediated clustering in suppressing Kv1.4 internalization, we first generated a PSD-95 mutant (C35S-PSD-95) that has been reported to bind to, but not cluster, Kv1.4 (22). The basis for this mutation lies in the conversion of those cysteine residues (positions 3 and 5) essential for PSD-95 clustering to serine residues. Next, we used a live cell assay to visualize, directly, the distribution of Kv1.4 in transfected HEK293 cells and thereby gauge the clustering activity of the wild-type and mutant PSD-95 constructs. In this approach, we tagged the amino terminus of Kv1.4 with a yellow variant of GFP (EYFP). The resulting construct (EYFP-Kv1.4) generated functional channels when expressed in HEK293 cells (Fig. 6, A and B). Although the current amplitudes were similar (peak conductance: wt-Kv1.4 ⫽ 19 ⫾ 7 pA/pF (n ⫽ 5); EYFP-Kv1.4 ⫽ 47 ⫾ 4 pA/pF (n ⫽ 4)), there was a slowing of inactivation (Fig. 6C), when compared with the wild-type (wt) channels (shown in Fig. 5). The inactivation rate (␶Inact) was 12-fold slower for EYFPtagged Kv1.4 channels (␶Inact EYFP-Kv1.4 ⫽ 1582 ⫾ 177 ms (n ⫽ 4); wt-Kv1.4, ␶Inact ⫽ 132 ⫾ 34 ms (n ⫽ 5). The activation kinetics (Fig. 6C) were also altered, i.e. they were 3-fold slower for EYFP-tagged versus wt-Kv1.4 channels (␶Act EYFP-Kv1.4 ⫽ 8.7 ⫾ 0.8 ms (n ⫽ 4), ␶Act wt-Kv1.4 ⫽ 2.7 ⫾ 0.5 ms (n ⫽ 6)). A slowing of channel inactivation is predicted on the basis of the “ball and chain” model for Kv1.4 inactivation (44) (see “Discussion”), and that is known to occur upon expression of similar GFP-tagged Kv1.4 constructs (23, 39). Most importantly, our data indicate that EYFP-Kv1.4 channels are faithfully assembled and functional upon targeting to the cell surface. Another distinguishing characteristic of Kv1.4 channels is their susceptibility to block by 4-aminopyridine (45), which did not differ for Kv1.4 currents in the presence or absence of PSD-95 (Fig. 6D). The transfection of HEK293 cells with EYFP-Kv1.4 results

FIG. 5. PSD-95 co-expression does not affect Kv1.4 current density or kinetics in HEK293 cells. Whole cell Kv1.4 currents in HEK293 cells transfected with Kv1.4 in the absence (A) or presence of co-transfected PSD-95 (B). Voltage-dependent outward K⫹ currents were elicited by voltage steps from a holding potential of ⫺80 mV in 20 mV increments between ⫺60 and ⫹60 mV (inset, voltage protocol). Each voltage step was 1500 ms long with an inter-pulse interval of 45 s. C, peak (open bars) and steady-state (black bars) current amplitudes for control (mock transfected) (n ⫽ 7), Kv1.4 (n ⫽ 11), and Kv1.4 ⫹ PSD95 (n ⫽ 5) expressing cells. Data are shown as mean ⫾ S.E. Peak currents were measured during steps to ⫹60 mV after the capacitative transient and normalized to cell capacitance to correct for differences in cell surface area. The steady-state amplitude was measured at the end of each voltage step. D, activation (␶Act) and inactivation (␶Inact) time constants for Kv1.4 currents in HEK293 cells expressing Kv1.4 alone (open bars) or with PSD-95 (black bars). Time constants were determined at ⫹60 mV by nonlinear least square-fit analysis to the following equation: I(t) ⫽ Imax{1-exp[ ⫺ t/Zact]}4 exp[⫺t/Zinact] (36). Currents were obtained in approximately 65% of cells tested.

1362


FIG. 6. Biophysical properties of extended yellow fluorescent proteintagged Kv1.4 (EYFP-Kv1.4) in HEK293 Cells. Representative whole cell currents in a control non-transfected cell (A) and an EYFP-Kv1.4 transfected cell (B) in response to voltage-clamp steps as in Fig. 5, but with an inter-pulse interval of 60 s. C, activation (␶Act) (open bars) and inactivation (␶Inact) (black bars) time constants were determined at ⫹60 mV as in Fig. 5. D, an example of the inhibition of Kv1.4 current by bath perfusion with 1 mM 4-aminopyridine. The peak Kv1.4 current during a step to ⫹40 mV was measured, every 20 s. The current for wt-Kv1.4 (filled circles) and EYFP-Kv1.4 (open squares) expressing cells, was then plotted with respect to the value at time 0 (I0). A control cell (wt-Kv1.4 transfected) without 4-AP (filled triangles) shows no decrease in current.

in the expression of diffuse fluorescence throughout the cell (Fig. 7A). The highest levels of fluorescence are seen in perinuclear regions and at the cell periphery. As noted elsewhere (39), this pattern is in marked contrast to the even fluorescence noted throughout the cytoplasm in cells expressing EYFP (Fig. 7B) or GFP (23) alone. Upon co-expression of EYFP-Kv1.4 with PSD-95 a distinctly different, punctate pattern of fluorescence is observed, consistent with the formation of clusters of Kv1.4 at the cell surface (Fig. 7C). Such aggregates are not observed when EYFP-Kv1.4 is co-expressed with C35S-PSD-95 (Fig. 7D), illustrating that PSD-95 is capable of mediating normal clustering interactions with EYFP-Kv1.4 in vivo that can be abrogated by site-directed mutagenesis of its amino-terminal cysteine residues. Having confirmed the inability of C35S-PSD-95 to mediate Kv1.4 clustering, we next examined its effect on the rate of internalization of Kv1.4 using the two-step periodation-biotinylation assay (Fig. 8, A and B). Surprisingly, co-expression of Kv1.4 with C35S-PSD-95 increased the rate of Kv1.4 internalization (k ⫽ 4.23 ⫾ 0.7 ⫻ 10⫺2 min⫺1; t1⁄2 ⫽ 16 min), to a value ⬎5 times faster than for Kv1.4 alone (Fig. 8B). These data suggest that clustering per se is both necessary and sufficient to suppress Kv1.4 endocytosis. DISCUSSION

In this paper we document evidence that the internalization of Kv1.4 K⫹ channels is suppressed by their interaction with PSD-95. Such an effect reflects the ability of PSD-95 to selfassociate and cluster Kv1.4 channels, directly, rather than through its ability to mediate indirect effects such as intracellular signaling. Our conclusions are based on our ability to faithfully detect Kv1.4 internalization and clustering in transfected cells. Transfection and antigen-competition experiments argue that our antibody is highly specific for Kv1.4 and that it behaves identically to an antibody generated previously using an identical peptide sequence and conjugation method (11). In rat brain membranes our antibody recognizes a single band of ⬃90 kDa, somewhat larger than the 73-kDa protein predicted from the amino acid sequence, or generated by in vitro translation of our Kv1.4 construct (data not shown). However, tissue-dependent differences in the size of Kv1.4 are well documented (11, 17, 46) and appear to arise through extensive post-translational modification especially N-linked glycosylation, or possibly the use of internal ribosome entry sites (46) (which are retained in our construct). The periodation-biotin hydrazide assay we used

FIG. 7. Clustering of EYFP-Kv1.4 upon co-expression with wild-type and mutant PSD-95. A, distribution of EYFP-Kv1.4 expressed in HEK293 cells. B, cells expressing EYFP vector alone. C, cells transfected with EYFP-Kv1.4 in the presence of wild-type PSD-95. Arrows indicate punctate clusters. D, cells transfected with EYFPKv1.4 in the presence of the clustering deficient mutant C35S-PSD-95. Confocal images were obtained as described under “Experimental Procedures.”

to monitor internalization has been employed previously, for example, in defining the internalization of proteins such as the cystic fibrosis transmembrane regulator (32). An advantage of this assay is that, owing to its dependence on both cell surface expression and extracellular glycosylation, it monitors only those Kv1.4 channels that are correctly processed and expressed in the plasma membrane (32, 40). Our strategy for determining clustering is based on a direct visualization of the spatial distribution of an EYFP-tagged Kv1.4 in HEK293 cells. A similar approach has recently been used to monitor the PSD-95-mediated clustering of GFP-tagged wild-type and mutant Kv1.4 in HEK293 and COS-1 cells (23). Our choice of EYFP was based on the greater potential flexibility it confers over GFP in two-color and energy transfer studies. Electrophysiologically, the properties of the expressed EYFP-Kv1.4 match those described for GFP-Kv1.4 (23, 39). This result is not surprising as EYFP is almost identical in structure to GFP and is attached at the same amino-terminal location in Kv1.4 as GFP. However, in addition to the chromophore, the GFP-Kv1.4 constructs made elsewhere also differ from our EYFP-Kv1.4 by the addition (39) or removal (23) of


FIG. 8. Kv1.4 internalization is not suppressed by a clusteringdeficient mutant C35S-PSD-95. Time course for the internalization of Kv1.4 in the presence of C35S-PSD-95. The expression of Kv1.4 at the cell surface of HEK293 cells co-transfected with Kv1.4 and C35S-PSD95, was determined as described in the legend to Fig. 3. A, immunoblots of corresponding cell surface and total Kv1.4 level pairs, obtained at different incubation times. B, time course for the internalization of Kv1.4 in the presence (filled circles) of C35S-PSD-95. Data were determined from densitometry of multiple blot pairs like those in A. Each data point is expressed as a normalized ratio of cell surface value to total lysate value, independent of the other time points and is expressed as the mean ⫾ S.E. from more than four independent transfection experiments. The solid curve is the best fit (nonlinear least squares) assuming first-order kinetics (see Fig. 3B, legend), with a first-order rate constant k ⫽ 4.23 ⫾ 0.7 ⫻ 10⫺2 min⫺1. Additional broken curves describe internalization kinetics for values of k that are the same as or 2- and 10-fold (k ⫽ 8.03 ⫻ 10⫺3 min⫺1, long dashes; k ⫽ 1.54 ⫻ 10⫺3 min⫺1, medium dashes; k ⫽ 8.03 ⫻ 10⫺2 min⫺1) greater than found experimentally with Kv1.4 in the absence of PSD-95 (Fig. 3A, i).

amino acids in the region of their attachment to Kv1.4. The loss of fast inactivation in EYFP-Kv1.4 has also been noted for GFP-Kv1.4 and is entirely consistent with the ability of mutations in the amino terminus to disrupt the normal role of this region in ball and chain inactivation. The blockade of fast inactivation appears to be due to the presence of the large amino-terminal GFP moiety as the residues comprising the ball were retained in our EYFP-Kv1.4. The residual slow inactivation seen in such mutants can be ascribed to a “C-type” mechanism involving residues in the extracellular S5 pore loop (39). The rates of membrane protein internalization can be extraordinarily high (25). In the absence of PSD-95, the rate of Kv1.4 internalization we detected is comparable to that for many other membrane proteins undergoing constitutive endocytosis (25, 32, 40). In this light, the ability of PSD-95 to suppress internalization, completely, is remarkable and its detailed mechanism unclear. Clustering appears to be obligatory for suppressing internalization. An alternative, that PSD-95 is manifesting its effects via intracellular signaling, seems unlikely. Although PSD-95 contains multiple proteinprotein interaction modules (notably an SH3 and three PDZ domains) that can act as a scaffold for proteins involved in intracellular signaling (42, 43), so does the clustering deficient C35S mutant, which does not suppress internalization of Kv1.4. Moreover, since C35S-PSD-95 can still associate with the membrane, presumably via its unfettered binding to Kv1.4 (22), it is unlikely that its inability to suppress Kv1.4 internal-

1363

FIG. 9. Models for suppression of Kv1.4 internalization by PSD-95. Cell-surface expression of Kv1.4 is envisaged to depend on at least two processes: insertion and endocytosis. Endocytosis is presumed to occur via delivery of Kv1.4 into an endosomal compartment from which channels may be either recycled to the cell surface, or sent to lysosomes for subsequent degradation (d). It is proposed that PSD-95 suppresses internalization without affecting total or cell surface steadystate levels by interfering with both insertion (a) and endocytosis (b). An alternative model (c) envisages suppression of internalization and recycling. In these models it is assumed that only free Kv1.4 (Kv1.4f) and not clustered Kv1.4 (Kv1.4c) is susceptible to endocytosis.

ization arises through spatial disruption of a membrane-associated signaling pathway. How clustering disrupts internalization is uncertain, in part, because the basis for clustering is unresolved. From studies in vitro (18, 20, 21) and in vivo (23), it is now well established that PSD-95 interacts, via its first and/or second PDZ domains, with a motif that includes the last four amino acids (ETDV) of the carboxyl terminus of Kv1.4. By itself this interaction is insufficient to cause channel clustering. Rather, mutagenesis studies, including those with heterologous constructs (21, 22, 41), suggest that clustering depends on residues located in the amino terminus of PSD-95 placed in an NH2-terminal location in the expressed protein. In the simplest model, clustering occurs through multimerization of PSD-95 mediated by NH2terminal head-to-head interactions (18, 21, 22, 47). Since a pair of cysteine residues at positions 3 and 5 appears essential for self-association of PSD-95, it was suggested that clustering occurred through disulfide bridges (21, 22). However, these same cysteine residues appear to undergo palmitoylation (41, 42) a reversible post-translational modification that is likely to target PSD-95 to the membrane and increase its availability for binding to proteins such as Kv1.4. Consequently, a more refined model would suggest that palmitoylated PSD-95 forms aggregates at the inner membrane surface, which act as a matrix for protein tethering (42). In either model, we envisage that the suppression of Kv1.4 internalization by PSD-95 simply reflects the difficulty of physically dissociating the supramolecular aggregate. A particularly intriguing observation is the extremely rapid rate of internalization of Kv1.4 in the presence of C35S, but not wild-type PSD-95, or Kv1.4 alone. This effect appears to be specific to C35S-PSD-95 since a binding deficient mutant of Kv1.4, lacking the ETDV motif (23), internalizes at a rate similar to wild-type Kv1.4 (data not shown). How can binding, but not clustering of the C35S-mutant to Kv1.4 facilitate internalization of Kv1.4? One possibility is that PSD-95, or a protein with which it interacts, contains an undisclosed motif for rapid internalization that is suppressed by membrane association and/or clustering (48). A direct examination of the rate of clearance of PSD-95 and its mutants would shed light on this phenomenon. In many situations the clustering of membrane proteins is a preliminary step in their internalization (25). Thus, the ability

1364


of clustering to suppress Kv1.4 internalization is somewhat surprising. However, the suppression of internalization may be a general phenomenon, and may be critical for maintaining channels in regions such as synapses that undergo very rapid membrane turnover events. Powerful evidence in support of this notion comes from recent data showing that rapsyn, a postsynaptic protein unrelated to MAGUKs, which clusters nicotinic receptors at the neuromuscular junction, also suppresses their internalization (49, 50). One of our most intriguing observations concerns the ability of PSD-95 to block the internalization of Kv1.4 without affecting its overall level of expression at the cell surface. This finding, confirmed through two entirely different assays, may provide further important insights into why clustering occurs. The level of expression of membrane proteins is determined by their relative rates of insertion into, and endocytosis from, the membrane (25). Thus, three mechanisms might explain why the expression levels of Kv1.4 were not higher in the presence of PSD-95 (Fig. 9). The first, a direct suppression of Kv1.4 protein synthesis by PSD-95 expression, was contradicted by immunoblots showing identical levels of Kv1.4 in lysates from whole cells transfected with or without PSD-95. Alternatively, PSD-95 could suppress the rate of insertion of Kv1.4 into the membrane. This model suggests a further novel role for PSD-95 where it acts as a sensor that recognizes, and through negative feedback, regulates the insertion and levels of membrane proteins at the cell surface. An equally intriguing third model is based on the possibility that internalized Kv1.4 can, like many other membrane proteins, be trafficked back to the cell surface as well as to a degradative pathway (24, 25). Since the suppression of internalization of Kv1.4 by PSD-95 would affect both paths, our data could easily be rationalized if the rate of recycling was faster than the rate of insertion or degradation. Any decrease in the clearance of Kv1.4 from the cell surface would then cause a concomitant drop in insertion via the recycling pathway. Resolution of these models will clearly require a greater understanding of the pathways of K⫹ channel trafficking. A final consideration is the possibility that, depending on the state of clustering, PSD-95 plays a general role in controlling neuronal physiology by differentially regulating the intracellular trafficking of ion channels. Such a notion is supported by the ability of phosphorylation to regulate clustering (51) and may explain how PSD-95 enhances the current density of Kir4.3 inward rectifier, but not Kv1.4, channels (52). In summary, our data provide strong evidence that PSD-95mediated clustering suppresses the internalization of Kv1.4. This observation reveals a novel role for PSD-95 and presumably other MAGUK’s, and suggests an intimate relationship between the lateral segregation, targeting, and trafficking of membrane proteins. REFERENCES 1. Miller, C. (1991) Science 252, 1092–1096 2. Aidley, D. J., and Stanfield, P. R. (1996) Ion Channels: Molecules in Action, Cambridge University Press, Cambridge, New York 3. Chandy K. G., and Gutman, G. A. (1995) Handbook of Receptors and Channels: Ligand and Voltage-gated Channels, pp. 1–71, CRC Press Inc., Boca Raton, FL 4. MacKinnon, R. (1991) Nature 350, 232–235

5. Wang, H., Kunkel, D. D., Martin, T. M., Schwartzkroin, P. A., and Tempel, B. L. (1993) Nature 365, 75–79 6. Sheng, M., Liao, Y. J., Jan, Y. N., and Jan, L. Y. (1993) Nature 365, 72–75 7. Li, M., Unwin, N., Stauffer, K. A., Jan, Y. N., and Jan, L. Y. (1994) Curr. Biol. 4, 110 –115 8. Rettig, J., Heinemann, S. H., Wunder, F., Lorra, C., Parcej, D. N., Dolly, J. O., and Pongs, O. (1994) Nature 369, 289 –294 9. Castellino, R. C., Morales, M. J., Strauss, H. C., and Rasmusson, R. L. (1995) Am. J. Physiol. 269, H385-H391 10. Shi, G., Nakahira, K., Hammond, S., Rhodes, K. J., Schechter, L. E., and Trimmer, J. S. (1996) Neuron 16, 843– 852 11. Sheng, M., Tsaur, M. L., Jan, Y. N., and Jan, L. Y. (1992) Neuron 9, 271–284 12. Veh, R. W., Lichtinghagen, R., Sewing, S., Wunder, F., Grumbach, I. M., and Pongs, O. (1995) Eur. J. Neurosci. 7, 2189 –2205 13. Cooper, E. C., Milroy, A., Jan, Y. N., Jan, L. Y., and Lowenstein, D. H. (1998) J. Neurosci. 18, 965–974 14. Sheng, M., and Wyszynski, M. (1997) Bioessays 19, 847– 853 15. Rasband, M. N., Trimmer, J. S., Schwarz, T. L., Levinson, S. R., Ellisman, M. H., Schachner, M., and Shrager, P. (1998) J. Neurosci. 18, 36 – 47 16. Tsaur, M. L., Sheng, M., Lowenstein, D. H., Jan, Y. N., and Jan, L. Y. (1992) Neuron 8, 1055–1067 17. Wickenden, A. D., Kaprielian, R., Parker, T. G., Jones, O. T., and Backx, P. H. (1997) J. Physiol. (Lond.) 504, 271–286 18. Kim, E., Niethammer, M., Rothschild, A., Jan, Y. N., and Sheng, M. (1995) Nature 378, 85– 88 19. Doyle, D. A., Lee, A., Lewis, J., Kim, E., Sheng, M., and MacKinnon. R. (1996) Cell 85, 1067–1076 20. Craven, S. E., El-Husseini, A. E., and Bredt, D. S. (1999) Neuron 22, 497–509 21. Hsueh, Y. P., Kim, E., and Sheng, M. (1997) Neuron 18, 803– 814 22. Hsueh, Y. P., and Sheng, M. (1999) J. Biol. Chem. 274, 532–536 23. Burke, N. A., Takimoto, K., Li, D., Han, W., Watkins, S. C., and Levitan, E. S. (1999) J. Gen. Physiol. 113, 71– 80 24. Koenig, J. A., and Edwardson, J. M. (1994) J. Biol. Chem. 269, 17174 –17182 25. Mukherjee, S., Ghosh, R. N., and Maxfield, F. R. (1997) Physiol. Rev. 77, 759 – 803 26. Parton, R. G., and Dotti, C. G. (1993) J. Neurosci. Res. 36, 1–9 27. Bailey, C. H., Chen, M., Keller, F., and Kandel, E. R. (1992) Science 256, 645– 649 28. Parton, R. G., Simons, K., and Dotti, C. G. (1992) J. Cell Biol. 119, 123–137 29. Shimkets, R. A., Lifton, R. P., and Canessa, C. M. (1997) J. Biol. Chem. 272, 25537–25541 30. Liu, J., Bangalore, R., Rutledge, A., and Triggle, D. J. (1994) Mol. Pharmacol. 45, 1198 –1206 31. Dargent, B., Jullien, F., and Couraud, F. (1995) Neurochem. 65, 407– 413 32. Prince, L. S., Tousson, A., and Marchase, R. B. (1993)Am. J. Physiol. 264, C491– 498 33. Jones, O. T., Bernstein, G. M., Jones, E. J., Jugloff, D. G., Law, M., Wong, W., and Mills, L. R. (1997) J. Neurosci. 17, 6152– 6164 34. Stuhmer, W., Ruppersberg, J. P., Schroter, K. H., Sakmann, B., Stocker, M., Giese, K. P., Perschke, A., Baumann, A., and Pongs, O. (1989) EMBO J. 8, 3235–3244 35. Laemmli, U. K. (1970) Nature 227, 680 – 685 36. Schlichter, L. C., Sakellaropoulos, G., Ballyk, B., Pennefather, P. S., and Phipps, D. J. (1996) Glia 17, 225–236 37. Barry, P. H., and Lynch, J. W. (1991) J. Membr. Biol. 121, 101–117 38. London, B., Wang, D. W., Hill, J. A., and Bennett, P. B. (1998) J. Physiol. 509.1, 171–182 39. Kupper, J. (1998) Eur. J. Neurosci. 10, 3908 –3912 40. Lisanti, M. P., Le Bivic, A., Sargiacomo, M., and Rodriguez-Boulan, E. (1989) J. Cell Biol. 109, 2117–2127 41. Topinka, J. R., and Bredt, D. S. (1998) Neuron 20, 125–134 42. Craven, S. E., and Bredt, D. S. (1998) Cell 93, 495– 498 43. Tezuka, T., Umemori, H., Akiyama, T., Nakanishi, S., and Yamamoto, T. (1999) Proc. Natl. Acad. Sci. U. S. A. 96, 435– 440 44. Hoshi, T., Zagotta, W. N., and Aldrich, R. W. (1990) Science 250, 533–538 45. Garcia, M. L., Hanner, M., Knaus, H. G., Koch, R., Schmalhofer, W., Slaughter, R. S., and Kaczorowski, G. J. (1997) Adv. Pharmacol. 39, 425– 471 46. Negulescu, D., Leong, L. E., Chandy, K. G., Semler, B. L., and Gutman, G. A. (1998) J. Biol. Chem. 273, 20109 –20113 47. Kim, E., and Sheng, M. (1996) Neuropharmacology 35, 993–1000 48. Perego, C., Vanoni, C., Villa, A., Longhi, R., Kaech, S. M., Frohli, E., Hajnal, A., Kim, S. K., and Pietrini, G. (1999) EMBO J. 18, 2384 –2393 49. Phillips, W. D., Vladeta, D., Han, H., and Noakes, P. G. (1997) Mol. Cell Neurosci. 10, 16 –26 50. Wang, Z. Z., Mathias, A., Gautam, M., and Hall, Z. W. (1999) J. Neurosci. 19, 1998 –2007 51. Cohen, N. A., Brenman J. E., Snyder, S. H., and Bredt, D. S. (1996) Neuron 17, 759 –767 52. Horio, Y., Hibino, H., Inanobe, A., Yamada, M., Ishii, M., Tada, Y., Satoh, E., Hata, Y., Takai, Y., and Kurachi, Y. (1997) J. Biol. Chem. 272, 12885–12888