Knockdown of ASIC1 and Epithelial Sodium Channel Subunits Inhibits Glioblastoma Whole Cell Current and Cell Migration*

High grade gliomas such as glioblastoma multiforme express multiple members of the epithelial sodium channel (ENaC)/Degenerin family, characteristically displaying a basally active amiloride-sensitive cation current not seen in normal human astrocytes or lower grade gliomas. Using quantitative real time PCR, we have shown higher expression of ASIC1, αENaC, and γENaC in D54-MG human glioblastoma multiforme cells compared with primary human astrocytes. We hypothesize that this glioma current is mediated by a hybrid channel composed of a mixture of ENaC and acid-sensing ion channel (ASIC) subunits. To test this hypothesis we made dominant negative cDNAs for ASIC1, αENaC, γENaC, and δENaC. D54-MG cells transfected with the dominant negative constructs for ASIC1, αENaC, or γENaC showed reduced protein expression and a significant reduction in the amiloride-sensitive whole cell current as compared with untransfected D54-MG cells. Knocking down αENaC or γENaC also abolished the high PK+/PNa+ of D54-MG cells. Knocking down δENaC in D54-MG cells reduced δENaC protein expression but had no effect on either the whole cell current or K+ permeability. Using co-immunoprecipitation we show interactions between ASIC1, αENaC, and γENaC, consistent with these subunits interacting with each other to form an ion channel in glioma cells. We also found a significant inhibition of D54-MG cell migration after ASIC1, αENaC, or γENaC knockdown, consistent with the hypothesis that ENaC/Degenerin subunits play an important role in glioma cell biology.

Gliomas are the most common primary tumors of the central nervous system. These tumors arise either from astrocytes or their progenitor cells (1). Gliomas are divided into four grades based on the degree of malignancy. Glioblastoma multiforme (GBM), 2 Grade IV, is the most frequently occurring, most invasive, and has the worst prognostic outcome with a median survival of approximately one year from diagnosis (2).
We have previously reported the presence of an amiloridesensitive current in glioblastoma cells that is not seen in normal astrocytes or low grade gliomas (3). Amiloride is a potassium sparing diuretic that inhibits sodium channels composed of subunits from the epithelial sodium channel (ENaC)/Degenerin (Deg) family. Amiloride-sensitive Na ϩ channels are essential for the regulation of Na ϩ transport into cells and tissues throughout the body. These channels are found in all body tissues; from epithelia, endothelia, osteoblasts, keratinocytes, taste cells, lymphocytes, and brain (4). Apart from the ENaCs, the ENaC/Deg family also includes acid-sensing ion channels (ASICs) which have been found predominantly in neurons (4 -6). Primary malfunctions of ENaC/Deg family members underlie or are involved in the pathophysiology of several human diseases such as salt-sensitive hypertension (7,8), pseudohypoaldosteronism type I (7), cystic fibrosis (9), chronic airway diseases (10,11), and flu (12).
The ENaC/Deg family subunits share the same structural topology. They all have short intracellular N and C termini, two transmembrane spanning domains, and a large extracellular cysteine-rich loop (4,5). There are five ENaC subunits termed ␣, ␤, ␥, ␦, and ⑀. Functional ion channels arise from a multimeric assembly of these subunits. The prototypical ENaC channel of the collecting duct principal cell is thought to be ␣␤␥ENaC (13,14). The ␣-ENaC subunit appears to be the core conducting element, whereas the ␤and ␥-ENaC subunits are associated with trafficking and insertion of the channel in the cell membrane (13,15,16). ASICs are homologous to ENaCs and are most prevalently expressed in the brain and nervous system (17)(18)(19), although they are also found in the retina (20 -22), testes (23), pituitary gland (24), lung epithelia (22), and bone and cartilage (25). Four ASIC genes have been identified so far, ASIC1-4. Of these, ASIC1-3 has multiple splice variants (19,22). The crystal structure of chicken ASIC1 has revealed it to be a homotrimer (26). ASICs differ from their ENaC counterparts in that they are transiently activated by extracellular acid (19) and are much less sensitive to inhibition by amiloride (27,28). Also ASIC1 is inhibited with high affinity by psalmotoxin 1 (PcTX-1), a 40-amino acid peptide found in the venom of the West Indies tarantula, Psalmopoeus Cambridgei (29). ASICs, because they are activated by acidic pH, have been suggested to play a role in chemical pain associated with increased tissue acidification as occurs in ischemia (30,31). They have also been implicated in touch sensation (32), taste (33), fearconditioning (6), and learning and memory (34).
Our laboratory has proposed that ENaC/Deg channels underlie the basally activated cation current measured in high grade glioma cells (3). We hypothesize that the channels forming this current pathway are composed of a mixture of ASIC and ENaC subunits. RNA profiling of a large number of GBM-derived cell lines and freshly resected tumors have revealed the presence of a myriad of ASIC/ENaC components (3). The basally active current seen in GBM cells can be significantly reduced by amiloride or benzamil (a higher affinity amiloride analog), both of which are inhibitors of the ENaC/Deg family of ion channels (3). PcTX1, a selective ASIC1 blocker, also effectively abolishes the basally active GBM current (35). We have previously shown that ENaC and ASIC subunits can form cross-clade interactions in a heterologous expression system (36). This study aims to probe the composition of the novel ENaC/Deg heteromer in a glioma cell line, D54-MG. Our study postulates that a change in GBM cell electrophysiological properties after subunit knockdown would be indicative of that subunit being a part of the GBM channel. We have sequentially knocked down different ENaC/Deg subunits from the D54-MG glioma cells and measured amiloride-sensitive whole cell current using patch clamp. We found that knocking down various ENaC/ Deg subunits significantly reduced the whole cell patch clamp current in glioma cells and changed the resting Na ϩ /K ϩ permeability of the these cells. After subunit knockdown, glioma cells showed a reduced cell migration as compared with control cells, consistent with our hypothesis that ENaC/Deg subunits play an important role in glioma cell pathophysiology.

EXPERIMENTAL PROCEDURES
Cell Culture-Experiments were performed on a gliomaderived cell line D54-MG (GBM, derived from a World Health Organization grade IV tumor; a kind gift of Dr. D. Bigner, Duke University, Durham, NC), primary human astrocytes (a kind gift of Dr. Yancey Gillipsie, University of Alabama at Birmingham, Birmingham, AL), and on CHO-K1 cells. Cells were maintained in continuous cell culture in 50:50 Dulbecco's modified Eagle's medium/F-12 medium (Invitrogen) supplemented with 10% fetal bovine serum (Hyclone, Logan, UT) and 1% penicillin/streptomycin (Invitrogen). For electrophysiological recording, cells were split 48 h before recording onto 35-mm dishes containing flame-sterilized coverslips. In the case of transfection, cells were split 24 h before the transfection experiment and then allowed to grow for 24 -48 h before use.
Isolation of Cellular RNA-Total cellular RNA was isolated from D54-MG cells and primary human astrocytes using RNeasy (Qiagen), according to the manufacturer's recommendations. RNA isolated 48 h post-transfection from CHO-K1 cells transfected with cDNAs for human ASIC1, ␣ENaC, ␥ENaC, and ␦ENaC was used as a positive control. RNA concentration was calculated based on the absorbance at 260 nm. RNA samples were stored at Ϫ20°C.
Measurement of Relative mRNA Levels Using Real Time PCR-Real time PCR to measure ASIC1 (Hs00241630_m1), ␣ENaC (Hs00168906_m1), ␥ENaC (Hs00168918_m1), and ␦ENaC (Hs00161595_m1) mRNA was performed using TaqMan One-Step RT-PCR Master Mix Reagents (Applied Biosystems, catalog no. 4309169) using the manufacturer's handbook as a reference (Relative Quantification; Applied Biosystems 7300/7500 Real Time PCR system; 2004). 18 S rRNA (Hs99999901_m1) was amplified as an internal control and used as a reference.
Generation of Dominant Negative cDNAs-Site-directed mutagenesis was performed using QuikChange XL II mutagenesis kit (Stratagene) to engineer premature stop codons in human ASIC1, human ␥ENaC, human ␣ENaC, and human ␦ENaC cDNA. Briefly, wild type constructs of each subunit (previously ligated at the N terminus to the enhanced green fluorescent protein (eGFP) or to enhanced yellow fluorescent protein (eYFP) at the C-terminal expression vector and shown to form functional ion channels (36), were subjected to PCR with sense and antisense primers containing the specific base change to result in the desired codon mutations. For ASIC1, a stop codon was introduced at Tyr-67, for ␥ENaC at Ser-155, for ␣ENaC at Glu-34, and for ␦ENaC at Ser-35. The reaction mixture was subjected to the cycling parameters according to the manufacturer's handbook; 1 cycle at 95°C for 1 min followed by 18 cycles sequentially at 95°C for 50 s, 60°C for 50 s, and 68°C for 1 min/ kilobase of plasmid length and finally followed by 1 cycle at 68°C for 7 min. PCR products were transformed, and Escherichia coli were plated on Luria Bertani (LB) plates with kanamycin (30 g/ml) and incubated overnight at 37°C. Single colonies were grown on LB medium with 30 g/ml kanamycin and incubated overnight at 37°C with constant shaking. DNA was isolated using the Qiagen miniprep kit (Qiagen), and the presence of the mutation was confirmed by sequencing (Heflin Genetics Center, University of Alabama at Birmingham). The colonies were grown in 250 ml of LB medium with antibiotics, and DNA was isolated using the Qiagen Maxiprep kit (Qiagen).
Cell Transfections-For transfecting dominant negative cDNAs, electroporation using the Bio-Rad Gene Pulser X-cell electroporator (Bio-Rad) was used. Cells are detached from the flask with 0.1 mM EDTA in phosphate-buffered saline and suspended at 1.3 ϫ 10 7 cells/ml in RPMI 1640 ϩ 10 mM dextrose and 0.1 mM dithiothreitol in a 4-mm sterile cuvette (0.3 ml/cuvette). 5-10 g of cDNA was added to the cuvette. The cuvette was then pulsed using a time lapse protocol of a single square wave pulse of 300 V for 35 ms and 960 microfarads as per the manufacturer's protocol, and the cells were returned to 60-mm dish in the incubator for 24 -48 h before use.
Stable ASIC1-GFP Cell Line-Post-transfection with ASIC1 ligated to GFP as described above, D54-MG cells were treated with 500 g/ml G418 to select for transfected cells. After antibiotic selection, GFP-positive cells were sterile-sorted by the Center for AIDS Research Fluorescent Activated Cell Sorting (FACS) facility at University of Alabama at Birmingham.
Total Membrane and Plasma Membrane Isolation-D54-MG cells stably transfected with ASIC1-GFP were grown on 100-mm tissue culture dishes. Cells were allowed to grow until 90 -100% confluent and then scraped and collected by centrifugation at 600 ϫ g for 5 min at 4°C. Total membranes and plasma membranes were isolated using a plasma membrane protein extraction kit (BioVision) according to the manufacturer's protocol. Total membrane and plasma membrane were dissolved in lysis buffer containing 20 mM Tris HCl, pH 8, 137 mM NaCl, 10% glycerol, 2 mM EDTA, 1% w/v CHAPS, and protease inhibitor (Complete, Roche Applied Science). Total membrane and plasma membrane protein concentration was measured using the BCA protein assay kit (Pierce).
Immunoblotting and Co-immunoprecipitation-For immunoblotting, cell lysates were assayed for protein concentration (BCA, Pierce), and then 50 g of protein was loaded in 2ϫ SDS sample buffer containing 5% ␤-mercaptoethanol into the wells of a 6 -8% SDS-PAGE gel (4% stacking gel). The gels were transferred onto polyvinylidene difluoride membranes and blocked with 5% nonfat dry milk in TBST buffer (100 mM Tris, pH 7.5, 150 mM NaCl, 0.1% Tween 20) for 1 h and then incubated overnight at 4°C with primary antibody. The blots were developed with horseradish peroxidase-conjugated secondary antibodies and enhanced chemiluminescence (SuperSignal West Pico, Pierce) and autoradiography. Band densities were assessed using Scanalytic software.
In co-immunoprecipitation experiments total membrane and plasma membrane fractions of D54-MG stably transfected with ASIC1-GFP was used. For control experiments CHO-K1 cells were transiently transfected with cDNA for ASIC1-GFP, and CHO-K1 and D54-MG cells were transiently transfected with CFP-CLC1. Cells were lysed using lysis buffer with 1% CHAPS and protease inhibitors (Complete, Roche Applied Science). 2 g of primary antibody was added to 500 g of cell lysate and incubated overnight at 4°C. 100 l of protein A-agarose beads were added and allowed to incubate for 4 h. The beads were then pelleted and washed 3-4 times with lysis buffer. To elute the immune complex, 50 l of elution buffer (0.1 M glycine-HCl, pH 2.5-3.0) was added to the beads and allowed to incubate on ice for 5 min and centrifuged at 2500 ϫ g for 1-3 min at 4°C. Supernatant was collected, and the pH was adjusted to phys-iological pH by adding 10 l of 1 M Tris. 10 -25 l of loading buffer was added to the eluate and loaded onto a 6% SDS gel. Immunoblotting was carried out as described above using the appropriate primary and secondary antibodies. All buffers used for preparing cell lysates, membranes, and subcellular fractions included protease inhibitors (Complete, Roche Applied Science).
Whole Cell Patch Clamp Recordings-Amiloride-sensitive whole cell currents were recorded in transfected and untransfected glioma cells. Cells were cultured on flame-sterilized coverslips and transfected 48 h before patch-clamping. Patch pipettes were made of thin-walled borosilicate glass using a P-97 Flaming/Browning Micropipette puller (Sutter Instruments) and typically had a resistance of 5-10 megaohms. Pipettes were filled with an electrolyte solution containing 100 mM potassium gluconate, 30 mM KCl, 10 mM NaCl, 20 mM HEPES, 0.5 mM EGTA, 4 mM ATP, pH 7.2. Current recordings were low pass-filtered at 2 kHz and digitized online at 10 kHz with a Digidata 1200 (Axon Instruments). pClamp 9.0 (Axon Instruments) was used to acquire and store data. Our standard bath solution consisted of 125 mM NaCl, 5 mM KCl, 1.2 mM MgSO 4 , 1 mM CaCl 2 , 1.6 mM Na 2 HPO 4 , 0.4 mM NaH 2 PO 4 , 10.5 mM glucose, and 32.5 mM HEPES acid; pH was adjusted to 7.4 with NaOH. The whole cell patch configuration was established as described previously (35). Currents were recorded by holding the cell at Ϫ60 mV and clamped sequentially to membrane potentials between Ϫ100 mV and ϩ80 mV in 20-mV increments.
To characterize the ion selectivity of the whole cell currents and calculate the relative permeability for Na ϩ and K ϩ , bath solutions were used as described previously (35), i.e. solutions A and B that contained Na ϩ and K ϩ , respectively, as the major cation. Bath solution A contained 140 mM NaCl, 5 mM KCl, 1 mM CaCl 2 , 1 mM MgCl 2 , 10 mM HEPES, and 5 mM dextrose. Bath solution B contained 140 mM KCl, 1 mM CaCl 2 , 1 mM MgCl 2 , 10 mM HEPES, and 5 mM dextrose. The pH of the solutions was adjusted with NaOH or KOH to 7.5. The solution of the entire chamber was perfused, whereas the cell was still in whole cell patch clamp mode.
Migration Assay-D54-MG cells were transfected with YFP-tagged dominant negative cDNAs for ASIC1, ␣ENaC, ␥ENaC, or with eYFP for the control group using electroporation as described above. Cells were transfected for 48 h before use and then detached (phosphate-buffered saline ϩ 1 mM EDTA) and counted using a hemocytometer. 50,000 YFP-labeled cells transfected with either test or control cDNAs were replated onto high throughput screening Fluoroblok TM (BD Bioscience) inserts coated on the underside with 5 g/ml vitronectin. High throughput screening Fluoroblok inserts have a special polyethylene terephthalate track-etched polymer coating on the membrane which blocks the passage of light through the membrane so only fluorescent cells that have migrated through the filter pores (8-m diameter) were visualized. Cells were allowed to migrate for 24 h and fixed with 4% paraformaldehyde and visualized using fluorescence. Because GFP and YFP fluorophores have overlapping emission spectra, YFP-tagged cells can be visualized using a fluorescein isothiocyanate filter.
YFP-expressing cells that had migrated through the pores to the other side of the filter were counted in five random fields per filter using a Nikon TE200V microscope equipped for epifluorescence. Three filters equal an n of 1.
Statistics-All experiments were repeated a minimum of three times. Differences between groups were assessed using standard statistical tests, primarily Student's t test. Data are presented as the means Ϯ 1 S.D. Significance was set at p Ͻ 0.05.

D54-MG Cells Show a Higher Expression of ASIC1, ␣ENaC, and ␥ENaC mRNA Compared with Primary Human Astrocytes-
We have previously shown that high grade gliomas such as GBMs express different subunits of the ENaC/Deg family (3). To see if a quantitative difference exists between the expression of ASIC1, ␣ENaC, ␥ENaC, and ␦ENaC mRNA levels in glioma cells and primary human astrocytes, we performed quantitative real time PCR on total mRNA obtained from D54-MG glioma cells and primary human astrocytes. Real time PCR was performed using Taq-Man One Step master mix (Applied Biosystems). 18 S rRNA was amplified as the internal control. Relative levels of mRNA for ASIC1, ␣ENaC, ␥ENaC, and ␦ENaC proportional to 18 S rRNA were compared between D54-MG cells and primary human astrocytes. We found that D54-MG cells have a significantly higher expression of ASIC1 (Fig. 1A), ␣ENaC (Fig. 1B), and ␥ENaC ( Fig. 1C) mRNA level compared with primary human astrocytes. ␦ENaC mRNA was not detected in either D54-MG cells or primary human astrocytes, whereas we could easily detect it in ␦ENaC-transfected CHO-K1 cells (data not shown). Even though we did not detect ␦ENaC mRNA in either astrocytes or in D54-MG cells, we did detect the protein expression in both the cell types (see Fig. 6A, primary astrocytes data are not shown). Therefore, we do not rule out that very low expression of ␦ENaC mRNA might occur in both primary human astrocytes and D54-MG cells, which was undetected using standard semiquantitative RT-PCR. Also it is possible that ␦ENaC mRNA may have a short half-life and is quickly degraded in situ. Among ASIC1, ␣ENaC, and ␥ENaC mRNA expression in D54-MG cells, we found ASIC1 levels to be severalfold higher than ␣ENaC or ␥ENaC levels, suggesting that the channel responsible for the amiloride-sensitive current seen in glioma cells is composed of ASIC1 subunit at the core of the channel in association with ENaC subunits.
Amiloride Inhibits the Whole Cell Current in the D54-MG Glioma Cell Line-We have previously shown that high grade glioma cells display a basally active cation current that is inhibited by the diuretic drug amiloride (3). This current is not present in low grade gliomas or normal astrocytes. Because we used D54-MG gliomaderived cell lines in our experiments, we wanted to confirm the presence of an amiloride-sensitive cation current in our cell line using whole cell patch clamp experiments.
D54-MG cells were whole cell patch-clamped, and the presence of a large inward current was detected. This inward current was significantly inhibited upon superfusion of 100 M amiloride. Representative traces are shown in Fig. 2A (before) and Fig. 2B (after the addition of 100 M amiloride). We next wanted to determine the IC 50 for amiloride in D54-MG cells. Conductance was measured as the chord between a clamp voltage of Ϫ160 mV and the reversal potential. The cells were then exposed to amiloride concentrations of 10, 30, 100, and 300 M. The conductance decrease was calculated as the conductance before amiloride minus the conductance after amiloride divided by the conductance before amiloride. Fig. 2D shows the average conductance decrease at each concentration. Fig. 2E shows the average conductance converted to a percentage of the maximum average reduction recorded in cells exposed to 300 M. This plot was used to calculate a K i for amiloride of 27.5 M.
In our previous publication we have shown that application of Psalmotoxin 1 completely eliminates glioma current (35), whereas there was a large amiloride-insensitive current seen with application of 100 M amiloride in our present study. To look for the effect of a higher concentration of amiloride on the D54-MG glioma current, we patch-clamped D54-MG cells before and after application of 1 mM amiloride (Fig. 2, F-H). We found that application of 1 mM amiloride further inhibited D54-MG whole cell current, and this inhibition was significantly more than the inhibition after application of 100 M amiloride (Fig. 2H). The fact that there is a further inhibition  of glioma cell whole cell current upon application of 1 mM amiloride after initial saturation of the inhibition at 100 M suggests that the glioma channel shows a bimodal inhibitory curve to amiloride.
To confirm that the current seen in D54-MG cells was sensitive to Psalmotoxin 1, we patch-clamped D54-MG cells before and after application of 10 nM PcTX-1. Fig. 3A shows the representative trace of the basal conductance of the cell. Fig. 3B shows that after exposure to 10 nM PcTx-1, the inward currents were abolished, indicating a complete inhibition of the inward sodium conductance. Fig. 3C shows the corresponding I/V curve. The complete elimination of current upon application of 1 mM amiloride or 10 nM PcTX-1 indicates that channel responsible for glioma current is made up of ENaC/Deg subunits and also that ASIC1 constitutes the central core of the channel.
Effect of Knocking Down ASIC1, ␥ENaC, and ␣ENaC on Glioma Whole Cell Current-Because we hypothesize that ENaC/ Deg subunits underlie the basally active amiloride-sensitive current seen in glioma cells, we next examined the effect of knocking down the various ENaC/Deg subunits upon glioma cell whole cell current. To determine whetherASIC1 is involved in the formation of the glioma ion channel, we knocked down ASIC1 using dominant negative (DN) mutation. We engineered a premature stop codon at Tyr-67 of ASIC1 cDNA and transfected the mutated cDNA into D54-MG glioma cells by electroporation. An equivalent construct has been used successfully to knockdown ␥ENaC in murine vascular smooth muscle cells (38). Fig. 4A shows the schematics of the ASIC1 DN construct. All the DN constructs were coupled to eGFP to facilitate identification of transfected cells for patch clamp recording. We have previously shown that the addition of the fluorescent tag has no effect on the whole cell current (36). Fig.  4B1 is a representative Western blot for ASIC1 protein expression in lysates from untreated D54-MG cells and in D54-MG cells transfected with the DN-ASIC1 cDNA construct. Quantification of Western blots shows that knocking down ASIC1 using the dominant negative mutation inhibited ASIC1 protein expression in D54-MG cells by 50 -60% (Fig. 4B2). To confirm for the specificity of dominant negative mutation in knocking down the protein of interest, we used cell lysates from DN-ASIC1-transfected D54-MG cells and blotted for ␥ENaC (Fig. 4C1). We found no difference in the expression of ␥ENaC in D54-MG cells and in D54-MG cells with ASIC1 knockdown (Fig. 4C2), confirming the specificity of dominant negative mutation. We next looked at the effect of knocking down ASIC1 on D54-MG whole cell current. Cells with visible green fluorescence were considered as transfected with dominant negative construct. Cells which were electroporated but did not exhibit fluorescence were considered as untransfected and used as control cells in our whole cell patch clamp studies. As additional controls, we used untransfected D54-MG cells and D54-MG cells transfected with eGFP or eYFP alone. We found no differences between the amiloride-sensitive whole cell currents in any of these control cells (comparative data not shown). Fig. 4D shows representative traces of whole cell patch clamp recording showing a typical whole cell current that is inhibited by the addition of 100 M amiloride to the bath solution in control cells (top panels) and marked by inhibition of current after ASIC1 knockdown (lower panels). Fig. 4E is an I/V curve showing significant reduction of whole cell patch clamp current after the addition of 100 M amiloride or knocking down ASIC1. Fig. 4F shows average maximal inward current at Ϫ100 mV for each of the conditions as in Fig. 4D. We found that knocking down ASIC1 completely abolished D54-MG whole cell current, and this was consistent with the inhibition seen after application of 1 mM amiloride (Fig. 2F) or 10 nM PcTX-1 (Fig. 3), suggesting that ASIC1 seems to form a key conduction element of the glioma channel.
We next wanted to check for the effect of knocking down other ENaC/Deg subunits on glioma whole cell current. Because we found higher mRNA expression for ␥ENaC and ␣ENaC in D54-MG cells compared with normal human astrocytes, we wanted to look for the effect that knocking down these FIGURE 3. Psalmotoxin 1 completely abolishes D54-MG whole cell current. Whole cell currents recorded from a D54-MG glioblastoma cell with normal intracellular and extracellular ionic concentrations for Na ϩ , K ϩ , Cl Ϫ , Ca 2ϩ , and pH. The voltage clamp potentials ranged from Ϫ160 to ϩ40 mV. The large negative potentials were used to maximize the inward currents. A shows the basal conductance of the cell. B shows that after exposure to 10 nM PcTx-1, the inward currents were abolished, indicating a complete inhibition of the inward sodium conductance. C, corresponding I/V curve; n ϭ 3. two subunits would have on glioma cell current. We knocked down ␥ENaC by constructing a dominant negative cDNA for ␥ENaC (eYFP-S155X) similar to the DN-ASIC1 construct and transfecting the mutated cDNA in D54-MG cells (Fig. 5A). Western blot for ␥ENaC showed a 60 -70% reduction in protein expression in D54-MG cells transfected with DN-␥ENaC compared with untransfected D54-MG cells (Fig. 5, B1 and B2). To check for the specificity of the dominant negative mutation in knocking down only ␥ENaC, we probed for the expression of ASIC1 (Fig. 5C1) and ␣ENaC (Fig. 5D1) protein levels in lysates from untransfected and transfected D54-MG cells and found no difference in the expression level of either ASIC1 (Fig. 5C2) or ␣ENaC (Fig. 5D2).
To study the effect of knocking down ␥ENaC on the glioma cell current, we patch-clamped control and ␥ENaC knocked-down D54-MG cells. D54-MG control cells exhibited a whole cell current that was inhibited by 100 M amiloride (Fig. 5E, representative traces, top panel). The amiloride-sensitive current seen in D54-MG cells was significantly inhibited after knocking down ␥ENaC in transfected cells as compared with control cells (representative traces, lower panel). Fig. 5F shows the corresponding I/V curve. Fig. 5G shows the average maximal inward current at Ϫ100 mV for each of the conditions illustrated in Fig. 5E.
We next knocked down ␣ENaC in D54-MG cells using a dominant negative cDNA for ␣ENaC (eYFP-E34X) (Fig. 6A) and found a 60% reduction of ␣ENaC protein levels as compared with control D54-MG cells, as shown by Western blot (Fig. 6, B1 and B2). Knocking down ␣ENaC did not change the expression of ASIC1 protein level (Fig. 6, C1 and C2), again confirming the specificity of the dominant negative mutation in knocking down the protein of interest. Similar to what we saw for ␥ENaC, knocking down ␣ENaC significantly inhibited the amiloride-sensitive whole cell current seen in D54-MG glioma cells (Fig. 6, D-F). Knocking down either ␥ENaC or ␣ENaC reduced the whole cell patch clamp current seen in D54-MG cells to a larger extent than after the addition of 100 M amiloride. Because knocking down either ␥ENaC or ␣ENaC significantly reduced the glioma cell conductance, both ␥ENaC and ␣ENaC seem to form key elements of the glioma channel.
Knocking Down ␦ENaC Does Not Change Glioma Cell Whole Cell Current-Although we did not detect ␦ENaC mRNA expression levels in GBM cells or in normal astrocytes as the protein was present, we wanted to test whether knocking down ␦ENaC would have an effect upon GBM whole cell currents. eYFP-DN-␦ENaC (S35X) cDNA was transfected into D54-MG glioma cells (Fig. 7A shows the schematics of DN-␦ENaC construct), and cells were whole cell patchclamped. We saw a 50% reduction in protein expression for ␦ENaC in cells transfected with the DN construct compared with untransfected control cells as shown by Western blot (Fig. 7, B1 and B2). Even though ␦ENaC protein expression was reduced in cells transfected with DN ␦ENaC cDNA, no change in whole cell patch clamp currents between control cells and cells with ␦ENaC knockdown was observed. Fig. 7C shows representative traces before (top panel) and after (lower panel) knocking down ␦ENaC in D54-MG cells. Fig. 7, D and E, show the corresponding I/V curve and average inward current at Ϫ100 mV, respec-tively, for each of the conditions shown in Fig. 7C. Because knocking down ␦ENaC causes no significant change in GBM whole cell current, we conclude that even though ␦ENaC is expressed in glioma cells, ␦ENaC is not a part of the glioma channel complex.
Glioma Cells Have a Higher K ϩ Selectivity over Na ϩ That Is Abolished after Knocking Down ␥ENaC and ␣ENaC-In a previous publication we showed that high grade glioma cells have a higher ionic selectivity for K ϩ over Na ϩ (35). In that study cells from three different high grade glioma cell lines (SK-MG, U251-MG, and U87-MG) were whole cell patch-clamped and, whereas in the patch clamp configuration, the bath solution was changed sequentially by perfusing the entire chamber with solutions containing Li ϩ , Ca 2ϩ , or K ϩ as the major cation instead of Na ϩ . In every case when Na ϩ was substituted with K ϩ as the major cation, a large increase in amiloride-sensitive inward current was observed. Using similar experimental conditions, we found a significant large increase in inward current in D54-MG cells when Na ϩ was substituted with K ϩ (Fig. 8). This increase in inward current was abolished in D54-MG cells in which ␥ENaC or ␣ENaC were knocked down. Cells with ␦ENaC knockdown showed a significant increase in inward current upon substituting K ϩ as the major cation similar to the effect seen in control cells. These data further suggest that knocking down ␥ENaC or ␣ENaC changes important electrophysiological properties of glioma cells in addition to reducing the whole cell current, whereas cells with ␦ENaC knocked down behave similarly to the control cells.
Co-immunoprecipitation of ASIC1 and ENaC Subunits in Glioma Cells-Our hypothesis is that different ENaC/Deg subunits interact with each other to form a channel complex in glioma cells, giving these cells their unique properties. To probe for this interaction we used co-immunoprecipitation to look for an interaction between ASIC1, ␥ENaC, and ␣ENaC in D54-MG cells. Endogenous glioma cells have a relatively low level of ENaC/Deg protein expression. To overcome this limitation, we overexpressed one of the putative subunits, ASIC1, and epitope-tagged it to allow for more sensitive immunoprecipitation and immunodetection. Total cell membrane and plasma membrane fractions of D54-MG cells stably transfected with ASIC1-GFP were obtained using the methods detailed above. Lysate from CHO-K1 cells transfected with ASIC1-GFP served as positive controls. Both in CHO-K1 (Fig. 9A) cells and in total membrane (Fig. 9B) and plasma membrane (Fig. 9C), isolates of D54-MG cells immunoprecipitating with GFP or ASIC1 antibodies and blotting with GFP pulled down GFP-ASIC1 at 100 kDa, consistent with the expected size of GFP tagged ASIC1. Furthermore, in both the membrane fractions of ASIC1-GFP-overexpressing D54-MG cells, immunoprecipitating with ␣ENaC or ␥ENaC antibodies and blotting with GFP also resulted in pulling down GFP-ASIC1 at 100 kDa, suggesting an interaction between ASIC1 and ␣ENaC and between ASIC1 and ␥ENaC in D54-MG glioma cells (Fig. 9, B and C).
One of the inherent limitations of overexpression studies is the occurrence of random or nonspecific associations. To rule out random association resulting from overexpression as a cause of ASIC1/ENaC interaction, we transfected D54-MG cells and CHO-K1 cells with cDNA for sarcolemmal Cl Ϫ channel, CLC1- The corresponding I/V curve (F) and average conductance at Ϫ100 mV (G) showing a significant inhibition of inward current after the addition of 100 M amiloride or after knocking down ␥ENaC. Control traces are the same as in Fig. 3; for ␥ENaC knockdown, n ϭ 6. Error bars Ϯ1 S.D. The asterisk indicates p Ͻ 0.05.
CFP, a protein unrelated to the ASIC/ENaC family, and looked for co-immunoprecipitation as a control for random association. Both in CLC1-CFP-transfected CHO-K1 cells and D54-MG cells, precipitating with GFP antibody and blotting for GFP resulted in pulling down CFP-tagged CLC1 but precipitating with ASIC1 antibody, and blotting for GFP did not show any interaction between ASIC1 and CLC1 (Fig. 9D), suggesting that the interaction between ASIC1 and ENaC subunits in glioma cells was not because of a random association resulting from overexpression.
We found that even though we had not transfected D54-MG cells with either ␣ENaC or ␥ENaC, both these subunits immunoprecipitated with ASIC1 in transfected D54-MG cells. This result combined with the electrophysiological data showing that knocking down these three subunits significantly reduces glioma cell whole cell current strongly suggests that these subunits come together to form a functional ion channel in glioma cells and that the active channel in the glioma cell contains (at the minimum) ASIC1, ␣ENaC, and ␥ENaC.
Knocking Down ASIC1, ␣ENaC, or ␥ENaC Inhibits GBM Cell Migration-Because ENaC/Deg family members are hypothesized to play an important role in glioma cell biology, we predict that knocking down the different ENaC/Deg subunits would have an inhibitory effect upon glioma cell migration. To study the migration properties of glioma cells before and after knocking down the different subunits, we used a Transwell migration assay using high throughput screening Fluoroblok inserts with an 8-m pore size (BD Bioscience). To allow for sufficient time for cell migration, we performed a 24-h migration assay. Because we found 24 h application of 100 M amiloride was toxic to cells, we substituted 100 M benzamil (an analogue of amiloride which inhibits D54-MG whole cell currents similarly to amiloride; supplemental Fig. 1) instead of amiloride in the migration assay study. Fig. 10 shows representative images of migrated D54-MG cells transfected with eYFP (Fig. 10A), transfected D54-MG cells with 100 M benzamil in the migration buffer (Fig. 10B), and migrated D54-MG cells with ASIC1 knockdown (Fig. 10C), ␣ENaC knockdown (Fig. 10D), or ␥ENaC knockdown (Fig. 10E). Fig. 10F shows the average number of cells counted per field for each of the conditions. We found that 100 M benzamil or knocking down ASIC1, ␣ENaC, or ␥ENaC had a significant inhibitory effect upon D54-MG cell migration, suggesting that ASIC1 and ENaC subunits play a role in glioma cell migration.

DISCUSSION
In our recent publication we showed that ASIC and ENaC subunits are capable of forming cross-clade heteromers (36). We found that expression of combinations of ASIC1 with ENaC subunits exhibited novel electrophysiological characteristics as compared with ASIC1 alone. That study suggested that heteromeric complexes of ASIC and ENaC subunits may contribute to the diversity of amiloride-sensitive cation conductance observed in a wide variety of tissues and cell types where co-expression of ASIC and ENaC subunits has been observed (36) in addition to differences due to posttranslational modification, e.g. phosphorylation (39 -43), carboxymethylation (44,45), and protease clipping (46,47). In this study we tested the hypothesis that the glioma cation current is comprised of a unique mixture of subunits from the ENaC/Degenerin superfamily of ion channels. Our aim was to assess the role of these different ENaC/Deg subunits in glioma biology and to probe the composition of these novel ENaC/Deg heteromers.
We used a dominant negative mutation strategy to knockdown ASIC1 and ENaC subunits, similar to the approach used by Jernigan and Drummond (38). A similar approach had been used to make N-terminal-truncated rat ␣-, ␤-, and ␥ENaC subunits, and it was shown to reduce I Na when co-expressed with wild type rat ␣-, ␤-, and ␥ENaC (4). Although the mechanism of action of dominant negative mutation is not fully understood, some possibilities are that the mutated subunit when overexpressed in cells displaces the wild type subunit from ion channel complexes and forms a non-functional ion channel (4,38). They also cause degradation of wild type subunits, as a reduction of the protein level is seen in dominant negative transfected cells. The loss of channel function can be explained by the fact that the mutated subunits are not full-length proteins and most likely never get folded into a functional subunit, are unstable, and therefore, are degraded (38,48). Ϫ100 mV for each condition as in Fig. 6B, showing similar inward current for both D54-MG control cells and in D54-MG cells with ␦ENaC knockdown. Amiloride significantly inhibited the inward current under both the conditions. Control traces are the same as in Fig. 3; for ␦ENaC knockdown, traces n ϭ 5. Error bars are Ϯ1 S.D. The asterisk indicates p Ͻ 0.05. FIGURE 8. The higher K ؉ selectivity of glioma cells is abolished after knocking down ␥ENaC and ␣ENaC. A significant large increase in inward current was seen in D54-MG cells when Na ϩ was substituted with K ϩ as the major cation. This increase in inward current was abolished in D54-MG cells in which ␥ENaC or ␣ENaC were knocked down. Cells with ␦ENaC knockdown showed a significant increase in inward current upon substituting K ϩ as the major cation similar to the effect seen in control cells. n ϭ 4 for each condition except for ␥ENaC DN, where n ϭ 6. Error bars are Ϯ1 S.D. The asterisk indicates p Ͻ 0.05. FIGURE 9. ENaC/Deg subunits interact with each other in glioma cells. A, lysate from CHO-K1 cells transfected with GFP ligated ASIC1 cDNA was immunoprecipitated (IP) with mouse anti-GFP or rabbit anti-ASIC1 antibody and blotted with mouse anti-GFP antibody. The Western blot shows that immunoprecipitating with either mouse anti-GFP or rabbit anti-ASIC1 antibody and blotting with mouse anti-GFP antibody pulled down GFP-ASIC1 at 100 kDa. Western blots of total membrane fractions (B) and plasma membrane fractions (C) isolated from D54-MG cells stably transfected with ASIC1-GFP and immunoprecipitated with mouse anti-GFP, rabbit anti-ASIC1, rabbit anti-␣ENaC, or rabbit anti-␥ENaC antibody show an interaction of ASIC1 with ␣ENaC and ␥ENaC in D54-MG cells. Non-immune mouse IgG immunoprecipitation and immunoprecipitation with only protein A-agarose beads showed that the anti-GFP antibody was specific. D, to rule out that ASIC1 and ENaC subunit interactions were because of a random association due to overexpression, CHO-K1 and D54-MG cells were transfected with an unrelated protein; CFP ligated CLC1. Lysates from both CHO-K1 and D54-MG cells immunoprecipitated (IP) with mouse anti-GFP antibody pulled down CFP-CLC1 upon blotting with mouse anti-GFP antibody, whereas immunoprecipitating D54-MG cell lysate with rabbit anti-ASIC1 antibody did not pull down CLC1 upon blotting with mouse anti-GFP antibody. n ϭ 3 for all the blots.
It has been shown that properties of ENaC/Deg channels vary with subunit composition (49). Comparing the properties of ENaC and ASIC channels with those of glioma conductance, the channel properties found in GBM cells are intermediate between those of ENaCs and ASICs, and we hypothesized that GBM channels are composed of a novel mixture of subunits of the ENaC/Deg family, giving the channels their unique electrophysiological characteristics. ENaC channels are highly sensitive to inhibition by amiloride with a K i amiloride for ␣␤␥ENaC of 0.1 M and a higher Na ϩ permeability than K ϩ with the Na ϩ /K ϩ selectivity being Ͼ50 (5). They are not inhibited by PcTX1 and are not acid-activated with the exception of ␦␤␥ENaC (50). ASIC channels, on the other hand, require a higher concentration of amiloride for inhibition; the K i amiloride for ASIC1 is 10 M and for ASIC2 is 26 M, and P Na ϩ /P K ϩ is 3 and 10 for ASIC1 and ASIC2, respectively (19). Both ASIC1 and ASIC2 are activated by low extracellular pH (19), and ASIC1 is highly sensitive to inhibition by PcTX-1 with a K i of 0.9 nM (29). The channels responsible for GBM current have a K i amiloride of 27.5 M (Fig. 2E), which is closer to that of ASICs than it is to ENaCs, and also like ASIC1, these channels are inhibited by PcTX-1 with a K i of 0.4 nM (35). But unlike ASICs, GBM cell currents are not acid-activated. Unlike both ENaCs and ASICs, the GBM conductance is more permeable to K ϩ than Na ϩ (35). Other transporters inhibited by amiloride are the Na ϩ /H ϩ exchanger (NHE) (51), but the fact that the NHE is a 1:1 exchanger for Na ϩ and H ϩ makes it very unlikely that the NHE plays a role in the amiloride-sensitive current seen in GBM cells. Also it has been shown that action of amiloride on inhibition of malignant glioma cell proliferation is independent of the Na ϩ /H ϩ exchanger (52). Typical ENaC channels are composed of ␣, ␤, and ␥ ENaC subunits and are sensitive to amiloride at submicromolar concentrations (14). Apart from these typical ENaC channel, atypical ENaC channels are present which show lower amiloride sensitivity (53,54) and have a conductance which is only 40 -50% amiloride-sensitive. These atypical channels include ␣-only channels or channels with ␦ENaC subunits (54). As we show in our study, amiloride-sensitive channels in GBM cells are composed of (at the minimum) ASIC1, ␣ENaC, and ␥ENaC subunits with lower sensitivity to amiloride, making them similar to the atypical ENaC channels. Because we hypothesize that ASIC1 forms the core of the channel in GBM cells, the bimodal inhibitory curve for amiloride that we saw could be the result of amiloride binding to different binding sites of ASIC1, with varying binding energy. It is also possible that there are multiple populations of channels with various posttranslational modifications, such as differential cleavage by proteases, which have similar gross electrophysiological characteristics but variations in amiloride binding. Detailed singlechannel analysis and kinetic modeling would be required to test this hypothesis, but it is consistent with the presented data.
We show that there is a higher expression of ASIC1, ␣ENaC, and ␥ENaC mRNA in D54-MG glioma cells compared with normal human astrocytes. It was recently shown that the higher tissue acidification seen in an experimental autoimmune encephalomyelitis (EAE) mouse model of multiple sclerosis increased the expression and function of ASIC1 in EAE mice brains (55). Although we did not test for tissue acidification in GBMs, we predict that glioma cells growing at a rapid rate would lead to tissue acidosis and a higher expression of ASIC1 similar to what was seen in experimental autoimmune encephalomyelitis. We also show a higher expression of ␣ENaC and ␥ENaC in GBM cells. Proteolytic cleavage of both these subunits is required to maximally activate ENaC channels (56). Even though we showed that knocking down ASIC1 significantly inhibited GBM whole cell current, glioma cells do not exhibit the typical acid induced ASIC1 current. One potential reason for the lack of acid activation may be that the channel is already maximally activated, and therefore, a further change in [H ϩ ] does not exert an additional stimulus. Another possibility is that the proton-sensitive region is masked by the presence of additional ENaC subunits.
One of the important prognostic factors in treatment of glioblastoma is secondary metastases, as these tumors are known to infiltrate widely throughout the brain. It has been shown that for cell migration to occur, the leading edge of the cell or lamellipodium needs to expand; movement of Na ϩ and H 2 O to the cell interior plays an important role in this process (57). We have shown previously that both amiloride and PcTX1 inhibits glioma cell swelling (58). We have also shown that amiloride inhibits glioma cell migration. In this study we show that knocking down ASIC1, ␣ENaC, or ␥ENaC inhibits glioma cell migration, possibly because of the inhibitory effect upon Na ϩ influx after channel knockdown and the corresponding inhibition of cell swelling required for lamellipodium expansion. Although much work is needed to fully understand the ENaC/ASIC cross-clade interactions in glioma cells and how these channels relate to each other on a biochemical and molecular level, this study represents the first evidence to show the presence of an endogenous functional amiloride-sensitive Na ϩ channel that is a ENaC/ASIC heteromer.