Wolframin Expression Induces Novel Ion Channel Activity in Endoplasmic Reticulum Membranes and Increases Intracellular Calcium*

Wolfram syndrome is an autosomal recessive neuro-degenerative disorder associated with juvenile onset non-autoimmune diabetes mellitus and progressive optic atrophy. The disease has been attributed to mutations in the WFS1 gene, which codes for a protein predicted to possess 9–10 transmembrane segments. Little is known concerning the function of the WFS1 protein (wolframin). Endoglycosidase H digestion, immunocytochemistry, and subcellular fractionation studies all indicated that wolframin is localized to the endoplasmic reticulum in rat brain hippocampus and rat pancreatic islet β-cells, and after ectopic expression in Xenopus oocytes. Reconstitution of wolframin from oocyte membranes into planar lipid bilayers demonstrated that the protein induced a large cation-selective ion channel that was blocked by Mg2+ or Ca2+. Inositol triphosphate was capable of activating channels in the fused bilayers that were similar to channel components induced by wolframin expression. Expression of wolframin also increased cytosolic calcium levels in oocytes. Wolframin thus appears to be important in the regulation of intracellular Ca2+ homeostasis. Disruption of this function may place cells at risk to suffer inappropriate death decisions, thus accounting for the progressive β-cell loss and neuronal degeneration associated with the disease.

protein (wolframin) possessing 9 -10 predicted transmembrane segments (6). However, no clear sequence similarities have been identified that might shed light on the function of the protein (7). Northern blot analysis has revealed prominent expression of wolframin mRNA in affected tissues, including the brain and pancreas (1,7). These expression sites correlate well with the atrophic changes associated with the syndrome.
It was suggested that wolframin may be a mitochondrial protein (8), because many of the clinical features characteristic of this syndrome are similar to defects in oxidative phosphorylation that are seen in mitochondrial diseases, such as MELAS (mitochondrial encephalomyopathy, lactic acidosis, and strokelike symptoms), Leber's hereditary optic neuropathy, and maternally inherited diabetes and deafness (mitochondrial diabetes). Supporting this hypothesis, deletions of mitochondrial DNA have been reported in some Wolfram patients (9). However, recent reports have provided direct evidence to refute this hypothesis (10,11). Various homozygous or compound heterozygous mutations, including nonsense, frameshift, deletion, insertion, and missense mutations were identified in Wolfram patients without revealing a clear relation between these mutations and phenotype (12). Although Wolfram syndrome is rare, heterozygous carriers have been reported as 26 times more likely to require hospitalization for depression and psychiatric illness, leading to the hypothesis that heterozygosity for WFS1 gene mutations may be a significant cause of psychiatric illness in the general population (13). An increased prevalence of diabetes mellitus in first degree relatives of Wolfram patients has also been reported (14). Recently, several single amino acid mutations associated with Wolfram's disease have been associated with type 2 diabetes in non-Wolfram patients (2,5).
A recent report has shown that wolframin is localized to the ER when overexpressed in cultured fibroblasts (11). This localization suggests that wolframin may play an as yet undefined role in membrane trafficking, secretion, processing, and/or regulation of ER calcium homeostasis. Rapid but controlled redistribution of intracellular calcium is central in signaling, secretion, and apoptosis in all eukaryotic cells (15)(16)(17)(18)(19)(20)(21)(22). The unfolded protein response and calcium regulation are well defined ER functions that have been shown to play prominent roles in cellular apoptosis (23)(24)(25). Calcium release and uptake by the ER are fundamental to calcium regulation but are thus far only partially understood (25).
Based on the pathophysiology of the syndrome, it appears reasonable that wolframin may be somehow involved in the survival pathways of pancreatic islet ␤-cells and neurons (26). The presence of 9 -10 predicted transmembrane segments in wolframin suggested that it might function in the transport of ions or other small molecules across membranes. In the present study, we conducted a series of experiments to test this hypothesis. Immunohistochemical studies of the rat pancreas confirmed that wolframin is primarily localized in the islet ␤-cells, present at low levels in ␣-cells and absent from the exocrine tissue. We demonstrate that ectopic expression of wolframin in Xenopus oocytes results in its strict localization to the endoplasmic reticulum, consistent with the subcellular localization of the native protein in mammalian cells. Expression of wolframin in oocytes produced an increase in cytosolic Ca 2ϩ and induced novel cation-selective channel activities in ER membranes. Given that Ca 2ϩ regulation by the ER is prominent in cellular apoptosis, these data suggest that wolframin may be involved in the regulation of ER-mediated cellular mortality and that a defect in this regulatory process is the cause of the progressive ␤-cell loss and neuronal degeneration associated with the disease.

Preparation of Anti-WFS1 Antiserum and Western Blot Analysis
Antibody directed against WFS1 was generated using a synthetic peptide corresponding to the 20 N-terminal amino acids of WFS1 coupled to keyhole limpet hemocyanin via an N-terminal cysteine residue. Immunization of rabbits and production of antiserum was conducted by Charles River Laboratories (Southbridge, MA) according to their standard protocols. IgG was purified from the rabbit antiserum using protein A-agarose. SDS-PAGE and immunoblotting were carried out after transfer of proteins to nitrocellulose membranes (MSI, Westborough, MA). 125 I-Labeled goat anti-rabbit or anti-mouse IgG (at 1:500 dilution, Amersham Biosciences) was used as the secondary antibody. Signals were visualized by autoradiography.

Expression of Human WFS1 in Xenopus Oocytes
Full-length human WFS1 cDNA was subcloned into the oocyte expression vector pSP64T. Xenopus laevis or Xenopus tropicalis oocytes were isolated and prepared for injection according to published methods (27). Briefly, stage V to VI oocytes were isolated from excised ovaries and treated with 2 mg/ml type II collagenase (Sigma) in modified Barth's saline (MBS) for 15 min at room temperature, washed extensively in MBS supplemented with 5% bovine serum albumin, and maintained in MBS plus bovine serum albumin at 18°C for all subsequent procedures. After an overnight incubation, healthy oocytes lacking a vitellin layer were selected using a dissecting microscope. WFS1 mRNA was prepared from BglII-linearized pSP64T. In vitro transcription was performed using SP6 RNA polymerase system (Promega). Oocytes were injected with 50 nl of water (sham) or 50 ng of cRNA and then incubated for 3 days at 18°C before analysis.
Endoglycosidase H Digestion-Xenopus oocyte intracellular membranes and rat brain hippocampal tissue lysate were digested with endoglycosidase H overnight at 37°C in a 25-l total reaction volume containing 0.5 liters of 3 M sodium acetate, pH 5.5, and 0.3 milliunits of enzyme (Roche Molecular Biochemicals). All membranes were analyzed by SDS-PAGE followed by immunoblotting as described above.

Confocal Immunofluorescence Microscopy and Immunohistochemistry
Frozen thin sections of oocyte were prepared as described previously (27), and HEK293 cells were obtained from American Type Culture Collection (ATCC). Thin sections and cell monolayers were analyzed by laser confocal immunofluorescence microscopy using protein A-agarosepurified anti-WFS1 antibody (1:100 dilution) and fluorescein-conjugated goat anti-rabbit IgG (1:500 dilution) (Cappel, Durham, NC). Rat pancreas sections were prepared, and immunohistochemical analysis was performed as described previously (28) with some modifications. Briefly, rats were killed by decapitation and pancreata were removed. Pancreata were fixed in 10% formalin and embedded in paraffin for serial sectioning (5 m thickness). Antigen unmasking was performed by treating the tissue sections in 10 mM sodium citrate buffer, pH 6.0, at 95°C for 5 min. Tissue sections were incubated with anti-WFS1 and mouse anti-BiP (Santa Cruz Biotechnology, Santa Cruz, CA) antibodies at 1:50 dilution for 1 h at room temperature in phosphate-buffered saline containing 2% goat serum, then with secondary antibodies (Alexa 594 goat anti-rabbit IgG, Alexa 488 goat anti-mouse IgG) (1:500 dilution; Molecular Probes). For detection of WFS1 in pancreatic islets, sections were processed by a modification of the indirect immunoperoxidase method (29), using mouse anti-insulin and anti-glucagon primary antibodies (1:500, Sigma) and the anti-WFS1 antibody followed by incubation with the secondary antibody as described. Fluorescence was detected using a Bio-Rad MRC-600 laser scanning confocal microscope.

Measurements of Intracellular Calcium in Oocytes
X. tropicalis oocytes were placed in a tissue chamber on the stage of an inverted fluorescence microscope (Axiovert 100M, Zeiss, Lambda 10-Z, Sutter Instrument Co.). Free cytosolic calcium was determined by incubating Xenopus oocytes with fura-2 AM (Molecular Probes) for 30 min at room temperature. Each oocyte was then exposed to 340 and 380 nm excitation wavelengths, respectively. The filter control and data acquisition were carried out using Slidebook, version 3.0.4.9 (Intelligent Imaging Innovations, Cambridge, MA). The excitation light passed through a 10ϫ objective lens to the oocyte through a dichroic mirror. All experiments were carried out at room temperature. Ca 2ϩ concentrations were determined by converting the 340 nm/380 nm fura-2 fluorescence ratios to free [Ca 2ϩ ] using in situ calibration with Ca 2ϩ -EGTA standards (Molecular Probes, Eugene, OR).

Preparation of Oocyte Microsomal Membranes
X. laevis oocytes were injected with cRNA from either wild-type or mutant (I269S, R456H) WFS cDNA, and ER membranes were prepared as described previously (30) with some modifications. Briefly, after collagenase treatment, oocytes were rinsed five times with 5 volumes of H/K buffer (90 mM KCl, 50 mM HEPES, pH 7.1). Stage V-VI oocytes were selected, and the volume of packed oocytes was measured. All residual H/K buffer was removed, and a volume of ice-cold sucrose/H/K (40% sucrose weight/volume in H/K buffer containing 3 mM MgCl 2 ) equal to the volume of the oocytes was added. All subsequent steps were performed at 4°C. The oocytes were gently broken open by passing the mixture through a 3.8-cm 19-gauge needle 10 times. The homogenate was centrifuged for 3 min at 3000 ϫ g. The cloudy gray supernatant was then gently mixed and transferred to a fresh centrifuge tube, and the pellet was discarded. This process was repeated three times. The resulting milky white supernatant was mixed gently and diluted with 0.3 volume of cold H/K buffer. The microsomal membranes were pelleted by centrifugation at 10,000 ϫ g for 10 min. The supernatant was discarded and the membranes were resuspended in H/K buffer containing 3 mM MgCl 2 and then either used directly for SDS-PAGE or stored at Ϫ80°C for bilayer experiments.

Reconstitution of Microsomal Membranes from Oocytes into Planar Lipid Bilayers and Channel Activity Measurements
Planar lipid bilayers were prepared from azolectin type II (Sigma) according to published methods (31). The lipids, 70% dioleylphosphatidylcholine and 30% dioleylphosphatidic acid, were dissolved in decane at 30 mg/ml and stored under nitrogen until use. Briefly, 2 l of lipid solution was applied to 0.25-mm orifice of a polystyrene cuvette (Warner Instruments, Hamden, CT), and the solvent was allowed to evaporate. The cuvette was then placed into a bilayer chamber and connected to a bilayer clamp (BC525-c; Warner Instruments) by Ag/ AgCl electrodes via agar bridges. Data were collected using CLAMPEX 8.0 (Axon Instruments, Foster City, CA), archived on videotape using a Neurocorder DR-484 (Neuro Data Instruments, Delaware Water Gap, PA), and analyzed using ORIGIN (Microcal, Amherst, MA) and CLAMPFIT (Axon Instruments). Bilayers were formed by spreading with a polished glass rod and allowed to thin to a capacitance of 0.4 F/cm 2 , at which point the noise was typically 0.2 pA and the leak conductance was 20 pS. The salt concentrations were initially 150 mM KCl in the cis and trans chamber and adjusted to a 450 mM in the cis chamber as indicated. Outward (positive) currents were defined as K ϩ moving cis to trans. All solutions were buffered to pH 7.0 with 10 mM K-HEPES. Oocyte microsomal membranes (10 g of protein) were added to the cis chamber with mixing. To vary calcium concentration in the bilayer, EGTA and CaCl 2 were added. The MilliQ water employed for buffers in these studies averaged 30 M Ca 2ϩ . When low calcium was required, 300 M EGTA was added to the buffer, reducing the free calcium to less than 1 M. High calcium was obtained by adding 100 M CaCl 2 to the buffer.

Animal Experiments
All experimental protocols were approved by the Animal Studies Committee and conducted according to the Guidelines for Animal Research at Washington University School of Medicine.

Generation and Characterization of Wolframin Antibody-
Rabbits were immunized with a synthetic peptide corresponding to the 20 N-terminal amino acids of human wolframin. Characterization of the wolframin antiserum was conducted by immunoblot analysis using membranes from Xenopus oocytes injected with wolframin mRNA. The anti-wolframin antibody detected a distinct band migrating at a molecular mass of ϳ100 kDa (Fig. 1A, lane 2). The observed band is consistent with the expected molecular mass of 100.29 kDa predicted from the hWFS1 cDNA. Specificity was confirmed by pre-incubation of the antibody with the wolframin immunizing peptide, which completely blocked the appearance of the 100-kDa band on the film (data not shown). The antibody also detected endogenous wolframin expressed in rat hippocampal tissue lysate ( Fig. 2A, Wolframin Is Localized Exclusively to the Endoplasmic Reticulum Membrane-Previously published data suggested that wolframin is localized to endoplasmic reticulum (ER) mem-branes when overexpressed in primary human fibroblasts (12). To confirm these data for endogenous wolframin and determine the subcellular distribution of wolframin after ectopic expression, total membranes from Xenopus oocytes (Fig. 1B, upper panel) and rat brain hippocampus (Fig. 1B, lower panel) were prepared. Digestion of these membranes with N-glycanase, an enzyme that cleaves N-linked oligosaccharides from proteins, or endoglycosidase H, an enzyme that cleaves only high mannose core oligosaccharides from glycosylated proteins, was conducted. Western blot analysis showed a shift in the electrophoretic mobility of wolframin after digestion with either glycosidase (Fig. 1B). These data indicate that the N-linked oligosaccharides of wolframin, in both native cells and after ectopic expression of the protein in oocytes, are not processed to a complex form. This result strongly suggests that wolframin is strictly confined to the endoplasmic reticulum membrane.
The intracellular localization of endogenous and ectopically expressed wolframin was confirmed by indirect immunofluorescence microscopic analysis of frozen sections of Xenopus oocytes injected with wolframin cRNA, HEK293 cells, and rat pancreas. In oocytes, wolframin exhibited a granular cytoplasmic distribution, suggesting intracellular membrane labeling ( Fig. 2A). Endogenous wolframin in HEK293 cells also demonstrated a granular cytoplasmic pattern, consistent with localization to intracellular membranes (Fig. 2C). Preadsorption of the WFS1 antibody with the immunizing peptide abolished labeling of pancreatic islet cells (Fig. 3, A-C). Interestingly, the ␤-cells of rat pancreatic islets stained strongly positive for wolframin (Fig. 3, D-F) and only slight staining was observed in the glucagon positive ␣-cells (Fig. 3, G-I). Double-labeling indirect immunofluorescence analysis using antibodies to wolframin and the ER marker, BiP, was consistent with the strict localization of wolframin to the ER (Fig. 3, J-L). These data confirm that wolframin is expressed predominantly, if not exclusively, in the endoplasmic reticulum in both native cell types and after ectopic expression in Xenopus oocytes.

Expression of Wolframin Induces Specific Ion Channel Activity in Planar Bilayers Derived from Endoplasmic Reticulum
Membranes of Xenopus Oocytes-The presence of predicted transmembrane segments, a very mild similarity to the pore region of the transient receptor potential (TRP)-caspasin channel family, and the lack of any other significant homology with a functional amino acid sequence suggested wolframin might be a novel cation channel in the ER membrane. Because of the strict localization of wolframin to the ER, it was necessary to use reconstitution into planar lipid bilayers to determine whether wolframin possessed channel activity. Microsomal membranes from control (sham-injected) and wolframin-expressing oocytes were studied in planar lipid bilayers using KCl salt solutions to determine how channel activity and its voltage dependence were affected by wolframin expression. The control membranes from four preparations, five preparations from wolframin-expressing oocytes, and two preparations expressing the mutant wolframin proteins were fused into bilayers a minimum of three times each, giving results that were qualitatively consistent. The magnitude of the currents from a fusion varied 5-fold, but several characteristics were reproducibly associated with wolframin expression and distinct from control membranes isolated from sham-injected oocytes. Single-channel records can be difficult to attain from isolated membrane fragments, and therefore the voltage dependence was determined from the mean current normalized to the mean current at zero applied potential in a KCl gradient, which allowed comparison between membrane preparations and individual fusions into bilayers.
A short segment representing the currents observed in control membranes is shown in Fig. 4. We employed all points histograms to characterize the multichannel recordings that were observed in the reconstituted bilayers. An extensive summation of data from four membrane preparations yielding 2-8 min each of recording in a 450/150 mM KCl gradient is shown under the "untreated" heading. Under this trace is the voltage dependence summarized as an average from eight control bilayers. This series demonstrates that our membrane preparation contains a number of channels and that the normalization process allows the incorporation of data from separate experiments. In a KCl gradient, the control microsomes are dominated by inward currents representing chloride movement across the bilayer. We used the voltage dependence of these currents to study calcium and IP 3 effects on these membrane preparations. When 100 M Ca 2ϩ was added to bilayers containing control membranes, the histogram indicated one open state with a conductance approximately double that of control membranes in the absence of Ca 2ϩ (Fig. 4, 100 M Ca 2ϩ ). The addition of 100 nM IP 3 also increased the bilayer currents, but with a small decrease in reversal potential (Fig. 4, 100 nM IP 3 ). These observations are consistent with previous studies on endoplasmic reticulum but do not exclude some contamination by other membrane components (32)(33)(34)(35).
When microsomal membranes from oocytes expressing wolframin were studied in a 450/150 KCl gradient, strikingly different patterns were observed. Although the inward current patterns previously described were present in all membrane preparations, outward currents not seen in the controls were observed (Fig. 5). Wolframin-expressing membranes frequently displayed a conductance 5-fold larger than the examples shown here, but they were unstable and difficult to analyze. The histogram under the Wolframin heading in Fig. 5 is a summation of 20 min of recording in 10 membranes that had moderate to low conductance after fusion with membranes from wol- framin-expressing oocytes. Routinely we saw multiple conductance levels in the outward currents and could identify two distinct phases of wolframin-associated channel activity. The large outward currents with multiple current levels shown at the top of Fig. 5 were the most distinctive wolframin-associated channels. This would occasionally develop into a second pattern, as shown in the second trace in Fig. 5, of smaller but much more regular channels. In bilayers with wolframin-associated channel activity, the outward currents occasionally alternated with inward currents identical to those that were seen in the control membrane (Fig. 5, Wolframin) but the two activity patterns were always separated by a period when the membrane was inactive. When 100 M Ca 2ϩ was added, there was a dramatic reduction of the wolframin-associated outward currents. The normalized slope conductance decreased from 3.19 Ϯ 0.193 nS to 0.116 Ϯ 0.002 nS (n Ͼ 3). Frequently the original currents were restored by the addition of EGTA to reduce total calcium to Ͻ1 M. Typically, treatment with 100 M Ca 2ϩ did not directly produce the control membrane inward currents; however, periods of channel activity identical to those in Fig. 4 were observed. When 100 nM IP 3 was added to membranes while the wolframin-associated channels were active, there was a uniform increase in conductance that corresponded to 8.78 Ϯ 0.14 nS (n ϭ 3) for the example shown. This was accompanied by a shift in E r to Ϫ6.77 mV, which is qualitatively similar to the increase in net outward current that was observed in membranes from sham-injected oocytes. Puromycin or signal peptides at 200 M and adenophostin A had no effect upon the wolframin-associated channel activity (36,37). The addition of 100 M Ca 2ϩ dramatically reduced the wolframin-associated outward currents in planar lipid bilayers (Fig. 6). In many trials the reduction in the normalized slope conductance was 90% or greater. Calcium also uniformly shifted the reversal potential in a 450/150 KCl gradient to the left (Fig. 6A). These effects were partially reversed by 500 M EGTA, which increased the currents but did not return the reversal potential to that seen in control membranes (Fig. 4). Divalent cation block of monovalent cation currents has previously been described to include both Ca 2ϩ and Mg 2ϩ in cells (38). Traditionally Mg 2ϩ is included in buffers used in the isolation of intracellular membranes. In the presence of 2 mM Mg 2ϩ , the wolframin-associated currents showed a mild voltagedependent rectification that disappeared upon removal of the Mg 2ϩ (Fig. 6B). This rectification was time-dependent, and instantaneous currents showed no sign of the rectification (Fig.  6B). In some preparations it was possible to observe a voltagedependent effect of Mg 2ϩ directly. This appeared as a rapid reduction of open channels upon application of positive potentials. By averaging multiple records for each voltage level, it was possible to generate a time course approximating a mac- FIG. 5. Channel activity of microsomal membranes from wolframin-expressing oocytes. Microsomal membranes isolated from wolframin injected oocytes were allowed to fuse with 0.4 F/ cm 2 bilayers in a 450/150 mM KCl gradient. The upper panel is a 3-s record typical of the low conductance wolframin preparations. This preparation was characterized by the all points currents histogram in the Wolframin column. This histogram was obtained from 20 min of recording on 10 membrane preparations. The histograms for 100 M Ca 2ϩ and 100 nM IP 3 columns represent 3-4 min of recording from two preparations of membranes. The voltage dependence of these currents were obtained by averaging the current for 100 ms at each applied potential. The slope conductance and reversal potential for each were determined by linear regression as follows: wolframin roscopic inactivation time course for the Mg 2ϩ effect and to clearly demonstrate the voltage dependence. A consistent characteristic of the wolframin-associated channel activity was this susceptibility to inhibition by the divalent cations Mg 2ϩ and Ca 2ϩ . In the next section we show that this was lost in a wolframin point mutant.
Several mutations in the WFS1 gene have been associated with Wolfram syndrome (39 -41). The existence of these mutations provides a means of testing the possible physiological relevance of the observed cation channel activity induced by wolframin in oocyte membranes. If the observed oocyte channel activity is relevant to the physiological function of wolframin, we would predict that this activity should be altered by mutations associated with the disease. One mutant protein containing an amino acid substitution within a putative transmembrane domain of wolframin (R456H) and a second mutant protein containing an amino acid substitution within the soluble N-terminal region of wolframin (I296S) were studied in planar lipid bilayers. The mutant cRNAs produced proteins of the expected molecular weight after injection into Xenopus oocytes (data not shown). The channel activity observed in oocyte membranes expressing the mutants was distinct from that observed for wild-type wolframin-expressing oocytes or control oocytes. The mutant proteins induced voltage-dependent currents with the E r shifted to 0.040 Ϯ 0.008 mV for I296S and Ϫ14.5 Ϯ 2.8 mV for R456H (see Fig. 7). Distinct from wild-type wolframin, where 100 M Ca 2ϩ reduced the cation selective currents by 90% (Fig. 6A), the membranes expressing the I296S mutant proteins showed no detectable inhibition by the addition of 100 M Ca 2ϩ (data not shown).
Effect of Wolframin Expression on Intracellular Calcium Levels-These studies indicate that either wolframin is a novel ER cation channel or, alternatively, regulates ER channel activities. Because ER cation channels are involved in regulating cytosolic calcium levels in response to various stimuli (42), the effect of wolframin expression on intracellular calcium levels was measured in oocytes. Fig. 8 illustrates the effect of wolframin expression on cytosolic free calcium concentrations in X. tropicalis oocytes. Fura-2 staining revealed that the [Ca 2ϩ ] i of wild-type wolframin-expressing oocytes was 646 Ϯ 185 nM (n ϭ 18), a value significantly greater (p ϭ 0.05) than the values observed in sham-injected oocytes (155 Ϯ 50 nM, n ϭ 26), Glut1-expressing oocytes (140 Ϯ 60 nM, n ϭ 14) or oocytes expressing the R456H wolframin mutant (325 Ϯ 55 nM, n ϭ 16) (Fig. 8). The Glut1-expressing oocytes served as a control for the possible nonspecific effects of ectopic membrane protein expression.
Effect of IP 3 on Channel Activity in Endoplasmic Reticulum Membranes from Wolframin-expressing Oocytes-The IP 3 increases the conductance of the endoplasmic reticulum and nuclear membranes to allow the release of Ca 2ϩ for intracellular signaling (43). We observed that the application of 100 nM IP 3 to microsomal membranes from either sham or wolframinexpressing oocytes was followed by an increase of slope conductance and changes of reversal potential that reflect the activation of the IP 3 receptors in our membrane preparations (Figs. 4 and 5). In control membranes, IP 3 increased the mean Expression of WFS1 protein increases free cytosolic calcium levels in Xenopus oocytes. Intracellular calcium levels in oocytes expressing C-Glut1, controls, mutant R456H, or WFS1 were measured by loading oocytes with fura-2-AM as described under "Materials and Methods." Fluorescence was monitored with an inverted fluorescence microscope assisted with a computer-driven program system. These results represent the means Ϯ S.E. of experiments in which the operator did not know which experimental groups individual oocytes belonged to until after the compilation of the results. Each experimental group included 10 -20 oocytes, and the experiments were repeated four times. C-Glutl cRNA and water-injected oocytes served as controls. Statistical analysis was conducted using analysis of variance, which indicated that only the wolframin group was significantly different (p ϭ 0.05) from Glut-1, control, or the R456H mutant.
conductance from 326 to 1200 pS (Fig. 9A). Membranes overexpressing wolframin exhibited a conductance increase of 5590 pS (Fig. 9A). To compare the IP 3 -dependent currents from sham and wolframin membranes, we used the normalized current-voltage relationships shown in Figs. 4 and 5 to subtract the basal currents from those observed at each voltage after IP 3 addition. The resulting voltage dependence of sham and wolframin membranes is shown in Fig. 9B. The reversal voltage is identical for both membranes, but the slope conductance is much larger in the membranes from oocytes expressing heterologous wolframin. DISCUSSION Wolfram syndrome is an autosomal recessive neurodegenerative disorder associated with juvenile onset non-autoimmune diabetes and progressive bilateral optic atrophy (1, 2). The WFS1 gene responsible for Wolfram syndrome was discovered by a positional cloning strategy (7). However, no clear sequence similarities have been identified in the public data bases that might shed light on the function of the protein. To gain insight into the function of the WFS1 protein, we studied its expression, processing, and localization in Xenopus oocytes, hippocampal tissue, and cultured cells using a polyclonal antibody directed against the N terminus of human wolframin. Our results indicate that wolframin is an endoglycosidase H-sensitive membrane glycoprotein confined to the ER in both native tissues and after ectopic expression in Xenopus oocytes. Additionally, our immunohistochemical studies on rat pancreas revealed a staining pattern for wolframin representative of ␤-cells of pancreatic islets. Because diabetes mellitus is recognized as a major clinical hallmark of Wolfram syndrome, these results suggest that the ␤-cell atrophy associated with this disease is likely to be a direct consequence of defective wolframin expression.
Membrane proteins located in the ER are often involved in protein processing, protein folding, or calcium homeostasis. Disrupting these processes can lead to cellular apoptosis (23,44). It is possible that abnormal apoptosis is responsible for the non-inflammatory cellular degeneration that is observed in Wolfram syndrome (4). The amino acid sequence of wolframin lacks calcium or ATP binding motifs and contains no homology to heat shock proteins or GroEL, thereby providing no evidence for a direct involvement of wolframin in protein folding or calcium sequestration (45). Instead, our experimental evidence suggests that wolframin is associated with ion channel activity in the ER membrane. A distinguishing characteristic of the wolframin-dependent currents was sensitivity to Mg 2ϩ and Ca 2ϩ that was reduced in the Wolfram syndrome point mutants. Similar cation block in transient receptor potential vallinoid-like (TRPV) channels results from channel occupancy at low concentrations of the divalent cations (46). Recently published gel filtration and cross-linking studies of wolframin in RIN and COS cells indicate that it exists as an oligomer, most likely a tetramer, in cellular membranes (6). Many voltage gated cation channels have been shown to be tetrameric (47). We were not able to show directly that the WFS1-dependent channels conducted Ca 2ϩ . However, overexpression of wildtype wolframin in oocytes increased cytosolic [Ca 2ϩ ], consistent with increased release of calcium from the ER to the cytoplasm. Conversely, the point mutant forms of wolframin were not blocked by calcium and did not significantly increase cytosolic [Ca 2ϩ ] when expressed in oocytes. These data suggest that wolframin may be a novel ER calcium channel with properties similar to the TRPV channels (48).
In addition, we determined that IP 3 application to sham membranes generated smaller slope conductance than it produced in membranes from wolframin-expressing cells. Because the IP 3 -dependent currents had the same reversal potential in both membranes, the larger voltage dependence we observed in membranes after heterologous wolframin expression suggests a possible increase of IP 3 receptors or enhanced sensitivity to IP 3 .
Therefore, it remains possible that wolframin activates endogenous ER channels. Active secretory cells require rapid bi-directional ER calcium transfers to regulate cellular secretion and calcium homeostasis (49,50). In mammalian cells the regulation of IP 3 receptor sensitivity is complex, showing both a biphasic response to Ca 2ϩ and activation by ATP (51,52). Recently, calcium sensor proteins that activate IP 3 receptor gating have been described (53). It is possible that wolframin activates a parallel pathway to IP 3 -mediated Ca 2ϩ release as well as functioning as a calcium-sensitive potassium channel that maintains ER membrane potential during calcium release.
In conclusion, the observations reported in this study suggest that wolframin may serve directly as a novel ER calcium channel or, alternatively, as a regulator of ER calcium channel activity. Interestingly, WFS1 mutations associated with Wolfram's disease reduced the susceptibility to cation block. It is possible that wolframin-mediated regulation of intracellular calcium provides an important protective function in secretory cells that are dependent on the ER for calcium signaling. Mutations that disrupt the calcium inhibition of this protein may stimulate cellular mortality decisions originating in the endoplasmic reticulum, thus leading to abnormal neuronal and ␤-cell degeneration that are characteristic of Wolfram syndrome.