Differential Expression of Glutamate Receptor Subtypes in Rat Pancreatic Islets*

Immunocytochemistry was carried out on sections of rat pancreas to localize the expression of glutamate receptor subunits and the major pancreatic peptide hor- mones. Glutamate receptor expression was concentrated in pancreatic islets, and each islet cell type expressed different neuronal glutamate receptors of the (cid:97) -amino-3-hydroxy-5-methylisoxazole-4-propionic acid (AMPA) and kainate classes. AMPA receptor subunits were expressed in (cid:97) , (cid:98) , and pancreatic polypeptide cells, whereas kainate receptors were found predominantly in (cid:97) and (cid:100) cells. Patch clamp electrophysiology was used to measure the functional properties of islet cell glutamate receptors. L -glutamate and other glutamate receptor agonists evoked currents in islet cells that were blocked by the selective AMPA receptor antagonist, 6-cyano-7-nitroquinoxaline-2,3-dione and potentiated by cy- clothiazide in a manner indistinguishable from that of neuronal AMPA receptors. Activation of islet cell AMPA receptors produced steady-state cation currents that depolarized the cells an average of 20.7 (cid:54) 5.4 mV ( n (cid:53) 6). Currents mediated by functional kainate receptors were also observed in a line of transformed pancreatic (cid:97) cells. Thus, L -glutamate probably regulates islet physiology via actions at both AMPA and kainate receptor classes.

Pancreatic islets of Langerhans are self-contained miniature organs responsible for maintaining metabolic homeostasis through the release of the hormones insulin, glucagon, somatostatin, and pancreatic polypeptide (PP). 1 Control of hormone release involves complex interactions between circulating fuels and hormones, autocrine and paracrine regulation, and neuronal input. However, the final common pathway to glucagon and insulin release appears to be the electrical activity of individual ␣ or ␤ cells. Depolarization induced by changes in resting ion channel activity opens voltage-dependent calcium channels, increasing intracellular calcium and activating calciumdependent secretion mechanisms (1,2). Islet cell ion channels can be regulated by intracellular second messengers, G-proteins, and energy levels, but very few of them respond directly to extracellular stimuli. One mechanism for rapidly translating extracellular chemical signals into electrical signals is through ligand-gated ion channels. ␥-amino butyric acid (GABA)-A receptors have been observed in many pancreatic islet cells. Activation of these receptors inhibits glucagon secretion (3) and depolarizes some types of ␤ cell lines (4). Since the GABA synthesizing enzyme, glutamic acid decarboxylase, is found in ␤ cells (5), it has been postulated that GABA is co-released with insulin from the ␤ cells and that it may mediate communication among islet cells themselves (3).
Mechanisms whereby glutamate receptor activation may control hormone secretion from pancreas or preparations of pancreatic cells have recently been described. Glutamate was found to potentiate glucose-stimulated secretion of insulin from perfused pancreas via actions at ␣-amino-3-hydroxy-5-methylisoxazole-4-propionic acid (AMPA)-type glutamate receptors (6). Interestingly, AMPA receptor agonists were also shown to increase glucagon secretion from the same preparation (7). Preliminary descriptions of functional glutamate receptors on a ␤ cell line (8) and on isolated islet cells have also appeared (9). Such observations raise important questions about the role of glutamate receptors in regulating hormone secretion from islet cells. For example, it is unclear which islet cell types actually contain glutamate receptors, and the functional characterization of these receptors was insufficient to clearly classify the glutamate receptors that appear on these cells. If glutamate serves as a mediator of communication between islet cells or between the central nervous system and the endocrine pancreas, it is essential to precisely define the cellular localization and functional properties of these receptors.
Ionotropic glutamate receptors subserve a large proportion of excitatory neurotransmission in the central nervous system. Three classes of glutamate receptors exist named according to the agonists: N-methyl-D-aspartate (NMDA), AMPA, and kainate. Receptors of each class are comprised of heteromeric combinations of homologous subunits (10,11). NMDA receptors are made of NMDAR1 plus one of seven NMDAR2 variants. AMPA receptors consist of combinations of GluRA, GluRB, GluRC, and GluRD. The GluR5, GluR6, GluR7, KA1, and KA2 subunits form neuronal kainate receptors. The exact subunit configuration of receptors in neurons is uncertain, but we do know that different combinations of subunits expressed in host cells produce receptors having unique functional properties. For example, AMPA receptors lacking a GluRB subunit are permeable to calcium and have current-voltage (I-V) relations with strong inward rectification (12,13). When GluRB is present the I-V relation is more linear, and the channels have lower permeability to calcium. Functional kainate receptors are formed when individual GluR5 or GluR6 subunits are expressed alone (14,15), but they also can combine with KA2 to form heteromeric receptors (16). L-Glutamate is probably the natural agonist for AMPA and kainate receptors in the brain. Although AMPA appears selective for the AMPA class, kainate will activate both kainate and AMPA receptor classes. AMPA and kainate receptors can be distinguished based on the desensitization behavior of kainate-induced currents and their sensitivity to allosteric potentiators, such as cyclothiazide (17).
We have examined glutamate receptors in pancreatic islets using histochemical and physiological approaches. Immunocytochemistry was used to identify particular islet cell types that express glutamate receptor subunits, and patch clamp electrophysiology was used to characterize the functional glutamate receptors found on islet cells. We show for the first time that AMPA receptors consisting of GluRB, GluRC, and perhaps GluRA subunits are functionally expressed on ␣ and ␤ cells in pancreatic islets. In addition, ␣ cells express functional kainate receptors, whereas ␦ cells appear to express a kainate receptor subunit protein KA2, whose functional significance is uncertain. These results indicate that glutamate receptors subserve specialized roles in islet physiology.

MATERIALS AND METHODS
GluR5/6/7 antibody was purchased from Pharmingen. GluRA, GluRB/C, GluRD, and NMDAR1, antibodies were purchased from Chemicon. GluR6/7 and KA2 antibodies were a gift from Dr. Robert Wenthold. Guinea pig anti-insulin antiserum was purchased from ICN biochemicals. Guinea pig anti-glucagon and guinea pig anti-pancreatic polypeptide antisera were purchased from Linco Research. Sheep antisomatostatin antiserum was purchased from Cortex Biochem. Indocarbocyanine (Cy3)-conjugated donkey anti-rabbit IgG, ML grade; Cy3conjugated donkey anti-mouse IgG, IgM specific; fluorescein isothiocyanate (FITC)-conjugated donkey anti-guinea pig IgG, ML grade; FITC-conjugated donkey anti-sheep IgG, ML grade; normal rabbit IgG, and normal mouse IgG were purchased from Jackson Immu-noResearch. Alkaline phosphatase-labeled goat anti-rabbit IgG was purchased from DAKO. Triton X-100 Surfact-Amps and bicinchoninic protein assay reagents were purchased from Pierce. EM grade 16% paraformaldehyde solution and tissue-freezing medium were purchased from Electron Microscopy Sciences. Poly-Aqua/Mount was purchased from Poly Sciences, Inc. Collagenase P was purchased from Boehringer Mannheim. Matrigel was purchased from Collaborative Biomedical. RPMI 1640 was purchased from Life Technologies, Inc. AMPA, 6-cyano-7-nitroquinoxaline-2,3-dione (CNQX), and cyclothiazide were purchased from Research Biochemicals. NMDA, kainic acid, amphotericin B, leupeptin, and pepstatin A were purchased from Sigma. All other chemicals were of reagent grade or higher.
Islet Isolation and Cell Culture-Pancreatic islets of Langerhans were isolated from 200 -250-g male Sprague-Dawley rats using the collagenase digestion technique essentially as described (18). Islets were cultured at 37°C in RPMI 1640 containing 5 mM glucose, 10% fetal bovine serum, 100 IU/ml penicillin, and 100 g/ml streptomycin for up to 1 week on Matrigel-coated glass coverslips. When dissociated islet cells were used, they were prepared as described (19) and cultured for up to 2 days on 35-mm plastic dishes (Falcon). ␣TC-6 and ␣TC-9 cells were cultured as described previously (20).
For immunostaining, slides were rehydrated in PBS. Sections stained with the GluR5/6/7 antibody were treated according to the supplied protocol. For all other antibodies, sections were permeabilized in 0.2% Triton X-100 in PBS and blocked for 1 h in 5% normal donkey serum in PBS. The blocking solution was decanted, and the sections were incubated overnight at 4°C with affinity-purified anti-glutamate receptor antibodies diluted to the following concentrations in PBS containing 0.1% Triton X-100 and 1% bovine serum albumin: GluRA, 1 g/ml; GluRB/C, 0.5 g/ml; GluRD, 1 g/ml; GluR6/7, 0.5 g/ml; NMDAR1, 0.5 g/ml. As a control anti-glutamate receptor antibodies were omitted or replaced with normal IgG. Following the incubation, the sections were washed three times for 10 min per wash in PBS containing 0.1% Triton X-100. For detection of tissue-bound anti-glutamate receptor antibody the sections were incubated for 1 h with a 1:500 dilution of Cy3-conjugated donkey anti-rabbit IgG or donkey antimouse IgG (IgM-specific) in PBS containing 0.1% Triton X-100 and 1% bovine serum albumin. Sections were then washed three times for 15 min per wash with PBS containing 0.1% Triton X-100.
Double labeling of anti-glutamate receptor-stained pancreas sections for neuroendocrine cell markers was done by incubating sections for 2.5 h at room temperature with a 1:200 dilution of guinea pig anti-insulin, anti-glucagon, anti-pancreatic polypeptide antisera, or a 1:250 dilution of sheep anti-somatostatin antiserum diluted in PBS containing 0.1% Triton X-100 and 1% bovine serum albumin. Sections were washed three times for 10 min per wash with PBS containing 0.1% Triton X-100. For detection of tissue-bound antibody the sections were incubated for 1 h with a 1:250 dilution of FITC-conjugated donkey anti-guinea pig IgG or anti-sheep IgG. Following the final incubation, the sections were washed three times for 15 min per wash with PBS containing 0.1% Triton X-100. Sections were mounted with Aqua-Poly/Mount. Immunofluorescence images were obtained on a Zeiss LSM 410 confocal microscope using the 488-and 543-nm lines of a HeNe and ArKr laser, respectively. Individual images for Cy3 and FITC fluorescence were examined separately to evaluate the extent of co-localization. All images were optimized using the Zeiss LSM program and then transferred as TIFF files to a Silicon Graphics Indigo, where figures were assembled using SGI Showcase and printed using a Tektronix Phaser IISDX color printer.
For Western blots, islets, ␣TC-6/9 cells, or rat brain cortex tissue were extracted on ice in an extraction buffer that contained (in mM) 20 MOPS, pH 7.5, 500 NaCl, 1 phenylmethylsulfonyl fluoride, 0.001 leupeptin, 0.001 pepstatin A sonicated on ice twice for 15 s at 1-min intervals and centrifuged at 100,000 ϫ g for 1 h at 4°C. The supernatant fraction was discarded, and the pellet was rinsed twice with extraction buffer. The pellet was resuspended in a buffer containing (in mM) 20 MOPS, pH 7.5, 0.1% (w/v) SDS, 1 phenylmethylsulfonyl fluoride, 0.001 leupeptin, 0.001 pepstatin A. Protein concentrations were determined using the bicinchoninic acid assay method (21). Proteins were separated on 7.5% (w/v) polyacrylamide gels using the method described by (22) and transferred to nitrocellulose membranes. The membranes were blocked with 10% (w/v) nonfat dry milk in PBS for 1 h, washed 3 times for 10 min in PBS containing 0.05% (v/v) Tween 20, and probed for 1 h with anti-glutamate receptor antibodies diluted in PBS containing 0.5% (v/v) goat serum. The dilution factors for the anti-glutamate receptor antibodies were the same as those used for the immunocytochemistry studies. Following incubation with primary antibody the blots were washed 3 times for 10 min in Tris-buffered saline (TBS) solution, which contained (in mM) 25 Tris, pH 8.0, 137 NaCl, 2.7 KCl. The blots were then incubated with a 1:1000 dilution of alkaline phosphatase-labeled goat anti-rabbit IgG in TBS containing 0.05% (v/v) Tween 20 and 0.5% (v/v) goat serum, washed three times for 10 min in TBS, and washed once for 10 min in alkaline phosphatase buffer, which contained (in mM) 100 Tris, pH 8.0, 100 NaCl, 5 MgCl 2 . The blots were developed using nitro blue tetrazolium and 5-bromo-4-chloro-3-indolyl phosphate as substrates dissolved in alkaline phosphatase buffer.
Electrophysiology-Patch clamp electrophysiology was performed on intact and dissociated islets at room temperature using either an Axopatch 200 (Axon Instruments) or a Dagan Cornerstone (Dagan Corporation) amplifier. Signals were filtered using a Frequency Devices 8-pole Bessel filter, recorded on a Macintosh Quadra 800 or IIfx computer, and analyzed using Igor and Microsoft Excel computer programs. Voltage clamp recordings were made using either conventional whole cell mode or the perforated patch technique. All current clamp recordings were made with perforated patches. Only cells showing a combined seal/input resistance of greater than 1 G⍀ were included for current clamp studies, and measurements were made only after the series resistance had dropped to a stable value usually less than 20 M⍀. For conventional whole cell, patch pipettes were filled with (in mM) 135 CsCl, 1 MgCl 2 , 11 EGTA, 10 HEPES, pH 7.3. For perforated patch, pipettes were filled with (in mM) 10 KCl, 10 NaCl, 70 K 2 SO 4 , 2 MgCl 2 , 10 HEPES, pH 7.3, and 240 g/ml amphotericin B. The extracellular solution contained (in mM) 150 NaCl, 2.5 KCl, 2.5 CaCl 2 , 1 MgCl 2 , 5 glucose, 10 HEPES, pH 7.3. Ligands were dissolved in the extracellular solution except for those used to test for NMDA receptors. These solutions did not contain Mg 2ϩ . In most experiments ligands were applied by controlled superfusion using a multibarreled pipette. Rapid agonist application was performed as described (23).
AMPA and kainate receptor decay constants were calculated from curves fit to single exponential equations. Current-voltage relations were obtained by ramping the voltage from Ϫ80 mV to ϩ80 mV at a rate of 40 mV/s. Currents measured in the absence of agonist were digitally subtracted from those measured in the presence of agonist to yield the agonist-evoked I-V curve. Reversal potentials were determined from a least squares fit of the I-V relation to third order polynomial equations. Rectification ratios were calculated by dividing the conductance measured at ϩ50 mV by that measured at Ϫ50 mV.

RESULTS
Immunocytochemistry was used to examine the expression pattern of glutamate receptors in cryosections of whole pancreas. To identify the types of islet cell that expressed glutamate receptors, each section was also stained with an antibody specific for one of the major pancreatic islet hormones. An antibody against insulin stained the core of the islets and was used as marker for ␤ cells, an antibody that recognizes gluca-gon specifically stained ␣ cells, ␦ cells were labeled with a somatostatin antibody, and cells positive for pancreatic polypeptide were referred to as PP cells. Laser scanning confocal microscopy was used to identify cells in which glutamate receptor immunoreactivity co-localized with that of the islet hormones. Sections were made from three different adult rats, and except where noted the results did not differ between animals.
An antibody specific for the AMPA receptor subunits GluRB and GluRC (GluRB/C) strongly stained large portions of every islet examined but did not stain the surrounding pancreatic acinar tissue. Islets were labeled in the center and the mantle, but the cells in the core of the islet showed more intense staining than did the cells near the edges. Based on double labeling with antibodies against GluRB/C and hormone markers, GluRB/C subunits appeared in ␣, ␤, and PP cells but not in ␦ cells (Fig. 1). Although the levels of GluRB/C varied between Coexpression of GluRB/C with glucagon is also evident; however, the GluRB/C expression levels are lower in the islet mantle than they are in the core. GluRB/C does not appear to be coexpressed with somatostatin. GluRD staining was absent in islet cells.
islet cells in the core, we were unable to detect core cells that were positive for only insulin or GluRB/C. Similar co-localization of GluRB/C immunoreactivity with that of glucagon (Fig.  1B) and PP (not shown) was also apparent in the mantle of islets. GluRB/C antibody staining did not co-localize with immunoreactivity for somatostatin (Fig. 1C). An antibody specific for the AMPA receptor subunit GluRA also stained sections from two rats (not shown). However, the staining with the GluRA antibody was generally weaker than that of the GluRB/C antibody, and in one of the two rats staining could only be visualized using immunoperoxidase detection with a metal-enhanced diaminobenzidine substrate. In both rats the GluRA staining was restricted to the islet core. In the rat where immunofluorescent double labeling was possible, the GluRA staining co-localized only with insulin.
Kainate receptor subunits were also detected in these pancreatic sections and were concentrated in islets but not pancreatic acinar cells. Strong immunoreactivity for the kainate re-ceptor subunits, GluR6/7, was observed in ␣ cells but not in ␤ cells (Fig. 2, A and B). GluR6/7 immunoreactivity was not observed in ␦ cells (Fig. 2B, inset), nor was GluR6/7 protein detected in PP cells (not shown). To assure the specificity of the GluR6/7 labeling we also stained sections with a monoclonal antibody directed against the GluR5/6/7 subunits. This antibody was made against a different epitope of these kainate receptor subunits but stained the same population of cells recognized by the GluR6/7 polyclonal antibody. This supports the assertion that GluR6/7 staining represents kainate receptor subunits. Islet cells were also labeled with an antibody specific for the KA2 subunit of kainate receptors. This immunoreactivity appeared only in ␦ cells (Fig. 2C) and not in ␤ (Fig.  2C, inset), ␣, or PP cells (not shown).
Other glutamate receptor subunits were notably absent from the pancreatic islets. An antibody specific for the AMPA receptor subunit, GluRD (Fig. 1D) did not stain pancreas sections above background, indicating that the AMPA receptors in islets consist largely of GluRB, GluRC, and perhaps GluRA. Staining was not observed with an antibody specific for the NMDAR1 subunit of the NMDA receptor complex (Fig. 2D). This antibody is known to recognize four of the most prevalent NMDAR1 splice variants in brain. If other NMDAR1 splice variants are expressed in pancreatic islets they would not have been detected. Staining was not observed when normal rabbit IgG was used in place of receptor antibodies (Fig. 2D, inset).
AMPA and kainate subunit receptors were also detected in Western blots of crude membrane fractions from brain, islets, and islet cell lines (Fig. 3). GluRB/C immunoreactivity was similar in brain, islet, and ␣TC6 membranes. GluR6/7 and KA2 immunoreactivity was observed in brain and ␣TC6 membranes. The apparent molecular weight of these proteins was appropriate for glutamate receptor subunits. We were unable to detect GluR 6/7 or KA2 immunoreactivity in isolated islets, perhaps due to limiting amounts of islet tissue, the low proportion of ␣ and ␦ cells, or proteolytic breakdown of these antigens caused by the collagenase digestion technique.
To examine the electrophysiological consequences of glutamate receptor activation in pancreatic islet cells we used the patch clamp technique and examined the effects of glutamate receptor agonists on membrane potential and ionic currents in these cells. Current clamp measurements of membrane potential were made using the perforated patch technique and revealed a variety of cell types in our preparation. The average resting membrane potential of islet cells in 5 mM glucose (a concentration considered to be nonstimulating for ␤ cells) was Ϫ59 Ϯ 2.5 mV (n ϭ 22). Of seven cells challenged with 16.7 mM glucose, three depolarized and fired action potentials, three gave no response to glucose, and one was firing spontaneous action potentials that were inhibited by high glucose. Two other cells also fired spontaneous action potentials, but they were not tested with high glucose. The observations suggest that our isolated islet preparations contained ␣ and ␤ cells in addition to some cells that were more difficult to classify.
A subpopulation of islet cells (6 of 22) also responded to the application of 300 M L-glutamate with marked depolarization (Fig. 4A). The average glutamate-induced depolarization was 20.7 Ϯ 5.4 mV (n ϭ 6). In one case the depolarization evoked by L-glutamate was sufficient to induce firing of action potentials. One of the six glutamate receptor positive cells also depolarized in response to high glucose. We were unable to test the other five with high glucose. When we measured the current generated by 300 M L-glutamate in these cells under voltage clamp at Ϫ70 mV we found it was often little more than an increase in membrane noise. One gave a measurable inward current at Ϫ70 mV (amplitude ϭ 3.1 pA). In 12 cells measured under voltage clamp the average amplitude of inward current evoked by 300 M L-glutamate at Ϫ70 mV was 2.0 Ϯ 0.7 pA. We reasoned that rapid desensitization attenuated the currents induced by our relatively slow application of agonist, and therefore we measured L-glutamate currents in the presence of cyclothiazide, which blocks desensitization of neuronal AMPA receptors (24). Cyclothiazide strongly potentiated the steadystate current evoked by L-glutamate (Fig. 4B, average potenti-FIG. 3. Anti-glutamate receptor antibodies recognize the same proteins in brain, islets, and islet cell lines. Shown is a Western blot of crude membrane proteins from brain (25 g of total protein/lane), islets (25 g of total protein/lane), and ␣TC-6 cells (50 g of total protein/lane). Proteins were resolved on 7.5% (w/v) polyacrylamide gels, blotted to nitrocellulose, and probed with anti-glutamate receptor antibodies. Immunoreactivity was detected using alkaline phosphataselabeled goat anti-rabbit IgG with nitro blue tetrazolium and 5-bromo-4-chloro-3-indolyl phosphate as substrates. Anti-GluRB/C antibodies recognized proteins of the same apparent molecular weight in samples from brain, islets, and ␣TC-6 cells. Antibodies against GluR6/7 and KA2 recognized proteins of the same apparent molecular weight in both brain and ␣TC-6 cells. ␣TC-9 cell immunoreactivity was identical to that of ␣TC-6 cells. Molecular weights (left side) were determined from standards. ation of the steady state current ϭ 33 Ϯ 11-fold, n ϭ 8) and allowed visualization of the L-glutamate current in five of the six cells that were depolarized by 300 M L-glutamate alone (one was not tested under voltage clamp). The average inward current evoked by 300 M L-glutamate plus 50 M cyclothiazide in these five cells was 25.5 Ϯ 9.6 pA. Cyclothiazide did not evoke currents when applied to islet cells by itself (n ϭ 230).
Voltage clamp studies were undertaken to access the agonist response characteristics of a larger population of islet cells. To facilitate characterization of islet cell AMPA receptors many of the voltage clamp recordings were made in the presence of 50 M cyclothiazide. Under these conditions steady-state inward currents evoked by L-glutamate, kainate, or AMPA were observed in 63 of 230 islet cells examined (27%). The agonist response characteristics in the presence of cyclothiazide are summarized in Table I. A small proportion of islet cells were also responsive to 30 M GABA (Table I), and 8 of 11 cells that responded to GABA also expressed functional glutamate receptors. One cell showed inward current in response to co-application of 30 M NMDA and 10 M glycine. Since we saw this type of current in only one cell, we could not determine if the current was mediated by the NMDA class of glutamate receptors or if it arose from strychnine-sensitive glycine receptors (see "Discussion").
Currents evoked by L-glutamate showed properties quite similar to those of neuronal AMPA receptors. Currents were evoked by AMPA, kainate, or L-glutamate and were sensitive to the selective AMPA receptor antagonist, CNQX (Figs. 3C and 4C). Currents evoked by 300 M kainate plus 50 M cyclothiazide were reduced from 24.7 Ϯ 5.0 pA under control conditions to 3.2 Ϯ 0.6 pA (n ϭ 8) in the presence of 10 M CNQX. Currents elicited by L-glutamate or AMPA showed a mean reversal potential of Ϫ2.5 Ϯ 0.7 mV (n ϭ 16) in normal extracellular solution, indicative of nonselective cation channels. The I-V relation varied in shape with an average rectification ratio (G (ϩ50) /G (Ϫ50) ) of 0.66 Ϯ 0.09 (n ϭ 16). This value ranged from 0.1 (strong inward rectification) to 1.4 (outward rectification), suggesting that some functional AMPA receptors in islet cells may lack the GluRB subunit (12,13,25). Overall, the pharmacological and physiological properties of islet cell AMPA receptors differ little from those of their neuronal counterparts.
To study the rapidly desensitizing component of glutamate receptors in islet cells, we turned to cell lines derived from pancreatic ␣ cells, ␣-TC6, and ␣-TC9 (20,26). These cells were hardier and therefore more amenable to rapid agonist application than were native islet cells. The immunological and functional properties of glutamate receptors did not differ between cell lines, and the results from each were pooled. Rapid appli-cation of 1 mM L-glutamate to these cells often elicited currents showing a desensitizing component followed by a much smaller steady-state current (Fig. 5A). Currents evoked by 1 mM Lglutamate desensitized with a single exponential time constant of 12.3 Ϯ 2.3 ms (n ϭ 12). In most cells (8 of 12) currents induced by kainate did not show desensitization, and 50 M cyclothiazide prevented desensitization, strongly potentiating the response to L-glutamate and kainate (Fig. 5B). A subpopulation of these cells (4 of 12 cells) gave desensitizing currents, average time constant ϭ 14.6 Ϯ 4.3 ms (n ϭ 4), in response to rapid application of 1 mM kainate (Fig. 5C). In one cell kainateevoked desensitization was superimposed upon a steady-state current, which was probably mediated by co-expressed AMPA receptors. In other cells little or no steady-state kainate current was observed. Three cells showing small steady-state kainate current were also insensitive to potentiation by 50 M cyclothiazide (not shown). Neuronal and recombinant kainate receptors display the unique property of being desensitized by kainate and insensitive to cyclothiazide (24,27,28), whereas AMPA receptors respond to kainate with little or no desensitization and are strongly potentiated by cyclothiazide. These results support the view that kainate receptors in pancreatic islets are functional and may have some role in regulating islet physiology.

DISCUSSION
Using complementary approaches we have shown that pancreatic islet cells express glutamate receptors of the AMPA and kainate classes. Immunocytochemistry using double labeling techniques showed that AMPA receptor subunits were expressed in ␣, ␤, and PP cells but were generally absent from ␦ cells. Kainate receptor subunits were expressed in ␣ and ␦ cells and were not found in ␤ or PP cells. Patch clamp of islet cells revealed that functional AMPA-type receptors appeared on the plasma membrane in about 25% of the cells tested. A trans-  FIG. 5. Functional AMPA and kainate receptors are present on ␣TC-9 cells. A, rapid application of L-glutamate (holding voltage ϭ Ϫ60 mV) to these cells produced a desensitizing peak current followed by a much smaller steady state component. B, cyclothiazide (50 M) prevented observable desensitization and potentiated both the peak and steady-state current in this cell. C, shown is a rapidly desensitizing current evoked by kainate (holding voltage ϭ Ϫ60 mV) in a different ␣TC-9 cell. D, the L-glutamate current in this cell shows rapid desensitization as well, but the ratio of the peak and steady-state amplitude differed from that seen in cells expressing only AMPA receptors (see A and B above). Cyclothiazide had no effect on the currents in this cell (not shown).
formed line of ␣ cells also expressed functional kainate-type receptors that could be detected with rapid agonist application techniques. The expression of different glutamate receptor types on islet cells specialized to secrete different hormones suggests that communication within islets is mediated by machinery similar to that used by the central nervous system.
Unambiguous identification of receptor expression in pancreatic islets depends on the specificity of the antibodies used here. The NMDAR1, GluRA, GluRD, and GluRB/C antibodies have been extensively characterized using transfected cells and brain tissue (29 -31). The NMDAR1 antibody recognizes splice variants 1a, 1b, 2a, and 2b. While these are the major splice variants expressed in brain, it remains possible that others are expressed in islets. In our hands, the GluRB/C antibodies only stained proteins of the appropriate molecular weight in membranes isolated from brain, islets, and islet cell lines when present in sufficient quantities. The GluR6/7 and KA2 antibodies also show appropriate staining patterns in brain tissue (32) and recognize bands of the appropriate size in Western blots with membrane proteins from brain and ␣TC-6 cells. The low density of these subunits in islet cell membranes prevented detection of GluR6/7 and KA2 antigens in Western blot experiments. This may have been due to loss or damage of ␣ and ␦ cells as a result of the collagenase islet isolation technique. However, a monoclonal antibody selective for GluR5/6/7 subunits (33) strongly stained islet mantle cells supporting the localization of kainate receptor subunits in ␣ and ␦ cells.
Similar to the situation in the central nervous system, the role of the KA2 subunit in ␦ cells of islets is uncertain. We were unable to study putative KA2-containing receptors with electrophysiology, since ␦ cells are in low abundance in islets and no ␦ cell line was available. The apparent expression of KA2 in the ␣TC-6 cell line may result from dedifferentiation of these cells, since no KA2 was detected in native ␣ cells.
The current densities in isolated islet cells were low enough to make complete characterization of the receptors difficult. It is possible that islet isolation procedures, which involve proteolytic digestion of the pancreatic acinar tissue, could have damaged or destroyed some of the receptors before our patch clamp experiments. Even so, the steady-state currents were large enough to depolarize islets when agonist was applied slowly, long after desensitization was complete. If glutamate is released rapidly onto the receptors, similar to what happens at synapses in the central nervous system, the much larger desensitizing component of the response should have larger, albeit brief, depolarizing effects.
The functional glutamate receptors in islet cells have properties similar to those of receptors found in neurons or expressed in host cells from cloned subunits. Islet cell AMPA receptors respond to L-glutamate, AMPA, and kainate, are blocked by the competitive antagonist, CNQX, and are potentiated by cyclothiazide. These properties are shared by neuronal AMPA receptors. The current-voltage relation of islet cell AMPA receptors indicate that they are typical nonselective cation channels. Some cells showed noticeable inward rectification, which is a hallmark of AMPA receptors lacking a GluRB subunit. Such receptors would be expected to have higher permeability to divalent cations such as calcium. Because the currents were so small, it was difficult to determine the calcium permeability of these receptors with standard ion substitution studies. In one cell showing inward current we did detect a very small inward current at Ϫ70 mV after replacing all extracellular monovalent cations with Ca 2ϩ . The possibility that some AMPA receptors in islets flux Ca 2ϩ suggests that AMPA receptor activation could bypass the voltage-dependent Ca 2ϩ channels and directly influence hormone secretion by changing in-tracellular [Ca 2ϩ ]. Imaging of intracellular calcium in islet cells is currently in progress to address this possibility.
The staining pattern of GluRB/C subunits indicates that the vast majority of islet cells express AMPA receptor subunits. By contrast, our functional studies indicate that only a subpopulation of them actually have significant numbers of functional receptors on the plasma membrane. The proportion of cells showing functional receptors remained consistent between preparations and is not markedly different from that seen in the ␣-TC6 and ␣-TC9 cell lines or the insulinoma cell line, GKP3. 2 The perinuclear pattern of receptor subunit staining suggests that much of the protein is intracellular as opposed to on the plasma membrane (34). Similar observations have been made using these antibodies in brain sections (35). The intracellular location of glutamate receptor subunit proteins and the fact that only some islet cells express functional receptors suggest that the presence of functional receptors may be regulated by environmental or metabolic factors. Alternatively, the presence or absence of functional glutamate receptors could be a manifestation of diversity among cells of a given islet cell type. For example, heterogeneity in glucose sensitivity among ␤ cells has been described (36). This heterogeneity could extend to the presence or absence of functional L-glutamate receptors.
Although formally possible, it seems unlikely that glutamate receptors are functional only in one type of islet cell (␣, ␤, ␦, or PP). Some cells that were depolarized by L-glutamate showed spontaneous action potentials similar to those seen in ␣ cells, whereas other cells depolarized and fired action potentials in response to glucose, which is typical of ␤ cells. The fact that L-glutamate depolarizes islet cells via actions at AMPA receptors supports the view that AMPA receptor activation can modulate hormone secretion from pancreatic islets. In a perfused pancreas preparation Bertrand et al. have shown that activation of AMPA, but not NMDA receptors, can influence secretion of insulin (6) and glucagon (7). In these studies perfusion of agonists stimulated secretion of both hormones, which are generally considered to have opposing physiological actions. Since AMPA receptors are found on both glucagon-and insulin-secreting cells this result is not surprising. However, it seems unlikely that the natural mode of receptor activation involves changes in plasma glutamate levels because that would stimulate the release of two physiological antagonistic hormones. Recent reports also support the presence of functional glutamate receptors on islet cells and cell lines derived from transformed ␤ cells (8,9). Besides showing that glutamate receptor activation depolarizes some islet cells these workers also showed that receptor activation can increase intracellular calcium. The postulated presence of functional NMDA receptors on islet cells awaits confirmation. The pharmacology of the NMDA responses, in particular the effects of antagonists, were incompletely described, and even though AMPA and perhaps kainate receptor activation stimulated insulin secretion in islet cells, NMDA could not (9). NMDA receptor ligands were likewise unable to modulate hormone secretion in perfused pancreas preparations (6,7).
Islet cells have much in common with neurons. The most notable similarity involves the expression of proteins specialized for synaptic transmission. Islet cells contain presynaptic proteins associated with vesicular secretion such as synaptotagmin (37). They also express GABA-A (3, 4) and glutamate receptors, which are major postsynaptic receptors in the mammalian central nervous system. The presence of such specialized synaptic machinery in islets suggests that synaptic-like communication is important for normal islet physiology. This machinery may subserve communication between the central nervous system and islets. For example, GABA-containing nerve fibers have been detected in or near islets (38). Differences in electrical activity between isolated islets and those measured in vivo may be due to neuronal inputs that are present in vivo but lacking in vitro (39). Islet cell AMPA and kainate receptors would be well suited for translating neuronal inputs into electrical signals that modify secretory properties. It will be of interest to determine if the coeliac ganglia, which provide most of the innervation to islets contains glutamatergic output neurons.
Neurotransmitters may also subserve communication between islet cells themselves. It has been postulated that GABA is released from ␤ cells, recognized by ␣ and ␦ cell GABA-A receptors and mediates communication among ␣, ␦, and ␤ cells (Refs. 2 and 3, but see Ref. 38). Indeed, glutaminase immunoreactivity has also been detected in the periphery of islets, prompting the suggestion that some islet mantle cells may be manufacturing glutamate from glutamine and releasing it in particular situations (9). The oscillations of intracellular calcium and electrical activity in mouse islets (40 -43) could be entrained by synchronized release of glutamate from one cell type onto another. It will be of interest to determine if AMPA and kainate receptor activation play unique roles in modulating islet physiology. The possible presence in islets of other proteins associated with glutamatergic neurotransmission in the central nervous system such as metabotropic glutamate receptors and high affinity glutamate transporters must also be investigated.
Potential roles for glutamate receptors in some forms of diabetes and glucose intolerance are still uncertain, but the presence of these receptors offers another potential target for therapeutic intervention to fine tune insulin or glucagon secretion in those who cannot properly regulate these hormones. Additionally, our observations raise important questions about the role of ambient glutamate in regulating islet physiology. At a minimum, the actions of monosodium glutamate (MSG) should probably be re-evaluated because glutamate clearly depolarizes some islet cells under situations similar to what may occur when serum glutamate concentrations are elevated. The fact that some forms of diabetes are autoimmune disorders raises interesting parallels with recent work on the autoimmune nature of certain types of epilepsy. For example, Rasmussen's encephalitis may be an autoimmune disorder in which antibodies that recognize GluRC activate neuronal AMPA receptors, causing brain damage and seizures (44,45). The fact that the GluRC subunit is likely to be expressed in pancreatic islets and normally exposed to the immune system suggests that antibodies against these receptors could be generated without exposure of the central nervous system to the immune system. It will be of importance to determine if some forms of diabetes or other glucose intolerance involve immune system attack on glutamate receptors in pancreatic islets.