A Novel Variant of Ionotropic Glutamate Receptor Regulates Somatostatin Secretion From δ-Cells of Islets of Langerhans

Islets of Langerhans were isolated from male Wistar rats at 7–8 postnatal weeks by the collagenase digestion method combined with discontinuous Ficoll gradient centrifugation (23). Islets were then hand picked and suspended in a bicarbonate-buffered Hank’s solution supplemented with 0.2% BSA. For experimental procedures, the islets were maintained for 1 day and washed with a culture medium comprised of RPMI-1640, 10% fetal bovine serum, 5 mmol/l glucose, 100 units/ml penicillin, and 100 μg/ml streptomycin. Animal care was performed under the authors’ institutional guidelines and was approved by the institutional review board.

Site-specific rabbit polyclonal antibodies against rat GluR2, GluR3, GluR4, GluR5, GluR6, GluR7, KA1, and KA2 were raised as follows. The synthetic peptides corresponding to the residues based on the published amino acid sequences (24–30) 854–864 [(C)NINPSSSQNSQN] of GluR2, 859–869 [(C)QNFKPAPATNTQN] of GluR3, 395–405 [(C)KLVLIQDMPTL] of GluR4, 890–907 [(C)TKGKSSFTSILT] of GluR5, 406–429 [(C)WDPASGLNMTESQKGKPANITDSLSN] of GluR6, 859–871 [CSTVADEIRFSLT] of GluR7, 393–412 [(C)VAEGLSMDSRLYASNISDS] of KA1, and 392–411 [(C)YSNRTLAMNATTLDINLSQ] of KA2 were conjugated with keyhole-limpet hemocyanin with m-maleidobenzoyl-N-hydrosuccinimide ester and then injected into rabbits according to the standard immunization protocol. These site-specific antibodies were used after purification on an affinity column, on which the corresponding peptides were immobilized. (The GluR4 antibodies are abbreviated as GluR4N.) The specificity of the antibodies was verified through immunoblotting with cells expressing the corresponding receptors. Site-specific rabbit polyclonal antibodies against GluR1, -δ1, and -δ2 have been previously reported (31,32). For the preparation of antibodies specific to GluR4c variants, DNA fragments encoding V849–I885 were amplified by PCR and then cloned into the EcoRI site of expression vector pGEX3X (Amersham Pharmacia Biotech) to form glutathione S-transferase (GST) fusion plasmids. After transformation, the GST fusion proteins encoding V849–I885 were purified on a glutathione-Sepharose 4B column (Amersham) and then injected into a rabbit with complete adjuvant twice with a 2-week interval between injections. The mouse monoclonal antibodies against glucagon were obtained from Progen, the rat monoclonal antibodies against somatostatin and rabbit polyclonal antibodies against GluR4 COOH-terminals from Chemicon, the antibodies against GluR2/3 and KA2 from Upstate Biotechnology, the antibodies against NR1 from Synaptic Systems, the guinea pig polyclonal antibodies against insulin and rat pancreatic polypeptide from Biogenesis, and the Alexa Fluor 568-labeled anti-mouse IgG and Alexa Fluor 488-labeled goat anti-rabbit IgG from Molecular Probes.

Indirect immunofluorescence microscopy was performed using an Olympus FV-300 confocal laser microscope, as previously described (10).

RT-PCR was performed as previously described (10). Briefly, 30 temperature cycles were performed, with denaturation at 94°C for 30 s, annealing at 54°C for the common region of GluR4 variants (GluR4 common), 58°C for GluR4 COOH-terminal and 4c, and 57°C for GluR4d and GluR4s for 30 s, with an extension at 72°C for 1 min. The sequences of the oligonucleotides used as primers for each GluR4 variant were based on published sequences (21,22). The conditions for PCR were as follows: GluR4 common sense primer, 5′-CAACTGCTAGAAGAGCTTGA-3′ (bases 586–605), and antisense primer, 5′-CATATCTTGAATCAAGACTA-3′ (bases 1187–1206) (accession no. M38063); GluR4 COOH-terminal sense primer, 5′-CCCATAGAAAGTGCAGAAGACC-3′ (bases 1960–1981), and antisense primer, 5′-CAATGACAGCCAATCCCGAAC-3′ (bases 2672–2692) (accession no. M38063); GluR4c sense primer, 5′-CCCATAGAAAGTGCAGAAGACC-3′ (bases 1960–1981), and antisense primer, 5′-AGTAAATCTTAATACTTTCGG-3′ (bases 2636–2659) (accession no. S94371); GluR4d sense primer, 5′-TGTCAGGTCGGATTGTTGGAG-3′ (bases 1865–1885), and antisense primer, 5′-CACTTTCACCTATAGTACAGCG-3′ (bases 2577–2593) (accession no. U65993) (22); and GluR4s sense primer, 5′-TGTCAGGTCGGATTGTTGGAG-3′ (bases 1865–1885), and antisense primer, 5′-GCTTTGTTTCGTATGGCTTCAG-3′ (bases 2317–2338) (accession no. U65991) (22). For semiquantitative analysis of the expression of GluR4c-flip or -flop variants, the amplified products of GluR4c were treated with AvaI or DraI to specifically digest flip and flop variants. The resultant mixture was analyzed by PAGE, and DNA fragments specific to the flip or flop variants were quantified with National Institutes of Health Image (version 1.63). The DNA sequences were determined by a standard method (33).

The amplified product of GluR4c-flip was used as a hybridization probe to isolate cDNA clones. Rat islet cDNA libraries were prepared using a ZAP Express cDNA/Gigapack kit (Stratagene, La Jolla, CA) and poly AT tract mRNA from islets (∼50,000 pieces) as a template, which was prepared using a poly AT tract isolation system (Promega, Madison, WI). The amplified product of GluR4c-flip was used as a hybridization probe to isolate two overlapping cDNA clones from rat islet cDNA libraries. The full-length cDNA was then constructed. cDNA encoding GluR4c-flip was subcloned into the XhoI and BamHI sites of expression vector pcDNA3.1(−) (Invitrogen, San Diego, CA). The resultant construct, GluR4c-flip, was transfected into COS-7 cells by the lipofection method.

Isolated islets (∼500 pieces) were homogenized in a buffer containing 20 mmol/l 3-morpholinopropanesulfonic acid-Tris (pH. 7.0), 0.3 mol/l sucrose, 5 mmol/l EDTA, 5 μg/ml leupeptin, and 5 μg/ml pepstatin A. The homogenate was centrifuged at 800g for 10 min, and the supernatant was again centrifuged at 266,000g for 30 min. The particulate fraction was suspended in SDS buffer and then subjected to SDS gel electrophoresis, followed by immunoblotting with enhanced chemiluminescence (10).

Isolated islets (40 pieces per assay) were washed three times with Dulbecco’s modified Eagle’s medium, then incubated in Ringer solution containing 10 mmol/l HEPES (pH 7.4), 0.2% BSA, and glucose at the specified concentrations for 1 h at 37°C. The islets were then transferred to 350 μl of the Ringer solution containing glucose at the specified concentrations. In some experiments, l-glutamate or its analogs at the indicated concentrations were added. At the times indicated, 70 μl of the Ringer solution was carefully removed, and the amounts of somatostatin, insulin, and glucagon were determined with enzyme-linked immunosolvent assay kits obtained from Phoenix Pharmaceuticals, Amersham Pharmacia Biotech, and Yanaihara, respectively (10). l-Glutamate was determined by high-performance liquid chromatography on a RESOLVE C18 column (3.9 × 150 mm; Waters) and by fluorescence detection (10). The statistical significance of datasets was assessed using Student’s t test.

As the first step of this study, we examined the expression of various iGluRs in islet cells using site-specific antibodies. Overall our results confirmed the observations of Weaver et al. (12): 1) GluR2 immunoreactivity overlapped with glucagon or insulin, but not with somatostatin, confirming the presence of GluR2 subunits in α- and β-cells, but not in δ-cells; 2) coexpression of the kainate receptor subunits (i.e., KA2 and GluR5–7) was not observed, suggesting there was no functional kainate receptor in any islet cells; and 3) only background levels of immunoreactivity for GluR1, GluR3, KA1, and NR1 were observed, indicating that they are not expressed in islet cells or that their levels are under the detection limit of our assay. In addition, the expression of δ1 and δ2 receptors, a fourth class of iGluRs (34), was not detected in islets (data not shown).

However, one significant difference was found for GluR4. Weaver et al. (12) did not detect GluR4 expression using a commercial antibody that recognizes the COOH-terminal region, which was confirmed by our assay (Fig. 1A). On the other hand, our antibody exhibiting binding affinity to the common domain of the NH₂-terminal half of GluR4 (anti-GluR4N) intensely immunostained β-, δ-, and F-cells, and, to a lesser extent, α-cells (Fig. 1B). Similarly, Western blotting confirmed that the anti-GluR4N antibody recognized a single polypeptide of 93 kDa, which was slightly lower than brain GluR4. In contrast, the commercial antibody did not recognize any islet polypeptides (Fig. 1C). On RT-PCR analysis, a DNA fragment encoding the common region of the GluR4 variants from islets and brain was amplified, confirming the expression of GluR4 type receptor in islets (Fig. 1D). These results strongly suggest that a GluR4 variant(s), most probably GluR4 lacking the COOH-terminal region, is expressed in islet cells.

GluR4 has seven possible splicing variants, some of which have not yet been detected at the protein level (Fig. 2A) (21,22). We performed RT-PCR to determine the DNA sequence of the COOH-terminal region of GluR4 variants. A DNA fragment (733 bp) encoding the GluR4 COOH-terminal region was not amplified from islet cDNA (Fig. 2B), confirming the absence of mRNA of authentic GluR4 in islets, as described above. In contrast, a DNA 700-bp fragment was amplified from islet cDNA, and the DNA sequence turned out to be that of a GluR4c-flip, the brain counterpart of which is GluR4c-flop (Fig. 2B). The nucleotide sequence of a cDNA cloned from an islet cDNA bank completely matched that of the GluR4c-flip. Neither GluR4d nor GluR4s mRNA was detected (Fig. 2B). Semiquantitative RT-PCR confirmed the dominant expression of GluR4c-flip in islets, whereas both GluR4c-flip and -flop variants were expressed to similar extents in the brain (Fig. 2C).

Next we prepared an antibody that specifically recognizes a recombinant GluR4c-flip and a single islet polypeptide band corresponding to the mass of the GluR4c-flip (Fig. 2D). As expected, immunohistochemical analysis indicated that islet cells, especially β- and δ-cells, expressed GluR4c subunits (Fig. 2E). The immunological localization was consistent with that of GluR4N immunoreactivity (Fig. 1B). These results demonstrated that δ-cells express GluR4c-flip at both the mRNA and protein levels.

The presence of iGluR on δ-cells strongly suggested that δ-cells can receive l-glutamate signals and that the resultant increase in cytoplasmic [Ca²⁺] may trigger somatostatin secretion. To test this hypothesis, we first measured the secretion of somatostatin derived from isolated islets under various glucose conditions. When islets were incubated with 3.3 mmol/l glucose throughout, the islets released a background level of somatostatin (114.8 ± 26.9 pg · 40 islets⁻¹ · h⁻¹; n = 4) (Fig. 3A). When islets were first incubated with 3.3 mmol/l glucose and then transferred to 16.7 mmol/l glucose (i.e., high-glucose stimulation), an enhanced level of somatostatin release (200% of that with 3.3 mmol/l glucose throughout) was observed. Continuous incubation of islets with 16.7 mmol/l glucose evoked the release of some somatostatin (169.1 ± 40.7 pg · 40 islets⁻¹ · h⁻¹; n = 4). When the islets were first incubated with 16.7 mmol/l glucose and then transferred to 3.3 mmol/l glucose (i.e., low-glucose stimulation), somatostatin secretion was significantly enhanced, corresponding to 147% of that with 16.7 mmol/l glucose throughout. Thus, high- and low-glucose stimulation enhanced somatostatin secretion, as has been observed in previous studies (3,15). Under similar assay conditions, the secretion of insulin and glucagon was enhanced by high- and low-glucose stimulation, respectively, indicating a well-known islet cell-specific response to glucose (Fig. 3A).

Next we examined the effect of exogenous l-glutamate on somatostatin secretion and found that 1 mmol/l l-glutamate enhanced somatostatin secretion (321.8 ± 31.1 pg · 40 islets⁻¹ · h⁻¹; n = 4), which was inhibited up to 60% by 6-cyano-7-nitroquinoxaline-2,3-dione (CNQX) (Fig. 3B). AMPA and kainate also stimulated somatostatin release ∼1.5- to 1.9-fold in a CNQX-sensitive manner. Neither N-methyl-d-aspartate (NMDA) nor MK-801, antagonists of the NMDA receptor, affected somatostatin secretion (Fig. 3B). CNQX alone inhibited somatostatin secretion by 55%, suggesting that endogenous l-glutamate cosecreted with glucagon from α-cells participates in the inhibition (Fig. 3B). l-Glutamate inhibited glucagon secretion without affecting insulin secretion, as previously described (10,13) (Fig. 3C). Somatostatin secretion evoked by endo- and exogenous l-glutamate was strongly inhibited by EGTA-acetoxymethylester (EGTA-AM) (Fig. 3D). These results strongly suggested that stimulation of the AMPA-type receptor triggers somatostatin secretion via an increase in [Ca²⁺]. It is noteworthy that the stimulatory effect of AMPA, kainate, and l-glutamate and the inhibitory effect of CNQX were not observed when islets were first incubated at 3.3 mmol/l glucose and then transferred to 16.7 mmol/l glucose (Fig. 3E). Essentially the same results were observed when islets were incubated at 16.7 mmol/l glucose throughout. Subsequently, the islets were then transferred to 3.3 mmol/l glucose, causing essentially the same stimulation by l-glutamate (data not shown). These results indicated that glutamatergic signaling was not involved in somatostatin secretion after high-glucose stimulation.

Finally, we checked whether exogenous somatostatin in turn could inhibit the secretion of glucagon and l-glutamate from islets. The secretion of both l-glutamate and glucagon was enhanced by low-glucose stimulation, a finding that was consistent with a previous observation (10). Somatostatin at 100 nmol/l decreased the secretion of both glucagon and l-glutamate to control levels (Fig. 4). Secretion of l-glutamate and glucagon was also enhanced by the addition of isoproterenol at 1 μmol/l; this secretion was inhibited by somatostatin as well (data not shown).

Besides insulin, glucagon, and pancreatic polypeptides, islets secrete l-glutamate from α- and possibly F-cells, GABA and Zn²⁺ from β-cells, and somatostatin from δ-cells as paracrine-like modulators (3,4,10,15,35). Although the modes of action of these paracrine modulators are less characterized, we recently showed that α- and β-cells communicate with each other through l-glutamate and GABA, which act as intercellular transmitters, to precisely regulate their endocrine functions (4). In the present study, we obtained evidence that α- and δ-cells also communicate through l-glutamate- and somatostatin-mediated signaling pathways.

We first examined whether δ-cells express iGluRs and possess the ability to receive glutamate signals. iGluRs are ligand-gated cation channels that are classified into AMPA, kainate, NMDA, and δ-type receptors based on their primary structure and pharmacology (36,37). The AMPA receptor is a homomeric or heteromeric complex consisting of GluR1 to -4 subunits. We showed that a splicing variant, GluR4c-flip, was actually expressed and present in islet cells; δ-cells express only GluR4c-flip, and β-cells express both GluR4c-flip and GluR2. As to the expression of mGluR, among known mGluRs, only mGluR4 has been shown to be present in α-cells (13). Collectively, it is quite likely that GluR4c-flip is the only GluR expressed in δ-cells.

The GluR4c-flip has also been detected in brain. In brain, the flop type seems to be dominant, whereas the flip type is more abundant in islets, with only a trace level of the flop type. The significance of such selective expression of the flip type in islets is unknown at present, but it may be related to a specific demand of islet glutamatergic systems.

We further explored the relationship between l-glutamate and somatostatin signaling. Somatostatin is confined to secretory granules and secreted upon increase in intracellular [Ca²⁺] and cAMP (14,15). We have shown that l-glutamate triggers the exocytosis of somatostatin, which is dependent on an increase in intracellular [Ca²⁺]. The glutamate effect was mimicked by AMPA and kainate and blocked by CNQX, suggesting the activation of AMPA-type GluRs in δ-cells. Even in the absence of exogenous l-glutamate, CNQX partially inhibited somatostatin secretion, probably by blockading endogenous glutamatergic signals from islet α-cells (Fig. 3B). Because l-glutamate becomes active as an intercellular transmitter only under low-glucose conditions (10), the present observations explain the mechanism by which somatostatin is secreted from δ-cells under such conditions.

It is noteworthy that under high-glucose conditions, δ-cells were not responsive to l-glutamate (Fig. 3). Similar glucose-dependent changes in glutamate signal reception have been observed for l-glutamate-evoked GABA secretion from β-cells (10). Although we do not know the molecular mechanisms underlying the glucose-dependent change in glutamate signal receptiveness, we assume desensitization of GluR4c-flip occurs because of a change in the metabolic state. It is known that flip/flop variants differ from each other in their agonist sensitivity and desensitization rate (38). Another plausible explanation is that, like GluR2 in neurons (39), GluR4c-flip at the plasma membrane is recycled depending on the glucose conditions, causing a loss or gain of sensitivity to glutamate.

Altogether, secretion of somatostatin from δ-cells may be regulated through two distinct signaling pathways: glucose-dependent gating of the ATP-sensitive potassium channels under high-glucose conditions, which is similar to insulin secretion from β-cells (3), and glutamate receptor-mediated signaling under low-glucose conditions, which is similar to GABA secretion from β-cells. Because l-glutamate is cosecreted with glucagon in islets under low-glucose conditions (4), the present results suggest the occurrence of l-glutamate- and somatostatin-mediated bidirectional signaling between α- and δ-cells. Thus, taking the previous observations into consideration, we propose that α-, β-, and δ-cells, as well as F-cells, may communicate with each other using l-glutamate, GABA, and somatostatin as intercellular transmitters (Fig. 5). l-Glutamate may act as a primary signaling molecule and thereby trigger secondary peptidergic and GABAergic responses in islets. Very recently, it has been shown that l-glutamate may act as an autocrine signaling molecule by way of mGluR4 on α-cells (13). These three distinct signaling pathways must regulate glucagon secretion from α-cells. In addition, recent evidence indicates that the vesicular inhibitory amino acid transporter, which is responsible for vesicular storage of GABA or glycine, is expressed in α-cells (40) and associated with glucagon-containing secretory granules (41). Thus either GABA or glycine may be costored and cosecreted with l-glutamate and glucagon and then may participate in the paracrine signaling shown above, although secretion of GABA or glycine from α-cells remains to be detected. Further studies, especially ones to reveal the components of the paracrine signal network at the molecular level and real- time imaging of the signal transmission, will be helpful for understanding the overall features of the complex, but well-ordered, regulation of secretion of blood glucose-regulating hormones.

This study was supported by a Grant-in-Aid for Scientific Research on Priority Areas-Advanced Brain Science Project from the Ministry of Education, Science, Sports, and Culture of Japan; grants from the Novartis Foundation, the Umami Science Foundation (Y.M.), the Kazato Science Foundation (M.H.); and by a Korea Science and Engineering Foundation Grant (D.S.K.). M.H. and S.Y. were supported by research fellowships from the Japan Society for the Promotion of Science for Young Scientists.

We thank Dr. M. Otsuka for preparation of the immunogens.