A Neuronal Isoform of Nitric Oxide Synthase Expressed in Pancreatic β-Cells Controls Insulin Secretion

The insulin-secreting cell line INS-1 was cultured according to the method of Asfari et al. (33). Islets were isolated from adult male Wistar rats (Iffa Credo, Lyon, France) using collagenase digestion. They were separated from the exocrine tissue and collected after centrifugation on a Ficoll density gradient (34). After isolation, islets were immediately incubated for 1 h, in the absence of added arginine, at 37°C in Krebs-Ringer bicarbonate buffer, pH 7.4, containing 1 g/l bovine serum albumin (BSA) and 2.8 mmol/l glucose.

Total RNA from rat isolated islets and INS-1 cells were extracted with TRIzol reagent (Life Technologies, Rockville, MD). The integrity of the RNA and the absence of contaminating genomic DNA were verified after migration on agarose gel and ethidium bromide staining (data not shown). First-strand cDNA was synthesized from 10 μg of total RNA in the presence of both 3 μg of random hexanucleotide primers (Life Technologies) and 1 μg oligo(dT) (Life Technologies) using Superscript II RNase H–Reverse Transcriptase (Life Technologies). PCR was performed in the presence of Hi-Taq DNA polymerase (Quantum Biotechnologies, Montreuil-Sous-Bois, France) with pairs of the following primers: for nNOS, 5′-ATGGAAGAGAACACGTTTGGGGTT-3′ and 5′-TTAGCTTGGGAGACTGAGCCAGCT-3′; for eNOS, 5′-CTGGTATCCTCTTGGCGGCGCAAG-3′ and 5′-CCTCCACTAGGCCAGGACGGTTGG-3′; for β2-microglobulin, 5′-ATCTTTCTGGTGCTTGTCTC-3′, and 5′-AGTGTGAGCCAGGATGTAG-3′. For each reaction, an initial denaturation for 5 min at 94°C was followed by 40 cycles of a denaturing step at 94°C for 1 min, an annealing step at 60°C for 1 min, an elongation step at 72°C for 1 min, and a final extension at 72°C for 10 min polymerase chain reaction (PCR) products were separated on 1.5% agarose gel and visualized by ethidium bromide staining. For sequencing, PCR fragments were purified after migration on agarose gel with the QIAEX II Extraction Kit (Qiagen S.A., Courtaboeuf, France). Fragments were sequenced manually using [α-³⁵S]dCTP (ICN, Costa Mesa, CA) and the Thermosequenase Cycle Sequencing Kit (Amersham Life Science, Little Chalfont, U.K.). For total sequencing of constitutive NOS, overlapping fragments of 450–650 bp were amplified using primers shown in Fig. 2. Each fragment obtained by reverse transcription (RT) PCR was sequenced twice manually and once using an ABI PRISM 377 automatic sequencer with the dRhodamine Terminator Cycle Sequencing Ready Reaction Kit (PE Applied Biosystems, Foster City, CA).

Fresly isolated islets and INS-1 cells were homogenized in 20 mmol/l Tris lysis buffer, pH 7.4, containing 150 mmol/l NaCl, 1% Triton X-100, 0.5% Nonidet P-40, 2 mmol/l EDTA, 1 mmol/l phenylmethylsulfonyl fluoride, 10 μg/ml leupeptin, and 10 μg/ml aprotinin. Insoluble materials were removed by centrifugation. The protein concentration of the supernatant was measured using the Coomassie Protein Assay Reagent (Pierce, Rockford, IL). Proteins of islet, INS-1 (50 μg), and cerebellum (5 μg) extracts were fractionated on a 7.5% SDS-polyacrylamide gel and transferred to a nitrocellulose membrane (Amersham Life Science). Filters were first blocked with 5% dried skim milk in phosphate-buffered saline containing 0.1% Tween 20 (PBS-T), and then incubated overnight with a monoclonal anti-nNOS antibody (diluted 1:750, Transduction Laboratories, Lexington, KY). After three washings in PBS-T, the membrane was finally incubated with a horseradish peroxidase–conjugated anti-mouse antibody (diluted 1:3,000; Sigma-Aldrich, Steinheim, Germany). Immunoreactivity was detected using an enhanced chemiluminescence reaction (Amersham Life Science).

INS-1 cells were first incubated in Krebs-Ringer bicarbonate buffer (without added arginine) containing 5 mmol/l glucose in the absence or the presence of 5 mmol/l l-NAME for 1 h, and nNOS was immunoprecipitated from 1 mg INS-1 and cerebellar protein extract with a polyclonal anti-nNOS antibody overnight (Transduction Laboratories). Protein A Sepharose was added for 1 h, and beads were washed three times with 10 mmol/l Tris, pH 7.4, and 150 mmol/l NaCl. Proteins were separated on a 15% tricine polyacrylamide gel, transferred to nitrocellulose, and further incubated first with a sheep anti-cytochrome c antibody (diluted 1:2,500; Biogenesis) and then with a peroxidase-conjugated anti-sheep antibody (diluted 1:5,000; Sigma-Aldrich).

INS-1 cells were seeded on poly-l-lysine (Sigma-Aldrich) coated Lab-Tek Chamber Slide System. They were fixed with 2% paraformaldehyde in PBS for 10 min and permeabilized 5 min in Triton X-100 0.1%. After blocking in BSA 2%, the cells were incubated with a rabbit polyclonal anti-nNOS antibody directed against the NH₂-terminus or the COOH-terminus of the enzyme (diluted 1:100; Euro-Diagnostica, Malmö, Sweden) and a guinea pig polyclonal anti-insulin antibody (diluted 1:300; ICN) overnight. After several washings, a rhodamine-conjugated anti-rabbit antibody (diluted 1:100, Biosys, Compiègne, France) and a fluorescein isothyocyate (FITC)-conjugated anti-guinea pig antibody (diluted 1:100; Vector Laboratories, Burlingame, CA) were applied to the cells for 1 h. After three additional washings, the cells were finally mounted in Citifluor (Citifluor, London, U.K.) and observed with the Leica TCS 4D confocal microscope equipped with an argon-krypton laser (Westlar, Germany).

Freshly isolated islets were fixed with 2% paraformaldehyde in 100 mmol/l phosphate buffer, pH 7. After washing in PBS containing 50 mmol/l NH₄Cl, the islets were soaked with 7.5% gelatin in phosphate buffer for 30 min at 37°C, centrifuged 10 min at 10,000g, and cooled in ice. They were then cryoprotected in 2.3 mol/l sucrose for 16 h at 4°C, cut in small blocks 1-mm large, mounted on a microtome specimen holder, and frozen in liquid nitrogen. Sections of 1-μm thickness were prepared according to the method of Reggio and Boller (35) on an ultracryomicrotome equipped with a cryodevia CR21 (RMC MT7, Tucson, AZ). Islet sections were first blocked in PBS containing 10% fetal calf serum and then incubated with the anti-nNOS antibodies (diluted 1:100) and the anti-insulin antibody (diluted 1:500) for 30 min at room temperature in PBS–10% fetal calf serum. After several washings, immunoreactivity was revealed with the rhodamine-conjugated antibody (diluted 1:100) and the FITC-conjugated antibody (diluted 1:100) in the same buffer for another period of 30 min. After washings, the sections were finally mounted in Citifluor and observed with the Leica confocal microscope.

A total of 200 freshly isolated islets were fixed in a solution of 100 mmol/l phosphate buffer, pH 7, containing 2.5% paraformaldehyde and 0.1% glutaraldehyde for 1 h at room temperature. After several washings in PBS–50 mmol/l NH₄Cl, the islets were postfixed in 1% osmium tetroxide for 2 min. The islets were then dehydrated in an ascending series of ethanol and routinely embedded in LR White (Electron Microscopy Sciences, Fort Washington, PA). Ultrathin sections of 60 nm were cut using a Reichert ultramicrotome (Ultracut S, Vienna, Austria) and deposited on gold grids. Sections were first treated with PBS containing 10% fetal calf serum and then incubated with a rabbit anti-nNOS antibody (diluted 1:100) and a guinea pig anti-insulin antibody (diluted 1:500) in the same buffer overnight at 4°C. After several washings in PBS, anti-rabbit and anti-guinea pig antibodies labeled, respectively, with 10 and 5 nm gold particles (diluted 1:25; British Biocell, Cardiff, U.K.) were applied to the sections for 1 h at room temperature. Sections were rinsed in deionized water, stained with 2% uranyl acetate for 20 min, and then observed with a transmission electron microscope (Hitachi H-7,100, Düsseldorf, Germany).

The specificity of the immune reaction, for both immunofluorescence and electron microscopy, was tested by replacing the primary antibody with a nonimmune rabbit serum or by incubating the sections with only the secondary antibody.

We used adult male Wistar rats (Iffa Credo) weighing 340–380 g. They were fed a standard pellet diet (U.A.R., Epinay sur Orge, France) ad libitum and had free access to tap water. Rat pancreata were isolated according to the procedure previously described (36), transferred to a thermostated chamber (37.5°C), and perfused with Krebs-Ringer Bicarbonate buffer (without added arginine, unless otherwise stated) supplemented with 2 g/l BSA and a basal 5 mmol/l glucose concentration. The ionic composition of the buffer was as follows: 108 mmol/l NaCl, 1.19 mmol/l KH₂PO₄, 4.74 mmol/l KCl, 2.54 mmol/l CaCl₂, 1.19 mmol/l MgSO₄ 7H₂O, and 18.0 mmol/l NaHCO₃. Continuous bubbling with a mixture of 95% O₂ and 5% CO₂ ensured an adequate oxygen supply and a pH close to 7.35. Circulation of the perfusion medium was performed with a peristaltic pump. Perfusion pressure, measured with a water manometer (35–45 cm H₂O), was selected to obtain a pancreatic flow rate of 2.5 ml/min. The medium was not allowed to recirculate and a pressure limiter returned the part not accepted by the organ (due to vascular resistance) to the origin reservoir. Pancreatic effluents were collected and measured in graduated test tubes; experiments began after a 30-min stabilization period.

Insulin concentrations in samples were determined by a radioimmunological assay (37) with Novo rat insulin as standard. The sensitivity of our assay was 0.1 ng/ml.

N^G-nitro-l-arginine methyl ester, l-arginine hydrochloride, succinic acid monomethyl ester, (±) miconazole nitrate salt, clotrimazole, and SNP dihydrate were purchased from the Sigma Chemical Company (St. Louis, MO).

Insulin outputs, calculated by multiplying the hormone concentration (nanogram per milliliter) in the effluent by the flow rate (milliliter per minute), were plotted on the graphs as means ± SE. In the text and in the figures, they are also given as mean integrated data obtained by calculating the areas under the curves (AUCs) during 20 min (nanograms × 20-min AUC). Both kinetic and integrated data were submitted to analysis of variance followed by the multiple comparison test of Newman-Keuls.

We used the RT-PCR approach to identify the isoform of NOS expressed in INS-1 and islet cells. Because studies using low-stringent Southern blot failed to find other isoforms of NOS (38), we hypothesized that NOS present in insulin-secreting β-cells was one of the constitutive NOSs already cloned, namely either the endothelial isoform or the neuronal isoform. We performed RT-PCR with primers based on the sequence of rat nNOS and on the sequence of rat eNOS. As shown in Fig. 1, a single band at the predicted size was obtained with specific primers for rat nNOS, whereas no fragment was amplified with the primers specific for rat eNOS. Sequencing of this fragment showed 100% identity with the PDZ domain of the nNOS cloned in rat cerebellum. This result demonstrates the expression of nNOS mRNA in INS-1 cells, which also strongly suggests the presence of nNOS in islet β-cells.

Using RT-PCR amplification of overlapping fragments, we sequenced the entire coding region of the nNOS cDNA expressed in INS-1 and islet cells. As shown in Fig. 2, we identified four mutations in the islet nNOS sequence and one more in INS-1 cells (data not shown); all of these mutations were confirmed by three independent experiments. Islet nNOS displayed a 99.8% homology with the cerebellar nNOS, suggesting that the isoform present in islets is encoded by the nNOS gene. Three of the mutations modified the amino acid sequence in islet and INS-1 cells: an isoleucine was mutated to a valine at position 269, a proline to an alanine at position 953, and a phenylalanine to a serine at position 1008. This result indicates that islet and INS-1 nNOS are only slightly different from the nNOS cloned in rat cerebellum and confer some degree of specificity to the pancreatic isoform.

Expression of the INS-1 and islet nNOS protein was confirmed by Western blot analysis using a monoclonal antibody that specifically recognized nNOS. As shown in Fig. 3, we identified a protein of the same molecular mass as the nNOS protein in the cerebellar extract. However, nNOS appears to be less expressed in the β-cells than in the cerebellum. INS-1 and pancreatic islet cells synthesized the constitutive neuronal isoform of NOS.

To confirm the presence of nNOS in insulin-secreting INS-1 and islet β-cells, we performed double-immunofluorescence studies with a polyclonal anti-nNOS antibody and an anti-insulin antibody. NOS staining was performed in several islet sections and INS-1 monolayers. Figure 4 shows that nNOS immunoreactivity is localized in both islet β-cells (upper panels) and in INS-1 cells (lower panels). Furthermore, the NOS and the insulin immunofluorescence appear to be largely colocalized, indicating that NOS is associated with insulin secretory granules. Faint nNOS staining was noted in the nucleus upon direct observation of these cells. The localization of the staining was found identical when performed with an antibody directed against either the NH₂- or the COOH-terminus. NOS staining was also observed in glucagon-secreting α-cells and, to a small extent, in somatostatin-secreting δ-cells (data not shown). Taken together, our results demonstrate that the nNOS gene is transcribed and that the protein is synthesized in insulin-producing β-cells.

To determine the subcellular localization of nNOS in β-cells, we performed electron microscopy with the same anti-nNOS antibodies. Figure 5A shows a view of β-secretory granules at different stages of maturation identified by staining with anti-insulin antibody (small gold particles). nNOS immunoreactivity was strongly associated with the insulin secretory granules; staining was found in the dense core, in the halo, or even associated with the membrane of the vesicles (big gold particles). We also found nNOS immunoreactivity in the mitochondria (Fig. 5B) and in the nucleus (data not shown). The nuclear localization of nNOS was confirmed by a Western blot analysis on a nuclear extract with the monoclonal anti-nNOS antibody (data not shown).

The neuronal nature of NOS present in pancreatic β-cells prompted us to determine the respective roles of NO production and cytochrome c reductase activity in the regulation of insulin secretion induced by different secretagogues. Therefore, we investigated the effect of a blockade of nNOS with a competitive inhibitor (l-NAME) in the presence of a substitution treatment with a chemical NO donor (SNP) and/or an inhibitor of nNOS cytochrome c reductase activity. Indeed, if the two catalytic activities have been well characterized for brain nNOS (28,30), there are until now no data available concerning a possible nNOS nonoxidating activity in pancreatic β-cells. This possibility was checked by using miconazole as an inhibitor of cytochrome c reductase activity in studies with three different insulin secretagogues, namely glucose, arginine, and methyl-succinate; it was further substantiated with a second inhibitor, clotrimazole, in studies with glucose.

Because imidazole antimycotics, including miconazole, have been reported to inhibit voltage-gated Ca²⁺ channels in GH₃ and adrenal chromaffin cells (39) and because the same effect has been seen on Ca²⁺-dependent K⁺ channels in red cells (40), we first performed a dose-response study to select the concentration of miconazole that did not affect the voltage-gated Ca²⁺ channels and the ATP-dependent K⁺ (K_ATP) channels, two major determinants of stimulus-secretion coupling in pancreatic β-cells (41). The two depolarizing agents tested, 0.185 mmol/l tolbutamide and 5 mmol/l KCl (data not shown), each induced a significant increase in insulin secretion, unaffected by 5 and 10 μmol/l miconazole (Fig. 6A). The tolbutamide effect was slightly reduced by 15 μmol/l (NS) and more markedly by 25 μmol/l (P < 0.05) and 50 μmol/l (P < 0.01) of the inhibitor. Thus, we chose the 10 μmol/l concentration of miconazole, slightly higher than the 8 μmol/l Ki (IC₅₀) (32) estimated for the inhibition of brain NOS cytochrome c reductase activity and that did not affect either K_ATP channels or voltage-gated Ca²⁺ channels (Fig. 6B).

The change of glucose concentration from five to 11 mmol/l induced the classical biphasic insulin response (Fig. 7A), achieving 178.6 ± 19.9 ng × 20 min. Infusion of 10 μmol/l miconazole 15 min before and during the 11 mmol/l glucose administration did not affect either basal insulin release or the first phase of high glucose–induced insulin release; the second phase (the last 15 min) appeared slightly but significantly (P < 0.05) reduced by ∼20%. The 20-min integrated response, however, was only decreased by 13% (NS; 178.6 ± 19.9 and 154.6 ± 15.9 ng × 20 min, respectively, in the absence and the presence of miconazole). This was not the case for 10 μmol/l clotrimazole, which was completely ineffective (Fig. 7A). Increasing miconazole concentrations to 15 and 25 μmol/l miconazole (P < 0.001) resulted in significant reductions in the glucose secretory effect (Fig. 7D). Again, clotrimazole was ineffective in the 0.2–20 μmol range (data not shown).

Blockade of nNOS with 5 mmol/l l-NAME provoked major alterations of β-cell responses to 11 mmol/l glucose (Fig. 7B). The biphasic pattern was severely blunted, and the magnitude of insulin release was strongly potentiated to 730.6 ± 57.2 ng × 20 min (P < 0.001) (i.e., a fourfold increase that was found to be reduced by 75% with 10 μmol/l miconazole). The 20-min integrated insulin response was very close to that recorded with glucose alone (177.2 ± 10.6 ng), but the secretory pattern remained clearly different from that observed with the sugar alone. It must be emphasized that the l-NAME–potentiating effect is already strongly decreased (P < 0.001) with 5 μmol/l miconazole (Fig. 7D). Similar data were obtained with 10 μmol/l clotrimazole (Fig. 7B), which was more potent (Ki: 0.8 μmol/l) than miconazole, but displayed a lower relative efficiency with a maximal inhibitory effect at 20 μmol/l (Fig. 7D). A major observation is that simultaneous administration of 10 μmol/l miconazole and 30 μmol/l SNP corrected this abnormality (Fig. 7C) and re-established a progressively rising second phase that reached 19.8 ± 4.5 ng/min after 20 min high glucose (P < 0.01 vs. 8.4 ± 1.0 ng/min with miconazole alone). Thus, a normal biphasic response to glucose could only be obtained by a simultaneous treatment with SNP and miconazole. Similar data were obtained with 10 μmol/l clotrimazole; a clearly biphasic response developed, but again, this inhibitor was quantitatively less efficient (Fig. 7C).

We were able to immunoprecipitate cytochrome c with an anti-nNOS antibody in an INS-1 cells and a cerebellar extracts, which demonstrates within β-cells a direct interaction between nNOS and cytochrome c (Fig. 8). Moreover, this interaction was strongly reinforced by incubation of β-cells with 5 mmol/l l-NAME, suggesting that cytochrome c reduction by nNOS increases in the presence of the NOS antagonist.

In the presence of 5 mmol/l l-NAME, the monophasic response to 5 mmol/l arginine alone (107.0 ± 11.0 ng × 20 min) was very strongly potentiated into a biphasic one with a 20-min integrated insulin secretion reaching 736.4 ± 69.4 ng × 20 min (P < 0.001). This effect appeared markedly sensitive to 10 μmol/l miconazole, since this antifungal imidazole at a concentration ineffective on the β-cell response to the positively charged amino acid alone (Fig. 9A) restored the insulin release in the presence of l-NAME (Fig. 9B) to values not far from those induced by arginine alone (184.2 ± 19.7 ng × 20 min; P < 0.01 vs. arginine alone).

Because the pathway for conversion of l-arginine and l-ornithine to glutamate with further catabolism to 2-ketoglutarate and succinyl CoA in the tricarboxylic acid (TCA) cycle is present in pancreatic islets (42), we were prompted to investigate the succinate secretory effect with special attention to interactions with the glucose effects and the relative sensitivity of both secretagogues to miconazole. Administration of the methylated TCA cycle intermediate, methyl-succinate (5 mmol/l), in the presence of 5 mmol/l glucose resulted in a clear biphasic response (Fig. 10A) (223.1 ± 20.9 ng × 20 min AUC). This response was slightly higher (P < 0.05) than that recorded with the 6 mmol/l increase in glucose concentration, but appeared more sensitive to miconazole inhibitory effect (−50%) and was conversely less responsive to 300 μmol/l SNP (−23%) when compared with glucose (Fig. 10B). The two drugs appeared to be able to exert additive effects because the combined treatment resulted in an overall 80% reduction of the methyl-succinate secretory effect.

The increase in glucose concentration from 5 to 11 mmol/l and the simultaneous administration of 5 mmol/l methyl-succinate provoked a dramatic increase in insulin release to 1,200 ± 128 ng × 20 min, corresponding to a threefold increase versus the value expected for an additive effect (Fig. 11A). However, this potentiating effect appeared to be extremely sensitive to miconazole, which reduced the effect by 80% to 236.9 ± 29.4 ng × 20 min and, to a slightly lesser extent, to a high (300 μmol/l) SNP concentration (75%; 296.1 ± 40.5 ng × 20 min); both drugs were again able, when combined (Fig. 11B), to exert a synergistic inhibitory effect (89.3 ± 18.4 ng × 20 min).

Our present work clearly demonstrates that pancreatic insulin-secreting β-cells express a constitutive neuronal isoform of NOS and that the enzyme is able to exert two catalytic activities: the production of NO and the reduction of cytochrome c, both of which appear to be involved in the control of β-cell function.

Pancreatic β-cell nNOS displays 99.8% homology with cerebellar NOS and is probably encoded by the brain NOS I gene. Among the three amino acid mutations identified in our study, one deserves particular attention: the mutation in position 953 of a proline to an alanine may suppress a bend in the secondary structure of pancreatic β-cell nNOS, making the protein structurally different from the enzyme present in neurons. Such a difference could affect the binding of pancreatic β-cell nNOS to the various anti-nNOS antibodies used for its detection and hence account for the apparent discordant data reported in the literature concerning localization of this enzyme. If these small differences confer some specificity to the isoform present in β-cells, the enzyme should be added to the impressive list of molecular markers and functional properties shared by pancreatic β-cells and neurons (43). Furthermore, and interestingly, in neurons strong nNOS immunoreactivity has been found localized in synaptic vesicles and in the cristae of some large mitochondria (44). nNOS has long been considered to be mainly soluble due to the absence of a consensus sequence for NH₂-terminal myristoylation and palmitoylation present in eNOS (45), which targets it to the membrane. There is now growing evidence that nNOS is essentially particulate (46) as also apparent in our study. This is probably attributable to the presence in the NH₂-terminal domain of a PDZ motif that is unique to nNOS and participates in protein-protein interactions at the membrane level (47). The localization of nNOS mainly in insulin secretory granules at different stages of their maturation raises the question of the possibility that the enzyme could play a role in insulin maturation and/or the complex process of exocytosis. Along this same line, the presence of nNOS in the nucleus could be related to the amplifying effect of NO on Ca²⁺-induced gene transcription reported in neuronal cells (48).

In our previous and current studies, we show that nNOS exerts a negative control on insulin secretion, based on the stimulating effect of the enzyme antagonist l-NAME. This result has been challenged by studies showing that l-NAME, at concentrations between 5 and 20 mmol/l, is able to reduce whole-cell K_ATP currents (49,50) and has thereby been claimed to increase glucose-induced insulin secretion. However, even if such an effect cannot be excluded, a number of features argue that it should be considered of minor importance. First, l-NAME at high concentrations affects β-cell K_ATP current within seconds; this contrasts with the 5-min delay necessary for the antagonist to potentiate glucose- and arginine-induced insulin secretion (22,27); such a delay is, however, in agreement with the pro-drug status of l-NAME, which requires hydrolytic bioactivation into N^G-nitro-l-arginine, the rate of which depends on the nature of tissues (51). Second, l-NAME is able to potentiate the glucose and arginine secretory effects in the presence of a K_ATP channel opener, diazoxide (i.e., independently of K_ATP channels) (52). Third, we previously showed that l-NAME was able to significantly affect the biphasic pattern of insulin response to glucose at very low concentrations (0.1–0.5 mmol/l) (21,22). Finally, we tested 7-nitroindazole, another potent inhibitor more specific than l-NAME for brain NOS (53), which enhances glucose-induced insulin secretion in pancreatic islets (26) and dose-dependently (20–100 μmol/l) potentiates the biphasic insulin response to glucose in the isolated perfused rat pancreas (data not shown). Hence, there is convincing pharmacological evidence that l-NAME effects result from a specific blockade of constitutive NOS activity. However, our previous studies pointed out that decreased NO production only partly accounts for l-NAME–induced alteration of the glucose secretory effect (22). Indeed, the NO donor, SNP, at a concentration able to inhibit the insulin response to glucose alone, only partly corrected the l-NAME–induced alterations. Our original data show that two antifungal imidazoles, miconazole and clotrimazole, were able to consistently counteract l-NAME–induced alteration of the β-cell response to the sugar. From our dose-response study of the effect of miconazole on tolbutamide, KCl, glucose, and l-NAME plus glucose, it appears very unlikely that miconazole, at the concentration used, interferes with Ca²⁺ and K_ATP channels and/or other major Ca²⁺/calmodulin-regulated processes. Therefore, the potentiation of the β-cell response to glucose, after blockade of nNOS with l-NAME, might result from both a decreased NO production and an increased cytochrome c reductase activity. Indeed, the NADPH-derived electrons transferred to nNOS are mainly channeled by its reductase domain to cytochrome c due to blockade of the oxidating site by the nonmetabolizable arginine analog; such an increase in the respiratory chain activity should result into a rise in ATP production and subsequently in insulin secretion. This is strongly comforted by our demonstration that nNOS coimmunoprecipitates markedly with cytochrome c after exposure of β-cells with l-NAME, suggesting an increased velocity of cytochrome c reduction by nNOS. Our finding in β-cells is original in that such a direct molecular interaction and the cytochrome c reductase activity of brain NOS have been described with isolated purified proteins (31). Thus, the β-cells nNOS isoform is able to exert two catalytic activities, an NO production and a cytochrome c reductase activity, as shown for brain NOS. In addition, from the restoration of a normal biphasic secretory pattern upon simultaneous administration of the exogenous NO donor and miconazole (and to a lesser extent clotrimazole), we suggest that a normal balance between both activities is essential for a normal β-cell response to glucose.

As for arginine, we have previously shown that decreased NO production is of minor importance in the potentiation of the amino acid insulinotropic effect observed after nNOS blockade. The production of urea and ornithine from arginine by arginase is a very active pathway in pancreatic islets (42), and arginase has been shown to be inhibited by citrulline (54), the coproduct of and N^G-OH-l-arginine (55), the intermediate in NO synthesis by NOS (28,29). Hence, we were able to show that the potentiation of the arginine secretory effect in the presence of l-NAME resulted from the expected suppression of the two endogenous inhibitors by the NOS antagonist (27) and from a subsequent increased arginase activity. Furthermore, our data show that the inhibitor miconazole almost completely restored a normal response to the amino acid after nNOS blockade, suggesting a major role for a nNOS-related calmodulin-dependent cytochrome c reductase activity that interferes with increased arginase activity. This proposal is strongly supported by the demonstration of the presence of arginase II (different from the liver cytosolic isozyme) in the mitochondrial matrix of most tissues (56), where the enzyme is involved in the production of ornithine as a precursor of glutamate, which might enter the TCA cycle after further metabolism to succinyl CoA.

The TCA cycle intermediate succinate, which bypasses glycolysis when infused as a methylester, is known to mimic the glucose effect on insulin secretion (57) and to provide FADH₂ to site II of mitochondrial respiratory chain. The great stimulating effect of methyl-succinate, a potent “anaplerotic” secretagogue (57), is probably attributable to the interplay of different sets of metabolic coupling factors. The first set is sensitive to the high concentration of the exogenous NO donor, in agreement with the inhibitory effect of NO on cytochrome oxidase (58). Such an effect accounts for almost the entire glucose stimulating effect, which is also in accordance with the inhibition by NO of two glycolytic enzymes, phosphofructokinase (26) and glyceraldehyde phosphate dehydrogenase (59). The other set is insensitive to the high SNP concentration, including the recently described mitochondrially derived factor (60,61), but also a methyl-succinate–induced increase in nNOS cytochrome c reductase activity, which may account for the strong synergistic effect between glucose and methyl-succinate on β-cell function. Such an effect is likely to be more pronounced, as feeding the TCA cycle with succinate provides an essential factor for nNOS activation, NADPH, via the pyruvate-malate shuttle shown to be activated by methyl-succinate and proposed as the major means of generating cytosolic NADPH) (62). Furthermore, NOS has been found constitutively active in succinate-energized rat liver mitochondria (63).

If we cannot exclude an interplay of other calmodulin-regulated processes affected by NOS blockade with l-NAME, the involvement of a cytochrome c reductase activity is nevertheless strongly suggested by 1) our demonstration that nNOS is also present in β-cell mitochondria, 2) the direct molecular interaction between nNOS and cytochrome c, and 3) the neuronal nature of the β-cell enzyme, reductase domain of which has been shown in brain to be able to transfer electrons to cytochrome c at a rate much greater than the maximum rate of NO production (64). Moreover, we were able to detect an in vitro cytochrome c reductase activity for nNOS purified from INS-1 cells (32). This activity was weak and underestimated because of the low expression level and the very small amount of nNOS we recovered, but was very sensitive to the inhibitory effect of miconazole (∼39% at a concentration of 10 μmol/l, data not shown).

In summary, we report for the first time the expression of a constitutive neuronal isoform of NOS in rat pancreatic insulin-secreting β-cells, where it is mainly present in insulin secretory granules, in the nucleus, and notably in the mitochondria. Another original finding suggested by our data is that, as shown for brain nNOS, pancreatic β-cell nNOS is able to exert two different catalytic activities and that a normal balance between NO production and a nonoxidating calmodulin-dependent reductase activity of the enzyme appears to be required for a biphasic physiological response of β-cell to glucose. From our data, it might be concluded that dysfunction of pancreatic β-cell nNOS, resulting from an imbalance between oxidating and nonoxidating activities, could be considered a defect involved in the pathogenesis of certain situations characterized by β-cell dysfunction and hypersinsulinism.

This study was supported by grants from the Swiss National Science Foundation (no. 32-49755.96) to C.B.W. and from the Association pour la Recherche sur le Cancer (A.R.C.) to A.D.L.

We are very grateful to Valérie Montesinos for typing the manuscript and Michel Tournier, Jacqueline Boyer, and Chantal Clement for expert technical assistance. We also thank Marie Margout, Paul Paulet, and Thierry Pujol from the CRIC for assistance with the electron microscopy studies. The authors are indebted to Sharon Lynn Salhi for extensive correction of the manuscript.