The Actions of a Novel Potent Islet β-Cell–Specific ATP-Sensitive K⁺ Channel Opener Can Be Modulated by Syntaxin-1A Acting on Sulfonylurea Receptor 1 Free

The constructs pcDNA3–syntaxin-2 and pcDNA3–syntaxin-1A were from Richard Scheller (Genentech, San Francisco, CA), pGEX-4T-1–syntaxin-1A from W. Trimble (Hospital for Sick Children, Toronto, Canada), and pcDNA3–syntaxin-1B from G. Zamponi (University of Calgary, Calgary, Canada). The coding sequences corresponding to syntaxin-1A (amino acid 1–266, without transmembrane domain), syntaxin-1A–Habc domain (amino acid 1–160), and syntaxin-1A–H3 domain (amino acid 191–256) were amplified by PCR and cloned into pGEX-4T-1 expression vector for generation of glutathione S-transferase (GST) fusion proteins. GST fusion protein expression, purification, and thrombin (Sigma) cleavage were performed according to the manufacturer's instructions (Amersham Biosciences, Piscataway, NJ).

HEK-293 cells stably expressing human Kir6.2/SUR1 (BA8 cells) have been previously characterized (24,27), and cell culture and transfection were performed as previously described (33–35). For a detailed description, refer to the supplementary appendix, available online at http://dx.doi.org/10.2337/db07-0030.

Pancreatic islets were isolated from male Sprague-Dawley rats by collagenase digestion and dispersed into single cells by treatment with 0.05% trypsin in Ca²⁺- and Mg²⁺-free PBS as previously described (33,34). Islet cells were plated on glass coverslips in 35-mm tissue culture dishes and cultured overnight in RPMI-1640 medium (Invitrogen), containing 2.8 mmol/l glucose and supplemented with 7.5% FBS, before electrophysiological recordings.

Current recordings from single-islet β-cells and BA8 cells were performed using standard patch-clamp techniques in the whole-cell and inside-out configurations as previously described (33–35). Refer to the supplementary appendix for a detailed description.

BA8 cells were washed with 1× PBS (pH 7.4) and harvested in binding buffer (25 mmol/l HEPES, pH 7.4, 100 mmol/l KCl, 1.5% Triton X-100, 2 μmol/l pepstatin A, 1 μg/ml leupeptin, and 10 μg/ml aprotinin). The cells were lysed by sonication, and insoluble materials were removed by centrifugation (55,000g, 4°C, 30 min). The detergent extract (0.3 ml, 1.2 μg/μl protein) of BA8 cells was incubated for 2 h at 4°C with GST (as negative control) or GST–syntaxin-1A (500 pmol protein bound to glutathione agarose beads) in the absence or presence of indicated concentrations of NNC55-0462 or untagged syntaxin-1A and/or ATP. The beads were washed three times with binding buffer, the samples were separated on 10% SDS-PAGE and transferred to nitrocellulose membrane, and the proteins were identified with mouse anti-SUR1 antibody (1:750; a gift from J. Ferrer, University of Barcelona, Barcelona, Spain). For Western blotting analysis of syntaxin-1A, -1B, and -2 expression, whole-cell lysates were prepared from rat brain (as positive control), rat islet, syntaxin-1A–or syntaxin-1B–expressing HEK293 cells (as negative or positive control), and rat pancreatic acinar cells (as positive control) by sonication or with a homogenizer. Samples were separated by 12% SDS-PAGE, and proteins of interest were identified with specific antibodies (syntaxin-1A and -1B from Syntaxinaptic Systems [Goettingen, Germany] and syntaxin-2 from V. Olkkonen [Helsinki, Finland]).

Data analysis and curve fitting were performed using Origin 6.0 (Microcal Software, Northampton, MA). Data are means ± SE. Differences in means were compared using either one-way ANOVA, followed by Dunnett's post hoc test, or paired/unpaired Student's t test, with P < 0.05 considered as statistically significant. Concentration-response curves were fitted to the equation: y = {(A₁-A₂)/[1+(X/X₀)^ρ]} + A₂, where y is the K_ATP current at different concentrations of NNC55-0462, X is the NNC55-0462 concentration applied externally to the cells, X₀ is the NNC55-0462 concentration that produced half-maximal K_ATP channel activation, A₁ is the K_ATP current before NNC55-0462 was applied, A₂ is the maximal K_ATP current level activated by NNC55-0462, and ρ is the slope of the curve.

Previously, we had reported that syntaxin-1A inhibited Kir6.2/SUR1 currents by its actions on nucleotide binding fold-1 and -2 of SUR1 (33,34). Here, we examined the possibility that NNC55-0462, belonging to a class of thiadiazine dioxide K_ATP channel openers (Fig. 1) and acting on distinct domains in SUR1 (23,27), might influence the actions of syntaxin-1A on the nucleotide binding folds, or vice versa. To examine the direct functional effects of syntaxin-1A on K_ATP channels, we again first used BA8 cells, a HEK cell line stably expressing human Kir6.2/SUR1 (24,27) that does not contain endogenous syntaxin-1A (33), and dialyzed into the cytoplasm of these cells soluble GST–syntaxin-1A lacking its transmembrane domain. Here, external perfusion of NNC55-0462 dose-dependently activated Kir6.2/SUR1 K_ATP channels, effective starting at 0.1 μmol/l and achieving maximum effect at 1 μmol/l (Fig. 2A). A higher concentration of NNC55-0462 (3 μmol/l) did not elicit a further increase in channel amplitude; instead, it induced channel rundown. GST–syntaxin-1A (100 nmol/l) (Fig. 2B), when compared with GST control (Fig. 2A), reduced the ability of NNC55-0462 to open K_ATP channels. We analyzed the data in several ways to show that syntaxin-1A (10 and 100 nmol/l) reduced both drug potency (Fig. 2C) and efficacy (Fig. 2D).

Figure 2C examines drug potency, where dose-response curves were normalized to each group's maximal channel activation at 1 μmol/l NNC55-0462, which demonstrated a rightward shift in potency caused by GST–syntaxin-1A dialysis. This is further illustrated in Fig. 2E, showing the half-maximal effective concentration (EC₅₀) of NNC55-0462 in GST control group was 0.11 ± 0.02 μmol/l (n = 7), whereas GST–syntaxin-1A dialysis at 10 and 100 nmol/l increased it to 0.15 ± 0.05 μmol/l (n = 5, not significant) and 0.21 ± 0.04 μmol/l (n = 7, P < 0.05), respectively. Fig. 2D examines drug efficacy, where dose-response curves were normalized to the maximum channel activation observed with the GST-dialyzed control group, which demonstrated a downward shift in efficacy caused by GST–syntaxin-1A dialysis. This is further illustrated in Fig. 2E, where maximum current density activated by NNC55-0462 was 283 ± 14 pA/pF (n = 7) with GST dialysis, whereas syntaxin-1A dialysis at 10 and 100 nmol/l decreased it to 256 ± 28 pA/pF (n = 5, not significant) and 194 ± 12 pA/pF (n = 7, P < 0.01), respectively, with the latter being a significant reduction of 31 ± 4%.

To reiterate, these modulatory effects of syntaxin-1A on the effectiveness of NNC55-0462 on BA8 K_ATP channels are caused by direct interactions with the SUR1 protein because BA8 cells are devoid of endogenous syntaxin-1A and syntaxin-1A–interacting proteins (33). Note that the concentrations of GST–syntaxin-1A required likely represent an overestimation because of suboptimal targeting of soluble GST–syntaxin-1A to the sites of K_ATP channels on the plasma membrane.

To demonstrate the physiological implication of the BA8 study, we performed similar studies on rat islet β-cells. Shown in Fig. 3A and B, NNC55-0462 dose-dependently activated K_ATP channel activity in rat islet β-cells, effective starting at 0.03 μmol/l, with maximal effects reached at 1 μmol/l. Like BA8 cells, β-cell K_ATP channels displayed rundown on further perfusion with a supramaximal concentration of NNC55-0462 at 3 μmol/l. The β-cell currents were sensitive to tolbutamide inhibition applied in the presence of 3 μmol/l NNC55-0462, confirming the identity of the activated currents to be K_ATP currents (IK_ATP). K_ATP channel openers have negative allosteric effects on glibenclamide binding to SUR (36,37). In fact, such SUR1-specific K_ATP channel openers could displace ³H-glibenclamide binding (24), which, taken along with the high affinity of NNC55-0462 for SUR1, led us to use a higher concentration of tolbutamide to inhibit the channels.

Consistent with the BA8 results (Fig. 2), both potency (Fig. 3C) and efficacy (Fig. 3D) of NNC55-0462 in opening islet β-cell K_ATP channels were decreased by dialysis of GST–syntaxin-1A into the β-cell cytoplasm. Here, GST–syntaxin-1A (1 μmol/l) increased the EC₅₀ of NNC55-0462 from 0.050 ± 0.009 μmol/l (n = 10) to 0.11 ± 0.03 μmol/l (n = 7, P < 0.05) (Fig. 3E). With respect to efficacy, the maximum current density activated by NNC55-0462 was suppressed by GST–syntaxin-1A dialysis from 157 ± 8 pA/pF (n = 10) to 110 ± 4 pA/pF (n = 7, P < 0.01), which is a 30 ± 3% reduction (Fig. 3E). It is possible that soluble GST–syntaxin-1A lacking the transmembrane domain could bind a number of syntaxin-1A binding proteins in β-cells that might have independent effects on K_ATP channels or, alternatively, sequester syntaxin-1A (i.e., endogenous v-SNAREs on insulin granules) to reduce the amount of GST–syntaxin-1A available for K_ATP channel interaction. Although we could not rule out these possibilities, their contributions are not likely to be significant in our study because the results of dialyzed syntaxin-1A on β-cell K_ATP channel opener potency and efficacy were remarkably similar to those of BA8 cells, albeit the K_ATP channel density is much higher in the K_ATP-overexpressing BA8 cell line.

It is possible that other syntaxins in the β-cell (38) might have actions similar to that of syntaxin-1A on K_ATP channels. Syntaxin-1B and -2 exhibit the highest homology to syntaxin-1A (>90 and >70%, respectively) (39). Using Syntaxin isoform–specific antibodies and overexpressed proteins as controls, Fig. 4A shows that syntaxin-2 is abundant in rat islets, but not syntaxin-1B, the latter confirming a previous report (40). We therefore proceeded to explore the specificity of action of syntaxin-1A on NNC55-0462–induced β-cell K_ATP channel opening by comparing the effects of syntaxin-1A versus syntaxin-2.

BA8 cells were transiently transfected with green fluorescence protein (GFP) alone as control, or cotransfected with full-length (including transmembrane domain) syntaxin-1A or -2 to localize the expressed syntaxin to the plasma membrane, where K_ATP channels are located. Fig. 4B shows representative current-voltage relationships demonstrating that maximum current density activated by 1 μmol/l NNC55-0462 was specifically reduced in the syntaxin-1A–but not the syntaxin-2–transfected group. Activated currents were confirmed to be of I_KATP by their sensitivity to tolbutamide inhibition (Fig. 4C), applied in the presence of NNC55-0462 after maximal channel activation had occurred. Fig. 4D summarizes these results, showing that the maximum current density activated by NNC55-0462 (at −120 mV) was reduced by GFP/syntaxin-1A transfection to 206 ± 14 pA/pF (n = 10, P < 0.01) from 323 ± 14 pA/pF in the GFP control group (n = 9), which is a ∼36% reduction. The lack of effect of GFP/syntaxin-2 transfection on K_ATP channel opening (349 ± 17 pA/pF, n = 11, not significant), despite its high homology to syntaxin-1A, indicates that syntaxin-1A effects on K_ATP channels are isoform specific within the syntaxin family in β-cells.

We used the inside-out patch configuration on BA8 cells (Fig. 5) for patch-clamp recording to examine: 1) if NNC55-0462 can still open K_ATP channels when applied intracellularly, 2) if syntaxin-1A could influence these NNC55-0462 actions, and 3) which domain within syntaxin-1A (H3 versus Habc) effects these actions. Membrane patches were first exposed to 0 and 3 mmol/l Mg-ATP to confirm the ATP sensitivity of currents recorded. Patches were then held at 0.3 mmol/l Mg-ATP to produce partial channel inhibition and to prevent rapid channel rundown. NNC55-0462 (0.3 μmol/l) was applied in a 0.3 mmol/l Mg-ATP environment to induce channel opening, after which GST, GST–syntaxin-1A, GST–syntaxin-1A–Habc, or GST–syntaxin-1A–H3 was applied to the solution. NNC55-0462 was able to induce K_ATP channel opening when applied on the intracellular side of the membrane, achieving a remarkable >80% of maximum channel opening at 0.3 μmol/l (Fig. 5A–D). Application of GST (Fig. 5A) or GST–syntaxin-1A–Habc (300 nmol/l) (Fig. 5C) had no effect on NNC55-0462–mediated channel opening, whereas GST–syntaxin-1A (Fig. 5B) and GST–syntaxin-1A–H3 (300 nmol/l) (Fig. 5D) significantly suppressed the current activity activated by NNC55-0462. Figure 5E summarizes the data showing that the percentage of maximal current activated by NNC55-0462 in the presence of GST (92 ± 13%, n = 9) was not significantly different from NNC55-0462 alone (84 ± 11%), whereas NNC55-0462 plus GST–syntaxin-1A showed a current of 39 ± 6% (n = 9), which was a significant reduction of ∼52% (P < 0.01) from NNC55-0462 alone (91 ± 10%). Although the percentage of current activation by NNC55-0462 plus GST–syntaxin-1A–Habc (88 ± 10%, n = 6) was not significantly different from NNC55-0462 alone (93 ± 10%), NNC55-0462 plus GST–syntaxin-1A–H3 caused a similar (to GST–syntaxin-1A), ∼54% reduction of current (38 ± 7%, n = 6, P < 0.01) compared with NNC55-0462 alone (92 ± 6%). These results indicate that 1) NNC55-0462 can access its site of action on K_ATP channels when pplied intracellularly and that 2) the H3 domain of syntaxin-1A is the putative syntaxin-1A domain inhibiting K_ATP channels previously activated by this K_ATP channel opener.

We next examined whether the effect of syntaxin-1A is specific to this particular K_ATP channel opener compound. We chose to study the parent compound, diazoxide, which shares the same backbone structure as NNC55-0462 (Fig. 1) but interacts with SUR1 at slightly different domains. We repeated the experiment performed with NNC55-0462 on rat islet β-cells, but instead we perfused increasing concentrations of diazoxide (3–300 μmol/l). Similar to the NNC55-0462 study (Fig. 3), GST–syntaxin-1A (Fig. 6B) dialyzed into the cytoplasm of rat islet β-cells significantly reduced the potency and efficacy of diazoxide, compared with GST dialysis (Fig. 6A). Analysis of drug potency (Fig. 6C and E) shows the EC₅₀ of diazoxide was increased from 21 ± 5 μmol/l (n = 10) to 42 ± 7 μmol/l (n = 7, P < 0.05) by GST–syntaxin-1A dialysis. Analysis of drug efficacy (Fig. 6D and E) shows the maximum current density activated by diazoxide was reduced by GST–syntaxin-1A from 136 ± 4 pA/pF (n = 10) to 92 ± 5 pA/pF (n = 7, P < 0.01), which is a ∼32% inhibition. These results together suggest that the effect of syntaxin-1A on the opening of K_ATP channels by K_ATP channel openers is not specific to NNC55-0462 alone, but rather is shared among different dioxide compounds.

Because we showed that NNC55-0462 was effective when applied not only extracellularly but also intracellularly, the latter suggests the possibility that this K_ATP channel opener might share common binding domains with syntaxin-1A on SUR1 nucleotide binding folds to influence K_ATP channel opening. We therefore examined whether this K_ATP channel opener could affect syntaxin-1A–SUR1 binding interactions. GST–syntaxin-1A bound to glutathione agarose beads was able to pull down the same amount of SUR1 protein from BA8 cell lysate extract with increasing NNC55-0462 concentrations (up to 10 μmol/l) both in the absence (Fig. 7A) or presence of 0.3 mmol/l ATP (Fig. 7B) or 5 mmol/l ATP (Fig. 7C). ATP at 5 mmol/l appeared to reduce the amount of SUR1 pulled down by GST–syntaxin-1A at all concentrations of NNC55-0462 (Fig. 7C). A positive control with increasing concentrations of thrombin-cleaved syntaxin-1A (instead of NNC55-0462) showed that the assay used is indeed capable of detecting competitive binding inhibition (Fig. 7D). A negative control experiment showed that GST bound to beads could not pull down SUR1, representing nonspecific binding (Fig. 7E); however, glutathione agarose beads alone did bind a small amount of SUR1 (∼20% of that bound by GST–syntaxin-1A). This inability of NNC55-0462 to modulate syntaxin-1A binding to SUR1 suggests that this K_ATP channel opener does not likely bind the same domains that syntaxin-1A binds to in SUR1 at its nucleotide binding folds (33,34). The ability of syntaxin-1A to block this K_ATP channel opener action on K_ATP channels suggests that syntaxin-1A could modulate NNC55-0462 binding to the SUR1 protein. However, because of the lack of a radiolabeled NNC55-0462, we were unable to investigate this possibility.

The current study is the first to demonstrate an effect of the SNARE protein syntaxin-1A on the actions of NNC55-0462, a member of a novel class of thiadiazine dioxide K_ATP channel openers, and its parent compound diazoxide. Our results showed that syntaxin-1A reduces the potency and, more significantly, the efficacy of NNC55-0462 and diazoxide. In inside-out membrane patches, syntaxin-1A can also inhibit channels previously activated by NNC55-0462. Within the syntaxin family present in β-cells, these actions were specific to syntaxin-1A because syntaxin-2, which shares >70% homology with syntaxin-1A (39), had no effect on K_ATP channels. We further determined that the H3 domain, and not the Habc domain, is the putative syntaxin-1A domain that modulated the K_ATP channel opener's actions on the K_ATP channel. Because syntaxin-1A also influenced the K_ATP channel–opening actions of diazoxide, a nonspecific K_ATP channel opener, this would suggest that syntaxin-1A might also influence the other K_ATP channel openers’ actions on SUR2 proteins, which will require further study.

The isoform specificity of syntaxin-1A's inhibition on NNC55-0462–mediated K_ATP channel opening is in line with previous studies examining other syntaxin-1A–ion channel interactions. Studies with Lc- and N-type calcium channels showed similar results, with the basis of such specificity mapped out to two critical cysteine residues found in syntaxin-1A but not in syntaxin-2 (41). However, these cysteine residues on the transmembrane domain of syntaxin-1A are unlikely candidates for the basis of its actions toward K_ATP channel modulation because: 1) a soluble GST–syntaxin-1A fusion protein lacking the transmembrane domain inhibited NNC55-0462–mediated K_ATP channel opening as efficaciously as expression of the full-length syntaxin-1A, and 2) the soluble GST–syntaxin-1A–H3 domain induced a similar inhibition as syntaxin-1A, consistent with our previous study, which mapped out the nucleotide binding fold–interacting domain to be the H3 domain (33). Nonetheless, that report (41) does point out that differences as small as one to two amino acid residues are enough to produce isoform-specific action. Moreover, a study of syntaxin-1A versus syntaxin-1B, which shares 93% homology, shows that only syntaxin-1A is capable of mediating tonic voltage-dependent inhibition on N-type calcium channels (42). With regard to other K⁺ channels, we previously reported that syntaxin-1A, but not syntaxin-2, reduced Kv2.1 current density by direct actions on specific Kv2.1 cytoplasmic domains and by inhibiting channel trafficking (43).

Although the mechanisms of action of syntaxin-1A on calcium channel regulation have been well characterized, those involved in K_ATP channel modulation are less clear. We postulated three possibilities: 1) syntaxin-1A may influence these K_ATP channel openers’ interaction with SUR1, 2) conversely, these K_ATP channel openers may influence syntaxin-1A interactions with SUR1, or 3) syntaxin-1A may influence the mechanism of action of K_ATP channel openers by interfering with adenosine nucleotide binding and/or hydrolysis.

For the first possibility, syntaxin-1A interacting with the nucleotide binding folds may induce conformational changes to SUR1 that affect the affinity of NNC55-0462 to its binding sites. Unfortunately, NNC55-0462 could not be labeled to directly test this possibility. Nonetheless, direct competitive inhibition is unlikely because of the inability of increasing concentrations of NNC55-0462 to overcome the reduced efficacy caused by syntaxin-1A. The SUR1 domains critical to the actions of a closely related compound, NNC55–9216, were mapped out to transmembrane domains 8–11 and both nucleotide binding folds (24). Because syntaxin-1A can bind both nucleotide binding folds of SUR1 (33,34), such interactions may induce negative allosteric modulation onto the K_ATP channel opener binding sites of SUR1, thereby decreasing the K_ATP channel opener binding affinity. In addition, previous studies have determined the sites of action of diazoxide to be transmembrane domains 6–11 and nucleotide binding fold-1 (44). Therefore, it is not surprising that syntaxin-1A is capable of influencing the effects of both K_ATP channel openers because of potential overlapping binding or signal transduction sites. Further mutational studies will be required to elucidate these sites.

For the second possibility, NNC55-0462 may act to relieve syntaxin-1A inhibition on K_ATP channels, and thus increased levels of syntaxin-1A would reduce the effectiveness of this K_ATP channel opener. Our in vitro binding studies showed that increasing concentrations of NNC55-0462 did not affect syntaxin-1A binding to SUR1 protein, thus eliminating this possibility.

For the third possibility, the binding and activity of a wide range of K_ATP channel openers are dependent on the presence of Mg²⁺ and hydrolyzable ATP (45–47), and a postulated mechanism of action of these compounds is the acceleration of nucleotide hydrolysis by nucleotide binding folds (48). Syntaxin-1A may affect K_ATP channel openers by interfering with this mechanism, either via slowing the rate of nucleotide hydrolysis or by affecting the binding of Mg-ATP or Mg-ADP to their respective nucleotide binding folds. In support of this, NNC55-9216 required both nucleotide binding folds (24), whereas diazoxide required nucleotide binding fold-1 to exert their K_ATP channel opening action (44). Our in vitro binding study revealed reduced syntaxin-1A binding to SUR1 when the Mg-ATP concentration was increased to 5 mmol/l, which suggests competitive binding displacement at the nucleotide binding folds. Should syntaxin-1A influence the efficacy of these K_ATP channel openers via this postulated mechanism, further modulating effects may be observed with combinations of adenosine nucleotides and Mg²⁺. These complex interactions are not within the scope of the current study and will require much further investigation.

Our previous report on syntaxin-1A–overexpressing mice showed that a moderate increase of ∼30% of islet syntaxin-1A levels was sufficient to cause hyperglycemia from reduction in β-cell insulin exocytosis and inhibition of L-type Ca²⁺ current density (49). However, there was no effect on K_ATP or Kv2.1 channels, suggesting that these distinct components of the insulin secretory process exhibit differing sensitivity to syntaxin-1A modulation. This also suggests that moderate increases in syntaxin-1A levels beyond normal islet levels (which do not occur physiologically) are not likely to influence K_ATP channel opener actions.

The converse, however, may not be true—that is, when islet syntaxin-1A levels become reduced. Alemzadeh et al. (50) reported that 6 weeks of treatment with NN414, a SUR1-specific K_ATP channel opener of the same family as NNC55-0462, led to improvements in fasting and post–intraperitoneal glucose tolerance test glucose levels in Zucker obese diabetic rats but not in lean control rats, which was postulated to be attributable to lower insulin levels in lean animals during basal conditions and after NN414 treatment. We propose an alternative explanation. Islet syntaxin-1A levels are severely reduced in type 2 diabetic rodent models and human patients (percent of normal controls: ∼35% for obese Zucker rats, ∼40% for GK rats, and ∼20% for humans) (5–8). This may at least partly explain the increased effectiveness of K_ATP channel opener therapy in the Zucker rats (50) by enhancing K_ATP channel opener potency and efficacy on β-cells, whereas normal islet levels of syntaxin-1A would dampen the K_ATP channel opener actions. Because brain levels of syntaxin-1A are high and do not fluctuate with diabetes (6), the K_ATP channel opener would not affect neuronal SUR1 K_ATP channels, thereby increasing the drug's specificity. Normoglycemic control partially restored islet syntaxin-1A levels in GK rats (6), which could dampen the K_ATP channel opener effects that may be misconstrued as loss of drug effectiveness. These dynamics of islet syntaxin-1A expression occurring during the course of therapy may profoundly modulate K_ATP channel opener drug potency and efficacy on β-cells, rendering the drug more effective at stages of severe diabetes while dampening its actions as normoglycemic control is gradually attained. These exciting possibilities will require further experimental verification in diabetic rodent models.

H.Y.G. has received support from the Canadian Institutes for Health Research (CIHR MOP-69083) and the Heart and Stroke Foundation of Ontario (T-6064). B.N is supported by studentships from the Ontario Graduate Studentship and the Banting and Best Diabetes Center (University of Toronto). Y.H. is supported by a postdoctoral award from the Janssen-Ortho/Canadian Association of Gastroenterology.

We thank X. Gao for technical assistance in preparing the rat pancreatic islets and R. Tsushima and Y.-M. Leung for active discussion.