Sustained Exposure of L6 Myotubes to High Glucose and Insulin Decreases Insulin-Stimulated GLUT4 Translocation but Upregulates GLUT4 Activity

Anti-GLUT1 and anti-GLUT4 antisera were raised in rabbits to peptides encompassing 12 COOH-terminal residues of each protein (19); anti-GLUT3 was generated using the 19 COOH-terminal peptides (20). Anti-p38 MAPK, anti-insulin receptor β (C-19), anti-myc (9E10) antibodies, and anti-p85 subunit antibodies were purchased from Santa Cruz Biotechnology (Santa Cruz, CA). Phosphospecific antibodies to p38 MAPK, Jun NH₂-terminal kinase (JNK), and activating transcription factor (ATF)-2 fusion protein were from New England Biolabs (Mississauga, ON, Canada). Anti-phosphotyrosine and anti- IRS-1 antibodies were from Upstate Biotechnology (Lake Placid, NY). o-Phenylenediamine dihydrochloride (OPD reagent) was from Sigma. SB202190 and blasticidin-HCl were from Calbiochem (La Jolla, CA). Indinavir was a gift from Dr. Ralph Germinario (Sir Mortimer B. Davis Jewish General Hospital, Montreal, Canada) and Dr. Nava Bashan (Ben-Gurian University, Béer-Sheva, Israel).

L6 muscle cells stably expressing myc-tagged GLUT4myc (L6-GLUT4myc cells) were described previously (21,22). L6 stably expressing GLUT1myc were prepared by cotransfection of L6 myoblasts with pCX2-GLUT1myc and pSV2-bsr, a blasticidin S deaminase expression plasmid, and selected with blasticidin-HCl. Glass cloning cylinders (Bellco Glass, Vineland, NJ) were used to select individual colonies for expansion. Each clonal line was maintained in 2 μg/ml blasticidin-HCl and tested for GLUT1myc expression and the ability to differentiate into myotubes. GLUT1 or GLUT4-myc myoblasts were differentiated into myotubes, pretreated with 25 mmol/l glucose and 100 nmol/l insulin (high Glc/Ins) for 24 h, and deprived of serum for 5 h in 5 mmol/l glucose before incubation with inhibitors, followed by acute insulin challenge. Inhibitors were administered in DMSO, with maximum concentration of 0.1% vol/vol in the incubation medium and without effect on the parameters measured.

2-Deoxyglucose uptake measurements were carried out as described previously (18) for 5 min in HEPES-buffered saline containing 10 μmol/l 2-[³H]deoxyglucose (0.5 μCi/ml) in the absence of insulin and inhibitors, unless otherwise indicated. Nonspecific uptake was determined in the presence of 10 μ mol/l cytochalasin B and subtracted from all experimental values.

The amount of myc-tagged GLUT4 at the surface of intact cells was measured by an antibody-coupled colorimetric assay as validated previously (23). Briefly, monolayer of myotubes were exposed to anti-myc antibody (1:100) for 60 min, fixed with 4% paraformaldehyde for 10 min, and then incubated with peroxidase-conjugated donkey anti-mouse IgG (1:1,000) for 30 min, all at 4°C. Cells were washed extensively, and 1 ml of OPD reagent was added for 30 min at room temperature. The reaction was stopped by 0.25 ml of 3N HCl. The supernatant was collected, and the absorbance was measured at 492 nm. Nonspecific IgG binding, as measured by a peroxidase-conjugated anti-mouse IgG, was subtracted from experimental values.

Total p38 MAPK protein and p38 MAPK phosphorylation and activity were measured as previously described (18) with the following modifications. For p38 MAPK activity, monoclonal mouse anti-phospho-p38 MAPK antibody covalently linked to Sepharose beads (New England Biolab) (20 μl/condition) was used to immunoprecipitate activated p38 MAPK from 200 μg total cell lysate. The complex was washed four times with 1 ml wash buffer and twice with 1 ml kinase buffer and then incubated under constant agitation for 30 min at 30°C with 30 μl reaction mixture (kinase buffer containing 200 μmol/l ATP and 2 μg recombinant ATF-2). Reaction was stopped by 50 μl of 2× Laemmli sample buffer, and supernatant was resolved by 10% SDS-PAGE. Phospho-ATF-2 was determined by immunoblotting with anti-phospho-ATF-2 antibody at 1:500 dilution.

L6-GLUT4myc myoblasts were differentiated into myotubes in 10-cm plates, and total cellular membranes were prepared as described previously (24). Polyclonal primary antibody was used at 1:2,000 dilution.

Acute insulin stimulation (100 nmol/l for 5 min at 37°C) was followed by cell lysis, and whole-cell extracts were prepared from myotubes as described (18,25). Insulin receptor (IR) was immunoprecipitated from 500 μg protein, and the sample was boiled for 3 min before resolving the protein by 7.5% SDS-PAGE. IRS-1 was immunoprecipitated from 300 μg protein, and the immunoprecipitates were resolved by 7.5% SDS-PAGE. Tyrosine phosphorylation of IR and IRS-1 and the association of p85 with IRS-1 were determined by immunoblotting with anti-phosphotyrosine or anti-p85 antibodies at 1:2,000, 1:1,000, and 1:1,000 dilutions, respectively. Total IRS-1 content was determined in whole cell lysates, prepared by the same method as described above for detection of total cellular p38 MAPK.

PI 3-kinase activity associated with phosphotyrosine immunoprecipitates was measured in vitro toward phosphatidylinositol as described previously (24)

Statistical analysis was performed using either unpaired Student’s t test or ANOVA (Fisher, multiple comparisons), as applicable.

We first examined whether muscle cells were susceptible to the effect of preincubation with high glucose and insulin. L6-GLUT4myc myotubes were incubated for 24 h with 25 mmol/l glucose and 100 nmol/l insulin and then for 5 h in serum-free medium (5 mmol/l glucose). Subsequently, the rates of hexose uptake in the basal state and after an acute (30-min) insulin challenge were determined (Fig. 1A). High Glc/Ins pretreatment increased basal 2-deoxyglucose uptake by 40% (control, 25.2 ± 1.7; high Glc/Ins, 34.9 ± 1.3 pmol · min⁻¹ · mg⁻¹ protein; P < 0.005). In control cells, acute insulin stimulation elevated 2-deoxyglucose uptake by 2.3-fold (basal, 25.2 ± 1.7; insulin, 58.7 ± 4.9 pmol · min⁻¹ · mg⁻¹ protein; Δ, 33.5 pmol · min⁻¹ · mg⁻¹ protein; P = 0.0002). In contrast, myotubes that were pretreated with high Glc/Ins for 24 h had a significantly smaller response to the acute insulin challenge, showing only a 1.6-fold increase in 2-deoxyglucose uptake (basal, 34.9 ± 1.3; insulin, 56.9 ± 3.1 pmol · min⁻¹ · mg⁻¹ protein; Δ, 22.0 pmol · min⁻¹ · mg⁻¹ protein; P = 0.0002).

To elucidate the mechanism by which high Glc/Ins pretreatment altered glucose uptake, we measured surface GLUT4 levels under the various treatments. L6-GLUT4myc cells stably express GLUT4 tagged with an exofacial myc epitope (GLUT4myc). GLUT4myc segregates, cycles, and responds to insulin in a manner similar to endogenous GLUT4 (21–23). The amount of GLUT4myc incorporated into the plasma membrane was quantitated by immunologically labeling the myc epitope at the surface of intact cells. Acute insulin stimulation of untreated cells increased the amount of GLUT4myc at the cell surface by 2.3 ± 0.2-fold above basal (P = 0.01) (Fig. 1B). Preincubation of cells for 24 h with high Glc/Ins caused a 45% reduction in acute insulin-induced recruitment of GLUT4myc to the plasma membrane (basal, 0.9 ± 0.1; insulin, 1.7 ± 0.17; P < 0.005). However, high Glc/Ins treatment did not alter surface GLUT4myc under basal conditions (Fig. 1B), despite increasing 2-deoxyglucose uptake by 40% (Fig. 1A).

In Fig. 1, we show that high Glc/Ins pretreatment caused a significant elevation of basal glucose uptake without increasing GLUT4myc translocation. These results suggest that GLUT4 may be hyperactive under these conditions or that upregulation of another transporter may account for the increased glucose uptake. To assess these possibilities, GLUT1 and GLUT4 levels were determined in total membranes isolated from L6-GLUT4myc myotubes pretreated with high Glc/Ins for 24 h. Total GLUT1 content was increased by 2.5 ± 0.5-fold (P < 0.05) (Fig. 2A). In contrast, high Glc/Ins pretreatment did not alter total GLUT4 content (Fig. 2A).

Because of the rise in GLUT1 protein, it was important to compare the level of GLUT4myc to that of the endogenous GLUT1. In parental L6 myotubes, the molar ratio of surface GLUT1 to GLUT4 is 1.0:0.8, as determined by quantitative photolabeling using azi-trifluroethylbenzoyl-bis-mannose-yloxypropylamine (ATB-BMPA) (28). In L6-GLUT4myc myotubes, the myc-tagged transporter is significantly overexpressed relative to the endogenous GLUT4. Figure 2B shows the molar ratio of GLUT4myc to that of GLUT4 in other cells and tissues expressing this transporter. GLUT1 expression is the same in GLUT4myc cells as in parental L6 cells. Thus, in L6-GLUT4myc cells, the ratio of GLUT4myc to GLUT4 to GLUT1 is 124:1:1.2. Hence, GLUT4myc is vastly the predominant glucose transporter in L6-GLUT4myc myotubes, and it is likely to be the transporter responsible for the increase in glucose uptake caused by high Glc/Ins (Fig. 1A). To further support this hypothesis, we used indinavir, a protease inhibitor that was shown to inhibit GLUT4 but not GLUT1 transporters directly (29,30). Figure 3A shows that when indinavir was present during the 2-deoxyglucose transport assay, there was a dose-dependent inhibition of 2-deoxyglucose uptake in the GLUT4myc but not in the GLUT1myc myotubes; a similar dose-response curve was observed after the cells were pretreated with high Glc/Ins. Taken together, these results suggest that GLUT4myc is the predominant glucose transporter in these cells responsible for glucose influx. Moreover, the high Glc/Ins effect is via GLUT4.

It was then of interest to explore whether GLUT1 activity is similarly regulated in cells where GLUT1 is the predominant route for glucose uptake. To this end, we used L6 muscle cells overexpressing GLUT1myc, where GLUT1myc expression is controlled by a constitutively active cytomegalovirus promoter, and therefore its levels are unlikely to change biosynthetically with high Glc/Ins. High Glc/Ins preincubation did not elevate the basal level of glucose uptake in L6-GLUT1myc myotubes (control, 20.6; high Glc/Ins, 21.3 pmol · min⁻¹ · mg⁻¹ protein) (Fig. 3B), suggesting that GLUT1 activity cannot be upregulated by high Glc/Ins. Collectively, these results highlight that functionally, GLUT4myc is responsible for both basal and acute insulin-stimulated glucose uptake in L6-GLUT4myc myotubes, and the basal state activity of GLUT4, but not of GLUT1, increases upon preincubation of myotubes with high Glc/Ins.

We have previously reported that an acute insulin challenge increases the activity of translocated GLUT4, an effect that is reduced by pretreating cells with p38 MAPK inhibitors (17). Therefore, we hypothesized that the activity of p38 MAPK contributes to the elevated basal glucose uptake observed following high Glc/Ins pretreatment. L6-GLUT4myc myotubes that were pretreated for 24 h with high Glc/Ins were incubated with an inhibitor of p38 MAPK for 20 min before assaying glucose uptake. Figure 4 shows that SB202190 restored the basal rate of glucose uptake in cells pretreated with high Glc/Ins, suggesting that p38 MAPK may regulate the basal state activity of GLUT4 in L6-GLUT4myc myotubes preincubated with high Glc/Ins.

The abrogation by SB202190 of the high Glc/Ins-induced elevation in basal glucose uptake in GLUT4myc myotubes suggested the participation of p38 MAPK. Hence, we determined the effect of high Glc/Ins on p38 MAPK phosphorylation and kinase activity in these cells. Activation of p38 MAPK by cytokines (31) and insulin (17) correlates with the dual phosphorylation of the enzyme on threonine 180 and tyrosine 182. Using antibodies that recognize p38 MAPK only when phosphorylated on both of these residues, pretreatment of myotubes with high Glc/Ins elevated basal p38 MAPK phosphorylation by 80% (P < 0.05) (Fig. 5A). The cellular content of p38 MAPK was also elevated by 25% by the high Glc/Ins pretreatment (P < 0.05, Fig. 5B).

We then determined the effect of high Glc/Ins treatment on p38 MAPK protein kinase activity, using an in vitro kinase assay. Consistent with the results of p38 MAPK phosphorylation, preincubation of cells with high Glc/ Ins for 24 h increased basal p38 MAPK activity by 2.3 ± 0.3-fold (P < 0.05) (Fig. 5C). We also measured phosphorylation of JNK, another stress-activated MAPK, to determine if the effects of high Glc/Ins were specific for p38 MAPK. Treatment with high glucose and insulin for 24 h did not increase basal JNK phosphorylation (data not shown).

We demonstrate in Fig. 1 that pretreatment of myotubes with high Glc/Ins reduced subsequent insulin-stimulated GLUT4 translocation by 45%. This result could arise from alterations in the signals thought to regulate GLUT4 translocation or from defects in the fusion machinery at the plasma membrane. Hence, we analyzed the effect of high Glc/Ins on tyrosine phosphorylation of the IR, IRS-1, activation of PI 3-kinase activity, and phosphorylation of Akt/protein kinase B. High Glc/Ins pretreatment of L6-GLUT4myc myotubes for 24 h led to a 38% and a 35% reduction in insulin receptor protein content (as determined by immunoprecipitation from cell lysate) and insulin-stimulated (5 min) tyrosine phosphorylation of insulin receptor (P = 0.002 and 0.003), respectively (Fig. 6A). Insulin stimulation (5 min) increased IRS-1 phosphorylation by fourfold in control cells (P < 0.005) (Fig. 6B), which was diminished by 65% after 24 h of high Glc/Ins pretreatment (P < 0.005). Concomitantly, there was a 50% (P < 0.005) reduction in total cellular IRS-1 protein (Fig. 6B). High Glc/Ins for 24 h also reduced IRS-1- associated p85 by 80% (Fig. 6C). Similarly, PI 3-kinase activity associated with anti-phosphotyrosine immunoprecipitates was also reduced by 40% in pretreated cells (Fig. 6C). As expected, acute insulin caused a robust phosphorylation of Akt at Thr 308 and Ser 473 in control cells, which was reduced by 60% and 50%, respectively, after 24-h preincubation with high Glc/Ins (Fig. 6D). Therefore, starting at the level of IR, we observed a downregulation to acute insulin effect on IR, IRS-1 phosphorylation, phosphotyrosine associated PI 3-kinase activity, and Akt phosphorylation.

Insulin resistance is a key feature of type 2 diabetes. In particular, it has been argued that elevated levels of glucose and insulin are a major cause for the development of secondary insulin resistance, but the molecular mechanisms remain obscure (32). Therefore, establishment of in vitro models of high glucose and insulin state has been pursued to allow understanding of the molecular basis of acquired insulin resistance. Previous studies in 3T3-L1 adipocytes and primary cultured adipocytes have shown that prolonged exposure to high concentrations of insulin and glucose resulted in increased basal glucose uptake and a decrease in acute insulin-mediated glucose transport, the latter attributed to reduced insulin-stimulated GLUT4 translocation (3–6,33). However, it remained possible that the insulin-resistant state may also be associated with diminished GLUT4 activity.

To our knowledge, there are no comprehensive accounts of the basis of acquired insulin resistance in muscle cells using a high glucose and insulin model, yet muscle tissue is a primary determinant of glycemic control in vivo. The aim of this study was to examine the alterations in signaling pathways regulating GLUT4 activity and GLUT4 translocation when a cell culture of muscle origin was exposed to sustained high levels of glucose and insulin. A muscle cell line where GLUT4 is the predominant transporter, L6 myotubes overexpressing myc-tagged GLUT4, offers the opportunity to compare GLUT4 translocation with glucose uptake in intact cell preparations, allowing us to assess possible changes in GLUT4 activity.

We demonstrated in this study that 24 h of high glucose and insulin reduced insulin-stimulated GLUT4 translocation, with no significant change in cell surface GLUT4 at the basal state. This is consistent with previous reports of 50–100% reduction in GLUT4 translocation in 3T3-L1 adipocytes under similar conditions (4–6,33). To understand the mechanism of the translocation defect, we examined IR and IRS-1 phosphorylation, PI 3-kinase activity, and Akt phosphorylation. We found a 35% reduction in immunoprecipitable insulin receptor after 24 h of high insulin and glucose pretreatment in L6-GLUT4myc myotubes. The insulin-stimulated tyrosine phosphorylation of the receptor was similarly decreased by 35%. Similar observations have been made in insulin-resistant and diabetic humans (34,35). In agreement with Ricort et al. (5), we found that high Glc/Ins treatment also resulted in a significant reduction in insulin-stimulated IRS-1 tyrosine phosphorylation and total IRS-1 content. Reduced IRS-1 tyrosine phosphorylation has been reported in skeletal muscle of insulin-resistant Zucker rats (36) and in humans with type 2 diabetes (37,38) and obesity (39). Increased degradation of IRS-1 under chronic insulin treatment is one mechanism underlying insulin resistance (40). In our cellular model, a slight discrepancy between the percentage reduction of IRS-1 tyrosine phosphorylation (65% reduction) and reduction of IRS-1 content (50%) was observed. There are two possible explanations: first, the reduction in IRS-1 content was compounded with a reduction in IR phosphorylation and therefore activity, leading to a more significant reduction in IRS-1 phosphorylation than expected from the decrease in IRS-1 content. Second, there may be a separate defect in the insulin signaling pathway such that it impaired the tyrosine phosphorylation of the remaining IRS-1.

With the reduction in IRS-1 content and its phosphorylation after chronic exposure to high Glc/Ins, we predicted and observed a reduction in the acute insulin-dependent activation of PI 3-kinase. We also observed a 60% reduction in insulin-mediated Akt phosphorylation at threonine 308 and serine 473. Therefore, the defect in translocation may be explained in part by the alterations in signal transduction from IR to IRS-1 and then to PI 3-kinase and Akt, resulting in reduced GLUT4 translocation. In contrast, the reduction in GLUT4 translocation caused by high insulin and glucose using 3T3-L1 adipocytes was associated with reduced levels of total cellular GLUT4 (6,33).

In L6-GLUT4myc myotubes pretreated with high Glc/Ins, there was an increase in total cellular GLUT1 content, which could potentially account for the increase in basal glucose uptake. Indinavir, an HIV protease inhibitor, has been associated with insulin resistance in humans and animals (29,30). Using Xenopus oocytes transfected with GLUT4 or GLUT1, Murata et al. (29) showed that indinavir inhibits GLUT4- but not GLUT1-mediated glucose transport, suggesting this as the basis of insulin resistance in vivo. Accordingly, we found profound inhibition of both basal and insulin-stimulated 2-deoxyglucose uptake in GLUT4myc but not in GLUT1myc cells. Indinavir effectively inhibited glucose uptake in L6-GLUT4myc cells pretreated with high Glc/Ins, indicating that this pretreatment did not alter the affinity of glucose transporters to indinavir. The elevation in basal glucose uptake caused by pretreatment with high glucose and insulin was entirely abolished by indinavir. These results pointed to GLUT4myc as the major glucose transporter in L6-GLUT4myc cells, and suggest that the increase in GLUT1 protein had no significant contribution to the observed increase in basal glucose uptake caused by high Glc/Ins pretreatment. Because GLUT4myc also overrides the endogenous GLUT4, the results in this study refer to the activity and translocation of GLUT4myc, in both basal and high glucose/insulin conditions.

An insulin-stimulated increase in glucose uptake is the final outcome of any changes that affect the insulin signaling pathway. We found that in L6-GLUT4myc myotubes, 24-h exposure to high Glc/Ins resulted in a 40% increase in basal glucose uptake accompanied by a diminished net response to an acute insulin stimulation. Because the cell surface GLUT4 was not elevated at the basal state in the pretreated cells, the possibility that the intrinsic activity of GLUT4 was upregulated was explored.

An emerging body of literature suggests that GLUT4 translocation is probably not sufficient to fully account for the effect of insulin on glucose uptake (41). Differences are noted in the time course (17,42) and sensitivity to wortmannin (13,43) of insulin-dependent GLUT4 translocation and stimulation of glucose uptake in both 3T3-L1 adipocytes and GLUT4myc L6 myotubes. In each instance, substantial GLUT4 translocation occurred without commensurate increases in glucose uptake. Other studies have also alluded to a dissociation between GLUT4 translocation and the full insulin response of glucose uptake in vivo (14, 15). Recently, we and others observed that SB203580, a selective inhibitor of p38MAPK (16), diminished insulin-stimulated glucose uptake by 30–60% in L6 myotubes (17) and 3T3-L1 adipocytes (18) without reducing GLUT4 translocation. These studies have led to the suggestion that insulin may increase the intrinsic activity of GLUT4 via a p38 MAPK-dependent mechanism. Indeed, insulin activates p38 MAPK in L6-GLUT4myc myotubes (17), 3T3-L1 adipocytes (18,44,45), rat skeletal muscle (46,47), and vascular smooth muscle (48). The precise role of p38 MAPK in insulin-stimulated glucose uptake needs further evaluation. In 3T3-L1 adipocytes, Kayali et al. (44) demonstrated that p38 MAPK was activated by insulin (2.7-fold), and insulin-stimulated glucose uptake was inhibited in a dose-dependent manner by SB203580 and SB202190. However, the half-maximal inhibitory concentration (IC₅₀) for inhibition of glucose transport was determined to be >10 μmol/l, greater than the reported half-maximal stimulatory concentration (EC₅₀) for inhibition of p38 MAPK (∼0.5 μmol/l). Therefore, the authors concluded that p38 MAPK is not involved in insulin-stimulated glucose uptake in 3T3-L1 adipocytes. However, because SB203580 and SB202190 reduce insulin stimulation of glucose uptake by only 50% even at maximal concentrations, the IC₅₀ for inhibition of glucose uptake is closer to 1–2 μmol/l (17). This value is very close to the IC₅₀ for inhibition of phosphorylation of targets of p38 MAPK in intact cells (17). Therefore, the reduction in glucose uptake by SB202190 and SB203580 closely parallels that for p38 MAPK activity.

In the present study, we observed that with high Glc/Ins, basal-state glucose uptake was elevated without a concomitant increase in surface levels of GLUT4 and without contribution of GLUT1. We interpret these results to suggest that high Glc/Ins elevated the intrinsic activity of GLUT4. Therefore, we explored whether this elevation might be caused by a regulatory input via p38 MAPK. To this effect, we examined the status of p38 MAPK expression and activity in response to high Glc/Ins. This pretreatment caused a 25% gain in total cellular p38 MAPK, an 80% increase in p38 MAPK phosphorylation, and a 2.3-fold increase in the kinase activity measured in vitro. Moreover, when high Glc/Ins-pretreated L6 myotubes were treated briefly with SB202190, the elevation in basal glucose uptake was reversed to control levels. These results support the hypothesis that in response to the chronic high Glc/Ins environment, p38 MAPK is activated and may contribute to the increase in basal glucose uptake.

The magnitude of increase in p38 MAPK phosphorylation (85%) was greater than the increase in p38 MAPK content (25%), suggesting that there was also upregulation of the upstream kinases or downregulation of phosphatases, which would normally dephosphorylate p38 MAPK. Alterations in the p38 MAPK signal pathway is unlikely to be a generalized stress response, however, since there was no increase in phosphorylation of other known stress response kinases, such as JNK or the extracellular signal-regulated kinase (ERK) MAPK (results not shown). In contrast to basal p38 MAPK, high Glc/Ins treatment reduced the ability of insulin to acutely increase p38 MAPK phosphorylation, uncovering a correlation between reduced insulin-stimulated p38 MAPK activity and reduced insulin stimulation of glucose uptake in insulin-resistant states in muscle cells.

In conclusion, the results summarized in Table 1 suggest the following model of imposed insulin resistance in L6-GLUT4myc myotubes: high Glc/Ins for 24 h caused a defect in acute insulin-mediated IR and IRS-1 phosphorylation, PI 3-kinase activity, and Akt phosphorylation, leading ultimately to a significant reduction in insulin-mediated GLUT4 translocation. We hypothesize that basal glucose uptake is elevated as an adaptive response, probably as a result of elevated p38 MAPK phosphorylation and activity. However, an additional defect was introduced into the p38 MAPK signal pathway by the high Glc/Ins pretreatment, resulting in a decrease in acute insulin-mediated gain in p38 MAPK phosphorylation and activity. These results support the tenet that p38 MAPK may be a regulator of glucose uptake and is altered in states of insulin resistance.

This work was supported by a research grant from the Canadian Diabetes Association to A.K. C.H. was supported by a Canadian Pediatric Endocrinology Fellowship. R.S. was supported by a Doctoral Award from the Canadian Institute for Health Research. N.P. was supported by a studentship from the Banting and Best Diabetes Center at the University of Toronto. W.N. was supported by a fellowship from the Visiting Scientist Program at the Hospital for Sick Children.

We thank Dr. Phil Bilan for helpful discussions and careful review of the manuscript and Dr. Assaf Rudich for thoughtful suggestions.