Evidence for a Role of Superoxide Generation in Glucose-Induced β-Cell Dysfunction In Vivo

Female Wistar rats (Charles River), weighing 250–300 g, were cannulated as previously described (14). The jugular catheter served for infusion and the carotid catheter for blood sampling. Rats were allowed at least 2–3 days of postsurgery recovery before infusions. All procedures were approved by the University of Toronto Animal Care Committee.

Rats were randomized and infused for 48 h with one of the following: 1) saline, 2) a variable infusion of 37.5% glucose to achieve and maintain plasma glucose at 20–22 mmol/l (HG), 3) HG + low-dose TAU (l-TAU), 4) HG + medium-dose TAU (m-TAU), 5) HG + high-dose TAU (h-TAU), 6) HG + NAC, 7) HG + TPO, or 8) TAU, NAC, or TPO without HG infusion. TAU was given at three doses: 2.14 μmol · kg⁻¹ · min⁻¹, which reversed insulin resistance induced by high glucose (15); 2.76 μmol · kg⁻¹ · min⁻¹, which prevented β-cell dysfunction induced by 48-h fat infusion (16); and 5.52 μmol · kg⁻¹ · min⁻¹. NAC was given at 2.76 μmol · kg⁻¹ · min⁻¹ (equimolar to m-TAU), which also prevented fat-induced β-cell dysfunction (16). TPO was given at 2.41 μmol · kg⁻¹ · min⁻¹, which protected against ROS toxicity in experimental pancreatitis (17). All antioxidants (Sigma, St. Louis, MO) were dissolved in saline at pH = 7.4.

Pancreatic islets were isolated in overnight-fasted rats, after the 48-h infusions, using the Ficoll/Histopaque method (18).

The freshly isolated islets were preincubated for 1 h at 37°C in Krebs Ringer buffer containing 10 mmol/l HEPES (KRBH) and 2.8 mmol/l glucose and incubated in triplicate for 2 h at 37°C in KRBH at the following glucose concentrations: 2.8 mmol/l, to evaluate non-GSIS; 6.5 mmol/l, which is basal glucose in rats; 13 mmol/l, the upper physiological glucose level in rats; and 22 mmol/l, which is a maximum stimulatory concentration. Insulin was measured in the supernatant with a Linco radioimmunoassay kit. The islet pellets were subjected to acid-ethanol extraction for measurement of insulin content using the kit from Linco (19).

Islet ROS was measured with H₂DCF-DA (18), which detects most ROS including hydrogen peroxide, peroxyl radical, and peroxynitrite anion (20). H₂DCF-DA does not detect superoxide directly; however, superoxide can be detected indirectly via its ROS metabolites (21). Total superoxide was measured by hydroethidine (22), and mitochondrial superoxide was measured by MitoSOX (23).

Islet apoptosis was measured in dispersed islet cells using fluorescein isothiocyante–conjugated Annexin-5 (Sigma) and propidium iodide (Sigma). Annexin-5 detects membrane phosphotidylserine flip-flop once viability is compromised (24). Propidium iodide stains dead and necrotic cells. Cells stained with annexin-5 only were identified as apoptotic. After islet dissociation, performed as previously described (25), islet cells were loaded onto coverslips and incubated at 37°C in RPMI-1640 containing 2.8 mmol/l glucose for 2 h. Cells were then washed with KRBH buffer and incubated with annexin-5 (0.15 μg/ml) and propidium iodide (5 μmol/l) in KRBH for 20 min at 25°C. After washing, fluorescence was measured at 480/545 nm excitation and 510/610 nm emission for annexin and propidium iodide, respectively.

A set of rats was randomized into the following groups and infused for 96 h with one of the following: 1) saline, 2) a variable infusion of 37.5% glucose to achieve and maintain plasma glucose at 15 mmol/l (HG), 3) HG + m-TAU, or 4) HG + TPO. The m-TAU and TPO doses used were the same as those in the 48-h infusion study. Following the 96-h infusion, islets were isolated and ex vivo evaluation of GSIS performed as described above.

Rats were randomized and infused for 48 h with one of the following: 1) saline, 2) a variable infusion of 37.5% glucose to achieve and maintain plasma glucose at 20–22 mmol/l (HG), 3) HG + TPO, 4) TPO alone, or 5) mannitol to control for plasma osmolarity as in ref. 26, infused at a rate comparable with that of the glucose infusion (∼45 μl/min) in HG rats. After 48 h of infusion and overnight fasting, all groups were subjected to a two-step hyperglycemic clamp to evaluate GSIS in vivo.

At the end of 48 h, glucose infusion in the HG alone and HG + TPO groups was decreased to ∼5 μl/min for ∼75 min, which was required to achieve basal glucose without hypoglycemia. Basal insulin and C-peptide were measured at −20 and 0 min. At time = 0, an infusion of 37.5% glucose was started. Plasma glucose was maintained at 13 mmol/l by adjusting the rate of glucose infusion according to frequent (5–10 min) glycemic determinations. At 120 min, the glucose infusion was again raised to 22 mmol/l until the end of the experiment (time = 240 min). Samples for insulin and C-peptide were taken at regular intervals.

Plasma glucose was measured using a Beckman Analyzer II (Fullerton, CA). Radioimmunoassay kits (Linco, St. Charles, MO) specific for rat insulin and C-peptide were used.

The M/I index of insulin sensitivity was calculated by dividing the glucose infusion rate (GIR) by the plasma insulin level.

Regarding β-cell function, C-peptide response was taken as an index of absolute insulin secretion, as insulin secretion rate cannot be calculated in rats. This is due to the fact that rat C-peptide kinetics have not been determined because species-specific rat C-peptide is unavailable for injection. Insulin secretion in vivo has to be evaluated in the context of insulin sensitivity, since the normal β-cell compensates for insulin resistance. In normal subjects, the relationship between insulin sensitivity and insulin secretion is hyperbolic (27,28), i.e., the product of insulin sensitivity and insulin secretion is a constant defined as disposition index (DI) and considered a measure of β-cell function. We have shown that the relationship between C-peptide and M/I index is indeed hyperbolic in control rats (29). Therefore, we calculated DI during the last 40 min of each step of the hyperglycemic clamp.

At the end of the hyperglycemic clamp, rats were anesthetized with a ketamine:xylazine:acepromazine cocktail (87:1.7:0.4 mg/ml) and pancreases were removed, fixed overnight in 10% formalin, and stored in 70% ethanol. Samples were processed and embedded in paraffin within 5 days of collection.

Three sections (separated by 100 μm) were immunostained for insulin using an antibody from Biomeda (Foster City, CA) coupled with peroxidase detection. Relative β-cell area was determined from the ratio between areas occupied by insulin-positive cells and area occupied by total pancreatic tissue. β-Cell mass was determined by multiplying this ratio by pancreas weight. Individual β-cell area (index of size) was calculated by dividing insulin-positive areas by the number of nuclei within each area.

Sections (three per rat) adjacent to those used for insulin immunostaining were used to detect islet apoptotic nuclei with TUNEL (transferase-mediated dUTP nick-end labeling) staining (30).

Data are presented as means ± SE. One-way nonparametric ANOVA for repeated measurements followed by Tukey's t test was used to compare differences between treatments. Calculations were performed using SAS (Cary, NC).

Baseline fed plasma glucose and insulin were not different between groups before the onset of the 48-h infusions. In rats infused for 48 h with HG alone or in combination with antioxidants, glucose levels were elevated to ∼20–22 mmol/l by ∼6 h and were maintained at that level for the remainder of the 48-h period (not shown).

Figure 1A shows that at 48 h, co-infusion of TAU (all doses) or NAC tended to increase and TPO to significantly increase GIR compared with HG alone (P < 0.001). HG + TPO also increased GIR versus HG + l-TAU and HG + h-TAU (P < 0.05). Plasma insulin tended to decrease compared with HG with all antioxidants (Fig. 1B). The HG + TAU (all doses) and HG + NAC groups had significantly lower C-peptide levels than the group treated with HG alone at 48 h (Fig. 1C). Although HG + TPO tended to increase the ratio of C-peptide to insulin (index of insulin clearance) compared with HG alone, this was not significantly different (not shown). Co-infusion of all antioxidants with HG improved insulin sensitivity (Fig. 1D). HG + TPO increased the DI compared with HG alone. However, TAU (all doses) and NAC did not have any effect on DI (Fig. 1E). Non–glucose-infused groups are not shown in the graph because they had GIR = 0, and thus insulin sensitivity index (SI) and DI could not be calculated.

Insulin secretion at 2.8 mmol/l glucose did not differ between groups. Forty-eight–hour glucose infusion increased basal insulin secretion at 6.5 mmol/l (P < 0.001 HG vs. saline) but impaired GSIS at 13 and 22 mmol/l glucose (P < 0.001). All three doses of TAU did not prevent the increase in basal insulin secretion induced by glucose or the impairment in GSIS at 13 and 22 mmol/l glucose. In fact, h-TAU tended to accentuate the impairment in GSIS at 22 mmol/l glucose (Fig. 2A). Co-infusion of NAC, similar to TAU, did not prevent the effect of high glucose on either basal secretion or GSIS (Fig. 2B). Co-infusion of TPO did not affect insulin secretion at 6.5 mmol/l glucose but completely restored GSIS at both 13 and 22 mmol/l glucose (Fig. 2C).

Forty-eight–hour hyperglycemia drastically reduced islet insulin content. The addition of TPO doubled the insulin content (P < 0.05 HG + TPO vs. HG alone), but this was still profoundly reduced compared with saline (Fig. 2D).

Forty-eight–hour hyperglycemia increased intracellular ROS in islets compared with saline (P < 0.01). Co-infusion of m-TAU or NAC abolished the increase in ROS induced by HG (Figs. 3A and B). Co-infusion of TPO did not have a significant effect (Fig. 3C).

Forty-eight–hour glucose infusion elevated total islet superoxide (P < 0.05 vs. saline). TAU did not normalize superoxide (Fig. 4A). NAC tended to decrease superoxide, but this was not significant (Fig. 4B). In contrast, TPO prevented the increase in superoxide induced by prolonged glucose infusion (Fig. 4C).

Forty-eight–hour glucose infusion elevated islet mitochondrial superoxide, and co-infusion of TPO completely prevented this increase (Fig. 5).

The percentage of apoptotic cells was measured in dispersed islet cells from isolated islets using the fluorescent dyes annexin-5 and propidium iodide. Islet cells stained for annexin-5 only were identified as apoptotic. There was no detectable difference in islet cells apoptosis between groups (percentage of apoptotic cells: saline, 1.86 ± 0.08%; HG alone, 1.99 ± 0.17%; HG + TPO, 1.88 ± 0.20%; and TPO, 1.91 ± 0.18%; n = 3–4 per group).

Islet morphological studies were performed in the isolated pancreas after 48-h infusion and hyperglycemic clamps. β-Cell mass and individual area (i.e., size) were increased ∼1.5- to 2-fold compared with saline. β-Cell mass and size were similar in HG alone and HG + TPO (Table 1). The number of β-cells per section tended to increase in HG alone and HG + TPO, but these differences were not significant (saline, 3,119 ± 313; HG, 3,431 ± 551; HG + TPO, 3,312 ± 279; and TPO, 3,083 ± 337). Apoptotic cells were very rare without detectable differences between groups (not shown).

To further investigate whether TPO is effective in preventing glucose-induced β-cell dysfunction in vivo, a two-step hyperglycemic clamp was performed after 48-h infusions. A subset of rats were also infused with mannitol as a control for osmolarity of the glucose infusate, and the results were not different from those of saline.

Basal plasma glucose before the clamp was slightly lower in HG alone and HG + TPO groups (not visible because of graph scale). Plasma glucose was elevated to 13 mmol/l until 120 min and then to 22 mmol/l until 240 min, with no difference between groups (Fig. 6A).

In rats given HG alone, a lower GIR was necessary to clamp glucose at 22 mmol/l, indicating that the circulating insulin was inadequate to compensate for insulin resistance. Co-infusion of TPO completely restored GIR to the saline levels (Fig. 6B).

Basal insulin and C-peptide levels (−20 to 0 min) were higher in rats given HG alone and HG + TPO than in those given saline (P < 0.001). In response to increasing glucose levels, plasma insulin and C-peptide increased as expected. Absolute insulin and C-peptide levels were higher in rats given HG alone than in those given saline (P < 0.001) at 13 mmol/l and tended to be higher at 22 mmol/l. However, the response of insulin and C-peptide to the rise in glucose was blunted in HG compared with saline, suggesting β-cell dysfunction. TPO tended to decrease absolute insulin and C-peptide levels and increased the response of insulin and C-peptide to the rise in glucose (Figs. 6C and D).

The ratio of C-peptide to insulin (index of insulin clearance) was lower in rats given HG than in those given saline at 13 mmol/l glucose (saline 4.43 ± 0.26 vs. HG 3.08 ± 0.25; P < 0.05 vs. saline). This was partially prevented by the co-infusion of TPO (HG + TPO 3.65 ± 0.26; N.S. vs. saline or HG). Similar intergroup differences in C-peptide–to–insulin ratio failed to reach significance at basal and 22 mmol/l glucose.

Insulin sensitivity index (M/I index) was decreased in HG compared with saline at both 13 and 22 mmol/l glucose. Decreased insulin sensitivity was partially prevented by the co-infusion of TPO at 22 mmol/l glucose (Fig. 6E).

Although insulin and C-peptide levels were elevated in HG, the DI, which represents the ability of the β-cell to compensate for insulin resistance, was lower in rats given HG than in those given saline. TPO was able to completely restore DI during both steps of the clamp (Fig. 6F).

To investigate the effect of antioxidants in a more chronic, less extreme model of glucose-induced β-cell dysfunction, GSIS in isolated islets was assessed following 96-h hyperglycemia (plasma glucose ∼15 mmol/l) with/without TPO or TAU. Ninty-six–hour hyperglycemia increased insulin secretion at 6.5 mmol/l glucose but impaired GSIS at 22 mmol/l glucose. Insulin secretion at 13 mmol/l glucose was unaffected. Similar to our findings following 48-h glucose infusion, both antioxidants did not prevent the increase in basal insulin secretion induced by hyperglycemia, and only TPO was able to prevent the impairment in GSIS at 22 mmol/l glucose (Fig. 7).

In this study, the role of oxidative stress in glucose-induced β-cell dysfunction was investigated using ex vivo and in vivo models. Chronic hyperglycemia has been well documented to induce β-cell dysfunction in vitro (31–34) and ex vivo (35–37). Consistent with previous results, we found that β-cell function was impaired ex vivo in isolated islets and also in vivo, as assessed by the hyperglycemic clamp following 48 h of glucose infusion.

During the hyperglycemic clamp, absolute insulin and C-peptide levels were elevated in rats that received 48-h glucose infusion. However, insulin resistance was also induced in these animals. Insulin secretion in vivo has to be evaluated in the context of insulin sensitivity, since the normal β-cell compensates for insulin resistance by increasing insulin secretion, independent of plasma glucose. In subjects with normal glucose tolerance, insulin secretion and sensitivity are linked through a hyperbolic relationship (27,28), i.e., the product of insulin sensitivity and insulin secretion is a constant. This constant is defined as the DI and is a measure of β-cell function (including β-cell ability to compensate for insulin resistance). DI was decreased in glucose-infused rats, demonstrating that β-cell dysfunction was induced by 48-h hyperglycemia. Aside from a decreased DI, there was also a marked defect in the response of insulin and C-peptide to a rise in plasma glucose, which further demonstrates β-cell dysfunction in glucose-infused rats.

Increased insulin secretion was observed at basal glucose (∼6.5 mmol/l) both in isolated islets and in vivo before the onset of the hyperglycemic clamp. This increase was not observed in islets at 2.8 mmol/l, which suggests that it is likely not due to islet hypertrophy or hyperplasia, but to a shift in the sensitivity of the β-cell to glucose. The mechanisms whereby chronic glucose exposure increases the insulin secretory response to low glucose are unclear but may be linked to the well-described induction of glycolytic genes by glucose (38). From our findings, we postulate that this increase in insulin secretion at low glucose levels is likely not due to oxidative stress, since all three antioxidants did not prevent this effect.

β-Cell mass was increased by short-term glucose infusion as previously reported (39–41), mainly because of β-cell hypertrophy (also as reported previously at 48 h [39,40]). There was no detectable change in islet viability or apoptosis. This establishes the 48-h glucose infusion model as a model of β-cell dysfunction rather than loss.

The mechanism whereby prolonged exposure to high glucose impairs β-cell function is not completely understood and is probably due to oxidative stress–dependent and –independent pathways. The studies that have linked oxidative stress to glucose-induced β-cell dysfunction have mainly been performed in vitro (2–6) or in animal models of type 2 diabetes (9,35), which in addition to hyperglycemia have many other metabolic alterations that can affect β-cell function. Thus, it is important to study the effect of chronic hyperglycemia per se on β-cell function in vivo.

Unexpectedly, the antioxidants TAU and NAC, which we have previously shown to prevent lipid-induced β-cell dysfunction in vivo and in isolated islets (16) (same doses as in this study), were not able to prevent glucose-induced β-cell dysfunction in isolated islets. (The effects of TAU and NAC were not assessed by the hyperglycemic clamp becaue TAU and NAC did not show any effect on the DI at 48 h). One possible explanation may be that a higher dose of TAU and NAC is required to prevent glucose-induced than lipid-induced β-cell dysfunction. However, this is unlikely, since 1) co-infusion of TAU or NAC completely normalized H₂DCF-DA–measured ROS and 2) doubling the dose of TAU (compared with the dose infused with lipids) was still ineffective. Another more plausible explanation may be that different types of ROS are important for β-cell dysfunction induced by glucose and lipids. TAU is effective in scavenging toxic aldehydes. NAC contains an −SH group and thus is capable of reducing oxidized proteins. NAC is also a precursor of intracellular glutathione. It is therefore possible that toxic aldehyde generation and glutathione depletion play a lesser role in glucose- than lipid-induced β-cell dysfunction.

In contrast to TAU and NAC, the antioxidant TPO was completely effective in preventing β-cell dysfunction induced by prolonged hyperglycemia in isolated islets and in vivo. Interestingly, TPO did not decrease total ROS (as detected by H₂DCF-DA) induced by prolonged glucose infusion. This was not surprising because TPO is a superoxide dismutase mimetic, i.e., it converts superoxide to hydrogen peroxide. Instead, TPO did reduce total superoxide in islets of glucose-treated rats, unlike TAU or NAC. Therefore, our findings show that the effect of antioxidants to prevent glucose-induced β-cell dysfunction parallels their ability to normalize superoxide but not H₂DCF-DA–detected ROS, suggesting that superoxide generation is important in our in vivo model of glucose-induced β-cell dysfunction. Our findings are in accordance with an in vitro study by Krauss et al. (42), which showed that overexpression of Mn superoxide dismutase, but not glutathione peroxidase 1 (which decreases H₂O₂), prevented glucose-induced β-cell dysfunction.

In the β-cell, superoxide can be generated in various ways. One possible source of superoxide is the mitochondrial electron transport chain. During hyperglycemia, superoxide production is increased due to greater electron flow in the electron transport chain. Studies have shown that mitochondrial production of ROS can impair insulin secretion (42,43). Our findings show that glucose-induced β-cell dysfunction in vivo is indeed associated with increased mitochondrial superoxide levels, and this increase was completely prevented by the co-infusion of TPO.

Increased mitochondrial superoxide can affect β-cell function in several ways. Superoxide can decrease mitochondrial glucose oxidation. Superoxide can also activate uncoupling protein 2, resulting in decreased ATP produced by glucose oxidation and thus decreased GSIS (42). The fact that insulin content was only mildly improved by TPO suggests that after 48 h, TPO is mainly acting on the insulin secretory process, presumably by increasing glucose oxidation (44) or decreasing uncoupling (42) via decreased mitochondrial superoxide.

In addition to the mitochondria, the cytosol can also generate superoxide via the activation of NAD(P)H oxidase. β-Cells have been shown to express NAD(P)H oxidase (45), and components of this enzyme are elevated in islets of animal models of type 2 diabetes (46). Furthermore, glucose can activate NAD(P)H oxidase via protein kinase C (47). When NAD(P)H was inhibited in islets of db/db mice, insulin content was partially restored (46), suggesting that cytosolic superoxide may also be important in β-cell dysfunction. Here, we showed that the mitochondria are an important site of superoxide production in a model of glucose-induced β-cell dysfunction. However, further studies need to be performed to assess the role of cytosolic superoxide production.

It is recognized that our glucose infusion model is short-term and extreme in nature and that 48-h glucose infusion does not induce β-cell dysfunction in humans (48). However, there is evidence that glucose-induced β-cell dysfunction does occur in humans following exposure to hyperglycemia (13 mmol/l) for longer than 68 h (48). Thus, we wished to determine the effect of antioxidants in a less extreme, more chronic model of glucose-induced β-cell dysfunction that can be reproduced in humans. We showed that similar to 48-h infusion, TPO but not TAU prevented β-cell dysfunction induced by 96 h of hyperglycemia (plasma glucose ∼15 mmol/l). Further prolongation of in vivo infusion is not feasible, and more chronic glucose exposure requires in vitro studies (49) or studies in hyperglycemic but not hyperlipidemic diabetic models such as GK rats. Future studies in these models may reveal that not only TPO but also other antioxidants are beneficial, perhaps due to their effect on insulin gene transcription (2,10).

Previously, we have shown that TAU and NAC (15) (in doses similar to those in this study) were effective in preventing the decrease in insulin sensitivity induced by 6 h of hyperglycemia. In this study, we showed that following 48-h hyperglycemia, TAU and NAC increased insulin sensitivity in glucose-infused rats. Despite TAU and NAC improving insulin sensitivity, both antioxidants did not increase DI. Oxidative stress has been shown to play a role in both hyperglycemia-induced insulin resistance and hyperglycemia-induced β-cell dysfunction; it is thus unclear why TAU and NAC were effective in the former but not the latter. It is possible that mitochondrial superoxide is important for glucose-induced β-cell dysfunction, at least in the short term, whereas glucose-induced insulin resistance is due to cytosolic production of reactive aldehydes and to cytosolic glutathione depletion.

In vivo, metabolic compensation for insulin resistance involves both increased GSIS and decreased insulin clearance. The C-peptide–to–insulin ratio (index of insulin clearance) was lower at some points after prolonged hyperglycemia, consistent with a reduction in insulin clearance. Addition of TPO tended to reverse the decrease in C-peptide–to–insulin ratio, although not significantly, presumably because this ratio is not a good indicator of insulin clearance. Therefore, it cannot be excluded that a TPO-induced increase in insulin clearance stimulates β-cell compensation for insulin resistance, thus contributing to improved β-cell function in vivo. Nevertheless, data from isolated islets clearly show a direct β-cell effect of TPO.

In conclusion, our study demonstrates that β-cell dysfunction induced by 48- and 96-h glucose elevation in vivo is due to superoxide generation. Furthermore, this study suggests that superoxide dismutase mimetics are of potential interest to preserve β-cell function in type 2 diabetes.

This work was supported by Canadian Institutes of Health Research Grants MOP-69018 and NET-54012 to A.G. and MOP-12898 to M.B.W. C.T. was supported by a Banting and Best Diabetes Centre-Novo Nordisk studentship and an Ontario Graduate Scholarship.

The authors thank Loretta Lam for her excellent technical assistance.