Regulation of α-Cell Function by the β-Cell in Isolated Human and Rat Islets Deprived of Glucose: the “Switch-off” Hypothesis

Both normal male Wistar rats and rats made diabetic by injections of 80 mg/kg STZ were used (16). Two weeks were allowed to elapse after the STZ injection. On the day of pancreas removal, pancreata were cannulated for infusion of collagenase and then removed from the animal for processing, according to the method of Lacy and Kostianovsky (17). After isolation, the islets were hand picked and split into groups of 50 islets each, which were then placed into eight separate perifusion chambers for simultaneous study. The islets were placed in the chambers on top of mesh filters small enough to prevent islets from escaping into the effluent. Human islets were obtained from the Human Islet Distribution Core of the Puget Sound Blood Center in Seattle, Washington. When human islets were used, 100 islets were placed in each chamber. Perifusion was performed using Kreb’s Ringer buffer, 3-isobutryl-1-methylxanthine, regular insulin (Humulin, 300 μU/ml), and various concentrations of glucose (0, 3.0, 5.6, and 16.7 mmol/l). A solution containing 19 amino acids (alanine, 1.64 mmol/l; argininine, 0.70 mmol/l; asparagine, 0.15 mmol/l; citrulline, 0.35 mmol/l; glutamic acid, 0.45 mmol/l; glutamine, 1.88 mmol/l; glycine, 1.12 mmol/l; histadine, 0.29 mmol/l; isoleucine, 0.35 mmol/l; leucine, 0.61 mmol/l; lysine, 1.39 mmol/l; methionine, 0.18 mmol/l; ornithine, 0.26 mmol/l; phenylalanine, 0.31 mmol/l; proline, 1.31 mmol/l; serine, 2.13 mmol/l; threonine, 1.01 mmol/l; tryptophan, 0.28 mmol/l; and valine, 0.75 mmol/l) was used as an alternate stimulator for glucagon secretion. Insulin was measured by the method of Morgan and Lazarow (18), and glucagon was measured by the method of Harris et al. (19).

Results for each perifusion experiment were calculated as the mean ± SE of all eight lanes and considered as n = 1. Comparisons were by paired Student’s t test. P values <0.05 were considered significant.

Isolated islets were perifused with 5.6 mmol/l glucose for 60 min, and then the perifusate was switched to 0 mmol/l glucose for 30 min. In three separate experiments, no glucagon response was observed (Fig. 1A). In six other experiments, islets were first perifused with 3 mmol/l glucose for 60 min and then with 16.7 mmol/l glucose for 30 min. This was followed by a switch to perifusion with buffer containing no glucose. In these experiments, the expected rise in endogenous insulin secretion (0 min = 13 ± 3 and 30 min = 54 ± 8 μU/ml; P < 0.02) and fall in glucagon secretion (0 min = 54 ± 10 and 8 min = 38 ± 7 pg/ml; P < 0.02) was observed during the perifusion with 16.7 mmol/l glucose (Fig. 1B). When the perifusion was switched to a zero glucose concentration, there was a fall in insulin levels (30 min = 54 ± 8 and 60 min = 15 ± 3 μU/ml; P < 0.01) and a prompt rise in glucagon levels (peak change within 10 min after switch off = 61 ± 15 pg/ml [mean ± SE], n = 6, P < 0.01).

Human islets were perifused with 3 mmol/l glucose for 60 min, and then the perifusion was switched to 16.7 mmol/l glucose to characterize β-cell function. Upon exposure to the high glucose concentration, first-phase (0 min = 8 ± 1 and 3 min = 30 ± 9 μU/ml; n = 2) and second-phase insulin secretion was observed (Fig. 2A). Exposure of human islets to 5.6 mmol/l glucose for 60 min, followed by perifusion with 0 mmol/l glucose failed to elicit glucagon secretion (Fig. 2B). In another experiment, human islets were first perifused with 3 mmol/l glucose for 60 min, followed by perifusion with 16.7 mmol/l glucose for 30 min and then by a switch to 0 mmol/l glucose for 30 min. During the infusion with 16.7 mmol/l glucose, insulin secretion increased (0 min = 14 and 30 min = 58 μU/ml) and glucagon secretion decreased (0 min = 151 and 30 min = 132 pg/ml) (Fig. 2C). Upon switching to 0 mmol/l glucose, insulin levels fell and a glucagon response was observed (30 min = 132 and 35 min = 168 pg/ml).

In contrast to normal islets, isolated islets from rats made diabetic with STZ released a small amount of insulin during perifusion with 16.7 mmol/l glucose and had a glucagon response when the perifusate was switched from 16.7 to 0 mmol/l glucose (peak change within 7 min after switch off = 31 ±1 pg/ml, n = 3, P < 0.001) (Fig. 3). However, provision of exogenous insulin via perifusion during the last 15 min of the 16.7-mmol/l glucose infusion, followed by discontinuation of the exogenous insulin at the same time as the glucose perifusion was switched to 0 mmol/l glucose, elicited a significantly (P < 0.05) larger glucagon response (peak change within 7 min after switch off = 71 ± 11 pg/ml, n = 4, P < 0.01) (Fig. 4). This glucagon response did not occur when insulin was not switched off (Fig. 5) or when insulin was switched off without glucose deprivation (Figs. 6 and 7). However, a glucagon response to a solution of amino acids was observed, indicating functional α-cells (45 min = 38 ± 4 and 48 min = 237 ± 89 pg/ml; n = 2) (Fig. 7).

Isolated pancreatic islets, unlike islets in intact pancreata, characteristically do not release glucagon when they are exposed to very low concentrations of glucose. This has variously been attributed to lack of innervation or damage to α-cells during isolation procedures. Our studies were designed to examine the lack of α-cell responsiveness to low glucose levels in the context of regulation of the α-cell by insulin. We posited that the lack of periportal blood flow and delivery of insulin to downstream α-cells renders impossible the provision of an insulin switch-off signal during exposure of the islet to low glucose concentrations. As expected, we observed that the exposure of normal rat and human islets to a zero glucose concentration did not release glucagon. However, if we first perifused the islets with a high glucose concentration to stimulate endogenous insulin secretion and then followed this with discontinuation of the glucose infusion to provide an endogenous insulin switch-off signal, we observed a glucagon response. We noted that the glucagon response occurred before insulin levels in the effluent decreased. We speculate that this is an artifact caused by the perifusate washing away insulin that had collected in the capsules during the time taken to switch perifusate solutions at the 30-min time point. We also examined isolated islets obtained from STZ-induced diabetic Wistar rats. They secreted a small amount of insulin when they were first exposed to a high glucose concentration and released a small amount of glucagon when deprived of glucose. However, when exogenous insulin and high glucose were provided by perifusion before exposing the islets to a zero glucose concentration, switching off the glucose and insulin perifusate elicited a significantly greater glucagon response. Thus, in both normal and β-cell-deficient isolated islets, provision of an insulin switch-off signal utilizing either endogenous or exogenous insulin signaled the α-cells to respond to hypoglycemic conditions. In control experiments, during which the insulin perifusion was not switched off, no glucagon response to zero glucose concentration was observed. In all conditions where we observed a glucagon response, we also noted that glucagon levels rose no higher than “basal levels.” The question these results raise is whether the α-cell perceived that a new basal level had been set by prior exposure to a high glucose concentration and that the amount of glucagon released during zero glucose was the appropriate response. This obviously could not be the case in an in vivo situation wherein defense mechanisms intervene to stimulate glucagon release until normal glucose levels are reached. However, the isolated islet operates in an artificial situation in which it may have a programmed response unrelated to its basal level of secretion. That hypoglycemia was required for a glucagon response was shown in other experiments wherein physiologic or supraphysiologic levels of glucose were included in the perifusate. In these instances, the insulin switch-off signal was not sufficient to elicit glucagon secretion. Consequently, both low glucose concentrations and switching off α-cell exposure to insulin are required for the glucagon response. These conclusions are supported by the findings reported in the accompanying work by Zhou et al. (16), which describes in vivo experiments using STZ-administered rats receiving regional insulin infusions via the pancreaticoduodenal artery.

Evidence for the regulation of α-cell function by the secretory product of the β-cell has been previously provided by several research groups. Weir et al. (20) demonstrated that pancreata from rats made diabetic with STZ had enhanced glucagon secretion and that exogenous insulin could suppress this enhancement. Stagner and Samols (21) reported that retrograde perfusion of a constant glucose concentration increased mean glucagon secretion from dog pancreas and explained this phenomenon by suggesting that it was due to the prevention of insulin in the islet portal circulation from reaching the α-cell downstream. Maruyama et al. (22) perfused anti-insulin serum in rat pancreas and observed a significant rise in glucagon secretion, whereas nonimmune guinea pig serum had no effect. They concluded that insulin maintains an ongoing restraint on α-cell secretion and that loss of this inhibition by insulin may account for the hyperglucagonemia observed in insulin-deficient states. These initial observations were followed by a series of in vivo and in vitro experiments (23–30) that reinforced the hypothesis of β-cell regulation of α-cell function. More recently, McCrimmon et al. (31) have reported reversal of the hypoglycemia-specific defect in glucagon secretion in the diabetic BB rat through the use of a combination of a noninsulin glucose-lowering agent (5-aminoimidazole-4-carboxamide [AICAR]) and phlorizin. This combination induced moderate and equivalent hypoglycemia in both diabetic and nondiabetic animals in the absence of marked hyperinsulinemia. Glucagon responses were improved during the hypoglycemia caused by these drugs, and the improvement was attenuated by infusion of exogenous insulin. The authors concluded that α-cell glucagon secretion and response to hypoglycemia are not defective in this diabetic model if intraislet hyperinsulinemia is prevented. Banarer et al. (15) reported a decreased glucagon response to hypoglycemia during tolbutamide infusion in normal subjects, suggesting that enhanced β-cell secretion of insulin dampens the glucagon response to hypoglycemia. These latter two reports provided strong evidence supporting the switch-off hypothesis. Our studies, which fully support the switch-off hypothesis, more directly assessed this issue by providing insulin to and then removing it from islets from β-cell-depleted diabetic animals.

The role of the central nervous system as a regulator of glucagon secretion is a critically important one. The degree to which the central nervous system is required for normal glucagon responses in vivo during hypoglycemia has recently been reviewed by Taborsky et al. (1). The authors considered three different mechanisms for regulation of the α-cell response to hypoglycemia: direct stimulation of glucagon secretion by low glucose concentrations, local effects of endogenous insulin secretion on neighboring α-cells, and circulating epinephrine as well as autonomic inputs to the α-cells via sympathetic and parasympathetic nerves. They reviewed studies (32–37) that have examined the relative importance of the autonomic nervous system in the regulation of glucagon secretion during hypoglycemia through the use of surgical techniques and pharmacologic agents. One of the strongest arguments that the central nervous system may play the dominant role in regulating α-cell responses to hypoglycemia has been the failure to observe glucagon secretion from isolated islets when they are exposed to very low glucose concentrations. The results described in this work show for the first time that isolated islets can secrete glucagon during glucose deprivation. This result clearly demonstrates that the central nervous system and circulating epinephrine are not required for the glucagon response to hypoglycemia. This finding is consistent with the observation made by Diem et al. (38) that human recipients of ectopically placed and denervated pancreas transplants have intact glucagon responses to hypoglycemia, even during infusion of intravenous propranolol, which blocks a possible glucagon-stimulatory contribution of circulating epinephrine.

This study was supported by National Institutes of Health Grant RO1 DK-39994 (to R.P.R.).