Decrease in β-Cell Mass Leads to Impaired Pulsatile Insulin Secretion, Reduced Postprandial Hepatic Insulin Clearance, and Relative Hyperglucagonemia in the Minipig

The study was designed to examine the effect of a decrease in β-cell mass on the pattern of insulin secretion. First, we established the parameters for pulsatile insulin detection in the pig (protocol 1). Second, we investigated the effects of a decreased β-cell mass achieved by a pharmacological dose of alloxan on pulsatile insulin secretion. The dynamics of insulin secretion were studied in response to a mixed meal (protocol 2) and during a hyperglycemic clamp (protocol 3). Two animals (one male and one female) completed protocol 1, seven animals (five males and two females) completed protocol 2, and 5 animals completed protocol 3. Finally, we measured β-cell mass in the pigs included in the studies. In addition, four pigs started protocol 2 or 3 but did not complete them, as sampling catheters were lost. When available, the data obtained from these pigs was included in the correlations shown in Figs. 2 and 9.

All studies were approved by the Scottish Home Office (equivalent to the Animal Use and Care Committee in the U.S.). A total of 12 Gottingen minipigs (11) (9 males and 3 females) that were 1–3 years old and weighed 20–32 kg were studied. The animals were obtained from Ellegaards Gottingen Minipigs (Gottingen, Denmark), where they were kept in a barrier unit for the first year. In Edinburgh, the pigs were housed under controlled conditions, 12-h light/12-h dark, fed twice daily a total of 450 g of Minipigs Maintenance Diet (9.28 MJ/kg; SDS, Edinburgh, U.K.), 13.8% protein, 51.2% nitrogen-free extract (equivalent to carbohydrate content), 2.5% fat, 12.7% fiber, and 12% water and allowed free access to water. Body temperature, hematocrit, and weight were recorded once a week, and the animals were exercised on a daily basis. The pigs were studied at least 10 days after surgery when they were eating normally, had normal stools, were maintaining weight, and had normal hemoglobin and leukocyte count and normal renal and liver function. Before the study, all pigs were acclimatized to the laboratory and staff.

The technique of surgical implantation of the portal vein sampling catheter (protocol 1 only), arterial catheters, and jugular venous catheters (protocols 1, 2, and 3) was adapted from a canine model described previously (7). In brief, with the pigs under general anesthesia, the portal vein was isolated for a distance of 3–4 cm cranial to the pancreaticoduodenal venous tributary, and a catheter was inserted and advanced cephalic until its tip was located at the bifurcation of the portal vein. A catheter was implanted similarly in the splenic vein, and both catheters exited the left flank and tunneled to the intrascapular area.

The left jugular vein and carotid artery were exposed by cut-down and were catheterized, and the free ends of the catheters tunneled to the left dorsal interscapular area. The catheters were filled carefully to capacity with heparin (10,000 IU/ml) and color-labeled for later identification. The catheters were protected by a vest. Ampicillin (20 mg/kg; SmithKline Beecham, Dublin, Ireland) was given intravenously pre- and postoperatively for infection prophylaxis. Buprenorphine 0.01 mg/kg was given by epidural infusion after induction of anesthesia as analgesic cover for 24 h. This was supplemented by a bolus of Zenecarp 4 mg/kg (Parke Davis, Eastleigh, U.K.) administered intravenously at the end of the surgical procedure. A mix of buprenorphine (0.01 mg/kg) and Zenecarp (2 mg/kg) was given intravenously twice daily postoperatively until the animal had recovered fully from the procedure. Body temperature was measured daily for 7 days postoperatively, and antibiotics were administered only when there were signs of postoperative infection. Patency of catheters was checked every day. The catheter exteriorization site was cleaned regularly with hydrogen peroxide (4%), and Bactroban cream (SmithKline Beecham) was applied.

β-cell mass was reduced by intravenous injection of alloxan monohydrate (Sigma, Dorset, U.K.). The alloxan was refrigerated in a dessicator until used; all alloxan doses were made up from the same batch. The alloxan was weighed out immediately before administration in a 0.1N acetic acid solution (Sigma, Dorset, U.K.), pH 4.4. A dose of 80 mg/kg was diluted in the buffer to a volume of 1.2 ml stock solution/kg. Immediately after preparation, the solution was injected as a bolus through a 0.22-μm Millipore filter (Sigma, Dorset, U.K.) into the jugular vein over 1 min, immediately after which the animals were fed. All of the animals developed mild to moderate hyperglycemia (range 6–11 mmol/l) within the first 24 h after alloxan administration, but this was not treated with insulin until 48 h after alloxan administration. After 48 h, the animals were treated only with insulin (Velosulin, Novo Nordisk A/S, Bagsvaerd, Denmark) if the fasting plasma glucose was above 9 mmol/l (n = 1/7) or if the animal had glucosuria (n = 1/7, the same animal). Plasma glucose was measured daily and urinalysis was performed (Chemstrip, Boehringer Mannheim, U.K.) twice a week to check for glucose, ketones, pH, and protein. No pig developed ketonuria. The diabetic animal that received insulin had no insulin on the morning of the experiments.

Pancreata were retrieved at the end of the 8-week experimental series. The animals were killed with phenobarbital, and after exposure through a midline incision, the gastrointestinal block was removed in toto including the pancreatic tissue surrounding the caudal vena cava inferior and kept in ice-cold saline. The volume and the weight of the pancreas were determined, and the tissue then was dissected immediately on a cold cutting board and kept on ice to prevent autodigestion. Sequential 1-cm samples were fixed in 10% neutral-buffered formaldehyde and then embedded in paraffin. Sections from each block subsequently were immunostained for insulin (guinea pig anti-porcine, 1:1,000; Dako, Carpinteria, CA).

Insulin concentrations in plasma samples were measured in duplicate by an enzyme-linked immunosorbent assay (ELISA) (12) modified for porcine insulin. The assay is based on two monoclonal murine antibodies specific for intact human insulin. The operating range is from 5 to 2,000 pmol/l, recovery 85–120%, and the assay is linear up to 2,000 pmol/l. Glucagon samples were measured by radioimmunoassay as described previously (13). Glucose was measured by the glucose oxidase technique using a Beckman Glucose Analyser (Fullerton, CA).

Deconvolution of insulin secretion from plasma insulin concentration profiles requires knowledge of the decay characteristics (i.e., mono vs. biexponential decay values) in the sampling pool of interest. We established these measures by administration of a bolus of 80 mU of human insulin injected over 30 s via a catheter implanted in the splenic vein while endogenous insulin secretion was suppressed by a somatostatin infusion (Bachem Ltd., Essex, U.K.) 0.1 μg · kg^–1 · min^–1. Insulin (Actrapid, Novo Nordisk A/S) was dissolved in saline and 0.25% porcine albumin (a gift of Novo Nordisk). After the insulin bolus, blood was sampled from both the portal vein and the carotid artery catheters at 15-s intervals for 5 min, every minute for the next 20 min, and every 5 min for the next 25 min. At time 60 min, a constant insulin infusion (1 mU · kg^–1 · min^–1, Actrapid) was given via the same splenic vein catheter and continued for 65 min (to t = 125 min). At time 120 min, blood was sampled at 15-s intervals for 5 min, then every min for 15 min until 140 min, and thereafter every 5 min until 220 min (all samples collected simultaneously from the portal vein catheter and the artery). Throughout this protocol, blood glucose was measured every 5 min and a variable glucose infusion was given to maintain euglycemia.

Although the portal sampling site (versus the systemic circulation) enhances the detection of insulin pulses, it requires invasive intra-abdominal surgery to implant portal vein catheters. We recently reported in humans that use of a highly specific two-site ELISA insulin assay allows accurate detection of insulin pulses sampled from the systemic circulation (14). The purpose of protocol 1b was to extend this observation to the pig to verify use of a systemic sampling site for pulse detection. Portal vein and arterial catheters were implanted in two pigs. Blood was sampled at 1-min intervals simultaneously from the arterial and portal vein catheters in basal (fasting) conditions and during a hyperglycemic clamp (glucose 9 mmol/l) and after a mixed meal (hyperglycemic clamp and meal methods described below in protocols 2 and 3). These data were examined subsequently by deconvolution using criteria for portal versus arterial sampling obtained in protocol 1a.

Pigs (n = 7) were trained to eat a mixed meal that contained glucose in sugar-free jelly (U.S. equivalent Jell-O) mixed with Minipig Maintenance Diet. After a 16-h fast, pigs were placed in a restraining box and the sampling catheters were aspirated and flushed in preparation for sampling. A blood sample was obtained from the artery to determine the hematocrit. Saline 0.9% was infused via the jugular vein catheter at 30 ml/h throughout the study (from t = −40 to 210 min). At t = 0 min, a mixed meal consisting of 2.2 g/kg d-glucose in sugar-free jelly mixed with 50 g of Minipig Maintenance Diet was provided. Blood samples (1.0 ml) were obtained from an arterial catheter (n = 5) at 1-min intervals from −40 to 90 min, every 5 min from 90 to 160 min, and then every 10 min until 210 min for measurement of plasma insulin. Arterial blood samples also were obtained at 5-min intervals from −40 to 160 min and at 10-min intervals from 160 to 210 min for measurement of plasma glucose concentrations. Arterial blood samples were collected for measurement of plasma glucagon concentration at 30-min intervals throughout the study (−40 to 10).

The samples for insulin were drawn into chilled polyethylene tubes that contained EDTA, placed on ice, and centrifuged within 5 min of collection (2 min, 15,000 rpm). Plasma was separated into two tubes and frozen at −20°C for later analysis. The blood glucose samples were drawn into tubes that contained heparin and sodium fluoride, centrifuged, and measured immediately by the glucose oxidase technique (Beckman).

Pigs (n = 5) were prepared as in protocol 1. From −40 to 0 min, insulin secretion was evaluated in the fasting state at the prevailing fasting glucose concentration. From t = 0 to t = 70, the plasma glucose was increased to 8 mmol/l by a glucose infusion. Insulin secretion then was evaluated at steady-state hyperglycemia from t = 30 to t = 70 min. First-phase insulin secretion was evaluated in response to an intravenous glucose bolus (0.3 g/kg) given at t = 70 min. Blood samples (1.0 ml) for subsequent insulin assay were taken from the arterial catheter at 1-min intervals from t = −40 to t = 0 min and from t = 30 to t = 90 min. Blood samples for subsequent glucose measurement were taken every 5 min from the arterial catheter from t = 30 to t = 90 min.

Analysis of the insulin disappearance profiles best fit a single (arterial) or two-component (portal) exponential model, respectively. The biexponential decay observed at the portal vein is consistent with other regional sampling catheters proximate to the site of the secretory burst; the first decay component presumably reflected mixing of hormone within the regional venous blood and its movement through the regional sampling site by blood flow. The rate constant was obtained by interactive curve fitting. The parameters were estimated by Gauss-Newton iteration with confidence intervals for the precision of fit defined by Monte Carlo exploration of the joint parameter spaces.

A comparison of the corresponding arterial versus portal vein plasma insulin concentration profiles was made by inspection as well as cross-correlation analyses of pulse location and magnitude observed in the corresponding deconvolved insulin secretion profiles.

The mean fasting plasma glucose, insulin, and glucagon concentration was calculated in each pig before and after alloxan administration. The mean concentration (and/or area under the concentration curve by the trapezoidal rule) of each of these parameters also was measured during the clamp and after meal ingestion. The glucose/insulin concentration ratio was calculated in each pig 30 min after meal ingestion. The first-phase insulin response to glucose bolus injection was quantified by calculating the area under the curve of the plasma insulin concentration profile by the trapezoidal rule during the 12 min after glucose injection at the end of the hyperglycemic clamp. Pulsatile insulin secretion was quantified by use of a deconvolution method developed by Veldhuis and Johnson (15) and was specifically validated for this purpose in a canine model (7). In brief, the 1-min plasma (duplicate) insulin concentration time series were deconvolved by a multiparameter technique with the following assumptions. The arterial plasma insulin concentration in each animal, as monitored at frequent intervals, results from five determinable and correlated parameters: 1) a finite number of discrete insulin secretory bursts occurring at specific times and having 2) individual amplitudes (maximal rate of insulin secretion within a burst), 3) a common half-duration (duration of an algebraically Gaussian secretory profile), superimposed on a 4) basal time invariant insulin secretory rate, and 5) a monoexponential insulin disappearance model in the systemic circulation. The last parameter consisted of an estimated half-life of 2.8 min and a slow fractional compartment of 0.065 as measured directly with the same assays. Assuming the forgoing nominal insulin disappearance values, we estimated the numbers, locations, amplitudes, and half-duration of insulin secretory bursts, as well as a nonnegative basal insulin secretory rate for each data set using nonlinear least square fitting of the multiparameter convolution for each insulin time series. A modified Gauss-Newton quadratically convergent iterative technique was used with an inverse (sample variance) weighting function. Parameters were estimated until their predicted values and the total fitted variance both varied by <1 part in 100,000. Asymmetric, highly correlated joint variance spaces were calculated for each parameter by the Monte Carlo support plane procedure. Secretory rates are expressed as mass units of insulin (picomoles) released per unit of distribution volume (liters) per unit of time (minutes). The mass of hormone secreted per burst (time integral of the calculated secretory burst) thus is computed as picomoles of insulin released per liter of volume of distribution. The percentage of insulin delivered into the circulation in discrete insulin bursts was calculated as described previously (7).

Evidence of regular periodicity (oscillations) in insulin concentration was examined by two independent techniques, CORRCOS (16) and Fast Fourier Transform-Nonlinear Least Squares (17). Because they led to the same conclusion, we report the data obtained by CORRCOS only. CORRCOS is a fully automated analytical algorithm for quantitative detection and statistical characterization of rhythmic changes in a time series measurement of the feature of interest (here insulin concentration) (16).

The pseudo steady-state clearance rate of endogenously secreted insulin was calculated by the equation C (l/min) = S (pmol/min)/I (pmol/l), where C = insulin clearance rate, S = insulin secretion rate (quantified by deconvolution), and I = arterial plasma insulin concentration.

Pancreatic sections immunostained for insulin were evaluated to determine the percentage of pancreatic cross-sectional area that was positive for β-cells. Islets clearly were present in all slides. An average of 482 islets (range 181–777) were included in evaluations of each pig. The β-cell mass then was calculated by the product of the β-cell fraction of pancreas cross-sectional area and the pancreatic weight. Because pigs served as their own controls (studied before and after alloxan administration), it was not possible to determine β-cell mass before alloxan administration in the seven pigs that underwent protocols 2 and 3. Therefore, correlations of β-cell mass focused on β-cell mass after alloxan administration versus parameters of insulin secretion. Three pigs that did not receive alloxan (used in protocol 1) were available; therefore, these three pigs provided an estimate of normal β-cell mass in this minipig model and an estimate of percentage decrease in β-cell mass after alloxan administration in study pigs. In conclusion, pancreas was obtained from three pigs that did not receive alloxan and from seven pigs that did receive alloxan.

Data are presented as the mean and standard error of the mean. The impact of alloxan treatment on the stated parameters of insulin secretion was determined by using the two-tailed paired Student’s t test. Regression analysis was used to determine the relationship between variables. P < 0.05 was taken as evidence of statistical significance.

The plasma insulin concentration decay curves in the portal vein and the carotid artery after a bolus injection into a superior mesenteric vein are shown in Fig. 1A. The observed insulin concentration decay curve was monoexponential when sampling from the systemic circulation and biexponential when sampling from the portal vein. After discontinuation of the constant insulin infusion, the portal vein decay curve was monoexponential, the slope corresponding to the second component observed from the same sampling site after a bolus injection. The slow-phase half-life obtained from the portal vein sampling site therefore was calculated from the decay curve after constant insulin infusion, which was 2.8 min. This was used to calculate the first half-life after bolus injection, which equalled 0.8 min with a slow fractional component of 0.65. The calculated half-life for insulin in the systemic circulation was calculated both from the decay data after discontinuation of the constant insulin infusion (3.0 min) and from the decay data after the bolus insulin injection (2.8 min). The calculated effective volume of distribution for insulin after bolus injection into the splenic vein with sampling from the arterial catheter was 163 ml/kg body wt.

In both the fasting and the stimulated (meal or hyperglycemic clamp) states, distinct pulses in the concentration of insulin were apparent in both the arterial and the portal vein profiles. Figure 1B illustrates that pulses of insulin detected in the portal circulation (signified by the stars) correspond well to those detected in the arterial circulation in the conditions of the present studies, although as expected, the amplitude of these pulses is greater in portal vein.

Body weight and hepatic and renal biochemical parameters were unchanged after alloxan administration. As expected, administration of alloxan (n = 7 animals) resulted in a spectrum of changes in carbohydrate metabolism in individual pigs, ranging from glucose intolerance (n = 4) to diabetes (n = 3). The mean β-cell mass in the seven pigs that received alloxan was 41.5 ± 14.1% of that in control pigs and varied from 5.6 to 133% (Table 1; Fig. 2).

After alloxan administration, the mean fasting plasma glucose concentration (t = −40 to 0 min) was increased (7.7 ± 0.1 vs. 4.5 ± 0.1; P < 0.05) and increased further after meal ingestion (glucose concentration at 20 min 9.9 ± 1.4 vs. 6.3 ± 0.61 mmol/l; P < 0.02). Alloxan led to a modest decrease in the fasting plasma insulin concentration (55 ± 7.5 vs. 86 ± 17.3 pmol/l; P < 0.01). After the mixed meal, despite marked hyperglycemia, the rise in plasma insulin concentration was suppressed (insulin concentration at 20 min 276.4 ± 85.8 vs. 479.9 ± 125.6 pmol/l; P < 0.05), confirming the presence of a marked defect in meal-stimulated insulin secretion. Thirty minutes after meal ingestion, the glucose/insulin concentration ratio ([mmol/l × 10]/pmol) was markedly increased in pigs after alloxan treatment (P < 0.01). Alloxan treatment did not alter fasting plasma glucagon concentrations, which, given the corresponding hyperglycemia, suggests impaired suppression of glucagon secretion. However, after ingestion of the mixed meal, despite marked hyperglycemia after alloxan administration, the plasma glucagon concentration was increased compared with before alloxan administration (postprandial glucagon concentration 103 ± 6.3 vs. 92 ± 2.5 pg/ml; P < 0.01), emphasizing the failure of the α-cells to suppress glucagon release in response to hyperglycemia (Figs. 2 and 3).

Examination of the insulin concentration profiles before and after meal ingestion and alloxan administration (Fig. 4) indicated that the increased insulin secretion after a meal was accompanied normally by a marked increase in the amplitude of the insulin concentration pulses and that the extent of this increase was attenuated after alloxan administration. The calculated insulin secretion rate (Fig. 5) revealed deficient insulin secretion in the fasting state (∼50% reduction) and strikingly so after meal ingestion (∼80% reduction) after alloxan treatment. The posthepatic percentage of insulin derived from discrete insulin secretory bursts (∼60%) was unaffected by alloxan treatment and comparable to that reported previously when also measured by sampling from the systemic circulation (18). When the insulin concentration profiles were subjected to deconvolution analysis (Figs. 5 and 6), we observed a selective decrease in insulin burst mass (decreased by 54% in the fasting state and 63% after meal ingestion after alloxan administration). In contrast, there was no change in the frequency of pulsatile insulin secretion (Table 2).

When the insulin concentration data were examined for regular periodicity by CORRCOS, statistically significant regular oscillations were detected in five of the seven pigs in the fasting state both before and after alloxan administration but in only two of the seven pigs after meal ingestion. In the five pigs with oscillations present in the fasting state, the oscillation interval was unchanged before versus after alloxan administration (12.3 ± 1.7 vs. 12.9 ± 1.2 min).

The calculated insulin clearance rate (Fig. 7) for endogenously secreted insulin was slightly lower in the fasting state (P < 0.05) after alloxan administration. However, previous alloxan treatment lead to an ∼40% decline in insulin clearance rate after meal ingestion (P < 0.01). The insulin clearance rate was correlated with the amplitude of insulin pulse mass (r = 0.62, P < 0.001; Fig. 8).

Pigs that had been treated with alloxan after meal ingestion showed a strong positive correlation among the insulin secretion rate (r = 0.98, P < 0.01), the insulin pulse mass (r = 0.98, P < 0.001), and the 30-min postprandial insulin concentration (r = 0.6, P < 0.001) and the β-cell mass. However, there was no relationship between β-cell mass and insulin pulse interval (r = −0.14, P = NS). Inspection of the relationship between the fasting plasma glucose concentration or the 30-min postprandial glucose/insulin ratio versus the β-cell mass (Fig. 2) is instructive. These data suggest that there is a fairly wide range of β-cell mass that will suffice to accomplish glucose homeostasis. They also suggest that if the β-cell mass is decreased beyond a critical level (∼0.2 g in this porcine model), then there is insufficient insulin to retain glucose homeostasis, particularly in the fed state (Fig. 9).

Before the clamp, the patterns of the plasma glucose and insulin concentrations corresponded to those observed before meal ingestion in protocol 2. During the clamp, the glucose concentrations were comparable before and after alloxan administration (8.4 ± 0.5 vs. 9.8 ± 0.8 mmol/l). The glucose infusion rate required to achieve this (protocol time 30–70 min) was greater before versus after alloxan administration (12.5 ± 0.3 vs. 1.6 ± 0.7 mg · kg^–1 · min^–1; P < 0.01). Despite the comparable glucose concentrations during the clamp, the plasma insulin concentration was decreased after alloxan administration, suggesting a defect in insulin secretion (Fig. 10; Table 2).

Inspection of the individual insulin concentration profiles revealed an increase in the amplitude of insulin pulses after induction of hyperglycemia during the clamp and that this increment was attenuated after alloxan administration (Fig. 11). The clamp-stimulated insulin secretion rate in pigs after alloxan administration was only ∼10% of that before alloxan administration during the clamp. Deficient insulin secretion was attributable to a decreased insulin pulse mass (decreased to ∼10%) with no change in pulse frequency (Figs. 12 and 13). The pattern of insulin clearance rates was similar to that seen in the meal study (i.e., decreased after alloxan administration) but was not significant during the clamp (one of the five pigs had higher and four had lower insulin clearance rates; Table 2). The relationship between β-cell mass and insulin secretion was comparable to that observed after meal ingestion with a positive correlation between insulin secretion (r = 0.84, P < 0.05) and insulin pulse mass (r = 0.82, P < 0.05) but no relationship with insulin pulse interval.

As in the meal study, examination of the insulin concentration data for regular oscillations by CORRCOS revealed significant oscillations in most pigs (four of five) in the basal fasting state but only one of five pigs in the stimulated hyperglycemic clamp state. The oscillation interval was similar before and after alloxan administration in the fasting state (9.9 ± 1.2 vs. 8.9 ± 1.0 min).

First-phase insulin secretion in response to the intravenous glucose bolus given at comparable hyperglycemia (end of clamp) was decreased after alloxan administration (1.4 ± 0.6 vs. 11.6 ± 2.2 nmol · l^–1 · 12 min^–1; P < 0.001). There was a positive correlation between first-phase insulin secretion and the insulin pulse mass after alloxan administration (r = 0.63, P < 0.01).

In the present study, we sought to address the hypothesis that the defective pulsatile insulin secretion present in patients with type 2 diabetes can be recapitulated by a selective decrease in β-cell mass achieved by alloxan administration in a porcine model. The data reported here support this hypothesis.

Type 2 diabetes in humans is characterized by fasting hyperglycemia with an exaggerated postprandial glycemic excursion, both of which were observed in the present pig model. The insulin concentration in the systemic circulation generally has been reported as comparable in patients with type 2 diabetes and control subjects during fasting (1,5,18) but decreased compared with control subjects early after meal ingestion (1,16,19,20,21). We previously reported that increased insulin secretion after meal ingestion is achieved by an ∼600% amplification of insulin pulse mass (22). Because defective insulin secretion in patients with type 2 diabetes is most evident in the fed state (5) and to examine the effect of a partial decrease in β-cell mass on insulin secretion, we focused on pulsatile insulin release after ingestion of a mixed meal. In the present porcine model, we were able to demonstrate that a partial and selective β-cell deficiency results in decreased insulin secretion in response to meal ingestion that is due to a deficit in the quantity of insulin secreted with each insulin secretory burst while the insulin secretory burst frequency was unchanged. This closely mimics the abnormal pattern of pulsatile insulin release recently observed in patients with type 2 diabetes using the same insulin assay and deconvolution techniques (9). In the current study, we also examined the insulin concentration for regular periodicity. In the fasting state, regular oscillations were detected in the majority of animals with a frequency that was comparable to that observed in previous studies that used rhythm analysis under these fasting conditions (23,24,25), which is ∼30% greater than the pulse interval detected by deconvolution, presumably reflecting this degree of irregularity of insulin secretory bursts (deconvolution does not require a constant pulse interval in contrast to spectral analysis). However, we did not observe statistically significant regular oscillations in the majority of pigs after stimulation of insulin secretion by either enteral or intravenous glucose, which suggests greater irregularity of insulin oscillations after glucose stimulation. A primary assumption of spectral analysis methods is that there is stationarity in the data set. However, after glucose stimulation, aggressive nonstationarity is evident in the insulin concentration profiles, so it is not surprising that spectral methods are unable to accommodate these data sets. Some previous studies reported a disturbance of insulin concentration rhythms in patients or in relatives of patients with type 2 diabetes (24,25). However, a recent study that used the same ELISA insulin assay that we used here also did not report any disturbance in pulse rhythm by spectral analysis in patients with type 2 diabetes (26). Therefore, the issue of whether there is a rhythm disturbance (altered regular periodicity) in insulin concentrations in patients with type 2 diabetes remains to be resolved.

In a previous study of partial β-cell deficiency induced by streptozocin in baboons, an abnormal pattern of pulsatile insulin secretion was observed (27). However, the current study differs from that in several ways. First, in the study of baboons, several doses of streptozocin were used and insulin secretion was studied after intravenous glucose and arginine injections. The technique for examining pulses did not quantify pulse mass or the rate of insulin secretion. The mean fasting plasma glucose concentration in the treated baboons increased only from 4.6 to 5.8 mmol/l. Despite these differences, the reported decline of insulin pulse amplitude with no change in pulse frequency observed in the baboons after streptozocin administration is consistent with the decrease in pulse mass (and amplitude) and unchanged pulse frequency evident in the current porcine model. One year after hemipancreatectomy, humans have impaired glucose tolerance (28) and decreased first-phase insulin release (29). However, no information is available concerning pulsatile insulin release in these patients. A 64% pancreatectomy in dogs also led to a decrease in insulin secretion in response to arginine that was most pronounced at increased plasma glucose concentrations (i.e., glucose potentiation was decreased) (30). In humans after hemipancreatectomy and dogs after an approximately two-thirds pancreatectomy, both the fasting plasma glucose and the insulin concentrations (measured by conventional immunoassay) were normal, raising the question, “How does a decreased capacity for insulin secretion adapt to prevent the development of diabetes?” The possibilities were examined by Marincola et al. (31) in a study of dogs that had either sham surgery or ∼80% pancreatectomy. In these studies, pancreatic insulin output was measured directly by sampling from the portal vein. In the fasting state, insulin output was decreased to ∼30% of control values after 80% pancreatectomy, whereas the simultaneous plasma insulin concentration in the systemic circulation was comparable in the pancreatectomy and control animals. These data suggest that hepatic insulin clearance of endogenously secreted insulin was decreased by ∼60% in the dogs that underwent partial pancreatectomy.

In the present study, we also observed decreased clearance of endogenously secreted insulin in the setting of a decrease in β-cell mass. This approach is to be contrasted to measuring insulin clearance by infusing exogenous insulin, which has been shown to be unaffected by a partial pancreatectomy (30). We previously showed that partial suppression of insulin secretion achieved with a low-dose infusion of somatostatin reduces the clearance rate of endogenously secreted insulin (32). In that study, the insulin clearance rate was closely correlated to the insulin pulse amplitude. We speculated that the clearance rate of endogenously secreted insulin is related directly to the pulse and amplitude of insulin secretory bursts, so when the pulse amplitude decreases (as a result of decreased β-cell mass or somatostatin), the hepatic extraction of secrete insulin also decreases. We further hypothesized that this dual mechanism for the regulation of insulin delivery to the systemic circulation (modulation of insulin secretion rate with an associated modulation of the insulin clearance rate) serves to buffer the delivery of insulin to the systemic circulation. The present study lends support to these hypotheses. Of note, although the rate of insulin clearance measured in patients with type 2 diabetes has been reported to be unaltered when insulin is infused exogenously (in a nonpulsatile manner into the systemic circulation) (33,34), it is decreased when estimated endogenously (35). Because the defect of insulin secretion in patients with type 2 diabetes is characterized by a decreased pulse mass (9), these data support the notion that in this disease, the seemingly normal or near-normal insulin concentrations in the peripheral circulation may be due in part to a compensatory decrease in the rate of hepatic extraction of insulin as a result of the decreased pulse mass of secreted insulin. The present studies do not address the specific mechanism that accounts for the relationship between pulse amplitude and insulin clearance.

In the current study, we also observed hyperglucagonemia in the pigs that were treated with alloxan. Patients with type 2 diabetes also have abnormalities in glucagon secretion (36). Thus, in the fasting state, despite hyperglycemia, the glucagon concentration is not suppressed and after ingestion of a mixed meal is even higher than that seen in control subjects (37,38,39). The same pattern is observed in the present porcine model, which suggests that deficient β-cell mass alone and/or chronic hyperglycemia may lead to this defect in α-cell function. These data suggest that hyperglucagonemia in type 2 diabetes may be due to loss of paracrine inhibition of α-cell secretion by intra-islet insulin (40), although this postulate clearly merits further investigation.

From the present model, it is apparent that many of the characteristics of impaired insulin release present in patients with type 2 diabetes can be reproduced by a partial selective loss of β-cells. How are these different aspects of insulin release related, and what is the underlying mechanism? One possible link is the magnitude of the immediately secretable insulin pool. This pool has been recognized for many years as a physiological concept arising from the observed biphasic nature of insulin release in response to an acute glucose challenge (41). Grodsky (42) proposed that the first phase of insulin release was due to the exocytosis of insulin secretory granules docked at the plasma membrane with the later phase of release being related to the movement of undocked granules to the membrane thereafter. It has been possible only recently to visualize directly the granule traffic within secretory cells and to gain insights into the anatomical nature and regulation of the proposed immediately secretable pool. The immediately secretable insulin pool is now confirmed to consist of the insulin vesicles docked at the plasma membrane (43). Recent studies of granule traffic and exocytosis in single cells show that the net rate of movement of vesicles from the remote intracellular vesicle pool to the membrane-bound pool depends on the relative rate of vesicles being docked onto the membrane versus the rate of their release, in which some membrane-bound vesicles become undocked and return to the intracellular pools (44). In the chromaffin cell, acute stimulation of catecholamine secretion results in a rate of exocytosis that exceeds the maximal capacity of the cell to replete this pool from the remote pool so that the membrane-bound vesicles are at least transiently depleted (45). If the demand for secretion imposed on a secretory cell continued to exceed the capacity for the forward movement of vesicles from the large remote pool of granules to the cell membrane, then the membrane-bound “first-phase pool” would be depleted. Assuming that insulin release during a spontaneous insulin burst also is derived from the docked insulin secretory granules, one would predict that depletion of this pool would result in a concurrent decrease in both spontaneously pulsatile and first-phase insulin secretion.

Support of the above model is provided by the observation that both pulsatile insulin burst mass and first-phase insulin release were restored at least partially by a period of overnight inhibition of insulin release imposed by a somatostatin infusion in patients with type 2 diabetes (9). Culture of human islets at high glucose concentrations leads to depletion of insulin stores and decreased first-phase insulin release (46). Culture of human islets with high glucose concentrations along with an inhibitor of insulin secretion, diazoxide, results in a relative sparing of insulin stores and enhancement of first-phase insulin release (47) and insulin pulse amplitude (48). If available insulin stores are critical for secretion of insulin secretory bursts of an appropriate size, then chronic high glucose concentration in nondiabetic individuals would be expected to lead to diminished insulin secretion. When plasma glucose was sustained at ∼12 mmol/l in nondiabetic healthy humans for 36 h, insulin secretion was decreased and hepatic insulin clearance was increased (49). Because almost all insulin is secreted in insulin pulses, it is safe to infer that the decrease in insulin secretion in that study was characterized by decreased insulin pulses and that this also may have been the cause for the decreased hepatic insulin clearance. In the current study, the decreased number of β-cells present after alloxan administration is required to provide enhanced insulin secretion per cell as well as to be exposed to chronic increased glucose concentrations. The direct relationship between the extent that β-cell mass was reduced by alloxan and the glucose-stimulated insulin secretion rate (after meal or infusion) observed in the present study also is generally supported in this model. It also is possible that some of the alterations of β-cell function observed in the present porcine model are due to sustained injury to surviving β-cells by alloxan administration, although the toxic action of alloxan generally is believed to result in cell death or full recovery. Few data are available on insulin stores in islets from patients with type 2 diabetes, although, generally, these data suggest a loss of total insulin stores (50,51).

In summary, the present study reveals that a partial decrease in β-cell mass results in impaired insulin secretion particularly during enteral or intravenous glucose, which is mechanistically due to a selective decrease in insulin secretory burst mass. Whereas this defect results in a major decline in prehepatic insulin secretion, the effect on delivery of insulin to the systemic circulation is significantly offset by a corresponding fall in hepatic insulin clearance. Insulin secretion and insulin pulse mass were related directly to the β-cell mass. Data from this and previous studies suggest that the decrease in hepatic clearance of endogenously secreted insulin is due to the concurrent decrease in the amplitude of insulin concentration pulses presented to the liver. The specific mechanism to account for the correlation between portal vein insulin concentration amplitude and hepatic insulin clearance remains to be determined.

We acknowledge funding for this project by the Wellcome Trust (project grant) and a Novo Nordisk educational grant. Support for costs of the glucagon assays was made available by the European Association for the Study of Diabetes. L.L.K. was supported by a postdoctoral fellowship from the Danish Medical Research Council.

We acknowledge the excellent technical assistance of Jaqi Armstrong and Joan York (assistance with pig studies and insulin assays); Paula Azimi (with calculations); Christiane Jaminet and Angella Rombaux (glucagon assays); Don Henderson, Duncan McNeil, and Graham Grady for assistance with animal husbandry; and Kim Denkers for editorial assistance.