Mechanisms of Impaired Fasting Glucose and Glucose Intolerance Induced by a ∼50% Pancreatectomy

The current study was undertaken to examine the impact of a ∼50% partial pancreatectomy on insulin secretion and insulin action. A total of 14 mongrel dogs ∼1–3 years old weighing 20–24 kg were included in the current study. Seven dogs underwent a ∼50% pancreatectomy, and seven dogs underwent a sham surgical procedure (see details below). The seven dogs that underwent partial pancreatectomy were included in all three study protocols. All seven of the dogs that underwent sham surgery were included in protocols 1 and 2, and six of the sham-operated dogs were included in protocol 3, one being excluded because of clotted catheters.

Dynamics of insulin secretion were studied in response to oral glucose ingestion (protocol 1) and a hyperglycemic clamp (protocol 2). Insulin action was examined by the hyperinsulinemic-euglycemic clamp and isotope dilution to quantify glucose turnover (protocol 3). The hematocrit and body weight on the day of study and the timing of the studies in relation to surgery are shown in Table 1. The studies were approved by the animal care and use committee.

After an overnight fast, the dogs were anesthetized, using thiobarbiturate. Once anesthesia was achieved, it was subsequently maintained with 1.5–2.5% halothane in 3 l of oxygen and 2 l of nitric oxide per minute. After induction of anesthesia, a midline incision was performed, and the pancreas was exposed by elevating the stomach and spreading the duodenal mesentery. The dorsal pancreas was then subsequently mobilized by division of splenic arterial and venous tributaries and divided by cautery distal to the gastroduodenal vein. During this process the distal pancreas becomes cyanotic, thereby demarcating the minimum resection allowable. The ventral pancreas was then carefully dissected from the inferior pancreatoduodenal artery and vein, leaving the structure intact, and excised at a point distal to the pancreatic duct. Both raw edges of the pancreas were subsequently oversewn with 3-0 silk sutures. The initial extent of the pancreatectomy was determined by surgeons estimate (∼50%). To determine the actual extent of pancreatectomy, the surgically resected pancreas and the pancreatic remnant (removed at autopsy) were weighed, and the actual percentage of pancreatic resection was determined. The mean extent of pancreatectomy was 52 ± 2%. The sham surgery procedure involved the same dissection, with mobilization of the pancreas, but no pancreas removal. After the pancreatectomy or sham procedure, all dogs underwent implantation of a portal vein sampling catheter, portal vein flow probe, and arterial sampling catheter. Full descriptions of these procedures have been described previously in detail (14). In short, the portal vein sampling catheter was placed into the portal vein at the bifurcation of the portal vein in the liver. A Doppler flow probe was placed around the portal vein, and the probe cable and the portal catheter were tunneled subcutaneously to a pocket near the scapula. An arterial sampling catheter was also placed into the carotid artery and similarly tunneled into the subcutaneous pocket. All dogs recovered to their preoperative body weight and maintained body weight through the study period (Table 1). Partially pancreatectomized and sham-operated dogs ate all provided food (same for each group) and had normal stools and recovered their preoperative hematocrit (Table 1). The three protocols were carried out in the same order in all partially pancreatectomized and sham-operated dogs, and the same interval after surgery (protocol 1 at 14 days postoperation, protocol 2 at 21 days postoperation, and protocol 3 at 28 days postoperation).

After an overnight fast (16 h), dogs (n = 14) were placed in a laboratory sling. Portal vein and arterial sampling catheters as well as the flow probe cable were exteriorized from the subcutaneous pocket after use of a local anesthetic as previously described (14). The flow probe was connected to the transducer to continuously record corresponding portal vein blood flow levels. Normal saline was infused through the foreleg infusion catheter at 30 ml/h throughout the study (t = 0–150 min). Blood (∼1 ml) was then sampled from the portal vein catheter at 1-min intervals for measurement of the plasma insulin concentrations at basal (t = 0–60 min) and after a 30-g glucose ingestion (t = 61–150 min). Additionally, blood samples (∼1 ml) were also obtained at 10-min intervals from the arterial catheter for corresponding measurements of plasma insulin concentrations. All samples for plasma insulin were taken into ice-cold glass tubes containing EDTA, immediately cold-centrifuged, and stored at −20°C until analyzed. Plasma glucose levels were measured in additional 0.5-ml blood samples, using the glucose oxidase method (Beckman Instruments, Fullerton, CA), collected at 10-min intervals from the arterial catheter during the entire study (t = 0–150 min).

Dogs (n = 13) were prepared for the study as in protocol 1. After the baseline period (t = 0–40 min), glucose (20% dextrose) was infused via the foreleg catheter to clamp arterial glucose levels at ∼8 mmol/l (t = 40–130 min). Portal vein blood samples (∼1 ml) for subsequent plasma insulin measurements were obtained at 1-min intervals at baseline (t = 0–40 min) and during steady-state hyperglycemia (t = 60–130 min). Additionally, blood samples (∼1 ml) were also obtained at 10-min intervals from the arterial catheter for corresponding measurements of plasma insulin. Blood samples for glucose measurements were taken from the arterial catheter at 10-min intervals at baseline (t = 0–40 min) and every 5 min during glucose infusion (t = 41–130 min)

Dogs (n = 13) were prepared for the study as described in protocols 1 and 2. A primed (3 mg/kg) continuous (3 mg · kg⁻¹ · h⁻¹) infusion of [6,6-²H₂]glucose (Cambridge Isotope Laboratories, Andover, MA) was delivered into the foreleg catheter for 3 h during the basal period (t = −180 to 0 min) and for 3 h during insulin infusion (t = 0–180 min). The isotope was confirmed to be >99% pure by high-performance liquid chromatography. A continuous (nonprimed) infusion of 0.6 mU · kg⁻¹ · min⁻¹ insulin (Humulin; Eli Lilly, Indianapolis, IN) was started at t = 0 and continued until t = 180 min into the foreleg catheter, and the glucose concentrations were maintained within fasting levels via exogenous glucose infusion (20% dextrose) labeled with [6,6-²H₂]glucose to avoid non–steady-state errors in measurement of glucose turnover. Glucose disposal (R_d) and hepatic glucose release were determined in each dog at baseline and during the final hour of the hyperinsulinemic-euglycemic clamp as previously described (15). Blood was sampled from the arterial catheter twice at baseline (at −30 and 0 min) and every 10 min during the clamp (0–180 min) for determination of insulin concentrations and [6,6-²H₂]glucose enrichments. Additionally, blood samples for glucose measurements were taken from the arterial catheter at baseline (at −30 and 0 min) and every 5 min during the hyperinsulinemic-euglycemic clamp (t = 0–180 min).

Insulin levels in plasma were measured in triplicate by a validated radioimmunoassay (14). In short, the primary antibody (no. 1040; Linco Research, St. Louis, MO) has been determined to have 100% cross-reactivity to dog insulin. The operating range of the assay was determined to be from 15 to 968 pmol/l, using 50 μl plasma samples. The intra- and interassay coefficients of variation were 5 and 15%, respectively.

The portal vein plasma insulin concentrations were subjected to multiparameter deconvolution analysis, as previously described in detail (14). Briefly, the multiparameter deconvolution technique used in this study was based on the following five assumptions: 1) there is a finite number of discrete insulin secretory bursts occurring at specific times and having 2) individual amplitudes and 3) a common half-duration (50% of the time interval of a pulse), which is superimposed on 4) a basal time-invariant insulin secretory rate, and 5) there is a biexponential insulin disappearance model in the portal vein, consisting of half-lives of 0.2 and 3 min and a fractional slow-component amplitude of 0.065. The mean rates of insulin clearance were calculated at baseline (0–40 min), during oral glucose ingestion (60–150 min), and during hyperglycemic clamp (60–130 min) studies in each dog, using the equation: C (l/min) = S (pmol/min)/I (pmol/l), where C is the insulin clearance rate, S is the insulin secretion rate (by deconvolution), and I is the arterial plasma insulin concentration.

All data are the means ± SE. ANOVA was used to determine differences in concentrations and deconvolved secretion parameters in control and pancreatectomized animals. Regression analysis was used to examine relationships between variables. P < 0.05 was considered significant.

Body weight and hematocrit returned to preoperative levels in both groups of animals by the time of the first study (protocol 1) 2 weeks later. Body weight and hematocrit were comparable between groups (partial pancreatectomy and sham surgery) at the time of each study. Pancreatectomy resulted in elevated fasting plasma glucose in partially pancreatectomized compared with sham-operated dogs (113 ± 4 vs. 102 ± 2 mg/dl at 2 weeks after surgery, P < 0.05). This fasting glucose concentration was relatively stable for the first 3 weeks after partial pancreatectomy but then declined, but it was still increased versus sham surgery by 4 weeks (Table 1).

At 2 weeks postpancreatectomy, despite IFG, mean fasting arterial insulin levels in partially pancreatectomized dogs were no different than sham surgery controls (76 ± 5 vs. 70 ± 10 pmol/l for partial pancreatectomy vs. sham surgery, P > 0.05). However, at 2 weeks post–partial pancreatectomy, the basal insulin secretion rate in partially pancreatectomized dogs was diminished by ∼50% compared with sham-operated dogs (1.2 ± 0.3 vs. 2.4 ± 0.4 pmol · kg⁻¹ · min⁻¹, P < 0.05) (Table 2). This deficit in basal insulin secretion was caused by a ∼50% decrease in insulin pulse mass (P < 0.05), with no change in pulse frequency (Table 2), and it was concealed at the arterial sampling site by decreased hepatic insulin clearance (Table 2).

After glucose ingestion, arterial plasma glucose levels increased in partially pancreatectomized dogs to a higher peak (209 ± 17 vs. 151 ± 12 mg/dl, t = 110 min, P < 0.05) (Fig. 1) that was more prolonged (187 ± 16 vs. 139 ± 8, t = 70–150 min, P < 0.05) (Fig. 1) than in sham-operated dogs. Coincident with the peak postprandial plasma glucose concentrations (t = 110 min), despite higher plasma glucose levels in partially pancreatectomized dogs, the arterial (310 ± 45 vs. 277 ± 46 pmol/l, P > 0.05) (Fig. 1) and portal vein (509 ± 36 vs. 568 ± 148, P > 0.05) (Fig. 1) insulin concentrations were comparable in partially pancreatectomized and sham-operated dogs, respectively. Consequently, mean arterial (298 ± 31 vs. 243 ± 31 pmol/l, P > 0.05) (Fig. 1) and portal vein (566 ± 88 vs. 495 ± 79, P > 0.05) (Fig. 1) insulin concentrations during the entire 90-min period after glucose ingestion (t = 70–150 min) were also comparable between the two groups. Deconvolution analysis of 1-min portal vein insulin concentrations revealed that the increase in insulin secretion rate after glucose ingestion was accomplished by an approximately four- to sixfold increase in insulin pulse mass with no change in the frequency of pulsatile insulin secretion (Table 2, Figs. 3 and 4). Despite the increased plasma glucose levels in partially pancreatectomized dogs after glucose ingestion, insulin secretion was slightly, though not significantly, lower in partially pancreatectomized versus sham-operated dogs (Fig. 4, Table 2), implying defective glucose-mediated insulin secretion. To characterize this further at matched glucose concentrations, we performed hyperglycemic clamps.

During the basal period (t = 0–40 min), plasma glucose and insulin concentrations and insulin secretion were comparable to those observed in fasting conditions before the meal study (vida supra). By design, plasma glucose levels were matched during the clamp (t = 60–130 min, 140 ± 3 vs. 138 ± 5 mg/dl, P > 0.05) (Fig. 2). The glucose infusion rate required to maintain steady-state hyperglycemia was ∼30% in partially pancreatectomized dogs compared with sham surgery controls (3.5 ± 0.4 vs. 11.3 ± 2.3 mg · kg⁻¹ · min⁻¹, P < 0.01). Steady-state arterial (161 ± 36 vs. 330 ± 94 pmol/l, P < 0.01) (Fig. 2) and portal vein (364 ± 94 vs. 868 ± 321, P < 0.01) (Fig. 2) insulin concentrations during the clamp were decreased in partially pancreatectomized dogs because of a ∼70% deficit in insulin secretion. The latter was caused by a comparable deficit in insulin pulse mass (123 ± 51 vs. 596 ± 184 pmol/pulse, P < 0.05) (Table 2, Figs. 5 and 6), with no change in pulse interval (Table 2).

The calculated insulin clearance rate (of endogenously secreted insulin) was decreased in partially pancreatectomized dogs compared with controls (Table 2) at baseline (0.4 ± 0.1 vs. 0.7 ± 0.1 l/min, P < 0.05) and during oral glucose ingestion (0.4 ± 0.1 vs. 0.9 ± 0.2, P < 0.05). Insulin clearance was directly correlated with insulin pulse mass at baseline, after oral glucose ingestion, and during the hyperglycemic clamp (P < 0.05 for all three) (Fig. 7).

By design, plasma insulin and glucose concentrations were matched in partially pancreatectomized and sham-operated dogs during the hyperinsulinemic-euglycemic clamp (Fig. 8). Similarly, [6,6-²H₂]glucose percent enrichment remained at steady state during the clamp (Fig. 9). Whole-body insulin sensitivity, assessed by the mean glucose infusion rate required to maintain euglycemia in the final hour of the clamp, was decreased by ∼45% in partially pancreatectomized compared with sham-operated dogs (3.9 ± 0.9 vs. 6.8 ± 1.5 mg · kg⁻¹ · min⁻¹, P < 0.05) (Fig. 8). This impaired insulin sensitivity in partially pancreatectomized animals was caused by a selective deficit in insulin-stimulated glucose disposal (27.3 ± 3 vs. 42.2 ± 11.2 μmol · kg⁻¹ · min⁻¹, P < 0.05) (Fig. 9). In contrast, insulin-mediated suppression of hepatic glucose release (7.2 ± 0.8 vs. 7.2 ± 0.4 μmol · kg⁻¹ · min⁻¹) (Fig. 9) was unchanged. In the partially pancreatectomized dogs, insulin sensitivity measured isotopically or by the glucose infusion rate required to maintain euglycemia correlated with both insulin pulse mass and insulin pulse amplitude measured in the same dogs during the hyperglycemic clamp (Fig. 10).

We report that a ∼50% β-cell deficit caused by partial pancreatectomy in dogs leads to both IFG and IGT (IFG/IGT). Furthermore, we report that a ∼50% deficit in β-cell mass leads to defective glucose-induced insulin secretion and extrahepatic insulin resistance, reproducing the findings in humans with IFG/IGT (16–19). These findings support an important role for defective β-cell mass in the pathogenesis of IFG/IGT and subsequently type 2 diabetes.

The pattern of postprandial insulin secretion in partially pancreatectomized dogs is similar to that previously reported in humans with IFG/IGT (17,20). After glucose ingestion the initial increment in insulin secretion (first 30 min) was both delayed and inappropriate given the prevailing plasma glucose levels. Interestingly, this deficit was apparent from insulin concentrations in the portal vein sampling site but not the arterial sampling site, reflecting decreased hepatic insulin extraction after partial pancreatectomy (vida infra). Subsequently (60–120 min after glucose ingestion), partially pancreatectomized dogs had postprandial hyperinsulinemia compared with controls because of both marked hyperglycemia and impaired hepatic insulin clearance reminiscent of the postprandial glucose profile in humans with IFG/IGT (17,20).

To permit measurement of glucose-mediated insulin secretion in partially pancreatectomized and sham-operated dogs at comparable glucose concentrations, we also performed hyperglycemic clamps. Under these conditions glucose-stimulated insulin secretion was ∼70% defective in partially pancreatectomized versus sham-operated dogs because of an ∼80% deficit in insulin pulse mass and amplitude, with no change in pulse frequency. These findings recapitulate both the pattern and extent of the defect in insulin secretion previously reported in humans with type 2 diabetes (21). A similar defect in insulin secretion can be accomplished by culturing human islets at a glucose concentration of ∼11 mmol/l, leading to depletion of insulin stores, whereas both the deficit in insulin stores and insulin secretory burst mass are prevented if the islets are concurrently cultured at the same glucose concentration with a potassium channel opener that inhibits insulin secretion (22,23). These findings imply that the defective glucose-mediated insulin secretion in type 2 diabetes may be caused at least in part by a deficit in the docked and primed pool of insulin secretory granules. This postulate gains support from the complete reversal of defective glucose-mediated insulin secretion in type 2 diabetes when insulin secretion is temporarily inhibited overnight by somatostatin (21). In the current study, decreasing β-cell mass by ∼50% by surgical pancreatectomy presumably immediately increased demand per β-cell by approximately twofold, a demand that was presumably further increased by the insulin resistance induced by the ∼50% partial pancreatectomy as well as the modest increase in fasting glucose (IFG). In these in vivo studies, as in humans with IFG/IGT, it is not possible to distinguish whether the defect in glucose-mediated insulin secretion is only a consequence of normal secretion by fewer β-cells or is caused by defective insulin secretion by each β-cell. Although we postulate that the decrease in glucose-mediated insulin secretion in the partially pancreatectomized dogs may be caused by depletion of the pool of insulin available for immediate secretion, future studies with islets isolated from partial pancreatectomy versus sham surgery pancreas will be required to resolve these issues.

In the current study, we also observed that a ∼50% partial pancreatectomy caused insulin resistance, which was attributed to a ∼40% decrease in insulin-stimulated glucose disposal. Humans with IFG and/or IGT are insulin resistant compared with healthy control subjects (18), an observation often attributed to obesity (19). However, when both obese and nonobese glucose-intolerant individuals were examined compared with corresponding weight-matched healthy control subjects, the degree of insulin resistance was similar in both obese and lean individuals with IFG/IGT (24). The current study implies that the increment in insulin resistance in IFG/IGT over body mass–matched individuals with normal fasting glucose and glucose tolerance may be attributable in part to an abnormal pattern of insulin secretion consequent to a deficit in β-cell mass. There are other lines of evidence in favor of a role of a defect in β-cell mass contributing to insulin resistance. Humans with type 1 diabetes have a degree of insulin resistance comparable to humans with type 2 diabetes when matched for age and BMI (25). A number of factors have been implicated in the pathogenesis of this increment in insulin resistance in diabetes. These include defective pulsatile insulin secretion (26–29) and glucose toxicity (30–32). We did find a relationship between insulin sensitivity (measured isotopically or by glucose infusion rate during the hyperinsulinemic-euglycemic clamp) and the insulin pulse mass (and amplitude) that lends some support to a potentially important role of the pattern of insulin secretion influencing insulin sensitivity. The defect in pulsatile insulin secretion after 50% partial pancreatectomy was documented here by portal vein sampling, revealing that the liver is exposed to very different insulin concentration profiles under both basal and glucose-stimulated conditions after partial pancreatectomy. Although blood volume constraints prevented simultaneous 1-min sampling from the peripheral circulation, 10-min sampling from the peripheral circulation under basal conditions were not different between partial pancreatectomy and sham surgery, implying that any differences in exposure of extrahepatic tissue to insulin oscillations would likely be modest. Furthermore, these would be embedded within the much higher insulin concentrations arising from the constant infusion of insulin in the systemic circulation during the clamp. One possible explanation is that there is a signal (metabolic, cytokine, or perhaps neural) from liver to muscle that is influenced by the pattern of insulin presented to the liver, although at this point this notion has to be purely speculative. Of course an association does not prove causality, and it is possible that the observed association simply reflects two independent consequences of the decreased β-cell mass in partially pancreatectomized dogs. The contribution of glucose toxicity toward partial pancreatectomy–induced insulin resistance would seem to be minimal because elevations in fasting plasma glucose levels were quite subtle (∼10% above sham surgery controls) compared with levels (∼200–300% above controls) used to demonstrate hyperglycemia-induced insulin resistance in vivo (31,32).

The partial recovery of fasting glucose concentrations in partially pancreatectomized dogs 4 weeks postoperatively implies some degree of β-cell regeneration or adaptive increased β-cell function after partial pancreatectomy in dogs, both of which have been documented in rodents (7,33,34). Presumably, there was only limited overall pancreas regeneration during the ∼6 weeks between partial pancreatectomy and euthanasia in the current studies because the surgeon’s estimate of the extent of pancreatic resection and the measured mass of pancreas remnant at euthanasia corresponded closely. Moreover, previous studies in dogs maintained for 12 months after ∼50% pancreatectomy showed progressive hyperglycemia developing into diabetes (11). The current studies were not designed to examine β-cell regeneration but were focused on the impact of an acute decrement in β-cell mass on insulin secretion. A limitation of studies addressing regeneration after partial pancreatectomy is the fact that the extent of pancreatic volume left in situ after partial pancreatectomy is only an estimate. This could perhaps be overcome by use of a sophisticated scanning approach (for example computed tomography or magnetic resonance imaging) pre- and postoperatively, but these techniques were not used in the current studies.

The near-normal fasting glucose concentration in partially pancreatectomized dogs by the time of the hyperinsulinemic-euglycemic clamp studies explains the fact that there was no measurable difference in endogenous hepatic glucose release in partially pancreatectomized dogs at baseline. Because the fasting glucose concentration is directly related to endogenous hepatic glucose release in the fasting state (35), we would have presumably observed increased basal hepatic glucose release when the dogs had higher fasting glucose concentrations, although this may have been appropriate for the decreased portal vein insulin concentrations in the fasting state. We observed impaired insulin-mediated glucose disposal with normal insulin-mediated suppression of hepatic glucose release in partially pancreatectomized dogs. Previous studies combining the hyperinsulinemic clamp technique and isotope dilution method to quantify hepatic glucose release have been confounded by errors attributed to non–steady state and a contaminant in commercially available titrated glucose preparations (15,36). Both of these were avoided in the current study by use of stable glucose isotopes and labeling of the glucose infusion with isotope to minimize non–steady-state errors. An important distinction between the reduction of β-cell mass accomplished by partial pancreatectomy and that which occurs in type 1 or type 2 diabetes is that α-cells are also removed in equal proportion. Partial pancreatectomy in dogs decreases glucagon secretion, and the glucagon secretory capacity in these animals correlates with the α-cell mass (9). Similar findings were reported in healthy living relatives of individuals with type 1 diabetes after hemipancreatectomy (37). A limitation of the current studies is that because of the large blood volumes required to characterize pulsatile insulin secretion, we did not have sufficient blood available to measure glucagon concentrations, but we can predict from prior published canine studies that glucagon secretion is decreased after partial pancreatectomy (9). It is therefore possible that with a β-cell deficit comparable to that observed in the current studies and an intact α-cell mass, dogs would have also had documented impaired insulin-mediated suppression of hepatic glucose release.

In the current study, a partial pancreatectomy led to diminished clearance of endogenously secreted insulin. We have previously reported that the amplitude of insulin pulses directed to the liver is related to hepatic insulin clearance, implying that the islet dictates the rate of systemic insulin delivery through the dual mechanisms of the insulin secretion rate and the pulsatile mode of insulin secretion (8,38–40). The current data support this hypothesis. Consistent with the current canine model of IGT/IFG, in humans with IGT (41) and type 2 diabetes (42), hepatic clearance of endogenously secreted insulin is diminished. One implication of decreased hepatic insulin clearance under conditions of attenuated insulin secretion is that systemic insulin concentrations are relatively preserved, despite partial β-cell failure. This adaptive change may have contributed to the formerly widely held impression that deficient insulin secretion is a late factor in the pathogenesis of type 2 diabetes. More importantly, it likely offers some protection from life-threatening hyperglycemia or hypoglycemia by buffering systemic insulin delivery over a broad range of insulin secretion.

In summary, ∼50% pancreatectomy in dogs resulted in IFG, IGT, and insulin resistance, metabolic abnormalities previously documented in humans with IFG/IGT (16–19). Moreover, the deficit in insulin secretion in partially pancreatectomized dogs was comparable both quantitatively and qualitatively to that in humans with type 2 diabetes, with a ∼70% deficit in insulin pulse mass (21). Interestingly, the deficit in pulse mass was related to both decreased hepatic insulin clearance and decreased insulin sensitivity. These data support an important role of decreased β-cell mass in the pathogenesis of IFG/IGT and subsequently type 2 diabetes in humans. These data support strategies that seek to prevent the decline in β-cell mass to prevent type 2 diabetes. The data also support the concept that defective β-cell function may play an important role in the impaired insulin sensitivity present in type 2 diabetes.

Funding for these studies was provided by National Institutes of Health Grant DK61539 and the Larry Hillblom Foundation.

We gratefully acknowledge the suggestions of our colleagues in the Larry Hillblom Islet Research Center, Anil Bhushan, Alexandra Butler, Leena Haataja, Kathrin Maedler, and Tatyana Gurlo.