Abnormal Glucose Handling by the Kidney in Response to Hypoglycemia in Type 1 Diabetes

We studied 12 patients with type 1 diabetes and 14 normal subjects. Among the patients with diabetes, six were women and six were men, with a mean (± SD) age of 31 ± 5 years. The duration of diabetes was 15 ± 4 years. All patients were taking insulin through either multiple daily injections (n = 9) or continuous subcutaneous delivery with the assistance of a pump (n = 3). The average (mean ± SD) total dose of insulin was 1.2 ± 0.5 units/kg daily, with a mean HbA_1c value of 8.0 ± 0.5% (normal value, <6.5). No patients had clinical evidence of autonomic neuropathy; severe hypoglycemia within the preceding 4 weeks; or a blood glucose concentration (ascertained by self-monitoring four to six times daily) <68 mg/dl (3.8 mmol/l) 3 days before the experiment. Patients were not receiving medication other than insulin, and none had acute illness or evidence of hepatic or renal dysfunction, hypertension, or vascular diseases. Among the normal subjects, eight were men and six were women, with a mean age of 33 ± 5 years. None had a history of diabetes in first-degree relatives or conditions that might affect glucose regulation or hormonal responses to hypoglycemia. Characteristics of all 24 individuals are summarized in Table 1. The study was approved by the University Hospital review board and the State University of New York–Stony Brook Human Investigation Committee, and all subjects gave written informed consent.

All subjects were admitted to the University Hospital General Clinical Research Center at 6:00 p.m. the evening before the study, and they consumed a standard meal; no food was allowed after dinner. For 3 days before the study, all subjects had abstained from alcohol and had been on a weight-maintaining diet containing at least 200 g of carbohydrate per day. Patients with diabetes had the intermediate- and long-acting insulin discontinued 24 h before the experiment and received only subcutaneous injections of regular insulin at breakfast, lunch, and dinner. At 12 a.m. (midnight), a continuous intravenous insulin infusion was started and adjusted to maintain blood glucose concentration between 80 and 120 mg/dl (4.4 and 6.6 mmol/l) according to a scale using hourly capillary blood glucose measurements. The insulin infusion rate was fixed 2 h before the commencement of the study each morning. All experiments were carried out between 7:00 a.m. and 3:00 p.m. after a 10- to 12-h overnight fast, and each subject was randomly assigned to a single experimental protocol.

To ensure a standardized hypoglycemic stimulus and to factor out the independent effects of hyperinsulinemia, one-step hypoglycemic- and euglycemic-clamp procedures analogous to those described by Simonson et al. (16) were performed in two separate experiments in normal subjects and diabetic patients alike. Two cannulas were inserted, one in each antecubital vein, for infusion of insulin, glucose, para-aminohippurate, and stable isotopes, and a third cannula was inserted in a dorsal hand vein for blood sampling. The hand was placed in a heated (65°C) box to arterialize the venous blood. Primed continuous infusions of [6,6-²H₂]glucose and [2-¹³C]glycerol (Cambridge Isotope Laboratories, Andover, MA), together with a continuous infusion of para-aminohippurate (Merck & Co, West Point, PA), were started in all subjects at the beginning of each experiment. During the 180-min equilibration period, the subjects were transported to the radiology suite and had the left renal vein catheterized under fluoroscopy through the right femoral vein using local anesthesia (lidocaine 2%). The position of the catheter tip was ascertained by injecting iodinated contrast material, and the catheter was then infused continuously with heparinized saline solution (4.0 units/min).

After baseline determinations, all subjects were given continuous intravenous infusions of insulin (0.5 mU · kg⁻¹ · min⁻¹), and target plasma glucose values were achieved by varying the rate of an infusion of 10% glucose in water with added [6,6-²H₂]glucose to keep plasma glucose isotope enrichments constant, as described by Butler et al. (17). Plasma glucose was measured at the bedside at 10-min intervals with a glucose analyzer (Beckman, Fullerton, CA). Plasma glucose concentrations were stabilized between 100 and 140 mg/dl (5.5 and 7.7 mmol/l) in patients with diabetes, and between 90 and 105 mg/dl (5.0 and 5.8 mmol per liter) in normal subjects during 40 min (baseline) before the induction of either hypoglycemia or euglycemia with hyperinsulinemia. The plasma glucose concentration was reduced to ∼55 mg/dl (∼3.0 mmol/l) over a period of 180 min by combining insulin infusion with a reduction in the rate of exogenous glucose infusion. Plasma glucose concentrations were then maintained at that level for a further 40 min (hypoglycemia phase). In the euglycemic clamp studies, plasma glucose concentrations between 90 and 120 mg/dl (5.0 and 6.6 mmol/l) were achieved over a period of 180 min by combining insulin infusion with exogenous glucose infusion. Plasma glucose concentrations were then maintained at that level for a further 40 min (euglycemia phase). During all studies, vital signs, cognitive functions, and peripheral pulses were monitored frequently; all subjects were resting quietly in bed with free access to water but not to food. Normal saline (0.9% sodium chloride) was given at the rate of 75 ml/h to all subjects for the duration of the experiment.

Blood samples were taken simultaneously from the arterialized hand vein and the left renal vein at 10-min intervals during the 40-min baseline euglycemic period and again during the final 40 min of the hyperinsulinemic clamp (either hypoglycemia or euglycemia phase of the study period) for measurements of blood glucose and glycerol concentrations and isotope enrichments; plasma para-aminohippurate, free fatty acids, and insulin; and C-peptide, glucagon, and catecholamines. Renal plasma flow was estimated by para-aminohippurate clearance rates and corrected to renal blood flow by the (1-hematocrit) factor. Net renal glucose balance was determined using arterial-renal vein blood glucose concentration differences. Calculations of the rates of systemic renal glucose production and utilization and of glycerol conversion to glucose were based on systemic and renal dilution and fractional extraction of isotopes according to previously described methods (7).

In all subjects, symptoms of hypoglycemia were assessed at 15-min intervals, and the subjects were asked to rate if these were mild, moderate, or severe. The symptoms were pounding heart, shakiness, sweating, headache, difficulty in thinking, or slowed thinking. Heart rates were measured continuously and averaged over 30-s intervals; blood pressure and peripheral pulses in lower extremities were examined at 30-min intervals. Urine samples for measurements of glucose were obtained at the beginning and at the end of each study in all patients with diabetes.

Blood glucose and glycerol concentrations were measured by enzymatic assays (18), and isotope enrichments were measured by gas chromatography–mass spectrometry, as previously described (9). Plasma para-aminohippurate was measured by a colorimetric assay (19), and plasma insulin (20), C-peptide (21), and glucagon (22) were measured by double-antibody radioimmunoassays. Plasma free fatty acids (after alcohol-heptane extraction [23]) and catecholamines (24) were measured by high-performance liquid chromatography. All of the samples from each subject were analyzed in a single assay. The intra-assay variations for blood glucose and glycerol assays were <5%; for isotope enrichment, determinations were <3%. The intra-assay variations for para-aminohippurate assay and for insulin, C-peptide, and glucagon assays were <10%, and for free fatty acids, epinephrine, and norepinephrine, they were 15, 17, and 12%, respectively.

The blood glucose concentrations, renal glucose balance (net uptake or net output), systemic and renal glucose and glycerol kinetics, and hormone responses in all four groups of individuals studied were compared by analysis of variance with repeated-measures design when appropriate. Baseline values for each parameter were defined as the means of all five values obtained during the 40-min interval preceding the hyperinsulinemic clamp period. Hypoglycemia- and euglycemia-phase values for the same parameters were defined as the means of all five values obtained during the final 40-min interval of the hypoglycemic or euglycemic period, respectively. All statistical tests were two-sided.

The mean blood glucose concentrations during the hypoglycemic- and euglycemic-clamp procedures were comparable in the patients with diabetes and the normal subjects, and during each study the target blood glucose of ∼54 mg/dl (∼3.0 mmol/l) was reached (Figs. 1A and B). The rates of endogenous (whole-body) glucose production were equally decreased during the euglycemic clamp procedures, though the compensatory rise in endogenous glucose production during the hypoglycemic clamp procedures in normal subjects (Fig. 1D) was not observed in patients with diabetes (Fig. 1C).

The mean (± SE) renal glucose balance changed from output to net uptake during the euglycemic clamp procedures and to a higher net output during the hypoglycemic clamp procedures (from –0.8 ± 0.4 to 1.4 ± 0.5 and to −4.5 ± 1.3 μmol · kg⁻¹ · min⁻¹, respectively). However, in patients with diabetes, the small renal glucose output changed to an uptake during the euglycemic clamp procedures (from –0.3 ± 0.5 to 4.5 ± 0.7 μmol · kg⁻¹ · min⁻¹), which was greater than in control subjects (P < 0.05). Also, the renal glucose output during comparable hypoglycemic clamp procedures was small and not statistically significant (–0.4 ± 0.3 μmol · kg⁻¹ · min⁻¹) (Fig. 2). After insulin infusion, changes in net renal glucose balance in the euglycemic periods were due to concomitant increases in renal glucose utilization and decreases in renal glucose production documented both in normal subjects and in patients with diabetes. During hypoglycemia, however, the twofold increase in renal glucose production rates, which contributed to net renal glucose output in normal subjects, was not seen in patients with diabetes (Table 2).

The mean plasma free fatty acid, blood glycerol concentration, and whole-body glycerol turnover rates—all indexes of adipose tissue lipolysis—decreased significantly in all euglycemic clamp procedures and increased significantly during all hypoglycemic clamp procedures. Although the increments were consistently greater in normal subjects than in patients with diabetes, differences between groups were not statistically significant (Table 3).

In the baseline, rates of whole-body glycerol conversion to glucose, an index of systemic gluconeogenesis, and rates of renal glycerol conversion to glucose, an index of renal gluconeogenesis, were twofold higher in patients with diabetes (P = 0.02), though these showed equal percent decreases during the euglycemic clamp. The increments in systemic and renal glycerol gluconeogenesis during the hypoglycemic clamp procedures were attenuated in patients with diabetes when compared with normal subjects, but these differences did not reach statistical significance.

Elevations in the plasma insulin concentrations were comparable during all clamp procedures. The plasma C-peptide levels were much lower in the patients with diabetes than in normal subjects in the baseline, consistent with the diagnosis of type 1 diabetes. As expected, there was a substantial decline in the plasma C-peptide levels in the normal subjects after insulin administration in all clamp procedures (Table 4). Also, in the patients with diabetes, the plasma glucagon concentrations were lower than in the normal subjects and remained low during the hypoglycemic clamp procedures, whereas in the normal subjects, the plasma glucagon concentrations increased by ∼50% (Table 4).

The plasma epinephrine and norepinephrine concentrations did not change significantly during euglycemic hyperinsulinemia in either group during any of the studies (Table 4). In the normal subjects, the plasma epinephrine concentrations increased approximately eightfold during hypoglycemia, and the plasma norepinephrine concentrations nearly doubled, whereas during comparable hypoglycemia in the patients with diabetes, the plasma epinephrine and norepinephrine concentrations increased by only threefold and 50%, respectively (Table 4). In the patients with diabetes, markedly blunted catecholamine responses to hypoglycemia were also documented across the kidney by measuring renal norepinephrine spillover, an index of renal autonomic sympathetic activity (Fig. 3). In the normal subjects, the mean renal plasma norepinephrine balance increased during hypoglycemia, from an output of 0.23 ± 0.07 to 1.37 ± 0.29 μg · kg⁻¹ · min⁻¹ (from 1.36 ± 0.42 to 8.10 ± 1.71 pmol · kg⁻¹ · min⁻¹), but during equivalent hypoglycemia in the patients with diabetes, the renal norepinephrine balance only increased from 0.16 ± 0.08 to 0.65 ± 0.31 μg · kg⁻¹ · min⁻¹ (from 0.95 ± 0.47 to 3.84 ± 1.83 pmol · kg⁻¹ · min⁻¹, P = 0.02, for the comparison of the changes from baseline between groups). The net renal norepinephrine balance did not change during euglycemic hyperinsulinemia in either group during any of the studies.

Both the patients with diabetes and the normal subjects developed mild symptoms of hypoglycemia during the hypoglycemic clamp procedures (data not shown). The mean heart rates did not change significantly during the periods of hypoglycemia in either the patients with diabetes (at baseline, 82 ± 6 beats/min; during the final 40 min of hypoglycemia, 92 ± 7 beats/min) or the normal subjects (at baseline, 80 ± 5 beats/min; during the final 40 min of hypoglycemia, 88 ± 6 beats/min).

We found that adrenergic stimulation of renal glucose output in response to hypoglycemia is impaired and contributes to defective counterregulation in patients with long-standing type 1 diabetes and no clinical evidence of classical autonomic neuropathy. This impairment would be expected to make them more susceptible to severe and prolonged episodes of hypoglycemia. Because patients with type 1 diabetes cannot dissipate exogenously administered insulin and lack the glucagon secretion necessary to rapidly reverse the insulin-suppressed hepatic glucose production, activation of the autonomic nervous system becomes their main defense against hypoglycemia (25). As a result, the attenuation and correction of the fall in blood glucose concentration depend heavily on adrenergic stimulation of gluconeogenesis, a process shared by the liver and kidney (8,25). Although epinephrine also stimulates hepatic glycogen degradation, gluconeogenesis is the predominant source of endogenous glucose production during prolonged hypoglycemia (25). Recurrent episodes of hypoglycemia have been shown to reduce autonomic and symptomatic responses to subsequent hypoglycemia (15,16), thus increasing the risk of iatrogenic and asymptomatic hypoglycemia. Considering that our patients did not report any episodes of hypoglycemia requiring either hospitalization or the assistance of others during the 4-week period preceding the study, and that there were no capillary blood glucose measurements <3.8 mmol/l (68 mg/dl) 3 days before the experiment, the subnormal adrenergic responses observed in our studies are consistent with the concept of “elevated” glycemic threshold for epinephrine secretion, i.e., lower blood glucose concentrations are required to elicit autonomic and symptomatic responses to hypoglycemia (15,16,25). This threshold abnormality is characteristically seen in patients with type 1 diabetes with >10 years’ duration, and it is particularly common in those who are C-peptide–negative and have adequate HbA_1c levels. The susceptibility to recurrent hypoglycemia is believed to be secondary to undetected episodes of nocturnal hypoglycemia, which occur frequently during intensive insulin therapy (26,27,28).

Interest in the potential role of the kidney in the regulation of glucose metabolism has been revived since our original report (7). In humans, postabsorptive renal glucose production and utilization rates have been estimated in the range of 5–28% of whole-body glucose turnover by various groups of investigators (8,9,10). A more recent study has concluded that this variability is partially due to methodological noise and that the renal contribution to endogenous glucose production lies between 4 and 18% (29). In our experience, with studies performed in a heterogeneous group of healthy male and female volunteers (n = 42) combining left renal vein catheterization and arterialized hand vein sampling with infusion of [6,6-²H₂]glucose, postabsorptive renal glucose production averages 1.7 μmol · kg⁻¹ · min⁻¹ (∼14% of endogenous glucose production), with a standard deviation of 4.16 and a coefficient of variation of 2.44. These results are in close agreement with those obtained by others (8,9,10,29) and further support the notion that because quantitation of renal glucose flux is technically difficult and variable, highly sensitive methods and a large number of subjects must be used in studies designed to evaluate renal glucose production and utilization rates. These data also confirm that the contribution of the kidney to glucose appearance rates in postabsorptive humans is negligible compared with that of the liver, although observations that renal glucose production increases substantially with prolonged fasting (6), in type 2 diabetes (30), in insulin-induced hypoglycemia (11,12), and during the anhepatic phase of liver transplantation (31) strongly suggest that the kidney plays a more important role in glucose homeostasis than previously appreciated.

There is little or no controversy regarding the fact that the liver is the primary site of glucose production in most circumstances. In this study, for example, the relative contribution of the kidney to endogenous glucose production during hypoglycemia in normal subjects did not exceed 35% (i.e., 5.5 of 15.0 μmol · kg⁻¹ · min⁻¹), and net glucose output increased to a maximum of 4.5 μmol · kg⁻¹ · min⁻¹. Hence, the inadequate compensatory rise in renal glucose production in diabetic patients represents a small abnormality in comparison with that of the liver. Hepatic glucose production, which accounted for ∼65% (i.e., 9.5 of 15.0 μmol · kg⁻¹ · min⁻¹) of endogenous glucose production during hypoglycemia in normal subjects, was responsible for nearly 100% (∼6.0 μmol · kg⁻¹ · min⁻¹) of glucose appearing in the circulation in diabetic patients under a comparable degree of hypoglycemia. Therefore, our studies uncover an abnormality in glucose handling by the kidney, in addition to that of the liver, in response to hypoglycemia and further reaffirm that the liver represents the predominant organ responsible for the compensatory rise in endogenous glucose production during hypoglycemia in both normal subjects and patients with diabetes. In theory, however, the body’s ability to prevent and correct hypoglycemia without the full contribution of the kidney is incomplete and sufficient to impair normal blood glucose recovery.

The findings that plasma glucagon remained essentially unchanged, and that elevations in both plasma catecholamines and renal norepinephrine overflow were markedly attenuated during sustained hypoglycemia, indicate that the lack of compensatory rise in endogenous (hepatic and renal) glucose production observed in patients with diabetes was secondary to insufficient hormonal stimulation of glucose production. Specifically, our data demonstrate that both adrenomedullary (circulating epinephrine) and sympathetic neural (renal norepinephrine overflow) responses to a given level of hypoglycemia are reduced early in these patients. These results are in agreement with those reported by others (32,33) and further suggest that insufficient stimulation of renal, in addition to hepatic, gluconeogenesis due to reduced autonomic responses provides incomplete protection against hypoglycemia in patients with type 1 diabetes. The extent to which the lack of glucose release into the circulation by the kidney in these experimental conditions of mild-to-moderate hypoglycemia is the result of limited substrate supply remains unclear. Nonetheless, these abnormalities are analogous to other defects in the glucose counterregulatory responses (34), and with no evidence of intrinsic renal damage, they may be reversed by scrupulous avoidance of hypoglycemia.

Demonstration that net glucose output by the kidney, which represents an additional physiological mechanism in the defense against hypoglycemia, is impaired in patients with diabetes is of considerable significance. Hypoglycemia is the limiting factor and the most feared complication in the management of diabetes (2), and it is associated with substantial morbidity, particularly in patients on intensive insulin therapy (1). To significantly reduce the occurrence of hypoglycemia in these patients, we must learn how to avoid and compensate for the compromised glucose counterregulatory responses. Our findings indicate that protection against the fall in blood glucose concentration in patients with diabetes is largely accomplished by the activation of the autonomic nervous system and relies partly on stimulation of renal gluconeogenesis. This may represent an adaptation to the loss of glucagon secretion during hypoglycemia. Therefore, besides improving our ability to deliver more physiological insulin replacement and implementing strategies that aim to reduce the conventional risk factors (34,35), preservation of renal blood flow and function may be critical in restoring normal glucose counterregulation in patients with type 1 diabetes.

This work was supported by grants from the National Institutes of Health (DK-49861 and RR-10710), the American Diabetes Association, and the Juvenile Diabetes Foundation International.

We are indebted to Professor Geraldo Chini for his insightful comments and suggestions.