Introduction

Although hypoglycaemia is frequently observed in Type 1 diabetic patients, our knowledge of human brain energy metabolism during hypoglycaemia is limited in these patients. Animal studies have revealed that during severe hypoglycaemia, the concentration of key metabolites such as phosphocreatine (PCr) and ATP decrease [1]. In contrast, moderate hypoglycaemia as encountered frequently in Type 1 diabetic patients does not change intracerebral ATP or PCr concentrations, at least in non-diabetic subjects [2]. A short-term decrease of blood glucose down to approximately 1.4 mmol/l does not affect cerebral metabolism of phosphate compounds and intracellular pH (pHi) [2]. The effect of hypoglycaemia on brain energy metabolism has not yet been reported for Type 1 diabetic patients.

By combining non-invasive in vivo 31P nuclear magnetic resonance (NMR) spectroscopy with hyperinsulinaemic hypoglycaemic clamp tests, the present study aims to examine cerebral energy metabolism during moderate, insulin-induced hypoglycaemia, and to compare the results of Type 1 diabetic patients with those of healthy subjects.

Subjects and methods

Subjects

Male Type 1 diabetic patients (n=6; age 28±2 years; BMI 22.6±0.7 kg/m2; mean diabetes duration 13±2.5 years) with good metabolic control (HbA1c 6.8±0.3%), and healthy subjects matched for age and BMI (control; n=6; age 26±1 years; BMI 24.2±0.7 kg/m2; HbA1c 5.0±0.1%) were studied. Full written informed consent was obtained prior to each study and the study protocol was approved by the local ethics committee. The Type 1 diabetic patients were undertaking normoglycaemic (“functional”) insulin therapy, which relies on the separate substitution of prandial and basal insulin with multiple daily insulin injections. Patients were instructed to avoid hypoglycaemia (blood glucose <3.4 mmol/l) by meticulous adherence to their pre-set algorithms and by avoiding unusual strenuous exercise. By reviewing the blood glucose profiles (at least five measurements per day) it was ensured that the patients were only included if they had no history of severe hypoglycaemic episodes for at least 4 weeks before the study. None of the patients had experienced hypoglycaemic coma.

The diabetic patients were instructed to omit NPH or Zn insulin and to correct their blood glucose with regular insulin for 24 hours before the study. At 00.00 hours on the study day, the patients were admitted to the metabolic ward, and catheters were inserted into antecubital veins of both arms for blood sampling and glucose/insulin infusions. During the night, plasma glucose concentrations were kept between 4.4 and 6.7 mmol/l with a variable intravenous insulin infusion. Healthy subjects were admitted at 08.00 hours on the study day, and catheters were inserted as described above. At 08.45 hours (0 min), a hyperinsulinaemic (1.5 mU·kg−1·min−1) hypoglycaemic (~3.0 mmol/l) clamp test was started. At 10.45 hours (120 min), the insulin infusion was stopped and glucose was infused to increase plasma glucose to above 5.6 mmol/l and to maintain this concentration for the following 120 min. Plasma glucose was monitored every 5 minutes between 0 and 240 min. 31P NMR spectroscopy of the brain was performed before the clamp test (−45 to 0 min), during hypoglycaemia (80 to 120 min) and during the subsequent recovery period (200 to 240 min). Blood samples for measurement of hormones and metabolites were drawn before and after each NMR spectroscopic measurement.

Analytical methods

Plasma glucose concentrations were measured using glucose oxidase (Glucose Analyzer II; Beckman, Fullerton, Calif., USA). HbA1c, insulin, C-peptide, glucagon, growth hormone, cortisol and epinephrine were quantified as described previously [3].

In vivo 31P NMR spectroscopy

For NMR spectroscopy, patients lay in a 3-Tesla spectrometer (80 cm Medspec-DBX; Bruker Medical, Ettlingen, Germany) in the supine position, with their heads resting on a surface coil 10 cm in diameter. 31P NMR spectra were obtained by a pulse-and-acquire method (repetition time 5 seconds) from the occipital lobe of the brain, since this location allowed comfortable placement of the coil. To localise the signal from the region of interest and suppress the signal coming from superficial tissues and the skull, the flip angle of 250-µs rectangular excitation pulses was set to about 180° in the coil plane. Peak positions and intensities were read manually from the spectra using the software supplied by Bruker. The chemical shift of the peak of intracellular inorganic phosphate (Pi) relative to PCr was used to calculate pHi, while intracellular magnesium ([Mg2+]i) was calculated from the difference of chemical shifts of the α-ATP and β-ATP peaks [4].

Statistical analysis

All data are presented as means ± SD. Paired t tests or repeated measurement ANOVA followed by post hoc testing (Bonferroni’s t test) were used for statistical comparisons within different groups. Statistical comparisons between control subjects and Type 1 diabetic patients were performed using the unpaired Student’s t test. Differences were considered statistically significant at a p value of less than 0.05.

Results

At baseline, plasma concentrations of glucagon, cortisol, growth hormone and epinephrine (Table 1) were not different in Type 1 diabetic patients from those in healthy subjects. During the hyperinsulinaemic hypoglycaemic clamp test, plasma glucose decreased to approximately 3.0 mmol/l within 40 min and there was no difference between the groups (mean glucose concentration 30–120 min: control 3.1±0.02 vs diabetic patients 2.9±0.09 mmol/l; NS). While plasma insulin concentrations were higher in diabetic patients at baseline (p<0.01), no difference was observed during and after hypoglycaemia. During hypoglycaemia, the peak concentrations of counter-regulatory hormones were markedly reduced in diabetic patients compared to those in healthy controls (p<0.001 for all hormones).

Table 1 Plasma concentrations of glucose and counter-regulatory hormones, and intracerebral parameters before, during and after insulin-induced hypoglycaemia in six healthy and six Type 1 diabetic volunteers
Full size table

The 31P NMR spectroscopy revealed a decreased PCr:γ-ATP ratio at baseline (p<0.01; Fig. 1) but normal pHi and [Mg2+]i in diabetic patients. During hypoglycaemia, pHi increased in both groups (p<0.05) and was higher in diabetic patients (p<0.05), while [Mg2+]i decreased in both groups (p<0.05), there being no difference between the two values. The PCr:γ-ATP ratio remained lower in diabetic patients throughout the experiment (Table 1).

Fig. 1
figure 1

A typical 31P spectrum measured in the occipital lobe of a volunteer. The signal of phosphocreatine is taken as the chemical shift reference (0 parts per million [ppm]). Pi, intracellular inorganic phosphate

Full size image

Discussion

We report for the first time the 31P NMR spectroscopy measurement of energy-rich phosphates, pHi and [Mg2+]i during hypoglycaemia in the occipital lobe of Type 1 diabetic patients and a matched control group.

The major finding of our study is a lower PCr:γ-ATP ratio in diabetic patients, which we observed throughout the experiment. Chronically decreased PCr:γ-ATP ratios have been previously found in patients suffering from brain tumours, migraine syndromes and mitochondrial dysfunction [5]. A possible explanation for reduced PCr:γ-ATP ratios is therefore a diabetes-induced reduction of creatine kinase activity [6] or mitochondrial transcription [7]. Previous studies using 1H NMR spectroscopy have revealed a reduced N-acetyl:creatine ratio and an increase in myo-inositol and choline metabolites in Type 1 diabetic patients [8]. Taken together, these studies support the contention that brain metabolism could be chronically altered in patients with Type 1 diabetes.

In our study, the PCr:γ-ATP ratio remained unchanged in both healthy volunteers and Type 1 diabetic patients. Our findings are consistent with a previous study reporting no changes in energy-rich phosphate compounds during short-term hypoglycaemia in healthy subjects [2]. PCr serves as a rapidly available energy store, and the PCr:γ-ATP ratio is acutely reduced during metabolic insults such as severe hypoglycaemia or hypoxia [1]. The stable PCr:γ-ATP ratio reported could be due to the relatively mild degree of hypoglycaemia feasible in human experimental research, since animal studies as well as in vitro experiments show that plasma glucose has to be lowered to well below 1.4 mmol/l to disrupt brain energy metabolism and induce a decrease in PCr [1, 9]. However, we did find at least indirect evidence to suggest that subtle changes in energy metabolism take place during hypoglycaemia as outlined below.

Firstly, [Mg2+]i decreased in both groups. Such a decrease has been previously described in patients with defective mitochondrial respiration or ischaemic brain injury. It was proposed that reduced [Mg2+]i could be a physiological response to impaired energy metabolism, since the amount of free energy released by ATP hydrolysis is increased and the equilibrium constant of the creatine kinase reaction is reduced [10].

Secondly, we observed an increase in pHi during hypoglycaemia. While both increased and decreased pHi have been reported in animal studies, no change in pHi has been observed in healthy subjects during hypoglycaemia [2]. Performing NMR spectroscopy at 3 Tesla enabled us to obtain spectra with higher resolution and to detect a small but statistically significant increase. With regard to energy metabolism, an increase in pHi would help to facilitate glycolysis, since protons are powerful inhibitors of phosphofructokinase [11].

The limitation of our study is that non-localised 31P NMR spectroscopy does not allow absolute quantification of metabolites, and therefore inter-individual comparison of the concentrations of phosphorus-containing metabolites is not possible. However, the ratios of metabolites can be compared, and the time course can be followed under dynamic conditions, as done in this study. Thus, we were able to demonstrate alterations of energy-rich phosphate ratios in Type 1 diabetic patients compared with matched controls.

To ensure similar plasma glucose concentrations in the two groups at the beginning of the clamp test, we applied an overnight insulin infusion to the diabetic patients. Therefore, plasma insulin concentrations were higher in these patients at baseline, but similar during and after hypoglycaemia. An acute effect on the observed pattern of energy-rich compounds seems unlikely, since insulin substitution always results in peripheral hyperinsulinaemia, and since brain glucose uptake is independent of plasma insulin concentrations.

In conclusion, we found decreased cerebral PCr:γ-ATP ratios in Type 1 diabetic patients at baseline. Clinical hypoglycaemia did not affect the ratio of these energy-rich phosphates in non-diabetic or diabetic subjects. However, changes observed in pHi and [Mg2+]i do suggest subtle changes in energy metabolism.