Human Brain Glycogen Metabolism During and After Hypoglycemia

Five healthy volunteers (four men and one woman age 43 ± 13 years, BMI 25 ± 3 kg/m², means ± SD) on no medications participated in a paired experiment after giving informed consent using procedures (Fig. 1) approved by the University of Minnesota Institutional Review Board. Subjects were studied on two separate occasions separated by at least 1 week, with each subject serving as their own control. Subjects reported to the General Clinical Research Center (GCRC) at 6:00 p.m. after starting a fast at 2:00 p.m. Intravenous catheters were placed antegrade in contralateral arms for [¹³C]glucose infusion and blood sampling. At 7:00 p.m., a bolus of [1-¹³C]glucose was given to rapidly raise blood glucose enrichment. A continuous infusion was then given at a variable rate to maintain blood glucose 25% above basal levels to minimize endogenous hepatic glucose production and achieve stable glucose enrichments, because postprandial insulin levels are known to suppress hepatic glucose output (28). Blood glucose was measured on an automatic glucose meter (OneTouch SureStep; Lifescan, Milpitas, CA). Additional samples were collected hourly and immediately frozen for the later determination of isotopic enrichment of plasma glucose by gas chromatography–mass spectroscopy (GC-MS) as described previously (15). A total of 186 g of [1-¹³C]glucose (Isotec, Miamisburg, OH, and Cambridge Isotope Laboratories, Andover, MA; prepared as 20% weight/volume d-glucose in water with 99% isotopic enrichment) was administered for 11.1 ± 1.2 h to prelabel glycogen. Subjects were then given an unlabeled glucose infusion for 1.6 ± 0.6 h to wash-out [¹³C]glucose from blood so that [1-¹³C]glucose removed from glycogen would not be replenished by plasma [1-¹³C]glucose.

After the wash-out period, subjects were transferred to the Center for Magnetic Resonance Research (CMRR) for a baseline scan, which was omitted in one subject. They were prepared for a 2-h hyperinsulinemic clamp study by retrograde placement of a third intravenous catheter into a foot to provide venous access for arterialized blood sampling (29) while the subject was in the magnet. The two-arm intravenous catheters were used for administration of glucose, insulin (2 mU · kg⁻¹ · min⁻¹) and potassium phosphate (4 mEq/h). Glucose (20% dextrose in water) was titrated to achieve target blood glucose of 45 mg/dl (2.5 mmol/l) on one occasion and 95 mg/dl (5.3 mmol/l) on the other occasion, in random order. Blood was obtained every 5 min for immediate measurement of glucose using an autoanalyzer (Analox Instruments, Lunenburg, MA). During the hypoglycemic clamps, blood was also sampled every 30 min for the later determination of glucagon, catecholamines, growth hormone, and cortisol. Four [1-¹³C]glycogen NMR spectra were acquired using methods described below while blood glucose levels were clamped at the target levels, starting 4.9 ± 1.3 h after the end of the [1-¹³C]glucose infusion. Subjects were then removed from the magnet, the insulin and potassium infusions were stopped, glucose was administered to bring the blood glucose to 95 mg/dl, and subjects were fed a regular meal. Additional spectroscopic measurements of [¹³C]glycogen were obtained at ∼23, 28, 38, and 46 h after the start of the [¹³C]glucose infusion.

Five healthy volunteers (four men and one woman, age 43 ± 14 years, BMI 26 ± 4 kg/m², means ± SD) on no medications participated in a paired experiment after giving informed consent using procedures (Fig. 2) approved by the University of Minnesota Institutional Review Board. Subjects were studied on two occasions separated by at least 1 month. Subjects reported at 7:00 a.m. to the GCRC in the fasting state. Two intravenous catheters were placed antegrade in contralateral arms for administration of glucose, insulin, and KPhos. Each volunteer underwent a 2-h hyperinsulinemic-euglycemic clamp on one occasion and a hyperinsulinemic-hypoglycemic clamp on the other, in random order, as described above. After the clamp, the insulin and potassium infusions were stopped and glucose was administered to bring the blood glucose to 95 mg/dl. Thirty minutes after the end of the clamp, [1-¹³C]glucose infusion was started with an initial bolus to rapidly raise blood glucose enrichment. The two-arm intravenous lines were used for [¹³C]glucose infusion and blood sampling. The infusion rate was adjusted to maintain blood glucose 25% above basal levels. A total of 820–1,420 g of [1-¹³C]glucose (25–50% isotopic enrichment) was administered over 22–54 h. During the infusion, additional blood samples were collected and frozen for the later determination of isotopic enrichment of plasma glucose by GC-MS. Subjects were transferred to the CMRR to acquire [1-¹³C]glycogen NMR spectra at ∼4, 8, 12, 22, 29, 35, 46, 53, 60, 70, and 80 h after the start of the [1-¹³C]glucose infusion. During the infusion, subjects were fed isocaloric, low-carbohydrate meals designed to minimize the impact of dietary carbohydrate on [¹³C]glucose enrichment.

All measurements were performed on a 4-tesla, 90-cm bore magnet (Oxford Magnet Technology, Oxford, U.K.) with an INOVA console (Varian, Palo Alto, CA). Subjects were positioned supine on the patient bed with the occipital lobe just above the ¹H/¹³C surface coil (30). Subjects wore earplugs to minimize exposure to gradient noise and were positioned in the coil holder using cushions to minimize head movement.

The [1-¹³C]glycogen NMR signal localized in a 7 × 5 × 6 cm³ voxel in the occipital lobe was acquired as described previously (4,9,10). Each spectrum/data point presented here was acquired over 30 min. The amount of ¹³C label in the C1 position of glycogen was quantified by the external reference method (9,10). The [1-¹³C]glycogen concentrations were divided by the plasma glucose isotopic enrichment to correct for differences in isotopic enrichment between subjects and to determine the newly synthesized glycogen concentrations.

A model of glycogen metabolism (4) was fitted to the time courses of [¹³C]glycogen using the software SAAM II (The SAAM Institute, Seattle, WA). Data from the euglycemic clamp studies of each subject were used for modeling, together with their blood glucose isotopic enrichment time courses as input function. Glycogen synthase (V_syn) and phosphorylase (V_phos) rates were set to be equal and brain glycogen concentration was set to be constant. Thus, the fitted variables were total glycogen concentration (Glyc) and turnover rate V_syn = V_phos. The cerebral metabolic rate of glucose (CMR_glc) in the human brain was assumed to be 0.4 μmol · g⁻¹ · min⁻¹ = 24 μmol · g⁻¹ · h⁻¹ (15) and the glucose-6-phosphate concentration 0.1 μmol/g (14). Sensitivity analysis indicated that the results were not affected over large ranges of both of these variables (18–30 μmol · g⁻¹ · h⁻¹ for CMR_glc and up to 1 μmol/g for glucose-6-phosphate concentration). Concentration and rate estimates are reported as means ± SD.

In the utilization study, summary statistics and paired t tests were used to compare within-subject differences in plasma glucose, glucose enrichment, and glycogen on hypoglycemic versus euglycemic study days, at each baseline and during clamps. Repeated measures ANOVA was used to compare euglycemic measures to hypoglycemic measures of glycogen, for both within-clamp measures and after-clamp measures. In the supercompensation study, summary statistics and paired t tests were used to compare within-subject differences in plasma glucose, plasma insulin, and the 4-h glycogen measurement on hypoglycemic versus euglycemic study days.

Brain glycogen of five healthy volunteers was prelabeled via an overnight intravenous infusion of [1-¹³C]glucose before a euglycemic or hypoglycemic clamp study in the scanner (Fig. 1). Average plasma glucose levels and isotopic enrichments during the overnight infusion were not significantly different for the euglycemia versus hypoglycemia study days, leading to equal glycogen prelabeling before both clamp studies (Table 1). Target levels for blood glucose during the clamps were reached in ∼30 min after starting the insulin infusion (Fig. 1). Average blood glucose concentration during the hypoglycemic clamps was 57.2 ± 9.7 mg/dl and during the euglycemic clamps 95.1 ± 3.3 mg/dl. The hypoglycemic glucose concentration was slightly above our target level mainly because of one subject who did not require glucose infusion and stayed above 60 mg/dl during the clamp period. Average blood glucose during the hypoglycemic clamps of the other four subjects was 53.3 ± 4.7 mg/dl. Counterregulation during hypoglycemia was demonstrated by measurement of serum glucagon, growth hormone, cortisol, and catecholamines (Fig. 3). The residual plasma glucose enrichment after the chase with [¹²C]glucose (Fig. 1) during the hypoglycemic clamps tended to be higher than that during the euglycemic clamps (22 ± 14 vs. 11 ± 4%, P = 0.13, paired t test) likely because of mobilization of [¹³C]-labeled hepatic glycogen during hypoglycemia. However, considering the approximately twofold higher glucose concentrations during the euglycemic clamps, the level of cerebral [¹³C]glucose available for incorporation into glycogen was equal between the hypoglycemic and euglycemic clamps, which was also apparent from the residual glucose peaks in the spectra (Fig. 4 A).

The glycogen signal was stable during the euglycemic clamps, while it decreased during the hypoglycemic clamps (Fig. 4,A), indicating mobilization of glucose units from glycogen during moderate hypoglycemia. Average glycogen integrals, each normalized to first clamp measure, during euglycemia were higher than during hypoglycemia (0.98 ± 0.05 vs. 0.87 ± 0.08, P < 0.0001, repeated measures ANOVA) (Fig. 4 B). Glycogen utilization was confirmed by a higher rate at which newly synthesized glycogen levels decreased during hypoglycemia (0.12 ± 0.05 μmol · g⁻¹ · h⁻¹) than during euglycemia (0.03 ± 0.06 μmol · g⁻¹ · h⁻¹, P < 0.02, paired t test). This label wash-out rate during euglycemia was the same as we previously observed after an 11-h prelabeling period (4). To further analyze the consistency between our previous observations during euglycemia and slight hyperglycemia and this study, we fitted a model of glycogen turnover to the time courses of [¹³C]glycogen obtained during the euglycemia studies of the five volunteers. This resulted in estimates of glycogen content of 4.3 ± 0.2 μmol/g and turnover rate (V_syn = V_phos) of 0.18 ± 0.01 μmol · g⁻¹ · h⁻¹, indicating a turnover time constant of 24 h, in excellent agreement with prior results (4).

Newly synthesized glycogen levels after the clamp (at 23, 28, 38, and 46 h time points) were not different for euglycemia versus hypoglycemia studies (P = 0.64, repeated measures ANOVA). Note that the “newly synthesized glycogen” levels refer to measured [¹³C]glycogen levels divided by the isotopic enrichment of plasma glucose during the preclamp infusions; therefore, they do not necessarily reflect new glycogen synthesized after the clamps. Based on the lower [¹³C]glycogen levels at the end of the hypoglycemic clamps, one might expect the glycogen measurements after hypoglycemia to also be lower than those after euglycemia. However, the average plasma glucose isotopic enrichment at the end of the hypoglycemic clamps was 13 ± 11% (as opposed to 2 ± 1% at the end of the euglycemic clamps) and this enriched glucose could have been incorporated into glycogen once blood glucose levels were rescued, thereby equalizing [¹³C]glycogen levels in the following scans. This effect would have been augmented by glycogen supercompensation after hypoglycemia. Therefore, we investigated if glycogen synthesis was increased after hypoglycemia in the next set of experiments.

In this experiment, label incorporation from intravenous [1-¹³C]glucose into brain glycogen was measured after a euglycemic or hypoglycemic clamp (Fig. 2). Target levels for blood glucose during the clamps were reached in 40–60 min after starting the insulin infusion (Fig. 2). Average blood glucose concentration was 52.9 ± 4.8 mg/dl (means ± SD between subjects) during the hypoglycemic clamps and 92.4 ± 2.3 mg/dl during the euglycemic clamps. Counterregulation during hypoglycemia was demonstrated by measurement of serum glucagon, growth hormone, cortisol, and catecholamines (Fig. 3). Average plasma glucose levels during the [1-¹³C]glucose infusion were 115 ± 8 mg/dl and average insulin levels 40 ± 12 mU/l, with no difference between the euglycemia and hypoglycemia studies (P = 0.38 for glucose levels and 0.51 for insulin, paired t test) (Fig. 5,A and B). Steady ¹³C isotopic enrichment levels in blood glucose were achieved during the long infusions as demonstrated by data obtained in one subject in Fig. 5,C. We fitted a model of glycogen turnover to the time courses of [¹³C]glycogen obtained during the euglycemia studies of the five volunteers. This resulted in estimates of glycogen content of 3.5 ± 0.1 μmol/g and turnover rate (V_syn = V_phos) of 0.20 ± 0.01 μmol · g⁻¹ · h⁻¹, in agreement with the results of the glycogen utilization study and our prior published results for euglycemia and slight hyperglycemia (4). The newly synthesized glycogen levels were higher after hypoglycemia than after euglycemia across all time points during and after the ¹³C-glucose infusions (Fig. 6, P ≤ 0.01 paired t test for each subject separately), indicating increased glycogen synthesis after hypoglycemia. The glycogen synthesis rate can be estimated from the initial rate of label incorporation when ¹³C enrichment of glycogen is negligible and the labeling kinetics primarily represents synthesis. The first glycogen data point obtained from each volunteer at 4 h was used for this purpose. The synthesis rate of glycogen was 0.25 ± 0.03 μmol · g⁻¹ · h⁻¹ after euglycemia and 0.32 ± 0.05 μmol · g⁻¹ · h⁻¹ after hypoglycemia (P < 0.02, paired t test). The difference between newly synthesized glycogen levels increased steadily over time during the [¹³C]glucose infusion (Fig. 6) reaching ∼1 μmol/g at 34 h, indicating a net synthesis of ∼1 μmol/g glycogen occurred over this time period.

Here we present the first evidence for glycogen utilization during, and supercompensation after, moderate hypoglycemia in the healthy human brain. Using ¹³C NMR, we found that brain glycogen content decreased by ∼15% during modest hypoglycemia, whereas it was unchanged under isoinsulinemic euglycemia. Our data also indicate that brain glycogen content increased after a period of modest hypoglycemia but did not change after isoinsulinemic euglycemia in a second group of healthy volunteers.

In the utilization experiment, we detected glycogen mobilization by an increased ¹³C label wash-out from prelabeled glycogen during hypoglycemia relative to euglycemia. The ¹³C label was incorporated into glycogen mainly via turnover, as net synthesis does not occur at the euglycemia and slight hyperglycemia we utilized during prelabeling (4). Hence, the total and labeled glycogen levels were equal before the hypoglycemic and euglycemic clamps. We designed the study with an ∼1- to 2-h [¹²C]glucose infusion after the ¹³C to chase the [¹³C]glucose from blood, such that any [¹³C]glucose removed from glycogen during the clamps would not be replenished by [¹³C]glucose from the blood, increasing our chances to detect glycogen mobilization. Ideally, one would turn over all glycogen molecules before the clamp and keep the isotopic enrichment of the blood constant and equal to that of glycogen (31) during the clamps such that the glycogenolysis rate would equal the rate of label wash-out from glycogen. However, it takes 3–5 days to turn over the total human brain glycogen pool (4) and it is very difficult to keep blood isotopic enrichments constant during hypoglycemia based on our prior experience, making this experimental design unfeasible in humans. The isotopic enrichment of glucose during hypoglycemia in our studies tended to be higher than during euglycemia (22 vs. 11%), which might have even reduced the difference in [¹³C]glycogen levels between the hypoglycemic and euglycemic clamps. We do not expect this to be a factor because the [¹³C]glucose levels available for incorporation into glycogen were comparable during the two clamps considering the higher glucose levels during euglycemia. In theory, the increased label wash-out from glycogen during hypoglycemia may have been because of increased turnover; however, this possibility is highly unlikely considering the known reciprocal regulation of glycogen synthase and phosphorylase (22).

Our observations demonstrate that the human brain employs mechanisms of hypoglycemia response that are similar to those in the rodent brain (23,,–26). The [¹³C]glycogen signal decreases with a rate of ∼64%/h in the rat brain at ∼1.5 mmol/l blood glucose (23) and ∼10%/h in the human brain at ∼3 mmol/l blood glucose, indicating a mobilization rate commensurate with the severity of hypoglycemia. The rat study implied that glycogen was not mobilized until brain glucose levels were zero (23); however, in the current study brain glucose was 0.6–0.8 μmol/g based on reported glucose transport parameters for the human brain (5,32). At these glucose levels, hexokinase is 92–94% saturated (K_M = 50 μmol/l), whereas it is 95–97% saturated at euglycemia (1–1.5 μmol/g brain glucose). This slight desaturation of hexokinase may have been enough to trigger glycogen mobilization to supplement the glucose-6-phospate deficit. Alternatively, a more general brain stress response may have been operative, involving the activation of brainstem catecholaminergic neurons, shown to occur with hypoglycemia (33,–35). In particular norepinephrine is very effective in increasing glycogen breakdown (22) and may do so in the absence of a significant glucose deficit. Interestingly, after treatment with a glycogen phosphorylase inhibitor to increase brain glycogen content, neuronal function is prolonged during severe hypoglycemia in rats (36), providing further evidence that the brain may rely on glycogen stores to augment reduced energy delivery during hypoglycemia.

We only utilized the data from the euglycemia studies to fit a glycogen metabolic model because the model assumes the data were collected under steady-state conditions, which was not true during hypoglycemia. With the euglycemia data, we obtained glycogen content and turnover values in agreement with our previous findings (4). To roughly estimate the glycogenolysis rate during hypoglycemia we used the formula d[¹³C-glycogen]/dt = IE_glc × V_syn − IE_glyc × V_phos, using the average blood IE_glc (isotopic enrichment of free glucose) and assuming constant IE_glyc (isotopic enrichment of glycogen) during the 2-h clamps. IE_glyc was ∼40% based on the measured [¹³C]glycogen level relative to total glycogen (4). Because brain glucose isotopic enrichment closely follows the blood glucose isotopic enrichment, IE_glc was set equal to the average isotopic enrichment measured in the blood during the hypoglycemic clamps, 22%. The net glycogenolysis rate (V_phos − V_syn) could then be estimated by investigating two limiting conditions, with V_syn set to 0 or to the turnover rate of glycogen, 0.18 μmol · g⁻¹ · h⁻¹. This way we estimated a glycogenolysis rate of 0.22–0.30 μmol · g⁻¹ · h⁻¹, that is, that 0.4–0.6 μmol/g glycogen was mobilized during the 2-h hypoglycemic clamp. This glycogenolysis rate still constitutes a very small fraction (∼1%) of CMR_glc (if CMR_glc does not change under hypoglycemia) and shows that the blood supplies the majority of glucose utilized by the brain during moderate hypoglycemia. This was the case even during severe hypoglycemia in rats where glycogen was shown to supplement a small glucose deficit (23).

In the supercompensation experiment, we observed a higher synthesis rate for human brain glycogen after hypoglycemia versus euglycemia. This higher rate could not be because of any differences in insulin levels (23,31) as these were the same in the paired studies (Fig. 5 B). Clearly, some of this synthesis had to replenish the glycogen mobilized during the 2-h hypoglycemia. However, because a net synthesis of ∼1 μmol/g glycogen occurred during 34 h of [¹³C]glucose infusion and only 0.4–0.6 μmol/g glycogen was mobilized during the prior hypoglycemic clamp, our data indicate that glycogen content was higher after moderate hypoglycemia. It would be ideal to observe [¹³C]glycogen levels higher than the normal glycogen levels (3–4 μmol/g) to confirm supercompensation; however, this would require even longer experimental periods in humans than the 4 days in this study.

It is unlikely that the glycogen content of the brain can increase many-fold because of the restriction of brain volume within the skull and water retention by glycogen. However, up to fourfold increases above basal brain glycogen content have been observed (37) and can likely be accommodated because of the low glycogen content of the brain. Glycogen supercompensation has been observed after multiple metabolic stressors in the rodent brain, such as hypoxia (38), hypoglycemia (37), ischemia (37), brain injury (39), and sleep deprivation (3), and in other tissues, such as the muscle after exercise (40). Furthermore, supercompensation of muscle glycogen after its depletion with exercise is augmented with repeated bouts of exercise, that is, in exercise-trained rodents and humans (40). Therefore, glycogen supercompensation may be a protective measure taken by the affected tissue in preparation for the next bout of metabolic stress (38).

Our data that indicate supercompensation of human brain glycogen are in agreement with similar NMR studies in rats (23). Furthermore, glycogen supercompensation in the hypothalamus and cortex after recurrent glucopenia was demonstrated recently in a rat model of hypoglycemia-associated autonomic failure (HAAF) (41). Although these observations suggest that glycogen supercompensation may be involved in the development of hypoglycemia unawareness, recent experiments by Herzog et al. (24) failed to demonstrate glycogen supercompensation in the cortex, cerebellum and hypothalamus in awake rats 6 and 24 h after hypoglycemia. They attributed this failure to confirm the prior rat NMR study (23) to anesthesia and severe hyperglycemia levels used for the NMR experiments. However, our current data, that also indicate supercompensation, were acquired with awake humans maintained at mild hyperglycemia (115 mg/dl) during the [¹³C]glucose infusion after the clamps (Fig. 5 A). We suggest that the variability among measurements in animals studied at different time points might be the reason that Herzog et al. did not observe supercompensation after hypoglycemia in the rat model. In NMR experiments, time courses are monitored in individual subjects, thereby facilitating the observation of treatment effects relative to extraction studies where all data points are obtained from different animals. Indeed, in the Herzog et al. study, even though the cortical glycogen levels were almost doubled in cortex after recurrent versus acute hypoglycemia (∼7 vs. 4 μmol/g), this difference was not statistically significant, likely because of the large variance between animals.

Taken together, our observations demonstrate that brain glycogen is a dynamic source of energy and provide the first support for the hypothesis that brain glycogen may be used to offset the loss of substrate that occurs in humans during hypoglycemia. They further demonstrate increased brain glycogen synthesis after moderate hypoglycemia in humans and indicate glycogen supercompensation. The potential involvement of glycogen in the development of HAAF in humans, and specifically if supercompensated glycogen provides additional substrate to the brain during subsequent hypoglycemic episodes, need to be investigated in future studies.

This work was supported by the National Institutes of Health Grant R01 NS035192 (to E.R.S.). The CMRR is supported by National Center for Research Resources (NCRR) Biotechnology Research Resource Grant P41RR008079 and Neuroscience Center Core Blueprint Award P30NS057091. The GCRC is supported by NCRR Grant M01RR00400.

Parts of this study were presented in abstract form at the Scientific Meeting of the International Society for Magnetic Resonance in Medicine, Berlin, Germany, 19–25 May 2007; at the 67th Scientific Sessions of the American Diabetes Association, Chicago, Illinois, 22–26 June 2007; at the Perugia 2007 Hypoglycemia Symposium, Perugia, Italy, 12–15 May 2007; and at the 69th Scientific Sessions of the American Diabetes Association, New Orleans, Louisiana, 5–9 June 2009.

We thank the nurses and medical assistants of the GCRC for their enthusiastic support of the glucose infusion studies, the staff of the CMRR for maintaining and supporting the NMR system, Gregor Adriany for invaluable help with the NMR coil, and Alexander Shestov, Felipe Barros, and Gerry Dienel for discussions and comments on our work.