Attenuation of Counterregulatory Responses to Recurrent Hypoglycemia by Active Thalamic Inhibition: A Mechanism for Hypoglycemia-Associated Autonomic Failure

Nine healthy young adults (four women/five men, mean ± SD age 25 ± 3 years, BMI 23.7 + 2.8 kg/m²) gave their written informed consent to participate in this study, which was approved by the Washington University Human Research Protection Office and conducted at the Washington University General Clinical Research Center (GCRC) and Neurology-Neurosurgery ICU PET Research Facility. All had negative medical histories and normal physical examinations, normal fasting plasma glucose concentrations, normal serum creatinine concentrations and hematocrits, and normal electrocardiograms. None had a personal or first-degree relative history of diabetes or psychiatric or neurological conditions. These were different subjects than in our previous study (14).

Subjects were admitted to the GCRC and studied over 2 consecutive days after overnight fasts. On the morning of day 1, two intravenous catheters were inserted into arm veins (for infusions) and one intravenous catheter was inserted into a dorsal hand vein with that hand kept in an ∼55°C plexiglas box (for arterialized venous sampling). Hyperinsulinemic (2.0 mU · kg⁻¹ · min⁻¹ regular human insulin; Novo Nordisk, Bagsværd, Denmark), euglycemic (90 mg/dl [5.0 mmol/l] for ∼2 h), and then hypoglycemic (55 mg/dl [3.0 mmol/l]) for ∼2 h) clamps—variable infusions of 20% dextrose based on plasma glucose determinations (YSI Glucose Analyzer 2; Yellow Springs Instruments, Yellow Springs, OH) every 5 min—were performed on the mornings of day 1 and day 2. [¹⁵O]water PET measurements of regional cerebral blood flow (see below) were performed four times at 15-min intervals during the second hour of euglycemia and four times at 15-min intervals during the second hour of hypoglycemia on both days. Arterialized venous samples for the analytes described below were drawn, blood pressures and heart rates were recorded, and symptom scores were determined every 15 min, and the electrocardiogram was monitored throughout. To quantitate symptoms, subjects were asked to score (from 0 [none] to 6 [severe]) six neurogenic (autonomic) symptoms: heart pounding, shaky/tremulousness, nervous/anxious (adrenergic), sweaty, hungry, and tingling (cholinergic); and six neuroglycopenic symptoms: difficulty thinking/confused, tired/drowsy, weak, warm, faint, and dizzy (15). After completion of the glucose clamps/PET studies on day 1, interprandial plasma glucose concentrations were held at ∼55 mg/dl (3.0 mmol/l), with variable intravenous insulin infusions until they were increased to 90 mg/dl (5.0 mmol/l) at the start of the clamps on the morning of day 2.

Plasma glucose concentrations were measured with a glucose oxidase method (YSI Glucose Analyzer 2). Plasma epinephrine and norepinephrine concentrations were measured with a single isotope derivative (radioenzymatic) method (16), and plasma insulin (17), C-peptide (17), glucagon (18), pancreatic polypeptide (19), cortisol (20), and growth hormone (21) were measured with radioimmunoassays. Serum nonesterified fatty acid (22) and blood lactate (23) concentrations were measured with enzymatic techniques.

Relevant systemic variables during euglycemia on days 1 and 2 and during hypoglycemia on days 1 and 2 were contrasted with a t test for paired data. P values <0.05 were considered to indicate significant differences.

PET studies were performed with a Siemens/CTI (Knoxville, TN) ECAT EXACT HR 47 tomograph located in the Neurology-Neurosurgery Intensive Care Unit PET Research Facility (24). Subjects were positioned in the tomograph such that the entire brainstem was included within the 15-cm axial field of view. As a result, the vertex of the brain was not included within the study area. An individual transmission scan was obtained for each subject and used for subsequent attenuation correction of emission scan data. Regional cerebral blood flow was measured with a 40-s emission scan after bolus intravenous injection of 50 mCi of [¹⁵O]water (1 Ci = 37 GBq) using an adaptation of the Kety autoradiographic method (25–27). PET scans were acquired in the two-dimensional mode (interslice septa extended).

PET images were reconstructed using filtered back projection and measured attenuation. Each subject's attenuation correction was co-registered to each emission scan before the final reconstruction in order to eliminate spatial variation in attenuation from each individual's difference images. Reconstructed images were smoothed with a three-dimensional Gaussian filter to a resolution of 10 mm full-width at half-maximum. Next, each subject's images were co-registered to each other (28), and the mean of all of that subject's images was co-registered to a standard mean blood flow image in Talairach atlas space (29). All individual images were normalized using mean whole-brain counts. The four euglycemic and the four hypoglycemic images for each subject were combined into single state images for each day to increase statistical precision (30). Images of regional cerebral blood flow (CBF) differences between euglycemia and hypoglycemia on day 1 and on day 2 were created by image subtraction. For the group data, these two subtraction images, each representing the regions of brain activation during acute hypoglycemia on each day, were examined for differences with SPM99 (31) using a fixed effects model with multiple comparisons correction for the entire volume. T values were computed at each voxel, and clusters of at least 90 voxels (1 resel) with uncorrected P < 0.005 were identified. The statistical significance of these clusters was computed and corrected for the number of possible clusters of their size and magnitude within the brain volume (32,33). This analysis determines areas where differences in CBF occur without any prespecification of regions of interest. Location of regions of significant change is expressed in x, y, and z stereotactic coordinates in millimeters with positive values representing right, anterior, and superior, respectively (29).

Each subject also underwent structural magnetic resonance imaging of the brain with a 1.5 T system (Magnetom Sonata, Siemens, Erlangen, Germany). A three-dimensional magnetization-prepared rapid gradient echo (MPRAGE) sequence (1900/1,100/3.9 repetition time/inversion time/echo time (ms); flip angle 15°) was acquired in a sagittally oriented plane (160-mm thick, 128 partitions, 256 mm field of view) and reconstructed into a 256 × 256 matrix (1 × 1 × 1.25 mm pixels).

Plasma insulin concentrations were raised substantially, and plasma glucose targets were met during the euglycemic and hypoglycemic clamps on days 1 and 2 (Table 1). Glucose infusion rates required to maintain both the euglycemic and the hypoglycemic clamps were lower, following interval hypoglycemia, on day 2 (Table 1). Plasma C-peptide concentrations were suppressed, during hypoglycemic clamps, on both days (Table 1).

Symptom scores and plasma epinephrine, glucagon, pancreatic polypeptide, and cortisol concentrations during hypoglycemia were reduced, following interval hypoglycemia, on day 2, as were the required glucose infusion rates and blood lactate concentrations (Table 1). Serum nonesterified fatty acid concentrations were suppressed under these hyperinsulinemic conditions.

The regional CBF patterns during euglycemia on day 2, following interval hypoglycemia, were not different from those on day 1 (data not shown). The only brain region that showed a significantly greater increase in CBF during hypoglycemia on day 2 compared with day 1 was the thalamus (corrected P = 0.004) (Fig. 1). The size of the significant day-2 cluster was 207 voxels. The region of CBF change was centered in the left pulvinar (x = −14, y = −30, and z = 7) and included the dorsal medial and paraventricular nuclei of the thalamus bilaterally (Fig. 1) (34).

These data again document that recent antecedent hypoglycemia—here ∼12–14 h of interval interprandial hypoglycemia—produces in healthy humans (10) a model of HAAF in diabetes (1–3) including the key feature of HAAF, an attenuated sympathoadrenal response to subsequent hypoglycemia. Compared with the responses on day 1, the adrenomedullary plasma epinephrine response and the largely sympathetic neural symptomatic response (4) (among other responses) to hypoglycemia were reduced on day 2, following interval hypoglycemia. Thus, the phenomenon we sought to explain mechanistically, HAAF in diabetes, was produced experimentally in these nondiabetic individuals.

We have used the technique of measuring changes in regional CBF with PET to identify regions of increased synaptic activity during hypoglycemia (13). This approach requires that the physiological coupling between increases in neuronal activity and increases in regional CBF remains intact under these conditions. We have previously shown that this coupling is preserved with arterial hypoglycemia of 2.3 mmol/l (41 mg/dl) (35). We measured regional CBF on 2 consecutive days with interval hypoglycemia. Repeat PET CBF studies show no significant differences within the same scanning session (36). Furthermore, [¹⁸F]deoxyglucose PET studies of metabolism performed 1 to 18 days apart show no changes in thalamic uptake normalized for whole brain counts, as was done here (37,38). In the present study, the regional CBF patterns during euglycemia on day 2 were not different from these on day 1. Therefore, the observed difference in the regional CBF response to hypoglycemia is unlikely to be an order effect. Areas of increased CBF primarily reflect the increased metabolic activity of synaptic fields, not the spike firing rate of neuronal cell bodies (39,40). Thus, the definitive finding of this study was a significantly greater increase in synaptic activity in the dorsal midline thalamus during recurrent hypoglycemia with attenuated counterregulatory responses compared with initial hypoglycemia with robust counterregulatory responses. The center of this region is virtually identical (x = −14, y = −30, and z = 7) to the region of increase during initial hypoglycemia (x = −11, y = −23, z = 6) that we identified in our previous study as part of a brain network that mediates autonomic responses to various stimuli and includes the dorsal medial and paraventricular nuclei of the thalamus bilaterally (14). The finding of greater synaptic activity associated with attenuated counterregulatory responses indicates that the dorsal midline thalamus plays an active inhibitory role in reducing the sympathoadrenal and symptomatic responses to hypoglycemia when previous hypoglycemia has occurred, the key feature of HAAF in diabetes.

Sympathoadrenal (and glucagon) responses to systemic hypoglycemia are directly stimulated by reduced local hypothalamic glucose concentrations (41–44). These responses are reduced by both lesions of the ventromedial hypothalamus and by maintenance of high extracellular hypothalamic glucose concentrations (41,42). They can be mimicked by selective hypothalamic cellular glucopenia with systemic euglycemia (43). In chronically systemically hypoglycemic rats, the epinephrine responses to selective hypothalamic cellular glucopenia with systemic euglycemia are attenuated. This observation is consistent with inhibition of the normal hypothalamic response to hypoglycemia in response to antecedent hypoglycemia (44).

Given the role of the hypothalamus as the central integrator of the sympathoadrenal response to hypoglycemia (41–44), the novel finding reported here extends the regulation of the sympathoadrenal response to include the dorsal midline thalamus. To our knowledge, there is no precedent for a role of the dorsal midline thalamus in the regulation of the sympathoadrenal response to hypoglycemia. However, that region of the thalamus includes the paraventricular nucleus of the thalamus (PVNT), and there is evidence that activation of the PVNT mediates habituation of the pituitary-adrenocortical response to recurrent restraint stress in rats (45). Therefore, given the present finding, it is reasonable to suggest that increased PVNT activation might mediate habituation of the sympathoadrenal response to recurrent hypoglycemia in humans, the key feature of HAAF in diabetes. Indeed, given the neuroanatomical connections of the dorsal midline thalamus and their physiological functions (46–54), we could postulate a cerebral network, including the medial prefrontal cortex, orbito-medial prefrontal cortex, hippocampus, amygdala, periaqueductal gray, and thalamus, regulating the hypothalamic response to recurrent hypoglycemia. Clearly, however, the documentation and definition of such a cerebral network awaits further study.

This construct of the pathophysiology of HAAF is consistent with the findings of Cranston et al. (55). Using [¹⁸F]deoxyglucose and PET, they found that deoxyglucose uptake decreased in the subthalamic region of the brain during hypoglycemia in patients with type 1 diabetes and hypoglycemia unawareness, whereas it was little changed in patients with hypoglycemia awareness. Since that subthalamic region is centered in the hypothalamus, it may well have been inhibited by the increased thalamic activity documented in the present study.

To unequivocally establish the causal nature of thalamic activation in HAAF will require demonstration that direct inhibition or ablation of dorsal medial thalamic nuclei prevents attenuation of counterregulatory responses to recurrent hypoglycemia. Surgically implanted electrodes in the thalamus, currently used to treat intractable tremor, provide a means to further investigate this mechanism and may in the future offer a novel approach to treatment of HAAF (56).

This study was supported, in part, by National Institutes of Health (NIH) Grants R37 DK27085, MO1 RR00036, and P60 DK20579, by Washington University (St. Louis, MO), and by the University of North Carolina. A.M.A. was supported, in part, by NIH Grant K12 DK63683.

The authors acknowledge the assistance of the staff of the Washington University General Clinical Research Center in the performance of this study; the technical assistance of Krishan Jethi, Michael Orasco, Gene Wade Sherrow, Michael Morris, Zina Lubovich, Sharon Travis, and Sharon O’Neil; and the assistance of Janet Dedeke in the preparation of this manuscript. We also thank Lori McGee-Minnich, Lennis Lich, the Washington University Medical School Cyclotron Staff, and the East Building Magnetic Resonance Imaging Facility staff for their assistance.