Increased Brain Transport and Metabolism of Acetate in Hypoglycemia Unawareness

Abstract

Context:

Intensive insulin therapy reduces the risk for long-term complications in patients with type 1 diabetes mellitus (T1DM) but increases the risk for hypoglycemia-associated autonomic failure (HAAF), a syndrome that includes hypoglycemia unawareness and defective glucose counterregulation (reduced epinephrine and glucagon responses to hypoglycemia).

Objective:

The objective of the study was to address mechanisms underlying HAAF, we investigated whether nonglucose fuels such as acetate, a monocarboxylic acid (MCA), can support cerebral energetics during hypoglycemia in T1DM individuals with hypoglycemia unawareness.

Design:

Magnetic resonance spectroscopy was used to measure brain transport and metabolism of [2-¹³C]acetate under hypoglycemic conditions.

Setting:

The study was conducted at the Yale Center for Clinical Investigation Hospital Research Unit, Yale Magnetic Resonance Research Center.

Patients and Other Participants:

T1DM participants with moderate to severe hypoglycemia unawareness (n = 7), T1DM controls without hypoglycemia unawareness (n = 5), and healthy nondiabetic controls (n = 10) participated in the study.

Main Outcome Measure(s):

Brain acetate concentrations, ¹³C percent enrichment of glutamine and glutamate, and absolute rates of acetate metabolism were measured.

Results:

Absolute rates of acetate metabolism in the cerebral cortex were 1.5-fold higher among T1DM/unaware participants compared with both control groups during hypoglycemia (P = .001). Epinephrine levels of T1DM/unaware subjects were significantly lower than both control groups (P < .05). Epinephrine levels were inversely correlated with levels of cerebral acetate use across the entire study population (P < .01), suggesting a relationship between up-regulated brain MCA use and HAAF.

Conclusion:

Increased MCA transport and metabolism among T1DM individuals with hypoglycemia unawareness may be a mechanism to supply the brain with nonglucose fuels during episodes of acute hypoglycemia and may contribute to the syndrome of hypoglycemia unawareness, independent of diabetes.

Clinical trial data demonstrate the benefits of intensive insulin therapy in reducing the risk for long-term complications in patients with type 1 diabetes (T1DM) (1, 2). However, intensive insulin therapy is accompanied by an increased risk of severe hypoglycemia and hypoglycemia unawareness, an inability to sense low blood glucose levels (3). Many patients with T1DM do not achieve target glucose levels because the immediate risk of acute hypoglycemia outweighs the long-term benefits of tight glycemic control (3–5).

Hypoglycemia unawareness has been attributed to both a downward shift in central nervous system (CNS)-triggered sympathoadrenal responses to a lower glucose threshold and consequently a loss of the adrenergic symptoms and to adaptations in the cerebral cortex to allow normal function under hypoglycemic conditions. Both of these mechanisms are induced by recurring hypoglycemic events, a concept known as hypoglycemia-associate autonomic failure (HAAF), and can contribute to a recurring cycle of increasingly severe hypoglycemia (6, 7).

Multiple mechanisms have been proposed to explain hypoglycemia unawareness, including the use of nonglucose fuels such as monocarboxylic acids (MCAs; ie, lactate, ketones, acetate) in the cerebral cortex during periods of hypoglycemia (7, 8). Our group previously demonstrated increased brain transport and use of acetate at steady state in intensively treated T1DM individuals compared with matched nondiabetic controls, using magnetic resonance spectroscopy (MRS) during a hyperinsulinemia-hypoglycemic clamp study with concurrent infusion of [2-¹³C]acetate (8). The current study examines brain MCA use in a rigorously characterized group of T1DM patients with hypoglycemia unawareness and compares them with both T1DM control subjects without hypoglycemia unawareness (ie, intact awareness of hypoglycemia) and nondiabetic controls. The inclusion of a T1DM control group provided the opportunity to investigate whether hypoglycemia unawareness is an adaptation to recurrent hypoglycemia or a function of diabetes per se. To better characterize the pathophysiology of HAAF, we examined the relationship between the adaptations in cortical acetate metabolism in T1DM during hypoglycemia and counterregulatory responses. Finally, in contrast to the previous ¹³C-acetate study, we measured the dynamic time course of ¹³C labeling of glutamine and glutamate generated during hypoglycemia in diabetic patients and for the first time determined the absolute rates of brain [2-¹³C]acetate metabolism using mathematical modeling.

Materials and Methods

Recruitment and eligibility

Three groups of study participants between 18 and 55 years of age were recruited, including the following: 1) individuals with T1DM with a clinical history of recurrent severe hypoglycemia (defined as hypoglycemia requiring assistance from another person) within the prior year and moderate to severe hypoglycemia unawareness as measured by a Ryan Hypoglycemia Score (HYPO) greater than 423 (9) (T1DM/unaware), and this group included 4 individuals recruited from islet cell transplant programs; 2) a T1DM control group with infrequent hypoglycemia, no episodes of severe hypoglycemia within the prior year, and intact hypoglycemia awareness as defined by a HYPO score less than 423 (T1DM control); and 3) a nondiabetic control group with normal blood glucose less than 100 mg/dL and normal hemoglobin A1c (nondiabetic control). The HYPO is a composite measure of frequency, severity of hypoglycemia (defined as a blood glucose < 54 mg/dL), and degree of hypoglycemia unawareness (9), and HYPO scores 423–1047 are considered indicative of moderate hypoglycemia unawareness, and HYPO scores greater than 1047 are consistent with severe hypoglycemia unawareness.

Exclusion criteria were smoking, active alcohol or substance abuse, creatinine greater than 1.5 mg/dL, untreated proliferative retinopathy, active infection (including hepatitis, tuberculosis, etc), liver function tests greater than 1.5 times the upper limit of normal, hemoglobin less than 11 g/dL (females) or less than 12 g/dL (males), leukopenia, uncontrolled psychiatric disease, significant cardiac disease, use of systemic glucocorticoids, positive pregnancy test or breast-feeding. All participants underwent a thorough screening evaluation, including a complete history and physical, laboratory testing, and an electrocardiogram to ensure that they met protocol criteria.

Written informed consent was obtained from each participant after the study protocol was carefully explained and questions answered. The protocol was approved by the Human Investigations Committee at Yale University School of Medicine and the Institutional Review Board at the University of Miami (Miami, Florida) and the University of Minnesota (Minneapolis, Minnesota).

Experimental protocol

Diabetic participants were admitted to the Yale Center for Clinical Investigation Hospital Research Unit the evening before the study and fasted overnight until the end of the study the following day. At 9:00 pm, an iv catheter was inserted into an antecubital vein for infusion of insulin (regular human insulin; Novo Nordisk) and dextrose to maintain normoglycemia overnight. The following morning, an additional antecubital iv line was placed in the contralateral antecubital vein and kept patent with normal saline. Upon arrival at the Magnetic Resonance Research Center, baseline levels of glucose, lactate, insulin, catecholamines, and glucagon were drawn. After the subjects were positioned on the gurney of the magnetic resonance (MR) scanner, a hyperinsulinemic-hypoglycemic clamp was initiated (2 mU/kg · min). Plasma glucose was checked every 5 minutes, allowed to gradually fall to 3.1 mM (range 2.8–3.3 mM), and maintained at that level with a variable rate of 20% dextrose infusion. Once plasma glucose was maintained at 3.1 mM for 10–15 minutes, a bolus of [2-¹³C]acetate 6 mg/kg · min was administered for 5 minutes followed by a continuous infusion of 3 mg/kg · min for a total of 120 minutes. MR spectra were obtained, and serial plasma samples were drawn for glucose, lactate, and hormone levels and plasma measurements of ¹³C enrichment. After completion of the [2-¹³C]acetate infusion, insulin infusion was discontinued, 20% dextrose infusion was increased, and the participant was fed.

Measurements of hormones and metabolites

Plasma glucose and lactate were measured enzymatically using glucose and lactate oxidase, respectively (Yellow Springs Instruments). Plasma glucagon was measured using a double-antibody RIA (Millipore). Epinephrine and norepinephrine were measured by HPLC (ESA). One subject (T1DM control) transiently had an extremely high peak epinephrine level that coincided in time with experiencing severe backache that required a pause in the MRS acquisition and to briefly take the subject out of the scanner. The subject's epinephrine data were excluded because these were likely more representative of stress experienced due to pain than hypoglycemia counterregulation. Plasma ¹HMRS data were acquired on a 500-MHz Bruker Avance spectrometer (Bruker Instruments). Untreated plasma (200 μL) was diluted to a total volume of 600 μL, in a pH-buffered solution containing a formate concentration standard (2 mM). The percent ¹³C enrichment of the plasma acetate was calculated using integrals of the downfield satellite (1.77 ppm) and the central peak (1.9 ppm) of [2-¹³C]acetate in the ¹HMR spectrum. The plasma concentration of acetate was calculated using the acetate peaks and the formate standard.

MRS acquisition

MRS data were acquired on a 4 Tesla whole-body magnet (Bruker Instruments). Subjects lay supine in the magnet, with the head on a radiofrequency MR probe consisting of one 8.5-cm diameter ¹³C circular coil and 2 ¹H quadrature coils for ¹H acquisition and decoupling. After tuning, acquisition of scout images, shimming with FASTERMAP procedure (10), and calibration of the decoupling power, ¹³C MRS spectra were averaged every 5.3 minutes during the infusion of [2-¹³C]acetate. MR spectra were acquired using an adiabatic ¹³C-[¹H] polarization transfer sequence optimized for detection of glutamate and glutamine in the C4 position, including image-selected in vivo spectroscopy for 3-dimensional localization (11). The spectroscopic voxel (4.5 × 4 × 5 cm³) was located at the midline of the occipital-parietal lobe.

MRS spectral analysis

Data were added in running averages of 15.9 minutes (ie, three 5.3 min blocks) and prepared for analysis with −2 Hz per 6 Hz Lorentzian-to-Gaussian conversion and 8-fold zero filling followed by Fourier transformation. Spectral fitting was used to adjust basis sets of metabolites derived from ¹³C measurements in solutions of metabolites at 37°C and pH 7. The basis set consisted of glutamate C4 and C3, glutamine C4 and C3, aspartate C3, N-aspartyl aspartate C3 and C6, and creatine C2. Glutamate and glutamine were quantified relative to the average amplitude of C3 and C6 of N-aspartyl aspartate, which was not labeled over the time course of the study and was therefore assumed to represent 1.1% natural abundance of 11 μmol/g (12, 13). Percent ¹³C enrichments of glutamate and glutamine were calculated using literature values for concentrations for brain glutamate (9.8 μmol/g) and glutamine (4.2 μmol/g) (14).

Steady-state metabolic modeling

The ratio of acetate use to the astrocytic tricarboxylic acid cycle (TCA) cycle rate (CMR_ac/V_TCAA) was calculated based on steady state percent ¹³C enrichments of glutamate and glutamine C4. A series of differential equations describing a 2-compartment model was used, as previously described (8). Values of the glial TCA cycle rate (V_TCAA) and the rate of glutamine/glutamate neurotransmitter cycling (V_cyc) were considered similar as during euglycemia and thus used from a previous, comparable [2-¹³C]acetate study from our group (15). The assumed values of V_TCAA and V_cyc do not influence the calculation of absolute rates of glial acetate metabolism (CMR_ac) because these calculations use steady-state percent enrichments of glutamate and glutamine, which are independent of the metabolic rates. Possible differences in V_TCAA were investigated with the dynamic metabolic model.

Dynamic metabolic modeling

Using the time courses of ¹³C-enrichment of glutamate and glutamine C4 and steady-state enrichments of glutamate and glutamine C3, metabolic rates were determined by fitting a 2-compartment model of astroglia and neuronal metabolism (16). The individual plasma time courses of acetate enrichment were used as input for the model. Mass and isotopic flows from [2-¹³C]acetate to brain glutamate and glutamine were expressed as coupled differential equations with CWave 3.0 (17) running in Matlab (Mathworks). The equations (Supplemental Table 1, published on The Endocrine Society's Journals Online web site at http://jcem.endojournals.org) were solved using a first-order Runge-Kutta algorithm, and least-squares optimization was achieved using a Levenberg-Marquardt algorithm.

Statistics

Comparisons between 3 groups were carried out by 1-way ANOVA with Tukey honestly significant difference tests for post hoc analysis. ANOVA for repeated measures was used to analyze changes in epinephrine and glucagon levels between euglycemia and hypoglycemia. Pearson's correlation coefficient was calculated to analyze the relation between plasma peak epinephrine levels and steady-state glutamine C4 percent ¹³C enrichment. All statistical analyses were carried out using the SPSS 19 software package (IBM). P < .05 was considered as statistically significant and data are presented as mean ± SD.

Results

Subject characteristics are presented in Table 1. T1DM subjects were grouped into those with moderate to severe hypoglycemia unawareness (T1DM/unaware) as defined by the HYPO score greater than 423 (n = 7) and T1DM control subjects with intact hypoglycemia awareness as defined by a HYPO score less than 423 (n = 5).

Table 1.

Subject Characteristics

	T1DM/Unaware (n = 7)	T1DM Control (n = 5)	Control (n = 10)	P Value
Age, y	44.7 ± 11.1	39.2 ± 11.5	34.7 ± 11.7	>.05
Race	Cauc (7)	Cauc (4), AA (1)	Cauc (10)	n.a.
Gender	Male (4), female (3)	Male (4), female (1)	Male (6), female (4)	n.a.
Years since dx	30.9 ± 12.4	12.1 ± 10.5	n.a.	.021
BMI	24.1 ± 3.7	26.7 ± 3.0	26.1 ± 3.0	>.05
HbA1c, %	6.6 ± 1.1	7.6 ± 1.3	5.2 ± 0.4a	<.001
HYPO mean (range)	1804 (458–4315)	70.2 (0–351)	n.a.	.011

Abbreviations: AA, African American; BMI, body mass index; Cauc, Caucasian; dx, diagnosis; HbA1c, glycosylated hemoglobin; n.a., not applicable. P values were acquired using 1-way ANOVA for comparisons between 3 groups, and P values were acquired using unpaired t test for comparisons between 2 groups (HYPO score and years since diagnosis).

^aPost hoc analysis revealed HbA1c in the nondiabetic control group to be significantly different from both diabetic groups (P = .026 and P < .001). There was no significant difference for HbA1c between both diabetic groups (P = .101).

Substrate and counterregulatory hormones

Plasma glucose levels achieved and maintained during the hypoglycemic period were not significantly different in the 3 groups (T1DM/unaware 54.5 ± 1.5 mg/dL; T1DM control 56.7 ± 2.5 mg/dL; nondiabetic control 54.8 ± 1.5 mg/dL, P > .05). Plasma lactate levels at baseline were also indistinguishable (T1DM/unaware 0.9 ± 0.3 mM;T1DM control 0.8 ± 0.2 mM; nondiabetic control 1.0 ± 0.6 mM, P > .05) and rose similarly during hypoglycemia (at 120 min after hypoglycemia, T1DM/unaware lactate 1.7 ± 0.4 mM; T1DM control 1.8 ± 0.4 mM; nondiabetic control 1.6 ± 0.5 mM, P > .05). The counterregulatory epinephrine response to hypoglycemia is illustrated in Figure 1. The T1DM/unaware group showed no change in epinephrine levels in response to hypoglycemia, whereas both control groups displayed a significant increase (both P < .05). Plasma glucagon levels did not change in any of the experimental groups (basal vs hypoglycemia: 44.5 ± 3.8 vs 30.1 ± 9.8 pg/mL in T1DM/unaware; 47.5 ± 6.6 vs 45.5 ± 5.1 pg/mL in T1DM control; and 53.7 ± 19.1 vs 60.4 ± 23.4 pg/mL in nondiabetic control subjects, P < .05).

Figure 1.

Counterregulatory epinephrine response in basal (B) and hypoglycemic (H) conditions presented as mean and SD. *, P < .05 compared with the T1DM group for effect of hypoglycemia (ANOVA for repeated measures).

Plasma and brain acetate concentrations

Plasma acetate levels and percent enrichment reached steady state values within 5 minutes in the 2 T1DM groups and in controls. There were no significant differences in steady-state concentrations or enrichments between groups (T1DM/unaware: 0.7 ± 0.3 mM and 91.9% ± 3.6%; T1DM control: 0.8 ± 0.2 mM and 93.7% ± 1.3%; nondiabetic control: 0.7 ± 0.3 mM and 93.5% ± 1.6% P > .05). The concentration of brain acetate was determined by measuring the ¹³C concentration of brain acetate and dividing the concentration by the plasma acetate percent enrichment. The 1-way ANOVA revealed group differences for brain acetate (P = .046). Tukey post hoc tests showed trends toward higher brain acetate in the T1DM/unaware group (0.05 ± 0.01 μmol/g) compared with the nondiabetic control group (0.03 ± 0.02 μmol/g, P = .062) and compared with the T1DM control group (0.03 ± 0.02 μmol/g, P = .097).

¹³C MRS measurements of glutamate and glutamine labeling

Representative brain ¹³C MR spectra from a T1DM and nondiabetic control subject during hypoglycemia and concurrent infusion of [2-¹³C]acetate are shown in Figure 2. The T1DM subject with impaired awareness (Figure 2, top panel) had significantly greater ¹³C labeling of glutamine C4 compared with the nondiabetic control (Figure 2, middle panel). The bottom panel in Figure 2 shows the difference spectrum, illustrating the increased resonance intensity and ¹³C labeling of the glutamine and glutamate C4 resonances.

Figure 2.

¹³C MR spectra acquired during steady state from a hypoglycemia-unaware T1DM subject (top panel) and control subject (middle panel). The bottom spectrum is the difference of the latter two and illustrates the higher level of ¹³C labeling in the T1DM/unaware subject.

At steady state, brain C4 glutamine ¹³C percent enrichments were significantly higher among T1DM subjects with severe unawareness compared with both control groups (ANOVA: P = .001; T1DM/unaware 12.4% ± 3.4%, T1DM controls 7.7% ± 2.0%, P = .003; nondiabetic controls 6.9% ± 2.3%, P = .003), consistent with increased transport of acetate across the blood-brain barrier and metabolism of acetate within glial cells in the brain (Figure 3A). In contrast, there was no difference in the percent ¹³C enrichment of glutamine C4 between the T1DM controls and nondiabetic controls (P = .844). Glutamate ¹³C enrichments were also significantly higher in unaware T1DM subjects compared with the T1DM control subjects but not compared with healthy control subjects (ANOVA: P = .019; T1DM/unaware 3.6% ± 1.0%, T1DM controls 2.1% ± 0.6%, P = .016; nondiabetic controls 2.8% ± 0.8%, P = .141). Taken together, these data provide evidence indicating that the changes in brain fuel metabolism observed in T1DM/unaware subjects are a function of recurrent severe hypoglycemia and independent of diabetes per se.

Figure 3.

¹³C percent enrichment (P.E.) of glutamine C4 at steady state (∼90–120 min) from individual subjects of each study group (A). Hypoglycemia-unaware T1DM subjects (gray symbols) showed significantly higher levels of ¹³C P.E. of glutamine C4 than both control groups. ¹³C percent enrichment data and best fit of the 2-compartment model from a hypoglycemia-unaware T1DM subject (B) and a control subject (C) during the infusion of [2-¹³C]acetate to estimate the absolute brain acetate consumption (CMR_ac) during hypoglycemia are shown (see Table 2). Closed symbols (●) represent glutamine C4 and open symbols (○) represent glutamate C4. Data are presented as individual points and group mean ± SD.

Steady state metabolic modeling

We calculated the acetate consumption as a function of the astroglial TCA cycle rate (CMR_ac/V_TCAA) from the steady state ¹³C enrichment of glutamate, glutamine, and plasma acetate using the formula introduced in Mason et al. [(8), equation 2]. CMR_ac/V_TCAA was significantly higher in T1DM/unaware (0.36 ± 0.10) as compared with both T1DM (0.20 ± 0.0, P = .002) and nondiabetic control groups (0.19 ± 0.07, P = .004).

Dynamic metabolic modeling

Figure 3, B and C, show time courses and fitted curves for labeling of the C4 carbons of glutamate and glutamine in a hypoglycemia-unaware T1DM subject and a nondiabetic control subject, respectively. The T1DM/unaware group showed higher labeling of both metabolites, indicating greater transport and metabolism of [2-¹³C]acetate, than the control groups. Metabolic modeling of the whole time course during [2-¹³C]acetate infusion allowed us to calculate the CMR_ac, the V_TCAA, and the V_cyc. The modeling showed that the CMR_ac was approximately 50% higher in T1DM/unaware subjects compared with both control groups (Table 2). The glial TCA cycle and neurotransmitter cycling rates, on the other hand, were similar across the 3 groups. Acetate was the major substrate for the glial TCA cycle in both the unaware T1DM and control groups.

Table 2.

Brain Acetate Concentrations and Metabolic Modeling Results

	T1DM/Unaware	T1DM Control	Control	P Value
Acetate, μmol/g	0.053 ± 0.013	0.030 ± 0.023	0.031 ± 0.018	.046
CMRac, μmol/g · min	0.036 ± 0.009a	0.024 ± 0.007	0.025 ± 0.006	.020
VTCAA, μmol/g · min	0.051 ± 0.015	0.050 ± 0.032	0.040 ± 0.020	.577
Vcyc, μmol/g · min	0.30 ± 0.74	0.37 ± 0.80	0.35 ± 0.11	.393

Brain acetate concentrations and metabolic modeling results after iv infusion of [2-¹³C]acetate in hypoglycemia-unaware T1DM subjects (n = 7), T1DM controls (n = 5), and controls (n = 10) as mean ± SD. The variables are as follows: acetate, brain [2-¹³C]acetate; CMR_ac, acetate consumption rate in brain; V_TCAA, astroglial TCA cycle rate. P values consist of 1-way ANOVA for comparison between 3 groups.

^aSignificantly different from T1DM control and control groups.

Relationship between acetate metabolism and peak epinephrine response

To further investigate the mechanisms contributing to hypoglycemia unawareness, the peak epinephrine response to hypoglycemia (an indirect measure of the level of hypoglycemia awareness) was plotted against CNS MCA transport and metabolism (as measured by glutamine C4 labeling) across the entire study population. As shown in Figure 4, there is a negative relationship (r = 0.7, P < .001) between these 2 measures, lending further support for the concept that up-regulated brain MCA use is related to HAAF as measured by defective epinephrine counterregulation.

Figure 4.

Correlation analysis of steady-state glutamine ¹³C percent enrichment (P.E.) and peak epinephrine response during hypoglycemia. Symbols represent individual data of hypoglycemia unaware T1DM subjects (gray), control subjects (black), and T1DM control subjects (open symbols) (Pearson's r = 0.7, P < .001).

Discussion

In the present study, we used a rigorous scoring tool (HYPO) to identify T1DM individuals with moderate to severe hypoglycemia unawareness and recurrent bouts of severe hypoglycemia and compared them with T1DM control subjects with intact hypoglycemia awareness and nondiabetic controls (9). We demonstrated increased brain transport and metabolism of acetate in intensively treated T1DM patients with hypoglycemia unawareness and reduced epinephrine responses to hypoglycemia as compared with T1DM and nondiabetic control subjects. The T1DM controls and nondiabetic controls were indistinguishable with regard to their rates of brain acetate consumption (CMR_ac, Table 2). Both control groups had a vigorous epinephrine response to hypoglycemia (Figure 1), in striking contrast to the T1DM unaware group. These findings strengthen the hypothesis that increased brain transport and metabolism of MCAs is a cortical response to antecedent hypoglycemia, rather than a function of diabetes per se. This alteration provides the brain greater availability of alternative fuels during episodes of hypoglycemia.

¹³C-acetate provides distinct advantages in the study of MCA brain transport and metabolism. Because acetate is solely metabolized in astroglia, the use of ¹³C-acetate allowed us to dissect the specific role of astroglia in the syndrome of HAAF (18, 19). In addition, the endogenous production of acetate is negligible, resulting in a very stable plasma profile during an infusion study and thus allowing more straightforward analysis of the data. However, plasma concentrations of acetate under physiological conditions are generally very low (<0.1 mM). The most likely substrate for brain metabolism under hypoglycemic conditions is lactate because its plasma concentration is about 10 times that of acetate during hypoglycemia (20). Studies using ¹³C MRS and arteriovenous differences have shown that at normal physiological levels, lactate has the potential to provide up to 10% of the brain's energy needs under euglycemic conditions in control subjects (21, 22). Although recent results suggest other roles for brain lactate in addition to possibly acting as an alternative substrate (23, 24). Nevertheless, transport of all MCAs across the blood-brain barrier occurs via the monocarboxylic acid transporter 1 (MCT1) (25, 26). Although MCT1 transport kinetics differ somewhat for various MCAs, the increased transport of [2-¹³C]acetate in hypoglycemia-unaware subjects noticed in this study should be largely applicable to other MCAs (lactate, ketones).

In contrast to transport characteristics of MCAs, the metabolic fate of MCAs after crossing the blood-brain barrier can be completely different. Brain metabolism is highly compartmentalized among neurons and astroglial cells (27–29). For example, neurons and astroglia have characteristically different TCA cycles rates (16, 28). Also, both cell types do not metabolize all substrates to the same degree. As a case in point, due to transport characteristics, acetate is almost exclusively oxidized in astroglia (15, 18, 19).

The 2-compartment model that was used to analyze the acetate metabolism through kinetics of glutamine and glutamate ¹³C labeling provided estimates of the CMR_ac. Among individuals with T1DM with moderate to severe hypoglycemia unawareness, CMR_ac was approximately 50% higher than both T1DM control subjects with intact hypoglycemia awareness and nondiabetic control subjects (Table 2). These data suggest that the loss of awareness is linked to a cortical alteration that increases the use of acetate. It is likely that the higher capacity to use acetate, and other monocarboxylic acids, may play a neuroprotective role by allowing diversion of the limited supply of glucose for critical neuronal functions during episodes of hypoglycemia. As has been shown by in vivo MRS in humans and animal models, glial glucose oxidation can account for 15%–20% of total brain glucose oxidation (28). Therefore, replacement with acetate or lactate can provide sufficient additional glucose availability to support neuronal function during hypoglycemia.

At the low levels of plasma acetate concentrations used in the present study, blood-brain transport is effectively unidirectional (toward brain) because the brain acetate levels in all groups were low. Because plasma acetate levels were comparable in all groups, the increased acetate metabolism in the unaware T1DM group can occur only if transport was increased. We compared cerebral acetate to see whether the increased acetate metabolism in the unaware T1DM group was driven by higher brain acetate levels as a consequence of higher transport activity or alternatively whether the enzymes involved in converting acetate to acetyl-CoA were also up-regulated. Based on the trend toward higher brain acetate concentration in the unaware group, it appears the main change is increased transport activity instead of increased activity of enzymes involved in acetate metabolism.

In the T1DM unaware subjects, acetate accounted for approximately 50% of glial metabolism, which based on previous studies is 10%–12% of total brain metabolism (28). This 10%–12% is in the range of the reduction in glucose metabolism reported at moderate levels of hypoglycemia in control subjects (30). Based on recent studies of human brain lactate transport and assuming a similar up-regulation of transport in the T1DM subjects as for acetate, plasma lactate at approximately 1 mM could provide approximately 13% of brain energy metabolism, which is sufficient to compensate for reduced energy from glucose (21).

In contrast to the acetate consumption, no difference across groups was found between V_TCAA and neurotransmitter cycling rates (Table 2), indicating that the modest hypoglycemic condition did not affect the intrinsic metabolic rate of astroglia. Changes in TCA and neurotransmitter cycle rates across the study groups may become apparent at lower levels of plasma glucose than could safely be achieved in this study. At present we do not know whether hypoglycemia is needed to expose the metabolic adaptation because the up-regulation of MCA transport and metabolism may be constitutive. However, hypoglycemia appeared required to reveal increased MCA metabolism in studies of the animal model of HAAF (24, 31).

To further explore the relationship between hypoglycemia unawareness (the neurocognitive aspect of HAAF localized within the cortex) and the peripheral epinephrine response mediated via the hypothalamus, we plotted cortical MCA transport/metabolism (as measured by brain glutamine C4 labeling) vs peak epinephrine response to hypoglycemia. As shown in Figure 4, when combining data from all groups, there is a striking negative correlation between ¹³C labeling of glutamine and the peak epinephrine response. The negative relationship between these 2 variables, ie, lower epinephrine levels associated with higher levels of labeled glutamine in the cerebral cortex, suggests that the increased use of MCAs (that likely maintain cerebral energetics during episodes of hypoglycemia) may be a link between impaired counterregulation and unawareness. However, the design of the present study does not allow concluding that the relation between the increased cortical MCA use and HAAF is causal or the metabolic adaptation is secondary to HAAF.

Several other factors have been proposed as potential contributors to the development of HAAF. The possibility that the up-regulation of blood-to-brain glucose transport/and or metabolism contributes to the development of HAAF has led to conflicting results. Some studies have shown preserved or increased cerebral glucose uptake and/or metabolism during hypoglycemia (32–36) and hyperglycemic conditions (37), whereas other studies have not (38–40). Likewise, expanded glycogen stores in astrocytes after antecedent episodes of hypoglycemia have been suggested as an alternative fuel source during subsequent episodes of hypoglycemia based on rodent studies (41). However, a recent study in humans using ¹³C MRS demonstrated lower brain glycogen content in individuals with T1DM and hypoglycemia unawareness compared with controls, concluding that supercompensated glycogen stores do not contribute to the development of hypoglycemia unawareness (42).

Many of the hypotheses of HAAF focus on alterations in cerebral fuel delivery, storage, and use; however, the downstream mechanisms responsible for the disruption of normal counterregulation and consequent hypoglycemia unawareness remain elusive. Overall, there are likely multiple potential CNS mediators of HAAF; however, several studies support the hypothesis that the brain compensates for recurrent hypoglycemia by increasing the use of nonglucose substrates (MCAs) and that this may contribute to HAAF (8, 31). Each MCA may play a unique role in cerebral energetics and possibly in the development of HAAF by virtue of their different transport characteristics (Michaelis constant, maximal rate), their different cellular targets (glia vs neurons), and their relative plasma concentrations (15, 18, 20, 25, 43).

The limited sensitivity of ¹³C MRS restricts the in vivo application to relatively large volumes. As such, we cannot confirm that the increased transport and metabolism of acetate is ubiquitous in every region of the brain. It is conceivable that increased transport capacity across endothelial MCT1 as adaptation to hypoglycemia would occur over the whole blood-brain barrier. However, whether transport of plasma MCAs is indeed increased to the same degree across the brain is unknown.

Although the difference was not statistically significant, the T1DM-unaware subjects happened to be on average older than the other subjects. Brain metabolism has been shown not to drastically change across the age range studied (44, 45); we therefore do not expect the age of the subjects to have affected the conclusions.

The T1DM-unaware group had a longer disease duration compared with the T1DM control subjects at the time of study, which weakens proposing a causal link between the metabolic adaptations observed in the brain and the presence/degree of HAAF. The argument can be made that the 3 recognized causes of HAAF are all related to recent, acute events: recent antecedent hypoglycemia, prior exercise, and sleep (7); thus, disease duration does not play an important role. However, the adaptations seen in cortical brain metabolism of acetate could be unrelated to HAAF and be an epiphenomenon related to disease duration. The design of the present study does not allow for clearly separating the brain adaptations observed from disease duration and presence/degree of hypoglycemia unawareness. However, Spearman correlation coefficient analysis across all T1DM subjects and within separate T1DM groups did not reveal significant correlations between years since diagnosis and level of glutamine C4 ¹³C enrichment (all T1DM: n = 12, r = 0.49, P = .11; T1DM unaware: n = 7, r = −0.41, P = .36; T1DM control: n = 5, r = 0.6, P = .29). A ¹³C-acetate infusion study in the clinical model of HAAF [a non-T1DM subject made hypoglycemia unaware through an acute hypoglycemic episode (39)] would be required to reveal whether the acute induction of hypoglycemia unawareness is paralleled by increased acetate uptake and metabolism in the brain. To the best of our knowledge, such data do not exist.

Increased transport via MCT1 is necessary to support the increased brain metabolism of plasma acetate. For acutely induced HAAF to be related to increased blood-brain barrier transport of MCAs would require a mechanism other than increased concentration of the MCT1 protein. The time required to assemble more of the MCT1 protein appears incompatible with a fast induction of increased MCA transport. However, regulation of luminal vs internalized presence of MCT1 in human endothelial cells of the blood-brain barrier would be compatible with a relatively fast regulation of transport capacity for plasma MCAs (46, 47).

In conclusion, our ¹³C MRS data confirm that the cortical brain becomes accustomed to recurrent hypoglycemia by increasing the transport and metabolism of MCAs as an alternative fuel source during subsequent hypoglycemia, and this alteration may play a role in HAAF. Mathematical modeling of ¹³C-labeling kinetics revealed that astroglial TCA cycle rate and neurotransmitter cycling were not altered in individuals with hypoglycemia unawareness. The comparison of carefully selected T1DM subjects with and without hypoglycemia unawareness strongly suggests that HAAF is not a function of type 1 diabetes itself but rather is mainly the consequence of recurrent hypoglycemia. It is possible that the longer duration of diabetes among the T1DM-unaware group may have contributed to these alterations as well. The cortical metabolic adaptations parallel the hormonal (dys)function across all subjects, suggesting a link between up-regulated brain use of alternative fuels and the syndrome of HAAF, although this relationship is not necessarily causal. Sparing glucose for critical neuronal function may be a useful adaptation during episodes of acute hypoglycemia. However, in the long term, this brain adaptation has the detrimental effect of masking awareness of falling blood glucose concentration, thus potentiating the severity of hypoglycemia.