Infusions of ¹³C-Labeled Substrates

Solutions containing either [2,4-¹³C₂]BHB (1.5 to 4.5mol/L in deionized water, pH adjusted to 7.0, 99 atom%) or sodium [2-¹³C]acetate (2.0mol/L in deionized water, pH adjusted to 7.0, 99 atom%) (Cambridge Isotopes, Andover, MA, USA) were infused. Bolus-variable rate infusions were employed to achieve rapid and approximately constant increases in blood BHB or acetate concentrations and enrichments. [2,4-¹³C₂]BHB was infused in three steps (0 to 5minutes, 0.60mmol/minute per kg; 5 to 10minutes, 0.30mmol/minute per kg; thereafter, 0.12mmol/minute per kg). [2-¹³C]acetate was infused in four steps (0 to 15seconds, 6.25mmol/minute per kg; 15seconds to 4minutes, 0.875mmol/minute per kg; 4 to 8minutes, 0.50mmol/minute per kg; thereafter, 0.25mmol/minute per kg).

Preparation of Tissue Extracts and Blood Plasma

Frozen cortical tissue (150 to 200mg) was ground with 0.1M HCl/methanol (2:1 vol/wt) at low temperature using an ethanol/dry-ice bath followed by extraction with ice-cold ethanol.¹¹ [2-¹³C]glycine was added during tissue extraction as an internal concentration reference. Supernatants were clarified by centrifugation, passed through Chelex-100 to remove paramagnetic metal ions, lyophilized, and resuspended in 600μL of a phosphate-buffered (100mM, pH 7) deuterium oxide solution containing 3-trimethylsilyl[2,2,3,3-D₄]-propionate (0.5mM). Samples were loaded into 5mm nuclear magnetic resonance (NMR) tubes for analysis.

Blood plasma was centrifuged through Nanosep filters (10-kDa) to remove macromolecules. The ¹³C isotopic enrichment of plasma glucose and the concentration and ¹³C enrichment of plasma acetate were determined from filtered (10-kDa) samples using ¹H NMR.¹¹

Nuclear Magnetic Resonance Spectroscopy

¹H NMR spectra of plasma and brain extracts were acquired on a Bruker AVANCE NMR spectrometer at 500.13MHz (Bruker Instruments) under fully relaxed pulsing conditions. The ¹³C enrichments of glucose-C1 (5.2p.p.m.) and acetate-C2 (1.9p.p.m.) in ¹H NMR spectra were calculated as the ratio of ¹³C satellites to the total resonance intensity. ¹H-[¹³C]NMR spectra of brain tissue extracts were acquired using a pulse sequence incorporating adiabatic pulses for ¹H and ¹³C excitation.¹² Briefly, two subspectra are acquired, one in which the magnetization of ¹H atoms bound to ¹³C by one-bond coupling is inverted in phase relative to ¹H atoms bound to ¹²C atoms, and the other in which the ¹³C inversion pulse is not applied; this spectrum reflects total (¹³C+¹²C) composition from which total concentration is determined. Subtraction of one spectrum from the other (difference spectrum) yields ¹H atoms bound only to ¹³C. ¹³C decoupling radiofrequency is applied to both during spectral acquisition to simplify the spectra and increase the signal-to-noise ratio. After measurement of total amino acid concentrations in the extracts by ¹H-[¹³C]NMR, amino acids were separated using AG1-X8 anion exchange chromatography and samples re-measured to determine the fractional enrichments. The isotopic ¹³C enrichments of Glu-C3,C4, γ-aminobutyrate (GABA)-C2,C3,C4, Asp-C3, lactate (Lac)-C3, alanine -C3, [2,4-¹³C₂]β-hydroxybutyrate (BHB)-C2,C4, and Gln-C3,C4 in the separated fractions were determined as the ratio of ¹³C-bound ¹H to total ¹H(¹²C+¹³C) for each carbon position.

Analysis of Plasma Glucose and Acetate

The concentration of glucose in plasma samples was determined by using a Beckman Glucose Analyzer II (Beckman Coulter Inc., Brea, CA, USA). The ¹³C isotopic enrichment of plasma glucose and the concentration and ¹³C enrichment of plasma acetate were determined from filtered (10-kDa) samples using ¹H NMR (500.13MHz). ¹H-[¹³C]NMR spectra of plasma samples were also obtained at the conclusion of [2-¹³C]acetate infusions to assess potential ¹³C labeling in glucose.¹³

Metabolic Modeling

The metabolic fluxes were determined by fitting the two-compartment metabolic model (neuron and astroglia) depicted in Figure 1 to the time courses of ¹³C- enriched amino acids measured ex vivo or in vivo from timed or continuous infusions of [2,4-¹³C₂]BHB, respectively. For the awake rat study (ex vivo) the fitted enrichment time courses (target data) included Glu-C4,C3 and Gln-C4 with blood glucose-C1 and blood plasma BHB-C2,C4 used as input data (driver). For the halothane-anesthetized rat study (in vivo), the fitted enrichment data included Glu-C4, Gln-C4, and (Glu+Gln)-C3 (not resolved in vivo), with blood glucose-C1 and brain BHB-C4,C2 used as drivers. Plasma glucose labeling was assumed to be equally labeled at C1 and C6. The metabolic model includes the metabolism of the blood-borne infused ¹³C-labeled substrates, as well as endogenous glucose, all of which contribute to the determined metabolic rates. Mass and isotope flows from [2,4-¹³C₂]BHB into neuronal and glial Glu and Gln pools were written as coupled differential equations in the CWave software package¹⁴ and run in MATLAB 7.0 (Mathworks, Natick, MA, USA). The equations were solved using a first/second-order Runge–Kutta algorithm and data were fitted using a Levenberg–Marquardt algorithm. The entire Gln pool was assigned to the glial compartment, whereas Glu was divided between neuronal (90%) and astroglial (10%) compartments. The iterated rates included neuronal BHB oxidation (V_AcCoA-kbN), pyruvate oxidation (V_pdhN), and α-ketoglutarate (α-KG)/Glu exchange (V_x), astrocytic TCA cycle flux (V_tcaA) and the ratio of BHB oxidation-to-TCA cycle flux, V_AcCoA-kbA/V_tcaA. The calculated rates included V_tcaN (=V_pdhN+V_AcCoA-kbN), V_AcCoA-kbA (=V_AcCoA-kbA/V_tcaA × V_tcaA) and V_cyc (=V_cyc/V_tcaN × V_tcaN). The mass and isotope balance equations used to assess ketone body metabolism are presented in Jiang et al.⁴

Open in a separate window

Figure 1

Schematic depiction of the major metabolic pathways of ¹³C isotopic label flow from [2,4-¹³C₂]BHB and the astroglial substrate, [2-¹³C]acetate. The continuous rate of metabolism of unlabeled glucose in neurons and astrocytes through pyruvate dehydrogenase, as represented by the rates, V_pdhN or V_pdhA, respectively, serves as constant dilution fluxes in these compartments. The cerebral metabolic rate for ketone body utilization (CMR_kb) is related to the rate of acetyl-CoA oxidation from ketone bodies (V_AcCoA-kb) as (CMR_kb)/2=V_AcCoA-kb. Ketone bodies and acetate are transported through membranes from blood to brain by monocarboxylic acid transporter (MCT) proteins. Subscripts on metabolite abbreviations (e.g., Ac₂, αKG₄, Glu₄, GABA₂, Gln₄) refers to the initial (followed) carbon atom position labeled by ¹³C. AcAc, acetoacetate; Ac-CoA, acetyl-Coenzyme A; α-KG, α-ketoglutarate; BHB, β-hydroxybutyrate; GABA, gamma-aminobutyrate; Gln₄, glutamine-C4; Glu₄, glutamate-C4; lac, lactate; OAA, oxaloacetate; pyr, pyruvate; suc, succinate; V_Ac, rate of astroglial acetate oxidation; V_AcCoA-kbN, V_AcCoA-kbA, respective rates of neuronal and astroglial acetyl-CoA oxidation from ketone bodies; V_cyc, rate of glutamate-glutamine neurotransmitter cycle; V_Eff, efflux rate of glutamine; V_GS, rate of glutamine synthesis; V_PC, rate of pyruvate carboxylation; V_tcaN, V_tcaA; respective rates of neuronal and glial TCA cycles.

To improve reliability of flux estimates, a fixed value of the ratio, V_cyc/V_tcaN, was used to constrain the model fits to the ¹³C time courses.^{6, 15} V_cyc/V_tcaN was calculated from the steady-state ¹³C enrichments of brain Glu- and Gln-C4 after a prolonged (~2hours) intravenous infusion of [2-¹³C]acetate in separate groups of 36-hour fasted anesthetized (0.50±0.22, n=6) and awake rats (0.49±0.05, n=4). The neuronal Glu-C4 enrichment was calculated from the measured tissue enrichment and corrected by subtraction of the contributions arising from metabolism of blood plasma glucose and BHB labeling owing to systemic metabolism of [2-¹³C]acetate as described in Jiang et al⁴ and Supplementary Information. The enrichment of brain alanine-C3 was assumed to reflect brain pyruvate and its labeling from plasma glucose and plasma BHB-C4 ¹³C enrichment was used directly. The individual animal values were used as input in fitting the in vivo and ex vivo time courses, rather than their mean, to allow the inter-subject variance in V_cyc/V_tcaN to be reflected in the fitted parameter estimates for each group.

For the condition of PB-induced isoelectricity, where a full time course was deemed impractical owing to the very slow rate of label turnover, the rate of ketone body oxidation was approximated by ¹³C label trapped in the major amino acid pools as represented by Glu, GABA, Asp, and Gln over the 20minutes of [2,4-¹³C₂]BHB infusion according to:

where [Glu], [GABA], [Asp], and [Gln] are the total tissue concentrations, BHB₂ is the plasma ¹³C fractional enrichment of BHB-C2, and Glu_i, GABA_i, Asp_i, and Gln_i are the brain tissue ¹³C fractional enrichments at the specified carbon atom and Δt is the infusion time (20minutes). The fractional enrichments of Glu₂ and Asp₂ (not measured) were assumed to equal Glu₃ and Asp₃, respectively, and are accounted for in the terms, 2Glu₃ and 2Asp₃. This is an expected outcome related to the symmetry of succinate and label scrambling at C2 and C3, which is the precursor of oxaloacetate C2 and C3 (Asp_2,3) and α-KG C2 and C3 (Glu_2,3) in the TCA cycle. In astrocytes the enrichment of Glu-C4 was assumed to be equal to Gln-C4. Tissue amino acid distributions between neurons and astrocytes were assigned respectively^{6, 7} as Glu (90:10), GABA (100:0), Asp (90:10), and Gln (0:100). Note that the sum of the two equations will give the total BHB oxidation independent of the assumed distribution of metabolites. This method of rate approximation, which assumes linearity of the rate of total ¹³C label accumulation in brain amino acids when measured over a sufficiently short time period,¹⁶ will reflect a minimum rate owing to losses from ¹³C-labeled metabolites diffusing away from the sampled tissue (e.g., ¹³CO₂).

Labeling of Cortical Amino Acids from [2,4-¹³C₂]BHB in Awake and Anesthetized Rats

After uptake into the brain from blood [2,4-¹³C₂]BHB is metabolized in neuronal and astroglia mitochondria to two molecules of acetyl-CoA, which enter the TCA cycle labeling α-KG at C4 (first turn) followed by C3, C2, and carbonyls in subsequent TCA cycle turns (Figure 1). Isotope ¹³C/¹²C exchange between mitochondria-derived α-KG and the large cytosolic Glu pool leads to enrichment of Glu-C4 and other carbon positions. Because glucose, the other major substrate supporting the TCA cycle, is unlabeled, the rate of brain amino acid labeling reflects the combined (weighted) flows of the ¹³C-labeled ketone bodies and unlabeled glucose through the TCA cycle. Previous studies have shown that [2,4-¹³C₂]BHB is metabolized predominantly in neurons, reflecting their relatively larger brain volume fraction and TCA cycle rate compared with glial cells, resulting in a pattern of brain amino acid labeling similar to that produced by the oxidation [1-¹³C]glucose.^{3, 4}

Figure 2 depicts representative ¹H-[¹³C]NMR difference spectra of the cortex of rats infused with [2,4-¹³C₂]BHB under different levels of neural activity. In awake animals (Figure 2A) infused with [2,4-¹³C₂]BHB cortical ¹³C labeling of Glu-C4,C3, Gln-C4,C3, GABA-C2,C3,C4, and Asp-C3 rose progressively with infusion time, with Glu and Gln-C4 attaining nearly constant enrichment by ~60minutes. In halothane-anesthetized rats (Figure 2B), the magnitude and pattern of ¹³C labeling from [2,4-¹³C₂]BHB was similar to that seen in awake animals, with extensive initial labeling of Glu-C4, followed by Glu-C3, Gln-C4, and Asp-C3. Figure 2C depicts a ¹H-[¹³C]NMR difference spectrum for a typical PB-isoelectric rat infused with [2,4-¹³C₂]BHB for 20minutes. The ¹³C labeling in Glu-C4,C3, and Gln-C4 appeared similar, at first view, to the other two groups, driven by the higher blood concentration and fractional enrichment of BHB levels in the PB-isoelectric rats. No significant differences (P≥0.2) were seen between awake and PB-isoelectric rats in cortical levels of Glu, Gln, GABA, Asp, alanine, or lactate (Table 2).

Open in a separate window

Figure 2

Stacked plots of ¹H-[¹³C] difference spectra depicting cortical ¹³C metabolite labeling from the infused [2,4-¹³C₂]BHB measured ex vivo in awake rats (A), in vivo in a halothane-anesthetized rat (B), and ex vivo during pentobarbital-induced isoelectricity (C). Intensities were scaled to the total creatine signal. Ala₃, alanine-C3; Asp₃, aspartate-C3; BHB₄, β-hydroxybutyrate-C4; gaba₂, gamma-aminobutyrate-C2; gaba₃, gamma-aminobutyrate-C3; Gln₃, glutamine-C3; Gln₄, glutamine-C4; Glu₃, glutamate-C3; Glu₄, glutamate-C4; Lac₃, lactate-C3; p.p.m., parts-per-million; time (in minutes) appears to the right of spectra.

Table 2

Concentrations and percentage ¹³C enrichments of cortical amino acids and BHB in fasted rats during [2,4-¹³C₂]BHB infusion under awake (100 and 15-minute infusions) and pentobarbital-induced isoelectric (20-minute infusion, PB-isoelectric) rats

	Glutamate	Glutamine	GABA	Aspartate	Alanine	Lactate	BHB
Awake
Conc. (μmol/g) (n=15)	13.39±1.34	7.10±0.63	2.52±0.30	3.90±0.30	0.75±0.14	2.87±0.80	0.57±0.16
100-Minute infusion (n=3)
C4 %FE	39.40±1.91	32.87±2.72	24.72±2.86	—			52.1±11.4
C3 %FE	22.88±1.39	—	16.39 ±1.52	28.36±2.16	11.61±1.47	7.59±0.38
C2 %FE	—		27.94±1.53	—	—	—	—
15-Minute infusion (n=3)
C4 %FE	24.56±0.93	13.99±0.59	6.97±0.20	—			63.0±5.9
C3 %FE	4.42±0.44	—	3.08±0.45	11.43±0.78	2.22±1.22	1.55±0.53
C2 %FE	—	—	15.42±0.45	—	—	—	—
Halothanea
Conc. (μmol/g) (n=6)	11.9±0.3	7.9±1.2	1.4±0.5	3.2±0.8	0.8±0.2	2.8±0.5	0.7±0.3
100-Minute infusion (n=6)
C4 %FE	30.6±2.9	25.0±2.5	—	—			64±7
C3 %FE	20.2±3.3	16.7±4.2	6.90±2.2	14.9±2.7	6.1±1.7	4.5±1.1
C2 %FE	—	—	11.9±2.2	—	—	—	—
Isoelectric
Conc. (μmol/g) (n=4)	12.91±0.81	7.60±0.82	2.01±0.03	3.34±0.26	0.78±0.06	2.96±0.46	1.03±0.44
20-Minute infusion (n=4)
C4 %FE	26.30±2.71	8.81±0.53	0.67±0.09	—			77.0±5.1
C3 %FE	3.58±0.63		0.59±0.48	9.20±1.99	1.48±0.54	1.22±0.62
C2 %FE	—		9.69±1.97	—	—	—

Open in a separate window

BHB, β-hydroxybutyrate; Conc., concentration; GABA, gamma-aminobutyrate. Values reflect mean±s.d. ¹³C enrichments (%) reflect excess ¹³C after subtraction of the 1.1% natural abundance and were divided by the blood plasma [2,4-¹³C₂]BHB enrichment for the respective animal. Dash (—), value not determined. Note that the plasma BHB concentration was 34% higher and infusion time longer in PB-isoelectric (20minutes) compared with awake rats (15minutes); for an approximate comparison of these values for an equal plasma BHB concentration and time, the PB-isoelectric fractional enrichments should be multiplied by (1−0.34) × 15/20 = ~0.50.

^aValues for the halothane group were taken from Table 2 of Jiang et al.⁴

Rates of Cortical BHB and Glucose Oxidation

Individual results for ¹³C enrichment in Glu and Gln at different time points after isotope loading are shown in Figure 3. Metabolic rates of neuronal BHB and glucose oxidation and the Glu–Gln cycle were estimated for the awake rats by fitting a constrained two-compartment neuron-astrocyte metabolic model (Figure 1) to the time courses of Glu-C4,C3, and Gln-C4 ¹³C enrichment constructed from the cortical extract data. Table 3 summarizes the cortical metabolic rate estimates for the three different activity conditions. Figures 3A and B shows these time courses and the best-fits of the metabolic model to the data, along with Monte-Carlo error histograms of the fluxes (Figure 3C). Neuronal acetyl-CoA oxidation from BHB (V_AcCoA-kbN) accounted for 36% of neuronal TCA cycle flux (V_tcaN) in the awake state (V_AcCoA-kbN, 0.40±0.05; V_tcaN, 1.12±0.17μmol/g per minute), the remaining 64% contributed by glucose (pyruvate). A sensitivity analysis was performed (Figure 3D) to assess the influence of the assumption of a fixed value for the pyruvate carboxylase flux (V_PC) on the calculated values of V_AcCoA-kbN and V_pdhN, finding negligible impact (<7%) over a range of ±25% of the nominal value of V_PC.

Open in a separate window

Figure 3

Time courses of ¹³C labeling of cortical glutamate (Glu) and glutamine (Gln) during [2,4-¹³C₂]BHB infusion in awake rats measured ex vivo with best-fits (solid lines) of the metabolic model. (A) Glutamate-C4,C3 percentage ¹³C enrichment verses infusion time. (B) Glutamine-C4 percentage ¹³C enrichment versus infusion time. Data shown reflect fractional enrichments as measured, without prior normalization by plasma ¹³C-BHB enrichment. (C) Monte-Carlo uncertainty distributions for V_AcCoA-kbN and V_pdhN based on 1,000 simulations. (D) Sensitivity of V_AcCoA-kbN and V_pdhN to changes in the assumed rate of astroglial anaplerosis (V_PC) of 25% above or below the nominal value (0.17μmol/g per minute). Within the probable range of V_PC, effects on the neuronal rates were small (<3.5%). BHB, β-hydroxybutyrate; V_AcCoA-kbN, rate of acetyl-CoA oxidation from ketone bodies in neurons; V_cyc, rate of glutamate-glutamine neurotransmitter cycle; V_PC, rate of pyruvate carboxylation via pyruvate carboxylase; V_pdhN, rate of pyruvate oxidation in neurons via pyruvate dehydrogenase; V_tcaN, rate of TCA cycle in neurons.

Table 3

Contributions of ketone bodies and glucose to cortical metabolic rates in fasted and [2,4-¹³C₂]BHB-infused rats under different states of neural activity

	VtcaN	VAcCoA-kbN	VpdhN	VtcaA	VAcCoA-kbA	VpdhA	Vcyc
Awake	1.12 (±0.17)	0.40 (±0.05)	0.72 (±0.13)	0.49 (±0.17)	0.14 (±0.05)	0.35 (±0.12)	0.55 (±0.09)
aAnesthetized (1% halothane)	0.75 (±0.25)	0.42 (±0.14)	0.33 (±0.12)	0.24 (±0.03)	0.06 (±0.04)	0.19 (±0.04)	0.37 (±0.13)
bIsoelectric (pentobarbital)		0.23 (±0.03)			0.04 (±0.003)		~0

Open in a separate window

BHB, β-hydroxybutyrate; V_tcaN, neuronal TCA cycle rate; V_AcCoA-kbA; rate of astroglial acetyl-CoA oxidation from BHB; V_AcCoA-kbN, rate of neuronal acetyl-CoA oxidation from BHB; V_cyc, rate of the glutamate-glutamine cycle; V_pdhA, rate of astroglial pyruvate oxidation through pyruvate dehydrogenase; V_pdhN, rate of neuronal pyruvate oxidation through pyruvate dehydrogenase; V_tcaA, astroglial TCA cycle rate.

^aRe-analysis of data reported by Jiang et al;⁴ s.d.'s of fluxes reported for halothane (n=6) and PB-isoelectric (n=5) animals reflect the inter-animal, group s.d., whereas for awake animals, the s.d. was derived by Monte-Carlo simulation.

^bFlux under isoelectric conditions estimated from ¹³C labeling in glutamate, GABA, aspartate, and glutamine carbons at 20minutes as described in Materials and Methods.

In animals previously measured in vivo under halothane anesthesia,⁴ we refitted the same model used in the analysis of the awake data to the time courses from this study finding that ketone body oxidation accounted for more than half (55%) of neuronal TCA cycle flux under anesthesia (V_AcCoA-kbN, 0.37±0.08; V_tcaN, 0.66±0.12μmol/g per minute; Table 3).

Because glucose metabolism is substantially slowed under PB-isoelectric conditions,^{5, 17} rendering the time to attain an isotopic steady state impractical to achieve, the initial rates of ketone body oxidation in neurons and astrocytes were estimated from ¹³C accumulated in the major amino acids during the 20-minute infusion (see Materials and Methods; Patel et al.⁶) Using this approach V_AcCoA-kbN and V_AcCoA-kbA at isoelectricity were 0.23±0.03μmol/g/min and 0.04±0.01μmol/g/min, respectively. Because full time courses and/or steady-state enrichments were not measured, V_tcaN could not be determined directly. To ensure that the initial rate estimate based on label trapping (PB-isoelectric condition) and model-derived estimates of V_AcCoA-kbN gave similar values, justifying between-group comparisons, we also estimated V_AcCoA-kbN for awake rats using equation (1) by summation of ¹³C concentrations (in μmol/g) of detected carbon atoms of Glu, GABA, and Asp measured at 15minutes, which provides a lower bound estimate of the true rate (e.g., see Lu et al¹⁶ for a discussion of label trapping using brief ¹⁴C-glucose infusion) and that derived by modeling of the full turnover curve. Using this approach, V_AcCoA-kbN was 0.36μmol/g per minute, slightly lower than the model-derived rate of 0.40μmol/g per minute.

Evidence for Differing Roles of Glucose and Ketone Bodies in the Support of Neural Activity

Although ketone bodies are readily oxidized in the brain, previous studies have shown that their uptake (unlike glucose) is not increased during neuronal activation. For example, acoustic stimulation in unanesthetized rats results in increased uptake of ¹⁴C-labeled glucose but not ¹⁴C-labeled BHB in auditory sensory regions.¹⁸ During early postnatal development ¹⁴C-BHB uptake remains relatively uniform throughout the brain, in contrast to ¹⁴C-glucose uptake which is highly heterogeneous, and does not follow the postnatal age-dependent pattern of increased glucose utilization in sensory regions.^{19, 20, 21, 22, 23} This is consistent with in vitro findings using adult brain slices, which show that ketone bodies alone do not support evoked neuronal firing.²⁴ Together, these observations suggest that ketone bodies do not support oxygen consumption associated with increased functional demand in the adult brain. Significantly, in cultured cerebellar neurons, which are mostly glutamatergic, TCA cycle activity and the release of Glu during depolarization is supported by glucose²⁵ but not BHB.²⁶ These results, in combination with our in vivo findings, support a potential role for glucose in glutamatergic neurotransmission that cannot be replaced by ketone bodies.

Comparison to Previous Studies of Awake and Anesthetized Rats

Our finding that BHB supported a decreasing fraction of cortical neuronal oxidation with increasing neural activity is supported by previous studies of awake and anesthetized hyperketonemic rats. In awake rats fasted for 2 days ([BHB]_plasma ≈1.62mM) brain consumption of ketone bodies (BHB and acetoacetate (AcAc)) measured by arteriovenous differences accounted for 13.4% of total oxidation. An autoradiography study of 2-day fasted awake rats ([BHB]_plasma=1.27mM) using 3-[¹⁴C]BHB²⁷ found a rate of ~0.05 to 0.06μmol/g per minute for the combined frontal/parietal cortex. In the awake hyperketonemic rats ([BHB]_plasma=9.1mM) of our study, the sum of neuronal and astroglial ketone oxidation, V_{AcCoA-kb(N+A),} accounted for ~33% of total acetyl-CoA oxidation. Assuming the rate of ketone oxidation scales proportionately with plasma BHB concentration, V_{AcCoA-kb(N+A)} would be 0.06 to 0.12μmol/g per minute (or ≈4% to 8% of acetyl-CoA oxidation) for [BHB]_plasma of 1 to 2mM, which is in reasonable agreement with prior studies.

In contrast to awake rats, studies of barbiturate anesthetized and hyperketonemic (48-hour fast) adult rats consistently found higher ketone body contributions to total brain substrate oxidation, with values of 18% and 29% of total oxidation for [BHB]_plasma of 0.64mM and 2mM, respectively.^{28, 29} In a study that compared 24-hour and 48-hour fasted rats under PB anesthesia, including an acute infusion of DL-BHB in 24-hour fasted rats ([BHB]_plasma=0.83, 2.05, and 5.52mM, respectively), Ruderman et al³⁰ found that ketone bodies accounted for 28%, 33%, and 68% respectively of total acetyl-CoA oxidation. For the halothane-anesthetized rats in the present study ([BHB]_plasma=6.6mM), 48% of acetyl-CoA oxidation was fueled by ketone bodies, which is in reasonable agreement with the above studies.

Neuroenergetics of Glutamate Neurotransmission During Hyperketonemia

Our findings strongly suggest that the neuronal TCA cycle in isoelectric cortex was fully supported by ketone bodies in the hyperketonemic rats at the blood BHB level of ~12mM, despite their euglycemic state. This finding indicates that ketone bodies as fuels can provide complete energetic support of basal (non-signaling) processes in neurons when available in sufficient concentration. At the higher activity level (awake condition), ketone body oxidation was submaximal at the plasma levels achieved in our initial measurements (9.1mM), but rate saturation was observed for [BHB]_plasma >17mM, accounting for ~62% of V_tcaN (Figure 6), or ~59% of total substrate oxidation of neurons and astrocytes. Significantly, the peak rate of ketone body utilization, 1.75 × V_{AcCoA-kb(N+A)}/2=0.47μmol/g per minute, is approximately one-third of the V_max-like activity of BHB dehydrogenase (~1.2 to 1.4μmol/g per minute after correction to a physiologic temperature of 37°C from 25°C assuming Q₁₀=2), the least enzymatically active of the three pathway enzymes supporting ketone body oxidation,^{31, 32, 33, 34} as measured in rat cortex homogenate in vitro.³¹ Because brain AcAc concentration determines the rate of ketone body oxidation,³¹ and brain BHB levels at the rate saturation of ~1.7mM (Figure 5) were comparable with the in vitro K_m of BHB dehydrogenase for BHB (~2mM),^{35, 36} the saturation of this rate is not readily explained, but compartmentation and mitochondrial transport, or physiologic concentrations of cofactors (e.g., NAD⁺ (nicotinamide adenine dinucleotide)) might be involved. Interestingly, the estimated maximum rate of ketone body oxidation in the awake rat cortex of ~59% is similar to the global average of 58% to 60% measured by arteriovenous difference under conscious sedation in three obese subjects after 38 to 41 days of starvation,³⁷ extreme adaptive conditions where transport is not likely to be limiting for brain ketone consumption.

Open in a separate window

Figure 6

Estimated fraction of total neuronal oxidation supported by ketone bodies and glucose for saturating levels of ketone bodies under pentobarbital (PB)-isoelectric and awake conditions. V_AcCoA-kbN, rate of neuronal acetyl-CoA utilization from ketone bodies; V_pdhN, neuronal pyruvate dehydrogenase flux (i.e., rate of acetyl-CoA utilization from glucose).

Implications of Ketone body Oxidation for Models of Neuroenergetics Involving Glucose

The maximum rate of ketone body oxidation by the brain provides a test of certain bioenergetics models of glutamatergic neurotransmission involving glucose metabolism. By fueling cortical energy needs with ketone bodies to the maximum extent possible, neuronal processes not strictly dependent on glucose oxidation were effectively removed, giving the relationship between ΔV_pdhN and ΔV_cyc that could be compared with model predictions. This is possible only under the saturating hyperketonemic conditions, a distinction not apparent under non-fasted or glucose-infused conditions where V_pdhN≈V_tcaN.

Different mechanistic models have been proposed to explain the experimentally observed relationship between ΔV_tcaN and ΔV_cyc including one that couples astroglial glycolytic ATP to extracellular Glu/Na⁺ uptake and Gln synthesis,¹⁰ and another that relates elements of the malate-Asp shuttle in neurons to formation of neurotransmitter Glu from Gln via Gln–Glu cycling.^{25, 26, 38} In the models involving obligatory coupling of glial glycolysis (and transfer of the lactate produced by glycolysis to the neurons), the predicted relationship between ΔV_pdhN and ΔV_cyc ranges from 1:1 to 2:1, depending on whether Gln synthesis also requires glycolytic ATP, or the extent to which glial glucose oxidation is altered by activity.³⁹ An alternative model of Glu neurotransmitter synthesis from astroglial Gln was described in Hertz et al,³⁸ and more recently by Lund et al,²⁶ which would produce a flux ratio of 1:1 between ΔV_pdhN and ΔV_cyc (i.e., 0.5:1 between neuronal glucose oxidation and glutamate cycling). In this model the synthesis of neurotransmitter Glu from Gln is indirect, with mitochondrial formation of α-KG from Gln (deamidation by glutaminase in the mitochondrial membrane, Glu entry into the mitochondrial matrix by exchange with Asp, and transamination to α-KG), efflux of α-KG to cytoplasm in exchange with malate, and transamination of α-KG back to Glu. Because the transfer of reducing equivalents to mitochondria via malate is stoichiometrically related to the formation of NADH from glucose (1 glucose: 2 pyruvate: 2 NADH: 2 malate), one transmitter Glu produced from Gln will correlate with the flow of one reducing equivalent per pyruvate molecule produced from glucose, leading to a relationship between ΔV_pdhN and ΔV_cyc of 1:1 in the nerve terminal. This ratio would be ~1.2-to-1 if astrocytic de novo Glu synthesis (~15–30% of V_cyc) replaces Glu oxidized in neurons. Notably, under saturating conditions of ketone use, glucose need only provide one molecule of pyruvate for de novo synthesis of one molecule of Glu (ketone bodies can provide the acetyl-CoA) and one molecule of NADH (malate) for exchange-mediated efflux of α-KG from mitochondrial matrix to cytoplasm for Glu synthesis, giving a ratio of 1:1 between ΔV_pdhN and ΔV_cyc. Thus, in this model the flux ratio as it relates to neurons would reflect glucose metabolic support (in the form of reducing equivalents from malate) of Glu transmitter synthesis/release involving a ‘pseudo' malate-Asp shuttle mechanism,³⁸ which would not be involved in the oxidation of ketone bodies. An important role for neuronal glycolysis as a source of the reducing equivalents is supported by our recent findings of significant activity-dependent glucose phosphorylation in nerve terminals in vivo.⁴⁰ For the awake rat cortex oxidizing ketone bodies at the maximum (saturating) rate of ~0.7μmol/g per minute, V_pdhN would be 1.12−0.70=0.42μmol/g per minute. With V_cyc=0.55μmol/g per minute, V_pdhN:V_cyc ~0.76:1, which is lower than the 2:1 ratio found between ΔV_tcaN and ΔV_cyc, reflecting the substantial contribution of BHB to total neuronal metabolism under saturating conditions.

Because total glucose utilization was not measured in the present study, we cannot say whether the reduction in V_pdhN with increasing V_AcCoA-kbN corresponded to reduced or unchanged glucose utilization (e.g., net lactate formation followed by efflux). Reduced brain glucose utilization and increased lactate efflux is seen during prolonged fasting-induced hyperketonemia in humans³² and rats,²⁷ suggesting that a larger glycolytic component than we estimate based on pyruvate oxidation during maximum ketone use, cannot be fully excluded.

The authors thank Xiaoxian Ma and Bei Wang for preparation of the animals, Dr. Graeme Mason for providing the CWave modeling software, and Dr. Peter Herman for his help in implementing the electrophysiological recordings.

Author Contributions

GMIC performed the ex vivo experiments, analyzed the data, and significantly contributed to preparation of the manuscript and study design. LJ performed the in vivo experiments. DLR contributed to study design and data interpretation. KLB conceived the study, coordinated experiments, analyzed and interpreted the data, and wrote the manuscript.