Glutamatergic Neurotransmission and Neuronal Glucose Oxidation are Coupled during Intense Neuronal Activation

Abstract

¹³C nuclear magnetic resonance (NMR) experiments have previously shown that glutamatergic neurotransmitter flux (V_{cycle(Glu/Gln)}) changes proportionately with neuronal glucose oxidation (CMR_glc(ox)N) in the nonactivated cortex of anesthetized rats. Positron Emission Tomography measurements of glucose and oxygen uptake during sensory stimulation had shown that the incremental glucose utilization is greater than oxygen leading to the suggestion that the energy required for stimulated neuronal activity arises from nonoxidative glucose metabolism. In this study, the authors used spatially localized ¹H-observed, ¹³C-edited NMR spectroscopy during an infusion of [1,6–¹³C₂]glucose to assess the relationship between changes in V_{cycle(Glu/Gln}) and glucose utilization (CMR_glc(ox)N and CMR_glc(nonox)) during the intense cortical activity associated with bicuculline-induced seizures. Metabolic fluxes were determined by model-based analysis of the ¹³C-enrichment time courses of glutamate-C4 and glutamine-C4 (CMR_glc(ox)N, V_{cycle(Glu/Gln)}) and lactate-C3 (CMR_glc(nonox)). The exchange rate between α-ketoglutarate and glutamate was found to be significantly faster than TCA cycle flux both for control (41 μmol·g⁻¹·min⁻¹; 95% CI, 5 to 109 μmol·g⁻¹·min⁻¹) and during seizures (21 μmol·g⁻¹·min⁻¹; 95% CI, 4.4 to 51.8 μmol·g⁻¹·min⁻¹). During seizures, total glucose utilization (CMR_{glc(ox+nonox)}) increased substantially (466% between 0 and 6 minutes; 277% between 6 and 55 minutes). Glucose oxidation (CMR_glc(ox)N) also increased (214%; from 0.26 ± 0.02 to 0.57 ± 0.07 μmol·g⁻¹·min⁻¹) but to a lesser degree, resulting in a large increase in cortical lactate concentration. V_{cycle(Glu/Gln)} increased 233% (from 0.22 ± 0.04 to 0.52 ± 0.07 μmol·g⁻¹·min⁻¹), which was similar to the increase in glucose oxidation. The value of V_{cycle(Glu/Gln)} and CMR_glc(ox)N obtained here lie on the line predicted in a previous study. These results indicate that neuronal glucose oxidation and not total glucose utilization is coupled to the glutamate/glutamine cycle during intense cortical activation.

Nuclear magnetic resonance (NMR) spectroscopy with ¹³C isotopes has provided strong evidence for the coupling of the glutamate/glutamine cycle to glucose energetics in the cerebral cortex in vivo (Magistretti et al., 1999; Sibson et al., 1998. 1999). These studies have shown that the glutamate/glutamine cycle flux (V_{cycle(Glu/Gln)}) is comparable in magnitude to the neuronal cerebral metabolic rate of glucose oxidation (≈80% of CMR_glc(ox)N) in the cortices of resting humans (Shen et al., 1999) and of anesthetized rats (Sibson et al., 1997, 1998, 2001). In anesthetized rats V_{cycle(Glu/Gln)} and CMR_glc(ox)N change proportionately (≈1:1 ratio) over a substantial range of neuronal activity above isoelectricity (Sibson et al., 1998). The observed relationship is consistent with the calculations of the energetics of specific ion flows associated with glutamatergic neurotransmission (Attwell and Laughlin, 2001). The proportionality between V_{cycle(Glu/Gln)} and CMR_glc(ox)N supports a proposed coupling mechanism (Magistretti et al., 1999; Pellerin and Magistretti, 1994), in which glucose uptake is stoichiometrically coupled to glutamate neurotransmitter cycling between neurons and astrocytes.

To date, quantitative measurements of V_{cycle(Glu/Gln)} as a function of CMR_glc(ox)N have only been reported for the nonactivated cortex of the anesthetized rat (Sibson et al., 1998), although they are consistent with results in the resting human (Shen et al., 1999). It has been long known that the ratio of oxygen to glucose uptake (CMRo₂/CMR_glc) is lower than the theoretically expected value of 6:1 for the complete oxidation of glucose (Siesjo, 1978) during increased neuronal activity associated with seizures. Measurements of glucose and oxygen uptake using Positron Emission Tomography (PET) in human subjects during sensory stimulation also show a greater consumption of glucose than oxygen (Fox and Raichle, 1986; Fox et al., 1988; Mintun et al. 2002). Consistent with these findings, an increase in lactate was observed in the human visual cortex during visual stimulation using ¹H magnetic resonance spectroscopy (Prichard et al., 1991). These earliest results, which showed the largest degree of uncoupling, led to the proposal that the energy required for neuronal activity during stimulation above awake resting levels arises primarily from nonoxidative glucose metabolism (Fox et al., 1988; Gjedde, 1993). Because there is considerable electrical activity in the awake resting state (Sibson et al., 1998; Shulman and Rothman, 1998) where there is quantitative coupling between V_{cycle(Glu/Gln)} and CMR_glc(ox)N, this proposal implies that a different energetic mechanism must operate in activated states.

The present study investigated whether the observed proportionality between ΔV_{cycle(Glu/Gln)} and ΔCMR_glc(ox)N in the nonstimulated state holds during intense neuronal activation. We used spatially localized ¹H-observed, ¹³C-edited (POCE) NMR spectroscopy (de Graaf et al., 2003) with an intravenous infusion of [1,6–¹³C₂]glucose to determine V_{cycle(Glu/Gln)} and CMR_glu(ox)N in the rat cortex during bicuculline-induced seizures. Metabolic fluxes were determined by fitting a two-compartment (neuron–astroglia) metabolic model to the time courses of the ¹³C-labeled metabolites (Mason et al., 2003). We found that the relationship between ΔV_{cycle(Glu/Gln)} and ΔCMR_glu(ox)N reported previously in the nonstimulated state also applies during intense neuronal activation. Our results show that oxidative rather than nonoxidative glycolysis is coupled to the glutamate/glutamine cycle during intense neuronal activation. Thus, changes in V_{cycle(Glu/Gln)} and CMR_glc(ox)N during intense cortical activation follows the same relationship as reported previously in the nonactivated cortex, providing strong evidence for a singular coupling mechanism between these fluxes at the molecular level.

MATERIALS AND METHODS

Animal preparation

Animal experiments were performed in accordance with protocols approved by the Yale Animal Care and Use Committee. Wistar rats (160 to 180 g), fasted overnight, were anesthetized with 2% to 3% halothane in 30% O₂ / 67% to 68% N₂O, tracheotomized, and ventilated. The left femoral artery was cannulated for continuous monitoring of arterial blood pressure and intermittent sampling of blood for the measurement of glucose and gases. Both femoral veins were cannulated for the infusion of [1,6–¹³C₂]glucose and administration of bicuculline or saline. Body temperature was maintained near 37°C using a heating pad connected to a temperature-regulated circulating water bath. A pair of carbon fiber EEG leads was positioned in the scalp for monitoring seizure activity using a Grass polygraph in bicuculline-treated animals. The output of the amplifier was digitized using an analog-to-digital converter interfaced to a personal computer and monitored using LabView (National Instruments, Austin, TX, U.S.A.). After surgery, halothane was reduced to 1% to sustain an adequate blood pressure.

Animals were placed in a small plastic cradle and a surface coil was placed on the scalp. A layer of Saran Wrap was placed between the coil and scalp. The cradle/probe was inserted into the magnet and field homogeneity was optimized. After the acquisition of baseline spectra, an intravenous infusion of [1,6-¹³C₂]glucose (99 atom %; Cambridge Isotopes, Andover, MA, U.S.A.) was begun using a protocol described previously (Fitzpatrick et al., 1990). [1,6-¹³C₂]Glucose was dissolved in water and administered as a bolus (48 mg in 0.27 mL in 15 seconds) followed by a manual stepped reduction in infusion rate every 30 seconds for the next 8 minutes, whereupon the rate was constant until the time of in situ brain freezing. This protocol raises plasma glucose rapidly (<1 minute) and maintains a nearly constant level and enrichment thereafter.

Before induction of seizures, animals were immobilized with D-tubocurarine chloride (initial dose of 0.25 mg/kg, intraperitoneally with supplements of 0.10 mg/kg every 40 minutes). Blood was withdrawn (1.5 mL) into heparinized saline 2 minutes after the beginning of the glucose infusion in order to limit the peak rise in arterial blood pressure during the seizures and re-infused as needed to maintain blood pressure. Seizures were induced by injection of bicuculline (1 mg/kg dissolved in water, intravenously) 4 minutes after beginning the ¹³C-glucose infusion. The glucose infusion preceded the injection of bicuculline by 4 minutes in order to prelabel the brain pools of extra cellular and intracellular glucose to minimize label dilution in pyruvate by rapid metabolism of the initially unlabeled glucose and glycogen to lactate. Additional bicuculline was administered, as needed, to maintain electrographic seizure activity. An equivalent volume of saline was injected at 4 minutes of [1,6–¹³C₂]glucose infusion in control animals. The [1,6–¹³C₂]glucose was infused for 55 minutes in rats receiving bicuculline and for approximately 85 minutes in rats receiving saline. A shorter experimental time was used for bicuculline-treated rats because steady state labeling was achieved more quickly than in control animals and the physiologic condition can deteriorate during extended status epilepticus seizures. In addition, control and bicuculline-treated rats were also infused with [1,6–¹³C₂]glucose on the bench for selected time points to estimate the exchange rate (V_x) between α-ketoglutarate and glutamate. Arterial blood samples were taken periodically for the analysis of plasma glucose concentration and ¹³C percent enrichment. At the end of the experiment, the brain was frozen in situ in liquid nitrogen while mechanical ventilation was continued (Pontén et al., 1973).

In vivo nuclear magnetic resonance spectroscopy

In vivo experiments were performed on a 7 T horizontal bore Bruker magnet interfaced to a Bruker AVANCE spectrometer (Bruker Instruments, Billerica, MA, U.S.A.) equipped with a 12 cm diameter actively shielded gradient coil insert (190 mT/m in 200 μs). The operating frequencies for ¹H and ¹³C were 300.30 and 75.5 MHz, respectively. POCE NMR spectra were acquired using a 13-mm-diameter ¹H surface coil and two orthogonal 21-mm-diameter ¹³C coils driven in quadrature (de Graaf et al., 2003). The exact position of the spectroscopic volume-of-interest was determined by acquiring three orthogonal fast gradient-echo images of the brain (Haase et al., 1986). Automated adjustments of first and second order shims within a volume of 7 × 7 × 7 mm³ using the noniterative FASTMAP algorithm (Gruetter, 1993) resulted in water ¹H line widths of 11 to 13 Hz. Three-dimensional localization was achieved with a combination of outer-volume suppression, image selective in vivo spectroscopy, and slice-selective signal excitation techniques. Water suppression was achieved using SWAMP (de Graaf and Nicolay, 1998) with adiabatic modulated pulses. POCE NMR spectra were obtained every 45 seconds from a localized volume of 7 × 4 × 7 mm³.

Processing of in vivo nuclear magnetic resonance data

Free induction decays were zero-filled to 8,192 data points, apodized (2.0 Hz Gaussian line broadening), Fourier transformed, and phase corrected. The POCE difference spectrum was obtained by subtracting the ¹³C-inverted spectrum (¹²C-¹³C) from the noninverted (¹²C + ¹³C) spectrum leaving only the 2 x ¹³C-coupled ¹H resonances. Spectra were summed over 3-minute intervals to increase the signal-to-noise ratio. Resonance intensities were measured using the LCmodel algorithm (de Graaf et al., 2003; Provencher, 1993), which was used to determine the percent ¹³C enrichments and the concentration of metabolites.

Analysis of plasma glucose

Glucose concentration (Glucose Analyzer II, Beckman Instruments) and isotopic ¹³C enrichment were determined in the plasma from arterial blood samples obtained periodically during the ¹³C-glucose infusion. For NMR measurements, the plasma was mixed with D₂O and large macromolecules were removed by passage of the samples through a centrifugal filter (10-kd cut off, Nanosep, Gelman Laboratory, MN, U.S.A.). The percentage ¹³C enrichment of plasma glucose C1α was determined from the fully relaxed ¹H NMR spectrum obtained at 11.7 T AM-500 (Bruker AVANCE NMR spectrometer, Bruker Instruments). The enrichment of glucose C1α at 5.23 ppm was calculated by dividing the intensities of the two ¹³C satellites by the total (¹²C + ¹³C) intensity (Kalderon et al., 1989).

Preparation of cortical extracts and separation of amino acids

Metabolites were extracted from frozen frontoparietal cortical tissue (150 to 200 mg) as previously described (Patel et al., 2001). Briefly, the frozen tissue was pulverized with ice-cold 0.1 mol/L HCl in methanol (1:2 wt/vol) in a dry-ice/ethanol bath and transferred to a wet-ice bath. A known quantity of [2–¹³C]glycine (50 μL of 10 mmol/L solution) was added as an internal concentration reference to correct for losses during the extraction procedure. The tissue powder was homogenized with ice-cold ethanol (1:6 wt/vol; 60% ethanol: 40% deionized water) until no visible pieces remained. The homogenate was clarified by centrifugation (20,000g). The supernatant was lyophilized, and re-suspended in 500 μL of a phosphate-buffered (100 mmol/L, pH 7) deuterium oxide (Cambridge Isotopes) solution containing 0.5 mmol/L TSP (3-trimethylsilyl[2,2,3,3-D₄]-propionate), which served as a chemical shift reference. After quantification of total amino acid levels in the tissue extracts using POCE NMR, amino acids were separated using AG 1-X8 anion exchange resin (200 to 400 mesh, 1.4 mL bed volume; Bio-Rad Laboratories, Hercules, CA, U.S.A.). Glutamine and GABA were eluted with de-ionized water and glutamate was eluted with 2 mol/L acetic acid (Patel et al., 2001). The pH of both fractions was reduced to approximately 3 before lyophilization to reduce the concentration of acetic acid. Both fractions were lyophilized, and resuspended in phosphate-buffered deuterium oxide containing 0.5 mmol/L TSP for further NMR analysis.

Nuclear magnetic resonance analysis of cortical extracts

Fully relaxed POCE NMR spectra of extracts were acquired at 11.7 T (Fitzpatrick et al., 1990) using a pulse sequence that incorporated adiabatic pulses for ¹H and ¹³C excitation (de Graaf et al., 2003). The concentrations of metabolites were determined relative to known concentration of [2–¹³C]glycine added during tissue extraction as an internal reference. The isotopic ¹³C enrichments of C3 and C4 carbons of glutamate and glutamine were calculated from the ratio of the areas of these resonances in the POCE difference spectrum (¹³C only) and the nonedited spectrum (¹²C + ¹³C) of the separated glutamate and glutamine fractions.

Estimation of the exchange rate between α-ketoglutarate and glutamate

Isotope exchange between mitochondrial α-ketoglutarate and cytosolic glutamate is the major pathway for ¹³C labeling of brain glutamate during oxidative metabolism of ¹³C-labeled glucose. The rate of exchange (V_x) in relation to the TCA cycle flux (V_TCA) affects the rate at which label appears at glutamate-C4 and C3 and can influence the apparent value of TCA cycle flux depending on the value of the ratio (V_x/V_TCA) (Gruetter et al., 2001; Mason et al., 1992). We did not have sufficient spectral resolution in the current study for accurate quantification of glutamate and glutamine-C3 owing to extensive overlap (Fig. 3, 4). Therefore, the value of V_x was determined from an ex vivo ¹³C-labeling experiment whereby the carbon-13 atom enrichments were determined in tissue extracts after complete chromatographic separation of glutamate and glutamine. Cortical extracts were prepared from control and bicuculline-treated rats receiving an infusion of [1,6–¹³C]glucose for three different periods (control group: 7, 17, and 85 minutes; seizure group: 7, 17, and 55 minutes; for 7 and 17 minutes, n = 3 per time point). The last time point in each group (n = 6 each) represents the terminal point from the in vivo experiments. The ex vivo measured percent ¹³C enrichments, which consisted of glutamate-C4 and C3, glutamine-C4 and C3, and aspartate-C3, were normalized by the corresponding end-point plasma glucose-C1 as follows:

$${\mathrm{F}\mathrm{E}}_{\mathrm{n}\mathrm{o}\mathrm{r}\mathrm{m}}=({\mathrm{F}\mathrm{E}}_{\mathrm{a}\mathrm{a}}/{\mathrm{F}\mathrm{E}}_{\mathrm{g}\mathrm{l}\mathrm{c}})\times 100$$

(1)

where FE_norm and FE_aa are the normalized and measured percent ¹³C enrichments of the different carbon atom positions, and FE_glc is the end-point plasma glucose-C1 % ¹³C enrichment. A two-compartment metabolic model (neuron and astroglia) depicted in Fig. 1 and Table 1 (Gruetter et al., 2001; Mason et al., 2003; Shen et al., 1999; Sibson et al., 2001) was fitted simultaneously to all the measured ¹³C-enrichment time courses. The input function was the mean blood glucose-C1 enrichment. Best fits of the model to the experimental data were determined using a simulated annealing algorithm (Press et al., 1992) with iteration of V_TCAn, V_x, V_Gln, and V_dilLac. The uncertainty in V_x and other parameter estimates was determined by Monte-Carlo simulation as described by Mason et al. (1995). Briefly, 100 simulated noisy data sets were created from the least squares fit by adding random noise with the same Gaussian distribution as observed in the original data. For Monte-Carlo simulations of the GluC4, GlnC3 and AspC3 data sets with either V_x = 41 and Vx = 1, the random noise standard deviations were 6.14, 8.61, 5.24 and 8.66, 6.29, 6.66 units of percent ¹³C enrichment, respectively. Each of the simulated data sets was analyzed using the same fitting procedure as performed with the original data and the 100 sets of fitted parameters recorded.

Fig. 1.

Schematic diagram of the major pathways of ¹³C isotopic labeling of cerebral metabolites from [1,6–¹³C₂]glucose. The ¹³C-labeled glucose is metabolized in neurons and astroglia to lactate-C3 (Lac C3) / pyruvate-C3 (Pyr C3) by glycolysis and enters their respective TCA cycles through acetyl-CoA-C2 (AcCoA C2). AcCoA C2 labels α-ketoglutarate-C4 (α-KG C4) and, through fast enzyme-mediated exchange (V_x), glutamate-C4 (Glu C4). Glutamate released from glutamatergic neurons is transported into astroglia and converted to glutamine-C4 (Gln C4) by glutamine synthetase. Gln C4 is released from astroglia and transported back to neurons where hydrolysis to Glu C4 occurs by phosphate-activated glutaminase. Further metabolism of α-KG C4 (Glu-C4) in the TCA cycle labels C2 and C3 of Glu and Gln (not shown). AcCoA C2, acetylCoA-C2; Gln C4, glutamine-C4; Glu C4, glutamate-C4; α-KG C4, α-ketoglutarate-C4; Lac C3, lactate-C3; Pyr C3, pyruvate-C3; V_cycle, V_{cycle(Glu/Gln)}; V_TCAa, Astrocytic TCA cycle flux; V_TCAn, neuronal TCA cycle flux.

Table 1.

Coupled differential equations used for the modeling of the ¹³C time courses

Mass balance equations

d(Glcbrain)/dt = Vmaxin(Glcblood)/(Kmin + Glcblood) – Vmaxout(Glcbrain)/(Kmout + Glcbrain) – CMRglc

d(Glun)/dt = (dGlua)/dt = d(KGn)/dt = d(KGa)/dt = d(Aspn)/dt = d(Aspa)/dt = d(OAAn)/dt = d(OAAa)/dt = 0

Isotope balance equations

d(GlcbrainC1*)/dt = Vmaxin(Glcblood,C1*)/(Kmin + Glcblood) – Vmaxout(Glcbrain,C1*)/(Kmout + Glcbrain) – (CMRglc)(Glcbrain,C1*/Glcbrain)

d(L3*)/dt = CMRglc(Glcbrain,C1*/Glcbrain) + VdilLac(0) – (VTCAn + Vpdha + Vana + VdilLac)(L3*/L)

d(Glun4*)/dt = Vcycle(Gln4*/Gln) + Vx(KGn4*/KGn) – (Vcycle + Vx)(Glun4*/Glun)

d(Glua4*)/dt = Vcycle(Glun4*/Glun) + (Vx + Vana)(KGa4*/KGa) – (VGln + Vx)(Glua4*/Glua)

d(Gln4*)/dt = VGln(Glua4*/Glua) + VdilGln(0) – (Vcycle + Vefflux) (Gln4*/Gln)

d(KGa4*)/dt = Vpdha(L3*/L) + Vx(Glua4*/Glua) – (Vpdha + Vx)(KGa4*/KGa)

d(KGa3*)/dt = Vx(Glua3*/Glua) + VTCAa(OAAa2*/OAAa) – (Vpdha + Vx)(KGa3*/KGa)

d(Glua3*)/dt = Vcycle(Glun3*/Glun) + (Vx + Vana)(KGa3*/KGa) – (VGln + Vx)(Glua3*/Glua)

d(Gln3*)/dt = VGln(Glua3*/Glua) + VdilGln(0) – (Vcycle + Vefflux) (Gln3*/Gln)

d(Glun3*)/dt = Vcycle(Gln3*/Gln) + Vx(KGn3*/KGn) – (Vcycle + Vx)(Glun3*/Glun)

d(KGn4*)/dt = VTCAn(L3*/L) + Vx(Glun4*/Glun) – (Vx + VTCAn)(KGn4*/KGn)

d(KGn3*)/dt = Vx(Glun3*/Glun) + VTCAn(OAAn2*/OAAn) – (Vx + VTCAn)(KGn3*/KGn)

d(Aspn3*)/dt = Vx[(OAAn2*/OAAn) – (Aspn3*/Aspn)]

d(Aspa3*)/dt = Vx[(OAAa2*/OAAa) – (Aspa3*/Aspa)]

d(OAAn2*)/dt = Vx(Aspn3*/Aspn) + 0.5 VTCAn[(KGn4*/KGn) + (KGn3*/KGn)] – (Vx + VTCAn)(OAAn2*/OAAn)

d(OAAa2*)/dt = 0.5 (Vpdha – Vana)((KGa4*/KGa) + (KGa3*/KGa)) + Vx(Aspa3*/Aspa) + 0.5Vana(L3*/L) – (Vx + VTCAa)(OAAa2*/OAAa)

Neuronal and astroglial pools of KG and OAA were assumed to be 0.1 μmol/g, as used in the previous study (Gruetter et al., 2001; Mason et al., 1992).

Glc, glucose; L, lactate; Glu, glutamate; Gln, glutamine; Asp, aspartate; OAA, oxaloacetate; KG, α-ketoglutarate; V_TCA, TCA cycle flux; V_ana(= V_efflux), anaplerotic flux; V_Gln, glutamine synthesis rate; V_cycle, glutamate/glutamine cycle flux; V_x, exchange rate between α-KG and glutamate. Subscript “n” and “a” stand for neuronal and glial, respectively. Numeric subscripts represent the position of ¹³C in the corresponding molecule.

Determination of metabolic fluxes in vivo

The metabolic fluxes were determined by fitting a comprehensive two-compartment (neuron and astroglia) metabolic model to the in vivo time course of labeling of glutamate-C4 and glutamine-C4. The metabolic model is depicted schematically in Fig. 1 and essentially follows previous reports (Gruetter et al., 2001; Mason et al., 2003; Shen et al., 1999; Sibson et al., 2001). Mass and ¹³C isotope flows from [1,6–¹³C]glucose to the major amino acids (glutamate, glutamine, and aspartate) in neurons and astroglia were expressed as a series of coupled differential equations (Table 1) and solved using a Runge-Kutta algorithm (Burden et al., 1981). The analysis was implemented using the CWave software package (Mason et al., 2003). The measured resonance intensities in POCE spectra were normalized to the final spectrum and converted to percent ¹³C enrichment according to the following equation:

$${\mathrm{F}\mathrm{E}}_{\mathrm{t}}=({\mathrm{I}}_{\mathrm{t}}/{\mathrm{I}}_{\mathrm{f}})\times {\mathrm{F}\mathrm{E}}_{\mathrm{f}}$$

(2)

where I_t and I_f are the measured resonance intensities at time t and at the end of the experiment respectively, and FE_f is the measured end-point percent enrichment in the cortical extract. Neuronal TCA cycle flux (V_TCAn) and glutamine synthesis rate (V_Gln) was determined from the best fits of the model to individual data sets. The cerebral metabolic rate of neuronal glucose oxidation (CMR_glc(ox)N) and glutamate/glutamine cycling flux (V_{cycle(Glu/Gln)}) were calculated from V_TCAn and V_Gln, respectively.

Because of the change in lactate and glutamine levels during seizures, the ¹³C enrichment of glutamate-C4 and glutamine-C4 achieved in the 4 minutes before bicuculline administration (prelabeling period) was subtracted from the subsequent enrichments after seizure induction (t = 0). The concentration of total glutamine, which increased during the seizure, was fitted by the model. Glutamine was assigned to the astroglial compartment, whereas glutamate was divided between neurons (90%) and astroglia (10%). The anaplerotic flux (V_ana) was assumed to be 20% of V_gln and astroglial TCA cycle flux (V_TCAa) to be 15% of the total TCA cycle flux (Sibson et al., 2001) in cortex of both control and seizing animals. The increase in glycolytic flux (ΔCMR_glycolysis) at the onset of seizures, which represents nonoxidized glucose utilization, was calculated from the initial rate of accumulation of lactate (1/2 x ΔL/At₀→_1min). During seizures the steady state rate of lactate efflux (V_efflux(Lac)) was estimated by the Michaelis-Menten relation,

$${\mathrm{V}}_{\mathrm{e}\mathrm{f}\mathrm{f}\mathrm{l}\mathrm{u}\mathrm{x}(\mathrm{L}\mathrm{a}\mathrm{c})}=[\mathrm{L}]\times {\mathrm{T}}_{max}/([\mathrm{L}]+{\mathrm{K}}_{\mathrm{m}})$$

(3)

where [L] is the concentration of lactate at steady state, and T_max and K_m are the kinetic parameters of lactate transport (0.75 μmol·g⁻¹·min⁻¹ and 2.5 mmol/L, respectively) (Pardridge, 1983). Total glucose utilization (CMR_glc(tot)) during seizures was calculated as the sum of the oxidized and nonoxidized components according to the following relations:

$$\begin{array}{rl}{\mathrm{C}\mathrm{M}\mathrm{R}}_{\mathrm{g}\mathrm{l}\mathrm{c}(\mathrm{t}\mathrm{o}\mathrm{t})}=& {\mathrm{C}\mathrm{M}\mathrm{R}}_{\mathrm{g}\mathrm{l}\mathrm{c}(\mathrm{o}\mathrm{x})\mathrm{N}}+{\mathrm{C}\mathrm{M}\mathrm{R}}_{\mathrm{g}\mathrm{l}\mathrm{c}(\mathrm{o}\mathrm{x})\mathrm{A}}+1/2\mathrm{\Delta }\mathrm{L}/\mathrm{\Delta }\mathrm{t}\\ & +1/2{\mathrm{V}}_{\mathrm{e}\mathrm{f}\mathrm{f}\mathrm{l}\mathrm{u}\mathrm{x}(\mathrm{L}\mathrm{a}\mathrm{c})}\end{array}$$

(4)

where CMR_glc(ox)A = 1/2V_ana + 1/2V_TCAa. The value of V_efflux(Lac) was assumed to be negligible (≈0) in control animals and at seizure onset, and ΔL/Δt = 0 during seizures when lactate had reached a constant level.

Statistical analysis

The nonparametric Wilcoxon two-sample test was used to determine the significance of statistical differences between seizure and control groups for comparison between plasma glucose concentrations, isotopic ¹³C-enrichments, and the measured metabolic rates, CMR_gcl(ox)N and V_{cycle(Glu/Gln)}. The statistical analyses were carried out using SAS statistical software.

RESULTS

Effects of seizures on physiologic variables and the EEG

Mean arterial blood pressure increased form a baseline level of 85 to 95 mm Hg to 145 to 160 mm Hg after bicuculline treatment, and fell within approximately 5 minutes to 125 to 135 mm Hg. In both groups, blood gases were regulated within normal physiologic limits (P_aco₂ = 30 to 45 mm Hg, P_ao₂ > 100 mm Hg) with few exceptions (Table 2).

Table 2.

Effects of bicuculline-induced seizures on systemic physiologic variables (Pco₂ and Po₂)

	Time (min)
	0	5	25	55	85
Pco2 (mm Hg)
Control	44 ± 8	44 ± 1	38 ± 11	36 ± 7	37 ± 5
Seizure	42 ± 6	34 ± 11	47 ± 12	37 ± 13
Po2 (mm Hg)
Control	176 ± 65	165 ± 40	158 ± 15	138 ± 66	167 ± 26
Seizure	160 ± 29	190 ± 23	118 ± 29	175 ± 62

The values represent mean ± SD.

Cortical electrical activity as recorded by the EEG was increased within 15 seconds of bicuculline injection (Fig. 2A). A pattern of continuous, high-amplitude spiking was followed by periods of intermittent bursts and suppression. Additional doses of bicuculline were administered as needed (Fig. 2E) in order to maintain electrographic seizure activity during the 55 minutes of the ¹³C-labeling experiment.

Fig. 2.

Effects of bicuculline treatment on the EEG. EEG signals were recorded from subcutaneous carbon fiber leads, amplified (Grass polygraph), digitized, and viewed on a personal computer. The depicted EEG traces were recorded (A) 3 to 5 minutes, (B) 5 to 7 minutes, (C) 9 to 11 minutes, (D) 11 to 13 minutes, and (E) 15 to 17 minutes after the start of [1,6–¹³C₂]glucose infusion. The initial injection of bicuculline (arrow) occurred 4 minutes after beginning of the [1,6–¹³C₂]glucose infusion. Repeat doses of bicuculline were administered as needed (e.g., trace in E), to maintain electrographic seizures.

Plasma glucose concentration and enrichment

The average basal levels of plasma glucose measured before the infusion of labeled glucose were similar (P = 0.36; Table 3) in the control (5.7 ± 1.2 mmol/L, mean ± SD) and bicuculline-treated (6.3 ± 1.3 mmol/L) groups. Infusion of the ¹³C-glucose led to a rapid increase in plasma glucose concentration to similar levels (P = 0.83) in both groups (control, 11.4 ± 1.4 mmol/L, n = 6; seizures, 11.2 ± 1.1 mmol/L, n = 6; averaged over the entire period of infusion), which remained nearly constant throughout the infusion (Table 3). The ¹³C isotopic enrichment of blood plasma glucose C1α increased from its natural abundance level of 1.1% to approximately 50% within the first minute of infusion in both groups and remained constant thereafter (Table 3). The final average enrichment of the seizure group (47% ± 1%, n = 6) was slightly less (P = 0.07) than nontreated controls (51% ± 3%, n = 6).

Table 3.

Stability of total plasma glucose concentration (mmol/L) and [1-¹³C]glucose percent ¹³C enrichment in control and bicuculline-treated rats

	Time (min)
	0	1	5	25	55	85
Concentration
Control	5.7 ± 1.2	12.9 ± 2.4	10.8 ± 1.6	9.9 ± 1.6	10.4 ± 1.4	12.9 ± 2.3
Seizure	6.3 ± 1.3	12.3 ± 3.4	12.0 ± 3.6	10.7 ± 3.2	9.9 ± 2.7
Enrichment (%)
Control		55.1 ± 8.3	48.0 ± 6.0	49.7 ± 5.4	51.3 ± 5.0	52.0 ± 5.9
Seizure		46.2 ± 6.4	48.8 ± 8.0	47.2 ± 6.9	46.1 ± 6.5

Plasma glucose concentration was determined using a Beckman Glucose Analyzer II. Plasma glucose-C1α ¹³C enrichment was measured using ¹H NMR. The percent ¹³C enrichment of glucose C1α was calculated by dividing the intensities of the two ¹³C satellites by the total intensity (¹²C + ¹³C). Values are mean ± SD.

Effects of seizures on cortical metabolite concentrations

During seizures the intensities of the lactate, glutamine, and alanine resonances in ¹H-NMR spectra increased significantly (Fig. 3). Quantitation of these resonances in cortical tissue extracts at 55 minutes of seizures (Table 4) revealed significant increases in lactate (303%), glutamine (29%), GABA (76%), alanine (159%), and succinate (80%) and a decrease in aspartate (–21%) compared with control animals. No significant change was observed in glutamate between the two groups. Similar changes in these metabolites during bicuculline induced-seizures were reported by Chapman et al. (1977), although the decrease in glutamate seen by those investigators was not detected in the present study. The increase in lactate and alanine is most likely related to increased production and accumulation of their common precursor, pyruvate (Chapman et al., 1977), due to the initially greater rate of glucose utilization over mitochondrial oxidation of pyruvate. Total amino acid levels increased (Table 4) by approximately 3.6 μmol/g, which was dominated by the increase in glutamine and GABA during the 55 minutes of seizures, consistent with an increase in anaplerotic flux (Siesjo 1978; Chapman et al., 1984) and the expected increase in blood and brain ammonia (Blennow et al., 1978; Chapman et al., 1977). An increase in GABA concentration has been reported previously during seizures (Chapman et al., 1977) possibly reflecting a response to heightened cortical excitability.

Fig. 3.

Representative in vivo NMR spectra obtained from control (A), and bicuculline-treated (B) rat cortex. Spectra in panels A and B were recorded in the periods between 60 and 85 minutes and 30 and 55 minutes, respectively, after beginning the infusion of [1,6–¹³C₂]glucose. Ala, alanine; Cho, choline; Cr, creatine; GABA, γ-aminobutyric acid; Gln, glutamine; Glu, glutamate; Lac, lactate; NAA, n-acetyl aspartate; PCr, phosphocreatine.

Table 4.

Concentrations (μmol/g) of cortical metabolites in control and bicuculline-treated rats

Group	Glutamate	Glutamine	GABA	Aspartate	Alanine	Succinate	Lactate
Control	11.8 ± 1.8	6.5 ± 1.4	2.1 ± 0.5	3.6 ± 0.8	0.6 ± 0.2	0.9 ± 0.2	3.2 ± 0.7
Seizure	11.5 ± 1.1	8.4 ± 1.6	3.7 ± 0.5	2.8 ± 0.3	1.6 ± 0.2	1.7 ± 0.4	13.0 ± 4.9
		(29%)	(76%)	(–21%)	(159%)	(80%)	(303%)

The concentrations of metabolites were determined in cortical extracts by addition of a known concentration of [2-¹³C]glycine as an internal standard. The values represent mean ± SD. The numbers in parentheses represent the percentage changes in metabolite concentrations with respect to control.

Table 5.

Percent ¹³C enrichment of cortical glutamate and glutamine in control and seizing rats after infusion of [1,6-¹³C₂]glucose

	Glutamate  C3	Glutamate  C4	Glutamine  C3	Glutamine  C4	Glucose  C1
Control	21.97 ± 2.07	31.6 ± 3.0	19.5 ± 2.8	23.4 ± 3.2	52.0 ± 5.9
Seizure	19.68 ± 1.69	27.9 ± 1.7	15.2 ± 3.7	20.4 ± 2.4	46.1 ± 6.5

The isotopic ¹³C enrichment was calculated from the ratio of the area in the POCE difference spectrum (¹³C only) and the nonedited spectrum (¹²C + ¹³C) of the separated glutamate and glutamine fractions. Values are mean ± SD.

¹³C isotopic labeling of cortical metabolites in vivo

The in vivo time courses of ¹³C labeling of cortical metabolites in control and bicuculline-treated rats during infusion of [1,6–¹³C₂]glucose are shown in Fig. 4. In both groups, prominent ¹³C isotopic labeling was detected in glutamate- and glutamine-C4 and unresolved carbon atoms, C2 and C3, with lesser intensities of aspartate-C3 and GABA-C4 (top traces in Fig. 4). In contrast to the control animals where cortical lactate levels were low, during seizures lactate-C3 (1.35 ppm) was detected in the first spectrum and the level remained elevated throughout the experiment (Figs. 4 and 6C). Glutamate-C4 (2.35 ppm) was detected within 6 minutes of the beginning of the [1,6–¹³C₂]glucose infusion in both groups, whereas glutamine-C4 (2.46 ppm) lagged that of glutamate-C4 by 5 to 10 minutes. The C2 and C3 resonances of glutamate and glutamine increased throughout the infusion. In addition to these peaks, a smaller resonance from GABA-C4 was seen in both groups. However, the intensity of the ¹³C labeled GABA-C4 resonance was greater in seizing rats as compared to control.

Fig. 4.

Representative time courses of ¹³C-labeling of cortical metabolites in control (A) and bicuculline-treated (B) rats. The midpoint time of spectrum acquisition is depicted to the left of each spectrum. The POCE difference spectrum was obtained by subtracting the ¹³C-inverted (¹²C – ¹³C) spectrum from the noninverted (¹²C + ¹³C) spectrum leaving only the ¹³C-coupled ¹H resonances, at a relative intensity of 2 x ¹³C. Ala3, alanine-C3; Asp3, aspartate-C3; Lac3, lactate-C3; GABA4, γ-aminobutyric acid-C4; Gln4, glutamine-C4; Glu4, glutamate-C4; Glx2, (glutamate-C2 + glutamine-C2); Glx3, (glutamate-C3 + glutamine-C3).

Exchange rate between α-ketoglutarate and glutamate

The exchange flux between α-ketoglutarate and glutamate (V_x) was estimated from the ¹³C enrichment time courses of glutamate-C4 and C3 and aspartate-C3 in cortical extracts obtained ex vivo (Fig. 5A). The time courses were approximated by measurements of the enrichments at three different infusion times, two chosen to be early in the enrichment curve (7 and 17 minutes) and the third to be at or near isotopic steady state. The best fit of the two-compartment metabolic model to the enrichment data yielded an average value of 41 μmol·g⁻¹·min⁻¹ (95% CI, 5 to 109 μmol·g⁻¹·min⁻¹) and V_x/V_TCAn = 79. An estimate of the interanimal variation of V_x by Monte-Carlo simulations yielded a broad non-Gaussian distribution skewed toward higher values of V_x (Fig. 5B). When the metabolic model was fitted to the experimental data by assuming a lower rate of exchange, V_x = 1 μmol·g⁻¹·min⁻¹, the fit was generally poor, with a value of overall least square SD of 14 compared with 7 for V_x = 43 and undershooting glutamate-C4 and overshooting aspartate-C3. The calculated minimum value of V_x was always found to be greater than V_TCAn (V_x/V_TCAn ≥ 5.4), although higher values could not be excluded because of the loss in sensitivity of the fit to higher values of V_x. The value of V_x was determined similarly during seizures (data not shown), which yielded a mean value of 21 μmol·g⁻¹·min⁻¹ (95% CI, 4.4 to 51.8 μmol·g⁻¹·min⁻¹) and V_x/V_TCA of 37.

Fig. 5.

Calculation of the exchange rate (V_x) between α-ketoglutarate and glutamate using ex vivo measured ¹³C labeling time courses of amino acids. (A) The ex vivo measured percent ¹³C enrichments of glutamate-C3, glutamate-C4 and aspartate-C3 were normalized with plasma glucose enrichment using Eq. 1. The data shown at 85 minutes were taken from the in vivo measurements, in which the brain was frozen in situ with liquid N₂ after the conclusion of experiments and brain metabolites were extracted as described in Materials and Methods. Scatter in the data represent standard deviations at individual time points. Solid lines represent the best fit of experimental data (•) to the metabolic model with the resulting value of V_x = 41 μmol·g⁻¹·min⁻¹. The broken lines are drawn by fitting the data with a value of V_x equal to 1 μmol·g⁻¹·min⁻¹. (B) Histogram showing the distribution of exchange rates obtained by Monte-Carlo simulation. The curve represents a polynomial fit to the distribution of V_x.

Glucose oxidation and glutamate/glutamine cycling during seizures

Seizures led to faster labeling of both glutamate and glutamine C4 from [1,6–¹³C₂]glucose as compared to saline-treated controls (Fig. 6), consistent with an activity-dependent increase in glucose oxidation. A nearly proportional increase above the respective control values was observed between CMR_glc(ox)N (214%, P = 0.005, seizures: 0.57 ± 0.07 versus 0.26 ± 0.03 μmol·g⁻¹·min⁻¹, control) and V_{cycle(Glu/Gln)} (233%, P = 0.005; seizures: 0.52 ± 0.07 versus 0.22 ± 0.04 μmol·g⁻¹·min⁻¹, control) during seizures (Table 6). CMR_glc(ox)N was plotted against V_{cycle(Glu/Gln)} along with data reported from a previous study of the anesthetized, nonactivated rat cortex (Sibson et al., 1998) that were reanalyzed using the two-compartment model described in the present study (Fig. 7). As shown in Fig. 7, within the measurement error, the experimental data obtained in the present study lie on the line predicted from the earlier study. Thus, the linearity between V_{cycle(Glu/Gln)} and CMR_glc(ox)N reported for graded anesthesia extends to intense activation, at least for seizures lasting 1 hour or less.

Fig. 6.

Typical in vivo time course of labeling of (A) glutamate-C4, (B) glutamine-C4, and (C) lactate-C3 from control and seizing cortices. The measured in vivo intensity of glutamate-C4 and glutamine-C4 was converted to percent ¹³C enrichment using the relation in Eq. 2 and further normalized with plasma glucose enrichment using Eq. (1). In seizure group glucose infusion preceded the injection of bicuculline by 4 minutes to prelabel the brain pools of extracellular and intracellular glucose to minimize label dilution in pyruvate by rapid metabolism of the initially unlabeled glucose and glycogen to lactate. The respective percent ¹³C enrichments measured at 4 minutes were subtracted from the time courses to adjust for the small amount of basal (preseizure) ¹³C labeling. The lines represent the best fit of the metabolic model to the experimental data (• = control, ▴ = seizure). The lactate concentration (○) was based on the intensity of total creatine (3.0 ppm) assumed to be 10.4 μmol/g. The ¹³C enrichment (•) of lactate-C3 was calculated from the ratio of ¹³C in the POCE difference spectrum (2 x ¹³C) and the nonedited (¹²C + ¹³C) subspectrum.

Fig. 7.

Correlation between the rate of neuronal glucose oxidation (CMR_glc(ox)N) and glutamate neurotransmitter cycling (V_{cycle(Glu/Gln)}). The plot combines data from the present study (“O” circles) and the data reanalyzed from Sibson et al. (1998) (▴ = pentobarbital, ♦ = α-chloralose, ▀ = morphine). Line represents the linear regression analysis of Sibson et al. data with slope 0.89 (95% CI 0.67, 1.10) and y-intercept 0.08 (95% CI 0.03, 0.08). The data obtained from the current study (o) lie on the predicted line. H, halothane; HS, halothane plus seizures.

Table 6.

Effects of seizures on cerebral cortical metabolic fluxes (μmol·g·min⁻¹)

	CMRglc(ox)N	CMRglc(ox)tot	ΔCMRglc(nonox)	CMRglc(tot)	Vcycle(Glu/Gln)	ΔCMRglc(ox)N/ΔVcycle(Glu/Gln)	VdilLac
Control	0.26 ± 0.02	0.34 ± 0.03	—	0.34 ± 0.03*	0.22 ± 0.04	—	0.37 ± 0.07†
Seizure onset	nd	nd	0.89 ± 0.40	1.57	nd	nd	nd
During seizure	0.57 ± 0.07	0.67 ± 0.07	0.30 ± 0.02	0.97	0.52 ± 0.07	1.02 ± 0.18	0.70 ± 0.10‡
Percent of control	214	—	—	466§, 277 ‖	233	—	—

The metabolic fluxes were calculated by fitting the time course of ¹³C labeling of glutamate-C4 and glutamine-C4 to a mathematical model as described in Materials and Methods. In both groups V_TCAa was assumed to be 0.15 x V_TCA(tot) and V_ana was 0.2 x V_Gln. The value of ΔCMR_glc(nonox) is the increase in nonoxidative glycolysis after bicuculline injection and was determined by fitting the lactate time course to a single exponential function at seizure onset and using the Michaelis-Menten equation (Eq. 3) during seizures. The values represent the mean ± SEM.

*

This calculation assumed that CMR_glc(nonox) is negligible in control rats.

†

V_dilLac in control animals was determined by iteration.

‡

V_dilLac during seizures was calculated from the measured fractional enrichments using the relation

$${\mathrm{V}}_{\mathrm{d}\mathrm{i}\mathrm{l}\mathrm{L}\mathrm{a}\mathrm{c}}=\frac{[1-({\mathrm{F}\mathrm{E}}_{\mathrm{L}\mathrm{a}\mathrm{c}}/{\mathrm{F}\mathrm{E}}_{\mathrm{G}\mathrm{l}\mathrm{c}})]}{{\mathrm{F}\mathrm{E}}_{\mathrm{L}\mathrm{a}\mathrm{c}}/{\mathrm{F}\mathrm{E}}_{\mathrm{G}\mathrm{l}\mathrm{c}}}{\mathrm{V}}_{\mathrm{T}\mathrm{C}\mathrm{A}},$$

where FE_Lac and FE_Glc are the steady-state ¹³C enrichments of [3-¹³C]lactate and [1,6-¹³C₂]glucose, respectively.

§‖

Percentage change over control of glucose consumption at seizure onset§ and during seizures^‖.

nd, not determined in the present study.

Total glucose utilization during seizures

Lactate levels increased rapidly in the cortex after bicuculline injection (Figs. 4B, 6C, and Table 4). The increase in lactate during the seizure represents the excess glucose (and astroglial glycogen), which is metabolized to pyruvate through the glycolytic pathway but not oxidized in the TCA cycle. The rate of total glucose utilization (CMR_glc(tot)) attained its maximum value (1.57 μmol·g⁻¹·min⁻¹) at the seizure onset and decreased to 0.97 μmol·g⁻¹·min⁻¹ (62% of peak value) during the period when lactate levels had stabilized (Table 6). These values are somewhat lower than those reported in a previous study of bicuculline seizures in the rat cortex (Borgstrom et al., 1976), although the percentage changes were similar. Differences in absolute values between the two studies may reflect differences in depth of anesthesia or seizure intensity.

DISCUSSION

Although there is a broad consensus that glucose is the primary fuel for supporting brain function, there has been considerable debate on whether the energy required for neurotransmission is derived from glycolysis alone or the complete oxidation of glucose. Our previous findings of a near 1:1 proportionality between incremental changes in V_{cycle(Glu/Gln)} and CMR_glc(ox)N in nonactivated cortex of anesthetized rats suggested that neuronal glucose oxidation supported the activity of glutamatergic neurons. Because these experiments were conducted on nonstimulated, anesthetized animals the question remained about the nature of the coupling at higher levels of activity. If the observed proportionality reflects a general molecular mechanism, coupling neuronal activity to energy metabolism, then it would be expected for transitions of cortical activity between any two states, independent of the absolute baseline values. The present finding that V_{cycle(Glu/Gln)} and CMR_glc(ox)N increase proportionately during the intense neuronal activation of seizures supports this mechanism. The present data, combined with results from a previous study (Sibson et al., 1998), reveal a highly linear relationship between neuronal glucose oxidation and glutamatergic neurotransmitter cycling over a substantial range of neuronal activity (Fig. 7).

Relationship between glutamate/glutamine cycling and neuronal glucose oxidation

During seizures, changes in V_{cycle(Glu/Gln)} (233%) and CMR_glc(ox)N (214%) were proportional. The ratio (ΔCMR_glc(ox)N/ΔV_{cycle(Glu/Gln)}) of 1.02 was within the 95% CI (0.67 to 1.06) obtained from the reanalysis of previously reported data for the anesthetized nonactivated rat cortex (Sibson et al., 1998).

Analysis of energy expenditure processes during excitatory signaling in rat cerebral cortex led Attwell and Laughlin (2001) to postulate that energy consumption should increase proportionately with the firing rate of glutamatergic pyramidal neurons. Bicuculline-induced seizures, which are characterized by intense bursts of neuronal activity, represent an extreme case of rapid pyramidal cell firing (Schiller, 2002; Siesjo, 1978). Recently Smith et al. (2002) reported that during somatosensory activation the changes measured in neuronal spiking frequency (Δν/ν) and oxygen consumption (ΔCMR_glc(ox)N/CMR_glc(ox)N) are proportional. Thus, the finding that ΔCMR_glc(ox)N ≈ ΔV_{cycle(Glu/Gln)}, whether measured at states of lower neuronal activity or the heightened neuronal activity of seizures (present study), indicates that neuronal activity, neurotransmitter cycling flux, and oxidative glucose metabolism are coupled linearly over a substantial range of activity. The linkage to oxidative process during the electrographically defined periods of burst and suppression (Fig. 2) are not entirely clear, although it is likely that oxidative processes continue to replete energy stores during these apparently “silent” periods.

Influence of the α-ketoglutarate/glutamate exchange rate (Vx) on the measured fluxes

Isotopic labeling of cytosolic glutamate occurs principally through the decarboxylate exchangers present in the inner mitochondrial membrane (Griffin et al., 2000), although other pathways may exist (e.g., Kamzolova et al., 2000). As described by Mason et al. (1992), the α-ketoglutarate/glutamate exchange rate (V_x) reflects the rate of isotope movement between α-ketoglutarate and glutamate, which involves membrane carriers and/or transamination reactions. Accurate determination of V_TCA requires knowledge of V_x. An estimate of V_x can be determined by simultaneously fitting the time courses of the relevant labeled species; i.e., glutamate (C4 and C3) and aspartate (C3). Only when V_x is not explicitly determined (and assumed to be some finite value) will there be uncertainty in the calculated value of V_TCA; e.g., if it is assumed that V_x » V_TCA but in fact V_x = V_TCA, then the actual value of V_TCA would be underestimated by at most 47% (Gruetter et al., 2001). Studies of the perfused heart and isolated heart mitochondria (Griffin et al., 2000; Kamzolova et al., 2000) indicate that the α-ketoglutarate/malate exchanger is the primary route of ¹³C labeling of the cytosolic glutamate pool and this exchanger is likely to be of similar importance in brain. We have shown in previous studies that V_x is several times faster than TCA cycle flux in the mature rat (Mason et al., 1992) and human cortex (Mason et al., 1995) but appears slower in the immature (postnatal day 10) rat cortex (Novotny et al., 2001). However, in all cases that we have reported, V_x/V_TCA > 1. With V_x >V_TCA, the majority of ¹³C label passes through the cytosolic glutamate pool en route through the TCA cycle. This interpretation is consistent with earlier findings using ¹⁴C-labeled isotopes (Balazs and Haslam, 1965; Haslam and Krebs, 1963) but is in contrast to more recent studies by Gruetter et al., who reported the fluxes to be equal or nearly so in rat brain (i.e., V_x /V_TCA ≈ 1) (Choi et al., 2002; Henry et al., 2002) or even <1 in a preliminary report of the mouse brain (Nehiri et al., 2003). Under these conditions, glutamate becomes a less efficient isotope trap, reducing the apparent rate of C4 labeling while increasing C3 labeling. The calculated value of V_TCA from glutamate-C4 becomes dependent on the relative value of V_x. Measurement of time courses of both glutamate-C3 and C4 and simultaneously fitting both carbon positions (C3 and C4) circumvents this issue.

In the present study, the overlap of glutamate and glutamine-C3 in the ¹H spectrum precluded the assessment of V_x using the in vivo spectra. This fact, and the reported observations by Henry et al. (2002) that the determination of V_x is highly sensitive to relatively small errors in the glutamate-C3 labeling curves measured in vivo, led us to determine V_x using the ¹³C enrichments of glutamate-C4, C3 and aspartate-C3 measured in the cortical extract ex vivo (Fig. 5). Fitting the same two-compartment metabolic model to the ex vivo ¹³C-enrichment data gave a high value of V_x in the cortex of both control (41 μmol·g⁻¹·min⁻¹, 95% CI, 5 to 109 μmol·g⁻¹·min⁻¹) and bicuculline-treated rats (21 μmol·g⁻¹·min⁻¹, 95% CI, 4.4 to 51.8 μmol·g·⁻¹·min⁻¹, data not shown). In contrast to the good fit obtained for V_x » V_TCA, a poor fit was obtained for glutamate-C4 and aspartate-C3 under the assumption that V_x ≈ V_TCA (Fig 5, broken line). The difference in the calculated value of V_TCA based on the best-fit value of V_x (45 μmol·g⁻¹·min⁻¹) was 8% less than that determined by allowing V_x equal to V_TCA. Thus, under the conditions of this study the value of V_x, whether greater than or equal to V_TCA, introduced only a small uncertainty in V_TCA.

In studies of the perfused heart, the reported range of values for V_x (see Sherry et al., 1998) depend on substrate. Most information of the exchange process at the molecular level is derived from heart studies, which differ from brain in terms of substrate preference (e.g., fatty acids versus glucose) and relative importance of the malate-aspartate shuttle and the activities of its carriers. Studies directed specifically at brain mitochondria should help to resolve this issue.

Influence of non–steady-state behavior during seizures on the measured fluxes

A study of bicuculline seizures by Meldrum and Nilsson (1976) showed that cerebral oxygen consumption (CMRo₂) was increased by 267% at 1 minute of bicuculline seizures and decreased by approximately 15% (225% of control) after 1 hour. The increase in CBF was similar to the change in CMRo₂ observed during bicuculline seizures and both parameters change rapidly to attain new steady-state levels. The short period over which metabolic fluxes stabilize is consistent with the measured lactate time courses, which reached 80% of its plateau value within 6 minutes of seizure onset. The TCA cycle flux determination from the turnover curve is weighted most heavily to the period of time over which glutamate-C4 undergoes one turnover (≈10 to 15 minutes). A modest reduction in the rate of the metabolic fluxes at later times during seizures would have lesser influence on the fluxes calculated from the time course data (Mason et al., 2003). If we assume that the rate information is spread equally (uniform sensitivity) across the full 55 minute duration of the seizures, then the 15% reduction in CMRo₂ as reported by Meldrum and Nilsson (1976) would correspond to a 15% underestimate in our flux estimate in the first 20 minutes of seizures.

Coupling between astroglial glutamate uptake and glucose utilization

An activity-based coupling mechanism that is consistent with the experimentally determined 1:1 proportionality between ΔV_{cycle(Glu/Gln)} and ΔCMR_glu(ox)N has been advanced by Magistretti et al. (1999). Neuronally released glutamate is cleared from synaptic regions mainly by uptake into surrounding astroglia by the Na+-dependent glutamate transporters, GLT1 and GLAST (Cholet et al., 2001; Rothstein et al., 1996). Glutamate uptake and cotransported sodium ions (glutamate:Na+ = 1:3) activate the Na⁺/K⁺-ATPase, which stimulates glycolysis producing two molecules of ATP (net) and lactate per molecule of glucose. Lactate released from astroglia would be transported to neurons and oxidized there. Because two molecules of ATP would be required to restore the Na+ gradient by Na⁺/K⁺-ATPase (3 Na⁺ : 1 ATP) and metabolize glutamate to glutamine by glutamine synthetase (glutamate→glutamine : 1 ATP) for each molecule of glutamate entering astroglia, in principle astroglial glycolysis could provide this energy. Although this mechanism provides an explanation for the experimentally observed flux proportionality, other possibilities exist. As discussed by Dienel and Hertz (2001), the oxidation of glucose or glutamate in astroglia would be more than sufficient to produce the small amount of ATP required to support the astroglial metabolic reactions in the glutamate/glutamine cycle. Although evidence that Na⁺/K⁺-ATPase uses glycolytic ATP over that produced by oxidation has been reported from studies of some cellular systems (Balaban and Bader, 1984; Jung et al., 2002; Proverbio and Hoffman, 1977), no direct evidence has been reported for such an interaction in the CNS.

Our results indicate that during seizures, neuronal glucose oxidation (CMR_glc(ox)N) and not total glucose utilization (i.e., CMR_glc(ox) + CMR_glc(nonox)) is proportional to V_{cycle(Glu/Gln)} (Table 6). Under nonactivated conditions, glucose oxidation is >90% complete so that CMR_glc(total) ≈ CMR_glc(ox). In contrast during intense neuronal activation excess lactate production is observed such that total glucose utilization exceeds glucose oxidation in the early period of the seizures (Table 6). Shulman et al. (2001) have proposed that excess lactate production and the decreased oxygen-to-glucose ratio during intense activation could be explained by increased nonoxidative glucose consumption to support rapid neuronal firing. The increased fraction of glucose uptake above that supporting oxygen consumption (nonoxidative glycolysis) would be effluxed from the brain as lactate. The increased ATP required by astroglia for restoration of the Na+ gradient and conversion of glutamate to glutamine (2 ATP) due to the increased glutamate/glutamine cycling could be met by glucose shunting through glycogen or by increased glycolysis. Because the rate of neuronal oxidation for a specific glutamate release would be the same, independent of the path taken by glucose in the astroglia, the extra lactate/pyruvate would not be oxidized and thus would accumulate as lactate. The increased fraction of glucose consumed either glycolytically or shunted through glycogen (and thus the level of lactate observed) would depend on the intensity and frequency of neuronal firing. The increased nonoxidative contribution was proposed to satisfy the need for rapid ATP production so as to meet the power requirement of millisecond neuronal firing. Although evidence for increased astroglial glycogen turnover during brain activation has been reported (Swanson et al., 1992), nonoxidative glycolytic production of ATP could be supplied equally well by excess glycolysis, so the mechanism remains an open question. However these explanations indicate that the excess lactate produced by the nonoxidative pathway must be effluxed totally in order for oxidative glycolysis to be the universal mechanism.

Energetics of oxidative and nonoxidative glucose metabolism during seizures

The rapid rise of lactate during seizures is well described in the literature (Blennow and Siesjo, 1985; Blennow et al., 1978; Chapman et al., 1977; Folbergrova et al., 1981). Lactate rise resulting from nonoxidative glucose utilization was greatest immediately after seizure onset and declined after several minutes. In the present study excess lactate production (nonoxidative glycolysis), when expressed in glucose equivalents, contributed 57% (100 × 0.89 / 1.57) of CMR_glc(tot) after seizure onset and declined to 31% (100 × 0.30 / 0.97) as lactate levels stabilized (Table 6). The calculated net contribution of ATP produced by glycolysis during seizures is small in comparison to that produced by mitochondrial oxidative phosphorylation (Table 7). In the absence of glycogen shunting glycolysis (CMR_glc(ox)tot + CMR_glc(nonox)) could account for a maximum of approximately 11% of ATP production at seizure onset while declining to 7% when glucose and oxygen consumption are balanced. However, if excess lactate reflects glycogen shunting of the metabolized glucose in the astroglia, then net glycolytic ATP production would be only approximately 6%, equivalent to the net ATP produced by glycolysis for the complete oxidation of glucose.

Table 7.

Contribution of ATP supplied by glycolysis (nonoxidative pathway) versus mitochondrial oxidative phosphorylation during seizures

Seizure period	Glycolysis (nonox)	Glucose oxidation
Seizure onset	11.4%	88.6%
During seizure	7.4%	92.6%

ATP produced by nonoxidative glycolysis is calculated from the increase in rate of glycolysis at the onset of seizure (0.89 μmol·g·min⁻¹) and half of the lactate efflux during the seizure period (0.30 μmol·g·min⁻¹). The ATP produced by glucose oxidation was based on the value of CMR_glc(ox)tot during seizures (0.67 μmol·g·min⁻¹). The number of ATP molecules produced by glycolysis (nonoxidative pathway) and glucose oxidation is taken as 2 and 36, respectively. The fraction of energy contributed from non-oxidative glycolysis was calculated by dividing ATP produced by nonoxidative glycolysis with the total expected yield of ATP (nonoxidative + oxidative glycolysis). Data are reported as percent of total.

A common interpretation of the lactate concentration is that it reflects the total amount of nonoxidative glucose consumption, either by glycolysis alone or by shunting through glycogen. This interpretation would be valid during the initial rise of lactate after a sudden increase in activity (at the onset of seizures in the present experiment) but not at steady state when the concentration of lactate levels off. Lactate is a substrate for the membrane-bound monocarboxylate transporter (Magistretti and Pellerin, 1996). When lactate levels off and the rate of pyruvate oxidation is constant, lactate efflux through the monocarboxylate transporter must balance excess (nonoxidized) glycolysis. Thus, the lactate concentration reflects this balanced condition serving to drive efflux through the monocarboxylate transporter until the rate of efflux balances its synthesis. The rate of lactate efflux from brain, as determined by the Michaelis-Menten constants for lactate transport and the lactate concentration, was related to excess glucose uptake in the rat and human brain (Shulman et al., 2001). Although more data are needed to evaluate this mechanism quantitatively, it appears likely that the excess glucose consumed during cortical stimulation provides the rapid energy needed to power the millisecond firing of neurons and is effluxed as lactate. This mechanism accounts for the excess of glucose uptake and supports the present observation of neuronal glucose oxidation being linear with glutamate neurotransmitter release over the measured range of brain activity.

Implications for functional activation studies

Functional magnetic resonance imaging measures the change in image intensity associated with increases in CBF and CMRo₂ during brain activation (Ogawa et al., 1990). Several recent studies have provided evidence that the increases in the functional magnetic resonance imaging signal are associated with increases in neuronal activity (Logothetis et al., 2001; Smith et al., 2002). Smith et al. (2002) have shown that the increase in neuronal oxygen consumption in the rat somatosensory cortex during forepaw stimulation was proportional to the average increase in ensemble neuronal firing. Our findings show that even under the most intense state of cortical activation, the increase in glutamate/glutamine cycling flux is proportional to the increase in the rate of neuronal glucose oxidation but not to the increase in total glucose utilization. In contrast to what had previously been suggested (Fox and Raichle, 1986; Fox et al., 1988; Gjedde, 1993), nonoxidative glycolytic processes during seizures (and by extrapolation to other activated states) do not appear to supply the energy required during neuronal activity.