In vivo detection of cortical GABA turnover from intravenously infused [1-13C]D-glucose

Authors

  • Jehoon Yang,

    1. Molecular Imaging Branch, National Institute of Mental Health, Bethesda, Maryland, USA
    Search for more papers by this author
  • Charles Q. Li,

    1. Molecular Imaging Branch, National Institute of Mental Health, Bethesda, Maryland, USA
    Current affiliation:
    1. School of Engineering and Applied Sciences, University of Pennsylvania, Philadelphia, Pennsylvania
    Search for more papers by this author
  • Jun Shen

    Corresponding author
    1. Molecular Imaging Branch, National Institute of Mental Health, Bethesda, Maryland, USA
    • Molecular Imaging Branch, National Institute of Mental Health, Bldg. 10, Rm. 2D51A, 9000 Rockville Pike, Bethesda, MD 20892-1527
    Search for more papers by this author

  • This article is a US Government work and, as such, is in the public domain in the United States of America.

Abstract

In this study [2-13C] γ-aminobutyric acid (GABA) was spectrally resolved in vivo and detected simultaneously with [4-13C]glutamate (Glu) and [4-13C]glutamine (Gln) in the proton spectra obtained from a localized 40 μL voxel in rat neocortex with the use of an adiabatic 1H-observed, 13C-edited (POCE) spectroscopy method and an 89-mm-bore vertical 11.7 Tesla microimager. The time-resolved kinetics of 13C label incorporation from intravenously infused [1-13C]glucose into [4-13C]Glu, [4-13C]Gln, and [2-13C]GABA were measured after acute administration of gabaculine, a potent and specific inhibitor of GABA-transaminase. In contrast to previous observations of a rapid turnover of [2-13C]GABA from [1-13C]glucose in intact rat brain, the rate of 13C incorporation from [1-13C]glucose into [2-13C]GABA in the gabaculine-treated rats was found to be significantly reduced as a result of the blockade of the GABA shunt. Magn Reson Med 53:1258–1267, 2005. Published 2005 Wiley-Liss, Inc.

γ-Aminobutyric acid (GABA) is the most prevalent neurotransmitter with a primary inhibitory synaptic action. In mammals, GABA is found almost exclusively in the brain. GABA is a constituent of intermediary metabolism. It is directly synthesized from glutamate (Glu) in a single α-decarboxylation step catalyzed by Glu decarboxylase (GAD) (1). GABA is catabolized into succinic semialdehyde by GABA-transaminase (GABA-T), and then into succinate by succinic semialdehyde dehydrogenase (SSADH) (1). The metabolism of GABA forms the GABA shunt, which circumvents two steps of the TCA cycle (i.e., α-ketoglutarate dehydrogenase complex and succinyl-CoA synthase) in GABAergic neurons. In the cortex, the concentration of GABA is approximately 1 mM, which is about 1000 times that found in monoamines. The inhibitory neurotransmitter GABA is released by GABAergic neurons and acts primarily on GABAA and GABAB receptors located at axonal or dendritic output synapses to increase Cl or K+ conductance, respectively. Its transmission is terminated by high-affinity GABA transporters located on GABAergic neurons as well as surrounding glial cells (2). GABA taken up by the glial cells is converted into glutamine (Gln) and recycled back to GABAergic neurons for repletion of the carbon skeleton lost due to neurotransmitter GABA release (3).

In vivo MRS measurement of 13C labeling of Glu and Gln from intravenously infused [1-13C]glucose ([1-13C]Glc) or other 13C-labeled substrates has become a prominent tool for studying brain oxidative metabolism and Glu-Gln cycling between glutamatergic neurons and glial cells (4–6). The 13C labels flow from [1-13C]Glc via glycolysis to [2-13C]acetyl CoA, which enters the TCA cycle. During the first turn of the TCA cycle, 13C labels [4-13C]Glu through its exchange reaction with mitochondrial [4-13C]α-ketoglutarate. The 13C labels flow further to [4-13C]Gln via mainly the Glu-Gln cycle. In GABAergic neurons, GAD converts [4-13C]Glu directly into [2-13C]GABA. Therefore, the turnover of [2-13C]GABA is a direct measure of the activity of GAD and the GABA shunt, which is closely related to the metabolic aspect of the GABAergic function (1, 7, 8). Several compounds, such as gabaculine and vigabatrin, are irreversible inhibitors of GABA-T. On acute administration of GABA-T inhibitors, the GABA shunt is rapidly inhibited, causing substantial accumulation of both cytosolic and vesicular GABA in the brain (9, 10) and increased GABA release (9–12). The disruption of the recycling of endogenous carbon skeleton in GABAergic neurons caused by the acute inhibition of the GABA shunt is accompanied by a decrease in the total concentration of Glu and Gln. 13C NMR studies have shown that astrocytic Gln acts as the primary precursor for GABA following acute GABA-T inhibition (3, 13, 14). Since astrocytes do not synthesize GABA (1), a direct in vivo measurement of the turnover of [2-13C]GABA from [1-13C]Glc following acute GABA-T inhibition should be useful for characterizing the metabolic flux between astrocytes and GABAergic neurons in the living brain.

However, there are some major technical challenges in measuring the kinetics of [2-13C]GABA turnover from [1-13C]Glc in vivo. GABA turns over rapidly with normal GABA-T activity in the intact brain (8, 14–16). Acute GABA-T inhibition rapidly causes a dramatic reduction in the rate of turnover of 13C-labeled GABA (14), which makes it impossible to follow its temporal changes in vivo using direct 13C MRS due to the low concentration of 13C-labeled GABA. Inverse detection of 1H bound to 13C using the 1H-observed, 13C-edited (POCE) spectroscopy method offers greatly improved sensitivity (17–20). The proton triplet bound to [2-13C]GABA resonates at 2.30 ppm. It overlaps with the Glu-4 triplet at 2.35 ppm and, at low field strength, with Gln-4 triplet at 2.46 ppm, both of which are also labeled by 13C during [1-13C]Glc infusion. Thus, to reliably determine the turnover kinetics of [2-13C]GABA, it is important to spectrally resolve [2-13C]GABA from overlapping [4-13C]Glu in the crowded proton spectra using very high magnetic field strength. In this paper, we report the in vivo measurement of [2-13C]GABA turnover in the proton spectra of the rat brain following acute gabaculine treatment and subsequent intravenous infusion of [1-13C]Glc. To maximize sensitivity and to accommodate the narrow bore size of a low-cost vertical 89-mm-bore 11.7 Tesla microimager, an adiabatic POCE method with single-shot spatial localization was used for a concentric 1H/13C surface coil system. The purpose of this report is to demonstrate that at 11.7 Tesla with optimized B0 homogeneity and heteronuclear decoupling, [2-13C]GABA can be spectrally resolved from [4-13C]Glu and [4-13C]Gln, allowing, for the first time, direct in vivo measurement of cerebral [2-13C]GABA turnover from [1-13C]Glc. A preliminary account of spectral resolution of [2-13C]GABA from neighboring [4-13C]Glu at 11.7 Tesla in the POCE spectra acquired from the intact rat brain (with no inhibition of GABA-T) was previously reported as an abstract (19).

MATERIALS AND METHODS

All of the experiments were performed on a Bruker microimaging spectrometer (Bruker Biospin, Billerica, MA) interfaced to an 11.7 Tesla 89-mm-bore vertical magnet (Magnex Scientific, Abingdon, UK) located in an unshielded room. The spectrometer is equipped with a 57-mm i.d. gradient (maximum gradient strength: 3 G/mm; rise time: 100 μs) for in vivo experiments, and a 2.5-mm broadband inverse (BBI) gradient probe (5 G/mm on the z-axis) for high-resolution in vitro NMR spectroscopy experiments. For in vivo experiments, the Z0 coil for the deuterium lock was used to compensate transient field drift resulted from switching gradients. Preemphasis was adjusted with no significant cross preemphasis terms found during a 1–1000 ms period after a 50% 500-ms test gradient applied to each gradient axis was switched off. The in vivo proton and POCE experiments were performed with the use of an in-house-built concentric surface 1H (circular, diameter =15 mm)/13C (square, 25 × 25 mm2) RF coil system mounted on an integrated in-house-built animal-handling system capable of rat head fixation, body support, physiology maintenance, coil tuning, and RF shielding.

In Vivo 1H and 1H/13C Spectroscopy

Male Sprague-Dawley rats (150–180 g, N = 8) were fasted for 24 hr, with free access to drinking water. They were then studied to measure [2-13C]GABA turnover in the neocortex, as approved by the NIMH Animal Care and Use Committee. The rats were orally intubated and ventilated with a mixture of 70% N2O/30% O2 and 1.5% isoflurane. The left femoral artery was cannulated to measure arterial blood gases (pO2 and pCO2), pH, and plasma glucose concentration (using a blood analyzer; Bayer Rapidlab 860, East Walpole, MA, USA), and to monitor arterial blood pressure. Two femoral veins were also cannulated: one for intravenous infusion of α-chloralose (initial dose: 80 mg/kg supplemented with a constant infusion of 26.7 mg/kg/hr throughout the experiment), and one for intravenous infusion of [1-13C]Glc. After the animals were prepared for surgery, isoflurane was discontinued and pancuronium bromide was administered (1–3 mg/kg, i.v.) to maintain immobilization. Rectal temperature was maintained at 37.5°C ± 0.5°C using an external pump for water circulation (BayVoltex, Modesto, CA, USA). Arterial blood pO2 was maintained at 150–170 mmHg, pCO2 at 32–43 mmHg, mean blood pressure at 180 ± 30 mmHg, and plasma pH at 7.35–7.40. The end-tidal CO2, tidal pressure of ventilation, and heart rate were also monitored. Three-slice (coronal, horizontal, and sagittal) scout rapid acquisition with relaxation enhancement (RARE) images (FOV = 2.5 cm, slice thickness = 1 mm, TR/TE = 200/15 ms, rare factor = 8, 128 × 128 data matrix) were used to position the rat inside the magnet bore such that the gradient isocenter was about 0–1 mm posterior to bregma, and the center of the spectroscopy voxel (4.5 × 2.5 × 4.5 mm3 or 4 × 2.5 × 4 mm3) was localized in the cortex. The rat brain was shimmed as previously described using the fast automatic shimming technique by mapping along projections (FASTMAP) and five linear acquisitions for up to third order noniterative, efficient slice shimming (FLATNESS) methods (21). Usually a 9–13 Hz half-height line width for the metabolites was obtained from the ∼50 μL spectroscopy voxel after shimming. Prior to acute inhibition of GABA-T, a short-TE spectrum using a previously described 3D localization method (21–23) was acquired from a 4.5 × 2.5 × 4.5 mm3 spectroscopy voxel (NS = 128, TR/TE = 2700 ms/15 ms). Then the basal GABA concentration in the same voxel was measured using a 1D doubly selective homonuclear polarization transfer method (24). Briefly, the thermal equilibrium GABA-4 signal at 3.0 ppm and the overlapping creatine, glutathione, and macromolecules were suppressed first using the chemical shift selective (CHESS) method. The signal of interest (the GABA-4 peak at 3.02 ppm) was subsequently regenerated anew from the thermal equilibrium GABA-3 signal at 1.91 ppm via homonuclear polarization transfer. Single-shot spatial localization was achieved using a 90° sinc (five-lobe, 0.5 ms) for slice-selective excitation along the x-dimension, and four 180° slice-selective adiabatic hyperbolic secant pulses (μ = 5, 1% truncation, 2 ms) were used for spatial localization along the y- and z-dimensions. The J evolution between GABA-2 and GABA-3 was refocused using a 15-ms numerically optimized doubly selective 180° refocusing pulse (25) during the preparation of GABA-3, 4 antiphase coherence. The GABA-4 spin state prior to data acquisition is (Iy + Iy + 4IySzSz + 4Iy′SzSz)/2, which represents the full intensity of the outer two lines of the GABA-4 triplet (26). The doubly selective 180° pulse generates negligible excitation of the macromolecules at 1.72 ppm, leading to complete macromolecule suppression. The intense signal from the N-acetylaspartate (NAA) methyl group at 2.02 ppm in the edited GABA spectra serves as an internal reference for zero-order phase and concentration.

Gabaculine (100 mg/kg, 0.6 cc, i.v.; BIOMOL Research Laboratories) was administered after the baseline spectra were acquired. A transient ∼20% decrease in the mean arterial blood pressure (MABP) was observed after gabaculine injection, which was reversed quickly and did not occur again. One hour after gabaculine administration, the elevated cerebral GABA level was verified by the 1D homonuclear polarization method described above. Immediately after the acquisition of the edited GABA spectrum intravenous infusion of 99% enriched [1-13C]Glc (Cambridge Isotope Labs, Andover, MA, USA) was started. The infusion protocol consists of an initial bolus of 162 mg/kg/min of 1.1 M [1-13C]Glc in the first 5 min followed by constant-rate infusion of the same glucose solution at 62.8 mg/kg/min. The total plasma glucose level was rapidly raised to ∼15 mM (measured 10 min after the start of glucose infusion). Additional sampling of arterial blood was performed at 30-, 60-, 90-, and 120-min intervals. The plasma glucose level and [1-13C]Glc fractional enrichment (FE) were maintained at mean values of ∼15 mM and ∼94%, respectively. The pulse sequence for adiabatic POCE spectroscopy (Fig. 1) consists of proton outer volume suppression (OVS) using hyperbolic secant pulses (2 ms, μ = 5, 1% truncation) along the x (10 mm slab), -x (10 mm slab), y (3 mm slab), −y (5 mm slab), z (10 mm slab), and −z (10 mm slab) directions. Water suppression was accomplished using sinc (five-lobe, 15 ms) and Gaussian (15 ms, 1% truncation) pulses based on the CHESS method. The CHESS and OVS pulses were interleaved and repeated six times and three times, respectively. Immediately after the sixth CHESS pulse, a 500 μs 90° adiabatic half-passage (AHP) pulse was applied for nonselective excitation followed by six 180° adiabatic full-passage (AFP) pulses (hyperbolic secant, 1.5 ms, μ = 5, 1% truncation) for 3D spatial localization via adiabatic slice-selective refocusing with a pair of identical AFP pulses per dimension, as described previously (21–23), to localize a 4 × 2.5 × 4 mm3 spectroscopy voxel (40 μL). Broadband 13C-editing was achieved by applying an additional 180° adiabatic tanh/tan (23) inversion pulse (2 ms, ζ = 10, κ = 20, R = 80, frequency sweep width = 40000 Hz, γB1min = 1101 Hz) to the carbon channel. The theoretical delay between the center of the carbon pulse and the start of data acquisition (τ, see Fig. 1) is approximately 0.5/1JHC. τ was empirically optimized to be 3.9 ms using a phantom sample containing 5 mM [4-13C]Glu with a pH of 7.0. The empirically optimized τ is very close to the theoretical value of 3.94 ms (1JHC = 127 Hz in [4-13C]Glu). Broadband adiabatic decoupling was achieved with the use of a decoupling scheme based on the tanh/tan pulses (3 ms, ζ = 10, κ = 20, R = 150, frequency sweep width = 50000 Hz, γB2min = 1005 Hz) (20, 23). The carrier frequency for both the 13C editing and decoupling pulses was centered at 38.5 ppm. The broadband adiabatic decoupling pulse train was executed for a duration of 180 ms at the start of data acquisition. We verified the effective inversion and decoupling over the 20–57 ppm region covering 13C-labeled metabolites from [3-13C]lactate ([3-13C]Lac) to [2-13C]Glu experimentally using the 5 mM [4-13C]Glu phantom sample by shifting the carrier frequency of the carbon channel over the designated frequency range. The carbon pulse was switched off during odd-numbered scans, and on during even-numbered scans. The odd- and even-numbered scans were stored in separate data blocks. Protons bound to 13C were obtained by subtracting the even-numbered scans from the odd-numbered scans (TR/TE = 2700/22 ms). The shimming procedure was repeated periodically during the measurement of the kinetics of 13C label incorporation to maintain optimal B0 homogeneity.

Figure 1.

Adiabatic POCE pulse sequence. A 90° adiabatic half-passage pulse (AHP, 0.5 ms) was applied for nonselective excitation, followed by six 180° hyperbolic secant pulses (1.5 ms, μ = 5, 1% truncation) for 3D spatial localization via adiabatic slice-selective refocusing with a pair of identical hyperbolic pulses per dimension. 13C-editing was achieved by applying an additional 180° adiabatic tanh/tan inversion pulse (2 ms, ζ = 10, κ = 20, R = 80, frequency sweep width = 40000 Hz, γB1min = 1101 Hz) to the carbon channel. τ ≅ 0.5/1JHC. The carbon pulse was switched off during odd-numbered scans and on during even-numbered scans. Broadband 13C decoupling was achieved with the use of a decoupling scheme based on the tanh/tan pulses (3 ms, ζ = 10, κ = 20, R = 150, frequency sweep width = 50000 Hz, γB2min = 1005 Hz), which cover the 20–57 ppm region. The odd- and even-numbered scans were stored in separate data blocks. Protons bound to 13C were obtained by subtracting the even-numbered scans from the odd-numbered ones. Note that for clarity, the preceding CHESS water suppression pulses and OVS scheme were not drawn. ▪: slice-selection gradients; ░: crusher gradients. Inset: Coronal spin-echo anatomical image. The location of the spectroscopy voxel is indicated by a rectangular box.

In Vitro 1H and 1H/13C Spectroscopy

Immediately after the in vivo POCE data were acquired, we arrested the metabolism of the rat brain using a microwave fixation system (model TMW-6402C; Muromachi Kikai Co., Tokyo, Japan) that inactivates enzymatic processes in about 1 s (27). The rat was subsequently frozen in liquid nitrogen. The brain tissue corresponding to the spectroscopy voxel was removed and weighed. Extraction using perchloric acid (PCA; 12% solution, 3–4 mL/g of sample) was performed as described previously (24). Briefly, the brain extracts were centrifuged at 13000 rpm for 20 min at 4°C. The filtered supernatant was neutralized to pH = 7.4 using KOH or NaOH, centrifuged again, and filtered with a 0.2 μm Acrodisc syringe filter after it was passed through a 1-mL column of Chelex 100 resin. The samples were then repeatedly lyophilized and redissolved in D2O (pH = 7.4). PCA extracts of the plasma sample were prepared in a similar manner. A high-resolution version of the adiabatic POCE pulse sequence was also implemented using the 2.5-mm BBI probe (NS = 4096, TR/TE = 6000/10 ms). In the high-resolution version of the adiabatic POCE pulse sequence, OVS, water suppression, and spatial localization were deleted. The high-resolution PCA extracts/D2O spectra were acquired using added 3-(trimethylsilyl) propionic-2,2,3,3-d4 acid (TSP-d4) as a chemical shift reference standard. To determine the 13C FE of plasma glucose, we used a standard pulse-acquire experiment to obtain 1D proton spectra of the plasma extracts without decoupling (NS = 512, TR/TE = 6000/0 ms). The relative intensities of the α-anomeric protons of D-glucose bound to 12C and 13C were measured to calculate the 13C FE of plasma glucose.

The total concentrations of Glu, Gln, and GABA were determined using the LCModel approach (28). In vitro spectra of the following 18 metabolites were acquired according to the instructions described in the LCModel manual (http://www.s-provencher.com/pages/lcm-manual.shtml): alanine, aspartate (Asp), creatine (Cr), GABA, glucose, Glu, Gln, glutathione, lactate, myo-inositol, glycine, glycerophosphorylcholine, NAA, N-acetylaspartylglutamate (NAAG), phosphocreatine (PCr), phosphocholine, phosphoethanolamine, and taurine. In addition, basis spectra from [1-13C]Glc and the macromolecule baseline were also collected (28). We accounted for the fast T1 relaxation of the methylene protons of Cr and PCr in the LCModel analysis by removing the constraints on the intensity relationship between their methylene and methyl protons. LCModel analysis of the POCE subspectra with the 13C editing pulse switched off yielded the total concentrations of Glu, Gln, and GABA. The concentration of total creatine (8.5 μmol/g ww) was used as an internal reference standard. We analyzed the intensities of [4-13C]Glu, [4-13C]Gln, and [2-13C]GABA signals in the edited in vivo POCE time-course spectra using the MATLAB curve-fitting toolbox (The MathWorks, Inc., Natick, MA) in the frequency domain after we subtracted the even-numbered POCE subspectra from the corresponding odd-numbered ones. The FEs of [4-13C]Glu in the brain PCA extracts were used to quantify the in vivo edited POCE spectra acquired at the endpoint of the infusion experiment (29).

RESULTS

In Vivo 1H Spectroscopy

A typical in vivo short-TE proton spectrum acquired from the spectroscopy voxel in the rat brain prior to gabaculine injection (TR/TE = 2700/15 ms, 4.5 × 2.5 × 4.5 mm3, NS = 256, LB = −4 Hz, GB = 0.4) is shown in Fig. 2. The voxel (∼50 μL) is predominantly located in the rat neocortex, with minimal contributions from the corpus callosum, as shown in Fig. 1. The inset shows the position of the spectroscopy voxel marked on a spin-echo coronal image of the rat brain. The short-TE in vivo 1H spectrum demonstrates excellent spectral resolution and sensitivity. The phosphocreatine methylene peak at 3.93 ppm, and the creatine methylene peak at 3.92 ppm are clearly resolved. The Glu-4 signal at 2.35 ppm, the Gln-4 signal at 2.46 ppm, and the GABA-2 signal (and macromolecules underneath) signal at 2.30 ppm are also resolved spectrally. The signal from the NAAG methyl group at 2.05 ppm is also visible in the proton spectrum. The lactate signal at 1.32 ppm remains very low throughout the experiment, indicating excellent physiological conditions. The basal concentrations of the metabolites were obtained from LCModel analysis of the in vivo 1H spectra acquired prior to gabaculine administration.

Figure 2.

In vivo 1H short-TE spectrum prior to gabaculine injection (TR/TE = 2700/15 ms, 4.5 × 2.5 × 4.5 mm3, NS = 256, AQ = 206 ms). The PCr methylene peak at 3.93 ppm, and Cr methylene peak at 3.92 ppm are clearly resolved. The Glu-4 signal at 2.35 ppm, Gln-4 signal at 2.46 ppm, and GABA-2 (and macromolecules underneath) signal at 2.30 ppm are also resolved. The signal from the NAAG methyl group at 2.05 ppm is also visible in the proton spectrum. The spectrum was phased using zero-order phase only without any baseline correction. Glx-2: Glu-2 + Gln-2.

Figure 3 shows the edited GABA spectra before and 1 hr after gabaculine injection using the 1D doubly selective homonuclear polarization transfer method (TR/TE = 2700/68 ms, 4.5 × 2.5 × 4.5 mm3, NS = 256, LB = 5 Hz). The edited GABA doublet at 3.02 ppm and a clean baseline were obtained, indicating excellent suppression of the overlapping creatine methyl resonance at 3.03 ppm and glutathione cysteinyl methylene resonance at 2.87–2.96 ppm and outer volume signals. The negative peak at 2.35 ppm was the Glu-4 signal that resulted from partial polarization transfer from the Glu-3 signal at 2.11 pm. The doubly selective homonuclear polarization transfer method also completely suppresses macromolecules that resonate at 3.0 ppm (24). As shown in Fig. 3, after acute administration of gabaculine (100 mg/kg, 0.6 mL, i.v.) the signal of the edited GABA doublet at 3.02 ppm acquired from the α-chloralose anesthetized rat brain significantly increased, as expected. Based on the previously validated GABA quantification method using the GABA-to-NAA ratio (24), the basal concentration of GABA was determined to be 1.1 ± 0.1 μmol/g wet weight (mean ± SD, N = 8) in this study. [NAA] = 10.4 ± 0.5 μmol/g wet weight (mean ± SD, N = 8), as determined from the LCModel analysis, was used to calculate the GABA concentration. One hour after gabaculine injection, [GABA] was found to be 2.4 ± 0.3 μmol/g wet weight (mean ± SD, N = 8).

Figure 3.

Edited GABA spectra before (lower trace) and 1 hr after (upper trace) gabaculine injection using the 1D doubly selective homonuclear polarization transfer method (TR/TE = 2700/68 ms, 4.5 × 2.5 × 4.5 mm3, NS = 256, LB = 5 Hz, same vertical scale). The edited GABA doublet at 3.02 ppm and a clean baseline are obtained, indicating excellent suppression of the overlapping Cr methyl resonance at 3.03 ppm and the glutathione cysteinyl methylene resonance at 2.87–2.96 ppm, and outer volume signals. The negative peak at 2.35 ppm was the Glu-4 signal that resulted from a partial polarization transfer from the Glu-3 signal at 2.11 pm. Acute administration of gabaculine (100 mg/kg, 0.6 mL, i.v.) causes a significant increase in the intensity of the GABA doublet at 3.02 ppm from the basal 1.1 ± 0.1 μmol/g wet weight to 2.4 ± 0.3 μmol/g wet weight (mean ± SD, N = 8).

In Vivo 1H/13C Spectroscopy

Figure 4a shows a typical set of time-resolved POCE subspectra with the 13C editing pulse switched off following the onset of intravenous [1-13C]Glc infusion. As described in Materials and Methods, [1-13C]Glc infusion was started approximately 1 hr after acute gabaculine treatment. Each spectrum (NS = 128) was extracted from the interleaved acquisition of successive 256 scans over an 11.5-min interval. The POCE data acquisition was started every 15 min. The 3.5-min interval of dead time was used to periodically reshim the spectroscopy voxel to ensure that optimal B0 homogeneity was maintained throughout the data acquisition process. The spectra were processed using Lorentzian-Gaussian transformation with GB = 0.3, LB = −3 Hz. As shown in Fig. 4a, the spectrally resolved GABA-3 signal at 1.91 ppm, and the GABA-2 signal at 2.30 ppm increased significantly over the time period of the 132-min spectroscopic measurements. The apparent increase in the intensity of the signals in the 3.2–4.0 ppm range can be explained by the increase in brain glucose levels following the onset of intravenous [1-13C]Glc infusion. A significant decrease in the aspartate resonances at 2.79 ppm is also appreciable. The POCE difference spectra following the onset of intravenous [1-13C]Glc infusion are shown in Fig. 4b. The [4-13C]Glu signal at 2.35 ppm is always visible in the first spectrum, the acquisition of which was initiated at approximately the time the glucose reached the right femoral vein. The much weaker signals from [4-13C]Gln at 2.46 ppm and [2-13C]GABA at 2.30 ppm were detected subsequently. At the in vivo spectral resolution achieved at 11.7 Tesla, the [2-13C]GABA signal at 2.30 ppm is spectrally resolved from the neighboring [4-13C]Glu signal at 2.35 ppm, which allows the the turnover kinetics of [2-13C]GABA to be determined from intravenous infused [1-13C]Glc. As shown in the POCE difference spectra acquired at the 126-min time interval, [2-13C]Asp at 3.88 ppm, [2-13C]Glx at ∼3.8 ppm, creatine methyl group+[4-13C]GABA at ∼3.0 ppm, NAA methyl group + [3-13C]Glx at ∼2.1 ppm, and [3-13C]Lac at 1.32 ppm were also detected. The small [3-13C]Lac signal is consistent with the low lactate concentration shown in Figs. 2 and 4a. Note that due to the marked reduction in the cerebral level and labeling of aspartate accompanied by acute GABA-T inhibition (14) (Fig. 4a), the weaker [3-13C]Asp signals at 2.66 and 2.79 ppm were not distinguishable from noise in the 40 μL POCE spectra averaged over 11.5 min. This is consistent with the reduced total aspartate intensity as shown by the reduced Asp-3 intensity at 2.79 ppm in Fig. 4a. The total concentration of aspartate is reduced by 61% at the end of the in vivo spectroscopy measurement based on LCModel analysis of the POCE subspectra with the 13C editing pulse switched off. A comparison with conventional wideband alternating-phase low-power technique for zero-residual splitting (WALTZ-4) decoupling (γB2 = 625 Hz, duration of nominal 90° pulse length = 400 μs) is also shown in Fig. 5, where the in vivo edited POCE spectra were acquired over the 120–132-min interval from a 55 μL voxel. 13C editing and decoupling were achieved by the use of a sech pulse (2 ms, μ = 3, 1% truncation) and the WALTZ-4 scheme, respectively.

Figure 4.

In vivo time courses of (a) POCE subspectra with the 13C editing pulse switched off, and (b) POCE difference spectra following intravenous [1-13C]Glc infusion (1.1 M, see text for a description of the glucose infusion protocol), which was initiated approximately 1 hr after acute gabaculine treatment (100 mg/kg, 0.6 cc, i.v.). Each POCE spectrum was acquired from a 4.0 × 2.5 × 4.0 mm3 spectroscopy voxel shown in the Fig. 1 inset and averaged over 11.5 min (TR/TE = 2700/22 ms, NS = 256, GB = 0.3, LB = −3 Hz). The POCE data acquisition was started every 15 min. The time scale was referenced to the start of [1-13C]Glc infusion. In a, the spectrally resolved GABA-3 signal at 1.91 ppm and the GABA-2 signal at 2.30 ppm increase significantly over time. In b, the [2-13C]GABA signal at 2.30 ppm is spectrally resolved from the [4-13C]Glu signal at 2.35 ppm. At the 126-min time interval, [2-13C]Asp at 3.88 ppm, [2-13C]Glx at ∼3.8 ppm, Cr+[4-13C]GABA at ∼3.0 ppm, NAA + [3-13C]Glx at ∼2.1 ppm, and [3-13C]Lac at 1.32 ppm were also detected. In both aand b, the time-domain data were truncated by 50% before they were Fourier transformed to enhance the SNR, at the expense of the apparent spectral resolution.

Figure 5.

In vivo edited POCE spectra acquired over the 120–132-min (NS = 256) interval following [1-13C]Glc infusion. The results were from separate studies using the same experimental protocol as in Fig. 4b except that 1) 13C editing was achieved with the use of a sech pulse (2 ms, μ = 3, 1% truncation), 2) 13C decoupling was performed using a WALTZ-4 scheme composed of rectangular pulses (γB2 = 625 Hz, AQ = 196 ms), and 3) the localized volume was 55 μL. Lower trace: rat #1; upper trace: rat #2.

Figure 6a shows the time courses of the total concentrations of Glu, Gln, and GABA determined from the LCModel analysis of the POCE subspectra with the 13C editing pulse switched off. The total concentration of Glu undergoes a steady decline after acute gabaculine administration. At the endpoint of the in vivo spectroscopic measurement, a 30% reduction in total Glu concentration was measured. In contrast, the change in the total concentration of Gln was much smaller. It decreased initially following acute gabaculine administration. The decrease in total Gln concentration was partially reversed at the endpoint of our in vivo spectroscopic measurement. The most dramatic change was the concentration of GABA, which increased linearly following the acute administration of gabaculine. The concentrations of [4-13C]Glu, [4-13C]Gln, and [2-13C]GABA were determined based on the MATLAB curve-fitting analysis of the in vivo POCE difference spectra and the FE([4-13C]Glu) from the high-resolution spectra of brain PCA extracts, as described in Methods and Materials. The FEs of [4-13C]Glu, [4-13C]Gln, and [2-13C]GABA are plotted in Fig. 6b. At the end point of the in vivo spectroscopy measurement, the FEs of [4-13C]Glu and [4-13C]Gln reached 35% ± 2% and 33% ± 3% (mean ± SD, N = 8), respectively. Consistent with previous in vivo 13C NMR studies of the Glu-Gln cycling between glutamatergic neurons and astrocytes (4–6), the turnover of [4-13C]Gln following acute inhibition of GABA-T by gabaculine was slower than that of [4-13C]Glu. The most striking observation is the low FE of GABA. Although the total concentration of GABA increased to 6.3 ± 0.6 μmol/g wet weight (mean ± SD, N = 8) at the endpoint of the in vivo spectroscopy measurement, the corresponding FE was merely 12% ± 2% (mean ± SD, N = 8), which is significantly lower than those of [4-13C]Glu and [4-13C]Gln.

Figure 6.

a: The time courses of the total concentrations of Glu, Gln, and GABA determined following acute GABA-T inhibition by intravenous administration of gabaculine (100 mg/kg, 0.6 cc) using the LCModel analysis of the POCE subspectra with the 13C editing pulse switched off. The GABA concentrations measured using the 1D doubly selective homonuclear polarization transfer method before and 1 hr after gabaculine injection are also plotted. b: FEs of [4-13C]Glu, [4-13C]Gln and [2-13C]GABA following gabaculine treatment and intravenous infusion of [1-13C]Glc. At the endpoint of the in vivo spectroscopy measurement, the total concentration of GABA increased to 6.3 ± 0.6 μmol/g wet weight (mean ± SD, N = 8); the FEs of [4-13C]Glu, [4-13C]Gln, and [2-13C]GABA increased to 35% ± 2%, 33% ± 3%, and 12% ± 2% (mean ± SD, N = 8), respectively. The time scale is referenced to the start of [1-13C]Glc infusion. Gabaculine was administered at −66 min. All time courses were fitted using polynomial or multiexponential curves for better visualization, except for the concentration of GABA (a), which was fitted by least-square linear regression.

DISCUSSION

To the best of our knowledge, the 1H/13C POCE pulse sequence proposed here (see Fig. 1) is the first method to achieve both adiabatic 13C editing and single-shot adiabatic 3D localization. The adiabatic POCE method with single-shot 3D spatial localization, combined with the use of surface transceiver coils, allowed us to minimize the sensitivity loss that is inherent in other in vivo MRS methods for detecting protons bound to 13C (e.g., Refs. 17, 20, 29, and30). To overcome the effects of B1 inhomogeneity on water suppression and OVS, the CHESS and OVS schemes were repeated six times and three times, respectively. Using 1D imaging with no water suppression, we determined that the OVS schemes alone were able to suppress the in vivo outer volume water signal by a factor of 30:1 along the x dimension, 35:1 along the y dimension, and 26:1 along the z dimension, respectively. When the adiabatic PRESS localization was switched on, the outer volume water signal was suppressed to the noise level both in vivo and in a phantom sample of distilled water. The water suppression scheme with six repetitions of CHESS generally suppressed water in vivo by a factor of >5000–10000:1, which was sufficient for the purpose of the current study. Since spatial localization was achieved by the use of hyperbolic secant pulses, which have a broad bandwidth of 11200 Hz, the largest localization error caused by chemical shift displacement of protons was approximately 0.5 mm (using the proton chemical shift difference between [2-13C]Glu and [3-13C]Lac and the x or z dimension). For the signals of interest ([2-13C]GABA, [4-13C]Glu, and [4-13C]Gln) the frequency separation between [2-13C]GABA and [4-13C]Gln in the proton spectra is 80 Hz, which corresponds to a maximum chemical shift displacement of less than 1% along the x and z dimensions. The chemical shift displacement along the y dimension is even smaller (by a factor of 1.6). Although it is not described in the text in detail, we also rigorously validated the adiabatic 13C editing scheme using three phantom samples containing [1-13C]Glc, [4-13C]Glu, and [2-13C]proprionate, respectively. Based on the largest residual peak in each phantom sample after extensive signal averaging, the suppression factor for proton signals bound to 12C was determined to be in the range of 1500:1–2000:1, far exceeding the in vivo requirements.

The bandwidth of complete WALTZ-4 decoupling of 1JCH is approximately 2 × γB2. To measure the turnover time courses of [2-13C]GABA, [4-13C]Glu, and [4-13C]Gln, the 13C bandwidth that must be completely decoupled for 1JCH is only 4 ppm. Therefore, WALTZ-4 decoupling with γB2 set to 625 Hz, which corresponds to a decoupling bandwidth of approximately 10 ppm, is sufficient for this purpose (see Fig. 5). The decoupling bandwidth has been expanded dramatically by the use of adiabatic decoupling pulses (17, 20). To decouple the entire 13C spectrum (from 20 to 57 ppm, including 13C-labeled metabolites from [3-13C]Lac to [2-13C]Glu, but excluding [1-13C]Glc, which resonates at ∼95 ppm), the γB2 of the WALTZ-4 decoupling sequence must be set at approximately 2300 Hz, which leads to excessive heating. A dramatic reduction in the requirement of peak γB2 amplitude (and therefore RF power deposition) can be realized by the use of adiabatic decoupling pulses. In this study, we found that the adiabatic decoupling scheme (20) based on a 3-ms tanh/tan pulse (ζ = 10, κ = 20, R = 150, frequency sweep width = 50000 Hz) with γB2 set to 1005 Hz produced optimal decoupling results over the intended 20–57 ppm range. With the use of our experimental setup, the rat body temperature was maintained at 37.5°C ± 0.5°C. When either of the decoupling schemes was used, the [1-13C]Glc at ∼95 ppm in the carbon spectra was not decoupled. As shown in Fig. 5, when only [4-13C]Glu, [4-13C]Gln and [2-13C]GABA are of interest, a simple WALTZ decoupling is sufficient. The lower RF power deposition at γB2 = 625 Hz from narrowband WALTZ decoupling makes it possible to use a shorter recycle delay while maintaining normal body temperature.

The spectral resolution of [2-13C]GABA and [4-13C]Glu in the in vivo proton spectra of the rat brain is achieved in this study by the use of an 11.7 Tesla spectrometer and optimal B0 homogeneity. The spectral separation of [2-13C]GABA from the overlapping [4-13C]Glu in the proton spectra has not been possible at field strengths lower than 11.7 Tesla (e.g., Refs. 17, 20, and31). An elegant semiselective POCE approach was previously proposed in the context of separating [4-13C]Glu from [3-13C]Glu in the proton spectra at 3 Tesla (18). The 13C chemical shift difference between [2-13C]GABA and [4-13C]Glu, which is 1 ppm, could also be exploited to selectively edit [2-13C]GABA and [4-13C]Glu using the same semiselective POCE approach. However, the chemical shift difference between [2-13C]GABA and [4-13C]Glu at 11.7 Tesla and below is much smaller than that between [4-13C]Glu and [3-13C]Glu at 3 Tesla. With the semiselective POCE method, a longer 13C chemical shift evolution time is required. Since the simultaneous evolution of the proton-proton J couplings is not refocused, further signal loss is expected for proton multiplets bound to 13C if the overall TE is increased. Alternatively, protons bound to 13C can be spectrally resolved based on the corresponding 13C chemical shift differences using 2D 1H/13C NMR spectroscopy methods (30, 32). The 2D 1H/13C heteronuclear multiple quantum coherence (HMQC) method (30) loses half the intensity of the 13C-labeled proton signals due to multiple quantum filtering. The heteronuclear single quantum coherence (HSQC) methods (32) can retain the full intensity of protons bound to 13C, but require more RF pulses than the HMQC method, and therefore are more susceptible to B1 inhomogeneity produced by the surface coils used here. The spatial constraints imposed by our 89-mm-core vertical magnet do not allow the use of other RF coil designs for rat studies. As shown in Fig. 4b, the alternative approaches for separation of [2-13C]GABA from [4-13C]Glu are not necessary at the field strength of 11.7 Tesla. It is also possible to detect protons bound to 13C-labeled Glu without decoupling (33). However, for dilute metabolites such as [2-13C]GABA, the effects of undecoupled long-range 1H-13C couplings could easily cause complete overlapping between, e.g., [2, 4-13C2]Glu or [2, 3, 4-13C3]Glu and [2-13C]GABA. It should also be pointed out that in the proton spectra with complete heteronuclear decoupling, the intensity of [4-13C]Glu at 2.35 ppm is the sum of all isotopomers containing [4-13C]Glu. This is very different from direct 13C detection, where homonuclear 13C-13C couplings are not affected by proton decoupling and lead to different spectral patterns for different isotopomers. With inverse detection, remote 13C also splits 1H via long-range heteronuclear couplings when 13C decoupling is switched off. When 13C decoupling is switched on, not only are the one-bond 1H-13C couplings removed, but the much weaker long-range 1H-13C couplings also collapse with adequate 13C decoupling. Therefore, there are no other isotopomer patterns in fully decoupled, indirectly detected 1H spectra. This applies to all other 13C-labeled metabolites as well.

Gabaculine is a potent and specific inhibitor of GABA-T. One hour after acute injection of gabaculine (100 mg/kg, 0.6 cc, i.v.), the cerebral GABA concentration was raised from 1.1 ± 0.1 μmol/g wet weight (mean ± SD, N = 8) to 2.4 ± 0.3 μmol/g wet weight (mean ± SD, N = 8), as determined using the selective homonuclear polarization transfer method. As shown in Fig. 6a, the total concentration of brain GABA increased linearly over the period of the in vivo experimental measurement. The rate of GABA accumulation following acute intravenous administration of 100 mg/kg of gabaculine was determined to be 0.026 ± 0.004 μmol/g ww/min (mean ± SD, N = 8). Figure 6b is consistent with the following relationship between the FEs of [4-13C]Glu, [4-13C]Gln, and [2-13C]GABA being maintained throughout the time course of [1-13C]glucose infusion: FE([4-13C]Glu) > FE([4-13C]Gln) > FE([2-13C]GABA). Since [4-13C]Glu is the metabolic precursor of [4-13C]Gln via the Glu-Gln cycle (4–6), the precursor–product relationship dictates that FE([4-13C]Glu) > FE([4-13C]Gln), which is consistent with the in vivo 1H/13C NMR spectroscopy results reported here. The turnover of GABA was previously measured in studies that used ex vivo methods with administration of labeled precursors. Without inhibiting GABA-T, the turnover times of GABA measured are severalfold shorter than those of Glu (8, 15, 16, 34, 35). With [1-13C]Glc infusion and no GABA-T inhibition, it has been consistently found that FE([2-13C]GABA) > FE([2-13C]Glu). This has been considered to be evidence of the compartmentation of Glu, since the majority of Glu is located in glutamatergic neurons, the turnover of which is much slower than that of the small Glu pool in GABAergic neurons, which acts as the direct metabolic precursor of GABA. With infusion of [1-13C]Glc approximately 1 hr after acute gabaculine treatment, we found that FE([4-13C]Gln) > FE([2-13C]GABA), which is consistent with the concept that astrocytic Gln is the primary metabolic precursor of GABA after acute GABA-T inhibition (1, 3, 13, 14).

After the gabaculine injection, both the total pool size of GABA and the FE of GABA changed over time. Mathematically, the FE of GABA reaches a plateau when Δ([2-13C]GABA)/Δt = FE([2-13C]GABA) * Δ([GABA])/Δt. That is, when the rate of label incorporation into GABA reaches a certain fraction of the rate of increase in the total GABA pool size, the rate of increase in the FE of GABA becomes zero. In contrast to the rapid label incorporation from [1-13C]Glc into [2-13C]GABA that has been observed repeatedly in the intact brain, our results obtained approximately 1 hr after acute gabaculine treatment showed a significantly reduced rate of label incorporation into [2-13C]GABA from intravenously infused [1-13C]Glc (e.g., the rate of label incorporation into [2-13C]GABA over the 90–120-min interval was estimated to be ∼0.2 μmol/g/hr from Fig. 6 by assuming a constant FE of 11% and negligible GABA release). A previous ex vivo study has reported that acute inhibition of GABA-T by vigabatrin rapidly leads to a significant reduction in the rate of label incorporation from [1-13C]Glc into [2-13C]GABA (14), which is in excellent agreement with the results of the current study. In the intact brain, the GABA shunt activity allows rapid incorporation of 13C labels from [1-13C]Glc through the small GABAergic Glu precursor pool into [2-13C]GABA. During acute inhibition of the GABA shunt, astrocytic Gln becomes the primary source of the carbon skeleton of the accumulating GABA pool. Following acute GABA-T inhibition using gabaculine, and subsequent infusion of [1-13C]Glc, the accumulation of [2-13C]GABA is mainly a measure of the trafficking of 13C labels from astrocytic [4-13C]Gln to GABAergic neurons, which, as revealed by this study, is much slower than the rate of label incorporation from [1-13C]Glc in the intact brain. Therefore, in vivo measurement of the turnover of [2-13C]GABA from [1-13C]Glc after acute GABA-T inhibition may be used to quantify the metabolic flux from astrocytic Gln to GABA, which is part of the Gln-GABA cycle between astrocytes and GABAergic neurons (1, 3, 13). Both the unlabeled precursor astrocytic Gln (the precursor of which is the large Glu pool in glutamatergic neurons via the Glu-Gln cycling pathway) and the accumulated unlabeled GABA prior to [1-13C]Glc infusion contribute to the observed significant isotope dilution of [2-13C]GABA.

CONCLUSIONS

In conclusion, we have demonstrated for the first time that [2-13C]GABA can be spectrally resolved in vivo from [4-13C]Glu at 11.7 Tesla in rat brain following the intravenous infusion of [1-13C]Glc using a single-voxel, adiabatic POCE spectroscopy method. Using this method, we determined the time-resolved kinetics of 13C label incorporation from intravenously infused [1-13C]Glc into [4-13C]Glu, [4-13C]Gln, and [2-13C]GABA simultaneously following acute GABA-T inhibition from a 40 μL voxel located in the rat neocortex, at a temporal resolution of 11.5–15 min. Previous studies have shown that the turnover of [2-13C]GABA from [1-13C]Glc in normal rats with no inhibition of GABA-T is much faster than that of [4-13C]Glu and [4-13C]Gln. In contrast, in this study we found that the FE of [2-13C]GABA in gabaculine-treated rats was much lower than that of [4-13C]Glu or [4-13C]Gln, and it increased at a much reduced rate. It is also possible to use simulated annealing-based metabolic modeling (4, 36) to analyze the measured time courses (Fig. 6b) and quantify the metabolic flux from astrocytes to GABAergic neurons following acute GABA-T inhibition (the simultaneously changing size of metabolic pools of Glu, Gln, and GABA can be taken into account, as illustrated previously (37, 38)); however, that is beyond the scope of this report. Finally, the sensitivity of the current method can be doubled by the use of the more expensive [1,6-13C2]glucose instead of the [1-13C]glucose used here.

Acknowledgements

The authors thank Drs. Steven Provencher and Su Xu for assistance with the LCModel implementation, Dr. Steve Li for construction of the 1H/13C RF probe, and the anonymous reviewers for helpful suggestions.

Ancillary