Neurochemical Research, Vol. 21, No. 9. 1996, pp. 1031-1041

The Rate of Turnover of Cortical GABA from [1J3ClGlucose Is Reduced in Rats Treated with the GABA- Transaminase Inhibitor Vigabatrin ( /-Vinyl GABA)*

D. Manor, n,s D. L. Rothman, z G. F. Mason, 3 F. Hyder, 4 0 . A. C. Petroff, ~ and K. L. Behar n

(Accepted March 6. 1996)

Brain GABA levels rise and plateau following prolonged administration of the irreversible GABA- transaminasc inhibitor vigabatrin ('y-vinylGABA). Recently it has been shown that increased GABA levels reduces GAD~ protein, one of two major isoforms of glutamic acid decarboxylase (GAD). The effects of GABA elevation on GABA synthesis were assessed in vivo using 'H and "C-edited NMR spectroscopy. Rates of turnover of cortical glutamate and GABA from intrave- nously administercd [l-L~C]glucose were measured in o~-chloralosc anesthetized rats 24 hours after recciving vigabatrin (500 mg/kg, i.p.) and in non-treated controls. GABA concentration was in- creased 2-fold at 24 hours (from 1.3 + 0.4 to 2.7 _+ 0.9 ~tmol/g) and GABA-T activity was inhibited by 60%. Tricarboxylic acid cycle flux was not affccted by vigabatrin treatment compared to non-treatcd rats (0.47 ___ 0.19 versus 0.52 _+ 0.18 [tmol/g, respectively). GABA-C2 fractional enrichment (FE) measured in acid cxtracts rose more slowly in vigabatrin-treated compared to non- treated rats, reaching >90% of the glutamate FE after 3 hours. In contrast, GABA FE > glutamate FE in non-treated rats. A metabolic model consisting of a single glutamate pool failcd to account for the rapid labeling of GABA from glutamate. Metabolic modelling analysis bascd on two (non- communicating) glutamate pools revealed a -70% decrease in the rate of GABA synthesis fol- lowing vigabatrin-treatment, from 0.14 (non-treated) to 0.04 p.mol/g/min (vigabatrin-treated). These findings, in conjunction with the previously reported differential effects of elevated GABA on the GAD isoforms, suggests that GADs7 may account for a major fraction of cortical GABA synthesis in the a-chloralose anesthetized rat brain in vivo.

KEY WORDS: GABA; gamma-aminobutyric acid; glutamic acid; compartmentation; in vivo NMR spectros- copy; glutamic acid decarboxylasc; GABA transaminase.

I N T R O D U C T I O N

Gamma-aminobutyr ic acid (GABA) is the major in- hibitory neurotransmitter in rat and human cortex (1-5).

k Departments of Neurology, 21ntcrnal Medicine and "Molecular Bio- physics and Biochemistry, Yale University School of Medicine, New llavcn, Connecticut 06520, and ~Centcr for NMR Research and De- velopment, University of Alabama, Birmingham, Alabama 35294.

s Address reprint requests to: I)r. David Manor, Magnetic Resonance Center, P.O. Box 208043, Yale University School of Mcdicine, New Haven, Connecticut 06520-8043.

" Special issue dedicated to Dr. l lerman Bachclard.

1031

Extensive studies in animals have shown that G A B A has a pivotal role in suppressing the origin and spread of seizure activity (3-5). The level o f G A B A in synaptic terminals and in the extracellular fluid is dependent on the functioning o f a metabolic cycle between neurons and gila (6-10). Although this cycle is critical for main- taining normal GABAerg ic function, the in vivo regu- lation o f this cycle, as well as, the fraction o f total G A B A involved are unclear.

G A B A is synthesized from glutamate by the cata- lyric action o f glutamic acid decarboxylase (GAD). The activity of GAD is believed to be primarily responsible

0364-3190/96/0'9(10-1031 $09.50.0 ~ 1996 Plenum Publishing Corporation

1032 Manor, Rothman, Mason, Hyder, Petroff, and Behar

for regulating the concentration of GABA in vivo (5,11- 13) through the pyridoxal-5'P dependent interconversion of holo- (active) and apoenzyme (inactive) forms (13,14). In vitro, the proportion of GAD in the apo or holo state at any one time can be affected by the levels of several different modulators (14). Studies of GAD in vitro have shown that GABA is not effective as a feedback inhibitor due to the high concentration of GABA required for in- activation of GAD (I~ -- 16 mM) relative to the concen- tration of GABA (1-2 mM) in cerebral cortex. In vivo, the finding that GABA accumulation in the first few hours following GABA-transaminase (GABA-T) inhibition is linear with time (15) indicates that the concentration of GABA is not within the K~ region of GAD. Although highly specific GABA-T inhibitors such as vigabatrin (~- vinyl GABA) (16) or gabaculine (17,18) do not affect GAD activity in vitro, the observation that GAD activity is decreased following GABA-T inhibition (16,19) has not been readily explained.

Although total GAD activity is reduced following inhibition of GABA-T, the percentage reduction is small compared with that of GABA-T. Recent studies have revealed that GAD is composed of two major isoforms (65 kD and 67 kD proteins, (20,21)) which are the prod- ucts of two different genes (21). The two isoforms show different regional and subcellular distributions (22-25), as well as, differences in their response to certain effec- tors including GABA and degree of saturation by pyri- doxal phosphate (20,26-28). The level of GAD67 protein is sensitive to GABA concentration and decreases as GABA levels rise after a lag time of about 10 hours (27). Therefore, the time-dependent decrease in GAD67 fol- lowing GABA elevation may explain the decrease in to- tal GAD activity following prolonged GABA-T inhibition in vivo. However, the function of the GABA pool synthesized by GAD67 is not presently known.

We have used in vivo nuclear magnetic resonance (NMR) spectroscopy to measure the concentration of GABA in vivo in normal animal and human cerebral cortex (15,29) and in epilepsy patients treated with vi- gabatrin (30). In vivo lH NMR spectra obtained from the occipital lobes of epileptic patients revealed that GABA levels rise to a maximum level of 2-fold with increasing dose of vigabatrin (up to 3 g/day) above which there is no further increase of GABA (31). Fur- thermore, after several months on constant vigabatrin treatment human brain GABA gradually declined toward pretreatment levels in some patients (31). The clinical findings suggest that a prolonged increase in GABA lev- els may reduce GABA synthesis in human brain.

In the present study we have applied in vitro and in vivo localized magnetic resonance spectroscopy of rat

brain, in conjunction with 13C-labeled glucose, to inves- tigate the effect of an increase of GABA concentration on the turnover of the cortical GABA pool and the use of the latter as an indicator of GAD activity in vivo.

EXPERIMENTAL PROCEDURE

Animal Preparation. Male Sprague-Dawley rats (200-300 g), were injected with vigabatrin (0.5 g/kg, i.p.) 24 hours prior to the NMR study. Non-vigabatrin treated animals served as controls. Food was withheld for all animals 24 hours prior to the study although access to drinking water was not restricted. Rats were tracheotomized and ventilated (70% N20/30% 02) under 1-1.5% enflurane or halo- thane anesthesia. Core temperature was maintained near 37~ by use o f a heating pad using recireulated warm water. A femoral artery and vein were cannulated for measurement of arterial blood gases (pOz, pCO2), pH, mean pressure, and glucose concentration, and the intra- venous infusion of ~3C-labeled glucose. The scalp was removed for the placement of the surface transceiver coil. After surgery, anesthesia was maintained with alpha-chloralose administered via an intraperi- toneal catheter (80 mg/kg, initial dose supplemented with 20 mg/kg every 30 min) and the inhalation anesthetic was discontinued. Venti- lation was adjusted to maintain arterial blood pO2 between 120-150 mmHg and pCO z between 25-35 mmHg. Mean arterial blood pressure was between 80-100 mmHg.

Experimental Protocol. Animals were infused intravenously with [1-~3C]glucose using an infusion protocol designed to increase plasma glucose rapidly and maintain a nearly constant level of 10-13 mM with a constant t3C isotopic fractional enrichment (32). Blood samples were obtained every 20-30 minutes for measurement of plasma glu- cose concentration and the plasma was frozen for subsequent deter- mination of the glucose-C 1 fractional enrichment (see below). For both vigabatrin treated and non-treated rats, the isotopic turnover of gluta- mate was determined from the in vivo time course of glutamate-C4 labeling from [1-~3C]glucose. Animals were removed from the magnet at various times into the labeled glucose infusion (60, 90, 120, 180 min) and their brains were frozen in situ with liquid N 2. In order to characterize the early appearance of t3C-label in GABA, several ad- ditional groups of rats were infused with [1-~3C]glucose for lesser times (8, 15, and 30 min.) but these animals were not studied in vivo. For these animals a single value of ~3C-fractional enrichment for glu- tamate-C4 and GABA-C2 was obtained from the brain extract at the specified time. Following in situ freezing and storage at -90~ the frozen brains were exposed and a portion of the cortex (50-150 mg), located 2 mm behind the bregma (center of NMR detected volume), was removed and weighed. Frozen brain tissue was extracted in HCl/methanol and perchloric acid and dissolved in I)20 as described previously (15). A known amount of 3-trimethylsilyl-[2,2,3,3-ZH]-pro - pionate (TSP) was added as a chemical shift reference and concentra- tion standard.

In a separate group of animals GABA accumulation in the brain was observed for 2-2.5 h following the acute administration of the potent GABA-T inhibitor gabaculine (100 mg/kg, i.a.). This study was performed to obtain an independent estimate of the rate of GABA synthesis (12,15).

NMR Spectroscopy in Vivo. NMR spectra in vivo were obtained using a highly modified 7 Tesla Bruker Biospec NMR spectrometer (Bruker Instruments) interfaced to a 20 cm diameter horizontal-bore magnet (Oxford Magnet Technologies) with actively shielded gradi-

Cortical GABA Turnover Is Reduced Following Vigabatrin 1033

Fig, 1. MRI of the brain region used for spectroscopic measurements. Gradient echo images of coronal and sagittal slices of the rat brain (TR, 0.2s; TE, 20 ms; resolution, 156 x 156 gm; slice thickness, 500 gm). Spectroscopic measurements were obtained from the indicated vohLme within the cortex 2 mm posterior to the bregma (b). Dashed lines mark the positions of the orthogonal slices. A 2D ISIS sequence was used to localize the signal to a 6 • 6 mm (x,z) column. RF pulse power was adjusted (along the y-axis) to acquire most of the signal from within the first 3 mm below the brain surface. Brain tissue for acid extraction was obtained from the same region. Ant., anterior.

ents. The NMR probe consisted of concentric 8 mm diameter ~H (300.84 MHz) and 20 mm diameter ~3C (75.65 MHz) surface coils. The inner ~H surface coil was centered 2 mm posterior to the bregma. Gradient-echo water ~H images were acquired in the coronal and sag- ittal planes to visualize cerebral anatomy (Fig. 1) and small adjust- ments of position were made to locate the selected region at the magnet isocenter. A volume of 6 • 6 • 4 mm 3 (Fig. 1) was selected for localized optimization of field homogeneity using a STEAM pulse sequence (33,34).

In vivo spectra were obtained from a localized volume of cortex using the ISIS technique (35) to preselect the volume of interest. A 6 • 6 mm 2 column was selected along the x and z-axes (z-axis is parallel to magnet Z0-axis ) using 2D ISIS. The column was centered in the brain midline and the radio frequency pulse power was adjusted using the water signal profile along the y-axis to achieve maximum signal intensity from the uppermost 3 mm of the cortex.

GABA Editing in ~H NMR Spectra. GABA was measured in ~H NMR spectra using a J-editing pulse sequence as described previously (15,36) and modified to include spatial localization. Briefly, the pulse sequence consisted of a spin-echo using a hard excitation pulse fol- lowed by a 2-2 refocusing pulse. The TE time was 76 ms, where TE/2

= 1/4 JHH of the GABA C2 and C4 triplets, thus nulling the signal of these resonances during acquisition. In alternate scans the resonance of the GABA C2 at 1.9 ppm was inverted using a DANTE pulse-train positioned symmetrically around the refocusing pulse. When the two subspectra are subtracted (with DANTE and without DANTE), only resonances J-coupled to the resonance at 1.9 ppm remains. In practice, in the absence of macromolecule suppression techniques, J-editing with the DANTE at 1.9 ppm yields both the GABA C4 methylene protons and a macromolecule peak coresonating at 3.0 ppm (37). For basal GABA levels, macromolecules contribute at most 40% to the intensity of the edited 3.0 ppm signal (15) depending on the bandwidth of the DANTE inversion. Therefore, GABA levels measured in vivo were normalized to the intensity of creatine and expressed as the dif- ference between vigabatrin-treated and non-treated animals.

Glutamate ~C-Editing in ~H NMR Spectra. The isotopic turnover of glutamate was measured in ~H spectra using a proton-observed, carbon-edited (POCE) pulse sequence (32,36). This sequence is used to detect ~H resonances coupled to ~3C. 1H spectra were acquired with a spin echo pulse sequence (TE = 20 ms, TR = 2 s) using either a hard or adiabatic 90 ~ excitation pulse followed by a 2-2 refocusing pulse. Two balanced crusher gradients were used in each half of the spin-echo sequence to eliminate non-refocusing magnetization. In al- ternate scans ~3C-coupled 'H resonances were inverted using a 180~ pulse placed 4 ms (1/2Jc_H) after the 90~ excitation pulse, centered between the resonances of glutamate C4 and C3 in the ~3C spectrum. 13C composite pulse decoupling was used in both scans during acqui- sition. Subtraction of the t3C inverted from the non-inverted spectrum yielded only the proton resonances coupled to ~3C. 13C inversion and decoupling power was optimized to 40 and 5 W, respectively using a solution of [2-~3C]acetate. The POCE edited NMR spectrum was phase corrected between 4.0 and 0.5 ppm, and baseline corrected (zero- and first-order corrections) between 2.6 and 1.8 ppm. The t3C-labeled glu- tamate was quantitated by measurement of the peak height of the C4 methylene proton resonance at 2.35 ppm in the edited spectra.

~3C Enrichment of Glutamate and GABA in Acid Extract. High resolution tH NMR spectra of brain extracts were obtained at 360.13 MHz using a Bruker Instruments AM-360 NMR spectrometer with a previously described POCE pulse sequence (32). For determination of metabolite concentrations the spectra were zero filled to 32 K points followed by Lorentzian-to-gaussian multiplication (parameters: LB = -0 .75, GB = 0.1, acquisition time, 1.36 s). The peak areas of GABA- C2, glutamate-C4 and total creatine were measured after zero- and first-order baseline correction. Concentrations were calculated assuming a creatine concentration in rat brain cortex of 10 gmol/g wet weight. ~3C enrichments of glutamate-C4 and GABA-C2 in tH NMR spectra were determined from the ratio of the height of the center peaks of the respective multiplet in the difference spectrum (containing only t3C-coupled ~H resonances) to the non-inverted ~H sub-spectrum (containing the sum of uC and ~zC bonded ~H resonances). Based on spectra of pure GABA and glutamate in solution there was minimal resonance overlap at the position of the multiplet lines used for the measurement. A repetition time of 21 see was used to avoid saturation due to incomplete T t relaxation.

Enzymatic Assay of GABA-T Activity. GABA-T activity was measured in crude homogenates of rat brain tissue. Approximately 100-150 mg of frozen brain tissue removed from the fronto-parietal cortex was homogenized in ice cold de-ionized water (10:1 vol/wt). The suspension was centrifuged at 900 g at 4~ for 15 rain. GABA- T activity in the supernatant was determined at 30~ using a glutamate dehydrogenase coupled spectrophotometric assay (38,39).

Determination of Tricarboxylic Acid Cycle and GABA Synthesis Rates. A mathematical model of glucose metabolism (40,41) was rood-

1034 Manor, Rothman, Mason, Hyder, Petroff, and Behar

A t/f glucose

B glucose

; glutamate glutamate

Fig. 2. Two models were tested to determine the rate of GABA syn- thesis. Schematic representation of the metabolic flux from glucose, through the TCA cycle (TCA), into glutamate and GABA. (A) One- compartment model in which the total glutamate pool is the precursor of GABA. (B) Two-compartment model in which GABA is synthe- sized in a separate, small metabolic compartment, which contains only a fraction of total cortical glutamate. The second model is represen- tative of GABA being synthesized in GAD-containing GABAergic neurons. It should be noted that GABA degradation by GABA-T (GT) occurs in all neurons as well as in glial cells (not shown here) and the latter may contribute to the precursor glutamate pool through gluta- mate and glutamine exchange.

ified to incorporate GABA synthesis. The tricarboxylic acid cycle rate (V,r and the GABA synthesis rate were determined from a fit of the model to time courses of glutamate-C4 and GABA-C2 labeling. A Runge-Kutta method was used to solve the differential equations, and best fits were determined by iteration of parameters using a simplex algorithm. Two models were tested to determine the rate of GABA synthesis (Fig. 2). In the first model (one-compartment model) the fractional enrichment of the total glutamate pool (measured directly with NMR) was specified as the precursors enrichment and the rate of GABA synthesis was iterated in order to find the best fit to the time course of enrichment of GABA-C2. In the second model (two-com- partment model), the isotopic turnover of the glutamate pool that serves as the precursor for GABA was treated as instantaneous. The two-compartment model provides a minimum estimate of both the GABA synthesis rate and the fractional inhibition of cortical GABA synthesis in rats treated with vigabatrin.

The TCA cycle rate was determined from the time course of enrichment of glutamate-C4 measured directly in vivo. In contrast, the rate of labeling of GABA (and the rate of GABA synthesis) was de- termined ex vivo from measurements of acid-extracted brain tissue taken at the final time points of glucose infusions varying between 8 and 183 min. The direct measurement of ~3C incorporation into GABA-C2 in vivo using the current methods was not feasible because

c o n t r o l v i g a b a t r i n

cr Cr

GABA C4 B GA,A c4

4 3.5 3 2.5 4 3.$ 3 2.5 pprn ppm

Fig. 3. GABA is increased in ~H-edited spectra in vivo at 24 hours following vigabatrin. Edited 1H spectra of GABA-C4 obtained in vivo of non-treated (A,B) and vigabatrin-treated (C,D) animals. Spectra A and C represent the J-modulated subspectra in which GABA and other triplets with similar coupling constants are suppressed. Spectra B and D represent difference spectra (i.e., edited spectra) obtained between the subspectra shown in A and C (without DANTE) and one where a DANTE was applied to invert the GABA multiplets (not shown). Dif- ference spectra (scaling factor, X3) show that GABA C2 peaks dou- bled in intensity. No difference was observed in the total creatine signal (Cr) indicating that the increase in the edited GABA resonance was due to an increase in GABA concentration. TE, 76 ms.

of spectral overlap between GABA-C2 and glutamate-C4 and insuf- ficient signal-to-noise per unit time due to the relatively low level of GABA.

Because glutamate-C4 and GABA-C2 were not resolved in vivo, the TCA cycle rate was calculated from the glutamate-C4 enrichment time course data after subtraction of the contribution of labelled GABA-C2 (i.e., (Glu4* - GABA*)), based on the measured ex vivo time course of GABA-C2 turnover. The degree of spectral overlap between glutamate-C4 (2.35 ppm) and GABA-C2 (2.29 ppm) on glu- tamate-C4 enrichment was estimated using 50 mM solutions of GABA and glutamate in water. GABA-C2 was found to contribute 40% of its peak amplitude (based on a line-broadening of 15 Hz) to the glu- tamate-C4 peak at 2.35 ppm. The final resulting amplitude of (Glu* - GABA*) was scaled to match the fractional enrichment of Glu-C4 measured in the acid extract for that particular animal and the time course measured in vivo was scaled accordingly. The corrected time courses were used to determine V,oa. A steady state dilution of 25% was assumed for cortical glutamate-C4, using the following equation:

% dilution = [1 - ((Glu4*/Glu4)/(1/2 x GIc~*/Glcl))] • 100,

where Glu4*/Glu4 and Glct*/Glc~ were measured in acid extracts from a larger, but separate, group of ct-chloralose anesthetized rats. These animals were infused with [1-13C]glucose over a period of 3 hrs, by which time glutamate-C4 had reached an isotopic steady state.

R E S U L T S

GABA Elevation Following Vigabatrin. G A B A lev-

e ls w e r e i n c r e a s e d in J - e d i t e d s p e c t r a m e a s u r e d in ce r -

e b r a l c o r t e x in v i v o t w e n t y f o u r h o u r s f o l l o w i n g a s i n g l e

d o s e o f v i g a b a t r i n (500 m g / k g ) c o m p a r e d to n o n - t r e a t e d

Cortical GABA Turnover Is Reduced Following Vigabatrin 1035

control

A Glu C4 ,~ InC4 /'\ ,GABAC2 II

vigabatrin

C Giu c4 ""1 / N GABA CZ ]1

Glu C4 , ,~

B o, . ,c 2 D o,.,c2

[~ GInC, f~! ( /~ GIxC3 [ ~ GInC4 I l l ~ GIxC3

2.6 2.5 2.4 2.3 2.2 2.1 2 2.6 2.S 2.4 2.3 2,Z Z,1 Z pprn ppm Fig. 4. High resolution 13C-edited, ~H NMR spectra of acid extracts of vigabatrin-treated and non-treated rat cortex. (A,B) Acid extract of non- treated rat cerebral cortex. (C,D) Acid extract of rat cerebral cortex 24 hours after vigabatrin treatment (500 mg/kg, i.p.) following a 30 min infusion of [1J3C]glucose. The intensities of resonances in spectra A and C represent the total concentration of the compounds shown whereas spectra B and D represents the ~3C-labelled compounds only. Vertical scale in B and D was increased 3 times. Vigabatrin treatment lead to an increase in the ~H multiplets of GABA-C2, -C3, and -C4 (C) but no change in glutamate (compare Glu-C4 and C3 in spectra A and C). Lactate levels were low in the extracts (not shown) indicating minimal autolysis which can artificially increase GABA concentration. The trtmcated peak at 2.02 ppm is the methyl group of N-acetylaspartate; Gin, glutamine, Glu, glutamate, GABA, ~-aminobutyric acid, Glx, glutamate + glutamine. Propylene glycol, used as a vehicle for the a-chloralose (doublet at 1.13 ppm, not shown) was observed in extract and in vivo spectra.

animals (Figs. 3 and 4). The GABA-to-creatine peak in- tensity ratio 6 increased from 0.34 ___ 0.03 (n = 5) to 0.51 ___ 0.08 (n = 6) after vigabatrin-treatment. The change in the GABA-to-creatine ratio (AGABA/Cr = 0.17 +__ 0.10) corresponded to an average increase in GABA concentration of 1.7 _ 1.0 ~tmol/g (2-fold), which was similar to the change in GABA concentration determined in the acid extracted cortical tissue of 1.4 _ 0.9 (non-treated, 1.3 _ 0.4 ~tmol/g, n = 18; vigabatrin treated, 2.7 __+ 0.9 ~tmol/g, n = 21). Glutamate concen- trations determined in brain extracts were not signifi- cantly different between vigabatrin-treated (9.1 ___ 2.7 ~tmol/g) and non-treated (9.2 + 1.3 ~tmol/g) animals. GABA-T activity measured in vitro was reduced in the vigabatrin-treated rat cortex at 24 hours to 40% of non- treated control levels (non-treated, 10.8 + 4.0 ~tmol/g/h, n = 4; vigabatrin-treated, 4.1 + 2.3 ~tmol/g/h, n = 5).

Effects of Vigabatrin on Tricarboxylic Acid Cycle Flux in Vivo. Intravenous infusion of [1-13C]glucose led to the labelling of glutamate-C4 and C3 methylene pro- tons in localized 13C-edited ~H N M R spectra obtained from rat cortex (Fig. 5). The time course o f x3C labelling o f cortical glutamate from [1J3C]glucose is shown in

6 The GABA-to-ereatine ratio was corrected for the difference in num- ber of protons between the GABA C4 methylene and creatine methyl groups (multiplication by 3/2) and assuming that the creatine peak intensity represents 10 larnol/g wet weight.

A NAA

B

Glu C4 Glu C4 u C3

3.5 3 2.5 2 1.5 1 0.5 ppm

Fig. 5. Localized ~3C-edited, IH NMR spectrum showing glutamate ~3C labeling from [1-13C]glucose in rat cortex in vivo. A Short TE subspectrtun (TE, 20 ms; TR, 2 s; number of scans, 128; accumulation time, 4.3 min) which represents total concentration (~3C + ~2C). B. The ~3C-edited difference spectrum (scaled • shows the 13C-labeled, ~H resonances of glutamate-C4 and -C3. The spectrtun was obtained 80 rain after the start of the [1J3C]glucose infusion. Cr, total creatine; NAA, N-acetylaspartate; PG, propylene glycol (used as a vehicle for the a-chloralose).

Fig. 6 for a vigabatrin-treated and a non-treated rat. Glu- tamate-C4 approached an isotopic steady state at a rate

1036 Manor , Rothman, Mason, Hyder , Petroff, and Behar

A 0.8] control

[ o.1 oiO Oo.: / <vt.. "(y!" , .

B 0.8] v igabatr in 0.71

t N 0 .5

3 o41 /o E E .o W �9 ~ e ' -

:~ ~-" 0.Z (Vtc a = 0 . 4 0 i~mole/g/min)

~ o so 1oo ,5o 260

t ime (min)

Fig. 6. Comparison of the cortical rate of ~3C isotopic turnover of glutamate-C4 from [1-13C]glucose between non-treated and vigabatrin- treated rats in vivo. Time courses of the change in the ~3C-labeled fraction of glutamate-C4 measured in vivo using localized ~3C-edited, ~H NMR. (A), Non-treated (control). (B) Vigabatrin-treated (B). Lines through the points represent the best fit of the data to a mathematical metabolic model yielding the TCA cycle rates (Vto~). The ~3C fractional enrichment of glutamate-C4, which was measured in acid extracts of cortical tissue frozen in situ at the end of the infusion, was used to convert peak amplitude data measured in vivo to fractional enrichment. Values shown were also normalized to the ~3C fractional enrichment of glucose-C1 in blood plasma.

that was independent o f vigabatr in. The in v ivo tricar-

boxyl ic acid (TCA) cycle rate, as derived from the t ime courses o f the 13C isotopic turnover o f g lu tamate-C4 us- ing metabol ic mode l l ing analysis (40,41), was unchan-

ged at 24 hours fol lowing vigabatr in (vigabatrin-treated,

0.47 + 0.19 ~tmol/g/min (n = 6) versus non-treated, 0.52 ___ 0.18 ~tmol/g/min (n = 6)).

Effects of Vigabatrin on GABA and Glutamate Turnover. The t ime course of G A B A - C 2 label ing was de termined from brain extracts o f rats infused with [1- ~3C]glucose for periods o f 8, 15, 30, 60, 90, 120 and 180 minutes (Fig. 7). In order to min imize the effects o f var- iat ions in the absolute fractional enr ichment o f the glu-

tamate precursor pool be tween animals , the fractional

enr ichments of G A B A and glutamate were plotted as their ratio against the t ime o f the 13C labeled glucose

infus ion (Fig. 7A). In non- t reated animals , the G A B A -

C2 to glutamate-C4 enr ichment ratio rose quickly, and was > uni ty at the earliest t imes measured (8 min) . In

A 1.4. 1.2'

~r 1'

~ ~ 0.8' E

~ (~ 0.6 4.w

uu u. 0.2.

S 1-

,- , 0.8-

"= ~o.z ---o---control

[] GVG

50 1 O0 150 200 t (rain)

Fig. 7. Turnover of GABA is decreased at 24 hours following a single dose of vigabatrin. A. The ratio of ~3C fractional enrichment (FE) of GABA-C2 to glutamate-C4 ((FE GABA-C2)/(FE GLU-C4)) were measured from spectra of brain extracts following [1-13C]glucose in- fusions of various durations. In control animals (circles. Dashed line is an exponential fit) the enrichments of GABA and glutamate are already similar after 8 minutes of infusions. In vigabatrin-treated rats (squares, solid line is an exponential fit) the FE of GABA reaches that of glutamate only after 3 hours of infusion. B. The time courses of ~3C labeling of GABA-C2 for non-treated (circles) and vigabatrin- treated (squares) animals were fitted by the 2 compartment model as described in the text. Dashed and solid lines represent the fit to the non-treated and vigabatrin-treated rats, respectively. The calculated rates of GABA synthesis (in gmol/g/min) were 0.137 in non-treated and 0.045 in vigabatrin-treated rats. Points represent mean + SD of 2 to 6 measurements.

contrast to the rapid turnover of G A B A in non-treated rats, the fractional enrichment of GABA-C2 in vigabatrin-

treated rats lagged that o f glutamate C4 and the two ap- proached the same level (i.e., GABA-C2 FE/glutamate-C4

FE = 1) only after more than 180 minutes of the [1- ~3C]glucose infusion (Fig. 7A). In both treatment groups the concentrations of G A B A and glutamate remained con- stant during the glucose infusion.

Determination of the Rate of GABA Synthesis and the Effects of Vigabatrin. The rate o f G A B A synthesis was determined by mathemat ica l mode l ing o f the N M R x3C isotopic label ing data. Initially, the rate o f G A B A

synthesis was determined by us ing the average rate o f glutamate label ing de termined in vivo as the enr ichment

o f the glutamate precursor pool (see Fig. 2). This anal-

Cortical GABA Turnover Is Reduced Following Vigabatrin 1037

ysis assumes that the total glutamate pool measured with NMR serves as the precursor of GABA measured in the extract. Because the relationship between the rates of the TCA cycle and GABA synthesis in vivo is unknown and may not be independent, the rate of GABA synthesis was determined from a subset of the individual time courses for those animals with a value of Vtca within +__ 2SD of the group mean. When the data were fitted to this model, the rate of GABA synthesis in vigabatrin- treated rats was 0.035 gmol/g/min while the rate ap- proached infinity in non-treated rats. The unrealistically high rate of GABA synthesis found for the single glu- tamate pool model in the non-treated animals arises from the near equivalence of the fractional enrichments of GABA-C2 and glutamate-C4 measured at early times during the infusion (Fig. 7A). In order to estimate the lowest possible rate of GABA synthesis that would be consistent with a single glutamate pool model, the time course of GABA labelling was compared with simulated time courses based on a single glutamate pool model using a range of GABA synthesis rates between 0.1 and 10 X Vtca. A fit of the simulated time course to the actual data required the rate of GABA synthesis to be greater than or equal to the TCA cycle flux, which is 0.8-2 times the reported V~ax of GAD (13). Therefore, the data were fitted to a metabolic model whereby GABA is labeled from a smaller (more rapidly turning over) pool of glu- tamate not in chemical exchange with the large pool of glutamate measured with NMR. In the extreme case, where the turnover of the precursor glutamate pool is allowed to be instantaneous due to its small size, the turnover of glutamate-C4 can be ignored and the GABA- C2 enrichment time course analyzed directly (Fig. 7B). This analysis provides a minimum estimate of the rate of GABA synthesis and the difference between the two treatment groups. Based on this model the best fit to the GABA enrichment data yielded a rate of GABA synthe- sis of 0.14 gmol/g/min in non-treated and 0.04 gmol/g/min in the vigabatrin-treated rats. The calculated fractional enrichment of GABA-C2 at steady state in non-treated rats approached 66% (34% dilution) of that predicted if precursor glutamate was derived solely from plasma glucose (i.e,, 1/2 • glucose-C1 FE). The FE of GABA-C2 in vigabatrin-treated rats was 80% (20% di- lution) of its potential maximum. In both cases, the di- lution of ~3C labelling indicates the presence of a significant inflow into GABA of unlabelled substrate (see discussion in reference (40)).

In order to obtain an independent estimation of the rate of GABA synthesis, the rise of GABA concentration was measured in vivo following the acute administration of the potent GABA-T inhibitor gabaculine (100 mg/kg,

i.a.) (15). For these animals under ~-chloralose anesthe- sia (data not shown), GABA increased linearly in edited 1H spectra in vivo after an initial lag time of 50 minutes. Based on the GABA concentration in the acid extract of the in situ frozen brain after 60-90 rain, GABA in- creased at an average rate of 0.03 + 0.01 ~tmol/g/min (n = 3), providing a lower, but independent, estimate of the rate of GABA synthesis.

DISCUSSION

Determination of GABA Synthesis from ~3C Isotopic Turnover. The present study shows that the rate of GABA synthesis in the rat cortex was reduced 3-fold at 24 hours following partial inhibition of GABA degra- dation. We based this conclusion on the ex vivo time courses of ~3C incorporation into glutamate-C4 and GABA-C2, which demonstrated a decreased rate of turn- over of GABA from glutamate following vigabatrin treatment, i.e. the fractional enrichments of the two me- tabolites equilibrated much earlier in non-treated than in vigabatrin-treated animals. The following factors were considered in the derivation of the rate of GABA syn- thesis from the data:

i) The TCA Cycle Flux. Due to the fast isotopic exchange between a-ketoglutarate and glutamate, the time course of isotopic enrichment of glutamate-C4 re- flects the TCA cycle flux (40,41). With all other para- meters constant, equilibration of GABA and glutamate isotopic enrichments will occur earlier at lower TCA cy- cle rates which slows down glutamate labeling. It is pos- sible that variations in the rate of glucose oxidative metabolism could affect GABA synthesis. Therefore, we deliberately chose a dose of vigabatrin that would induce only a moderate elevation of GABA and which was sim- ilar to that measured in brains of epilepsy patients treated with vigabatrin (30,31). We found no significant differ- ence in the TCA cycle rates between vigabatrin-treated and non-treated animals.

ii) GABA and Glutamate Concentration. High doses of vigabatrin or gabaculine reduce glutamate lev- els in the first few hours after administration (15,42) but this change may be transient because normal levels of glutamate have been reported following prolonged treat- ment (61). The dose of vigabatrin used in the present study (500 mg/kg, i.p.) had no effect on glutamate con- centration measured 24 hours after administration. The higher concentration of GABA in vigabatrin-treated rats would be expected to increase the GABA turnover time. In order to correct for the difference in pool sizes be- tween the two groups, the mean concentration of GABA

1038 Manor, Rothman, Mason, Hyder, Petroff, and Behar

was used in the analysis of the time courses. GABA concentration did not vary with time of infusion in either treatment group. To first order, variations in GABA con- centration within a treatment group would decrease the precision of the calculated rate of synthesis without bias to higher or lower values.

iii) Precursor Pool Enrichment. Findings from other laboratories and the present work suggest that only a fraction of the total glutamate pool serves as the pre- cursor for GABA in vivo. First, the enrichment of GABA-C2 has been reported to exceed glutamate-C4 in 13C spectra of guinea-pig cortical slices perfused with [1-~3C]glucose (43). In the present study the fractional enrichment of GABA-C2 in non-treated animals tended to be greater than that of glutamate-C4 for some early time points (Fig. 7A). However, the significance of this observation was questionable because glutamate and GABA in the extract were not purified so that the pos- sibility of unlabeled species co-resonant with glutamate- C4 could not be excluded. Second, the calculated rate of cortical GABA synthesis based on a one-pool meta- bolic model was unreasonably high in non-treated rats. A minimum estimate of GABA synthesis was deter- mined for animals in both treatment groups by assuming that the turnover rate of the glutamate precursor pool was instantaneous. This procedure also provided a min- imum estimate of the fractional difference in the rate of GABA synthesis in both groups of animals, As dis- cussed below, a comparison of the GABA synthesis rate obtained from 13C isotopic turnover and GABA accu- mulation after gabaculine-induced inhibition of GABA- T, supports the existence of a small glutamate pool as the precursor of GABA. Our results are consistent with a metabolic modelling analysis of postmortem changes in GABA specific radioactivity in rat forebrain following a short pre-labelling from [U-14C]glucose which indi- cated that the glutamate precursor pool leading to GABA synthesis was only --2% of total glutamate (44).

An independent estimate of the rate of GABA syn- thesis was needed in order to estimate the size and turn- over time of the glutamate precursor pool. Therefore, rats were treated with gabaculine, which is a more potent inhibitor of GABA-T than vigabatrin (IC5o = 1.8 gM and 350 gM, respectively (45)) at a dose which is known to give 50-100% inhibition of GABA-T (46,47). We have previously shown that within the first several hours after gabaculine administration, GABA concentration rises linearly with time suggesting that GABA-T inhi- bition and rising GABA levels do not affect the rate of GABA synthesis acutely (15). Assuming that the rate of GABA accumulation following gabaculine represents at minimum half the rate of GABA synthesis, due to in-

complete inhibition of GABA-T, then the rate of GABA synthesis would be at least 0.06 gmol/g/min, which is similar to the rate determined from 13C isotopic labeling when the analysis was based on a precursor glutamate pool of negligible size. In contrast, when the total glu- tamate pool was used as the precursor to GABA, ex- tremely high rates (approaching infinity) of GABA synthesis were obtained, which is clearly inconsistent with the gabaculine-based measurement and previous studies (6,15,44,49).

Relationship Between GABA Synthesis and TCA Cycle Flux. The rate of GABA synthesis of 0.14 gmol/g/min determined in the present study is similar to the rate of 0.13 gmol/g/min estimated from postmortem changes in specific activity of 14C-labelled GABA from [U-14C]glucose in unanesthetized rats (44). In contrast to previous studies, ours is the first measurement of GABA synthesis made under conditions of a constant (and known) 13C enrichment of plasma glucose in vivo per- mitting application of metabolic modelling to obtain fluxes directly from the time courses of 13C isotopic en- richment. Previous reports of GABA synthesis have been obtained from postmortem preparations (44,48) un- der conditions where several of the known effectors of GAD are changing in the direction of stimulation (e.g., decreased ATP, increased P~ and H + (28,49)) or in brain slices (6) where synaptic metabolism and function are markedly depressed. Therefore, it is not clear how GABA synthesis under these conditions relates to the brain in vivo. Notably, the rate of GABA synthesis measured in the a-chloralose anesthetized brain was 26% of the TCA cycle flux, which is much greater than previous estimates of 8-10% (6,44). However, the TCA cycle flux in cortex of rats anesthetized with e~-chlora- lose (0.52 gmol/g/min) is 3-fold lower than under ni- trous oxide (1.58 gmol/g/min (32)), where the latter is believed to have little or no effect on oxygen consump- tion and TCA cycle flux (see refs in 32,50). Therefore, the near agreement between the value of GABA synthe- sis measured in the present study and previous reports from unanesthetized rats suggest that GABA synthesis was affected to a lesser degree by the anesthetic than the TCA cycle flux. An apparent lack of respiratory con- trol of GABA shunt flux has been shown in isolated rat brain mitochondria (51) suggesting that GABA synthesis and GAD regulation may operate separately of mito- chondrial respiration in GABAergic neurons.

Evidence for Compartmentation of GABA Metabo- lism in Vivo. Much evidence has been presented for met- abolic compartmentation of GABA and glutamate metabolism in the CNS (6,10,43,48,52-59). Rapid iso- topic labeling of GABA relative to glutamate is consis-

Cortical GABA Turnover Is Reduced Following Vigabatrin 1039

tent with compartmentation of glutamate into a small rapidly turning over pool of precursor glutamate and a larger pool not serving in GABA synthesis (10,43,59). The higher relative GABA-C2/glucose-C1 enrichment obtained from [1-13C]glucose in vivo compared to brain slices (43) suggests that GABA metabolism depends on the presence of physiological cortical activity. This is consistent with the observation in brain slices that an increase in metabolic activity through KC1 depolariza- tion leads to increased labeling of GABA from glucose (55). Metabolic trafficking between GABAergic neurons and adjacent glia is also believed to play a major role in GABA metabolism, both though uptake and oxidation of GABA released from GABAergic neurons and the release of glutamine and/or other intermediates for re- uptake into the neuron (7-10). Studies using brain cortex slices (43) and cultures of neurons and glia (57,58) have shown that glutamine or compounds (e.g., [2- ~3C]acetate) which specifically label glutamine in astro- cytes, are good precursors for GABA synthesis suggesting an important role for astrocytes in GABA metabolism.

Immunohistochemical studies have shown that GABA is concentrated in those neurons containing GAD and that putatively GABAergic neurons contain high GABA and low glutamate (59,60). In our study we could not resolve the fast turning-over, small glutamate pre- cursor pool between GABAergic nerve terminals and the small pool believed to exist in glia (6,7,9,59). However, the sum of the small glutamate pools in both types of cells must be low enough to be labeled quickly without having a major influence on the total rate of glutamate turn over observed in vivo.

The control of GABA synthesis and the regulation of the extracellular GABA concentration is believed to be determined by the activity of GAD (13). Thus, in order to explain the limited elevation of GABA follow- ing GABA-T inhibition, it is reasonable to assume that GAD activity is also reduced as GABA rises. Studies using vigabatrin, as well as, other GABA-T inhibitors, have indicated that the reduction in GAD activity lags the inhibition of GABA-T and that it is not due to a direct effect of the inhibitor on GAD (16,61). However, the at most 25-30% reduction of total GAD activity measured in vitro could not suffice to compensate for the 60-80% inhibition of GABA-T (16,19,61). Our re- sults show that the rate of GABA synthesis is reduced to - 3 0 % of the non-treated control value in vivo within 24 hours following GABA-T inhibition. Our result that the reduction of GABA synthesis is similar to the inhi- bition of GABA-T is consistent with a previous study which reported that for a constant level of GABA-T in-

hibition, the additional elevation of GABA concentration after 24 hours was minimal (16). The disagreement be- tween the large in vivo inhibition of GAD compared with the small decline in total extracted GAD activity may be reconciled by the recent discovery of two major isoforms of GAD in brain of 65 and 67 kD (21,22), of which one of them (GAD67) is sensitive to GABA con- centration (26,27). Rimvall and Martin have shown that for a dose of vigabatrin which produces similar levels of GABA elevation or GABA-T inhibition as in the present study, GAD67 protein was reduced by more than 50% at 24 hours, whereas GAD65 activity was unchan- ged (26). Taken together with the in vivo findings, these results suggest that in the ~x-chloralose anesthetized rat cortex (with potentially reduced function-related synap- tic activity), GAD67 may account for the major fraction of GABA synthesis in vivo.

ACKNOWLEDGMENTS

The authors were supported, in part, by grants from the National Institutes of Health HD32573, AA10121, DK27121, N532126. GFM was supported by the National Center for Research Resources NIH Grant RR-07723. The authors wish to thank Dr. Robert Felberg for measurements of GABA-transaminase activity and Dr. Richard H. Mattson, Dr. Jolm Krystal, and Dr. Robert Shulman for their support and encouragement. We also wish to thank Terry Nixon, chief engi- neer, for spectrometer development and maintenance at optimum per- formance and Peter Brown for construction of the surface coil probe.

REFERENCES

1. Roberts, E. 1986. Failure of GABAergic inhibition a key to local and global seizures. Adv. Neurol. 44:319-341.

2. Krnjevic, K. 1987. GABAergic inhibition in the neocortex. J Mind Behav. 8:537-547.

3. Roberts, E. 1988. The establishment of GABA as a neurotrans- mitter. Pages 151 , in Squires, R. (ed.), GABA and Benzodiaze- pine Receptors, CRC Press.

4. McCormick, D. A. 1989. GABA as an inhibitory neurotransmitter in human cerebral cortex. J Neurophysiol. 62:1018-1027.

5. Meldrum, B. S. 1989. GABAergic mechanisms in the pathogene- sis and treatment of epilepsy. Br. J. Clin. Pharmacol. 27 (Suppl 1):3S-11S.

6. Balazs, R., Machiyama, Y., Hammond, B. J., Julian, T., and Rich- ter, D. 1970. The operation of the ~/-aminobutyrate bypath of the tricarboxylic acid cycle in brain tissue in vitro. Biochem. J. 116: 445M67.

7. Reubi, J. C. Van den Berg, C. J., and Cuenod, M. 1978. Glutamine as a precursor for the GABA and glutamate transmitter pools. Neurosci Lert. 10:171-174.

8. Shank, R. P., and Campbell, G. L. 1984. c~-ketoglutarate and mal- ate uptake and metabolism by synaptosomes: further evidence for an astrocyte-to-neuron metabolic shuttle. J. Neurochem. 42:1153- 1161.

9. Paulsen, R. E., Odden, E., and Fonnum, F. 1988. Importance of glutamine for ~-aminobutyric acid synthesis in rat neostriatum in vivo. J. Neurochem. 51:1294-1299.

1040 Manor, Rothman, Mason, Hyder, Petroff, and Behar

10. Shank, R. P., Leo, G. C., and Zielke, H. R. 1993. Cerebral met- abolic compartmentation as revealed by nuclear magnetic reso- nance analysis of D-[1J3C]glucose metabolism. J. Neurochem. 61: 315-323.

11. Roberts, E. 1974. Gamma amino butyric acid and nervous system function--a perspective. Biochem. Pharmacol. 23:2637-49.

12. Bemasconi, R., Maitre, L., Martin, P., and Raschdorf, F. 1982. The use of inhibitors of GABA-transaminase for the determination of GABA turnover in mouse brain regions: an evaluation of ami- nooxyacetic acid and gabaculine. J. Neurochem. 38:57~56.

13. Martin, D. L., and Rimvall, K. 1993. Regulation of ~/-aminobu- tyric acid synthesis in the brain. J. Neurochem. 60:395-407.

14. Martin, D. L. 1987. Regulatory properties of brain glutamate de- carboxylase. Cell Mol. Neurobiol. 7:237-253.

15. Behar, K. L., and Boehm, D. 1994. Measurement of GABA fol- lowing GABA-transaminase inhibition by gabaculine: a ~H and 31p NMR spectroscopic study of rat brain in vivo. Magn. Reson. Med. 31:660-667.

16. Jung, M. J., Lippert, B., Metcalf, B. W., Bohlen, P., and Schech- ter, P. J. 1977. y-Vinyl GABA (4-amino-hex-5-enoic acid), a new selective irreversible inhibitor of GABA-T: effects on brain GABA metabolism in mice. J. Neurochem. 29:797-802.

17. Kobayashi, K., Miyazawa, S., and Endo, A. 1977. Isolation and inhibitory activity of gabaculine, a new potent inhibitor of gamma- aminobutyrate aminotransferase produced by a Streptomyces. FEBS Lett. 76:207-210.

18. Matsui, Y., and Deguchi, T. 1977. Effects of gabaculine, a new potent inhibitor of gamma-arninobutyrate transaminase, on the brain ganmaa-aminobutyrate content and convulsions in mice. Life Sci. 20:291-295.

19. Neal, M. J., and Shah, M. A. 1990. Development of tolerance to the effects of vigabatrin (~/-vinyl-GABA) on GABA release from rat cerebral cortex, spinal cord and retina. Br. J. Pharmacol. 100: 324-328.

20. Martin, D. L. 1993. Short-term control of GABA synthesis in brain. Prog. Biophys. Mol. Biol. 60:17-28.

21. Erlander, M. G., Tillakaratne, N. J. K., Feldblum, S., Patel, N., and Tobin, A. J. 1991. Two genes encode distinct glutamate de- carboxylases. Neuron. 7:91-100.

22. Feldblum, S., Erlander, M. G., and Tobin, A. J. 1993. Different distributions of GAD65 and GAD67 mRNAs suggest that the two glutamate decarboxylases play distinctive functional roles. J. Neu- rosci. Res. 34:68%706.

23. Esclapez, M., Tillakaratne, N. J. K., Kanfrnan, D., Tobin, A. J., and Houser, C. 1994. Comparative localization of two forms of glutamic acid decarboxylase and their mRNAs in rat brain sup- ports the concept of functional differences between the forms. J. Neurosci. 14:1834-1855.

24. Hendrickson, A., Tillakaratne, N. J. K., Mehra, R., Esclapez, M., Erickson, A., Vician, L., and Tobin, A. J. 1994. Differential lo- calization of two glutamic acid decarboxylases (GAD65 and GAD67 ) in adult monkey visual cortex. J. Comp. Neurol. 343:56(~ 581.

25. Li, L., Jiang, J., Hagopian, W., Karlsen, A., Skelly, M., Baskin, D., and Lernmark, A. 1995. Differential detection of rat islet and brain glutamic acid decarboxylase (GAD) isoforms with sequence- specific peptide antibodies. J. Histochem. Cytochem. 43:53-59.

26. Rimvall, K., and Martin, D. L. 1994. The level of GAD67 protein is highly sensitive to small increases in intraneuronal ~/-amino- butyric acid levels. J. Neurochem. 62:1375-1381.

27. Rimvall, K., Sheikh, S. N., and Martin, D. L. 1993. Effects of increased -/-aminobutyric acid levels on GAD67 protein and mRNA levels in rat cerebral cortex. J. Neurochem. 60:714-720.

28. Martin, S. B., and Martin, D. L. 1979. Stimulation by phosphate of the activation of glutamate apodecarboxylase by pyridoxal-5'- phosphate and its implications for the control of GABA synthesis. J. Neurochem. 33:1275-1283.

29. Rothman, D. L., Petroff, 0 . A. C., Behar, K. L., and Mattson, R. H. 1993. Localized ~H NMR measurements of ~/-aminobutyric acid in human brain in vivo. Proc. Natl. Acad. Sci. 90:5662-5666.

30. Petroff, O. A. C., Rothman, D. L., Behar, K. L., and Mattson, R. H. 1995. Initial observations on effect of vigabatrin on in vivo 1H spectroscopic measurements of ~/-aminobutyric acid, glutamate, and glutamine in human brain. Epilepsia. 36:457-464.

3 l. Petroff, O. A. C., Rothman, D. L., Behar, K. L., and Mattson, R. H. 1995. Human brain GABA levels rise following initiation of vigabatrin therapy, but fail to rise further with increasing dose. Neurol., in press.

32. Fitzpatrick, S. M., Hetherington, H. P., Behar, K. L., and Shulman, R. G. 1990. The flux from glucose to glutamate in the rat brain in vivo as determined by 1H-observed, ~3C-edited NMR spectros- copy. J. Cereb. Blood Flow Metab. 10:170-179.

33. Gruetter, R. 1993. Automatic, localized in vivo adjustment of all first and second-order shim coils. Magn. Reson. Med. 29:804-811.

34. Frahn, J., Merboldt, K., and Hfiinicke, W. 1987. Localized proton spectroscopy using stimulated echoes. J. Magn. Reson. 72:502- 508.

35. Ordridge, R. J., Connelly, A., and Lob_man, J. A. B. 1986. Im- age selected in vivo spectroscopy (ISIS): a new technique for spatially selective NMR spectroscopy. J. Magn. Reson. 66:283- 294.

36. Rothman, D. L., Behar, K. L., Hetherington, H. P., den Hollander, J. A., Bendall, M. R., Petroff, O. A. C., and Shulman, R. G. 1985. ~H-Observe/~3C-decoupled spectroscopic measurements of lactate and glutamate in the rat brain in vivo. Proc. Natl. Acad. Sci. 82: 1633 1637.

37. Behar, K. L., and Ogino, T. 1993. Characterization of macromol- ecule resonances in the IH NMR spectrum of rat brain. Magu. Reson. Med. 30:38-44.

38. Clark, J. B., and Lai, J. C. K. 1989. Glycolytic, tricarboxylic acid cycle and related enzymes in brain. Pages 233-281, in Boulton, A., Baker, G., and Butterworth, R. (eds.), Neuromethods, Humaua Press, Clifton, New Jersey.

39. Walsh, J., and Clark, J. B. 1976. Studies on the control of 4- aminobutyrate metabolism in 'synaptosomal' and free rat brain mitochondria. Biochem. J. 160: I47-57.

40. Mason, G. F., Rothman, D. L., Behar, K. L., and Shulman, R. G. 1992. NMR determination of the TCA cycle rate and c~-ketoglu- tarate/glutamate exchange rate in rat brain. J. Cereb. Blood Flow Metab. 12:434-447.

41. Mason, G. F., Gruetter, R., Rothman, D. L., Behar, K. L., Shul- man, R. G., and Novotny, E. J. 1995. Simultaneous determination of the rates of the TCA cycle, glucose utilization, et-ketoglutar- ate/glutamate exchange, and glutamine synthesis in human brain by NMR. J. Cereb. Blood Flow Metab. 15:12-25.

42. Paulsen, R. E., and Fonnum, F. 1988. Regulation of transmitter y-aminobutyric acid (GABA) synthesis and metabolism illustrated by the effect of ~-vinyl GABA and hypoglycemia. J. Neurochem. 50:1151-1157.

43. Badar-Goffer, R. S., Bachelard, H. S., and Morris, P. G. 1990. Cerebral metabolism of acetate and glucose studied by J3C-NMR spectroscopy: a technique for investigating metabolic compart- mentation in the brain. J. Neurochem. 266:133-139.

44. Patel, A. J., Johnson, A. L., and Balazs, R. 1974. Metabolic com- partmentation of glutamate associated with the formation of ~/- aminobutyrate. J. Neurochem. 23:1271-1279.

45. Loscher, W. 1980. Effects of inhibitors of GABA transaminase on the synthesis, binding, uptake, and metabolism of GABA. J. Neurochem. 34:1603-1608.

46. Schechter, P. J., Tranier, Y., and Grove, J. 1979. Gabaculine and isogabaculine: in vivo biochemistry and pharmacology in mice. Life Sci. 24:1173-1182.

47. Rando, R. R., and Bangerter, F. W. 1977. The in vivo inhibition of GABA transaminase by gabaculine. Biochem. Biophys. Res. Commun. 76:1276-1281.

Cortical GABA Turnover Is Reduced Following Vigabatrin 1041

48. Patel, A. J., Balazs, R., Richter, D. 1970. Contribution of the GABA bypath to glucose oxidation, and the development of com- partmentation in the brain. Nature 226:116(~1161.

49. Miller, L. P., Walters, J. R., and Martin, D. L. 1977. Postmortem changes implicate adenine nucleotides and pyridoxal-5'-phosphate in regulation of brain glutamate decarboxylase. Nature 266:847-848.

50. Carlsson, C., Hagerdal, M., and Siesjo, B. K. 1976. The effect of nitrous oxide on oxygen consumption and blood flow in the cer- ebral cortex of the rat. Acta Anaesth. Scand. 20:91-95.

51. Lopes-Cardozo, M., and Albers, R. W. 1979. Relationship be- tween the 4-aminobutyrate bypath and the oxidation of 2-oxoglu- tarate in rat brain mitochondria. J. Neurochem. 33:1259-1265.

52. Van den Berg, C. J., and Garfinkel, D. 1971. A simulation study of brain compartments: metabolism of glutamate and related sub- stances in mouse brain. Biochem. J. 123:211-218.

53. Van den Berg, C. J., Krzalic, L. J., Mela P., and Waelsch, H. 1970. Compartmentation of glutamate metabolism in brain: evi- dence for the existence of two different tricarboxylic acid cycles in brain. Biochem. J. 113:281-290.

54. Garfinkel, D. 1966. A simulation study of the metabolism and compartmentation in brain of glutamate, aspartate, the Krebs cy- cle, and related metabolites. J. Biol. Chem. 241:3918-3929.

55. Badar-Goffer, R. S., Ben-Yoseph, O., Bachelard, H. S., and Mor-

ris, P. G. 1992. Neuronal-Glial metabolism under depolarizing conditions: a ~3C-NMR study. Biochem. J. 282:225-230.

56. Bachelard, H. S., and Badar-Goffer, R. S. 1993. NMR spectros- copy in neurochemistry. J. Neurochem. 61:412-429.

57. Westergaard, N., Petersen, U., and Schousboe, A. 1995. Glutamate and glutamine metabolism in cultured GABAergic neurons studied by ~3C NMR spectroscopy may indicate compartmentation and mitochondrial heterogeneity. Neurosci. Lett. 185:24-28.

58. Sonnewald, U., Westergaard, N., Schousboe, A., Svendsen, J. S., Unsgard, G., and Petersen, S. B. 1993. Direct demonstration by 13C-NMR that glutamine from astrocytes is a precursor for GABA synthesis in neurons. Neurochem Int. 22:19-29.

59. Ottersen, O. P., Zhang, N., and Walberg, F. 1992. Metabolic eom- partmentation of glutamate and glutamine: morphological evi- dence obtained by quantitative immunocytochemistry in rat cerebellum. Neurosci. 46:519-534.

60. Storm-Mathisen, J., Leknes, A. K., Bore, A. T., Vaaland, J. L., Edminson, P., Haug, F. M. S., and Ottersen, O. P. 1983. First visualization of glutamate and GABA in neurones by immuno- cytochemistry. Nature 301:517-520.

61. Perry, T. L., Kish, S. J., and Hansen, S. 1979. Gamma-Vinyl GABA: effects of chronic administration on the metabolism of GABA and other amino compounds in rat brain. J Neurochem. 32:1641-1645.