Cerebral Glutamine Metabolism under Hyperammonemia Determined in vivo by Localized 1 H and 15 N NMR Spectroscopy

Introduction

In brain, glutamine synthetase (GS) activity is predominantly localized in the astrocytes (Martinez-Hernandez et al, 1977; Norenberg and Martinez-Hernandez, 1979). It is responsible for the synthesis of glutamine (Gln) from glutamate (Glu) cleared from the synaptic cleft, thus having a major role in neurotransmission. This glutamine is then shuttled to neurons where it can be deaminated, completing the so-called glutamate—glutamine cycle (Berl et al, 1968; Clarke et al, 1970; Gruetter, 2004; McKenna, 2007; Shank and Aprison, 1981; Zwingmann and Leibfritz, 2003). In addition, in conditions of hyperammonemia, GS is used for detoxification of brain ammonia (Cooper and Plum, 1987; Kanamori et al, 1993; Kanamori and Ross, 1993). In this process, glutamine is accumulated in the brain, while glutamate concentration remains constant. Thus, glutamine accumulation requires de-novo synthesis from glucose, with involvement of both glutamate dehydrogenase (GDH) and aspartate aminotransferase (Berl et al, 1962; Shen et al, 1998).

Because brain GS is an integral part of the glutamate—glutamine cycle (Berl et al, 1968; Clarke et al, 1970; Shank and Aprison, 1981) and is also functioning as a detoxification pathway for cerebral ammonia (Cooper and Plum, 1987; Kanamori and Ross, 1993; Kanamori et al, 1993), developments of new methods for the noninvasive measurement of the GS activity became important. Several in vitro studies were performed to measure the GS activity in rat brain homogenates (Butterworth et al, 1988) or in cultured rat and mouse brain astrocytes (Yudkoff et al, 1986). However, knowledge of in vivo GS activity is necessary because glial—neuronal interaction has a crucial role in the glutamate—glutamine cycle (Erecinska and Silver, 1990). ¹³N-labeled ammonia has been used to study the cerebral uptake and metabolism of ammonia in conscious rat (Cooper et al, 1979). However, the rate of GS was not determined, because the isotope enrichment of the precursor ammonia and the physical size of the compartment were not known (Cooper et al, 1979; Kanamori and Ross, 1993). A potential drawback of ¹³N is its short half-life (10 minutes) which limits its usefulness to experiments of < 1 hour (Cooper et al, 1979). In vivo Magnetic Resonance Spectroscopy (MRS) is a useful technique for the continuous and noninvasive measurement of the concentrations and synthesis rates of individual metabolites within precisely defined areas in the brain. In particular, in vivo noninvasive ¹³C MRS has been used to study metabolic fluxes, enzymes activities in the living brain of animals and humans (Duarte et al, 2011; Gruetter et al, 2001). ¹⁵N MRS is an alternative approach to ¹³C MRS in studying glutamate—glutamine metabolism and can in principle provide a more direct interpretation than ¹³C studies. The wide chemical shift dispersion of ¹⁵N provides resolution of important metabolites in glutamine—glutamate cycle. Sparse in vivo ¹⁵N studies and utilization of different protocols for infusion of ¹⁵N-enriched substrates lead to a diversity of results and limited their integration on mathematical models of compartmentalized brain intermediary metabolism.

From these previous studies (Kanamori et al, 1993, 1996; Kanamori and Ross, 1993; Shen et al, 1998), discrepancies have been noticed concerning the GS flux (V_syn) and the net glutamine accumulation (also called de-novo glutamine synthesis rate; Shen et al, 1998). These divergences could be intrinsically bound to the employment of different experimental protocols and assumptions used to calculate the metabolic fluxes. Therefore, there is a need to acquire additional experimental data during the same experiment, to minimize the number of assumptions to be done on mathematical modeling.

A few past in vivo MRS studies under ammonia infusion used either in vivo ¹H (Fitzpatrick et al, 1989) or unlocalized ¹⁵N spectroscopy (Kanamori et al, 1993, 1996; Kanamori and Ross, 1993; Shen et al, 1998) but never combined and used in vitro brain extracts (biochemical assays or high-resolution nuclear magnetic resonance (NMR)) for absolute quantification. These studies mainly focused on the incorporation of ¹⁵N of ammonia into [5-¹⁵N]Gln and measured the apparent GS flux (Kanamori et al, 1993, 1996; Kanamori and Ross, 1993; Shen et al, 1998).

The incorporation of ¹⁵N into [2-¹⁵N]Gln + Glu may provide further insight into the net flux through GDH (V_GDH) (Kanamori and Ross, 1995). However, due to the large chemical shift dispersion between [5-¹⁵N]Gln and [2-¹⁵N]Gln + Glu, V_syn and V_GDH were never simultaneously determined. Kanamori and Ross (1995) measured [2-¹⁵N]Gln + Glu (but not [5-¹⁵N]Gln) during ¹⁵N-enriched ammonia infusion and estimated V_GDH.

In this context, we used a new acquisition strategy combined with an appropriate mathematical model of compartmentalized cerebral glutamine metabolism and determined, in the same experiment, an important number of metabolic fluxes with minimal assumptions. For that purpose, we developed a new ¹⁵N pulse sequence to simultaneously detect, for the first time [5-¹⁵N]Gln and [2-¹⁵N]Gln + Glu in vivo. Together with in vivo localized ¹⁵N spectra, we additionally measured for the first time in the same experiment localized ¹H spectra for a direct measurement of the net glutamine accumulation. From our knowledge, the combination of localized ¹⁵N MRS and ¹H MRS in interleaved mode has not been reported previously under ¹⁵N ammonia infusion. Consequently, ¹H MRS allowed to follow up the total Gln accumulation during ammonia infusion and thereby to directly measure the net glutamine accumulation without any assumptions. Mathematical modeling of ¹H and ¹⁵N MRS data provided reliable estimations of V_syn, the apparent neurotransmission rate V_nt, V_GDH, and the net glutamine accumulation (V_syn– V_nt) under ¹⁵N-labeled ammonia infusion in the rat brain. Moreover, we performed a direct in vivo absolute quantification of total Gln, [5-¹⁵N]Gln and [2-¹⁵N]Glu + Gln in the same experiment.

Materials and methods

Magnetic Resonance Spectroscopy Measurements

All ¹H and ¹⁵N MRS data were acquired interleaved on a Varian Direct Drive (Palo Alto, CA, USA) console interfaced to an actively shielded 9.4 T magnet (Magnex Scientific, Oxford, UK) with a 31-cm horizontal bore. The magnet was equipped with 12 cm inner-diameter actively shielded gradient sets giving a maximum gradient of 400 mT/m in 120 μs. A home-built 14 mm-diameter ¹H quadrature surface coil combined with a single 5-loop 10 mm ¹⁵N coil was used as transceiver. The static field homogeneity was adjusted using first- and second-order shims using an echo-planar version of FASTMAP (Gruetter and Tkac, 2000). Localizer images were obtained in the coronal planes using a multislice fast spin echo protocol with echo time (TE)/repetition time (TR) = 60/5,000 ms, a slice thickness of 1 mm and an in-plane resolution of 94 μm.

¹H spectra were obtained using the ultra-short TE (TE/TR= 2.8/4,000 ms, 4,096 complex data points, SW (spectral width) = 7 kHz, 160 averages) SPin ECho, full Intensity Acquired Localized technique (SPECIAL) (Mlynarik et al, 2006, 2008). The voxel of interest (VOI) for ¹H MRS was placed in the center of the brain (5 × 7 × 7 mm³). The reproducibility of the voxel placement was based on anatomical landmarks using the localizer images. After first- and second-order shimming, the typical linewidth of water resonance at TE = 2.8 ms was 16 Hz. The water signal was suppressed by a series of seven 25 ms asymmetric variable power radio frequency (RF) pulses with optimized relaxation delays, i.e., VAPOR (Tkac et al, 1999). The water suppression pulses were interleaved with three modules of outer volume saturation (OVS) as described previously (Tkac et al, 1999). To compensate for the magnetic field drift, spectra were collected in blocks of 16 averages, which were stored separately in the memory and corrected for the relative shift in frequency and potential variations in phase before summation.

¹⁵N spectra were acquired using a new pulse sequence based on the previously published Single-shot Inversion Recovery based Non-Echo (SIRENE) sequence (Choi et al, 2000). Briefly, in the conventional SIRENE sequence adiabatic pulses (HS8 modulation) are applied to invert the Z magnetization in two x-slices adjacent to the VOI. Then, the magnetization in these two slices follows an inversion recovery trajectory and is minimized by adjusting the inversion time. Residual magnetization outside the VOI was further suppressed by using six OVS pulses applied with a nominal 90° flip angle. The thickness of the upper and lower saturated y-slices was different to adapt the geometry of the rat head. An adiabatic half passage (AHP) pulse was then applied for excitation. In addition, bilevel WALTZ-16 was applied on the ¹H channel for Nuclear Overhauser Effect (NOE) and decoupling during acquisition. Due to the large chemical shift difference between [5-¹⁵N]Gln and [2-¹⁵N]Glu + Gln (~70 p.p.m.) in ¹⁵N spectra, when using the conventional sequence with the carrier frequency placed on the [5-¹⁵N]Gln, the second signal of interest ([2-¹⁵N]Glu + Gln) was only partially excited by the AHP RF pulse. Consequently, conventional SIRENE sequence was modified in our study for ¹⁵N nuclei by adding a second AHP RF pulse. The two ¹⁵N signals of interest were acquired separately in an interleaved mode using two adiabatic excitation pulses with opposite frequency modulations. Figure 1A shows the evolution of the Z magnetization after the first AHP RF pulse with the carrier frequency placed on the [5-¹⁵N]Gln. As can be seen, after the AHP RF pulse, [5-¹⁵N]Gln was fully excited (no longitudinal magnetization left), but the reduction of the Z magnetization of [2-¹⁵N]Glu + Gln was <10%. After 3 seconds, the second AHP RF pulse was applied with the carrier frequency on [2-¹⁵N]Glu + Gln (Figure 1B). Similarly, the second AHP RF pulse flipped the magnetization of [2-¹⁵N]Glu + Gln into the transverse plane, whereas the Z magnetization of [5-¹⁵N]Gln was only slightly reduced by this pulse. This interleaved excitation scheme allows to selectively excite the two resonances of interest in the same effective TR. In addition, in our new sequence the adiabatic pulses (HS8 modulation) used to invert the Z magnetization in two x-slices adjacent to the VOI were eliminated; and consequently, the localization was performed by using only the OVS modules. Radio frequency power for excitation, NOE, and decoupling was carefully calibrated using a small sphere (3 mm diameter) containing 99% ¹⁵N-enriched glycine (Gly, Cambridge Isotopes) placed at the center of the ¹⁵N linear coil as an external reference as described previously (Gruetter et al, 1994, 2003). These calibrations were used to establish minimal power necessary for decoupling. Unlocalized ¹⁵N spectra were acquired in the first hours of infusion (256 scans, TR = 6 seconds), followed by a localized spectrum (VOI = 5 × 7 × 7 mm³, 256 scans) used for quantification. The robustness of localization was verified from the absence of the Gly signal in the in vivo spectra.

OPEN IN VIEWER

Figure 1

Description of the Z magnetization after the two AHP RF pulses with opposite frequency modulations used interleaved in the new sequence. (A) The evolution of the Z magnetization after the first AHP RF pulse with the carrier frequency placed on the [5-¹⁵N]Gln. As can be seen, after the AHP RF pulse, [5-¹⁵N]Gln was fully excited (no longitudinal magnetization left), but the reduction of the Z magnetization of [2-¹⁵N]Glu + Gln was < 10%; (B) After 3 seconds, the second AHP RF pulse was applied with the carrier frequency on [2-¹⁵N]Glu + Gln and flipped the magnetization of [2-¹⁵N]Glu + Gln into the transverse plane, whereas the Z magnetization of [5-¹⁵N]Gln was only slightly reduced by this pulse. [5-¹⁵N]Gln and [2-¹⁵N]Gln + Glu correspond to 5N Gln and 2N Glx in the Figure, respectively. AHP, adiabatic half passage; RF, radio frequency.

¹⁵N-enriched compounds are characterized by a relative long T₁ relaxation time (longitudinal relaxation time). To optimize the TR in the ¹⁵N experiments, the in vitro T₁ relaxation time of [2-¹⁵N]Glu was measured using an adiabatic inversion recovery experiment. The inversion time was varied in the range from 0.1 to 20 seconds. To minimize effects of partial saturation, spectra acquired with different inversion time were collected with the same relaxation delay. The T₁ relaxation curves were fitted with nonweighted two-parameter single exponential functions based on the Levenberg-Marquardt algorithm, fitting M(0) and T₁. For each fit, the correlation coefficients reflecting the quality of the least squares fit versus the original data were computed. A typical standard error of the fitted T₁ was ~5%. The in vitro T₁ relaxation time of [2-¹⁵N]Glu was 4.0 ± 0.3 seconds.

Metabolic Modeling

The neuroglial two-compartment metabolic model used in the present study is shown in Figure 2. Ammonia (NH₃+NH₄⁺) passes the blood—brain barrier and enters glial cells at a rate V_in. In astrocytes, ammonia and glutamate are converted to glutamine through GS, thereby labeling the Gln at the position 5N (Figure 2A). Gln is released by the astrocytes and transported into the neurons where it is converted to glutamate by glutaminase, which completes the glutamate—glutamine cycle (Figure 2A). For the labeling of [2-¹⁵N]Glu + Gln (Figure 2B), ammonia is incorporated in glial glutamate through GDH, then glutamate is converted to glutamine by GS. Glutamine is transported into neurons and converted to glutamate by neuronal glutaminase. Since most ammonia is detoxified by the GS reaction (Du Ruisseau et al, 1957), which is restricted to astrocytes (Martinez-Hernandez et al, 1977; Norenberg and Martinez-Hernandez, 1979), and little by GDH (Cooper and Plum, 1987), in the present study we assumed that the astrocyte rather than the neuron is uniquely responsible for ammonia detoxification in the brain.

OPEN IN VIEWER

Figure 2

Neuroglial two-compartment metabolic model: (A) for [5-¹⁵N]Gln: ammonia (NH3 + NH4 +) passes the blood—brain barrier (BBB) and enters glial cells at a rate V_in. In astrocytes, ammonia and glutamate are converted to glutamine through glutamine synthetase, thereby labeling the Gln at the position 5N. Gln is released by the astrocytes and transported into the neurons where it is converted to glutamate by glutaminase, which completes the glutamate-glutamine cycle; (B) for [2-¹⁵N]Gln + Glu: ammonia is incorporated in glutamate through glutamate dehydrogenase, then glutamate is converted to glutamine by glutamine synthetase. Glutamine is transported into neurons and converted to glutamate by neuronal glutaminase. Glial glutaminase was neglected.

In neurons, the exchange of ¹⁵N from glutamate in interaction with the tricarboxylic acid cycle happens through transamination reactions, while GDH is essentially taking place in the astrocytes (Aoki et al, 1987; Zaganas et al, 2001). We modeled the exchange of ¹⁵N from the position 2 of neuronal glutamate to the position 2 of aspartate through transamination, assuming a neuronal aspartate pool of 2 μmol/g, as measured by ¹H NMR spectroscopy and in agreement with previous reports (Kunz et al, 2010; Xin et al, 2010). This interaction has the role of a transient dilution of the neuronal [2-¹⁵N]Glu pool. We further varied the exchange flux V_x between glutamate and the neuronal tricarboxylic acid cycle through transamination from 0 to 1 μmol/g per minute, which represents a typical range for this chemical exchange in the anesthetized rat brain (Duarte et al, 2011). Varying V_x in this range did not affect the labeling curves of [2-¹⁵N]Glx significantly and had very low impact on the other determined metabolic fluxes. The model was therefore simplified and the exchange with the neuronal tricarboxylic acid cycle was neglected. We failed to observe any other amino acids to become enriched. A substantial enrichment of aspartate that occurs at ~2 mmol/L (almost exclusively in neurons) would be observable in the ¹⁵N spectra.

The time courses of total Gln measured by ¹H MRS, [5-¹⁵N]Gln, and [2-¹⁵N]Glu + Gln were used as experimental data to which the model was fitted as follows. The differential equations derived for [Gln], [5-¹⁵N]Gln, and [2-¹⁵N]Glu + Gln (see Appendix) were fitted simultaneously using SAAM II environment (The SAAM Institute, Seattle, WA, USA). Since no direct measurement of the dynamic labeling of brain ammonia was possible, V_in was assumed 0.23 μmol/g per minute (Cooper et al, 1979; Cooper and Plum, 1987; Shen et al, 1998). In addition, we assumed a negligible efflux from the Gln pool compared with V_syn (Kanamori and Ross, 1993; Kanamori et al, 1996; Shen et al, 1998). For the fit of [2-¹⁵N]Glu + Gln, we assumed that V_GDH was mainly in the glial compartment. This assumption was verified by measuring the FE ratio (([2-¹⁵N]Gln)FE/([2-¹⁵NGlu])FE) in high-resolution ¹⁵N brain extract spectra at a pH of ~9. The time evolution of the total Gln, [5-¹⁵N]Gln, and [2-¹⁵N]Glu + Gln was fitted simultaneously. The linear fit of the time evolution of the total Gln provided a measure of the net glutamine accumulation (V_syn–V_nt) as well as the initial Gln concentration (Gln(0)). In addition, from the fit of the time evolution of [5-¹⁵N]Gln we obtained V_syn and from these two fluxes (V_syn–V_nt and V_syn) the neurotransmission rate (V_nt) was calculated. Finally, using our model the fit of [2-¹⁵N]Glu + Gln enabled the measurement of V_GDH (see Appendix, equations (18), (19) and (23)). Results were generally presented as mean ± s.d. (n = 6).

In addition, using the values obtained for the different fluxes, the FEs of [2-¹⁵N]Glu and [2-¹⁵N]Gln were simulated for the end of the experiment and their ratio was compared with the high-resolution NMR measurements obtained on brain extracts.

Results

During the in vivo experiments, the physiological variables were maintained within normal physiology (pH= 7.41 ± 0.04; PaCo₂ = 40.0 ± 1.8 mmHg, O₂ saturation >99.8%). The plasma ammonia concentration increased from 0.08 ± 0.02 to 0.95 ± 0.08 mmol/L and remained approximately constant during the experiment. Plasma urea concentration was 7.6 ± 0.8 mmol/L before the ammonium chloride infusion, increased linearly on infusion of ammonium cloride and reached a final concentration of 29.9 ± 10.8 mmol/L. The brain ammonia concentration measured at the end of each experiment in brain extracts was 3.8 ± 0.8 μmol/g in agreement with previous measurements (Kanamori et al, 1993, 1996; Kanamori and Ross, 1993). Because ammonia diffuses across the blood—brain barrier mainly as neutral NH₃ and brain pH (7.1) is lower than the blood pH (7.4), brain ammonia level at equilibrium is normally 1.5 to 3 times that of the blood, which is typically maintained under hyperamonemia (Cooper and Plum, 1987). At the end of the infusion period, we determined a brain ammonia/blood ammonia ratio of 3.8 ± 0.8 (mean ± s.d.) that was in the range of previous reports (Cooper and Plum, 1987).

In general, ¹H spectra exhibited excellent signal-to-noise ratio allowing easy separation of Gln from Glu. The increase in the total Gln pool at different time points during infusion was thus apparent in the ¹H spectra (Figure 3). The total Gln (0) concentration (before starting the infusion) was 2.3 ± 0.4 μmol/g, which increased to 17.7 ± 4.0 μmol/g at the end of the infusion. As can be seen in Figures 3B and 3C, the increase of total Gln started immediately after the infusion and continued to increase linearly over time. The total concentrations of all other metabolites remained within < 10% fluctuation in standard deviation of the mean concentration over the ammonium chloride infusion (i.e., the N-acetylaspartate concentration before the ammonium chloride infusion was 7.1 ± 0.5 μmol/g and at the end of infusion was 7.6 ± 0.5 μmol/g). Although lactate concentration was observed to correlate with the total Gln increase (R² = 0.90), it never exceeded the physiological range of 2.5 to 3 μmol/g.

OPEN IN VIEWER

Figure 3

(A) Representative in vivo ¹H spectra acquired at 9.4T in the rat brain during ammonium chloride infusion (TE/TR = 2.8/4,000 ms, 160 averages, voxel size of 5 × 7 × 7 mm³); (B) Detailed time course of total Gln acquired in vivo in the first 2 hours of ¹⁵N ammonium chloride infusion; (C) The time course of total Gln acquired in vivo in one rat more than −12 hours of ¹⁵N ammonium chloride infusion. Phosphocholine (PCho), glycerophosphocholine (GPC), N-acetylaspartate (NAA), lactate (Lac), creatine (Cr), phosphocreatine (PCr), taurine (Tau), myo-inositol (Ins), glutamate (Glu), glutamine (Gln), aspartate (Asp), γ-aminobutyrate (GABA). TE, echo time; TR, repitition time.

To compare the new acquisition sequence with the conventional SIRENE sequence, Figure 4A shows a series of in vivo unlocalized ¹⁵N spectra acquired in the rat brain during ammonium chloride infusion using the new pulse sequence. It can be observed that the [5-¹⁵N]Gln peak (–271 p.p.m.) was detected within 25 minutes, whereas the [2-¹⁵N]Gln + Glu peak (–342 p.p.m.) was detected during the second scan (i.e., after 25 minutes of ammonium chloride infusion). With the conventional pulse sequence (Figure 4B), the [5-¹⁵N]Gln peak was also visible in the first and all subsequent scans; however, the [2-¹⁵N]Gln + Glu peak started to appear only from the fourth scan. Compared with the conventional SIRENE sequence, the new pulse sequence using the two AHP RF pulses allowed the detection of [2-¹⁵N]Gln + Glu peak ~ 50 minutes earlier due to a better excitation and consequently an increased signal-to-noise ratio.

OPEN IN VIEWER

Figure 4

(A) A representative series of in vivo unlocalized ¹⁵N spectra acquired using the new sequence with two AHP RF pulses with opposite frequency modulations at 9.4T in a rat brain at different time points (two spectra were acquired per TR); (B) Representative spectra acquired using the conventional sequence (one spectrum was acquired per TR); (C) Localized in vivo ¹⁵N spectrum acquired in a rat brain (a) during ammonium chloride infusion versus an unlocalized spectrum acquired in the same rat (b) at 25 minutes of interval. The performance of the OVS-based localization is proved by the suppression of the ¹⁵N Gly signal, which originates from the sphere placed in the center of the ¹⁵N coil. The ¹⁵N chemical shifts were referenced to nitromethane. The unmarked resonance in the figures represents the urea signal. AHP, adiabatic half passage; RF, radio frequency; TR, repitition time.

To assess the performance of the OVS localization technique, Figure 4C shows an example of a localized (a) in vivo ¹⁵N spectrum versus an unlocalized one (b) acquired at 25 minutes of interval in the same rat during ammonium chloride infusion. As can be seen from the Figure, the performance of the OVS-based localization is proved by the suppression of the ¹⁵N Gly signal, which originates from the sphere placed in the center of the ¹⁵N coil. In addition, by comparing the two resonances of [5-¹⁵N]Gln we can notice that there are very small differences, suggesting that almost the entire [5-¹⁵N]Gln signal comes from the brain and not from surrounding tissues.

Using the external reference method described previously for ¹³C data (Gruetter et al, 2003), we were able to perform a direct in-vivo absolute quantification for [5-¹⁵N]Gln and [2-¹⁵N]Gln + Glu. Consequently, at the end of infusion absolute concentrations of [5-¹⁵N]Gln and [2-¹⁵N]Gln + Glu were 13.3 ± 2.9 and 9.6 ± 3.5 μmol/g, respectively, in accordance with reported in vitro brain extracts determinations (biochemical assays or high-resolution NMR; Kanamori et al, 1993, 1996; Kanamori and Ross, 1993; Shen et al, 1998).

The time courses of total Gln, [5-¹⁵N]Gln, and [2-¹⁵N]Gln + Glu were highly reproducible in all six rats. The application of the model to the in vivo data showed an excellent fit (Figure 5). Using the model presented (Figure 2), we obtained from the ¹H data a net Gln accumulation (V_syn– V_nt) of 0.033 ± 0.001 μmol/g per minute (Figure 5A). By fitting in vivo [5-¹⁵N]Gln and [2-¹⁵N]Glu + Gln time courses (Figure 5B), V_syn was 0.30 ± 0.050 μmol/g per minute, V_GDH was 0.029 ± 0.002 μmol/g per minute, and the plasma NH₃ FE was 91 ± 5%. Finally, the apparent neurotransmission rate, V_nt, was 0.26 ± 0.030 μmol/g per minute.

OPEN IN VIEWER

Figure 5

The time courses and corresponding fits of total Gln (A), [5-¹⁵N]Gln and [2-¹⁵N]Gln + Glu (B) from one rat (duration of ammonium chloride infusion −7 hours). [2-¹⁵N]Gln + Glu corresponds to [2-¹⁵N]Glx in the Figure.

To determine the predominant cellular localization of GDH, we measured the ratio between the FEs of [2-¹⁵N]Gln and [2-¹⁵N]Glu in brain extracts. From high-resolution measurements on brain extracts, we calculated ([2-¹⁵N]Gln)FE/([2-¹⁵N]Glu)FE = 1.48 ± 0.04 which was in agreement with the simulations done for the FEs of [2-¹⁵N]Glu and [2-¹⁵N]Gln based on the values obtained for the different fluxes.

Discussion

In the present study, we directly and simultaneously measured for the first time V_syn, V_nt, V_GDH, and the net Gln accumulation under ¹⁵N-labeled ammonium chloride infusion in the rat brain. This was possible by using a new acquisition strategy: in vivo localized ¹⁵N MRS interleaved with in vivo ¹H MRS. In this context, from the modeling point of view this new acquisition strategy required fewer assumptions to determine metabolic rates.

By acquiring in vivo ¹H spectra, we followed the total Gln accumulation during ammonium chloride infusion. Unlike to previous studies, we directly measured for the first time the net Gln accumulation without any assumptions. Previous measurements of net Gln accumulation (i.e., de-novo glutamine synthesis; Shen et al, 1998) assumed that the increase of glutamine pool size during hyperammonemia is attributed to increased pyruvate carboxylase and GDH activities. Moreover, these studies assumed that pyruvate carboxylase is equal to GDH under hyperammonemia and that together these fluxes account for the net transport of ammonia.

In addition to the acquisition of in vivo ¹H spectra, we now simultaneously measured, for the first time, in vivo [5-¹⁵N]Gln and [2-¹⁵N]Gln + Glu time courses during infusion of ¹⁵N-enriched ammonium chloride. The new ¹⁵N pulse sequence proposed in the present study, which uses the two AHP RF pulses, allowed the detection of [2-¹⁵N]Gln + Glu peak ~ 50 minutes earlier compared with the conventional SIRENE sequence, due to a better excitation and consequently an increased signal-to-noise ratio. Moreover, by acquiring localized ¹⁵N spectra and by using the external reference method, we performed a direct in vivo absolute quantification for [5-¹⁵N]Gln and [2-¹⁵N]Gln + Glu. Finally, we directly quantified total Gln, [5-¹⁵N]Gln, and [2-¹⁵N]Glu + Gln in the same experiment in vivo without the need of in vitro brain extracts.

The linear and continuous increase of total Gln during the period of ammonium chloride infusion observed in our in vivo ¹H MRS data implies increased anaplerosis (Berl et al, 1962; Shen et al, 1998; Zwingmann, 2007), which appears to be coupled to the ammonia detoxification pathway. As can be seen in Figures 3B and 3C, the increase of total Gln started immediately after the infusion and was linear over the infusion time, in contrast to previous studies (Rothman et al, 1999; Shen et al, 1998), suggesting that no delay in Gln accumulation was observed. The total Gln concentration measured at the end of the infusion is comparable to previous studies based on in vitro brain extract determinations (Kanamori et al, 1993, 1996; Kanamori and Ross, 1993; Shen et al, 1998). Using our metabolic model, Gln(0) (initial total glutamine concentration measured under standard physiological conditions) was also fitted and the obtained value (2.4 ± 0.3 μmol/g) is in agreement with the measured concentration of 2.3 ± 0.4 μmol/g. Contrary to previous studies under ammonia infusion, we continuously quantified the total Gln increase over time, which allowed us to directly measure net Gln accumulation without any assumptions. The obtained V_syn–V_nt of 0.033 ± 0.001 μmol/g per minute was twofold lower than the previously reported value under hyperammonemia measured indirectly by extrapolation using ¹⁵N MRS (Shen et al, 1998). The main source of discrepancy between our study and the previous one may be related to the fact that the previous study required modeling extrapolation to estimate net Gln accumulation.

The pyruvate carboxylation flux measured using ¹³C MRS under normoammonimic conditions is well known to be high in the conscious rat (Oz et al, 2004) and nearly absent under isoelectricity (Choi et al, 2002; Sibson et al, 1998), whereas under light α-chloralose anesthesia was measured to be 0.07 μmol/g per minute (Duarte et al, 2011). In the present study, the measurement of the pyruvate carboxylation flux gave a lower limit, which is likely to be higher due to pyruvate recycling and glutamine efflux. Consequently, the de-novo glutamine synthesis measured in our study under hyperammonemia might be assumed to be equal to the sum of the pyruvate carboxylation flux measured under normoammonimic conditions (Duarte et al, 2011) and the net Gln accumulation measured in our study under hyperammonemia, i.e., 0.07 + 0.033 = 0.103 μmol/g per minute. This estimated value under hyperammonemia for the de-novo glutamine synthesis is similar to previously reported unidirectional measurements under hyperammonemia using ¹³C MRS (Sibson et al, 2001) and indicated a significant increase in anaplerotic flux during hyperammonemia supporting the hypothesis that the anaplerotic flux is coupled to nitrogen removal from the brain (ammonia detoxification) under hyperammonemia conditions (Berl et al, 1962; Sibson et al, 2001; Zwingmann, 2007).

The incorporation of ¹⁵N of ammonia into [5-¹⁵N]Gln gave a V_syn of 0.30 ± 0.050 μmol/g per minute, which was higher than previous published data under hyperammonemia ranging from 0.06 to 0.20 μmol/g per minute obtained when using ¹⁵N MRS (Kanamori et al, 1993, 1996; Kanamori and Ross, 1993; Shen et al, 1998) and lower than 0.29 to 0.43 μmol/g per minute (Sibson et al, 1997, 2001) obtained when using ¹³C MRS. As discussed previously, the main source of discrepancy between our study and the previous ones might be related to the increased number of assumptions used for the modeling of the previous studies. Together with V_syn, we also fitted the plasma NH₃ FE which was 91 ± 5%, consistent with experimental measurements (Shen et al, 1998). The apparent GS rate obtained in the present study under hyperammonemia conditions 0.30 ± 0.050 μmol/g per minute (see Results) was higher than the value of 0.18 μmol/g per minute reported under normoammonimic conditions using ¹³C MRS (Duarte et al, 2011), supporting the role of GS as a detoxification pathway for ammonia in the brain.

From the values of V_syn–V_nt of 0.033 ± 0.001 μmol/g per minute and V_syn of 0.30 ± 0.050 μmol/min per gram, we obtained a V_nt of 0.26 ± 0.030 μmol/g per minute. Our results (see Results) indicate that under hyperammonemia conditions the neurotransmission rate represents ~80% of the flux through the GS which is comparable to the values reported previously using ¹⁵N MRS or ¹³C MRS (Shen et al, 1998; Sibson et al, 1997, 2001). Since in the present study, we are assuming a negligible efflux from the Gln pool compared with V_syn based on previous reports (Kanamori and Ross, 1993; Kanamori et al, 1996; Shen et al, 1998), the obtained neurotransmission rate V_nt represents an upper limit. By considering a Gln efflux of 0.07 μmol/g per minute (Duarte et al, 2011), we obtain a lower limit for V_nt of 0.19 μmol/g per minute which is similar with the value reported under normoammonimic conditions using ¹³C MRS (Duarte et al, 2011).

Although [2-¹⁵N] Glu and [2-¹⁵N] Gln cannot be separated in ¹⁵N spectra in vivo, the application of the model (Figure 2B; equations (18), (19) and (23) from Appendix) provided a V_GDH with a high precision. Consequently, the incorporation of ¹⁵N into [2-¹⁵N]Gln + Glu alone, provided a V_GDH of 0.029 ± 0.002 μmol/g per minute, where V_syn and V_nt needed to model the exchange between [2-¹⁵N] Glu and [2-¹⁵N] Gln were obtained from the simultaneous fit of the position 5N and total Gln. To maintain a constant Glu concentration with increasing Gln, glutamate needs to be synthesized from tricarboxylic acid intermediates. To determine the predominant cellular localization of GDH, we performed high-resolution measurements on brain extracts and consequently measured the ratio between the FEs of [2-¹⁵N]Glu and [2-¹⁵N]Gln (see Results). From these measurements, we obtained ([2-¹⁵N]Gln)FE > ([[2-¹⁵N]Glu)FE which showed that net glutamate synthesis results from higher GDH activity in the glial compartment. Note that previous studies assumed that this replenishment takes place primarily in the glial compartment in hyperammonemic conditions (Shen et al, 1998). The high glial GDH flux obtained in our study is in agreement with higher specific activity in astrocytes relative to neurons (Schmitt and Kugler, 1999; Zaganas et al, 2001). In the presented model, a unidirectional flux was assumed for GDH in the astrocytes. When including a reversible flux for the GDH reaction, the reversible component of GDH was always converging to zero in the fitting process. In addition, our results also showed that under hyperammonemia V_GDH≈ (V_syn–V_nt), meaning that the rate of breakdown of Glu to glial 2-oxoglutarate was negligible. To our knowledge, there was only one previous study which measured V_GDH under ¹⁵N ammonia infusion (Kanamori and Ross, 1995) reporting a V_GDH value of 0.018 to 0.020 μmol/min per gram for an infusion rate of 3.3 mmol/kg per hour, which is consistent with our results considering that our infusion rate was higher (4.5 mmol/kg per hour) and thus likely to stimulate glutamine synthesis.

The incorporation of ¹⁵N of ammonia takes place mainly into [5-¹⁵N]Gln through V_syn and into [2-¹⁵N]Glu + Gln through V_GDH. Since V_GDH obtained in our study represents only 10% from the obtained V_syn, it supports the concept of GS as a major detoxification pathway for ammonia in the brain.

We conclude that it is feasible to combine localized in vivo ¹⁵N with ¹H MRS and with an appropriate mathematical model to directly, simultaneously and reliably detect for the first time along with V_syn, the net glutamine accumulation, V_nt and V_GDH under ammonia infusion in the in vivo rat brain. From the modeling point of view, this new acquisition strategy required fewer assumptions to determine metabolic rates. In addition, this new strategy allowed a robust absolute quantification of total Gln, [5-¹⁵N]Gln and [2-¹⁵N]Gln + Glu in the same experiment. Our results showed an increase of GS and net glutamine accumulation under hyperammonemia, supporting the concept of their implication in cerebral ammonia detoxification. In addition, under hyperammonemia conditions the neurotransmission rate represents ~80% of the flux through the GS. The higher ([2-¹⁵N]Gln)FE compared with ([[2-¹⁵N]Glu)FE showed that net glutamate synthesis results from higher GDH activity in the glial compartment. Our results also showed that under hyperammonemia the rate of breakdown of Glu to glial 2-oxoglutarate was negligible because V_GDH≈ (V_syn–V_nt). Finally, the role of GS as an important detoxification pathway for ammonia in the brain was also supported by the fact that, in our study, the obtained V_GDH was smaller than V_syn.