SEARCH

SEARCH BY CITATION

Keywords:

  • polarization transfer;
  • 13C NMR spectroscopy;
  • localization;
  • brain metabolism

Abstract

  1. Top of page
  2. Abstract
  3. THEORY
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. CONCLUSION
  8. Acknowledgements
  9. APPENDIX
  10. REFERENCES

Localized 13C NMR spectra were obtained from the rat brain in vivo over a broad spectral range (15–100 ppm) with minimal chemical-shift displacement error (<10%) using semi-adiabatic distortionless enhancement by polarization transfer (DEPT) combined with 1H localization. A new gradient dephasing scheme was employed to eliminate unwanted coherences generated by DEPT when using surface coils with highly inhomogeneous B1 fields. Excellent sensitivity was evident from the simultaneous detection of natural abundance signals for N-acetylaspartate, myo-inositol, and glutamate in the rat brain in vivo at 9.4 T. After infusion of 13C-labeled glucose, up to 18 13C resonances were simultaneously measured in the rat brain, including glutamate C2, C3, C4, glutamine C2, C3, C4, aspartate C2, C3, glucose C1, C6, N-acetyl-aspartate C2, C3, C6, as well as GABA C2, lactate C3, and alanine C3. 13C-13C multiplets corresponding to multiply labeled compounds were clearly observed, suggesting that extensive isotopomer analysis is possible in vivo. This unprecedented amount of information will be useful for metabolic modeling studies aimed at understanding brain energy metabolism and neurotransmission in the rodent brain. Magn Reson Med 50:684–692, 2003. © 2003 Wiley-Liss, Inc.

13C NMR spectroscopy is a powerful tool to study compartmentalized metabolism in the brain (1). In particular, the measurement of 13C labeling time courses during infusion of 13C-labeled glucose and their analysis with a metabolic model has the potential to yield quantitative metabolic fluxes through a number of important pathways, including glycolysis, tricarboxylic acid (TCA) cycle, glutamate-glutamine cycle, pyruvate carboxylase, and malate-aspartate shuttle. Such metabolic modeling becomes more robust as the number of simultaneously detected 13C labeling time courses increases, because fewer assumptions are necessary in the model. Therefore, metabolic modeling studies ideally require broadband detection of many metabolites with excellent temporal and spectral resolution.

Most initial in vivo NMR studies of rodents using direct13C detection have been performed without localization (2–4). However, localization is desirable to detect spectra from a well-defined region of the brain and to avoid contamination by signals from noncerebral tissue. In addition, the lack of localization in directly detected 13C NMR spectra requires the measurement of a natural abundance background spectrum, introducing potential subtraction errors (due to motion or linewidth changes) as an additional experimental complication and reducing the sensitivity of the experiment. Recently, a nonecho localization method based on 13C outer volume suppression (OVS) provided 13C NMR spectra where doubly labeled isotopomers of glutamate and glutamine could be clearly identified (5). However, direct 13C localization invariably leads to a significant chemical-shift displacement error, especially at high B0 magnetic field, so that only a portion of the 13C spectrum is usually detected. It is therefore not surprising that localized broadband detection of the entire 13C spectrum has not been achieved using direct localization of the 13C magnetization.

An elegant solution to reduce the chemical-shift displacement error takes advantage of the much smaller range of 1H chemical shifts. In this scheme, localization is performed on the 1H magnetization and the localized magnetization is subsequently transferred to carbon using polarization transfer sequences like distortionless enhancement by polarization transfer (DEPT) (6), INEPT (7), or J-cross-polarization (8). The use of DEPT combined with proton localization was first proposed with a volume coil in phantom experiments (9) and was reported to be feasible when using a surface coil as a transceiver (10). Subsequent studies reported the use of full 3D localization in the human brain in vivo using surface coils (11). Incorporation of adiabatic pulses was shown to improve efficiency of polarization transfer when using surface coils with inhomogeneous RF fields (12, 13). Finally, polarization transfer sequences combined with 1H-localization were used to measure 13C-label incorporation in human brain metabolism during 13C-labeled glucose infusion (11, 13–16) and in the monkey brain (17). To the best of our knowledge, however, the use of polarization transfer to detect localized 13C NMR spectra has not been reported in the rat brain in vivo.

Therefore, the aims of the present study were: 1) to achieve localized and broadband detection of 13C NMR spectra in the rat brain with minimum chemical-shift displacement error using polarization transfer; 2) to demonstrate the excellent sensitivity and spectral resolution using this technique at 9.4 T; and 3) to evaluate the possibility of detecting simultaneously singly and multiply labeled isotopomers in vivo. Overall, these methodological advances resulted in an unprecedented sensitivity and a spectral quality hitherto achievable only in extracts.

THEORY

  1. Top of page
  2. Abstract
  3. THEORY
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. CONCLUSION
  8. Acknowledgements
  9. APPENDIX
  10. REFERENCES

Gradient-Dephased Semi-adiabatic DEPT Sequence

In principle, the DEPT sequence (6) requires precisely calibrated RF pulses to maximize efficiency and minimize unwanted coherence pathways (Fig. 1a). However, in vivo NMR measurements often use surface coils with inhomogeneous B1 fields for better sensitivity. In addition, at high B0 magnetic fields, homogeneous RF fields are difficult to achieve even with volume coils. Under these conditions, well-defined flip angles cannot be obtained over the entire volume of interest using conventional RF pulses, especially if the size of the voxel is comparable to the dimensions of the coil.

thumbnail image

Figure 1. Semi-adiabatic 1H-localized polarization transfer. a: the original DEPT sequence. b: the modified sequence used in the present study. The 13C part of DEPT was replaced by a segmented adiabatic 0° BIR-4 pulse. Dephasing gradients (1 ms duration, 46 mT/m) where added between the RF pulses in DEPT to eliminate unwanted, offset-dependent coherences. The flip angle of last 1H pulse was set to a nominal 45° to detect signals simultaneously from the CH, CH2, and CH3 groups.

Download figure to PowerPoint

The main consequence of imperfect RF pulses is to reduce the polarization transfer efficiency, thereby decreasing the sensitivity of the experiment. To limit signal loss, the 13C part of DEPT can be replaced by a segmented 0° BIR-4 (13), resulting in the scheme shown in Fig. 1b. The principal advantage of the added BIR-4 pulse is that the first and second segment of the BIR-4 pulse achieve constant 90° and 180° angles, provided that the adiabatic condition is met (18, 19). The last segment is necessary to refocus the nonlinear phase generated by the first two segments. The use of a segmented BIR-4 was particularly advantageous in our study for the 13C part of the pulse sequence because of the small size of the 13C coil, which resulted in a highly inhomogeneous B1 field. In contrast, square pulses can be kept for the proton part of DEPT provided that the B2 field is reasonably homogeneous over the localized volume, as with the quadrature 1H coil used in the present study (20).

Another consequence of imperfect RF pulses is the creation of unwanted coherences whose intensity may be dependent on frequency offset. In particular, imperfections in the 1H 180° refocusing pulse result in a signal intensity that is dependent on the 1H frequency offset. Indeed, it can be shown (see calculation in Appendix) that if the 1H refocusing pulse angle is assumed to be 2α (instead of 180°), the resulting coherence after polarization transfer is:

  • equation image(1)

The first term corresponds to the detected magnetization under ideal condition (i.e., 2α = 180°). The third term IzSy corresponds to antiphase magnetization and is canceled when using 1H decoupling during the acquisition time. The second term corresponds to detected magnetization that is dependent on the 1H frequency offset ω when 2α deviates from the nominal 180° value. One well-known remedy to eliminate unwanted coherences generated by imperfect 180° pulses is the EXORCYCLE phase cycle (21). However, because DEPT contains two refocusing pulses, a minimum of 16 phase cycle steps are necessary to eliminate unwanted coherences (10). ISIS localization requires an additional 8 steps phase cycle, resulting in at least 128 steps. Such long phase cycle schemes can be detrimental in vivo because possible movement or linewidth changes can cause imperfect signal cancellation.

Therefore, we propose here to eliminate unwanted coherences by using gradient pulses (“spoilers”) during the interpulse delays instead of using phase cycling (Fig. 1b). Because the second gradient pulse acts simultaneously on 1H and 13C transverse coherences, only three gradient pulses are necessary for the two 180° refocusing pulses. The combination of DEPT with gradient spoiling renders phase cycling of either 180° pulse unnecessary, thus reducing the length of the phase cycle scheme by a factor of 16.

Finally, accurate localization requires the elimination of any signal coming from the initial 13C magnetization, which does not experience 1H localization pulses, and is therefore not localized. In the original DEPT sequence, elimination of this directly excited signal was achieved by phase cycling the last proton pulse Θ[±y] together with the receiver [±x]. In the semi-adiabatic DEPT described here, the 0° BIR-4 returns directly excited 13C magnetization back to the z axis. In addition, signal that does not experience the 1H localization pulses is eliminated by the receiver phase cycle during ISIS. Therefore, it was not necessary to apply a phase cycle to the last 1H pulse of DEPT in this scheme.

To summarize, the new gradient-dephased semi-adiabatic DEPT sequence proposed in here has been designed to provide accurate 3D localization of 13C signals while minimizing sensitivity losses and offset-dependent unwanted coherences generated in the presence of inhomogeneous RF fields, and while minimizing the length of the phase cycle.

MATERIALS AND METHODS

  1. Top of page
  2. Abstract
  3. THEORY
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. CONCLUSION
  8. Acknowledgements
  9. APPENDIX
  10. REFERENCES

Instrumentation

All experiments were performed on a 9.4T/31cm horizontal bore magnet interfaced to an Inova console (Varian, Palo Alto, CA). The gradient insert (11 cm internal diameter) reached 300 mT/m in 500 μs and was equipped with a customized set of second-order shim coils (Magnex Scientific, Oxford, UK) capable of generating shim fields of 0.01 mT/cm2/A with a 4A shim power supply (Resonance Research, Billerica, MA). The coil assembly consisted of a 1H quadrature surface coil and a 13C linearly polarized surface coil built according to a previously described design (20). The maximum γB1/2π at the coil center was 10 kHz for the 13C coil and 6 kHz for the proton coil. To reduce interference between the 13C and 1H channel, two custom-manufactured proton bandpass filters at 400 MHz (Trilithic, Indianapolis, IN) were inserted between each proton coil and the quadrature hybrid and a 13C lowpass filter at 100 MHz (Trilithic) was inserted into the 13C path between the RF coil and the T/R switch.

Pulse Sequence

In all experiments, multislice RARE images (3 × 3 cm2, matrix 256 × 128, 11 slices, TR = 5 sec, TE = 60 ms, 8 echoes per excitation) were used to position the voxel. Localized shimming was performed using FASTESTMAP (22), resulting in an 18–23 Hz full width at half height of the water resonance in a 400 μl voxel (9 × 5 × 9 mm3).

The pulse sequence consisted of 1H localization followed by semi-adiabatic DEPT (13) (Fig. 1). Localization was performed using 3D-ISIS (hyperbolic secant inversion pulses (23), 2 ms duration, 8 kHz bandwidth) combined with outer volume suppression (OVS). The OVS module consisted of six slice-selective excitation pulses (HS20 (24), bandwidth 42 kHz at 1 ms duration) followed by dephasing gradients. The slice thickness was 15 mm along X (1 ms pulse duration); 15 mm along Z (1.5 ms pulse duration); 12 mm (2 ms pulse duration) and 3.4 mm (1 ms pulse duration) along Y, below and above the voxel, respectively. The carrier frequency was placed at 3.07 ppm for all 1H localization RF pulses. The shift in voxel position caused by chemical-shift displacement error was largest for the H1 of α-glucose at 5.23 ppm (2.16 ppm frequency offset from the carrier, or 865 Hz at 9.4 T, leading to a displacement error of ∼10% of the voxel dimension). Different chemical-shift displacement errors for ISIS and for OVS slices also caused signal loss off-resonance, the effect of which can be calibrated in phantom experiments.

For polarization transfer, the last pulse of DEPT was set to 45° to allow simultaneous detection of CH, CH2, and CH3 resonances. Square pulses of duration 50, 100, and 25 μs were used on the 1H channel for the 90, 180, and 45° pulse, respectively, corresponding to a γB1/2π of 5 kHz. The interpulse delay τ in DEPT was chosen equal to 3.8 ms, optimal for JCH = 130 Hz (only glucose has significantly different JCH values equal to 160–170 Hz as stated previously (25)). The 1H power was calibrated at the coil center on a sphere containing 99% enriched formic acid (26). This calibration was used to set the optimal 1H RF power on the voxel based on phantom precalibration measurements at the same voxel position relative to the 13C coil center. A segmented 0° BIR-4 (2 ms duration) was used on the 13C channel to minimize signal loss due to B1 inhomogeneity. The carrier frequency was placed at 49.3 ppm for 13C and at 3.07 ppm for 1H. WALTZ-16 composite pulse decoupling was applied during the acquisition time (205 ms, 8192 complex points) with a nominal γB1/2π of 2 kHz. Gradient spoiling (46 mT/m, 1 ms duration) was applied between RF pulses in DEPT to eliminate unwanted coherences. Performance of the sequence and precalibration of 1H power was determined in a phantom containing 200 mM of 99% enriched [1-13C]glucose.

Animal Preparation

Animal experiments were carried out in accordance with the guidelines for the care and use of laboratory animals and were approved by the Institutional Animal Care and Use Committee (IACUC) of the University of Minnesota. Male Sprague-Dawley rats, fasted overnight, were intubated and ventilated with a 70/30 N2O/O2 mixture and 1.8% isoflurane. Both femoral veins and femoral arteries were cannulated. The right and left veins were used to infuse glucose and α-chloralose, respectively. Arterial blood was drawn to determine blood gases and plasma glucose concentrations. After completion of surgery, isoflurane was discontinued and replaced by intravenous α-chloralose (bolus 40 mg/kg, then infusion at 25.4 mg/kg/hr). The animal was placed prone in a custom-built stereotaxic holder including a bite-bar and ear barrels. Body temperature was maintained at 37°C with hot water circulation regulated by feedback from a rectal temperature probe. Physiological parameters were maintained within normal limits (pH = 7.35–7.45, pCO2 = 30–45 mmHg, pO2 > 100 mmHg) by small adjustments of respiration rate and volume.

Glucose Infusion

After adjustment of NMR parameters, a bolus of 99% enriched 13C-labeled glucose was infused i.v. at an exponentially decaying rate over 5 min, followed by a continuous infusion of 70% enriched glucose as previously described (27). This protocol was designed in bench experiments and verified to result in a rapid rise of glucose 13C isotopic enrichment from 1.1 to 70%. Plasma glucose was measured every 15–20 min and maintained constant at ∼16 mM (corresponding to a brain glucose concentration of 4.5 μmol/g according to Ref. 28) by small adjustments of the glucose infusion rate. 13C NMR spectra were collected in blocks of 5.4 min over 7 hr. [1-13C]Glucose was purchased from Isotec (Miamisburg, OH); [1,6-13C2]glucose from Cambridge Isotope Laboratories (Andover, MA); and [U-13C6]glucose from Martek (Columbia, MD).

Data Processing

NMR signals were averaged and saved on disk in blocks of 128 scans (repetition time 2.5 sec) corresponding to an acquisition time of 5.4 min. Spectra averaged over longer periods of time were obtained by adding 5.4 min blocks after frequency correction to correct for the drift of the B0 magnetic field. The frequency drift was determined using the glutamate C4 peak at 34.4 ppm and frequency correction was applied directly in the time domain. For natural abundance spectra, no frequency correction was used. Fourier transform was performed with filtering parameters specified in the figure captions. No baseline correction was applied.

RESULTS

  1. Top of page
  2. Abstract
  3. THEORY
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. CONCLUSION
  8. Acknowledgements
  9. APPENDIX
  10. REFERENCES

The elimination of off-resonance effects induced by semi-adiabatic polarization transfer in inhomogeneous RF fields was validated in phantom studies. As expected from the theoretical analysis (Eq. [1]), the use of DEPT with a surface coil resulted in an oscillating excitation profile when the decoupler 1H frequency was varied (Fig. 2a). These oscillations were completely suppressed by gradient spoiling (Fig. 2b). This demonstrates the efficiency of gradient spoiling to eliminate unwanted pathways generated by imperfect RF pulses.

thumbnail image

Figure 2. Decoupler off-resonance effects of polarization transfer when using a surface coil as a transceiver. The C1 resonance of β-D-glucose at 96.7 ppm was measured as a function of the 1H decoupler offset (10 Hz increments) in a phantom containing 200 mM of 99% enriched [1-13C]glucose using the sequence depicted in Fig. 1b. a:Without gradient spoiling, the detected signal oscillates as a function of 1H frequency offset. b:With gradient spoiling, oscillations are suppressed.

Download figure to PowerPoint

Spectra acquired in the rat brain in vivo using the sequence depicted in Fig. 1b showed substantial sensitivity gains compared to previous localized 13C NMR studies of rat brain (5, 29) (Fig. 3). For example, natural abundance 13C NMR signals of metabolites were detected for the first time in the rat brain (Fig. 3a). Several resonances from natural abundance glutamate, myo-inositol, and NAA (corresponding to a 13C concentration of ∼0.1 μmol/g) were clearly observed. Reliable localization of the NMR signals to the brain over the entire spectral range of 8 kHz was evident from the complete absence of lipid signal at 30 ppm, of lipid glycerol resonances at 61 and 69 ppm, and of olefinic carbon resonances at 130 ppm, all of which arise from skin and muscles surrounding the brain. After infusion of [1,6-13C2]glucose, up to 18 resonances were detected (Fig. 3b). These included signals from the C1 of α- and β-glucose at 92.9 and 97.8 ppm, the C6 of α- and β-glucose at 61.7 and 61.8 ppm, and signals from all CHn resonances of glutamate (at 55.9, 34.4, and 27.9 ppm), glutamine (at 55.2, 31.9, and 27.1 ppm), and aspartate (at 53.2 and 37.5 ppm). In addition, weaker resonances from NAA, GABA, alanine, and lactate were detected.

thumbnail image

Figure 3. In vivo 13C spectra from a 400 μl volume in the rat brain, acquired using the modified DEPT sequence depicted in Fig. 1b. a: Natural abundance spectrum acquired for 3.5 hr (5120 scans, TR 2.5 sec). b: Spectrum recorded for 1.8 hr (2560 scans, TR = 2.5 sec) during an infusion of 70% enriched [1,6-13C2]glucose. Acquisition was started 1.8 hr after the beginning of glucose infusion. c: Expansion of b showing the detection of doubly labeled isotopomers of glutamate, glutamine, and aspartate. Note the triplet of glutamate (Glu-C3T) at 27.9 ppm, corresponding to glutamate labeled simultaneously at the C2, C3, and C4 positions. Processing consisted of zero-filling, filtering (5 Hz Lorentzian line broadening for (a) and 2 Hz Lorentzian-to-Gaussian resolution enhancement for (b) and (c)), and fast Fourier transform. No baseline correction was applied.

Download figure to PowerPoint

The linewidth of 13C singlet resonances were typically 5–6 Hz (0.05 ppm). The high spectral resolution in the spectrum can be appreciated from the observation that the signals from the C6α and C6β glucose resonances were resolved in vivo despite a chemical shift difference of only 0.14 ppm (14 Hz at 9.4T). An expansion of the spectral region at 55 ppm and at 30 ppm (Fig. 3c) shows that with such a narrow linewidth the fine structure due to 13C-13C couplings was readily observed. For example, the glutamate C3 resonance at 27.9 ppm is a superposition of a singlet (Glu-C3S), a doublet (Glu-C3D), and a triplet (Glu-C3T). The detection of these multiply labeled amino acids following infusion of [1-13C] or [1,6-13C2] glucose reflected the cycling of 13C label through the TCA cycle.

Comparing the effect of infusing [1-13C]glucose, [1,6-13C2]glucose, and [U-13C]glucose (all infused with a fixed isotopic enrichment of 70%) on the detection of the 13C-13C isotopomers of glutamate and glutamine clearly indicated a strong influence of the type of 13C-labeled precursor being administered (Fig. 4). With [1-13C]glucose (Fig. 4a), the spectrum was dominated by singlets, with small contributions from the doublets. After infusion of [1,6-13C2]glucose (Fig. 4b), the relative proportion of multiplets in glutamate and glutamine compared to the singlets was increased due to the expected 2-fold increase in the isotopic enrichment of [2-13C]acetyl-CoA entering the TCA cycle. In this case, a triplet, Glu-C3T, was clearly detected at 27.9 ppm. Lastly, when infusing [U-13C6]glucose (Fig. 4c), the glutamate and glutamine C4 resonances were further split by the coupling (J45 = 53 Hz) to the C5 (carboxyl) carbons in glutamate and glutamine which were labeled from the C2 and C5 carbons of glucose.

thumbnail image

Figure 4. Effect of the infused substrate on the labeling pattern. Shown are the spectral regions containing glutamate and glutamine C4 and C3 in vivo after infusion of (a) [1-13C]glucose, (b) [1,6-13C2]glucose, and (c) [U-13C6]glucose (volume 400 μl). The isotopic enrichment of the infusate was 70% in all cases. Each spectrum corresponds to 1.8 hr of acquisition (2560 scans, TR = 2.5 sec) started 4.5 hr after the beginning of glucose infusion. Mild Lorentzian-to-Gaussian resolution enhancement was used.

Download figure to PowerPoint

Using the proposed methodology, labeling time courses of many resonances were detected simultaneously in the same animal over extended periods of time at a high temporal resolution of 5 min (Fig. 5a). During 7 hr of infusion of [1,6-13C2]glucose, glutamate C4 was labeled first along with the C2 and C3 of aspartate, followed by glutamine C4 and the C2 and C3 carbons of glutamate and glutamine. We also noted the labeling of the NAA C2, C3, and C6 resonances at a much slower rate (Fig. 5b), as well as the detection of label in the GABA C2, C3, and C4 resonances with a lower temporal resolution (not shown).

thumbnail image

Figure 5. Time course of 13C label incorporation during an infusion of [1,6-13C2]glucose. The stack plot of an expansion of the 13C spectra shows (a) the first 90 min with a temporal resolution of 5.4 min (128 scans each, TR = 2.5 sec), showing label incorporation into glucose C6, glutamate C4, C3, C2, glutamine C4, C3, C2, and aspartate C2 and C3; (b) the subsequent 5.5 hr with a temporal resolution of 22 min, showing the slower labeling of NAA C2, C3, and C6. Mild Lorentzian-to-Gaussian resolution enhancement was applied.

Download figure to PowerPoint

The improved sensitivity was used to detect 13C label in smaller volumes of 48 μl matched to the somatosensory cortex (Fig. 6). Resonances from glucose C6, glutamate C2, C3, C4, and glutamine C2, C3, C4, and aspartate C2, C3 were readily detected in 11 min, indicating the feasibility of combining time-resolved direct-detected 13C spectroscopy with functional activation of somatosensory cortex in the rat brain.

thumbnail image

Figure 6. Detection of 13C label in the rat somatosensory cortex during infusion of 70% enriched [1,6-13C2]glucose. The VOI (48 μl) is shown on transverse and sagittal RARE images. 13C resonances of glutamate, glutamine, and aspartate, as well as the C6 resonance of glucose, can be clearly identified. The spectrum was acquired in 11 min (256 scans, TR 2.5 sec) after 2 hr of glucose infusion. Processing consisted of 7 Hz Lorentzian line-broadening and zero-filling (128 k points). Shown without baseline correction.

Download figure to PowerPoint

DISCUSSION

  1. Top of page
  2. Abstract
  3. THEORY
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. CONCLUSION
  8. Acknowledgements
  9. APPENDIX
  10. REFERENCES

This study demonstrates that efficient localization covering the entire range of 13C resonances can be achieved in the rat brain in vivo. The chemical-shift displacement was minimized by using proton localization followed by polarization transfer, taking advantage of the smaller chemical-shift range of 1H spectra. For most detected metabolites, such as glutamate, glutamine, NAA, aspartate, and GABA, whose proton chemical-shift falls in the range 1.9–4.3 ppm, the chemical-shift displacement error was less than 6% of the voxel dimension in each spatial direction. The largest chemical-shift displacement error occurred for the H1 of α-glucose at 5.23 ppm and did not exceed 10% of the voxel dimension. Accurate localization over the entire spectral range was evident from the complete absence of lipid signals at 30, 63, and 69 ppm, as well as the elimination of the olefinic carbon resonances at 130 ppm (not shown), all of which arise from surrounding adipose and muscle tissues.

For the localization of carbon resonances with small or absent coupling to protons (JCH ∼0) such as the carboxyl resonances around 180 ppm (e.g., the C1 and C5 of glutamate), obviously polarization transfer cannot be used and localization methods based on the 13C magnetization are required, like outer volume suppression (5) or ISIS (30). 13C resonances with very short relaxation times are also not easily detected using polarization transfer methods (due to relaxation during polarization transfer) and nonecho localization methods have been proposed for the measurement of glycogen (5).

Although polarization transfer greatly reduced the chemical-shift displacement error, deviations of the RF pulses from the nominal flip angle resulted in less than maximal efficiency and selection of unwanted coherence transfer pathways. In particular, an imperfect 1H 180° RF pulse created unwanted coherences that resulted in an excitation profile that oscillated as a function of 1H frequency offset (Fig. 2a). Precise calibration of RF pulses can be achieved even with surface coils, provided that the RF field remains reasonably homogeneous across the voxel chosen for spectroscopy (11, 13). Alternatively, if this cannot be achieved, e.g., when the voxel dimensions are large compared to the coil, as was the case in the present study, unwanted coherences can be eliminated by phase cycling the 180° refocusing pulses for both nuclei (10). However, significant signal loss can occur in regions where the flip angles deviate from their nominal value. To optimize sensitivity, we replaced the 13C part of DEPT by a segmented 0° BIR-4. This resulted in optimal pulse angles for the 13C channel and eliminated the need for precise 13C power calibration, which was critical given the small size of the 13C coil (11 mm diameter). We furthermore introduced the use of dephasing gradients to eliminate the need for phase cycling 180° pulses on either 13C or 1H channels, thereby greatly reducing the number of steps in the phase cycle.

The combination of 1H localization and polarization transfer at 9.4 T resulted in the localized detection of the entire 13C spectrum (with the exception of resonances from carbons that have no proton attached, such as carboxyl COOH groups) with excellent sensitivity, as demonstrated by the first simultaneous detection of natural abundance signals of NAA, glutamate and myo-inositol in the rat brain in vivo as illustrated in Fig. 3a. The C1,3 resonance of myo-inositol was detected at 72.1 ppm in Fig. 3a with a signal-to-rms noise ratio of 5, consistent with the fact that natural abundance signals were observed routinely in the rat brain in vivo in less than 1 hr. The use of one of the highest B0 fields available to date for this type of in vivo studies was certainly an important factor contributing to the increased sensitivity. Another important element was the combination of a circularly polarized 1H coil and a linearly polarized 13C coil used (20), which retained a high sensitivity for 13C combined with a relatively homogeneous B2 field for 1H localization, decoupling, and polarization transfer. Finally, optimal adjustment of all first- and second-order shims using an automated procedure such as FAST(EST)MAP (22) was crucial to attain optimal sensitivity and spectral resolution, especially at high B0 magnetic field and for the larger volumes used in 13C NMR (31).

After infusion of 13C-labeled glucose, the new sequence allowed for the simultaneous localized detection of up to 18 resonances ranging from glucose C1 to lactate C3 over an 80 ppm bandwidth. This represents a substantial improvement over previous studies using 13C detection in the rat brain, which either measured the whole spectrum without localization (3, 4), or detected signal from a smaller spectral range with localization (5).

The excellent sensitivity and spectral resolution further resulted in the detection of multiply labeled metabolites (isotopomers) appearing in the spectrum as multiplets. Spectral resolution has been shown to increase with a higher magnetic field B0 due to increased chemical-shift dispersion (32). Again, excellent shimming was necessary to resolve homonuclear couplings, as these decrease (when expressed in ppm) with B0. In the present study a 13C linewidth of 5–6 Hz was routinely achieved for a 400 μl volume. This linewidth was at most 2 Hz above the achievable minimum in rat brain due to residual higher-order shim terms (n ≥ 3) as estimated from the line width of water in this volume. Nonetheless, such a small linewidth allowed the routine detection of the 13C-13C coupled resonances with J-coupling constants ranging from 31–54 Hz, as illustrated in Figs. 3 and 4. Many studies, especially in brain extracts (33–37) and in the perfused heart (38, 39), have shown the usefulness of isotopomer analysis to investigate metabolic pathways and substrate selection. Our results suggest that isotopomer analysis is now feasible in vivo in the rat brain for several metabolites, including glutamate, glutamine, and aspartate, provided that adequate fitting methods are used.

The localization performance, sensitivity, and spectral resolution demonstrated in the present study are expected to considerably extend the scope of 13C NMR studies in the rat brain and animal brain in general. For example, in vivo metabolic studies of the brain to date have used only a fraction of the information potentially available in the 13C spectrum. Measurements have generally focused on one (typically glutamate C4) or two labeling time courses, be it glutamate C4 and glutamate C3 without measuring glutamine (40), or glutamate C4 and glutamine C4 without measuring the C3 (4). In contrast, the detection of time-resolved 13C labeling time courses of more than 15 resonances in vivo is likely to provide a much more comprehensive assessment of metabolism than previously achieved, together with a potential reduction in the number of assumptions in the metabolic modeling. Recent reports have emphasized the importance of including sufficient information in the metabolic modeling to derive quantitative metabolic fluxes (14, 27, 29). For example, measuring the C3 labeling time course of glutamate and glutamine, in addition to the C4 time course, allows the measurement of the exchange rate between 2-oxoglutarate and glutamate, Vx. A number of recent studies have shown that Vx is most likely slow in the brain, on the order of the TCA cycle flux, and may change with the level of brain activity. In addition, the detection of the C2 labeling time course of glutamine can be used to determine the rate of anaplerosis through pyruvate carboxylase. The detection of aspartate and GABA provide additional information that will be of primary importance to further investigate metabolic compartmentation. In addition, the feasibility to detect 13C labeling in the rat somatosensory cortex (Fig. 6) emphasizes the possibility to measure multiple 13C labeling time courses during functional activation.

CONCLUSION

  1. Top of page
  2. Abstract
  3. THEORY
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. CONCLUSION
  8. Acknowledgements
  9. APPENDIX
  10. REFERENCES

In conclusion, localized, broadband 13C NMR spectroscopy at high magnetic field can detect 13C resonances over an 80 ppm bandwidth with less than 10% chemical-shift displacement error when using 1H-localization combined with polarization transfer. Excellent sensitivity and spectral resolution was achieved at 9.4 T in the rat brain. We conclude that metabolic information is now accessible in vivo that hitherto had to be obtained in extracts and that numerous labeling time courses in the brain during an infusion of a 13C-labeled substrate can be measured simultaneously in small, functionally specialized areas of the brain. These results enhance the role of 13C NMR spectroscopy as a valuable tool to further investigate models of brain metabolism and compartmentation in health, activation, and disease.

Acknowledgements

  1. Top of page
  2. Abstract
  3. THEORY
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. CONCLUSION
  8. Acknowledgements
  9. APPENDIX
  10. REFERENCES

We thank Professors Michael Garwood and Kamil Ugurbil for helpful discussions and encouragement, and Sarah Crawford for technical assistance.

APPENDIX

  1. Top of page
  2. Abstract
  3. THEORY
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. CONCLUSION
  8. Acknowledgements
  9. APPENDIX
  10. REFERENCES

Effect of an Imperfect 180° Refocusing Pulses in DEPT

The following calculation shows that an imperfect 1H 180° refocusing pulse in DEPT results in a signal that oscillates as a function of the 1H decoupler frequency offset. To simplify the calculation, an AX spin system was considered and all pulses were assumed to be at their nominal angle, except for the proton 180° pulse, for which the flip angle was set to 2α. The 13C observed frequency was assumed to be on resonance (no chemical-shift evolution at carbon frequency) and the effect of homonuclear coupling, JHH, coupling was neglected. The last 1H pulse of DEPT (Θ) was set to the optimal value of 90° for the AX system.

The first proton 90° pulse tilts the initial Iz magnetization to the transverse plane:

  • equation image(A1)

The magnetization evolves under 1H chemical-shift (ω) and J-coupling during the first delay τ = 1/2 J:

  • equation image(A2)

after the first 13C 90° pulse (Fig. 1a) this coherence is converted to:

  • equation image(A3)

and the application of the nominal 180° pulse, assumed here to be 2α, results in:

  • equation image(A4)

Evolution under chemical-shift and J-coupling during the second delay τ gives:

  • equation image(A5)

Following the 90math image pulse, the multiple-quantum term IySy leads to undetectable magnetization and was not further considered in the calculation. Finally, inversion by the 180° 13C pulse and J-coupling evolution during the last delay τ results in Eq. [1]:

  • equation image

REFERENCES

  1. Top of page
  2. Abstract
  3. THEORY
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. CONCLUSION
  8. Acknowledgements
  9. APPENDIX
  10. REFERENCES