SPECTROSCOPIC METHODOLOGY Full Papers
Magnetic Resonance in Medicine 73:451–458 (2015)
Detection of Glutamate, Glutamine, and Glutathione by Radiofrequency Suppression and Echo Time Optimization at 7 Tesla Li An,1* Shizhe Li,1 James B. Murdoch,2 Maria Ferraris Araneta,1 Christopher Johnson,1 and Jun Shen1 Purpose: To achieve detection of glutamate (Glu), glutamine (Gln), and glutathione (GSH) by minimizing the N-acetylaspartate (NAA) multiplet signals at 2.49 ppm using a echo time (TE) -optimized PRESS pulse sequence and a novel Jsuppression radiofrequency pulse. Methods: Using density matrix simulations, a PRESS sequence with (TE1, TE2) ¼ (69, 37) ms and an inserted 90 Jsuppression pulse were found to minimize the NAA multiplet at 2.49 ppm. Results: NAA phantom experiments confirmed the successful suppression of the NAA multiplet at 2.49 ppm. A study of eight healthy volunteers found both Glu and Gln to be significantly higher in gray matter (GM) dominant medial prefrontal cortex voxels than in white matter (WM) dominant right frontal cortex voxels. Time-course 1H spectra acquired during intravenous [U-13C6]glucose infusion showed gradually changing Glu C4 and Gln C4 proton resonance signals in a spectral pattern predicted by numerical simulations. Conclusion: Reliable detection of Glu, Gln, and GSH was achieved. Glu and Gln levels were significantly higher in frontal lobe GM than in frontal lobe WM. It is feasible to use the proposed proton MR spectroscopy method to measure the kinetics of 13C incorporation into Glu and Gln during infusion of 13C labeled glucose. Magn Reson Med 73:451–458, 2015. C 2014 Wiley Periodicals, Inc. V Key words: glutamate; glutamine; glutathione; J-suppression; TE-optimization
INTRODUCTION Glutamate (Glu) is the main excitatory neurotransmitter in the central nervous system and glutamine (Gln) is synthesized by glutamine synthetase from Glu and ammonia. Both Glu and Gln participate in the Glu-Gln cycle in which neuronal Glu released into the synaptic cleft is taken up by astrocytes, converted to Gln, and cycled back to neurons for replenishment of Glu. Studies in humans and animals have found altered levels of Glu and Gln in 1 National Institute of Mental Health, National Institutes of Health, Bethesda, Maryland, USA. 2 Toshiba Medical Research Institute USA, Mayfield Village, Ohio, USA.
*Correspondence to: Li An, Ph.D., Building 10, Room 3D46, 10 Center Drive, MSC 1216, Bethesda, MD 20892-1216. E-mail: [email protected]
Received 13 November 2013; revised 6 January 2014; accepted 8 January 2014 DOI 10.1002/mrm.25150 Published online 28 February 2014 in Wiley Online Library (wileyonlinelibrary. com). C 2014 Wiley Periodicals, Inc. V
several neurological and psychiatric diseases such as schizophrenia (1,2), depression (3,4), and epilepsy (5). Glutathione (GSH) is an antioxidant and detoxifier, and was implicated in several neurological diseases related to oxidative stress such as stroke (6) and multiple sclerosis (7). In proton MR spectroscopy (1H MRS), the C4 proton resonances of Glu (2.35 ppm), Gln (2.45 ppm), and the glutamyl moiety of GSH (2.54 ppm) overlap each other at low- and mid-fields (1.5–4.7 Tesla [T]). In contrast, at high fields such as 7T, the C4 proton resonances of Glu, Gln, and GSH are spectrally resolved, which makes the detection of Glu straightforward. However, the Gln multiplet at 2.45 ppm and the GSH multiplet at 2.54 ppm are still overlapped by the generally much larger signals from the aspartyl moiety of N-acetyl-aspartate (NAA) around 2.49 ppm. Despite this signal overlap, Gln and GSH can still be quantified using linear combination modeling, but small errors in NAA quantification and modeling of the NAA aspartyl multiplet would result in significant errors in quantification of Gln and GSH. Choi et al (8) have proposed using a echo time (TE) -optimized PRESS (point resolved spectroscopy) sequence at 7T to detect Glu, Gln, and GSH, taking advantage of the chemical shift offsets (9–11) to suppress overlapping signals from the aspartyl moiety of NAA. In this work, we propose suppressing spectral interference from the aspartyl moiety of NAA by a selective radiofrequency (RF) pulse placed at the resonance frequency of the NAA aspartyl CH proton at 4.38 ppm, thereby altering the J-evolution of the NAA aspartyl CH2 multiplet at 2.49 ppm. The flip angle of this suppression pulse along with the TE values of the PRESS sequence can be optimized to suppress the NAA aspartyl multiplet at 2.49 ppm in a single shot. This new sequence has been used at 7T to quantify Glu, Gln, and GSH in frontal lobe gray matter (GM) and white matter (WM) of human volunteers. Preliminary application of our method to detect Glu and Gln turnover from intravenously infused [U-13C6]glucose using proton MRS is also demonstrated. METHODS Suppression Pulse and TE Optimization A 20 ms sinc-Gauss RF pulse with its frequency targeting 4.38 ppm at 7T was inserted at the midpoint between the two refocusing pulses of a PRESS sequence. Because the NAA aspartyl CH proton at 4.38 ppm is J-coupled to the NAA aspartyl CH2 protons, this suppression pulse
indirectly affects the NAA aspartyl multiplet at 2.49 ppm. Different flip angles of the suppression pulse result in different NAA multiplet signals at 2.49 ppm. Density matrix simulation programs were developed using the GAMMA Cþþ library (12) to compute metabolite spectra for different values of TE1, TE2, and the flip angle of the suppression pulse. The objective of this optimization was to minimize the NAA multiplet at 2.49 ppm while retaining near-maximum peak amplitudes for C4 proton resonances of Glu, Gln, and GSH. The standard deviation of NAA signals between 2.44 and 2.54 ppm to the NAA singlet peak amplitude ratio was used as the criterion for determining the amount of NAA multiplet signals at 2.49 ppm. The simulations used chemical shift and Jcoupling constants from Govindaraju et al (13) for all metabolites except the glutamyl moiety of GSH, for which the refined chemical shift and J-coupling constants from Choi et al (8) were used. Experimental RF waveforms were used in density matrix simulations along with three-dimensional (3D) spatial localization using 201 201 201 spatial points (14). The excitation pulse was an amplitude-modulated pulse with duration ¼ 4.5 ms, FWHH (full width at half height) bandwidth ¼ 3.1 kHz, and maximum B1 ¼ 18.6 mT (15). The two refocusing pulses were also amplitudemodulated and had duration ¼ 8.0 ms, bandwidth ¼ 2.0 kHz, and maximum B1 ¼ 18.6 mT. The simulations showed that TE1 ¼ 69 ms, TE2 ¼ 37 ms, and suppression pulse flip angle ¼ 90 resulted in a minimum ratio of 0.20% between the standard deviation of NAA signals in the 2.44–2.54 ppm range and the NAA singlet peak amplitude, while retaining excellent spectral resolution and peak amplitude for Glu, Gln, and GSH.
An et al.
chemical shift difference between the NAA singlet peak and the water resonance frequency (DC component of the spectrum) was measured. If the phantom temperature matched the normal human body temperature of 37 C, the chemical shift difference would be 2.66 ppm (16). Because the phantom temperature was higher than 37 C at the beginning and went down slowly, the chemical shift difference was initially smaller than 2.66 ppm, then slowly approached 2.66 ppm as more scans were performed. When the chemical shift difference reached 2.66 ppm, the corresponding FID data were saved for more refined postprocessing. The refined postprocessing started with frequency- and phase-correction on each of the 16 transients to align the data. The aligned FIDs were then averaged over different transients for each channel. The 32-channel FID averages were merged into a combined single-channel metabolite FID by a generalized least square method (17). A combined water reference FID was similarly computed from the water-unsuppressed FIDs, and it was subsequently used to correct the phase errors in the combined metabolite FID caused by zero-order eddy currents (18). The corrected FID was apodized with a 3 Hz Gaussian function and then Fourier transformed into the frequency domain. The resulting spectrum was processed with an in-house developed linear combination fitting program that uses numerically simulated FID of NAA as the basis set and a Levenberg-Marquardt least square minimization algorithm. The simulated NAA FID was scaled, apodized using a Voigt lineshape, frequency shifted, zero-order phase corrected, and Fourier transformed to the frequency domain to fit the spectral data between 1.8 and 3.3 ppm.
NAA Phantom Experiments
In Vivo Experiments in the Frontal Lobe GM and WM Regions
MRS data of an NAA phantom were acquired on a Siemens 7T scanner equipped with a 32-channel receiver head coil to confirm that TE1 ¼ 69 ms, TE2 ¼ 37 ms, and suppression pulse flip angle ¼ 90 would lead to minimal NAA multiplet signals at 2.49 ppm. A spherical phantom containing 50 mM NAA solution with pH ¼ 7.0 at 22 C was made in-house. Before placing it into the scanner, the phantom was preheated to 42 C and covered with foam wraps to keep it warm. After a localizer scan, a 2 2 2 cm3 voxel in the center of the phantom was prescribed for the subsequent MRS scans. First- and secondorder B0 shim coefficients were adjusted for the voxel using a FASTMAP sequence. The parameters for the modified PRESS sequence were repetition time (TR) ¼ 2.5 s, TE1 ¼ 69 ms, TE2 ¼ 37 ms, suppression pulse flip angle ¼ 90 , spectral width ¼ 4000 Hz, number of data points ¼ 2048, and number of transients ¼ 16. Water suppression was accomplished using eight RF pulses of 350 Hz bandwidth. In addition, a single unsuppressed water FID (free induction decay) was collected at the end of the MRS scan. After the FIDs were acquired, a spectrum was reconstructed immediately using the averaged FID from the channel yielding the strongest FID signals. This MRS scan was performed repeatedly with the corresponding spectrum reconstructed right after the scan was finished. In each reconstructed spectrum, the
Eight healthy volunteers (four women and four men; age ¼ 32 6 11 years), all gave informed consent in accordance with procedures approved by our local institutional review board, were scanned on a Siemens 7T scanner to measure levels of Glu, Gln, GSH, and other metabolites in frontal lobe GM and WM. The 3D T1-weighted MPRAGE (magnetization prepared rapid gradient echo) images acquired with TR ¼ 3 s, TE ¼ 3.9 ms, matrix ¼ 256 256 256, and resolution ¼ 1 1 1 mm3 were used to localize the MRS voxel and perform tissue segmentation of the voxel with in-house developed software. For each subject, MRS data were collected from two 2 2 2 cm3 voxels, one in the GM dominant medial prefrontal cortex and the other in the WM dominant right frontal cortex. The pulse sequence parameters were similar to those of the phantom experiments except that the number of transients was increased to 128 and eight interleaved unsuppressed water acquisitions were performed, one after every 15 water-suppressed transients. Spectra were reconstructed from the FIDs using a procedure similar to the refined postprocessing used in the phantom experiments. To quantify metabolite concentrations, the spectral data were fitted with simulated basis sets of NAA, NAAG, g-aminobutyric acid (GABA), Glu, Gln, GSH, aspartate (Asp), creatine þ phosphocreatine (tCr), glycerophosphocholine þ phosphocholine (tCho), and
Detection of Glu, Gln, and GSH at 7T
FIG. 1. Pulse shape and frequency profile of the 90 J-suppression RF pulse.
myo-inositol (mI) using our linear combination fitting program. Cramer-Rao lower bounds (CRLB) were also computed (19). In the fitting process, the metabolite basis sets were scaled, apodized using a Voigt lineshape, frequency shifted, zero-order phase corrected, Fourier transformed to the frequency domain, and then combined with a spline baseline with eight control points to fit the in vivo spectral data between 1.8 and 3.3 ppm.
were numerically simulated using 13C chemical shift values and 1H-13C J-coupling constants reported by de Graaf (20). An ideal excitation pulse and two experimental refocusing pulses were used in a 2D spatial localization calculation with 201 201 spatial points to simulate the spectra accurately and efficiently.
RESULTS Carbon-13 Labeled Glucose Infusion Experiment To demonstrate the feasibility of applying this technique to measuring the labeling kinetics of Glu and Gln, a female volunteer was recruited for the 13C-labeled glucose infusion experiment following procedures approved by our local institutional review board. Before the scans, two antecubital veins of the subject were cannulated, one for infusing [U-13C6]glucose and the other for withdrawing blood to monitor glucose levels. A baseline MRS scan was performed before the infusion started using the same pulse sequence and sequence parameters as the previous study of eight healthy volunteers. The baseline scan lasted 6 min. The infusion of [U-13C6]glucose (20% w/w) started after the baseline scan at a bolus infusion rate of 900 mL/h followed by an exponential decay to the rate of 100 mL/h at the 15th minute of infusion. The subsequent infusion rate was adjusted to keep glucose levels at 160–200 mg/dL. MRS scans were performed repeatedly, and between scans, the resonance frequency was adjusted on the scanner to correct for small frequency drifts. A total of 10 MRS scans were performed during 82 min of infusion. Each set of MRS data was processed in the same way as the data in the previous study of eight healthy volunteers. The largest peaks in Glu and Gln 1H spectra correspond to the C4 protons. Moreover, 13C is incorporated into the C4 sites of Glu and Gln in the first turn of the tricarboxylic acid cycle during [U-13C6]glucose infusion. Hence, we focused our attention on the C4 proton signals to see if spectrally resolved 13C-labeled Glu and Gln C4 proton signals can be observed during the infusion. Before the in vivo 13C-labeled glucose infusion experiment, spectra of Glu, [4-13C]Glu, Gln, and [4-13C]Gln
The pulse shape and frequency profile of the 90 suppression pulse are plotted in Figure 1. The frequency profile between 620 Hz is very flat as the ratio of the longitudinal magnetization Mz to the equilibrium magnetization M0 at 620 Hz is only 0.0025. This ensures that the suppression effects of this RF pulse are virtually unchanged when scanner frequency drifts are within 620 Hz during an MRS scan, a condition that was met in our in vivo experiments. The frequency profile is also very flat (0.999 < Mz/M0 1) when the frequency offset is outside the range of 6150 Hz. The C2 proton resonances of Glu, Gln, and GSH at 3.75 ppm are 187 Hz away from the center frequency of the suppression pulse (4.38 ppm). Therefore, the C2, C3, and C4 proton resonances of Glu, Gln, and GSH are virtually undisturbed by the suppression pulse when scanner frequency drifts are within 630 Hz. Three sets of numerically simulated spectra for NAA, Glu, Gln, and GSH are plotted in Figure 2 using a concentration ratio of 1 : 0.7 : 0.15 : 0.15. The first set of spectra was computed using TE values proposed by Choi et al (8). The NAA multiplet signals at 2.49 ppm are relatively small but still comparable to the Gln peak at 2.45 ppm and the GSH peak at 2.54 ppm. The second set of spectra was computed using our proposed TE values of TE1 ¼ 69 ms and TE2 ¼ 37 ms but without the Jsuppression pulse. The NAA multiplet at 2.49 ppm is a large negative peak. For the third set of spectra, a 90 Jsuppression pulse was used along with the proposed TE values. The NAA multiplet signals are very small compared with the Gln and GSH peaks, which is advantageous for accurate quantification of Gln and GSH. The ratios of the standard deviation of NAA signals between 2.44 and 2.54 ppm to the NAA singlet peak amplitude
An et al.
FIG. 2. Numerically simulated spectra of NAA, Glu, Gln, and GSH with a concentration ratio of 1 : 0.7 : 0.15 : 0.15 using TE values proposed by Choi et al (8), our proposed TE values without the J-suppression pulse, and our proposed TE values with the J-suppression pulse. The spectra were broadened to a linewidth of 9 Hz for the NAA singlet.
were found to be 1.07%, 1.13%, and 0.20% for the three set of spectra, respectively. The spectrum of the NAA phantom is plotted in Figure 3, along with the numerically simulated NAA spectrum and the difference between the two. It can be seen that the spectrum from the phantom experiment agrees very well with the simulated spectrum. This result confirms that our proposed pulse sequence yields very low NAA multiplet signals at 2.49 ppm. For the MRS study of eight healthy volunteers, tissue segmentation results were as follows: voxels in the medial prefrontal cortex contained (64 6 6)% GM, (33 6 6)% WM, and (4 6 2)% CSF, whereas voxels in the right frontal cortex contained (18 6 7)% GM, (81 6 7)% WM, and (1 6 1)% CSF. The mean SNR of the NAA peak was 285 in the prefrontal cortex and 236 in the right frontal cortex. The SNR was computed as the ratio of the NAA peak amplitude to the standard deviation (SD) of the noise measured between 9 and 12 ppm in a spectrum generated by directly Fourier transforming the combined metabolite FID into the frequency domain without any line broadening or filtering. Spectra from the eight volunteers are displayed in Figure 4, and linear combination fitting plots for the last volunteer are displayed in Figure 5. The height of each spectrum was normalized by the total area of the NAA and NAAG peaks, a sum that is slightly higher in GM than in WM (21). In both medial prefrontal cortex and right frontal cortex, the Glu, Gln, and GSH peaks have well-defined shapes and are adequately separated. On average, the Glu peak at 2.35 ppm is noticeably larger in the medial prefrontal cortex than in the right frontal cortex, consistent with previous literature reports of substantially higher concentration of Glu in GM than in WM (22). A similar ratio is apparent for the Gln peak at 2.45 ppm. It is visually noticeable that the tCr and tCho peaks are broader in the right frontal cortex than in the medial
prefrontal cortex. The linewidths of tCr, tCho, and NAA were 9.9 6 0.9, 9.5 6 1.0 and 9.1 6 0.9 Hz, respectively, in the medial prefrontal cortex, compared with 11.6 6 0.7, 12.3 6 0.8, and 10.2 6 0.8 Hz in the right frontal cortex. On the other hand, water linewidth was similar in the two regions, which was 11.6 6 0.9 Hz in the medial prefrontal cortex and 11.1 6 1.8 Hz in the right frontal cortex. Quantification results from the eight volunteers for Glu, Gln, GSH, NAA, NAAG, tCr, and tCho are given in Table 1. The concentration values of GABA, Asp, and mI are not listed because, at the chosen TE, they are prone
FIG. 3. Comparison of experimental and numerically simulated NAA spectra. The experimental spectrum was obtained by measuring a NAA phantom at 37 C using the modified PRESS sequence with the J-suppression pulse (TR ¼ 2.5 s, TE1 ¼ 69 ms, TE2 ¼ 37 ms, suppression pulse flip angle ¼ 90 , spectral width ¼ 4000 Hz, number of data points ¼ 2048, and number of transients ¼ 16). The spectrum was apodized with Gaussian line broadening of 3 Hz. The simulated spectrum was obtained by density matrix simulation and fitted to the phantom spectrum by phase correction, frequency shift, and line broadening.
Detection of Glu, Gln, and GSH at 7T
FIG. 4. Stack plots of spectra from the prefrontal cortex and right frontal cortex of eight healthy volunteers acquired using the modified PRESS sequence with the J-suppression pulse (TR ¼ 2.5 s, TE1 ¼ 69 ms, TE2 ¼ 37 ms, suppression pulse flip angle ¼ 90 , spectral width ¼ 4000 Hz, number of data points ¼ 2048, and number of transients ¼ 128). Metabolite peaks downfield from 3.7 ppm were partially suppressed by the water suppression pulses that had a bandwidth of 350 Hz. Gaussian line broadening of 3 Hz was applied to all spectra.
to systematic modeling errors such as errors in modeling the baseline. In the medial prefrontal cortex, the ratios of Glu, Gln, and GSH to tCr were found to be 1.17 6 0.07, 0.25 6 0.03, and 0.21 6 0.02, respectively. In the right frontal cortex, the corresponding values were 1.06 6 0.09, 0.20 6 0.04, and 0.27 6 0.03. The ratios of Glu and Gln to tNAA were also computed. The Glu/tNAA ratio was 0.72 6 0.05 in the medial prefrontal cortex and 0.47 6 0.04 in the right frontal cortex. Corresponding values for Gln/ tNAA were 0.15 6 0.03 and 0.09 6 0.02. Both of the Glu/ tNAA and Gln/tNAA ratios were significantly higher (two tailed t-test, P < 0.001) in the medial prefrontal cortex than in the right frontal cortex. Because tNAA is slightly
higher in GM than in WM, it can be concluded that the absolute concentrations of Glu and Gln are significantly higher in frontal lobe GM than in frontal lobe WM, which agrees with previous findings from neurochemical analysis of brain tissues (23,24). Numerically simulated spectra of Glu, [4-13C]Glu, Gln, and [4-13C]Gln are displayed in Figure 6. Due to magnetization transfer within the strongly coupled glutamate spin system, the proton C4 peak at 2.35 ppm in the Glu spectrum is asymmetrically shifted in the [4-13C]Glu spectrum, that is, having only one major peak at 2.56 ppm plus some small resonances around 2.14 ppm that are mixed in with the C3 proton resonances. In the weak
An et al.
FIG. 5. Linear combination fitting plots for one healthy volunteer, whose spectra are displayed on the bottom row in Figure 4. Spectral data between 1.8 and 3.3 ppm were used in the data fitting.
coupling limit, the proton Glu C4 peak should split into two symmetrical peaks, one at 2.56 ppm and the other at 2.14 ppm, a result of the 127 Hz J-coupling constant between the C4 carbon and protons. However, [4-13C]Glu is not a simple weak coupling system because the two C4 protons are coupled to each other and to the C3 protons. Density matrix simulations were used to elucidate the [4-13C]Glu spectrum after two refocusing pulses and a relatively long evolution time at 7T. The appearance of the [4-13C]Gln spectrum is similar to the Glu case: the Gln peak at 2.45 ppm splits into a major peak at 2.66 ppm and some small resonance signals at 2.24 ppm. Magnetization transfer within strongly coupled spin systems during the relatively long echo time is known (25). We have verified by density matrix simulations that approximately symmetrical 13C satellite peaks are obtained for both [4-13C]Glu and [4-13C]Gln at a higher field strength such as 28T. Time-course 1H spectra from the healthy volunteer during [U-13C6]glucose infusion are displayed in Figure 7. The spectrum acquired before the start of infusion is plotted on top and labeled as “Baseline.” Spectra acquired during the infusion are plotted sequentially below the baseline spectrum with their scan start time and finish
time (relative to infusion start time) labeled beside the corresponding spectrum. The stack plots of spectra in Figure 7 show that the Glu peak at 2.35 ppm ([4-12C]Glu) became smaller as infusion progressed. At the end of the experiment, that is, 76–82 min after infusion started, the [4-12C]Glu peak was less than half of the original peak in the baseline spectrum. The [4-12C]Gln peak at 2.45 ppm also became smaller during infusion. From numerical simulations, we know that the major peak of [4-13C]Glu is located at 2.56 ppm, which overlaps the GSH peak. In Figure 7, this compound peak around 2.56 ppm, originally containing only GSH signal, became larger during the course of [U-13C6]glucose infusion. This indicates that the [4-13C]Glu multiplet at 2.56 ppm was growing as more and more 12C atoms were replaced by 13C atoms at the C4 site of Glu. DISCUSSION The J-suppression pulse used in this work is reminiscent of widely used J-editing pulses. However, the J-suppression pulse is not used to edit the signal of the metabolite of interest as in conventional two-step J-editing. Instead, the J-suppression pulse is used to suppress unwanted NAA aspartyl signals in a single shot through complex
Table 1 Metabolite Ratios in the GM Dominant Medial Prefrontal Cortex and WM Dominant Right Frontal Cortex of Eight Healthy Volunteers Medial prefrontal cortex
Glu Gln GSH NAA NAAG tCr tCho
Metaboliteratios (/[tCr]) 1.1760.07 0.2560.03 0.2160.02 1.5260.14 0.1260.04 1 0.3060.03
CRLB (%) 0.860.1 4.860.7 4.860.5 0.560.1 3.061.6 0.560.1 0.660.1
Right frontal cortex Metaboliteratios (/[tCr]) 1.0660.09 0.2060.04 0.2760.03 1.9060.18 0.3560.09 1 0.3560.03
CRLB (%) 1.260.1 8.861.7 5.460.9 0.560.1 1.460.5 0.760.1 0.760.1
FIG. 6. Numerically simulated spectra of Glu, [4-13C]Glu, Gln, and [4-13C]Gln. The Gln/Glu ratio was set to 0.22. The spectra were broadened to a singlet width of 9 Hz.
Detection of Glu, Gln, and GSH at 7T
FIG. 7. Time-course 1H spectra from a voxel in the medial prefrontal cortex of a healthy volunteer during [U-13C6]glucose infusion. The pulse sequence was the same as the one used to scan the eight volunteers (Fig. 4). Gaussian line broadening of 3 Hz was applied to all spectra.
J-coupling interactions without affecting the signals of the metabolites of interest, that is, C4 proton resonances of Glu, Gln, and GSH. In the case of the aspartyl moiety of NAA, there is a large difference between JH2–H3 (3.9 Hz) and JH2–H3’ (9.8 Hz), as well as a large JH3–H3’ (15.6 Hz) (13). The overall action of the J-suppression pulse and the chemical shift effects effectively reduces contribution from the NAA aspartyl CH2 multiplet
around 2.49 ppm. The chemical shift induced spatial inhomogeneity of J-coupling effects is mainly determined by the bandwidth of the two refocusing pulses (9–11). A different bandwidth for the two refocusing pulses may lead to a different optimal flip angle for the J-suppression pulse. Compared with the technique used by Choi et al (8), our approach suppressed the NAA multiplet signals at 2.49 ppm much more thoroughly (Fig. 2). The NAA phantom experiments confirmed both the effectiveness of our method in suppressing the NAA multiplet signals at 2.49 ppm, as well as the accuracy of our density matrix simulation program. Because our approach does not solely depends on the spatial offsets created by chemical shift differences, we can afford to use refocusing pulses with a higher bandwidth of 2.0 kHz compared with 1.4 kHz in Choi et al (8). As such, the overall chemical shift displacement is reduced by 30% for the same voxel size. Our density matrix simulation programs used 3D localization with 201 201 201 spatial points to compute basis sets for all experiments except the [U-13C6]glucose infusion experiment. On a 2.6 GHz laptop computer, our simulation program with 3D localization took 60 min to compute a five-spin metabolite such as Glu. For the [U-13C6]glucose infusion experiment, [4-13C]Glu and [4-13C]Gln are six spin systems (long-range proton Jcouplings to Glu and Gln C5 are inconsequential here and omitted) and other 13C isotopomer spin systems such as [3,4-13C2]Glu have seven or more spins. Simulations with 3D localization generally take a long time. Because the spatial inhomogeneity of J-coupling effects is mainly caused by the limited bandwidth of the two refocusing pulses, an ideal excitation pulse without localization combined with experimental refocusing pulses with 2D localization can yield simulation results very similar to those with 3D localization. Therefore, we used an ideal excitation pulse and experimental refocusing pulses with 201 201 spatial points to compute the spectra for the [U-13C6]glucose infusion experiment. On the laptop computer, it took 2 min to compute the Glu or Gln spectrum and 20 min to compute the [4-13C]Glu or [4-13C]Gln proton spectrum. Quantification results for Glu, Gln, and GSH in the medial prefrontal cortex obtained in this study, that is, Glu/tCr ¼ 1.17 6 0.07, Gln/tCr ¼ 0.25 6 0.03, and GSH/ tCr ¼ 0.21 6 0.02, agree well with the values reported in Choi et al (8), that is, Glu/tCr ¼ 1.16 6 0.11, Gln/ tCr ¼ 0.31 6 0.05, and GSH/tCr ¼ 0.23 6 0.07, which were measured under similar experimental conditions (B0 ¼ 7T, TR ¼ 2.5 s, TE ¼ 100 ms). In their study (8), only one voxel in the medial prefrontal cortex of each volunteer was measured. In our study, two voxel locations in each volunteer—one in the medial prefrontal cortex and the other in the right frontal cortex—were examined so we could compare metabolite concentration values in frontal lobe GM and WM. The spectral pattern of the time-course 1H spectra acquired during [U-13C6]glucose infusion matches our numerical simulations very well. As infusion progressed, more C4 12C atoms in Glu and Gln were replaced by 13C. As a result, the [4-12C]Glu (2.35 ppm) and [4-12C]Gln (2.45 ppm) peaks became smaller and the [4-13C]Glu (2.56 ppm)
signals grew larger. Although a detailed and quantitative analysis of the time-course data requires arterial input function and is beyond the scope of this work, this experiment shows that it is feasible to use our proposed pulse sequence to measure dynamic 13C incorporation into Glu and Gln during intravenous infusion of 13C labeled glucose or other substrates (e.g., the glia-specific substrate [2-13C]acetate). For future work, more [U-13C6]glucose and [2-13C]acetate infusion studies will be conducted on healthy volunteers and patients, and more sophisticated postprocessing algorithms will be used to quantify isotopic steady state or dynamic time-course data. CONCLUSIONS Numerical simulations, phantom experiments, and an in vivo study of eight healthy volunteers demonstrated that the TE-optimized PRESS sequence with its J-suppression pulse minimized the NAA aspartyl multiplet signals at 2.49 ppm while retaining excellent spectral resolution and peak amplitude for Glu, Gln, and GSH. In the medial prefrontal cortex, the Glu, Gln, and GSH ratios to tCr were found to be 1.17 6 0.07, 0.25 6 0.03, and 0.21 6 0.02, respectively. In the right frontal cortex, the corresponding values were 1.06 6 0.09, 0.20 6 0.04, and 0.27 6 0.03. The Glu and Gln ratios to tNAA were found to be significantly higher in the medial prefrontal cortex than in the right frontal cortex. As such, we determined that Glu and Gln concentrations are significantly higher in frontal lobe GM than in frontal lobe WM. Numerical simulations and a preliminary 13C-labeled glucose infusion study demonstrated the feasibility of using the proposed pulse sequence to measure dynamic Glu and Gln labeling processes during infusion of 13C labeled glucose. Conventional proton-observed carbon-edited experiments require an additional carbon channel (26,27). Previously, extraction of 13C labeling of metabolites using protononly MRS was hampered by spectral overlap (28) except at very high magnetic field strength (29). The proposed 7T pulse sequence allows sufficient spectral resolution for detecting 13C labeling of glutamate and glutamine in the human brain in vivo. ACKNOWLEDGMENTS This work was supported by the intramural programs of the NIH and NIMH. REFERENCES 1. Theberge J, Al-Semaan Y, Williamson PC, Menon RS, Neufeld RWJ, Rajakumar N, Schaefer B, Densmore M, Drost DJ. Glutamate and glutamine in the anterior cingulate and thalamus of medicated patients with chronic schizophrenia and healthy comparison subjects measured with 4.0-T proton MRS. Am J Psychiatry 2003;160:2231–2233. 2. Marsman A, van den Heuvel MP, Klomp DWJ, Kahn RS, Luijten PR, Pol HEH. Glutamate in schizophrenia: a focused review and meta-analysis of H-1-MRS studies. Schizophrenia Bull 2013;39: 120–129. 3. Auer DP, Putz B, Kraft E, Lipinski B, Schill J, Holsboer F. Reduced glutamate in the anterior cingulate cortex in depression: an in vivo proton magnetic resonance spectroscopy study. Biol Psychiatry 2000; 47:305–313. 4. Hasler G, van der Veen JW, Tumonis T, Meyers N, Shen J, Drevets WC. Reduced prefrontal glutamate/glutamine and gamma-aminobutyric acid levels in major depression determined using proton magnetic resonance spectroscopy. Arch Gen Psychiatry 2007;64:193–200.
An et al. 5. Pfund Z, Chugani DC, Juhasz C, Muzik O, Chugani HT, Wilds IB, Seraji-Bozorgzad N, Moore GJ. Evidence for coupling between glucose metabolism and glutamate cycling using FDG PET and H-1 magnetic resonance spectroscopy in patients with epilepsy. J Cereb Blood Flow Metab 2000;20:871–878. 6. Zimmermann C, Winnefeld K, Streck S, Roskos M, Haberl RL. Antioxidant status in acute stroke patients and patients at stroke risk. Eur Neurol 2004;51:157–161. 7. Shukla VKS, Jensen GE, Clausen J. Erythrocyte glutathione peroxidase deficiency in multiple-sclerosis. Acta Neurol Scand 1977;56:542–550. 8. Choi CH, Dimitrov IE, Douglas D, Patel A, Kaiser LG, Amezcua CA, Maher EA. Improvement of resolution for brain coupled metabolites by optimized H-1 MRS at 7 T. NMR Biomed 2010;23:1044–1052. 9. Slotboom J, Mehlkopf AF, Bovee WMMJ. The effects of frequencyselective Rf pulses on J-coupled spin-1/2 systems. J Magn Reson A 1994;108:38–50. 10. Yablonskiy DA, Neil JJ, Raichle ME, Ackerman JJH. Homonuclear J coupling effects in volume localized NMR spectroscopy: pitfalls and solutions. Magn Reson Med 1998;39:169–178. 11. Maudsley AA, Govindaraju V, Young K, Aygula ZK, Pattany PM, Soher BJ, Matson GB. Numerical simulation of PRESS localized MR spectroscopy. J Magn Reson 2005;173:54–63. 12. Smith SA, Levante TO, Meier BH, Ernst RR. Computer-simulations in magnetic-resonance - an object-oriented programming approach. J Magn Reson A 1994;106:75–105. 13. Govindaraju V, Young K, Maudsley AA. Proton NMR chemical shifts and coupling constants for brain metabolites. NMR Biomed 2000;13: 129–153. 14. An L, Li S, Wood ET, Reich DS, Shen J. N-acetyl-aspartyl-glutamate detection in the human brain at 7 Tesla by echo time optimization and improved Wiener filtering. Magn Reson Med 2014;72:903–912. 15. Murdoch JB, Lent AH, Kritzer MR. Computer-optimized narrow-band pulses for multislice imaging. J Magn Reson 1987;74:226–263. 16. Cady EB, Dsouza PC, Penrice J, Lorek A. The estimation of local brain temperature by in-vivo H-1 magnetic-resonance spectroscopy. Magn Reson Med 1995;33:862–867. 17. An L, van der Veen JW, Li SZ, Thomasson DM, Shen J. Combination of multichannel single-voxel MRS signals using generalized least squares. J Magn Reson Imaging 2013;37:1445–1450. 18. Klose U. In vivo proton spectroscopy in presence of eddy currents. Magn Reson Med 1990;14:26–30. 19. Cavassila S, Deval S, Huegen C, van Ormondt D, Graveron-Demilly D. Cramer-Rao bound expressions for parametric estimation of overlapping peaks: influence of prior knowledge. J Magn Reson 2000;143:311–320. 20. de Graaf RA. In vivo NMR spectroscopy - principles and techniques. West Sussex, England: John Wiley and Sons Ltd; 2007. 21. Kreis R, Ernst T, Ross BD. Absolute quantitation of water and metabolites in the human brain. 2. Metabolite concentrations. J Magn Reson B 1993;102:9–19. 22. Hurd R, Sailasuta N, Srinivasan R, Vigneron DB, Pelletier D, Nelson SJ. Measurement of brain glutamate using TE-averaged PRESS at 3T. Magn Reson Med 2004;51:435–440. 23. Perry TL, Hansen S, Berry K, Mok C, Lesk D. Free amino acids and related compounds in biopsies of human brain. J Neurochem 1971; 18:521–528. 24. Petroff OAC, Pleban LA, Spencer DD. Symbiosis between in-vivo and in-vitro NMR-spectroscopy - the Creatine, N-acetylaspartate, glutamate, and gaba content of the epileptic human brain. Magn Reson Imaging 1995;13:1197–1211. 25. Thomas MA, Kumar A. Two-dimensional spin-echo multiple-quantum transitions in strongly coupled spin systems - calculation of spectra. J Magn Reson 1984;56:479–509. 26. Pan JW, Stein DT, Telang F, et al. Spectroscopic imaging of glutamate C4 turnover in human brain. Magn Reson Med 2000;44:673–679. 27. de Graaf RA, Rothman DL, Behar KL. State of the art direct C-13 and indirect H-1-[C-13] NMR spectroscopy in vivo. A practical guide. NMR Biomed 2011;24:958–972. 28. Boumezbeur F, Besret L, Valette J, Vaufrey F, Henry PG, Slavov V, Giacomini E, Hantraye P, Bloch G, Lebon V. NMR measurement of brain oxidative metabolism in monkeys using C-13-labeled glucose without a C-13 radiofrequency channel. Magn Reson Med 2004;52:33–40. 29. Xu S, Yang J, Shen J. Measuring N-acetylaspartate synthesis in vivo using proton magnetic resonance spectroscopy. J Neurosci Methods 2008;172:8–12.