Detection of 13C labeling of glutamate and glutamine in human brain by proton magnetic resonance spectroscopy

A proton magnetic resonance spectroscopy (MRS) technique was used to measure 13C enrichments of glutamate and glutamine in a 3.5 × 1.8 × 2 cm3 voxel placed in the dorsal anterior cingulate cortex of five healthy participants after oral administration of [U-13C]glucose. Strong pseudo singlets of glutamate and glutamine were induced to enhance the signal strength of glutamate and glutamine. This study demonstrated that 13C labeling of glutamate and glutamine can be measured with the high sensitivity and spatial resolution of 1H MRS using a proton-only MRS technique with standard commercial hardware. Furthermore, it is feasible to measure 13C labeling of glutamate and glutamine in limbic structures, which play major roles in behavioral and emotional responses and whose abnormalities are involved in many neuropsychiatric disorders.

www.nature.com/scientificreports/ 13 C labeling of both Glu and Gln from an area in the limbic system. Specifically, using oral administration of 13 C-labeled glucose, we will show that 13 C labeling of Glu and Gln can be measured with high precision from the dorsal anterior cingulate cortex (dACC), a limbic region involved in cognition and motor control but is beyond the reach of conventional 13 C MRS that relies on surface coils. It is hoped that the demonstration of measuring 13 C-labeling of Glu and Gln with the high sensitivity and spatial resolution of proton MRS using commercial scanners and RF coils will generate interest in further improving MRS technology and greatly facilitate the adoption of 13 C MRS strategies for probing energy metabolism and glutamatergic neurotransmission in clinical research. Figure 1 shows the calculated spectra of Glu, 13 C satellites of Glu ([ 13 C]Glu), Gln, 13 C satellites of Gln ([ 13 C] Gln), Asp, 13 C satellites of Asp ([ 13 C]Asp), GABA, and 13 C satellites of GABA ([ 13 C]GABA). As shown in Fig. 1, the spectra of Glu and Gln at TE = 56 ms are dominated by their respective H4 and H2 pseudo singlets. The 13 C satellite signals of Glu H4 in the proton channel, [ 13 C]Glu, are resulted from the large one-bond scalar coupling between the H4 pseudo singlet and the 13 C label at C4. As both carbons of the acetyl CoA are 13 C-labeled after administration of uniformly labeled glucose 15 , [4,[5][6][7][8][9][10][11][12][13] C]Glu was used as the starting point for spectral fitting of Glu 13 C satellites in this study. Similarly, [4,[5][6][7][8][9][10][11][12][13] C]Gln and [3,[4][5][6][7][8][9][10][11][12][13] C]Asp were chosen as the starting points for fitting the 13 C satellites of Gln H4 and Asp H3, respectively. The [4,[5][6][7][8][9][10][11][12][13] C]Glu, [4,[5][6][7][8][9][10][11][12][13] C]Gln and [3,4-13 C] Asp spectra in Fig. 1 were generated using one-bound and long-range 1 H-13 C coupling constants reported in the literature 16 except for the long-range 1 H-13 C coupling constants involving the carboxylic carbons which, to the best of our knowledge, were not available. Instead, the corresponding long-range 1 H-13 C coupling constants involving C4 (C3 for Asp) were used as their substitutes. The actual values of the long-range 1 H-13 C couplings used in the spectral model are not important as the spectral fitting program fits the in vivo spectra by adjusting the lineshape and linewidth of the 13 C satellites to account for changes in B 0 inhomogeneity and additional 1 H-13 C scalar couplings. Furthermore, although 31 C label is transferred from glutamate C4 to GABA C2 during GABA formation, [2-13 C]GABA was omitted in spectral fitting because the spectra of [2-13 C]GABA and GABA are very weak compared to the resonance signals of Glu H4, Gln H4, and their 13 C satellites, as shown in Fig. 1. Note that the 13 C satellite spectra of Glu H4 and Gln H4 in Fig. 1 are highly asymmetrical and dominated by a single downfield satellite peak.

Results
Time-course 1 H spectra from the dACC of the five participants are displayed in Fig. 2. The spectra are highly consistent with the spectral patterns by the numerical simulations. The Glu H4 peak at 2.34 ppm dropped dramatically after oral administration of [U-13 C]glucose and, correspondingly, the peak at 2.56 ppm significantly increased due to the rise of the downfield 13 C satellite signals of Glu H4. Figure 3 displays the time-course spectra of a representative participant and corresponding fitted spectra of Glu, Gln, and their 13 C satellites. In the fitted Gln time-course spectra, the drop in peak amplitude of Gln H4 after oral administration of [U-13 C]glucose can be clearly seen. Meanwhile, the rise of the peak for Gln 13 C satellites (vertically scaled up by a factor of 4) after oral administration of [U-13 C]glucose is apparent. The spectra and corresponding fits for the pre-13 C MRS scan and the last post-13 C scan of the participant are displayed in Fig. 4 with good fitting results. The spline baseline obtained by fitting the pre-13 C spectrum was labeled as baseline 1 . To account for the participant repositioning www.nature.com/scientificreports/ (see "Methods" section), a second much weaker baseline (baseline 2 ) was used, which was determined when fitting the post-13 C spectrum. The total baseline for the post-13 C spectrum was the sum of baseline 1 and baseline 2. Figure 5 displays the plots of 13 C enrichments of Glu C4 and Gln C4 vs. time after oral administration of [U-13 C]glucose for all five participants. The behavior of the time course of 13 C-labled Glu and Gln shown in Fig. 5 is consistent with them approaching maximum 13 C labeling as expected from the time scale of cerebral Glu and Gln turnover 1 . Similar behavior was also observed in our previous direct 13 C MRS experiments in the carboxylic/amide spectral region following oral administration of [U-13 C]glucose 17 .
Metabolite ratios (/[tCr]) in the dACC of the five participants quantified from the pre-13 C spectrum of each participant are given in Table 1. The results are highly consistent with our earlier 1 H-only MRS study of the same brain region using the same pulse sequence 13 . Using the 12.6 mL (3.5 × 1.8 × 2 cm 3 ) voxel size and 10 min scan time, the CRLB values for Glu and Gln were found to be 1.6 ± 0.2% for Glu and 3.2 ± 0.4% for Gln, indicating excellent precision. 13 C enrichments of Glu C4 and Gln C4 for the five participants computed from the last two post-13 C spectra of each participant, acquired at 113 ± 9 min after oral administration of [U-13 C]glucose, were found to be 64 ± 5% for Glu and 40 ± 10% for Gln.
Supplementary Figure S1 online shows the time-course difference spectra of the participant in Fig. 3, which were obtained by subtracting the pre-13 C spectrum from the post-13 C spectra. Due to repositioning of the  Figure S2 online displays the reprocessed time-course spectra of a participant in the previous study 11,12 , in which 1 H spectra with TE = 106 ms were acquired from the prefrontal cortex of eight healthy participants after intravenous infusion of [U-13 C 6 ]glucose. In the fitted Gln time-course spectra, the Gln H4 peak is relatively weak. As a result, the gradual drop in peak amplitude of Gln H4 and the gradual rise in peak amplitude of Gln 13 C satellites (vertically scaled up by a factor of 4) after intravenous infusion of [U-13 C 6 ]glucose are also very weak, which leads to relatively large errors in the computed 13 C enrichment values of Gln C4. Supplementary Figure S3 online displays the plots of 13 C enrichments of Glu and Gln vs. time after intravenous infusion of [U-13 C]glucose for all eight healthy participants in the previous study. The 13 C enrichment curves of Glu are relatively smooth and consistent with each other, indicating reliable measurement of 13 C labeling of Glu. However, the 13 C enrichment curves of Gln have large variations within each curve and between different participants. The end-point 13 C enrichments of Glu and Gln for the eight participants computed from the last two post-13 C spectra of each participant were found to be 51 ± 9% for Glu and 43 ± 16% for Gln. The last two post-13 C spectra for the eight participants were acquired at 77 ± 9 min after intravenous infusion of [U-13 C] glucose using the TE = 106 ms pulse sequence 11 . In comparison, a more reliable measurement of 13 C-labeling of Gln was achieved using TE = 56 ms as evidenced by the higher Gln peaks in Fig. 3 and less scattered Gln 13 C enrichment curves in Fig. 5.   This method used both the pre-13 C and post-13 C MRS data to compute 13 C enrichments of Glu and Gln. Because the participants exited the scanner after the pre-13 C scan for oral administration of glucose and reentered the scanner for acquisition of the post-13 C spectra, the pre-13 C and post-13 C spectra generally had small differences in metabolite linewidth and lineshape, as well as in the spectral baseline. In the previous work (TE = 106 ms) 12 , the participants stayed in the magnet during the entire scan. Frequency shift, zero-order phase, and line-broadening of the pre-13 C spectrum were adjusted to fit each post-13 C spectrum before generating a difference spectrum. The 13 C-labeled Glu and Gln concentrations were obtained by fitting the difference spectrum. In the current study, the pre-13 C and post-13 C spectra generally do not match very well due to repositioning of the participants after oral glucose administration outside the magnet. As shown by Supplementary Fig. S1, subtraction of the pre-13 C spectrum from the post-13 C spectra caused significant subtraction errors. A novel post-processing method was developed in this work. The metabolite concentration ratios (/[tCr] + 3[tCho]) of acetate (Ace), NAA, N-acetylaspartylglutamate (NAAG), GABA, Glu, Gln, glutathione (GSH), Asp, total creatine (tCr), total choline (tCho), taurine (Tau), myo-inositol (mI), and scyllo-inositol (sI) obtained by fitting the pre- 13 C spectrum were used as constraints when fitting the post-13 C spectra. Meanwhile, the spline baseline obtained from the pre-13 C spectrum was also used in fitting the post-13 C spectrum, along with an additional much weaker  Figure 4. In vivo spectra and corresponding fitted spectra for the pre-13 C MRS scan and the last post-13 C scan of the participant in Fig. 3. Baseline 1 is the baseline in the pre-13 C spectrum, which is a spline baseline with 13 knots. The baseline in the post-13 C spectrum is baseline 1 + baseline 2 , in which baseline 2 is a much weaker spline baseline with 8 knots.
Scientific Reports | (2022) 12:8729 | https://doi.org/10.1038/s41598-022-12654-y www.nature.com/scientificreports/ baseline. This approach of using the prior information from the pre-13 C spectrum in the fitting of each post-13 C spectrum avoids spectrum subtraction and hence the resultant subtraction errors. The turnover of NAA and GSH in brain is known to be much slower than that of Glu and Gln. Therefore, 13 C labeling of NAA and GSH, in addition to GABA, was omitted from our spectral model. In this study, 13 C labeling of Glu and Gln was extracted using information from both the decrease in the pseudo singlet signals and the increase in their 13 C satellite signals. The relationship between the parent Glu H4 signal and its 13 C satellites is established by full density matrix simulations. In Fig. 1, the 13 C satellite signals are much weaker than their parent signals for the same concentration. Therefore, the contribution from the 13 C satellites to spectral fitting results is less than the corresponding parent signals. Our current approach simplifies the spectral fitting process by omitting the 13 C labeling of GABA and adjusting the lineshape and linewidth to account for changes in B 0 inhomogeneity and additional 13 C labels. An alternative and more sophisticated approach is to use the actual 1 H- 13 C couplings to compute the lineshape and linewidth of all signals since the outcome of the entire process of 13 C labeling of Glu and Gln is determined by very few kinetic parameters such as the tricarboxylic acid cycle rate and the Glu-Gln cycle rate. This more sophisticated approach will be developed to improve the current technique. The high sensitivity of the proton MRS method employed in this study also suggests that it may be possible to use the current strategy to measure the initial rates of 13 C incorporation into Glu and Gln during intravenous infusion of 13  The data acquired from the previous study 11,12 were reprocessed using the new post-processing method. However, the 13 C enrichment value of Gln was still highly scattered due to the small Gln resonance signals at TE = 106 ms. Using the current sequence (TE = 56 ms), the Gln H4 peak is at least 61% higher than that of the previous sequence (TE = 106 ms) 13 . Therefore, a more precise measurement of the end-point 13 C enrichment of Gln was achieved in this study, which is evidenced by the more consistent 13 C enrichment curves of Gln.
Although the current study used a 7 Tesla scanner to resolve Glu and Gln H4 signals in the 1 H MRS spectra, spectral resolution of Glu and Gln H4 signals at 3 Tesla is also achievable 18 . In principle, it is possible to use a  www.nature.com/scientificreports/ similar strategy to measure 13 C labeling of Glu C4 and possibly Gln C4 using 1 H MRS on the more prevalent 3 Tesla scanners. Research along this direction is currently in progress in our laboratory. Previous studies have demonstrated quantification of the Glu-Gln neurotransmitter cycling flux between neurons and astroglia using direct 13 C MRS by measuring 13 C labeling of Glu and Gln at isotopic steady state following administration of 13 C-labeled acetate or by measuring dynamic turnover of Glu and Gln following administration of 13 C-labeled glucose, lactate, or β-hydroxybutyrate 1,19 . Therefore, it is possible to use protononly MRS techniques to quantify the Glu-Gln neurotransmitter cycling flux with much higher spatial resolution and from brain regions inaccessible to surface coils, e.g., from limbic structures which play a major role in many neuropsychiatric disorders. As shown by Fig. 5, the transient isotopic steady state was approached but not attained in this study. Future studies should measure the initial rates, or the entire time course, or delay the time window of sampling the turnover curves after oral glucose administration to capture the transient isotopic steady state.
In summary, a proton-only MRS technique that induces intense Glu and Gln H4 singlets at TE = 56 ms was used to measure 13 C enrichments of Glu and Gln in the dACC of five healthy participants after oral administration of [U-13 C]glucose. A novel post-processing method was developed, in which the metabolite ratios and spline baseline obtained from fitting the pre-13 C spectrum were used in the fitting of the post-13 C spectra to compute the 13 C enrichments of Glu C4 and Gln C4. At 113 ± 9 min after oral administration of [U-13 C]glucose, the endpoint 13 C enrichment of Glu C4 was found to be 64 ± 5% (n = 5) and that of Gln C4 was found to be 40 ± 10% (n = 5). This technique offers a novel option to study Glu neurotransmission in the human brain with the high sensitivity and spatial resolution of 1 H MRS using standard commercial equipment. 13 C labeling in brain regions inaccessible to surface coils can also be investigated using proton MRS.

Methods
Five healthy participants (two females, three males; age = 34 ± 12 years) were recruited for the study. Written informed consent was obtained from the participants before the study following the procedures approved by the Institutional Review Board (IRB) of the National Institute of Mental Health (NIMH; NCT00109174). The 13 C enriched glucose solution was prepared by the National Institutes of Health (NIH) Clinical Center Pharmacy Department. All experimental protocols and methods were performed in accordance with the guidelines and regulations of NIH MRI Research Facility. Experiments were carried out on a Siemens Magnetom 7 Tesla scanner equipped with a 32-channel receiver head coil. The participants underwent overnight fasting before the MRS study.
In each scan session, the participant was first scanned to acquire the pre-13 C MRS data. T 1 -weighted magnetization prepared rapid gradient echo (MPRAGE) images were acquired with repetition time (TR) = 3 s, TE = 3.9 ms, matrix = 256 × 256 × 256, and resolution = 1 × 1 × 1 mm 3 . Based on the MPRAGE images, the MRS voxel with a size of 3.5 × 1.8 × 2 cm 3 was placed in the dACC of the participant. The first-and second-order B 0 shimming coefficients were adjusted, achieving water linewidths of 11.1 ± 0.4 Hz. The pre-13 C MRS scan was subsequently performed using a previously described pulse sequence 13 . The main component of the pulse sequence was a point resolved spectroscopy sequence (PRESS) with a 10 ms truncated 180° Gaussian pulse applied at 2.12 ppm. The pulse sequence parameters were: TR = 2.2 s, TE = 56 ms, spectral width = 4000 Hz, number of data points = 1024, number of averages = 264, number of unsuppressed water signal averages = 2, and total scan time = 10 min.
After the pre-13 C MRS scan was finished, the participant exited the scanner and was orally administered 20% w/w 99% enriched [U-13 C]glucose solution at a dosage of 0.75 g [U-13 C]glucose per kg of body weight following procedures described in our previous study of carbonic anhydrase-catalyzed 13 C magnetization transfer and references therein 17 . After a rest period, the participant reentered the scanner. The MPRAGE images were repeated, based on which the MRS voxel was placed at the same location and with the same size as in the pre-13 C MRS scan. Post-13 C MRS scans were repeatedly performed, each lasted 5 min (number of averages = 132). B 0 shimming coefficients were adjusted before each MRS scan.
The pre-13 C MRS data were processed first and the process was similar to that of the previous work 13 . Briefly, the raw free induction decay (FID) data were reconstructed into the pre-13 C spectrum after going through the necessary steps that include multi-channel data combination 20 , eddy current correction 21 , Bloch-Siegert phase shift correction 22 , frequency drift correction 23 , and Fourier transform. The reconstructed pre-13 C spectrum was fitted in the range of 1.8-3.4 ppm by linear combination of numerically computed basis spectra of Ace, NAA, NAAG, GABA, Glu, Gln, GSH, Asp, tCr, tCho, Tau, mI, and sI, as well as a spline baseline with 13 knots. Chemical shifts and coupling constants were obtained from the literature for GABA 24 , GSH 14 , and the rest of the metabolites 25 . The fitting program was developed and improved in-house and was based on the Levenberg-Marquardt least square minimization algorithm. After the metabolite concentrations in arbitrary unit were obtained from the fitting, we computed the metabolite ratios, which were defined as the concentration of a metabolite divided by the sum of concentration of tCr and three times the concentration of tCho. The combined concentration, [tCr] + 3[tCho], weighs approximately equally the intensities of the tCr and tCho singlet peaks and is less prone to errors than using either [tCr] or [tCho] alone.
The post-13 C spectra acquired after oral administration of [U-13 C]glucose were reconstructed in the same way as the pre-13 C spectrum. For the subsequent fitting process, additional basis spectra of 13 C-labeled Glu, Gln, and Asp were simulated. Additional 13 C chemical shifts and 1 H-13 C and 13 C-13 C coupling constants were obtained from Ref. 16 with the exception of long-range 1 H-13 C couplings involving the carboxylic carbons. As shown in Fig. 1, contribution from GABA is minimal. Hence, 13 C-labeled GABA was not included in spectral fitting. When fitting each post-13 C spectrum, the metabolites ratios (/[tCr] + 3[tCho]) of Ace, NAA, NAAG, GABA, GSH, Asp, tCr, tCho, Tau, mI, and sI were fixed to the pre-13 C values. However, the linewidths and lineshape of the metabolites in a post-13 C spectrum were allowed to be different from those of the pre- 13  www.nature.com/scientificreports/ repositioning, B 0 shimming, and additional 1 H-13 C couplings caused changes in the linewidths and lineshape. The sum of metabolite ratios of Glu and 13 C-labeled Glu was constrained to be the same as the metabolite ratio of Glu obtained from the pre-13 C spectrum. The same constraints were also applied to Gln and Asp. Because the spectral baseline in the post-13 C spectrum was expected to be slightly different from the spectral baseline in the pre-13 C spectrum due to participant repositioning and B 0 shimming, the spectral baseline in each of the post-13 C spectrum was approximated by the sum of the spline baseline in the pre-13 C spectrum and another much weaker spline baseline with 8 knots.
After the metabolite concentrations were obtained by fitting each post-13 C spectrum, the 13 C enrichment of Glu C4 for the post-13 C spectrum was computed as the ratio of the concentration of its 13 C satellites to the total concentration of Glu.
For comparison purposes, the MRS data from the previous study 11,12 were reprocessed using the new postprocessing method as described above. The MRS data were acquired using the pulse sequence 11 with TE = 106 ms from the prefrontal cortex of eight healthy participants (5 females and 3 males, age = 37 ± 8 years). The pulse sequence used the following parameters: voxel size = 2 × 2 × 2 cm 3 , TR = 2.5 s, TE = 106 ms, J-suppression pulse frequency = 4.38 ppm, J-suppression pulse flip angle = 90°, spectral width = 4000 Hz, number of data points = 2048, number of averages = 128, and total scan time = 5.5 min for each individual spectrum.