In Vivo Measurement of Carbon-13 Labeling of Glutamate and Glutamine in Human Brain Using Proton Magnetic Resonance Spectroscopy

A single-step spectral editing 1 H magnetic resonance spectroscopy (MRS) technique was used to measure fractional enrichments of glutamate and glutamine in the dorsal anterior cingulate cortex of ve healthy volunteers after oral administration of [U- 13 C]glucose. Strong pseudo singlets of glutamate and glutamine were induced at an echo time of 56 ms using an always-on editing pulse at 2.12 ppm. At 113 ± 9 minutes after oral administration of [U- 13 C]glucose, fractional enrichment of glutamate was found to be 64 ± 5% with 1.7% within-subject coecient of variation (CV) and fractional enrichment of glutamine was found to be 40 ± 10% with 11% within-subject CV. This study demonstrated that 13 C labeling of both glutamate and glutamine can be measured with the high sensitivity and spatial resolution of proton MRS using a proton-only MRS technique with standard commercial hardware. Furthermore, it is now feasible to measure 13 C 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.


Introduction
Noninvasive in vivo detection of 13 C labeling of glutamate (Glu) and glutamine (Gln) is a powerful tool for investigating Glu and Gln metabolism and neurotransmission in the brain 1 . Two types of 13 C magnetic resonance spectroscopy (MRS) techniques have been widely used: direct 13

C MRS and indirect 1 H-[ 13 C]
MRS. Due to the much lower sensitivity of 13 C nuclei, direct 13 C MRS generally requires surface coils and very large tissue volume to achieve adequate signal-to-noise ratio (SNR). Hence, only the neocortex is accessible to direct 13 C MRS experiments. The indirect 1 H-[ 13 C] MRS techniques make use of the 1 H-13 C coupling and difference spectroscopy to detect signals from protons bound to 13 C 2-9 . Because of the larger gyromagnetic ratio of 1 H compared to 13 C, these indirect 13 C detection techniques have higher sensitivity compared to the direct 13 C detection techniques. To perform either direct 13 C or indirect 1 H-[ 13 C] MRS, broadband magnetic resonance imaging (MRI) machines equipped with heteronuclear capabilities are required. The RF coil assembly needs to be high-e ciency non-volume 13 C and 1 H coils, which is a non-standard device and not commercially available, to enhance sensitivity and make heteronuclear nuclear Overhauser enhancement/decoupling feasible. Because of these technical barriers, 13 C MRS of human brain has been largely con ned to only a few research groups with very limited clinical applications.
Attempts have been made to circumvent the hardware limitation of 13 C MRS by measuring the changes in short echo time (TE) 1 H MRS spectra caused by incorporation of 13 C labels into brain amino acids 10 .
However, it has been di cult to reliably separate Glu and Gln in the crowded 1 H MRS spectra. With the incorporation of 13 C labels, the short TE 1 H MRS spectra become even more complex. Due to these di culties, quanti cation of 13 C labeling of both Glu and Gln using 1 H MRS in the human brain has not been reported.
Recently, our laboratory developed a 7 T single-step spectral editing method which can reliably measure Glu and Gln at TE = 56 ms 11 . By placing an editing pulse at 2.12 ppm, which targets the H3 protons of Glu and Gln, the H4 protons of Glu and Gln form intense pseudo singlets at TE = 56 ms. The editing pulse here is used to alter the J-evolution of the strongly coupled spins such that Glu H4 and Gln H4 form intense pseudo singlet peaks at a relatively short TE of 56 ms. Without using an editing pulse, Glu H4 and Gln H4 form pseudo singlet peaks at a much longer TE of 100-110 ms 12 . We found that the Gln H4 pseudo singlet induced by the editing pulse at TE = 56 ms is at least 61% more intense than the Gln H4 pseudo singlet formed naturally at TE = 106 ms 11 .
In this work, we demonstrate the feasibility of measuring 13 C fractional enrichments of Glu and Gln using the single-step spectral editing 1 H MRS technique (TE = 56 ms) with a commercial proton-only head coil at 7 T. Since this is a proton-only technique, no broadband scanner or any custom-made hardware is necessary. In addition, because of the much higher sensitivity of proton MRS, we will also demonstrate, for the rst time, detection of 13 C labeling of Glu and Gln from an area in the limbic system. Speci cally, 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 structure 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 greatly facilitate the adoption of 13  The basis spectrum of [2-13 C]GABA was not included in the basis set because the differences between the basis spectra of [2-13 C]GABA and GABA are vanishingly small compared to the resonance signals of Glu H4, Gln H4, and their 13 C satellites, as shown in Fig. 1. Due to magnetization transfer within the strongly coupled spin system, the 13 C satellite spectra of Glu H4 and Gln H4 at TE = 56 ms are highly asymmetrical and dominated by a single down eld peak. It can be shown using density matrix simulation that the two 13 C satellites of Glu H4 and Gln H4 at TE = 56 ms become symmetrical at much higher eld strength (data not shown).
Time-course 1 H spectra from the dACC of the ve subjects are displayed in Fig. 2. The spectra are highly consistent with the spectral patterns predicted 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 signi cantly increased due to the rise of the down eld 13 C satellite signals of Glu H4. Figure 3 displays the time-course spectra of subject 1 and corresponding tted spectra of Glu, Gln and their 13 C satellites. In the tted Gln time-course spectra, the drop in peak amplitude of Gln H4 after oral administration of [U-13 C]glucose can be clearly seen. The spectra and corresponding ts for the pre-13 C MRS scan and the last post-13 C scan of subject 1 are displayed in Fig. 4. The spectral model ts the in vivo spectra very well. The spline baseline obtained by tting the pre-13 C spectrum is labelled as baseline 1 . The total baseline for the post-13 C spectrum is the sum of baseline 1 and a much weaker baseline 2 which was determined when tting the post-13 C spectrum.
Metabolite ratios (/[tCr]) in the dACC of the ve subjects quanti ed from the pre-13 C spectrum of each subject 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 11 . 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. Table 2 gives the fractional enrichments of Glu H4 and Gln H4 for the ve subjects computed from the last two post-13 C spectra of each subject. The last two post-13 C spectra were acquired at 113 ± 9 min after oral administration of [U-13 C]glucose and each scan lasted 5 min. The fractional enrichments were found to be 64 ± 5% with 1.7% within-subject coe cient of variation (CV) for Glu and 40 ± 10% with 11% within-subject CV for Gln.  Table 2 Fractional enrichments of Glu H4 and Gln H4 in the dACC of the ve healthy volunteers computed from the last two post-13 C spectra of each subject, which were acquired at 113 ± 9 min after oral administration of [U-13 C]glucose. For each post-13 C MRS scan, the number of averages was 132 and the total scan time was 5 min.
Fractional enrichment (%) Within-subject CV (%) Glu H4 64 ± 5 1.7 Gln H4 40 ± 10 11 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 linewidths and lineshape, as well as in the spectral baseline. In our previous work (TE = 106 ms), the frequency shift, zero th -order phase, and line-broadening of the pre-13 C spectrum were adjusted to t a post-13 C spectrum before generating the difference spectrum 15 . The 13 C-labelled Glu and Gln concentrations were obtained by tting the difference spectrum. Due to the subtraction of the pre-13 C and post-13 C spectra, the noise level in the difference spectrum was ampli ed. In the difference spectrum, the unlabeled Gln H4 peak at 2.44 ppm was very weak with its 13 C satellite signals barely discernible. Therefore, the 13  Ace, NAA, NAAG, GABA, GSH, Asp, tCr, tCho, Tau, mI, and sI obtained by tting the pre-13 C spectrum were used as constraints when tting the post-13 C spectra. Meanwhile, the spline baseline obtained from the pre-13 C spectrum was also used in tting the post-13 C spectrum, along with an additional much weaker baseline. This approach of using the prior information from the pre-13 C spectrum in the tting of each post-13 C spectrum avoids spectrum subtraction and hence the corresponding noise ampli cation. Additionally, the Gln signals acquired using the current sequence (TE = 56 ms) were at least 61% higher than the Gln signals obtained by the previous sequence (TE = 106 ms) 11 . Therefore, a much higher precision for measuring the fractional enrichment of Gln was achieved in this study, which is evidenced by its 11% within-subject CV.
The turnover of NAA and GSH in brain is known to be orders of magnitude slower than that of Glu and Gln. 13 C labeling of NAA and GSH in our experiment therefore was ignored. Furthermore, this study used [4,5-13 C]Glu to represent all 13 C isotopomers of Glu with a 13 C label at C4. This approximation is justi ed because the much greater one-bond 1 H-13 C scalar coupling ( 1 J HC = 127 Hz) splits Glu H4 into an uneven doublet (see Fig. 1) whereas 13 C labels at other positions lead to line-broadening of the satellite signals due to the much smaller long-range 1 H-13 C scalar couplings.
Our in vivo data in Fig. 5 clearly show the sizable effect of Gln isotopic dilution which leads to a smaller Gln fractional enrichment than that of Glu 16 . Gln dilution is primarily resulted from exchange between astroglial Gln and unlabeled Gln in blood across the blood-brain barrier 17 as well as oxidation of shortand medium-chain free fatty acids and branched chain amino acids preferentially in astroglia 18,19 . The existence of Gln dilution sensitizes the Gln turnover to the effect of intercompartmental Glu-Gln cycling ux 16,20 . Using the standard small pool approximation in kinetic analysis, the Glu-Gln cycling ux V cyc at isotopic steady state following administration of [U-13 C]glucose is given by Eq. (8) in Ref. 20 : where V Gln(dil.) is the Gln dilution ux, and f Glu ss and f Gln ss represent fractional enrichments of Glu H4 and Gln H4 at isotopic steady state, respectively. Using the mean literature value of V Gln(dil.) (0.18 µµολ/g/min, averaged from the reported range of 0.14-0.22 µµολ/g/min, Shen et al. 16 and references therein) and our end point fractional enrichments of Glu H4 and Gln H4, V cyc in the dACC of the healthy subjects was estimated to be 0.32 ± 0.13 µµολ/g/min (mean ± SD, n = 5). Interestingly, the estimated V cyc in the dACC of the ve subjects closely matches most of the literature V cyc values which were measured from brain regions dominated by the neocortex 16 . The high activities of Glu neurotransmission in the dACC is consistent with its purported glutamatergic role in many brain functions and neuropsychiatric disorders.
Although the current study used a 7 T scanner to resolve Glu and Gln H4 protons in the 1 H MRS spectra, spectral resolution of Glu and Gln H4 signals at 3 T is also achievable 21 . In principle, it is possible to use a similar strategy to measure 13 C labeling of Glu H4 and possibly Gln H4 using 1 H MRS on the prevalent 3 T scanners. Research along this direction is currently in progress in our laboratory.
Previous studies have demonstrated quanti cation of the Glu-Gln neurotransmitter cycling ux between neurons and astroglia using direct 13 C MRS by measuring fractional enrichments of Glu C4 and Gln C4 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 b-hydroxybutyrate 1 . Therefore, it is possible to use the 1 H-only MRS technique demonstrated in this study to quantify the Glu-Gln neurotransmitter cycling ux 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).
In summary, a recently developed single-step spectral editing technique that induces intense Glu and Gln H4 singlets at TE = 56 ms was used to measure fractional enrichments of Glu and Gln in the dACC of ve healthy volunteers after oral administration of [U-13 C]glucose. A new post-processing method was developed, in which the metabolite ratios and spline baseline obtained from tting the pre-13 C spectrum were used in the tting of the post-13 C spectra to compute the fractional enrichments of Glu H4 and Gln H4. At 113 ± 9 min after oral administration of [U-13 C]glucose, the fractional enrichment of Glu H4 was found to be 64 ± 5% with 1.7% within-subject CV and the fractional enrichment of Gln H4 was found to be in the human brain with the high sensitivity and spatial resolution of 1 H MRS using standard commercial equipment. Brain regions inaccessible to surface coils can now be investigated using the method described in this study.

Methods
Five healthy volunteers (two females, three males; age = 34 ± 12 years) were recruited for the study. In each scan session, the subject was rst 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 subject. The rst-and second-order B 0 shimming coe cients were adjusted, achieving water linewidths of 11.1 ± 0.4 Hz. The pre-13 C MRS scan was subsequently performed using the single-step spectral editing pulse sequence 11 .
The main component of the pulse sequence was a point resolved spectroscopy sequence (PRESS) with an always-on spectral editing pulse between the two 180° refocusing pulses. The editing pulse was a truncated Gaussian pulse with a duration of 10 ms, and it was applied at 2.12 ppm. The pulse sequence parameters were: TR = 2.2 s, TE = 56 ms, editing pulse frequency = 2.12 ppm, editing pulse ip angle = 180°, 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 nished, the subject exited from 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 14 and references therein. After a rest period, the subject 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 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 coe cients were adjusted before each MRS scan.
The pre-13 C MRS data were processed rst and the process was similar to that of the previous work 11 .
Brie y, 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 22 , eddy current correction 23 , Bloch-Siegert phase shift correction 24 , frequency drift correction 25 , and Fourier transform. The reconstructed pre-13 C spectrum was tted in the range of 1.8-3.4 ppm by linear combination of numerically computed basis spectra of acetate (Ace), N-acetyl-aspartate (NAA), Nacetylaspartylglutamate (NAAG), γ-aminobutyric acid (GABA), Glu, Gln, glutathione (GSH), aspartate (Asp), total creatine (tCr), total choline (tCho), taurine (Tau), myo-inositol (mI), and scyllo-inositol (sI), as well as a spline baseline with 13 knots. The tting program was developed in-house and was based on the Levenberg-Marquardt least square minimization algorithm. After the metabolite concentrations in arbitrary unit were obtained from the tting, we computed the metabolite ratios, which were de ned as the concentration of a metabolite divided by the sum of concentration of tCr and three times the concentration of tCho. of Ace, NAA, NAAG, GABA, GSH, Asp, tCr, tCho, Tau, mI, and sI were xed 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 C spectrum because subject repositioning and B 0 shimming caused changes in the linewidths and lineshape. The sum of metabolite ratios of Glu and its 13 C satellites 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 subject 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 tting each post-13 C spectrum, the 13 C fractional enrichment of Glu H4 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. Meanwhile, the fractional enrichment of Gln H4 for each post-13 C spectrum was similarly computed. The within-subject CV 26 of the fractional enrichment values were estimated from the last two post-13 C spectra for each subject, which were acquired at 113 ± 9 min after oral administration of [U-13 C]glucose. The 13 C enrichment of Glu H4 is expected to have approximately reached its maximum value at this stage 1,14 . Declarations