PMCCPMCCPMCC

Search tips
Search criteria 

Advanced

 
Logo of nihpaAbout Author manuscriptsSubmit a manuscriptHHS Public Access; Author Manuscript; Accepted for publication in peer reviewed journal;
 
Magn Reson Med. Author manuscript; available in PMC 2012 September 1.
Published in final edited form as:
PMCID: PMC3139742
NIHMSID: NIHMS264106

Measurement of glycine in the human brain in vivo by 1H-MRS at 3T: Application in brain tumors

Abstract

Glycine (Gly) is a key metabolic intermediate required for the synthesis of proteins, nucleic acids and other molecules, and its detection in cancer could therefore provide biologically relevant information about the growth of the tumor. Here we report measurement of Gly in human brain and gliomas by an optimized point-resolved spectroscopy (PRESS) sequence at 3T. Echo time dependence of the major obstacle, myo-inositol (mI) multiplet, was investigated with numerical simulations, incorporating the 3D volume localization. The simulations indicated that a subecho pair (TE1, TE2) = (60, 100) ms permits detection of both Gly and mI with optimum selectivity. In vivo validation of the optimized PRESS was conducted on the right parietal cortex of five healthy volunteers. Metabolite signals estimated from LCModel were normalized with respect to the brain water signal and the concentrations were evaluated assuming the total creatine concentration at 8 mM. The Gly concentration was estimated as 0.6±0.1 mM (mean±SD, n=5), with a mean Cramér-Rao lower bound of 9±1%. The PRESS sequence was applied to measure the Gly levels in patients with glioblastoma multiforme. Metabolite concentrations were obtained using the water signal from the tumor mass. The study revealed that a subset of human gliomas contains glycine levels elevated 1.5–8 fold relative to normal.

Keywords: Glycine, Glioblastoma, 1H-MRS, 3T, Point-resolved spectroscopy (PRESS), Human brain, Echo time optimization

INTRODUCTION

Non-invasive techniques to detect and quantify metabolic features of human tumor tissue have outstanding clinical potential in cancer imaging. Magnetic resonance spectroscopy (MRS) to detect lactate, choline and a few other abundant metabolites is commonly used to differentiate between benign and malignant lesions in the brain. However, MRS at 1.5 T is limited in scope because of the relatively low spectral resolution. For example, glycine (Gly) is difficult to quantify by 1H-MRS primarily due to its relatively low concentration and the spectral overlap with myo-inositol (mI) normally at much higher concentrations. Gly has two uncoupled spins resonating at 3.55 ppm, while the four mI resonances at 3.52 and 3.61 ppm are scalar coupled and give rise to an abundant multiplet in the proximity of the Gly resonance (1). The J evolution of mI resonances during the echo time can be exploited for differentiation between the Gly and mI signals. Several studies have been reported for measurement of Gly in the healthy human brain in recent years. Approaches at intermediate field strengths includes 2D J-resolved MRS at 3T (2), TE-averaging at 4T (3), and triple refocusing at 3T (4). These methods all require modifications to the vendor-supplied standard sequences and/or post-data processing. Given that the 3T field strength is increasingly available for in vivo spectroscopy, direct measurement of Gly using the standard sequences at this field strength could greatly facilitate clinical applications of 1H-MRS. For 7T, optimized short-TE and long-TE approaches were reported recently (5,6).

Gly is an inhibitory neurotransmitter (7) and coagonist at glutamatergic N-methyl-D-aspartate receptors (8) and abnormal Gly levels have been implicated in neuropsychiatric disorders (8,9). Alterations in Gly levels are implicated in several diseases of the central nervous system. A recent MRS study reported elevation of brain Gly levels following oral Gly administration as a potential treatment for schizophrenia (10). Early studies of high-resolution MRS in brain biopsies reported that Gly levels are elevated in malignant tumors (1116). The studies showed unmasked detections of increased Gly levels in glioblastoma multiforme (GBM) and recurrent astrocytomas compared to that in normal tissues and low-grade gliomas. In-vivo observations of increased Gly levels in tumors have been reported in recent years (1719). High-grade gliomas showed higher Gly levels than low-grade gliomas and thus Gly was proposed as a biomarker for brain tumor malignancy (18,19). Gly was also observed to be elevated in central neurocytomas (17). Although human brain Gly can be directly observed in vivo by 1H-MRS when severe excess of Gly is present such as in nonketotic hyperglycinemia (20,21), precise measurement of moderately increased or decreased Gly levels remains as a challenge. Given the implications of Gly in several diseases, development of noninvasive methods for measuring brain Gly levels with improved precision is of high significance.

The current paper reports differentiation between Gly and mI by means of TE optimization of point-resolved spectroscopy (PRESS) at 3T. The subecho times were optimized with density matrix simulations that include the slice selection of the radio-frequency (RF) and gradient pulses. The performance of the Gly detection was validated in phantoms and in the healthy brain. Preliminary clinical application in patients with GBM is demonstrated.

MATERIALS AND METHODS

Experiments were carried out on a 3.0 T whole-body Philips scanner with dual QUASAR gradient coils (Philips Medical Systems, Best, The Netherlands). An integrated body coil was used for RF transmission and an 8-channel phased-array head coil for signal reception. A PRESS sequence was used for measuring brain metabolite levels. Volume localization was obtained using a 9.8-ms amplitude/frequency-modulated 90° RF pulse (bandwidth = 4.22 kHz) and 13.2-ms amplitude-modulated 180° RF pulses (bandwidth = 1.25 kHz), at a maximum available RF field intensity of 13.5 μT.

PRESS subecho times were optimized with density matrix simulations for detection of Gly at 3T. The time evolution of the density operators of mI and Gly proton spins was calculated numerically incorporating the 90° and 180° slice-selective RF and gradient pulses. A product-operator-based transformation matrix, which represents the coherence evolution during an NMR action, was created for each slice-selective RF pulse (carrier at 3.0 ppm) and used to calculate the density operator at the onset of the data acquisition for various echo time sets (4). Spectra of mI and Gly were calculated for the PRESS first and second echo times TE1 and TE2 between 20 – 200 ms with 2 ms increments. The density matrix simulations were performed using in-house Matlab programs (The MathWorks Inc., Natick, MA, USA). Published chemical shift and coupling constants were used in the simulation (1).

The optimized PRESS echo time was tested on two spherical phantoms (inner diameter = 6 cm; pH = 7.2); one with Cr (35 mM) and mI (25 mM), and another with Gly (1 mM), mI (5 mM), Cr (7 mM), N-acetylaspartate (NAA) (10 mM), and glutamate (Glu) (10 mM). Phantom spectra were obtained from a 2×2×2 cm3 voxel, using a TR of 15 s (> 5T1).

In vivo experiments included the validation of the optimized PRESS sequence in the healthy human brain and its application in brain tumors. The protocol was approved by the Institutional Review Board of the University of Texas Southwestern Medical Center. Written informed consent was obtained prior to the scans. In vivo validation of the PRESS was conducted in five healthy volunteers. Following survey imaging scans, T1-weighted images (MP-RAGE) were acquired. A 2×2×2 cm3 voxel was positioned in the right parietal cortex. The application of the Gly-optimized PRESS in brain tumors was undertaken in 12 patients with GBM, as determined by histological analyses of biopsies according to the World Health Organization (WHO) criteria. Following the survey and MP-RAGE scans, T2-weighted fluid-attenuated inversion recovery (T2w FLAIR) images were acquired to identify tumor regions. A 2×2×2 cm3 voxel was positioned at the center of the tumor mass for MRS data acquisition. First and second-order shimming was carried out, using FASTMAP (22), for a 5×5×5 cm3 volume that included the MRS voxel. Data acquisition parameters included TR = 2.0 s, sweep width = 2500 Hz, and 2048 sampling points. Spectra were acquired in 32 blocks, each with 4 averages (scan time 4.3 min). A 64-step phase cycling scheme was employed (four steps for each RF pulse of PRESS). A variable-flip-angle water suppression subsequence was applied prior to the PRESS sequence (23). The carriers of the volume selective RF pulses were set at 2.7 ppm for water suppressed data acquisitions. An unsuppressed water signal was acquired using the same gradient scheme for eddy current compensation using RF carrier at the water resonance. In addition, an unsuppressed short-TE water signal was acquired using STEAM (TE, TM) = (18, 21) ms and TR = 20 s for use as reference in metabolite quantification.

The multi-block data were processed individually for correction of eddy current and frequency drift artifacts using an in-house Matlab program. Residual eddy current effects were removed using the unsuppressed brain water signal (24). The frequency drifts were corrected using a prominent singlet (2.01 or 3.21 ppm) as a reference. Data were apodized with a 1-Hz exponential function prior to Fourier transformation. LCModel software was used for spectral fitting (25), using numerically-calculated spectra of 21 metabolites as a basis set. The basis set included spectra of Cr, NAA, Gly, mI, Glu, Gln (glutamine), Lac (lactate), Ala (alanine), Ace (acetate), Asp (aspartate), Eth (ethanolamine), GABA (γ-aminobutyric acid), GSH (glutathione), PE (phosphorylethanolamine), Ser (serine), sI (scyllo-inositol), Tau (taurine), NAAG (N-acetylaspartylglutamate), GPCPC (glycerophosphorylcholine + phosphorylcholine), Glc (glucose) and fCho4 (4 coupled spins of free choline). Given the potential differences in T1 and T2 between resonances within a metabolite, the subgroups of metabolites were treated as separate systems in the fitting according to their coupling connections. These included Cr (methyl and methylene groups), NAA (acetyl and aspartate moieties), NAAG (acetyl, aspartyl, and glutamate moieties), and GSH (glycine, cysteine, and glutamate moieties). Concentrations of these metabolites were estimated from their dominant resonances. The spectral fitting was conducted between 0.5 – 4.1 ppm, using a default spline baseline option. Cramér-Rao lower bounds (CRLB) were returned as percentage standard deviation of the fits.

Metabolite levels in normal brain and tumors, Cm, were estimated from LCModel estimates of metabolite signals, LCMm, and the short-TE, long-TR STEAM water peak area, Sw, using an equation

equation M1
[1]

where TEdiff was the TE difference between the metabolite and STEAM water signal acquisitions (160 − 18 = 142 ms) and TRm was the repetition time of metabolite acquisitions (2 s). Published metabolite T2 and T1 values were used (2628); T2 = 150, 250, and 200 ms for Cr, NAA, and other metabolites, respectively, and T1 = 1.2 for mI, Glu and Gln, and 1.5 for other metabolites. An adjustment factor β was set to give the mean concentration of total Cr (tCr) of the five healthy subjects as 8 mM (29) and applied for estimating metabolite concentrations in tumors, assuming that metabolite T1s and T2s are identical between normal tissue and tumors.

RESULTS

Figure 1 displays numerically-calculated PRESS spectra of mI (brown, thin) and mI+Gly (blue, thick) vs. the first and second subecho times TE1 and TE2 between 50 – 110 ms, for a concentration ratio of [mI]/[Gly] = 5. The spectral pattern and signal intensity of mI depended on the subecho times due to the J coupling effects, while the Gly peak was the same at all echo times, ignoring T2 relaxation effects. For TE1 and TE2 in 50 – 70 ms, the maximum amplitude of the mI multiplet appeared at the Gly ~3.55 ppm resonance, making it difficult to resolve Gly from the mI background signal. The mI multiplet intensity at ~3.55 was decreased at total echo time ~ 160 ms. At (TE1, TE2) = (60, 100) ms, the maximum amplitude of mI was shifted to 3.62 ppm and as a result, the spectral difference between mI and mI+Gly became pronounced. The mI multiplet was dependent on TE1 and TE2 for a constant total echo time. When TE1 was greater than TE2, the mI peak amplitude at 3.62 ppm became similar to that at 3.55 ppm and eventually the mI multiplet exhibited a uniform low intensity between 3.52 – 3.62 ppm at (TE1, TE2) = (110, 50) ms. A Gly peak may be easily detectable at this condition, but mI may not be measured reliably due to its relatively low intensity. Thus, (TE1, TE2) = (60, 100) ms may be optimal for measuring both Gly and mI and was selected for the present study.

FIG. 1
Calculated PRESS spectra of mI+Gly (thick/blue) and mI (thin/brown), at 3T, are displayed vs. subecho times TE1 and TE2 for 50 – 110 ms with 10 ms increments. The spectra are scaled with a concentration ratio of [Gly]:[mI] = 1:5. The spectral ...

Figure 2 presents the phantom validation for the performance of the optimized PRESS echo time. A STEAM sequence with (TE, TM) = (18, 21) ms, whereby the mI signal modulation due to J coupling effects was negligible, reproduced closely a 90°-acquired spectrum. For phantom-1, a STEAM spectrum showed a mI multiplet at ~3.55 ppm with amplitude 31% with respect to the Cr 3.03-ppm singlet at linewidth of 3 Hz. Following the PRESS (TE1, TE2) = (60, 100) ms, the mI signal between 3.45 – 3.65 ppm was reduced to 5% relative to the Cr peak. The lineshape of this mI multiplet agreed well with the simulation result in Fig. 1. For phantom-2, which contained Gly at [mI]/[Gly] = 5, Gly was not readily detectable in a STEAM spectrum due to the large mI signal. However, the small Gly singlet at 3.55 ppm was clearly revealed in the optimized-PRESS spectrum. The composite signal pattern of Gly and mI agreed well with the calculated spectra in Fig. 1. An LCModel analysis of this phantom spectrum gave a concentration ratio of [Cr]:[mI]:[Gly] = 7:4.8:1. The MRS-measured concentration ratio reproduced closely the prepared concentration ratio (i.e., 7:5:1). The measured ratio of mI was slightly lower than the prepared value most likely due to the mI T2 shorter than the Cr T2 in the phantom solution (0.8 and 1.6 s, respectively, as measured from phantom-1). The mI multiplet at 4.06 ppm was smaller in the PRESS spectra than in the STEAM spectra, mainly due to the effects of J evolution. The Glu multiplet was also reduced in the PRESS spectrum due to the J coupling effects, giving a C4-proton multiplet at 2.35 ppm with amplitude ~10% with respect to the Cr 3.03 ppm peak.

FIG. 2
In vitro spectra at PRESS (TE1, TE2) = (60, 100) ms, obtained from Phantom-1 with mI (25 mM) and Cr (40 mM), and Phantom-2 with Gly, mI, Cr, NAA and Glu at a concentration ratio of 1:5:8:10:10, at 3T, are shown together with short-TE STEAM spectra. Singlet ...

Figure 3 presents signal-to-noise ratio (SNR) dependence of Gly estimation by the optimized PRESS sequence for the healthy brain. Spectra are shown for various number of signal averages (NSA) in the left column and their LCModel fitting results (concentration estimates and CRLB) are plotted vs. NSA in the right column for Gly, mI, tCr, NAA and GPCPC. With TR = 2 s and voxel size = 8 ml, the predominant signals of Cr, NAA and GPCPC was measurable with CRLB less than 5% even at NSA = 8, but the concentration estimates were not constant until NSA = 30. The dependence of the concentration estimates and CRLB on NSA was more pronounced for mI and Gly. The CRLB of mI was less than 20% for NSA ≥ 8. The mI concentration estimates were varied at small NSAs and became constant after NSA ~ 80. For Gly, CRLB less than 20% was achievable only when NSA was greater than 50. After this, Gly CRLB was further decreased with increasing NSA and eventually became ≤ 8 at NSA > 110. Figures 3e and 3f present LCModel fitting results from a basis set with or without Gly, respectively. While the fit with Gly in the basis function gave noise-level residuals at ~3.55 ppm, a fit without Gly produced large residuals at the Gly resonance, indicating that a peak at 3.55 ppm is primarily attributed to Gly.

FIG. 3
(a) In vivo brain spectra are presented vs. number of signal averages (NSA). Vertical dotted lines are drawn at 3.55 and 3.62 ppm. (b) Voxel (2×2×2 mm3) positioning in the occipital cortex. (c) Concentration estimates and (d) LCModel spectral ...

Figure 4 displays in vivo brain spectra and the LCModel fits from a healthy volunteer and a GBM patient, together with the individual signals of Gly, mI, Lac, Ala, Glu and Gln. The GBM patient was scanned for two regions where the T2w-FLAIR intensity was enhanced. The in vivo spectra were well reproduced by the fits, with small residuals between 1.0 – 4.1 ppm. The residuals between 3.45 – 3.65 ppm were equivalent to the noise levels, indicating that the signals were primarily attributed to Gly and mI. The spectra from the GBM patient both exhibited spectral patterns different from the normal-brain spectral pattern. Moreover, the GBM spectra were quite different from each other. Table 1 shows metabolite concentrations of tCr, NAA, GPCPC, Gly, mI, Lac and Ala, estimated with Eq. [1]. The Gly and mI concentrations in the normal brain were estimated to be 0.6±0.1 and 4.3±0.4 (mean±SD, n=5), respectively. The Gly signal in Fig. 4b was much larger than in Fig. 4a. The Gly level in this tumor mass was estimated as 3.3 mM, approximately 6-fold higher than the mean Gly level in the healthy brain. The mI level in this region was measured to be lower compared to the normal brain tissue. In contrast, the spectrum from another tumor mass, Fig. 4c, exhibited a normal Gly level and a substantially increased mI level. The mI level in this tumor mass was estimated to be 12.3 mM, ~3-fold higher than the normal level. Lac and Ala were also measurable in the tumor spectra, but undetectable in the normal brain. The Lac and Ala concentrations were estimated to be ~3 mM and ~1 mM for both tumor regions, with mean CRLBs of ~8% and ~18%, respectively. Further, Glu and Gln levels were altered in a tumor mass (Fig. 4c), whilst the other tumor mass (Fig. 4b) showed no difference compared to normal tissues. For tCr, NAA and GPCPC, the signals exhibited a well-known tumor spectral pattern, i.e., a GPCPC signal greater than tCr and NAA signals. The tCr level in the tumor mass in the parietal region (Fig. 4c) was estimated to be 11.1 mM, ~1.4 times higher compared to the normal brain. The tumor mass for the spectrum in Fig. 4b showed T1w-MRI enhancement following a gadolinium injection, but the tumor mass for Fig. 4c did not, as observed in a regular clinical scan in this patient.

FIG. 4
In vivo brain spectra from (a) a healthy volunteer and (b, c) a multi-focal GBM patient are shown together with the LCModel fits and individual signals of Gly, mI, Lac, Ala, Glu and Gln. Voxel positioning (2×2×2 cm3) is shown in the images. ...
Table 1
Metabolite concentrations of 5 normal subjects (mean±SD) and eight GBM patients are tabulated. The concentrations were estimated using short-TE water signals as a normalization reference and assuming a normal-brain tCr concentration at 8 mM. The ...

Among the 12 GBM patients enrolled in the present study, Gly levels were elevated in 8 patients. Figure 5 displays in vivo brain spectra from a normal subject (Fig. 5a) and seven patients with GBM (Figs. 5b–h), following the normalization with respect to the short-TE water signals. Compared to the normal-tissue spectrum, spectra from tumors all exhibited abnormal spectral patterns, which include increases in Gly, GPCPC, Lac, Ala and lipids (Lip) and decreases in tCr and NAA. The Gly levels in Fig. 5b – 5f were higher than the normal level (by factors of 3.1, 3.8, 8.3, 1.9 and 2.0, respectively). The GBM patients of Figs. 5g and 5h exhibited normal Gly levels but abnormal concentrations for other metabolites. Increased GPCPC/tCr and GPCPC/NAA ratios may be representative abnormalities in metabolic profiles in GBM, as shown in Fig. 5. The concentrations of tCr and NAA appeared to be lower than normal in most of the spectra from GBM patients, leading to elevation of GPCPC/tCr and GPCPC/NAA ratios. However, for patients for Figs. 5b, 5c and 5f, the increases in these concentration ratios resulted from decreased tCr and NAA levels, with the GPCPC level about the same as or slightly smaller than the normal level. In Fig. 5b, a large negative doublet of the Lac CH3 protons was detected at 1.31 ppm. The Lac level was estimated to be 12.6 mM, much higher than the normal level (< 1 mM). In Figs. 5c and 5g, large positive signals appeared at 1.3 ppm, due to increased Lip levels, but the spectral patterns at the resonance were slightly different between the two spectra due to the signals of Lac and Ala. In Fig. 5c, relatively small negative signals of Lac and Ala were overlapped with the large Lip signal, but with their negative polarity, the Lac and Ala signals were well differentiated from Lip signals using the LCModel spectral fitting, giving Lac and Ala levels at 4.9 and 1.7 mM, respectively, with CRLB of 9% for both. The Lac and Ala levels were not measurable in Fig. 5g due to the insufficient SNR and the presence of a large, broad lipid signal. In Figs. 5e and 5f, for which the Gly levels were estimated to be 2-fold higher than the normal level, the spectral patterns at 3.5 – 3.65 ppm were different due to different mI concentrations in these patients. The mI concentration was similar to the normal level in Fig. 5e, but lower in Fig. 5f. In these two spectra, Lac-Lip composite signals were resolved, giving Lac levels as 7.0 and 5.4 mM, respectively. The Ala levels were negligible in these patients. In the GBM patients for Figs. 5d and 5h, Lac and Ala signals were both pronounced, indicating their concentrations much higher than normal.

FIG. 5
In vivo brain spectra from a normal subject (a) and 7 patients with GBM are shown on top of LCModel fits, together with voxel positioning (size 2×2×2 cm3). Spectra were all acquired with PRESS (TE1, TE2) = (60, 100) ms, TR = 2 s, and NSA ...

DISCUSSION

Tumor tissue can be differentiated from surrounding normal tissue based on the appearance of alterations in metabolic activity or the accumulation of specific small-molecule metabolites. So far, clinical techniques in metabolic imaging are limited to the analysis of molecules with concentrations in the range of several mM, significantly restricting the overall view of tumor metabolism in vivo. Advanced techniques like high-field MRS promise to bring out more detail about the involvement of less-abundant metabolites in tumor biology and therapy. Here we developed an MRS method to detect Gly, a key intermediary molecule that connects glycolysis to multiple other metabolic processes, including maintenance of one-carbon pools and the production of proteins and nucleic acids for cell growth and proliferation. Evidence suggests that production of Gly is under oncogenic control, emphasizing its importance in tumor biology (30). This study presents a resolution improvement for measurement of Gly at 3T, using an optimized PRESS subecho times (TE1, TE2) = (60, 100) ms, and its performance for the normal and tumorous brain. This echo time pair provides a mI multiplet pattern that has reduced amplitude at the Gly 3.55 ppm resonance and a measurable strength at an adjacent frequency of 3.62 ppm. This enables Gly detection at 3T, with improved precision compared to short-TE conventional MRS methods. The attenuation of macromolecule signals at the long TE is beneficial for improving the specificity of the small Gly signal, but the efficiency of mI measurement may be reduced.

Compared to a prior triple-refocusing study at 3T (4), which gave CRLB of Gly at 13%, the mean CRLB of ~9% of the present study is a marked improvement in that our scan time was ~2-fold shorter (4.5 min vs. 10 min) for a 3-fold smaller voxel size (8 ml vs. 22 ml). This is likely due to the SNR enhancement by a multi-channel phased-array coil and partially to reduced T2 signal loss (TE = 160 ms vs. 200 ms). The Gly CRLB of the present study is lower than that of a prior 2D NMR study at 3T (2) and is quite comparable to those of prior 7T studies (~7%) (5,6). The mean Gly concentration in the normal brain of the present study is slightly higher than or comparable to those of prior 3T studies, and lower than or similar to prior measures at 7T. The Gly concentrations in the healthy brain, reported in these several studies and the present study, appear to be 0.5 – 1 mM with reference to tCr at 8 mM, depending on the sequences and the brain regions.

The current paper presents metabolite concentrations obtained using the short-TE brain water signals as a reference for normal brain and tumors. The metabolite estimation in this study is valid only when the water concentration is not very different between brain regions and between normal tissue and tumors. For the data from healthy volunteers, since the metabolite levels were calculated using an adjustment factor β in Eq. [1] which was set for the normal tCr level at 8 mM, the quantification method is essentially equivalent to directly using tCr as a reference without the need of brain water data. However, this method may not be applicable in tumors since the tCr level is altered drastically, requiring alternative reference signals for metabolite estimation. Although uncertainties in metabolite estimates can be theoretically minimized employing an external reference signal such as from a phantom, provided the low reproducibility of this external referencing method shown in a multi-center study (31), referencing with respect to brain water signals may be a realistic means of estimating metabolite levels in tumor data (32). The present study reports abnormal Gly levels in tumors with respect to the Gly levels in the occipital region of the healthy brain. However, precise evaluation of the concentration changes requires comparison of metabolite measures from matched brain regions when a regional distribution of metabolite levels exists.

Precise estimation of Gly may be hampered by the spectral overlap with the J-coupled resonances of free Cho (fCho4) and threonine (Thr), in addition to mI. A model spectrum of fCho4 was included in the basis function for the LCModel analysis, but a Thr spectrum was not included. The signal of fCho4 was not detectable in most of our data, primarily due to insufficient SNR for this low-concentration metabolite (33), whose estimates were essentially zero with CRLB > 100%. In addition to its CH proton resonance at 3.58 ppm in the proximity of the Gly resonance, Thr has a CH3 proton resonance at 1.31 ppm, overlapped with the Lac CH3 proton resonance. Since the J coupling strengths of the CH3 protons with their coupling partners are about the same between Lac and Thr, the CH3 resonances exhibit similar multiplet pattern irrespective of echo time and thus difficult to differentiate unless spectral editing is applied (34). Inclusion of Lac and Thr in the basis set resulted in Thr levels of 2–3 mM in some tumor data, higher than the normal brain level (0.5 – 1 mM) (34,35). Thr was not included in the basis set for the tumor data analyses of the present study, thus the Lac measures may contain considerable contaminations from Thr when Thr is elevated in GBM (13,36). In addition, D-glucose has a coupled resonance at 3.51 ppm. The LCModel estimates of glucose were low (< 0.5 mM) with large CRLB (> 200).

The proposed sequence allows co-detection of Lac and Ala CH3 proton resonances against the overlapping signals of the Lip CH2-chain proton resonances. The coherence evolution under J = 7 Hz during TE = 160 ms results in a fraction of negative inphase coherences, giving negative doublets at 1.31 and 1.47 ppm, respectively, while the Lip signal remains positive. Since the spectral distances between the coupling partners are relatively large in these metabolites, partial-volume artifacts (37,38) may occur substantively when the BW of the PRESS 180° pulse is not sufficiently large. For the 180° pulse (BW = 1.25 kHz) of the present study, with the RF pulse carrier at 2.7 ppm, the slice of the Lac CH3 resonance is displaced by 14% with respect to the planned slice, resulting in no signal from part of the planned voxel and contamination signals from outside the planned voxel. The slice mismatch of the Lac 1.31 and 4.1 ppm resonances is 29% relative to the slice thickness, leading to four compartments of 50%, 21%, 21% and 8% relative to the entire 2D volume, between which the coherence evolution differs due to the different effects of the 180° pulses on the coupled resonances. When the metabolite concentrations are homogeneous within the localized volumes, the signal loss due to this inhomogeneous J evolution, which is ~55% compared to the infinite BW, can be properly accounted for using a volume localized model spectrum for data analysis, as in the present study. However, when the Lac levels are inhomogeneous, which is likely the case in most of tumors, the in vivo spectral pattern may be different than the calculation due to the different contributions from the four compartments.

Although it has been extensively demonstrated in ex vivo MRS studies in tumor biopsies, elevation of Gly levels in human brain tumors in vivo has been reported in a few studies in recent years (1719). These in vivo studies suggested an increase in Gly abundance in high-grade gliomas compared to low-grade gliomas, and proposed that Gly may be a biomarker of tumor malignancy. In the previous in vivo studies, MRS data were acquired at both short (~30 ms) and long echo times (~140 ms) and subsequently the spectral difference at ~3.55 ppm between the two echo times, which arises largely from the strong dependence of the mI multiplet on TE due to the J coupling effects, were analyzed to obtain the Gly portion of the composite signal. Detection of small changes in Gly concentrations may require precise evaluation of the contribution of the macromolecule resonances to the signal at ~3.55 ppm, especially at short TE. However, as demonstrated in our study, Gly and mI signals can be differentiated quite reliably when a properly refined echo time is employed, enabling detection of moderate alterations of Gly levels in tumor conditions. The echo time proposed in this study is quite long, so signal loss due to T2 relaxation is substantive. However, small signals are often better resolved at optimized long TE, benefiting from the attenuated complex baseline signals of macromolecules, as indicated by low CRLB (~9%) for SNR ≈ 6 for Gly in normal brain. Among the 12 GBM patients enrolled in this study, 8 patients showed increased Gly levels and the remaining patients exhibited the normal level. This result implies that elevated Gly levels occur only in a subset of human gliomas. The biological basis for this subset is unclear at present, but may be related to specific molecular signatures that favor the accumulation of intracellular glycine. With its improved ability for Gly detection, the proposed PRESS sequence may have extended applications for which alterations of Gly levels are small or the levels are even decreased relative to its normal level.

ACKNOWLEDGMENTS

This work was supported by a research grant funded by the National Institute of Health (RC1NS0760675), research grants funded by the Cancer Prevention Research Institute of Texas (CPRIT) (RP101243-P04 and HIRP100437), and a resource grant funded by the National Center for Research Resources (P41 RR002584).

REFERENCES

1. Govindaraju V, Young K, Maudsley AA. Proton NMR chemical shifts and coupling constants for brain metabolites. NMR Biomed. 2000;13:129–153. [PubMed]
2. Schulte RF, Boesiger P. ProFit: two-dimensional prior-knowledge fitting of J-resolved spectra. NMR Biomed. 2006;19:255–263. [PubMed]
3. Prescot AP, de BFB, Wang L, Brown J, Jensen JE, Kaufman MJ, Renshaw PF. In vivo detection of brain glycine with echo-time-averaged 1H magnetic resonance spectroscopy at 4.0 T. Magn Reson Med. 2006;55:681–686. [PubMed]
4. Choi C, Bhardwaj PP, Seres P, Kalra S, Tibbo PG, Coupland NJ. Measurement of glycine in human brain by triple refocusing 1H-MRS in vivo at 3.0T. Magn Reson Med. 2008;59:59–64. [PubMed]
5. Gambarota G, Mekle R, Xin L, Hergt M, van der Zwaag W, Krueger G, Gruetter R. In vivo measurement of glycine with short echo-time 1H MRS in human brain at 7 T. MAGMA. 2009;22:1–4. [PubMed]
6. Choi C, Douglas D, Hawesa H, Jindal A, Storey C, Dimitrov I. Measurement of glycine in human prefrontal brain by point-resolved spectroscopy at 7.0 tesla in vivo. Magn Reson Med. 2009;62:1305–1310. [PubMed]
7. Legendre P. The glycinergic inhibitory synapse. Cell Mol Life Sci. 2001;58:760–793. [PubMed]
8. Eulenburg V, Armsen W, Betz H, Gomeza J. Glycine transporters: essential regulators of neurotransmission. Trends Biochem Sci. 2005;30:325–333. [PubMed]
9. Javitt DC. Glutamate and Schizophrenia: Phencyclidine, N-Methyl-d-Aspartate Receptors, and Dopamine-Glutamate Interactions. Int Rev Neurobiol. 2007;78:69–108. [PubMed]
10. Kaufman MJ, Prescot AP, Ongur D, Evins AE, Barros TL, Medeiros CL, Covell J, Wang L, Fava M, Renshaw PF. Oral glycine administration increases brain glycine/creatine ratios in men: a proton magnetic resonance spectroscopy study. Psychiatry Res. 2009;173:143–149. [PMC free article] [PubMed]
11. Gill SS, Thomas DG, Van Bruggen N, Gadian DG, Peden CJ, Bell JD, Cox IJ, Menon DK, Iles RA, Bryant DJ, et al. Proton MR spectroscopy of intracranial tumours: in vivo and in vitro studies. J Comput Assist Tomogr. 1990;14:497–504. [PubMed]
12. Bruhn H, Michaelis T, Merboldt KD, Hanicke W, Gyngell ML, Hamburger C, Frahm J. On the interpretation of proton NMR spectra from brain tumours in vivo and in vitro. NMR Biomed. 1992;5:253–258. [PubMed]
13. Florian CL, Preece NE, Bhakoo KK, Williams SR, Noble M. Characteristic metabolic profiles revealed by 1H NMR spectroscopy for three types of human brain and nervous system tumours. NMR Biomed. 1995;8:253–264. [PubMed]
14. Kinoshita Y, Yokota A. Absolute concentrations of metabolites in human brain tumors using in vitro proton magnetic resonance spectroscopy. NMR Biomed. 1997;10:2–12. [PubMed]
15. Martinez-Bisbal MC, Marti-Bonmati L, Piquer J, Revert A, Ferrer P, Llacer JL, Piotto M, Assemat O, Celda B. 1H and 13C HR-MAS spectroscopy of intact biopsy samples ex vivo and in vivo 1H MRS study of human high grade gliomas. NMR Biomed. 2004;17:191–205. [PubMed]
16. Lehnhardt FG, Bock C, Rohn G, Ernestus RI, Hoehn M. Metabolic differences between primary and recurrent human brain tumors: a 1H NMR spectroscopic investigation. NMR Biomed. 2005;18:371–382. [PubMed]
17. Yeh IB, Xu M, Ng WH, Ye J, Yang D, Lim CC. Central neurocytoma: typical magnetic resonance spectroscopy findings and atypical ventricular dissemination. Magn Reson Imaging. 2008;26:59–64. [PubMed]
18. Hattingen E, Lanfermann H, Quick J, Franz K, Zanella FE, Pilatus U. 1H MR spectroscopic imaging with short and long echo time to discriminate glycine in glial tumours. MAGMA. 2009;22:33–41. [PubMed]
19. Davies NP, Wilson M, Natarajan K, Sun Y, MacPherson L, Brundler MA, Arvanitis TN, Grundy RG, Peet AC. Non-invasive detection of glycine as a biomarker of malignancy in childhood brain tumours using in-vivo 1H MRS at 1.5 tesla confirmed by ex-vivo high-resolution magic-angle spinning NMR. NMR Biomed. 2010;23:80–87. [PubMed]
20. Choi CG, Lee HK, Yoon JH. Localized proton MR spectroscopic detection of nonketotic hyperglycinemia in an infant. Korean J Radiol. 2001;2:239–242. [PMC free article] [PubMed]
21. Huisman TA, Thiel T, Steinmann B, Zeilinger G, Martin E. Proton magnetic resonance spectroscopy of the brain of a neonate with nonketotic hyperglycinemia: in vivo-in vitro (ex vivo) correlation. Eur Radiol. 2002;12:858–861. [PubMed]
22. Gruetter R. Automatic, localized in vivo adjustment of all first- and second-order shim coils. Magn Reson Med. 1993;29:804–811. [PubMed]
23. Ogg RJ, Kingsley PB, Taylor JS. WET, a T1- and B1-insensitive water-suppression method for in vivo localized 1H NMR spectroscopy. J Magn Reson B. 1994;104:1–10. [PubMed]
24. Klose U. In vivo proton spectroscopy in presence of eddy currents. Magn Reson Med. 1990;14:26–30. [PubMed]
25. Provencher SW. Estimation of metabolite concentrations from localized in vivo proton NMR spectra. Magn Reson Med. 1993;30:672–679. [PubMed]
26. Mlynarik V, Gruber S, Moser E. Proton T1 and T2 relaxation times of human brain metabolites at 3 Tesla. NMR Biomed. 2001;14:325–331. [PubMed]
27. Traber F, Block W, Lamerichs R, Gieseke J, Schild HH. 1H metabolite relaxation times at 3.0 tesla: Measurements of T1 and T2 values in normal brain and determination of regional differences in transverse relaxation. J Magn Reson Imaging. 2004;19:537–545. [PubMed]
28. Choi C, Coupland NJ, Bhardwaj PP, Kalra S, Casault CA, Reid K, Allen PS. T2 measurement and quantification of glutamate in human brain in vivo. Magn Reson Med. 2006;56:971–977. [PubMed]
29. Tkac I, Oz G, Adriany G, Ugurbil K, Gruetter R. In vivo 1H NMR spectroscopy of the human brain at high magnetic fields: metabolite quantification at 4T vs. 7T. Magn Reson Med. 2009;62:868–879. [PMC free article] [PubMed]
30. Nikiforov MA, Chandriani S, O'Connell B, Petrenko O, Kotenko I, Beavis A, Sedivy JM, Cole MD. A functional screen for Myc-responsive genes reveals serine hydroxymethyltransferase, a major source of the one-carbon unit for cell metabolism. Mol Cell Biol. 2002;22:5793–5800. [PMC free article] [PubMed]
31. Keevil SF, Barbiroli B, Brooks JC, Cady EB, Canese R, Carlier P, Collins DJ, Gilligan P, Gobbi G, Hennig J, Kugel H, Leach MO, Metzler D, Mlynarik V, Moser E, Newbold MC, Payne GS, Ring P, Roberts JN, Rowland IJ, Thiel T, Tkac I, Topp S, Wittsack HJ, Wylezinska M, Zaniol P, Henriksen O, Podo F. Absolute metabolite quantification by in vivo NMR spectroscopy: II. A multicentre trial of protocols for in vivo localised proton studies of human brain. Magn Reson Imaging. 1998;16:1093–1106. [PubMed]
32. Tong Z, Yamaki T, Harada K, Houkin K. In vivo quantification of the metabolites in normal brain and brain tumors by proton MR spectroscopy using water as an internal standard. Magn Reson Imaging. 2004;22:1017–1024. [PubMed]
33. Barker PB, Breiter SN, Soher BJ, Chatham JC, Forder JR, Samphilipo MA, Magee CA, Anderson JH. Quantitative proton spectroscopy of canine brain: in vivo and in vitro correlations. Magn Reson Med. 1994;32:157–163. [PubMed]
34. Choi C, Coupland NJ, Kalra S, Bhardwaj PP, Malykhin N, Allen PS. Proton spectral editing for discrimination of lactate and threonine 1.31 ppm resonances in human brain in vivo. Magn Reson Med. 2006;56:660–665. [PubMed]
35. Marjanska M, Henry PG, Ugurbil K, Gruetter R. Editing through multiple bonds: threonine detection. Magn Reson Med. 2008;59:245–251. [PubMed]
36. Kotitschke K, Jung H, Nekolla S, Haase A, Bauer A, Bogdahn U. High-resolution oneand two-dimensional 1H MRS of human brain tumor and normal glial cells. NMR Biomed. 1994;7:111–120. [PubMed]
37. Slotboom J, Mehlkope AF, Bovee WMMJ. The effects of frequency-selective RF pulses on J-coupled spin-1/2 systems. J Magn Reson A. 1994;108:38–50.
38. Kaiser LG, Young K, Matson GB. Numerical simulations of localized high field 1H MR spectroscopy. J Magn Reson. 2008;195:67–75. [PMC free article] [PubMed]