The MEGA-PRESS sequence is especially designed for the detection of γ-aminobutyric acid (GABA), but also enables an estimation of the so-called ‘Glx’ concentration. This Glx concentration is a combined measure of glutamate and its precursor glutamine, from which the concentrations are difficult to disentangle due to overlapping resonance frequencies. However, some authors claim that this 3.7 ppm Glx peak constitutes predominantly glutamate (see for instance Ref [1-3]), which is attractive as this enables the measurement of both major neurotransmitters glutamate and GABA with the same sequence. In this phantom study, we therefore tested this claim by assessing associations of the actual concentrations of glutamate, glutamine, and the chemically related GABA on the resulting ‘Glx’ measurements with MEGA-PRESS sequences. Additionally, the spectra were also analyzed with LCModel, which estimates the metabolite concentrations by a linear combination of individual metabolite spectra, rather than by fitting one peak, which is a commonly applied method to estimate the Glu concentration.

Phantoms: Nine 350 mL phantom solutions were made out of a stock solution, including physiological concentrations of creatine (7.8 mM), myo-Inositol (5.9 mM), N-acetylaspartate (10.3 mM) and choline (1.7 mM). Glutamate, glutamine and GABA were added to the solutions in different concentrations, varying from 0-18 mM, 0-9 mM, and 0-3.5 mM for glutamate, glutamine, and GABA, respectively. The pH of these solutions was set to 7.2 using NaOH.

Measurements: MRS was performed on the day of phantoms preparation and included a MEGA-PRESS (TE/TR 68/2000 ms, 320 averages, with editing pulses at 1.9 (ON) and 7.46 ppm (OFF) interleaved in 40 blocks, MOIST water suppression, voxel size 3x3x3 cm³) on a 3T Philips Achieva TX MR scanner with a 32 channel head coil. Each phantom was measured twice, while the temperature was controlled to 36.5-38°C.

Analysis: Gannet (version 2.0) was applied to estimate the ‘Glx peak’ in the MEGA-PRESS difference spectrum (Figure 1A) using a double Gaussian fit^[4]. The MEGA-PRESS difference spectrum was also analyzed with LCModel (version 6.3-1B, Figure 1B). Concentrations were considered relative to creatine (Gannet) or N-acetylasparatate (LCModel). The mean concentrations of the repeated measurements were considered for subsequent analyses. To assess the relative contributions of the glutamate, glutamine, and GABA concentrations to the Glx concentration, linear regression analysis was applied with the measured ‘Glx peak’ as dependent variable, and preset metabolite concentrations as independent variables. To assess the accuracy, the Pearson correlation coefficient was calculated between the actual and measured concentrations of glutamate and glutamine as measured with LCModel. Finally, for the reproducibility, the mean coefficient of variation (CV) was calculated over the two subsequent measurements.

The regression analysis showed significant associations of both glutamate and glutamine with the Glx peak measured with MEGA-PRESS, with comparable regression coefficients (Table 1, Figure 2). The individual LCModel estimates showed a high correlation with the prepared glutamate and glutamine concentrations, with R2>0.96 in all measurements. The CVs of the glutamine measurements (6.9%) were approximately twice as high as those of the glutamate (3.2%) or Glx measurements (3.2%).The results show that both glutamate and glutamine contribute strongly to the Glx peak measured with a MEGA-PRESS sequence. The in vivo glutamate concentration is roughly twice the glutamine concentration^[5], suggesting that approximately two-third of the Glx estimate represents glutamate, while the remaining portion is glutamine. This result contradicts a previous in vivo study, that concluded that the Glx peak in MEGA-PRESS constitutes mainly glutamate, based on a comparison between the Glx measure from MEGA-PRESS and glutamate and Glx measures from CT-PRESS^[6]. However, as the glumate and Glx concentration are strongly correlated, the independent contribution of glutamate to Glx cannot be estimated in vivo. Estimations by LCModel for glutamate and glutamine showed high correlations to the actual concentrations and low CVs, as was shown previously^[7]. A high reproducibility of these glutamate measurements was also shown in vivo, albeit worse for glutamine measurements^[8]. In conclusion, measuring the ‘Glx’ peak does not yield a reliable and independent estimate of exclusively glutamate , whereas a reliable estimate of this concentration can be obtained by analyzing the MEGA-PRESS spectra with LCModel.