Schizophrenia is a chronic mental disorder characterized by disturbances in perception, thought, emotion, cognition, and behavior. Over the course of development, interactions between genes and environment may affect neural systems and give rise to several clinical symptoms. The pathology in schizophrenia includes abnormalities in neurotransmission, brain structure, and function. Proton magnetic resonance spectroscopy (1H-MRS) is a non-invasive neuroimaging technique used for measurement of regional brain metabolites that possibly reflect the status of neuronal and glial functions. In schizophrenia an increasing number of 1H-MRS studies have been conducted suggesting abnormal neurometabolism. However, many 1H-MRS studies yielded ambivalent results. 1H-MRS is a relatively new technique and there are many drawbacks with respect to the validity and reliability of 1H-MRS measurements.
The main goal of this literature review was to understand the variation of 1H-MRS findings in schizophrenia. We highlight the differential effects of age, gender, schizophrenia subtype (e.g. paranoid type), stage of disease progression (e.g. first episode patients) and medication status on neurometabolite concentrations measured by 1H-MRS in schizophrenia. Another aim was to trace inconsistent findings of studies back to methodological differences. To allow a better understanding of the reviewed issues and findings, a short summary of the technique and measured metabolites, as a comprised view of major neurophysiologic abnormalities in schizophrenia will be provided in the following.
Magnetic resonance spectroscopy
1H-MRS is a magnetic resonance tool that specifically enables non-invasive in vivo differentiation and quantification of small chemical compounds based on different resonance frequencies (Dager et al. 2008). The proton (1H), phosphorus (31P), and carbon (13C) nuclei are the most often used atomic nuclei in MRS studies given their magnetic properties. 1H-MRS is the most frequently applied measurement as the 1H nucleus forms the most sensitive (after the unstable and radioactive 3H isotope), and most abundant, naturally-occurring nucleus (Gillies 1992; Rothman et al. 1999). The higher sensitivity of 1H-MRS makes it better suited for examining specific small brain regions, as opposed to 31P-MRS which necessitates larger voxel and longer scan times for adequate signal to noise ratio (SNR) in regions of similar sizes (Stanley et al. 2000).
The spectroscopy signal is acquired by using either ‘single voxel spectroscopy’ (SVS) or by using ‘multi voxel spectroscopic imaging’ (MRSI). The former measures resonances exclusively in a predetermined region (voxel) and has the advantages of a more homogenous magnetic field and thus better spectral resolution per voxel despite a decreased spatial resolution. MRSI, on the other hand, is able to measure different spectra in parallel and gives a better SNR ratio for multiple voxels of interest (VOIs) since the signal from each voxel is averaged for the total data collection (Drost et al. 2002). As voxel sizes are generally smaller in MRSI than in SVS, partial volume effects (i.e. combining two or more tissue types within a single voxel) have less impact on the assessment of metabolites (Stanley et al. 2000). Yet, the spatial resolution of MRSI is still quite low as compared to conventional MRI techniques with typical voxel size around 1–1.5 cc, and chemicals must exist in concentrations higher than 100 μM to be measured (Novotny et al. 2003). In sum, SVS is superior for a detailed analysis of specific regions, whereas MRSI is preferentially used when the VOI is of unknown origin and when there is a need for a quick assessment of several regions (Drost et al. 2002; Hammen and Stefan 2004; Morche 2005).
The most commonly applied localization method for sequence acquisition is ‘Point Resolved Spectroscopy’ (PRESS) and ‘Stimulated Echo Acquisition Mode’ (STEAM). With STEAM a shorter echo time (TE) can be achieved and the localization is more precise, yet PRESS has an intrinsic higher signal to noise ratio (Drost et al. 2002). The external magnetic field strength determines the location of different resonance signals, hence differences in tesla (T) result in different resonances. To standardize between different magnetic field strengths, the 1H-MRS signal is transformed to a frequency spectrum. The position of the signal peaks on the x-axis is expressed as ‘chemical shifts’ (shift in resonance frequency that is unique to a given molecule) in units of parts per million (ppm) and therefore called the ppm-axis. Consequently, inter- and intra-individual comparisons can be made (Wobrock et al. 2005; Dager et al. 2008). The measurement units can be expressed as ratios, absolute values (AVs) or arbitrary units.
Higher field strengths and short TEs (< 30 ms) heighten the SNR and narrow peak widths on the ppm-axis, resulting in improved spectral resolution and sensitivity, thereby enabling the detection of overlapping metabolites (Walter 2005). After selection of the VOI the homogeneity of the field is adjusted and optimized by a process called ‘shimming’ [through adjusting direct currents in the gradient coils and 20–25°C shim coils] to maximize the field homogeneity and thus the spectral resolution (Stanley et al. 2000). The much higher intensity of water (because of its abundance) and lipid molecules (because of their hydrogen content) as compared to the measured metabolites disturbs the 1H-MR-signal and must be suppressed (Frahm et al. 1989; Walter 2005; Wobrock et al. 2005). Consequently, water suppression techniques are used that apply frequency specific radio frequency pulses that convert and eliminate disturbing signals (Bertolino and Weinberger 1999). Lipid suppression is only used when there is a need to counteract outer-volume lipid contaminations. Long TEs decrease signals from lipid and macromolecules and thus enhance the validity of metabolite quantification, whereas short TEs are able to differentiate metabolites undetected by long TEs (Drost et al. 2002). Mostly, long TEs are used together with MRSI and short TEs together with SVS, and the former when a higher spatial resolution is required whereas the latter serves best when metabolites with overlapping peaks on the ppm-axis need to be analyzed (Stanley et al. 2000).
Given that metabolites are differentially distributed in cerebrospinal fluid (CSF), white matter (WM) and gray matter (GM), an important factor when assessing metabolites, is to correct for differences in tissue composition of the VOI. Otherwise concentrations might suffer from over and underestimations and thereby lead to so-called partial volume effects.
The general quality of 1H-MRS data depends consequently on the magnetic field strength, the size and composition of the brain region of interest, the acquisition time and the concentration of metabolites (Auer et al. 2001; Drost et al. 2002; Kreis 2004; Wobrock et al. 2005).
Major 1H-MRS metabolites: regional variability and other considerations
Neurometabolites measured by 1H-MRS include N-acetyl-aspartate (NAA), creatine (Cr), choline (Cho), myo-Inositol (mI), lactate (Lac), glutamate (Glu), glutamine (Gln), and Gamma-amino butyric acid (GABA). Glx describes the combined spectra of Glu and Gln, because of their overlapping peaks on the ppm-axis, which appear between 2.1 and 2.5 ppm and between 3.72 and 3.82 ppm (Kreis et al. 1993; Bertolino and Weinberger 1999; Dager et al. 2008). After being released by neurons, Glu is taken up into glial cells and rapidly converted to Gln, which is subsequently transported back to neurons and reconverted to Glu (Watts et al. 2005). The three multiplets resonances (2.31 ppm, 1.9 ppm, and 3.0 ppm) of the main inhibitory neurotransmitter of the brain – GABA – overlap with the Glx, NAA, and Cr spectra. Although careful post-processing may help to differentiate GABA, Glu, and Gln, often high field strengths (> 4.7 T) and short TEs together with sophisticated editing techniques are required to obtain valid results for GABA spectra (Govindaraju et al. 2000; Drost et al. 2002; Novotny et al. 2003; Hammen and Stefan 2004; van Elst et al. 2005; Bogner et al. 2010).
The peaks of total Cr (tCr) at 3.0 and 3.9 ppm comprise phosphocreatine and Cr, which can hardly be separated even when using magnetic resonance devices with magnetic field strengths higher than 1.5 T. tCr is found to be relatively constant in the general population yet higher concentrations were seen in the cerebellum and in regional GM (Pouwels and Frahm 1998; Drost et al. 2002; Yucel et al. 2007; Brief et al. 2009; Ongur et al. 2009). Despite these variations in tCr concentration, and despite the fact that tCr has been found to differ in schizophrenia (Ongur et al. 2009) it is still often taken as internal reference for 1H-MRS metabolite quantification.
Although there are still discussions on the exact role of NAA (Clark 1998; Barker 2001), it is found mainly around mature neuronal cell bodies, axons, and dendrites within the central nervous system, not in glia (Simmons et al. 1991; Urenjak et al. 1993) and with highest concentrations in pyramidal glutamatergic neurons (Moffett and Namboodiri 1995). NAA contributes 75–85% to the total NAA (tNAA) signal with a main peak at 2.01 ppm and a smaller peak around 2.6 ppm, which is only observed with short TE. The other 15–25% of the tNAA signal is accounted for by N-acetylaspartylglutamate, with highest concentrations in parietal and occipital WM (Pouwels and Frahm 1998; Hammen and Stefan 2004).
Cho's peak at 3.22 ppm is comprised of various signals including Cho, phosphocholine, and glycerophosphocholine. The latter two have been ascribed the most important contribution to the signal. The signal is assumed to be predominantly related to phospholipid membrane turnover and metabolism, as seen in myelination or inflammation and related diseases (Bertolino and Weinberger 1999; Hammen and Stefan 2004).