Examples of 1H-MRS voxel placement in the left hippocampus (2.7 × 1.5 × 1 cm) are shown on the right; the images are displayed per radiological convention (ie, the right side of the image is the participant's left side). On the left side are 2 examples of an 1H-MRS spectrum. The black line is a spectrum (640 averages) obtained from the left hippocampus voxel, and the red line is an overlay of the spectral fit. Cho indicates choline; Cr, creatine; Glx, glutamate and glutamine; and NAA, N-acetylaspartate.
A, Hippocampal volumetric deficits in patients with schizophrenia are compared with those from healthy controls. One significant cluster was found using a small-volume correction for the left hippocampus. The single cluster shown is located in the dentate gyrus extending posterolaterally to the cornu ammonis and parahippocampal gyrus, with maximum intensity peak (t = 4.47; familywise error [FWE]–corrected P = .03; cluster extent [ie, number of voxels] kE = 46; Montreal Neurological Institute [MNI] coordinates: x = −18, y = −30, z = −9). The data were corrected for multiple comparisons using FWE-corrected P < .05. The cluster was overlaid on the MRIcron ch2bet template. B, A negative correlation is seen between the glutamate and glutamine to creatine ratio (Glx/Cr) and the left hippocampal VBM measure in patients with schizophrenia. Two significant clusters were found using a small-volume correction for the left hippocampus. The first cluster (top row) had a maximum intensity peak located in the dentate gyrus (t = 4.81; FWE-corrected P = .02; kE = 110; MNI coordinates: x = −26, y = −32, z = −4). The second cluster (bottom row) had a maximum intensity peak located in the cornu ammonis (t = 4.96; FWE-corrected P = .04; kE = 110; MNI coordinates: x = −36, y = −15, z = −16). The 2 clusters do not overlap. The data were corrected for multiple comparisons using FEW-corrected P < .05. Clusters were overlaid on the MRIcron ch2bet template. C, Scatterplot for the relationship between Glx/Cr and volumetric beta weight extractions from 1-sample t tests of VBM in patients with schizophrenia, controlling for age, smoking status (packs per day), and total intracranial volume (r = −0.35). The numbers adjacent to the slices indicate x, y, and z coordinates in MNI convention for sagittal, coronal, and axial slices, respectively. The color bar indicates the t values.
The horizontal lines represent the mean value. Cr indicates creatine; Glx, glutamate and glutamine.
For each group, the solid line represents the linear regression fit across all participants, and the dashed lines represent the 95% confidence bands for the regression line. Data were controlled for age, smoking status (packs per day), and voxel gray and white matter content (healthy controls: r = 0.39, P = .04; patients with schizophrenia: r = −0.19, P = .21). Cr indicates creatine; Glx, glutamate and glutamine; and NAA, N-acetylaspartate.
eTable 1. Demographics and clinical measures
eTable 2. Hippocampal neurometabolites in subjects with CRLB <20%
eTable 3. Hippocampal neurometabolites in chronic schizophrenia, medication naïve patients and first episode patients
eFigure 1. Glutamate phantom spectrum
eFigure 2. Glx/Cr ratios showing individual measurements and means (horizontal lines) in the hippocampus in patients with schizophrenia and healthy controls
Customize your JAMA Network experience by selecting one or more topics from the list below.
Kraguljac NV, White DM, Reid MA, Lahti AC. Increased Hippocampal Glutamate and Volumetric Deficits in Unmedicated Patients With Schizophrenia. JAMA Psychiatry. 2013;70(12):1294–1302. doi:10.1001/jamapsychiatry.2013.2437
Alterations in glutamatergic neurotransmission have been postulated to be a key pathophysiologic mechanism in schizophrenia.
To evaluate hippocampal volumetric measures and neurometabolites in unmedicated patients with schizophrenia and the correlations between these markers. Our a priori hypothesis was that glutamate levels would negatively correlate with hippocampal volume in schizophrenia.
Design, Setting, and Participants
Combined 3-T structural magnetic resonance imaging and single-voxel proton magnetic resonance spectroscopy study at the Department of Psychiatry and Behavioral Neurobiology, University of Alabama at Birmingham, of 27 unmedicated patients with schizophrenia and 27 healthy controls.
Main Outcomes and Measures
Hippocampal volumetric measures and neurometabolites, and the correlations between volumetric measurements and neurometabolites.
Hippocampal volumetric deficits, increased ratios of hippocampal glutamate and glutamine to creatine (Glx/Cr), and a loss of correlation between hippocampal N-acetylaspartate (NAA)/Cr and Glx/Cr in patients with schizophrenia were found. Significant correlations between hippocampal volumetric measures and Glx/Cr were also found in patients with schizophrenia but not healthy controls.
Conclusions and Relevance
Our findings support the theory that alterations in hippocampal glutamate levels potentially account for structural deficits in the hippocampus observed in schizophrenia neuroimaging studies.
The glutamate system is complex; it includes multiple modulating cotransmitters, interacting receptors, and multisystem synapses. The regional characteristics of neuronal architecture, local circuitry, and signaling mechanisms suggest that important aspects of glutamate metabolism are modulated at a local level.1 Imbalances in glutamatergic neurotransmission and metabolism are proposed as key pathomechanisms in schizophrenia.1,2N-methyl-d-aspartate receptor hypofunction on γ-aminobutyric acid neurons could result in disinhibition of postsynaptic glutamate neurons, excessive glutamate release, and neurotoxicity. Although excitotoxicity may not be the cause of schizophrenia, it could contribute to the course of the disorder (see Konradi and Heckers3).
Decreased hippocampal volume is one of the most robust findings in schizophrenia. It is present in first-episode patients and appears to modestly progress in chronic disease.4 Although no change in the number of hippocampal neurons is found,5,6 alterations in the packing density of neurons,7,8 decreased cross-sectional area of neuronal cell bodies,9,10 altered pyramidal cell shapes,11 and decreased dendritic density12 are reported. Olney and Farber13 suggested that alterations in hippocampal glutamatergic neurotransmission potentially account for histological abnormalities and are the basis for the structural deficits observed in imaging studies.
Proton magnetic resonance spectroscopy (1H-MRS) allows for in vivo measurements of neurometabolites, such as glutamate and N-acetylaspartate (NAA), a putative marker of neuronal integrity.14,15 Proton magnetic resonance spectroscopy studies of the hippocampus in patients with schizophrenia have documented reductions in levels of NAA.16,17 Glutamate measurements have been less common because of difficulties in quantifying this metabolite at low magnetic fields, and results have varied.18-21 In healthy participants, studies have reported on a positive correlation between glutamate and NAA,22,23 which is perhaps not surprising given that glutamate and NAA metabolisms are linked via the glutamate/glutamine cycle and tricarboxylic acid cycle.14 We previously compared hippocampus metabolites with anterior cingulate cortex metabolites in a large group of medicated patients with schizophrenia and healthy controls. We replicated the finding of positive correlation between glutamate and glutamine (Glx) and creatine (Cr) (ie, the Glx/Cr ratio) and between NAA and Cr (the NAA/Cr ratio) in healthy controls. In schizophrenia, this correlation was absent only in the hippocampus.24 We postulated the ratio between Glx/Cr and NAA/Cr in the hippocampus to be a potential biomarker for schizophrenia that captures the pathophysiology of schizophrenia better than either neurometabolite alone.24
The goal of our study was to examine volumetric and neurometabolite alterations in the hippocampus in patients with schizophrenia. We performed structural magnetic resonance imaging and 1H-MRS for 27 unmedicated patients with schizophrenia and 27 healthy controls. Based on our own data24 and the existing literature,4 we tested the following hypotheses: (1) hippocampal volumetric deficits are present in patients with schizophrenia, (2) hippocampal Glx/Cr is altered in patients with schizophrenia, (3) the correlation between Glx/Cr and NAA/Cr is absent in patients with schizophrenia, and (4) Glx/Cr and Glx/Cr:NAA/Cr correlate with decreased hippocampal volumetric measures in patients with schizophrenia.
Patients with schizophrenia were recruited from the emergency department and outpatient psychiatry clinics at the University of Alabama at Birmingham. Healthy controls, matched on age, sex, and parental occupation, were recruited through advertisements. Approval by the local institutional review board was obtained. Each patient with schizophrenia completed an Evaluation to Sign Consent form before written informed consent was obtained.25
Diagnoses were established by a review of the medical records and by the consensus of 2 clinicians, and then confirmed with the Diagnostic Interview for Genetic Studies.26 Exclusion criteria were major medical/neurologic conditions, substance abuse within 6 months prior to enrollment, pregnancy, and head injury with loss of consciousness. Healthy controls could not have a first-degree relative with an Axis I disorder.
In this analysis, we included 27 patients with schizophrenia who were unmedicated for at least 2 weeks prior to enrollment and 27 healthy controls. All participants were recruited for a prospective study exploring imaging markers of treatment response to risperidone. The neurometabolite measurements of some of the healthy controls have been reported earlier.24,27
The Brief Psychiatric Rating Scale (BPRS)28 and its positive and negative subscales were used to assess symptom severity. Cognitive function was characterized using the Repeatable Battery for the Assessment of Neuropsychological Status (RBANS).29
Imaging was performed on a 3-T head-only scanner (Magnetom Allegra; Siemens Medical Solutions), equipped with a circularly polarized transmit/receive head coil. A high-resolution structural scan was acquired using the 3-dimensional T1-weighted magnetization-prepared rapid acquisition gradient-echo sequence (repetition time [TR]/echo time [TE]/inversion time [TI], 2300/3.93/1100 milliseconds; flip angle, 12°; 256 × 256 matrix; and 1-mm isotropic voxels).
Data were processed using Statistical Parametric Mapping 8 (SPM8; Wellcome Trust Centre for Neuroimaging) in MATLAB version 7.11.0 (Mathworks). Each participant’s T1-weighted magnetization-prepared rapid acquisition gradient-echo image was segmented into gray matter (GM), white matter (WM), cerebrospinal fluid (CSF), skull, and scalp using the “new segment” routine in SPM8.
The DARTEL (diffeomorphic anatomical registration through exponentiated Lie algebra) toolbox for SPM8 was used to improve registration of magnetic resonance imaging scans (www.fil.ion.ucl.ac.uk/~john/misc/VBMclass10.pdf).30 This method uses a subject-specific template that we created by averaging images from 39 patients with schizophrenia and 39 healthy controls. The resulting flow fields created by DARTEL were used to generate native space and spatially normalized (in Montreal Neurological Institute space), modulated (to adjust for volume changes during normalization), resliced (1.5-mm isotropic voxels), and smoothed (6-mm full-width at half maximum) GM and WM images. Using native-space masks, we calculated the total GM, WM, and CSF volumes and the total intracranial volume (which is the sum of the GM, WM, and CSF volumes) by use of a script adapted from John’s SPM Gems (http://www-personal.umich.edu/~nichols/JohnsGems.html) in MATLAB. To localize areas of volumetric deficits, the results were visually inspected by 2 raters (N.V.K. and D.M.W.) and identified by consensus based on the atlas of hippocampal anatomy by Duvernoy.31
A series of sagittal, coronal, and axial T1-weighted anatomical scans (gradient-recalled echo sequence; TR/TE = 250/3.48 milliseconds; flip angle, 70°; 5-mm slice thickness; 1.5-mm gap; and 512 × 512 matrix) were acquired for spectroscopic voxel placement. Data were collected from a voxel in the hippocampus (2.7 × 1.5 × 1 cm) (Figure 1). To control for head tilt, slices were realigned to midline anatomical landmarks in transverse and coronal orthogonal planes. To facilitate prescription of hippocampal volume, axial magnetic resonance images were obtained in an orientation tilted along the long axis of the hippocampus, as viewed in the sagittal images (Figure 1). The voxel was placed in the left hippocampus such that the amount of GM was maximized while avoiding major vessels. Manual shimming was performed to optimize field homogeneity across the voxel. Chemical shift selective pulses were used to suppress the water signal. Water-suppressed spectra were acquired using the point-resolved spectroscopy sequence (with TR/TE = 2000/80 milliseconds to optimize the Glx signal and minimize macromolecule contribution, a spectral bandwidth of 1200 Hz, 1024 points, and 640 averages [21 minutes, 20 seconds]). To limit the effects of nicotine intoxication or withdrawal, patients were allowed, but not encouraged, to smoke up to 1 hour prior to image acquisition.
We determined the volume of GM, WM, and CSF in the 1H-MRS voxels by segmenting participants’ structural scans in SPM8. The 1H-MRS voxel images, created from the 1H-MRS raw data headers, were used to mask each of the tissue classes, and the volumes were calculated in MATLAB. Voxel segmentation was not successful for 2 patients with schizophrenia. The 1H-MRS data were analyzed in jMRUI version 3.0. The residual water peak was removed using the Hankel-Lanczos singular values decomposition filter. Spectra were quantified in the time domain using the AMARES (advanced method for accurate, robust, and efficient spectral fitting) algorithm.32 Prior knowledge derived from in vitro and in vivo metabolite spectra was included in the model. A phantom solution of 20mM glutamate in buffer (30mM sodium hydrogen carbonate and 30mM sodium carbonate; pH, 7.1) was imaged using the parameters from the in vivo study. The resulting spectrum was quantified and used to fit the in vivo data (eFigure 1 in Supplement). The AMARES prior knowledge model consisted of peaks for NAA, choline (Cho), Cr, and the C4 resonance of glutamate (Glu). Glutamate was modeled as triplet (large peak with 2 small outer wings). The amplitudes of NAA, Cho, Cr, and the center Glu peak were estimated by use of the algorithm. The amplitudes of the Glu outer wings were set at a fixed ratio relative to the center peak based on the Glu phantom spectrum. The relative phases of NAA, Cho, Cr, and the center Glu peak were fixed at 0, and the phases of the Glu outer wings were fixed at the phase determined from the Glu phantom spectrum. The line width of NAA was estimated by the algorithm, and the line widths of the remaining peaks were set to be equal to that of NAA. The frequencies of NAA, Cho, Cr, and the center Glu peak were estimated by use of AMARES. The Glu outer wing frequencies were set at fixed shifts relative to the center peak based on the Glu phantom spectrum. All peak shapes were fixed at Lorentzian. Cramér-Rao lower bounds, used as an estimate of uncertainty, were calculated for each peak. The exclusion criteria were (1) a line width of the magnitude signal during manual shimming of greater than 25 Hz at full-width at half maximum; (2) a Cramér-Rao lower bound of greater than 25%, and (3) failure of the fitting algorithm. No spectra were excluded based on these criteria, and NAA, Cho, and Glx were quantified with respect to Cr. As previously reported, the coefficients of variability of the MRS measurements for 1 healthy control were 2.87% (NAA/Cr), 11.93% (Glx/Cr), and 2.15% (Cho/Cr).27
Statistical analyses were performed using SPSS 12.0 for Microsoft Windows (SPSS Inc). For analyses of group differences in demographic and clinical variables, independent t tests and χ2 analyses were performed. Differences in metabolites were investigated with multivariate analyses of covariance, using disease state as the between-group factor and age, smoking status, and voxel GM and WM content as covariates. Planned contrasts were created to assess group differences in individual metabolites. P < .05 was considered statistically significant.
Statistical parametric maps of GM content within the hippocampus mask for patients with schizophrenia were compared with those for healthy controls using an independent-sample t test. To evaluate correlations of GM content within the hippocampus mask and neurometabolites, we performed regression analyses, corrected for age, smoking status (packs per day), and total intracranial volume in each group separately. Using the results from this statistical parametric map (P < .001), we extracted the first eigenvariate from the residual of GM, controlling for age, smoking status (packs per day), and total intracranial volume for each individual. The eigenvariates were used to correlate with Glx/Cr and to create a scatterplot, allowing for inspection of variance between these variables. In an exploratory analysis, we performed regression for both RBANS and BPRS values on hippocampal volumes and neurometabolites. All statistical parametric mapping analyses were controlled for multiple comparisons by implementing a small-volume correction, confined to the left hippocampus from the Automated Anatomical Labeling atlas implemented in the Wake Forest University PickAtlas software version 2,33,34 and corrected for multiple comparisons using a familywise error (FWE)–corrected P < .05.
Groups did not differ with regard to sex, age, or parental occupation. More patients with schizophrenia than healthy controls were smokers. Healthy controls scored significantly higher on than RBANS than did patients with schizophrenia. Of the 27 patients with schizophrenia, 16 had previously been treated with antipsychotics. The remaining 11 patients had no prior exposure to antipsychotics; 5 were first-episode patients, and the others had not received prior treatment for a prolonged period after illness onset (eTable 1 in Supplement).
Analyses revealed 1 cluster of decreased hippocampal voxel-based morphometry (VBM) measure in patients with schizophrenia (t = 4.47; P = .03; cluster extent [ie, number of voxels] kE = 46; Montreal Neurological Institute coordinates: x = −18, y = −30, z = −9). The area of maximum deficit was mapped to the dentate gyrus extending posterolaterally to the cornu ammonis (CA) and parahippocampal gyrus (Figure 2A).
There were no group differences in GM, WM, or CSF content in the MRS voxel (GM: F52 = 1.350, P = .25; WM: F52 = 0.489, P = .49; CSF: F52 = 1.232, P = .27). The mean (SD) signal to noise ratio of spectra did not differ between groups: 11.77 (1.08) for healthy controls and 11.39 (1.77) for patients with schizophrenia (t = 0.966, P = .34). However, the mean (SD) full-width at half maximum was higher for patients with schizophrenia, 8.86 (1.99) Hz, than for healthy controls, 7.93 (1.16) Hz (t = −2.13, P = .04).
The results of overall multivariate analyses of covariance for neurometabolites were significant at the trend level (Wilks λ = 0.853; F = 2.351; P = .09). Planned contrasts showed no group differences in NAA/Cr, Glx/Cr was elevated in patients with schizophrenia (P = .02) (Figure 3; eFigure 2 in Supplement), and Glx/Cr:NAA/Cr was increased (P < .01) (Table; eTable 2 in Supplement). We identified correlations between Glx/Cr and NAA/Cr in healthy controls (r = 0.39, P = .04) but not in patients with schizophrenia (r = −0.19, P = .21) (Figure 4). Illness duration did not correlate with neurometabolite levels (NAA/Cr: r = 0.38, P = .14; Glx/Cr: r = −0.34, P = .17; Cho/Cr: r = 0.46, P = .09) or hippocampal VBM measure (P > .05). In previously medicated patients, the interval since last antipsychotic treatment did not correlate with neurometabolites (NAA/Cr: r = 0.33, P = .17; Glx/Cr: r = −0.12, P = .38; Cho/Cr: r = 0.26, P = .24). There was no difference in hippocampal VBM measure (P > .05) or neurometabolites between previously medicated and medication-naive patients or between first-episode and chronic patients (eTable 3 in Supplement).
We found 2 clusters of negative correlation between hippocampal VBM measure and Glx/Cr in patients with schizophrenia (t = 4.81; FWE-corrected P = .02; kE = 110; Montreal Neurological Institute coordinates: x = −26, y = −32, z = −4 vs t = 4.96; FWE-corrected P = .04; kE = 110; Montreal Neurological Institute coordinates: x = −36, y = −15, z = −16) but not in healthy controls (Figure 2B and C); NAA/Cr did not correlate with the hippocampal VBM measure in patients with schizophrenia or healthy controls (P > .05). Unexpectedly, we did not find correlations between Glx/Cr:NAA/Cr and the hippocampal VBM measure in either group.
Neither the BPRS total nor the positive and negative subscales correlated with VBM measures in patients with schizophrenia (FWE-corrected P > .05). We did not find correlations between VBM and RBANS total score in either group (FWE-corrected P > .05).
We did not find correlations in patients with schizophrenia between Glx/Cr and BPRS (BPRS total score: r = 0.18, P = .46; BPRS positive subscales: r = −0.08, P = .73; BPRS negative subscales: r = 0.25, P = .29) or NAA/Cr and BPRS (BPRS total score: r = 0.25, P = .29; BPRS positive subscales: r = 0.05, P = .82; BPRS negative subscales: r = −0.3; P = .90). The RBANS total scores did not correlate with Glx/Cr (healthy controls: r = 0.29; P = .19; patients with schizophrenia: r = 0.35, P = .15) or NAA/Cr (healthy controls: r = −0.23, P = .57; patients with schizophrenia: r = −0.03, P = .90).
The findings of our study include (1) GM volumetric deficits in the hippocampus, (2) increased Glx/Cr but no difference in NAA/Cr, (3) loss of correlation between Glx/Cr and NAA/Cr, and (4) negative correlations between Glx/Cr and hippocampal VBM measure in unmedicated patients with schizophrenia. To our knowledge, we are the first to report increased Glx/Cr and a possible link between hippocampal glutamate abnormalities and volumetric deficits in a relatively large group of unmedicated patients.
A decreased hippocampal volume is present in first-episode patients and appears to modestly progress in chronic disease.4 Voxel-based morphometry studies showed GM deficits in the medial temporal lobe in more than 50% of studies,35 with total hippocampal volume deficits of 8%.4 Although these findings are widely replicated, regional specificity has been less explored. The volumetric deficits in our study were most prominent in the dentate gyrus with posterolateral extensions to the CA and parahippocampal gyrus. Similar to our study, the studies by Narr et al36,37 reported pronounced posterior hippocampal volume reductions in 2 groups of patients with schizophrenia and their unaffected twins. In first-episode patients, Narr et al38 reported volume loss in the anterior, but not the posterior, hippocampus, as well as perihippocampal CSF increases surrounding anterolateral and midbody hippocampal regions. Anterior volume reductions without posterior hippocampal volume loss in male patients with chronic schizophrenia have also been reported.39
We found increased hippocampal Glx/Cr in unmedicated patients with schizophrenia but no alterations in NAA/Cr. Only one study21 explored NAA and Glx in the medial temporal lobe in a heterogeneous group of 34 patients with first-episode psychosis that included affective psychoses (some medicated patients and some medication-naive patients). Wood et al21 reported no differences in NAA or Glx levels among these patients when compared with healthy controls. Although the unaltered NAA levels in the study by Wood et al21 are in agreement with our results, the differences in the findings for Glx between their study and ours may be due to differences in patient populations or sample size.
Examining Glx in other areas of the brain, Kegeles et al40 reported elevated Glx levels in the medial prefrontal cortex in 16 unmedicated patients with schizophrenia, and de la Fuente-Sandoval et al41 reported the same in the dorsal caudate of 18 antipsychotic-naive patients. However, Tibbo et al42 did not find alterations in the glutamate levels of unmedicated patients with first-episode psychosis.
Few studies reported hippocampal glutamate measurements in medicated patients. Van Elst et al19 found higher glutamate levels in 21 medicated, acutely psychotic, hospitalized patients. Lutkenhoff et al20 failed to identify such differences in 9 patients with schizophrenia or their unaffected co-twins, and Olbrich et al18 did not detect a change in glutamate levels in 9 medicated, first-episode patients. Similarly, we did not find elevated Glx/Cr in 46 stable, medicated patients with schizophrenia.24
Taken together, elevated levels of glutamate in patients with schizophrenia may be present in acute psychosis but can possibly be corrected if successfully treated with antipsychotics, thus underscoring the importance of continuity of treatment. Increases in hippocampal regional cerebral blood flow, hypothesized to be secondary to elevated synaptic glutamate levels, have been reported to be associated with psychosis, indirectly suggesting a link between abnormal glutamatergic neurotransmission and clinical presentation.43 Assuming elevated Glx/Cr to be excitotoxic, it is unclear why progressive hippocampal volume loss is not more pronounced over the course of the illness or to what extent antipsychotic medications may counteract the progression of volume loss. Our assumptions are rather speculative and need more definite, prospective, longitudinal structural and neurometabolite assessments over the course of antipsychotic treatment before conclusions can be drawn.
Some previous studies have identified decreased hippocampal NAA levels in patients with schizophrenia,44-47 but some have not.18,20 A recent meta-analysis48 showed significantly decreased levels of NAA. Our lack of detecting abnormalities in NAA may be due to lack of power.
Correlations between glutamate and NAA have consistently been reported in healthy individuals.22,23 As in our previous study,24 Glx/Cr and NAA/Cr were correlated in healthy controls. In addition, we replicated our finding of a loss of correlation between Glx/Cr and NAA/Cr in the hippocampus of patients with schizophrenia,24 this time in unmedicated patients. Correlation coefficients between metabolites were similar in both our present study and our previous study24 (healthy controls: r = 0.39 vs r = 0.40; patients with schizophrenia: r = −0.19 vs r = −0.02). The replication of this finding in unmedicated patients is important because it shows that it is not purely a phenomenon related to antipsychotic medications. It also confirms that it is not merely a reflection of illness severity because we found it in both stable and acutely ill patients. In an attempt to quantify the extent of loss of correlation, we calculated Glx/Cr:NAA/Cr to explore their potential as a biomarker. Group differences were more robust than with either neurometabolite alone, indicating Glx/Cr:NAA/Cr may be a more sensitive marker differentiating between groups than either Glx/Cr or NAA/Cr alone. In a similar approach, Savic et al49 reported that, in comparison with NAA/Cr, Glx/NAA was superior in correctly identifying epileptic regions in neocortical epilepsy. Although Savic et al49 did not explicitly explore correlations between Glx/Cr and NAA/Cr, their findings suggest that a loss of correlation between metabolites could be present in neocortical epilepsy, implying that it may not be specific for schizophrenia. Further studies will be helpful to clarify illness specificity.
Olney and Farber13 suggested that subtle alterations in hippocampal glutamatergic neurotransmission potentially account for histological abnormalities that cause observed structural deficits. We found a negative correlation between Glx/Cr and hippocampal VBM measure in patients with schizophrenia, suggesting that glutamatergic excitotoxicity may be associated with hippocampal volumetric deficits. Interestingly, though, VBM measures that negatively correlate with Glx/Cr were more extensive than VBM measures that were significantly decreased in patients with schizophrenia when compared with healthy controls. Few studies combined structural magnetic resonance imaging and 1H-MRS to explore connections between hippocampal volume and neurometabolites. With NAA being an assumed marker of neuronal integrity, a positive correlation between hippocampal volume and NAA levels would be expected. We did not find correlations between NAA/Cr and hippocampal VBM measure in patients with schizophrenia or healthy controls. Like us, most investigators have failed to identify a correlation between NAA and hippocampal volume, suggesting that volume deficits and NAA reductions are independent surrogate markers in schizophrenia.50-53
Klär et al46 support this assumption by reporting negative correlations between hippocampal volume and NAA. They state that their finding is not congruent with the view of NAA as a marker for neuronal integrity, and they hypothesize that NAA may mediate the synthesis pathways of glutamatergic neurotransmission via N-acetylaspartate-glutamate.46 Contrary to our hypothesis, we did not find a correlation between Glx/Cr:NAA/Cr and the hippocampal VBM measure; Glx/Cr:NAA/Cr may measure subtle alterations in glutamatergic neurotransmission that do not lead to structural deficits in the hippocampus.
Our results must be interpreted in the context of several limitations. Association does not necessarily imply a causative relationship. Also, we planned to match participants on sex, age, and smoking status, factors that are reported to influence hippocampal neurometabolites.54,55 Matching on smoking status was incomplete; therefore, we used smoking status as a covariate in our analyses.
To investigate volumetric alterations, we chose a VBM approach using the DARTEL algorithm, which is more sensitive and more able to identify more localized morphologic alterations than conventional VBM.56 Although this method does not allow for the calculation of absolute volumes in native space, it was shown to be of equal sensitivity in detecting group differences in the hippocampus when compared with region-of-interest analyses with manual tracings for patients with schizophrenia, depression, or Alzheimer dementia.56-59 In addition, it has the advantage that it is not operator dependent and is less time intensive than manual tracings. Areas of volumetric deficits were identified by visual inspection, and given the spatial resolution, we did not narrow findings to CA subfields.
We acknowledge that this method is not ideal. However, we believe that it is informative nonetheless, especially given that the dentate gyrus has been implicated as a key hippocampal structure affected by altered glutamatergic neurotransmission in schizophrenia.1 To investigate neurometabolites, we used 1H-MRS. Test-retest reliability in healthy controls was comparable to previous reports,60 but we were unable to estimate measures in unmedicated patients with schizophrenia because of the need to medicate them as soon as possible.
We recognize that the remaining uncertainty about the retest reliability in our sample is a potential confounding factor, as is the higher full-width at half maximum in patients with schizophrenia. Furthermore, we calculated neurometabolites normalized to Cr; data acquisition parameters did not allow for the calculation of absolute metabolite levels. Few studies reported alterations in Cr levels in schizophrenia, which thus questions the validity of referencing neurometabolites to Cr.61,62
In a meta-analytic approach, we did not detect significant Cr abnormalities across studies in the hippocampus in patients with schizophrenia, although heterogeneity between studies was significant.48 Although we cannot rule out that Cr may be altered in this sample, we conclude that our findings more likely reflect Glx abnormalities than Cr abnormalities. Furthermore, the MRS voxel was not exclusively confined to the hippocampus; metabolites from adjacent structures included in the voxel may have affected the findings. Also, the Glx peak does not reflect synaptic glutamate alone but is a combination of neuronal, glial, and synaptic glutamate, glutamine, and γ-aminobutyric acid in the voxel. It is therefore not possible to equate Glx with glutamatergic neurotransmission.
A technical strength of our study that partially compensates for this is the use of an echo time of 80 milliseconds, which appears to optimize the glutamate signal by diminishing the macromolecule contribution and reducing the glutamine contribution, but elevations in the level of glutamine cannot be ruled out,60 especially since our phantom did not include glutamine.
Future studies explicitly investigating individual glutamate and glutamine peaks at high field strength are needed to clarify how much of the Glx abnormalities reported are contributable to glutamine abnormalities. In summary, our findings show that hippocampal volumetric deficits correlate with elevated Glx/Cr in unmedicated patients with schizophrenia, supporting the theory that an altered hippocampal glutamate level potentially accounts for the structural deficits in the hippocampus observed in neuroimaging studies.
Corresponding Author: Adrienne C. Lahti, MD, Department of Psychiatry and Behavioral Neurobiology, University of Alabama at Birmingham, SC 501, 1530 Third Ave S, Birmingham, AL 35294-0017 (firstname.lastname@example.org).
Published Online: October 9, 2013. doi:10.1001/jamapsychiatry.2013.2437.
Author Contributions: Dr Lahti had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study concept and design: White, Lahti.
Acquisition of data: All authors.
Analysis and interpretation of data: All authors.
Drafting of the manuscript: Kraguljac, White, Lahti.
Critical revision of the manuscript for important intellectual content: White, Reid.
Statistical analysis: Kraguljac, White, Reid.
Obtained funding: Lahti.
Administrative, technical, or material support: White.
Study supervision: White.
Conflict of Interest Disclosures: Over the past 2 years, Dr Lahti has received research funds from the National Institutes of Health (grant RO1 MH081014) and an investigator-initiated research grant from Pfizer. No other disclosures were reported.
Funding/Support: This work was supported by the National Institutes of Health (grant RO1 MH081014).
Role of the Sponsor: The funding agency had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; and preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.
Additional Contributions: We thank Jan den Hollander, PhD, at the University of Alabama at Birmingham, who provided invaluable input in MRS data acquisition methods.