Glutamate Levels Predict Psychosis Symptoms

Glutamatergic dysfunction in schizophrenia and psychotic disorders are gradually being acknowledged [1,2,3,4,5,6], but clinical trials of glutamatergic compounds have overall been negative [7]. However, glutamatergic dysfunction and the relationship with clinical symptoms might change over time, and, thus, glutamatergic compounds may mainly be beneficial at specific illness stages.

Schizophrenia and psychotic disorders typically have a clinical onset around the age of twenty [8]. This overlaps with the period of prefrontal cortex maturation that is dependent on glutamatergic inputs [9, 10], suggesting glutamatergic dysfunction as a contributing cause of schizophrenia [11]. In support, an increasing number of in vivo studies using proton magnetic resonance imaging (1H-MRS) has reported abnormalities in brain levels of glutamate in schizophrenia, and, recently, meta-analyses have been published [1,2,3, 12]. Findings indicate that glutamate levels in anterior cingulate cortex (ACC) and nearby regions are decreased [1, 2, 12] whereas glutamatergic metabolites in the subcortical areas thalamus and basal ganglia are increased when comparing patients in a broad age-range with healthy controls (HC) [2, 3]. Studies of antipsychotic-naïve or minimally-treated first-episode patients with psychosis (FEP), on the other hand, suggest a phase-specific pattern in the early course of illness with increased brain levels of glutamate in either thalamus or ACC in patients having a subsequent poor treatment response [4,5,6, 13]. It is therefore likely that glutamate levels change over the course of illness, but this has only been sparsely investigated. A recent four-year follow-up study that investigated FEP within the first two years of illness onset reported a decrease of glutamate levels over time in ACC but not thalamus in both patients and HC [14]. Antipsychotic treatment also appears to decrease glutamate levels in the subcortical regions thalamus and striatum, but not hippocampus, in initially antipsychotic-naïve patients receiving first-line treatment [4, 6, 15]. In contrast, glutamate levels in ACC seem unaffected after 6 weeks and 6 months treatment [4, 15, 16]. Thus, there is a need for longitudinal studies of initially antipsychotic-naïve patients followed-up after longer-term treatment to gain insight into changes in glutamate levels over time as well as the impact of antipsychotic treatment.

Theories of glutamatergic dysfunction have received particular attention by capturing all the three major symptom dimensions of schizophrenia that is positive and negative symptoms and cognitive deficits [17,18,19]. The theory is mainly based on pharmacological challenge studies, genetic- and post mortem studies [19] and needs further validation in patient groups. A recent meta-analysis of 1H-MRS studies reported greater overall symptom severity in patients with lower frontal and higher basal ganglia glutamate levels, but direct associations with positive and negative symptoms were not addressed [2]. For positive symptoms, individual studies tend to find higher symptomatology in patients with higher levels of glutamatergic metabolites in FEP [6, 20,21,22,23] as well as in the chronic patient populations [24, 25]. For negative symptoms, findings are less consistent [5, 6, 13, 21, 26] but greater negative symptom severity in FEP may be associated with higher glutamate levels [5, 13], whereas the reverse pattern has been observed in the chronic phase [27]. For cognitive deficits, pharmacological challenge studies and preclinical models imply that glutamatergic dysfunction especially affects the domains attention, spatial working memory (SWM), and executive function [28,29,30,31,32]. In support, we previously reported that lower glutamate levels in dACC were associated with impaired performance in tests of attention and spatial working memory but not intelligence in antipsychotic-naïve FEP [33]. Other studies have found that both too high [25, 34,35,36] and too low glutamate levels [37,38,39,40,41] may impair different cognitive domains, but the involvement of glutamate levels in different brain regions on specific cognitive domains as well as the impact of antipsychotics and illness trajectory remains inadequately understood [42].

Last, a range of factors may affect cerebral glutamate levels including age and sex [2, 12, 43], and recent research also points toward the importance of adjustment for measurement quality parameters [1]. Furthermore, abnormalities in other metabolites such as reduced frontal myo-inositol [44] and thalamic N-acetylaspartate (NAA) [45] have also been reported in patients with schizophrenia, but the trajectories of these metabolites over two years in initially antipsychotic-naïve patients have not been published.

To address these unknowns, we assessed glutamate levels in dACC and left thalamus in initially antipsychotic-naïve patients with FEP and HC and followed the participants up over two years to relate brain levels of glutamate with positive and negative symptoms and the cognitive domains attention and SWM.

Based on meta-analyses and our previous studies [1,2,3,4, 33], we tested the primary hypothesis that glutamate levels in dACC would be reduced in FEP whereas thalamic glutamate levels initially would be higher but reduced after two years when compared with HC. Our second hypotheses explored if lower glutamate levels in dACC was related to a higher degree of cognitive deficits, if higher glutamate levels in left thalamus were associated with a higher symptom burden of positive symptoms, and if the relation with negative symptoms was altered over the two years. Last, we explored the effect of sex, age at study inclusion, smoking, and dose of antipsychotic compound on glutamate levels.

Participants and study design

We included 65 antipsychotic-naïve patients with FEP and 57 HC in the Pan European Collaboration on Antipsychotic Naïve Schizophrenia II (PECANSII) study [4, 46] described in detail in the supplementary information. Participants were invited for four visits (baseline, six weeks, six months, and two years) all including a Magnetic Resonance Imaging (MRI) scan, and clinical as well as cognitive assessments as shown in Fig. 1.

Glutamate levels at baseline, after six weeks and six months have been published previously in overlapping samples [4, 33].

The study was approved by the Committee on Biomedical Research Ethics (H-3-2013-149) and all participants provided written informed consent prior to enrollment in the study. All methods in the study were performed in accordance with the relevant guidelines and regulations. Patients were recruited from in- and outpatient units in the Capital Region of Denmark and the diagnosis was assessed with Schedules for Clinical Assessment in Neuropsychiatry [47] according to the ICD-10 criteria for schizophrenia (DF20.x), schizoaffective disorder (DF25.x) or non-organic psychosis (DF22.x, DF28, or DF29). Antipsychotic-naïve and central stimulant-naïve status (lifetime) was reported by the patients and validated by medical records. HC were recruited via online advertisement (www.forsøgsperson.dk). Exclusion criteria are described in the Supplementary Methods.

Clinical and cognitive assessment

The Positive and Negative Syndrome Scale (PANSS) [48] and Global Assessment of Functioning scale [49] (social and occupational functioning score (GAF-F)) were used to assess psychopathology and level of function, respectively. Cognitive function was assessed using selected tasks from the Cambridge Neuropsychological Test Automated Battery [50, 51] that was A’ from the rapid visual information processing (RVP) as a measure of sustained visual attention and the strategy score from the SWM task.

Magnetic spectroscopy acquisition, quantification, and quality control

MRI was acquired on a 3.0T Philips scanner (Achieva, Philips Healthcare, Eindhoven, The Netherlands) using a 32-channel head coil (Invivo, Orlando, Florida, USA). First, a structural T1-weighted scan covering the whole brain was acquired for correct anatomical placement of the spectroscopic voxels and tissue segmentation (TR: 10 ms; TE: 4.6 ms; flip angle: 8°; voxel size: 0.79*0.79*0.80 mm3). Second, two point-resolved spectroscopy (PRESS) sequences were acquired (TR 3000 ms, TE 30 ms, 128 averages with MOIST water-suppression, 7 min per scan) together with an inbuild unsuppressed water reference scan in a 2.0 × 2.0 × 2.0 cm3 voxel in dACC (Brodmann area 24 and 32) and a 2.0 × 1.5 × 2.0 cm3 voxel in left thalamus (Supplementary Fig. S1A, C). The magnetic resonance spectroscopy (MRS) spectra were fitted in the spectral range 0.2–4.0 ppm using LCModel version 6.3–1 L [52] that estimated levels of glutamate, glutamine, glutamate+glutamine (glx), N-acetylaspartate (NAA), total creatine (PCr + Cr), choline, and myo-inositol. Illustrative glutamate spectra are provided in Supplementary Fig. S1B, D for dACC and left thalamus, respectively. Metabolite levels were quantified in Institutional Units (IU) by adjusting the water-scaled metabolite values for partial-volume cerebrospinal fluid, and gray and white matter using the formula previously described [4, 33] and specified in the Supplementary Information together with the quality control procedure, minimum reporting standards for MRS [53] in Supplementary Table S1, and MRS quality measures in Supplementary Tables S2 and S3.

The MRS data in dACC and left thalamus at each visit after quality control are summarized in Fig. 2.

Demographic and clinical characteristics were compared between groups using t-tests, X2, or linear mixed models. SWM strategy scores were log-transformed to conform to normality and multiplied by −1 before statistical analyses so that a higher score indicated better performance.

The trajectory of glutamate and glx over the two years was analyzed using a linear mixed model with group (FEP versus HCs) and time (0, 6, 26, and 96 weeks) as independent variables with adjustment for the nuisance variables age, sex, and smoking status, as well as the scanning quality parameters Cramér-Rao lower bound (CRLB) and Full width at half maximum (FWHM) that differed significantly between FEP and HC (Supplementary Table S2). A significant group*time interaction indicated a different trajectory between groups and was followed-up by post hoc tests at the individual visits. Insignificant interactions were removed, and persistent differences between FEP and HC (main effect of group) or changes over time in both groups (main effect of time) were then evaluated with the p-level set to p < 0.025 to adjust for two comparisons (dACC and left thalamus). Explorative main effects of sex, age of participants at study inclusion, and number of cigarettes per day are reported as well.

Similar mixed models were used to explore if the relationship was changed over the two years between glutamate levels and the positive as well as negative score from PANSS, and with antipsychotic dose.

For cognitive performance, a similar mixed model also including a group*time*cognitive score interaction first evaluated if there was a significant different trajectory of the association between glutamate levels and cognitive performance in FEP compared with HC. This analysis was followed up by separate mixed models for FEP and HC, respectively.

A general linear model was used to investigate associations between glutamate levels and aripiprazole serum levels after six weeks.