Introduction

Psychotic disorders such as schizophrenia (SZ) and bipolar disorder with psychotic features (BD) are characterized by hallucinations, disorganized thought, impaired emotional and motivational processes, and cognitive dysfunction. Researchers have sought to identify the pathophysiology of these disorders for decades without great success. The development of novel neuroimaging techniques holds promise for the identification of biomarkers for psychosis based on brain function, structure, and metabolites. One promising direction of research has to do with the “disconnection syndrome” observed in psychotic disorders and characterized by abnormal within-network connectivity and between-network anticorrelation [1]. The presence of a default mode network (DMN), including the medial prefrontal cortex (MPFC), posterior cingulate cortex (PCC), and bilateral inferior parietal lobules (IPL), is one of the most reliable observations from neuroimaging studies. The DMN is active during internally oriented processing and is downregulated during task-related processing [2]. Failure to suppress DMN activity during tasks in various psychiatric disorders has been observed [3], as well as a reduced anti-correlation between DMN and task-related brain networks at rest in schizophrenia [4] and an opposite trend in major depressive disorder [5]. However, the cellular and molecular mechanisms underlying this phenomenon are poorly understood.

At the cellular level, neuronal activity is regulated by multiple neurochemical processes including cycling of glutamate (Glu) and GABA, the major excitatory and inhibitory neurotransmitters in the central nervous system. There has been considerable interest in the role of glutamatergic neurotransmission in the pathophysiology of psychosis [6]. Determining the nature of glutamatergic abnormalities in psychosis may prove useful in understanding symptomatology and in developing novel treatment strategies. Combined functional MRI (fMRI) and magnetic resonance spectroscopy (MRS)studies have demonstrated a role for regional neurotransmitter modulation of brain functional activity and connectivity in multiple brain regions [7, 8]. In previous studies, we found that neurotransmitter concentrations in the posterior cingulate cortex (PCC), a key component of the DMN, modulate DMN deactivation induced by an n-back working memory task and this association increases with cognitive load [9] and investigated the resting-state functional connectivity (rsFC) of DMN and task positive networks (TPNs) [10]. We studied the triple relationship between PCC neurotransmitters, resting-state FC and DMN deactivation during working memory tasks and found that the internetwork resting-state FC may mediate the relationship between neurotransmitters representing excitation-inhibition balance and task induced DMN deactivation. More recently, we measured Glu and GABA from medial prefrontal cortex (MPFC), another component of DMN, and dorsolateral prefrontal cortex (DLPFC), a critical node in frontoparietal executive control network (CN), as well as the functional connectivity between them under resting state and during Sternberg task, an alternative working memory paradigm [11]. We confirmed a relationship between MPFC neurotransmitters, FC between DMN-CN, and task-evoked DMN deactivation. On the other hand, it is striking that DLPFC neurotransmitter levels did not show significant association with functional MRI measures. These findings imply that MPFC and DLPFC transmission have an asymmetric relationship with task-related activation and inter-network functional connectivity, with MPFC neurotransmission playing a more important role. Recent studies by other groups have reported disrupted correlations between neurotransmitter levels and brain BOLD signals. The associations between cortical neurotransmitters and BOLD activations evoked by the Stroop [12] and working memory [13] tasks were found to be altered in SZ. A few studies have compared the relationship of resting-state connectivities in the ACC [14,15,16] and hippocampus [17] and neurotransmitter levels in SZ and HC and found alterations of this relationship in both regions. In another work, we have identified a relationship between brain energy metabolism and functional connectivity in psychotic disorders [18]. Specifically, we discovered significant correlations between MPFC energy metabolism rate and its functional connectivity with long-distance and large-scale brain networks in healthy controls, and a breakdown of this relationship in individuals with psychotic disorders.

Our findings regarding neurochemical characteristics of DMN have several important implications for DMN dysfunction in psychiatric disorders. In the current report, we extended our previous work [11] to first episode psychosis (FEP) patients and their first-degree unaffected relatives (siblings). Working with FEP patients, we can obviate long-term medication effects and the toxic effects of chronic psychosis. In parallel with our previous work [10, 11], we focused on the measurements of MPFC neurotransmitters and DMN (MPFC)-CN (DLPFC) rsFC. We hypothesized that abnormal excitatory and inhibitory neurotransmission would be associated with the well-known rsFC abnormalities in FEP.

Methods

Participants

43 subjects, including 15 FEP patients, 10 relatives (RL) and 18 healthy controls (HC), were recruited in the current study. 15 patients with all DSM-5 non-affective psychotic disorder diagnoses were allowed. Participants were men and women between the ages of 18–30 under the care of McLean OnTrack, an outpatient specialty program for those with new onset of psychosis within the past 12 months and with at most once hospitalization. At this stage, patients are minimally-treated during acute stabilization but have not been on long-term pharmacotherapy. We excluded first generation antipsychotic medications, clozapine, as well as NMDA-related medications such as memantine and amantadine. We did not exclude GABA-ergic medications since benzodiazepines are so commonly used in the care of FEP patients and it would have limited recruitment even further if we had excluded these medications. Medication doses were stable for 12 weeks at the time of the study. Patients who meet criteria for a DSM-5 substance use disorder, including tobacco and alcohol in the past 3 months were excluded. 10 first-degree unaffected relatives, all siblings, both men and women with no DSM diagnoses of any psychotic disorder (schizophrenia spectrum or bipolar disorder with psychotic features), psychiatric hospitalizations, nor psychoactive medication were enrolled. 18 control participants with no history of medical, neurological, psychiatric, substance use problems, and not taking medications and no history of the same in first degree relatives were also recruited from the community. Exclusion criteria for patients included history of head injury with loss of consciousness, seizure disorder, and MRI contraindications. Exclusion criteria for controls included the same as well personal history of psychiatric diagnosis or treatment, and first degree relative with a history of a psychotic disorder. All procedures were approved by the Mass General Brigham IRB and all participants provided written informed consent.

Clinical assessment

Clinical and community functioning were assessed using the Positive and Negative Syndrome Scale (PANSS), the Young Mania Rating Scale (YMRS), the Montgomery-Asberg Depression Rating Scale (MADRS) and the Multnomah Community Ability Scale (MCAS). The MCAS measures functioning in multiple domains including social interest/effectiveness, independence in daily living, and instrumental role functioning. The North American Adult Reading Test (NAART) was used as a measure of premorbid IQ. Information regarding current psychiatric medication use was collected by patient report, and chlorpromazine (CPZ) equivalents were calculated according to [19].

MRI

The following MRS and fMRI scans were performed on a Siemens 3 T Trio scanner using a 32-channel volume coil: a high-resolution anatomical scan (5 min), resting-state fMRI (6 min × 2 runs), followed by MRS (20 min for localized shimming, MEGA-PRESS and semi-LASER). Resting-state functional connectivity MPFC and DLPFC ROIs, MRS voxels, and representative MR spectra of semi-LASER and MEGA-PRESS of MPFC were presented in Fig. 1.

Fig. 1: Study overview.
figure 1

Resting-state functional connectivity MPFC and DLPFC ROIs (a) MRS voxels (b) and representative MR spectra of semi-LASER (c) and MEGA-PRESS (d) of MPFC.

Full size image

Single-voxel proton MRS was acquired at MPFC (30 × 20 × 30 mm3). A MEGA-PRESS [20, 21] optimized in house for GABA detection [22] was acquired with the following parameters: TE/TR = 68/3000 ms and total 192 averages (scan time = 10 min); the editing pulses applied alternatively at frequency of 1.9 or 7.5 ppm interleaved with the averages. A semi-LASER sequence [23, 24] with optimized TE = 28 ms [25] and TR = 3 s and 64 averages (scan time = 4 min) was used to measure Glu and Gln from the same regions as MEGA-PRESS. Fastmap shimming [26] was performed before MRS scans to ensure the full widths at half maxima (FWHM) of water resonance <12 Hz. A VAPOR (VAriable pulse Power and Optimized Relaxation delays) module [27] was utilized in both sequences to achieve water suppression. Spectra with unsuppressed water were also acquired for eddy-current correction and quantification reference.

Resting-state fMRI was collected using a single-shot gradient-echo echo-planar imaging sequence. The imaging parameters are: TR/TE = 2.5 s/30 ms; FA = 78°; slice thickness/gap = 4/0 mm; 39 slices; FOV = 220 × 220 mm2 (3.44 × 3.44 mm2 in-plane resolution). Two runs of resting-state fMRI were performed, with 124 time points each run for 5 min. The subjects were instructed to keep the eyes open and think of nothing in particular during the scan.

The anatomical data were acquired for structural reference using a MP-RAGE sequence (256 × 256 × 122 matrix size; 1 × 1 × 1.28 mm3 spatial resolution; TI/TR/TE = 1100/1600/2.25 ms; flip angle = 12°).

We also collected DTI data in the same scans. DTI data were acquired by axial, single-shot spin echo EPI sequence (TR = 6230 ms; TE = 84 ms; FA = 90°; FOV = 1540 × 1540 mm2; matrix = 1120 × 1120; slice thickness=3 mm; slice interval=3 mm; consecutive scanning of 45 slices). One control image without diffusion weighting and 6 images with noncolinear diffusion directions at b = 1000 sec/mm2 were acquired.

Processing

The processing of MRS and rsfMRI followed the previous strategies described in [11]. Briefly, MRS averages were saved individually for all spectra and phase and frequency corrections were applied using FID-A [28]. Quantifications of metabolite concentrations were performed with LCModel [29] with eddy current correction using water as reference. Home-simulated basis spectra of metabolites were used in both MRS sequences. An experimental acquired macromolecule (MM) basis set was used for semi-LASER. The GABA signal measured by MEGA-PRESS contained a MM co-editing contribution which was removed using the MM basis set [30]. The partial volume effect was corrected with the same strategy as in [11]. We also calculated the Glu and GABA concentrations without normalization of gray and white matter (Fig. S2). To confirm the measurement consistency between MEGA-PRESS and semi-LASER, we also performed the correlation between Glx measured by semi-LASER and MEGA-PRESS. The two measures well agree with each other with r = 0.81, p < 0.001 (Fig. S1).

Preprocessing of the rsfMRI data was carried out in DPARSFA [31] and included removing first 5 time points, slice-timing correction, realignment, head motion correction, Gaussian spatial smoothing (FWHM = 4 mm), linear detrending, and spatial normalization to MNI space with a resampling resolution of 3 × 3 × 3 mm3. Nuisance signal regressions and band-pass temporal filtering (passband: 0.01–0.1 Hz) were also performed. The nuisance regressors included Friston’s 24 head motion parameters (6 head motion parameters, 6 head motion parameters one time point before, and the 12 corresponding squared items) [32] and 5 principal components of principal time courses extracted from respective white matter and cerebral spinal fluid (CSF) fluctuations using the anatomical CompCor (aCompCor) method [33]. To minimize the influence of head motion in the rsFC analysis, data points with the instantaneous head motion, defined by framewise displacement (FD) [34] was larger than 0.25 mm, as well as the one time points before and after these “bad” time points were used as regressors in the nuisance regression. Subjects with mean FD > 0.3 mm were excluded from analysis. The global signal regression (GSR) was performed. The MPFC and DLPFC ROIs obtained from the working memory activation/deactivation map in the previous study [11] were used to extract functional connectivity between the two regions. The ROI information including MNI coordinates, volumes etc were presented in Table S1. The importance of including both type of analyses, with and without GSR in rsFC has been shown in [5], therefore the analysis without GSR was also performed.

Statistics

All statistical analyses were performed using IBM SPSS Statistics Version 25. One-way ANOVA was performed to compare the differences of MPFC GABA and Glu concentrations as well as the MPFC-DLPFC functional connectivity between the FEP, RL and HC groups. Bonferroni post hoc tests were performed for significant findings from the one-way ANOVA.

Pearson correlation coefficients were computed between MPFC-DLPFC functional connectivity and MPFC GABA and Glu levels within each group. The age of the participants was within a very narrow range so it was not included as a coefficient in the analysis. Gender was not controlled in the analysis either as our previous study did not show gender as a significant contributor [11]. The relation between GM percentage and measured metabolite concentrations is non-linear and therefore not appropriate to be included as a coefficient. Instead, we calculated the metabolite concentrations in MPFC and DLPFC in pure gray matter with the information of tissue percentages and the prior knowledge of metabolite WM/GM concentration ratios in the same way as our previous publication [11] and suggested by others [35, 36].

Results

Neurotransmitters in MPFC

Two subjects (one HC, one RL) were excluded because of the low SNR of the MEGA-PRESS spectrum caused by the eddy current. The spectral quality metrics, including Cramer-Rao lower bounds (CRLBs), signal of noise ratio (SNR) and the linewidth full-width at half-maximum (FWHM) of the three subject groups, as well as the tissue percentages of the MPFC voxel, are presented in Table S2. All these measures indicate the high MRS data quality stably acquired at our 3 T clinical scanner. There is no significant difference in these measures between the three groups by one-way ANOVA.

Regarding the MPFC neurotransmitters (Fig. 2), a significant group-wise difference was observed on the Glu concentrations (F = 5.6, p = 0.004). Bonferroni post hoc tests showed an elevated Glu level in the MPFC in FEP (9.6 ± 0.6 mM) compared to the control group (8.8 ± 0.6 mM). The Glu level of the relative group (9.1 ± 0.7 mM) is between FEP and HC. There was no significant difference in GABA between the three groups (1.04 ± 0.11 mM, 1.01 ± 0.09 mM and 1.06 ± 0.13 mM for FEP, RL and HC groups respectively).

Fig. 2: Neurotransmitter levels.
figure 2

Comparison of Glu and GABA concentrations in MPFC between FEP, relative and control groups.

Full size image

MPFC-DLPFC resting-state functional connectivity

One subject in the FEP group was excluded because of motions (mean FD > 0.3 mm). Based on MPFC and DLPFC ROIs obtained from the working memory activation/deactivation map in our previous study [11], we extracted rsFC between these two regions; this rsFC is typically an anti-correlation (Fig. 3). A significant group-wise difference in MPFC-DLPFC rsFC was observed in the one-way ANOVA (F = 7.2, p = 0.002). The Bonferroni post hoc test showed the functional anti-correlation between MPFC and DLPFC in FEP was significantly reduced to −0.097 ± 0.184 compared to the control group (−0.250 ± 0.096). The difference between RL (0.302 ± 0.138) and FEP was also significant (p = 0.005).

Fig. 3: MPFC-based resting state functional connectivity.
figure 3

Group-wise resting-state functional connectivity with seed on MPFC of FEP, RL and HC respectively; the comparison of MPFC-DLPFC resting-state functional connectivity between the three groups.

Full size image

Relationship between MPFC neurotransmitters and MPFC-DLPFC anti-correlation

The group-wise correlations were presented in Fig. 4. In the control group we replicated the previous finding [11] that GABA and Glu concentrations in the MPFC were significantly associated with anti-correlation between DMN and CN at rest: R2 = 0.42, p = 0.005 (GABA) and R2 = 0.34, p = 0.01 (Glu). The relationship between MPFC neurotransmitter concentrations and functional anti-correlation was absent in FEP: R2 = 7.1e–4, p = 0.93 (GABA) and R2 = 0.011, p = 0.71 (Glu) and in RL: R2 = 0.05, p = 0.56 (GABA) and R2 = 0.13, p = 0.34 (Glu). The removal of GSR reduced the functional anti-correlation but didn’t change the associations with neurotransmitters in different groups (Fig. S3).

Fig. 4: Scatterplots.
figure 4

Correlations between the MPFC-DLPFC resting-state connectivity and MPFC GABA and Glu concentrations respectively in FEP, RL and HC groups.

Full size image

DTI

There was no significant difference in global FA between the groups. Furthermore, including the mean FA values as covariates did not change the results of the correlation analysis between neurotransmitters and FC in any group.

Discussion

The prefrontal cortex plays a key role in higher-order cognition and emotional processing through its regulation of cortical and limbic regions and its involvement in higher-order executive function [37, 38]. We do not yet have a detailed picture of how neurochemicals modulate cognition and behavior in psychiatric diseases, one emerging principle is that there is a balance and coordinated activity between excitatory glutamatergic and inhibitory GABAergic neurons. The imbalance of cortical cellular excitation and inhibition could give rise to many abnormalities observed in psychosis [39,40,41,42]. In the current study we measured Glu and GABA levels using dedicated semi-LASER and MEGA-PRESS MRS sequences respectively in the human brain at 3 T. We observed elevated Glu levels in the MPFC in FEP, without a significant change in MPFC GABA. Glutamatergic dysfunction is known to occur in psychosis. The increased Glu observed in the cingulate cortex in our study is consistent with many previous reports. For example, a study by Egerton et al. [43] showed that the Glu/Cr ratio in the ACC of 15 patients with FEP was significantly higher compared to patients in remission. Another MRS study using a 3 T MR scanner revealed that absolute Glu concentrations in the dorsal-caudate of antipsychotic naïve FEP patients were elevated compared with HC subjects [44]. A recent study by our group, using TE-averaged MRS, a different sequence to measure Glu on 4 T, also observed an elevated Glu in the MPFC in FEP [45]. There are several other lines of evidence that implicate the glutamatergic system in the pathophysiology of schizophrenia. Dissociative anesthetics such as phencyclidine (PCP) and ketamine block the glutamate N-methyl-d-aspartate receptor (NMDAR), and the administration of these drugs produces a state in adults that is phenomenologically similar to schizophrenia [46]. Ketamine also increases prefrontal glutamatergic metabolites in human [46, 47] and in rat [48]. On the other hand, a recent review and meta-analysis did not show any significant group differences in MPFC GABA between individuals with schizophrenia and health controls [49]. This is in line with our current findings, although not with our previous report of elevated GABA in schizophrenia [50].

The reduced anticorrelations between DMN and positive task-related activation networks has been widely observed [4, 51,52,53]. DMN is characterized by increased activation during task-free resting periods but becomes deactivated during externally-oriented processing; the latter engages the frontoparietal executive control network (CN) [54], one critical note of which is the DLPFC. There is a report of an association between magnitude of MPFC-DLPFC anti-correlation and working memory capacity for young adults [55], as well as findings of reduced MPFC-DLPFC resting-state anti-correlation and reduced working memory capacity in schizophrenia patients [4] and older healthy subjects [56]. In our previous study, we found that MPFC neurotransmitter concentrations modulate not only the MPFC deactivation during a working memory task, but also the MPFC-DLPFC anticorrelation during resting state and working memory task in heathy young adults [11]. To follow up on our previous study, we targeted the MPFC and DLPFC here to measure their rsFC. The loss of MPFC-DLPFC anticorrelation in FEP in the current study is in line with the literature and may reflect insufficient segregation of internally and externally focused states and disturbance of cognition in psychotic disorders [53].

We find that MPFC Glu modulates the MPFC-DLPFC connectivity positively in healthy young adults, i.e., higher MPFC Glu level predicts weaker anticorrelation, as well as an opposite trend of the correlation between GABA and the MPFC-DLPFC connectivity. This finding replicates our previous study [11] in a new cohort of healthy subjects and in line with other studies [57,58,59,60]. In FEP, we found that both GABA-FC and Glu-FC correlations have broken down. The results of this study suggest that the relationship between GABA and Glu and inter-network functional connectivity between DMN and CN is significantly reduced in FEP. Given the elevated Glu level which may reflect an abnormal glutamatergic system in FEP, it appears that the neurotransmission abnormality is accompanied by a breakdown of normal circuit modulation. On the other hand, though we find no MPFC GABA abnormalities in FEP, the GABA system is known to be abnormal in schizophrenia [61, 62] and an abnormal GABA system could lead to disturbances in FC in FEP. Recent studies have reported a disrupted correlation between regional neurotransmitter levels and task evoked BOLD activation in SZ [12, 13]. Furthermore, a few studies have compared the relationship of rsFC in the ACC [14, 15] and hippocampus [17] and neurotransmitter levels in SZ and HC and found alterations of this relationship in both regions. To our knowledge, the current investigation is the first one to study the association of DMN neurotransmitters and internetwork connection between DMN and task activated networks in psychotic disorders. According to our previous study, MPFC neurotransmitters could be more involved in orchestrating between-network brain activity at rest and during task performance than DLPFC neurotransmitters [11], suggesting a special role for MPFC in this system.

Although there’s substantial evidence of abnormalities in white matter in FEP [63], our current findings cannot be explained by white matter abnormalities. The DTI findings in frontal brain regions of FEP are inconsistent [64,65,66,67]. This picture could be caused by highly variable DTI data acquisition, processing and analysis protocols, and the heterogeneity of subject populations [68]. In the current study, we did not find any statistically significant difference in DTI between HC and FEP. Including the mean FA values as covariate did not significantly change the correlation between neurochemical levels and FC. The absence of DTI findings strengthens our interpretation that neurotransmission abnormalities account for rsFC abnormalities in FEP.

We were able to study a small sample of unaffected siblings as a RL group in this study. These individuals were not siblings of the patients in the study, but they all have siblings with FEP themselves. The RL group provides the advantage of examining psychosis risk without the confounding factors of medication and disease effects. In this study, Glu levels and the rsFC results in the RL group were somewhere between those of the FEP and the HC groups and this is broadly consistent with other observations in the literature. This pattern suggests that our findings reflect the pathophysiology of FEP and cannot be fully explained by confounding factors.

Potential limitations of the current study include a relatively small sample size because of the narrow inclusion criteria; thus patients with different diagnoses were not separated into different groups. The observation of correlation was in the context of a small sample and a “noisy” signal. The correlation directions could vary across studies, as different brain regions are studied. Future studies with a significantly larger sample size and an improved MRSI measurement may further validate the correlations observed in the previous and current studies. Additionally, patients were taking medication. In the current study we did not uncover a relationship between CPZ equivalents and neurotransmitter levels or functional connectivity, suggesting that the dose of antipsychotic medication did not have an effect on our measures. Nonetheless, we cannot rule out medication effects and this needs to be addressed in future studies. Thirdly, the GABA measurement is challenging, and it is subject to MM co-editing. The degree of MM signal contamination varies with sequences and parameters [30, 69, 70]. The symmetric editing method [30] can suppress the co-edited MM signal but it was shown that the MM-suppressed GABA signal is very sensitive to frequency drift [71]. On the other hand, previous studies [72, 73] suggested that the MM concentrations in cortical regions of healthy adults are very stable with respect to age and gender. Therefore, individual differences in the contaminated GABA levels are likely to reflect differences in GABA itself. Lastly, using 1H-MRS to detect synaptic Glu is an unsolved issue. The 1H MRS signal arises from all “free” Glu and not only that at the synapse. It cannot distinguish Glu packed in synaptic vesicles (20–30% of total) from the metabolic pool in cytosol (~50%) [74]. Therefore, total tissue levels of metabolites may not solely reflect neurotransmission. And it should be noted that it is Glu signal instead of Glx that was reported here, as most Glu signal quantified using semi-LASER comes from glutamate. However the glutamate signal is not completely resolved from glutamine, and we cannot rule out the contributions from glutamine or other metabolites.

In the current study, we observed that the relationship between MPFC neurotransmitter concentrations and the functional anti-correlation seen in healthy people was compromised in FEP patients. Coordinated activity within and differential activity between DMN and CN is a critical feature of brain organization. These findings imply that DMN glutamatergic neurotransmission dysfunction has an adverse impact on the interaction between DMN and CN in FEP, and it is important to better understand this dysregulation. They also provide opportunities for developing novel treatment strategies for psychotic disorders.