Are working memory and glutamate concentrations involved in early‐life stress and severity of psychosis?

Abstract Objective Occurrences of early‐life stress (ELS) are associated with the severity of psychotic symptoms and working memory (WM) deficits in patients with psychosis (PSY). This study investigated potential mediation roles of WM behavioral performance and glutamate concentrations in prefrontal brain regions on the association between ELS and psychotic symptom severity in PSY. Method Forty‐seven patients with PSY (established schizophrenia, n = 30; bipolar disorder, n = 17) completed measures of psychotic symptom severity. In addition, data on ELS and WM performance were collected in both patients with PSY and healthy controls (HC; n = 41). Resting‐state glutamate concentrations in the bilateral dorsolateral prefrontal cortex (DLPFC) and anterior cingulate cortex (ACC) were also assessed with proton magnetic resonance spectroscopy for both PSY and HC groups. t tests, analyses of variance, and regression analyses were utilized. Results Participants with PSY reported significantly more ELS occurrences and showed poorer WM performance than HC. Furthermore, individuals with PSY displayed lower glutamate concentrations in the left DLPFC than HC. Neither ELS nor WM performance were predictive of severity of psychotic symptoms in participants with PSY. However, we found a significant negative correlation between glutamate concentrations in the left DLPFC and ELS occurrence in HC only. Conclusion In individuals with PSY, the current study found no evidence that the association between ELS and psychotic symptoms is mediated by WM performance or prefrontal glutamate concentrations. In HC, the association between ELS experience and glutamate concentrations may indicate a neurometabolite effect of ELS that is independent of an illness effect in psychosis.


S IG NIFIC ANT OUTCOME S
Our findings of reduced glutamate concentrations in the left dorsolateral prefrontal cortex and poorer working memory performance in patients with psychosis (including both schizophrenia and bipolar disorder) provide evidence for illness-related changes.
The significant relationship between early-life stress experience and glutamate concentrations in the left dorsolateral prefrontal cortex in healthy controls indicates preliminary evidence for general neurobiological effect of ELS.

LI M ITATI O N S
The early-life stress measure used in this study does not directly ask participants about abusive and neglectful experiences during childhood although it is known that such experiences are highly prevalent in individuals with a psychotic disorder. This may account for the lack of a relationship between early-life stress occurrences and psychotic symptom severity in our findings. However, it is also known that other environmental risk factors are implicated in the onset and severity of psychosis.
Due to the relatively small sample, our analyses were not sensitive to weaker associations among early-life stress occurrences, working memory performance, glutamate concentrations, and psychotic symptom severity.

| INTRODUC TI ON
Psychotic disorders are debilitating conditions that are primarily characterized by positive symptoms (including delusions, hallucinations, and disorganized thoughts), negative symptoms (such as avolition, alogia, or apathy), and cognitive deficits (such as social cognition and working memory impairments). The latter can occur prior to the diagnosis of a psychotic disorder and can worsen with illness progression (Fusar-Poli et al., 2010;MacDonald & Schulz, 2009). Cognitive deficits are widely considered to be core symptoms of schizophrenia (SZ; Gold, 2004) and bipolar disorder (BD; Bortolato, Miskowiak, Köhler, Vieta, & Carvalho, 2015;Martinez-Aran & Vieta, 2015) and are also associated with reductions in working memory performance. Working memory deficits are one of the main neurocognitive impairments found in subjects with first-episode psychosis (Kim et al., 2011;Seidman, 2010) and in patients at the established stage (Genevsky, Garrett, Alexander, & Vinogradov, 2010). Deficits in working memory are important given the role of working memory in information retention and manipulation in order to perform abilities such as speech and multitasking (Vöhringer et al., 2013) and involve the recruitment of the dorsolateral prefrontal cortex (DLPFC; Glahn et al., 2005;Jalbrzikowski et al., 2018;Roberts, Libby, Inhoff, & Ranganath, 2018). Furthermore, working memory deficits are associated with impaired daily functioning in patients with a psychotic disorder (Vesterager et al., 2012).
In addition, recent evidence suggests that the occurrence of ELS also impacts on neurocognitive function, such as working memory performance in both individuals with PSY and HC (Dauvermann & Donohoe, 2019a;Goodman, Freeman, & Chalmers, 2019;Vargas et al., 2019) as well as brain activation and connectivity in HC (Philip et al., 2016;Philip et al., 2013). However, little is known about the underlying mechanism between the environmental risk factor of ELS and working memory function in individuals with PSY.
The glutamate hypothesis of schizophrenia (Coyle, 2012) and the N-methyl-D-aspartate receptor (NMDA-R) hypofunction model (Coyle, 1996) posit that aberrant glutamate systems (Krystal et al., 1994) may be implicated in the pathophysiology of schizophrenia and psychosis. Proton magnetic resonance spectroscopy ( 1 H-MRS) studies examined both the glutamate hypothesis of schizophrenia and the NMDA-r hypofunction model by measuring resting-state glutamate levels in prefrontal brain regions in individuals with SZ and BP. Such studies reported inconsistent findings of glutamate levels in the dorsolateral prefrontal cortex (DLPFC) in patients with SZ (Poels et al., 2014) with studies reporting (a) increased glutamate levels , (b) decreased glutamate levels, (Ćurčić-Blake et al., 2017) or (c) no difference in glutamate levels (Yoo et al., 2009) in the DLPFC compared to healthy controls (HC). For glutamate levels in the ACC in patients with SZ, there are further inconsistent findings with the majority of studies demonstrating no difference in concentrations of glutamate or glutamate-related metabolites, such as glutamine or glutamate and glutamine combined (Glx) concentrations between chronic patients with SZ and HC (Bustillo et al., 2011;Kraguljac, Reid, White, Hollander, & Lahti, 2012;Reid et al., 2010;Rowland et al., 2013;Shirayama et al., 2010). However, a small number of studies also reported decreased concentrations of glutamate between patients with SZ and HC (Tayoshi et al., 2009;Théberge et al., 2003Théberge et al., , 2004. The pattern of findings in patients with BD differs from patients with SZ, where individuals with BD consistently showed higher Glx levels in the DLPFC and ACC when compared to HC (Öngür et al., 2008;Soeiro-de-Souza et al., 2016).
Focusing on working memory deficits, it has been proposed that altered glutamate regulation is one of the main neurobiological pathways underlying working memory impairments in psychosis based on both preclinical and clinical evidence (Dauvermann, Lee, & Dawson, 2017;Dauvermann et al., 2014;Javitt, 2007). Further support for the role of glutamate during working memory in individuals with PSY comes from a recent study using functional 1 H-MRS (Jelen et al., 2019).
In the last few years, emerging evidence proposes a role of stress, in particular social stress, in aberrant glutamatergic transmission in the PFC in rodents (Schiavone et al., 2020;Zhang, Hernández, Vázquez-Juárez, Chay, & Barrio, 2016), which has also been in shown in HC who have experienced ELS (Duncan et al., 2015). Importantly, it has been shown that stress, in particular chronic social stress, resulted in both deficits in working memory and glutamatergic dysregulation in the PFC in rodents (Onaolapo et al., 2019;Shao et al., 2015;Yuen et al., 2011). However, to date no study has examined the association between ELS experience and severity of psychotic symptoms in individuals with PSY when considering glutamate concentrations as a potential neurobiological mechanism, including working memory function (Aas et al., 2012;Janssen et al., 2004).

| Aims of the study
In this study, we investigated whether there is a relationship between the experience of ELS and severity of psychotic symptoms in PSY. In addition, we examined whether working memory performance and glutamate concentrations during resting state in the bilateral DLPFC and the ACC as measured with 1 H-MRS mediate this relationship between the ELS experience and severity of psychotic symptoms.

| Participants
Forty-seven participants with psychosis (SZ = 30, BD = 17) and 41 HC were recruited for the study. PSY and HC were recruited from the Royal Edinburgh Hospital, associated hospitals, and the Scottish Mental Health Research Register (http://www.smhrn.org.uk/).

Diagnosis of SZ and BD was based on interview using the Structured
Clinical Interview for DSM-IV (First, Spitzer, Gibbon & Williams, 2002). Inclusion criteria included (a) diagnosis of established SZ or BD, and (b) no acute psychotic symptoms at the time of the scan.
Exclusion criteria included (a) history of any major psychiatric illness other than SZ or BD, (b) history of severe brain injury, (c) history of a neurological disorder, and (d) dependency or harmful use of alcohol or drugs during the last 12 months. Also, HC were excluded if they had a family history of SZ or BD. All participants provided written informed consent. The study was approved by the local Research Ethics Committee (09/MRE00/81).

| Demographic measures
Participants provided information regarding their age, sex, and education status. Participants were also asked regarding medication use (antipsychotic medication, mood stabilizers, and antidepressant medication).
Smoking status has not been recorded.

| Clinical measures
The Positive and Negative Syndrome Scale (PANSS; Kay, Fiszbein, & Opler, 1987) is a 30-item standardized clinical interview to rate the presence and severity of positive and negative symptoms in PSY.
The measure consists of the following scales: positive symptoms, negative symptoms, general symptoms, and total score. Items were rescored to 0 ("Absent") to 6 ("Extreme") for symptoms during the past week with higher scores indicating greater severity. Internal reliability for the three subscales of positive symptoms, negative symptoms, and general symptoms has been reported as adequate (Cronbach's α .73-.83; Kay et al., 1987). Previous studies have indicated adequate validity of the PANSS (Kay et al., 1987). Symptom rating took place within 1 week of the 1 H-MRS acquisition.

| Environmental questionnaires
The Childhood Life Events Questionnaire (CLEQ; http://bdrn.org/) is a 13-item verbally administered measure of ELS shared by the Bipolar Disorder Research Network (BDRN [Upthegrove et al., 2015]). All participants were asked whether they had experienced any or all of a list of 12 adverse events prior to the age of 16 years as previously reported (Barker et al., 2016;Neilson et al., 2017). This list includes death of a parent, death of a sibling, death of a friend, parental separation, parental divorce, admission to hospital, hospitalization of parent, visible deformity, teenage pregnancy or fatherhood, imprisoned parent and suspension from school. Responses ranged from 0 ("No") to 1 ("Yes"). While no questions regarding childhood abuse or neglect were asked, participants were given the opportunity to provide this information by the final item which is an open-ended question: "Are there any other significant life events you experienced as a child that are not mentioned above?"

| Working memory data
All participants performed the verbal "2-back" task. They were presented with a sequence of single capital letters. The experimental block design consisted of (a) the baseline or "0-back" condition; (b) the "1-back" condition; and (c) the "2-back" condition. Full experimental details are presented in the Appendix S1.

| Structural MRI acquisition and analysis
Brain scans were collected using a 3 T Siemens Verio MRI scanner using a manufacturer-supplied 12-element matrix head coil at the Clinical Research Imaging Centre (CRIC), the Queen's Medical Research Institute, Edinburgh, UK. After a sagittal localizer, T1-weighted magnetization-prepared rapid-acquisition gradient echo (MPRAGE) MR images were obtained using TR = 2,300 ms, TE = 2.98 ms, and TI = 900 ms, (flip angle = 9, FOV = 256 mm × 256 mm) with an isotropic voxel resolution of 1 mm, parallel to AC-PC plane.

Magnetic resonance spectroscopy data acquisition
We acquired MRS point-resolved selective spectroscopy (PRESS) spectra with voxel placements in the ACC, right DLPFC, and left DLPFC. The voxels were shimmed using the Siemens advance mode. We acquired a water unsuppressed spectra with 16 averages and a water suppressed spectra with 128 averages. The TE was set to 80ms and the TR was set to 3,000 ms. The phase cycling was set to the Siemens 16 EXOR-cycle mode and the bandwidth was 2,500 Hz with oversampling enabled.
This followed a standardized protocol, with navigation steps to determine the coronal slice for placement. Cortical feature identification was used to designate the voxel center, followed by rotations of the voxel in the transverse and sagittal views to obtain the final placement. The location of the ACC voxel is shown in Figure 1a and the placement of the left DLPFC voxel is shown in Figure 1b. Details for the voxel placement are presented in the Appendix S1 and in a previous paper (Thomson et al., 2016). An example of a spectrum acquired in the dACC in a HC ( Figure 1c).

Metabolite measurement
Spectral metabolite quantification was performed in LCModel version 6.2 (http://www.s-prove ncher.com/pages /lcmod el.shtml) using a gamma_press_te80_123mhz basis set provided at http://s-prove ncher -com for use with the Siemens 3T Verio scanner. LCModel obtains maximum-likelihood estimates of metabolite concentrations and their uncertainties (Cramer-Rao lower bounds; CRLB). The quality of the spectra obtained and the specificity of the measurement of a given metabolite were evaluated using the CRLB measure of uncertainty. This is a measure of the specificity of the peak in the spectrum associated with a given metabolite. Only metabolite measurements that were associated with a fitting error (CRLB) of <15% were included in the analysis.
The raw spectra were read into the graphical user interface for LCModel and processed through LC Model with eddy current correction enabled and internal water reference was used to give the metabolite in institutional units. For each voxel the metabolite values derived from LCModel were corrected for voxel cerebrospinal fluid (CSF) content as per the equation: where Met CI is the metabolite in institutional units and corrected for partial volume effects, Met I is the internal water scaled metabolite value given by LC model, and F CSF is the fractional CSF occupancy of the voxel. The fractional CSF volume was determined from the segmentation of the T 1 -weighted scan and the voxel placement and rotation noted at scan time. The T1-weighted images were segmented into gray matter, white matter, and CSF maps using SPM8 (Statistical Parametric Mapping; http://fil.ion.ucl.ac.uk/spm/). The CSF volume in each voxel was extracted using a c-script that sampled the SPM segmentation maps at the native space location noted for each voxel at time of scan.

| Clinical and environmental data
A series of t tests were conducted to examine differences between groups on measures, such as PANSS total scores, PANSS positive symptoms, PANSS negative symptoms, PANSS general scores, ELS data, and glutamate concentrations (separately in the right DLPFC, left DLPFC, and ACC). In following analyses, a multivariate ANOVA (MANOVA) was conducted to examine group differences in glutamate concentrations in the bilateral DLPFC and ACC. Multiple linear regressions assessed whether ELS occurrences, WM performance, or glutamate metabolite concentrations (separately in the bilateral DLPFC and ACC) predicted psychotic symptom severity by introducing these predictors systematically across blocks. Statistical significance was set at a p value below .05. Pearson's product-moment correlations were computed to examine the relationships between ELS occurrence, WM behavioral performance, glutamate metabolite concentrations, and psychotic symptom severity. p values were adjusted for multiple comparison using the Hochberg method in R Studio (RStudio Team, 2018).

| Working memory data
Behavioral performance was calculated using the sensitivity index performance on the 0-back, 1-back, and 2-back conditions between the groups was compared in SPSS using independent t tests and ANOVAs.

| Glutamate concentrations
Analysis of covariance (ANCOVA) was performed to examine the hypothesized differences in glutamate concentrations in each region between the PSY and HC groups. Age and sex were entered in as covariates for all comparisons.

| Demographic, clinical, and working memory data
Demographic and clinical data are presented in Table 1 for SZ and BD subgroups are presented in the Appendix S1 (Table S1).

Psychotic symptom severity
Positive and Negative Syndrome Scale severity in PSY was assessed with comparable severity (Table 1). In addition, independent t tests  Table S1).

| Early-life stress
The difference between the HC and the SZ groups (p = .001) in ELS scores, with the SZ group scoring higher (see Table S1). There was no significant group difference between the BD and HC groups (p = .267) and the BD and SZ groups (p = .272) on ELS scores.

| Working memory performance
To examine group differences in behavioral performance on the N-back task, a series of t tests were conducted (HC vs. PSY). There was no significant difference on the 0-back performance (t (74)  To examine any differences in N-back performance between subgroups, a series of one-way ANOVAs (HC vs. SZ vs. BD) were conducted. Regarding behavioral performance on the 0-back task, we did not find any significant differences between the three groups (F (2,76) = 1.31, p = .277, η 2 = .035). Regarding the 1-back task, a significant difference was observed (F (2,76) = 5.121, p = .008, η 2 = .123).
Post hoc analyses revealed that the SZ group performed significantly poorer (p = .012) than the HC group. There was no significant difference between the SZ and BD groups (p = .970) or the BD and HC groups (p = .092). On the 2-back task, a significant group difference was also found (F (2,76) = 3.357, p = .040, η 2 = .084). Post hoc analyses revealed that the SZ group scored significantly lower (p = .049) than the HC group. There were no significant differences

TA B L E 1 Demographic and clinical details
between the SZ and BD groups (p = .961) or between the BD and HC groups (p = .228).

| Glutamate concentrations
The number of participants who met the criterion of data quality is presented in Table 2. A series of one-way ANCOVAs (PSY vs. HC) were conducted to examine whether glutamate concentrations in the three brain areas varied according to group, while controlling for age and sex. The results of the testing found a significant difference in glutamate concentration in the left DLPFC between groups (F (1, 63) = 5.742, p = .020, partial η 2 = .084), with the PSY group reporting lower concentrations of glutamate than the HC group (see Table 2). There were no significant differences between the PSY and HC groups with regards to glutamate concentration in the right DLPFC (p = .324) or ACC (p = .971). Extension of the inclusion criteria (glutamate standard deviation <30%) resulted in no significant difference in glutamate concentrations in the bilateral DLPFC or ACC between the PSY and HC groups (p > .05). These analyses were repeated for subgroups (SZ vs. BD vs. HC). No significant difference was found for the right or left DLPFC or ACC (see Appendix S1; Table S2).

| Multivariate analyses for psychotic symptom severity differences between clinical group and earlylife stress occurrences
To examine differences in psychotic symptom severity between both clinical groups (SZ and BD) and ELS experience, reported ELS experience was dichotomized into absent levels (e.g., individuals who reported no ELS experience; SZ = 4, BD = 4) and present levels

| D ISCUSS I ON
The main aims of this study were to examine the association between the occurrence of ELS and severity of psychotic symptoms in PSY. Furthermore, we studied whether glutamate concentrations in prefrontal brain regions would mediate this relationship between ELS experience and psychotic symptom severity.
One of the main findings was that patients with PSY experienced significantly higher occurrences of ELS than HC, which is consistent with previous research (Bailey et al., 2018;Varese et al., 2012). This trainings (based on altered working memory function). In our additional analyses focusing on both SZ and BP, we found that SZ reported higher ELS exposure when compared to HC, whereas BD showed comparable levels to HC. This latter finding contrasts previous meta-analyses, which suggested that ELS occurrences were significantly higher among BD compared to HC (Etain et al., 2010;Macmillan et al., 2004). A possible reason for this discrepancy is the choice of ELS measure of the CLEQ in this study when compared to the Childhood Trauma Questionnaire (CTQ) among others (Dauvermann et al., 2017).
Contrary to our hypothesis, the occurrence of ELS was not significantly associated with the severity of psychotic symptoms in patients with PSY. This contrasts previous studies in which positive correlations between these two measures were revealed (Janssen et al., 2004;Kay et al., 1987;Öngür et al., 2008). experiences in this study. Support for this interpretation comes from a previous meta-analysis which reported that, dependent on adversity type, between 26% and 39% of participants with PSY reported experiencing trauma as a child (Bonoldi et al., 2013). However, this figure was just 12.2% for individuals with PSY in this study. Despite the established association between childhood trauma and severity of psychotic symptoms (Agnew-Blais & Danese, 2016;Bailey et al., 2018;Bechdolf et al., 2010;Day et al., 1987;Gershon et al., 2013;Holtzman et al., 2013;Lataster et al., 2012;Mayo et al., 2017;Os et al., 2010;Shannon et al., 2011;Thompson et al., 2009;Üçok et al., 2015;Varese et al., 2012;Walder et al., 2014), it is also known that other environmental risk factors, such as cannabis use, migration and urbanicity are highly correlated with psychosis (Murray et al., 2017;Neilson et al., 2017;Newbury et al., 2018). Recently, it has been suggested that several measurements for the assessment of ELS should be used to cover different aspects of ELS (Vargas et al., 2019).
In the current study, ELS experiences were not significantly associated with working memory performance despite finding that individuals with PSY had significantly increased ELS experience and reduced working memory performance relative to HC. These group differences for both the low and high difficulty levels during working memory are consistent with meta-analyses and systematic reviews repeatedly confirming this (Forbes, Carrick, McIntosh, & Lawrie, 2009;Frydecka et al., 2014;Hamilton et al., 2009). We speculate that the lack of a significant relationship between ELS and working memory in our study may be due to the working memory data utilized. In other words, it is possible that a neurobiological measure of working memory function in the form of blood-oxygen-level-dependent (BOLD) response of brain activity as measured with functional magnetic resonance imaging (fMRI) may be required to reveal a potential mediation role of reduced working memory function on the relationship between ELS and severity of psychotic symptoms. Support for this interpretation comes from inconsistently reported relationships between ELS experience and working memory function across studies (Dauvermann & Donohoe, 2019a;Goodman et al., 2019;Vargas et al., 2019) when behavioral performance is considered. In contrast, widely established findings of altered BOLD responses of the DLPFC in individuals with PSY when compared to HC during working memory (Glahn et al., 2005;Owen, McMillan, Laird, & Bullmore, 2005) emphasize the greater reliability to study the role of working memory on the severity of symptoms, while the laterality of the DLPFC underlying working memory function seems to be variable across studies. In addition, a recent study linking glutamate concentrations as well as glutamate + glutamine (Glx) levels during a working memory task adds further support to the potential relevance of using brain function during working memory in individuals with psychosis (Jelen et al., 2019). Emerging evidence for utilizing brain activation data during working memory to study this association between ELS exposure and working memory also emphasizes the recruitment of brain activation of the DLPFC in HC with the experience of childhood trauma (Philip et al., 2013(Philip et al., , 2016Teicher, Samson, Anderson, & Ohashi, 2016) and rodents following chronic stress exposure (Onaolapo et al., 2019;Shao et al., 2015;Yuen et al., 2011).  (Frey et al., 2007;Michael et al., 2003;Michael, Erfurth, & Pfleiderer, 2009;Taylor, 2014), but significant differences in Glx between individuals with SZ (regardless of medication status) and HC have not been observed (Kaminski et al., 2020 (Bustillo et al., 2011;Kraguljac et al., 2012;Reid et al., 2010;Rowland et al., 2013;Shirayama et al., 2010).
A limitation of the previous research is that many studies interpret glutamate, Gln, and Glx concentrations interchangeably, so BD-and SZ-related increases (Gigante et al., 2012;Théberge et al., 2003) that are specific to one of these indicators (e.g., Gln or Glx) but not others  (Zwanzger et al., 2013) and no change in glutamate levels (Houtepen et al., 2017). To our knowledge, this is the first study providing preliminary evidence of a link between ELS exposure and altered glutamate concentrations in the DLPFC in humans, which is supporting impairments of glutamatergic neurotransmission previously reported in rodents after chronic social stress (Aas et al., 2012;Janssen et al., 2004). However, we did not observe a significant relationship between ELS experience and glutamate concentrations in individuals with PSY. This observed association in HC only is comparable to a stronger relationship between the experience of ELS and deficits in neurocognitive function (including working memory) in HC than in individuals with PSY (Malarbi, Abu-Rayya, Muscara, & Stargatt, 2017;Philip et al., 2016;Vargas et al., 2019). It has been suggested that such a finding could be due to the fact that the effect of ELS cannot be isolated from other known implicated factors on cognitive function and severity of psychotic symptoms in PSY, such as genetic risk factors, current and acute stress levels (Dauvermann & Donohoe, 2019b), and medication effects (Dauvermann & Donohoe, 2019a;Vargas et al., 2019). However, given the strong link between ELS and increased likelihood of developing PSY and the support of aberrant glutamatergic transmission in the DLPFC in PSY is it also possible that the lack of significant findings in individuals with PSY was limited by methodological shortcomings, including sample size.
Future research studying relationships between ELS, cognitive function, and neurobiological markers in PSY and other stress-sensitive psychiatric disorders should focus on overcoming limitations of a variety of neurocognitive batteries and ELS measures (Dauvermann & Donohoe, 2019a;Vargas et al., 2019) that are challenging to interpret. Emerging evidence on combining ELS, cognitive, endocrine, cytokine, and neuroimaging data will lead to greater insight into the interrelationships and therefore etiology of PSY. However, the funder had no role in study design, data collection, analysis, or interpretation or the writing of the report. The Dr

ACK N OWLED G M ENTS
Mortimer and Theresa Sackler Foundation also offered financial support for imaging aspects of this study.

CO N FLI C T O F I NTE R E S T
The author SML has received financial support for research, in the past 3 years, from Janssen, in relation to research. He has also received fees for advisory panels and/or educational meetings from Janssen and

DATA AVA I L A B I L I T Y S TAT E M E N T
The data that support the findings of this study are available from the corresponding author upon reasonable request.