Longitudinal trajectories of anterior cingulate glutamate and subclinical psychotic experiences in early adolescence: the impact of bullying victimization

Previous studies reported decreased glutamate levels in the anterior cingulate cortex (ACC) in non-treatment-resistant schizophrenia and first-episode psychosis. However, ACC glutamatergic changes in subjects at high-risk for psychosis, and the effects of commonly experienced environmental emotional/social stressors on glutamatergic function in adolescents remain unclear. In this study, adolescents recruited from the general population underwent proton magnetic resonance spectroscopy (MRS) of the pregenual ACC using a 3-Tesla scanner. We explored longitudinal data on the association of combined glutamate-glutamine (Glx) levels, measured by MRS, with subclinical psychotic experiences. Moreover, we investigated associations of bullying victimization, a risk factor for subclinical psychotic experiences, and help-seeking intentions, a coping strategy against stressors including bullying victimization, with Glx levels. Finally, path analyses were conducted to explore multivariate associations. For a contrast analysis, gamma-aminobutyric acid plus macromolecule (GABA+) levels were also analyzed. Negative associations were found between Glx levels and subclinical psychotic experiences at both Times 1 (n = 219, mean age 11.5 y) and 2 (n = 211, mean age 13.6 y), as well as for over-time changes (n = 157, mean interval 2.0 y). Moreover, effects of bullying victimization and bullying victimization × help-seeking intention interaction effects on Glx levels were found (n = 156). Specifically, bullying victimization decreased Glx levels, whereas help-seeking intention increased Glx levels only in bullied adolescents. Finally, associations among bullying victimization, help-seeking intention, Glx levels, and subclinical psychotic experiences were revealed. GABA+ analysis revealed no significant results. This is the first adolescent study to reveal longitudinal trajectories of the association between glutamatergic function and subclinical psychotic experiences and to elucidate the effect of commonly experienced environmental emotional/social stressors on glutamatergic function. Our findings may deepen the understanding of how environmental emotional/social stressors induce impaired glutamatergic neurotransmission that could be the underpinning of liability for psychotic experiences in early adolescence.


Subclinical psychotic experiences
Subclinical psychotic experience data were obtained longitudinally at Times 1 and 2.
Four questions administered in the Diagnostic Interview Schedule for Children (DISC-C), 5 which were also used in later studies, 6 were used in the current study to assess subclinical psychotic experiences: (a) "Have other people ever read your mind?"(b) "Have you ever had messages sent just to you through television or radio?"(c) "Have you ever thought that people are following you or spying on you?"; and (d) "Have you heard voices other people can't hear?"Three responses of "Yes, definitely," "Maybe," and "No, never" are possible, and scored as 2 points, 1 point, and 0 points, respectively.[9]

Bullying victimization
Bullying victimization data obtained at Time 1 were used in this study.Children were asked two questions: "How often have you been bullied at school in the past two months?"and "How often have you been bullied outside school in the past two months?"Similarly, primary parents were asked the following question: "How often has your child been bullied in the past two months?"The participants were asked to choose one of five responses: "never," "once or twice in two months," "twice or three times a month," "once a week," and "several times a week." 4,10 ll answers other than "never" were interpreted as being bullied. 11Because the reproducibility of the evaluation for bullying victimization is low between children and parents, similarly to our previous study, disclosure by either a child or a primary parent of being bullied resulted in a categorization as a bullied victim. 11A dummy variable was defined for the bullying victimization variable, with 1 indicating a bullied victim and 0 indicating a non-bullied victim.

Help-seeking intention
Help-seeking intention data obtained at Time 1 were used in the current study.Helpseeking intentions were assessed using the same method as in our previous study, 12 which was similar to that used in a large Australian study intended to identify a child with depression. 13We called the child Taro, which is one of the most popular boy names in Japan.The description is as follows: "For the last several weeks, Taro has been feeling unusually sad.He is tired all the time and has trouble sleeping at night.Taro doesn't feel like eating and has lost weight.He can't keep his mind on his studies, and his grades have dropped.He puts off making any decisions, and even day-to-day tasks, such as studying and extracurricular activities, seem too much for him.His parents and teachers are very concerned about him."This vignette was written to satisfy the diagnostic criteria for major depression according to the Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition (DSM-IV) and International Statistical Classification of Diseases and Related Health Problems: 10th revision (ICD-10).Children were asked whether they would seek help from others if they were in the same situation as the boy in the vignette.The response options were, "I would consult someone immediately" and "I would wait and see without consulting anyone."Here, we did not specify the source of help, because the focus of this study was the intention to seek help from anyone and not the source of help.A dummy variable was defined for the help-seeking intention variable, with 0 indicating being non-helpseeking and 1 indicating being help-seeking.

Intelligence quotient
In the present analysis, two subtests (information, picture completion) of the Wechsler Intelligence Scale for Children -Third Edition (WISC-III) 14 were collected and used for calculation of IQ at Time 1. Our rationale for only using the two subtests of WISC-III is as follows.The full version of the WISC-III was conducted for 28 children one year after the initial survey using the two subtests.Using multiple regression analysis with full IQ as a dependent variable and the results of the two subtests as independent variables, a formula for estimating IQ from the two subtests was created and the estimated IQ explained 78% of the variance of the full IQ. 4

Supplementary Method 3: Power analysis
Using G*Power 3.1.9.6, 15 a priori statistical power analyses were performed to estimate the minimum sample size.The minimum sample size for correlational analysis was estimated at 84 with effect size ρ = 0.30, alpha = 0.05 (two-tailed), and power = 0.8.The minimum sample size for repeated measures analysis of variance (ANOVA) was estimated at 98 with effect size f = 0.25, alpha = 0.05, power = 0.8, number of groups = 2, and number of measurements = 2.

Supplementary Method 4: Statistical analysis for GABA+ levels
First, we investigated the association of gamma-aminobutyric acid (GABA) plus macromolecule (GABA+) levels with subclinical psychotic experiences at each time point.Partial Spearman's correlation between GABA+ levels and subclinical psychotic experiences was assessed, adjusted for age at MRI scanning, sex, SES, and IQ.In addition, we investigated whether the GABA+ level changes and changes in subclinical psychotic experiences between the two time points were associated.Partial Spearman's correlation between GABA+ level changes and changes in subclinical psychotic experiences between the two time points was assessed, adjusted for MRI scan interval, sex, SES, and IQ.Multiple testing correction was performed using false discovery rate (FDR) analysis using p.adjust in R 4.0.5.
Second, the effects of bullying victimization and help-seeking intention on GABA+ levels at Times 1 and 2 were investigated.GABA+ levels were adjusted for age, sex, SES, and IQ in a linear regression model.After checking whether all the assumptions were met, a threefactor mixed-design ANOVA was conducted with age-sex-SES-IQ-adjusted GABA+ levels as the dependent factor, bullying victimization and help-seeking intention as between-subjects factors, and time as a within-subjects factor.
Finally, we performed path analysis to determine the relationships among variables including bullying victimization, help-seeking intention, age-sex-SES-IQ-adjusted GABA+ levels, and subclinical psychotic experiences using partial least squares structural equation modeling (PLS-SEM), which is a non-parametric analysis technique.All subjects enrolled in the ANOVA study were included.Basically, in this analysis, paths from bullying victimization to GABA+ levels and paths from GABA+ levels and subclinical psychotic experiences were regarded as the main connections, and the moderating effects of help-seeking intention on paths from bullying victimization to GABA+ levels were also investigated.We created a time lagged model, where latent variables at two time points were separately included, and a latent change score model, where latent baseline (intercept) variables and latent change (slope) variables were included.The PLS-SEM analyses were implemented using the SmartPLS 4.0 software, which allows us to estimate direct and moderating (interaction) effects.Bootstrapping method with 5,000 random resamples was used.The fit of the PLS-SEM model was evaluated using standardized root mean residual (SRMR), which is provided by the SmartPLS 4.0 software.An SRMR value of less than 0.10 indicates an acceptable fit to the model. 16Indirect effects were also assessed using PLE-SEM models.

PLS-SEM analysis
Path analyses were performed to determine the relationships among variables including bullying victimization, help-seeking intention, age-sex-SES-IQ-adjusted GABA+ levels, and subclinical psychotic experiences using PLS-SEM.First, a time lagged model was created, and the PLS-SEM test revealed that the current model fitted the data well (n = 156, SRMR = 0.000) (Supplementary Fig. 6a).Specifically, the path coefficient (PC) from subclinical psychotic experiences at Time 1 to subclinical psychotic experiences at Time 2 was significant (PC = 0.32, p = 0.000), whereas the PC from bullying victimization to GABA+ levels at Time 1 (PC = −0.47,p = 0.45), PC from bullying victimization to GABA+ levels at Time 2 (PC = −0.23,p = 0.60), PC from GABA+ levels at Time 1 to GABA+ levels at Time 2 (PC = 0.094, p = 0.24), PC from GABA+ levels at Time 1 to subclinical psychotic experiences at Time 1 (PC = 0.064, p = 0.47), PC from GABA+ levels at Time 1 to subclinical psychotic experiences at Time 2 (PC = 0.021, p = 0.81), and PC from GABA+ levels at Time 2 to subclinical psychotic experiences at Time 2 were non-significant (PC = −0.11,p = 0.10).In addition, help-seeking intention did not significantly moderate the path from bullying victimization to GABA+ levels at Time 1 (PC = 0.21, p = 0.75) or to GABA+ levels at Time 2 (PC = 0.32, p = 0.49).No total indirect effects or specific indirect effects were found.
Next, a latent change score model was created, while the PLS-SEM test revealed that the current model fitted the data well (n = 156, SRMR = 0.085) (Supplementary Fig. 6b).
Specifically, the PC from baseline GABA+ levels to differences in GABA+ levels (PC = 1.0, p = 0.000) and PC from baseline subclinical psychotic experiences levels to differences in subclinical psychotic experiences levels (PC = 0.97, p = 0.000) were significant, whereas PC from bullying victimization to baseline GABA+ levels (PC = −0.31,p = 0.50), PC from bullying victimization to differences in GABA+ levels (PC = 0.039, p = 0.63), PC from baseline GABA+ levels to baseline subclinical psychotic experiences (PC = −0.057,p = 0.43), PC from baseline GABA+ levels to differences in subclinical psychotic experiences (PC = −0.10,p = 1.0), and PC from differences in GABA+ levels to differences in subclinical psychotic experiences were nonsignificant (PC = 0.069, p = 1.0).In addition, help-seeking intention did not significantly moderate the path from bullying victimization to baseline GABA+ levels (PC = 0.36, p = 0.46) or the path from bullying victimization to differences in GABA+ levels (PC = −0.013,p = 0.86).No total indirect effects or specific indirect effects were found.