Association between increased anterior cingulate glutamate and psychotic-like experiences, but not autistic traits in healthy volunteers

Despite many differences, autism spectrum disorder and schizophrenia spectrum disorder share environmental risk factors, genetic predispositions as well as neuronal abnormalities, and show similar cognitive deficits in working memory, perspective taking, or response inhibition. These shared abnormalities are already present in subclinical traits of these disorders. The literature proposes that changes in the inhibitory GABAergic and the excitatory glutamatergic system could explain underlying neuronal commonalities and differences. Using magnetic resonance spectroscopy (1H-MRS), we investigated the associations between glutamate concentrations in the anterior cingulate cortex (ACC), the left/right putamen, and left/right dorsolateral prefrontal cortex and psychotic-like experiences (Schizotypal Personality Questionnaire) and autistic traits (Autism Spectrum Quotient) in 53 healthy individuals (26 women). To investigate the contributions of glutamate concentrations in different cortical regions to symptom expression and their interactions, we used linear regression analyses. We found that only glutamate concentration in the ACC predicted psychotic-like experiences, but not autistic traits. Supporting this finding, a binomial logistic regression predicting median-split high and low risk groups for psychotic-like experiences revealed ACC glutamate levels as a significant predictor for group membership. Taken together, this study provides evidence that glutamate levels in the ACC are specifically linked to the expression of psychotic-like experiences, and may be a potential candidate in identifying early risk individuals prone to developing psychotic-like experiences.

pated in this study.See supplementary materials for recruitment and inclusion criteria.The study was approved by the medical research ethics committee of the Technical University of Munich.All subjects gave written informed consent in accordance with the Declaration of Helsinki.
All participants completed a German version 59 of the Schizotypal Personality Questionnaire (SPQ) 60 capturing psychotic-like experiences; as well as the Autism Spectrum Quotient (AQ) 61 assessing autistic traits (Table 1, Fig. S1; see supplementary material for details on correlations between AQ and SPQ).
1 H-MRS analysis.The MRS data were analyzed using Osprey version 2.4.0, an all-in-one software for stateof-the-art processing and quantitative analysis of in-vivo MRS data 62 .As eddy-current correction had already been performed during the automatic pre-processing of the scanner, this was excluded from the automatic preprocessing in Osprey.Subsequently, Osprey executed the remaining processing steps appropriate to the provided data, including frequency-and-phase alignment, water removal, frequency referencing, and initial phasing.
As the TE of our MRS data varied between and within the individual voxels to ensure the shortest acquisition time, we simulated six different PRESS basis sets for each TE (36, 37, 38, 39, 40, 41) with a bandwidth of 2000 Hz and 1024 pts using MARSS in INSPECTOR version 11-2021 63 .Based on recommendations in the LCModel manual 64 , we selected the following metabolites: alanine, aspartate, creatine (Cr), GABA, glucose, glutamate (Glu), glutamine, glutathione, glycerophosphocholine, lactate, myoinositol, NAA, N-acetyl-aspartylglutamate (NAAG), phosphocholine, phosphocreatine (PCr), scyllo-inositol and taurine.To this basis set we added the default macromolecular and lipid components provided by Osprey.
To analyze our data in Osprey, we utilized the LCModel 65 implementation for fitting and quantification.The T1-weighted images were segmented in grey matter (GM), white matter (WM), and cerebrospinal fluid (CSF) using SPM12 within Osprey.The processing followed standard parameters, with a metabolite fit range of 0.5 to 4.0 ppm and a water fit range of 2.0 to 7.4 ppm.The knot spacing used was 0.4 ppm.Finally, Osprey estimates the tissue and relaxation corrected molal concentration according to the Gasparovic method 66 .See supplementary materials for a detailed workflow diagram (Supplementary Materials Fig. S2A).
Osprey provides data quality outcomes, including signal-to-noise ratio (Cr SNR; ratio between amplitude of Cr peak and standard deviation of detrended noise in the range of − 2 to 0 ppm 62 ), linewidth for water [full-width half-maximum (FWHM) of single-Lorentzian fit to H2O reference peak 62 ] and Cramer-Rao lower bounds (CRLB) for each estimated metabolite.All spectra were visually inspected for artifacts.An exemplary fit of the spectra can be seen in Fig. 1A-E.Spectral exclusion criteria were therefore either visual failure of the fitting algorithm, a resultant FWHM > 13 Hz in the ACC and DLPFC or FWHM > 10 Hz in the putamen, with thresholds chosen according to recommendations for B0-shimming provided by Juchem et al. 67 , or a CRLB > 20% of glutamate concentration.
1 H-MRS glutamate levels and spectral quality.We measured the tissue and relaxation corrected molal concentration of glutamate (mol/kg) in five voxels of interest, the ACC, the left putamen, right putamen, left DLPFC and right DLPFC.For the ACC and the right DLPFC, the fit of the glutamate spectra were appropriate for all participants.For the left DLPFC, one subject had to be excluded due to visual failure of the fit.Although the visual inspection was appropriate, but generally showing more variation in the residuals, for the left and right putamen, 15 subjects from the left putamen and 20 from the right putamen were excluded due to FWHM > 10 Hz. Results are presented in Table 2. Due to the substantial loss of data in the putamen, we did not include glutamate levels derived from the putamen in any of the following analysis.www.nature.com/scientificreports/Association between glutamate and clinical scores.To understand the contributions of glutamate concentrations in different cortical regions to symptom expression, we first conducted two linear regression models with psychotic-like experiences and autistic traits as outcome variables and levels of glutamate derived from the ACC, left DLPFC and right DLPFC as predictor variables.Furthermore, we corrected for age and sex.
To investigate non-linear relationships between clinical traits and glutamate concentrations, we dichotomized the clinical scores and performed logistic regression analyses.This is important as there has been criticism of the hypothesis of a continuous development of neurobiological changes in psychosis from sub-threshold symptoms to severe psychotic symptoms 20,68,69 .To dichotomize the clinical scores, we performed a median split, defining a high and low risk group separately for psychotic-like experiences (median = 75) and autistic traits (median = 22).See Fig. 2 for distributions of glutamate by group, and Supplementary Table S3 for demographic and clinical differences.Using a logistic regression, we investigated whether any predictor variables (i.e., levels of glutamate derived from the ACC, left DLPFC and right DLPFC) significantly contributed to a change in the response variable (i.e., low vs high psychotic-like experiences) from 0 to 1 (i.e., from low to high psychotic-like experiences).Also, the logistic regressions were corrected for age and sex.The significance level was set to p < 0.05.Bonferroni multiple comparison corrections were applied, corrected significance levels are reported together with the results.Analyses were performed in R using the stats package version 4.0 70 .

Results
Association between psychotic-like experiences and glutamate concentration in the anterior cingulate cortex.We first fitted two multiple linear regression models to test if absolute glutamate concentrations in the ACC, left DLPFC and right DLPFC predicted symptom scores.The first fitted regression model was: psychotic-like experiences ~ Glu DLPFC_R + Glu DLPFC L + Glu ACC + age + sex.The overall regression was not significant (R2 = 0.15, F(5,46) = 1.63, p = 0.17).Importantly, however we found that levels of glutamate in the ACC significantly predicted psychotic-like experiences (β = 9.72, 95% CI [2.17, 17.27] p = 0.013) (Fig. 3).Autistic traits however were not predicted by levels of glutamate.
Supporting this finding, the binomial logistic regression (PLE-group ~ Glu DLPFC_R + Glu DLPFC L + Glu ACC + age + sex) revealed that while holding all other predictor variables constant, the odds of an individual belonging to the high psychotic-like experiences group increased by 33.1% (β = 0.49, 95% CI [0.01, 2.70], p = 0.014) for a one-unit increase in ACC glutamate.We did not see this effect of glutamate on autistic traits.The effects of the linear and logistic regressions both survived Bonferroni multiple comparison correction (level of significance, p < 0.025, correcting for two tests each).As many studies use Glx (glutamate + glutamine) to measure glutamate, we computed the same analysis using Glx, and replicated our results, which are presented in the Supplementary Materials.Also see Supplementary Table S1 for an overview on levels of Glx.Furthermore, we also computed the same analysis using a novel RStudio based MRS analysis tool-spant 71 .Again, the results are similar, although only marginally significant, and are presented in the Supplementary Materials, together with a detailed description of the method and analysis (Supplementary Table S2, Supplementary Figs.S2B, S3).

Discussion
In this study, we explored if and how glutamate concentrations in the ACC and left and right DLPFC and putamen are associated with autistic traits and psychotic-like experiences in a sample of healthy individuals.We found that glutamate concentrations in the ACC predicted psychotic-like experiences in healthy individuals, but they did not predict autistic traits.Supporting this finding, we furthermore found that glutamate concentrations in the ACC contributed significantly to the determination of group status (i.e., low psychotic-like experiences vs high psychotic-like experiences) when applying a median split to the clinical data.This was not the case for autistic traits.Regarding the glutamate concentrations in the DLPFC and the putamen, however, we found that whereas the DLPFC did not yield any significant results, levels of glutamate could not be reliably estimated in the putamen.www.nature.com/scientificreports/Association between ACC glutamate and psychotic-like experiences.ACC glutamate levels have been associated with psychotic-like experiences in healthy individuals 72 , symptoms and structural changes in high-risk individuals 73 , symptoms in first episode psychosis [73][74][75] and chronic psychosis 76 .Importantly, our results provide support for both a dimensional and a non-linear relationship, showing a positive linear relationship between the severity of psychotic-like experiences and ACC glutamate across the complete sample; importantly however, ACC glutamate is also predictive of whether or not an individual classifies as a high-risk versus a lowrisk individual, indicating qualitative differences in the neuropathology of the different disease stages 69 .This however needs to be explored in future research.
In general, evidence on altered levels of ACC glutamate in the schizophrenia spectrum is inconsistent with some studies reporting higher levels 31,74 , while others find reductions 29,30 or no differences 77,78 .Although Modinos and colleagues 78 did not find different levels of glutamate in the ACC between individuals with high schizotypy and low schizotypy (schizotypy and psychotic-like experiences have overlapping meanings, and are two concepts to describe the liability for psychosis 79,80 ), they found that increased grey matter volume in ACC was negatively related to ACC glutamate.This supports the view that alterations in glutamatergic levels may explain structural changes which are associated with the development of psychotic-like experiences.The association in this, but also in other studies 72 , however, suggests that decreasing levels of glutamate in the ACC are associated with specific pathologies in psychosis, whereas our study indicates that higher levels of ACC glutamate are associated with the development of psychotic-like experiences.Our results correspond to findings from Demro and colleagues 81 who reported that increased subclinical symptoms of grandiosity were linked to increased levels of glutamate in the ACC.Egerton and colleagues 82 found higher levels of glutamate/creatine-ratio in the ACC in symptomatic first episode psychosis patients compared to those in remission.In a different study, Egerton and colleagues 83 found that higher levels of glutamate metabolites in the ACC predicted treatment response, indicating that higher baseline levels are associated with poorer treatment response, while lower levels of ACC glutamate are predictive of improvements across positive and negative symptoms as well as general functioning in first episode psychosis patients.Relatedly, Godlewska and colleagues 27 reported lower levels of glutamate in the ACC in early psychosis patients compared to healthy controls.Importantly, however, the majority of patients included in this study were medicated, which confirms results from a recent meta-analysis 84 on the impact of anti-psychotic treatment on frontal glutamate levels.The authors found that antipsychotics were linked to a significant decrease in frontal Glx levels in both first episode and chronic schizophrenia patients, with the effect being stronger in first episode psychosis patients.As our results indicate that ACC glutamate levels are predictive of individuals belonging to the high psychotic-like experience group, the results are in line with the theory that increased glutamate is a sign of an acute psychotic or prodromal phase which builds up to a first episode of psychosis and is linked to inflammatory processes 31,32 .The theory further suggests that a hypofunction of NMDA receptors reduces the activity of inhibitory GABAergic interneurons, on which they are located.This activity reduction increases glutamatergic neurotransmission of pyramidal cells, and may contribute to the development of, especially, positive symptoms during an acute episode of psychosis 32,85 .
Lack of effect for other regions and autistic traits.Surprisingly, our data did not reveal any associations between the DLPFC and psychotic-like experiences.Generally, the literatures present an inconsistent picture regarding glutamate concentration in different regions.For example, although Godlewska and colleagues 27 reported alterations in glutamate in the ACC in early psychosis patients compared to healthy controls, they did not find differences in the putamen and DLPFC.However, especially the literature with regard to subclinical schizotypal traits is sparse, and mainly concentrates on ACC and hippocampal glutamate 72,78 .To our knowledge, only one other study investigated prefrontal glutamate in individuals with high schizotypy compared to low schizotypy 86 , in which the authors reported a reduction of glutamate in high schizotypy.
Furthermore, our data did not reveal any associations between glutamate concentrations and autistic traits despite several studies showing an interaction, which is at odds with results reported in patients suffering from ASD (e.g., 37,38,40,[42][43][44][45] ).Interestingly, studies exploring psychotic-like experiences and autistic traits together found similar patterns between both subclinical scores.For example, results by Ford and colleagues 87 suggest that social disorganization is associated with increased glutamate/GABA + ratio in the right superior temporal region across both the autistic and schizotypal spectrum.Whereas, Kondo and colleagues 88 report that autistic and schizotypal traits were associated with the Glutamine + Glutamate/GABA ratio in the auditory cortex but not in the frontal areas, including ACC, DLPFC and inferior frontal cortex.
Based on these studies and the fact that psychotic-like experiences and autistic traits are highly correlated, this might not be surprising at all.Our results however clearly indicate no such relationship between glutamate and autistic traits.We hypothesize that the relationship between changes in glutamate and autistic traits is mediated by psychotic-like experiences, which is rooted in their high correlation.To provide some initial evidence we computed an exploratory non-parametric causal mediation model, which is described in the supplementary materials.Indeed, our results revealed that the association between ACC glutamate and autistic traits is fully mediated by psychotic-like experiences.This may suggest that the high correlation between the two subclinical concepts results has a confounding on the relationship, and that underlying and often unassessed psychoticlike experiences may be the driving factor in the relationship between glutamatergic changes and autistic traits.However, more research is needed to explore this hypothesis.
Consistency of results using different measures of glutamate and a novel toolbox.In addition to our analysis using the LCModel implementation in Osprey and extracting absolute water and tissue corrected levels of glutamate (mol/kg), we also replicated our results using Glx as well as a novel r-implemented toolbox (spant 71 ).The replication of our results using different types of concentrations and methods further increases confidence in our results.Nevertheless, studies should focus on consistent ways of measuring metabolite differences, as it is not only the choice of location for voxel placement that differs grossly among studies but also the choice of metabolite representation (e.g., glutamate/GABA ratio, absolute glutamate, glutamate/creatine ratio, etc.) and selection of basis sets.

Limitations
This study has several limitations.First, we did not analyze levels of GABA in the voxels of interest, as no spectral editing was applied to the MRS sequence.Future studies should try to measure both metabolites reliably in order to understand interactions between neurotransmitters and regions.Second, due to high values of FWHM, glutamate concentrations estimated in the left and right putamen had to be excluded from the analysis.One possible explanation for this is that subcortical voxels are usually noisier.Higher resolution (e.g., 7 T) or spectral editing of the sequence should be applied to improve data quality.Third, we were unable to recruit participants with highly increased scores along both spectra.One possible explanation is that many individuals with highly increased scores would already have a clinical psychiatric diagnosis, which was an exclusion criterion in our study.Nevertheless, follow-up studies should recruit larger samples allowing for a larger spectrum.Fourth, our sample does not include many individuals with low SPQ and high AQ scores, and vice versa, which would be necessary in order to clearly differentiate between clinical traits.Yet, our sample represents the distribution of subclinical symptoms within the general population as confirmed by similar correlations of the two scores compared to previous studies 22,89 .Bigger samples would be favorable to also investigate the clinical subscores (e.g., positive-like experiences, negative-like experiences).

Conclusion
Taken together, this study shows that changes in glutamate in the ACC are associated with the manifestation of psychotic-like experiences but not autistic traits, which indicates that an imbalance in the glutamatergic neurotransmitter system involving the ACC may contribute to the development of psychotic-like experiences specifically.These results suggest that glutamate levels could potentially serve as a promising indicator for identifying individuals at an early stage who are susceptible to experiencing psychotic-like episodes.

Figure 1 .
Figure 1.Voxel placement and representative fitted MRS spectra.Placement of MRS voxels and 1 H-MRS spectrum fitted by Osprey of the (A) ACC, (B) DLPFC right, (C) DLPFC left, (D) putamen right, and (E) putamen left.The colors indicate the areas covered by the subjects' individually placed MRS voxels.The individual voxels were standardized with SPM12, overlapped in MRIcroGL and visualized in FSLeyes.ACC, anterior cingulate cortex; DLPFC, dorsolateral prefrontal cortex.

Figure 2 .
Figure 2. Glutamate distribution after a median split of SPQ and AQ.Differences of the glutamate levels in the ACC, left and right DLPFC after a (A) Median split of psychotic-like experiences with a median of SPQ = 75.(B) Median split of autistic traits with a median of AQ = 22.

Figure 3 .
Figure 3. Visualization of association between levels of glutamate in ACC and psychotic-like experiences, when correcting for age and sex.