Longitudinal alterations in fronto-striatal glutamate are associated with functioning during inhibitory control in autism spectrum disorder and obsessive compulsive disorder

Background Autism spectrum disorder (ASD) and obsessive compulsive disorder (OCD) are neurodevelopmental disorders with overlapping symptomatology. Both show deficits in inhibitory control, which are associated with altered functioning and glutamate concentrations in the fronto-striatal circuitry. These parameters have never been examined together. Here we, for the first time, used a multi-center, longitudinal approach to investigate fronto-striatal functioning during an inhibitory control task and its association with fronto-striatal glutamate concentrations across these two disorders. Methods 74 adolescents with ASD (24) or OCD (15) and controls (35) aged 8-17 were recruited across three sites of the European TACTICS consortium. They underwent two magnetic resonance imaging (MRI) sessions with a one-year interval. This included proton magnetic resonance spectroscopy (1H-MRS; n=74) and functional MRI during an inhibitory control task (n=57). We used linear mixed effects models to investigate, over time, the relationship between fronto-striatal functioning and glutamate concentrations across these groups and continuous measures of overlapping compulsivity symptoms. Results During failed inhibitory control, in OCD increased striatal glutamate was associated with increased neural activation of ACC, an effect that decreased over time. During successful inhibitory control, higher ACC glutamate was positively associated with striatal activation in OCD and compulsivity across time. ACC glutamate levels decreased over time in the ASD group compared to controls, while striatal glutamate decreased over time, independent of diagnosis. Conclusions Significant differences in fronto-striatal glutamate were observed in ASD and OCD, affecting functional activity during failed- and successful inhibitory control differently, especially in OCD, with effects changing over time.


Introduction
Although autism spectrum disorder (ASD) and obsessive compulsive disorder (OCD) are two separate neurodevelopmental disorders with distinct diagnostic characteristics (1), they are highly comorbid and a comparison of symptoms has suggested that there are more similarities than differences between them (2)(3)(4). Individuals with these disorders typically show compulsive behaviors, which are defined as the repetitive, irresistible urge to perform certain behaviors or thoughts, and diminished control over this urge (5). Compulsive behaviors are associated with deficits in inhibitory control in tasks such as the stop-signal task (3,6). Fronto-striatal areas are known to be involved in inhibitory control, and are strongly regulated by the neurotransmitter glutamate (7)(8)(9). Structural, functional and neurochemical imaging studies show alterations across fronto-striatal brain regions in both ASD and OCD, which suggest a possible shared underlying mechanism affecting compulsive behaviors (10)(11)(12).
Compulsive behaviors are highly associated with deficits in inhibitory control, which in turn is regulated by the fronto-striatal circuitry, where particularly the anterior cingulate cortex (ACC) is crucial for exerting inhibitory control, and the striatum is thought to be important driving them (7,10,(12)(13)(14)(15)(16).
In studies using the stop signal task in ASD and OCD, there has been inconsistent results. Some studies have found no behavioral differences in ASD nor in OCD (17)(18)(19), while others have found worse performance in participants with OCD (5,6,(20)(21)(22), demonstrating deficits in inhibitory control. However, these differences are more commonly found in adults with OCD than children (23). Altered activity in fronto-striatal areas during response inhibition has been found in both disorders as well (18,24,25) and some studies found altered functional activity despite not finding behavioral differences in response inhibition compared to controls (26,27).
However, in a previous study using a partly overlapping sample of the current study, no behavioral or neural alterations were found during inhibitory control in participants with ASD and OCD (19).
Several studies suggest that altered glutamate concentrations in the fronto-striatal regions may be associated with deficits in inhibitory control. For instance, altered glutamate concentrations have been linked to repetitive behaviors and compulsivity (7,28), which seem to differ in individuals with ASD and OCD compared to controls across development. A metaanalysis of studies investigating fronto-striatal glutamate using proton magnetic resonance spectroscopy ( 1 H-MRS) in neurodevelopmental disorders reported that increased glutamate concentrations in striatum seemed to be present across both disorders (7). In the ACC, on the other hand, glutamate concentrations were often higher in children and adolescents with these disorders while in adults the opposite pattern was found, with lower concentrations compared to controls, suggesting a developmental shift (7).
In a study investigating glutamate concentrations and neural functioning during inhibitory control in an ADHD population, decreased ACC glutamate was associated with increased activity in the striatum (9). This strongly suggests that investigating the interplay between glutamate and functional activity during inhibitory control is an important step for understanding the mechanistic underpinnings of behaviors across neurodevelopmental disorders. In a study including the first time of measure (T1) of the participants in this study, increased ACC glutamate was found in both ASD and OCD, and a positive association between ACC glutamate and compulsivity was found (8). In the current study we followed part of this multi-center, multimodal sample up (T2). With this longitudinal data we aim to investigate whether fronto-striatal glutamatergic alterations and functioning during inhibitory control are stable across (atypical) neurodevelopment. In addition, we were interested in the relation between these parameters and expected fronto-striatal glutamatergic alterations to be negatively associated with neural activation during inhibitory control across ASD and OCD.

Participants
We included 74 participants (ASD = 24, OCD = 15, controls = 35) for the longitudinal 1 H-MRS analysis, who were between 8 and 16 years old at the first time of measurement, and between 9 and 17 years at the second measurement. The paper regarding T1 of our sample included a total amount of n=133 participants (Naaijen et al., 2017). Reasons for drop-out for this longitudinal study were loss of interest or quality issues regarding ) in the multicenter study COMPULS, part of the TACTICS consortium (www.tactics-project.eu). Another site was excluded due to too few participants surviving quality control (N=3). The inclusion criteria were IQ > 70, ability to speak and comprehend the native language of the location of recruitment and being of Caucasian descent (for further details, see (29)). To confirm DSM-IV-TR (30) diagnoses of ASD and OCD, we used the Autism Diagnostic Interview-Revised (ADI-R) (31) (and Children´s Yale Brown Obsessive Compulsive Scale (CYBOCS) (32) for ASD and OCD respectively. Control participants were confirmed to not score in the clinical range for any diagnoses using the Child Behavior Checklist (CBCL) and the Teacher Report Form (TRF) (33). Compulsive behaviors were measured using the Repetitive Behavior Scale -Revised (RBS-R) (34). Information on medication use was collected on the measurement days via parental report. Participants were asked to abstain from stimulant medication 48 hours before scanning. Ethical approval for the study was obtained for all centers separately and participants and their parents gave written informed consent for participation.

Stop-Signal Task
To measure inhibitory control participants completed a modified visual version of the stop-signal task (SST) (35) during an fMRI session. For details of the design of the task, see Figure 1. For details of behavioral measures and results of the behavioral analysis, see the supplementary material.

Image Acquisition
Participants were familiarized with the MRI settings and practice of the SST using a dummy scanner at T1. The data were acquired from the three study locations, all using 3 Tesla scanners (Siemens Details on the structural, functional and 1 H-MRS scan parameters can be found in Table 1.  Figure 2. Across all sites, the proton spectra were acquired using a point resolved spectroscopy sequence (PRESS) with the chemically selective suppression (CHESS) water suppression technique (38), details can be seen in Table 1.

Imaging Analysis
fMRI. As per regular praxis, the first five volumes from each scan were removed to account for equilibration effects. Head movement correction was performed by realigning to the middle volume (MCFLIRT; (39)). A Gaussian kernel with full width at half maximum (FWHM) of 6 mm 8 was used for grand mean scaling and spatial smoothing. ICA-AROMA (40,41) was then used to remove signal components related to secondary-head motion artefacts, subsequently followed by nuisance regression to remove signal from CSF and white matter (WM), and high-pass filtering (100 sec). These images were then co-registered to each participants' anatomical scan using boundary-based registration within FSL-FLIRT (42). The anatomical scans were spatially normalized using a 12-parameter affine registration to MNI152 standard space, using FSL-FNIRT (43). The images were then brought into standard space by applying the resulting warp fields to the concatenated functional image.
1 H-MRS. Glutamate concentrations were estimated using Linear Combination Model (LCModel) (44,45). Example fitted spectra for both ACC and striatum can be seen in Figure 2. As different tissues contain different amounts of water, correction for tissue percentage and partial volume effects was calculated using the formula:

Statistical analyses
All analyses were performed using the R-software package (46) unless otherwise described.
We investigated fronto-striatal glutamate concentrations, neural activation and behavioral responses during inhibitory control separately. We tested whether glutamate concentrations (in either ACC or striatum) were associated with diagnosis, RBS-R total, RBS-R compulsivity subscale scores, time of measure (T1 and T2) and their possible interactions using linear mixed effects models (lme4 package (47)). Age, sex and site were added as covariates of noninterest and participant as a random factor to account for within subject variability across time. As age and sex did not affect the results or increased the fit of the model they were removed from further analyses.
Neural activation during inhibitory control was analyzed using SPM12 (Statistical Parametric Mapping release 12, https://www.fil.ion.ucl.ac.uk/spm/). For the whole brain analysis of fMRI during the stop-task, the first level models included two contrasts of interest; (1) failed stop -successful go, to isolate failed inhibitory control and (2) successful -failed stop, to isolate successful inhibitory control. For the second level of analysis looking at differences across groups and times of measure, t-contrasts were applied to these contrast maps. Covariates of non-interest were age, sex and site, to control for its possible effects We additionally investigated whether the BOLD response in ACC and striatum during response inhibition was associated with glutamate concentrations in the same areas using  (47).

Demographics
No differences were found between the groups in age, IQ or sex. Both the RBS-R total score and all sub-scores were significantly different between controls and diagnostic groups where controls expectedly scored lower. There were no differences in the RBS total and sub-scores between T1 and T2. Table 2 shows an overview of the demographics and clinical variables of the largest subsample used.

Continuous measures.
No significant associations were found between the RBS total score, RBS compulsivity score and glutamate in ACC and striatum at either of the time-points separately nor longitudinally.

Stop Signal Task
All groups showed common patterns of brain activation during successful inhibitory control and failed inhibitory control, where there was activation in areas typically associated with inhibitory control, such as ACC and striatum (Figure 4). No significant differences in neural activation between groups nor timepoints nor their interaction was found in our contrasts of interest (all pvalues > 0.05). See the supplemental material for behavioral results.

Fronto-striatal functioning and glutamate
To investigate our main hypothesis regarding the association between fronto-striatal functioning and neurochemistry, we combined the 1 H-MRS and fMRI data by extracting both beta values from our regions of interest as well as glutamate concentrations. in ACC glutamate when compulsivity scores are higher). No effects were found for the ASD group or with RBS-R total scores (all p-values > 0.05). A visual representation of these results can be seen in Figure 6.

Discussion
This is the first study that used a multi-center, longitudinal, transdiagnostic approach to investigate the associations of compulsivity with fronto-striatal glutamate concentrations and functioning during inhibitory control in a childhood/adolescent cross-disorder population.
Most interestingly, in our multi-modal analysis we found effects of fronto-striatal We further found a larger decrease in ACC glutamate over time in ASD compared to controls. Previous studies investigating children with ASD have shown higher glutamate concentrations in ACC (49)(50)(51), while studies looking at adults with ASD have found both lower and higher glutamate concentrations in ACC compared to controls (7,52). Our finding may therefore reflect changes in development into adulthood in ASD being different from development in controls. We found no such differences in the OCD group, although they did not significantly differ from the ASD group either, and previous studies with OCD have shown inconsistent results (7). This may be due to a larger heterogeneity in the disorder, and future studies considering possible subtypes of OCD may successfully disentangle such differing results. However, the previous study investigating an overlapping sample of our first wave measures (however, larger) at the first time-point (T1) found increased ACC glutamate in both ASD and OCD (8). Here we found no such differences over time, but we found effects of glutamate on neural activation during inhibitory control changing over time. This suggests there may also be differences in glutamate alterations during development between these disorders from childhood into adolescence. This, however, needs further investigation.
In the striatum we found that glutamate decreased over time in all groups. This is in line with the study that found no group differences in striatal glutamate during the first time of measure (8). Alterations in metabolite concentrations during development are known to occur in controls as well (53), and our finding may reflect such development in striatum, independent of a clinical diagnosis.
Our recent longitudinal TACTICS study on inhibitory control in ASD and OCD found improvements in SSRT over time, regardless of primary diagnosis (58). In our partly overlapping subsample in the current study, as shown in the supplementary material, males performed better than females. The fact that we did not find a general improvement may, however, be due to a lack of power and/or to a larger proportion of males in this subsample.
We found no whole-brain differences during inhibitory control in participants with ASD and OCD, which is also in concordance with the main findings from Gooskens and colleagues (58). However, other studies with similar behavioral results still found altered brain activation during inhibitory control (17,25,27).
Strengths of this study are combining categorical and dimensional analyses, as well as a longitudinal approach to investigate the relation between compulsivity with glutamate. Our study is also further strengthened by its multimodal nature, and investigating ASD and OCD together. There were also some limitations. Firstly, the OCD group was smaller than the ASD group, which may have led to less power and the possibility of false negatives. However, we still found significant associations with glutamate concentrations affecting functional activity in OCD.
Secondly, the percentage GM in the striatum decreased over time, suggesting worse voxel placement. However, these changes were not different across diagnostic groups and therefore probably did not affect our main findings. We additionally had a large overall loss of sample size, due to lack of longitudinal data in several participants. There are also difficulties performing multicenter studies, where data quality may differ across sites. However, we did manage to control for these effects in our models and our results may not have been affected by left-over site effects.
The time between measures in this study was about one year. For future studies we suggest using an even longer time period between measures, extending into older ages, for an increased understanding of further development in ASD, OCD and compulsive behavior.
In conclusion, we found significant associations of increased glutamate concentrations in striatum, with decreased functional activity in ACC during failed inhibitory control in OCD, and that over time this effect changed differently in OCD compared to controls.
We also found that ACC glutamate had a larger effect on striatal neural activation in participants with increased compulsive behavior (indicated by both the diagnostic and continuous finding) during successful inhibitory control. These results should be replicated in an independent sample, but this study has given new insights into the alterations of glutamate in ASD and OCD during development in adolescence, and its role in functional activity. to the grants and relationships noted earlier.    controls. The asterisks in B indicate significant difference in striatal glutamate over time independently of diagnosis. Plots were made using ggplot2 (54) and in-house adapted violin plots (55). Note: this figure shows raw data-points, not model estimates.