Neural Correlations Between Executive Function and Social Function in Autism Spectrum Disorder: A Mediation Analysis of fMRI

Anyi Zhang Peking University Sixth Hospital https://orcid.org/0000-0003-0673-9554 Le Shi Peking University Sixth Hospital Xiao Lin Peking University Sixth Hospital Suhua Chang Peking University Sixth Hospital Peng Li Peking University Sixth Hospital Yanping Bao Peking University Wei Yan Peking University Sixth Hospital Li Yang Peking University Sixth Hospital Jie Shi Peking University Jiahui Deng Peking University Sixth Hospital Lin Lu Peking University Sixth Hospital Jiajia Liu (  liujiajia_sdu@bjmu.edu.cn ) Peking University https://orcid.org/0000-0001-5548-8777


Background
Autism spectrum disorder (ASD) is a set of neurodevelopmental disorders that is characterized by an early-onset of social de cits, restricted interests and repetitive behaviors (1). Previously, children with ASD were always diagnosed at the age of 3-4 years, while, the atypical development of social function might be identi ed at 6-12 months old (1). The prevalence of ASD has increased from 0.01% in 1931 to nearly 2% in 2020 (2,3). Social function is essential for development of relationships and tness to the environment and emerges gradually and dynamically through childhood and adolescence (4). As we live in a largely social constructed world, the social interaction with others was very important due to its contributions for survival and reproductive tness (5). While, the inappropriate or absence of social behaviors in ASD patients resulted in signi cant disability. Disruptions of social skills and behaviors might be related to psychological stress, low self-esteem, and de cits in executive function (EF) (4,6). In addition to the core clinical characteristics in ASDs, de cit in EF was also described as a prevalence feature (7). Executive function comprises a set of higher cognitive functions that mediate goal-directed activities(8). The cognitive components that are involved in EF include inhibition, emotional control, exibility, working memory, set shifting, planning, and organization (9). Studies previously described that de cit in EF might affect the mental and physical health, school life and employment, cognitive development and social process (10). Evidence has shown that executive dysfunction might contribute to the dysfunction of other main characteristics of ASD individual, including social interaction, adaptive behavior, and school performance (11)(12)(13)(14).
ASD patients might have di culty inhibiting information that would interrupt e cient social and cognitive behavior (15,16). Moreover, EF was shown to impact adaptation and internalization in patients with ASD(6). ASD patients with typica social de cits often fail to adjust their behavior to the environment, which might be explained by executive dysfunction, thereby impacting many aspects across the lifespan including reading ability in childhood, resilience in adolescents, and social interaction in adulthood (17)(18)(19). Moreover, emotional control is an important ability in social communication. ASD individuals with executive dysfunction might present impairments in emotional control, which further contributes to social de cits (20). It has also been reported that improvement of EF might be a possible intervention target for the improvement in social functions in ASDs(6). The underlying mechanism that explains assessing this relationship, however, remains unclear.
Advances in brain imaging researches have helped us understand the neural mechanisms for many disorders, including ASD. In ASD, abnormalities that are associated with social de cits were mainly located in the "social brain," including the amygdala, insula, inferior frontal gyrus, posterior superior temporal sulcus, temporo-parietal junction and dorsomedial prefrontal cortex (21,22). According to a previous meta-analysis summarizing the fMRI characteristics of 655 ASD children, amygdala impairment was reported as a core abnormality in ASD patients that might contribute to social de cit, and dysfunctions in facial recognition (23). Additionally, neural-imaging evidences have suggested that executive dysfunction in ASD individuals might originate from the atypical activation and connectivity of the insular, cerebellum, right middle temporal cortex (24,25). Aberrant striatal function has been reported in autistic brain, as the two important components: caudate and putamen were involved in many cognitive processes (rewarding, reinforcement learning etc.) social interaction and stereotype behaviors(26). One of the potential targets for ASD patients reported in the previous study were stratum, which was functionally related to the social and rewarding (27). As a result, the neuro-imaging provided a crucial way to identify the underlying neural mechanism that mediated the effect of EF on social function in patients with ASD. Few studies have evaluated the underlying neural mechanisms that mediate the effect of EF on social function in patients with ASD, thus leaving a pressing need to better understand the neural bases of this association.
Moreover, there are limited pharmacological treatment for ASD patients, it is urgent to expand the intervention options and potential brain targets (27). In the present study, we (1) evaluated the association between executive dysfunction and social de cits in ASD patients and (2) identi ed the underlying neural mechanisms for this association by performing a regions-of-interests (ROIs)-based resting-state functional connectivity (FC) analysis to better inform potential interventions and therapeutic targets.

Participants
Participants were recruited from the Autism Brain Imaging Data Exchange II (ABIDEII) database, which was established to further promote scienti c discovery of the brain connectome in ASD(28). The ABIDEII database is publicly available and consists of data from 19 international sites, that share phenotypic characteristics, anatomical imaging, and resting-state functional magnetic resonance imaging (fMRI) of a total of 521 ASD patients and 593 healthy controls(28). The inclusion criteria for this study were the following: (1) patients diagnosed with ASD using the Diagnostic and Statistical Manual of Mental Disorders, 5th Edition, and con rmed by the Autism Diagnostic Observation Schedule (ADOS) and the Autism Diagnostic Interview, Revised (ADI-R), (2) ASD patients who were assessed for the social function using the Social Responsiveness Scale (SRS), and (3) ASD patients who were assessed for EF using the Behavior Rating Inventory of Executive Function (BRIEF). Participants were excluded if they lacked the following data: gender, age, handedness, full intelligence quotient (IQ), SRS score, or BRIEF score. Moreover, subjects were excluded if they presented excessive head motions (beyond 2 mm translation or 2° rotation) during the scan. A total of 186 patients with ASD, 5-18 years old (10.25 ± 2.72 years old), were included in this study, which came from ve datasets: Georgetown University, San Diego State University, New York University Langone Medical Center (samples 1 and 2), and Kennedy Krieger Institute.

Measures
Social Responsiveness Scale. The parent-reported SRS is a widely used measure to screen social function in ASD individuals aged 4-18 years old (29). Previous studies found that SRS scores were highly related to ADI-R scores (30,31). The SRS is a 65-item 4point Likert scale, from 0 (never true) to 3 (almost true). The SRS total score ranges from 0 to 195. The scale measures ve clusters of social function: social awareness, social cognition, social communication, social motivation and autistic mannerisms (32). We used the SRS total T-score in the mediation analyses. According to Bruni, SRS T-score higher than 76 was considered as severe impairment in social function (32).
Behavior Rating Inventory of Executive Function. Executive function was measured using the parent-reported BRIEF. The BRIEF has 86 items on a 3-point Likert scale, ranging from 1 (never) to 3 (often). The BRIEF is helpful for understanding children's behaviors in different environments (33). The BRIEF consists of two broader indices (Behavioral Regulation Index [BRI] and Metacognition Index [MI]), and eight subscales (inhibition, shift, emotional control, initiate, working memory, plan/organize, and organization of materials). The T-score of Global Executive Composite (GEC) was used in the mediation analyses. Patients with a T score ≥ 65 represented clinically signi cant EF problems, and higher scores meant poorer EF (34).
Full IQ. The tests that were used to assess full IQ included Different Ability Scale-2nd edition (DAS-II), Wechsler Abbreviated Scale of Intelligence (WASI), Wechsler Abbreviated Scale of Intelligence-2nd edition (WASI-II), and Wechsler Intelligence Scale of Children -4th and 5th editions (WISC-IV, WISC-V). Consistency among these IQ tests was identi ed in previous studies (35,36).

Magnetic resonance imaging data preprocessing
Regions-of-interests selection. In our study, we selected bilateral amygdala, putamen and caudate as the core ROIs. According to a previous meta-analysis, core brain abnormalities in ASD patients were the amygdala, caudate and putamen, based on resting-state fMRI (37). Studies previously reported that the amygdala is an important hub in many cognitive processes (e.g., emotion, social, reward, etc.)(38, 39). Extensive evidence also showed that amygdala abnormalities in ASD affect social behavior (40), Therefore, we chose the amygdala as an ROI. Additionally, the caudate and putamen are two key components of the striatum. Striatum abnormalities correlated with the severity of social de cits in ASD patients, and the striatum has been implicated in ASD pathogenesis (41). Thus, the caudate and putamen were also included as ROIs. Altogether, the six summarized ROIs were left amygdala, right amygdala, left caudate, right caudate, left putamen and right putamen, which were selected and the coordinate were extracted from the Anatomical Automatic Labeling (AAL) atlas (67×71×67) (42).
Imaging Preprocessing. We extracted the resting-state fMRI of each patients for further analysis. The detailed sequence parameters were shown in TableS1 (available online). Resting-state fMRI data preprocessing was conducted using the Data Processing and Analysis for (Resting-State) Brain Imaging (DPABI) toolbox in MATLAB (43). The rst four volumes of each subject were removed, and the remaining volumes were then slice-timing corrected, head motion realigned (participants with excessive head motion were excluded), spatially normalized to standard Montreal Neurological Institute (MNI) stereotaxic space, resampled to 3 × 3 × 3 mm 3 , smoothened with a 6 mm full-width at half-maximum (FWHM) Gaussian kernel and bandpass ltered (0.01Hz-0.1Hz).
FC measure. Next, we calculated the ROIs-based resting-state FC in all participants using the DPABI toolbox as well. Whole brain FC was calculated by correlating the time series of the ROIs and the rest brain regions in AAL 90 atlas, and we extracted the z transformed FC to con rm normality (6×89 brain regions with 534 edges for all subjects) using the DPABI toolbox.

Statistical analysis
We rstly analyzed the correlation of executive dysfunction and social de cit in ASD children with ASD using the Pearson correlation analysis with Bonferroni correction after controlling covariates, including age, gender, full IQ, and current medication.
Next, whether FC correlated with executive dysfunction and social de cits separately. Brain regions that were related to social de cits and executive dysfunction were assessed using Pearson correlation analysis. A General Linear Model (GLM) that controlled covariates (i.e., gender, full IQ, handedness, and current medication [0 for not currently taking medication, 1 for currently taking medication]) were analyzed to further verify the correlation. To address multiple comparisons, we also controlled the False Discovery Rate (FDR). The FCs with p < 0.05 (FDR corrected) were considered as signi cant.
We also identi ed shared FC links that simultaneously correlated with EF and social function in ASD patients. We applied model 4 with 5,000 bias-corrected bootstrap resamples in PROCESS to explore whether the shared FC links signi cantly mediated the association between executive dysfunction (independent variable) and social de cits (output variable) in ASD children (44). The 95% con dence interval (CI) didn't contain 0 was considered as signi cant in the mediation analysis.

Demographic characteristics
The nal participants in our study included 186 ASD children, aged 5-18 years old (mean = 10.25 years, standard deviation [SD] = 2.73 years), from ve datasets. Most of them were boys (n = 154, 82.8%). The participants' full IQs ranged from 63 to 149 (mean = 107.54, SD = 17.32). The children in our study were mostly right-handed (78%, 145 of 186). The prevalence of severe social de cits was 50.5% (94 of 186) among the participants. The prevalence of executive dysfunction was 66.7% (124 of 186; Table 1). Correlation between social function and EF in ASD After controlling covariates including age, gender, full IQ, and current medication, social functions in ASD children were positively associated with the EF. Higher scores on the BRIEF BRI, MI, and GEC in children with ASD were related to higher SRS scores. As shown in Table 2, a positive relationship was found between almost all subscales of the SRS and BRIEF. However, correlations between social motivation and inhibition, social cognition, social communication, social motivation and autistic mannerism with organization were not signi cant. Functional connectivity links associated with social de cits Fourteen FC links were associated with social function, evaluated by the SRS, in ASD children (Table 3, Fig. 1). The Pearson correlation analysis results were consistent with the results of the General Linear Models. We totally found fourteen FC links associated with social function in ASD individuals, including nine links for positive association that means children with an increased FC in those links would presented with a higher level of social de cit, and ve links for negative association that means the increased FC in those links were mostly founded in children with lower level of social de cit. Details were shown in Table 3 and Fig. 1. Correlations between SRS scores and FC links were analyzed using Pearson correlation analysis and General Linear Model analysis after controlling covariates, including age, gender, full IQ, handedness, and current medication.

Functional connectivity links associated with executive dysfunction
In the correlation analysis, we found that executive dysfunction in patients with ASD was associated with alterations of FC between 27 brain regions and 21 FC links after controlling covariates. As shown in Table 4 and Fig. 2, more effect regions were located at left hemisphere. In the Pearson correlation analysis, there were eleven FC links negatively related to the executive dysfunction in ASD individuals which means that children with an increased connectivity in those links would present a lower risk for executive function. While, there were thirteen FC links negatively related to the executive dysfunction in the GLM. In addition, the rest eight FC links were positively related to the executive dysfunction, that the increased connectivity might contribute to a decreased executive function, and higher risk for executive dysfunction. Details were shown in Table 4 and Fig. 2. Overlapping FC links for the association between social de cit and executive dysfunction and mediation analysis According to the correlation analysis, the shared FC links that were associated with SRS scores and BRIEF scores simultaneously in patients with ASD included the right amygdala and right inferior frontal gyrus opercular part, left caudate and right supramarginal gyrus, left putamen and right superior frontal gyrus medial orbital, right putamen and left gyrus rectus, and right putamen and left paracentral lobule (Fig. 3). We then assessed whether these ve FC links signi cantly mediated the effect of executive dysfunction on social de cits in patients with ASD using mediation analysis. We found one FC link that mediated the association between SRS scores and BRIEF scores. As shown in Fig. 4, connectivity between the right amygdala and right inferior frontal gyrus mediated the effects of BRIEF scores on SRS scores. The mediation effects of other FC links are shown in Fig. S1 (available online).

Discussion
The present study explored the association between social function and EF in children with ASD and found that the social de cit was positively associated with the executive dysfunction. We further identi ed the shared underlying brain FC link between right amygdala and right inferior frontal gyrus might mediate the effect of EF on social function. These ndings provided a new insight for understanding the neuro mechanisms of the co-occurrence of social de cit and executive dysfunction in ASD patients, and further inform the intervention target.
The present results con rmed that social de cits and executive dysfunction were positively related to each other in children with ASD, and social de cits in ASD children were predominately accompanied by executive dysfunction. These results were similar to previous ndings, which indicated that components of EF might contribute to social communication in ASD patients (6, 14, 45, 46). The two major components of EF were positively related to the social awareness, social cognition, and social communication in the present study. The association between organization of material in EF and social cognition, communication, motivation, autistic mannerisms, emotional control in EF and social awareness in social de cit, inhibition in EF and social motivation in social function were not statistically signi cant. As described in a previous study, the organization of materials domain was not signi cantly different between children with ASD and typically developing children (47). As a result, abnormalities in the organization of materials might not be a predictor of ASD or social function in children with ASD. Studies that evaluated inhibitory control in ASD children reported mixed results. A review reported that inhibitory control ability in ASD would predict future scholastic, occupational, and social interaction performance (10). Further studies are needed to clarify the association between inhibition and autistic mannerisms. The shift domain re ects the ability to change from one situation to another, and it was closely related to higher-order restricted and repetitive behaviors and interests(48). In the present study, we did not nd a statistically signi cant correlation between shift and autistic mannerisms.
The present results indicated that brain regions that are related to social de cits in children with ASD were mostly located in the supramarginal gyrus, which is also involved in visual experiences (49). The supramarginal gyrus and angular gyrus are two components of the inferior parietal cortex, which is an important hub in the default mode network (50). Alterations of activation of the supramarginal gyrus, inferior partial cortex, and default mode network were commonly reported in children with ASD (51,52). Alterations of FC between the supramarginal gyrus and caudate were previously reported to be involved in other cognition processes, including hyperarousal symptoms in insomnia patients and secondary language vocabulary acquisition (53,54). In our study, it might also be related to social function in ASD children. Moreover, the supramarginal gyrus might be implicated in failure of the emotional self-other distinction in children with ASD (55). Numerous imaging studies suggested that FC abnormalities in the supramarginal gyrus are related to emotional processes, such as interoception, empathy, and attentional processing (55)(56)(57). Social-emotional dysfunctions are core symptoms in patients with ASD. Interventions or training that target emotional recognition and understanding might in turn impact social interaction in ASD patients (20,58). Other brain regions in the present study that were related to social function in ASD children included the frontal lobe and limbic lobe. Previous studies reported that the inferior frontal gyrus was abnormally activated in ASD children during emotional-social expression (59,60). Atypical FC in the limbic lobe and cingulate was also observed in ASD patients(61). The cingulate is an important component of the limbic lobe. It is a white matter joint of the frontal lobe, parietal lobe, and temporal lobe, and this brain region is related to EF, emotion, and pain(62). The hippocampus is another brain region in the limbic lobe that is related to memory and relational recoding(63). Changes in FC between the limbic lobe and amygdala and between the limbic lobe and putamen in the present study were consistent with a previous study that showed that social de cits in ASD patients might be related to atypical amygdala-cingulate connectivity(64).
Executive dysfunction is a common symptom in ASD patients. Brain regions that are associated with impairments in EF were predominantly located in the frontal lobe, including the superior part, inferior part, and middle part. Several studies provided evidence that the frontal lobe is a key brain region for EF. Damage to the frontal lobe might contribute to executive dysfunction in many disorders(65-68). The frontal lobe is involved various cognitive and emotional processes. Developmental changes in frontal lobe function occur during aging, which can impact executive control, decision-making, working memory, and inhibition(69). Individuals with ASD might exhibit the prolonged heterogeneity of frontal lobe dysfunction (70). In the present study, abnormal FC links were found between the frontal lobe and amygdala, caudate, and putamen in ASD children. Consistent with a previous study, alterations of frontal-striatal connectivity contribute to de cits in facial viewing, which is another core characteristic of children with ASD (71). The left gyrus rectus is a part of the inferior frontal gyrus (IFG) and involved in social interaction (72). This region was signi cantly associated with EF in children with ASD in the present study and previous studies (72,73). Correlations were found between the right amygdala and left superior occipital gyrus and between the left caudate and supramarginal gyrus in ASD children who presented executive dysfunction, which was previously reported in autistic subjects (74,75). According to Simard et al., the occipital gyrus is related to visual and cognitive processing, such as the identi cation of correct answers in a reasoning task (74).
Functional connectivity links that signi cantly mediated the association between social de cits and executive dysfunction included the right amygdala and right inferior frontal gyrus. The amygdala and inferior frontal gyrus were reported to play a role in social and emotional processes(76, 77). Atypical frontal lobe development and function have been described in many psychological disorders. The frontal lobe is involved in many cognitive progresses, including emotion, personality, self-awareness, social function, and attention(78). Abnormal frontal lobe-amygdala connectivity has been previously reported in ASD patients, which could contribute to social de cits (79). The atypical activation of these brain regions would prevent appropriate reactions to social or emotional stimuli (76). The results of a previous neuroimaging meta-analysis suggested that subjects with ASD exhibited aberrant activation in the bilateral inferior frontal gyrus and bilateral amygdala during face processing(80). Executive function was associated with the amygdala, the dorsolateral prefrontal cortex, and connectivity between core regions, including the amygdala and thalamus(81). Accumulating evidence suggests changes in activation of the amygdala and inferior frontal gyrus in ASD subjects and associations between FC and clinical characteristics. The present study found that alterations of FC between these two regions mediated the correlation between two characteristics of ASD-social de cits and executive dysfunction-which might reveal a neural mechanism of the pathogenesis of ASD. Furthermore, our ndings focused on the behavioral associated amygdalacircuits, which might expand the treatment routes for ASD subjects.
The present study has several limitations. First, all participants were ASD subjects. The co-occurrence of social de cits and executive dysfunction can also be observed in the general population of children. Unknown is whether the mediation effect we found is also appliable to typically developing children with social de cits and executive dysfunction. Second, we used various evaluations of social function and EF in children with ASD. We chose the SRS and BRIEF because correlations between SRS and BRIEF scores have been widely described in the literature. Further studies that evaluate these associations using other measurements are needed. Third, the majority of the participants were boys. The prevalence of autism is 4-5 times higher in males than in females.
In conclusion, we found a signi cant association between social de cits and executive dysfunction in ASD children. We also explored the mediation effect of FC links and found that alterations of FC between the amygdala and inferior frontal gyrus signi cantly mediated the correlation between clinical symptoms, which may further illustrate the underlying mechanism of ASD.

Availability of data and materials
All data used and analyzed in the current study are available from the open-source database, which is available at the website http://fcon_1000.projects.nitrc.org/indi/abide/abide_II.html.     Mediation effect of shared FC links for the association between SRS scores and BRIEF scores. SRS, Social Responsiveness Scale; BRIEF, Behavior Rating Inventory of Executive Function; eff, effect; SE, standard error; CI, con dence interval.