Reduced beta connectivity during emotional face processing in adolescents with autism

Autism spectrum disorder (ASD) is a neurodevelopmental disorder in which deficits in social cognition are one of the defining features. The ability to process facial expressions is critically important for social cognition; facial expressions are signals from the social environment and deficits in the ability to accurately perceive and process emotional expressions play a critical role in difficulties with social interactions. While it is generally understood that individuals with ASD experience difficulties with social cues, the current literature on emotional face processing in ASD has yielded inconsistent results, with some authors finding deficits in emotional processing [1‐5], whereas others report no deficits [4, 6‐8]. Magnetoencephalography (MEG) is a functional neuroimaging technique that allows for the examination of neural network synchronization. We used this approach to investigate the functional network connectivity underlying emotional face processing in ASD, which plays a crucial role in the social cognitive difficulties that are a hallmark of ASD.

Functional magnetic resonance imaging (fMRI) studies have reported atypical activation of social brain networks during emotional face processing in adults with ASD [9‐11], while event-related potential and MEG studies have demonstrated abnormal neural responses in youths and adults with ASD, relative to controls [12‐16]. Increasing evidence indicates that atypical structural and functional connectivity contribute to cognitive difficulties in ASD. Diffusion tensor imaging (DTI) indicates aberrant white matter connectivity in autism [17] (see [18] for review). Studies of functional interactions among brain areas using hemodynamic neuroimaging (fcMRI) also suggest altered intrinsic and task-dependent network connectivity (see [19] for review). Collectively, such findings lend increasing credence to the view that altered development of brain connectivity may be associated with social and cognitive difficulties in ASD [20‐23].

Phase synchronization of neural oscillations among brain areas has been proposed as a mechanism supporting communication in distributed neural networks underlying cognition and perception [24, 25]. Disruption of normal oscillatory network coherence is associated with various neurological and neuropsychiatric conditions [26], and accumulating evidence points to this as being critical for understanding neurodevelopmental disorders, including ASD [27, 28]. Studies of phase coherence among scalp electroencephalography (EEG) electrodes have indicated abnormal functional connectivity in ASD [29]. Reduced interregional MEG synchronization has also been reported in children with ASD during the performance of an executive set-shifting task [30]. Reduced interregional coordination of coupling between low- and high-frequency MEG oscillations has been reported among task-relevant brain areas during a face perception task in ASD [31]. It is not clear, however, whether phase synchronization of neuromagnetic oscillations in distributed networks is also atypical in individuals with ASD during social cognition.

Only a few fMRI studies have examined functional connectivity during affective face processing in ASD. Decreased connectivity between the fusiform gyri, other cortical areas (bilateral posterior cingulate gyri, left cuneus) and subcortical (left amygdala) structures during face processing in adults with ASD, relative to controls, has been observed [32]. Furthermore, atypical connectivity between the right amygdala and frontal and temporal regions has been shown in adolescents with ASD, compared to controls, in response to happy, sad and angry faces [33]. MEG affords both a unique and good combination of spatial and temporal resolution and is a direct measure of neural activity. In addition to both high spatial and temporal resolution, MEG is also silent; an advantage when considering studies in ASD given that sensory issues are known to be a symptom of the disorder. To date, there have been few MEG studies examining functional connectivity during the perception of emotional faces in ASD. One study reported reduced task-dependent local connectivity within the fusiform gyri, which was proportional to reduced long-range connectivity between the fusiform and left precuneus, left inferior frontal gyrus and left anterior cingulate cortex [31]. These results contrast previous hypotheses about increased local functional connectivity at the expense of reduced long-range connectivity [34‐36]. Furthermore, despite the importance of studying a paediatric population in order to understand how atypical patterns of connectivity arise in a neurodevelopmental disorder, a limited number of studies has addressed these issues in the developing population and the current literature remains largely focused on adults [32, 33].

Emotional face processing remains an effective method of assessing social cognition in ASD, as the ability to process emotional faces efficiently is integral to social cognition and successful social interactions. Happy and angry facial expressions were chosen as emotional expressions of interest, with the former acting as definitely positive displays of emotion and the latter as negative affective stimuli. Although some studies have reported intact processing of happy emotional information in individuals with ASD [37], investigating the neural connectivity underlying the processing of positive affect in ASD remains an important topic of examination, as individuals with ASD have been shown to have deficits in deriving social reward from happy faces [38]. We also employed angry, rather than fearful, faces as the processing of anger appears to involve a greater understanding of social norms, which has been found to be impaired in many individuals with ASD [39, 40]. Anger, while still considered a basic emotion, is usually exhibited in response to transgressions by another individual and both children with ASD and, to a greater extent, typically developing children can identify social reasons for anger [41]. Furthermore, individuals with ASD have shown difficulty in processing angry faces [42, 43]. Hence, focusing on the network connectivity underlying angry face processing in adolescents with ASD is an effective approach for understanding atypical affective processing and, subsequently, social cognitive deficits in ASD.

The present study investigated the perception of both neutral and emotional faces in adolescents with and without ASD to investigate the hypothesis that interregional network synchronization is atypical in ASD during the processing of emotional information in faces. Although the relations between neural connectivity and ASD as measured by oscillatory neural synchronization remain unclear, in light of accumulating evidence indicating impaired long-range connectivity in ASD (see [21] for review), we hypothesized that reduced long-range connectivity in networks related to emotional processing would be observed in adolescents with ASD.

We present the first demonstration of increased neuromagnetic interregional beta-band synchronization during face processing. We also provide the first evidence for reduced interregional beta phase synchronization during emotional face perception in adolescents with ASD, specifically to angry faces. These results are consistent and support previous findings. A reduction in beta phase synchronization only in response to angry faces may reflect the self-conscious nature of anger as an emotion, requiring the understanding of social norms and context, with which individuals with ASD struggle [41, 42, 65‐67]. In contrast, seemingly typical processing of happy affect in individuals with ASD was observed, which may be attributed to the greater frequency of encountering and familiarity with happy faces [10, 37]. The pattern of normative beta-band synchronization was predominantly characterized by connectivity between visual brain areas and ‘higher-order’ processing areas. This likely represents the transfer of information from brain areas relevant for vision to other task-relevant regions, including those responsible for face perception and affective processing. The pattern of reduced connectivity in adolescents with ASD did not reflect the normative pattern of task synchronization, but rather indicated reduced synchrony involving regions relevant for face perception and affective processing (the fusiform and insula, respectively). Accordingly, we interpret reduced task-dependent network synchronization in ASD as reflecting reduced communication in brain networks that contribute to social cognition. The fact that this reduction occurred within the first 200 ms and involved the fusiform and insula, as well as association areas such as the supramarginal gyrus and frontal areas, demonstrates the very rapid and complex processing of emotional faces is impaired in adolescents with ASD.

While the ability to perceive basic emotions is largely in place by six years of age, accuracy in discriminating between facial expressions, as well as recognition of more complex emotions, develops with age, continuing into the adolescent period [68‐72]. Maturation of emotional processing for unique emotions is staggered across development, with the ability to identify happy expressions maturing earliest while accurate identification of negative emotions, such as anger and fear, mature at later ages, suggesting that negative emotions require more complex processing [70]. Evidence that different emotions develop distinctly is found in reports of a significant increase in sensitivity towards anger from adolescence to adulthood, in contrast to fear, which increases linearly from late childhood into adulthood [73]. A developmental phase at around 11 years of age has been found, during which emotional processing abilities undergo marked improvement, suggesting greater demands on neural areas implicated in emotional processing in early adolescence [74]. The present results suggest that the adolescents with ASD are still impaired in the acquisition of the more difficult emotional expression, anger.

The reduced task-dependent synchronization to angry faces included the right fusiform gyrus, and in particular synchronization between the right insula and other brain areas. Further investigation of task-dependent beta connectivity involving the right insula using graph analysis revealed reduced task-dependent connectivity strength, clustering and eigenvector centrality in adolescents with ASD. These results indicate that ASD may be associated with a reduced ability to marshal the very rapid and complex network communication of brain regions critical for emotional processing; such functional disconnection may contribute to deficits in emotional face processing prevalent in this population. Cognitive and perceptual performance requires selection and functional integration of task-relevant neuronal populations, which may be distributed across brain areas [75]. Multiple lines of evidence indicate that neural synchronization is a mechanism mediating such communication dynamics in the brain [25, 28]. The dynamics of neural synchronization supporting various cognitive and perceptual processes can be effectively imaged in source space using MEG [76]. It is also clear that such task-dependent neuromagnetic synchronization is relevant for understanding individual differences in cognitive ability [77], as well as cognitive difficulties in clinical child populations [78].

Previous MEG research has implicated beta-band connectivity in visual perception [79]. Recent theories synthesizing results across studies and modalities have suggested that beta-band connectivity is particularly pertinent for establishing long-range communication among brain regions, and especially relevant for feedback interactions among brain areas [51, 80]. In light of this, the reduced beta synchronization in the ASD group in the present study could reflect inadequate re-entrant processing in task-dependent networks, involving areas such as the right insula, leading to an inability to effectively extract emotional information from faces.

Current literature implicates the insula as an interface between the frontal and limbic regions, playing a role in assigning emotional salience to perceived events (see [64] for a review) [81‐83]. Furthermore, atypical insula activity in individuals with ASD may underlie difficulties in emotional awareness of one’s own self, as well as others [84]. The anterior insula is recruited as part of a larger valence network responsible for attributing emotional salience [85], and is more likely to be hypoactive during social cognitive tasks in ASD [86]. This region also exhibits functional connectivity with the anterior cingulate cortex, a region widely implicated in emotion processing that also shows hypoactivity during social tasks [86‐88]. Decreased intrinsic functional connectivity between both the anterior and posterior insular cortices and other brain regions implicated in emotional processing has been noted in youths with ASD [83]. These findings are in concert with our current results and support the notion that altered beta-band synchronization between the insula and other neural regions is associated with affective processing deficits in ASD.

Of further interest was the insular lateralization in our findings. In typically developing individuals, emotional stimuli have been shown to elicit bilateral anterior and mid-insula activation, with left-hemisphere dominance in the anterior and mid-insula towards positively valenced stimuli, and bilateral activation in both regions towards negative stimuli [64]. Previously, a high degree of coupled activity between the left and right insula regions was reported in fMRI of typical adults [89], while decreased interhemispheric connectivity between these regions has been noted in ASD [90]. Only the right insula was found to be a hub of significantly reduced connectivity in the ASD group, showing reduced connections with areas including the right fusiform, right inferior temporal gyrus and superior frontal regions. Reduced early insula beta-band connectivity with the fusiform is of particular interest given the fusiform’s relevance for face processing; these results are in line with previous reports showing fusiform hypoactivation during emotional face processing in ASD [91].

A popular hypothesis is that reduced long-range functional connectivity and increased local connectivity plays a critical role in ASD [34‐36]. A body of literature supports the idea of long-range underconnectivity in ASD, however, a review of fcMRI studies in this field found that studies supporting the underconnectivity theory share methodological similarities that may account for the results [19]. Despite such findings, a recent MEG study examining task-related local connectivity through phase-amplitude coupling in the fusiform regions has found both local and long-range underconnectivity in ASD [31]. Furthermore, this study indicated that the strength of local reduction in connectivity was proportional to the reduction in long-range connectivity and was associated with the severity of autistic symptomology, lending credence to the theory of general underconnectivity in ASD [31]. The present study provides a concordant set of findings, and suggests that reductions in task-dependent long-range connectivity may contribute to difficulties in social cognition in ASD.

A limitation of the present study is that although group differences in beta-band synchronization were observed during the perception of angry faces, our approach of using NBS for characterizing network connectivity differences did not support direct analysis of interactions. Specifically, a group-by-emotion interaction could not be addressed in the current study design. Accordingly, it is possible that adolescents with ASD may express atypical network connectivity during the processing of happy and neutral faces, but that the current investigation was insufficiently powered to reveal these differences. A further potential limitation of the present study was the lack of behavioural corroboration for the observed between-group differences in neural networks. However, comparable performance indicates that differences in brain connectivity cannot be attributed to performance. Studies in adults have also found differences in functional connectivity between those with and without ASD without any behavioural group differences in happy and angry faces [33]. These results suggest that while behavioural performance is comparable, individuals with ASD may be processing emotional faces through the recruitment of alternate neural patterns of functional connectivity. Employing tasks with greater emotional and/or cognitive load may reveal behavioural differences with corresponding neural effects. While the accuracy limits of MEG in localizing deep brain sources are still evolving, multiple studies have shown MEG to be effective at localizing deep sources [88, 92] and interregional connectivity involving deep sources [93], thus our findings in this regard were not unexpected. Lastly, our clinical and control groups were not matched for IQ, with our ASD participants showing significantly lower IQ scores relative to controls. Future studies may examine emotional face processing focusing on individuals with ASD with normal IQs or including IQ-matched controls (see [94] for a discussion of this issue).

The present study provides the first evidence for reduced MEG beta-band synchronization to emotional faces in adolescents with ASD. During processing of angry faces, we demonstrated reduced task-dependent connectivity in graph theory through measurements of strength, clustering and eigenvector centrality, peaking between 170 and 200 ms in the right insula in the ASD group. These findings suggest that difficulties in emotional face perception in those with ASD may be associated with reduced communication among task-dependent brain regions, particularly involving the right insula; a region critical for affective processing. Future studies should explore these findings throughout childhood and thus study the development of these connectivity patterns associated with emotional face processing in ASD.