Aberrant Striatal Functional Connectivity in Children with Autism

Introduction

Human brain development is characterized by progressive remodeling of the brain’s structural and functional architecture. Resting state functional connectivity (FC) approaches vividly capture these developmental changes. Consistent with histopathological and structural MRI findings (1;2), diffuse patterns of local FC are gradually replaced by a distributed architecture, dominated by focal long-range connections (3-6). Concurrently, the strength of FC between cortical and subcortical regions progressively decreases (6), suggesting decreased subcortical influence over cortical functioning with age. Disturbances in these maturational processes are increasingly implicated in neurodevelopmental conditions such as autism spectrum disorders (ASD) (7).

Models of ASD as a brain dysconnection syndrome (8-13) have emerged from analyses of abnormalities in cortico-cortical connections. Studies of FC in ASD have found evidence of reduced long-range FC in various cortical circuits (14-24). Since development is characterized by an overall increase in long-range functional connections (3-5;25), these studies suggest dysmaturation as a core feature of ASD pathophysiology. Examples of ASD hypoconnectivity include findings of reduced FC in fronto-parietal circuits supporting working memory (26), in medial-wall circuitry implicated in social cognition (27-29), and in a left inferior frontal gyrus-based network supporting language processes (10). Although specific affected circuits vary depending on the process examined and approach (15-23;30;31), the common denominator has been evidence of abnormal cortico-cortical circuitry. Partly, this may reflect a greater focus on higher-order cognitive and social processes, largely subserved by cortical regions, found to be abnormal in individuals with ASD (32;33).

Here, we draw attention to subcortical-cortical interactions as a potential locus of dysfunction in ASD. We focus on striatal circuits that contribute to maturational changes in subcortical-cortical FC (6). Although infrequently examined, abnormal subcortical-cortical FC has been detected in ASD. Specifically, a resting-state positron-emission-tomography study found weaker correlations in glucose consumption between frontal regions and both striatum and thalamus in young adults with ASD compared to typical adults (TA) (34). In contrast, in a task-based fMRI study, increased striatal FC with frontal, parietal, and occipital lobes was reported in adults with ASD during a sensorimotor task, relative to TA (35). These divergences may reflect differences in imaging modalities, design (i.e., presence vs. absence of a task), analyses, or small sample sizes.

While direct examinations of striatal FC in ASD are limited, ASD-related structural and functional impairments have been documented, particularly, but not exclusively (36), in the caudate. Morphological studies have generally reported increased caudate volume in ASD (37;38). Individuals with ASD showed striatal hypoactivation relative to healthy controls in studies of facial expression imitation (39) and cognitive flexibility (40). On the other hand, caudate hyperactivation was evident in a study of sensorimotor processes in ASD (41). Regardless of the direction of the aberrant striatal activation (hypo- or hyper-), cortical components of striatal circuits, such as dorsolateral prefrontal cortex, dorsal anterior cingulate cortex, and inferior parietal sulcus, exhibited the same type of aberrant activation (39-41). Overall, this literature suggests abnormal striatal-cortical circuits in ASD and highlights the need for their direct investigation in this population.

Using resting state fMRI (R-fMRI), we previously established that seed-based FC analyses delineate detailed maps of striatal circuitry (42). We employed seeds in three caudate and three putamen regions, adapted from a meta-analysis of the task-based literature (43). Independent applications of the same approach yielded similar results (44;45). Accordingly, we used R-fMRI to comprehensively examine striatal functional circuitry in school-age children with ASD compared to typically developing children (TDC).

Consistent with the notion that ASD are characterized by neurodevelopmental dysmaturity (7;10;32;46-49), initial R-fMRI studies of adults with ASD demonstrated reduced long-range FC in large-scale networks (3;e.g., 5;50). This resembles TDC functional architecture rather than that of TA (3;5;25), suggesting developmental immaturity in ASD. Given 1) our focus on children, 2) evidence of increased striatal-cortical FC in TDC relative to TA (6), and 3) initial findings of increased striatal-cortical FC in ASD (35), we predicted that children with ASD would exhibit immature FC development. Specifically, we anticipated that caudate and putamen circuits in ASD would show stronger and more diffuse FC with paralimbic, associative and sensory cortex compared to age-matched TDC. Based on the frequent findings of inappropriate patterns of co-activation in task-based studies (32;33), we also expected “ectopic” FC in ASD – FC between regions that are not typically functionally connected in pediatric or adult samples. Additionally, we conducted secondary analyses to contrast striatal FC in TDC and TA included in our earlier striatal study (42). In preparation for large-scale longitudinal studies, such analyses allow interpreting the findings of aberrant striatal FC in children with ASD in the context of possible deviance from typical development.

Methods

Results

Discussion

Our seed-based resting state fMRI examination of striatal functional architecture in school-age children with ASD revealed a widespread pattern of excessive functional connectivity (FC) in striatal-cortical circuitry, relative to TDC. Excessive striatal FC was evident between nearly all striatal regions examined and a variety of heteromodal associative and limbic cortices previously implicated in the pathophysiology of ASD, including the right STG and insular cortex. We also found evidence of striatal functional hyperconnectivity with the pons, thus expanding the scope of functional alterations implicated in ASD. Secondary FC analyses focusing on this brainstem area revealed broad patterns of ASD-related hyperconnectivity with bilateral insular cortices.

One of the challenges facing fMRI studies of neurodevelopmental disorders is ascertaining the pathophysiological processes underlying increased or decreased FC. Functional hyperconnectivity in clinical populations is increasingly interpreted as reflecting delayed or stunted maturational processes. For example, increased local and reduced long-range FC- characteristic of immature development- has been reported to be widespread in Tourette’s disorder (along with two anomalous circuits) (67) and was also found in ADHD (68). Here, the only evidence suggesting developmental immaturity was ASD-related increases in FC between VSi and right insula/STG. Beyond this single example, the remaining ASD-related differences were suggestive of ectopic FC – patterns of inappropriate, rather than residual FC. Most aberrant FC observed in ASD was with regions not correlated with the striatal seeds for either TDC or TA. The idea of ectopic FC is consistent with the task-based literature, which has shown that participants with ASD exhibit inappropriate patterns of task-activation in sensory (e.g., occipital cortex), motor (e.g., supplemental motor cortex) and heteromodal association areas (e.g., ventromedial prefrontal cortex) not recruited by the specific tasks examined in neurotypical populations (11;32). Although intriguing, this model remains speculative until validated by cross-sectional and longitudinal studies beginning in early development. These studies will elucidate whether increased striatal FC is specific to school-age children or extends to younger and older children with ASD. For young children with ASD, no FC studies are currently available, while for older children, findings of mostly cortico-cortical decreased FC have been reported (e.g., 10;69), albeit with exceptions (28;35;70).

In our data, ASD-related differences in the FC of insular cortex were prominent, being evident in both striatal and brainstem analyses. These findings are timely; a recent meta-analysis of functional imaging studies of autism highlighted ASD-related insula hypoactivation in studies examining social processing (30), prompting calls for increased attention to insula dysfunction in models of autism (71). In considering the role of the insula in ASD, a logical step has been to focus on the anterior insula (AI) which is implicated in empathy (72), task-control (73), and more broadly, in the ability to integrate visceral, autonomic processes to guide behavior (74-76). However, recent work suggests that AI function may be more directly related to alexithymia, a phenomenon that is associated with, but distinct from, impaired empathy (77;78).

Our findings underscore the importance of extending the scope of ASD-related insular abnormalities beyond the AI. A meta-analysis of 1768 task-based functional imaging studies highlighted the heterogeneity of insular cortex and identified four functionally distinct areas (76). These are ventral and dorsal AI areas, associated with empathy and cognitive functions respectively, a posterior region associated with sensorimotor processes, and an intermediate region involved in olfactory-gustatory functions (76). Considering the potential role of insular circuits in ASD beyond empathy deficits may provide a more comprehensive model of ASD pathology. For example, posterior insular circuitry, implicated in bodily sensory processes (79), may play a role in the ASD-related abnormalities in sensory perception and integration (80). Further, it is necessary to understand the interactions among the different insular functional divisions, rather than focusing on any region or circuit in isolation. A recent examination of autistic traits in TA illustrates this point (30). In that study, the severity of autistic traits was directly related to the extent to which FC between mid-insula and pregenual ACC resembled that of the ventral AI (low autistic traits) as opposed to posterior insular regions (high autistic traits; 30).

Another cortical region hyperconnected with the striatum in ASD was the right STG, particularly its posterior division, which exhibited hyperconnectivity with both right VSi and vrP. Posterior divisions of the right STG (pSTG) have been implicated in perception of intentional movements such as eye gaze (81), salience of acoustic cues (82;83), prediction of reward based on other agents’ strategies (84), and live face-to-face interactions (85). Consistent with a role of pSTG in these processes, fMRI studies of ASD have demonstrated abnormal activation of the pSTG during tasks involving perception of intentional movements, particularly in the right hemisphere (86-88). Recent work showed that reductions in white matter volume and hypoactivation of the pSTG during a Stroop task positively correlated with levels of autistic traits in TA (89). ASD-related abnormalities in temporal gyri gray matter have also been noted (90). Increased FC between STG and frontal cortex in ASD emerged from a study of language processes (91), while R-fMRI studies of ASD are beginning to reveal STG abnormalities. For example, initial R-fMRI work has revealed ASD-related increases in FC between STG and PCC (28) and increased regional homogeneity of temporal cortices including STG (92). Here, we extend the range of STG dysconnectivity in ASD, by demonstrating excessive FC with striatal sub-regions. Given the role of the VSi in reward mechanisms, abnormal FC between these regions may underlie the hypothesized deficiency of social rewards in ASD (93).

We detected robust abnormalities in striato-brainstem and brainstem-cortical FC, based in the pons. The pons comprises a number of nuclei that support sensory and motor cranial nerves innervating the face and the reticular formation - a net of cell bodies and bundles of axons that extends from the spinal cord to the thalamus that is essential for heart beat, respiration, alertness, sleep and a wide array of basic sensory and motor processes (94). Noting the early ontogeny of brainstem development and evidence of increased prevalence of putative ASD cases associated with early exposure to teratogens affecting brainstem development (95), models of autism have posited brainstem abnormalities (66). Evidence from neuropathology (96) and electrophysiology (97) also implicate the brainstem in ASD; morphometric studies of individuals with ASD have revealed reduced grey and white matter in the pons, and brainstem more broadly (90;e.g., 98). Despite these convergent lines of evidence, brainstem dysfunction has been rarely considered in imaging studies of ASD. We hope our findings will encourage its inclusion in future studies.

Our findings should be interpreted in light of limitations. Although our sample size is on-par with functional imaging studies of ASD, larger samples are needed to account for the marked heterogeneity characteristic of ASD. Along with sample size, failure to detect robust relationships between FC and symptom severity may reflect our failure to sample symptoms beyond the cardinal diagnostic domains. Nevertheless, we tentatively identified relationships between right vrP/STG and dcP/brainstem FC and RRB scores. Future work will explore these brain/behavior relationships in greater depth. Given the role of striatal- and brainstem-based circuits in the generation of eye movements related to attention control, reported abnormal in ASD (e.g., 41), future studies should record eye movements during R-fMRI. Additionally, we included too few females to examine group by sex interactions. Incomplete coverage of the cerebellum in some participants prevented full analyses of striatal-cerebellar functional interactions. Finally, even though cortico-striatal circuits are mediated by the thalamus, we did not examine the thalamus directly because of current limitations on thalamic functional parcellation. Future work building on recent findings (99;100) should examine striatal-thalamic-cortico circuitry, taking advantage of ultra-high field imaging.

In summary, school-age children with ASD exhibited prominent ectopic FC affecting striatal interactions with cortical and subcortical regions, most notably including the insula and brainstem. Our results support broadening the focus of examination in ASD to regions involved in early stages of brain development, such as the brainstem, insula, and striatum.

Supplementary Material