Neuroimaging studies consistently reveal cortical anatomical alterations in individuals with autism spectrum disorder (ASD). A large-scale ENIGMA study (3222 individuals, 1571 with ASD) showed increased cortical thickness in the frontal cortex and decreased thickness in the temporal cortex in individuals with ASD. Interestingly, these alterations extend beyond clinical diagnoses to the general population, supporting a continuum model of ASD, where autistic tendencies exist on a spectrum across the population. This continuum is further evidenced by the dimensional relationship between cortical thickness in frontal and parietal regions and genetic risk for ASD in the general population, which overlaps with cortical alterations in clinically diagnosed individuals. To understand the brain-wide mechanisms underlying ASD, this study examines the role of brain network architecture, focusing on hubs—regions with many connections serving as information relay centers—which have been implicated in various psychiatric disorders. Prior research has shown that disorder-related cortical alterations are often more pronounced in hub regions, potentially due to their high metabolic activity and connections to multiple brain networks. This study investigates the relationship between brain network architecture (specifically hubs and disease epicenters), ASD-related cortical alterations, and cortical correlates of polygenic risk for ASD within this continuum model. The researchers hypothesize that hub regions show a selective vulnerability mirroring ASD-related cortical alterations and polygenic risk correlates. They further investigate if polygenic risk for ASD is predictable from structural brain networks and whether the predictive models identify regions that coincide with ASD disease epicenters.

Previous research has extensively documented cortical anatomical alterations in individuals with ASD, including studies showing increased cortical thickness in frontal regions and decreased thickness in temporal regions. These findings extend beyond individuals with clinical diagnoses, supporting the notion of ASD as a continuum within the general population, with varying levels of autistic traits. Studies have explored the dimensional relationship between cortical thickness and genetic risk for ASD, finding overlaps between the general population and clinical samples. The investigation of underlying brain network architecture, especially hubs, has proven valuable in understanding other psychiatric disorders. Hub regions, characterized by high connectivity, are often disproportionately affected in brain disorders, possibly due to their high metabolic demands and widespread network participation. Identifying disease epicenters—regions whose network architecture plays a central role in the disorder's whole-brain manifestation—offers additional insights. This study builds on this existing research by explicitly examining the interplay between brain network architecture, cortical alterations in ASD, and polygenic risk for ASD within a continuum model.

This study utilized data from two publicly available databases: the Autism Brain Imaging Data Exchange (ABIDE) database, providing clinical data (560 male subjects: 266 with ASD, 294 controls), and the Pediatric Imaging, Neurocognition, and Genetics (PING) study, offering data from a general population sample (391 individuals). Polygenic risk scores (PRS) for ASD were computed for the PING dataset using GWAS data from 18,381 individuals with ASD and 27,969 controls. Cortical thickness was measured for both datasets using the CIVET processing pipeline, after rigorous quality control. Structural connectivity matrices were derived from diffusion MRI data in the PING dataset using the FSL pipeline, identifying 62 regions of interest (ROIs). Hub regions were identified using degree centrality. General linear models (GLMs) were used to assess ASD-related cortical alterations (group differences in cortical thickness between ASD and control groups in the ABIDE dataset) and the association of polygenic risk for ASD with cortical thickness (in the PING dataset). The interaction of age and these effects were also examined. Spin permutation tests were employed to evaluate regional overlap between the cortical difference maps and centrality maps. Finally, connectome predictive modeling (CPM) was used to predict PRS-ASD from structural connectivity data, and the top predictors were analyzed to determine if they overlapped with identified disease epicenters of ASD. Power analyses were conducted for both datasets.

The study revealed a strong association between brain network organization and ASD-related cortical alterations. ASD-related cortical thickness differences, as well as cortical correlates of polygenic risk for ASD, were significantly more pronounced in hub regions compared to non-hub regions (p < 0.05). Furthermore, the cross-sectional progression of ASD-related cortical alterations and polygenic risk-related cortical correlates showed stronger associations with cortical hubs (p < 0.05). Connectome predictive modeling successfully predicted polygenic risk for ASD from structural connectivity (r = 0.30, p < 0.0001), explaining approximately 9% of the variance. Two key brain regions (left inferior parietal and left supramarginal gyri) emerged as top predictors and were also identified as disease epicenters of ASD. These findings held true even after correcting for multiple testing using ComBat.

The findings support a critical role for brain network architecture in ASD, demonstrating its influence on both ASD-related cortical alterations and cortical correlates of polygenic risk. The stronger involvement of hub regions in these alterations is consistent with previous research on the role of hubs in brain disorders. The identification of disease epicenters, consistent with prior literature, further underscores the significance of network architecture. The successful prediction of polygenic risk for ASD from structural connectivity strongly supports the continuum model of ASD, suggesting that brain network organization reflects the spectrum of autistic traits extending beyond clinical diagnoses. The overlap between top predictive regions and identified disease epicenters reinforces this concept, highlighting the shared neural underpinnings across the continuum. The implicated regions, including the left inferior parietal and supramarginal gyri, play crucial roles in various cognitive functions, including sensory integration, language processing, social perception, and executive attention, linking network architecture to specific behavioral impairments observed in ASD.

This study demonstrates a crucial role of brain network architecture in ASD, bridging clinical diagnoses and polygenic risk within a continuum model. The findings highlight the importance of considering both polygenic risk scores and multi-modal neuroimaging to understand the interplay between genetic risk and brain alterations in ASD. Future research should investigate the underlying biological mechanisms linking network architecture to ASD, and explore the potential of connectome-based predictive modeling for personalized interventions.

The study primarily focused on male subjects, limiting the generalizability to female populations. The reliance on publicly available datasets may introduce limitations related to data heterogeneity and potential biases in data acquisition and processing. Further longitudinal studies are needed to confirm the observed cross-sectional findings and understand the developmental trajectory of these network alterations.