Sex-Modulated Atypical Resting-State Functional Connectivity in Autism: An Independent Component Analysis
Objectives: (1) To explore the atypical connectivity hypothesis of autism using functional connectivity (FC) analysis on resting-state functional magnetic resonance imaging (fMRI) data, using a non-biased, model-free, data-driven approach: independent component analysis (ICA); (2) To explore if biological sex modulates the characteristics of resting-state FC in autism.
Methods: Resting-state fMRI data (3T, 13 minutes with TR=1302 ms, eyes closed) from 117 individuals in four groups (males with autism N=25, neurotypical males N=33, females with autism N=30, neurotypical females N=29), matched for age, IQ and in-scanner head-motion, was analyzed with ICA using the Group ICA Toolbox (running under Matlab). We investigated 29 components, out of a total of 59 estimated independent components, that showed a high goodness-of-fit with known resting-state networks (for the remaining components, 17 were labelled undetermined and 13 likely reflected aliasing noise signal or artefact). To test group differences in within-component FC, two-way factorial ANCOVAs (factor 1: diagnosis [autism, neurotypical], factor 2: sex [male, female]) were performed for the 29 component maps of interest, with age and average frame-wise displacement as nuisance covariates. Using permutation tests, we further compared group differences in FC between all 46 non-artefactual components.
Results: For within-component FC, main effects of diagnosis and sex were found in both directions, on almost all components that were examined. In brief, we found functional decoupling between the precuneus and the anterior cingulate in the autism group, as well as decreased coupling between the basal ganglia and medial prefrontal cortex. However, diagnosis-by-sex interactions were almost ubiquitously seen in regions other than those showing main effects. For between-component FC, the default mode network components were more functionally connected to other components in the male autism group compared to the male neurotypical group. This difference, however, was not apparent in the female group.
Conclusions: Results revealed evidence for atypical but heterogeneous connectivity in high-functioning adults with autism compared to neurotypical adults. The directionality of differences in within-component FC varied substantially with components. In addition, biological sex significantly modulated the effect of diagnosis in most components. Overconnectivity between the default mode network components and other components was found in autism, but only in males. Overall these indicate a complicated picture of atypical connectivity in autism, which at the same time substantially differed between males and females. More fine-grained descriptions on patterns of atypical connectivity in different subgroups of autism are needed, as opposed to an over-simplified view that general hypo- or hyper-connectivity marks the atypical neurobiology of autism.