Analysis of the Resting-State fMRI Data of the EU-AIMS Longitudinal European Autism Project (LEAP)

Background: Studies applying resting-state functional magnetic resonance imaging (R-fMRI) to investigate functional connectivity in autism spectrum disorders (ASD) have revealed mixed results. Interpretability of findings is further limited by the use of small sample sizes and the focus on a limited number of resting-state networks (RSNs). In this study we used R-fMRI data from 486 participants of the multicentre Longitudinal European Autism Project (LEAP) to investigate ASD-related changes in functional connectivity of 20 large-scale RSNs.

Objectives: To provide a comprehensive characterization of ASD-related alterations in functional connectivity within and between 20 RSNs in a large ASD cohort using 1) classical case-control comparisons and 2) dimensional analyses based on autism severity scores.

Methods: Good quality R-fMRI data was available for 265 ASD participants and 221 controls (age range: 7-30 years). We derived 20 RSNs by applying independent component analysis (FSL Melodic) to R-fMRI data from 75 control participants. Next, we obtained the subject-specific RSNs for each of the remaining participants (FSL dual regression) and compared functional connectivity within these 20 RSNs between the control group and ASD group (categorical analysis). In addition, we correlated functional connectivity within the 20 RSNs to the total score of the social responsiveness scale (SRS) across all participants (to investigate functional connectivity across the ASD continuum; dimensional analysis). In both analyses we applied permutation testing (N=5000; FSL Randomise) and corrected for multiple comparisons (i.e., correcting for testing 20 RSNs using Bonferroni). In addition, we computed Pearson correlations between the timeseries of the 20 RSNs to investigate categorical and dimensional ASD-related changes in between-network connectivity (5000 permutations, FDR-corrected). In all analyses we corrected for effects of scan site, age, gender and head motion.

Results: We obtained 20 RSNs that included common sensory, motor, default-mode, and task-related networks (Figure 1). We did not observe differences in functional connectivity within the 20 RSNs between the ASD and control group. Yet we did observe a significant association in the dimensional analysis: functional connectivity within the medial motor network increased at higher SRS scores (i.e., higher ASD severity scores). Furthermore, the between-network analysis revealed that functional connectivity between the lateral visual network and the auditory, lateral motor, and somatomotor network was decreased in the ASD compared to control group (Table 1).

Conclusions: Our findings suggest that the integration of information from different sensory modalities is disturbed in ASD, which might lead to the abnormalities in sensory processing in ASD. The increased functional connectivity within the motor network might further be related to the repetitive motor behaviors in patients with ASD. Importantly, the relatively limited amount of ASD-related alterations in functional connectivity observed in this large cohort points to a potential large heterogeneity among ASD patients and warrants caution when generalizing results from small sample R-fMRI studies in ASD to the general population. Accordingly, future work will focus on stratification of ASD patients and the investigation of developmental effects in ASD.