Are There Reproducible and Replicable Connectivity Patterns in the Intrinsic ASD Connectome?

Background:

Heterogenous reports of hypo- and/or hyper-connectivity affecting different neural circuits have cast doubts on the validity and utility of connectivity models of ASD. Recently the field has begun to amass the samples and approaches necessary to test robustness and reproducibility of findings. This is a critical step towards the testing and refinement of connectivity models.

Objectives:

To examine, using multiple and relatively large samples, the extent to which patterns of mixed hypo- and hyper- connectivity in ASD are consistent, are represented by distinct connectivity subtypes, and associated to specific functional neuronal systems, clinical impairments and biological sex.

Methods:

Three resting state fMRI (R-fMRI) studies of large-datasets examining the whole brain architecture will be presented. Study 1 examined functional connectivity cortical and subcortical units in an independent male-only sample selected from ABIDE II using the same criteria as the original ABIDE I report (N=298 ASD and N=339 typically developing controls [TC]). Study 2 (collaboration with T Yeo’s laboratory at U Singapore) applied a Bayesian model inge identification of latent distinct, but possibly overlapping, connectivity factors in a sample of both males and females (F) selected from ABIDE II and GENDAAR (N=306 ASD, N=348 TC). Associations with clinical measures were analyzed using canonical correlations. Study 3 examined the reproducibility of diagnosis, sex and sex by diagnostic interaction effects across multiple R-fMRI metrics in the same dataset (ABIDE I and II: N= 444 [82 F] ASD, N=576 TC [166 F]) analyzed with different pipelines and across independent samples after removing batch effects (GENDAAR: N=87 ASD [44 F], N=89 TC [53 F]; EUAIMS: N=173 ASD [43 F]; N=153 TC [48 F]). Analyses were conducted controlling for micromotion at the individual and group levels and using stringent statistical correction.

Results:

Study 1: consistent with prior results in ABIDE I sample, analyses revealed whole-brain pattern of coexisting hypo and hyper-connectivity in ASD relative to TC. Hypo-connectivity characterized cortico-cortical regions albeit with a lesser extent than previous findings. Hyperconnectivity affected subcortico-cortical circuits with primary somatomotor and unimodal association units being mostly affected. Study 2: Bayesian analyses revealed three dissociable whole-brain connectivity factors, each with coexisting ASD-related hypo and hyper connectivity. Importantly, most ASD participants expressed multiple factors with different degree of probability; each factor each was differently associated to core ASD, comorbid symptoms or sex. Study 3: While statistically significant effects of diagnosis, sex and their interactions identified in the ABIDE II dataset were largely robust to the preprocessing pipeline, their replicability across samples varied with by R-fMRI metrics and particularly for diagnostic effects.

Conclusions:

Mosaic patterns of co-existing hypo- and hyper- connectivity characterize ASD. Their variation in the specific network and metrics affected underscore the relevance of emerging data driven approaches accounting for heterogeneity. Findings that differential combination of brain functional connectivity represent each individual are consistent with models of continuous ASD heterogeneity. Challenges in identifying reproducible sex differences in the neural basis of ASD likely reflect the relatively limited availability of data in affected females and identifies a gap needing greater attention.