Hypoconnectivity of insular resting-state networks in adolescents with Autism Spectrum Disorder

Abstract

Autism Spectrum Disorder (ASD) is characterized by deficits in social interaction and communication. The anterior insula (AI) participates in emotional salience detection; and the posterior insula (PI) participates in sensorimotor integration and response selection. Meta-analyses have noted insula-based aberrant connectivity within ASD. Given the observed social impairments in ASD and the role of the insula in social information processing (SIP), investigating functional organization of this structure in ASD is important. We investigated differences in resting-state functional connectivity (RSFC) using fMRI in male youths with (N=13; mean=14.6 years; range: 10.2-18.0 years) and without ASD (N=17; mean=14.5 years; range: 10.0-17.5 years). With seed-based FC measures, we compared RSFC in insular networks. Hypoconnectivity was observed in ASD (AI-superior frontal gyrus (SFG); AI-thalamus; PI-inferior parietal lobule (IPL); PI-fusiform gyrus (FG); PI-lentiform nucleus/putamen). Using the Social Communication Questionnaire (SCQ) to assess social functioning, regression analyses yielded negative correlations between SCQ scores and RSFC (AI-SFG; AI-thalamus; PI-FG; PI-IPL). Given the insula's connections to limbic regions, and its role in integrating external sensory stimuli with internal states, atypical activity in this structure may be associated with social deficits characterizing ASD. Our results suggest further investigation of the insula's role in SIP across a continuum of social abilities is needed.

Keywords: Anterior insula; Autism spectrum disorder; Neurodevelopment; Posterior insula; Resting-state functional connectivity; Salience network; Social communication questionnaire.

Fig. 1.

A–E – Anterior Insula Seed. (A–B) A significant difference (B, p = 1.2 * 10⁻⁴) in resting-state functional connectivity (RSFC) between the anterior insula (AI) and the superior frontal gyrus (SFG, *) was observed between the two groups. This difference was noted bilaterally (SFG coordinates: left −25, 43, 22; right 28, 45, 31). The ASD group had lower RSFC as compared to the non-clinical sample. (C–D) A significant difference (D, p = 1.7 * 10⁻⁴) in RSFC between the AI and the thalamus (**) were observed between the two groups (thalamus coordinates: 12, −16, 11). The ASD group had lower RSFC as compared to the non-clinical sample. (E) The green highlights the AI (‡) seed that was used for analysis. (Error bars are ± 1SE. ASD = blue, Control = green).

Fig. 2.

A–F – Posterior Insula Seed. (A–B) A significant difference (B, p=1.4 * 10⁻⁴) in RSFC between the posterior insula (PI) and the left fusiform gyrus (FG, *) in the temporal cluster were observed between the two groups (FG coordinates: −45, −50, −16). The non-clinical group had increased RSFC as compared to the ASD sample. (C–D) A significant difference (D, p = 4.2 * 10⁻⁴) in RSFC between the PI and the right lentiform nucleus/putamen (**) was observed between the two groups (Lentiform nucleus/putamen coordinates: 28, 0, −1). The non-clinical group had higher RSFC as compared to the ASD sample. (E-F) Significant differences (F, p = 4.5 * 10⁻⁴) in RSFC were observed between the PI and both the left and right inferior parietal lobule (IPL, †; coordinates: −53, −40, 39; 55, −40, 39, respectively). (G) The green highlights the PI (‡) seed that was used for analysis. (Error bars are ± 1SE. ASD = blue, Control = green).

Fig. 3.

A–D – SCQ and RSFC. A regression analysis using SCQ scores as a continuous variable was performed. SCQ scores were significantly correlated with RSFC (z-scores) between: (A) AI to SFG (R² = 0.253; p = 0.007), (B) AI to thalamus (R² = 0.244; p = 0.009), (C) PI to FG (R² = 0.380; p = 0.001), and (D) PI to IPL (R² = 0.237; p = 0.010).

Fig. 4.

A–C – Age (mean-centered) and RSFC in the ASD group. Mean-centered age significantly correlated with RSFC in the PI of the ASD group. The correlations were: (A) PI to FG (R²=0.746; p=0.003), (B) PI to lentiform nucleus/putamen (R²=0.812; p=0.001), and (C) PI to IPL (R²=0.723; p=0.005). All correlations p < 0.01 (Bonferroni correction: p = 0.05/5) are shown.