The present work represents an initial application of a protocol developed to examine the neural correlates of autistic traits as indexed by the SRS-A in neurotypical individuals using resting state fMRI. As hypothesized, we found that individual differences in autistic traits in a non-clinical population were related to pgACC FC. Several studies have linked individual behavioral performance to patterns of FC observed during rest (e.g., 34). The present work extends this by relating informant-based measures of social skills directly to pgACC connectivity.
Consistent with one of our two predictions, we found that autistic traits were related to pgACC connectivity with insula, although this relationship was specifically limited to the anterior mid-insula (aMI) rather than anterior insula per se. Discussing the implications of this finding requires considering the pattern of pgACC connectivity with the various insula sub-regions. The insula can be subdivided broadly into anterior, mid-insula and posterior insula, with the anterior insula comprising ventral (vAI) and dorsal regions (35
). The posterior insula, commonly involved in primary representations of visceral and somatic sensation (32
), showed no connectivity with pgACC, in agreement with a recent report (22
). In contrast, the anterior and mid-insula sub-regions, implicated in maintaining higher order representations of sensation and emotion (20
), exhibited a significant pattern of pgACC FC along an anterior/posterior gradient, extending from positive to negative connectivity. At the anterior-most extent of the insula, the ventral anterior insula (vAI), along with a broader network of structures commonly implicated in social cognition, such as the retrosplenial complex, was positively connected with the pgACC. In contrast, at the posterior extent of this gradient, the posterior mid-insula (pMI) was negatively connected with the pgACC, along with a network of structures implicated in somatic and self-focused attention (primary somatosensory cortices, superior parietal cortices and primary motor cortices). Between these, an intervening aMI region exhibited variable patterns of pgACC FC across participants. Thus, the pattern of differential pgACC connectivity in anterior and middle insula sub-regions appears to reflect differential involvement of the vAI and pMI in social cognition and somatic and self-focused attention, respectively.
These findings of differentiable mid-insula sub-regions are noteworthy when considered in the context of current models of insula organization, which tend to treat the mid-insula as a single sub-region (37
). Specifically, we found that the aMI appears to function as a transition zone between the vAI, which is related to social cognition, and the pMI, which is involved in somatic attention. As indicated by our secondary analyses, when a substantial portion of the aMI showed a positive connectivity pattern similar to that observed in vAI, low levels of autistic traits were observed. However, when the valence of FC in aMI resembled that of the pMI (i.e., was negatively correlated with pgACC), we observed elevated levels of autistic traits. These results have broad implications for future efforts to relate inter-individual differences in behavior to functional connectivity. Specifically, they highlight the need to interrogate connections exhibiting a high degree of variability across participants, as opposed to limiting analyses to a priori connections of interest, or to only those showing consistent patterns of connectivity across subjects.
We did not detect a significant relationship between SRS-A scores and pgACC FC with PCC. The PCC is strongly implicated in the development of theory of mind (38
), and consistent patterns of PCC hypofunction have been noted in ASD (18
). Further, initial studies of FC in ASD reported decreased connectivity between PCC and pgACC (14
). Our study used a non-clinical screened sample which may not recapitulate brainbehavior relationships encountered in clinically diagnosed individuals. Future work with affected individuals will help to clarify this divergence from the literature.
Finally, we also found that increased SRS-A ratings were related to increased FC between pgACC and higher order sensory processing regions (superior parietal gyrus, lateral occipital cortex and angular gyrus). These regions are often abnormally hyperactive in individuals with ASD (18
). As such, although not hypothesized, our findings suggest the need to consider abnormal patterns of connectivity beyond hypoconnectivity alone.
In considering the imaging approach employed in the present work, it is important to note that while SRS-A scores could also be related to task activation-based approaches, combining the SRS-A with resting state fMRI has several advantages. First, measuring resting-state FC avoids potential confounds associated with task-based approaches such as practice or floor/ceiling effects and sensitivities to experiment design parameters (9
). Second, resting-state data can simultaneously delineate entire multiple networks which, are usually only partially observable in task-based studies depending on the selected contrasts (8
). Despite concerns about the unconstrained nature of the resting state, patterns of intrinsic connectivity during rest have been found to be remarkably replicable across individuals and across research labs (39
). Furthermore, Shehzad et al. (24
) demonstrated the substantial quantitative stability of inter-individual differences in resting-state FC measures. Specifically, moderate (ICC > 0.5) to high (ICC 0.7 – 0.95) short- and long-term test-retest reliability was found for measures of resting state FC (24
). The present work provides further support for the applicability of resting state fMRI approaches in studies of inter-individual differences in FC, by demonstrating moderate to high test-retest reliability for key components of the pgACC network.
Our findings should be interpreted in light of several limitations. First, seed-based approaches to mapping FC require a priori selection of a seed of interest. We selected the pgACC due to its prominence in models of ASD-related abnormalities and its emergence in our recent meta-analysis of ASD. Future work will entail investigating the intrinsic connectivity of other regions of interest potentially implicated in the social impairments associated with ASD (18
). Second, we maximized the representativeness of individual parameter estimates by combining data from all available scans. Since 20% of the sample did not have three available scans, we considered the possibility that our results might have been influenced by differences in number of scans. However, we accounted for potential differences at the group level by including the number of scans as a covariate. We also repeated analyses limited to a single scan per participant and to the 20 participants who provided three scans; neither differed substantially from our main findings (see Supplementary Figure 1
). Finally, significant controversy continues to surround the interpretation of negative relationships in functional connectivity (40
). Thus, while patterns of positive and negative connectivity differentiate insula sub-regions, caution should be taken when interpreting the meaning of negative relationships.
In summary, the present work demonstrates the utility of resting state fMRI for mapping functional connectivity in relation to the SRS-A, a continuous measure of autistic traits in the general population. This approach led us to identify pgACC connectivity with the aMI as a candidate marker of social competence in a neurotypical sample. Application of this method in future studies examining patients with ASD should allow confirmation of whether this circuit is a locus of dysfunction in autism that is dimensionally related to the severity of autistic traits.