Bias Towards within-Network Functional Connectivity Among Toddlers with ASD during Resting State

Friday, May 12, 2017: 5:00 PM-6:30 PM
Golden Gate Ballroom (Marriott Marquis Hotel)
M. C. Datko1, M. V. Lombardo2,3, L. T. Eyler1, K. Pierce1 and E. Courchesne1, (1)University of California, San Diego, San Diego, CA, (2)University of Cambridge, Cambridge, United Kingdom, (3)University of Cyprus, Nicosia, Cyprus
Background: While a small number of studies in younger participants suggest that atypical network connectivity patterns observed in adults with ASD begin much earlier in development (Dinstein et al., 2011; Lombardo et al., 2015; Shen et al., 2016), no studies have specifically investigated how the complex changes in within- and between-network connectivity dynamics previously observed in typically developing (TD) infants and toddlers (Gao et al., 2016) may differ in ASD. Of particular importance are examinations of resting state networks that develop in early life, such as primary sensory networks, as well as the default mode network (DMN), which is thought to be the cornerstone of self-referential processing and has previously been shown to have atypical connectivity in adults with ASD.

Objectives:  Using resting state fMRI, we compared brain network connectivity between a group of ASD toddlers (N=47, mean age=26.5, range=14.2-44.1) and matched group of TD toddlers (N=47, mean age=26.5, range=13.2-44.5). We first compared connectivity between nodes across several networks on a node-to-node basis. We then compared the balance of within-network to between-network connectivity. Finally, we tested whether the developmental trajectories of brain connectivity differ between ASD and TD groups over the age range of our participants.

Methods: With resting state fMRI time series extracted from the set of 264 spherical ROIs (Power et al., 2011), we conducted mass univariate testing using the Network Based Statistic method (Zalesky et al., 2010). Next, using resting state fMRI time series extracted from a set of 333 ROIs based on a functional cortical parcellation map (Gordon et al. 2014), we defined within-network connectivity (WNC) for each network as the average of correlations between each pair of seeds belonging to a network. We defined between-network connectivity (BNC) for each network as the average correlation between each seed of that network and each other seed in all other networks. The Within-Network Bias (WNB) for each network was then calculated as the percentage of the total strength of all of its connections (WNC+BNC) attributed to the strength of its WNC. Finally, we tested for interactions between age and WNB across networks and groups, using a GLM approach.

Results:  Using NBS, we found a graph component consisting of 25 connections between 24 seed ROIs in which ASD showed significantly lower connectivity compared to TD (FWER-corrected p=0.0368). The 24 ROIs belonged to the following networks: 8 in DMN, 7 in salience, 4 in visual, 2 in cingulate-operculum, 2 in auditory, and 1 in subcortical (Figure 1). ASD showed significantly higher WNB across all networks compared to TD (p=0.006, Figure 2). In a series of follow-up GLMs examining the interaction of participant age and WNB for each individual network, there was a significant interaction between group and age for the auditory network (p=0.0428).

Conclusions: We found evidence for hypoconnectivity between DMN, visual, and salience networks, and a global bias towards within-network connectivity in ASD toddlers. Our results demonstrate developmental differences of large-scale brain network connectivity between typically developing toddlers and those in the pre- or early clinical stages of ASD.