Atypical measures of diffusion at the gray‐white matter boundary in autism spectrum disorder in adulthood

Abstract Autism spectrum disorder (ASD) is a highly complex neurodevelopmental condition that is accompanied by neuroanatomical differences on the macroscopic and microscopic level. Findings from histological, genetic, and more recently in vivo neuroimaging studies converge in suggesting that neuroanatomical abnormalities, specifically around the gray‐white matter (GWM) boundary, represent a crucial feature of ASD. However, no research has yet characterized the GWM boundary in ASD based on measures of diffusion. Here, we registered diffusion tensor imaging data to the structural T1‐weighted images of 92 adults with ASD and 92 matched neurotypical controls in order to examine between‐group differences and group‐by‐sex interactions in fractional anisotropy and mean diffusivity sampled at the GWM boundary, and at different sampling depths within the superficial white and into the gray matter. As hypothesized, we observed atypical diffusion at and around the GWM boundary in ASD, with between‐group differences and group‐by‐sex interactions depending on tissue class and sampling depth. Furthermore, we identified that altered diffusion at the GWM boundary partially (i.e., ~50%) overlapped with atypical gray‐white matter tissue contrast in ASD. Our study thus replicates and extends previous work highlighting the GWM boundary as a crucial target of neuropathology in ASD, and guides future work elucidating etiological mechanisms.


Quality Assessment
Initially, three raters blind to the diagnosis visually inspected the image quality for the surface models derived by FreeSurfer v6.0.0 of 266 structural T1-weighted images.
The image quality of the scans was rated on the basis of the following three categories: (1) good quality, i.e., no visible reconstruction errors or artifacts; scans of this category were accepted as is and no manual edits were performed; (2) with visible reconstruction errors in pial and/or white matter surface; for scans in this category manual edits were performed and scans were subsequently re-preprocessed and reassessed (n=78, i.e., 29% of the total sample; out of which n=28 did not improve after manual editing and were thus excluded); or (3) with gross anatomical abnormalities or severe acquisition or motion artifacts; scans of this category were excluded prior to the analysis (n=15, i.e., 6% of the total sample). Based on this quality assessment, the quality of the Diffusion Tensor Imaging (DTI) data of the remaining n=223 participants was visually inspected with all participants with a T1-weighted image of good quality also having a good quality DTI scan. Manual edits were only performed in regard to the T1-weighted images, as eddy current and motion correction of the DTI scans was already implemented in the framework of the preprocessing pipeline. After the quality assessments, we further excluded 39 participants in order to obtain two groups that were matched on age and full-scale IQ. This led to a final sample size of N=184 participants. Supplementary Table S1 (Beck & Steer, 1987); BAI: Beck Anxiety Inventory (Beck & Steer, 1993); OCI-R: Obsessive-Compulsive Inventory-Revised (Abramowitz & Deacon, 2006;Huppert et al., 2007); a based on N=178 (n=89 ASD, n=89 TDC); b based on N=176 (n=88 ASD, n=88 TDC); c based on N=176 (n=89 ASD, n=87 TDC); d based on N=177 (n=88 ASD, n=89 TDC); e based on N=179 (n=88 ASD, n=91 TDC). Supplementary Table S5. Clusters with a significant group-by-sex interaction in fractional anisotropy (FA) at the gray-white matter boundary (0%) and different sampling depths within the gray matter and superficial white matter

Supplementary Figures
Supplementary Figure S1. Nested Model Comparison comparing a general linear model without vs.
with a group-by-sex interaction term (N=184). Clusters with improvement of model fit by inclusion of a group-by-sex interaction term in the total sample (N=184) for (A) fractional anisotropy (FA), (B) mean diffusivity (MD), both sampled at the gray-white matter boundary, i.e., at 0% cortical thickness, and for (C) gray-white matter tissue contrast (GWC).
Supplementary Figure S2. Nested Model Comparison comparing a general linear model without vs.
with a group-by-sex interaction term (N=177). Clusters with improvement of model fit by inclusion of a group-by-sex interaction term in all participants, for whom the total score in the Beck Depression Inventory (BDI) was available (N=177), for (A) fractional anisotropy (FA), (B) mean diffusivity (MD), both sampled at the gray-white matter boundary, i.e., at 0% cortical thickness, and for (C) gray-white matter tissue contrast (GWC).
Supplementary Figure S3. Nested Model Comparison comparing a general linear model without vs.
with the total severity of depressive symptoms (i.e., total score in the Beck Depression Inventory [BDI]) as covariate (N=177). Clusters with improvement of model fit by inclusion of the total score in the BDI as covariate in all participants, for whom BDI total score was available (N=177), for (A) fractional anisotropy (FA), (B) mean diffusivity (MD), both sampled at the gray-white matter boundary, i.e., at 0% cortical thickness, and for (C) gray-white matter tissue contrast (GWC). Regions of increased and decreased MD in female individuals with autism spectrum disorder (ASD) compared to typically developing (TD) female controls at the gray-white matter (GWM) boundary (0%), at different cortical thickness (CT) projection fractions within the gray matter (i.e., 30% and 60% CT, sampled from the GWM boundary into the thickness of the cortical ribbon), and within the superficial white matter (sampled at absolute distances of -1mm and -2mm below the GWM boundary).  Figure S18. Association between atypical diffusion and reduced gray-white matter tissue contrast in autism spectrum disorder with clinical symptom severity. Correlation Plot depicting Pearson correlation coefficients between participants' full-scale intelligence quotient (IQ), total scores in the distinct questionnaires (i.e., Beck Depression Inventory (BDI), Autism Spectrum Quotient (AQ), Empathy Quotient (EQ), Systemizing Quotient (SQ), Beck Anxiety Inventory (BAI), and Obsessive-Compulsive Inventory-Revised (OCI-R)), and mean cluster values of the clusters obtained in the main effect of group analyses (autism spectrum disorder (ASD) vs. typically developing controls (TDC)) for fractional anisotropy (FA) and mean diffusivity (MD), sampled at the gray-white matter boundary (i.e., FA0 and MD0), as well as gray-white matter tissue contrast (GWC). Multiple testing was addressed by false discovery rate (FDR)-adjustment of p-values (one-tailed, p < .05). Correlations were based on N=169 participants (n=83 ASD and n=86 TDC), for whom the complete set of questionnaire information was available. Abbreviations: mOFC= medial orbitofrontal cortex; DLPFC= dorsolateral prefrontal cortex; mFG= middle frontal gyrus; sFG= superior frontal gyrus; lOFC= lateral orbitofrontal cortex.