Animal care and husbandry
This study used male wild type (WT), Shank3+/-, and Shank3-/- littermate three-month-old rats. Shank3-deficient rats were generated using zinc-finger nucleases on the outbred Sprague-Dawley background, as previously described (10). All rats were kept under veterinary supervision in a 12 h reverse light/dark cycle at 22±2°C. Animals were pair-caged with food and water available ad libitum. All animal procedures were approved by the Institutional Animal Care and Use Committee at the Icahn School of Medicine at Mount Sinai.
MRI
All imaging was performed by the BioMedical Molecular Imaging Institute using a Bruker Biospec 70/30 7 T scanner with a B-GA12S gradient insert (gradient strength 440 mT/m and slew rate 3444 T/m/s). A Bruker 4-channel rat brain phased array was used for all data acquisition in conjunction with a Bruker volume transmit 86-cm coil. All rats were imaged on a heated bed and respiration was monitored continuously until the end of the scan. The animals were anesthetized using isoflurane anesthesia (3% induction and 1.5% maintenance). After a three-plane localizer, a field map was acquired and the rat brain was shimmed using Mapshim software. A DTI protocol was acquired with a Pulsed Gradient Spin Echo – Echo-planar imaging (EPI) sequence with the following parameters: repetition time (TR) = 5000 ms, echo time (TE) = 22.6 ms, 4 segments, 30 gradient directions with b-value = 1000 s/mm2 and 5 B0’s, field of view (FOV) = 25 mm, matrix = 128x128, slice thickness = 1 mm, skip = 0, 6 averages, total acquisition time = 1 hr. The voxel size was 0.195 x 0.195 x 1 mm3 (1000 µm-thick). A high resolution T2 anatomical scan was obtained with a 3D Rapid Acquisition with Relaxation Enhancement (RARE) sequence with a RARE factor of 8, TR = 777 ms, effective TE = 52 ms, FOV = 30 mm x 27.25 mm x 30 mm, matrix size 256 x 256 x 128. The voxel size was 0.117 x 0.117 x 0.234 mm3 (234 µm-thick).
MRI region-based analytical pipeline with manual editing
A magnetic resonance imaging processing pipeline was used to perform semi-automated nonbiased brain segmentation, while blinded to genotype (WT: N = 6, Shank3+/-: N = 10, Shank3-/-: N = 10) (11). The pipeline is composed of six major steps: rigid registration of images to each other, generation of a whole-brain mask for each image, averaging of all images, creation of a whole-brain mask for this averaged image, segmentation of the average mask by regions of interest (ROIs), parcellation propagation of the segmented mask to individual subjects, and ROI-based statistics for the individual images. The deformation necessary to warp each subject’s image to the average was used to calculate the volume of the ROIs. After each mask was generated, it was improved manually in ITK-SNAP (www.itksnap.org). The segmentation into ROIs was determined by a template that was hand-segmented into 32 brain regions, listed in Supplementary Table 1.
Automatically generated segmented masks for the individual images that did not closely match the average segmented mask (two WT and one Shank3-/- T2 mask) and individual data point outliers, defined as being outside 1.5 times the interquartile range, were excluded from the analysis. The whole-brain masks were used to determine whole-brain measures. Mean voxel intensity for each ROI and across the whole-brain was measured in both the T2 and DTI images and the volume of each ROI and whole-brain was calculated from the T2 images. Only white matter structures were included in the DTI analysis (see Supplementary Table 1). For each independent variable, if the distribution was nonparametric according to the Shapiro-Wilk’s test, a Kruskal-Wallis test was performed, and if the distribution was normally distributed, a two-way ANOVA was used. Due to the fact that many comparisons were made across ROIs, the output was then assessed for the ability to survive a correction for multiple comparisons with a Bonferroni correction. Genotype was the only between-groups factor. If an effect had a significant nominal p-value (0.05), pairwise comparisons were made. If the data were parametric, a Tukey HSD test was used to compare the individual means. If the data were nonparametric, a Dunn’s test was used for pairwise comparisons. An adjustment of the p-values was made to account for the additional comparisons. The effect size of each volumetric change was measured with a Cohen’s d. Custom scripts written in the R statistical programming environment were used for the statistical analyses (R Development Core Team, 2006).