We conducted MRIs on N = 131 children between 0.5 and 12 years of age: n = 32 children with AS and n = 99 neurotypical (NT) control participants. Total brain volumes (TBV) were generated in all scans between 0.5 and 12 years. White matter (WM) and gray matter (GM) volumes were generated in the subset of scans acquired between 1 and 12 years, owing to the ability to automatically segment between WM/GM tissues in MRI scans acquired over age 1 year. The final WM and GM analyses sample included N = 94 children: n = 28 AS (19 M, 9 F) and n = 66 age- and sex-matched NT controls (54 M, 12 F). Of these, 25 participants (13 AS; 12 NT) had a second longitudinal MRI scan, yielding a total of 119 scans in the WM and GM analyses (41 AS scans; 78 NT scans). T1- and T2-weighted structural MRI scans (1mm3 voxels) were acquired at UNC on 3T Siemens TIM Trio scanners with a 12-channel head coil. Children with AS were sedated using IV administration of propofol in the hospital MR suite. A pediatric anesthesiologist and certified registered nurse anesthetist implemented the standard hospital pediatric sedation protocol after consent from the parent/guardian and screening for contraindications. All images were reviewed for head motion and image quality by a reviewer blinded to genotype. Only images that passed quality control were used in the analysis. T1 and T2 images underwent registration, transformation to stereotactic space, and segmentation of total brain, WM, and GM volumes [23,24,25].

AS participants were confirmed with a chromosomal microarray showing a chromosomal deletion of the 15q11.2-13 region on the maternal allele, with deletion sizes ranging from 4.7 to 7.6 MB included in the WM and GM analyses. We excluded subjects with significantly larger deletions that extended beyond this region to ensure our WM findings specifically reflect the impact of typical AS deletions. NT participants were enrolled and had no first-degree relatives with a psychiatric diagnosis [26]. All subjects were excluded for the presence of: (a) diagnosis or physical signs of known genetic conditions or syndromes other than AS (e.g., significant dysmorphology, asymmetry on physical exam), significant medical or neurological conditions affecting growth, development or cognition (e.g., CNS infection, diabetes, tuberous sclerosis, congenital heart disease), or sensory impairments such as significant vision or hearing loss (or evidence of during the course of the study); (b) a history of significant perinatal adversity, exposure to in-utero neurotoxins (including alcohol, illicit drugs, selected prescription medications), or a history of maternal gestational diabetes; (c) contraindication for MRI (pacemaker, vascular stents, metallic ear tubes, other metal implants or braces); (d) families whose predominant home language is not English; and (e) children who were adopted. Parents of AS and NT individuals provided informed consent, and the institutional review board approved the research protocol.

The Ube3a mouse model was created in Dr. A. Beaudet’s laboratory [27]. This model represents specific UBE3A loss-of-function rather than the larger 15q11.2-13 deletions typically found in AS patients. While this model may not capture all aspects of AS genetics, it allows us to isolate UBE3A’s specific role in neurodevelopment. AS model mice (Ube3am–/p+), which maternally inherit the Ube3a knock-out allele, were generated by mating wild-type (Ube3am+/p+) male mice to female mice with paternal inheritance of the Ube3a knock-out allele (Ube3am+/p−, paternal null model), which themselves are phenotypically normal [27, 28]. Table 1 details the number of mice, their sex distribution, ages, and litter information for each experiment.

WT and AS mice at P14 were deeply anesthetized using isoflurane and transcardially perfused with heparinized saline (0.9% NaCl, 0.1% heparin), immediately followed by 10% neutral buffered formalin (NBF) containing 10% Gadoteridol (0.5 M) (Gd, ProHance, Bracco Diagnostics Inc.). The mice were then decapitated, and their heads were postfixed in 10% NBF overnight. The following day, mouse heads were transferred to 0.5% Gd in PBS for 14 days. Before imaging, samples were placed in custom-designed MR-compatible tubes and submerged in Fomblin (Fomblin perfluoropolyether, Ausimont, Thorofare, NJ, USA) to minimize susceptibility artifacts and prevent sample dehydration.

Diffusion Tensor Imaging (DTI) was performed using a 9.4 Tesla scanner at the UNC Small Animal Imaging Facility (BioSpec 9.4/30 USR; Bruker Biospin). DTI was chosen because P14 mouse brains do not display robust T1 and T2 contrast between gray and white matter. Diffusion data was obtained using a multi-shot EPI sequence with the following parameters: TE = 30.896 ms, TR = 180 ms, BW = 400 kHz, FOV = 18 mm × 12 mm × 9 mm, matrix size = 240 × 160 × 120, Number of shots = 16, voxel resolution = 75 × 75 × 75 µm3, δ = 4 ms, Δ = 16 ms, Number of b0 images = 6, Shell = 3, b-value = 2000 (30 directions), 4000 (30 directions), 6000 (60 directions) s/mm2, partial Fourier = 1.11 (homodyne reconstruction for TE shortening), scan time = 12 h.

The diffusion data underwent preprocessing using the FSL eddy package to correct for eddy current distortions and subject motion artifacts [29]. Post-correction, Fractional Anisotropy (FA) was calculated using DSI Studio (RRID: SCR_009557) [30]. To spatially normalize the DTI images, each subject’s FA image was first registered to a reference subject’s FA map using rigid-body registration. A study-specific template was then generated using ANTs SyN non-linear registration. A brain mask was drawn on this template, and the inverse transform was applied to estimate subject-level brain masks. These masks were manually refined using ITK-SNAP and employed for skull stripping to improve spatial normalization. The skull-stripped FA images were then registered to the Developmental Common Coordinate Framework (DevCCF) at P14 [31] using ANTs SyN non-linear registration. To calculate brain volumes, the total number of voxels in each subject’s brain mask was multiplied by the volume of a single voxel. To calculate WM volumes, the DevCCF FA template was thresholded with FA > 0.3 [32, 33] and inverse transformed to each subject’s FA map.

To identify myelin basic protein (MBP), we used a rat monoclonal antibody (Abcam Cat#ab7349, RRID: AB_305869) raised against the full-length protein corresponding to cow MBP. This antibody binds to a region defined by amino acids 82–87 (DENPVV).

For UBE3A identification we used a mouse monoclonal antibody (Sigma-Aldrich, Cat#SAB1404508, RRID: AB_10740376). This antibody was developed against a peptide sequence common to all three isoforms of mouse UBE3A (amino acids 315–415 for isoforms 1 & 3, and 336–436 for isoform 2) and has been extensively validated [14, 34, 35].

Animals were anesthetized with isoflurane, followed by decapitation and extraction of forebrain tissue, excluding the olfactory bulb, cerebellum, and brainstem. The tissue was homogenized in a lysis buffer containing Tris-HCl (25mM, pH 7.4), 1% SDS, and 1mM EDTA with protease inhibitor. Protein concentrations were determined with the BCA assay, and protein (30 µg for P14 and 20 µg for P45) was loaded onto a 20% polyacrylamide-SDS gel. Proteins were then transferred onto a 0.2 μm nitrocellulose membrane. After blocking with Intercept Blocking Buffer (LiCor) for 1 h, the membranes were incubated overnight at 4 °C with primary antibodies: rat anti-MBP (1:1,000) and mouse anti-GAPDH (1:5,000; Sigma Cat#MAB374; RRID: AB_2107445). Following washing with PBS-Tween (PBS with 0.1% Tween™ 20), membranes were incubated in HRP-conjugated secondary antibodies, washed in PBS-Tween, and chemiluminescence imaging was performed. MBP protein band intensity was normalized to GAPDH band intensity. Protein expression in WT and Ube3a mutants was represented as a fraction of protein level in WT mice.

Mice were anesthetized with sodium pentobarbital (60 mg/kg i.p.) and subsequently transcardially perfused, starting with a rapid flush with PBS (0.1 M, pH 7.3), followed by 10 min of 4% freshly depolymerized paraformaldehyde in phosphate buffer (pH 7.3). Brains were then postfixed overnight at 4 °C in the same fixative solution, cryoprotected in 30% sucrose in PBS, and sectioned at 50 μm using a sliding microtome. Free-floating sections underwent an initial methanol permeabilization step (2 × 15 min in 50% methanol in PBS) followed by a second permeabilization step: 1 h at 37 °C in a solution comprising 2.3% glycine, 20% DMSO, and 0.2% Triton X-100 in PBS. Subsequently, sections were preincubated for 30 min in 5% DMSO/0.1% Triton X-100/1% BSA in PBS and incubated overnight at 37 °C with primary antibody (MBP, 1:2,000 in PBS with 5% DMSO, 0.1% Triton X-100, 1% BSA, 0.2% Tween-20, and 1% heparin). The primary antibody was visualized using a secondary antibody conjugated with Alexa Fluor dye. Sections were counterstained with DAPI to reveal nuclei, and scanning was performed using a Slideview VS200 slide scanner (Olympus, Hamburg, Germany). Analysis was conducted using the QuPath software package (RRID: SCR_018257).

Mice were anesthetized with sodium pentobarbital (60 mg/kg i.p.) and perfused with a solution containing 2% glutaraldehyde, 2% paraformaldehyde, and 0.2% picric acid in 0.1 M phosphate buffer (pH 6.8). Following perfusion, brains were promptly removed and postfixed overnight at 4 °C in the same fixative. Subsequently, the brains were sectioned to a thickness of 50 μm using a vibratome. The sections were processed for reduced osmium following the Knott protocol [34, 36, 37]. In brief, the sections were washed in cacodylate buffer (0.1 M, pH 7.4) and post-fixed for 40 min in a solution of 1.5% potassium ferrocyanide and 1% osmium tetroxide, followed by an hour in 1% osmium tetroxide alone. After rinsing in water, the sections were incubated for 40 min in 1% uranyl acetate in water, rinsed again in water, dehydrated in an ethanol series, and finally infiltrated and embedded in resin (Spurr’s low viscosity epoxy with ERL-4221, Electron Microscopy Sciences, Hatfield, PA; cat. No. 14300). The embedded sections were flat mounted between sheets of ACLAR® fluoropolymer (Electron Microscopy Sciences, Hatfield, PA; cat. No. 50425) within glass slides. Small chips of the corpus callosum (body region) were affixed to plastic blocks, sectioned en face at ~ 60 nm, collected on 300 mesh nickel grids, and coated and contrasted with uranyl acetate and Sato’s lead. Grids were imaged at a voltage of 120 kV using a Technai 12 D230 transmission electron microscope running SerialEM [38, 39]. We acquired large 50–60 μm by 50–60 μm montages at 6500x (close to 1 nm per pixel) for quantification. Axons and myelin were manually traced using FIJI [40, 41]. The relative myelin thickness around an axon (g-ratio) was calculated as the √(AxonArea/MyelinArea). Quantitative analyses included 1,500–2,000 axons per animal at P16 and 1,000–1,500 axons per animal at P30.

Differences in human MRI brain volumes (TBV, WM, GM) between the AS and NT groups were tested using a mixed effects model for repeated measures while covarying for the fixed effects of age, sex, scanner, and group x age interaction. Random effects included the individual subjects’ age at scan. All statistical analyses of MRI data were performed using SAS JMP software.

Early nutrition and maternal care are known to influence brain development, specifically growth and myelination. To account for these factors, we sampled AS mice and their WT littermates at various ages, ensuring litters were culled to a size of 5–7 mice (Table 1). Statistical analyses were performed using GraphPad Prism 9 (RRID: SCR_002798), including unpaired two-tailed t-tests to compare brain volumes for mouse MRI-DTI, MBP protein levels, and the percentage of myelinated axons.