Participants
Twenty-seven children with PWS and 28 typically developing children were included. Inclusion criteria for PWS group were (1) genetically confirmed PWS, (2) age 6–23 years, and (3) no neurological or psychiatric history. Inclusion criteria for the healthy control group were (1) age 6–23 years, and (2) no neurological or psychiatric history. All children with PWS participated in the Dutch PWS Cohort Study [21]. The study was approved by the Medical Ethics Committee of the Erasmus Medical Center Rotterdam, Rotterdam, The Netherlands. Written informed consent was obtained in all cases from the caregivers and children older than 12 years and informed assent in children younger than 12 years.
MRI acquisition
Children were introduced to a mock scanner before they underwent the MRI scan. Caregivers were given the option to stay in the room with the MRI scanner, close to their child.
All images were acquired on a 3T GE 750 Discovery MRI scanner (General Electric, Milwaukee, WI, USA), using a dedicated 8-channel head coil. Following 3-plane localizing and coil intensity calibration scans, a high-resolution T1-weighted inversion recovery fast spoiled gradient recalled (IR FSPGR) sequence was obtained with the following parameters: TR = 10.3 ms, TE = 4.2 ms, TI = 350 ms, NEX = 1, flip angle = 16°, readout bandwidth = 20.8 kHz, matrix 256 × 256, imaging acceleration factor of 2, and an isotropic resolution of 0.9 × 0.9 × 0.9 mm3 (duration; 5 min 40 s). Resting-state fMRI (rs-fMRI) utilized a gradient-echo blood oxygen level dependent (BOLD) EPI sequence with a TR = 2000 ms, TE = 30 ms, flip angle = 85°, matrix 64 × 64, and voxel resolution of 3.6 ×3.6 × 4.0 mm3. The duration of the rs-fMRI was 160 TRs (5 min 20 s). The children were asked to keep their eyes closed during the rs-fMRI sequence and to think about nothing in particular. All MRI images were reviewed by a qualified radiologist and no gross brain abnormalities were identified.
Twenty-two children with PWS and 27 healthy controls had both structural MRI and resting state fMRI data available, suitable for manual segmentation functional connectivity analysis, respectively.
Data processing
Manual segmentation
All MRI images were coded to ensure investigators’ blindness to subject identification and diagnosis. Manual segmentation of the hypothalamus, mammillary bodies, and pituitary gland were performed with the FreeSurfer 5.3 image analysis suite. Tracing was performed slice by slice (1-mm thick) in coronal, axial, and sagittal orientations. Thus, the anatomic boundaries were viewed from different orientations. The hypothalamus segmentation boundaries used were based on the boundaries used by Klomp et al. [21] and Lemaire et al. [22]. The anterior boundary of the hypothalamus was defined by the first coronal slice after the anterior commissure (AC) and the posterior boundary by the last slice where the mammillary bodies were visible. The dorsal boundary consisted of the AC-PC (posterior commissure) plane in the transverse section or the hypothalamic sulcus and ventral border of the lamina terminalis. The lateral boundaries were determined by the white matter bundles. The boundaries of the mammillary bodies were defined anterior by the first coronal slice in which the mammillary bodies were visible as part of the hypothalamus and posterior by the very last slice in which they were still visible. The floor of the third ventricle and the cerebral spinal fluid (CSF)-filled suprasellar cistern defined the dorsal and ventral boundaries, respectively. Laterally the boundaries consisted of the darker gray of the hypothalamus or the bundle of white matter that surrounds it.
We measured pituitary volume instead of pituitary height, as the latter measure only weakly correlates with pituitary volume [23]. The boundaries of the pituitary gland were defined anteriorly and ventrally by the sphenoid sinus, posteriorly by the dorsum sellae, dorsally by the diaphragma sellae, and laterally by the cavernous sinuses. The infundibular stalk was excluded from the segmentations. For each subject, it was assessed whether the posterior pituitary bright spot (PPBS) was present. The length of the pituitary stalk was defined and measured as the distance from the tip of the infundibular recesses to the junction of the pituitary stalk and the pituitary gland, along the course of the pituitary stalk.
Resting-state functional connectivity
All rs-fMRI data were preprocessed by the FMRIB’s Diffusion Toolbox (FDT) from the FMRIB’s freely available Software Library (FSL, http://www.fmrib.ox.ac.uk/fsl). First, the first four volumes of rs-fMRI scan of each participant were discarded to allow for T1-equilibration effects. Each subject’s functional images were motion-corrected with MCFLIRT, spatially smoothed with an 8-mm spatial filter (White et al. 2001) and denoised using FSL’s ICA-based X-noiseifier (FIX) tool [24], regressing out six motion parameters. Individual hypothalamus masks, obtained from the manual segmentations, were registered into functional space using FLIRT, and the individual fMRI time series within the hypothalamus ROI were extracted. Individual whole-brain voxel-wise regression analyses were performed with FSL’s FEAT using the time series within the hypothalamus as the independent variable, controlling for the white matter and CSF time series. The obtained z-transformed individual images were then registered to the high-resolution structural image, and both structural and functional images were registered into standard space using the Montreal Neurological Institute T1 template (MNI52) using FSL’s FLIRT. We then generated group mean connectivity maps as well as tested for group differences (PWS vs. control, DEL vs. control, mUPD vs. control and DEL vs. mUPD), using FSL’s randomize permutation-testing tool with 5000 permutations, excluding white matter and CSF voxels, and threshold-free cluster enhancement (TFCE), accounting for age and gender effects and setting p value at <0.05. The minimal number of voxels for a cluster to be considered was set at 50. Family-wise error (FWE) was used to correct for multiple comparisons. Fisher’s z-transformed partial correlation coefficients were estimated.
Hormones and assays
Blood samples for the assessment of IGF-1, thyroid stimulating hormone (TSH), T4, free T4 (fT4), T3, and reverse T3 (rT3) were collected prior to start with GH treatment, in the morning, after overnight fasting. After centrifugation, the samples were immediately frozen at –20 °C until assayed.
Serum IGF-I levels were measured in one central laboratory using a immunometric technique on an Advantage Automatic Chemiluminescence System (Nichols Institute Diagnostics, San Juan Capistrano, CA). The intraassay coefficient of variation was 4%, and the interassay coefficient of variation was 6%. Because of age and sex dependency, IGF-I levels were transformed into standard deviation scores (SDS) [25].
Plasma levels of rT3 were measured by established radioimmunoassay procedures, as described previously [10]. Serum TSH, free T4, T4, and T3 levels were measured by Vitros Eci technology (Ortho-Clinical Diagnostics, Amersham, UK) [10]. The within-assay coefficient of variation (CV) was 3–7% and the between-assay CV 5–10%. Thyroid hormone SDSs were calculated with data from a control group comprising 500 healthy Dutch children (aged 1.4–18 years) from the Rotterdam region, The Netherlands, participating in a study to assess reference values.
The fasting blood samples were taken for the analysis of maximal adrenocorticotropic hormone (ACTH) levels after overnight single-dose metyrapone test, after start of treatment with GH (for detailed description, see [8]). Plasma ACTH levels were measured with an immunoradiometric assay (Bio-International, Gif-sur-Yvette, France) with a minimal detection level of 1.1 pmol/liter. Glucose levels were measured with the Hitachi 917 (Hitachi Device Development Center, Tokyo, Japan), detecting glucose levels between 0 and 42 mmol/liter.
Statistical analyses of volumetric data
Intrarater reliability of the volume measurements, assessed by intraclass correlation coefficient (ICC) in the five brains was as follows: the hypothalamus (ICC = 0.97), mammillary bodies (ICC = 0.73), and pituitary gland (ICC = 0.99).
Interrater reliability of the volume measurements, assessed by intraclass correlation coefficient (ICC), for five randomly selected patients by two raters (AL and SD) who were both blinded to subject identification and diagnosis was as follows: the hypothalamus (ICC = 0.93), mammillary bodies (ICC = 0.76), and pituitary gland (ICC = 0.99).
Volumes of the hypothalamus, mammillary bodies, and pituitary gland and pituitary stalk length were exported to SPSS (version 21, IBM Corporation, Armonk, NY, USA) for statistical analyses. Groups were compared using a nonparametric Mann-Whitney or Kruskal-Wallis analysis of variance test and a chi-square, where appropriate. For significant group effects only, post hoc Mann-Whitney tests were performed for pairwise comparisons with a Bonferroni correction of p = .05/3 (DEL, mUPD, and control). For interaction effects between pituitary volume and age, nonparametric Quade’s rank analysis of covariance was performed.