A total of 28 patients with RTT aged 2–27 years were recruited in this study. All participants were enrolled and followed up at the Joint Clinic for RTT in National Taiwan University Hospital (NTUH), Taiwan. RTT was diagnosed based on the revised diagnostic criteria [38] by a senior pediatric neurologist (Lee W.T.). The control group included 32 sex-matched participants whose ages were paired with individuals with RTT within 6 months of range. Only one male subject was enrolled in each group (Table 1). All participants in the control group completed an interview by pediatric neurologists, which included detailed developmental history, birth history, family history, and neurological examinations. Participants in the control group displayed typical behavioral and neurological development. Individuals with any neurological diseases (e.g., psychiatric diagnoses, history of neurological impairment, neuropsychiatric conditions, and clinical evidence of a genetic disorder) were excluded. Remaining subjects were suitable for participation in the MRI study. To avoid motion-related artifacts, individuals with RTT and younger neurotypical individuals (aged ≤ 4 years) were routinely sedated using chloral hydrate before performing MRI. Older neurotypical individuals were also sedated if there were motion artifacts after repeating scanning. The 28 patients with RTT were classified into a younger group (aged < 10 years) and an older group (aged > 10 years) (Table 1), and the differences between the two age groups were compared in this study.

Parents and caregivers were requested to complete developmental and clinical questionnaires. The motor function of all participants with RTT was evaluated using the Peabody Developmental Motor Scales-second edition (PDMS-2). All evaluations were performed by the same occupational therapist to ensure quality and consistency. This study was approved by the institutional review board of NTUH in Taipei, Taiwan (201510011RINC). The request to complete the questionnaires and participate in the imaging studies was approved by the internal review board.

All scans were acquired using a Siemens Tim Trio 3 T scanner with a 32-channel head coil at NTUH. Because potential motion artifacts and general reconstruction issues may occur in the MRI scans of nonsedated individuals with RTT with severe disability, all patients with RTT or younger neurotypical individuals were sedated during MRI in this study, and the heart rate and blood O2 saturation were continuously monitored using pulse oximetry.

The MRI protocol consisted of sagittal T1-weighted and axial fast spin-echo T2-weighted imaging. High-resolution T1-weighted imaging was performed using 3D magnetization-prepared rapid gradient-echo (MPRAGE), which was further used for anatomical reference and calculations of gray matter volume. Transverse sections were obtained parallel to the anterior and posterior commissure line with a thickness of 1.0 mm3. The total number of slices per slab was 208 with repetition time (TR) = 2000 ms, echo time (TE) = 2.98 ms, flip angle = 9°, and field of view (FOV) read = 256 × 192 mm2. The fast spin-echo sequence was acquired from 56 contiguous axial T2-weighted images with TR = 9650 ms, TE = 103 ms, FOV = 200 × 200 mm2, and slice thickness = 2.5 mm.

Imaging processing and analysis were completed in the following steps. All scans passed a successful quality control test for gross structural abnormalities before data processing. The actual coordinates were adjusted using displayed methods on the Statistical Parametric Mapping 12 software package (http://www.fil.ion.ucl.ac.uk/spm/software/spm12/) that was run through in MATLAB 2016 (Mathworks, USA) to avoid artifacts, poor directionality, and insufficient image quality.

For calculating the gray matter volume, the FreeSurfer software package (v.6.0.0–64 bit; Martinos Center for Biomedical Imaging, Boston, MA, USA) was used, which created a three-dimensional model of the cortical surface and cortical thickness along with area measurements. The gray matter volume in this study indicated the cortical gray volume and did not include the subcortical gray matter and cerebellum. Segmentation of the gray and white matter volume was performed using the FreeSurfer “recon-all” pipeline (http://surfer.nmr.mgh.harvard.edu) based on intensities and continuity of information from the full-head image to establish representations of the boundary of the gray/white matter (surface) and the pial surface. For reconstruction and estimation, motion correction for better segmentation in the original imaging was performed before the procedure to ensure the accuracy of T1-weighted images. Although the FreeSurfer software includes motion correction, the removal of nonbrain tissue, and automated intensity normalization for surface- and intensity-based segmentation of the cortex, FreeSurfer segmentations would be visually checked and manually corrected or excluded for all individuals by a neuroradiologist with experience in segmentation after completing the scans. If there was incorrect segmentation, the neuroradiologist would perform manual correction before further processing. To reduce motion artifacts, all participants would be comforted for at least 30 min before scanning. For participants with RTT or younger neurotypical individuals, we would continue to comfort them after sedation and lying on the MRI machine. After scanning, for those with motion artifacts, which resulted in incorrect segmentations, we would repeat all scans on the same day or another day to obtain better images. For some individuals, the scans had to be repeated at least two to three times to acquire better images for analysis. For almost all healthy control individuals, the scans can be completed at the first time because no motion artifacts were observed and there was good brain volume segmentation. Although all participants with RTT were sedated during MRI processing, there was also no significant difference in brain volume segmentation for sedative (individuals with RTT) and nonsedative individuals (healthy controls) in our analysis after scanning, despite the failure rate being much higher in individuals with RTT due to motion artifacts.

The brain tissue was automatically transformed from the topological surface to the Talairach space. Volumetric structures were segmented for the tessellation of the gray matter–white matter boundary [14, 17]. Subsequently, the intensity gradients, automated topology, and deformed surface corrections were applied to optimize the locations of the gray/white and gray/cerebrospinal fluid borders for the greatest shift in intensity [15, 16]. The automatic labeling method provided similar results as those of manual labeling in previous studies and demonstrated low reproducibility errors and high precision [15, 25].

The intracranial volume and gray matter and white matter volumes were acquired from the abovementioned procedure. Moreover, considering the feasibilities, consistency, and accuracy that were required for investigating the gray matter volume, Desikan–Killiany–Tourville (DKT) cortical labeling atlases (rh.aparc.DKTatlas.stats/ lh.aparc.DKTatlas.stats) were selected for volumetric data analysis and manual correction [12] to investigate local changes in the gray matter. Of 31 cortical regions in DKT atlases, 24 were selected and subdivided into four lobar groups, and the differences between the two age groups were compared. The volumes of the frontal lobe included those of the superior frontal gyrus, middle frontal gyrus (rostral part and caudal part), inferior frontal gyrus (pars orbitalis, pars opercularis, and pars triangularis), precentral gyrus, and orbitofrontal gyrus (medial and lateral division). The parietal lobe included the superior parietal lobule, inferior parietal lobule, supramarginal gyrus, postcentral gyrus, paracentral gyrus, and precuneus. The temporal lobe included the superior temporal gyrus, middle temporal gyrus, inferior temporal gyrus, transverse temporal gyrus, and fusiform gyrus. The occipital lobes encompassed the lingual gyrus, pericalcarine cortex, cuneus cortex, and lateral occipital cortex [28].

PDMS-2 is a reliable and valid tool used to evaluate motor functions [10], it was administered to evaluate the gross and fine motor skills objectively [18, 32]. PDMS-2 is a motor development program comprising the following six subtests: reflexes, stationary, locomotion, object manipulation, grasping, and visual–motor integration. The items of reflexes were excluded because reflexes are inhibited at the cortical level in normal development. For each subtest task, performance criteria were specified and scored on a 3-point scale, from 0 to 2, where 0 indicates bad and 2 implies mastered. For administration, a basal level was determined when a score of 2 was obtained on three consecutive items, and a ceiling level was determined when a score of 0 was obtained on three consecutive items.

The Rett Syndrome Behavior Questionnaire (RSBQ) evaluates typical pathological behaviors and clinical symptoms in individuals with RTT [5, 26, 36]. A total of 45 items were categorized into the following eight domains/subscales: (1) general mood, (2) breathing problems, (3) body rocking and expressionless face, (4) hand behaviors, (5) repetitive face movements, (6) nighttime behaviors, (7) fear/anxiety, and (8) walking/standing. Each item was scored on a Likert scale of 0–2 based on how well the description fitted the patient’s behavior.

The Rett Syndrome Severity Scale (RSSS) includes the following seven domains: (1) frequency and manageability of seizures, (2) respiratory irregularities, (3) scoliosis, (4) ability to walk, (5) hand use, (6) speech, and (7) sleep. Each item was scored on a Likert scale of 0–3 based on the probability that the description portrays the severity of symptom. Total scores were then categorized into three degrees, viz., mild (0–7), moderate (8–14), and severe (15–21) [27, 34].

All data were expressed as number or mean ± standard deviation. The difference in the scores of RSBQ and RSSS and the performance differences in the five subtests of PDMS-2 among younger and older individuals with RTT were compared using independent t-tests.

Imaging data were analyzed for the total intracranial volume (TIV), total cortical gray matter volume, and cerebral white matter volume. An independent t-test was used to analyze the differences in intracranial volume, cerebral white matter volume, cortical gray matter volume, and cortical gray matter volume in four cerebral lobes between the RTT group and control group. To further analyze the lobe differences between the RTT group and control group, the one-way analysis of covariance (one-way ANCOVA) was used and adjusted by the covariate (TIV). Moreover, to determine the association between brain volume and age changes, the volume changes were incorporated into a linear regression model with age and age squared for developmental trajectories as follows:

$$\text=\text0+\text1\left(\text\right)+\text2\left(\text\right)+\text3\left(\text\right)+\text4\left(}^\text\right)+\text5\left(}^\text\right).$$

In the analysis, we investigated the effect of age and group and also the interaction between group and age. Finally, the general linear model was applied to investigate group differences in age-related structural changes. The linear model represented the relationships among cortical gray matter volume in four cerebral lobes and the following input variables: age, age squared, group, age-by-group interaction, and age-squared-by-group interaction. These variables were corrected using Benjamini–Hochberg false discovery rate (FDR) for multiple comparisons in q-value. These relationships were compared with the average values and slopes of both groups. A Levene’s test was used to confirm the homogeneity of the variance of both RTT and control groups before implementing the one-way ANCOVA and general linear model.

The Levene’s test revealed a greater p value than the critical value (0.05), which supported the hypothesis of the Levene’s test that the variances were equal. The Bonferroni correction was also applied to minimize their effects on the study variables.

All statistical analyses were conducted using IBM SPSS Statistics 20 (SPSS Inc., Chicago, Illinois, USA). A p value of < 0.05 was considered statistically significant.

The families of patients with RTT in Taiwan and the Rett Syndrome Association in Taiwan were actively involved and supported throughout this study process, and they participated in interviews, surveys, and other data collection. We anticipate that they will be aware of the effect of MECP2 variants on brain development, especially the long-term effect on brain development.