Syndrome specific neuroanatomical phenotypes in girls with Turner and Noonan Syndromes

Abstract

Background:

Turner syndrome (TS) and Noonan syndrome (NS) are distinct genetic conditions with highly similar physical and neurodevelopmental phenotypes. TS is caused by X-chromosome absence, whereas NS results from genetic mutations activating the Ras mitogen-activated protein kinase (RAS-MAPK) signaling pathway. Previous neuroimaging studies in TS and NS have shown neuroanatomical variations relative to typically developing (TD) individuals, a standard comparison group when initially examining a clinical group of interest. However, none of these studies included a second clinical comparison group, limiting their ability to identify syndrome-specific neuroanatomical phenotypes.

Methods:

In this study, we compared the behavioral and brain phenotypes of 37 girls with TS, 26 girls with NS, and 37 TD girls, all 5–12 years of age, using univariate and multivariate data-driven analyses.

Results:

We found divergent neuroanatomical phenotypes between groups, despite high behavioral similarities. TS was associated with smaller whole-brain cortical surface area (SA) (p=<0.0001) whereas NS was associated with smaller whole-brain cortical thickness (CT) (p=.013) relative to TD. TS was associated with larger subcortical volumes (left amygdala, p=0.002; right hippocampus, p=0.002) whereas NS was associated with smaller subcortical volumes (bilateral caudate, p≤0.003; putamen, p<0.001; pallidum, p<0.001; right hippocampus, p=0.015). Multivariate analyses also showed diverging brain phenotypes in terms of SA and CT, with SA outperforming CT at group separation.

Conclusion:

TS and NS have syndrome-specific brain phenotypes, despite their behavioral similarities. Our observations suggest that neuroanatomical phenotypes better reflect the different genetics etiologies of TS and NS and may be superior biomarkers relative to behavioral phenotypes.

INTRODUCTION

Turner syndrome (TS) and Noonan syndrome (NS) are rare genetic conditions affecting several organ systems and are associated with high neuropsychiatric morbidity (Cardoso et al., 2004; Perrino et al., 2018). TS arises from a partial or complete congenital loss of the second X-chromosome occurring in 1:2000–2500 live female births (Gravholt, 2005). NS is an autosomal dominant disorder caused by germline mutations encoding components of the Ras mitogen-activated protein kinase (RAS-MAPK) pathway (Tartaglia and Gelb, 2005; Rauen, 2013; Roberts et al., 2013) occurring in 1:1000–2500 live births (Roberts et al., 2013). However, although etiologically distinct, TS and NS have highly similar physical and behavioral phenotypes (Green et al., 2017) to the extent that NS is alternatively known as male or pseudo-TS. Clinically, both TS and NS present with short stature, craniofacial dysmorphic features, hearing loss, congenital heart anomalies, ophthalmologic and gastrointestinal issues, and delayed/absent pubertal development (Roberts et al., 2013) — factors that can influence brain development. Behaviorally, both disorders are associated with impairments in executive function (Hong et al., 2009; Lepage et al., 2011; Pierpont et al., 2015) and social cognition (Hong et al., 2011; Lepage et al., 2013a; Alfieri et al., 2014). Compared to typically developing (TD) individuals, both have higher prevalence rates for attention deficit and hyperactivity disorder (ADHD) (Green et al., 2015; Pierpont et al., 2015; Perrino et al., 2018), autism spectrum disorder (ASD) (Marco and Skuse, 2006; Geoffray et al., 2021), and mood and anxiety disorders (Cardoso et al., 2004; Alfieri et al., 2021). Behavioral similarities between TS and NS are likely a result of the high polygenicity and substantial genetic overlap (Pettersson et al., 2016; Brainstorm Consortium et al., 2018) of behavioral phenotypes. Contrasting TS and NS at the neuroanatomical level offers a unique opportunity to test whether the high similarity observed between the behavioral phenotypes of these two conditions extends to their neuroanatomical phenotypes (Green et al., 2015; Mallard et al., 2021).

Independent lines of structural magnetic resonance imaging (sMRI) research on TS and NS suggest that in contrast to behavioral phenotypes, the two present syndrome-specific neuroanatomical phenotypes. In subcortical structures, females with TS show enlarged amygdala (Murphy et al., 1993; Kesler et al., 2004; Green et al., 2016) and striatal volumes (Molko et al., 2004; Lepage et al., 2013b, 2013c). Conversely, in individuals with NS the amygdala appears unaffected and striatal volume is smaller (Johnson et al., 2019). Both conditions are also associated with alterations in cortical morphology. Individuals with TS show smaller cortical surface area (SA) in parieto-occipital regions and larger cortical thickness (CT) in parieto-temporal regions (Raznahan et al., 2010; Lepage et al., 2013b, 2013c; Green et al., 2014); preliminary data for NS show a combination of larger and smaller SA and CT spanning regions across all lobes (Johnson et al., 2019). However, prior studies have varied widely on participant inclusion criteria (age, sex, pubertal status, and mosaicism). Moreover, they have seldomly included a clinical group. By only comparing to TD, several physical features (e.g. cardiac issues) can act as confounding factors given their impact on brain and behavior. Thus, limiting inferences of syndromic neuroanatomical specificity and genetic downstream effects on neuroanatomy. By contrasting TS and NS we can overcome several of these confounders and robustly test whether brain phenotypes are syndrome-specific, rather than a general consequence of adverse genetic effects.

In this study, we contrast TS and NS on (1) behavioral phenotypes (cognitive, emotion, and socio-communicative) and (2) neuroanatomical phenotypes (SA, CT, and subcortical volumes) using univariate and multivariate analyses. Due to differences in their genetic etiologies (Panizzon et al., 2009), organizing principles (Chen et al., 2013), and polygenicity degree (Matoba et al., 2020) we conducted separate analyses of SA and CT. We acquired cognitive, behavioral, and high-resolution sMRI data in a consistent manner from girls with TS, NS, and TD. The groups matched for age and pubertal status. We hypothesized convergence between TS and NS at the behavioral level but divergence at the neuroanatomical level.

MATERIALS AND METHODS

Participants

In this study we included girls aged 5-12 years with TS (n=35), with NS (n=26), and TD (n=37). Medications and pubertal status of participants were recorded and can be found in Table 1. Subjects with TS and typically developing (TD) were recruited over a five-year period (2013-2017) from a larger study examining longitudinal brain changes associated with TS. Individuals with TS were recruited through the national Turner Syndrome Society, the Turner Syndrome Foundation, a local network of physicians, and advertisements on the Stanford University School of Medicine website. The TD group was recruited through local newspapers and parent networks. Subjects with NS were recruited over a three-year period (2016-2019) through the Noonan Syndrome Foundation, NS social media groups, and physician referrals from relevant specialties nationwide. TS and TD participants were selected to group-match NS on age and prepubertal status. Exclusion criteria: participants with any history of neurological disorders known to affect cognitive development or brain structure (e.g. seizures, or a diagnosis of gross structural malformations), known diagnoses of major psychiatric disorders, any MRI contraindications, premature birth (<34 weeks), birth weight <2000g, mosaic or uncommon structural karyotypes (TS). Only individuals with PTPN11 or SOS1 mutation (supported by prior genetic testing) were included (NS). The study was conducted under the ethical and safety guidelines set forth by the Stanford University School of Medicine Institutional Review Board, which approved the study. Written informed consent was obtained from all parents or legal guardians for their child’s participation in the study. A complementary written assent was obtained for participants age seven or older.

Table 1.

Participant Demographic Information

	TD (n= 37)	TS (n= 35)	NS (n= 26)	TS-NS	TS-TD	NS-TD
Age (range)	9.2 (7.0 , 11.7)	9.2 (6.3 , 11.8)	8.4 (5.1 , 11.9)	x2(2)=3.0507, p=0.2175
Genetic mutation	-	Monosomic (n= 35)	PTPN11 (n= 20) SOS1 (n= 6)	-	-	-
Tanner Stage	≤2	≤2 (n=34)	≤2	-	-	-
Growth Hormone	0	32	1	-	-	-
Stimulants	0	5	2	-	-	-
SSRIs	0	3	2	-	-	-
other psychoactivea	0	4	2	-	-	-
FSIQ	110.6 (87 , 144)	91.5 (54 , 116)	92.4 (63 , 122)	t(50.8)=−0.214, p=0.832	t(67.2) = 6.11, p < 0.0001	t(45.2) = 4.98, p <0.0001
VIQ	111.9 (79 , 144)	103.9 (65 , 150)	96.7 (71 , 121)	t(58.9) = 2.05, p = 0.056	t(69.6) = 2.24, p = 0.056	t(60.1) = 4.47, p < 0.0001

^a.Clonidine (n=4), Guanfacine (n=1), Remeron (n=1)

Data acquisition

Imaging data of subjects in all groups (TS, NS, and TD) were collected from the Stanford University Lucas Center for Imaging, using a GE Healthcare Discovery 3.0 T whole-body MR system (GE Medical Systems, Milwaukee, WI) and a standard 8-channel head coil. Further details on participant MRI preparation, the pulse sequence and image quality control can be found in the Supplementary.

Morphometric analysis

We used bias field correction methods available with SPM8 (http://www.fil.ion.ucl.ac.uk/spm) for scans. Next, we used the FreeSurfer image analysis suite, version 5.3 (http://surfer.nmr.mgh.harvard.edu/) for cortical reconstruction and volumetric segmentation of cortical and subcortical structures (details on the FreeSurfer pipeline can be found in the Supplementary). The cerebral cortex was parcellated into 68 regions of interest (ROIs) based on gyral and sulcal structure following the Desikan-Killiany atlas (Fischl et al., 2004; Desikan et al., 2006). We calculated volumetric data for subcortical structures, mean SA and CT for each ROI.

Behavioral assessments

We assessed global cognitive ability using the Full-Scale IQ (FSIQ) as well as verbal intelligence quotients (VIQ) from either Wechsler Preschool and Primary Scale of Intelligence III (WPPSI-III; Wechsler, 2002) or Wechsler Intelligence Scale for Children IV (WISC-IV; Wechsler, 2003) according to participant age. We used the Behavior Assessment System for Children, 2^nd Edition Parent Rating Scale (BASC-2 PRS; Reynolds and Kamphaus, 2004) to measure psychopathology symptoms in aggression, anxiety, attention problems, atypicality, conduct problems, depression, hyperactivity, somatization, and withdrawal. We assessed social skills with the Social Responsiveness Scale (SRS; Constantino et al., 2003). A valid 65-item parental rating scale assessing a continuum of social impairment severity, including subthreshold variation levels of autistic-like traits. Higher scores in the BASC-2 and the SRS indicate higher symptom severity.

Statistical analysis

Mean group differences.

We assessed between-group differences (TS, NS, TD) in age using the Kruskal-Wallis test (due to non-homogeneous variances) and FSIQ/VIQ using independent samples two-tailed t-tests. Normality plots from continuous variables showed variables were normally distributed. To control for the effects of age and brain volume differences, we regressed both age (mean-centered) and total brain volume (TBV) on volume, SA, and CT prior to statistical analyses. We compared mean general subcortical and cortical anatomical differences (subcortical volume, total SA, and total CT) across the groups using ANOVA and subsequent pairwise comparisons using Tukey HSD. For individual ROIs, we compared between-group means of volume (subcortical structures), SA, and CT using ANOVA. The dependent variable was either volume (subcortical ROIs), CT, or SA and the between-group factor was diagnosis. We controlled for multiple comparisons using the false discovery rate (FDR; Benjamini and Hochberg, 1995). Significant mean differences post-FDR correction were followed up by planned comparisons (TS-TD, NS-TD, TS-NS) with independent two-tailed t-tests. All statistical analyses were performed in R-Studio (4.0.5).

Principal Component Analysis (PCA).

We used PCA to test for differential patterns of neuroanatomical phenotypes across the groups in SA and CT. PCA is a data-driven unsupervised feature reduction method that decreases the dimensionality of the data while incurring minimal information loss. It does so by extracting the primary sources of variation in the form of new variables or principal components. We selected the number of principal components using the scree plot “elbow” method (Abdi and Williams, 2010). Additionally, to quantitatively test which of the two cortical brain measures (SA, CT) presented tighter and better-separated groups on the PCAs, we used a K-nearest neighbors (KNN) classification algorithm with K = 3, combined with k-fold cross-validation using k = 5 (details on the KNN analyses can be found in the Supplementary). Brain visualizations of principal components using “Simple Brain Plot” (Scholtens et al., 2021), and the KNN classification were done in MATLAB (The Math Works, Inc., 2021a).

RESULTS

Demographic, medical, and cognitive characteristics

Table 1 shows demographic, medical, and cognitive data for all groups. All subjects in the three groups were female. There were no significant age differences between the groups, nor between clinical groups in FSIQ and VIQ (see Table 1).

Similar behavioral profiles for TS and NS

We found similar profiles for TS and NS relative to TD on behavioral measures (see Figs. 1-2). We did not find significant differences between TS and NS in subscales measuring attention (t(50.1)=0.283, p=0.779), hyperactivity (t(46.3)=0.763, p=0.449), anxiety (t(50.8)=1.21, p=0.466), and atypicality (t(45)=0.670, p=0.507), but individuals with NS reported higher mean scores for aggression (t(38.2)=2.99, p=0.015), depression (t(35.8)=2.43, p=0.04), somatization (t(44.4)=4.9, p<0.0001) and withdrawal (t(42.8)=2.67, p=0.022) relative to TS. We also did not find significant differences in subscales assessing social awareness (t(48)=0.607, p=0.547), social cognition (t(42.9)=0.891, p=0.378), social motivation (t(38.3)=0.968, p=0.339), social communication (t(43.1)=0.673, p=0.504) and restricted repetitive behaviors (t(48.3)=0.210, p=0.834). Scores for both TS and NS ranged between the mild and severe ranges and were significantly higher than those in the TD group (Supplementary Tables S1 and S2).

Fig. 1.

Behavioral phenotypes: T-scores from the Behavior Assessment System for Children, 2nd Edition Parent Rating Scale (BASC-2 PRS) subscales are similar between girls with Turner syndrome (TS) and Noonan syndrome (NS). X axis: T-scores; y-axis: BASC-2 PRS subscales. TS is displayed in orange. NS is displayed in gold. Typically developing (TD) are displayed in blue. Scores ranging from 60-69 are considered to be “At Risk”; Scores above 70 are considered to be of clinical significance.

Fig 2.

Behavioral phenotypes: T-scores from the Social Responsiveness Scale (SRS) subscales are similar between girls with Turner syndrome (TS) and Noonan syndrome (NS). X axis: T-scores; y-axis: SRS subscales. TS is displayed in orange. NS is displayed in gold. Typically developing (TD) are displayed in blue. Scores above 76 are considered severe and of clinical significance.

Diverging neuroanatomical profiles in TS and NS

Whole brain SA is smaller in TS, whereas CT and subcortical regional volumes are smaller in NS.

We found that SA was smaller in TS relative to TD (p < 0.0001) and NS (p = 0.002) but did not find significant differences in SA between NS and TD (p = 0.216). In contrast, we did not find significant differences in CT between individuals with TS and TD (p = 0.324) but found that CT was smaller in NS relative to both TD (p = 0.013) and TS (p < 0.0001). We did not find significant differences in total subcortical volume between individuals with TS and TD (p = 0.122) but observed smaller total subcortical volume in NS relative to both TD (p = 0.0003) and TS (p < 0.0001). We did, however, find volume differences in individual subcortical structures between clinical groups (TS, NS) and TD (see next section for details).

TS and NS affect subcortical structures differently.

We found individuals with TS presented larger volumes in the bilateral hippocampi and left amygdala relative to TD (Fig. 3) and to the NS group (Table S1). In contrast, we found individuals with NS presented smaller volumes in the right hippocampus as well as in bilateral caudate, putamen, and pallidum relative to the TD (Fig. 3) and TS groups (Table S3).

Fig 3.

A. Differences in regional SA. P values (FDR-corrected) for independent samples t-tests planned comparisons of Turner syndrome (TS) and Noonan syndrome (NS) relative to typically developing (TD) controls of surface area (SA) from parcellated brain regions according to the Desikan-Killiany atlas (Desikan et al., 2006). Relative to TD, larger mean SA ROIs are shaded in orange and smaller mean SA ROIs are shaded in blue. B. TS and NS affect subcortical structures in different ways. Left amygdala volume is larger in TS, whereas bilateral striatum structures volumes are smaller in NS. TS and NS affect the hippocampi in divergent ways. Bilateral hippocampi volumes are larger in TS, whereas right hippocampus volume is smaller in NS. We used a log-based color scale to depict significant between-group differences (FDR corrected p-values) of subcortical structures volumes. The color scale was generated from the negative log of the p value (e.g., the negative log (p) of 0.05 = 1.3). Smaller volumes relative to typically developing (TD) individuals are represented in cold colors with deeper shades of blue indicating smaller p-values. Larger volumes relative to TD are represented in warm colors with brighter shades of red indicating smaller p-values.

Regional SA is smaller in girls with TS but larger in girls with NS.

We found differences in SA across several cortical regions between TS and NS subjects relative to TD (Fig. 4) and to each other (Table S4). Overall, consistent with total SA mean differences, individuals with TS presented smaller SA relative to TD, particularly in parietal (bilateral postcentral gyrus, precuneus gyrus, and superior parietal lobule), and occipital (bilateral lingual gyrus, cuneus gyrus, pericalcarine sulcus, and right lateral occipital gyrus) regions, and a few frontal (right caudal middle frontal cortex and pars opercularis) and temporal (right middle temporal cortex and left fusiform gyrus) regions. SA was larger in TS relative to TD in the temporal lobe (right bank of the superior temporal sulcus, right parahippocampal gyrus, left superior temporal gyrus, and bilateral insula). We also found differences in SA for a few ROIs in the NS group relative to TD, but, in contrast to the TS group, the directions of the differences were more variable. In the NS group we observed differences in SA in frontal (larger SA: right superior frontal gyrus; smaller SA: left caudal middle frontal cortex), temporal (larger SA: right bank of the superior temporal sulcus, right parahippocampal gyrus, bilateral superior temporal gyrus; smaller SA: right entorhinal cortex), parietal (larger SA of bilateral supramarginal gyrus and left inferior parietal lobule) and occipital (smaller SA of the left cuneus gyrus) regions compared to TD.

Fig 4.

A. Differences in regional SA. P values (FDR-corrected) for independent samples t-tests planned comparisons of Turner syndrome (TS) and Noonan syndrome (NS) relative to typically developing (TD) controls of surface area (SA) from parcellated brain regions according to the Desikan-Killiany atlas (Desikan et al., 2006). Relative to TD, larger mean SA ROIs are shaded in orange and smaller mean SA ROIs are shaded in blue. B. Regional SAs were, at large, smaller in individuals with TS and larger in individuals with NS. We used a log-based color scale to depict significant between-group differences (FDR corrected p-values) in SA. The color scale was generated from the negative log of the p value (e.g., the negative log (p) of 0.05 = 1.3). Smaller regional SAs relative to typically developing (TD) individuals are represented in cold colors with deeper shades of blue indicating smaller p-values. Larger regional SAs relative to TD are represented in warm colors with brighter shades of red indicating smaller p-values.

TS and NS show different profiles of alterations in regional CT relative to TD.

Significant mean differences of CT were observed between clinical (TS and NS) and TD groups across several ROIs (Fig. 5) and between the two clinical groups (Table S5). Overall, we observed a profile of larger CT across regions in the TS group relative to TD, particularly in the frontal (left pars opercularis and left lateral orbitofrontal gyrus; but smaller in the bilateral isthmus cingulate) and temporal (bilateral banks of the superior temporal sulcus, superior temporal gyrus, left inferior temporal gyrus, and right middle temporal cortex; but smaller in bilateral parahippocampal gyrus, right entorhinal cortex, and pericalcarine gyrus) lobes. In contrast, we observed a profile of smaller CT across all lobes in the NS group relative to TD. We found smaller CT mainly in the frontal (bilateral pars triangularis, pars opercularis, precentral gyrus, right superior frontal gyrus, and left rostral and caudal middle frontal gyri) and parietal (right precuneus gyrus, paracentral gyrus, superior parietal and inferior lobes, and bilateral supramarginal gyrus) lobes.

Fig 5.

A. Differences in regional CT. P values (FDR-corrected) for independent samples t-tests planned comparisons of Turner syndrome (TS) and Noonan syndrome (NS) relative to typically developing (TD) controls of cortical thickness (CT) from parcellated brain regions according to the Desikan-Killiany atlas (Desikan et al., 2006). Relative to TD, larger mean CT ROIs are shaded in orange and smaller mean CT ROIs are shaded in blue. B. TS and NS show different profiles of alterations in regional CTs relative to typically developing (TD) individuals. We used a log-based color scale to depict significant between-group differences (FDR corrected p-values) in CT. The color scale was generated from the negative log of the p value (e.g., the negative log (p) of 0.05 = 1.3). Smaller regional CTs relative to typically developing (TD) individuals are represented in cold colors with deeper shades of blue indicating smaller p values. Larger regional CTs relative to TD are represented in warm colors with brighter shades of red indicating smaller p-values.

Principal component analysis results show that parieto-occipital regions primarily drive between-group profiles that distinguish TS, NS, and TD groups.

In line with the univariate analyses results, PCA results mirrored the effect of TS on SA and of NS on CT. The scree plot of regional SA (see Fig. S1) showed that most of the variance was explained by the first two principal components (PC1 and PC2), leading to 20.7% of cumulative explained variance (Fig. 6A). SA in the following cortical regions displayed the top five highest loadings in PC1 (in descending order): right precuneus gyrus, right superior temporal gyrus, left postcentral gyrus, left precuneus gyrus, and right superior parietal lobule. In PC2, SA in the following cortical regions displayed the highest loadings (in descending order): left pericalcarine sulcus, right pericalcarine sulcus, left cuneus, and left and right orbitofrontal gyri. Projections of all loadings of PC1 and PC2 on their specific brain region’s SA can be found in Fig. 6B.

Fig 6.

A. Biplot of first (PC1) and second (PC2) principal components of the PCA on regional Surface Area (SA). PC1 is represented along the x-axis and explains 13% of the variance. PC2 is represented along the y-axis and explains 7.7% of the variance. Ellipses are drawn at 95% confidence levels. Turner syndrome (TS) is displayed in orange. Noonan syndrome (NS) is displayed in gold. Typically developing (TD) are displayed in blue. B. Variable (SA) loadings (weight) on PC1 are projected onto the corresponding brain regions on the top row. Variable (SA) loadings (weight) on PC2 are projected onto the corresponding brain regions on the bottom row. Red colors indicate positive loadings; blue colors indicate negative loadings. Color bars and loading ranges are displayed to the right of each row. Regions were defined following the Desikan-Killiany atlas parcellation (Desikan et al. 2006).

The scree plot for CT (Fig. S2) showed the data concentrated around a single dominant dimension, as exemplified by a significant drop in explained variance after the first component (PC1= 29.3%) (see Fig. 7A - PC1 and PC2 are plotted for illustrative purposes). The following cortical regions’ CT displayed the top five highest loadings in PC1 (in descending order): right inferior parietal lobule, right superior parietal lobule, left superior frontal gyrus, right supramarginal gyrus, and right precuneus gyrus. Projections of all loadings of PC1 on their specific brain region’s CT can be found in Figure 7B.

Fig 7.

A. Biplot of first (PC1) and second (PC2) principal components of the PCA on Cortical Thickness (CT). PC1 is represented along the x-axis and explains 29.3% of the variance. PC2 is represented along the y-axis and explains 5.6% of the variance. Ellipses are drawn at 95% confidence levels. Turner syndrome (TS) is displayed in orange. Noonan syndrome (NS) is displayed in gold. Typically developing (TD) are displayed in blue. B. Variable (CT) loadings (weight) on PC1 are projected onto the corresponding brains region. Red colors indicate positive loadings; blue colors indicate negative loadings. Color bars and loading ranges are displayed to the right of each row. Regions were defined following the Desikan-Killiany atlas parcellation (Desikan et al. 2006).

Visual inspection of the loadings in Figs. 6A-7A suggested that SA was more informative than CT in distinguishing between TS, NS, and TD. Average sensitivity and specificity derived from the KNN classification algorithm with k-fold cross-validation (K = 3; including PC1 and PC2 for consistency across measures) showed that both measures where higher for SA, with 22% higher sensitivity (SA:0.581 ± 0.043 vs CT: 0.478 ± 0.160) and 7% higher specificity (SA: 0.791 ± 0.021 vs CT: 0.739 ± 0.080) compared to CT. Parameter perturbation experiments where we repeated the classification analysis but setting K=4-5 replicated this finding (Table S6).

DISCUSSION

In this study, we investigated whether TS and NS, two genetically different yet physically and behaviorally similar syndromes, have divergent effects on the developing brain. This was done across two levels: behavioral and neuroanatomical. We found that TS and NS have qualitatively (affected ROIs/structures) and quantitatively (smaller or larger) distinct neuroanatomical phenotypes that include differences in subcortical volumes, SA, and CT. Our data-driven approach showed that group separation was driven primarily by a hub of parietal regions and by SA (relative to CT). Methodological differences between prior studies have limited direct comparisons of neuroanatomical phenotypes across syndromes. However, by simultaneously contrasting these two syndromes to TD and to each other on consistently acquired and processed neural data from a large, age- and pubertal status-matched, female-only cohort, we were able to show syndrome-specific effects on brain morphology while accounting for potential confounders influencing brain and behavior.

As predicted, we found similarities between TS and NS at the behavioral level (Figs. 1,2). These findings align with previously described attention/hyperactive-impulsive and socio-communicative deficits in TS and NS (Hong et al., 2011; Lepage et al., 2013a; Alfieri et al., 2014; Green et al., 2015; Pierpont et al., 2015; Perrino et al., 2018), underscoring how, even in the context of known causal genetic mutation, behavioral phenotypes are non-specific. This non-specificity has challenging implications for the use of behaviors as potential markers for targeted therapeutic interventions.

Results from the univariate analyses supported our hypothesis of diverging neural phenotypes, showing different profiles of affected regions in TS and NS. Qualitatively, syndromes differed in affected cortical measures (TS affected mostly cortical SA whereas NS mostly affected CT) and regions (TS affected occipital-parietal lobes and NS affected precentral gyri and frontal regions). Quantitatively, for SA, individuals with TS showed a profile of smaller SA versus individuals with NS who showed larger SA (Fig. 4). For CT we observed the opposite pattern, with only a few regions affected (larger) in individuals with TS, and several smaller in individuals with NS (Fig. 5). The contrasting findings for SA and CT in our clinical groups likely result from different underlying biological mechanisms. Abnormalities of subcortical and cortical neuroanatomy in TS are likely attributed to the lack of expression of genes escaping inactivation (about 20-30%) in females with two X chromosomes. These genes likely play a role in brain development as the X chromosome is highly enriched in genes expressed in neural tissue (Nguyen & Disteche, 2006; Hong et al. 2014; Raznahan & Disteche, 2021). In NS, cortical abnormalities might reflect perturbations in the ratio of glia versus neurons, resulting from downstream effects of increased SHP2 protein activity. Increased SHP2 activity is associated with increased neurogenesis over astrogenesis resulting in a decrease in astrocytes (Gauthier et al., 2007). Similarly, reduced CT has also been observed in mouse models of Neurofibromatosis type 1 (NF-1; Zhu et al., 2001) and individuals with NF-1 (Barkovich et al., 2018) suggesting that thinning of the cortex could be a pathognomonic neural marker of Rasopathies.

Nonetheless, we observed a few regions where SA and CT were similarly affected (qualitatively and quantitatively) in TS and NS. Two of these converging regions (superior temporal sulcus and temporoparietal junction) have been linked to social cognition aspects, such as the perception of biological motion, (Allison et al., 2000; Grossman et al., 2000), theory of mind, and intention and agency representation (Gobbini et al., 2007; Hein and Knight, 2008). Moreover, some of these regions (left superior temporal and right parahippocampal gyri) have been linked to autistic traits in the general population (Jiao et al., 2010; Scheel et al., 2011; Cauvet et al., 2019). Temporal regions, and particularly their SA, are highly influenced by additive effects of single-nucleotide polymorphisms (SNPs) (Chen et al., 2015). Thus, convergence in this context may reflect high sensitivity to the combined effects of many common SNPs and non-heritable genetic variation (including rare mutations). Alternatively, these regions may be indicative of commonalities in the downstream effects of distinct biological mechanisms.

Findings from our multivariate data-driven analysis supported our hypothesis of diverging neural phenotypes in TS and NS. Group separation (TS, NS, and TD) was driven mostly by parieto-occipital regions, and primarily driven by SA of parieto-occipital regions affected in TS but not in NS. Most of these regions showed smaller SA, a robust finding in TS (Brown et al., 2004; Green et al., 2014). In contrast, we found that CT-based group separation was driven by fronto-parieto-occipital regions affected in NS. Parieto-occipital regions are highly influenced by genetic effects (heritable) but only a small fraction of this heritability is explained by common genetic variability (Chen et al., 2015). Thus, it is theorized that these regions are highly sensitive to rare genetic variation (Chen et al., 2015). Results from the KNN classification algorithm confirmed that SA was more informative than CT for group separation. An explanation for this finding is that SA best differentiates these groups due to its higher heritability but lower polygenicity relative to CT (van der Meer et al., 2020). Alternatively, it may also reflect the disproportionate impact X-chromosome has on the SA of parietal regions (Mallard et al., 2021) and/or susceptibility to dosage effects, as parieto-occipital regions also seem affected in females with fragile X syndrome (Lee et al., 2021).

The present findings suggest that regional morphometric measures, such as SA and CT, are, not only sensitive to, but also specifically vulnerable to different genetic abnormalities, underscoring their potential as disease-specific neuroanatomical markers. In animal models of NS and NF-1, statins (and mitogen-activated protein kinase (MEK) inhibitors in NS), normalize excitatory synaptic function and long-term potentiation deficits, and improve learning and memory deficits (Li et al., 2005; Lee et al., 2014). This success has not been consistently replicated on behavioral measures in human clinical trials (Chabernaud et al., 2012; van der Vaart et al., 2013; Payne et al., 2016). However, preliminary findings on human NF-1 trials suggest statins normalize function of both cellular mechanisms and neural structure and connectivity associated with social functions (Chabernaud et al., 2012; Stivaros et al., 2018). Therefore, future studies would benefit from including neuroanatomical and multiparametric imaging-derived measures, in addition to behavioral measures, when assessing treatment effects.

This study is not without limitations. First, while our research presents larger cohorts of neurogenetic conditions (TS and NS) than previous studies, particularly for NS (Johnson et al., 2019), these samples remain relatively small. However, the concern for lack of power at detecting differences, particularly when accompanied by strict correction for multiple comparisons as done here, is partly mitigated by the large effect sizes of both syndromes on brain structure. Second, our findings are limited to TS and NS. Third, while we used a KNN-based classifier to compare the differentiating abilities of SA and CT rather than for solving a classification problem per se, sensitivity and specificity for SA and CT were moderate. Additionally, generalizability is limited as KNN setups are highly sample dependent.

In this study, we provide evidence of robust neuroanatomical phenotypes intrinsic to TS and NS that, unlike behavioral phenotypes, are syndrome-specific such that distinct cortical regions and subcortical structures are differentially affected. Our findings are strengthened by our matched cohort and study design, making it unlikely that the differing morphological patterns found are due to differences in data acquisition or preprocessing, or to confounders that affect brain development. These findings demonstrate that different genetic aberrations have different downstream effects on the brain. These effects are more distinct at the neural level compared to the behavioral level.