The pediatric obesity rate has increased explosively over the past 30 years, reaching up to 10% worldwide (Gerhart et al., 2022). This is a critical public health issue, since obesity is associated with far-reaching and costly implications, such as type II diabetes and heart disease (Flegal et al., 2010). Brain circuitry involved in motivation, reward, and cognitive control has been identified to explore neurobiological underpinnings of childhood obesity (Bruce et al., 2010).

Prader-Willi Syndrome (PWS) is a typical genetic obesity disease caused by the loss of paternal gene expression located in chromosome 15q11-q13 (Bittel and Butler, 2005). It is characterized by variable and typical phenotypes, affecting one in 15,000–20,000 live births (Soni et al., 2008). The predominant clinical manifestations of PWS are hypotonia and poor suck before the neonatal stage; as age increases, these transition to developmental delay and an insatiable appetite, showing hyperphagia as well as morbid and even life-threatening obesity without external restriction of food intake (Butler et al., 2006; Angulo et al., 2015). In addition, other phenotypes, such as hypogonadism, temperature instability, high pain threshold, hypersomnia, mild to moderate intellectual disability, and avariable range of social and behavioral difficulties, and psychiatric disturbances also occur in children with PWS (Dykens and Kasari, 1997; Clarke et al., 2002; Holland et al., 2003).

These diverse symptoms indicate a widespread developmental abnormality in the brain structure of children with PWS. Eating behavior is known to be regulated by high levels of cognitive control and satiety mechanisms based on brain subcortical reward and motivational neurocircuitry involving the amygdala, hippocampus, caudate nucleus, putamen, pallidum, nucleus accumbens (NAc), septal nucleus and thalamus (Orsi et al., 2011; Kirouac, 2015). As the predominant endocrine gland, the hypothalamus is responsible for sensing nutrients and regulating neuropeptides such as leptin, ghrelin, and insulin (Volkow et al., 2011; Angulo et al., 2015). Additionally, deep cerebellar nuclei (DCN) are associated with motor control, hypotonia, and satiation networks for sensing nutrient infusion in the gut and releasing dopamine (Blanco-Hinojo et al., 2021; Low et al., 2021). All of these brain nuclei comprehensively contribute to the recognition and regulation of homeostatic and hedonic eating behavior.

Previous neuroimaging studies of PWS have focused on cortical reward and inhibitory circuities and found functional deficits in multiple related regions, such as the dorsolateral and medial prefrontal cortex, orbitofrontal cortex, right anterior cingulate cortex, temporal lobe, and insula (Holsen et al., 2006; Xu et al., 2017). However, only a few studies have paid attention to the alteration of brain nuclei, let alone in children with PWS (Honea et al., 2012; Xu et al., 2017; Manning et al., 2018). Honea et al. (2012) found lower gray matter volume (GMV) in the right hippocampus in young adults with PWS. Lukoshe et al. (2017) reported no volumetric differences in the hypothalamus and mammillary bodies in children with PWS. In cerebellar studies, decreased whole cerebellar and relative lobular volume ratios in posterior inferior lobules in PWS individuals compared with normal controls were reported (Honea et al., 2012; Lukoshe et al., 2013, 2017; Yamada et al., 2020). In conclusion, due to the limited number of cerebral and cerebellar nuclei investigated in previous findings, a systematic and comprehensive study of the brain nuclei in children with PWS is needed.

To date, segmentation methods of subcortical nuclei for PWS have been concentrated on manual segmentation and voxel-based morphometry (VBM). However, there are still several knowledge gaps. First, based on voxel-wise comparison, VBM is prone to registration artifacts in the deep gray matter and may not be suitable for the analysis of cerebral nuclei. Second, arbitrary smoothing makes VBM insensitive to boundary location, and it is difficult to detect subtle atrophy of nuclei. Third, relatively small studies such as those involving PWS run the risk of false-positive findings or nonsignificant results using hypothesis-free segmentation methods (Patenaude et al., 2011; van den Bogaard et al., 2011; Nemmi et al., 2015; Zonneveld et al., 2019). Subcortical shape analysis can provide a direct, purely local measure of geometric changes by analyzing differences in boundary vertex locations, which has been applied to Parkinson’s disease and Huntington’s disease to find cerebral nuclei atrophy (van den Bogaard et al., 2011; Nemmi et al., 2015). Meanwhile, the shape analysis method has been validated in pediatric studies and makes it a promising technique for cerebral nuclei segmentation in children with PWS (Ortega et al., 2019; Lidauer et al., 2021). Moreover, it is well known that the hypothalamus is an indispensable nucleus in eating control, and because of its small size, the conventional segmentation method relies on manual delineation (Baroncini et al., 2012). A recent study accomplished auto-segmentation of 10 subregions of the hypothalamus by a deep convolutional neural network, which has been verified in several diseases (Billot et al., 2020; Chao et al., 2022; Lee et al., 2022), so the method was used in our article to investigate alterations in the hypothalamic subunits of PWS and OB subjects.

In summary, children with PWS and obesity share similar alterations in the cerebral cortex, which are mainly caused by acquired plasticity of the brain, while alterations in brain nuclei are closely related to inheritance (Roshchupkin et al., 2016). The current study aims to analyze the differences in brain nuclei between PWS, OB, and control children through multiple convincing analysis methods to reveal neural mechanisms involved in extreme hyperphagia and dietary restraint in PWS and obesity. We hypothesized that children with PWS would exhibit genetically influenced impairment in brain nuclei compared with the control group, which was not found in children with obesity. To test our hypothesis, we performed high-resolution T1-weighted magnetic resonance imaging (MRI) in a three-group study and implemented a systematic quantitative study on PWS brain nuclei using auto-segmentation techniques (Figure 1), which included shape analysis for cerebral nuclei, automated segmentation approach for subregions of the hypothalamus, and region of interest (ROI)-based analysis for DCN.

The study was approved by the Institutional Review Board at the University of Florida. We certify that all applicable institutional and governmental regulations concerning the ethical use of human volunteers were followed during this research.

In the current study, we investigated alterations in brain nuclei in children with PWS and obesity with systematic and combined methods, including shape analysis for cerebral nuclei, automated segmentation methods for the hypothalamus, and VBM for DCN. The results demonstrated that PWS individuals showed significant atrophy in the cerebral nuclei, hypothalamus, and cerebellum compared with control and OB subjects. We also found no significant difference in these brain nuclei between OB and control children. Our results suggest that although PWS and obesity share predominant clinical manifestations (e.g., overweight and hyperphagia) and similar alterations in cortical volume (Xu et al., 2017), impairment in brain nuclei may be very different. Thus, our findings indicate that alterations in brain nuclei might be associated with eating disorders, cognitive disability, and endocrine abnormalities that characterize PWS. Additionally, alterations in brain nuclei were found in PWS individuals during childhood, which might be due to the heritability of subcortical structural shape and the influence of imprinted genes (Roshchupkin et al., 2016), while the relationship between manifestations and brain nuclei in obese children still needs further investigation.

In the shape analysis, atrophy was found in the bilateral thalamus, pallidum, hippocampus, amygdala, NAc, and right caudate in individuals with PWS. The amygdala is the vital nucleus connecting the cortex and hypothalamus, which play an important role in the regulation of food intake by learned and motivational cues (Gottfried et al., 2003; Killgore et al., 2003; Kringelbach et al., 2003; Hinton et al., 2006). The amygdala itself possesses steady-state regulation by maintaining the balance between food satisfaction and a high level of insulin that inhibits eating (Areias and Prada, 2015). Animal studies have suggested that rodents with amygdala lesions showed insulin resistance and typical PWS phenotypes, such as hyperphagia, elevated weight, and hyperinsulinemia (King et al., 1994, 1996; Thaler et al., 2012). Therefore, the volumetric abnormality of the amygdala may be a predictor of such dysfunctions.

Previous studies have reported smaller hippocampal volumes in PWS patients than in normal controls (Honea et al., 2012; Lukoshe et al., 2013; Manning et al., 2018). Consistent with these findings, our results further investigated the shape variation in detail, revealing that the variation was mainly contributed by specific subregions, such as the dorsal, ventral, and lateral of field 1/3 of Ammon’s horn and subiculum. Hippocampal-dependent mnemonic functions play a dominant role in feeding behavior through external sensory food-relevant and internal energy-relevant information (Kanoski and Grill, 2017). Neurons in the dorsal hippocampus receive visuospatial information from the cortex, the ventral hippocampus primarily forms a bidirectional modulation of olfactory cues, and the ventral and lateral hippocampus receive gastrointestinal and other visceral information, all of which determine where, when, what, and how much to eat (Webster et al., 1991; De La Rosa-Prieto et al., 2009; Fanselow and Dong, 2010; Kanoski and Grill, 2017). Patients with bilateral hippocampal lesions show deficits in establishing episodic memory, leading to much more frequent eating. Although the mechanism of hippocampal reduction is unclear, one possible explanation is that hormone signal disturbance caused by abnormal genetic expression influences the formation and maintenance of new memories, including synaptic plasticity and neurogenesis (Shanley et al., 2001; During et al., 2003; Diano et al., 2006; Moult and Harvey, 2008).

The neurons in the NAc regulate motivation for feeding behavior and are activated or inhibited by hormones such as glucose-like-protein-1 or leptin, the dysfunction of which results in excessive food intake (Kenny, 2011; Alhadeff et al., 2012). The subregions of the NAc serve distinct functions; for example, the rostral part contains a hedonic hotspot where “liking” is enhanced for food rewards, while the caudal part contains opioid coldspot where positive hedonic impact is suppressed (Castro et al., 2015). Although the mechanism of NAc volumetric alteration is unclear, the differential performance in PWS and OB may suggest that the reduction is caused by genetically induced cytokine and brain metabolites.

It is well known that limbic cortico-striatal-pallido-thalamic loops are involved in emotional cognition, reward-guided decision-making, and goal-related behavior (Burton et al., 2015; Manning and Holland, 2015). The caudate nucleus plays a dominant role in cognitive function, and its dorsolateral part mediates memory processing and consolidation (Grahn et al., 2008; White, 2009). The ventral pallidum mediates the hedonic influence of food rewards and salience of motivational incentives, particularly through the caudal and rostral parts (Castro et al., 2015). In our study, subregional changes in the thalamus, pallidum and caudate nucleus were noted by the shape analysis method, providing finer neuroimaging evidence than previous findings. Honea et al. (2012) and Kim et al. (2020) reported a larger thalamic volume in obese individuals and a smaller white matter volume in PWS; Hayashi et al. (1992) found coagulative necrosis with nerve cell loss scattered throughout the pallidum; and a small GMV in the caudate nucleus was reported in both PWS and obese patients (Ogura et al., 2011; Kim et al., 2020). Although the genesis of volumetric alteration is unclear, our finding that specific subregions in PWS were much smaller than control and OB groups suggests genetically induced impairment may not be compensated by the acquired neuroadaptive reward system.

However, no morphological change was found in the cerebral nuclei of obese children, according to our study. Previous findings on the relationship between cerebral nuclei volume and obesity are inconsistent. Some studies have reported an enlarged amygdala, hippocampus, thalamus, putamen, and pallidus in obese individuals (Widya et al., 2011; Bernardes et al., 2018; Kim et al., 2020). However, Bobb et al. (2014) reported that higher BMI was associated with a lower volume of the amygdala, and a study with a large sample size of 12,087 subjects from the UK Biobank proved that the percentage of total body fat was negatively associated with subcortical GMVs in men, including the thalamus, caudate nucleus, putamen, globus pallidus, hippocampus, and NAc (Dekkers et al., 2019). From these contradictory evidences, we suggest that there might be sophisticated neurobiological interactions between obesity and cerebral nuclei under physiological and pathological conditions.

Hypothalamic dysfunction is well known to be responsible for the majority of PWS phenotypes, such as hyperphagia, hypersomnia, and multiple endocrine abnormalities (Angulo et al., 2015). However, limited by segmentation methods, most of the studies have not found significantly hypothalamic structural abnormality (Honea et al., 2012; Lukoshe et al., 2017). Recently, using the same Hypothalamic Subunits tool as in this manuscript, Brown et al. (2022) reported PWS adults had smaller volumes of whole hypothalamus and its subunits except for right a-iHyp, than that in obese and control adults. Consistent with the results of PWS adults, atrophy of the bilateral hypothalamus in PWS children was found in our study, specifically contributed by the bilateral in fTub andposHyp, right supTub and left a-iHyp subregions. The hypothalamic paraventricular nucleus, dorsomedial nucleus, and lateral hypothalamic area of the superior tubular hypothalamus, as well as the ventromedial nucleus and infundibular nucleus, comprise the feeding pathway, which facilitates or inhibits feeding through sensing cues, including hormone levels (leptin and ghrelin), visceroceptive inputs (such as food taste), olfactory cues, and glucose-sensitive neurons (Saper and Lowell, 2014). Previous studies have reported that large lesions centered in the ventromedial nucleus lead to hyperphagia and obesity, and neurons in the infundibular nucleus, paraventricular nucleus and lateral hypothalamus nucleus are affected in postmortem hypothalamic tissue specimens from PWS patients (Saper and Lowell, 2014; Correa-da-Silva et al., 2021). Meanwhile, no hypothalamic atrophy was found in OB in our study, which is consistent with a previous study with semiautomated segmentation of the hypothalamus of 338 subjects, indicating that hypothalamic volume is not associated with obesity (Thomas et al., 2019). The relationship between volumetric alterations in the hypothalamus and obesity needs to be further investigated.

The cerebellum plays an irreplaceable role in regulating satiety and meal termination, mainly by DCN and lateral-DCN neurons (Simerly and DiLeone, 2021). The dentate nucleus has been demonstrated to participate in a widespread functional network, including the striatum, hypothalamus, thalamus, sensorimotor and associative cortices, associated with cognition, learning, and sensorimotor coordination (Habas, 2010). The interposed nucleus assists the motor thalamus, mediates eye movement, and responds to tactile stimulation (Judd et al., 2021). The fastigial nucleus receives histaminergic and orexinergic afferents from the hypothalamus and directly projects to the hypothalamus, visceral-related nuclei/regions in the medullary reticular formations and the limbic system to regulate feeding and emotion (Zhang et al., 2016). The dysfunction of DCN is greatly associated with the phenotype of PWS. Low et al. (2021) observed a difference in DCN between individuals with PWS and control participants with fMRI responding to food images. Moreover, mice with a knockout of the Fmr1 gene showed a reduced volume of DCN and displayed the restricted, repetitive behaviors characteristic of PWS (Ellegood et al., 2010; Wilkes and Lewis, 2018). Meanwhile, as we know, no studies have been reported about DCN alterations in obese children. Consistent with previous studies, we reported a decreased volume of DCN in children with PWS, which was not found in OB. Our findings suggested that the dysfunction of DCN contributed to the developmental behavioral characteristics in PWS and provided evidence that DCN plays an important role in regulating satiety and meal termination in terms of genetics.

There are several limitations that need to be discussed here. First, even though the sample size was small in our study, leading sex and age were not perfectly matched, it had sufficient power to find between-group differences. Meanwhile, as a rare disease, large age ranges and small sample sizes are acceptable in PWS neuroimaging studies (Manning et al., 2018; Kim et al., 2020). Second, structural change in the two major PWS subtypes was not investigated because they do not exhibit clinical differences, except for hypopigmentation (Butler, 1990). Third, the slice number was fixed to 160 slices for all subjects, which was an alternative acquisition mode to the fixed slice thickness, causing the slice thickness to vary with the field of view in z direction. Though all images were resampled to the same voxel size before segmentation, the variation of slice thickness might have potential impact on the precision of segmentation. Fourth, Hypothalamic Subunits tool has not been widely validated on pediatric analysis, except for a recent preprint which was conducted by Voldsbekk et al. (2022) on young samples (age range 5–21 years). Although the quality control procedure was performed manually by an experienced radiologist for every subject in our study, the validation needs to be further investigated. Fifth, unlike subcortical nuclei and hypothalamus, the volume of DCN was calculated by adding total GMV and WMV within each ROI. As discussed previously, VBM methods have limitations in brain nuclei, though we have used the SUIT atlas for better registration. Finally, no significant results were found between OB and the control. Thus, the meaning of different kinds of changes in adult obesity could not be explained by our current study. In the future, the relationship between genetic subtypes and structural change could be investigated to examine the influence of heritable differences involving PWS on brain nuclei.

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

The studies involving human participants were reviewed and approved by Institutional Review Board at the University of Florida. Written informed consent to participate in this study was provided by the participants’ legal guardian/next of kin.

NW and MX conceived the study, performed the formal analysis, and wrote the original draft. NW pursued the implementation of the methodology, conducted the data processing and experiments, generated and validated the results. HY conducted quality control of raw data and the segmentation. All authors contributed to the article and approved the submitted version.

The study was supported by the Capital’s Funds for Health Improvement and Research (2020-4-7073), Capital Foundation of Medical Development, and Open access publication fees were supported by Capital Medical University.