Neural Correlates of Obesity and Inflammation in Children and Adolescents with Congenital Adrenal Hyperplasia

Abstract

Introduction

Patients with classical congenital adrenal hyperplasia (CAH) exhibit an increased prevalence of obesity from childhood including central adiposity and inflammation. There is also an emerging affected brain phenotype in CAH, with decreased cortico-limbic gray matter volumes and white matter abnormalities. We aimed to study the relationship between brain structure, obesity, and inflammation in children and adolescents with CAH compared to controls.

Methods

27 CAH (12.6 ± 3.4 y, 16 females) and 35 control (13.0 ± 2.8 y, 20 females) participants had MRI of gray matter regions of interest (prefrontal cortex [PFC], amygdala, hippocampus) and white matter microstructure (fornix, stria terminalis [ST]). Anthropometric measures and lab analytes were obtained. Relaimpo analyses (relative importance for linear regression; percent variance) identified which brain structures were most different between groups. Subsequent regressions further quantified the magnitude and direction of these relationships. Correlations analyzed relationships between brain structure, obesity, and inflammation in the context of CAH status.

Results

PFC (13.3% variance) and its superior frontal (SF) subregion (14%) were most different between CAH and controls for gray matter; ST (16%) for white matter. Patients with CAH had lower caudal middle frontal (β = −0.56 [−0.96, −0.15]) and superior frontal (β = −0.58 [−0.92, −0.25]) subregion volumes, increased orientation dispersion index in the fornix (β = 0.56 [0.01, 1.10]) and ST (β = 0.85 [0.34, 1.36]), and decreased fractional anisotropy in the fornix (β = −0.91 [−1.42, −0.42]) and ST (β = −0.83 [−1.34, −0.33]) (all p’s < 0.05) indicating axonal disorganization, reduced myelin content, and/or higher microglial density within the affected white matter tracts. For the full cohort, SF was correlated with MCP-1 (r = −0.41), visceral adipose tissue (r = −0.25), and waist-to-height ratio (r = −0.27, all p’s < 0.05); ST was correlated with MCP-1 (r = 0.31) and TNF-α (r = 0.29, all p’s < 0.05); however, after adjusting for CAH status, almost all correlations were attenuated for significance.

Conclusions

Relationships among key brain structures, body composition, and inflammatory markers in pediatric patients with CAH could be largely driven by having CAH, with implications for obesity and neuroinflammation in this high-risk population.

Introduction

Classical congenital adrenal hyperplasia (CAH) due to 21-hydroxylase deficiency is the most common primary adrenal insufficiency in pediatric patients, affecting 1 in 15,000 live births. CAH is characterized by cortisol and aldosterone deficiencies, along with androgen excess secondary to disrupted steroid biosynthesis. Fetuses with classical CAH are exposed to an altered intrauterine environment from the 7th gestational week through pregnancy, from when the adrenal gland is formed, with virilized external genitalia in newborn females [1, 2]. However, fetal programming during early stages of development is not understood in CAH, including its role in predisposition to postnatal disease. Children and adolescents with CAH exhibit a higher prevalence of obesity, with an earlier adiposity rebound and metabolic morbidity beginning at a prepubertal age [3–6]. Adolescents and young adults with CAH can also exhibit increased abdominal adiposity, with a higher ratio of pro-inflammatory visceral-to-subcutaneous adipose tissue compared to controls [7] and an increased waist-to-height ratio [8]. The underlying pathophysiology for these observations of increased obesity and adiposity in patients with CAH needs to be better understood [9].

There is an emerging brain structural phenotype in pediatric and adult patients with CAH that includes affected regions that are important for the regulation of behavior: decreased total intracranial volume (i.e., whole brain), smaller volumes for prefrontal cortex (PFC) and medial temporal lobe (amygdala and hippocampus), and affected white matter tracts connecting key gray matter regions [10–14] (shown in Fig. 1). Expression of androgen and glucocorticoid receptors throughout the brain could mediate effects from fluctuating hormone levels on structural brain development in CAH from an early age [10, 11], with emerging evidence of differences in brain structure in early life, including smaller thalamic regions in infants [15, 16]. Importantly, the PFC is associated with self-control and the ability to make informed decisions, and individuals with obesity exhibit a weakened ability of the PFC and limbic system (e.g., amygdala, hippocampus, basal ganglia, and associated white matter neurocircuitry) to regulate ingestive behaviors and dietary decision-making [17, 18], which can influence obesogenic behaviors. White matter microstructure affected in patients with CAH include outgoing tracts from the hippocampus (fornix) and amygdala (stria terminalis) that are part of the limbic system and important for memory and emotional regulation [11, 12, 14]. In otherwise healthy children and adolescents, structural morphology of the PFC and amygdala has been associated with both obesity and dietary self-control [19]. However, there is little known about the relationship of affected brain structures with obesity in patients with CAH.

Fig. 1.

Left panel: one representative sagittal slice of an adolescent patient with classical CAH depicting white matter tracts of interest. The fornix is the main outgoing tract (dark white) from the hippocampus, while the stria terminalis is the main outgoing tract (dark black) from the amygdala. Right panels: two representative sagittal slices from an adolescent patient with classical CAH depicting gray matter regions of interest: amygdala, hippocampus, and subregions of the prefrontal cortex.

As well, white matter microstructural integrity can be affected by obesity-induced, chronic, low-grade systemic inflammation [14, 20, 21]. Obesity can be considered a form of chronic low-grade systemic inflammation. Studies concerning other conditions such as type 2 diabetes mellitus exhibit changes in white matter microstructural integrity, specifically axonal demyelination, in adolescents with obesity compared to controls [22]. Although there is little known regarding the relationship between obesity and white matter microstructure in patients with CAH, there are rodent models that suggest that obesity induces neuroinflammation in the hypothalamus, amygdala, and hippocampus, causing a decreased number of newly generated neurons in these regions [23–25].

Therefore, changes and activation of brain regions connected to regulation and reward seeking in children and adolescents with CAH could affect their dietary decision-making and risk for obesity. Furthermore, these structural changes in the brain could also be a consequence of obesity and associated inflammation. However, the role of brain structural abnormalities and neuroinflammation in the setting of obesity in youth with CAH remains yet to be identified. Thus, we aimed to study the relationship between key brain regions, body composition, and inflammatory markers in children and adolescents with classical CAH.

Methods

Study Population

This was a prospective cross-sectional study of 27 children and adolescents with classical CAH due to 21-hydroxylase deficiency (age 12.6 ± 3.4 years, 16 females) and 35 healthy controls (age 13.0 ± 2.8 years, 20 females) similar for age, sex, and Tanner stage (as performed by a pediatric endocrinologist) [26] (Table 1). Patients with CAH had either the salt-wasting (n = 25) or simple-virilizing form (n = 2) diagnosed by positive newborn screen (n = 12), biochemically (n = 9), and/or by CYP21A2 genotype (n = 6). Exclusionary criteria for all participants included prenatal drug or alcohol exposure, premature birth, serious medical illness (other than CAH), eating disorders, serious food allergies, psychotropic medication, psychiatric or developmental disorders, and significant neurological conditions. Participants were also screened for factors that would prevent the proper and safe usage of magnetic resonance imaging (MRI), including irremovable ferrous materials (e.g., braces), need for hearing aids, uncorrectable vision impairments (e.g., color blindness or blind spots), or claustrophobia. All participants were required to be able to read and speak English to understand the study tasks. This study was approved by the Children’s Hospital Los Angeles (CHLA) Institutional Review Board, study number CHLA-14-00191. Written informed consent was obtained from all parents or legal guardians and/or participants older than 14 years old. All minors up to 14 years of age gave assent.

Table 1.

Study population characteristics for CAH and control youth

	CAH (n = 27)	Controls (n = 35)	Group difference, p
Age, years	12.6±3.4	13.0±2.8	0.62
Biological sex, n
Female	16	20	0.87
Male	11	15
BMI z-score	1.61±0.86	0.80±0.95	0.001
Ethnicity, n
Hispanic	11	20	0.28
Non-Hispanic	16	15
Race, n
White	19	24	0.64
Black	1	3
Asian	2	3
Native American	0	1
Unknown	5	4
CAH Form, n
Salt-wasting	24	–	–
Simple-virilizing	3
CAH medications
Glucocorticoid dose, mg/m2/day	15.1±2.9	–	–
Fludrocortisone dose, mg/day	0.10±0.04
CAH analytes
17-OHP, ng/dL	2,125 [736, 4,324]	–	–
nmol/L	64.39 [22.3, 131.02]
Testosterone, ng/dL	2 [0.7, 6.4]
nmol/L	0.07 [0.02, 0.22]
Androstenedione, ng/dL	75 [31, 94]
nmol/L	2.62 [1.08, 3.28]
Plasma renin activity, ng/mL/h and µg/L/h	3.46 [1.12, 5.41]
Highest newborn 17-OHP, ng/dL	15,300 [7,600, 25400]
nmol/L	463.59 [230.28, 769.62]

Mean ± SD; median [IQR].

Anthropometric measures obtained in all participants included waist circumference (cm; between the lowest rib and iliac crest), hip circumference (cm; at the level of the greater trochanter), weight (kg; using an electronic scale), and height (cm; using a digital stadiometer). Waist-to-hip ratio, waist-to-height ratio (WHtR), and body mass index (BMI) were calculated. BMI z-scores were derived using Center for Disease Control Growth Chart data [27]. MRI was performed (Achieva 3-Tesla; Philips Healthcare, Cleveland, OH, USA) to assess abdominal subcutaneous adipose tissue (SAT) and visceral adipose tissue (VAT) volume as a proportion of total abdominal adipose tissue volume (proportional SAT and VAT).

All participants had a fasting morning blood draw for measurement of the following analytes: leptin, monocyte chemoattractant protein-1 (MCP-1), resistin, tumor necrosis factor-α (TNF-α), plasminogen activator inhibitor-1 (PAI-1), interleukin-6 (IL-6), and adiponectin (Multiplex Human Metabolic Hormone and Human Adipokine Magnetic Bead Panels, Millipore Sigma, Burlington, MA, USA). For patients with CAH, additional hormone analytes measured prior to medication doses included 17-hydroxyprogesterone (17-OHP), androstenedione, testosterone, and plasma renin activity (PRA), all by liquid chromatography tandem mass spectrometry (Quest Diagnostics Nichols Institute, San Juan Capistrano, CA, USA) (Table 1).

For patients with CAH, we also recorded the newborn screen 17-OHP result (n = 21; 28,401.2 ± 44,610.0 ng/dL or 859.4 ± 1,349.9 nmol/L) measured on filter-paper blood specimens (California Department of Public Health Services; conversion to serum 17-OHP concentration [ng/dL] = whole blood units [nM] × 66) [28] and/or the confirmatory serum 17-OHP (n = 20; 23,795.4 ± 45,204.9 ng/dL or 720.1 ± 1,367.9 nmol/L) collected at diagnosis. For analyses, the higher measurement between the newborn screen and confirmatory serum 17-OHP at diagnosis was utilized (Table 1). Medication doses were recorded for patients with CAH, including glucocorticoid dose (mg/m²/day) as hydrocortisone equivalents (prednisone dose was multiplied by 5 and dexamethasone dose was multiplied by 80) [4] and fludrocortisone dose (mg/day) (Table 1).

Neuroimaging: Gray Matter

T1- and T2-weighted MRI acquisition (3T, Siemens Prisma) and settings, structural MRI preprocessing, and quality control methods were followed as previously described [10]. All scans were reviewed by a radiologist for incidental findings, with a Type 1 Chiari malformation found in one subject with CAH and an arachnoid cyst in another two subjects. Analyses were performed including these subjects as sensitivity analyses did not yield group differences when we excluded the 3 patients [10]. Of note, central nervous system anomalies have been previously reported in CAH [10, 12]. Additionally, 1 patient with CAH had medial temporal lobe data that were not useable and were therefore excluded from analyses involving this brain region.

Structural image processing and whole-brain segmentation with automated labeling of different neuroanatomical structures were obtained using FreeSurfer v6.0 [29, 30]. Previously identified gray matter regions of interest included amygdala, hippocampus, and five a priori subregions of the PFC (superior frontal, rostral middle frontal, caudal middle frontal, lateral orbitofrontal, medial orbitofrontal) for both the right and left hemispheres [10] (shown in Fig. 1). For the purposes of analyzing PFC volume with the amygdala and hippocampus, the total PFC variable was computed as the sum of its five subregions.

Neuroimaging: White Matter

Multi-shell diffusion-weighted image (DWI) acquisition and tractography were performed as previously described [14]. White matter tracts of interest included the fornix and stria terminalis which have been previously identified as differing between patients with CAH and controls [14] (shown in Fig. 1). DWI parameters and image collection to enhance echo planar imaging distortion correction were followed [14]. Ten of the 62 participants were excluded from white matter tract analysis due to incidental findings and motion artifacts.

DWI data were fitted to two models to probe white matter microstructural integrity: traditional diffusion-tensor imaging (DTI) and the novel neurite orientation dispersion and density imaging (NODDI). DTI measures extracellular water diffusion only with main outcomes including fractional anisotropy (FA) and mean diffusivity (MD). FA and MD are typically inversely related, with a higher FA indicating increased myelination and/or more densely packed axons, along with increased microstructural integrity. In contrast, a higher MD indicates decreased microstructural integrity. NODDI measures both intracellular and extracellular water diffusion and can more accurately estimate the biophysical properties of the brain [31]. NODDI outcomes include neurite density index (NDI; intraneurite diffusion), orientation dispersion index (ODI; extraneurite diffusion), and isotropic water fraction (ISO; representing cerebrospinal fluid) [32]. While most neuroimaging has utilized DTI modeling, the inclusion of NODDI outcomes increases our understanding of microstructural integrity [33].

Statistical Analysis

All analyses were performed in R v4.1.1 [34]. The sum of left and right hemisphere volumes was computed for each gray matter region of interest, as well as for white matter tracts of interest, with no significant evidence of laterality in this cohort noted previously [10, 14].

Preliminary analysis was performed to assess the distribution of brain, body composition, and inflammatory markers and their univariable relationships with belonging to the CAH group. All brain volume measurements for major regions and subregions appeared normally distributed in both CAH and control groups. Large violations of normality (heavy right skew) were found and subsequently log-transformed for: IL-6, adiponectin, 17-OHP, testosterone, androstenedione, and the highest newborn 17-OHP. PRA was also right-skewed, but a square root transformation was found to best transform the variable to normality.

With the exception of study population characteristics, all p-values reported for between-group comparisons were corrected for the false discovery rate (i.e., to reduce the chance of false positives that may arise due to testing multiple variables). T tests were performed to evaluate group differences in body composition and inflammatory markers between CAH and control groups.

Brain Regions and Tracts: Relaimpo and Regression Analyses

Because of the large number of subregions and tracts under consideration, we used a statistical approach called relaimpo analyses to identify the most important regressors before proceeding to regression model building. Relaimpo analysis is based on the R package “Relative Importance of Linear Regression” (R v2.2-3; Lindemann, Merenda, and Gold [lmg]) [35] and was used to identify which regressors (brain region or white matter tract) were most different between CAH and control groups by computing the r_lmg²  contribution of each predictor on the outcome (i.e., CAH vs. control) in the regression model, averaged over all possible sequential orderings. Specific gray matter regions and white matter tracts were included in the analysis [10, 14] with relaimpo estimating the percent variance, which is equivalent to the average r_lmg²  over the different linear regressions for all regions or tracts that could be adjusted for. A greater percent variance represents a greater difference between the two groups for that region or tract, with the most different regressor having the highest percent variance. The r_lmg²  analyses were adjusted for age, sex, Tanner stage, and (for gray matter) whole-brain volume.

Following relaimpo analyses, we performed regression analyses focusing on how gray and white matter variables varied based on CAH in order to quantify the magnitude and direction of the relationships suggested by the relaimpo. Regression β-coefficients reflect standardized effects of CAH versus controls (i.e., the average difference in brain outcome between CAH and controls, scaled by the standard deviation), with 95% confidence intervals.

Brain, Body Composition, Inflammatory Markers: Correlations

Correlations were performed across all participants between the most different gray matter regions and white matter tracts, with inflammatory markers and body composition variables. Correlations were then visualized using the ggcorrplot package in R to help compare the magnitude of correlation coefficients among all pairwise variable combinations. All correlations were adjusted for age, sex, Tanner stage, and (for gray matter) whole-brain volume. To help determine whether these correlations were spuriously caused by CAH, we then performed correlations on regression-adjusted variables. Briefly, brain regions, inflammatory markers, and body composition variables were treated as outcomes in a separate regression model with CAH as an independent variable. The residuals were saved for each regression model as the “regression-adjusted” version of that variable. These regression-adjusted variables were used in a subsequent correlation analysis to determine the degree of relationship among the variables that was not due to CAH and BMI z-score.

Results

Body Composition and Inflammatory Markers: Group Differences

For group differences in body composition and inflammatory markers, patients with CAH had higher WHtR (CAH 0.56 ± 0.11, controls 0.46 ± 0.06; p = 0.002), proportional VAT (CAH 0.07 ± 0.03, controls 0.05 ± 0.03; p = 0.03), proportional SAT (CAH 0.35 ± 0.13, controls 0.25 ± 0.14; p = 0.01), and BMI z (CAH 1.61 ± 0.86, controls 0.80 ± 0.95; p = 0.004) compared to controls. As well, patients with CAH exhibited higher MCP-1 (CAH 93 ± 37 pg/mL, controls 62 ± 23; p = 0.003) and leptin (CAH 14,627 ± 11,231 pg/mL, controls 7,182 ± 7,119; p = 0.01) compared to controls (Table 2).

Table 2.

Group differences in body composition and inflammatory markers

	CAH (n = 27)	Controls (n = 35)	p
Body composition
WHtR	0.56±0.11	0.46±0.06	0.002
BMI z	1.61±0.86	0.80±0.95	0.004
Proportional VAT	0.07±0.03	0.05±0.03	0.029
Proportional SAT	0.25±0.14	0.35±0.13	0.01
Inflammatory markers
MCP-1, pg/mL	93±37	62±23	0.003
Leptin, pg/mL	14,627±11,231	7,182±7,119	0.012
TNF-α, pg/mL	3.70±1.31	3.30±1.56	0.4
Resistin, µg/mL	0.03±0.02	0.03±0.02	>0.9
IL-6, pg/mL	1.78±1.81	2.28±1.98	0.4
Adiponectin, µg/mL	3.69±1.01	3.57±1.01	0.7
PAI-1, µg/mL	0.03±0.01	0.03±0.01	0.4

Mean ± SD.

CAH, congenital adrenal hyperplasia; WHtR, waist-to-height ratio; BMI z, body mass index z-score; VAT, visceral adipose tissue; SAT, subcutaneous adipose tissue; MCP-1, monocyte chemoattractant protein-1; TNF-α, tumor necrosis factor-alpha; IL-6, interleukin-6; PAI-1, plasminogen activator inhibitor-1.

Brain Regions and Tracts: Relaimpo and Regression Analyses

The PFC was the most different gray matter region between CAH and control groups (percent variance 13.3%; shown in Fig. 2a) compared to the amygdala and hippocampus (0.2% and 0.4%, respectively). For subregions of the PFC, the superior frontal (14%) and caudal middle frontal (6%) were the most different between groups in terms of percent variance (shown in Fig. 2a) compared to the other subregions [i.e., rostral middle frontal (3%), medial orbitofrontal (3%), and lateral orbitofrontal (4%)].

Fig. 2.

Relaimpo (relative importance for linear regression) analyses to identify the brain regions that are most different between CAH and control participants. a For gray matter regions of interest, the prefrontal cortex (PFC) was the most different region between CAH and control groups (percent variance = 13%; left panel). Of the PFC subregions, the caudal middle (6%) and superior frontal (14%) were the most different between groups (right panel). b For white matter tracts, the fornix and stria terminalis were analyzed using two modeling approaches: diffusion tensor imaging (fractional anisotropy [FA] and mean diffusivity [MD]) and neurite orientation dispersion and density imaging (intracellular diffusion [NDI], isotropic water fraction [ISO], and orientation diffusion index [ODI]). For the fornix, ISO (14%), fractional anisotropy (11%), and ODI (9.6%) were the most different between groups. For the stria terminalis, ODI (16%) and fractional anisotropy (12%) were the most different between groups.

For the fornix white matter tract, ISO (14%), FA (11%), and ODI (9.6%) were most different between CAH and control groups in terms of percent variance (shown in Fig. 2b) compared to other measures (MD [7%] and NDI [0.1%]). For the stria terminalis white matter tract, ODI (16%) and FA (12%) were most different between groups (shown in Fig. 2b) compared to other measures (ISO [4%], NDI [3%], and MD [1%]).

To quantify the magnitude and direction of these relationships, we used standardized linear regression model coefficients, adjusting for age, sex, Tanner stage, and (for gray matter) whole-brain volume (Table 3). Youth with CAH (compared to controls) had lower volumes (i.e., negative directionality) in the caudal middle frontal (β = −0.56 [−0.96, −0.15]) and superior frontal (β = −0.58 [−0.92, −0.25]) subregions. Youth with CAH also showed higher values of fornix ISO (β = 0.82 [0.30, 1.33]), fornix ODI (β = 0.56 [0.01, 1.10]), and stria terminalis ODI (β = 0.85 [0.34, 1.36]), and lower values of fornix FA (β = −0.91 [−1.42, −0.42]) and stria terminalis FA (β = −0.83 [−1.34, −0.33]).

Table 3.

Effect of CAH on brain regions

	Prefrontal cortex subregions		White matter tracts
	caudal middle	superior frontal	fornix FA	fornix ISO	fornix ODI	stria terminalis FA	stria terminalis ODI
CAH	−0.56** (−0.96, −0.15)	−0.58*** (−0.92, −0.25)	−0.91*** (−1.42, −0.42)	0.82** (0.30, 1.33)	0.56* (0.01, 1.10)	−0.83** (−1.34, −0.33)	0.85** (0.34, 1.36)

β-coefficients (95% confidence interval) are reported. *p ≤ 0.05; **p ≤ 0.01; ***p ≤ 0.001.

Regressions were adjusted for age, sex, Tanner stage, and (for gray matter) whole-brain volume.

CAH, congenital adrenal hyperplasia.

Brain, Body Composition, Inflammatory Markers: Correlations

Correlations were then performed across all participants between the most different gray matter regions and white matter tracts, inflammatory markers, and body composition variables (shown in Fig. 3a). For the gray matter regions, the caudal middle was correlated with MCP-1 (r = −0.44, p < 0.001). As well, the superior frontal was correlated with MCP-1 (r = −0.41, p = 0.001), proportional VAT (r = −0.25, p = 0.048) and WHtR (r = −0.27, p = 0.03).

Fig. 3.

Correlation analyses between key brain regions of interest (gray matter = caudal middle, superior frontal; white matter = fornix FA, fornix ODI, fornix ISO, stria terminalis ODI, stria terminalis FA), inflammatory markers (TNF-α, MCP-1), and body composition variables (SAT, VAT, WHtR). Brain regions of interest across the heatmaps were identified by relaimpo analyses as being the most different between CAH and control groups. Only inflammatory markers and body composition variables that had significant group differences between CAH and control groups were included in the figure. a The first heatmap includes all participants, adjusting for age, sex, Tanner stage, and whole-brain volume. b The second heatmap was additionally adjusted for CAH. Most relationships that were significant in the first heatmap became attenuated, suggesting that CAH could be the explanation for the significant correlations seen in the entire cohort.

For the white matter tracts, fornix FA was correlated with TNF-α (r = −0.39, p = 0.004) and WHtR (r = −0.40, p = 0.002). Fornix ODI was correlated with MCP-1 (r = 0.28, p = 0.04), proportional SAT (r = 0.28, p = 0.04), and WHtR (r = 0.38, p = 0.004). Stria terminalis FA was correlated with TNF-α (r = −0.35, p = 0.01). Stria terminalis ODI was correlated with TNF-α (r = 0.29, p = 0.04) and MCP-1 (r = 0.31, p = 0.02).

CAH Adjusted Correlations

Next, correlations were further adjusted for CAH status to determine its contribution to observed relationships (shown in Fig. 3b). After partialling out the effect of CAH, the only remaining significant correlations were between the caudal middle and MCP-1 (r = −0.33, p = 0.01), the superior frontal and MCP-1 (r = −0.26, p = 0.04), the fornix FA and TNF-α (r = −0.34, p = 0.01), and the stria terminalis FA and TNF-α (r = −0.28, p = 0.048). All other previously significant correlations were attenuated.

BMI z Adjusted Correlations

Given that BMI z differences were observed between study groups, and that patients with CAH have a higher prevalence of obesity compared to controls, sensitivity analyses were performed to address possible effects of weight status on correlation analyses by adjusting for BMI z. The vast majority of significant correlations for the full cohort were not affected after adjusting for BMI z. Some correlations gained significance (fornix ISO and stria terminalis ODI with WHtR; r = 0.33, p = 0.01 for both), while others lost significance (superior frontal with proportional VAT, fornix FA with WHtR, fornix ODI with MCP-1 and WHtR, and stria terminalis with TNF-α; all p’s > 0.05).

CAH Clinical Features: Correlations

Correlations were analyzed between clinical features for participants with CAH (17-OHP ln, testosterone ln, androstenedione ln, PRA square root, and glucocorticoid and fludrocortisone dosage) with gray matter regions, white matter tracts, body composition, and inflammatory markers as identified above. The only significant correlation was between BMI z and PRA square root (r = 0.51, p = 0.007). All other variables were not correlated with clinical features for participants with CAH (p > 0.05 for all).

Discussion

In children and adolescents with classical CAH due to 21-hydroxylase deficiency, we found disease-specific relationships between brain regions affected in CAH with body composition (WHtR) and inflammatory markers (MCP-1 and TNF-α). Key brain structures that were most different between CAH and control groups held significant negative correlations between PFC subregions (caudal middle and superior frontal) and MCP-1 (for both), as well as WHtR and VAT (superior frontal). White matter tracts were correlated with MCP-1 and TNF-α (for the fornix and stria terminalis), along with WHtR and SAT (for the fornix). However, when these correlations were adjusted for CAH status, the majority were attenuated for significance, suggesting that these relationships may be largely driven by having CAH.

Key brain structures affected in CAH comprise an emerging neural phenotype, with variations in structure and function potentially contributing to obesogenic, appetitive behaviors [18, 36]. Specifically, the PFC is involved in higher-order executive functions and the regulation of important limbic system regions for reward-seeking behaviors. Children and adolescents with CAH exhibit both smaller total PFC and subregion volumes (superior frontal, caudal middle frontal, and lateral orbitofrontal) compared to controls [10]. These regions notably overlap with the dorsolateral and medial PFC areas, which are key contributors to dietary self-regulation [36, 37]. The negative correlations between PFC subregions with both body composition and inflammatory markers seen in our study could therefore be meaningful in this patient population. Although more research is needed to fully understand the etiology of structural brain changes in CAH, these relationships in our study could potentially carry implications for these patients who have higher cardiometabolic risk throughout their lifetime [10, 14–16, 38, 39].

In addition, compromised white matter tracts in the limbic system, specifically the major output tracts for the hippocampus (fornix) and amygdala (stria terminalis), are noted in the brain phenotype of patients with CAH [14]. We found that patients with CAH had increased ODI values (positive directionality; fornix, stria terminalis), as well as reduced FA values (negative directionality; fornix, stria terminalis) compared to controls, indicating axonal disorganization, reduced myelin content, and/or higher microglial density within the affected white matter tracts. In addition, a higher WHtR and TNF-α correlated with FA reductions in the fornix, and WHtR, MCP-1, and TNF-α correlated with ODI increases in both the fornix and stria terminalis. These associations between white matter microstructural integrity, central adiposity (WHtR), and inflammatory markers could indicate an effect of obesity-induced neuroinflammation in CAH. Microstructural integrity can be affected by obesity as a form of chronic, low-grade systemic inflammation [23–25]. Rodent models involving neuroinflammation in the limbic system exhibit associations between white matter hyperintensities and increased visceral adipose tissue cytokines (including MCP-1) which recruit macrophages and damage myelin sheaths [23–25, 38–40]. We and others have found that there is an association between body composition and markers of white matter demyelination in the limbic system [24, 41, 42], with an additional relationship between white matter and inflammatory markers noted in our CAH patient population. Affected limbic white matter tracts can additionally impact dietary decision-making and appetite control [43], which may be important in patients with CAH who already exhibit disrupted cognitive processes (e.g., working and episodic memory) in adulthood [11, 12, 14, 19, 44–47].

There are several limitations to consider for our study. Our sample size was relatively small, and thus the associations detected may reflect only the strongest effects among variables. As well, we studied mostly patients with salt-wasting CAH, limiting our ability to examine disease severity over differing phenotypes. Although we found a positive correlation between PRA and WHtR, more correlations with clinical factors may be found with a much larger sample size. Analyses for subgroup-specific (CAH vs. control) correlations are planned for a larger cohort. Lastly, functional brain imaging would further elucidate the role of affected limbic structures in the dietary behaviors of patients with CAH.

We conclude that relationships between key brain regions affected in pediatric patients with CAH and markers of central obesity and inflammation could largely be driven by having CAH. This could have implications for the development of affected cognitive processes and their relationship to the increased obesity observed in patients with CAH. Further study of the limbic system and its role in functional dietary behaviors is merited in this high-risk population.

Acknowledgments

We gratefully thank our patients and their families who participated in this study. We also thank Norma Martinez, Christina Koppin, Michelle Canales, Eva Gabor, Heather Ross, Anisa Azad, and Emily Waters for their assistance with coordination. We thank Lilit Baronikian at the USC Metabolic Core for measuring the inflammatory marker concentrations. REDCap was supported by the Southern California Clinical and Translational Science Institute (SC CTSI).

Statement of Ethics

This study was approved by the Institutional Review Board at CHLA, CHLA-14-00191. Written informed consent was obtained from parents or legal guardians in accordance with The Code of Ethics of the World Medical Association.

Conflict of Interest Statement

M.E.G. has a research contract with Novo Nordisk; receives consultant fees from Adrenas, Eton Pharmaceuticals, Neurocrine Biosciences, Novo Nordisk, and Pfizer; receives royalties from McGraw-Hill and UpToDate; and serves on a data safety monitoring board for Ascendis. M.S.K. receives research support from Neurocrine Biosciences, Spruce Biosciences, and Diurnal and receives royalties from UpToDate.

Funding Sources

K01MH1087610 and R03HD090308 (NIH/NICHD to M.M.H.), K23HD084735 and R03HD101718 (NIH/NICHD to M.S.K.), and UL1TR000130 (CHLA CTU to M.S.K. and NCATS to T.AP.). This work was supported by grant UL1TR001855 from the National Center for Advancing Translational Science (NCATS) of the U.S. National Institutes of Health, a Conte Center Grant and Neuroimaging Core (5P50MH094258-5388) (to M.M.H. and T.A.P.), the CARES Foundation (to M.E.G. and M.S.K.), and the Grace Nixon Foundation and Abell Foundation (to M.E.G.). REDCap was supported by the Southern California Clinical and Translational Science Institute. The contents of this work are solely the responsibility of the authors and do not necessarily represent the official views of the National Institutes of Health.

Author Contributions

M.S.K., T.A.P., M.E.G., and M.M.H. obtained funding for this study. M.S.K. and M.M.H. conceived of and designed the study. M.S.K., D.L.C., N.R.F., S.L., and M.M.H. collected the data. D.L.C., S.L., and M.M.H. analyzed the brain imaging data. T.A.P. performed the statistical analyses. M.S.K., T.A.P., D.L.C., N.R.F., S.L., and M.M.H. interpreted and drafted the article. M.S.K., T.A.P., D.L.C., N.R.F., S.L., C.Y.W., M.E.G., and M.M.H. critically revised the article.

Funding Statement

K01MH1087610 and R03HD090308 (NIH/NICHD to M.M.H.), K23HD084735 and R03HD101718 (NIH/NICHD to M.S.K.), and UL1TR000130 (CHLA CTU to M.S.K. and NCATS to T.AP.). This work was supported by grant UL1TR001855 from the National Center for Advancing Translational Science (NCATS) of the U.S. National Institutes of Health, a Conte Center Grant and Neuroimaging Core (5P50MH094258-5388) (to M.M.H. and T.A.P.), the CARES Foundation (to M.E.G. and M.S.K.), and the Grace Nixon Foundation and Abell Foundation (to M.E.G.). REDCap was supported by the Southern California Clinical and Translational Science Institute. The contents of this work are solely the responsibility of the authors and do not necessarily represent the official views of the National Institutes of Health.

Data Availability Statement

Restrictions apply to the availability of data generated or analyzed during this study to preserve patient confidentiality or because they were used under license. The corresponding author will, on request, detail the restrictions and any conditions under which access to some data may be provided.