Volumetric brain alterations associated with a high-fat, high-fructose obesogenic diet: Insights from cross-sectional and longitudinal MRI in mice

Abstract

Childhood obesity is associated with altered brain structure in humans, but the causal role of obesogenic diets and the longitudinal trajectory of these changes remain poorly understood. This study investigates the impact of a high-fat, high-fructose (HFHF) diet on brain volumetrics in mice using cross-sectional and longitudinal MRI. Mice were fed either a control chow or HFHF diet starting at 6 weeks of age. Structural MRI and voxel-based morphometry (VBM) were performed at 14 weeks (after 8 weeks of diet) and 6 months to quantify total brain and regional volumes (neocortex, hippocampus, subregions). Body weights were monitored throughout. HFHF-fed mice exhibited significant weight gain. Cross-sectionally, obese mice showed reduced total brain volume and increased neocortical volumes (mid- and posterior regions, alongside persistent posterior hippocampal decreases). Longitudinally, obese mice displayed an increase of total brain volume by 6 months, contrasting with age-associated decrease in controls. VBM revealed obesity-specific reductions in basal forebrain, sensory cortices, and hippocampal subfields, distinct from normative aging. Early-life obesogenic diets induce dynamic, region-specific brain alterations in mice, including increased neocortical and decreased hippocampal volumes. Longitudinal MRI captures dynamic volumetric trajectories providing insights into how obesogenic diets disrupt neurodevelopment.

1. Introduction

The global obesity pandemic continues to escalate at an alarming rate, particularly affecting children and adolescents [1–4]. This multi-factorial disease, influenced by complex interactions between genetic and environmental factors [5], poses significant challenges for therapeutic intervention. Despite recent advancements in pharmacological approaches for weight management, a substantial proportion of patients fail to achieve meaningful or sustained weight reduction or prevent the neurodevelopment consequences associated with obesity [6]; [7]. Thus, there is an urgent need to better understand how early-life obesity impacts brain development and to identify potential targets for intervention.

Robust animal models, particularly mouse models of diet-induced obesity, are critical for elucidating the biological mechanisms linking metabolic stress to brain structural changes [8–10]. Mouse models fed high-fat high-fructose (HFHF) diets closely mimic key aspects of human childhood obesity, including excess body weight, metabolic dysregulation, and low-grade systemic inflammation, that have been implicated in altering brain development. While some of these models are commonly used in non-alcoholic steatohepatitis (NASH) research, their value for investigating obesity-related neurodevelopmental effects is increasingly recognized. Our study confirms the translational relevance of the HFHF mouse model by demonstrating significant weight gain consistent with obesity, providing a controlled platform to examine diet-brain interactions during critical developmental windows.

Voxel-based morphometry (VBM) is a powerful neuroimaging technique widely used to quantify structural brain differences associated with neurological and psychiatric conditions, as well as typical neurodevelopment and aging [11–15]. Human neuroimaging studies also have consistently demonstrated that obesity is associated with structural alterations in both gray and white matter. A large-scale population study from the UK Biobank (n = 12,087; ages 45–76 years) reported reduced volumes in subcortical gray matter regions including the caudate, thalamus, hippocampus, and globus pallidus, along with widespread changes in white matter microstructural coherence [16]. Additional VBM studies have shown that individuals with higher body mass index (BMI) exhibit smaller volumes in cortical regions involved in cognition and reward processing, such as the frontal cortex and anterior cingulate cortex [17–19]. Importantly, these neuroanatomical alterations are not limited to adults. Pediatric studies have reported reduced gray matter volumes in the prefrontal cortex, thalamus, Para hippocampal gyrus/amygdala, and sensorimotor cortices in children with obesity [13, 20–22]. These findings raise critical concerns that childhood obesity may disrupt normal brain maturation, potentially increasing vulnerability to cognitive and behavioral dysfunction.

Animal studies further support and extend these findings by enabling experimental control over diet exposure and developmental timing. Rodent models of high-fat or high-sugar diet exposure have revealed regional brain volume reductions in the hippocampus, frontal cortex, and cerebellum, often accompanied by neuroinflammatory and cognitive changes [23]; [24]. These models are especially useful for investigating the long-term neuroanatomical consequences of diet-induced obesity and metabolic syndrome under controlled conditions. However, only very few studies have longitudinally assessed how early-life exposure to obesogenic diets affects brain structural development and whether such alterations persist or progress into adulthood.

To address this gap, our study employs structural MRI and VBM analysis to characterize volumetric changes following early-life HFHF diet exposure in a commercially available mouse model. By assessing both immediate (14 weeks old, 8-week HFHF diet) and longer term (6 months follow up) effects, we aim to investigate how early dietary environments change cortical and subcortical brain structures and whether these alterations persistent to young adulthood. Understanding these relationships will advance knowledge of the neurodevelopmental risks associated with childhood obesity and may inform development of targeted strategies to mitigate its impact on brain health.

2. Materials and methods

2.1. Animals

C57BL/6NTac male mice were purchased from Taconic Bioscience (La Jolla, CA, USA). A total of 20 obese and 15 control C57BL/6NTac male mice were included in this study. All mice were maintained on a 12/12-h light/dark cycle with bedding and cage enrichment, free access to food and water, and in a temperature-controlled environment (22 °C). Obese mice were fed a HFHF obesogenic diet (Research Diet #D09100310, containing 40 % kcal from fat, 22 % kcal from fructose, and 2 % cholesterol) starting at 6 weeks of age and housed at reduced density. Control mice were housed in the same location, also at reduced density, and fed the NIH-31 M chow diet (5 % kcal from fat).

After 8 weeks on their respective diets (14 weeks of age), all mice were shipped, and group housed (5 per cage) at the University of Maryland School of Medicine in an Association for Assessment and Accreditation of Laboratory Animal Care International-accredited facility. Obese mice continued the research diet, while control mice were fed the regular chow diet (LabDiet #5053, 5 % dietary fat, 3.42 kcal/g). All research procedures were conducted in compliance with National Institutes of Health guidelines for animal care and were approved by the Institutional Animal Care and Use Committee at the University of Maryland School of Medicine (Protocol #00822007). Body weights of all mice were measured and recorded upon arrival (Day 0), on Days 10, 20, 30, 40, 50, and on the day of the MRI experiment.

2.2. MRI data acquisition

All mice were subjected to MRI brain scans at age of 14 weeks (baseline) and 6-month follow up (post-6month). A total of 5 mice were scanned per day, in an interleaved manner between control and obese groups, to prevent batch effects. MR scans were conducted using a horizontal bore Bruker Biospec 7T 70/30 MR Scanner (Bruker Biospin MRI GmbH, Germany) and a Bruker Paravision 6.0 console. A Bruker 72 mm linear-volume coil and a Bruker ¹H four-element surface coil array served as transmitter and receiver. Each mouse received an intraperitoneal injection of dexmedetomidine (0.03 mg/kg) immediately before imaging, and anesthesia was maintained with 0.3 % isoflurane. Animal respiration rate and body temperature were monitored using a MR-compatible small-animal monitoring and gating system (SA Instruments, Inc., New York, USA). The animals’ body temperature was maintained at 37–38.5 °C by circulating warm water in a bath.

Structural T2-weighted (T2w) images covering the whole brain were obtained using a 2D Rapid Acquisition with Relaxation Enhancement (RARE) pulse sequence along the coronal direction. Imaging parameters were as follows: repetition Time (TR) = 2500 ms, echo time (TE) = 30 ms, number of averages = 6, echo train length = 4, percent phase field of view = 100 %. The acquisition matrix was 112 x 112 (112 phase-encoding steps). Images were reconstructed by the scanner to a 150 x 150 display matrix through zero-filling interpolation, which preserved the effective resolution but provided smoother images. This yielded an in-plane voxel size of 0.12 × 0.12 mm² and a field of view of 18 × 18 mm². Slice thickness was 0.5 mm with no gap; and total number of slices = 26. The total acquisition time was 7 min.

2.3. MRI data processing

All T2-weighted (T2w) images were preprocessed using a standardized pipeline to ensure accurate morphological comparisons across subjects. Initially, bias field correction was applied to each T2w image to minimize low-frequency intensity nonuniformities caused by magnetic field inhomogeneities, thereby enhancing tissue contrast [25]. Following this, brain extraction was performed to remove non-brain tissues such as the skull, scalp, and dura, improving the accuracy of downstream segmentation. Tissue segmentation was then carried out using the active contour (snake) algorithm implemented in ITK-SNAP [26], which allows semi-automated delineation of tissue classes based on image intensity and edge information. Segmentation included classification into gray matter (GM), white matter (WM), and cerebrospinal fluid (CSF). All brain extraction and tissue segmentation outputs were visually inspected, and manual corrections were performed when necessary to ensure anatomical accuracy.

A study-specific template was generated using the ANTs template-building pipeline (buildtemplateparallel.sh) by iteratively aligning the bias-corrected, brain-extracted T2w images from the control group at the baseline time point. Images from the control baseline group were specifically selected to construct the template in order to represent a healthy, unbiased anatomical reference, minimizing potential confounding effects from pathological or longitudinal changes associated with the obese condition.

Subsequently, each individual preprocessed T2w image was spatially normalized and nonlinearly registered to the study-specific template using the Symmetric Normalization (SyN) algorithm implemented in ANTs [27], which provides high-accuracy diffeomorphic registration. The resulting deformation fields were applied to warp each subject’s image into the common template space, enabling ROI-based and voxel-wise comparisons of brain morphology across subjects within the VBM framework.

To improve anatomical registration and enable more reliable voxel-wise comparisons, high-resolution isotropic T2w images (0.12 x 0.12 × 0.12 mm³) were generated from the original anisotropic data (0.12 x 0.12 × 0.5 mm³) using ECLARE, an advanced super-resolution reconstruction method developed by Prince’s group at Johns Hopkins University [28,29]. The original T2w images were limited by large slice thickness and anisotropic voxel dimensions, constrained the quality of downstream analyses and standard interpolation methods were insufficient to address these limitations. ECLARE was chosen for its proven ability to enhance resolution in anisotropic datasets, thereby improving the structural images quality. The same processing procedures applied to the original data were applied to the ECLARE-derived isotropic T2w images. Fig. 1 displays the study-specific brain templates generated from both the original and high-resolution datasets. The high-resolution template provided enhanced anatomical detail, particularly in the horizontal and sagittal views, facilitating improved visualization and delineation of brain structures. All subsequent processing steps and analyses were performed using the high-resolution isotropic T2-weighted images to enable more accurate volumetric comparisons.

Fig. 1.

Template brain images generated with and without ECLARE interpolation. (A) Template constructed from images before interpolation (anisotropic resolution: 0.12 x 0.12 × 0.5 mm³; (B) Template constructed from images after interpolation using ECLARE (isotropic resolution: 0.12 x 0.12 × 0.12 mm³. ECLARE stands for Efficient Cross-planar Learning for Anisotropic Resolution Enhancement.

2.4. ROI-based and voxel-wise volumetric estimation

We performed volumetric analysis at both the region-of-interest (ROI) level and voxel-wise to assess structural differences cross-sectionally and longitudinally. Deformation fields obtained from nonlinear registration to the study-specific template were used to compute Jacobian determinant maps (J-maps), which quantify local volume changes relative to the template. A Jacobian value (J) greater than 1 indicates local expansion, a value less than 1 indicates shrinkage, and a value equal to 1 reflects no change. To restrict the analysis to gray matter (GM), the J-maps were modulated using a binary GM segmentation of the template, thereby masking out non-GM regions. The natural logarithm of the modulated Jacobian maps (logJ maps) was then computed to linearize the distribution and stabilize variance across voxels. Both the modulated J-maps and logJ maps were spatially smoothed using a Gaussian kernel with full-width at half-maximum (FWHM) equal to twice the voxel size to improve signal-to-noise ratio and account for inter-subject anatomical variability.

ROI-based volumetric measurements were obtained by averaging the modulated J values of all voxels within each ROI. The ROIs included binary masks for the whole brain, neocortex and its subdivisions (anterior/frontal, middle, and posterior), as well as the hippocampus and its subdivisions (anterior-dorsal, posterior-dorsal, and posterior-ventral). Except for the whole-brain mask, all ROIs were manually delineated on the template brain using ITK-SNAP.

2.5. Statistical analysis

Body weight data were analyzed using a Linear Mixed Model (LMM) to account for repeated measurements on individual mouse. The model included Group (Control vs. Obese), Time, and their interaction (Group × Time) as fixed effects. Following the LMM, a post-hoc analysis was conducted to perform pairwise comparisons between the groups at each time point. To correct multiple comparisons and reduce the risk of Type I error, a Bonferroni correction was applied to the p-values. Statistical significance was set at p < 0.05.

Volumes of brain regions of interest (ROIs) were analyzed using a two-way repeated-measures ANOVA with Group (Control vs. Obese) as the between-subject factor and Timepoint (baseline vs. 6-month follow-up) as the within-subject factor. Partial eta-squared (η²) was reported as a measure of effect size for main effects and interactions. Significant interactions were followed by post-hoc analyses to clarify the direction and magnitude of differences. Post-hoc comparisons were conducted to investigate: (1) between-group differences at each timepoint, using independent t-tests with Cohen’s d as the effect size, and (2) within-group longitudinal changes across timepoints, using paired t-tests with Cohen’s d calculated manually. All post-hoc p-values were not adjusted for multiple comparisons. Statistical significance was set at p < 0.05.

For voxel-wise statistical analysis, smoothed and modulated logJ maps were used. FSL’s randomise tool was employed to perform permutation-based non-parametric testing with threshold-free cluster enhancement (TFCE), providing robust inference without relying on parametric assumptions. Group comparisons were assessed both cross-sectionally and longitudinally, and statistical significance was corrected for multiple comparisons using family-wise error (FWE) correction. A threshold of TFCE FWE-corrected p-value <0.05 was used to determine significance.

3. Results

3.1. Body weights

At the baseline of measurements (Day 0 upon arrival at the UMB, after 8 weeks on an obesogenic diet high in fat and fructose), obese mice showed noticeable differences in appearance compared to chow-fed controls. The most prominent change was oily or greasy fur, consistent with observations in other high-fat diet models. Truncal obesity was also evident, as shown in Fig. 2A and B.

Fig. 2.

Body weight progression and representative appearance in control and diet-induced obese mice. (A) Representative photograph of a chow-fed control mouse. (B) Representative photograph of an obese mouse, characterized by a high-fat, high-fructose diet. (C) Body weight trajectories for control and obese cohorts measured over the study duration. Control group (N = 15); Obese group (N = 20).

Analysis of body weight using a mixed linear model revealed significant main effects of group and time, as well as a group × time interaction (Table 1 and Fig. 2C). At baseline, Obese animals already had higher body weight than Controls (Group effect: β = 3.614, p < 0.001). Body weight increased significantly over time across all animals (Time effect: β = 0.096 per day, p < 0.001). Importantly, the interaction term was significant (β = 0.015, p < 0.001), indicating that Obese animals gained weight at a faster rate than Controls.

Table 1.

Statistic analysis results for body-weight.

Mixed Linear Model
Variable	Coefficient	Std. Error	z-statistic	P-value
Intercept	28.754	0.711	40.423	<0.001***
Group[T.Obese]	3.614	0.941	3.841	<0.001***
Time	0.096	0.006	17.277	<0.001***
Group[T.Obese]:Time	0.015	0.007	2.025	<0.001***

Post-hoc Analysis
Time (days)	t-statistic	p-unc	p-corrected	Significance
0	3.841	<0.001***	<0.001***	Yes
10	4.107	<0.001***	<0.001***	Yes
20	4.361	<0.001***	<0.001***	Yes
30	4.596	<0.001***	<0.001***	Yes
40	4.806	<0.001***	<0.001***	Yes
50	4.984	<0.001***	<0.001***	Yes
180	4.799	<0.001***	<0.001***	Yes

Note: Linear Mixed model with a Bonferroni correction to account for multiple comparisons.

Post-hoc comparisons confirmed consistent and significant between-group differences across all time points examined (days 0–180, all p < 0.001 after Bonferroni correction), with Obese animals remaining heavier than Controls throughout the study.

3.2. ROI-based volumetric changes: cross-sectional and longitudinal

3.2.1. Whole brain volume

The Repeated Measures ANOVA revealed a significant main effect of time (p = 0.023, η² = 0.146) on whole brain volume, indicating a significant change over the 6-month period (Tables 2 and 3 and Fig. 3).

Table 2.

Repeated Measurement ANOVA results for regional brain volume.

ROI	Group p (η2)	Time p (η2)	Interaction p (η2)	Between-group p (baseline/6mo, d)	Within-group p (Control/Obese, d)
Wholebrain	0.201 (0.058)	0.036* (0.148)	0.041* (0.141)	0.967/0.588 (0.01/0.20)	0.008*/0.014* (−0.83/−0.83)
Neocortex	0.009* (0.222)	<0.001* (0.696)	0.696 (0.006)	0.572/0.065 (0.13/0.70)	< 0.001*/< 0.001* (1.42/1.52)
Anterior	0.901 (0.001)	0.714 (0.005)	0.818 (0.002)	0.467/0.805 (0.09/0.03)	0.167
Middle	0.039* (0.144)	<0.001* (0.610)	0.988 (<0.001)	0.461/0.177 (0.22/0.51)	< 0.001*/< 0.001* (1.08/1.39)
Posterior	0.005* (0.246)	<0.001* (0.776)	0.350 (0.031)	0.467/0.004* (0.21/1.16)	< 0.001*/< 0.001* (1.81/1.80)
Hippocampus	0.098 (0.095)	0.015* (0.193)	0.409 (0.024)	0.557/0.069 (0.22/0.69)	0.020*/0.267 (−0.67/−0.30)
AnteriorDorsal	0.561 (0.012)	0.001* (0.338)	0.363 (0.030)	0.327/0.945 (0.36/0.03)	0.009*/0.039* (−0.78/−0.59)
PosteriorDorsal	0.148 (0.073)	0.328 (0.034)	0.198 (0.058)	0.116/0.045* (0.83/0.77)	0.827/0.058 (−0.43/0.06)
PosteriorVentral	0.005* (0.249)	0.092 (0.098)	0.543 (0.013)	0.287/0.049* (0.19/0.75)	0.286/0.192 (−0.29/−0.35)

Notes: η² in parentheses next to main p-values is the cohen’s d of effect size.

Between-group post-hocs: baseline/6mo with Cohen’s d in parentheses.

Within-group post-hocs: Control/NASH with Cohen’s d in parentheses.

Table 3.

Repeated measures ANOVA results for brain subregion volumes.

ROI	Group p (η2)	Time p (η2)	Int. p (η2)	BG Base p (d)	BG 6mo p (d)	WG Ctrl p (d)	WG Obese p (d)
Whole brain	0.339 (0.028)	0.023 (0.146) *	0.054 (0.108)	0.026 (0.79) *	0.508 (0.23)	0.967 (0.01)	0.004 (0.74) **
Neocortex	0.054 (0.108)	0.000(0.722) ***	0.732 (0.004)	0.297 (0.40)	0.159 (0.50)	0.000(−1.42) ****	0.000(−1.73) ****
Ant. Neocortex	0.975 (0.000)	0.908 (0.000)	0.974 (0.000)	0.963 (0.02)	0.998 (0.00)	0.930 (0.02)	0.941 (0.02)
Mid. Neocortex	0.158 (0.060)	0.000 (0.631) ***	0.938 (0.000)	0.440 (0.28)	0.338 (0.35)	0.001 (−1.08) ***	0.000 (−1.51) ***
Post. Neocortex	0.026 (0.141) *	0.000 (0.781) ***	0.388 (0.023)	0.352 (0.33)	0.012 (0.89) *	0.000 (−1.81) ****	0.000 (−1.87) ****
Hippocampus	0.027 (0.139) *	0.003 (0.235) ***	0.672 (0.005)	0.239 (0.41)	0.051 (0.72) *	0.020 (0.67) *	0.058 (0.45)
Ant.-dors. Hipp.	0.952 (0.000)	0.000 (0.370) ***	0.520 (0.013)	0.682 (0.14)	0.725 (0.12)	0.009 (0.78) **	0.005 (0.72) **
Post.-dors. Hipp.	0.076 (0.092)	0.260 (0.038)	0.298 (0.033)	0.643 (0.16)	0.039 (0.77) *	0.116 (0.43)	0.871 (0.04)
Post.-vent. Hipp.	0.001 (0.268) **	0.016 (0.163) *	0.275 (0.036)	0.015 (0.87) *	0.081 (0.59)	0.286 (0.29)	0.034 (0.51) *

Note: ROI = Region of Interest; Ant. = Anterior; Mid. = Middle; Post. = Posterior; dors. = dorsal; vent. = ventral; Hipp. = Hippocampus; Int. = Interaction; BG = Between-group; Base = Baseline; 6mo = 6-month; WG = Within-group; Ctrl = Control; η² = eta squared; d = Cohen’s d. Significance:

^*p < 0.05,

^**p < 0.01,

^***p < 0.001,

^****p < 0.0001.

Bold = significant.

Fig. 3.

Whole brain volume changes in control and obese mice: cross-sectional and longitudinal assessments. (A) Representative whole brain mask used for volume estimation. (B) Whole brain volume changes, showing both cross-sectional differences between control and obesity groups, and longitudinal changes from baseline to post-6 months. Statistical significance is indicated by asterisks *. Group sizes: Control (N = 15); obese (N = 20).

Although the group × time interaction was not statistically significant (p = 0.054), post-hoc analysis revealed a notable between-group difference at baseline, where obese mice exhibited a smaller whole brain volume compared to controls mice (control: 465.31 ± 3.04 mm³; obese: 462.78 ± 3.34 mm³; p = 0.026, d = 0.79) following 8 weeks on a obesogenic diet. By the 6-month follow-up, no significant difference in whole brain volume was observed (control: 465.37 ± 3.87 mm³; obese: 466.24 ± 3.78 mm³; p = 0.508), indicating partial catch-up in the obese group.

Longitudinally, the obese group showed a significant increase in whole brain volume over time (ΔV = 3.46 mm³, p = 0.004, d = 0.74), whereas the control group remained stable (p = 0.967). These results suggest that early diet-induced reduction in whole brain volume in obese mice may be reversible over the study period.

3.2.2. Neocortex and subdivisions

Total neocortical volume decreased significantly over the 6 months in all animals (p < 0.001, η² = 0.722, Tables 2 and 3 and Fig. 4), with no significant difference between-group at either baseline (p = 0.297) or 6 months (p = 0.159). Post-hoc longitudinal testing confirmed comparable longitudinal volume loss in both control (ΔV = −3.37 mm³, p < 0.001, d = −1.42) and obese (ΔV = −3.13 mm³, p < 0.001, d = −1.73) animals, reflecting an overall age- or time-related decline in neocortical volume independent of diet.

Fig. 4.

Regional volume changes in neocortex and its subdivisions in control and obese mice: Cross-sectional and longitudinal assessments. (A) Neocortex subdivision masks overlapped on T2 brain template image, indicating anatomical regions used for regional volume estimation. (B) Regional volume changes, showing both cross-sectional differences between control and obesity groups, and longitudinal changes from baseline to post-6 months. Statistical significance is indicated by asterisks: ***p < 0.001; **p < 0.01; *p < 0.1. Group sizes: Control (N = 15); obese (N = 20).

Anterior neocortex volume showed no significant changes over time (baseline: p = 0.963, 6months: p = 0.998) or between groups (control: p = 0.930; obese: p = 0.941), indicating stability of this region throughout the study.

Middle neocortex volume decreased significantly over time in both groups, (p < 0.001, η² = 0.631), with similar changes for control (ΔV = −1.97 mm³, p < 0.001, d = −1.08) and obese groups (ΔV = −1.93 mm³, p < 0.001, d = −1.51), but no between-group differences at any time point.

Posterior neocortex volume showed both significant time (p < 0.001, η² = 0.781) and group (p = 0.026, η² = 0.141) effects Notably, obese mice had smaller posterior neocortex volume than controls at 6 months (p = 0.012, d = 0.89), although baseline volumes were similar (p = 0.352). Both groups showed significant longitudinal volume decreases (Control: ΔV = −1.42 mm³, p < 0.001, d = −1.81; Obese: ΔV = −1.21 mm³, p < 0.001, d = −1.87), suggesting posterior neocortex is particularly sensitive to both age and dietary effects.

3.2.3. Hippocampus and subdivisions

Total hippocampal volume showed significant main effects of both group (p =0.027, η² = 0.139) and time (p = 0.003, η² = 0.235) (Tables 2 and 3 and Fig. 5). While there were no significant between-group differences at baseline (p = 0.239) or 6 months (p = 0.051), longitudinal analysis revealed a modest but significant volume increase in control group (ΔV = 0.39 mm³, p = 0.020, d = 0.67), with the obese group showing a non-significant trend (p = 0.058).

Fig. 5.

Regional volume changes in Hippocampus and its subdivisions in control and obese mice: Cross-sectional and longitudinal assessments. (A) Hippocampus subdivision masks overlapped on T2 brain template image, indicating anatomical regions used for regional volume estimation. (B) Regional volume changes, showing both cross-sectional differences between control and obese groups, and longitudinal changes from baseline to post-6 months. Statistical significance is indicated by asterisks: ***p < 0.001; **p < 0.01; *p < 0.1. Group sizes: Control (N = 15); obese (N = 20).

Anterior-dorsal hippocampus showed increased volume over time in both groups (Control: ΔV = 0.18 mm³, p = 0.009, d = 0.78; obese: ΔV = 0.13 mm³, p = 0.005, d = 0.72), with no group difference, indicating a consistent growth pattern.

Posterior-dorsal hippocampus showed significant smaller volume than controls at 6months (p = 0.039, d = 0.77), while no difference at baseline (p = 0.643). Neither group showed significant longitudinal changes.

Posterior-ventral hippocampus showed significant group differences at baseline (p = 0.015, d = 0.87) but not at the 6 months (p = 0.081). Longitudinally, the obese group exhibited a modest but significant increase (ΔV = 0.15 mm³, p = 0.034, d = 0.51), whereas the controls did not (p = 0.286).

3.3. Voxel-wise volumetric changes: cross-sectional and longitudinal

A voxel-wise volumetric analysis was performed using a two-way repeated measures ANOVA to identify significant localized brain volume changes and differences between groups. The analysis yielded several significant clusters, which are reported in detail below and summarized in Table 4 and Figs. 6 and 7. For this report, only clusters with a size of 100 voxels or greater were considered for discussion.

Table 4.

Two-way ANOVA results for voxel-wise volume differences between groups and volume changes over time.

Tests	Clust. Idx	Voxels	Vol (mm3)	Peak-p	Peak-p X	Peak-p Y	Peak-p Z	COG X	COG Y	COG Z	Anat. Region (Peak-p)
Main effect of time	1	5000	8.64	<0.001	−2.09	−1.11	−1.23	−0.269	−0.724	−0.00758	lateral thalamus
	2	2353	4.07	<0.001	0.674	−1.23	−2.91	−0.0306	−0.291	−2.53	Midbrain Reticular Nucleus
	3	2310	3.99	0.005	2.35	1.05	2.61	2.27	1.56	2.79	primary somatosensory cortex
	4	479	0.83	<0.001	1.51	−1.35	1.05	1.77	−0.877	1.32	caudate putamen
	5	356	0.62	0.016	2.83	1.17	−2.43	3.22	0.669	−1.72	Hippocampus dorsal
	6	314	0.54	<0.001	2.23	−1.11	0.0876	2.55	−1.3	−0.0832	Hippocampus ventral
	7	149	0.26	0.006	−0.406	0.0914	2.01	−0.154	0.297	2.15	septum
6mo-baseline (Control)	1	863	1.49	0.006	−2.09	−1.23	−0.392	−1.73	−1.21	0.279	lateral thalamus (L)
	2	522	0.90	0.008	−0.646	−0.749	−2.67	−0.966	−0.68	−2.4	Midbrain Reticular Nucleus (L)
	3	291	0.50	0.007	0.314	−0.629	−2.55	0.379	−0.673	−2.51	Midbrain Reticular Nucleus (R)
	4	190	0.33	0.008	1.87	−1.11	−0.512	1.78	−0.96	−0.143	lateral thalamus (R)
6mo-baseline (Obese)	1	1056	1.82	<0.001	−2.09	−1.23	−0.752	−1.79	−1.08	−0.139	lateral thalamus (L)
	2	1012	1.75	0.001	1.75	−1.11	−0.632	1.4	−1.1	−0.168	lateral thalamus (R)
	3	328	0.57	0.004	−0.766	−0.869	−2.55	−0.859	−0.534	−2.6	Midbrain Reticular Nucleus (L)
	4	300	0.52	0.001	0.674	−1.23	−2.91	0.544	−0.598	−2.45	Midbrain Reticular Nucleus (R)
	5	264	0.46	0.004	1.63	−0.869	1.05	1.69	−1.02	1.29	caudate (R)
	6	168	0.29	0.001	2.23	−1.23	0.0876	2.46	−1.34	−0.0371	Hippocampus (R)
Obese-Control (baseline)	1	1047	1.81	0.02	−3.05	2.25	−1.95	−3.83	1.76	−1.81	Primary visual cortex (L)
	2	812	1.40	0.03	−4.37	1.77	0.688	−3.58	1.76	1.5	primary somatosensory cortex (L)
	3	96	0.17	0.05	3.07	1.29	2.01	3.06	1.38	2.11	primary somatosensory cortex (R)
Obese-Control (6mo)	1	152	0.26	0.06	1.63	2.97	−0.752	1.44	2.98	−1.02	primary visual cortex (R)

Note: Clust. Idx = Cluster Index; Vol = Volume; Anat. = Anatomical; COG = Center of Gravity; 6mo = 6-month; L = left; R = right. All coordinates in millimeters (mm). Peak-p refers to the peak p-value location within each cluster.

Fig. 6.

Group Volumetric Difference maps between obesity and controls at Baseline and 6 Months. (A) Baseline comparison. (B) Post-6-month comparison. T-contrast maps with significant clusters were overlaid on the T2-template brain image. Blue color indicates voxels with significantly smaller volume in the obese group compared to controls. Statistical significance was defined as Family-wise Error (FWE) corrected p < 0.01. Group sizes: Control (N = 15); Obese (N = 20). (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)

Fig. 7.

Longitudinal Volumetric Difference maps. (A) Overall longitudinal volumetric changes across all mice. (B) Longitudinal volumetric changes within the control group. (C) Longitudinal volumetric changes within the obese group. T-contrast maps with significant clusters were overlaid on the T2-template brain image. Color blue indicates voxels with significantly reduced volume at Post-6month time point compared with baseline. Statistical significance was defined as Family-wise Error (FWE) corrected p < 0.05. Group sizes: Control (N = 15); Obese (N = 20). (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)

3.3.1. Cross-sectional

As shown in Table 4, we did not observe main effect of group and group × time interaction. Post-hoc analysis showed group difference at both baseline and 6month time points (Fig. 6). The analysis of the baseline data revealed several regions with significant volume differences between the Obese and Control groups (Fig. 6A). A significant cluster was found in the left Primary Visual Cortex (Cluster 1, volume = 1.81 mm³, p = 0.02), where the obese group showed a smaller volume compared to controls. A second significant cluster was identified in the left primary somatosensory cortex (Cluster 2, volume = 1.40 mm³, p = 0.03), where the obese group had a smaller volume. A smaller cluster in the right primary somatosensory cortex (Cluster 3, volume = 0.17 mm³, p = 0.05) also showed a smaller volume in the obese group.

At the 6-month time point, only one cluster showed a trend towards a significant difference between the two groups (Fig. 6B). This cluster in the right primary visual cortex (Cluster 1, volume = 0.26 mm³, p = 0.06) showed a smaller volume in the obese group, but this finding did not meet the statistical significance threshold of p < 0.05.

3.3.2. Longitudinal

The main effect of time revealed multiple clusters with significant volume change across both groups, regardless of group assignment (Fig. 7A). A large significant cluster was found in the lateral thalamus (Cluster 1, volume = 8.64 mm³, p < 0.001), indicating a widespread change over time. Other significant clusters were observed in the Midbrain Reticular Nucleus (Cluster 2, volume = 4.07 mm³, p < 0.001), primary somatosensory cortex (Cluster 3, volume = 3.99 mm³, p = 0.005), and caudate putamen (Cluster 4, volume = 0.83 mm³, p < 0.001). Smaller but significant clusters were also identified in the dorsal hippocampus (Cluster 5, volume = 0.62 mm³, p = 0.016) and ventral hippocampus (Cluster 6, volume = 0.54 mm³, p < 0.001).

The longitudinal post-hoc analysis for the Control group showed a significant volume decrease in several regions (Fig. 7B). The most significant finding was a cluster in the lateral thalamus (Cluster 1, volume = 1.49 mm³, p = 0.006). Significant volume loss was also found in two clusters within the Midbrain Reticular Nucleus (Cluster 2, volume = 0.90 mm³, p = 0.008; Cluster 3, volume = 0.50 mm³, p = 0.007).

The longitudinal post-hoc analysis for the Obese group also revealed significant clusters of volume decrease (Fig. 7C). Two significant clusters of volume loss were found in the lateral thalamus, one on the left (Cluster 1, volume = 1.82 mm³, p < 0.001) and one on the right (Cluster 2, volume = 1.75 mm³, p = 0.001). Significant volume loss was also found in two clusters within the Midbrain Reticular Nucleus (Cluster 3, volume = 0.57 mm³, p = 0.004; Cluster 4, volume = 0.52 mm³, p = 0.001). Additional significant clusters were identified in the caudate (Cluster 5, volume = 0.46 mm³, p = 0.004) and hippocampus (Cluster 6, volume = 0.29 mm³, p = 0.001).

3.4. Correlation between whole brain volume and body weight

To assess the relationship between whole brain volume and body weight, Pearson correlation analyses were performed across all groups and time points (Fig. 8). Overall, no significant correlation was observed between whole brain volume and body weight (r = 0.118, p = 0.33). When analyzing the groups and time points separately, the control group showed no significant correlation between whole brain volume and body weight at baseline (r = 0.080, p = 0.778) or at the 6-month time point (r = 0.099, p = 0.725). In contrast, the obese group exhibited a statistically significant positive correlation between whole brain volume and body weight at the 6-month time point (r = 0.502, p = 0.0241), although no significant correlation was found at baseline (r = −0.222, p = 0.348).

Fig. 8.

Correlation between whole brain volume and body weight within groups at baseline and 6month follow-up. (left) correlation at baseline; (right) correlation at post-6month.

4. Discussion

This study employed both cross-sectional and longitudinal MRI to investigate brain volumetric alterations in mice subjected to an obesogenic high-fat, high-fructose (HFHF) diet during a critical developmental window. Our findings reveal dynamic and heterogeneous changes in brain structure, reflecting both age-related volume decrease and alterations unique to the obese condition.

4.1. Whole brain volumetric dynamics

The initial finding of significantly smaller whole-brain volumes in obese mice at 14 weeks suggests an early developmental impact of the HFHF diet. This aligns with clinical neuroimaging studies showing that children with obesity often exhibit reduced global and regional brain volumes, including in gray matter-rich regions involved in higher-order cognition [13,21]. In our study, this early deficit was no longer evident at 6 months, as obese mice showed a significant longitudinal increase in brain volume: a pattern absent in control animals. This atypical volumetric expansion likely reflects a adaptive response, potentially driven by chronic neuroinflammation, gliosis, or vascular changes such as blood-brain barrier disruption and tissue edema, as observed in models of metabolic dysfunction [30–32].

Although the increased brain volume might appear compensatory, it is unlikely to reflect normative growth. Instead, it may indicate sustained metabolic stress triggering non-neuronal tissue changes, masking underlying structural deficits. These findings parallel reports in human studies, where abnormal brain maturation trajectories, including both reduced cortical thinning and altered white matter development, are linked to excess weight and systemic inflammation [22,33].

4.2. Regional volumetric patterns: Neocortex and subdivisions

At the regional level, the neocortex exhibited distinct longitudinal patterns, indicating differential vulnerability to aging and obesity-related stress. The frontal neocortex, preserved in both groups, may reflect resilience to early metabolic disruption, consistent with delayed prefrontal cortex maturation in children, which may render it less sensitive to early insults but vulnerable later [34]. Conversely, the mid-neocortex showed a consistent decline in volume across groups, likely representing normative age-related decrease.

The posterior neocortex, however, exhibited a unique pattern: obese mice showed greater volume at 6 months and less longitudinal decrease than controls. This echoes published findings that obesity can alter posterior cortical structures involved in sensory integration and reward processing [35]. The regional specificity could be explained by higher metabolic demands in posterior areas or differential susceptibility to obesity-induced neuroinflammation and altered vascular supply.

4.3. Hippocampus and subdivisions: A window into cognitive vulnerability

Hippocampal changes in obese mice also mirror findings in children with obesity, who frequently show reduced hippocampal volumes and impaired memory performance [36]. Our cross-sectional results revealed smaller posterior hippocampal volumes in obese mice at both time points, suggesting persistent structural deficits. Voxel-wise analyses further revealed localized decrease in hippocampal subregions and the substantia innominata—critical hubs for memory and attention—despite ROI-based findings of gross volume increases. These findings suggest that neuronal changes may require future experiments that utilize imaging technologies capable of capturing alterations at multiple spatial resolutions.

The hippocampus’s prolonged developmental trajectory [37] and high metabolic sensitivity [38] render it especially vulnerable to early-life nutritional insults. These findings reinforce the relevance of early dietary environments in shaping long-term cognitive function, with obesity-related disruptions potentially predisposing individuals to neurodevelopmental and neurodegenerative risks later in life.

4.4. Voxel-wise analysis: differentiating aging from obesity effects

Voxel-wise analyses across the whole brain provided a more nuanced view, helping to distinguish between normative aging and obesity-specific effects. While both control and obese mice displayed widespread longitudinal volume decreases in regions such as the thalamus, midbrain, and pons, likely reflecting normal aging, obese mice exhibited unique and localized reductions in key basal forebrain and hippocampal structures. These alterations may underlie functional deficits observed in obese individuals and animal models, particularly in domains of memory, attention, and reward sensitivity.

Notably, early gray matter reductions in sensory cortices (e.g., primary somatosensory and visual cortex) in obese mice resemble human studies where sensory processing regions are impacted by excess adiposity and low-grade inflammation [22]. Although some of these deficits resolved over time, their initial presence suggests transient disruption during critical windows of cortical maturation.

4.5. Overall implications and translational relevance

Together, our findings demonstrate that exposure to an obesogenic diet during early life induces complex, region-specific, and temporally dynamic brain changes. Our observation of reduced whole-brain volumes in diet-induced obese mice at 14 weeks aligns with findings of an almost 15 % decrease in brain weights of a genetic model of obesity [39], suggesting a shared neuroanatomical phenotype irrespective of the underlying cause excessive weight gain. Beyond alterations to the whole brain, reduced cortical volumes were also seen in mice fed high-fat diets, where the cerebral and somatosensory cortices display significant volume reductions, especially in male mice. [40]. While the decrease in brain volumes in these published studies is consistent with changes seen in our work, previous studies did not report a compensatory increase in total brain volume following extended obesogenic diets. This difference may be related to variations in diet composition. Notably, our control chow diet contained 5 % kcal from fat, compared to 18 % used by Patel et al. These findings highlight the importance of dietary composition in shaping neuroanatomical outcomes in obesity models. Nevertheless, the combination of early deficits, abnormal volumetric expansion, and persistent decrease in select regions supports the notion that childhood obesity can lead to long-lasting disruptions in brain development. These structural abnormalities may not be fully reversible, even with time, and could serve as early biomarkers of neurodevelopmental risk.

By modeling these effects in mice using longitudinal neuroimaging, we provide preclinical evidence that complements human studies and offers insight into how early metabolic stress alters brain trajectories. These findings underscore the importance of early intervention in childhood obesity, not only for physical health but also for preserving optimal brain development.

Limitations

Despite these significant findings, this study has several limitations. The use of a mouse model, while allowing for controlled dietary intervention, may not fully recapitulate the complexity of human obesity and associated brain pathologies. This commercial model also required shipping to our imaging facility, potentially inducing stress-related responses. Our study also only utilized male mice, preventing broader interpretation of our findings. Furthermore, while the HFHF diets resulted in significant body weight changes, the magnitude of these changes became smaller over time. The mechanism driving this observation remains unknown, although variability in diet-induced weight gain is mice is well documented [41]. Food intake in individual mice represents a significant predictor of body weight variability [42], so future studies involving diet-induced mouse models of obesity should incorporate measures to quantify food intake. Our MRI approach quantifies gross volumetric changes; however, it does not directly differentiate between changes in neuronal cell bodies, glial cells, vascular components, or extracellular fluid. Future studies should incorporate histological validation, such as neuronal counts, glial activation markers, and assessment of microvascular integrity, to elucidate the cellular basis of the observed volumetric changes. Furthermore, the absence of behavioral or cognitive assessments limits our ability to directly correlate volumetric changes with functional deficits in this model. Future research should integrate behavioral phenotyping to establish direct links between structural brain alterations and cognitive performance. The 6-month follow-up period captures subacute to chronic effects, but longer-term studies would be valuable to track disease progression and brain changes over extended periods.

Appendix A. Supplementary data

Supplementary data to this article can be found online at https://doi.org/10.1016/j.bbrc.2025.152774.