Experimental design and weight changes.

(A) For the first two months, all mice received a standard diet. From two to four months of age, mice were put on either a low-fat or a high-fat diet. From four to six months of age, mice were placed in intervention groups. T1w MRIs were collected at two, four, and six months. Behavioural tests (MWM and NOR) were performed after the last MRI scan. (B) Weekly measurements of weight plotted per group. The vertical dotted line indicates the onset of interventions. The x-axis denotes time in weeks, and the y-axis body weight in grams (g). Before interventions, trajectories reflect the weight of mice maintained on a low-fat diet (black) or a high-fat diet (red). After the onset of interventions (week ≥ 8), trajectories are shown separately for each intervention group (see legend). Significant weight differences (p < 0.05) compared against Control are denoted with an asterisk (*). (C) Model-estimated rates of weight change (g/week), stratified by intervention, genotype, and sex. Weight change was estimated using marginal contrasts derived from a four-way linear mixed effects model (Weight ∼ Week*Intervention*Genotype*Sex + (Week|ID)). Asterisks (*) indicate estimates with 95% confidence intervals excluding zero (p < 0.05), indicating a significant weight gain or loss. N = 156 mice, with 18 repeated weight measurements each.

Age-dependent group differences in brain volume change.

The t-statistic maps show the linear component of the group and age interaction, thresholded at 10% FDR and displayed at their peak t-values, overlaid on the population-average brain. Estimates were derived from a linear mixed-effects model of the log-transformed Jacobian determinant (∼ Group * poly(Age, 2) + Sex + (1 | ID)). N = 152 mice.

Trajectories of significant peak voxels identified for the genotype effect.

The figure shows peak voxels selected from the comparison between Tg_Control and WT_Control, which highlights the genotype effect on the hippocampus (A) and cerebellum (B). The left panels show the statistical maps of the interaction between the group and age in days, thresholded at 10% FDR, together with the corresponding t-values and voxel coordinates (from model: log-Jacobian ∼ Group*poly(Age,2) + Sex + (1|ID)). The middle panels show the percent change in volume from baseline in the selected peak voxels. The right panels show the percent volume change from baseline at the final time point; horizontal lines represent the group median, and individual data points are overlaid to illustrate distribution. FDR: False Discovery Rate.

Trajectories of significant peak voxels identified for the high-fat diet effect.

Significant peak voxels were selected from comparisons between WT_HighFat and WT_Control (top panels, A-B) and between Tg_HighFat and Tg_Control (bottom panel, A), highlighting genotype-specific effects of the obesity-inducing high-fat diet. Left panels: statistical maps of the interaction between group and age (from model: log-Jacobian ∼ Group*poly(Age,2) + Sex + (1|ID), 10% FDR corrected). Middle panels: longitudinal percent change in volume from baseline across all three time points. Right panels: percent volume change from baseline and median at the last time point. FDR: False Discovery Rate.

Trajectories of significant peak voxels identified for the combined interventions.

Significant peak voxels were selected from the left side of the cerebellum, from the comparison between WT_Both and WT_HighFat (top panel, A), and the left hippocampus from the comparison between Tg_Both and Tg_HighFat (bottom panel, A), highlighting effects of the combined exercise and diet intervention in each genotype. Panels show the statistical maps of the interaction between group and age (from model: log-Jacobian ∼ Group*poly(Age,2) + Sex + (1|ID), 10% FDR corrected, left panels), longitudinal percent change in volume from baseline across all three time points (middle panels), and percent volume change from baseline and the group median at the last time point (right panels). FDR: False Discovery Rate.

Intervention-associated effects across behavioural outcomes, multivariate, and gene enrichment analyses.

(A) Marginal contrasts for log-transformed Gallagher’s Cumulative Measurement (GCM), where lower GCM values indicate better learning. (B) Marginal contrasts for the number of pass-throughs, a measure of spatial memory retention; higher values indicate better performance. (C) Marginal contrasts for the percentage of time spent in the target quadrant of interest (where the platform was located during training); higher values indicate better performance. Points show contrasts comparing each intervention (HighFat, Diet, Exercise, and Both) to the Control, stratified by Genotype (WT, top; Tg, bottom) and Sex (Blue = Male, Red = Female). Error bars denote 95% confidence intervals; intervals crossing the zero line indicate non-significant differences. Tg = 3xTgAD mice. (D) shows the PLS-LV3 behavioural pattern that maximally covaries with the brain pattern (E). For the behavioural data (D), variables from the NOR (in green), the MWM (in yellow), and the treatment groups or demographics (in blue) are shown. The bars show how much each behaviour contributes to the pattern of LV3 (x-axis: loading values). Singular value decomposition estimated the size of the bars, whereas confidence intervals were estimated by bootstrapping. Bars with error bars that crossed the zero line are not significant (in gray). (E) Brain loading BSRs are overlaid on the population average. BSR thresholded to 95% confidence interval (N = 96 mice with complete data). (F) Group-wise distribution of brain scores (derived from PLS) and median, illustrates that individuals in the Diet and Exercise intervention are highly represented by the LV3 pattern. (G) Scatter plot showing significant modules identified with SGEA. Seven gene ontology (GO) terms were associated with top-ranked genes, defined as those most strongly correlated with the LV3 brain loadings, with an area under the curve > 0.5 and P < 0.05. The significant terms are: negative regulation of response to oxidative stress (GO:1902883), glucose homeostasis (GO:0042593), carbohydrate homeostasis (GO:0033500), cellular glucose homeostasis (GO:0001678), cellular response to hexose stimulus (GO:0071331), cellular response to glucose stimulus (GO:0071333), and positive regulation of fatty acid transport (GO:2000193). PLS: Partial Least Squares. SGEA: Spatial Gene Enrichment Analysis.