Addition of humanized APP to humanized APOE mouse model reduces brain size and increases the ratio of cortical representation

Abstract

INTRODUCTION:

Age, Apolipoprotein E4 (APOE4) genotype, and biological sex are major risk factors for late-onset Alzheimer’s disease (LOAD). Neuroimaging is central to its characterization, and preclinical mouse models enable controlled investigation of these factors. To date, humanized APOE4 has not recapitulated LOAD-relevant brain phenotypes. Given the central role of amyloid precursor protein (APP) in LOAD pathogenesis, incorporating humanized APP (hAPP) alongside humanized APOE (hAPOE) may therefore improve translational modeling of structural brain changes.

METHODS:

Aged mice (mean age = 23.25 months) carrying murine (m) or humanized (h) APP and either murine Apoe or hAPOE3/3, hAPOE3/4, or hAPOE4/4 underwent in-skull ex vivo volumetric MRI. Regional volumes were quantified in absolute terms and relative to total brain volume (TBV). Linear models included APP type, APOE genotype, and sex, with FDR correction applied within contrasts.

RESULTS:

Brain volumes were primarily determined by APP background, with hAPP globally reducing total and regional volumes relative to mAPP mice. Across hAPP models, hAPOE4/4 exhibited the greatest brain-wide reductions, which was mitigated by a single hAPOE3 allele. In contrast, mouse APP exerted modest effect in hAPOE, with hAPOE4 carriers exhibiting greater total volume without regional specificity. After TBV adjustment, hAPP mice exhibited subcortical vulnerability with relative cortical preservation. Females exhibited larger brain volumes than males, independent of APP or APOE genotype.

DISCUSSION:

These findings demonstrate that APP background is a primary driver of mouse brain volume, with hAPP producing global reductions amplified by the hAPOE4/4 genotype. In contrast, hAPOE4 effects in the mAPP background were modest and nonspecific, consistent with normative aging. Together, these results suggest that hAPP and hAPOE4 act synergistically, and that hAPOE4 alone is insufficient to recapitulate AD-relevant brain changes in mice. The hAPP/hAPOE4/4 model yields a brain-wide phenotype consistent with LOAD-associated volumetric patterns, whereas mAPP/hAPOE models may better reflect non-pathological aging.

1. INTRODUCTION

Sporadic, or late-onset, Alzheimer’s disease (LOAD) accounts for > 95% of AD cases and is driven by a combination of genetic and biological factors [1]. Pathological outcomes of AD include decreased brain glucose uptake, widespread and regionally selective patterns of atrophy, and reduced white matter myelination and structural coherence in AD patients relative to healthy individuals [2–4]. Regions showing atrophy early and prominently include the hippocampus and entorhinal cortex, which collectively form the medial temporal lobe system [5, 6]. These volumetric reductions are measured using magnetic resonance imaging (MRI) and have been reported in numerous studies, highlighting their ubiquity in AD patients and importance for measuring disease progression [7, 8]. These volumetric changes reflect both global brain scaling and regionally selective patterns of atrophy.

Transgenic preclinical AD mouse models used to date have been reported to replicate aspects of hallmark pathology features, including global brain atrophy, hippocampal atrophy, brain amyloidosis, white matter degradation, and inflammation [9–16]. Although these models incorporate aggressive genetic mutations found in humans, there have been no instances of AD in humans that include multiple such mutations. The magnitude of AD pathology generated in transgenic mice better reflects the pathological amplification and acceleration of familial genetic forms of AD, rather than the genetic profile and pathogenesis time course of LOAD.

Females and APOE4 carriers have the greatest clinical incidence and genetic risk for LOAD [17, 18]. To develop more clinically representative models, the MODEL-AD consortium generated humanized APOE (hAPOE) mice [19]. These mice, particularly hAPOE4/4, exhibit lower APOE protein levels, mitochondrial dysfunction, increased immune activation, demyelination, and endocrine profiles consistent with accelerated aging relative to hAPOE3/3 carriers [20–22]. Cross-sectional volumetric analyses further show age-related increases in total brain volume, regional sexual dimorphism, and modestly greater total brain volume in hAPOE4 carriers [23].

However, these changes do not culminate in LOAD-like pathology or cognitive deficits [24], suggesting that hAPOE4 alone is insufficient to recapitulate AD-relevant neuroimaging phenotypes and that additional genetic factors, including amyloid precursor protein (APP), may be necessary for translational relevance.

APP is a key contributor to amyloid-β plaque formation and clinical pathogenesis. Murine App, in contrast to human APP, does not result in plaque aggregation in the mouse brain. Existing APP-centric models are often amplified with multiple mutations to induce AD-like pathology, such as brain atrophy, axonal swelling, myelin and neuronal dysfunction, coupled with peripheral inflammation [25–27]. However, these models either utilize APP mutations leading to APP overexpression or familial AD mutations resulting in early pathology, common in early-onset AD, limiting translatability to LOAD [28–30]. Mice expressing the humanized isoform of murine APP (hAPP) do exhibit cognitive impairment and reduced long-term potentiation relative to wild-type mice [31]. However, the impact of APP expression on brain volumetric phenotypes, particularly in the absence of familial mutations, remains less well characterized.

Studies investigating the combined impact of humanized APP and APOE genotype on brain volumetric phenotypes are lacking. Addressing this gap, we investigated the impact of humanized APP on brain morphology using ex-vivo imaging in the presence of the three key LOAD risk factors, namely advanced age (23–25 months), biological sex, and humanized APOE genotype.

2. METHODS

2.1. ANIMAL USE

All procedures were performed following National Institutes of Health guidelines and approved by the University of Arizona Institutional Animal Care and Use Committee (IACUC). The following breeder pairs were obtained from Jackson Laboratory: mAPP/humanized APOE3 knock-in mice (strain #029018), mAPP/humanized APOE4 knock-in mice (strain #027894), and humanized APP/murine Apoe (strain #030898). Animals were bred in-house and maintained in our animal colony on a 14-h:10-h light: dark cycle and housed in ventilated cages. Adult male and female mAPP/hAPOE3/3, mAPP/hAPOE3/4, mAPP/hAPOE4/4, hAPP/mApoe, hAPP/hAPOE3/3, hAPP/hAPOE3/4, and hAPP/hAPOE4/4 mice (23–25 months old, C57/BL6 background) were used for these experiments. Food and water were available ad libitum. The combination of genotype and sex that underwent magnetic resonance imaging (MRI) is shown in Table 1.

Table 1:

Distribution of all the mice used for ex-vivo imaging by amyloid precursor protein (APP) type (murine or humanized) and apolipoprotein (APOE) genotype.

APP type	APOE genotype	Sex	Count
mAPP	hAPOE3/3	Female	9
mAPP	hAPOE3/3	Male	9
mAPP	hAPOE3/4	Female	6
mAPP	hAPOE3/4	Male	8
mAPP	hAPOE4/4	Female	6
mAPP	hAPOE4/4	Male	7
hAPP	hAPOE3/3	Female	12
hAPP	hAPOE3/3	Male	13
hAPP	hAPOE3/4	Female	11
hAPP	hAPOE3/4	Male	12
hAPP	hAPOE4/4	Female	12
hAPP	hAPOE4/4	Male	19
hAPP	mApoe	Male	12
hAPP	mApoe	Female	12

2.2. NECROPSY

Mice underwent transcardial perfusion with 1 X phosphate buffer saline (PBS) for 4–6 minutes, followed by perfusion with 4% paraformaldehyde (PFA) in PBS for an additional 4–6 minutes. The skulls were cleaned to remove soft tissue and stored in 4% PFA in PBS for 48 hours to allow for complete fixation. They were transferred to PBS solution with 0.01% sodium-azide after 48 hours for long-term storage.

2.3. MRI ACQUISITION

All MRI data were acquired on a Bruker Biospec 70/20 7.0T scanner (Bruker Biospin GmbH, Germany) running Paravision 360(v. 3.5). All mouse brains were prepared for ex-vivo imaging by vacuum desiccation in Fluorinert (FC-3283). The ex-vivo imaging protocol included a T2-weighted rapid acquisition with relaxation enhancement (RARE) sequence (TE: 30 ms; TR: 1800 ms; flip angle 180°; RARE factor: 8; number of averages: 2; FOV: 2.4 × 1.44 × 0.96 cm³; acquisition matrix 320×192×120; reconstructed voxel size 75μm; total acquisition time: 3 hours and 4 minutes), which was used for the present analyses. The mAPP/hAPOE mice used here have been previously analyzed by Raikes et al, however the analytical pipeline and statistical modeling are different for this study.

2.4. DATA PROCESSING

2.4.1. PRE-PROCESSING

All the pre-processing steps were carried out as described in Raikes et al, 2025. Briefly, a custom Apptainer container on the University of Arizona’s High-Performance Cluster was utilized for pre-processing and analyzing data. Visual quality control was carried out on T2-weighted images. Preprocessing steps included denoising, intensity scaling, bias field correction (N4BiasFieldCorrection; ANTs v. 2.5.4), and creation of native space brain masks using a previously defined template space brain mask [32–35].

2.4.2. ATLAS-BASED WORKFLOW

A minimum deformation template (MDT) was generated using 36 mice (3 mice per APP/APOE/sex combination) using the optimized_antsMultivariateTemplateConstruction pipeline [36] running ANTs (v. 2.5.4). Following template generation, all brains (n = 148) were independently non-linearly registered to the MDT and a final study-specific template brain and template mask was produced. The template brain was registered to the Allen Mouse Brain template supplied with Transbrain and the Transbrain atlas was inverse warped to native mouse brain space in a single interpolation step. The Transbrain atlas includes 68 regions labeled bilaterally and regional volumes were computed for each mouse in native space [37]. Additionally, total brain volume, including brain tissue and ventricles, was computed in native space by inverse warping the study-specific template space mask. All volumetric measurements were recorded in cubic millimeters (mm³).

2.5. STATISTICAL ANALYSIS

Regional brain volumes were analyzed using linear models fit independently for each anatomical region. Three sets of outcomes were analyzed: total brain volume (TBV), absolute regional volumes, and relative regional volumes adjusted for total brain volume. TBV and all regional volumes were log-transformed prior to analyses.

Animals were grouped according to APP type (murine vs. humanized) and APOE genotype, which was modeled as four independent groups (murine Apoe, hAPOE3/3, hAPOE3/4, and hAPOE4/4). Therefore, each combination of APP type and APOE genotype was represented as a single group (e.g., hAPP/hAPOE3/4, hAPP/mApoe).

2.5.1. Model specification

Three families of models were fit using ordinary least squares implemented in R (v. 4.5.0). Total brain volume was modeled as:

$$\text{log}\left(TBV\right)~\mathit{\text{Group}}+\mathit{\text{Sex}}+\mathit{\text{Age}}$$

Absolute regional volumes were modeled independently for each region r as:

$$\text{log}\left({\mathit{\text{AbsVolume}}}_r\right)~\mathit{\text{Group}}+\mathit{\text{Sex}}+\mathit{\text{Age}}$$

Finally, relative regional volumes were modeled in the same manner and included log-transformed TBV as a covariate:

$$\text{log}\left({\mathit{\text{AbsVolume}}}_r\right)~\mathit{\text{Group}}+\mathit{\text{Sex}}+\mathit{\text{Age}}+\text{log}\left(\mathit{\text{TBV}}\right)$$

In all models, group denotes APP type/APOE genotype groupings. Sex was included as a two-level categorical variable, and age was mean-centered (in days). No interaction terms were included directly in the models because no mice carried mApp/mApoe and therefore we could not fit a fully parameterized APP x APOE interaction model. For regional analyses, each model was fit once per region and hypothesis tests for group differences were evaluated using estimated marginal means and computed using contrast vectors specifying linear combinations of groupings and holding all other variables at their mean. Planned contrasts were used to address biologically motivated comparisons:

1. Differences between hAPOE genotypes under mouse App (e.g. mAPP/hAPOE4/4 – mAPP/hAPOE3/3)

2. Differences between hAPOE genotypes under humanized APP (e.g. hAPP/hAPOE4/4 – hAPP/hAPOE3/3)

3. The effect of replacing mAPP with hAPP in the presence of hAPOE (e.g. hAPP/hAPOE4/4 – mAPP/hAPOE4/4)

4. Difference-in-difference contrasts to determine if APP-associated differences differed by APOE genotype.

5. Sex differences independent of APP and APOE effects.

For regional analyses, statistical significance was determined using false discovery rate (FDR) correction across regions within each planned contrast. Differences were identified as statistically significant at FDR corrected p < 0.05.

All volumes were entered into models as log-transformed values. Therefore, groupwise differences were exponentiated and reported as percent differences. Relative regional models include log-transformed TBV as a covariate and therefore group differences are interpreted as percent difference after accounting for overall brain size.

As an exploratory analysis, we also assessed the relationship between total brain and regional volume with two behavioral indices (novel object recognition discrimination index and open field freezing time). These models assessed slope differences in group x volume interactions on behavior. For interpretability, behavioral data were z-scored, and volumes were log-transformed and z-scored prior to analyses and findings are reported at nominal p < 0.05 for hypothesis-generating purposes.

Results were plotted using ggplot and overlaid on brain maps using Transbrain.

3. RESULTS

3.1. Total brain volume (TBV)

After adjusting for age and sex, the relationship between APOE genotype and total brain volume differed by APP background (Figure 1). In mice carrying murine APP (mAPP), the presence of an hAPOE4 allele was associated with a 3% increase in TBV relative to hAPOE3/3 mice (p_FDR = 0.025–0.031; Figure 1B). Conversely, among humanized APP (hAPP) mice, hAPOE4/4 carriers exhibited 4% (compared to hAPOE3/3; p_FDR = 1.45 × 10⁻⁵) to 6% (compared to hAPOE3/4; p_FDR = 1.0 × 10⁻⁸) smaller brains than those carrying at least one hAPOE3 allele (Figure 1B).

Figure 1: Impact of amyloid precursor protein (APP) type and humanized apolipoprotein E4 (hAPOE4) carrier status on Total Brain Volume (TBV) showed reduced brain volume in humanized APP carriers relative to their murine counterparts.

Statistically significant differences (FDR-corrected; p < 0.05) are denoted in red. (A) TBV distribution by APP type and hAPOE4 carrier status (dashed line represents reference TBV for an adult C57/Bl6 mouse), and (B) as forest plot showing difference in TBV (%) between groups across different APP types and hAPOE4 allele carrier combinations.

Comparisons of APP backgrounds within each hAPOE genotype revealed that humanized APP reduced total brain volume in a genotype-dependent manner. Reductions ranged from 3.6% in hAPOE3/3 mice to 10.15% in hAPOE4/4 mice (p_FDR range = 5.6 × 10⁻¹⁸ – 0.00235; Figure 1B). Difference-in-difference analyses confirmed that the hAPP-associated reduction was greater in hAPOE4/4 background compared to hAPOE3/3 (mean difference = 6.8%; p_FDR = 7.1 × 10⁻⁶; Figure 1B).

3.2. Absolute and TBV-Adjusted Regional Effects of hAPOE Genotype in mouse APP (mAPP) Mice

In the mAPP background, hAPOE4 carriers exhibited a consistent, brain-wide tendency toward larger volumes relative to hAPOE3/3 mice, spanning all cortical and most subcortical regions (see Figure 2A for region-wise estimates and significance). This effect was most evident for hAPOE3/4 versus hAPOE3/3 in a subset of regions, particularly within somatosensory, temporal, hippocampal, striatal, and thalamic areas (p_FDR = 0.047–0.049; Figure 2D). In contrast, comparisons involving hAPOE4/4 did not yield spatially localized effects (Figure 2B–C), despite similar underlying directional trends.

Figure 2: In the presence of murine amyloid precursor protein (APP), each copy of the humanized apolipoprotein E4 (hAPOE4) allele is linked with increasing total brain volume (TBV).

Spatial distribution of regional volume differences (percentage; TBV-adjusted) linked with the effect of carrying at least one copy of hAPOE4 carrier status relative to hAPOE3/3, on a murine APP background. Greater volume is denoted in red. All results are FDR-corrected for multiple comparison (p < 0.05). (A) Cumulative region-by-region difference visualization as a heatmap, color coded for %age difference, statistical significance and brain region hierarchy. Left-side Panels (B, C and D) represent absolute (TBV-unadjusted) volume differences while right-side panels (E, F and G) represent relative (TBV-adjusted) volume differences.

After TBV adjustment, these differences were attenuated (Figure 2E–G), with no regional patterns emerging. Together, this diffuse and non-localized profile indicates that hAPOE4-related variation in the mAPP background primarily reflects global scaling of brain volume rather than regionally specific remodeling.

3.3. Absolute and TBV-Adjusted Regional Effects of hAPOE Genotype in hAPP Mice

In contrast to the mAPP background, where hAPOE4 carriers exhibited similar volumetric enlargement compared to the hAPOE3/3 mice, the hAPP background was characterized by similarity between hAPOE3/3 and hAPOE3/4 mice, with reductions observed only in the hAPOE4/4 animals (Figure 3A–D). In the hAPP mice, hAPOE4/4 was associated with a brain-wide reduction in absolute regional volumes relative to both hAPOE3/3 and hAPOE3/4 mice, spanning all cortical and nearly all subcortical regions (p_FDR = 4.7 × 10⁻ – 0.049), indicating a global effect.

Figure 3: In presence of humanized amyloid precursor protein (APP), two copies of humanized apolipoprotein E4 (hAPOE4) allele are associated with the lowest brain volume.

Results are represented as spatial distribution of regional volume differences (%) linked with the effect of hAPOE4 carrier status relative to hAPOE3, on a humanized APP background. All results are FDR-corrected for multiple comparison (p < 0.05). Greater volume is denoted in red and lower volume is denoted in blue. All results are FDR-corrected for multiple comparison (p < 0.05). (A) Cumulative region by region difference visualization as a heatmap, color coded for %age difference, statistical significance and brain region hierarchy. Left-side Panels (B, C, and D) represent absolute (TBV-unadjusted) volume differences while right-side panels (E, F and G) represent relative (TBV-adjusted) volume differences.

After TBV adjustment, these global differences were attenuated, and no regionally specific differences emerged between hAPOE3/3 and hAPOE3/4 mice, consistent with the diffuse, non-specific pattern observed in the absolute volumes (Figure 3D, G). In contrast, hAPOE4/4 mice exhibited a more regionally specific pattern, with smaller volumes localized to visual regions and hippocampal CA3, alongside relative enlargement of the ventral thalamus (p_FDR = 0.0206 – 0.05; Figure 3D). Additional focal reductions in posterior amygdalar nucleus and subiculum, and enlargement of the epithalamus, were observed relative to hAPOE3/4 mice (p_FDR = 0.0031–0.0258; Figure 3E). Together, these findings indicate that in the hAPP background, hAPOE4/4 drives a global reduction in brain volume that, after normalization, reveals selective regional differences, and is consistent with partial mitigation of these effects by a single hAPOE3 allele.

3.4. Absolute and TBV-Adjusted Regional Effects of APP Background Stratified by hAPOE Genotype

Replacing mAPP with hAPP resulted in a consistent reduction in regional volumes across all hAPOE genotypes, with a shift in the distribution of effects across genotypes (Figure 4A–D). In hAPOE3/3 mice, hAPP effects were heterogeneous, with both larger and smaller regional volumes observed (p_FDR = 3.1 × 10⁻¹¹ – 0.0302; Figure 4B). This pattern shifted with increasing hAPOE4 dose, such that hAPOE3/4 mice showed predominantly smaller volumes (p_FDR = 1.2 × 10⁻¹¹ – 0.0437; Figure 4C), and hAPOE4/4 mice exhibited a near-uniform reduction across regions (p_FDR = 1.6 × 10⁻²³ – 0.0206; Figure 4D). Across all genotypes, hAPP mice consistently exhibited larger epithalamic volumes (Figure 4A).

Figure 4: The presence of humanized amyloid precursor protein (hAPP) across different humanized apolipoprotein (hAPOE) genotypic combination results in smaller brains with greater representational impact on brain regions involved in cognition.

Results shown as spatial distribution of regional volume differences (%) linked with the effect of hAPP inclusion across each genotypic combination. Greater volume is denoted in red and lower volume is denoted in blue. All results are FDR-corrected for multiple comparison (p < 0.05). (A) Cumulative region-by-region difference visualization as a heatmap, color coded for %age difference, statistical significance and brain region hierarchy. Left-side Panels (B, C, and D) represent absolute (TBV-unadjusted) volume differences while right-side panels (E, F and G) represent relative (TBV-adjusted) volume differences.

After accounting for the globally smaller brains in the hAPP mice (Figure 1), a consistent regional pattern emerged (Figure 4E–G). Across genotypes, hAPP mice showed relatively greater cortical volumes alongside persistently smaller subcortical volumes, with this cortical–subcortical pattern most clearly defined in hAPOE4/4 mice (p_FDR = 2.4 × 10⁻¹² – 0.0487; Figure 4G). Notable exceptions included larger CA1 volumes, mixed thalamic effects, and persistent epithalamic enlargement across genotypes.

3.5. Interaction of humanized APP and humanized APOE on regional volume

Difference-in-slope analyses were conducted to assess whether the magnitude of hAPP-related volume reductions differed by hAPOE genotype (Figure 5). These contrasts indicated that the reduction associated with hAPP, relative to mAPP, was consistently greater in hAPOE4/4 mice than in hAPOE3/3 or hAPOE3/4 mice, spanning the majority of regions (p_FDR = 6.0 × 10⁻⁶ – 0.0463; Figure 5A–C). In contrast, differences between hAPOE3/3 and hAPOE3/4 mice were not observed (all p_FDR > 0.05; Figure 5D), reflecting their similar volumetric profiles within the hAPP background.

Figure 5: Humanized amyloid precursor protein (hAPP)-associated reduction across different humanized apolipoprotein (hAPOE) genotypic combinations showed greatest reductions in the hAPP/hAPOE4/4 carriers.

Results shown as spatial distribution of regional volume differences (%) linked with the effect of interaction of humanized APP and humanized APOE. Lower volume is denoted in blue. All results are FDR-corrected for multiple comparison (p < 0.05). (A) Cumulative region-by-region difference visualization as a heatmap, color coded for %age difference, statistical significance and brain region hierarchy. Left-side Panels (B, C and D) represent absolute (TBV-unadjusted) volume differences while right-side panels (E, F and G) represent relative (TBV-adjusted) volume differences.

After TBV adjustment, differences in hAPP effect magnitude were attenuated and did not exhibit regionally specific patterns (all p_FDR > 0.05; Figure 5E–G). However, the residual pattern was directionally consistent with prior observations, with hAPOE4 carrier mice showing relatively greater cortical reductions and comparatively smaller subcortical reductions than hAPOE3/3 mice. Together, these findings indicate that hAPOE4 primarily modulates the magnitude of hAPP-associated effects, with limited evidence for genotype-dependent regional specificity after normalization.

3.6. Sex differences

Independent of APP type or APOE genotype, female mice had larger brains than males (mean difference = 1.75%; p = 0.0005). Regionally, this was reflected by greater absolute volumes in females across most of the cortex and subcortex, with no regions showing larger volumes in males (p_FDR = 3.0 × 10⁻⁸ – 0.0394; Figure 6A–B).

Figure 6: Females had larger brain volume and cortical regions than males, after adjusting for amyloid precursor protein (APP) type and humanized apolipoprotein (hAPOE) genotype.

Results shown as spatial distribution of regional volume differences (%) linked with the effect of sex. Greater volume is denoted in red and lower volume is denoted in blue. All results are FDR-corrected for multiple comparison (p < 0.05). A) Region-by-region difference heatmap, color coded for %age difference, statistical significance and brain region hierarchy. Left side panel showing absolute volumes and C) Right side panel adjusted for differences in total brain volume (TBV).

After TBV adjustment, a regionally specific pattern was evident (Figure 6C). Females exhibited relatively greater cortical and thalamic volumes (p_FDR = 2.6 × 10⁻⁶ – 0.0191), whereas males showed relatively greater volumes in the hippocampus, amygdala, and striatum (p_FDR = 5.3 × 10⁻⁶ – 0.0424). Together, these findings are consistent with sex differences primarily reflecting global scaling of brain volume, with additional region-specific variation after normalization.

3.7. Association between behavior and regional brain volumes

Exploratory analyses identified six regions in which associations between regional volume and open field freezing time differed between hAPP/hAPOE3/3 and hAPP/hAPOE4/4 mice. Greater volumes in the striatum-like amygdalar nuclei (sAMY), posterior agranular insular area (AIp), hippocampal CA2, and pallidum (PAL) were associated with shorter freezing times in hAPP/hAPOE4/4 mice but longer freezing times in hAPP/hAPOE3/3 mice (p_FDR = 0.0502 – 0.096; Figure 7). These associations did not meet a strict p_FDR < 0.05 threshold but were retained for descriptive interpretation given their consistency across regions. No significant volumetric associations were observed for novel object recognition (NOR) discrimination index.

Figure 7: In humanized amyloid precursor protein (hAPP)/humanized apolipoprotein (hAPOE) mice, presence of two copies of the hAPOE4 allele induces freezing behavior.

Results presented as correlation analyses between freezing time and regional brain volumes (TBV-adjusted). All values are z-scored. After controlling for TBV, four regional volumes (in clockwise order; sAMY – Striatum-like amygdalar nuclei, CA2, PAL – Pallidum, and Alp - Agranular insular area) were negatively associated with hAPP/hAPOE4/4 carrier status (in red) and positively associated with hAPP/hAPOE3/3 carrier status (in blue). Only FDR-corrected (p < 0.1) results are shown.

4. DISCUSSION

Age, Apolipoprotein E4 (APOE4) genotype, and biological sex are major risk factors for late-onset Alzheimer’s disease (LOAD). Our findings indicate that hAPP is a primary driver of brain volume in this LOAD mouse model, producing global reductions that are amplified by the hAPOE4/4 genotype, whereas hAPOE4 alone was insufficient to recapitulate AD-relevant volumetric patterns. Sex differences were modest and independent of genotype, with females exhibiting slightly larger brain volumes.

Targeted replacement of murine APP with humanized APP induced a 3–10% decrease in total brain volume relative to murine APP/hAPOE counterparts, with the largest effect observed in hAPP/hAPOE4/4 mice. This gradient of effect was evident in absolute regional volumes. All hAPP mice exhibited smaller subcortical volumes than their mAPP counterparts. The cortical regions presented a more complex picture, where each copy of the hAPOE4 allele altered the balance of regional effects, with some cortical volumes changing from being greater in hAPP/hAPOE3/3 mice to being smaller in hAPP/hAPOE4 carriers. The extent of this shift was most pronounced in hAPOE4/4 mice.

Further, after accounting for total brain volume, hAPP/hAPOE mice exhibited relatively greater cortical and smaller subcortical volumes independent of hAPOE genotype, indicating that hAPP effects are regionally differentiated only after normalization for global brain size. Within this morphological framework, the combination of hAPP with hAPOE4/4 genotype most closely approximates a brain-wide phenotype consistent with LOAD-associated volumetric patterns. Additionally, hAPP/hAPOE4/4 mice exhibited smaller regional volumes compared to hAPP/hAPOE3/3 or hAPP/hAPOE3/4 mice, with few differences persisting after adjusting for total brain volume, indicating that hAPOE4 homozygosity primarily modulates the magnitude of the hAPP-associated effects. Our results also suggest that the presence of a single hAPOE3 allele may attenuate the magnitude of hAPP-associated volume reductions. Additional histological analyses will be necessary to confirm this hypothesis [38].

In mice with murine APP, the inclusion of one or more hAPOE4 alleles increased total brain volume above hAPOE3/3, without differences between hAPOE3/4 and hAPOE4/4 mice, consistent with previous findings on aging in this model as well as gene expression findings [23, 39]. These volumetric results in the mouse are consistent with human studies of normal aging, where APOE4 is associated with similar or greater global brain volumes compared to APOE3 without regionally specific effects [40–43]. An hAPOE4-associated increase in TBV suggests that, by itself, hAPOE4 alone does not produce an AD-like phenotype in these mice. Previous work on these mice, as well as other APOE targeted replacement mice (APOE-TR) have shown no cognitive deficits or AD-relevant pathological changes, consistent with low AD translational relevance [24, 44].

Across each hAPP and hAPOE genotype comparison, the most consistently decreased volumes were observed in thalamic (ATN, GENv, GENd, RT), amygdalar (Ald, LA), and hippocampal (CA1, CA2) regions, suggesting that these regions are potentially impacted by this combination of humanized AD risk genes. Notably, these regions are consistently identified as atrophic in clinical AD [45–47]. Collectively, these findings indicate that hAPOE4 primarily modulates APP-driven effects rather than independently recapitulating AD-relevant volumetric changes. Therefore, both humanized genes are necessary to recapitulate the effect of APOE4 on brain volumes observed in AD patients [48–50].

After accounting for both APP and APOE effects in our models, we observed sex differences, with female mice having 1.75% greater brain volume than males. Past work in the hAPOE model identifies sex as a major driver of regional brain volumes and past work in the background strain highlights sexual dimorphism in brain volume distributions between both gray and white matter as well as cortex and subcortex [51, 52]. The brain volumes in aged female hAPP/hAPOE mice reported herein do not, however, directly recapitulate the phenotype observed in women carrying the APOE4 allele [53, 54]. Further work will be necessary to target and translate the mechanisms of sex-specific vulnerabilities on brain volumetrics to this mouse model.

Open field freezing behavior was directionally consistent with observed brain volume differences across hAPP/hAPOE genotypes, whereas novel object recognition was not. Differences in the relationship between freezing and brain volumes were observed in six regions, including the amygdalar, hippocampal, visceral, and insular areas. These regions are canonically associated with anxiety behaviors in mice, which can be indexed by freezing behaviors [55–57]. Differential effects were observed, where greater volume was associated with less freezing in hAPP/hAPOE4/4 mice and greater freezing in hAPP/hAPOE3/3 mice. Anxiety is one of the most common psychiatric disorders in the elderly population and has been associated with preclinical stages of AD, where APOE4 carrier status has been linked to higher anxiety rates in mild dementia [58, 59].

While the hAPP/hAPOE mouse model did not exhibit overt AD pathology, these findings shed light on the previously unexplored influence of hAPP on total and regional brain volume, separately and in combination with APOE4. Across models incorporating age, sex, APP status, and APOE genotype, hAPP in combination with hAPOE4 was associated with global reductions in brain volume and regionally specific patterns after normalization. Humanized APP was the primary driver of these volume reductions, supporting its role in developing translational models of LOAD. The absence of pathology also suggests that a combination of advanced age, APP, and APOE4 is insufficient to model all aspects of LOAD. However, when considering brain volume as one of the key neuroimaging biomarkers, this combination of risk factors/genes more closely reflects the clinically observed volumetric reductions associated with APOE4 carriership compared to models with familial genes, hAPOE, or hAPP alone.

4.1. LIMITATIONS AND FUTURE DIRECTIONS

The current study focused on exploring the impact of humanized APP in an AD-risk mouse model at an advanced age in both sexes using neuroimaging. However, several limitations should be acknowledged. First, we did not have an aged background strain cohort that would provide a direct means of fully characterizing the independent and combined effects of hAPP and hAPOE effects. The mAPP/hAPOE mice analyzed herein had total brain volumes consistent with published TBV values for the adult mice with the background strain (mAPP/hAPOE: 499 mm³; C57/BL6; 535mm³) [60]. Future work could address the magnitude of hAPOE effects relative to background strain volumetrics.

Another limitation of this model is the lack of evidence for AD-linked amyloid pathology despite advanced age and female sex. Finally, the study design limits some interpretability. Our ex-vivo, fixed tissue, cross-sectional design at old age cannot distinguish between developmental and atrophic effects of hAPP. Future work is required to establish whether the observed volume differences reflect translationally relevant atrophy or developmental vulnerability. Additionally, we did not model the impact of fixation time or storage durations, and these may have introduced unmeasured confounding due to known fixation effects [61]. However, we applied the same fixation protocol to all brains, minimizing this confounding effect.

5. CONCLUSION

Outcomes of the analyses reported herein indicate that the addition of hAPP resulted in global reductions in brain volume, with smaller cortical and subcortical regions in an aged hAPOE mouse model in both sexes. These outcomes highlight the importance of modeling combined humanized AD risk genes when considering MRI-derived outcome measures. Coupling hAPP and hAPOE4/4 was associated with the largest reductions in total and regional brain volumes. Consistent with these findings, hAPOE4 alone was insufficient to recapitulate AD-relevant brain changes. While the modest degree of AD pathology is a significant limitation, these aged hAPP/hAPOE mice produce a brain imaging phenotype more consistent with clinically observed AD-related volumetric reductions.

Funding Statement

Research reported herein was supported by the National Institute on Aging (grants P01AG026572 and R01AG057931), the University of Arizona Center for Innovation in Brain Science to Roberta Diaz Brinton and S10 OD025016 to the University of Arizona TBIR.

Data availability statement

The datasets generated and/or analyzed during the current study are available from the corresponding author on reasonable request.