Characterizing molecular and synaptic signatures in mouse models of late-onset Alzheimer's disease independent of amyloid and tau pathology

Abstract

Introduction: MODEL-AD (Model Organism Development and Evaluation for Late-Onset Alzheimer's Disease) is creating and distributing novel mouse models with humanized, clinically relevant genetic risk factors to capture the trajectory and progression of late-onset Alzheimer's disease (LOAD) more accurately.

Methods: We created the LOAD2 model by combining apolipoprotein E4 (APOE4), Trem2*R47H, and humanized amyloid-beta (Aβ). Mice were subjected to a control diet or a high-fat/high-sugar diet (LOAD2+HFD). We assessed disease-relevant outcome measures in plasma and brain including neuroinflammation, Aβ, neurodegeneration, neuroimaging, and multi-omics.

Results: By 18 months, LOAD2+HFD mice exhibited sex-specific neuron loss, elevated insoluble brain Aβ42, increased plasma neurofilament light chain (NfL), and altered gene/protein expression related to lipid metabolism and synaptic function. Imaging showed reductions in brain volume and neurovascular uncoupling. Deficits in acquiring touchscreen-based cognitive tasks were observed.

Discussion: The comprehensive characterization of LOAD2+HFD mice reveals that this model is important for preclinical studies seeking to understand disease trajectory and progression of LOAD prior to or independent of amyloid plaques and tau tangles.

Highlights: By 18 months, unlike control mice (e.g., LOAD2 mice fed a control diet, CD), LOAD2+HFD mice presented subtle but significant loss of neurons in the cortex, elevated levels of insoluble Ab42 in the brain, and increased plasma neurofilament light chain (NfL). Transcriptomics and proteomics showed changes in gene/proteins relating to a variety of disease-relevant processes including lipid metabolism and synaptic function. In vivo imaging revealed an age-dependent reduction in brain region volume (MRI) and neurovascular uncoupling (PET/CT). LOAD2+HFD mice also demonstrated deficits in acquisition of touchscreen-based cognitive tasks.

Keywords: APOE4; Alzheimer's disease; LOAD; MODEL‐AD; TREM2; genetics; high‐fat diet; late‐onset Alzheimer's disease.

FIGURE 1

Longitudinal metabolic and behavioral phenotyping of mice on high‐fat diet. LOAD1 (APOE4/Trem2*R47H) and LOAD2 (hAbeta/APOE4/Trem2*R47H) animal strains differ in the App allele with a humanized Abeta1‐42 region (G601R, F606Y, R609H in the mouse gene, corresponding to amino acid positions 676, 681, 684 in the human APP locus) (A). Alignment of mouse (top; Uniprot ID P12023) and humanized (bottom; Uniprot ID P05067) APP amino acid sequences. White letters denote nonhomology. Red arrows indicate cleavage sites of processing enzymes. Yellow arrows denote sites of humanizing mutations in App allele (LOAD1, top, and LOAD2, bottom). (Cohort 1) Animals of an 18 month longitudinal cohort were assayed at 4, 12, and 18 months of age. Males and females, of LOAD1 and LOAD2 genotypes, fed either CD or HFD beginning at 2‐months of age were measured for body weight (B), fasted blood glucose (C), and frailty assay index score (D), as a measure of general animal health changes. Running wheel assay measured average animal activity time for 3 days and nights at the 18 month age timepoint (E). Spontaneous alternation behavioral assay was utilized to measure cognition longitudinally across ages at the 18‐month age timepoint (F) (Three‐way ANOVA [sex, genotype, diet effects]; * = p < 0.05). APP, amyloid precursor protein; CD, control diet; HFD, high‐fat diet.

FIGURE 2

Neuropathological assessment of brain tissue. Immunohistochemistry of brain tissue in the cortex and hippocampus from 18 month old animals stained for cell markers to reveal genotype‐ and diet‐driven differences in glial cell densities. Slices of brain hemispheres were stained with NeuN (neurons A,B) or IBA1 (microglia C,D) (representative cortical images shown from LOAD2 females with DAPI co‐stain) and counted relative to area. Astrocytes (GFAP) quantitated in the hippocampus of LOAD2 females fed either CD or HFD (E‐F). ThioS staining of brain tissue to visualize amyloid plaques (representative images shown from LOAD2 females) (G). Inlay: scaled image of 12 month B6J.APP‐SAA hyperamyloid positive controls. Linear regression analyses were performed to identify effect of each factor on neurofilament light‐chain (NfL) levels determined by ELISA testing in plasma derived from terminal, peripheral blood samples at 18 months of age (H). (NeuN = neuronal marker; ThioS = amyloid plaques; GFAP = astrocyte marker; IBA1 = microglial marker. Scale bar equals 100 µm).

FIGURE 3

A gene module associated with AD biomarkers is driven by age and high‐fat/high‐sugar diet. The lightyellow gene module was associated with advanced age (p < 0.001) and both genotypes on HFD (p < 0.05), while the turquoise module was primarily associated with age (p < 0.001) (A). Correlations between the turquoise and lightyellow module eigengenes. Lightyellow was significantly correlated with frailty score, body weight, NfL, and many plasma cytokines (IL‐1β, IL‐2, IL‐12p70, IL‐10, IL‐5, IL‐6, KC‐GRO), while the age‐driven turquoise module was correlated with behavioral assay (frailty score and body weights) and weakly correlated with a few plasma cytokines (IL‐2, IFNγ) and inflammatory cell counts (IBA1 and GFAP counts) (B). Positive correlation coefficients are shown in blue and negative correlations in red, proportional to color intensity and circle size, with frames for significant correlations (FDR < 0.05). AD‐related biological domain enrichment analysis in the age and HFD driven lightyellow module gene set using Fisher exact test, with the top six enriched GO terms within each enriched bidomain (C). Network of genes in each enriched biological domain and the lightyellow module (D). AD, Alzheimer's disease; HFD, high‐fat diet.

FIGURE 4

LOAD mice exhibit proteomics changes similar to human LOAD. Correlation coefficients between 18‐month‐old LOAD mouse models and 44 human proteomics co‐expression modules [Johnson et al Ravi's ref ⁹ ] (A). Modules in bold face were significantly correlated to one or more AD traits. Circles correspond to positive (blue) and negative (red) Pearson correlation coefficients for protein expression changes in LOAD mice (log fold change of LOAD strains vs. B6J) and human disease (log fold change for cases vs. controls). Color intensity and size of the circles are proportional to the Pearson correlation coefficient, with significant correlations (p < 0.05) framed. Five top enriched GO terms for proteins with common directional changes for 18‐month‐old LOAD2 mice on HFD and human AD cases (B). Protein module network with common directional changes for 18‐month‐old LOAD2 mice on HFD and human proteomics modules (C). Blue (red) nodes correspond to increased (reduced) protein abundance in both 18‐month‐old LOAD2 HFD mice compared to B6J mice and human AD cases versus controls. AD, Alzheimer's disease; HFD, high‐fat diet.

FIGURE 5

High‐fat diet reduces brain volume in multiple brain regions and alters plasma biomarkers. Volume statistics map for LOAD2 males at 4, 12, and 18 months. The significant brain areas were overlaid over gray scale subject template image. (The p‐values were converted into logarithmic scale between range −5 and −1. Volume statistics maps for Load2 female at 4, 12, and 18 months. The significant brain areas were overlaid over gray scale subject template image. The p‐values were converted into logarithmic scale between range −5 and −1.) (A). Bar plot showing whole brain volume for CD and HFD groups for male and female cohorts at age groups 4, 12, and 18 months, respectively (B). Longitudinal assessment of peripheral plasma from male and female LOAD2 animals provided CD or HFD measured levels of NfL (C) and Aβ species (40 and 42; panels D and E, respectively). Cytokines related to inflammation regulation were also measured and include TNF‐α (F), IFNγ (G), IL‐6 (H), IL‐2 (I), and IL‐1β (J). Two‐way ANOVA *p < 0.05, **p < 0.01, ***p < 0.001, ^**** p < 0.0001. A29c, cingulate cortex; cc, corpus callosum; cg, Cingulum; Cpu, striatum; FrA, frontal association area; IC, inferior colliculus; LO, lateral orbital cortex; M1, primary motor cortex; M2 secondary motor cortex; Pir, piriform cortex; S1BF Primary Somatosensory Cortex, Barrel Field; S1HL Primary Somatosensory Cortex, Hindlimb; S1, Primary Somatosensory Cortex; SPT, Septum; S2, Secondary Somatosensory Cortex; V1M, Primary Visual Cortex, Monocular area.

FIGURE 6

Brain biomarkers in a longitudinal cohort of LOAD2 mice fed a high‐fat diet. Whole brain lysates from 18‐month‐old animals were analyzed by ELISA for biomarkers and cytokines aligned with inflammation and disease progression. For changes related to sex, genotype, and diet measurements are provided for insoluble Aβ42 (A) and Aβ40 (B); soluble Aβ42 (C) and Aβ40 (D); IL‐1β (E); IL‐5 (F); IL‐4 (G); IL‐2 (H); KC‐GRO (I); IL‐12 (J); TNF‐α (K); and IL‐10 (L). Analysis by two‐way ANOVA *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001.

FIGURE 7

Neurovascular Uncoupling of LOAD1 and LOAD2 mouse models. The degree of neurovascular coordination in LOAD1 (A) and LOAD2 mouse models (B) conditioned on high‐fat diet (HFD), we performed uncoupling analysis. (Left) Uncoupling analysis chart in male (blue) and female (red) mice at 12 months, with many brain regions showing significant decreases in metabolism with increases in perfusion. LOAD2 animals aged to 18 months (C) were similarly analyzed. (Upper Right) Female and (Lower Right) Male p‐value males showing which regions were significantly different for perfusion, metabolism, and uncoupling. HFD, high‐fat diet.

FIGURE 8

Comprehensive validation of LOAD2 mouse model for preclinical drug testing. As a confirmation and extension of initial characterization data of the LOAD2 mouse model conditioned on high‐fat diet (HFD) to serve as a potential model for preclinical testing, independent cohorts were evaluated for disease trajectory of serial plasma biomarkers and cognitive testing. (A) Illustration of timeline and procedures; (B) plasma TNF‐α (pg/mL); (C) plasma IL‐6 (pg/mL); (D) plasma IL‐10 (pg/mL); (E) plasma IL‐1β (pg/mL); (F) plasma IFNy (pg/mL), (G) plasma IL‐5 (pg/mL); (H) plasma KC‐GRO (pg/mL); (I) plasma IL‐2 (pg/mL); (J) plasma Aβ40 (pg/mL); (K) plasma Aβ42 (pg/mL); (L) calculated Aβ 42:40 ratio in plasma; (M) learning curves of aged (14+ month) LOAD2 mice ± HFD in comparison to age‐ and sex‐matched WT controls during the acquisition phase of the touchscreen cognitive testing battery. Plasma cytokines and plasma Aβ were measured using MesoScale Discovery multiplex ELISA kits in accordance with the manufacturer's protocol. HFD, high‐fat diet.