APOE4 to APOE2 allelic switching in mice improves Alzheimer's disease-related metabolic signatures, neuropathology and cognition

Abstract

Compared to individuals carrying two copies of the ε4 allele of apolipoprotein E (APOE), ε2 homozygotes have an approximate 99% reduction in late-onset Alzheimer's disease (AD) risk. Here we develop a knock-in model that allows for an inducible 'switch' between risk and protective alleles (APOE4s2). Gene expression and proteomic analyses confirm that APOE4s2 mice synthesize E4 at baseline and E2 after tamoxifen administration. A whole-body allelic switch results in a metabolic profile resembling E2/E2 humans and drives AD-relevant alterations in the lipidome and single-cell transcriptome, particularly in astrocytes. Finally, when crossed to the 5xFAD background, astrocyte-specific E4 to E2 switching improves cognition, decreases amyloid pathology, lowers gliosis and reduces plaque-associated apolipoprotein E. Together, these data show that a short-term transition from APOE4 to APOE2 can broadly affect the cerebral transcriptome and lipidome, and that astrocyte-specific APOE replacement may be a viable strategy for future gene editing approaches to simultaneously reduce multiple AD-associated pathologies.

Fig. 1. APOE ‘switch mice’ efficiently transition from the expression of ApoE4 to ApoE2.

a, Schematic depicting the genetic construct for the APOE4 to APOE2 ‘switch’ mouse (APOE4s2) and the experimental design for the experiments performed in APOE4s2 mice crossed to a globally (‘G’) expressed ROSA26-CreER^T1 recombinase (APOE4s2^G). Homozygous floxed Cre⁻ littermates (APOE4s2) were used as controls. b,c, Allelic discrimination plots depicting a shift from APOE4 to APOE2 mRNA expression (b) in APOE4s2^G mice in both the brain (b) and liver (c) compared to Cre⁻ APOE4s2 littermates and ApoE2 and ApoE4 TR controls injected with TAM. d, Methodology used to semiquantitatively detect the ratio of peptides unique to E2 or E4 in the brain and plasma using liquid chromatography (LC)–MS/MS-based proteomic analysis. e,f, Ratio of unique E2 versus E4 peptides detected using LC–MS/MS-based proteomic analysis showing an efficient transition to E2 in both the brain (f) and plasma (g) of APOE4s2^G (n = 4), APOE4s2 (n = 4), E2-TR (n = 2) and E4-TR (n = 2) mice. Data are represented as mean values ± s.e.m. RFU, relative fluorescence unit.

Fig. 2. Whole-body APOE switching results in an ‘E2-like’ peripheral and cerebral metabolic profile.

a, Plasma TGs before the TAM-induced switch (E2-TR, n = 8; E4-TR, n = 8; APOE4s2, n = 8; APOE4s2G, n = 8; one-way analysis of variance (ANOVA) with Tukey’s multiple comparisons test). b, Plasma TGs after TAM (E2-TR, n = 8; E4-TR, n = 8; APOE4s2, n = 7; APOE4s2G n = 7; one-way ANOVA with Tukey’s multiple comparisons test). c, TGs in each lipoprotein fraction after switching (E2-TR, n = 3 pooled; E4-TR, n = 3 pooled; APOE4s2, n = 3 pooled; APOE4s2G, n = 3 pooled). d,e, ELISA quantification of ApoE in plasma before TAM (d) and after the TAM-induced switch (e) (E2-TR, n = 5; E4-TR, n = 5; APOE4s2, n = 6; APOE4s2G, n = 7; one-way ANOVA with Tukey’s multiple comparisons test). f, ELISA quantification of ApoE in whole-brain tissue after TAM (E2-TR, n = 3; E4-TR, n = 3; APOE4s2, n = 5; APOE4s2G, n = 5; one-way ANOVA with Tukey’s multiple comparisons test). g,h, Representative images showing GFAP (green) (g) or IBA1 (green) (h) and ApoE (red) in APOE4s2 controls (top) or APOE4s2^G brains (bottom). i, Volcano plot comparing brain lipids altered in APOE4s2^G mice versus Cre⁻ APOE4s2 controls (APOE4s2, n = 8; APOE4s2, n = 8; all female). The cutoff value for significance is log₁₀(P) > 1.3. j, Heatmap depicting the top ten most significantly changed lipids between APOE4s2^G and APOE4s2 controls. k,l, WGCNA analysis identified lipid modules associated with the switch from E4 to E2. k, Module-trait correlations between modules and APOE expression indicated by the numbers in the table, with P values shown in parentheses. l, Network plot of lipids identified in the blue module. a–f, For detailed statistical information, see Supplementary Table 3. Data represent mean ± s.e.m. of biological replicates. Individual data points have been plotted; in the box plots, the center line indicates the median; the box limits indicate the upper and lower quartiles; and the whiskers indicate the minimum and maximum. ***P < 0.001, ****P < 0.0001. i–l, Twenty-two classes of lipids were detected: CER; cardiolipin (CI); coenzyme (CO); diglyceride (DG); dimethyl phosphotidylehtanolamine (DMEPE); fatty acid (FA); ganglioside (GM); lysodimethylphosphatidylethanolamine (LDMEPE); LPC; lysophosphatidylethanolamine (LPE); lysophosphatidylglycerol (LPG); lysophosphatidylinositol (LPI); lysophosphatidylserine (LPS); (O-acyl)-ω-hydroxy FA (OAHFA); phosphatidic acid (PA); PC; PE; phosphatidylglycerol (PG); phosphatidylinositol (PI); phosphatidylserine (PS); sphingomyelin (SM); and TG. HDL, high-density lipoprotein. LDL, low-density lipoprotein.

Fig. 3. Full-body APOE allele switching drives AD-relevant transcriptomic changes in multiple CNS cell types.

a, Uniform manifold approximation and projection (UMAP) of all 85,701 cells analyzed, classified based on canonical gene expression markers into 12 cell types (APOE4s2^G, n = 3; APOE4s2, n = 3; E2-TR, n = 3, E4-TR, n = 2; all females). b, Number of DEGs between APOE4s2^G and Cre⁻ APOE4s2 controls according to cell type (P_adj = <0.001). c, Venn diagrams revealing commonality of DEGs in the APOE4s2^G versus APOE4s2 and E2-TR versus E4-TR datasets. d–g, Upregulated and downregulated GO terms associated with DEGs from astrocytes (d), oligodendrocytes (e), microglia (f) and endothelial cells (g). The cutoff value for significance was set at a false discovery rate (FDR)-adjusted P value of 0.05. h, Pie charts visualizing the categorical distribution of APOE switch-related GO terms into AD-related pathways. Upregulated (left) AD pathways account for 24%, while downregulated (right) AD pathways account for 21% of total downregulated GO terms. i, Overlap of the APOE switch transcriptomics dataset with five relevant gene lists from AD-related transcriptomic studies. DAA, disease-associated astrocyte; OPC, oligodendrocyte precursor cell; Transcription corep. binding, transcription corepressor binding; VLC, vascular leukocyte.

Fig. 4. An astrocyte-specific APOE4 to APOE2 switch also drives AD-relevant transcriptomic changes.

a, Graphic depicting astrocyte-specific APOE4s2^A mice with Aldh1l1-CreER^T2. b, Representative images showing colocalization of tdTomato Ai9 Cre reporter (red) and GFAP (green). c,d, Allelic discrimination plots measuring APOE4 versus APOE2 mRNA expression in astrocytes (c) sorted from other cell types in the brain, or from whole-brain homogenates (d), of APOE4s2^A mice versus APOE4s2 controls (E2-TR, n = 2; E4-TR, n = 2; APOE4s2^A, n = 6; APOE4s2, n = 6). e, Number of DEGs between APOE4s2^A and Cre⁻ APOE4s2 controls according to cell type (P_adj ≤ 0.001) (APOE4s2^A, n = 3; APOE4s2, n = 3). f, Overlap of the APOE switch transcriptomic dataset with five other gene lists from AD-related transcriptomic studies. g, Venn diagrams revealing the commonality of DEGs in the APOE4s2^A versus APOE4s2 and APOE4s2^G versus APOE4s2 datasets. h–k, GO analysis for downregulated and upregulated terms associated with DEGs from microglia (h), oligodendrocytes (i), astrocytes (j) and endothelial cells (k). The cutoff value for significance was set at an FDR-adjusted P < 0.05 and fold change ≥ 1.3.

Fig. 5. Astrocyte-specific APOE switching improves associative memory and decreases plaque load in a model of AD.

a, Schematic depicting the APOE4s2^A × 5xFAD cross to generate the 4s2^A/FAD mice. b–d, Percentage time freezing measured during fear conditioning (learning) (b), and the contextual (memory) (c) and cued memory test (d) tests, in 4s2^A/FAD mice compared to Cre⁻ controls 2 months after TAM (4s2^A/FAD, n = 16, 4s2/FAD, n = 16; **P < 0.01,***P < 0.001, ****P < 0.0001, using a repeated measures ANOVA). e, Representative images of Amylo-Glo (blue) and nucleus (red) staining in 4s2^A/FAD and 4s2/FAD mice. f, Quantification of the total percentage amyloid⁺ area in 4s2^A/FAD mice showed significant decreases in total amyloid measured according to the percentage of the amyloid⁺ area. The dashed line represents the mean amyloid percentage area in 4s2/FAD mice at the 6-month pre-switch time point (y = 0.118). **P < 0.01, using a two-tailed unpaired t-test. g,h, Heatmap (g) and quantification (h) of the percentage of amyloid⁺ area according to region in 4s2^A/FAD versus 4s2/FAD mice (4s2^A/FAD, n = 10; 4s2/FAD, n = 10; two-way ANOVA with multiple comparisons). i–l, ELISA quantification of Aβ₄₀ (i,j) or Aβ₄₂ (k,l) in soluble (PBS) or insoluble (5 M GuHCl) whole-brain fractions from 4s2^A/FAD mice compared to 4s2/FAD controls (4s2^A/FAD, n = 14; 4s2/FAD, n = 14; *P < 0.05, **P < 0.01, ****P < 0.0001, using a two-tailed unpaired t-test). m, Representative images of mOC31 anti-CAA (green) and CD31 (red) staining in the cortex of 8-month-old 4s2^A/FAD and 4s2/FAD mice. n, Quantification of CAA⁺ vessels normalized to total CD31⁺ area in 8-month-old 4s2^A/FAD mice and 4s2/FAD controls after the switch compared with 6-month-old mice before the switch (4s2^A/FAD 8 months, n = 8; 4s2/FAD 8 months, n = 8; 4s2/FAD 6 months, n = 8; *P < 0.05, **P < 0.01, using a two-way ANOVA with multiple comparisons). a–n, For detailed statistical information, see Supplementary Table 3. Data represent mean ± s.e.m. of biological replicates. Individual data points have been plotted. Sex is denoted by the shape of the symbol: females (circle), males (triangle). In the box plots, the center line indicates the median; the box limits indicate the upper and lower quartiles; and the whiskers indicate the minimum and maximum.

Fig. 6. Replacement of astrocyte E4 with E2 reduces plaque-associated gliosis.

a, Representative images of Amylo-Glo (blue), IBA1 (green), GFAP (orange) and nucleus (red) staining in 4s2^A/FAD mice and 4s2/FAD controls. b,c, Quantification of the percentage GFAP⁺ (b) or IBA1⁺ (c) area in 4s2^A/FAD mice versus 4s2FAD controls (4s2^A/FAD, n = 16; 4s2/FAD, n = 16; ***P < 0.001, ****P < 0.000, using a two-tailed unpaired t-test). d,e, Quantification of GFAP⁺ (d) or IBA1⁺ (e) percentage area according to region in 4s2^A/FAD and 4s2/FAD mice (4s2^A/FAD n = 10; 4s2/FAD n = 10; *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001, using a two-way ANOVA with multiple comparisons). f, Representative images of Amylo-Glo (blue), IBA1 (green), GFAP (orange) and nucleus (red) proximity and colocalization. g–i, Percentage area of colocalization of Amylo-Glo and GFAP (g), Amylo-Glo and IBA1 (h) or Amylo-Glo with GFAP and IBA1 (i) in 4s2^A/FAD and 4s2/FAD mice (4s2^A/FAD, n = 16, 4s2/FAD, n = 16; **P < 0.01, ***P < 0.001, using a two-tailed unpaired t-test). j–m, Histograms (j,l) and area under the curve (AUC) quantification of histograms (k,m) from plaque proximity analyses measuring the percentage area of GFA⁺ (j,k) and IBA1⁺ (l,m) cells from the plaque center in 4s2^A/FAD mice compared to 4s2/FAD controls (4s2^A/FAD, n = 8; 4s2/FAD, n = 8). a–m, For detailed statistical information, see Supplementary Table 3. Data represent mean ± s.e.m. of biological replicates. Individual data points have been plotted. Sex is denoted by the shape of the symbol: females (circle), males (triangle). In the box plots, the center line indicates the median; the box limits indicate the upper and lower quartiles; and the whiskers indicate the minimum and maximum. NS, not significant.

Fig. 7. Reactive microglial subpopulations and plaque-associated ApoE decrease after an astrocytic switch to APOE2.

a, Representative images of Amylo-Glo (blue), IBA1 (green), MHC-II (orange) and nuclei (red) staining. b–d, Quantification of total MHC-II⁺ percentage area (b), colocalized MHC-II⁺ and IBA1⁺ area (c), and colocalization of MHC-II⁺, IBA1⁺ and Amylo-Glo⁺ area (d) in 4s2^A/FAD and 4s2/FAD mice in the cortex and hippocampus (4s2^A/FAD, n = 16; 4s2/FAD, n = 16; *P < 0.0, **P < 0.01, ***P < 0.001, ****P < 0.0001 using a two-way ANOVA with multiple comparisons). e,f, Fold change of Tmem119, P2ry12 and Cx3cr1 (e) or Clec7a, Trem2 and APOE (f) gene expression normalized to 18S in the whole brain from 4s2^A/FAD mice compared to 4s2/FAD controls (4s2^A/FAD, n = 14; 4s2/FAD, n = 14; *P < 0.0, **P < 0.01 using a two-tailed unpaired t-test). g, Representative images of Amylo-Glo (blue), ApoE (green) and GFAP (orange) in 4s2^A/FAD mice and 4s2/FAD controls. h, Quantification of plaque-associated ApoE in 4s2^A/FAD mice relative to 4s2/FAD controls in both the cortex and hippocampus (4s2A/FAD, n = 16, 4s2/FAD, n = 16; *P < 0.05, **P < 0.01, using a two-way ANOVA with multiple comparisons). i,j, ELISA quantification of ApoE within the soluble (PBS) (i) and insoluble (5 M GuHCl) (j) brain fractions of 4s2/FAD and 4s2^A/FAD whole-brain tissue (4s2^A/FAD, n = 14; 4s2/FAD, n = 14; *P < 0.05, using a two-tailed unpaired t-test). k, Correlation of cognitive measures and neuropathological analyses. The red squares indicate a positive correlation; the blue squares indicate a negative correlation. a–k, For detailed statistical information, see Supplementary Table 3. Data represent mean ± s.e.m. of biological replicates. Individual data points have been plotted. Sex is denoted by the shape of the symbol: females (circle), males (triangle). In the box plots, the center line indicates the median; the box limits indicate the upper and lower quartiles; and the whiskers indicate the minimum and maximum.

Extended Data Fig. 1. Peaks for unique E2 and unique E4 peptides look similar in TR mice and APOE4s2 and APOE4s2^G mice.

a,d,g,j, Extracted-ion chromatogram (XIC) of peaks from unique E4 peptide LGADMEDVR in an E4-TR control sample (a) and an APOE4s2 sample expressing E4 protein (d) or peaks from unique E2 peptide LGADMEDVRGC in and E2-TR control sample (g) or an APOE4s2^G sample (j) expressing E2 protein. b,e,h,k, Mass spectra (MS) showing peaks with similar mass to charge ratios of unique E4 peptide LGADMEDVR in an E4-TR control sample (b) and an APOE4s2 sample expressing E4 protein (e) or peaks from unique E2 peptide LGADMEDVRGC in and E2-TR control sample (h) or an APOE4s2^G sample (k) expressing E2 protein. c,f,i,l, Tandem mass spectra (MS/MS) showing ion fragment peaks with similar mass to charge ratios of unique E4 peptide LGADMEDVR in an E4-TR control sample (c) and an APOE4s2 sample expressing E4 protein (f) or peaks from unique E2 peptide LGADMEDVRGC in and E2-TR control sample (i) or an APOE4s2^G sample (l) expressing E2 protein. m, Peptide sequences unique to ApoE2 or ApoE4 at amino acid residues 112 and 159 following trypsin digestion. Sequences were used to create a custom database of hApoE2 and hApoE4 peptides. LC-MS/MS data sets generated from plasma and brain samples were searched in against these sequences at both amino acid residues to generate a semi-quantitative ratio of E2:E4 in APOE4s2 and APOE4s2^G samples.

Extended Data Fig. 2. Peripheral and cerebral metabolic profiling of APOE4s2 and ApoE-TR mice and ApoE colocalization with glia.

a, Allelic discrimination plot of mRNA analysis showing efficient and sustained expression of APOE2 in APOE4s2^G mice at 72 h, 3 months, 6 months, and 12 months post switch compared to E2-TR, E4-TR, and APOE4s2 controls (n_E2-TR = 6, n_E4-TR = 6, n_APOE4s2 = 6, n_APOE4s2G = 6 for all timepoints).). b, Percent change in bodyweight recorded between mice on western diet (APOE4s2^G: n = 12 total, n = 6 female, n = 6 male; APOE4s2: n = 13 total, n = 6 females, n = 7 males; E2-TR: n = 18 total, n = 10 females, n = 8 males; E4-TR: n = 12 total, n = 9 females, n = 3 males; no significance; 2way ANOVA with multiple comparisons). c, ApoE quantified in SEC fractions corresponding to VLDL, IDL/LDL and HDL lipoprotein classes. d,f,h, Triglycerides (d), total cholesterol (f), and phospholipids (h) quantified in plasma from E2-TR, E4-TR, APOE4s2, and APOE4s2^G mice on normal chow (n_E2-TR = 3, n_E4-TR = 3, n_APOE4s2 = 5, n_APOE4s2G = 5; *p < 0.05, **p < 0.01, ****p < 0.0001; 2way ANOVA with multiple comparisons). e,g,i, Triglycerides (e), total cholesterol (g), and phospholipids (i) quantified in FPLC eluants corresponding to VLDL, LDL and HDL (n_E2-TR = 3 pooled, n_E4-TR = 3 pooled, n_APOE4s2 = 4 pooled, n_APOE4s2G = 4 pooled). j, Representative images of ApoE (red), NeuN (green), and DAPI (blue) staining in APOE4s2^G and APOE4s2 controls. Scale is 100um. k, Representative images of ApoE (red) and NeuN (green) with DAPI nuclear stain (blue) in E4-TR and E2-TR brains. Scale is 100um. l, Representative images of ApoE (red), IBA1 (green), and DAPI (blue) in E4-TR and E2-TR mice. Scale is 100um. m, Representative images of ApoE (red), GFAP (green), and DAPI (blue) in E4-TR and E2-TR brains. Scale is 100um. n, Quantification of ApoE with astrocytes, microglia, and neurons in the hippocampus. *p < 0.05, ***p < 0.001, 2way ANOVA with multiple comparisons. o, Heatmap depicting top 10 most significantly changed lipids between E2-TR and E4-TR controls. p, Volcano plot showing differentially expressed lipids betweenE4-TR vs. E2-TR mice (n_E2-TR = 8, n_E4-TR = 8; all females). (FDR adjusted p < 0.05 (log10(P) > 1.3)). For detailed statistical information, see supplementary table 3. Data represent mean ± s.e.m. of biological replicates. Individual data points plotted. Box plots show center line, median; box limits, upper and lower quartiles; whiskers, minimum and maximum.

Extended Data Fig. 3. scRNAseq cell-type assignment.

a, UMAP identifying a total of 41 unique clusters before cell-type assignment. b, UMAP showing relatively even distribution of clusters across genotypes. c, After cell-type assignment, plots show relatively even distribution of cell type across different genotypes. d, Violin plots showing expression of gene markers in each cell type. e, Gene expression in each cell-type cluster of cell-specific gene markers.

Extended Data Fig. 4. Plotted DEGs for all single cell comparisons.

a, Volcano plots depicting differentially expressed genes in astrocytes identified between APOE4s2^A vs. APOE4s2, APOE4s2^G vs. APOE4s2, and E2-TR vs. E4-TR comparisons. p = 0.001, log2FC = 1.0. b, Volcano plots depicting differentially expressed genes in oligodendrocytes identified between APOE4s2^A vs. APOE4s2, APOE4s2^G vs. APOE4s2, and E2-TR vs. E4-TR comparisons. p = 0.001, log2FC = 1.0. c, Volcano plots depicting differentially expressed genes in microglia identified between , APOE4s2^A vs. APOE4s2, APOE4s2^G vs. APOE4s2, and E2-TR vs. E4-TR comparisons. p = 0.001, log2FC = 1.0. d, Volcano plots depicting differentially expressed genes in endothelial cells identified between APOE4s2^A vs. APOE4s2, APOE4s2^G vs. APOE4s2, and E2-TR vs. E4-TR comparisons. p = 0.001, log2FC = 1.0.

Extended Data Fig. 5. Validation and plasma lipid profiling of astrocyte-selective APOE4s2^A mice.

a, Allelic discrimination plot of mRNA analysis from whole brain tissue showing efficient and sustained expression of APOE2 in APOE4s2^A mice at 72 h, 3 months, 6 months, and 12 months post switch compared to E2-TR, E4-TR, and APOE4s2 controls (n_E2-TR = 6, n_E4-TR = 6, n_APOE4s2 = 6, n_APOE4s2A = 6 for all timepoints) b, Ratio of unique E2 vs. E4 peptides detected using mass spec based proteomic analysis (n_E2-TR = 2, n_E4-TR = 2, n_APOE4s2 = 3, n_APOE4s2A = 3). c, Percent td-tomato Aldh1l1-Cre reporter positive area vs. tdTomato Aldh1l1-Cre reporter colocalized with GFAP staining in APOE4s2^A (65%) mice vs. APOE4s2 Cre-negative controls (4%%) (_APOE4s2A = 3, n_APOE4s2 = 3). d-e, Allelic discrimination plot from liver tissue (d) with ratios of relative fluorescence from SNPs associated with APOE2 vs. APOE4 showing an average of 81% APOE2 mRNA expression in the liver (e) (n_E2-TR = 6, n_E4-TR = 6, n_APOE4s2 = 6, n_APOE4s2A = 6). f, ApoE quantified by ELISA in SEC fractions corresponding to VLDL, IDL/LDL and HDL lipoprotein classes. g-i, Triglycerides (g), total cholesterol (h), and phospholipids (i) quantified in plasma from E2-TR, E4-TR, APOE4s2, and APOE4s2^A mice on normal chow (n_E2-TR = 3, n_E4-TR = 3, n_APOE4s2 = 5, n_APOE4s2G = 5; *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001; 2way ANOVA with multiple comparisons). j-l Triglycerides (j), total cholesterol (k), and phospholipids (l) quantified in FPLC eluants corresponding to VLDL, LDL and HDL lipoprotein classes (n_E2-TR = 3 pooled, n_E4-TR = 3 pooled, n_APOE4s2 = 4 pooled, n_APOE4s2G = 4 pooled). a-i, For detailed statistical information, see supplementary table 3. Data represent mean ± s.e.m. of biological replicates. Individual data points plotted. Box plots show center line, median; box limits, upper and lower quartiles; whiskers, minimum and maximum.

Extended Data Fig. 6. Sex stratified analysis of behavior showing improvements in associative, but not spatial learning and memory.

a-d, Percent time freezing measured during conditioning in 4s2^A/FAD vs. 4s2/FAD stratified by sex. Analyses showed only one significant sex difference where 4s2/FAD males froze more than 4s2/FAD females. (n_4s2A/FAD = 8 females; 8 males, n_4s2/FAD = 8 females; 8 males; *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001; Repeated measures ANOVA with multiple comparisons; symbols denote significance by sex(es). e, Latency to platform measured in 4s2^A/FAD compared to tamoxifen treated, Cre-negative 4s2/FAD controls over the 5-day hidden platform training trials (n_4s2A/FAD = 10, n_4s2/FAD = 10; two-tailed, paired t-test). f, Thigmotaxis (defined by 30 s of swimming around perimeter) in 4s2^A/FAD and 4s2/FAD controls during hidden platform trials (n_4s2A/FAD = 10, n_4s2/FAD = 10; two-tailed, paired t-test). g, Latency to platform by sex showing no sex differences in 4s2^A/FAD or 4s2/FAD controls. (n_4s2A/FAD = 5 females; 5 males, n_4s2/FAD = 5 females; 5 males; 2way ANOVA with repeated measures and multiple comparisons). h, There were no significant sex-based differences in thigmotaxis (n_4s2A/FAD = 5 females; 5 males, n_4s2/FAD = 5 females; 5 males 2way ANOVA with repeated measures and multiple comparisons). i-k, Number of platform crosses (i), occupancy of target quadrant (j), and occupancy of target quadrant compared to other quadrants (k) in the 72-hour memory probe. (n_4s2A/FAD = 5 females; 5 males, n_4s2/FAD = 5 females; 5 males; **p < 0.01, ***p < 0.001, ****p < 0.0001; two-tailed, unpaired t-test (I,k); 2way ANOVA with multiple comparisons (k)). l-n, Sex-based analyses of platform crosses (l), time in target quadrant (m), and frequency in quadrants relative to other quadrants (n) showed no sex differences (n_4s2A/FAD = 5 females; 5 males, n_4s2/FAD = 5 females; 5 males; *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001; 2way ANOVA with multiple comparisons. a-n. For detailed statistical information, see supplementary table 3. Data represent mean ± s.e.m. of biological replicates. Individual data points plotted. Sex is denoted by symbol shape: females (circle), males (triangle). Box plots show center line, median; box limits, upper and lower quartiles; whiskers, minimum and maximum.

Extended Data Fig. 7. Amyloid deposition by sex, plaque toxicity, and tight junction protein quantification.

a, Quantification of total amylo-glo+ area by sex showing significantly higher amyloid+ area in 4s2/FAD females than 4s2/FAD males (n_4s2A/FAD = 8 females; 8 males, n_4s2/FAD = 8 females; 8 males; **p < 0.01; 2way ANOVA with multiple comparisons). b, Representative image of amylo-Glo (blue) and nuclei (red) staining in 6-month-old (pre-switch) 4s2/FAD mice. Scale is 1 mm. c, Percent amyloid+ area quantified over the 2-month period of astrocytic E2 expression showing a significant increase in amyloid+ percent area in8-month-old 4s2/FAD mice. (n_4s2A/FAD-6mo = 8, n_4s2/FAD-8mo. = 16, n_{4s2A/FAD-8mo.} = 16; **p < 0.01; 2way ANOVA with multiple comparisons). d, Average plaque size at 8 mo. between 4s2/FAD controls and 4s2^A/FAD mice (n_4s2/FAD = 7, n_4s2A/FAD = 7; two-tailed, unpaired t-test). e-h, Sex based analyses ELISA quantification of soluble and insoluble Ab₄₂ and Ab₄₀ species show no sex differences but suggest decreases in amyloid could be driven by trends within one sex (n_4s2A/FAD = 7 females; 7 males, n_4s2/FAD = 7 females; 7 males; *p < 0.05, **p < 0.01, ***p < 0.001, 2way ANOVA with multiple comparisons). i, Representative image of proximity analysis used to quantify PSD-95 puncta in 3um increments up to 15um from the plaque. PSD-95 (green), Amylo-Glo (blue), nuclei (red). Scale is 100um. j, l, PSD-95 quantification in the cortex showing decreases in PSD-95 in 4s2^A/FAD mice starting at 12um from plaque center (j). Sex based analyses revealed that 4s2/FAD females had significantly more PSD-95 puncta than 4s2/FAD males farther from the plaque (l) (n_4s2A/FAD = 6 females; 6 males, n_4s2/FAD = 6 females; 6 males; **p < 0.01, ***p < 0.001, ****p < 0.0001; 2way ANOVA with multiple comparisons). k, m, PSD-95 quantification in the hippocampus (n_4s2A/FAD = 6 females; 6 males, n_4s2/FAD = 6 females; 6 males; 2way ANOVA with multiple comparisons). n, Representative images of CD31 (red), ZO1 (green), and DAPI (blue) staining. Scale is 100um. o, q, Quantification of tight junction protein ZO1 (o) with sex stratified analyses showing that 4s2^A/FAD females have higher ZO1 than 4s2^A/FAD males (q) (n_4s2A/FAD = 8 females; 8 males, n_4s2/FAD = 8 females; 8 males; **p < 0.01; 2way ANOVA with multiple comparisons). p, r, Quantification of total % area of CD31+ vessels (n_4s2A/FAD = 8 females; 8 males, n_4s2/FAD = 8 females; 8 males; 2way ANOVA with multiple comparisons). a-r, Detailed statistical information in supplementary table 3. Data represent mean ± s.e.m. of biological replicates. Individual data points plotted. Sex is denoted by symbol shape: females (circle), males (triangle). Box plots show center line, median; box limits, upper and lower quartiles; whiskers, minimum and maximum.

Extended Data Fig. 8. Amyloid-associated gliosis measures stratified by sex.

a,c, Quantification of total percent GFAP+ (a) or IBA1+ (c) area in 4s2^A/FAD and 4s2/FAD controls at 8 months compared to baseline measures of gliosis at 6 months (n_APOE4s2-6mo = 8, n_APOE4s2-8mo. = 16, n_{APOE4s2A-8mo.} = 16; **p < 0.01, ****p < 0.000, 2way ANOVA with multiple comparisons). b,d, Quantification by sex of total percent GFAP+ (a) or IBA1+ (c) area in 4s2^A/FAD and 4s2/FAD controls at 8 months compared to baseline measures of gliosis at 6 months (n_APOE4s2-6mo = 8, n_APOE4s2-8mo. = 8 females; 8 males, n_{APOE4s2A-8mo.} = 8 females; 8 males; *p < 0.05, **p < 0.01, ****p < 0.0001; 2way ANOVA with multiple comparisons). e-g, Sex stratified analyses for colocalization of GFAP+ astrocytes and plaque (e), IBA1+ microglia and plaque (f), and plaque with both GFAP+ astrocytes and IBA1+ microglia (g). (n_APOE4s2-6mo = 8, n_APOE4s2-8mo. = 8 females; 8 males, n_{APOE4s2A-8mo.} = 8 females; 8 males; *p < 0.05, **p < 0.01; 2way ANOVA with multiple comparisons). h,j, Representative images exemplifying plaque and glial cell proximity analyses in HALO software for proximity analysis of amyloid plaques and GFAP positive cells (h), or amyloid plaques and IBA1 positive cells (j). Scale is 100um. i-k, Plaque proximity analyses by sex (n_APOE4s2-6mo = 4, n_APOE4s2-8mo. = 4 females; 4 males, n_{APOE4s2A-8mo.} = 4 females; 4 males; *p < 0.05; 2way ANOVA with multiple comparisons). a-k, Detailed statistical information in supplementary table 3. Data represent mean ± s.e.m. of biological replicates. Individual data points plotted. Sex is denoted by symbol shape: females (circle), males (triangle). Box plots show center line, median; box limits, upper and lower quartiles; whiskers, minimum and maximum.

Extended Data Fig. 9. Sex based analyses of microglia sub-populations and plaque-associated ApoE.

a-c, Sex stratified quantification of cortical MHCII+ percent area (a) MHCII+ and IBA1+ percent area (b) or MHCII + , IBA1 + , and amylo-glo+ percent area (c) (n_APOE4s2-6mo = 8, n_APOE4s2-8mo. = 8 females; 8 males, n_{APOE4s2A-8mo.} = 8 females; 8 males; *p < 0.05, **p < 0.01, ***p < 0.001. 2way ANOVA with multiple comparisons). d-f, Sex stratified quantification of hippocampal MHCII+ percent area (d) MHCII+ and IBA1+ percent area (e) or MHCII + , IBA1 + , and amylo-glo+ percent area (f) (n_APOE4s2-6mo = 8, n_APOE4s2-8mo. = 8 females; 8 males, n_{APOE4s2A-8mo.} = 8 females; 8 males; *p < 0.05, **p < 0.01, ***p < 0.001, 2way ANOVA with multiple comparisons). g-h, Sex stratified analysis of homeostatic (g) and DAM (h) gene expression showed no sex differences between groups. i, Representative images of ApoE (green), amylo-glo (blue), and IBA1 (orange) staining. Scale is 100um. j-k, Quantification of percent area of total ApoE in the cortex (j) or cortical plaque-associated ApoE (k) with significant reduction in total cortical ApoE, as well as plaque-associated ApoE, appearing to be driven by 4s2^A/FAD females (n_APOE4s2-6mo = 8, n_APOE4s2-8mo. = 8 females; 8 males, n_{APOE4s2A-8mo.} = 8 females; 8 males; *p < 0.05. 2way ANOVA with multiple comparisons. l-m, Quantification of percent area of total ApoE in the hippocampus (l) or plaque-associated ApoE (m) (n_APOE4s2-6mo = 8, n_APOE4s2-8mo. = 8 females; 8 males, n_{APOE4s2A-8mo.} = 8 females; 8 males; 2way ANOVA with multiple comparisons). n-o, ELISA quantification of total brain soluble ApoE (n) and insoluble ApoE (o) by sex 4s2/FAD females had higher insoluble ApoE than males. (n_APOE4s2-6mo = 8, n_APOE4s2-8mo. = 8 females; 8 males, n_{APOE4s2A-8mo.} = 8 females; 8 males; *p < 0.05, **p < 0.01. 2way ANOVA with multiple comparisons. a-o, Detailed statistical information in supplementary table 3. Data represent mean ± s.e.m. of biological replicates. Individual data points plotted. Sex is denoted by symbol shape: females (circle), males (triangle). Box plots show center line, median; box limits, upper and lower quartiles; whiskers, minimum and maximum.