Alzheimer's disease-linked risk alleles elevate microglial cGAS-associated senescence and neurodegeneration in a tauopathy model

Abstract

The strongest risk factors for late-onset sporadic Alzheimer's disease (AD) include the ε4 allele of apolipoprotein E (APOE), the R47H variant of triggering receptor expressed on myeloid cells 2 (TREM2), and female sex. Here, we combine APOE4 and TREM2^R47H (R47H) in female P301S tauopathy mice to identify the pathways activated when AD risk is the strongest, thereby highlighting detrimental disease mechanisms. We find that R47H induces neurodegeneration in 9- to 10-month-old female APOE4 tauopathy mice. The combination of APOE4 and R47H (APOE4-R47H) worsened hyperphosphorylated tau pathology in the frontal cortex and amplified tauopathy-induced microglial cyclic guanosine monophosphate (GMP)-AMP synthase (cGAS)-stimulator of interferon genes (STING) signaling and downstream interferon response. APOE4-R47H microglia displayed cGAS- and BAX-dependent upregulation of senescence, showing association between neurotoxic signatures and implicating mitochondrial permeabilization in pathogenesis. By uncovering pathways enhanced by the strongest AD risk factors, our study points to cGAS-STING signaling and associated microglial senescence as potential drivers of AD risk.

Keywords: APOE; Alzheimer’s disease; R47H; TREM2; cGAS; inflammation; interferon; microglia; senescence; tau.

Figure 1.. APOE4-R47H induces neurodegeneration and synaptic loss in female tauopathy mice

(A) Representative images of 9–10-month-old female mouse brains stained with Sudan black; scale bar, 500μm. (B and C) Quantification of ventricle volume (B) and hippocampal volume (C) based on staining from (A). Each circle represents sum volume measurements of five to seven brain sections per animal. Two statistically significant outliers were removed for ventricle volume. **P = 0.0063, n.s. not significant. Kruskal-Wallis test, Dunn’s multiple comparisons test. Data are reported as boxplot with min. to max.; E4/+/+, n = 17; E4/R47H/+, n = 12; E4/+/+/P301S, n = 11; E4/R47H/+/P301S, n = 9. (D) Representative immunofluorescence images of 9–10-month-old female mouse hippocampal CA1 stratum radiatum labeled with anti-PSD95 (green); scale bar, 10μm. (E) Mean intensity of PSD95-positive puncta in the female hippocampal CA1 stratum radiatum. Each circle represents mean intensity of three to four brain sections per animal. **P = 0.0043, n.s. not significant. Two-way ANOVA, Tukey’s multiple comparisons test. Data are reported as boxplot with min. to max.; E4/+/+, n = 8; E4/R47H/+, n = 13; E4/+/+/P301S, n = 12; E4/R47H/+/P301S, n = 13. (F) Representative immunofluorescence images of 9–10-month-old female mouse hippocampal CA1 stratum radiatum labeled with anti-VGAT (green); scale bar, 10μm. (G) Mean intensity of VGAT-positive puncta in the female hippocampal CA1 stratum radiatum. Each circle represents mean intensity of three to four brain sections per animal. One statistically significant outlier was removed. **P = 0.0038, n.s. not significant. Kruskal-Wallis test, Dunn’s multiple comparisons test. Data are reported as boxplot with min. to max.; E4/+/+, n = 13; E4/R47H/+, n = 12; E4/+/+/P301S, n = 11; E4/R47H/+/P301S, n = 8. See also Figure S1.

Figure 2.. APOE4-R47H increases hyperphosphorylated tau in the frontal cortex, but not in the hippocampus, of P301S mice

(A) Representative immunofluorescence images of 9–10-month-old female mouse hippocampi labeled with anti-MC1 (red) and DAPI (blue); scale bar, 500μm. (B) Representative immunofluorescence images of 9–10-month-old female mouse hippocampal DG or CA3 subregions labeled with anti-MC1 (red); scale bars, 100μm or 50μm, as labeled. (C) Mean intensity of MC1 across the full hippocampus in female mice. Each circle represents mean intensity of three to four brain sections per animal. n.s. not significant. Mann-Whitney test. Data are reported as boxplot with min. to max.; E4/+/+/P301S, n = 11; E4/R47H/+/P301S, n = 9. (D and E) Mean intensity of MC1 within the DG (D) or CA3 (E) hippocampal subregions of female mice. Two statistically significant outliers were removed for each subregion. n.s. not significant. Unpaired two-tailed t-test. Data are reported as boxplot with min. to max.; E4/+/+/P301S, n = 10; E4/R47H/+/P301S, n = 8. (F) Western blots for AT8, HT7, and β-Actin using 9–10-month-old female mouse hippocampal tissue lysates. (G, H, I, and J) Western blot quantification of HT7 (G), AT8 (H), high molecular weight AT8 (I), and AT8/HT7 ratio (J) normalized to β-Actin in the female mouse hippocampus. Each circle represents one animal. n.s. not significant. Unpaired two-tailed t-test. Data are reported as mean ± S.E.M.; n = 6 mice per genotype. (K) Western blots for AT8, HT7, and β-Actin using 9–10-month-old female mouse frontal cortex tissue lysates. (L, M, N, and O) Western blot quantification HT7 (L), AT8 (M), high molecular weight AT8 (N), and AT8/HT7 ratio (O) normalized to β-Actin in the female mouse frontal cortex. Each circle represents one animal. *P < 0.05, **P < 0.01, ***P < 0.001. Unpaired two-tailed t-test. Data are reported as mean ± S.E.M.; n = 6 mice per genotype.

Figure 3.. APOE4-R47H elevates tauopathy-induced IFN-I signaling in female microglia

(A) Representative immunofluorescence images of the 9–10-month-old female mouse hippocampal CA3 subregion labeled with anti-IBA1 (green); scale bar, 20μm. (B) Mean intensity of IBA1 across the hippocampus of female mice. Each circle represents mean intensity of three to four brain sections per animal. One statistically significant outlier was removed. *P = 0.0374, ***P = 0.0002, ****P < 0.0001, n.s. not significant. Two-way ANOVA, Tukey’s multiple comparisons test. Data are reported as boxplot with min. to max.; E4/+/+, n = 14; E4/R47H/+, n = 12; E4/+/+/P301S, n = 11; E4/R47H/+/P301S, n = 8. (C) UMAP plot of 9–10-month-old female APOE4/APOE4 hippocampal immune cell nuclei split by genotype and colored according to subclusters. n = 3 mice per genotype. (D) Violin plots showing cell type markers split by immune cell subcluster to highlight cluster identity. (E) Quantification of cell ratio per immune cell subcluster within each genotype for 9–10-month-old female mice. Each circle represents one animal. *P < 0.05, **P < 0.01, ***P < 0.001. One-way ANOVA, Tukey’s multiple comparisons test. Data are reported as mean ± S.E.M.; n = 3 mice per genotype. (F) Dot plot showing top IPA upstream regulator predictions for IFN signaling molecules based on microglial subcluster 4 markers. (G) IPA upstream regulator IRF3 activated network and downstream predicted targets. (H) Dot plot showing IFN-stimulated genes enriched in microglial subcluster 4 and significantly upregulated in E4/R47H/+/P301S versus E4/+/+/P301S female hippocampal microglia. (I) Representative immunofluorescence images of 9–10-month-old female mouse hippocampal CA1 subregion labeled with anti-IBA1 (green) and anti-pSTAT1 (red); scale bar, 20μm. (J) Quantification of pSTAT1 intensity within IBA1 overlapping regions in the female mouse hippocampal CA1 subregion. Each circle represents the mean quantification of three to four brain sections per animal. *P = 0.0269, n.s. not significant. Two-way ANOVA, Tukey’s multiple comparisons test. Data are reported as boxplot with min. to max.; E4/+/+, n = 11; E4/R47H/+, n = 11; E4/+/+/P301S, n = 12; E4/R47H/+/P301S, n = 12. See also Figures S2–S3 and Table S1.

Figure 4.. APOE4-R47H exacerbates tauopathy-induced microglial cGAS-STING activation

(A) IPA cGAS activated network and downstream predicted targets. IPA predicts cGAS as an upstream regulator of IFN-high microglial subcluster 4 (activation z-score = 3.573, P < 0.0001). (B) Dot plot showing Cgas expression split by genotype in 9–10-month-old female mouse hippocampal microglia. (C and D) Western blot (C) and quantification (D) of cGAS normalized to β-Actin in 9–10-month-old female frontal cortex tissue lysates. Each circle represents one animal. *P < 0.05, n.s. not significant. Two-way ANOVA, Tukey’s multiple comparisons test. Data are reported as mean ± S.E.M.; n = 6 mice per genotype. (E and F) Western blot (E) and quantification (F) of cGAS normalized to β-Actin in 9–10-month-old female hippocampal tissue lysates. Each circle represents one animal. **P = 0.0015, n.s. not significant. Two-way ANOVA, Tukey’s multiple comparisons test. Data are reported as mean ± S.E.M.; n = 5 mice per genotype for P301S^- and n = 6 mice per genotype for P301S⁺. (G and H) Western blot (G) and quantification (H) of STING normalized to β-Actin in 9–10-month-old female hippocampal tissue lysates. Each circle represents one animal. **P = 0.0097, ****P < 0.0001, n.s. not significant. Two-way ANOVA, Tukey’s multiple comparisons test. Data are reported as mean ± S.E.M.; n = 6 mice per genotype. (I) Representative immunofluorescence images of 9–10-month-old female mouse hippocampal CA1 subregion labeled with anti-IBA1 (green) and anti-STING (red); scale bar, 20μm. (J) Quantification of STING intensity within IBA1 overlapping regions in the hippocampal CA1 subregion of female mice. Each circle represents the mean quantification of three to four brain sections per animal. **P = 0.0022, n.s. not significant. Two-way ANOVA, Tukey’s multiple comparisons test. Data are reported as boxplot with min. to max.; E4/+/+, n = 14; E4/R47H/+, n = 11; E4/+/+/P301S, n = 11; E4/R47H/+/P301S, n = 9. (K and L) Correlation scatterplot of microglial STING quantification compared to microglial pSTAT1 quantification (K) and brain ventricle volume quantification (L) in female mice. Simple linear regression. n = 18–19 P301S⁺ mice. See also Figures S4–S5.

Figure 5.. APOE4-R47H induces LTP deficits in female tauopathy mice and elevates tau-provoked cGAS-STING-IFN signaling in primary microglia

(A and B) Western blot (A) and quantification (B) of hTREM2 normalized to β-Actin in 9–10-month-old female frontal cortex tissue lysates. Each circle represents one animal. Data not significant. Two-way ANOVA, Tukey’s multiple comparisons test. Data are reported as mean ± S.E.M.; n = 6 mice per genotype. (C) Field excitatory postsynaptic potential (fEPSP) recordings from the CA1 hippocampal subregion of 7–8-month-old female mice. Schematic shows LTP measurement strategy. fEPSPs were recorded in the CA1 hippocampal subregion, and a TBS protocol was applied to the CA3 pathway to induce LTP (Schematic created with BioRender.com). An arrow denotes the time-point where TBS was applied. Each circle represents one mouse. One to two brain slices per mouse were used. Data are reported as mean ± S.E.M.; E4/CV/CV, n = 6; E4/R47H/CV, n = 7; E4/CV/CV/P301S, n = 10; E4/R47H/CV/P301S, n = 10. (D) LTP residual potentiation calculated from the mean fEPSP slope 50–60 minutes after TBS was applied. Each circle represents one mouse. One to two brain slices per mouse were used. **P = 0.0012, n.s. not significant. Two-way ANOVA, Tukey’s multiple comparisons test. Data are reported as mean ± S.E.M.; E4/CV/CV, n = 6; E4/R47H/CV, n = 7; E4/CV/CV/P301S, n = 10; E4/R47H/CV/P301S, n = 10. (E) Schematic showing bulk RNA-sequencing experimental design for E4/CV (APOE4/APOE4; hTREM2-CV/CV) and E4/R47H (APOE4/APOE4; hTREM2-R47H/CV) primary mouse microglia at baseline and after 18-hour 0N4R tau stimulation. Primary microglia were isolated from pups of both sexes. n = 3 biological replicates. Schematic created with BioRender.com. (F and G) GSEA showing hallmark pathways based on the top 500 upregulated and downregulated DEGs in E4/R47H versus E4/CV primary microglia at baseline (F) and after tau stimulation (G). (H) Volcano plot of RNA-seq comparing E4/R47H versus E4/CV tau-stimulated primary microglia. Red and blue dots represent significant DEGs (P < 0.05). (I) Dot plot showing top IPA upstream regulator predictions for E4/R47H versus E4/CV tau-stimulated primary microglia. (J) IPA upstream regulator IFNB1 activated network and downstream predicted targets. (K) Bar plot showing primary microglial secretion of IFNβ in conditioned media after 18-hour 0N4R tau stimulation. Data is normalized to cell count per well. Each circle represents one biological replicate. **P = 0.0029. Mann-Whitney test. Data are reported as mean ± S.E.M.; n = 12 biological replicates performed in 3 independent experiments. (L and M) Representative western blots (L) and quantification (M) of cGAS normalized to β-Actin in primary microglial lysates with or without 18-hour 0N4R tau stimulation. Each circle represents one biological replicate. **P = 0.0025, ****P < 0.0001. Two-way ANOVA, Tukey’s multiple comparisons test. Data are reported as mean ± S.E.M.; n = 6 biological replicates performed in 2 independent experiments. See also Figures S6–S7 and Table S6.

Figure 6.. APOE4-R47H induces p21⁺ senescence in tauopathy mice

(A) IPA activated canonical pathways based on DEGs (adjusted P value < 0.05, log₂FC ≥ 0.1 or ≤ −0.1) in E4/R47H/+/P301S versus E4/+/+/P301S hippocampal microglia from 9–10-month-old female mice. n = 3 mice per genotype. (B) Heatmap showing senescence pathway genes from (A) that are significantly altered in E4/R47H/+/P301S versus E4/+/+/P301S hippocampal microglia. (C, D, and E) Bar plots of CXCL10 (C), CCL5 (D), and CCL3 (E) chemokine concentrations in frontal cortex lysate from 9–10-month-old female mice. Chemokine secretion was normalized to total protein concentration in the lysate. Each circle represents one animal. *P < 0.05, **P < 0.01, ****P < 0.0001, n.s. not significant. Two-way ANOVA, Tukey’s multiple comparisons test. Data are reported as mean ± S.E.M.; n = 6 mice per genotype. (F) Representative immunofluorescence images of 9–10-month-old female mouse hippocampal CA3 subregion labeled with anti-IBA1 (green) and anti-p21 (red); scale bar, 20μm. (G) Quantification of p21 intensity within IBA1 overlapping regions in the hippocampal CA3 subregion of female mice. Each circle represents the mean quantification of three to four brain sections per animal. *P = 0.0330, n.s. not significant. Kruskal-Wallis test, Dunn’s multiple comparisons test. Data are reported as boxplot with min. to max.; E4/+/+, n = 10; E4/R47H/+, n = 10; E4/+/+/P301S, n = 10; E4/R47H/+/P301S, n = 12. (H and I) Correlation scatterplot of microglial p21 quantification compared to hippocampal MC1 quantification (H) and microglial STING quantification (I). Simple linear regression. n = 17 P301S+ mice. See also Table S1.

Figure 7.. APOE4-R47H amplifies BAX- and cGAS-dependent senescence in tau-treated microglia

(A) Heatmap showing mass spectrometry of secreted proteins in conditioned media from E4/R47H and E4/CV primary mouse microglia at baseline and after 72-hour 0N4R tau simulation. Genes displayed in the heatmap are SenMayo senescence-associated genes. n = 3 biological replicates. (B) Volcano plot comparing secreted proteins in E4/R47H versus E4/CV tau-stimulated primary microglia. Red and blue circles represent DEGs. Red and blue closed dots represent SenMayo senescence-associated proteins. Labels denote senescence-associated proteins significantly (P < 0.05) enriched in tau-stimulated E4/R47H microglia. n = 3 biological replicates. (C) Correlation scatterplot comparing secreted proteins from E4/R47H primary microglia after 72-hour tau simulation versus 72-hour etoposide treatment. Red and blue circles represent DEGs. Red and blue closed dots represent SenMayo senescence-associated proteins. R = 0.95, P < 0.0001. Simple linear regression. n = 3 biological replicates. (D) Representative immunofluorescence images of primary microglia labeled with anti-γH2AX (red) and DAPI (blue) after 72-hour tau stimulation; scale bar, 20μm. (E) Mean γH2AX intensity per cell in primary microglia after 72-hour tau exposure. Each circle represents mean intensity from six images per well. One statistically significant outlier was removed. ****P < 0.0001. Two-way ANOVA, Tukey’s multiple comparisons test. Data are reported as mean ± S.E.M.; n = 8 biological replicates performed in 2 independent experiments. (F) Heatmap showing proliferation genes that are significantly diminished in E4/R47H versus E4/CV tau-treated primary mouse microglia. n = 3 biological replicates. (G) BrdU proliferation ELISA in primary microglia at baseline, after 18-hour etoposide exposure, or after 18-hour tau stimulation. Each circle represents one biological replicate. **P = 0.0085, ***P = 0.0003, ****P < 0.0001. Kruskal-Wallis test, Dunn’s multiple comparisons test. Data are reported as mean ± S.E.M.; n = 16 biological replicates performed in 2 independent experiments. (H) Bar plot showing Cdkn1a normalized RNA counts after 18-hour tau stimulation in primary microglia. Each circle represents one biological replicate. **P = 0.0021. Unpaired two-tailed t-test. Data are reported as mean ± S.E.M.; n = 3 biological replicates. (I) Representative immunofluorescence images of primary microglia labeled with anti-p21 (green) and DAPI (blue) after 72-hour tau exposure, with or without BAX inhibition (BAI1) or cGAS inhibition (TDI6570); scale bar, 20μm. (J) Mean p21 intensity per cell in primary microglia after 72-hour tau exposure, with or without BAX inhibition (BAI1) or cGAS inhibition (TDI6570). Each circle represents mean intensity from six to eight images per well. One statistically significant outlier was removed. *P < 0.05, ****P < 0.0001. Kruskal-Wallis test, Dunn’s multiple comparisons test. Data are reported as mean ± S.E.M.; E4/CV/CV +Tau, n = 15 and E4/R47H/CV +Tau, n = 14 performed in 4 independent experiments; E4/R47H/CV +Tau +BAXi, n = 8 and E4/R47H/CV +Tau +cGASi, n = 7 performed in 2 independent experiments. (K) Illustration of proposed mechanism for senescence induction in APOE4-R47H tauopathy microglia. Tau and APOE4-R47H-related stressors increase BAX-mediated mitochondrial membrane permeabilization, causing mtDNA release and DNA damage. This activates cGAS-STING-IFN and downstream senescence. Inhibition of BAX or cGAS reduces p21⁺ senescence. Schematic created with BioRender.com. See also Table S7.